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 @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 -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{PDP-11 Options}
844 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
845 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
846 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
847 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
848 -mbranch-expensive -mbranch-cheap @gol
849 -munix-asm -mdec-asm}
851 @emph{picoChip Options}
852 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
853 -msymbol-as-address -mno-inefficient-warnings}
855 @emph{PowerPC Options}
856 See RS/6000 and PowerPC Options.
859 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78 @gol
860 -m64bit-doubles -m32bit-doubles}
862 @emph{RS/6000 and PowerPC Options}
863 @gccoptlist{-mcpu=@var{cpu-type} @gol
864 -mtune=@var{cpu-type} @gol
865 -mcmodel=@var{code-model} @gol
867 -maltivec -mno-altivec @gol
868 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
869 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
870 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
871 -mfprnd -mno-fprnd @gol
872 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
873 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
874 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
875 -malign-power -malign-natural @gol
876 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
877 -msingle-float -mdouble-float -msimple-fpu @gol
878 -mstring -mno-string -mupdate -mno-update @gol
879 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
880 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
881 -mstrict-align -mno-strict-align -mrelocatable @gol
882 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
883 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
884 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
885 -mprioritize-restricted-insns=@var{priority} @gol
886 -msched-costly-dep=@var{dependence_type} @gol
887 -minsert-sched-nops=@var{scheme} @gol
888 -mcall-sysv -mcall-netbsd @gol
889 -maix-struct-return -msvr4-struct-return @gol
890 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
891 -mblock-move-inline-limit=@var{num} @gol
892 -misel -mno-isel @gol
893 -misel=yes -misel=no @gol
895 -mspe=yes -mspe=no @gol
897 -mgen-cell-microcode -mwarn-cell-microcode @gol
898 -mvrsave -mno-vrsave @gol
899 -mmulhw -mno-mulhw @gol
900 -mdlmzb -mno-dlmzb @gol
901 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
902 -mprototype -mno-prototype @gol
903 -msim -mmvme -mads -myellowknife -memb -msdata @gol
904 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
905 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
906 -mno-recip-precision @gol
907 -mveclibabi=@var{type} -mfriz -mno-friz @gol
908 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
909 -msave-toc-indirect -mno-save-toc-indirect @gol
910 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
911 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
912 -mquad-memory -mno-quad-memory @gol
913 -mquad-memory-atomic -mno-quad-memory-atomic @gol
914 -mcompat-align-parm -mno-compat-align-parm @gol
915 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
916 -mupper-regs -mno-upper-regs}
919 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
921 -mbig-endian-data -mlittle-endian-data @gol
924 -mas100-syntax -mno-as100-syntax@gol
926 -mmax-constant-size=@gol
929 -mno-warn-multiple-fast-interrupts@gol
930 -msave-acc-in-interrupts}
932 @emph{S/390 and zSeries Options}
933 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
934 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
935 -mlong-double-64 -mlong-double-128 @gol
936 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
937 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
938 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
939 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
940 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
941 -mhotpatch=@var{halfwords},@var{halfwords}}
944 @gccoptlist{-meb -mel @gol
948 -mscore5 -mscore5u -mscore7 -mscore7d}
951 @gccoptlist{-m1 -m2 -m2e @gol
952 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
954 -m4-nofpu -m4-single-only -m4-single -m4 @gol
955 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
956 -m5-64media -m5-64media-nofpu @gol
957 -m5-32media -m5-32media-nofpu @gol
958 -m5-compact -m5-compact-nofpu @gol
959 -mb -ml -mdalign -mrelax @gol
960 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
961 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
962 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
963 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
964 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
965 -maccumulate-outgoing-args -minvalid-symbols @gol
966 -matomic-model=@var{atomic-model} @gol
967 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
968 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
969 -mpretend-cmove -mtas}
971 @emph{Solaris 2 Options}
972 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
976 @gccoptlist{-mcpu=@var{cpu-type} @gol
977 -mtune=@var{cpu-type} @gol
978 -mcmodel=@var{code-model} @gol
979 -mmemory-model=@var{mem-model} @gol
980 -m32 -m64 -mapp-regs -mno-app-regs @gol
981 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
982 -mfpu -mno-fpu -mhard-float -msoft-float @gol
983 -mhard-quad-float -msoft-quad-float @gol
984 -mstack-bias -mno-stack-bias @gol
985 -munaligned-doubles -mno-unaligned-doubles @gol
986 -muser-mode -mno-user-mode @gol
987 -mv8plus -mno-v8plus -mvis -mno-vis @gol
988 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
989 -mcbcond -mno-cbcond @gol
990 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
991 -mfix-at697f -mfix-ut699}
994 @gccoptlist{-mwarn-reloc -merror-reloc @gol
995 -msafe-dma -munsafe-dma @gol
997 -msmall-mem -mlarge-mem -mstdmain @gol
998 -mfixed-range=@var{register-range} @gol
1000 -maddress-space-conversion -mno-address-space-conversion @gol
1001 -mcache-size=@var{cache-size} @gol
1002 -matomic-updates -mno-atomic-updates}
1004 @emph{System V Options}
1005 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1007 @emph{TILE-Gx Options}
1008 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1009 -mcmodel=@var{code-model}}
1011 @emph{TILEPro Options}
1012 @gccoptlist{-mcpu=@var{cpu} -m32}
1015 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1016 -mprolog-function -mno-prolog-function -mspace @gol
1017 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1018 -mapp-regs -mno-app-regs @gol
1019 -mdisable-callt -mno-disable-callt @gol
1020 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1021 -mv850e -mv850 -mv850e3v5 @gol
1032 @gccoptlist{-mg -mgnu -munix}
1034 @emph{Visium Options}
1035 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1036 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1039 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1040 -mpointer-size=@var{size}}
1042 @emph{VxWorks Options}
1043 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1044 -Xbind-lazy -Xbind-now}
1047 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1048 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1049 -mfpmath=@var{unit} @gol
1050 -masm=@var{dialect} -mno-fancy-math-387 @gol
1051 -mno-fp-ret-in-387 -msoft-float @gol
1052 -mno-wide-multiply -mrtd -malign-double @gol
1053 -mpreferred-stack-boundary=@var{num} @gol
1054 -mincoming-stack-boundary=@var{num} @gol
1055 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1056 -mrecip -mrecip=@var{opt} @gol
1057 -mvzeroupper -mprefer-avx128 @gol
1058 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1059 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
1060 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
1061 -mclflushopt -mxsavec -mxsaves @gol
1062 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1063 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mthreads @gol
1064 -mno-align-stringops -minline-all-stringops @gol
1065 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1066 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1067 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1068 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1069 -mregparm=@var{num} -msseregparm @gol
1070 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1071 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1072 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1073 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1074 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
1075 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1076 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1077 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
1079 @emph{x86 Windows Options}
1080 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1081 -mnop-fun-dllimport -mthread @gol
1082 -municode -mwin32 -mwindows -fno-set-stack-executable}
1084 @emph{Xstormy16 Options}
1087 @emph{Xtensa Options}
1088 @gccoptlist{-mconst16 -mno-const16 @gol
1089 -mfused-madd -mno-fused-madd @gol
1091 -mserialize-volatile -mno-serialize-volatile @gol
1092 -mtext-section-literals -mno-text-section-literals @gol
1093 -mtarget-align -mno-target-align @gol
1094 -mlongcalls -mno-longcalls}
1096 @emph{zSeries Options}
1097 See S/390 and zSeries Options.
1099 @item Code Generation Options
1100 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1101 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1102 -ffixed-@var{reg} -fexceptions @gol
1103 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1104 -fasynchronous-unwind-tables @gol
1105 -fno-gnu-unique @gol
1106 -finhibit-size-directive -finstrument-functions @gol
1107 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1108 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1109 -fno-common -fno-ident @gol
1110 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1111 -fno-jump-tables @gol
1112 -frecord-gcc-switches @gol
1113 -freg-struct-return -fshort-enums @gol
1114 -fshort-double -fshort-wchar @gol
1115 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1116 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1117 -fno-stack-limit -fsplit-stack @gol
1118 -fleading-underscore -ftls-model=@var{model} @gol
1119 -fstack-reuse=@var{reuse_level} @gol
1120 -ftrapv -fwrapv -fbounds-check @gol
1121 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1122 -fstrict-volatile-bitfields -fsync-libcalls}
1126 @node Overall Options
1127 @section Options Controlling the Kind of Output
1129 Compilation can involve up to four stages: preprocessing, compilation
1130 proper, assembly and linking, always in that order. GCC is capable of
1131 preprocessing and compiling several files either into several
1132 assembler input files, or into one assembler input file; then each
1133 assembler input file produces an object file, and linking combines all
1134 the object files (those newly compiled, and those specified as input)
1135 into an executable file.
1137 @cindex file name suffix
1138 For any given input file, the file name suffix determines what kind of
1139 compilation is done:
1143 C source code that must be preprocessed.
1146 C source code that should not be preprocessed.
1149 C++ source code that should not be preprocessed.
1152 Objective-C source code. Note that you must link with the @file{libobjc}
1153 library to make an Objective-C program work.
1156 Objective-C source code that should not be preprocessed.
1160 Objective-C++ source code. Note that you must link with the @file{libobjc}
1161 library to make an Objective-C++ program work. Note that @samp{.M} refers
1162 to a literal capital M@.
1164 @item @var{file}.mii
1165 Objective-C++ source code that should not be preprocessed.
1168 C, C++, Objective-C or Objective-C++ header file to be turned into a
1169 precompiled header (default), or C, C++ header file to be turned into an
1170 Ada spec (via the @option{-fdump-ada-spec} switch).
1173 @itemx @var{file}.cp
1174 @itemx @var{file}.cxx
1175 @itemx @var{file}.cpp
1176 @itemx @var{file}.CPP
1177 @itemx @var{file}.c++
1179 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1180 the last two letters must both be literally @samp{x}. Likewise,
1181 @samp{.C} refers to a literal capital C@.
1185 Objective-C++ source code that must be preprocessed.
1187 @item @var{file}.mii
1188 Objective-C++ source code that should not be preprocessed.
1192 @itemx @var{file}.hp
1193 @itemx @var{file}.hxx
1194 @itemx @var{file}.hpp
1195 @itemx @var{file}.HPP
1196 @itemx @var{file}.h++
1197 @itemx @var{file}.tcc
1198 C++ header file to be turned into a precompiled header or Ada spec.
1201 @itemx @var{file}.for
1202 @itemx @var{file}.ftn
1203 Fixed form Fortran source code that should not be preprocessed.
1206 @itemx @var{file}.FOR
1207 @itemx @var{file}.fpp
1208 @itemx @var{file}.FPP
1209 @itemx @var{file}.FTN
1210 Fixed form Fortran source code that must be preprocessed (with the traditional
1213 @item @var{file}.f90
1214 @itemx @var{file}.f95
1215 @itemx @var{file}.f03
1216 @itemx @var{file}.f08
1217 Free form Fortran source code that should not be preprocessed.
1219 @item @var{file}.F90
1220 @itemx @var{file}.F95
1221 @itemx @var{file}.F03
1222 @itemx @var{file}.F08
1223 Free form Fortran source code that must be preprocessed (with the
1224 traditional preprocessor).
1229 @c FIXME: Descriptions of Java file types.
1235 @item @var{file}.ads
1236 Ada source code file that contains a library unit declaration (a
1237 declaration of a package, subprogram, or generic, or a generic
1238 instantiation), or a library unit renaming declaration (a package,
1239 generic, or subprogram renaming declaration). Such files are also
1242 @item @var{file}.adb
1243 Ada source code file containing a library unit body (a subprogram or
1244 package body). Such files are also called @dfn{bodies}.
1246 @c GCC also knows about some suffixes for languages not yet included:
1257 @itemx @var{file}.sx
1258 Assembler code that must be preprocessed.
1261 An object file to be fed straight into linking.
1262 Any file name with no recognized suffix is treated this way.
1266 You can specify the input language explicitly with the @option{-x} option:
1269 @item -x @var{language}
1270 Specify explicitly the @var{language} for the following input files
1271 (rather than letting the compiler choose a default based on the file
1272 name suffix). This option applies to all following input files until
1273 the next @option{-x} option. Possible values for @var{language} are:
1275 c c-header cpp-output
1276 c++ c++-header c++-cpp-output
1277 objective-c objective-c-header objective-c-cpp-output
1278 objective-c++ objective-c++-header objective-c++-cpp-output
1279 assembler assembler-with-cpp
1281 f77 f77-cpp-input f95 f95-cpp-input
1287 Turn off any specification of a language, so that subsequent files are
1288 handled according to their file name suffixes (as they are if @option{-x}
1289 has not been used at all).
1291 @item -pass-exit-codes
1292 @opindex pass-exit-codes
1293 Normally the @command{gcc} program exits with the code of 1 if any
1294 phase of the compiler returns a non-success return code. If you specify
1295 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1296 the numerically highest error produced by any phase returning an error
1297 indication. The C, C++, and Fortran front ends return 4 if an internal
1298 compiler error is encountered.
1301 If you only want some of the stages of compilation, you can use
1302 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1303 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1304 @command{gcc} is to stop. Note that some combinations (for example,
1305 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1310 Compile or assemble the source files, but do not link. The linking
1311 stage simply is not done. The ultimate output is in the form of an
1312 object file for each source file.
1314 By default, the object file name for a source file is made by replacing
1315 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1317 Unrecognized input files, not requiring compilation or assembly, are
1322 Stop after the stage of compilation proper; do not assemble. The output
1323 is in the form of an assembler code file for each non-assembler input
1326 By default, the assembler file name for a source file is made by
1327 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1329 Input files that don't require compilation are ignored.
1333 Stop after the preprocessing stage; do not run the compiler proper. The
1334 output is in the form of preprocessed source code, which is sent to the
1337 Input files that don't require preprocessing are ignored.
1339 @cindex output file option
1342 Place output in file @var{file}. This applies to whatever
1343 sort of output is being produced, whether it be an executable file,
1344 an object file, an assembler file or preprocessed C code.
1346 If @option{-o} is not specified, the default is to put an executable
1347 file in @file{a.out}, the object file for
1348 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1349 assembler file in @file{@var{source}.s}, a precompiled header file in
1350 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1355 Print (on standard error output) the commands executed to run the stages
1356 of compilation. Also print the version number of the compiler driver
1357 program and of the preprocessor and the compiler proper.
1361 Like @option{-v} except the commands are not executed and arguments
1362 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1363 This is useful for shell scripts to capture the driver-generated command lines.
1367 Use pipes rather than temporary files for communication between the
1368 various stages of compilation. This fails to work on some systems where
1369 the assembler is unable to read from a pipe; but the GNU assembler has
1374 Print (on the standard output) a description of the command-line options
1375 understood by @command{gcc}. If the @option{-v} option is also specified
1376 then @option{--help} is also passed on to the various processes
1377 invoked by @command{gcc}, so that they can display the command-line options
1378 they accept. If the @option{-Wextra} option has also been specified
1379 (prior to the @option{--help} option), then command-line options that
1380 have no documentation associated with them are also displayed.
1383 @opindex target-help
1384 Print (on the standard output) a description of target-specific command-line
1385 options for each tool. For some targets extra target-specific
1386 information may also be printed.
1388 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1389 Print (on the standard output) a description of the command-line
1390 options understood by the compiler that fit into all specified classes
1391 and qualifiers. These are the supported classes:
1394 @item @samp{optimizers}
1395 Display all of the optimization options supported by the
1398 @item @samp{warnings}
1399 Display all of the options controlling warning messages
1400 produced by the compiler.
1403 Display target-specific options. Unlike the
1404 @option{--target-help} option however, target-specific options of the
1405 linker and assembler are not displayed. This is because those
1406 tools do not currently support the extended @option{--help=} syntax.
1409 Display the values recognized by the @option{--param}
1412 @item @var{language}
1413 Display the options supported for @var{language}, where
1414 @var{language} is the name of one of the languages supported in this
1418 Display the options that are common to all languages.
1421 These are the supported qualifiers:
1424 @item @samp{undocumented}
1425 Display only those options that are undocumented.
1428 Display options taking an argument that appears after an equal
1429 sign in the same continuous piece of text, such as:
1430 @samp{--help=target}.
1432 @item @samp{separate}
1433 Display options taking an argument that appears as a separate word
1434 following the original option, such as: @samp{-o output-file}.
1437 Thus for example to display all the undocumented target-specific
1438 switches supported by the compiler, use:
1441 --help=target,undocumented
1444 The sense of a qualifier can be inverted by prefixing it with the
1445 @samp{^} character, so for example to display all binary warning
1446 options (i.e., ones that are either on or off and that do not take an
1447 argument) that have a description, use:
1450 --help=warnings,^joined,^undocumented
1453 The argument to @option{--help=} should not consist solely of inverted
1456 Combining several classes is possible, although this usually
1457 restricts the output so much that there is nothing to display. One
1458 case where it does work, however, is when one of the classes is
1459 @var{target}. For example, to display all the target-specific
1460 optimization options, use:
1463 --help=target,optimizers
1466 The @option{--help=} option can be repeated on the command line. Each
1467 successive use displays its requested class of options, skipping
1468 those that have already been displayed.
1470 If the @option{-Q} option appears on the command line before the
1471 @option{--help=} option, then the descriptive text displayed by
1472 @option{--help=} is changed. Instead of describing the displayed
1473 options, an indication is given as to whether the option is enabled,
1474 disabled or set to a specific value (assuming that the compiler
1475 knows this at the point where the @option{--help=} option is used).
1477 Here is a truncated example from the ARM port of @command{gcc}:
1480 % gcc -Q -mabi=2 --help=target -c
1481 The following options are target specific:
1483 -mabort-on-noreturn [disabled]
1487 The output is sensitive to the effects of previous command-line
1488 options, so for example it is possible to find out which optimizations
1489 are enabled at @option{-O2} by using:
1492 -Q -O2 --help=optimizers
1495 Alternatively you can discover which binary optimizations are enabled
1496 by @option{-O3} by using:
1499 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1500 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1501 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1504 @item -no-canonical-prefixes
1505 @opindex no-canonical-prefixes
1506 Do not expand any symbolic links, resolve references to @samp{/../}
1507 or @samp{/./}, or make the path absolute when generating a relative
1512 Display the version number and copyrights of the invoked GCC@.
1516 Invoke all subcommands under a wrapper program. The name of the
1517 wrapper program and its parameters are passed as a comma separated
1521 gcc -c t.c -wrapper gdb,--args
1525 This invokes all subprograms of @command{gcc} under
1526 @samp{gdb --args}, thus the invocation of @command{cc1} is
1527 @samp{gdb --args cc1 @dots{}}.
1529 @item -fplugin=@var{name}.so
1531 Load the plugin code in file @var{name}.so, assumed to be a
1532 shared object to be dlopen'd by the compiler. The base name of
1533 the shared object file is used to identify the plugin for the
1534 purposes of argument parsing (See
1535 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1536 Each plugin should define the callback functions specified in the
1539 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1540 @opindex fplugin-arg
1541 Define an argument called @var{key} with a value of @var{value}
1542 for the plugin called @var{name}.
1544 @item -fdump-ada-spec@r{[}-slim@r{]}
1545 @opindex fdump-ada-spec
1546 For C and C++ source and include files, generate corresponding Ada specs.
1547 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1548 GNAT User's Guide}, which provides detailed documentation on this feature.
1550 @item -fada-spec-parent=@var{unit}
1551 @opindex fada-spec-parent
1552 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1553 Ada specs as child units of parent @var{unit}.
1555 @item -fdump-go-spec=@var{file}
1556 @opindex fdump-go-spec
1557 For input files in any language, generate corresponding Go
1558 declarations in @var{file}. This generates Go @code{const},
1559 @code{type}, @code{var}, and @code{func} declarations which may be a
1560 useful way to start writing a Go interface to code written in some
1563 @include @value{srcdir}/../libiberty/at-file.texi
1567 @section Compiling C++ Programs
1569 @cindex suffixes for C++ source
1570 @cindex C++ source file suffixes
1571 C++ source files conventionally use one of the suffixes @samp{.C},
1572 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1573 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1574 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1575 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1576 files with these names and compiles them as C++ programs even if you
1577 call the compiler the same way as for compiling C programs (usually
1578 with the name @command{gcc}).
1582 However, the use of @command{gcc} does not add the C++ library.
1583 @command{g++} is a program that calls GCC and automatically specifies linking
1584 against the C++ library. It treats @samp{.c},
1585 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1586 files unless @option{-x} is used. This program is also useful when
1587 precompiling a C header file with a @samp{.h} extension for use in C++
1588 compilations. On many systems, @command{g++} is also installed with
1589 the name @command{c++}.
1591 @cindex invoking @command{g++}
1592 When you compile C++ programs, you may specify many of the same
1593 command-line options that you use for compiling programs in any
1594 language; or command-line options meaningful for C and related
1595 languages; or options that are meaningful only for C++ programs.
1596 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1597 explanations of options for languages related to C@.
1598 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1599 explanations of options that are meaningful only for C++ programs.
1601 @node C Dialect Options
1602 @section Options Controlling C Dialect
1603 @cindex dialect options
1604 @cindex language dialect options
1605 @cindex options, dialect
1607 The following options control the dialect of C (or languages derived
1608 from C, such as C++, Objective-C and Objective-C++) that the compiler
1612 @cindex ANSI support
1616 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1617 equivalent to @option{-std=c++98}.
1619 This turns off certain features of GCC that are incompatible with ISO
1620 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1621 such as the @code{asm} and @code{typeof} keywords, and
1622 predefined macros such as @code{unix} and @code{vax} that identify the
1623 type of system you are using. It also enables the undesirable and
1624 rarely used ISO trigraph feature. For the C compiler,
1625 it disables recognition of C++ style @samp{//} comments as well as
1626 the @code{inline} keyword.
1628 The alternate keywords @code{__asm__}, @code{__extension__},
1629 @code{__inline__} and @code{__typeof__} continue to work despite
1630 @option{-ansi}. You would not want to use them in an ISO C program, of
1631 course, but it is useful to put them in header files that might be included
1632 in compilations done with @option{-ansi}. Alternate predefined macros
1633 such as @code{__unix__} and @code{__vax__} are also available, with or
1634 without @option{-ansi}.
1636 The @option{-ansi} option does not cause non-ISO programs to be
1637 rejected gratuitously. For that, @option{-Wpedantic} is required in
1638 addition to @option{-ansi}. @xref{Warning Options}.
1640 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1641 option is used. Some header files may notice this macro and refrain
1642 from declaring certain functions or defining certain macros that the
1643 ISO standard doesn't call for; this is to avoid interfering with any
1644 programs that might use these names for other things.
1646 Functions that are normally built in but do not have semantics
1647 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1648 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1649 built-in functions provided by GCC}, for details of the functions
1654 Determine the language standard. @xref{Standards,,Language Standards
1655 Supported by GCC}, for details of these standard versions. This option
1656 is currently only supported when compiling C or C++.
1658 The compiler can accept several base standards, such as @samp{c90} or
1659 @samp{c++98}, and GNU dialects of those standards, such as
1660 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1661 compiler accepts all programs following that standard plus those
1662 using GNU extensions that do not contradict it. For example,
1663 @option{-std=c90} turns off certain features of GCC that are
1664 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1665 keywords, but not other GNU extensions that do not have a meaning in
1666 ISO C90, such as omitting the middle term of a @code{?:}
1667 expression. On the other hand, when a GNU dialect of a standard is
1668 specified, all features supported by the compiler are enabled, even when
1669 those features change the meaning of the base standard. As a result, some
1670 strict-conforming programs may be rejected. The particular standard
1671 is used by @option{-Wpedantic} to identify which features are GNU
1672 extensions given that version of the standard. For example
1673 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1674 comments, while @option{-std=gnu99 -Wpedantic} does not.
1676 A value for this option must be provided; possible values are
1682 Support all ISO C90 programs (certain GNU extensions that conflict
1683 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1685 @item iso9899:199409
1686 ISO C90 as modified in amendment 1.
1692 ISO C99. This standard is substantially completely supported, modulo
1693 bugs and floating-point issues
1694 (mainly but not entirely relating to optional C99 features from
1695 Annexes F and G). See
1696 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1697 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1702 ISO C11, the 2011 revision of the ISO C standard. This standard is
1703 substantially completely supported, modulo bugs, floating-point issues
1704 (mainly but not entirely relating to optional C11 features from
1705 Annexes F and G) and the optional Annexes K (Bounds-checking
1706 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1710 GNU dialect of ISO C90 (including some C99 features).
1714 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1718 GNU dialect of ISO C11. This is the default for C code.
1719 The name @samp{gnu1x} is deprecated.
1723 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1724 additional defect reports. Same as @option{-ansi} for C++ code.
1728 GNU dialect of @option{-std=c++98}. This is the default for
1733 The 2011 ISO C++ standard plus amendments.
1734 The name @samp{c++0x} is deprecated.
1738 GNU dialect of @option{-std=c++11}.
1739 The name @samp{gnu++0x} is deprecated.
1743 The 2014 ISO C++ standard plus amendments.
1744 The name @samp{c++1y} is deprecated.
1748 GNU dialect of @option{-std=c++14}.
1749 The name @samp{gnu++1y} is deprecated.
1752 The next revision of the ISO C++ standard, tentatively planned for
1753 2017. Support is highly experimental, and will almost certainly
1754 change in incompatible ways in future releases.
1757 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1758 and will almost certainly change in incompatible ways in future
1762 @item -fgnu89-inline
1763 @opindex fgnu89-inline
1764 The option @option{-fgnu89-inline} tells GCC to use the traditional
1765 GNU semantics for @code{inline} functions when in C99 mode.
1766 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1767 is accepted and ignored by GCC versions 4.1.3 up to but not including
1768 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1769 C99 mode. Using this option is roughly equivalent to adding the
1770 @code{gnu_inline} function attribute to all inline functions
1771 (@pxref{Function Attributes}).
1773 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1774 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1775 specifies the default behavior). This option was first supported in
1776 GCC 4.3. This option is not supported in @option{-std=c90} or
1777 @option{-std=gnu90} mode.
1779 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1780 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1781 in effect for @code{inline} functions. @xref{Common Predefined
1782 Macros,,,cpp,The C Preprocessor}.
1784 @item -aux-info @var{filename}
1786 Output to the given filename prototyped declarations for all functions
1787 declared and/or defined in a translation unit, including those in header
1788 files. This option is silently ignored in any language other than C@.
1790 Besides declarations, the file indicates, in comments, the origin of
1791 each declaration (source file and line), whether the declaration was
1792 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1793 @samp{O} for old, respectively, in the first character after the line
1794 number and the colon), and whether it came from a declaration or a
1795 definition (@samp{C} or @samp{F}, respectively, in the following
1796 character). In the case of function definitions, a K&R-style list of
1797 arguments followed by their declarations is also provided, inside
1798 comments, after the declaration.
1800 @item -fallow-parameterless-variadic-functions
1801 @opindex fallow-parameterless-variadic-functions
1802 Accept variadic functions without named parameters.
1804 Although it is possible to define such a function, this is not very
1805 useful as it is not possible to read the arguments. This is only
1806 supported for C as this construct is allowed by C++.
1810 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1811 keyword, so that code can use these words as identifiers. You can use
1812 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1813 instead. @option{-ansi} implies @option{-fno-asm}.
1815 In C++, this switch only affects the @code{typeof} keyword, since
1816 @code{asm} and @code{inline} are standard keywords. You may want to
1817 use the @option{-fno-gnu-keywords} flag instead, which has the same
1818 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1819 switch only affects the @code{asm} and @code{typeof} keywords, since
1820 @code{inline} is a standard keyword in ISO C99.
1823 @itemx -fno-builtin-@var{function}
1824 @opindex fno-builtin
1825 @cindex built-in functions
1826 Don't recognize built-in functions that do not begin with
1827 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1828 functions provided by GCC}, for details of the functions affected,
1829 including those which are not built-in functions when @option{-ansi} or
1830 @option{-std} options for strict ISO C conformance are used because they
1831 do not have an ISO standard meaning.
1833 GCC normally generates special code to handle certain built-in functions
1834 more efficiently; for instance, calls to @code{alloca} may become single
1835 instructions which adjust the stack directly, and calls to @code{memcpy}
1836 may become inline copy loops. The resulting code is often both smaller
1837 and faster, but since the function calls no longer appear as such, you
1838 cannot set a breakpoint on those calls, nor can you change the behavior
1839 of the functions by linking with a different library. In addition,
1840 when a function is recognized as a built-in function, GCC may use
1841 information about that function to warn about problems with calls to
1842 that function, or to generate more efficient code, even if the
1843 resulting code still contains calls to that function. For example,
1844 warnings are given with @option{-Wformat} for bad calls to
1845 @code{printf} when @code{printf} is built in and @code{strlen} is
1846 known not to modify global memory.
1848 With the @option{-fno-builtin-@var{function}} option
1849 only the built-in function @var{function} is
1850 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1851 function is named that is not built-in in this version of GCC, this
1852 option is ignored. There is no corresponding
1853 @option{-fbuiltin-@var{function}} option; if you wish to enable
1854 built-in functions selectively when using @option{-fno-builtin} or
1855 @option{-ffreestanding}, you may define macros such as:
1858 #define abs(n) __builtin_abs ((n))
1859 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1864 @cindex hosted environment
1866 Assert that compilation targets a hosted environment. This implies
1867 @option{-fbuiltin}. A hosted environment is one in which the
1868 entire standard library is available, and in which @code{main} has a return
1869 type of @code{int}. Examples are nearly everything except a kernel.
1870 This is equivalent to @option{-fno-freestanding}.
1872 @item -ffreestanding
1873 @opindex ffreestanding
1874 @cindex hosted environment
1876 Assert that compilation targets a freestanding environment. This
1877 implies @option{-fno-builtin}. A freestanding environment
1878 is one in which the standard library may not exist, and program startup may
1879 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1880 This is equivalent to @option{-fno-hosted}.
1882 @xref{Standards,,Language Standards Supported by GCC}, for details of
1883 freestanding and hosted environments.
1887 @cindex OpenACC accelerator programming
1888 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1889 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1890 compiler generates accelerated code according to the OpenACC Application
1891 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1892 implies @option{-pthread}, and thus is only supported on targets that
1893 have support for @option{-pthread}.
1895 Note that this is an experimental feature, incomplete, and subject to
1896 change in future versions of GCC. See
1897 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1901 @cindex OpenMP parallel
1902 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1903 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1904 compiler generates parallel code according to the OpenMP Application
1905 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1906 implies @option{-pthread}, and thus is only supported on targets that
1907 have support for @option{-pthread}. @option{-fopenmp} implies
1908 @option{-fopenmp-simd}.
1911 @opindex fopenmp-simd
1914 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1915 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1920 @cindex Enable Cilk Plus
1921 Enable the usage of Cilk Plus language extension features for C/C++.
1922 When the option @option{-fcilkplus} is specified, enable the usage of
1923 the Cilk Plus Language extension features for C/C++. The present
1924 implementation follows ABI version 1.2. This is an experimental
1925 feature that is only partially complete, and whose interface may
1926 change in future versions of GCC as the official specification
1927 changes. Currently, all features but @code{_Cilk_for} have been
1932 When the option @option{-fgnu-tm} is specified, the compiler
1933 generates code for the Linux variant of Intel's current Transactional
1934 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1935 an experimental feature whose interface may change in future versions
1936 of GCC, as the official specification changes. Please note that not
1937 all architectures are supported for this feature.
1939 For more information on GCC's support for transactional memory,
1940 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1941 Transactional Memory Library}.
1943 Note that the transactional memory feature is not supported with
1944 non-call exceptions (@option{-fnon-call-exceptions}).
1946 @item -fms-extensions
1947 @opindex fms-extensions
1948 Accept some non-standard constructs used in Microsoft header files.
1950 In C++ code, this allows member names in structures to be similar
1951 to previous types declarations.
1960 Some cases of unnamed fields in structures and unions are only
1961 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1962 fields within structs/unions}, for details.
1964 Note that this option is off for all targets but x86
1965 targets using ms-abi.
1967 @item -fplan9-extensions
1968 @opindex fplan9-extensions
1969 Accept some non-standard constructs used in Plan 9 code.
1971 This enables @option{-fms-extensions}, permits passing pointers to
1972 structures with anonymous fields to functions that expect pointers to
1973 elements of the type of the field, and permits referring to anonymous
1974 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1975 struct/union fields within structs/unions}, for details. This is only
1976 supported for C, not C++.
1980 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1981 options for strict ISO C conformance) implies @option{-trigraphs}.
1983 @cindex traditional C language
1984 @cindex C language, traditional
1986 @itemx -traditional-cpp
1987 @opindex traditional-cpp
1988 @opindex traditional
1989 Formerly, these options caused GCC to attempt to emulate a pre-standard
1990 C compiler. They are now only supported with the @option{-E} switch.
1991 The preprocessor continues to support a pre-standard mode. See the GNU
1992 CPP manual for details.
1994 @item -fcond-mismatch
1995 @opindex fcond-mismatch
1996 Allow conditional expressions with mismatched types in the second and
1997 third arguments. The value of such an expression is void. This option
1998 is not supported for C++.
2000 @item -flax-vector-conversions
2001 @opindex flax-vector-conversions
2002 Allow implicit conversions between vectors with differing numbers of
2003 elements and/or incompatible element types. This option should not be
2006 @item -funsigned-char
2007 @opindex funsigned-char
2008 Let the type @code{char} be unsigned, like @code{unsigned char}.
2010 Each kind of machine has a default for what @code{char} should
2011 be. It is either like @code{unsigned char} by default or like
2012 @code{signed char} by default.
2014 Ideally, a portable program should always use @code{signed char} or
2015 @code{unsigned char} when it depends on the signedness of an object.
2016 But many programs have been written to use plain @code{char} and
2017 expect it to be signed, or expect it to be unsigned, depending on the
2018 machines they were written for. This option, and its inverse, let you
2019 make such a program work with the opposite default.
2021 The type @code{char} is always a distinct type from each of
2022 @code{signed char} or @code{unsigned char}, even though its behavior
2023 is always just like one of those two.
2026 @opindex fsigned-char
2027 Let the type @code{char} be signed, like @code{signed char}.
2029 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2030 the negative form of @option{-funsigned-char}. Likewise, the option
2031 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2033 @item -fsigned-bitfields
2034 @itemx -funsigned-bitfields
2035 @itemx -fno-signed-bitfields
2036 @itemx -fno-unsigned-bitfields
2037 @opindex fsigned-bitfields
2038 @opindex funsigned-bitfields
2039 @opindex fno-signed-bitfields
2040 @opindex fno-unsigned-bitfields
2041 These options control whether a bit-field is signed or unsigned, when the
2042 declaration does not use either @code{signed} or @code{unsigned}. By
2043 default, such a bit-field is signed, because this is consistent: the
2044 basic integer types such as @code{int} are signed types.
2047 @node C++ Dialect Options
2048 @section Options Controlling C++ Dialect
2050 @cindex compiler options, C++
2051 @cindex C++ options, command-line
2052 @cindex options, C++
2053 This section describes the command-line options that are only meaningful
2054 for C++ programs. You can also use most of the GNU compiler options
2055 regardless of what language your program is in. For example, you
2056 might compile a file @file{firstClass.C} like this:
2059 g++ -g -frepo -O -c firstClass.C
2063 In this example, only @option{-frepo} is an option meant
2064 only for C++ programs; you can use the other options with any
2065 language supported by GCC@.
2067 Here is a list of options that are @emph{only} for compiling C++ programs:
2071 @item -fabi-version=@var{n}
2072 @opindex fabi-version
2073 Use version @var{n} of the C++ ABI@. The default is version 0.
2075 Version 0 refers to the version conforming most closely to
2076 the C++ ABI specification. Therefore, the ABI obtained using version 0
2077 will change in different versions of G++ as ABI bugs are fixed.
2079 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2081 Version 2 is the version of the C++ ABI that first appeared in G++
2082 3.4, and was the default through G++ 4.9.
2084 Version 3 corrects an error in mangling a constant address as a
2087 Version 4, which first appeared in G++ 4.5, implements a standard
2088 mangling for vector types.
2090 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2091 attribute const/volatile on function pointer types, decltype of a
2092 plain decl, and use of a function parameter in the declaration of
2095 Version 6, which first appeared in G++ 4.7, corrects the promotion
2096 behavior of C++11 scoped enums and the mangling of template argument
2097 packs, const/static_cast, prefix ++ and --, and a class scope function
2098 used as a template argument.
2100 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2101 builtin type and corrects the mangling of lambdas in default argument
2104 Version 8, which first appeared in G++ 4.9, corrects the substitution
2105 behavior of function types with function-cv-qualifiers.
2107 See also @option{-Wabi}.
2109 @item -fabi-compat-version=@var{n}
2110 @opindex fabi-compat-version
2111 Starting with GCC 4.5, on targets that support strong aliases, G++
2112 works around mangling changes by creating an alias with the correct
2113 mangled name when defining a symbol with an incorrect mangled name.
2114 This switch specifies which ABI version to use for the alias.
2116 With @option{-fabi-version=0} (the default), this defaults to 2. If
2117 another ABI version is explicitly selected, this defaults to 0.
2119 The compatibility version is also set by @option{-Wabi=@var{n}}.
2121 @item -fno-access-control
2122 @opindex fno-access-control
2123 Turn off all access checking. This switch is mainly useful for working
2124 around bugs in the access control code.
2128 Check that the pointer returned by @code{operator new} is non-null
2129 before attempting to modify the storage allocated. This check is
2130 normally unnecessary because the C++ standard specifies that
2131 @code{operator new} only returns @code{0} if it is declared
2132 @code{throw()}, in which case the compiler always checks the
2133 return value even without this option. In all other cases, when
2134 @code{operator new} has a non-empty exception specification, memory
2135 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2136 @samp{new (nothrow)}.
2138 @item -fconstexpr-depth=@var{n}
2139 @opindex fconstexpr-depth
2140 Set the maximum nested evaluation depth for C++11 constexpr functions
2141 to @var{n}. A limit is needed to detect endless recursion during
2142 constant expression evaluation. The minimum specified by the standard
2145 @item -fdeduce-init-list
2146 @opindex fdeduce-init-list
2147 Enable deduction of a template type parameter as
2148 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2151 template <class T> auto forward(T t) -> decltype (realfn (t))
2158 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2162 This deduction was implemented as a possible extension to the
2163 originally proposed semantics for the C++11 standard, but was not part
2164 of the final standard, so it is disabled by default. This option is
2165 deprecated, and may be removed in a future version of G++.
2167 @item -ffriend-injection
2168 @opindex ffriend-injection
2169 Inject friend functions into the enclosing namespace, so that they are
2170 visible outside the scope of the class in which they are declared.
2171 Friend functions were documented to work this way in the old Annotated
2172 C++ Reference Manual, and versions of G++ before 4.1 always worked
2173 that way. However, in ISO C++ a friend function that is not declared
2174 in an enclosing scope can only be found using argument dependent
2175 lookup. This option causes friends to be injected as they were in
2178 This option is for compatibility, and may be removed in a future
2181 @item -fno-elide-constructors
2182 @opindex fno-elide-constructors
2183 The C++ standard allows an implementation to omit creating a temporary
2184 that is only used to initialize another object of the same type.
2185 Specifying this option disables that optimization, and forces G++ to
2186 call the copy constructor in all cases.
2188 @item -fno-enforce-eh-specs
2189 @opindex fno-enforce-eh-specs
2190 Don't generate code to check for violation of exception specifications
2191 at run time. This option violates the C++ standard, but may be useful
2192 for reducing code size in production builds, much like defining
2193 @code{NDEBUG}. This does not give user code permission to throw
2194 exceptions in violation of the exception specifications; the compiler
2195 still optimizes based on the specifications, so throwing an
2196 unexpected exception results in undefined behavior at run time.
2198 @item -fextern-tls-init
2199 @itemx -fno-extern-tls-init
2200 @opindex fextern-tls-init
2201 @opindex fno-extern-tls-init
2202 The C++11 and OpenMP standards allow @code{thread_local} and
2203 @code{threadprivate} variables to have dynamic (runtime)
2204 initialization. To support this, any use of such a variable goes
2205 through a wrapper function that performs any necessary initialization.
2206 When the use and definition of the variable are in the same
2207 translation unit, this overhead can be optimized away, but when the
2208 use is in a different translation unit there is significant overhead
2209 even if the variable doesn't actually need dynamic initialization. If
2210 the programmer can be sure that no use of the variable in a
2211 non-defining TU needs to trigger dynamic initialization (either
2212 because the variable is statically initialized, or a use of the
2213 variable in the defining TU will be executed before any uses in
2214 another TU), they can avoid this overhead with the
2215 @option{-fno-extern-tls-init} option.
2217 On targets that support symbol aliases, the default is
2218 @option{-fextern-tls-init}. On targets that do not support symbol
2219 aliases, the default is @option{-fno-extern-tls-init}.
2222 @itemx -fno-for-scope
2224 @opindex fno-for-scope
2225 If @option{-ffor-scope} is specified, the scope of variables declared in
2226 a @i{for-init-statement} is limited to the @code{for} loop itself,
2227 as specified by the C++ standard.
2228 If @option{-fno-for-scope} is specified, the scope of variables declared in
2229 a @i{for-init-statement} extends to the end of the enclosing scope,
2230 as was the case in old versions of G++, and other (traditional)
2231 implementations of C++.
2233 If neither flag is given, the default is to follow the standard,
2234 but to allow and give a warning for old-style code that would
2235 otherwise be invalid, or have different behavior.
2237 @item -fno-gnu-keywords
2238 @opindex fno-gnu-keywords
2239 Do not recognize @code{typeof} as a keyword, so that code can use this
2240 word as an identifier. You can use the keyword @code{__typeof__} instead.
2241 @option{-ansi} implies @option{-fno-gnu-keywords}.
2243 @item -fno-implicit-templates
2244 @opindex fno-implicit-templates
2245 Never emit code for non-inline templates that are instantiated
2246 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2247 @xref{Template Instantiation}, for more information.
2249 @item -fno-implicit-inline-templates
2250 @opindex fno-implicit-inline-templates
2251 Don't emit code for implicit instantiations of inline templates, either.
2252 The default is to handle inlines differently so that compiles with and
2253 without optimization need the same set of explicit instantiations.
2255 @item -fno-implement-inlines
2256 @opindex fno-implement-inlines
2257 To save space, do not emit out-of-line copies of inline functions
2258 controlled by @code{#pragma implementation}. This causes linker
2259 errors if these functions are not inlined everywhere they are called.
2261 @item -fms-extensions
2262 @opindex fms-extensions
2263 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2264 int and getting a pointer to member function via non-standard syntax.
2266 @item -fno-nonansi-builtins
2267 @opindex fno-nonansi-builtins
2268 Disable built-in declarations of functions that are not mandated by
2269 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2270 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2273 @opindex fnothrow-opt
2274 Treat a @code{throw()} exception specification as if it were a
2275 @code{noexcept} specification to reduce or eliminate the text size
2276 overhead relative to a function with no exception specification. If
2277 the function has local variables of types with non-trivial
2278 destructors, the exception specification actually makes the
2279 function smaller because the EH cleanups for those variables can be
2280 optimized away. The semantic effect is that an exception thrown out of
2281 a function with such an exception specification results in a call
2282 to @code{terminate} rather than @code{unexpected}.
2284 @item -fno-operator-names
2285 @opindex fno-operator-names
2286 Do not treat the operator name keywords @code{and}, @code{bitand},
2287 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2288 synonyms as keywords.
2290 @item -fno-optional-diags
2291 @opindex fno-optional-diags
2292 Disable diagnostics that the standard says a compiler does not need to
2293 issue. Currently, the only such diagnostic issued by G++ is the one for
2294 a name having multiple meanings within a class.
2297 @opindex fpermissive
2298 Downgrade some diagnostics about nonconformant code from errors to
2299 warnings. Thus, using @option{-fpermissive} allows some
2300 nonconforming code to compile.
2302 @item -fno-pretty-templates
2303 @opindex fno-pretty-templates
2304 When an error message refers to a specialization of a function
2305 template, the compiler normally prints the signature of the
2306 template followed by the template arguments and any typedefs or
2307 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2308 rather than @code{void f(int)}) so that it's clear which template is
2309 involved. When an error message refers to a specialization of a class
2310 template, the compiler omits any template arguments that match
2311 the default template arguments for that template. If either of these
2312 behaviors make it harder to understand the error message rather than
2313 easier, you can use @option{-fno-pretty-templates} to disable them.
2317 Enable automatic template instantiation at link time. This option also
2318 implies @option{-fno-implicit-templates}. @xref{Template
2319 Instantiation}, for more information.
2323 Disable generation of information about every class with virtual
2324 functions for use by the C++ run-time type identification features
2325 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2326 of the language, you can save some space by using this flag. Note that
2327 exception handling uses the same information, but G++ generates it as
2328 needed. The @code{dynamic_cast} operator can still be used for casts that
2329 do not require run-time type information, i.e.@: casts to @code{void *} or to
2330 unambiguous base classes.
2332 @item -fsized-deallocation
2333 @opindex fsized-deallocation
2334 Enable the built-in global declarations
2336 void operator delete (void *, std::size_t) noexcept;
2337 void operator delete[] (void *, std::size_t) noexcept;
2339 as introduced in C++14. This is useful for user-defined replacement
2340 deallocation functions that, for example, use the size of the object
2341 to make deallocation faster. Enabled by default under
2342 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2343 warns about places that might want to add a definition.
2347 Emit statistics about front-end processing at the end of the compilation.
2348 This information is generally only useful to the G++ development team.
2350 @item -fstrict-enums
2351 @opindex fstrict-enums
2352 Allow the compiler to optimize using the assumption that a value of
2353 enumerated type can only be one of the values of the enumeration (as
2354 defined in the C++ standard; basically, a value that can be
2355 represented in the minimum number of bits needed to represent all the
2356 enumerators). This assumption may not be valid if the program uses a
2357 cast to convert an arbitrary integer value to the enumerated type.
2359 @item -ftemplate-backtrace-limit=@var{n}
2360 @opindex ftemplate-backtrace-limit
2361 Set the maximum number of template instantiation notes for a single
2362 warning or error to @var{n}. The default value is 10.
2364 @item -ftemplate-depth=@var{n}
2365 @opindex ftemplate-depth
2366 Set the maximum instantiation depth for template classes to @var{n}.
2367 A limit on the template instantiation depth is needed to detect
2368 endless recursions during template class instantiation. ANSI/ISO C++
2369 conforming programs must not rely on a maximum depth greater than 17
2370 (changed to 1024 in C++11). The default value is 900, as the compiler
2371 can run out of stack space before hitting 1024 in some situations.
2373 @item -fno-threadsafe-statics
2374 @opindex fno-threadsafe-statics
2375 Do not emit the extra code to use the routines specified in the C++
2376 ABI for thread-safe initialization of local statics. You can use this
2377 option to reduce code size slightly in code that doesn't need to be
2380 @item -fuse-cxa-atexit
2381 @opindex fuse-cxa-atexit
2382 Register destructors for objects with static storage duration with the
2383 @code{__cxa_atexit} function rather than the @code{atexit} function.
2384 This option is required for fully standards-compliant handling of static
2385 destructors, but only works if your C library supports
2386 @code{__cxa_atexit}.
2388 @item -fno-use-cxa-get-exception-ptr
2389 @opindex fno-use-cxa-get-exception-ptr
2390 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2391 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2392 if the runtime routine is not available.
2394 @item -fvisibility-inlines-hidden
2395 @opindex fvisibility-inlines-hidden
2396 This switch declares that the user does not attempt to compare
2397 pointers to inline functions or methods where the addresses of the two functions
2398 are taken in different shared objects.
2400 The effect of this is that GCC may, effectively, mark inline methods with
2401 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2402 appear in the export table of a DSO and do not require a PLT indirection
2403 when used within the DSO@. Enabling this option can have a dramatic effect
2404 on load and link times of a DSO as it massively reduces the size of the
2405 dynamic export table when the library makes heavy use of templates.
2407 The behavior of this switch is not quite the same as marking the
2408 methods as hidden directly, because it does not affect static variables
2409 local to the function or cause the compiler to deduce that
2410 the function is defined in only one shared object.
2412 You may mark a method as having a visibility explicitly to negate the
2413 effect of the switch for that method. For example, if you do want to
2414 compare pointers to a particular inline method, you might mark it as
2415 having default visibility. Marking the enclosing class with explicit
2416 visibility has no effect.
2418 Explicitly instantiated inline methods are unaffected by this option
2419 as their linkage might otherwise cross a shared library boundary.
2420 @xref{Template Instantiation}.
2422 @item -fvisibility-ms-compat
2423 @opindex fvisibility-ms-compat
2424 This flag attempts to use visibility settings to make GCC's C++
2425 linkage model compatible with that of Microsoft Visual Studio.
2427 The flag makes these changes to GCC's linkage model:
2431 It sets the default visibility to @code{hidden}, like
2432 @option{-fvisibility=hidden}.
2435 Types, but not their members, are not hidden by default.
2438 The One Definition Rule is relaxed for types without explicit
2439 visibility specifications that are defined in more than one
2440 shared object: those declarations are permitted if they are
2441 permitted when this option is not used.
2444 In new code it is better to use @option{-fvisibility=hidden} and
2445 export those classes that are intended to be externally visible.
2446 Unfortunately it is possible for code to rely, perhaps accidentally,
2447 on the Visual Studio behavior.
2449 Among the consequences of these changes are that static data members
2450 of the same type with the same name but defined in different shared
2451 objects are different, so changing one does not change the other;
2452 and that pointers to function members defined in different shared
2453 objects may not compare equal. When this flag is given, it is a
2454 violation of the ODR to define types with the same name differently.
2456 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2457 @opindex fvtable-verify
2458 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2459 feature that verifies at run time, for every virtual call, that
2460 the vtable pointer through which the call is made is valid for the type of
2461 the object, and has not been corrupted or overwritten. If an invalid vtable
2462 pointer is detected at run time, an error is reported and execution of the
2463 program is immediately halted.
2465 This option causes run-time data structures to be built at program startup,
2466 which are used for verifying the vtable pointers.
2467 The options @samp{std} and @samp{preinit}
2468 control the timing of when these data structures are built. In both cases the
2469 data structures are built before execution reaches @code{main}. Using
2470 @option{-fvtable-verify=std} causes the data structures to be built after
2471 shared libraries have been loaded and initialized.
2472 @option{-fvtable-verify=preinit} causes them to be built before shared
2473 libraries have been loaded and initialized.
2475 If this option appears multiple times in the command line with different
2476 values specified, @samp{none} takes highest priority over both @samp{std} and
2477 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2481 When used in conjunction with @option{-fvtable-verify=std} or
2482 @option{-fvtable-verify=preinit}, causes debug versions of the
2483 runtime functions for the vtable verification feature to be called.
2484 This flag also causes the compiler to log information about which
2485 vtable pointers it finds for each class.
2486 This information is written to a file named @file{vtv_set_ptr_data.log}
2487 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2488 if that is defined or the current working directory otherwise.
2490 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2491 file, be sure to delete any existing one.
2494 @opindex fvtv-counts
2495 This is a debugging flag. When used in conjunction with
2496 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2497 causes the compiler to keep track of the total number of virtual calls
2498 it encounters and the number of verifications it inserts. It also
2499 counts the number of calls to certain run-time library functions
2500 that it inserts and logs this information for each compilation unit.
2501 The compiler writes this information to a file named
2502 @file{vtv_count_data.log} in the directory named by the environment
2503 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2504 directory otherwise. It also counts the size of the vtable pointer sets
2505 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2506 in the same directory.
2508 Note: This feature @emph{appends} data to the log files. To get fresh log
2509 files, be sure to delete any existing ones.
2513 Do not use weak symbol support, even if it is provided by the linker.
2514 By default, G++ uses weak symbols if they are available. This
2515 option exists only for testing, and should not be used by end-users;
2516 it results in inferior code and has no benefits. This option may
2517 be removed in a future release of G++.
2521 Do not search for header files in the standard directories specific to
2522 C++, but do still search the other standard directories. (This option
2523 is used when building the C++ library.)
2526 In addition, these optimization, warning, and code generation options
2527 have meanings only for C++ programs:
2530 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2533 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2534 G++ to warn when it generates code that is probably not compatible with the
2535 vendor-neutral C++ ABI@. Since G++ now defaults to
2536 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2537 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2538 or an older compatibility version is selected (with
2539 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2541 Although an effort has been made to warn about
2542 all such cases, there are probably some cases that are not warned about,
2543 even though G++ is generating incompatible code. There may also be
2544 cases where warnings are emitted even though the code that is generated
2547 You should rewrite your code to avoid these warnings if you are
2548 concerned about the fact that code generated by G++ may not be binary
2549 compatible with code generated by other compilers.
2551 @option{-Wabi} can also be used with an explicit version number to
2552 warn about compatibility with a particular @option{-fabi-version}
2553 level, e.g. @option{-Wabi=2} to warn about changes relative to
2554 @option{-fabi-version=2}. Specifying a version number also sets
2555 @option{-fabi-compat-version=@var{n}}.
2557 The known incompatibilities in @option{-fabi-version=2} (which was the
2558 default from GCC 3.4 to 4.9) include:
2563 A template with a non-type template parameter of reference type was
2564 mangled incorrectly:
2567 template <int &> struct S @{@};
2571 This was fixed in @option{-fabi-version=3}.
2574 SIMD vector types declared using @code{__attribute ((vector_size))} were
2575 mangled in a non-standard way that does not allow for overloading of
2576 functions taking vectors of different sizes.
2578 The mangling was changed in @option{-fabi-version=4}.
2581 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2582 qualifiers, and @code{decltype} of a plain declaration was folded away.
2584 These mangling issues were fixed in @option{-fabi-version=5}.
2587 Scoped enumerators passed as arguments to a variadic function are
2588 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2589 On most targets this does not actually affect the parameter passing
2590 ABI, as there is no way to pass an argument smaller than @code{int}.
2592 Also, the ABI changed the mangling of template argument packs,
2593 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2594 a class scope function used as a template argument.
2596 These issues were corrected in @option{-fabi-version=6}.
2599 Lambdas in default argument scope were mangled incorrectly, and the
2600 ABI changed the mangling of @code{nullptr_t}.
2602 These issues were corrected in @option{-fabi-version=7}.
2605 When mangling a function type with function-cv-qualifiers, the
2606 un-qualified function type was incorrectly treated as a substitution
2609 This was fixed in @option{-fabi-version=8}.
2612 It also warns about psABI-related changes. The known psABI changes at this
2618 For SysV/x86-64, unions with @code{long double} members are
2619 passed in memory as specified in psABI. For example:
2629 @code{union U} is always passed in memory.
2633 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2634 @opindex Wctor-dtor-privacy
2635 @opindex Wno-ctor-dtor-privacy
2636 Warn when a class seems unusable because all the constructors or
2637 destructors in that class are private, and it has neither friends nor
2638 public static member functions. Also warn if there are no non-private
2639 methods, and there's at least one private member function that isn't
2640 a constructor or destructor.
2642 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2643 @opindex Wdelete-non-virtual-dtor
2644 @opindex Wno-delete-non-virtual-dtor
2645 Warn when @code{delete} is used to destroy an instance of a class that
2646 has virtual functions and non-virtual destructor. It is unsafe to delete
2647 an instance of a derived class through a pointer to a base class if the
2648 base class does not have a virtual destructor. This warning is enabled
2651 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2652 @opindex Wliteral-suffix
2653 @opindex Wno-literal-suffix
2654 Warn when a string or character literal is followed by a ud-suffix which does
2655 not begin with an underscore. As a conforming extension, GCC treats such
2656 suffixes as separate preprocessing tokens in order to maintain backwards
2657 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2661 #define __STDC_FORMAT_MACROS
2662 #include <inttypes.h>
2667 printf("My int64: %"PRId64"\n", i64);
2671 In this case, @code{PRId64} is treated as a separate preprocessing token.
2673 This warning is enabled by default.
2675 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2677 @opindex Wno-narrowing
2678 Warn when a narrowing conversion prohibited by C++11 occurs within
2682 int i = @{ 2.2 @}; // error: narrowing from double to int
2685 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2687 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses for
2688 non-constants the diagnostic required by the standard. Note that this
2689 does not affect the meaning of well-formed code; narrowing conversions
2690 are still considered ill-formed in SFINAE context.
2692 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2694 @opindex Wno-noexcept
2695 Warn when a noexcept-expression evaluates to false because of a call
2696 to a function that does not have a non-throwing exception
2697 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2698 the compiler to never throw an exception.
2700 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2701 @opindex Wnon-virtual-dtor
2702 @opindex Wno-non-virtual-dtor
2703 Warn when a class has virtual functions and an accessible non-virtual
2704 destructor itself or in an accessible polymorphic base class, in which
2705 case it is possible but unsafe to delete an instance of a derived
2706 class through a pointer to the class itself or base class. This
2707 warning is automatically enabled if @option{-Weffc++} is specified.
2709 @item -Wreorder @r{(C++ and Objective-C++ only)}
2711 @opindex Wno-reorder
2712 @cindex reordering, warning
2713 @cindex warning for reordering of member initializers
2714 Warn when the order of member initializers given in the code does not
2715 match the order in which they must be executed. For instance:
2721 A(): j (0), i (1) @{ @}
2726 The compiler rearranges the member initializers for @code{i}
2727 and @code{j} to match the declaration order of the members, emitting
2728 a warning to that effect. This warning is enabled by @option{-Wall}.
2730 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2731 @opindex fext-numeric-literals
2732 @opindex fno-ext-numeric-literals
2733 Accept imaginary, fixed-point, or machine-defined
2734 literal number suffixes as GNU extensions.
2735 When this option is turned off these suffixes are treated
2736 as C++11 user-defined literal numeric suffixes.
2737 This is on by default for all pre-C++11 dialects and all GNU dialects:
2738 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2739 @option{-std=gnu++14}.
2740 This option is off by default
2741 for ISO C++11 onwards (@option{-std=c++11}, ...).
2744 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2747 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2750 Warn about violations of the following style guidelines from Scott Meyers'
2751 @cite{Effective C++} series of books:
2755 Define a copy constructor and an assignment operator for classes
2756 with dynamically-allocated memory.
2759 Prefer initialization to assignment in constructors.
2762 Have @code{operator=} return a reference to @code{*this}.
2765 Don't try to return a reference when you must return an object.
2768 Distinguish between prefix and postfix forms of increment and
2769 decrement operators.
2772 Never overload @code{&&}, @code{||}, or @code{,}.
2776 This option also enables @option{-Wnon-virtual-dtor}, which is also
2777 one of the effective C++ recommendations. However, the check is
2778 extended to warn about the lack of virtual destructor in accessible
2779 non-polymorphic bases classes too.
2781 When selecting this option, be aware that the standard library
2782 headers do not obey all of these guidelines; use @samp{grep -v}
2783 to filter out those warnings.
2785 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2786 @opindex Wstrict-null-sentinel
2787 @opindex Wno-strict-null-sentinel
2788 Warn about the use of an uncasted @code{NULL} as sentinel. When
2789 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2790 to @code{__null}. Although it is a null pointer constant rather than a
2791 null pointer, it is guaranteed to be of the same size as a pointer.
2792 But this use is not portable across different compilers.
2794 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2795 @opindex Wno-non-template-friend
2796 @opindex Wnon-template-friend
2797 Disable warnings when non-templatized friend functions are declared
2798 within a template. Since the advent of explicit template specification
2799 support in G++, if the name of the friend is an unqualified-id (i.e.,
2800 @samp{friend foo(int)}), the C++ language specification demands that the
2801 friend declare or define an ordinary, nontemplate function. (Section
2802 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2803 could be interpreted as a particular specialization of a templatized
2804 function. Because this non-conforming behavior is no longer the default
2805 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2806 check existing code for potential trouble spots and is on by default.
2807 This new compiler behavior can be turned off with
2808 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2809 but disables the helpful warning.
2811 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2812 @opindex Wold-style-cast
2813 @opindex Wno-old-style-cast
2814 Warn if an old-style (C-style) cast to a non-void type is used within
2815 a C++ program. The new-style casts (@code{dynamic_cast},
2816 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2817 less vulnerable to unintended effects and much easier to search for.
2819 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2820 @opindex Woverloaded-virtual
2821 @opindex Wno-overloaded-virtual
2822 @cindex overloaded virtual function, warning
2823 @cindex warning for overloaded virtual function
2824 Warn when a function declaration hides virtual functions from a
2825 base class. For example, in:
2832 struct B: public A @{
2837 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2848 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2849 @opindex Wno-pmf-conversions
2850 @opindex Wpmf-conversions
2851 Disable the diagnostic for converting a bound pointer to member function
2854 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2855 @opindex Wsign-promo
2856 @opindex Wno-sign-promo
2857 Warn when overload resolution chooses a promotion from unsigned or
2858 enumerated type to a signed type, over a conversion to an unsigned type of
2859 the same size. Previous versions of G++ tried to preserve
2860 unsignedness, but the standard mandates the current behavior.
2863 @node Objective-C and Objective-C++ Dialect Options
2864 @section Options Controlling Objective-C and Objective-C++ Dialects
2866 @cindex compiler options, Objective-C and Objective-C++
2867 @cindex Objective-C and Objective-C++ options, command-line
2868 @cindex options, Objective-C and Objective-C++
2869 (NOTE: This manual does not describe the Objective-C and Objective-C++
2870 languages themselves. @xref{Standards,,Language Standards
2871 Supported by GCC}, for references.)
2873 This section describes the command-line options that are only meaningful
2874 for Objective-C and Objective-C++ programs. You can also use most of
2875 the language-independent GNU compiler options.
2876 For example, you might compile a file @file{some_class.m} like this:
2879 gcc -g -fgnu-runtime -O -c some_class.m
2883 In this example, @option{-fgnu-runtime} is an option meant only for
2884 Objective-C and Objective-C++ programs; you can use the other options with
2885 any language supported by GCC@.
2887 Note that since Objective-C is an extension of the C language, Objective-C
2888 compilations may also use options specific to the C front-end (e.g.,
2889 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2890 C++-specific options (e.g., @option{-Wabi}).
2892 Here is a list of options that are @emph{only} for compiling Objective-C
2893 and Objective-C++ programs:
2896 @item -fconstant-string-class=@var{class-name}
2897 @opindex fconstant-string-class
2898 Use @var{class-name} as the name of the class to instantiate for each
2899 literal string specified with the syntax @code{@@"@dots{}"}. The default
2900 class name is @code{NXConstantString} if the GNU runtime is being used, and
2901 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2902 @option{-fconstant-cfstrings} option, if also present, overrides the
2903 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2904 to be laid out as constant CoreFoundation strings.
2907 @opindex fgnu-runtime
2908 Generate object code compatible with the standard GNU Objective-C
2909 runtime. This is the default for most types of systems.
2911 @item -fnext-runtime
2912 @opindex fnext-runtime
2913 Generate output compatible with the NeXT runtime. This is the default
2914 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2915 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2918 @item -fno-nil-receivers
2919 @opindex fno-nil-receivers
2920 Assume that all Objective-C message dispatches (@code{[receiver
2921 message:arg]}) in this translation unit ensure that the receiver is
2922 not @code{nil}. This allows for more efficient entry points in the
2923 runtime to be used. This option is only available in conjunction with
2924 the NeXT runtime and ABI version 0 or 1.
2926 @item -fobjc-abi-version=@var{n}
2927 @opindex fobjc-abi-version
2928 Use version @var{n} of the Objective-C ABI for the selected runtime.
2929 This option is currently supported only for the NeXT runtime. In that
2930 case, Version 0 is the traditional (32-bit) ABI without support for
2931 properties and other Objective-C 2.0 additions. Version 1 is the
2932 traditional (32-bit) ABI with support for properties and other
2933 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2934 nothing is specified, the default is Version 0 on 32-bit target
2935 machines, and Version 2 on 64-bit target machines.
2937 @item -fobjc-call-cxx-cdtors
2938 @opindex fobjc-call-cxx-cdtors
2939 For each Objective-C class, check if any of its instance variables is a
2940 C++ object with a non-trivial default constructor. If so, synthesize a
2941 special @code{- (id) .cxx_construct} instance method which runs
2942 non-trivial default constructors on any such instance variables, in order,
2943 and then return @code{self}. Similarly, check if any instance variable
2944 is a C++ object with a non-trivial destructor, and if so, synthesize a
2945 special @code{- (void) .cxx_destruct} method which runs
2946 all such default destructors, in reverse order.
2948 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2949 methods thusly generated only operate on instance variables
2950 declared in the current Objective-C class, and not those inherited
2951 from superclasses. It is the responsibility of the Objective-C
2952 runtime to invoke all such methods in an object's inheritance
2953 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2954 by the runtime immediately after a new object instance is allocated;
2955 the @code{- (void) .cxx_destruct} methods are invoked immediately
2956 before the runtime deallocates an object instance.
2958 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2959 support for invoking the @code{- (id) .cxx_construct} and
2960 @code{- (void) .cxx_destruct} methods.
2962 @item -fobjc-direct-dispatch
2963 @opindex fobjc-direct-dispatch
2964 Allow fast jumps to the message dispatcher. On Darwin this is
2965 accomplished via the comm page.
2967 @item -fobjc-exceptions
2968 @opindex fobjc-exceptions
2969 Enable syntactic support for structured exception handling in
2970 Objective-C, similar to what is offered by C++ and Java. This option
2971 is required to use the Objective-C keywords @code{@@try},
2972 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2973 @code{@@synchronized}. This option is available with both the GNU
2974 runtime and the NeXT runtime (but not available in conjunction with
2975 the NeXT runtime on Mac OS X 10.2 and earlier).
2979 Enable garbage collection (GC) in Objective-C and Objective-C++
2980 programs. This option is only available with the NeXT runtime; the
2981 GNU runtime has a different garbage collection implementation that
2982 does not require special compiler flags.
2984 @item -fobjc-nilcheck
2985 @opindex fobjc-nilcheck
2986 For the NeXT runtime with version 2 of the ABI, check for a nil
2987 receiver in method invocations before doing the actual method call.
2988 This is the default and can be disabled using
2989 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2990 checked for nil in this way no matter what this flag is set to.
2991 Currently this flag does nothing when the GNU runtime, or an older
2992 version of the NeXT runtime ABI, is used.
2994 @item -fobjc-std=objc1
2996 Conform to the language syntax of Objective-C 1.0, the language
2997 recognized by GCC 4.0. This only affects the Objective-C additions to
2998 the C/C++ language; it does not affect conformance to C/C++ standards,
2999 which is controlled by the separate C/C++ dialect option flags. When
3000 this option is used with the Objective-C or Objective-C++ compiler,
3001 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3002 This is useful if you need to make sure that your Objective-C code can
3003 be compiled with older versions of GCC@.
3005 @item -freplace-objc-classes
3006 @opindex freplace-objc-classes
3007 Emit a special marker instructing @command{ld(1)} not to statically link in
3008 the resulting object file, and allow @command{dyld(1)} to load it in at
3009 run time instead. This is used in conjunction with the Fix-and-Continue
3010 debugging mode, where the object file in question may be recompiled and
3011 dynamically reloaded in the course of program execution, without the need
3012 to restart the program itself. Currently, Fix-and-Continue functionality
3013 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3018 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3019 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3020 compile time) with static class references that get initialized at load time,
3021 which improves run-time performance. Specifying the @option{-fzero-link} flag
3022 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3023 to be retained. This is useful in Zero-Link debugging mode, since it allows
3024 for individual class implementations to be modified during program execution.
3025 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3026 regardless of command-line options.
3028 @item -fno-local-ivars
3029 @opindex fno-local-ivars
3030 @opindex flocal-ivars
3031 By default instance variables in Objective-C can be accessed as if
3032 they were local variables from within the methods of the class they're
3033 declared in. This can lead to shadowing between instance variables
3034 and other variables declared either locally inside a class method or
3035 globally with the same name. Specifying the @option{-fno-local-ivars}
3036 flag disables this behavior thus avoiding variable shadowing issues.
3038 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3039 @opindex fivar-visibility
3040 Set the default instance variable visibility to the specified option
3041 so that instance variables declared outside the scope of any access
3042 modifier directives default to the specified visibility.
3046 Dump interface declarations for all classes seen in the source file to a
3047 file named @file{@var{sourcename}.decl}.
3049 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3050 @opindex Wassign-intercept
3051 @opindex Wno-assign-intercept
3052 Warn whenever an Objective-C assignment is being intercepted by the
3055 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3056 @opindex Wno-protocol
3058 If a class is declared to implement a protocol, a warning is issued for
3059 every method in the protocol that is not implemented by the class. The
3060 default behavior is to issue a warning for every method not explicitly
3061 implemented in the class, even if a method implementation is inherited
3062 from the superclass. If you use the @option{-Wno-protocol} option, then
3063 methods inherited from the superclass are considered to be implemented,
3064 and no warning is issued for them.
3066 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3068 @opindex Wno-selector
3069 Warn if multiple methods of different types for the same selector are
3070 found during compilation. The check is performed on the list of methods
3071 in the final stage of compilation. Additionally, a check is performed
3072 for each selector appearing in a @code{@@selector(@dots{})}
3073 expression, and a corresponding method for that selector has been found
3074 during compilation. Because these checks scan the method table only at
3075 the end of compilation, these warnings are not produced if the final
3076 stage of compilation is not reached, for example because an error is
3077 found during compilation, or because the @option{-fsyntax-only} option is
3080 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3081 @opindex Wstrict-selector-match
3082 @opindex Wno-strict-selector-match
3083 Warn if multiple methods with differing argument and/or return types are
3084 found for a given selector when attempting to send a message using this
3085 selector to a receiver of type @code{id} or @code{Class}. When this flag
3086 is off (which is the default behavior), the compiler omits such warnings
3087 if any differences found are confined to types that share the same size
3090 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3091 @opindex Wundeclared-selector
3092 @opindex Wno-undeclared-selector
3093 Warn if a @code{@@selector(@dots{})} expression referring to an
3094 undeclared selector is found. A selector is considered undeclared if no
3095 method with that name has been declared before the
3096 @code{@@selector(@dots{})} expression, either explicitly in an
3097 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3098 an @code{@@implementation} section. This option always performs its
3099 checks as soon as a @code{@@selector(@dots{})} expression is found,
3100 while @option{-Wselector} only performs its checks in the final stage of
3101 compilation. This also enforces the coding style convention
3102 that methods and selectors must be declared before being used.
3104 @item -print-objc-runtime-info
3105 @opindex print-objc-runtime-info
3106 Generate C header describing the largest structure that is passed by
3111 @node Language Independent Options
3112 @section Options to Control Diagnostic Messages Formatting
3113 @cindex options to control diagnostics formatting
3114 @cindex diagnostic messages
3115 @cindex message formatting
3117 Traditionally, diagnostic messages have been formatted irrespective of
3118 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3119 options described below
3120 to control the formatting algorithm for diagnostic messages,
3121 e.g.@: how many characters per line, how often source location
3122 information should be reported. Note that some language front ends may not
3123 honor these options.
3126 @item -fmessage-length=@var{n}
3127 @opindex fmessage-length
3128 Try to format error messages so that they fit on lines of about
3129 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3130 done; each error message appears on a single line. This is the
3131 default for all front ends.
3133 @item -fdiagnostics-show-location=once
3134 @opindex fdiagnostics-show-location
3135 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3136 reporter to emit source location information @emph{once}; that is, in
3137 case the message is too long to fit on a single physical line and has to
3138 be wrapped, the source location won't be emitted (as prefix) again,
3139 over and over, in subsequent continuation lines. This is the default
3142 @item -fdiagnostics-show-location=every-line
3143 Only meaningful in line-wrapping mode. Instructs the diagnostic
3144 messages reporter to emit the same source location information (as
3145 prefix) for physical lines that result from the process of breaking
3146 a message which is too long to fit on a single line.
3148 @item -fdiagnostics-color[=@var{WHEN}]
3149 @itemx -fno-diagnostics-color
3150 @opindex fdiagnostics-color
3151 @cindex highlight, color, colour
3152 @vindex GCC_COLORS @r{environment variable}
3153 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3154 or @samp{auto}. The default depends on how the compiler has been configured,
3155 it can be any of the above @var{WHEN} options or also @samp{never}
3156 if @env{GCC_COLORS} environment variable isn't present in the environment,
3157 and @samp{auto} otherwise.
3158 @samp{auto} means to use color only when the standard error is a terminal.
3159 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3160 aliases for @option{-fdiagnostics-color=always} and
3161 @option{-fdiagnostics-color=never}, respectively.
3163 The colors are defined by the environment variable @env{GCC_COLORS}.
3164 Its value is a colon-separated list of capabilities and Select Graphic
3165 Rendition (SGR) substrings. SGR commands are interpreted by the
3166 terminal or terminal emulator. (See the section in the documentation
3167 of your text terminal for permitted values and their meanings as
3168 character attributes.) These substring values are integers in decimal
3169 representation and can be concatenated with semicolons.
3170 Common values to concatenate include
3172 @samp{4} for underline,
3174 @samp{7} for inverse,
3175 @samp{39} for default foreground color,
3176 @samp{30} to @samp{37} for foreground colors,
3177 @samp{90} to @samp{97} for 16-color mode foreground colors,
3178 @samp{38;5;0} to @samp{38;5;255}
3179 for 88-color and 256-color modes foreground colors,
3180 @samp{49} for default background color,
3181 @samp{40} to @samp{47} for background colors,
3182 @samp{100} to @samp{107} for 16-color mode background colors,
3183 and @samp{48;5;0} to @samp{48;5;255}
3184 for 88-color and 256-color modes background colors.
3186 The default @env{GCC_COLORS} is
3188 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3191 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3192 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3193 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3194 string disables colors.
3195 Supported capabilities are as follows.
3199 @vindex error GCC_COLORS @r{capability}
3200 SGR substring for error: markers.
3203 @vindex warning GCC_COLORS @r{capability}
3204 SGR substring for warning: markers.
3207 @vindex note GCC_COLORS @r{capability}
3208 SGR substring for note: markers.
3211 @vindex caret GCC_COLORS @r{capability}
3212 SGR substring for caret line.
3215 @vindex locus GCC_COLORS @r{capability}
3216 SGR substring for location information, @samp{file:line} or
3217 @samp{file:line:column} etc.
3220 @vindex quote GCC_COLORS @r{capability}
3221 SGR substring for information printed within quotes.
3224 @item -fno-diagnostics-show-option
3225 @opindex fno-diagnostics-show-option
3226 @opindex fdiagnostics-show-option
3227 By default, each diagnostic emitted includes text indicating the
3228 command-line option that directly controls the diagnostic (if such an
3229 option is known to the diagnostic machinery). Specifying the
3230 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3232 @item -fno-diagnostics-show-caret
3233 @opindex fno-diagnostics-show-caret
3234 @opindex fdiagnostics-show-caret
3235 By default, each diagnostic emitted includes the original source line
3236 and a caret '^' indicating the column. This option suppresses this
3237 information. The source line is truncated to @var{n} characters, if
3238 the @option{-fmessage-length=n} is given. When the output is done
3239 to the terminal, the width is limited to the width given by the
3240 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3244 @node Warning Options
3245 @section Options to Request or Suppress Warnings
3246 @cindex options to control warnings
3247 @cindex warning messages
3248 @cindex messages, warning
3249 @cindex suppressing warnings
3251 Warnings are diagnostic messages that report constructions that
3252 are not inherently erroneous but that are risky or suggest there
3253 may have been an error.
3255 The following language-independent options do not enable specific
3256 warnings but control the kinds of diagnostics produced by GCC@.
3259 @cindex syntax checking
3261 @opindex fsyntax-only
3262 Check the code for syntax errors, but don't do anything beyond that.
3264 @item -fmax-errors=@var{n}
3265 @opindex fmax-errors
3266 Limits the maximum number of error messages to @var{n}, at which point
3267 GCC bails out rather than attempting to continue processing the source
3268 code. If @var{n} is 0 (the default), there is no limit on the number
3269 of error messages produced. If @option{-Wfatal-errors} is also
3270 specified, then @option{-Wfatal-errors} takes precedence over this
3275 Inhibit all warning messages.
3280 Make all warnings into errors.
3285 Make the specified warning into an error. The specifier for a warning
3286 is appended; for example @option{-Werror=switch} turns the warnings
3287 controlled by @option{-Wswitch} into errors. This switch takes a
3288 negative form, to be used to negate @option{-Werror} for specific
3289 warnings; for example @option{-Wno-error=switch} makes
3290 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3293 The warning message for each controllable warning includes the
3294 option that controls the warning. That option can then be used with
3295 @option{-Werror=} and @option{-Wno-error=} as described above.
3296 (Printing of the option in the warning message can be disabled using the
3297 @option{-fno-diagnostics-show-option} flag.)
3299 Note that specifying @option{-Werror=}@var{foo} automatically implies
3300 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3303 @item -Wfatal-errors
3304 @opindex Wfatal-errors
3305 @opindex Wno-fatal-errors
3306 This option causes the compiler to abort compilation on the first error
3307 occurred rather than trying to keep going and printing further error
3312 You can request many specific warnings with options beginning with
3313 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3314 implicit declarations. Each of these specific warning options also
3315 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3316 example, @option{-Wno-implicit}. This manual lists only one of the
3317 two forms, whichever is not the default. For further
3318 language-specific options also refer to @ref{C++ Dialect Options} and
3319 @ref{Objective-C and Objective-C++ Dialect Options}.
3321 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3322 options, such as @option{-Wunused}, which may turn on further options,
3323 such as @option{-Wunused-value}. The combined effect of positive and
3324 negative forms is that more specific options have priority over less
3325 specific ones, independently of their position in the command-line. For
3326 options of the same specificity, the last one takes effect. Options
3327 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3328 as if they appeared at the end of the command-line.
3330 When an unrecognized warning option is requested (e.g.,
3331 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3332 that the option is not recognized. However, if the @option{-Wno-} form
3333 is used, the behavior is slightly different: no diagnostic is
3334 produced for @option{-Wno-unknown-warning} unless other diagnostics
3335 are being produced. This allows the use of new @option{-Wno-} options
3336 with old compilers, but if something goes wrong, the compiler
3337 warns that an unrecognized option is present.
3344 Issue all the warnings demanded by strict ISO C and ISO C++;
3345 reject all programs that use forbidden extensions, and some other
3346 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3347 version of the ISO C standard specified by any @option{-std} option used.
3349 Valid ISO C and ISO C++ programs should compile properly with or without
3350 this option (though a rare few require @option{-ansi} or a
3351 @option{-std} option specifying the required version of ISO C)@. However,
3352 without this option, certain GNU extensions and traditional C and C++
3353 features are supported as well. With this option, they are rejected.
3355 @option{-Wpedantic} does not cause warning messages for use of the
3356 alternate keywords whose names begin and end with @samp{__}. Pedantic
3357 warnings are also disabled in the expression that follows
3358 @code{__extension__}. However, only system header files should use
3359 these escape routes; application programs should avoid them.
3360 @xref{Alternate Keywords}.
3362 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3363 C conformance. They soon find that it does not do quite what they want:
3364 it finds some non-ISO practices, but not all---only those for which
3365 ISO C @emph{requires} a diagnostic, and some others for which
3366 diagnostics have been added.
3368 A feature to report any failure to conform to ISO C might be useful in
3369 some instances, but would require considerable additional work and would
3370 be quite different from @option{-Wpedantic}. We don't have plans to
3371 support such a feature in the near future.
3373 Where the standard specified with @option{-std} represents a GNU
3374 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3375 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3376 extended dialect is based. Warnings from @option{-Wpedantic} are given
3377 where they are required by the base standard. (It does not make sense
3378 for such warnings to be given only for features not in the specified GNU
3379 C dialect, since by definition the GNU dialects of C include all
3380 features the compiler supports with the given option, and there would be
3381 nothing to warn about.)
3383 @item -pedantic-errors
3384 @opindex pedantic-errors
3385 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3386 requires a diagnostic, in some cases where there is undefined behavior
3387 at compile-time and in some other cases that do not prevent compilation
3388 of programs that are valid according to the standard. This is not
3389 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3390 by this option and not enabled by the latter and vice versa.
3395 This enables all the warnings about constructions that some users
3396 consider questionable, and that are easy to avoid (or modify to
3397 prevent the warning), even in conjunction with macros. This also
3398 enables some language-specific warnings described in @ref{C++ Dialect
3399 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3401 @option{-Wall} turns on the following warning flags:
3403 @gccoptlist{-Waddress @gol
3404 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3405 -Wc++11-compat -Wc++14-compat@gol
3406 -Wchar-subscripts @gol
3407 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3408 -Wimplicit-int @r{(C and Objective-C only)} @gol
3409 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3412 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3413 -Wmaybe-uninitialized @gol
3414 -Wmissing-braces @r{(only for C/ObjC)} @gol
3421 -Wsequence-point @gol
3422 -Wsign-compare @r{(only in C++)} @gol
3423 -Wstrict-aliasing @gol
3424 -Wstrict-overflow=1 @gol
3427 -Wuninitialized @gol
3428 -Wunknown-pragmas @gol
3429 -Wunused-function @gol
3432 -Wunused-variable @gol
3433 -Wvolatile-register-var @gol
3436 Note that some warning flags are not implied by @option{-Wall}. Some of
3437 them warn about constructions that users generally do not consider
3438 questionable, but which occasionally you might wish to check for;
3439 others warn about constructions that are necessary or hard to avoid in
3440 some cases, and there is no simple way to modify the code to suppress
3441 the warning. Some of them are enabled by @option{-Wextra} but many of
3442 them must be enabled individually.
3448 This enables some extra warning flags that are not enabled by
3449 @option{-Wall}. (This option used to be called @option{-W}. The older
3450 name is still supported, but the newer name is more descriptive.)
3452 @gccoptlist{-Wclobbered @gol
3454 -Wignored-qualifiers @gol
3455 -Wmissing-field-initializers @gol
3456 -Wmissing-parameter-type @r{(C only)} @gol
3457 -Wold-style-declaration @r{(C only)} @gol
3458 -Woverride-init @gol
3461 -Wuninitialized @gol
3462 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3463 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3466 The option @option{-Wextra} also prints warning messages for the
3472 A pointer is compared against integer zero with @code{<}, @code{<=},
3473 @code{>}, or @code{>=}.
3476 (C++ only) An enumerator and a non-enumerator both appear in a
3477 conditional expression.
3480 (C++ only) Ambiguous virtual bases.
3483 (C++ only) Subscripting an array that has been declared @code{register}.
3486 (C++ only) Taking the address of a variable that has been declared
3490 (C++ only) A base class is not initialized in a derived class's copy
3495 @item -Wchar-subscripts
3496 @opindex Wchar-subscripts
3497 @opindex Wno-char-subscripts
3498 Warn if an array subscript has type @code{char}. This is a common cause
3499 of error, as programmers often forget that this type is signed on some
3501 This warning is enabled by @option{-Wall}.
3505 @opindex Wno-comment
3506 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3507 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3508 This warning is enabled by @option{-Wall}.
3510 @item -Wno-coverage-mismatch
3511 @opindex Wno-coverage-mismatch
3512 Warn if feedback profiles do not match when using the
3513 @option{-fprofile-use} option.
3514 If a source file is changed between compiling with @option{-fprofile-gen} and
3515 with @option{-fprofile-use}, the files with the profile feedback can fail
3516 to match the source file and GCC cannot use the profile feedback
3517 information. By default, this warning is enabled and is treated as an
3518 error. @option{-Wno-coverage-mismatch} can be used to disable the
3519 warning or @option{-Wno-error=coverage-mismatch} can be used to
3520 disable the error. Disabling the error for this warning can result in
3521 poorly optimized code and is useful only in the
3522 case of very minor changes such as bug fixes to an existing code-base.
3523 Completely disabling the warning is not recommended.
3526 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3528 Suppress warning messages emitted by @code{#warning} directives.
3530 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3531 @opindex Wdouble-promotion
3532 @opindex Wno-double-promotion
3533 Give a warning when a value of type @code{float} is implicitly
3534 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3535 floating-point unit implement @code{float} in hardware, but emulate
3536 @code{double} in software. On such a machine, doing computations
3537 using @code{double} values is much more expensive because of the
3538 overhead required for software emulation.
3540 It is easy to accidentally do computations with @code{double} because
3541 floating-point literals are implicitly of type @code{double}. For
3545 float area(float radius)
3547 return 3.14159 * radius * radius;
3551 the compiler performs the entire computation with @code{double}
3552 because the floating-point literal is a @code{double}.
3555 @itemx -Wformat=@var{n}
3558 @opindex ffreestanding
3559 @opindex fno-builtin
3561 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3562 the arguments supplied have types appropriate to the format string
3563 specified, and that the conversions specified in the format string make
3564 sense. This includes standard functions, and others specified by format
3565 attributes (@pxref{Function Attributes}), in the @code{printf},
3566 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3567 not in the C standard) families (or other target-specific families).
3568 Which functions are checked without format attributes having been
3569 specified depends on the standard version selected, and such checks of
3570 functions without the attribute specified are disabled by
3571 @option{-ffreestanding} or @option{-fno-builtin}.
3573 The formats are checked against the format features supported by GNU
3574 libc version 2.2. These include all ISO C90 and C99 features, as well
3575 as features from the Single Unix Specification and some BSD and GNU
3576 extensions. Other library implementations may not support all these
3577 features; GCC does not support warning about features that go beyond a
3578 particular library's limitations. However, if @option{-Wpedantic} is used
3579 with @option{-Wformat}, warnings are given about format features not
3580 in the selected standard version (but not for @code{strfmon} formats,
3581 since those are not in any version of the C standard). @xref{C Dialect
3582 Options,,Options Controlling C Dialect}.
3589 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3590 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3591 @option{-Wformat} also checks for null format arguments for several
3592 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3593 aspects of this level of format checking can be disabled by the
3594 options: @option{-Wno-format-contains-nul},
3595 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3596 @option{-Wformat} is enabled by @option{-Wall}.
3598 @item -Wno-format-contains-nul
3599 @opindex Wno-format-contains-nul
3600 @opindex Wformat-contains-nul
3601 If @option{-Wformat} is specified, do not warn about format strings that
3604 @item -Wno-format-extra-args
3605 @opindex Wno-format-extra-args
3606 @opindex Wformat-extra-args
3607 If @option{-Wformat} is specified, do not warn about excess arguments to a
3608 @code{printf} or @code{scanf} format function. The C standard specifies
3609 that such arguments are ignored.
3611 Where the unused arguments lie between used arguments that are
3612 specified with @samp{$} operand number specifications, normally
3613 warnings are still given, since the implementation could not know what
3614 type to pass to @code{va_arg} to skip the unused arguments. However,
3615 in the case of @code{scanf} formats, this option suppresses the
3616 warning if the unused arguments are all pointers, since the Single
3617 Unix Specification says that such unused arguments are allowed.
3619 @item -Wno-format-zero-length
3620 @opindex Wno-format-zero-length
3621 @opindex Wformat-zero-length
3622 If @option{-Wformat} is specified, do not warn about zero-length formats.
3623 The C standard specifies that zero-length formats are allowed.
3628 Enable @option{-Wformat} plus additional format checks. Currently
3629 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3630 -Wformat-signedness -Wformat-y2k}.
3632 @item -Wformat-nonliteral
3633 @opindex Wformat-nonliteral
3634 @opindex Wno-format-nonliteral
3635 If @option{-Wformat} is specified, also warn if the format string is not a
3636 string literal and so cannot be checked, unless the format function
3637 takes its format arguments as a @code{va_list}.
3639 @item -Wformat-security
3640 @opindex Wformat-security
3641 @opindex Wno-format-security
3642 If @option{-Wformat} is specified, also warn about uses of format
3643 functions that represent possible security problems. At present, this
3644 warns about calls to @code{printf} and @code{scanf} functions where the
3645 format string is not a string literal and there are no format arguments,
3646 as in @code{printf (foo);}. This may be a security hole if the format
3647 string came from untrusted input and contains @samp{%n}. (This is
3648 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3649 in future warnings may be added to @option{-Wformat-security} that are not
3650 included in @option{-Wformat-nonliteral}.)
3652 @item -Wformat-signedness
3653 @opindex Wformat-signedness
3654 @opindex Wno-format-signedness
3655 If @option{-Wformat} is specified, also warn if the format string
3656 requires an unsigned argument and the argument is signed and vice versa.
3659 @opindex Wformat-y2k
3660 @opindex Wno-format-y2k
3661 If @option{-Wformat} is specified, also warn about @code{strftime}
3662 formats that may yield only a two-digit year.
3667 @opindex Wno-nonnull
3668 Warn about passing a null pointer for arguments marked as
3669 requiring a non-null value by the @code{nonnull} function attribute.
3671 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3672 can be disabled with the @option{-Wno-nonnull} option.
3674 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3676 @opindex Wno-init-self
3677 Warn about uninitialized variables that are initialized with themselves.
3678 Note this option can only be used with the @option{-Wuninitialized} option.
3680 For example, GCC warns about @code{i} being uninitialized in the
3681 following snippet only when @option{-Winit-self} has been specified:
3692 This warning is enabled by @option{-Wall} in C++.
3694 @item -Wimplicit-int @r{(C and Objective-C only)}
3695 @opindex Wimplicit-int
3696 @opindex Wno-implicit-int
3697 Warn when a declaration does not specify a type.
3698 This warning is enabled by @option{-Wall}.
3700 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3701 @opindex Wimplicit-function-declaration
3702 @opindex Wno-implicit-function-declaration
3703 Give a warning whenever a function is used before being declared. In
3704 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3705 enabled by default and it is made into an error by
3706 @option{-pedantic-errors}. This warning is also enabled by
3709 @item -Wimplicit @r{(C and Objective-C only)}
3711 @opindex Wno-implicit
3712 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3713 This warning is enabled by @option{-Wall}.
3715 @item -Wignored-qualifiers @r{(C and C++ only)}
3716 @opindex Wignored-qualifiers
3717 @opindex Wno-ignored-qualifiers
3718 Warn if the return type of a function has a type qualifier
3719 such as @code{const}. For ISO C such a type qualifier has no effect,
3720 since the value returned by a function is not an lvalue.
3721 For C++, the warning is only emitted for scalar types or @code{void}.
3722 ISO C prohibits qualified @code{void} return types on function
3723 definitions, so such return types always receive a warning
3724 even without this option.
3726 This warning is also enabled by @option{-Wextra}.
3731 Warn if the type of @code{main} is suspicious. @code{main} should be
3732 a function with external linkage, returning int, taking either zero
3733 arguments, two, or three arguments of appropriate types. This warning
3734 is enabled by default in C++ and is enabled by either @option{-Wall}
3735 or @option{-Wpedantic}.
3737 @item -Wmissing-braces
3738 @opindex Wmissing-braces
3739 @opindex Wno-missing-braces
3740 Warn if an aggregate or union initializer is not fully bracketed. In
3741 the following example, the initializer for @code{a} is not fully
3742 bracketed, but that for @code{b} is fully bracketed. This warning is
3743 enabled by @option{-Wall} in C.
3746 int a[2][2] = @{ 0, 1, 2, 3 @};
3747 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3750 This warning is enabled by @option{-Wall}.
3752 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3753 @opindex Wmissing-include-dirs
3754 @opindex Wno-missing-include-dirs
3755 Warn if a user-supplied include directory does not exist.
3758 @opindex Wparentheses
3759 @opindex Wno-parentheses
3760 Warn if parentheses are omitted in certain contexts, such
3761 as when there is an assignment in a context where a truth value
3762 is expected, or when operators are nested whose precedence people
3763 often get confused about.
3765 Also warn if a comparison like @code{x<=y<=z} appears; this is
3766 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3767 interpretation from that of ordinary mathematical notation.
3769 Also warn about constructions where there may be confusion to which
3770 @code{if} statement an @code{else} branch belongs. Here is an example of
3785 In C/C++, every @code{else} branch belongs to the innermost possible
3786 @code{if} statement, which in this example is @code{if (b)}. This is
3787 often not what the programmer expected, as illustrated in the above
3788 example by indentation the programmer chose. When there is the
3789 potential for this confusion, GCC issues a warning when this flag
3790 is specified. To eliminate the warning, add explicit braces around
3791 the innermost @code{if} statement so there is no way the @code{else}
3792 can belong to the enclosing @code{if}. The resulting code
3809 Also warn for dangerous uses of the GNU extension to
3810 @code{?:} with omitted middle operand. When the condition
3811 in the @code{?}: operator is a boolean expression, the omitted value is
3812 always 1. Often programmers expect it to be a value computed
3813 inside the conditional expression instead.
3815 This warning is enabled by @option{-Wall}.
3817 @item -Wsequence-point
3818 @opindex Wsequence-point
3819 @opindex Wno-sequence-point
3820 Warn about code that may have undefined semantics because of violations
3821 of sequence point rules in the C and C++ standards.
3823 The C and C++ standards define the order in which expressions in a C/C++
3824 program are evaluated in terms of @dfn{sequence points}, which represent
3825 a partial ordering between the execution of parts of the program: those
3826 executed before the sequence point, and those executed after it. These
3827 occur after the evaluation of a full expression (one which is not part
3828 of a larger expression), after the evaluation of the first operand of a
3829 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3830 function is called (but after the evaluation of its arguments and the
3831 expression denoting the called function), and in certain other places.
3832 Other than as expressed by the sequence point rules, the order of
3833 evaluation of subexpressions of an expression is not specified. All
3834 these rules describe only a partial order rather than a total order,
3835 since, for example, if two functions are called within one expression
3836 with no sequence point between them, the order in which the functions
3837 are called is not specified. However, the standards committee have
3838 ruled that function calls do not overlap.
3840 It is not specified when between sequence points modifications to the
3841 values of objects take effect. Programs whose behavior depends on this
3842 have undefined behavior; the C and C++ standards specify that ``Between
3843 the previous and next sequence point an object shall have its stored
3844 value modified at most once by the evaluation of an expression.
3845 Furthermore, the prior value shall be read only to determine the value
3846 to be stored.''. If a program breaks these rules, the results on any
3847 particular implementation are entirely unpredictable.
3849 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3850 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3851 diagnosed by this option, and it may give an occasional false positive
3852 result, but in general it has been found fairly effective at detecting
3853 this sort of problem in programs.
3855 The standard is worded confusingly, therefore there is some debate
3856 over the precise meaning of the sequence point rules in subtle cases.
3857 Links to discussions of the problem, including proposed formal
3858 definitions, may be found on the GCC readings page, at
3859 @uref{http://gcc.gnu.org/@/readings.html}.
3861 This warning is enabled by @option{-Wall} for C and C++.
3863 @item -Wno-return-local-addr
3864 @opindex Wno-return-local-addr
3865 @opindex Wreturn-local-addr
3866 Do not warn about returning a pointer (or in C++, a reference) to a
3867 variable that goes out of scope after the function returns.
3870 @opindex Wreturn-type
3871 @opindex Wno-return-type
3872 Warn whenever a function is defined with a return type that defaults
3873 to @code{int}. Also warn about any @code{return} statement with no
3874 return value in a function whose return type is not @code{void}
3875 (falling off the end of the function body is considered returning
3876 without a value), and about a @code{return} statement with an
3877 expression in a function whose return type is @code{void}.
3879 For C++, a function without return type always produces a diagnostic
3880 message, even when @option{-Wno-return-type} is specified. The only
3881 exceptions are @code{main} and functions defined in system headers.
3883 This warning is enabled by @option{-Wall}.
3885 @item -Wshift-count-negative
3886 @opindex Wshift-count-negative
3887 @opindex Wno-shift-count-negative
3888 Warn if shift count is negative. This warning is enabled by default.
3890 @item -Wshift-count-overflow
3891 @opindex Wshift-count-overflow
3892 @opindex Wno-shift-count-overflow
3893 Warn if shift count >= width of type. This warning is enabled by default.
3898 Warn whenever a @code{switch} statement has an index of enumerated type
3899 and lacks a @code{case} for one or more of the named codes of that
3900 enumeration. (The presence of a @code{default} label prevents this
3901 warning.) @code{case} labels outside the enumeration range also
3902 provoke warnings when this option is used (even if there is a
3903 @code{default} label).
3904 This warning is enabled by @option{-Wall}.
3906 @item -Wswitch-default
3907 @opindex Wswitch-default
3908 @opindex Wno-switch-default
3909 Warn whenever a @code{switch} statement does not have a @code{default}
3913 @opindex Wswitch-enum
3914 @opindex Wno-switch-enum
3915 Warn whenever a @code{switch} statement has an index of enumerated type
3916 and lacks a @code{case} for one or more of the named codes of that
3917 enumeration. @code{case} labels outside the enumeration range also
3918 provoke warnings when this option is used. The only difference
3919 between @option{-Wswitch} and this option is that this option gives a
3920 warning about an omitted enumeration code even if there is a
3921 @code{default} label.
3924 @opindex Wswitch-bool
3925 @opindex Wno-switch-bool
3926 Warn whenever a @code{switch} statement has an index of boolean type.
3927 It is possible to suppress this warning by casting the controlling
3928 expression to a type other than @code{bool}. For example:
3931 switch ((int) (a == 4))
3937 This warning is enabled by default for C and C++ programs.
3939 @item -Wsync-nand @r{(C and C++ only)}
3941 @opindex Wno-sync-nand
3942 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3943 built-in functions are used. These functions changed semantics in GCC 4.4.
3947 @opindex Wno-trigraphs
3948 Warn if any trigraphs are encountered that might change the meaning of
3949 the program (trigraphs within comments are not warned about).
3950 This warning is enabled by @option{-Wall}.
3952 @item -Wunused-but-set-parameter
3953 @opindex Wunused-but-set-parameter
3954 @opindex Wno-unused-but-set-parameter
3955 Warn whenever a function parameter is assigned to, but otherwise unused
3956 (aside from its declaration).
3958 To suppress this warning use the @code{unused} attribute
3959 (@pxref{Variable Attributes}).
3961 This warning is also enabled by @option{-Wunused} together with
3964 @item -Wunused-but-set-variable
3965 @opindex Wunused-but-set-variable
3966 @opindex Wno-unused-but-set-variable
3967 Warn whenever a local variable is assigned to, but otherwise unused
3968 (aside from its declaration).
3969 This warning is enabled by @option{-Wall}.
3971 To suppress this warning use the @code{unused} attribute
3972 (@pxref{Variable Attributes}).
3974 This warning is also enabled by @option{-Wunused}, which is enabled
3977 @item -Wunused-function
3978 @opindex Wunused-function
3979 @opindex Wno-unused-function
3980 Warn whenever a static function is declared but not defined or a
3981 non-inline static function is unused.
3982 This warning is enabled by @option{-Wall}.
3984 @item -Wunused-label
3985 @opindex Wunused-label
3986 @opindex Wno-unused-label
3987 Warn whenever a label is declared but not used.
3988 This warning is enabled by @option{-Wall}.
3990 To suppress this warning use the @code{unused} attribute
3991 (@pxref{Variable Attributes}).
3993 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3994 @opindex Wunused-local-typedefs
3995 Warn when a typedef locally defined in a function is not used.
3996 This warning is enabled by @option{-Wall}.
3998 @item -Wunused-parameter
3999 @opindex Wunused-parameter
4000 @opindex Wno-unused-parameter
4001 Warn whenever a function parameter is unused aside from its declaration.
4003 To suppress this warning use the @code{unused} attribute
4004 (@pxref{Variable Attributes}).
4006 @item -Wno-unused-result
4007 @opindex Wunused-result
4008 @opindex Wno-unused-result
4009 Do not warn if a caller of a function marked with attribute
4010 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4011 its return value. The default is @option{-Wunused-result}.
4013 @item -Wunused-variable
4014 @opindex Wunused-variable
4015 @opindex Wno-unused-variable
4016 Warn whenever a local variable or non-constant static variable is unused
4017 aside from its declaration.
4018 This warning is enabled by @option{-Wall}.
4020 To suppress this warning use the @code{unused} attribute
4021 (@pxref{Variable Attributes}).
4023 @item -Wunused-value
4024 @opindex Wunused-value
4025 @opindex Wno-unused-value
4026 Warn whenever a statement computes a result that is explicitly not
4027 used. To suppress this warning cast the unused expression to
4028 @code{void}. This includes an expression-statement or the left-hand
4029 side of a comma expression that contains no side effects. For example,
4030 an expression such as @code{x[i,j]} causes a warning, while
4031 @code{x[(void)i,j]} does not.
4033 This warning is enabled by @option{-Wall}.
4038 All the above @option{-Wunused} options combined.
4040 In order to get a warning about an unused function parameter, you must
4041 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4042 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4044 @item -Wuninitialized
4045 @opindex Wuninitialized
4046 @opindex Wno-uninitialized
4047 Warn if an automatic variable is used without first being initialized
4048 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4049 warn if a non-static reference or non-static @code{const} member
4050 appears in a class without constructors.
4052 If you want to warn about code that uses the uninitialized value of the
4053 variable in its own initializer, use the @option{-Winit-self} option.
4055 These warnings occur for individual uninitialized or clobbered
4056 elements of structure, union or array variables as well as for
4057 variables that are uninitialized or clobbered as a whole. They do
4058 not occur for variables or elements declared @code{volatile}. Because
4059 these warnings depend on optimization, the exact variables or elements
4060 for which there are warnings depends on the precise optimization
4061 options and version of GCC used.
4063 Note that there may be no warning about a variable that is used only
4064 to compute a value that itself is never used, because such
4065 computations may be deleted by data flow analysis before the warnings
4068 @item -Wmaybe-uninitialized
4069 @opindex Wmaybe-uninitialized
4070 @opindex Wno-maybe-uninitialized
4071 For an automatic variable, if there exists a path from the function
4072 entry to a use of the variable that is initialized, but there exist
4073 some other paths for which the variable is not initialized, the compiler
4074 emits a warning if it cannot prove the uninitialized paths are not
4075 executed at run time. These warnings are made optional because GCC is
4076 not smart enough to see all the reasons why the code might be correct
4077 in spite of appearing to have an error. Here is one example of how
4098 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4099 always initialized, but GCC doesn't know this. To suppress the
4100 warning, you need to provide a default case with assert(0) or
4103 @cindex @code{longjmp} warnings
4104 This option also warns when a non-volatile automatic variable might be
4105 changed by a call to @code{longjmp}. These warnings as well are possible
4106 only in optimizing compilation.
4108 The compiler sees only the calls to @code{setjmp}. It cannot know
4109 where @code{longjmp} will be called; in fact, a signal handler could
4110 call it at any point in the code. As a result, you may get a warning
4111 even when there is in fact no problem because @code{longjmp} cannot
4112 in fact be called at the place that would cause a problem.
4114 Some spurious warnings can be avoided if you declare all the functions
4115 you use that never return as @code{noreturn}. @xref{Function
4118 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4120 @item -Wunknown-pragmas
4121 @opindex Wunknown-pragmas
4122 @opindex Wno-unknown-pragmas
4123 @cindex warning for unknown pragmas
4124 @cindex unknown pragmas, warning
4125 @cindex pragmas, warning of unknown
4126 Warn when a @code{#pragma} directive is encountered that is not understood by
4127 GCC@. If this command-line option is used, warnings are even issued
4128 for unknown pragmas in system header files. This is not the case if
4129 the warnings are only enabled by the @option{-Wall} command-line option.
4132 @opindex Wno-pragmas
4134 Do not warn about misuses of pragmas, such as incorrect parameters,
4135 invalid syntax, or conflicts between pragmas. See also
4136 @option{-Wunknown-pragmas}.
4138 @item -Wstrict-aliasing
4139 @opindex Wstrict-aliasing
4140 @opindex Wno-strict-aliasing
4141 This option is only active when @option{-fstrict-aliasing} is active.
4142 It warns about code that might break the strict aliasing rules that the
4143 compiler is using for optimization. The warning does not catch all
4144 cases, but does attempt to catch the more common pitfalls. It is
4145 included in @option{-Wall}.
4146 It is equivalent to @option{-Wstrict-aliasing=3}
4148 @item -Wstrict-aliasing=n
4149 @opindex Wstrict-aliasing=n
4150 This option is only active when @option{-fstrict-aliasing} is active.
4151 It warns about code that might break the strict aliasing rules that the
4152 compiler is using for optimization.
4153 Higher levels correspond to higher accuracy (fewer false positives).
4154 Higher levels also correspond to more effort, similar to the way @option{-O}
4156 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4158 Level 1: Most aggressive, quick, least accurate.
4159 Possibly useful when higher levels
4160 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4161 false negatives. However, it has many false positives.
4162 Warns for all pointer conversions between possibly incompatible types,
4163 even if never dereferenced. Runs in the front end only.
4165 Level 2: Aggressive, quick, not too precise.
4166 May still have many false positives (not as many as level 1 though),
4167 and few false negatives (but possibly more than level 1).
4168 Unlike level 1, it only warns when an address is taken. Warns about
4169 incomplete types. Runs in the front end only.
4171 Level 3 (default for @option{-Wstrict-aliasing}):
4172 Should have very few false positives and few false
4173 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4174 Takes care of the common pun+dereference pattern in the front end:
4175 @code{*(int*)&some_float}.
4176 If optimization is enabled, it also runs in the back end, where it deals
4177 with multiple statement cases using flow-sensitive points-to information.
4178 Only warns when the converted pointer is dereferenced.
4179 Does not warn about incomplete types.
4181 @item -Wstrict-overflow
4182 @itemx -Wstrict-overflow=@var{n}
4183 @opindex Wstrict-overflow
4184 @opindex Wno-strict-overflow
4185 This option is only active when @option{-fstrict-overflow} is active.
4186 It warns about cases where the compiler optimizes based on the
4187 assumption that signed overflow does not occur. Note that it does not
4188 warn about all cases where the code might overflow: it only warns
4189 about cases where the compiler implements some optimization. Thus
4190 this warning depends on the optimization level.
4192 An optimization that assumes that signed overflow does not occur is
4193 perfectly safe if the values of the variables involved are such that
4194 overflow never does, in fact, occur. Therefore this warning can
4195 easily give a false positive: a warning about code that is not
4196 actually a problem. To help focus on important issues, several
4197 warning levels are defined. No warnings are issued for the use of
4198 undefined signed overflow when estimating how many iterations a loop
4199 requires, in particular when determining whether a loop will be
4203 @item -Wstrict-overflow=1
4204 Warn about cases that are both questionable and easy to avoid. For
4205 example, with @option{-fstrict-overflow}, the compiler simplifies
4206 @code{x + 1 > x} to @code{1}. This level of
4207 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4208 are not, and must be explicitly requested.
4210 @item -Wstrict-overflow=2
4211 Also warn about other cases where a comparison is simplified to a
4212 constant. For example: @code{abs (x) >= 0}. This can only be
4213 simplified when @option{-fstrict-overflow} is in effect, because
4214 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4215 zero. @option{-Wstrict-overflow} (with no level) is the same as
4216 @option{-Wstrict-overflow=2}.
4218 @item -Wstrict-overflow=3
4219 Also warn about other cases where a comparison is simplified. For
4220 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4222 @item -Wstrict-overflow=4
4223 Also warn about other simplifications not covered by the above cases.
4224 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4226 @item -Wstrict-overflow=5
4227 Also warn about cases where the compiler reduces the magnitude of a
4228 constant involved in a comparison. For example: @code{x + 2 > y} is
4229 simplified to @code{x + 1 >= y}. This is reported only at the
4230 highest warning level because this simplification applies to many
4231 comparisons, so this warning level gives a very large number of
4235 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4236 @opindex Wsuggest-attribute=
4237 @opindex Wno-suggest-attribute=
4238 Warn for cases where adding an attribute may be beneficial. The
4239 attributes currently supported are listed below.
4242 @item -Wsuggest-attribute=pure
4243 @itemx -Wsuggest-attribute=const
4244 @itemx -Wsuggest-attribute=noreturn
4245 @opindex Wsuggest-attribute=pure
4246 @opindex Wno-suggest-attribute=pure
4247 @opindex Wsuggest-attribute=const
4248 @opindex Wno-suggest-attribute=const
4249 @opindex Wsuggest-attribute=noreturn
4250 @opindex Wno-suggest-attribute=noreturn
4252 Warn about functions that might be candidates for attributes
4253 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4254 functions visible in other compilation units or (in the case of @code{pure} and
4255 @code{const}) if it cannot prove that the function returns normally. A function
4256 returns normally if it doesn't contain an infinite loop or return abnormally
4257 by throwing, calling @code{abort} or trapping. This analysis requires option
4258 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4259 higher. Higher optimization levels improve the accuracy of the analysis.
4261 @item -Wsuggest-attribute=format
4262 @itemx -Wmissing-format-attribute
4263 @opindex Wsuggest-attribute=format
4264 @opindex Wmissing-format-attribute
4265 @opindex Wno-suggest-attribute=format
4266 @opindex Wno-missing-format-attribute
4270 Warn about function pointers that might be candidates for @code{format}
4271 attributes. Note these are only possible candidates, not absolute ones.
4272 GCC guesses that function pointers with @code{format} attributes that
4273 are used in assignment, initialization, parameter passing or return
4274 statements should have a corresponding @code{format} attribute in the
4275 resulting type. I.e.@: the left-hand side of the assignment or
4276 initialization, the type of the parameter variable, or the return type
4277 of the containing function respectively should also have a @code{format}
4278 attribute to avoid the warning.
4280 GCC also warns about function definitions that might be
4281 candidates for @code{format} attributes. Again, these are only
4282 possible candidates. GCC guesses that @code{format} attributes
4283 might be appropriate for any function that calls a function like
4284 @code{vprintf} or @code{vscanf}, but this might not always be the
4285 case, and some functions for which @code{format} attributes are
4286 appropriate may not be detected.
4289 @item -Wsuggest-final-types
4290 @opindex Wno-suggest-final-types
4291 @opindex Wsuggest-final-types
4292 Warn about types with virtual methods where code quality would be improved
4293 if the type were declared with the C++11 @code{final} specifier,
4295 declared in an anonymous namespace. This allows GCC to more aggressively
4296 devirtualize the polymorphic calls. This warning is more effective with link
4297 time optimization, where the information about the class hierarchy graph is
4300 @item -Wsuggest-final-methods
4301 @opindex Wno-suggest-final-methods
4302 @opindex Wsuggest-final-methods
4303 Warn about virtual methods where code quality would be improved if the method
4304 were declared with the C++11 @code{final} specifier,
4305 or, if possible, its type were
4306 declared in an anonymous namespace or with the @code{final} specifier.
4308 more effective with link time optimization, where the information about the
4309 class hierarchy graph is more complete. It is recommended to first consider
4310 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4313 @item -Wsuggest-override
4314 Warn about overriding virtual functions that are not marked with the override
4317 @item -Warray-bounds
4318 @itemx -Warray-bounds=@var{n}
4319 @opindex Wno-array-bounds
4320 @opindex Warray-bounds
4321 This option is only active when @option{-ftree-vrp} is active
4322 (default for @option{-O2} and above). It warns about subscripts to arrays
4323 that are always out of bounds. This warning is enabled by @option{-Wall}.
4326 @item -Warray-bounds=1
4327 This is the warning level of @option{-Warray-bounds} and is enabled
4328 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4330 @item -Warray-bounds=2
4331 This warning level also warns about out of bounds access for
4332 arrays at the end of a struct and for arrays accessed through
4333 pointers. This warning level may give a larger number of
4334 false positives and is deactivated by default.
4338 @item -Wbool-compare
4339 @opindex Wno-bool-compare
4340 @opindex Wbool-compare
4341 Warn about boolean expression compared with an integer value different from
4342 @code{true}/@code{false}. For instance, the following comparison is
4347 if ((n > 1) == 2) @{ @dots{} @}
4349 This warning is enabled by @option{-Wall}.
4351 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4352 @opindex Wno-discarded-qualifiers
4353 @opindex Wdiscarded-qualifiers
4354 Do not warn if type qualifiers on pointers are being discarded.
4355 Typically, the compiler warns if a @code{const char *} variable is
4356 passed to a function that takes a @code{char *} parameter. This option
4357 can be used to suppress such a warning.
4359 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4360 @opindex Wno-discarded-array-qualifiers
4361 @opindex Wdiscarded-array-qualifiers
4362 Do not warn if type qualifiers on arrays which are pointer targets
4363 are being discarded. Typically, the compiler warns if a
4364 @code{const int (*)[]} variable is passed to a function that
4365 takes a @code{int (*)[]} parameter. This option can be used to
4366 suppress such a warning.
4368 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4369 @opindex Wno-incompatible-pointer-types
4370 @opindex Wincompatible-pointer-types
4371 Do not warn when there is a conversion between pointers that have incompatible
4372 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4373 which warns for pointer argument passing or assignment with different
4376 @item -Wno-int-conversion @r{(C and Objective-C only)}
4377 @opindex Wno-int-conversion
4378 @opindex Wint-conversion
4379 Do not warn about incompatible integer to pointer and pointer to integer
4380 conversions. This warning is about implicit conversions; for explicit
4381 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4382 @option{-Wno-pointer-to-int-cast} may be used.
4384 @item -Wno-div-by-zero
4385 @opindex Wno-div-by-zero
4386 @opindex Wdiv-by-zero
4387 Do not warn about compile-time integer division by zero. Floating-point
4388 division by zero is not warned about, as it can be a legitimate way of
4389 obtaining infinities and NaNs.
4391 @item -Wsystem-headers
4392 @opindex Wsystem-headers
4393 @opindex Wno-system-headers
4394 @cindex warnings from system headers
4395 @cindex system headers, warnings from
4396 Print warning messages for constructs found in system header files.
4397 Warnings from system headers are normally suppressed, on the assumption
4398 that they usually do not indicate real problems and would only make the
4399 compiler output harder to read. Using this command-line option tells
4400 GCC to emit warnings from system headers as if they occurred in user
4401 code. However, note that using @option{-Wall} in conjunction with this
4402 option does @emph{not} warn about unknown pragmas in system
4403 headers---for that, @option{-Wunknown-pragmas} must also be used.
4406 @opindex Wtrampolines
4407 @opindex Wno-trampolines
4408 Warn about trampolines generated for pointers to nested functions.
4409 A trampoline is a small piece of data or code that is created at run
4410 time on the stack when the address of a nested function is taken, and is
4411 used to call the nested function indirectly. For some targets, it is
4412 made up of data only and thus requires no special treatment. But, for
4413 most targets, it is made up of code and thus requires the stack to be
4414 made executable in order for the program to work properly.
4417 @opindex Wfloat-equal
4418 @opindex Wno-float-equal
4419 Warn if floating-point values are used in equality comparisons.
4421 The idea behind this is that sometimes it is convenient (for the
4422 programmer) to consider floating-point values as approximations to
4423 infinitely precise real numbers. If you are doing this, then you need
4424 to compute (by analyzing the code, or in some other way) the maximum or
4425 likely maximum error that the computation introduces, and allow for it
4426 when performing comparisons (and when producing output, but that's a
4427 different problem). In particular, instead of testing for equality, you
4428 should check to see whether the two values have ranges that overlap; and
4429 this is done with the relational operators, so equality comparisons are
4432 @item -Wtraditional @r{(C and Objective-C only)}
4433 @opindex Wtraditional
4434 @opindex Wno-traditional
4435 Warn about certain constructs that behave differently in traditional and
4436 ISO C@. Also warn about ISO C constructs that have no traditional C
4437 equivalent, and/or problematic constructs that should be avoided.
4441 Macro parameters that appear within string literals in the macro body.
4442 In traditional C macro replacement takes place within string literals,
4443 but in ISO C it does not.
4446 In traditional C, some preprocessor directives did not exist.
4447 Traditional preprocessors only considered a line to be a directive
4448 if the @samp{#} appeared in column 1 on the line. Therefore
4449 @option{-Wtraditional} warns about directives that traditional C
4450 understands but ignores because the @samp{#} does not appear as the
4451 first character on the line. It also suggests you hide directives like
4452 @code{#pragma} not understood by traditional C by indenting them. Some
4453 traditional implementations do not recognize @code{#elif}, so this option
4454 suggests avoiding it altogether.
4457 A function-like macro that appears without arguments.
4460 The unary plus operator.
4463 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4464 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4465 constants.) Note, these suffixes appear in macros defined in the system
4466 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4467 Use of these macros in user code might normally lead to spurious
4468 warnings, however GCC's integrated preprocessor has enough context to
4469 avoid warning in these cases.
4472 A function declared external in one block and then used after the end of
4476 A @code{switch} statement has an operand of type @code{long}.
4479 A non-@code{static} function declaration follows a @code{static} one.
4480 This construct is not accepted by some traditional C compilers.
4483 The ISO type of an integer constant has a different width or
4484 signedness from its traditional type. This warning is only issued if
4485 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4486 typically represent bit patterns, are not warned about.
4489 Usage of ISO string concatenation is detected.
4492 Initialization of automatic aggregates.
4495 Identifier conflicts with labels. Traditional C lacks a separate
4496 namespace for labels.
4499 Initialization of unions. If the initializer is zero, the warning is
4500 omitted. This is done under the assumption that the zero initializer in
4501 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4502 initializer warnings and relies on default initialization to zero in the
4506 Conversions by prototypes between fixed/floating-point values and vice
4507 versa. The absence of these prototypes when compiling with traditional
4508 C causes serious problems. This is a subset of the possible
4509 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4512 Use of ISO C style function definitions. This warning intentionally is
4513 @emph{not} issued for prototype declarations or variadic functions
4514 because these ISO C features appear in your code when using
4515 libiberty's traditional C compatibility macros, @code{PARAMS} and
4516 @code{VPARAMS}. This warning is also bypassed for nested functions
4517 because that feature is already a GCC extension and thus not relevant to
4518 traditional C compatibility.
4521 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4522 @opindex Wtraditional-conversion
4523 @opindex Wno-traditional-conversion
4524 Warn if a prototype causes a type conversion that is different from what
4525 would happen to the same argument in the absence of a prototype. This
4526 includes conversions of fixed point to floating and vice versa, and
4527 conversions changing the width or signedness of a fixed-point argument
4528 except when the same as the default promotion.
4530 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4531 @opindex Wdeclaration-after-statement
4532 @opindex Wno-declaration-after-statement
4533 Warn when a declaration is found after a statement in a block. This
4534 construct, known from C++, was introduced with ISO C99 and is by default
4535 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4536 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4541 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4543 @item -Wno-endif-labels
4544 @opindex Wno-endif-labels
4545 @opindex Wendif-labels
4546 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4551 Warn whenever a local variable or type declaration shadows another
4552 variable, parameter, type, class member (in C++), or instance variable
4553 (in Objective-C) or whenever a built-in function is shadowed. Note
4554 that in C++, the compiler warns if a local variable shadows an
4555 explicit typedef, but not if it shadows a struct/class/enum.
4557 @item -Wno-shadow-ivar @r{(Objective-C only)}
4558 @opindex Wno-shadow-ivar
4559 @opindex Wshadow-ivar
4560 Do not warn whenever a local variable shadows an instance variable in an
4563 @item -Wlarger-than=@var{len}
4564 @opindex Wlarger-than=@var{len}
4565 @opindex Wlarger-than-@var{len}
4566 Warn whenever an object of larger than @var{len} bytes is defined.
4568 @item -Wframe-larger-than=@var{len}
4569 @opindex Wframe-larger-than
4570 Warn if the size of a function frame is larger than @var{len} bytes.
4571 The computation done to determine the stack frame size is approximate
4572 and not conservative.
4573 The actual requirements may be somewhat greater than @var{len}
4574 even if you do not get a warning. In addition, any space allocated
4575 via @code{alloca}, variable-length arrays, or related constructs
4576 is not included by the compiler when determining
4577 whether or not to issue a warning.
4579 @item -Wno-free-nonheap-object
4580 @opindex Wno-free-nonheap-object
4581 @opindex Wfree-nonheap-object
4582 Do not warn when attempting to free an object that was not allocated
4585 @item -Wstack-usage=@var{len}
4586 @opindex Wstack-usage
4587 Warn if the stack usage of a function might be larger than @var{len} bytes.
4588 The computation done to determine the stack usage is conservative.
4589 Any space allocated via @code{alloca}, variable-length arrays, or related
4590 constructs is included by the compiler when determining whether or not to
4593 The message is in keeping with the output of @option{-fstack-usage}.
4597 If the stack usage is fully static but exceeds the specified amount, it's:
4600 warning: stack usage is 1120 bytes
4603 If the stack usage is (partly) dynamic but bounded, it's:
4606 warning: stack usage might be 1648 bytes
4609 If the stack usage is (partly) dynamic and not bounded, it's:
4612 warning: stack usage might be unbounded
4616 @item -Wunsafe-loop-optimizations
4617 @opindex Wunsafe-loop-optimizations
4618 @opindex Wno-unsafe-loop-optimizations
4619 Warn if the loop cannot be optimized because the compiler cannot
4620 assume anything on the bounds of the loop indices. With
4621 @option{-funsafe-loop-optimizations} warn if the compiler makes
4624 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4625 @opindex Wno-pedantic-ms-format
4626 @opindex Wpedantic-ms-format
4627 When used in combination with @option{-Wformat}
4628 and @option{-pedantic} without GNU extensions, this option
4629 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4630 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4631 which depend on the MS runtime.
4633 @item -Wpointer-arith
4634 @opindex Wpointer-arith
4635 @opindex Wno-pointer-arith
4636 Warn about anything that depends on the ``size of'' a function type or
4637 of @code{void}. GNU C assigns these types a size of 1, for
4638 convenience in calculations with @code{void *} pointers and pointers
4639 to functions. In C++, warn also when an arithmetic operation involves
4640 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4643 @opindex Wtype-limits
4644 @opindex Wno-type-limits
4645 Warn if a comparison is always true or always false due to the limited
4646 range of the data type, but do not warn for constant expressions. For
4647 example, warn if an unsigned variable is compared against zero with
4648 @code{<} or @code{>=}. This warning is also enabled by
4651 @item -Wbad-function-cast @r{(C and Objective-C only)}
4652 @opindex Wbad-function-cast
4653 @opindex Wno-bad-function-cast
4654 Warn when a function call is cast to a non-matching type.
4655 For example, warn if a call to a function returning an integer type
4656 is cast to a pointer type.
4658 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4659 @opindex Wc90-c99-compat
4660 @opindex Wno-c90-c99-compat
4661 Warn about features not present in ISO C90, but present in ISO C99.
4662 For instance, warn about use of variable length arrays, @code{long long}
4663 type, @code{bool} type, compound literals, designated initializers, and so
4664 on. This option is independent of the standards mode. Warnings are disabled
4665 in the expression that follows @code{__extension__}.
4667 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4668 @opindex Wc99-c11-compat
4669 @opindex Wno-c99-c11-compat
4670 Warn about features not present in ISO C99, but present in ISO C11.
4671 For instance, warn about use of anonymous structures and unions,
4672 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4673 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4674 and so on. This option is independent of the standards mode. Warnings are
4675 disabled in the expression that follows @code{__extension__}.
4677 @item -Wc++-compat @r{(C and Objective-C only)}
4678 @opindex Wc++-compat
4679 Warn about ISO C constructs that are outside of the common subset of
4680 ISO C and ISO C++, e.g.@: request for implicit conversion from
4681 @code{void *} to a pointer to non-@code{void} type.
4683 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4684 @opindex Wc++11-compat
4685 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4686 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4687 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4688 enabled by @option{-Wall}.
4690 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4691 @opindex Wc++14-compat
4692 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4693 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4697 @opindex Wno-cast-qual
4698 Warn whenever a pointer is cast so as to remove a type qualifier from
4699 the target type. For example, warn if a @code{const char *} is cast
4700 to an ordinary @code{char *}.
4702 Also warn when making a cast that introduces a type qualifier in an
4703 unsafe way. For example, casting @code{char **} to @code{const char **}
4704 is unsafe, as in this example:
4707 /* p is char ** value. */
4708 const char **q = (const char **) p;
4709 /* Assignment of readonly string to const char * is OK. */
4711 /* Now char** pointer points to read-only memory. */
4716 @opindex Wcast-align
4717 @opindex Wno-cast-align
4718 Warn whenever a pointer is cast such that the required alignment of the
4719 target is increased. For example, warn if a @code{char *} is cast to
4720 an @code{int *} on machines where integers can only be accessed at
4721 two- or four-byte boundaries.
4723 @item -Wwrite-strings
4724 @opindex Wwrite-strings
4725 @opindex Wno-write-strings
4726 When compiling C, give string constants the type @code{const
4727 char[@var{length}]} so that copying the address of one into a
4728 non-@code{const} @code{char *} pointer produces a warning. These
4729 warnings help you find at compile time code that can try to write
4730 into a string constant, but only if you have been very careful about
4731 using @code{const} in declarations and prototypes. Otherwise, it is
4732 just a nuisance. This is why we did not make @option{-Wall} request
4735 When compiling C++, warn about the deprecated conversion from string
4736 literals to @code{char *}. This warning is enabled by default for C++
4741 @opindex Wno-clobbered
4742 Warn for variables that might be changed by @code{longjmp} or
4743 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4745 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4746 @opindex Wconditionally-supported
4747 @opindex Wno-conditionally-supported
4748 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4751 @opindex Wconversion
4752 @opindex Wno-conversion
4753 Warn for implicit conversions that may alter a value. This includes
4754 conversions between real and integer, like @code{abs (x)} when
4755 @code{x} is @code{double}; conversions between signed and unsigned,
4756 like @code{unsigned ui = -1}; and conversions to smaller types, like
4757 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4758 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4759 changed by the conversion like in @code{abs (2.0)}. Warnings about
4760 conversions between signed and unsigned integers can be disabled by
4761 using @option{-Wno-sign-conversion}.
4763 For C++, also warn for confusing overload resolution for user-defined
4764 conversions; and conversions that never use a type conversion
4765 operator: conversions to @code{void}, the same type, a base class or a
4766 reference to them. Warnings about conversions between signed and
4767 unsigned integers are disabled by default in C++ unless
4768 @option{-Wsign-conversion} is explicitly enabled.
4770 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4771 @opindex Wconversion-null
4772 @opindex Wno-conversion-null
4773 Do not warn for conversions between @code{NULL} and non-pointer
4774 types. @option{-Wconversion-null} is enabled by default.
4776 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4777 @opindex Wzero-as-null-pointer-constant
4778 @opindex Wno-zero-as-null-pointer-constant
4779 Warn when a literal '0' is used as null pointer constant. This can
4780 be useful to facilitate the conversion to @code{nullptr} in C++11.
4784 @opindex Wno-date-time
4785 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4786 are encountered as they might prevent bit-wise-identical reproducible
4789 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4790 @opindex Wdelete-incomplete
4791 @opindex Wno-delete-incomplete
4792 Warn when deleting a pointer to incomplete type, which may cause
4793 undefined behavior at runtime. This warning is enabled by default.
4795 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4796 @opindex Wuseless-cast
4797 @opindex Wno-useless-cast
4798 Warn when an expression is casted to its own type.
4801 @opindex Wempty-body
4802 @opindex Wno-empty-body
4803 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
4804 while} statement. This warning is also enabled by @option{-Wextra}.
4806 @item -Wenum-compare
4807 @opindex Wenum-compare
4808 @opindex Wno-enum-compare
4809 Warn about a comparison between values of different enumerated types.
4810 In C++ enumeral mismatches in conditional expressions are also
4811 diagnosed and the warning is enabled by default. In C this warning is
4812 enabled by @option{-Wall}.
4814 @item -Wjump-misses-init @r{(C, Objective-C only)}
4815 @opindex Wjump-misses-init
4816 @opindex Wno-jump-misses-init
4817 Warn if a @code{goto} statement or a @code{switch} statement jumps
4818 forward across the initialization of a variable, or jumps backward to a
4819 label after the variable has been initialized. This only warns about
4820 variables that are initialized when they are declared. This warning is
4821 only supported for C and Objective-C; in C++ this sort of branch is an
4824 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4825 can be disabled with the @option{-Wno-jump-misses-init} option.
4827 @item -Wsign-compare
4828 @opindex Wsign-compare
4829 @opindex Wno-sign-compare
4830 @cindex warning for comparison of signed and unsigned values
4831 @cindex comparison of signed and unsigned values, warning
4832 @cindex signed and unsigned values, comparison warning
4833 Warn when a comparison between signed and unsigned values could produce
4834 an incorrect result when the signed value is converted to unsigned.
4835 This warning is also enabled by @option{-Wextra}; to get the other warnings
4836 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4838 @item -Wsign-conversion
4839 @opindex Wsign-conversion
4840 @opindex Wno-sign-conversion
4841 Warn for implicit conversions that may change the sign of an integer
4842 value, like assigning a signed integer expression to an unsigned
4843 integer variable. An explicit cast silences the warning. In C, this
4844 option is enabled also by @option{-Wconversion}.
4846 @item -Wfloat-conversion
4847 @opindex Wfloat-conversion
4848 @opindex Wno-float-conversion
4849 Warn for implicit conversions that reduce the precision of a real value.
4850 This includes conversions from real to integer, and from higher precision
4851 real to lower precision real values. This option is also enabled by
4852 @option{-Wconversion}.
4854 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
4855 @opindex Wsized-deallocation
4856 @opindex Wno-sized-deallocation
4857 Warn about a definition of an unsized deallocation function
4859 void operator delete (void *) noexcept;
4860 void operator delete[] (void *) noexcept;
4862 without a definition of the corresponding sized deallocation function
4864 void operator delete (void *, std::size_t) noexcept;
4865 void operator delete[] (void *, std::size_t) noexcept;
4867 or vice versa. Enabled by @option{-Wextra} along with
4868 @option{-fsized-deallocation}.
4870 @item -Wsizeof-pointer-memaccess
4871 @opindex Wsizeof-pointer-memaccess
4872 @opindex Wno-sizeof-pointer-memaccess
4873 Warn for suspicious length parameters to certain string and memory built-in
4874 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4875 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4876 but a pointer, and suggests a possible fix, or about
4877 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4880 @item -Wsizeof-array-argument
4881 @opindex Wsizeof-array-argument
4882 @opindex Wno-sizeof-array-argument
4883 Warn when the @code{sizeof} operator is applied to a parameter that is
4884 declared as an array in a function definition. This warning is enabled by
4885 default for C and C++ programs.
4887 @item -Wmemset-transposed-args
4888 @opindex Wmemset-transposed-args
4889 @opindex Wno-memset-transposed-args
4890 Warn for suspicious calls to the @code{memset} built-in function, if the
4891 second argument is not zero and the third argument is zero. This warns e.g.@
4892 about @code{memset (buf, sizeof buf, 0)} where most probably
4893 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4894 is only emitted if the third argument is literal zero, if it is some expression
4895 that is folded to zero, or e.g. a cast of zero to some type etc., it
4896 is far less likely that user has mistakenly exchanged the arguments and
4897 no warning is emitted. This warning is enabled by @option{-Wall}.
4901 @opindex Wno-address
4902 Warn about suspicious uses of memory addresses. These include using
4903 the address of a function in a conditional expression, such as
4904 @code{void func(void); if (func)}, and comparisons against the memory
4905 address of a string literal, such as @code{if (x == "abc")}. Such
4906 uses typically indicate a programmer error: the address of a function
4907 always evaluates to true, so their use in a conditional usually
4908 indicate that the programmer forgot the parentheses in a function
4909 call; and comparisons against string literals result in unspecified
4910 behavior and are not portable in C, so they usually indicate that the
4911 programmer intended to use @code{strcmp}. This warning is enabled by
4915 @opindex Wlogical-op
4916 @opindex Wno-logical-op
4917 Warn about suspicious uses of logical operators in expressions.
4918 This includes using logical operators in contexts where a
4919 bit-wise operator is likely to be expected.
4921 @item -Wlogical-not-parentheses
4922 @opindex Wlogical-not-parentheses
4923 @opindex Wno-logical-not-parentheses
4924 Warn about logical not used on the left hand side operand of a comparison.
4925 This option does not warn if the RHS operand is of a boolean type. Its
4926 purpose is to detect suspicious code like the following:
4930 if (!a > 1) @{ @dots{} @}
4933 It is possible to suppress the warning by wrapping the LHS into
4936 if ((!a) > 1) @{ @dots{} @}
4939 This warning is enabled by @option{-Wall}.
4941 @item -Waggregate-return
4942 @opindex Waggregate-return
4943 @opindex Wno-aggregate-return
4944 Warn if any functions that return structures or unions are defined or
4945 called. (In languages where you can return an array, this also elicits
4948 @item -Wno-aggressive-loop-optimizations
4949 @opindex Wno-aggressive-loop-optimizations
4950 @opindex Waggressive-loop-optimizations
4951 Warn if in a loop with constant number of iterations the compiler detects
4952 undefined behavior in some statement during one or more of the iterations.
4954 @item -Wno-attributes
4955 @opindex Wno-attributes
4956 @opindex Wattributes
4957 Do not warn if an unexpected @code{__attribute__} is used, such as
4958 unrecognized attributes, function attributes applied to variables,
4959 etc. This does not stop errors for incorrect use of supported
4962 @item -Wno-builtin-macro-redefined
4963 @opindex Wno-builtin-macro-redefined
4964 @opindex Wbuiltin-macro-redefined
4965 Do not warn if certain built-in macros are redefined. This suppresses
4966 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4967 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4969 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4970 @opindex Wstrict-prototypes
4971 @opindex Wno-strict-prototypes
4972 Warn if a function is declared or defined without specifying the
4973 argument types. (An old-style function definition is permitted without
4974 a warning if preceded by a declaration that specifies the argument
4977 @item -Wold-style-declaration @r{(C and Objective-C only)}
4978 @opindex Wold-style-declaration
4979 @opindex Wno-old-style-declaration
4980 Warn for obsolescent usages, according to the C Standard, in a
4981 declaration. For example, warn if storage-class specifiers like
4982 @code{static} are not the first things in a declaration. This warning
4983 is also enabled by @option{-Wextra}.
4985 @item -Wold-style-definition @r{(C and Objective-C only)}
4986 @opindex Wold-style-definition
4987 @opindex Wno-old-style-definition
4988 Warn if an old-style function definition is used. A warning is given
4989 even if there is a previous prototype.
4991 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4992 @opindex Wmissing-parameter-type
4993 @opindex Wno-missing-parameter-type
4994 A function parameter is declared without a type specifier in K&R-style
5001 This warning is also enabled by @option{-Wextra}.
5003 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5004 @opindex Wmissing-prototypes
5005 @opindex Wno-missing-prototypes
5006 Warn if a global function is defined without a previous prototype
5007 declaration. This warning is issued even if the definition itself
5008 provides a prototype. Use this option to detect global functions
5009 that do not have a matching prototype declaration in a header file.
5010 This option is not valid for C++ because all function declarations
5011 provide prototypes and a non-matching declaration declares an
5012 overload rather than conflict with an earlier declaration.
5013 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5015 @item -Wmissing-declarations
5016 @opindex Wmissing-declarations
5017 @opindex Wno-missing-declarations
5018 Warn if a global function is defined without a previous declaration.
5019 Do so even if the definition itself provides a prototype.
5020 Use this option to detect global functions that are not declared in
5021 header files. In C, no warnings are issued for functions with previous
5022 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5023 missing prototypes. In C++, no warnings are issued for function templates,
5024 or for inline functions, or for functions in anonymous namespaces.
5026 @item -Wmissing-field-initializers
5027 @opindex Wmissing-field-initializers
5028 @opindex Wno-missing-field-initializers
5032 Warn if a structure's initializer has some fields missing. For
5033 example, the following code causes such a warning, because
5034 @code{x.h} is implicitly zero:
5037 struct s @{ int f, g, h; @};
5038 struct s x = @{ 3, 4 @};
5041 This option does not warn about designated initializers, so the following
5042 modification does not trigger a warning:
5045 struct s @{ int f, g, h; @};
5046 struct s x = @{ .f = 3, .g = 4 @};
5049 In C++ this option does not warn either about the empty @{ @}
5050 initializer, for example:
5053 struct s @{ int f, g, h; @};
5057 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5058 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5060 @item -Wno-multichar
5061 @opindex Wno-multichar
5063 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5064 Usually they indicate a typo in the user's code, as they have
5065 implementation-defined values, and should not be used in portable code.
5067 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5068 @opindex Wnormalized=
5069 @opindex Wnormalized
5070 @opindex Wno-normalized
5073 @cindex character set, input normalization
5074 In ISO C and ISO C++, two identifiers are different if they are
5075 different sequences of characters. However, sometimes when characters
5076 outside the basic ASCII character set are used, you can have two
5077 different character sequences that look the same. To avoid confusion,
5078 the ISO 10646 standard sets out some @dfn{normalization rules} which
5079 when applied ensure that two sequences that look the same are turned into
5080 the same sequence. GCC can warn you if you are using identifiers that
5081 have not been normalized; this option controls that warning.
5083 There are four levels of warning supported by GCC@. The default is
5084 @option{-Wnormalized=nfc}, which warns about any identifier that is
5085 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5086 recommended form for most uses. It is equivalent to
5087 @option{-Wnormalized}.
5089 Unfortunately, there are some characters allowed in identifiers by
5090 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5091 identifiers. That is, there's no way to use these symbols in portable
5092 ISO C or C++ and have all your identifiers in NFC@.
5093 @option{-Wnormalized=id} suppresses the warning for these characters.
5094 It is hoped that future versions of the standards involved will correct
5095 this, which is why this option is not the default.
5097 You can switch the warning off for all characters by writing
5098 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5099 only do this if you are using some other normalization scheme (like
5100 ``D''), because otherwise you can easily create bugs that are
5101 literally impossible to see.
5103 Some characters in ISO 10646 have distinct meanings but look identical
5104 in some fonts or display methodologies, especially once formatting has
5105 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5106 LETTER N'', displays just like a regular @code{n} that has been
5107 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5108 normalization scheme to convert all these into a standard form as
5109 well, and GCC warns if your code is not in NFKC if you use
5110 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5111 about every identifier that contains the letter O because it might be
5112 confused with the digit 0, and so is not the default, but may be
5113 useful as a local coding convention if the programming environment
5114 cannot be fixed to display these characters distinctly.
5116 @item -Wno-deprecated
5117 @opindex Wno-deprecated
5118 @opindex Wdeprecated
5119 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5121 @item -Wno-deprecated-declarations
5122 @opindex Wno-deprecated-declarations
5123 @opindex Wdeprecated-declarations
5124 Do not warn about uses of functions (@pxref{Function Attributes}),
5125 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5126 Attributes}) marked as deprecated by using the @code{deprecated}
5130 @opindex Wno-overflow
5132 Do not warn about compile-time overflow in constant expressions.
5137 Warn about One Definition Rule violations during link-time optimization.
5138 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5141 @opindex Wopenm-simd
5142 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5143 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
5144 be used to relax the cost model.
5146 @item -Woverride-init @r{(C and Objective-C only)}
5147 @opindex Woverride-init
5148 @opindex Wno-override-init
5152 Warn if an initialized field without side effects is overridden when
5153 using designated initializers (@pxref{Designated Inits, , Designated
5156 This warning is included in @option{-Wextra}. To get other
5157 @option{-Wextra} warnings without this one, use @option{-Wextra
5158 -Wno-override-init}.
5163 Warn if a structure is given the packed attribute, but the packed
5164 attribute has no effect on the layout or size of the structure.
5165 Such structures may be mis-aligned for little benefit. For
5166 instance, in this code, the variable @code{f.x} in @code{struct bar}
5167 is misaligned even though @code{struct bar} does not itself
5168 have the packed attribute:
5175 @} __attribute__((packed));
5183 @item -Wpacked-bitfield-compat
5184 @opindex Wpacked-bitfield-compat
5185 @opindex Wno-packed-bitfield-compat
5186 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5187 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5188 the change can lead to differences in the structure layout. GCC
5189 informs you when the offset of such a field has changed in GCC 4.4.
5190 For example there is no longer a 4-bit padding between field @code{a}
5191 and @code{b} in this structure:
5198 @} __attribute__ ((packed));
5201 This warning is enabled by default. Use
5202 @option{-Wno-packed-bitfield-compat} to disable this warning.
5207 Warn if padding is included in a structure, either to align an element
5208 of the structure or to align the whole structure. Sometimes when this
5209 happens it is possible to rearrange the fields of the structure to
5210 reduce the padding and so make the structure smaller.
5212 @item -Wredundant-decls
5213 @opindex Wredundant-decls
5214 @opindex Wno-redundant-decls
5215 Warn if anything is declared more than once in the same scope, even in
5216 cases where multiple declaration is valid and changes nothing.
5218 @item -Wnested-externs @r{(C and Objective-C only)}
5219 @opindex Wnested-externs
5220 @opindex Wno-nested-externs
5221 Warn if an @code{extern} declaration is encountered within a function.
5223 @item -Wno-inherited-variadic-ctor
5224 @opindex Winherited-variadic-ctor
5225 @opindex Wno-inherited-variadic-ctor
5226 Suppress warnings about use of C++11 inheriting constructors when the
5227 base class inherited from has a C variadic constructor; the warning is
5228 on by default because the ellipsis is not inherited.
5233 Warn if a function that is declared as inline cannot be inlined.
5234 Even with this option, the compiler does not warn about failures to
5235 inline functions declared in system headers.
5237 The compiler uses a variety of heuristics to determine whether or not
5238 to inline a function. For example, the compiler takes into account
5239 the size of the function being inlined and the amount of inlining
5240 that has already been done in the current function. Therefore,
5241 seemingly insignificant changes in the source program can cause the
5242 warnings produced by @option{-Winline} to appear or disappear.
5244 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5245 @opindex Wno-invalid-offsetof
5246 @opindex Winvalid-offsetof
5247 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5248 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5249 to a non-standard-layout type is undefined. In existing C++ implementations,
5250 however, @code{offsetof} typically gives meaningful results.
5251 This flag is for users who are aware that they are
5252 writing nonportable code and who have deliberately chosen to ignore the
5255 The restrictions on @code{offsetof} may be relaxed in a future version
5256 of the C++ standard.
5258 @item -Wno-int-to-pointer-cast
5259 @opindex Wno-int-to-pointer-cast
5260 @opindex Wint-to-pointer-cast
5261 Suppress warnings from casts to pointer type of an integer of a
5262 different size. In C++, casting to a pointer type of smaller size is
5263 an error. @option{Wint-to-pointer-cast} is enabled by default.
5266 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5267 @opindex Wno-pointer-to-int-cast
5268 @opindex Wpointer-to-int-cast
5269 Suppress warnings from casts from a pointer to an integer type of a
5273 @opindex Winvalid-pch
5274 @opindex Wno-invalid-pch
5275 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5276 the search path but can't be used.
5280 @opindex Wno-long-long
5281 Warn if @code{long long} type is used. This is enabled by either
5282 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5283 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5285 @item -Wvariadic-macros
5286 @opindex Wvariadic-macros
5287 @opindex Wno-variadic-macros
5288 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5289 alternate syntax is used in ISO C99 mode. This is enabled by either
5290 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5291 messages, use @option{-Wno-variadic-macros}.
5295 @opindex Wno-varargs
5296 Warn upon questionable usage of the macros used to handle variable
5297 arguments like @code{va_start}. This is default. To inhibit the
5298 warning messages, use @option{-Wno-varargs}.
5300 @item -Wvector-operation-performance
5301 @opindex Wvector-operation-performance
5302 @opindex Wno-vector-operation-performance
5303 Warn if vector operation is not implemented via SIMD capabilities of the
5304 architecture. Mainly useful for the performance tuning.
5305 Vector operation can be implemented @code{piecewise}, which means that the
5306 scalar operation is performed on every vector element;
5307 @code{in parallel}, which means that the vector operation is implemented
5308 using scalars of wider type, which normally is more performance efficient;
5309 and @code{as a single scalar}, which means that vector fits into a
5312 @item -Wno-virtual-move-assign
5313 @opindex Wvirtual-move-assign
5314 @opindex Wno-virtual-move-assign
5315 Suppress warnings about inheriting from a virtual base with a
5316 non-trivial C++11 move assignment operator. This is dangerous because
5317 if the virtual base is reachable along more than one path, it is
5318 moved multiple times, which can mean both objects end up in the
5319 moved-from state. If the move assignment operator is written to avoid
5320 moving from a moved-from object, this warning can be disabled.
5325 Warn if variable length array is used in the code.
5326 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5327 the variable length array.
5329 @item -Wvolatile-register-var
5330 @opindex Wvolatile-register-var
5331 @opindex Wno-volatile-register-var
5332 Warn if a register variable is declared volatile. The volatile
5333 modifier does not inhibit all optimizations that may eliminate reads
5334 and/or writes to register variables. This warning is enabled by
5337 @item -Wdisabled-optimization
5338 @opindex Wdisabled-optimization
5339 @opindex Wno-disabled-optimization
5340 Warn if a requested optimization pass is disabled. This warning does
5341 not generally indicate that there is anything wrong with your code; it
5342 merely indicates that GCC's optimizers are unable to handle the code
5343 effectively. Often, the problem is that your code is too big or too
5344 complex; GCC refuses to optimize programs when the optimization
5345 itself is likely to take inordinate amounts of time.
5347 @item -Wpointer-sign @r{(C and Objective-C only)}
5348 @opindex Wpointer-sign
5349 @opindex Wno-pointer-sign
5350 Warn for pointer argument passing or assignment with different signedness.
5351 This option is only supported for C and Objective-C@. It is implied by
5352 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5353 @option{-Wno-pointer-sign}.
5355 @item -Wstack-protector
5356 @opindex Wstack-protector
5357 @opindex Wno-stack-protector
5358 This option is only active when @option{-fstack-protector} is active. It
5359 warns about functions that are not protected against stack smashing.
5361 @item -Woverlength-strings
5362 @opindex Woverlength-strings
5363 @opindex Wno-overlength-strings
5364 Warn about string constants that are longer than the ``minimum
5365 maximum'' length specified in the C standard. Modern compilers
5366 generally allow string constants that are much longer than the
5367 standard's minimum limit, but very portable programs should avoid
5368 using longer strings.
5370 The limit applies @emph{after} string constant concatenation, and does
5371 not count the trailing NUL@. In C90, the limit was 509 characters; in
5372 C99, it was raised to 4095. C++98 does not specify a normative
5373 minimum maximum, so we do not diagnose overlength strings in C++@.
5375 This option is implied by @option{-Wpedantic}, and can be disabled with
5376 @option{-Wno-overlength-strings}.
5378 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5379 @opindex Wunsuffixed-float-constants
5381 Issue a warning for any floating constant that does not have
5382 a suffix. When used together with @option{-Wsystem-headers} it
5383 warns about such constants in system header files. This can be useful
5384 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5385 from the decimal floating-point extension to C99.
5387 @item -Wno-designated-init @r{(C and Objective-C only)}
5388 Suppress warnings when a positional initializer is used to initialize
5389 a structure that has been marked with the @code{designated_init}
5394 @node Debugging Options
5395 @section Options for Debugging Your Program or GCC
5396 @cindex options, debugging
5397 @cindex debugging information options
5399 GCC has various special options that are used for debugging
5400 either your program or GCC:
5405 Produce debugging information in the operating system's native format
5406 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5409 On most systems that use stabs format, @option{-g} enables use of extra
5410 debugging information that only GDB can use; this extra information
5411 makes debugging work better in GDB but probably makes other debuggers
5413 refuse to read the program. If you want to control for certain whether
5414 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5415 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5417 GCC allows you to use @option{-g} with
5418 @option{-O}. The shortcuts taken by optimized code may occasionally
5419 produce surprising results: some variables you declared may not exist
5420 at all; flow of control may briefly move where you did not expect it;
5421 some statements may not be executed because they compute constant
5422 results or their values are already at hand; some statements may
5423 execute in different places because they have been moved out of loops.
5425 Nevertheless it proves possible to debug optimized output. This makes
5426 it reasonable to use the optimizer for programs that might have bugs.
5428 The following options are useful when GCC is generated with the
5429 capability for more than one debugging format.
5432 @opindex gsplit-dwarf
5433 Separate as much dwarf debugging information as possible into a
5434 separate output file with the extension .dwo. This option allows
5435 the build system to avoid linking files with debug information. To
5436 be useful, this option requires a debugger capable of reading .dwo
5441 Produce debugging information for use by GDB@. This means to use the
5442 most expressive format available (DWARF 2, stabs, or the native format
5443 if neither of those are supported), including GDB extensions if at all
5448 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5450 @item -ggnu-pubnames
5451 @opindex ggnu-pubnames
5452 Generate .debug_pubnames and .debug_pubtypes sections in a format
5453 suitable for conversion into a GDB@ index. This option is only useful
5454 with a linker that can produce GDB@ index version 7.
5458 Produce debugging information in stabs format (if that is supported),
5459 without GDB extensions. This is the format used by DBX on most BSD
5460 systems. On MIPS, Alpha and System V Release 4 systems this option
5461 produces stabs debugging output that is not understood by DBX or SDB@.
5462 On System V Release 4 systems this option requires the GNU assembler.
5464 @item -feliminate-unused-debug-symbols
5465 @opindex feliminate-unused-debug-symbols
5466 Produce debugging information in stabs format (if that is supported),
5467 for only symbols that are actually used.
5469 @item -femit-class-debug-always
5470 @opindex femit-class-debug-always
5471 Instead of emitting debugging information for a C++ class in only one
5472 object file, emit it in all object files using the class. This option
5473 should be used only with debuggers that are unable to handle the way GCC
5474 normally emits debugging information for classes because using this
5475 option increases the size of debugging information by as much as a
5478 @item -fdebug-types-section
5479 @opindex fdebug-types-section
5480 @opindex fno-debug-types-section
5481 When using DWARF Version 4 or higher, type DIEs can be put into
5482 their own @code{.debug_types} section instead of making them part of the
5483 @code{.debug_info} section. It is more efficient to put them in a separate
5484 comdat sections since the linker can then remove duplicates.
5485 But not all DWARF consumers support @code{.debug_types} sections yet
5486 and on some objects @code{.debug_types} produces larger instead of smaller
5487 debugging information.
5491 Produce debugging information in stabs format (if that is supported),
5492 using GNU extensions understood only by the GNU debugger (GDB)@. The
5493 use of these extensions is likely to make other debuggers crash or
5494 refuse to read the program.
5498 Produce debugging information in COFF format (if that is supported).
5499 This is the format used by SDB on most System V systems prior to
5504 Produce debugging information in XCOFF format (if that is supported).
5505 This is the format used by the DBX debugger on IBM RS/6000 systems.
5509 Produce debugging information in XCOFF format (if that is supported),
5510 using GNU extensions understood only by the GNU debugger (GDB)@. The
5511 use of these extensions is likely to make other debuggers crash or
5512 refuse to read the program, and may cause assemblers other than the GNU
5513 assembler (GAS) to fail with an error.
5515 @item -gdwarf-@var{version}
5516 @opindex gdwarf-@var{version}
5517 Produce debugging information in DWARF format (if that is supported).
5518 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5519 for most targets is 4. DWARF Version 5 is only experimental.
5521 Note that with DWARF Version 2, some ports require and always
5522 use some non-conflicting DWARF 3 extensions in the unwind tables.
5524 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5525 for maximum benefit.
5527 @item -grecord-gcc-switches
5528 @opindex grecord-gcc-switches
5529 This switch causes the command-line options used to invoke the
5530 compiler that may affect code generation to be appended to the
5531 DW_AT_producer attribute in DWARF debugging information. The options
5532 are concatenated with spaces separating them from each other and from
5533 the compiler version. See also @option{-frecord-gcc-switches} for another
5534 way of storing compiler options into the object file. This is the default.
5536 @item -gno-record-gcc-switches
5537 @opindex gno-record-gcc-switches
5538 Disallow appending command-line options to the DW_AT_producer attribute
5539 in DWARF debugging information.
5541 @item -gstrict-dwarf
5542 @opindex gstrict-dwarf
5543 Disallow using extensions of later DWARF standard version than selected
5544 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5545 DWARF extensions from later standard versions is allowed.
5547 @item -gno-strict-dwarf
5548 @opindex gno-strict-dwarf
5549 Allow using extensions of later DWARF standard version than selected with
5550 @option{-gdwarf-@var{version}}.
5552 @item -gz@r{[}=@var{type}@r{]}
5554 Produce compressed debug sections in DWARF format, if that is supported.
5555 If @var{type} is not given, the default type depends on the capabilities
5556 of the assembler and linker used. @var{type} may be one of
5557 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5558 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5559 compression in traditional GNU format). If the linker doesn't support
5560 writing compressed debug sections, the option is rejected. Otherwise,
5561 if the assembler does not support them, @option{-gz} is silently ignored
5562 when producing object files.
5566 Produce debugging information in Alpha/VMS debug format (if that is
5567 supported). This is the format used by DEBUG on Alpha/VMS systems.
5570 @itemx -ggdb@var{level}
5571 @itemx -gstabs@var{level}
5572 @itemx -gcoff@var{level}
5573 @itemx -gxcoff@var{level}
5574 @itemx -gvms@var{level}
5575 Request debugging information and also use @var{level} to specify how
5576 much information. The default level is 2.
5578 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5581 Level 1 produces minimal information, enough for making backtraces in
5582 parts of the program that you don't plan to debug. This includes
5583 descriptions of functions and external variables, and line number
5584 tables, but no information about local variables.
5586 Level 3 includes extra information, such as all the macro definitions
5587 present in the program. Some debuggers support macro expansion when
5588 you use @option{-g3}.
5590 @option{-gdwarf-2} does not accept a concatenated debug level, because
5591 GCC used to support an option @option{-gdwarf} that meant to generate
5592 debug information in version 1 of the DWARF format (which is very
5593 different from version 2), and it would have been too confusing. That
5594 debug format is long obsolete, but the option cannot be changed now.
5595 Instead use an additional @option{-g@var{level}} option to change the
5596 debug level for DWARF.
5600 Turn off generation of debug info, if leaving out this option
5601 generates it, or turn it on at level 2 otherwise. The position of this
5602 argument in the command line does not matter; it takes effect after all
5603 other options are processed, and it does so only once, no matter how
5604 many times it is given. This is mainly intended to be used with
5605 @option{-fcompare-debug}.
5607 @item -fsanitize=address
5608 @opindex fsanitize=address
5609 Enable AddressSanitizer, a fast memory error detector.
5610 Memory access instructions are instrumented to detect
5611 out-of-bounds and use-after-free bugs.
5612 See @uref{http://code.google.com/p/address-sanitizer/} for
5613 more details. The run-time behavior can be influenced using the
5614 @env{ASAN_OPTIONS} environment variable; see
5615 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5616 a list of supported options.
5618 @item -fsanitize=kernel-address
5619 @opindex fsanitize=kernel-address
5620 Enable AddressSanitizer for Linux kernel.
5621 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5623 @item -fsanitize=thread
5624 @opindex fsanitize=thread
5625 Enable ThreadSanitizer, a fast data race detector.
5626 Memory access instructions are instrumented to detect
5627 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5628 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5629 environment variable; see
5630 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5633 @item -fsanitize=leak
5634 @opindex fsanitize=leak
5635 Enable LeakSanitizer, a memory leak detector.
5636 This option only matters for linking of executables and if neither
5637 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5638 case the executable is linked against a library that overrides @code{malloc}
5639 and other allocator functions. See
5640 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5641 details. The run-time behavior can be influenced using the
5642 @env{LSAN_OPTIONS} environment variable.
5644 @item -fsanitize=undefined
5645 @opindex fsanitize=undefined
5646 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5647 Various computations are instrumented to detect undefined behavior
5648 at runtime. Current suboptions are:
5652 @item -fsanitize=shift
5653 @opindex fsanitize=shift
5654 This option enables checking that the result of a shift operation is
5655 not undefined. Note that what exactly is considered undefined differs
5656 slightly between C and C++, as well as between ISO C90 and C99, etc.
5658 @item -fsanitize=integer-divide-by-zero
5659 @opindex fsanitize=integer-divide-by-zero
5660 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5662 @item -fsanitize=unreachable
5663 @opindex fsanitize=unreachable
5664 With this option, the compiler turns the @code{__builtin_unreachable}
5665 call into a diagnostics message call instead. When reaching the
5666 @code{__builtin_unreachable} call, the behavior is undefined.
5668 @item -fsanitize=vla-bound
5669 @opindex fsanitize=vla-bound
5670 This option instructs the compiler to check that the size of a variable
5671 length array is positive.
5673 @item -fsanitize=null
5674 @opindex fsanitize=null
5675 This option enables pointer checking. Particularly, the application
5676 built with this option turned on will issue an error message when it
5677 tries to dereference a NULL pointer, or if a reference (possibly an
5678 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5679 on an object pointed by a NULL pointer.
5681 @item -fsanitize=return
5682 @opindex fsanitize=return
5683 This option enables return statement checking. Programs
5684 built with this option turned on will issue an error message
5685 when the end of a non-void function is reached without actually
5686 returning a value. This option works in C++ only.
5688 @item -fsanitize=signed-integer-overflow
5689 @opindex fsanitize=signed-integer-overflow
5690 This option enables signed integer overflow checking. We check that
5691 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5692 does not overflow in the signed arithmetics. Note, integer promotion
5693 rules must be taken into account. That is, the following is not an
5696 signed char a = SCHAR_MAX;
5700 @item -fsanitize=bounds
5701 @opindex fsanitize=bounds
5702 This option enables instrumentation of array bounds. Various out of bounds
5703 accesses are detected. Flexible array members and initializers of variables
5704 with static storage are not instrumented.
5706 @item -fsanitize=alignment
5707 @opindex fsanitize=alignment
5709 This option enables checking of alignment of pointers when they are
5710 dereferenced, or when a reference is bound to insufficiently aligned target,
5711 or when a method or constructor is invoked on insufficiently aligned object.
5713 @item -fsanitize=object-size
5714 @opindex fsanitize=object-size
5715 This option enables instrumentation of memory references using the
5716 @code{__builtin_object_size} function. Various out of bounds pointer
5717 accesses are detected.
5719 @item -fsanitize=float-divide-by-zero
5720 @opindex fsanitize=float-divide-by-zero
5721 Detect floating-point division by zero. Unlike other similar options,
5722 @option{-fsanitize=float-divide-by-zero} is not enabled by
5723 @option{-fsanitize=undefined}, since floating-point division by zero can
5724 be a legitimate way of obtaining infinities and NaNs.
5726 @item -fsanitize=float-cast-overflow
5727 @opindex fsanitize=float-cast-overflow
5728 This option enables floating-point type to integer conversion checking.
5729 We check that the result of the conversion does not overflow.
5730 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
5731 not enabled by @option{-fsanitize=undefined}.
5732 This option does not work well with @code{FE_INVALID} exceptions enabled.
5734 @item -fsanitize=nonnull-attribute
5735 @opindex fsanitize=nonnull-attribute
5737 This option enables instrumentation of calls, checking whether null values
5738 are not passed to arguments marked as requiring a non-null value by the
5739 @code{nonnull} function attribute.
5741 @item -fsanitize=returns-nonnull-attribute
5742 @opindex fsanitize=returns-nonnull-attribute
5744 This option enables instrumentation of return statements in functions
5745 marked with @code{returns_nonnull} function attribute, to detect returning
5746 of null values from such functions.
5748 @item -fsanitize=bool
5749 @opindex fsanitize=bool
5751 This option enables instrumentation of loads from bool. If a value other
5752 than 0/1 is loaded, a run-time error is issued.
5754 @item -fsanitize=enum
5755 @opindex fsanitize=enum
5757 This option enables instrumentation of loads from an enum type. If
5758 a value outside the range of values for the enum type is loaded,
5759 a run-time error is issued.
5761 @item -fsanitize=vptr
5762 @opindex fsanitize=vptr
5764 This option enables instrumentation of C++ member function calls, member
5765 accesses and some conversions between pointers to base and derived classes,
5766 to verify the referenced object has the correct dynamic type.
5770 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5771 @option{-fsanitize=undefined} gives a diagnostic message.
5772 This currently works only for the C family of languages.
5774 @item -fno-sanitize=all
5775 @opindex fno-sanitize=all
5777 This option disables all previously enabled sanitizers.
5778 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
5781 @item -fasan-shadow-offset=@var{number}
5782 @opindex fasan-shadow-offset
5783 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5784 It is useful for experimenting with different shadow memory layouts in
5785 Kernel AddressSanitizer.
5787 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5788 @opindex fsanitize-recover
5789 @opindex fno-sanitize-recover
5790 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5791 mentioned in comma-separated list of @var{opts}. Enabling this option
5792 for a sanitizer component causes it to attempt to continue
5793 running the program as if no error happened. This means multiple
5794 runtime errors can be reported in a single program run, and the exit
5795 code of the program may indicate success even when errors
5796 have been reported. The @option{-fno-sanitize-recover=} can be used to alter
5797 this behavior: only the first detected error is reported
5798 and program then exits with a non-zero exit code.
5800 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5801 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5802 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5803 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5804 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
5805 accepted, the former enables recovery for all sanitizers that support it,
5806 the latter disables recovery for all sanitizers that support it.
5808 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5810 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5813 Similarly @option{-fno-sanitize-recover} is equivalent to
5815 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5818 @item -fsanitize-undefined-trap-on-error
5819 @opindex fsanitize-undefined-trap-on-error
5820 The @option{-fsanitize-undefined-trap-on-error} instructs the compiler to
5821 report undefined behavior using @code{__builtin_trap} rather than
5822 a @code{libubsan} library routine. The advantage of this is that the
5823 @code{libubsan} library is not needed and is not linked in, so this
5824 is usable even in freestanding environments.
5826 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5827 @opindex fdump-final-insns
5828 Dump the final internal representation (RTL) to @var{file}. If the
5829 optional argument is omitted (or if @var{file} is @code{.}), the name
5830 of the dump file is determined by appending @code{.gkd} to the
5831 compilation output file name.
5833 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5834 @opindex fcompare-debug
5835 @opindex fno-compare-debug
5836 If no error occurs during compilation, run the compiler a second time,
5837 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5838 passed to the second compilation. Dump the final internal
5839 representation in both compilations, and print an error if they differ.
5841 If the equal sign is omitted, the default @option{-gtoggle} is used.
5843 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5844 and nonzero, implicitly enables @option{-fcompare-debug}. If
5845 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5846 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5849 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5850 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5851 of the final representation and the second compilation, preventing even
5852 @env{GCC_COMPARE_DEBUG} from taking effect.
5854 To verify full coverage during @option{-fcompare-debug} testing, set
5855 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
5856 which GCC rejects as an invalid option in any actual compilation
5857 (rather than preprocessing, assembly or linking). To get just a
5858 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5859 not overridden} will do.
5861 @item -fcompare-debug-second
5862 @opindex fcompare-debug-second
5863 This option is implicitly passed to the compiler for the second
5864 compilation requested by @option{-fcompare-debug}, along with options to
5865 silence warnings, and omitting other options that would cause
5866 side-effect compiler outputs to files or to the standard output. Dump
5867 files and preserved temporary files are renamed so as to contain the
5868 @code{.gk} additional extension during the second compilation, to avoid
5869 overwriting those generated by the first.
5871 When this option is passed to the compiler driver, it causes the
5872 @emph{first} compilation to be skipped, which makes it useful for little
5873 other than debugging the compiler proper.
5875 @item -feliminate-dwarf2-dups
5876 @opindex feliminate-dwarf2-dups
5877 Compress DWARF 2 debugging information by eliminating duplicated
5878 information about each symbol. This option only makes sense when
5879 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5881 @item -femit-struct-debug-baseonly
5882 @opindex femit-struct-debug-baseonly
5883 Emit debug information for struct-like types
5884 only when the base name of the compilation source file
5885 matches the base name of file in which the struct is defined.
5887 This option substantially reduces the size of debugging information,
5888 but at significant potential loss in type information to the debugger.
5889 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5890 See @option{-femit-struct-debug-detailed} for more detailed control.
5892 This option works only with DWARF 2.
5894 @item -femit-struct-debug-reduced
5895 @opindex femit-struct-debug-reduced
5896 Emit debug information for struct-like types
5897 only when the base name of the compilation source file
5898 matches the base name of file in which the type is defined,
5899 unless the struct is a template or defined in a system header.
5901 This option significantly reduces the size of debugging information,
5902 with some potential loss in type information to the debugger.
5903 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5904 See @option{-femit-struct-debug-detailed} for more detailed control.
5906 This option works only with DWARF 2.
5908 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5909 @opindex femit-struct-debug-detailed
5910 Specify the struct-like types
5911 for which the compiler generates debug information.
5912 The intent is to reduce duplicate struct debug information
5913 between different object files within the same program.
5915 This option is a detailed version of
5916 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5917 which serves for most needs.
5919 A specification has the syntax@*
5920 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5922 The optional first word limits the specification to
5923 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5924 A struct type is used directly when it is the type of a variable, member.
5925 Indirect uses arise through pointers to structs.
5926 That is, when use of an incomplete struct is valid, the use is indirect.
5928 @samp{struct one direct; struct two * indirect;}.
5930 The optional second word limits the specification to
5931 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5932 Generic structs are a bit complicated to explain.
5933 For C++, these are non-explicit specializations of template classes,
5934 or non-template classes within the above.
5935 Other programming languages have generics,
5936 but @option{-femit-struct-debug-detailed} does not yet implement them.
5938 The third word specifies the source files for those
5939 structs for which the compiler should emit debug information.
5940 The values @samp{none} and @samp{any} have the normal meaning.
5941 The value @samp{base} means that
5942 the base of name of the file in which the type declaration appears
5943 must match the base of the name of the main compilation file.
5944 In practice, this means that when compiling @file{foo.c}, debug information
5945 is generated for types declared in that file and @file{foo.h},
5946 but not other header files.
5947 The value @samp{sys} means those types satisfying @samp{base}
5948 or declared in system or compiler headers.
5950 You may need to experiment to determine the best settings for your application.
5952 The default is @option{-femit-struct-debug-detailed=all}.
5954 This option works only with DWARF 2.
5956 @item -fno-merge-debug-strings
5957 @opindex fmerge-debug-strings
5958 @opindex fno-merge-debug-strings
5959 Direct the linker to not merge together strings in the debugging
5960 information that are identical in different object files. Merging is
5961 not supported by all assemblers or linkers. Merging decreases the size
5962 of the debug information in the output file at the cost of increasing
5963 link processing time. Merging is enabled by default.
5965 @item -fdebug-prefix-map=@var{old}=@var{new}
5966 @opindex fdebug-prefix-map
5967 When compiling files in directory @file{@var{old}}, record debugging
5968 information describing them as in @file{@var{new}} instead.
5970 @item -fno-dwarf2-cfi-asm
5971 @opindex fdwarf2-cfi-asm
5972 @opindex fno-dwarf2-cfi-asm
5973 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5974 instead of using GAS @code{.cfi_*} directives.
5976 @cindex @command{prof}
5979 Generate extra code to write profile information suitable for the
5980 analysis program @command{prof}. You must use this option when compiling
5981 the source files you want data about, and you must also use it when
5984 @cindex @command{gprof}
5987 Generate extra code to write profile information suitable for the
5988 analysis program @command{gprof}. You must use this option when compiling
5989 the source files you want data about, and you must also use it when
5994 Makes the compiler print out each function name as it is compiled, and
5995 print some statistics about each pass when it finishes.
5998 @opindex ftime-report
5999 Makes the compiler print some statistics about the time consumed by each
6000 pass when it finishes.
6003 @opindex fmem-report
6004 Makes the compiler print some statistics about permanent memory
6005 allocation when it finishes.
6007 @item -fmem-report-wpa
6008 @opindex fmem-report-wpa
6009 Makes the compiler print some statistics about permanent memory
6010 allocation for the WPA phase only.
6012 @item -fpre-ipa-mem-report
6013 @opindex fpre-ipa-mem-report
6014 @item -fpost-ipa-mem-report
6015 @opindex fpost-ipa-mem-report
6016 Makes the compiler print some statistics about permanent memory
6017 allocation before or after interprocedural optimization.
6019 @item -fprofile-report
6020 @opindex fprofile-report
6021 Makes the compiler print some statistics about consistency of the
6022 (estimated) profile and effect of individual passes.
6025 @opindex fstack-usage
6026 Makes the compiler output stack usage information for the program, on a
6027 per-function basis. The filename for the dump is made by appending
6028 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6029 the output file, if explicitly specified and it is not an executable,
6030 otherwise it is the basename of the source file. An entry is made up
6035 The name of the function.
6039 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6042 The qualifier @code{static} means that the function manipulates the stack
6043 statically: a fixed number of bytes are allocated for the frame on function
6044 entry and released on function exit; no stack adjustments are otherwise made
6045 in the function. The second field is this fixed number of bytes.
6047 The qualifier @code{dynamic} means that the function manipulates the stack
6048 dynamically: in addition to the static allocation described above, stack
6049 adjustments are made in the body of the function, for example to push/pop
6050 arguments around function calls. If the qualifier @code{bounded} is also
6051 present, the amount of these adjustments is bounded at compile time and
6052 the second field is an upper bound of the total amount of stack used by
6053 the function. If it is not present, the amount of these adjustments is
6054 not bounded at compile time and the second field only represents the
6057 @item -fprofile-arcs
6058 @opindex fprofile-arcs
6059 Add code so that program flow @dfn{arcs} are instrumented. During
6060 execution the program records how many times each branch and call is
6061 executed and how many times it is taken or returns. When the compiled
6062 program exits it saves this data to a file called
6063 @file{@var{auxname}.gcda} for each source file. The data may be used for
6064 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6065 test coverage analysis (@option{-ftest-coverage}). Each object file's
6066 @var{auxname} is generated from the name of the output file, if
6067 explicitly specified and it is not the final executable, otherwise it is
6068 the basename of the source file. In both cases any suffix is removed
6069 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6070 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6071 @xref{Cross-profiling}.
6073 @cindex @command{gcov}
6077 This option is used to compile and link code instrumented for coverage
6078 analysis. The option is a synonym for @option{-fprofile-arcs}
6079 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6080 linking). See the documentation for those options for more details.
6085 Compile the source files with @option{-fprofile-arcs} plus optimization
6086 and code generation options. For test coverage analysis, use the
6087 additional @option{-ftest-coverage} option. You do not need to profile
6088 every source file in a program.
6091 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6092 (the latter implies the former).
6095 Run the program on a representative workload to generate the arc profile
6096 information. This may be repeated any number of times. You can run
6097 concurrent instances of your program, and provided that the file system
6098 supports locking, the data files will be correctly updated. Also
6099 @code{fork} calls are detected and correctly handled (double counting
6103 For profile-directed optimizations, compile the source files again with
6104 the same optimization and code generation options plus
6105 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6106 Control Optimization}).
6109 For test coverage analysis, use @command{gcov} to produce human readable
6110 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6111 @command{gcov} documentation for further information.
6115 With @option{-fprofile-arcs}, for each function of your program GCC
6116 creates a program flow graph, then finds a spanning tree for the graph.
6117 Only arcs that are not on the spanning tree have to be instrumented: the
6118 compiler adds code to count the number of times that these arcs are
6119 executed. When an arc is the only exit or only entrance to a block, the
6120 instrumentation code can be added to the block; otherwise, a new basic
6121 block must be created to hold the instrumentation code.
6124 @item -ftest-coverage
6125 @opindex ftest-coverage
6126 Produce a notes file that the @command{gcov} code-coverage utility
6127 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6128 show program coverage. Each source file's note file is called
6129 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6130 above for a description of @var{auxname} and instructions on how to
6131 generate test coverage data. Coverage data matches the source files
6132 more closely if you do not optimize.
6134 @item -fdbg-cnt-list
6135 @opindex fdbg-cnt-list
6136 Print the name and the counter upper bound for all debug counters.
6139 @item -fdbg-cnt=@var{counter-value-list}
6141 Set the internal debug counter upper bound. @var{counter-value-list}
6142 is a comma-separated list of @var{name}:@var{value} pairs
6143 which sets the upper bound of each debug counter @var{name} to @var{value}.
6144 All debug counters have the initial upper bound of @code{UINT_MAX};
6145 thus @code{dbg_cnt} returns true always unless the upper bound
6146 is set by this option.
6147 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6148 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6150 @item -fenable-@var{kind}-@var{pass}
6151 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6155 This is a set of options that are used to explicitly disable/enable
6156 optimization passes. These options are intended for use for debugging GCC.
6157 Compiler users should use regular options for enabling/disabling
6162 @item -fdisable-ipa-@var{pass}
6163 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6164 statically invoked in the compiler multiple times, the pass name should be
6165 appended with a sequential number starting from 1.
6167 @item -fdisable-rtl-@var{pass}
6168 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6169 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6170 statically invoked in the compiler multiple times, the pass name should be
6171 appended with a sequential number starting from 1. @var{range-list} is a
6172 comma-separated list of function ranges or assembler names. Each range is a number
6173 pair separated by a colon. The range is inclusive in both ends. If the range
6174 is trivial, the number pair can be simplified as a single number. If the
6175 function's call graph node's @var{uid} falls within one of the specified ranges,
6176 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6177 function header of a dump file, and the pass names can be dumped by using
6178 option @option{-fdump-passes}.
6180 @item -fdisable-tree-@var{pass}
6181 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6182 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6185 @item -fenable-ipa-@var{pass}
6186 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6187 statically invoked in the compiler multiple times, the pass name should be
6188 appended with a sequential number starting from 1.
6190 @item -fenable-rtl-@var{pass}
6191 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6192 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6193 description and examples.
6195 @item -fenable-tree-@var{pass}
6196 @itemx -fenable-tree-@var{pass}=@var{range-list}
6197 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6198 of option arguments.
6202 Here are some examples showing uses of these options.
6206 # disable ccp1 for all functions
6208 # disable complete unroll for function whose cgraph node uid is 1
6209 -fenable-tree-cunroll=1
6210 # disable gcse2 for functions at the following ranges [1,1],
6211 # [300,400], and [400,1000]
6212 # disable gcse2 for functions foo and foo2
6213 -fdisable-rtl-gcse2=foo,foo2
6214 # disable early inlining
6215 -fdisable-tree-einline
6216 # disable ipa inlining
6217 -fdisable-ipa-inline
6218 # enable tree full unroll
6219 -fenable-tree-unroll
6223 @item -d@var{letters}
6224 @itemx -fdump-rtl-@var{pass}
6225 @itemx -fdump-rtl-@var{pass}=@var{filename}
6227 @opindex fdump-rtl-@var{pass}
6228 Says to make debugging dumps during compilation at times specified by
6229 @var{letters}. This is used for debugging the RTL-based passes of the
6230 compiler. The file names for most of the dumps are made by appending
6231 a pass number and a word to the @var{dumpname}, and the files are
6232 created in the directory of the output file. In case of
6233 @option{=@var{filename}} option, the dump is output on the given file
6234 instead of the pass numbered dump files. Note that the pass number is
6235 computed statically as passes get registered into the pass manager.
6236 Thus the numbering is not related to the dynamic order of execution of
6237 passes. In particular, a pass installed by a plugin could have a
6238 number over 200 even if it executed quite early. @var{dumpname} is
6239 generated from the name of the output file, if explicitly specified
6240 and it is not an executable, otherwise it is the basename of the
6241 source file. These switches may have different effects when
6242 @option{-E} is used for preprocessing.
6244 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6245 @option{-d} option @var{letters}. Here are the possible
6246 letters for use in @var{pass} and @var{letters}, and their meanings:
6250 @item -fdump-rtl-alignments
6251 @opindex fdump-rtl-alignments
6252 Dump after branch alignments have been computed.
6254 @item -fdump-rtl-asmcons
6255 @opindex fdump-rtl-asmcons
6256 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6258 @item -fdump-rtl-auto_inc_dec
6259 @opindex fdump-rtl-auto_inc_dec
6260 Dump after auto-inc-dec discovery. This pass is only run on
6261 architectures that have auto inc or auto dec instructions.
6263 @item -fdump-rtl-barriers
6264 @opindex fdump-rtl-barriers
6265 Dump after cleaning up the barrier instructions.
6267 @item -fdump-rtl-bbpart
6268 @opindex fdump-rtl-bbpart
6269 Dump after partitioning hot and cold basic blocks.
6271 @item -fdump-rtl-bbro
6272 @opindex fdump-rtl-bbro
6273 Dump after block reordering.
6275 @item -fdump-rtl-btl1
6276 @itemx -fdump-rtl-btl2
6277 @opindex fdump-rtl-btl2
6278 @opindex fdump-rtl-btl2
6279 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6280 after the two branch
6281 target load optimization passes.
6283 @item -fdump-rtl-bypass
6284 @opindex fdump-rtl-bypass
6285 Dump after jump bypassing and control flow optimizations.
6287 @item -fdump-rtl-combine
6288 @opindex fdump-rtl-combine
6289 Dump after the RTL instruction combination pass.
6291 @item -fdump-rtl-compgotos
6292 @opindex fdump-rtl-compgotos
6293 Dump after duplicating the computed gotos.
6295 @item -fdump-rtl-ce1
6296 @itemx -fdump-rtl-ce2
6297 @itemx -fdump-rtl-ce3
6298 @opindex fdump-rtl-ce1
6299 @opindex fdump-rtl-ce2
6300 @opindex fdump-rtl-ce3
6301 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6302 @option{-fdump-rtl-ce3} enable dumping after the three
6303 if conversion passes.
6305 @item -fdump-rtl-cprop_hardreg
6306 @opindex fdump-rtl-cprop_hardreg
6307 Dump after hard register copy propagation.
6309 @item -fdump-rtl-csa
6310 @opindex fdump-rtl-csa
6311 Dump after combining stack adjustments.
6313 @item -fdump-rtl-cse1
6314 @itemx -fdump-rtl-cse2
6315 @opindex fdump-rtl-cse1
6316 @opindex fdump-rtl-cse2
6317 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6318 the two common subexpression elimination passes.
6320 @item -fdump-rtl-dce
6321 @opindex fdump-rtl-dce
6322 Dump after the standalone dead code elimination passes.
6324 @item -fdump-rtl-dbr
6325 @opindex fdump-rtl-dbr
6326 Dump after delayed branch scheduling.
6328 @item -fdump-rtl-dce1
6329 @itemx -fdump-rtl-dce2
6330 @opindex fdump-rtl-dce1
6331 @opindex fdump-rtl-dce2
6332 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6333 the two dead store elimination passes.
6336 @opindex fdump-rtl-eh
6337 Dump after finalization of EH handling code.
6339 @item -fdump-rtl-eh_ranges
6340 @opindex fdump-rtl-eh_ranges
6341 Dump after conversion of EH handling range regions.
6343 @item -fdump-rtl-expand
6344 @opindex fdump-rtl-expand
6345 Dump after RTL generation.
6347 @item -fdump-rtl-fwprop1
6348 @itemx -fdump-rtl-fwprop2
6349 @opindex fdump-rtl-fwprop1
6350 @opindex fdump-rtl-fwprop2
6351 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6352 dumping after the two forward propagation passes.
6354 @item -fdump-rtl-gcse1
6355 @itemx -fdump-rtl-gcse2
6356 @opindex fdump-rtl-gcse1
6357 @opindex fdump-rtl-gcse2
6358 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6359 after global common subexpression elimination.
6361 @item -fdump-rtl-init-regs
6362 @opindex fdump-rtl-init-regs
6363 Dump after the initialization of the registers.
6365 @item -fdump-rtl-initvals
6366 @opindex fdump-rtl-initvals
6367 Dump after the computation of the initial value sets.
6369 @item -fdump-rtl-into_cfglayout
6370 @opindex fdump-rtl-into_cfglayout
6371 Dump after converting to cfglayout mode.
6373 @item -fdump-rtl-ira
6374 @opindex fdump-rtl-ira
6375 Dump after iterated register allocation.
6377 @item -fdump-rtl-jump
6378 @opindex fdump-rtl-jump
6379 Dump after the second jump optimization.
6381 @item -fdump-rtl-loop2
6382 @opindex fdump-rtl-loop2
6383 @option{-fdump-rtl-loop2} enables dumping after the rtl
6384 loop optimization passes.
6386 @item -fdump-rtl-mach
6387 @opindex fdump-rtl-mach
6388 Dump after performing the machine dependent reorganization pass, if that
6391 @item -fdump-rtl-mode_sw
6392 @opindex fdump-rtl-mode_sw
6393 Dump after removing redundant mode switches.
6395 @item -fdump-rtl-rnreg
6396 @opindex fdump-rtl-rnreg
6397 Dump after register renumbering.
6399 @item -fdump-rtl-outof_cfglayout
6400 @opindex fdump-rtl-outof_cfglayout
6401 Dump after converting from cfglayout mode.
6403 @item -fdump-rtl-peephole2
6404 @opindex fdump-rtl-peephole2
6405 Dump after the peephole pass.
6407 @item -fdump-rtl-postreload
6408 @opindex fdump-rtl-postreload
6409 Dump after post-reload optimizations.
6411 @item -fdump-rtl-pro_and_epilogue
6412 @opindex fdump-rtl-pro_and_epilogue
6413 Dump after generating the function prologues and epilogues.
6415 @item -fdump-rtl-sched1
6416 @itemx -fdump-rtl-sched2
6417 @opindex fdump-rtl-sched1
6418 @opindex fdump-rtl-sched2
6419 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6420 after the basic block scheduling passes.
6422 @item -fdump-rtl-ree
6423 @opindex fdump-rtl-ree
6424 Dump after sign/zero extension elimination.
6426 @item -fdump-rtl-seqabstr
6427 @opindex fdump-rtl-seqabstr
6428 Dump after common sequence discovery.
6430 @item -fdump-rtl-shorten
6431 @opindex fdump-rtl-shorten
6432 Dump after shortening branches.
6434 @item -fdump-rtl-sibling
6435 @opindex fdump-rtl-sibling
6436 Dump after sibling call optimizations.
6438 @item -fdump-rtl-split1
6439 @itemx -fdump-rtl-split2
6440 @itemx -fdump-rtl-split3
6441 @itemx -fdump-rtl-split4
6442 @itemx -fdump-rtl-split5
6443 @opindex fdump-rtl-split1
6444 @opindex fdump-rtl-split2
6445 @opindex fdump-rtl-split3
6446 @opindex fdump-rtl-split4
6447 @opindex fdump-rtl-split5
6448 These options enable dumping after five rounds of
6449 instruction splitting.
6451 @item -fdump-rtl-sms
6452 @opindex fdump-rtl-sms
6453 Dump after modulo scheduling. This pass is only run on some
6456 @item -fdump-rtl-stack
6457 @opindex fdump-rtl-stack
6458 Dump after conversion from GCC's ``flat register file'' registers to the
6459 x87's stack-like registers. This pass is only run on x86 variants.
6461 @item -fdump-rtl-subreg1
6462 @itemx -fdump-rtl-subreg2
6463 @opindex fdump-rtl-subreg1
6464 @opindex fdump-rtl-subreg2
6465 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6466 the two subreg expansion passes.
6468 @item -fdump-rtl-unshare
6469 @opindex fdump-rtl-unshare
6470 Dump after all rtl has been unshared.
6472 @item -fdump-rtl-vartrack
6473 @opindex fdump-rtl-vartrack
6474 Dump after variable tracking.
6476 @item -fdump-rtl-vregs
6477 @opindex fdump-rtl-vregs
6478 Dump after converting virtual registers to hard registers.
6480 @item -fdump-rtl-web
6481 @opindex fdump-rtl-web
6482 Dump after live range splitting.
6484 @item -fdump-rtl-regclass
6485 @itemx -fdump-rtl-subregs_of_mode_init
6486 @itemx -fdump-rtl-subregs_of_mode_finish
6487 @itemx -fdump-rtl-dfinit
6488 @itemx -fdump-rtl-dfinish
6489 @opindex fdump-rtl-regclass
6490 @opindex fdump-rtl-subregs_of_mode_init
6491 @opindex fdump-rtl-subregs_of_mode_finish
6492 @opindex fdump-rtl-dfinit
6493 @opindex fdump-rtl-dfinish
6494 These dumps are defined but always produce empty files.
6497 @itemx -fdump-rtl-all
6499 @opindex fdump-rtl-all
6500 Produce all the dumps listed above.
6504 Annotate the assembler output with miscellaneous debugging information.
6508 Dump all macro definitions, at the end of preprocessing, in addition to
6513 Produce a core dump whenever an error occurs.
6517 Annotate the assembler output with a comment indicating which
6518 pattern and alternative is used. The length of each instruction is
6523 Dump the RTL in the assembler output as a comment before each instruction.
6524 Also turns on @option{-dp} annotation.
6528 Just generate RTL for a function instead of compiling it. Usually used
6529 with @option{-fdump-rtl-expand}.
6533 @opindex fdump-noaddr
6534 When doing debugging dumps, suppress address output. This makes it more
6535 feasible to use diff on debugging dumps for compiler invocations with
6536 different compiler binaries and/or different
6537 text / bss / data / heap / stack / dso start locations.
6540 @opindex freport-bug
6541 Collect and dump debug information into temporary file if ICE in C/C++
6544 @item -fdump-unnumbered
6545 @opindex fdump-unnumbered
6546 When doing debugging dumps, suppress instruction numbers and address output.
6547 This makes it more feasible to use diff on debugging dumps for compiler
6548 invocations with different options, in particular with and without
6551 @item -fdump-unnumbered-links
6552 @opindex fdump-unnumbered-links
6553 When doing debugging dumps (see @option{-d} option above), suppress
6554 instruction numbers for the links to the previous and next instructions
6557 @item -fdump-translation-unit @r{(C++ only)}
6558 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6559 @opindex fdump-translation-unit
6560 Dump a representation of the tree structure for the entire translation
6561 unit to a file. The file name is made by appending @file{.tu} to the
6562 source file name, and the file is created in the same directory as the
6563 output file. If the @samp{-@var{options}} form is used, @var{options}
6564 controls the details of the dump as described for the
6565 @option{-fdump-tree} options.
6567 @item -fdump-class-hierarchy @r{(C++ only)}
6568 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6569 @opindex fdump-class-hierarchy
6570 Dump a representation of each class's hierarchy and virtual function
6571 table layout to a file. The file name is made by appending
6572 @file{.class} to the source file name, and the file is created in the
6573 same directory as the output file. If the @samp{-@var{options}} form
6574 is used, @var{options} controls the details of the dump as described
6575 for the @option{-fdump-tree} options.
6577 @item -fdump-ipa-@var{switch}
6579 Control the dumping at various stages of inter-procedural analysis
6580 language tree to a file. The file name is generated by appending a
6581 switch specific suffix to the source file name, and the file is created
6582 in the same directory as the output file. The following dumps are
6587 Enables all inter-procedural analysis dumps.
6590 Dumps information about call-graph optimization, unused function removal,
6591 and inlining decisions.
6594 Dump after function inlining.
6599 @opindex fdump-passes
6600 Dump the list of optimization passes that are turned on and off by
6601 the current command-line options.
6603 @item -fdump-statistics-@var{option}
6604 @opindex fdump-statistics
6605 Enable and control dumping of pass statistics in a separate file. The
6606 file name is generated by appending a suffix ending in
6607 @samp{.statistics} to the source file name, and the file is created in
6608 the same directory as the output file. If the @samp{-@var{option}}
6609 form is used, @samp{-stats} causes counters to be summed over the
6610 whole compilation unit while @samp{-details} dumps every event as
6611 the passes generate them. The default with no option is to sum
6612 counters for each function compiled.
6614 @item -fdump-tree-@var{switch}
6615 @itemx -fdump-tree-@var{switch}-@var{options}
6616 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6618 Control the dumping at various stages of processing the intermediate
6619 language tree to a file. The file name is generated by appending a
6620 switch-specific suffix to the source file name, and the file is
6621 created in the same directory as the output file. In case of
6622 @option{=@var{filename}} option, the dump is output on the given file
6623 instead of the auto named dump files. If the @samp{-@var{options}}
6624 form is used, @var{options} is a list of @samp{-} separated options
6625 which control the details of the dump. Not all options are applicable
6626 to all dumps; those that are not meaningful are ignored. The
6627 following options are available
6631 Print the address of each node. Usually this is not meaningful as it
6632 changes according to the environment and source file. Its primary use
6633 is for tying up a dump file with a debug environment.
6635 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6636 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6637 use working backward from mangled names in the assembly file.
6639 When dumping front-end intermediate representations, inhibit dumping
6640 of members of a scope or body of a function merely because that scope
6641 has been reached. Only dump such items when they are directly reachable
6644 When dumping pretty-printed trees, this option inhibits dumping the
6645 bodies of control structures.
6647 When dumping RTL, print the RTL in slim (condensed) form instead of
6648 the default LISP-like representation.
6650 Print a raw representation of the tree. By default, trees are
6651 pretty-printed into a C-like representation.
6653 Enable more detailed dumps (not honored by every dump option). Also
6654 include information from the optimization passes.
6656 Enable dumping various statistics about the pass (not honored by every dump
6659 Enable showing basic block boundaries (disabled in raw dumps).
6661 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6662 dump a representation of the control flow graph suitable for viewing with
6663 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6664 the file is pretty-printed as a subgraph, so that GraphViz can render them
6665 all in a single plot.
6667 This option currently only works for RTL dumps, and the RTL is always
6668 dumped in slim form.
6670 Enable showing virtual operands for every statement.
6672 Enable showing line numbers for statements.
6674 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6676 Enable showing the tree dump for each statement.
6678 Enable showing the EH region number holding each statement.
6680 Enable showing scalar evolution analysis details.
6682 Enable showing optimization information (only available in certain
6685 Enable showing missed optimization information (only available in certain
6688 Enable other detailed optimization information (only available in
6690 @item =@var{filename}
6691 Instead of an auto named dump file, output into the given file
6692 name. The file names @file{stdout} and @file{stderr} are treated
6693 specially and are considered already open standard streams. For
6697 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6698 -fdump-tree-pre=stderr file.c
6701 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6702 output on to @file{stderr}. If two conflicting dump filenames are
6703 given for the same pass, then the latter option overrides the earlier
6707 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6708 and @option{lineno}.
6711 Turn on all optimization options, i.e., @option{optimized},
6712 @option{missed}, and @option{note}.
6715 The following tree dumps are possible:
6719 @opindex fdump-tree-original
6720 Dump before any tree based optimization, to @file{@var{file}.original}.
6723 @opindex fdump-tree-optimized
6724 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6727 @opindex fdump-tree-gimple
6728 Dump each function before and after the gimplification pass to a file. The
6729 file name is made by appending @file{.gimple} to the source file name.
6732 @opindex fdump-tree-cfg
6733 Dump the control flow graph of each function to a file. The file name is
6734 made by appending @file{.cfg} to the source file name.
6737 @opindex fdump-tree-ch
6738 Dump each function after copying loop headers. The file name is made by
6739 appending @file{.ch} to the source file name.
6742 @opindex fdump-tree-ssa
6743 Dump SSA related information to a file. The file name is made by appending
6744 @file{.ssa} to the source file name.
6747 @opindex fdump-tree-alias
6748 Dump aliasing information for each function. The file name is made by
6749 appending @file{.alias} to the source file name.
6752 @opindex fdump-tree-ccp
6753 Dump each function after CCP@. The file name is made by appending
6754 @file{.ccp} to the source file name.
6757 @opindex fdump-tree-storeccp
6758 Dump each function after STORE-CCP@. The file name is made by appending
6759 @file{.storeccp} to the source file name.
6762 @opindex fdump-tree-pre
6763 Dump trees after partial redundancy elimination. The file name is made
6764 by appending @file{.pre} to the source file name.
6767 @opindex fdump-tree-fre
6768 Dump trees after full redundancy elimination. The file name is made
6769 by appending @file{.fre} to the source file name.
6772 @opindex fdump-tree-copyprop
6773 Dump trees after copy propagation. The file name is made
6774 by appending @file{.copyprop} to the source file name.
6776 @item store_copyprop
6777 @opindex fdump-tree-store_copyprop
6778 Dump trees after store copy-propagation. The file name is made
6779 by appending @file{.store_copyprop} to the source file name.
6782 @opindex fdump-tree-dce
6783 Dump each function after dead code elimination. The file name is made by
6784 appending @file{.dce} to the source file name.
6787 @opindex fdump-tree-sra
6788 Dump each function after performing scalar replacement of aggregates. The
6789 file name is made by appending @file{.sra} to the source file name.
6792 @opindex fdump-tree-sink
6793 Dump each function after performing code sinking. The file name is made
6794 by appending @file{.sink} to the source file name.
6797 @opindex fdump-tree-dom
6798 Dump each function after applying dominator tree optimizations. The file
6799 name is made by appending @file{.dom} to the source file name.
6802 @opindex fdump-tree-dse
6803 Dump each function after applying dead store elimination. The file
6804 name is made by appending @file{.dse} to the source file name.
6807 @opindex fdump-tree-phiopt
6808 Dump each function after optimizing PHI nodes into straightline code. The file
6809 name is made by appending @file{.phiopt} to the source file name.
6812 @opindex fdump-tree-forwprop
6813 Dump each function after forward propagating single use variables. The file
6814 name is made by appending @file{.forwprop} to the source file name.
6817 @opindex fdump-tree-copyrename
6818 Dump each function after applying the copy rename optimization. The file
6819 name is made by appending @file{.copyrename} to the source file name.
6822 @opindex fdump-tree-nrv
6823 Dump each function after applying the named return value optimization on
6824 generic trees. The file name is made by appending @file{.nrv} to the source
6828 @opindex fdump-tree-vect
6829 Dump each function after applying vectorization of loops. The file name is
6830 made by appending @file{.vect} to the source file name.
6833 @opindex fdump-tree-slp
6834 Dump each function after applying vectorization of basic blocks. The file name
6835 is made by appending @file{.slp} to the source file name.
6838 @opindex fdump-tree-vrp
6839 Dump each function after Value Range Propagation (VRP). The file name
6840 is made by appending @file{.vrp} to the source file name.
6843 @opindex fdump-tree-all
6844 Enable all the available tree dumps with the flags provided in this option.
6848 @itemx -fopt-info-@var{options}
6849 @itemx -fopt-info-@var{options}=@var{filename}
6851 Controls optimization dumps from various optimization passes. If the
6852 @samp{-@var{options}} form is used, @var{options} is a list of
6853 @samp{-} separated option keywords to select the dump details and
6856 The @var{options} can be divided into two groups: options describing the
6857 verbosity of the dump, and options describing which optimizations
6858 should be included. The options from both the groups can be freely
6859 mixed as they are non-overlapping. However, in case of any conflicts,
6860 the later options override the earlier options on the command
6863 The following options control the dump verbosity:
6867 Print information when an optimization is successfully applied. It is
6868 up to a pass to decide which information is relevant. For example, the
6869 vectorizer passes print the source location of loops which are
6870 successfully vectorized.
6872 Print information about missed optimizations. Individual passes
6873 control which information to include in the output.
6875 Print verbose information about optimizations, such as certain
6876 transformations, more detailed messages about decisions etc.
6878 Print detailed optimization information. This includes
6879 @samp{optimized}, @samp{missed}, and @samp{note}.
6882 One or more of the following option keywords can be used to describe a
6883 group of optimizations:
6887 Enable dumps from all interprocedural optimizations.
6889 Enable dumps from all loop optimizations.
6891 Enable dumps from all inlining optimizations.
6893 Enable dumps from all vectorization optimizations.
6895 Enable dumps from all optimizations. This is a superset of
6896 the optimization groups listed above.
6900 omitted, it defaults to @samp{optimized-optall}, which means to dump all
6901 info about successful optimizations from all the passes.
6903 If the @var{filename} is provided, then the dumps from all the
6904 applicable optimizations are concatenated into the @var{filename}.
6905 Otherwise the dump is output onto @file{stderr}. Though multiple
6906 @option{-fopt-info} options are accepted, only one of them can include
6907 a @var{filename}. If other filenames are provided then all but the
6908 first such option are ignored.
6910 Note that the output @var{filename} is overwritten
6911 in case of multiple translation units. If a combined output from
6912 multiple translation units is desired, @file{stderr} should be used
6915 In the following example, the optimization info is output to
6924 gcc -O3 -fopt-info-missed=missed.all
6928 outputs missed optimization report from all the passes into
6929 @file{missed.all}, and this one:
6932 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
6936 prints information about missed optimization opportunities from
6937 vectorization passes on @file{stderr}.
6938 Note that @option{-fopt-info-vec-missed} is equivalent to
6939 @option{-fopt-info-missed-vec}.
6943 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
6947 outputs information about missed optimizations as well as
6948 optimized locations from all the inlining passes into
6954 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
6958 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
6959 in conflict since only one output file is allowed. In this case, only
6960 the first option takes effect and the subsequent options are
6961 ignored. Thus only @file{vec.miss} is produced which contains
6962 dumps from the vectorizer about missed opportunities.
6964 @item -frandom-seed=@var{number}
6965 @opindex frandom-seed
6966 This option provides a seed that GCC uses in place of
6967 random numbers in generating certain symbol names
6968 that have to be different in every compiled file. It is also used to
6969 place unique stamps in coverage data files and the object files that
6970 produce them. You can use the @option{-frandom-seed} option to produce
6971 reproducibly identical object files.
6973 The @var{number} should be different for every file you compile.
6975 @item -fsched-verbose=@var{n}
6976 @opindex fsched-verbose
6977 On targets that use instruction scheduling, this option controls the
6978 amount of debugging output the scheduler prints. This information is
6979 written to standard error, unless @option{-fdump-rtl-sched1} or
6980 @option{-fdump-rtl-sched2} is specified, in which case it is output
6981 to the usual dump listing file, @file{.sched1} or @file{.sched2}
6982 respectively. However for @var{n} greater than nine, the output is
6983 always printed to standard error.
6985 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
6986 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
6987 For @var{n} greater than one, it also output basic block probabilities,
6988 detailed ready list information and unit/insn info. For @var{n} greater
6989 than two, it includes RTL at abort point, control-flow and regions info.
6990 And for @var{n} over four, @option{-fsched-verbose} also includes
6994 @itemx -save-temps=cwd
6996 Store the usual ``temporary'' intermediate files permanently; place them
6997 in the current directory and name them based on the source file. Thus,
6998 compiling @file{foo.c} with @option{-c -save-temps} produces files
6999 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7000 preprocessed @file{foo.i} output file even though the compiler now
7001 normally uses an integrated preprocessor.
7003 When used in combination with the @option{-x} command-line option,
7004 @option{-save-temps} is sensible enough to avoid over writing an
7005 input source file with the same extension as an intermediate file.
7006 The corresponding intermediate file may be obtained by renaming the
7007 source file before using @option{-save-temps}.
7009 If you invoke GCC in parallel, compiling several different source
7010 files that share a common base name in different subdirectories or the
7011 same source file compiled for multiple output destinations, it is
7012 likely that the different parallel compilers will interfere with each
7013 other, and overwrite the temporary files. For instance:
7016 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7017 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7020 may result in @file{foo.i} and @file{foo.o} being written to
7021 simultaneously by both compilers.
7023 @item -save-temps=obj
7024 @opindex save-temps=obj
7025 Store the usual ``temporary'' intermediate files permanently. If the
7026 @option{-o} option is used, the temporary files are based on the
7027 object file. If the @option{-o} option is not used, the
7028 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7033 gcc -save-temps=obj -c foo.c
7034 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7035 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7039 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7040 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7041 @file{dir2/yfoobar.o}.
7043 @item -time@r{[}=@var{file}@r{]}
7045 Report the CPU time taken by each subprocess in the compilation
7046 sequence. For C source files, this is the compiler proper and assembler
7047 (plus the linker if linking is done).
7049 Without the specification of an output file, the output looks like this:
7056 The first number on each line is the ``user time'', that is time spent
7057 executing the program itself. The second number is ``system time'',
7058 time spent executing operating system routines on behalf of the program.
7059 Both numbers are in seconds.
7061 With the specification of an output file, the output is appended to the
7062 named file, and it looks like this:
7065 0.12 0.01 cc1 @var{options}
7066 0.00 0.01 as @var{options}
7069 The ``user time'' and the ``system time'' are moved before the program
7070 name, and the options passed to the program are displayed, so that one
7071 can later tell what file was being compiled, and with which options.
7073 @item -fvar-tracking
7074 @opindex fvar-tracking
7075 Run variable tracking pass. It computes where variables are stored at each
7076 position in code. Better debugging information is then generated
7077 (if the debugging information format supports this information).
7079 It is enabled by default when compiling with optimization (@option{-Os},
7080 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7081 the debug info format supports it.
7083 @item -fvar-tracking-assignments
7084 @opindex fvar-tracking-assignments
7085 @opindex fno-var-tracking-assignments
7086 Annotate assignments to user variables early in the compilation and
7087 attempt to carry the annotations over throughout the compilation all the
7088 way to the end, in an attempt to improve debug information while
7089 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7091 It can be enabled even if var-tracking is disabled, in which case
7092 annotations are created and maintained, but discarded at the end.
7093 By default, this flag is enabled together with @option{-fvar-tracking},
7094 except when selective scheduling is enabled.
7096 @item -fvar-tracking-assignments-toggle
7097 @opindex fvar-tracking-assignments-toggle
7098 @opindex fno-var-tracking-assignments-toggle
7099 Toggle @option{-fvar-tracking-assignments}, in the same way that
7100 @option{-gtoggle} toggles @option{-g}.
7102 @item -print-file-name=@var{library}
7103 @opindex print-file-name
7104 Print the full absolute name of the library file @var{library} that
7105 would be used when linking---and don't do anything else. With this
7106 option, GCC does not compile or link anything; it just prints the
7109 @item -print-multi-directory
7110 @opindex print-multi-directory
7111 Print the directory name corresponding to the multilib selected by any
7112 other switches present in the command line. This directory is supposed
7113 to exist in @env{GCC_EXEC_PREFIX}.
7115 @item -print-multi-lib
7116 @opindex print-multi-lib
7117 Print the mapping from multilib directory names to compiler switches
7118 that enable them. The directory name is separated from the switches by
7119 @samp{;}, and each switch starts with an @samp{@@} instead of the
7120 @samp{-}, without spaces between multiple switches. This is supposed to
7121 ease shell processing.
7123 @item -print-multi-os-directory
7124 @opindex print-multi-os-directory
7125 Print the path to OS libraries for the selected
7126 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7127 present in the @file{lib} subdirectory and no multilibs are used, this is
7128 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7129 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7130 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7131 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7133 @item -print-multiarch
7134 @opindex print-multiarch
7135 Print the path to OS libraries for the selected multiarch,
7136 relative to some @file{lib} subdirectory.
7138 @item -print-prog-name=@var{program}
7139 @opindex print-prog-name
7140 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7142 @item -print-libgcc-file-name
7143 @opindex print-libgcc-file-name
7144 Same as @option{-print-file-name=libgcc.a}.
7146 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7147 but you do want to link with @file{libgcc.a}. You can do:
7150 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7153 @item -print-search-dirs
7154 @opindex print-search-dirs
7155 Print the name of the configured installation directory and a list of
7156 program and library directories @command{gcc} searches---and don't do anything else.
7158 This is useful when @command{gcc} prints the error message
7159 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7160 To resolve this you either need to put @file{cpp0} and the other compiler
7161 components where @command{gcc} expects to find them, or you can set the environment
7162 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7163 Don't forget the trailing @samp{/}.
7164 @xref{Environment Variables}.
7166 @item -print-sysroot
7167 @opindex print-sysroot
7168 Print the target sysroot directory that is used during
7169 compilation. This is the target sysroot specified either at configure
7170 time or using the @option{--sysroot} option, possibly with an extra
7171 suffix that depends on compilation options. If no target sysroot is
7172 specified, the option prints nothing.
7174 @item -print-sysroot-headers-suffix
7175 @opindex print-sysroot-headers-suffix
7176 Print the suffix added to the target sysroot when searching for
7177 headers, or give an error if the compiler is not configured with such
7178 a suffix---and don't do anything else.
7181 @opindex dumpmachine
7182 Print the compiler's target machine (for example,
7183 @samp{i686-pc-linux-gnu})---and don't do anything else.
7186 @opindex dumpversion
7187 Print the compiler version (for example, @code{3.0})---and don't do
7192 Print the compiler's built-in specs---and don't do anything else. (This
7193 is used when GCC itself is being built.) @xref{Spec Files}.
7195 @item -fno-eliminate-unused-debug-types
7196 @opindex feliminate-unused-debug-types
7197 @opindex fno-eliminate-unused-debug-types
7198 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7199 output for types that are nowhere used in the source file being compiled.
7200 Sometimes it is useful to have GCC emit debugging
7201 information for all types declared in a compilation
7202 unit, regardless of whether or not they are actually used
7203 in that compilation unit, for example
7204 if, in the debugger, you want to cast a value to a type that is
7205 not actually used in your program (but is declared). More often,
7206 however, this results in a significant amount of wasted space.
7209 @node Optimize Options
7210 @section Options That Control Optimization
7211 @cindex optimize options
7212 @cindex options, optimization
7214 These options control various sorts of optimizations.
7216 Without any optimization option, the compiler's goal is to reduce the
7217 cost of compilation and to make debugging produce the expected
7218 results. Statements are independent: if you stop the program with a
7219 breakpoint between statements, you can then assign a new value to any
7220 variable or change the program counter to any other statement in the
7221 function and get exactly the results you expect from the source
7224 Turning on optimization flags makes the compiler attempt to improve
7225 the performance and/or code size at the expense of compilation time
7226 and possibly the ability to debug the program.
7228 The compiler performs optimization based on the knowledge it has of the
7229 program. Compiling multiple files at once to a single output file mode allows
7230 the compiler to use information gained from all of the files when compiling
7233 Not all optimizations are controlled directly by a flag. Only
7234 optimizations that have a flag are listed in this section.
7236 Most optimizations are only enabled if an @option{-O} level is set on
7237 the command line. Otherwise they are disabled, even if individual
7238 optimization flags are specified.
7240 Depending on the target and how GCC was configured, a slightly different
7241 set of optimizations may be enabled at each @option{-O} level than
7242 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7243 to find out the exact set of optimizations that are enabled at each level.
7244 @xref{Overall Options}, for examples.
7251 Optimize. Optimizing compilation takes somewhat more time, and a lot
7252 more memory for a large function.
7254 With @option{-O}, the compiler tries to reduce code size and execution
7255 time, without performing any optimizations that take a great deal of
7258 @option{-O} turns on the following optimization flags:
7261 -fbranch-count-reg @gol
7262 -fcombine-stack-adjustments @gol
7264 -fcprop-registers @gol
7267 -fdelayed-branch @gol
7269 -fforward-propagate @gol
7270 -fguess-branch-probability @gol
7271 -fif-conversion2 @gol
7272 -fif-conversion @gol
7273 -finline-functions-called-once @gol
7274 -fipa-pure-const @gol
7276 -fipa-reference @gol
7277 -fmerge-constants @gol
7278 -fmove-loop-invariants @gol
7280 -fsplit-wide-types @gol
7285 -ftree-copy-prop @gol
7286 -ftree-copyrename @gol
7288 -ftree-dominator-opts @gol
7290 -ftree-forwprop @gol
7300 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7301 where doing so does not interfere with debugging.
7305 Optimize even more. GCC performs nearly all supported optimizations
7306 that do not involve a space-speed tradeoff.
7307 As compared to @option{-O}, this option increases both compilation time
7308 and the performance of the generated code.
7310 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7311 also turns on the following optimization flags:
7312 @gccoptlist{-fthread-jumps @gol
7313 -falign-functions -falign-jumps @gol
7314 -falign-loops -falign-labels @gol
7317 -fcse-follow-jumps -fcse-skip-blocks @gol
7318 -fdelete-null-pointer-checks @gol
7319 -fdevirtualize -fdevirtualize-speculatively @gol
7320 -fexpensive-optimizations @gol
7321 -fgcse -fgcse-lm @gol
7322 -fhoist-adjacent-loads @gol
7323 -finline-small-functions @gol
7324 -findirect-inlining @gol
7328 -fisolate-erroneous-paths-dereference @gol
7330 -foptimize-sibling-calls @gol
7331 -foptimize-strlen @gol
7332 -fpartial-inlining @gol
7334 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7335 -frerun-cse-after-loop @gol
7336 -fsched-interblock -fsched-spec @gol
7337 -fschedule-insns -fschedule-insns2 @gol
7338 -fstrict-aliasing -fstrict-overflow @gol
7339 -ftree-builtin-call-dce @gol
7340 -ftree-switch-conversion -ftree-tail-merge @gol
7345 Please note the warning under @option{-fgcse} about
7346 invoking @option{-O2} on programs that use computed gotos.
7350 Optimize yet more. @option{-O3} turns on all optimizations specified
7351 by @option{-O2} and also turns on the @option{-finline-functions},
7352 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7353 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7354 @option{-ftree-loop-distribute-patterns},
7355 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7356 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7360 Reduce compilation time and make debugging produce the expected
7361 results. This is the default.
7365 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7366 do not typically increase code size. It also performs further
7367 optimizations designed to reduce code size.
7369 @option{-Os} disables the following optimization flags:
7370 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7371 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7372 -fprefetch-loop-arrays}
7376 Disregard strict standards compliance. @option{-Ofast} enables all
7377 @option{-O3} optimizations. It also enables optimizations that are not
7378 valid for all standard-compliant programs.
7379 It turns on @option{-ffast-math} and the Fortran-specific
7380 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7384 Optimize debugging experience. @option{-Og} enables optimizations
7385 that do not interfere with debugging. It should be the optimization
7386 level of choice for the standard edit-compile-debug cycle, offering
7387 a reasonable level of optimization while maintaining fast compilation
7388 and a good debugging experience.
7390 If you use multiple @option{-O} options, with or without level numbers,
7391 the last such option is the one that is effective.
7394 Options of the form @option{-f@var{flag}} specify machine-independent
7395 flags. Most flags have both positive and negative forms; the negative
7396 form of @option{-ffoo} is @option{-fno-foo}. In the table
7397 below, only one of the forms is listed---the one you typically
7398 use. You can figure out the other form by either removing @samp{no-}
7401 The following options control specific optimizations. They are either
7402 activated by @option{-O} options or are related to ones that are. You
7403 can use the following flags in the rare cases when ``fine-tuning'' of
7404 optimizations to be performed is desired.
7407 @item -fno-defer-pop
7408 @opindex fno-defer-pop
7409 Always pop the arguments to each function call as soon as that function
7410 returns. For machines that must pop arguments after a function call,
7411 the compiler normally lets arguments accumulate on the stack for several
7412 function calls and pops them all at once.
7414 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7416 @item -fforward-propagate
7417 @opindex fforward-propagate
7418 Perform a forward propagation pass on RTL@. The pass tries to combine two
7419 instructions and checks if the result can be simplified. If loop unrolling
7420 is active, two passes are performed and the second is scheduled after
7423 This option is enabled by default at optimization levels @option{-O},
7424 @option{-O2}, @option{-O3}, @option{-Os}.
7426 @item -ffp-contract=@var{style}
7427 @opindex ffp-contract
7428 @option{-ffp-contract=off} disables floating-point expression contraction.
7429 @option{-ffp-contract=fast} enables floating-point expression contraction
7430 such as forming of fused multiply-add operations if the target has
7431 native support for them.
7432 @option{-ffp-contract=on} enables floating-point expression contraction
7433 if allowed by the language standard. This is currently not implemented
7434 and treated equal to @option{-ffp-contract=off}.
7436 The default is @option{-ffp-contract=fast}.
7438 @item -fomit-frame-pointer
7439 @opindex fomit-frame-pointer
7440 Don't keep the frame pointer in a register for functions that
7441 don't need one. This avoids the instructions to save, set up and
7442 restore frame pointers; it also makes an extra register available
7443 in many functions. @strong{It also makes debugging impossible on
7446 On some machines, such as the VAX, this flag has no effect, because
7447 the standard calling sequence automatically handles the frame pointer
7448 and nothing is saved by pretending it doesn't exist. The
7449 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7450 whether a target machine supports this flag. @xref{Registers,,Register
7451 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7453 Starting with GCC version 4.6, the default setting (when not optimizing for
7454 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
7455 @option{-fomit-frame-pointer}. The default can be reverted to
7456 @option{-fno-omit-frame-pointer} by configuring GCC with the
7457 @option{--enable-frame-pointer} configure option.
7459 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7461 @item -foptimize-sibling-calls
7462 @opindex foptimize-sibling-calls
7463 Optimize sibling and tail recursive calls.
7465 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7467 @item -foptimize-strlen
7468 @opindex foptimize-strlen
7469 Optimize various standard C string functions (e.g. @code{strlen},
7470 @code{strchr} or @code{strcpy}) and
7471 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7473 Enabled at levels @option{-O2}, @option{-O3}.
7477 Do not expand any functions inline apart from those marked with
7478 the @code{always_inline} attribute. This is the default when not
7481 Single functions can be exempted from inlining by marking them
7482 with the @code{noinline} attribute.
7484 @item -finline-small-functions
7485 @opindex finline-small-functions
7486 Integrate functions into their callers when their body is smaller than expected
7487 function call code (so overall size of program gets smaller). The compiler
7488 heuristically decides which functions are simple enough to be worth integrating
7489 in this way. This inlining applies to all functions, even those not declared
7492 Enabled at level @option{-O2}.
7494 @item -findirect-inlining
7495 @opindex findirect-inlining
7496 Inline also indirect calls that are discovered to be known at compile
7497 time thanks to previous inlining. This option has any effect only
7498 when inlining itself is turned on by the @option{-finline-functions}
7499 or @option{-finline-small-functions} options.
7501 Enabled at level @option{-O2}.
7503 @item -finline-functions
7504 @opindex finline-functions
7505 Consider all functions for inlining, even if they are not declared inline.
7506 The compiler heuristically decides which functions are worth integrating
7509 If all calls to a given function are integrated, and the function is
7510 declared @code{static}, then the function is normally not output as
7511 assembler code in its own right.
7513 Enabled at level @option{-O3}.
7515 @item -finline-functions-called-once
7516 @opindex finline-functions-called-once
7517 Consider all @code{static} functions called once for inlining into their
7518 caller even if they are not marked @code{inline}. If a call to a given
7519 function is integrated, then the function is not output as assembler code
7522 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7524 @item -fearly-inlining
7525 @opindex fearly-inlining
7526 Inline functions marked by @code{always_inline} and functions whose body seems
7527 smaller than the function call overhead early before doing
7528 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7529 makes profiling significantly cheaper and usually inlining faster on programs
7530 having large chains of nested wrapper functions.
7536 Perform interprocedural scalar replacement of aggregates, removal of
7537 unused parameters and replacement of parameters passed by reference
7538 by parameters passed by value.
7540 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7542 @item -finline-limit=@var{n}
7543 @opindex finline-limit
7544 By default, GCC limits the size of functions that can be inlined. This flag
7545 allows coarse control of this limit. @var{n} is the size of functions that
7546 can be inlined in number of pseudo instructions.
7548 Inlining is actually controlled by a number of parameters, which may be
7549 specified individually by using @option{--param @var{name}=@var{value}}.
7550 The @option{-finline-limit=@var{n}} option sets some of these parameters
7554 @item max-inline-insns-single
7555 is set to @var{n}/2.
7556 @item max-inline-insns-auto
7557 is set to @var{n}/2.
7560 See below for a documentation of the individual
7561 parameters controlling inlining and for the defaults of these parameters.
7563 @emph{Note:} there may be no value to @option{-finline-limit} that results
7564 in default behavior.
7566 @emph{Note:} pseudo instruction represents, in this particular context, an
7567 abstract measurement of function's size. In no way does it represent a count
7568 of assembly instructions and as such its exact meaning might change from one
7569 release to an another.
7571 @item -fno-keep-inline-dllexport
7572 @opindex fno-keep-inline-dllexport
7573 This is a more fine-grained version of @option{-fkeep-inline-functions},
7574 which applies only to functions that are declared using the @code{dllexport}
7575 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7578 @item -fkeep-inline-functions
7579 @opindex fkeep-inline-functions
7580 In C, emit @code{static} functions that are declared @code{inline}
7581 into the object file, even if the function has been inlined into all
7582 of its callers. This switch does not affect functions using the
7583 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7584 inline functions into the object file.
7586 @item -fkeep-static-consts
7587 @opindex fkeep-static-consts
7588 Emit variables declared @code{static const} when optimization isn't turned
7589 on, even if the variables aren't referenced.
7591 GCC enables this option by default. If you want to force the compiler to
7592 check if a variable is referenced, regardless of whether or not
7593 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7595 @item -fmerge-constants
7596 @opindex fmerge-constants
7597 Attempt to merge identical constants (string constants and floating-point
7598 constants) across compilation units.
7600 This option is the default for optimized compilation if the assembler and
7601 linker support it. Use @option{-fno-merge-constants} to inhibit this
7604 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7606 @item -fmerge-all-constants
7607 @opindex fmerge-all-constants
7608 Attempt to merge identical constants and identical variables.
7610 This option implies @option{-fmerge-constants}. In addition to
7611 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7612 arrays or initialized constant variables with integral or floating-point
7613 types. Languages like C or C++ require each variable, including multiple
7614 instances of the same variable in recursive calls, to have distinct locations,
7615 so using this option results in non-conforming
7618 @item -fmodulo-sched
7619 @opindex fmodulo-sched
7620 Perform swing modulo scheduling immediately before the first scheduling
7621 pass. This pass looks at innermost loops and reorders their
7622 instructions by overlapping different iterations.
7624 @item -fmodulo-sched-allow-regmoves
7625 @opindex fmodulo-sched-allow-regmoves
7626 Perform more aggressive SMS-based modulo scheduling with register moves
7627 allowed. By setting this flag certain anti-dependences edges are
7628 deleted, which triggers the generation of reg-moves based on the
7629 life-range analysis. This option is effective only with
7630 @option{-fmodulo-sched} enabled.
7632 @item -fno-branch-count-reg
7633 @opindex fno-branch-count-reg
7634 Do not use ``decrement and branch'' instructions on a count register,
7635 but instead generate a sequence of instructions that decrement a
7636 register, compare it against zero, then branch based upon the result.
7637 This option is only meaningful on architectures that support such
7638 instructions, which include x86, PowerPC, IA-64 and S/390.
7640 Enabled by default at @option{-O1} and higher.
7642 The default is @option{-fbranch-count-reg}.
7644 @item -fno-function-cse
7645 @opindex fno-function-cse
7646 Do not put function addresses in registers; make each instruction that
7647 calls a constant function contain the function's address explicitly.
7649 This option results in less efficient code, but some strange hacks
7650 that alter the assembler output may be confused by the optimizations
7651 performed when this option is not used.
7653 The default is @option{-ffunction-cse}
7655 @item -fno-zero-initialized-in-bss
7656 @opindex fno-zero-initialized-in-bss
7657 If the target supports a BSS section, GCC by default puts variables that
7658 are initialized to zero into BSS@. This can save space in the resulting
7661 This option turns off this behavior because some programs explicitly
7662 rely on variables going to the data section---e.g., so that the
7663 resulting executable can find the beginning of that section and/or make
7664 assumptions based on that.
7666 The default is @option{-fzero-initialized-in-bss}.
7668 @item -fthread-jumps
7669 @opindex fthread-jumps
7670 Perform optimizations that check to see if a jump branches to a
7671 location where another comparison subsumed by the first is found. If
7672 so, the first branch is redirected to either the destination of the
7673 second branch or a point immediately following it, depending on whether
7674 the condition is known to be true or false.
7676 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7678 @item -fsplit-wide-types
7679 @opindex fsplit-wide-types
7680 When using a type that occupies multiple registers, such as @code{long
7681 long} on a 32-bit system, split the registers apart and allocate them
7682 independently. This normally generates better code for those types,
7683 but may make debugging more difficult.
7685 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7688 @item -fcse-follow-jumps
7689 @opindex fcse-follow-jumps
7690 In common subexpression elimination (CSE), scan through jump instructions
7691 when the target of the jump is not reached by any other path. For
7692 example, when CSE encounters an @code{if} statement with an
7693 @code{else} clause, CSE follows the jump when the condition
7696 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7698 @item -fcse-skip-blocks
7699 @opindex fcse-skip-blocks
7700 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7701 follow jumps that conditionally skip over blocks. When CSE
7702 encounters a simple @code{if} statement with no else clause,
7703 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7704 body of the @code{if}.
7706 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7708 @item -frerun-cse-after-loop
7709 @opindex frerun-cse-after-loop
7710 Re-run common subexpression elimination after loop optimizations are
7713 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7717 Perform a global common subexpression elimination pass.
7718 This pass also performs global constant and copy propagation.
7720 @emph{Note:} When compiling a program using computed gotos, a GCC
7721 extension, you may get better run-time performance if you disable
7722 the global common subexpression elimination pass by adding
7723 @option{-fno-gcse} to the command line.
7725 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7729 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7730 attempts to move loads that are only killed by stores into themselves. This
7731 allows a loop containing a load/store sequence to be changed to a load outside
7732 the loop, and a copy/store within the loop.
7734 Enabled by default when @option{-fgcse} is enabled.
7738 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7739 global common subexpression elimination. This pass attempts to move
7740 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7741 loops containing a load/store sequence can be changed to a load before
7742 the loop and a store after the loop.
7744 Not enabled at any optimization level.
7748 When @option{-fgcse-las} is enabled, the global common subexpression
7749 elimination pass eliminates redundant loads that come after stores to the
7750 same memory location (both partial and full redundancies).
7752 Not enabled at any optimization level.
7754 @item -fgcse-after-reload
7755 @opindex fgcse-after-reload
7756 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7757 pass is performed after reload. The purpose of this pass is to clean up
7760 @item -faggressive-loop-optimizations
7761 @opindex faggressive-loop-optimizations
7762 This option tells the loop optimizer to use language constraints to
7763 derive bounds for the number of iterations of a loop. This assumes that
7764 loop code does not invoke undefined behavior by for example causing signed
7765 integer overflows or out-of-bound array accesses. The bounds for the
7766 number of iterations of a loop are used to guide loop unrolling and peeling
7767 and loop exit test optimizations.
7768 This option is enabled by default.
7770 @item -funsafe-loop-optimizations
7771 @opindex funsafe-loop-optimizations
7772 This option tells the loop optimizer to assume that loop indices do not
7773 overflow, and that loops with nontrivial exit condition are not
7774 infinite. This enables a wider range of loop optimizations even if
7775 the loop optimizer itself cannot prove that these assumptions are valid.
7776 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7777 if it finds this kind of loop.
7779 @item -fcrossjumping
7780 @opindex fcrossjumping
7781 Perform cross-jumping transformation.
7782 This transformation unifies equivalent code and saves code size. The
7783 resulting code may or may not perform better than without cross-jumping.
7785 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7787 @item -fauto-inc-dec
7788 @opindex fauto-inc-dec
7789 Combine increments or decrements of addresses with memory accesses.
7790 This pass is always skipped on architectures that do not have
7791 instructions to support this. Enabled by default at @option{-O} and
7792 higher on architectures that support this.
7796 Perform dead code elimination (DCE) on RTL@.
7797 Enabled by default at @option{-O} and higher.
7801 Perform dead store elimination (DSE) on RTL@.
7802 Enabled by default at @option{-O} and higher.
7804 @item -fif-conversion
7805 @opindex fif-conversion
7806 Attempt to transform conditional jumps into branch-less equivalents. This
7807 includes use of conditional moves, min, max, set flags and abs instructions, and
7808 some tricks doable by standard arithmetics. The use of conditional execution
7809 on chips where it is available is controlled by @option{-fif-conversion2}.
7811 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7813 @item -fif-conversion2
7814 @opindex fif-conversion2
7815 Use conditional execution (where available) to transform conditional jumps into
7816 branch-less equivalents.
7818 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7820 @item -fdeclone-ctor-dtor
7821 @opindex fdeclone-ctor-dtor
7822 The C++ ABI requires multiple entry points for constructors and
7823 destructors: one for a base subobject, one for a complete object, and
7824 one for a virtual destructor that calls operator delete afterwards.
7825 For a hierarchy with virtual bases, the base and complete variants are
7826 clones, which means two copies of the function. With this option, the
7827 base and complete variants are changed to be thunks that call a common
7830 Enabled by @option{-Os}.
7832 @item -fdelete-null-pointer-checks
7833 @opindex fdelete-null-pointer-checks
7834 Assume that programs cannot safely dereference null pointers, and that
7835 no code or data element resides there. This enables simple constant
7836 folding optimizations at all optimization levels. In addition, other
7837 optimization passes in GCC use this flag to control global dataflow
7838 analyses that eliminate useless checks for null pointers; these assume
7839 that if a pointer is checked after it has already been dereferenced,
7842 Note however that in some environments this assumption is not true.
7843 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7844 for programs that depend on that behavior.
7846 Some targets, especially embedded ones, disable this option at all levels.
7847 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7848 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7849 are enabled independently at different optimization levels.
7851 @item -fdevirtualize
7852 @opindex fdevirtualize
7853 Attempt to convert calls to virtual functions to direct calls. This
7854 is done both within a procedure and interprocedurally as part of
7855 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7856 propagation (@option{-fipa-cp}).
7857 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7859 @item -fdevirtualize-speculatively
7860 @opindex fdevirtualize-speculatively
7861 Attempt to convert calls to virtual functions to speculative direct calls.
7862 Based on the analysis of the type inheritance graph, determine for a given call
7863 the set of likely targets. If the set is small, preferably of size 1, change
7864 the call into a conditional deciding between direct and indirect calls. The
7865 speculative calls enable more optimizations, such as inlining. When they seem
7866 useless after further optimization, they are converted back into original form.
7868 @item -fdevirtualize-at-ltrans
7869 @opindex fdevirtualize-at-ltrans
7870 Stream extra information needed for aggressive devirtualization when running
7871 the link-time optimizer in local transformation mode.
7872 This option enables more devirtualization but
7873 significantly increases the size of streamed data. For this reason it is
7874 disabled by default.
7876 @item -fexpensive-optimizations
7877 @opindex fexpensive-optimizations
7878 Perform a number of minor optimizations that are relatively expensive.
7880 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7884 Attempt to remove redundant extension instructions. This is especially
7885 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7886 registers after writing to their lower 32-bit half.
7888 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7889 @option{-O3}, @option{-Os}.
7891 @item -fno-lifetime-dse
7892 @opindex fno-lifetime-dse
7893 In C++ the value of an object is only affected by changes within its
7894 lifetime: when the constructor begins, the object has an indeterminate
7895 value, and any changes during the lifetime of the object are dead when
7896 the object is destroyed. Normally dead store elimination will take
7897 advantage of this; if your code relies on the value of the object
7898 storage persisting beyond the lifetime of the object, you can use this
7899 flag to disable this optimization.
7901 @item -flive-range-shrinkage
7902 @opindex flive-range-shrinkage
7903 Attempt to decrease register pressure through register live range
7904 shrinkage. This is helpful for fast processors with small or moderate
7907 @item -fira-algorithm=@var{algorithm}
7908 @opindex fira-algorithm
7909 Use the specified coloring algorithm for the integrated register
7910 allocator. The @var{algorithm} argument can be @samp{priority}, which
7911 specifies Chow's priority coloring, or @samp{CB}, which specifies
7912 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7913 for all architectures, but for those targets that do support it, it is
7914 the default because it generates better code.
7916 @item -fira-region=@var{region}
7917 @opindex fira-region
7918 Use specified regions for the integrated register allocator. The
7919 @var{region} argument should be one of the following:
7924 Use all loops as register allocation regions.
7925 This can give the best results for machines with a small and/or
7926 irregular register set.
7929 Use all loops except for loops with small register pressure
7930 as the regions. This value usually gives
7931 the best results in most cases and for most architectures,
7932 and is enabled by default when compiling with optimization for speed
7933 (@option{-O}, @option{-O2}, @dots{}).
7936 Use all functions as a single region.
7937 This typically results in the smallest code size, and is enabled by default for
7938 @option{-Os} or @option{-O0}.
7942 @item -fira-hoist-pressure
7943 @opindex fira-hoist-pressure
7944 Use IRA to evaluate register pressure in the code hoisting pass for
7945 decisions to hoist expressions. This option usually results in smaller
7946 code, but it can slow the compiler down.
7948 This option is enabled at level @option{-Os} for all targets.
7950 @item -fira-loop-pressure
7951 @opindex fira-loop-pressure
7952 Use IRA to evaluate register pressure in loops for decisions to move
7953 loop invariants. This option usually results in generation
7954 of faster and smaller code on machines with large register files (>= 32
7955 registers), but it can slow the compiler down.
7957 This option is enabled at level @option{-O3} for some targets.
7959 @item -fno-ira-share-save-slots
7960 @opindex fno-ira-share-save-slots
7961 Disable sharing of stack slots used for saving call-used hard
7962 registers living through a call. Each hard register gets a
7963 separate stack slot, and as a result function stack frames are
7966 @item -fno-ira-share-spill-slots
7967 @opindex fno-ira-share-spill-slots
7968 Disable sharing of stack slots allocated for pseudo-registers. Each
7969 pseudo-register that does not get a hard register gets a separate
7970 stack slot, and as a result function stack frames are larger.
7972 @item -fira-verbose=@var{n}
7973 @opindex fira-verbose
7974 Control the verbosity of the dump file for the integrated register allocator.
7975 The default value is 5. If the value @var{n} is greater or equal to 10,
7976 the dump output is sent to stderr using the same format as @var{n} minus 10.
7980 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7981 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7982 values if it is profitable.
7984 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7986 @item -fdelayed-branch
7987 @opindex fdelayed-branch
7988 If supported for the target machine, attempt to reorder instructions
7989 to exploit instruction slots available after delayed branch
7992 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7994 @item -fschedule-insns
7995 @opindex fschedule-insns
7996 If supported for the target machine, attempt to reorder instructions to
7997 eliminate execution stalls due to required data being unavailable. This
7998 helps machines that have slow floating point or memory load instructions
7999 by allowing other instructions to be issued until the result of the load
8000 or floating-point instruction is required.
8002 Enabled at levels @option{-O2}, @option{-O3}.
8004 @item -fschedule-insns2
8005 @opindex fschedule-insns2
8006 Similar to @option{-fschedule-insns}, but requests an additional pass of
8007 instruction scheduling after register allocation has been done. This is
8008 especially useful on machines with a relatively small number of
8009 registers and where memory load instructions take more than one cycle.
8011 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8013 @item -fno-sched-interblock
8014 @opindex fno-sched-interblock
8015 Don't schedule instructions across basic blocks. This is normally
8016 enabled by default when scheduling before register allocation, i.e.@:
8017 with @option{-fschedule-insns} or at @option{-O2} or higher.
8019 @item -fno-sched-spec
8020 @opindex fno-sched-spec
8021 Don't allow speculative motion of non-load instructions. This is normally
8022 enabled by default when scheduling before register allocation, i.e.@:
8023 with @option{-fschedule-insns} or at @option{-O2} or higher.
8025 @item -fsched-pressure
8026 @opindex fsched-pressure
8027 Enable register pressure sensitive insn scheduling before register
8028 allocation. This only makes sense when scheduling before register
8029 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8030 @option{-O2} or higher. Usage of this option can improve the
8031 generated code and decrease its size by preventing register pressure
8032 increase above the number of available hard registers and subsequent
8033 spills in register allocation.
8035 @item -fsched-spec-load
8036 @opindex fsched-spec-load
8037 Allow speculative motion of some load instructions. This only makes
8038 sense when scheduling before register allocation, i.e.@: with
8039 @option{-fschedule-insns} or at @option{-O2} or higher.
8041 @item -fsched-spec-load-dangerous
8042 @opindex fsched-spec-load-dangerous
8043 Allow speculative motion of more load instructions. This only makes
8044 sense when scheduling before register allocation, i.e.@: with
8045 @option{-fschedule-insns} or at @option{-O2} or higher.
8047 @item -fsched-stalled-insns
8048 @itemx -fsched-stalled-insns=@var{n}
8049 @opindex fsched-stalled-insns
8050 Define how many insns (if any) can be moved prematurely from the queue
8051 of stalled insns into the ready list during the second scheduling pass.
8052 @option{-fno-sched-stalled-insns} means that no insns are moved
8053 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8054 on how many queued insns can be moved prematurely.
8055 @option{-fsched-stalled-insns} without a value is equivalent to
8056 @option{-fsched-stalled-insns=1}.
8058 @item -fsched-stalled-insns-dep
8059 @itemx -fsched-stalled-insns-dep=@var{n}
8060 @opindex fsched-stalled-insns-dep
8061 Define how many insn groups (cycles) are examined for a dependency
8062 on a stalled insn that is a candidate for premature removal from the queue
8063 of stalled insns. This has an effect only during the second scheduling pass,
8064 and only if @option{-fsched-stalled-insns} is used.
8065 @option{-fno-sched-stalled-insns-dep} is equivalent to
8066 @option{-fsched-stalled-insns-dep=0}.
8067 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8068 @option{-fsched-stalled-insns-dep=1}.
8070 @item -fsched2-use-superblocks
8071 @opindex fsched2-use-superblocks
8072 When scheduling after register allocation, use superblock scheduling.
8073 This allows motion across basic block boundaries,
8074 resulting in faster schedules. This option is experimental, as not all machine
8075 descriptions used by GCC model the CPU closely enough to avoid unreliable
8076 results from the algorithm.
8078 This only makes sense when scheduling after register allocation, i.e.@: with
8079 @option{-fschedule-insns2} or at @option{-O2} or higher.
8081 @item -fsched-group-heuristic
8082 @opindex fsched-group-heuristic
8083 Enable the group heuristic in the scheduler. This heuristic favors
8084 the instruction that belongs to a schedule group. This is enabled
8085 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8086 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8088 @item -fsched-critical-path-heuristic
8089 @opindex fsched-critical-path-heuristic
8090 Enable the critical-path heuristic in the scheduler. This heuristic favors
8091 instructions on the critical path. This is enabled by default when
8092 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8093 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8095 @item -fsched-spec-insn-heuristic
8096 @opindex fsched-spec-insn-heuristic
8097 Enable the speculative instruction heuristic in the scheduler. This
8098 heuristic favors speculative instructions with greater dependency weakness.
8099 This is enabled by default when scheduling is enabled, i.e.@:
8100 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8101 or at @option{-O2} or higher.
8103 @item -fsched-rank-heuristic
8104 @opindex fsched-rank-heuristic
8105 Enable the rank heuristic in the scheduler. This heuristic favors
8106 the instruction belonging to a basic block with greater size or frequency.
8107 This is enabled by default when scheduling is enabled, i.e.@:
8108 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8109 at @option{-O2} or higher.
8111 @item -fsched-last-insn-heuristic
8112 @opindex fsched-last-insn-heuristic
8113 Enable the last-instruction heuristic in the scheduler. This heuristic
8114 favors the instruction that is less dependent on the last instruction
8115 scheduled. This is enabled by default when scheduling is enabled,
8116 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8117 at @option{-O2} or higher.
8119 @item -fsched-dep-count-heuristic
8120 @opindex fsched-dep-count-heuristic
8121 Enable the dependent-count heuristic in the scheduler. This heuristic
8122 favors the instruction that has more instructions depending on it.
8123 This is enabled by default when scheduling is enabled, i.e.@:
8124 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8125 at @option{-O2} or higher.
8127 @item -freschedule-modulo-scheduled-loops
8128 @opindex freschedule-modulo-scheduled-loops
8129 Modulo scheduling is performed before traditional scheduling. If a loop
8130 is modulo scheduled, later scheduling passes may change its schedule.
8131 Use this option to control that behavior.
8133 @item -fselective-scheduling
8134 @opindex fselective-scheduling
8135 Schedule instructions using selective scheduling algorithm. Selective
8136 scheduling runs instead of the first scheduler pass.
8138 @item -fselective-scheduling2
8139 @opindex fselective-scheduling2
8140 Schedule instructions using selective scheduling algorithm. Selective
8141 scheduling runs instead of the second scheduler pass.
8143 @item -fsel-sched-pipelining
8144 @opindex fsel-sched-pipelining
8145 Enable software pipelining of innermost loops during selective scheduling.
8146 This option has no effect unless one of @option{-fselective-scheduling} or
8147 @option{-fselective-scheduling2} is turned on.
8149 @item -fsel-sched-pipelining-outer-loops
8150 @opindex fsel-sched-pipelining-outer-loops
8151 When pipelining loops during selective scheduling, also pipeline outer loops.
8152 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8154 @item -fsemantic-interposition
8155 @opindex fsemantic-interposition
8156 Some object formats, like ELF, allow interposing of symbols by the
8158 This means that for symbols exported from the DSO, the compiler cannot perform
8159 interprocedural propagation, inlining and other optimizations in anticipation
8160 that the function or variable in question may change. While this feature is
8161 useful, for example, to rewrite memory allocation functions by a debugging
8162 implementation, it is expensive in the terms of code quality.
8163 With @option{-fno-semantic-interposition} the compiler assumes that
8164 if interposition happens for functions the overwriting function will have
8165 precisely the same semantics (and side effects).
8166 Similarly if interposition happens
8167 for variables, the constructor of the variable will be the same. The flag
8168 has no effect for functions explicitly declared inline
8169 (where it is never allowed for interposition to change semantics)
8170 and for symbols explicitly declared weak.
8173 @opindex fshrink-wrap
8174 Emit function prologues only before parts of the function that need it,
8175 rather than at the top of the function. This flag is enabled by default at
8176 @option{-O} and higher.
8178 @item -fcaller-saves
8179 @opindex fcaller-saves
8180 Enable allocation of values to registers that are clobbered by
8181 function calls, by emitting extra instructions to save and restore the
8182 registers around such calls. Such allocation is done only when it
8183 seems to result in better code.
8185 This option is always enabled by default on certain machines, usually
8186 those which have no call-preserved registers to use instead.
8188 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8190 @item -fcombine-stack-adjustments
8191 @opindex fcombine-stack-adjustments
8192 Tracks stack adjustments (pushes and pops) and stack memory references
8193 and then tries to find ways to combine them.
8195 Enabled by default at @option{-O1} and higher.
8199 Use caller save registers for allocation if those registers are not used by
8200 any called function. In that case it is not necessary to save and restore
8201 them around calls. This is only possible if called functions are part of
8202 same compilation unit as current function and they are compiled before it.
8204 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8206 @item -fconserve-stack
8207 @opindex fconserve-stack
8208 Attempt to minimize stack usage. The compiler attempts to use less
8209 stack space, even if that makes the program slower. This option
8210 implies setting the @option{large-stack-frame} parameter to 100
8211 and the @option{large-stack-frame-growth} parameter to 400.
8213 @item -ftree-reassoc
8214 @opindex ftree-reassoc
8215 Perform reassociation on trees. This flag is enabled by default
8216 at @option{-O} and higher.
8220 Perform partial redundancy elimination (PRE) on trees. This flag is
8221 enabled by default at @option{-O2} and @option{-O3}.
8223 @item -ftree-partial-pre
8224 @opindex ftree-partial-pre
8225 Make partial redundancy elimination (PRE) more aggressive. This flag is
8226 enabled by default at @option{-O3}.
8228 @item -ftree-forwprop
8229 @opindex ftree-forwprop
8230 Perform forward propagation on trees. This flag is enabled by default
8231 at @option{-O} and higher.
8235 Perform full redundancy elimination (FRE) on trees. The difference
8236 between FRE and PRE is that FRE only considers expressions
8237 that are computed on all paths leading to the redundant computation.
8238 This analysis is faster than PRE, though it exposes fewer redundancies.
8239 This flag is enabled by default at @option{-O} and higher.
8241 @item -ftree-phiprop
8242 @opindex ftree-phiprop
8243 Perform hoisting of loads from conditional pointers on trees. This
8244 pass is enabled by default at @option{-O} and higher.
8246 @item -fhoist-adjacent-loads
8247 @opindex fhoist-adjacent-loads
8248 Speculatively hoist loads from both branches of an if-then-else if the
8249 loads are from adjacent locations in the same structure and the target
8250 architecture has a conditional move instruction. This flag is enabled
8251 by default at @option{-O2} and higher.
8253 @item -ftree-copy-prop
8254 @opindex ftree-copy-prop
8255 Perform copy propagation on trees. This pass eliminates unnecessary
8256 copy operations. This flag is enabled by default at @option{-O} and
8259 @item -fipa-pure-const
8260 @opindex fipa-pure-const
8261 Discover which functions are pure or constant.
8262 Enabled by default at @option{-O} and higher.
8264 @item -fipa-reference
8265 @opindex fipa-reference
8266 Discover which static variables do not escape the
8268 Enabled by default at @option{-O} and higher.
8272 Perform interprocedural pointer analysis and interprocedural modification
8273 and reference analysis. This option can cause excessive memory and
8274 compile-time usage on large compilation units. It is not enabled by
8275 default at any optimization level.
8278 @opindex fipa-profile
8279 Perform interprocedural profile propagation. The functions called only from
8280 cold functions are marked as cold. Also functions executed once (such as
8281 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8282 functions and loop less parts of functions executed once are then optimized for
8284 Enabled by default at @option{-O} and higher.
8288 Perform interprocedural constant propagation.
8289 This optimization analyzes the program to determine when values passed
8290 to functions are constants and then optimizes accordingly.
8291 This optimization can substantially increase performance
8292 if the application has constants passed to functions.
8293 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8295 @item -fipa-cp-clone
8296 @opindex fipa-cp-clone
8297 Perform function cloning to make interprocedural constant propagation stronger.
8298 When enabled, interprocedural constant propagation performs function cloning
8299 when externally visible function can be called with constant arguments.
8300 Because this optimization can create multiple copies of functions,
8301 it may significantly increase code size
8302 (see @option{--param ipcp-unit-growth=@var{value}}).
8303 This flag is enabled by default at @option{-O3}.
8307 Perform Identical Code Folding for functions and read-only variables.
8308 The optimization reduces code size and may disturb unwind stacks by replacing
8309 a function by equivalent one with a different name. The optimization works
8310 more effectively with link time optimization enabled.
8312 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8313 works on different levels and thus the optimizations are not same - there are
8314 equivalences that are found only by GCC and equivalences found only by Gold.
8316 This flag is enabled by default at @option{-O2} and @option{-Os}.
8318 @item -fisolate-erroneous-paths-dereference
8319 @opindex fisolate-erroneous-paths-dereference
8320 Detect paths that trigger erroneous or undefined behavior due to
8321 dereferencing a null pointer. Isolate those paths from the main control
8322 flow and turn the statement with erroneous or undefined behavior into a trap.
8323 This flag is enabled by default at @option{-O2} and higher.
8325 @item -fisolate-erroneous-paths-attribute
8326 @opindex fisolate-erroneous-paths-attribute
8327 Detect paths that trigger erroneous or undefined behavior due a null value
8328 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8329 attribute. Isolate those paths from the main control flow and turn the
8330 statement with erroneous or undefined behavior into a trap. This is not
8331 currently enabled, but may be enabled by @option{-O2} in the future.
8335 Perform forward store motion on trees. This flag is
8336 enabled by default at @option{-O} and higher.
8338 @item -ftree-bit-ccp
8339 @opindex ftree-bit-ccp
8340 Perform sparse conditional bit constant propagation on trees and propagate
8341 pointer alignment information.
8342 This pass only operates on local scalar variables and is enabled by default
8343 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8347 Perform sparse conditional constant propagation (CCP) on trees. This
8348 pass only operates on local scalar variables and is enabled by default
8349 at @option{-O} and higher.
8352 @opindex fssa-phiopt
8353 Perform pattern matching on SSA PHI nodes to optimize conditional
8354 code. This pass is enabled by default at @option{-O} and higher.
8356 @item -ftree-switch-conversion
8357 @opindex ftree-switch-conversion
8358 Perform conversion of simple initializations in a switch to
8359 initializations from a scalar array. This flag is enabled by default
8360 at @option{-O2} and higher.
8362 @item -ftree-tail-merge
8363 @opindex ftree-tail-merge
8364 Look for identical code sequences. When found, replace one with a jump to the
8365 other. This optimization is known as tail merging or cross jumping. This flag
8366 is enabled by default at @option{-O2} and higher. The compilation time
8368 be limited using @option{max-tail-merge-comparisons} parameter and
8369 @option{max-tail-merge-iterations} parameter.
8373 Perform dead code elimination (DCE) on trees. This flag is enabled by
8374 default at @option{-O} and higher.
8376 @item -ftree-builtin-call-dce
8377 @opindex ftree-builtin-call-dce
8378 Perform conditional dead code elimination (DCE) for calls to built-in functions
8379 that may set @code{errno} but are otherwise side-effect free. This flag is
8380 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8383 @item -ftree-dominator-opts
8384 @opindex ftree-dominator-opts
8385 Perform a variety of simple scalar cleanups (constant/copy
8386 propagation, redundancy elimination, range propagation and expression
8387 simplification) based on a dominator tree traversal. This also
8388 performs jump threading (to reduce jumps to jumps). This flag is
8389 enabled by default at @option{-O} and higher.
8393 Perform dead store elimination (DSE) on trees. A dead store is a store into
8394 a memory location that is later overwritten by another store without
8395 any intervening loads. In this case the earlier store can be deleted. This
8396 flag is enabled by default at @option{-O} and higher.
8400 Perform loop header copying on trees. This is beneficial since it increases
8401 effectiveness of code motion optimizations. It also saves one jump. This flag
8402 is enabled by default at @option{-O} and higher. It is not enabled
8403 for @option{-Os}, since it usually increases code size.
8405 @item -ftree-loop-optimize
8406 @opindex ftree-loop-optimize
8407 Perform loop optimizations on trees. This flag is enabled by default
8408 at @option{-O} and higher.
8410 @item -ftree-loop-linear
8411 @opindex ftree-loop-linear
8412 Perform loop interchange transformations on tree. Same as
8413 @option{-floop-interchange}. To use this code transformation, GCC has
8414 to be configured with @option{--with-isl} to enable the Graphite loop
8415 transformation infrastructure.
8417 @item -floop-interchange
8418 @opindex floop-interchange
8419 Perform loop interchange transformations on loops. Interchanging two
8420 nested loops switches the inner and outer loops. For example, given a
8425 A(J, I) = A(J, I) * C
8430 loop interchange transforms the loop as if it were written:
8434 A(J, I) = A(J, I) * C
8438 which can be beneficial when @code{N} is larger than the caches,
8439 because in Fortran, the elements of an array are stored in memory
8440 contiguously by column, and the original loop iterates over rows,
8441 potentially creating at each access a cache miss. This optimization
8442 applies to all the languages supported by GCC and is not limited to
8443 Fortran. To use this code transformation, GCC has to be configured
8444 with @option{--with-isl} to enable the Graphite loop transformation
8447 @item -floop-strip-mine
8448 @opindex floop-strip-mine
8449 Perform loop strip mining transformations on loops. Strip mining
8450 splits a loop into two nested loops. The outer loop has strides
8451 equal to the strip size and the inner loop has strides of the
8452 original loop within a strip. The strip length can be changed
8453 using the @option{loop-block-tile-size} parameter. For example,
8461 loop strip mining transforms the loop as if it were written:
8464 DO I = II, min (II + 50, N)
8469 This optimization applies to all the languages supported by GCC and is
8470 not limited to Fortran. To use this code transformation, GCC has to
8471 be configured with @option{--with-isl} to enable the Graphite loop
8472 transformation infrastructure.
8475 @opindex floop-block
8476 Perform loop blocking transformations on loops. Blocking strip mines
8477 each loop in the loop nest such that the memory accesses of the
8478 element loops fit inside caches. The strip length can be changed
8479 using the @option{loop-block-tile-size} parameter. For example, given
8484 A(J, I) = B(I) + C(J)
8489 loop blocking transforms the loop as if it were written:
8493 DO I = II, min (II + 50, N)
8494 DO J = JJ, min (JJ + 50, M)
8495 A(J, I) = B(I) + C(J)
8501 which can be beneficial when @code{M} is larger than the caches,
8502 because the innermost loop iterates over a smaller amount of data
8503 which can be kept in the caches. This optimization applies to all the
8504 languages supported by GCC and is not limited to Fortran. To use this
8505 code transformation, GCC has to be configured with @option{--with-isl}
8506 to enable the Graphite loop transformation infrastructure.
8508 @item -fgraphite-identity
8509 @opindex fgraphite-identity
8510 Enable the identity transformation for graphite. For every SCoP we generate
8511 the polyhedral representation and transform it back to gimple. Using
8512 @option{-fgraphite-identity} we can check the costs or benefits of the
8513 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8514 are also performed by the code generator ISL, like index splitting and
8515 dead code elimination in loops.
8517 @item -floop-nest-optimize
8518 @opindex floop-nest-optimize
8519 Enable the ISL based loop nest optimizer. This is a generic loop nest
8520 optimizer based on the Pluto optimization algorithms. It calculates a loop
8521 structure optimized for data-locality and parallelism. This option
8524 @item -floop-unroll-and-jam
8525 @opindex floop-unroll-and-jam
8526 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8527 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8528 The unrolled dimension (counting from the most inner one) can be changed
8529 using the @option{loop-unroll-jam-depth} parameter. .
8531 @item -floop-parallelize-all
8532 @opindex floop-parallelize-all
8533 Use the Graphite data dependence analysis to identify loops that can
8534 be parallelized. Parallelize all the loops that can be analyzed to
8535 not contain loop carried dependences without checking that it is
8536 profitable to parallelize the loops.
8538 @item -fcheck-data-deps
8539 @opindex fcheck-data-deps
8540 Compare the results of several data dependence analyzers. This option
8541 is used for debugging the data dependence analyzers.
8543 @item -ftree-loop-if-convert
8544 @opindex ftree-loop-if-convert
8545 Attempt to transform conditional jumps in the innermost loops to
8546 branch-less equivalents. The intent is to remove control-flow from
8547 the innermost loops in order to improve the ability of the
8548 vectorization pass to handle these loops. This is enabled by default
8549 if vectorization is enabled.
8551 @item -ftree-loop-if-convert-stores
8552 @opindex ftree-loop-if-convert-stores
8553 Attempt to also if-convert conditional jumps containing memory writes.
8554 This transformation can be unsafe for multi-threaded programs as it
8555 transforms conditional memory writes into unconditional memory writes.
8558 for (i = 0; i < N; i++)
8564 for (i = 0; i < N; i++)
8565 A[i] = cond ? expr : A[i];
8567 potentially producing data races.
8569 @item -ftree-loop-distribution
8570 @opindex ftree-loop-distribution
8571 Perform loop distribution. This flag can improve cache performance on
8572 big loop bodies and allow further loop optimizations, like
8573 parallelization or vectorization, to take place. For example, the loop
8590 @item -ftree-loop-distribute-patterns
8591 @opindex ftree-loop-distribute-patterns
8592 Perform loop distribution of patterns that can be code generated with
8593 calls to a library. This flag is enabled by default at @option{-O3}.
8595 This pass distributes the initialization loops and generates a call to
8596 memset zero. For example, the loop
8612 and the initialization loop is transformed into a call to memset zero.
8614 @item -ftree-loop-im
8615 @opindex ftree-loop-im
8616 Perform loop invariant motion on trees. This pass moves only invariants that
8617 are hard to handle at RTL level (function calls, operations that expand to
8618 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8619 operands of conditions that are invariant out of the loop, so that we can use
8620 just trivial invariantness analysis in loop unswitching. The pass also includes
8623 @item -ftree-loop-ivcanon
8624 @opindex ftree-loop-ivcanon
8625 Create a canonical counter for number of iterations in loops for which
8626 determining number of iterations requires complicated analysis. Later
8627 optimizations then may determine the number easily. Useful especially
8628 in connection with unrolling.
8632 Perform induction variable optimizations (strength reduction, induction
8633 variable merging and induction variable elimination) on trees.
8635 @item -ftree-parallelize-loops=n
8636 @opindex ftree-parallelize-loops
8637 Parallelize loops, i.e., split their iteration space to run in n threads.
8638 This is only possible for loops whose iterations are independent
8639 and can be arbitrarily reordered. The optimization is only
8640 profitable on multiprocessor machines, for loops that are CPU-intensive,
8641 rather than constrained e.g.@: by memory bandwidth. This option
8642 implies @option{-pthread}, and thus is only supported on targets
8643 that have support for @option{-pthread}.
8647 Perform function-local points-to analysis on trees. This flag is
8648 enabled by default at @option{-O} and higher.
8652 Perform scalar replacement of aggregates. This pass replaces structure
8653 references with scalars to prevent committing structures to memory too
8654 early. This flag is enabled by default at @option{-O} and higher.
8656 @item -ftree-copyrename
8657 @opindex ftree-copyrename
8658 Perform copy renaming on trees. This pass attempts to rename compiler
8659 temporaries to other variables at copy locations, usually resulting in
8660 variable names which more closely resemble the original variables. This flag
8661 is enabled by default at @option{-O} and higher.
8663 @item -ftree-coalesce-inlined-vars
8664 @opindex ftree-coalesce-inlined-vars
8665 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8666 combine small user-defined variables too, but only if they are inlined
8667 from other functions. It is a more limited form of
8668 @option{-ftree-coalesce-vars}. This may harm debug information of such
8669 inlined variables, but it keeps variables of the inlined-into
8670 function apart from each other, such that they are more likely to
8671 contain the expected values in a debugging session. This was the
8672 default in GCC versions older than 4.7.
8674 @item -ftree-coalesce-vars
8675 @opindex ftree-coalesce-vars
8676 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8677 combine small user-defined variables too, instead of just compiler
8678 temporaries. This may severely limit the ability to debug an optimized
8679 program compiled with @option{-fno-var-tracking-assignments}. In the
8680 negated form, this flag prevents SSA coalescing of user variables,
8681 including inlined ones. This option is enabled by default.
8685 Perform temporary expression replacement during the SSA->normal phase. Single
8686 use/single def temporaries are replaced at their use location with their
8687 defining expression. This results in non-GIMPLE code, but gives the expanders
8688 much more complex trees to work on resulting in better RTL generation. This is
8689 enabled by default at @option{-O} and higher.
8693 Perform straight-line strength reduction on trees. This recognizes related
8694 expressions involving multiplications and replaces them by less expensive
8695 calculations when possible. This is enabled by default at @option{-O} and
8698 @item -ftree-vectorize
8699 @opindex ftree-vectorize
8700 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8701 and @option{-ftree-slp-vectorize} if not explicitly specified.
8703 @item -ftree-loop-vectorize
8704 @opindex ftree-loop-vectorize
8705 Perform loop vectorization on trees. This flag is enabled by default at
8706 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8708 @item -ftree-slp-vectorize
8709 @opindex ftree-slp-vectorize
8710 Perform basic block vectorization on trees. This flag is enabled by default at
8711 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8713 @item -fvect-cost-model=@var{model}
8714 @opindex fvect-cost-model
8715 Alter the cost model used for vectorization. The @var{model} argument
8716 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8717 With the @samp{unlimited} model the vectorized code-path is assumed
8718 to be profitable while with the @samp{dynamic} model a runtime check
8719 guards the vectorized code-path to enable it only for iteration
8720 counts that will likely execute faster than when executing the original
8721 scalar loop. The @samp{cheap} model disables vectorization of
8722 loops where doing so would be cost prohibitive for example due to
8723 required runtime checks for data dependence or alignment but otherwise
8724 is equal to the @samp{dynamic} model.
8725 The default cost model depends on other optimization flags and is
8726 either @samp{dynamic} or @samp{cheap}.
8728 @item -fsimd-cost-model=@var{model}
8729 @opindex fsimd-cost-model
8730 Alter the cost model used for vectorization of loops marked with the OpenMP
8731 or Cilk Plus simd directive. The @var{model} argument should be one of
8732 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8733 have the same meaning as described in @option{-fvect-cost-model} and by
8734 default a cost model defined with @option{-fvect-cost-model} is used.
8738 Perform Value Range Propagation on trees. This is similar to the
8739 constant propagation pass, but instead of values, ranges of values are
8740 propagated. This allows the optimizers to remove unnecessary range
8741 checks like array bound checks and null pointer checks. This is
8742 enabled by default at @option{-O2} and higher. Null pointer check
8743 elimination is only done if @option{-fdelete-null-pointer-checks} is
8746 @item -fsplit-ivs-in-unroller
8747 @opindex fsplit-ivs-in-unroller
8748 Enables expression of values of induction variables in later iterations
8749 of the unrolled loop using the value in the first iteration. This breaks
8750 long dependency chains, thus improving efficiency of the scheduling passes.
8752 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8753 same effect. However, that is not reliable in cases where the loop body
8754 is more complicated than a single basic block. It also does not work at all
8755 on some architectures due to restrictions in the CSE pass.
8757 This optimization is enabled by default.
8759 @item -fvariable-expansion-in-unroller
8760 @opindex fvariable-expansion-in-unroller
8761 With this option, the compiler creates multiple copies of some
8762 local variables when unrolling a loop, which can result in superior code.
8764 @item -fpartial-inlining
8765 @opindex fpartial-inlining
8766 Inline parts of functions. This option has any effect only
8767 when inlining itself is turned on by the @option{-finline-functions}
8768 or @option{-finline-small-functions} options.
8770 Enabled at level @option{-O2}.
8772 @item -fpredictive-commoning
8773 @opindex fpredictive-commoning
8774 Perform predictive commoning optimization, i.e., reusing computations
8775 (especially memory loads and stores) performed in previous
8776 iterations of loops.
8778 This option is enabled at level @option{-O3}.
8780 @item -fprefetch-loop-arrays
8781 @opindex fprefetch-loop-arrays
8782 If supported by the target machine, generate instructions to prefetch
8783 memory to improve the performance of loops that access large arrays.
8785 This option may generate better or worse code; results are highly
8786 dependent on the structure of loops within the source code.
8788 Disabled at level @option{-Os}.
8791 @itemx -fno-peephole2
8792 @opindex fno-peephole
8793 @opindex fno-peephole2
8794 Disable any machine-specific peephole optimizations. The difference
8795 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8796 are implemented in the compiler; some targets use one, some use the
8797 other, a few use both.
8799 @option{-fpeephole} is enabled by default.
8800 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8802 @item -fno-guess-branch-probability
8803 @opindex fno-guess-branch-probability
8804 Do not guess branch probabilities using heuristics.
8806 GCC uses heuristics to guess branch probabilities if they are
8807 not provided by profiling feedback (@option{-fprofile-arcs}). These
8808 heuristics are based on the control flow graph. If some branch probabilities
8809 are specified by @code{__builtin_expect}, then the heuristics are
8810 used to guess branch probabilities for the rest of the control flow graph,
8811 taking the @code{__builtin_expect} info into account. The interactions
8812 between the heuristics and @code{__builtin_expect} can be complex, and in
8813 some cases, it may be useful to disable the heuristics so that the effects
8814 of @code{__builtin_expect} are easier to understand.
8816 The default is @option{-fguess-branch-probability} at levels
8817 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8819 @item -freorder-blocks
8820 @opindex freorder-blocks
8821 Reorder basic blocks in the compiled function in order to reduce number of
8822 taken branches and improve code locality.
8824 Enabled at levels @option{-O2}, @option{-O3}.
8826 @item -freorder-blocks-and-partition
8827 @opindex freorder-blocks-and-partition
8828 In addition to reordering basic blocks in the compiled function, in order
8829 to reduce number of taken branches, partitions hot and cold basic blocks
8830 into separate sections of the assembly and .o files, to improve
8831 paging and cache locality performance.
8833 This optimization is automatically turned off in the presence of
8834 exception handling, for linkonce sections, for functions with a user-defined
8835 section attribute and on any architecture that does not support named
8838 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8840 @item -freorder-functions
8841 @opindex freorder-functions
8842 Reorder functions in the object file in order to
8843 improve code locality. This is implemented by using special
8844 subsections @code{.text.hot} for most frequently executed functions and
8845 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8846 the linker so object file format must support named sections and linker must
8847 place them in a reasonable way.
8849 Also profile feedback must be available to make this option effective. See
8850 @option{-fprofile-arcs} for details.
8852 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8854 @item -fstrict-aliasing
8855 @opindex fstrict-aliasing
8856 Allow the compiler to assume the strictest aliasing rules applicable to
8857 the language being compiled. For C (and C++), this activates
8858 optimizations based on the type of expressions. In particular, an
8859 object of one type is assumed never to reside at the same address as an
8860 object of a different type, unless the types are almost the same. For
8861 example, an @code{unsigned int} can alias an @code{int}, but not a
8862 @code{void*} or a @code{double}. A character type may alias any other
8865 @anchor{Type-punning}Pay special attention to code like this:
8878 The practice of reading from a different union member than the one most
8879 recently written to (called ``type-punning'') is common. Even with
8880 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8881 is accessed through the union type. So, the code above works as
8882 expected. @xref{Structures unions enumerations and bit-fields
8883 implementation}. However, this code might not:
8894 Similarly, access by taking the address, casting the resulting pointer
8895 and dereferencing the result has undefined behavior, even if the cast
8896 uses a union type, e.g.:
8900 return ((union a_union *) &d)->i;
8904 The @option{-fstrict-aliasing} option is enabled at levels
8905 @option{-O2}, @option{-O3}, @option{-Os}.
8907 @item -fstrict-overflow
8908 @opindex fstrict-overflow
8909 Allow the compiler to assume strict signed overflow rules, depending
8910 on the language being compiled. For C (and C++) this means that
8911 overflow when doing arithmetic with signed numbers is undefined, which
8912 means that the compiler may assume that it does not happen. This
8913 permits various optimizations. For example, the compiler assumes
8914 that an expression like @code{i + 10 > i} is always true for
8915 signed @code{i}. This assumption is only valid if signed overflow is
8916 undefined, as the expression is false if @code{i + 10} overflows when
8917 using twos complement arithmetic. When this option is in effect any
8918 attempt to determine whether an operation on signed numbers
8919 overflows must be written carefully to not actually involve overflow.
8921 This option also allows the compiler to assume strict pointer
8922 semantics: given a pointer to an object, if adding an offset to that
8923 pointer does not produce a pointer to the same object, the addition is
8924 undefined. This permits the compiler to conclude that @code{p + u >
8925 p} is always true for a pointer @code{p} and unsigned integer
8926 @code{u}. This assumption is only valid because pointer wraparound is
8927 undefined, as the expression is false if @code{p + u} overflows using
8928 twos complement arithmetic.
8930 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8931 that integer signed overflow is fully defined: it wraps. When
8932 @option{-fwrapv} is used, there is no difference between
8933 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8934 integers. With @option{-fwrapv} certain types of overflow are
8935 permitted. For example, if the compiler gets an overflow when doing
8936 arithmetic on constants, the overflowed value can still be used with
8937 @option{-fwrapv}, but not otherwise.
8939 The @option{-fstrict-overflow} option is enabled at levels
8940 @option{-O2}, @option{-O3}, @option{-Os}.
8942 @item -falign-functions
8943 @itemx -falign-functions=@var{n}
8944 @opindex falign-functions
8945 Align the start of functions to the next power-of-two greater than
8946 @var{n}, skipping up to @var{n} bytes. For instance,
8947 @option{-falign-functions=32} aligns functions to the next 32-byte
8948 boundary, but @option{-falign-functions=24} aligns to the next
8949 32-byte boundary only if this can be done by skipping 23 bytes or less.
8951 @option{-fno-align-functions} and @option{-falign-functions=1} are
8952 equivalent and mean that functions are not aligned.
8954 Some assemblers only support this flag when @var{n} is a power of two;
8955 in that case, it is rounded up.
8957 If @var{n} is not specified or is zero, use a machine-dependent default.
8959 Enabled at levels @option{-O2}, @option{-O3}.
8961 @item -falign-labels
8962 @itemx -falign-labels=@var{n}
8963 @opindex falign-labels
8964 Align all branch targets to a power-of-two boundary, skipping up to
8965 @var{n} bytes like @option{-falign-functions}. This option can easily
8966 make code slower, because it must insert dummy operations for when the
8967 branch target is reached in the usual flow of the code.
8969 @option{-fno-align-labels} and @option{-falign-labels=1} are
8970 equivalent and mean that labels are not aligned.
8972 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8973 are greater than this value, then their values are used instead.
8975 If @var{n} is not specified or is zero, use a machine-dependent default
8976 which is very likely to be @samp{1}, meaning no alignment.
8978 Enabled at levels @option{-O2}, @option{-O3}.
8981 @itemx -falign-loops=@var{n}
8982 @opindex falign-loops
8983 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8984 like @option{-falign-functions}. If the loops are
8985 executed many times, this makes up for any execution of the dummy
8988 @option{-fno-align-loops} and @option{-falign-loops=1} are
8989 equivalent and mean that loops are not aligned.
8991 If @var{n} is not specified or is zero, use a machine-dependent default.
8993 Enabled at levels @option{-O2}, @option{-O3}.
8996 @itemx -falign-jumps=@var{n}
8997 @opindex falign-jumps
8998 Align branch targets to a power-of-two boundary, for branch targets
8999 where the targets can only be reached by jumping, skipping up to @var{n}
9000 bytes like @option{-falign-functions}. In this case, no dummy operations
9003 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9004 equivalent and mean that loops are not aligned.
9006 If @var{n} is not specified or is zero, use a machine-dependent default.
9008 Enabled at levels @option{-O2}, @option{-O3}.
9010 @item -funit-at-a-time
9011 @opindex funit-at-a-time
9012 This option is left for compatibility reasons. @option{-funit-at-a-time}
9013 has no effect, while @option{-fno-unit-at-a-time} implies
9014 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9018 @item -fno-toplevel-reorder
9019 @opindex fno-toplevel-reorder
9020 Do not reorder top-level functions, variables, and @code{asm}
9021 statements. Output them in the same order that they appear in the
9022 input file. When this option is used, unreferenced static variables
9023 are not removed. This option is intended to support existing code
9024 that relies on a particular ordering. For new code, it is better to
9025 use attributes when possible.
9027 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9028 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9033 Constructs webs as commonly used for register allocation purposes and assign
9034 each web individual pseudo register. This allows the register allocation pass
9035 to operate on pseudos directly, but also strengthens several other optimization
9036 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9037 however, make debugging impossible, since variables no longer stay in a
9040 Enabled by default with @option{-funroll-loops}.
9042 @item -fwhole-program
9043 @opindex fwhole-program
9044 Assume that the current compilation unit represents the whole program being
9045 compiled. All public functions and variables with the exception of @code{main}
9046 and those merged by attribute @code{externally_visible} become static functions
9047 and in effect are optimized more aggressively by interprocedural optimizers.
9049 This option should not be used in combination with @option{-flto}.
9050 Instead relying on a linker plugin should provide safer and more precise
9053 @item -flto[=@var{n}]
9055 This option runs the standard link-time optimizer. When invoked
9056 with source code, it generates GIMPLE (one of GCC's internal
9057 representations) and writes it to special ELF sections in the object
9058 file. When the object files are linked together, all the function
9059 bodies are read from these ELF sections and instantiated as if they
9060 had been part of the same translation unit.
9062 To use the link-time optimizer, @option{-flto} and optimization
9063 options should be specified at compile time and during the final link.
9067 gcc -c -O2 -flto foo.c
9068 gcc -c -O2 -flto bar.c
9069 gcc -o myprog -flto -O2 foo.o bar.o
9072 The first two invocations to GCC save a bytecode representation
9073 of GIMPLE into special ELF sections inside @file{foo.o} and
9074 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9075 @file{foo.o} and @file{bar.o}, merges the two files into a single
9076 internal image, and compiles the result as usual. Since both
9077 @file{foo.o} and @file{bar.o} are merged into a single image, this
9078 causes all the interprocedural analyses and optimizations in GCC to
9079 work across the two files as if they were a single one. This means,
9080 for example, that the inliner is able to inline functions in
9081 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9083 Another (simpler) way to enable link-time optimization is:
9086 gcc -o myprog -flto -O2 foo.c bar.c
9089 The above generates bytecode for @file{foo.c} and @file{bar.c},
9090 merges them together into a single GIMPLE representation and optimizes
9091 them as usual to produce @file{myprog}.
9093 The only important thing to keep in mind is that to enable link-time
9094 optimizations you need to use the GCC driver to perform the link-step.
9095 GCC then automatically performs link-time optimization if any of the
9096 objects involved were compiled with the @option{-flto}. You generally
9097 should specify the optimization options to be used for link-time
9098 optimization though GCC tries to be clever at guessing an
9099 optimization level to use from the options used at compile-time
9100 if you fail to specify one at link-time. You can always override
9101 the automatic decision to do link-time optimization at link-time
9102 by passing @option{-fno-lto} to the link command.
9104 To make whole program optimization effective, it is necessary to make
9105 certain whole program assumptions. The compiler needs to know
9106 what functions and variables can be accessed by libraries and runtime
9107 outside of the link-time optimized unit. When supported by the linker,
9108 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9109 to the compiler about used and externally visible symbols. When
9110 the linker plugin is not available, @option{-fwhole-program} should be
9111 used to allow the compiler to make these assumptions, which leads
9112 to more aggressive optimization decisions.
9114 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9115 compiled with @option{-flto}, the generated object file is larger than
9116 a regular object file because it contains GIMPLE bytecodes and the usual
9117 final code (see @option{-ffat-lto-objects}. This means that
9118 object files with LTO information can be linked as normal object
9119 files; if @option{-fno-lto} is passed to the linker, no
9120 interprocedural optimizations are applied. Note that when
9121 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9122 but you cannot perform a regular, non-LTO link on them.
9124 Additionally, the optimization flags used to compile individual files
9125 are not necessarily related to those used at link time. For instance,
9128 gcc -c -O0 -ffat-lto-objects -flto foo.c
9129 gcc -c -O0 -ffat-lto-objects -flto bar.c
9130 gcc -o myprog -O3 foo.o bar.o
9133 This produces individual object files with unoptimized assembler
9134 code, but the resulting binary @file{myprog} is optimized at
9135 @option{-O3}. If, instead, the final binary is generated with
9136 @option{-fno-lto}, then @file{myprog} is not optimized.
9138 When producing the final binary, GCC only
9139 applies link-time optimizations to those files that contain bytecode.
9140 Therefore, you can mix and match object files and libraries with
9141 GIMPLE bytecodes and final object code. GCC automatically selects
9142 which files to optimize in LTO mode and which files to link without
9145 There are some code generation flags preserved by GCC when
9146 generating bytecodes, as they need to be used during the final link
9147 stage. Generally options specified at link-time override those
9148 specified at compile-time.
9150 If you do not specify an optimization level option @option{-O} at
9151 link-time then GCC computes one based on the optimization levels
9152 used when compiling the object files. The highest optimization
9155 Currently, the following options and their setting are take from
9156 the first object file that explicitely specified it:
9157 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9158 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9159 and all the @option{-m} target flags.
9161 Certain ABI changing flags are required to match in all compilation-units
9162 and trying to override this at link-time with a conflicting value
9163 is ignored. This includes options such as @option{-freg-struct-return}
9164 and @option{-fpcc-struct-return}.
9166 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9167 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9168 are passed through to the link stage and merged conservatively for
9169 conflicting translation units. Specifically
9170 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9171 precedence and for example @option{-ffp-contract=off} takes precedence
9172 over @option{-ffp-contract=fast}. You can override them at linke-time.
9174 It is recommended that you compile all the files participating in the
9175 same link with the same options and also specify those options at
9178 If LTO encounters objects with C linkage declared with incompatible
9179 types in separate translation units to be linked together (undefined
9180 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9181 issued. The behavior is still undefined at run time. Similar
9182 diagnostics may be raised for other languages.
9184 Another feature of LTO is that it is possible to apply interprocedural
9185 optimizations on files written in different languages:
9190 gfortran -c -flto baz.f90
9191 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9194 Notice that the final link is done with @command{g++} to get the C++
9195 runtime libraries and @option{-lgfortran} is added to get the Fortran
9196 runtime libraries. In general, when mixing languages in LTO mode, you
9197 should use the same link command options as when mixing languages in a
9198 regular (non-LTO) compilation.
9200 If object files containing GIMPLE bytecode are stored in a library archive, say
9201 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9202 are using a linker with plugin support. To create static libraries suitable
9203 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9204 and @command{ranlib};
9205 to show the symbols of object files with GIMPLE bytecode, use
9206 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9207 and @command{nm} have been compiled with plugin support. At link time, use the the
9208 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9209 the LTO optimization process:
9212 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9215 With the linker plugin enabled, the linker extracts the needed
9216 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9217 to make them part of the aggregated GIMPLE image to be optimized.
9219 If you are not using a linker with plugin support and/or do not
9220 enable the linker plugin, then the objects inside @file{libfoo.a}
9221 are extracted and linked as usual, but they do not participate
9222 in the LTO optimization process. In order to make a static library suitable
9223 for both LTO optimization and usual linkage, compile its object files with
9224 @option{-flto} @option{-ffat-lto-objects}.
9226 Link-time optimizations do not require the presence of the whole program to
9227 operate. If the program does not require any symbols to be exported, it is
9228 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9229 the interprocedural optimizers to use more aggressive assumptions which may
9230 lead to improved optimization opportunities.
9231 Use of @option{-fwhole-program} is not needed when linker plugin is
9232 active (see @option{-fuse-linker-plugin}).
9234 The current implementation of LTO makes no
9235 attempt to generate bytecode that is portable between different
9236 types of hosts. The bytecode files are versioned and there is a
9237 strict version check, so bytecode files generated in one version of
9238 GCC do not work with an older or newer version of GCC.
9240 Link-time optimization does not work well with generation of debugging
9241 information. Combining @option{-flto} with
9242 @option{-g} is currently experimental and expected to produce unexpected
9245 If you specify the optional @var{n}, the optimization and code
9246 generation done at link time is executed in parallel using @var{n}
9247 parallel jobs by utilizing an installed @command{make} program. The
9248 environment variable @env{MAKE} may be used to override the program
9249 used. The default value for @var{n} is 1.
9251 You can also specify @option{-flto=jobserver} to use GNU make's
9252 job server mode to determine the number of parallel jobs. This
9253 is useful when the Makefile calling GCC is already executing in parallel.
9254 You must prepend a @samp{+} to the command recipe in the parent Makefile
9255 for this to work. This option likely only works if @env{MAKE} is
9258 @item -flto-partition=@var{alg}
9259 @opindex flto-partition
9260 Specify the partitioning algorithm used by the link-time optimizer.
9261 The value is either @samp{1to1} to specify a partitioning mirroring
9262 the original source files or @samp{balanced} to specify partitioning
9263 into equally sized chunks (whenever possible) or @samp{max} to create
9264 new partition for every symbol where possible. Specifying @samp{none}
9265 as an algorithm disables partitioning and streaming completely.
9266 The default value is @samp{balanced}. While @samp{1to1} can be used
9267 as an workaround for various code ordering issues, the @samp{max}
9268 partitioning is intended for internal testing only.
9269 The value @samp{one} specifies that exactly one partition should be
9270 used while the value @samp{none} bypasses partitioning and executes
9271 the link-time optimization step directly from the WPA phase.
9273 @item -flto-odr-type-merging
9274 @opindex flto-odr-type-merging
9275 Enable streaming of mangled types names of C++ types and their unification
9276 at linktime. This increases size of LTO object files, but enable
9277 diagnostics about One Definition Rule violations.
9279 @item -flto-compression-level=@var{n}
9280 @opindex flto-compression-level
9281 This option specifies the level of compression used for intermediate
9282 language written to LTO object files, and is only meaningful in
9283 conjunction with LTO mode (@option{-flto}). Valid
9284 values are 0 (no compression) to 9 (maximum compression). Values
9285 outside this range are clamped to either 0 or 9. If the option is not
9286 given, a default balanced compression setting is used.
9289 @opindex flto-report
9290 Prints a report with internal details on the workings of the link-time
9291 optimizer. The contents of this report vary from version to version.
9292 It is meant to be useful to GCC developers when processing object
9293 files in LTO mode (via @option{-flto}).
9295 Disabled by default.
9297 @item -flto-report-wpa
9298 @opindex flto-report-wpa
9299 Like @option{-flto-report}, but only print for the WPA phase of Link
9302 @item -fuse-linker-plugin
9303 @opindex fuse-linker-plugin
9304 Enables the use of a linker plugin during link-time optimization. This
9305 option relies on plugin support in the linker, which is available in gold
9306 or in GNU ld 2.21 or newer.
9308 This option enables the extraction of object files with GIMPLE bytecode out
9309 of library archives. This improves the quality of optimization by exposing
9310 more code to the link-time optimizer. This information specifies what
9311 symbols can be accessed externally (by non-LTO object or during dynamic
9312 linking). Resulting code quality improvements on binaries (and shared
9313 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9314 See @option{-flto} for a description of the effect of this flag and how to
9317 This option is enabled by default when LTO support in GCC is enabled
9318 and GCC was configured for use with
9319 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9321 @item -ffat-lto-objects
9322 @opindex ffat-lto-objects
9323 Fat LTO objects are object files that contain both the intermediate language
9324 and the object code. This makes them usable for both LTO linking and normal
9325 linking. This option is effective only when compiling with @option{-flto}
9326 and is ignored at link time.
9328 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9329 requires the complete toolchain to be aware of LTO. It requires a linker with
9330 linker plugin support for basic functionality. Additionally,
9331 @command{nm}, @command{ar} and @command{ranlib}
9332 need to support linker plugins to allow a full-featured build environment
9333 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9334 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9335 to these tools. With non fat LTO makefiles need to be modified to use them.
9337 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9340 @item -fcompare-elim
9341 @opindex fcompare-elim
9342 After register allocation and post-register allocation instruction splitting,
9343 identify arithmetic instructions that compute processor flags similar to a
9344 comparison operation based on that arithmetic. If possible, eliminate the
9345 explicit comparison operation.
9347 This pass only applies to certain targets that cannot explicitly represent
9348 the comparison operation before register allocation is complete.
9350 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9352 @item -fcprop-registers
9353 @opindex fcprop-registers
9354 After register allocation and post-register allocation instruction splitting,
9355 perform a copy-propagation pass to try to reduce scheduling dependencies
9356 and occasionally eliminate the copy.
9358 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9360 @item -fprofile-correction
9361 @opindex fprofile-correction
9362 Profiles collected using an instrumented binary for multi-threaded programs may
9363 be inconsistent due to missed counter updates. When this option is specified,
9364 GCC uses heuristics to correct or smooth out such inconsistencies. By
9365 default, GCC emits an error message when an inconsistent profile is detected.
9367 @item -fprofile-dir=@var{path}
9368 @opindex fprofile-dir
9370 Set the directory to search for the profile data files in to @var{path}.
9371 This option affects only the profile data generated by
9372 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9373 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9374 and its related options. Both absolute and relative paths can be used.
9375 By default, GCC uses the current directory as @var{path}, thus the
9376 profile data file appears in the same directory as the object file.
9378 @item -fprofile-generate
9379 @itemx -fprofile-generate=@var{path}
9380 @opindex fprofile-generate
9382 Enable options usually used for instrumenting application to produce
9383 profile useful for later recompilation with profile feedback based
9384 optimization. You must use @option{-fprofile-generate} both when
9385 compiling and when linking your program.
9387 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9389 If @var{path} is specified, GCC looks at the @var{path} to find
9390 the profile feedback data files. See @option{-fprofile-dir}.
9393 @itemx -fprofile-use=@var{path}
9394 @opindex fprofile-use
9395 Enable profile feedback-directed optimizations,
9396 and the following optimizations
9397 which are generally profitable only with profile feedback available:
9398 @option{-fbranch-probabilities}, @option{-fvpt},
9399 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9400 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9402 By default, GCC emits an error message if the feedback profiles do not
9403 match the source code. This error can be turned into a warning by using
9404 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9407 If @var{path} is specified, GCC looks at the @var{path} to find
9408 the profile feedback data files. See @option{-fprofile-dir}.
9410 @item -fauto-profile
9411 @itemx -fauto-profile=@var{path}
9412 @opindex fauto-profile
9413 Enable sampling-based feedback-directed optimizations,
9414 and the following optimizations
9415 which are generally profitable only with profile feedback available:
9416 @option{-fbranch-probabilities}, @option{-fvpt},
9417 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9418 @option{-ftree-vectorize},
9419 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9420 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9421 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9423 @var{path} is the name of a file containing AutoFDO profile information.
9424 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9426 Producing an AutoFDO profile data file requires running your program
9427 with the @command{perf} utility on a supported GNU/Linux target system.
9428 For more information, see @uref{https://perf.wiki.kernel.org/}.
9432 perf record -e br_inst_retired:near_taken -b -o perf.data \
9436 Then use the @command{create_gcov} tool to convert the raw profile data
9437 to a format that can be used by GCC.@ You must also supply the
9438 unstripped binary for your program to this tool.
9439 See @uref{https://github.com/google/autofdo}.
9443 create_gcov --binary=your_program.unstripped --profile=perf.data \
9448 The following options control compiler behavior regarding floating-point
9449 arithmetic. These options trade off between speed and
9450 correctness. All must be specifically enabled.
9454 @opindex ffloat-store
9455 Do not store floating-point variables in registers, and inhibit other
9456 options that might change whether a floating-point value is taken from a
9459 @cindex floating-point precision
9460 This option prevents undesirable excess precision on machines such as
9461 the 68000 where the floating registers (of the 68881) keep more
9462 precision than a @code{double} is supposed to have. Similarly for the
9463 x86 architecture. For most programs, the excess precision does only
9464 good, but a few programs rely on the precise definition of IEEE floating
9465 point. Use @option{-ffloat-store} for such programs, after modifying
9466 them to store all pertinent intermediate computations into variables.
9468 @item -fexcess-precision=@var{style}
9469 @opindex fexcess-precision
9470 This option allows further control over excess precision on machines
9471 where floating-point registers have more precision than the IEEE
9472 @code{float} and @code{double} types and the processor does not
9473 support operations rounding to those types. By default,
9474 @option{-fexcess-precision=fast} is in effect; this means that
9475 operations are carried out in the precision of the registers and that
9476 it is unpredictable when rounding to the types specified in the source
9477 code takes place. When compiling C, if
9478 @option{-fexcess-precision=standard} is specified then excess
9479 precision follows the rules specified in ISO C99; in particular,
9480 both casts and assignments cause values to be rounded to their
9481 semantic types (whereas @option{-ffloat-store} only affects
9482 assignments). This option is enabled by default for C if a strict
9483 conformance option such as @option{-std=c99} is used.
9486 @option{-fexcess-precision=standard} is not implemented for languages
9487 other than C, and has no effect if
9488 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9489 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9490 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9491 semantics apply without excess precision, and in the latter, rounding
9496 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9497 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9498 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9500 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9502 This option is not turned on by any @option{-O} option besides
9503 @option{-Ofast} since it can result in incorrect output for programs
9504 that depend on an exact implementation of IEEE or ISO rules/specifications
9505 for math functions. It may, however, yield faster code for programs
9506 that do not require the guarantees of these specifications.
9508 @item -fno-math-errno
9509 @opindex fno-math-errno
9510 Do not set @code{errno} after calling math functions that are executed
9511 with a single instruction, e.g., @code{sqrt}. A program that relies on
9512 IEEE exceptions for math error handling may want to use this flag
9513 for speed while maintaining IEEE arithmetic compatibility.
9515 This option is not turned on by any @option{-O} option since
9516 it can result in incorrect output for programs that depend on
9517 an exact implementation of IEEE or ISO rules/specifications for
9518 math functions. It may, however, yield faster code for programs
9519 that do not require the guarantees of these specifications.
9521 The default is @option{-fmath-errno}.
9523 On Darwin systems, the math library never sets @code{errno}. There is
9524 therefore no reason for the compiler to consider the possibility that
9525 it might, and @option{-fno-math-errno} is the default.
9527 @item -funsafe-math-optimizations
9528 @opindex funsafe-math-optimizations
9530 Allow optimizations for floating-point arithmetic that (a) assume
9531 that arguments and results are valid and (b) may violate IEEE or
9532 ANSI standards. When used at link-time, it may include libraries
9533 or startup files that change the default FPU control word or other
9534 similar optimizations.
9536 This option is not turned on by any @option{-O} option since
9537 it can result in incorrect output for programs that depend on
9538 an exact implementation of IEEE or ISO rules/specifications for
9539 math functions. It may, however, yield faster code for programs
9540 that do not require the guarantees of these specifications.
9541 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9542 @option{-fassociative-math} and @option{-freciprocal-math}.
9544 The default is @option{-fno-unsafe-math-optimizations}.
9546 @item -fassociative-math
9547 @opindex fassociative-math
9549 Allow re-association of operands in series of floating-point operations.
9550 This violates the ISO C and C++ language standard by possibly changing
9551 computation result. NOTE: re-ordering may change the sign of zero as
9552 well as ignore NaNs and inhibit or create underflow or overflow (and
9553 thus cannot be used on code that relies on rounding behavior like
9554 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9555 and thus may not be used when ordered comparisons are required.
9556 This option requires that both @option{-fno-signed-zeros} and
9557 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9558 much sense with @option{-frounding-math}. For Fortran the option
9559 is automatically enabled when both @option{-fno-signed-zeros} and
9560 @option{-fno-trapping-math} are in effect.
9562 The default is @option{-fno-associative-math}.
9564 @item -freciprocal-math
9565 @opindex freciprocal-math
9567 Allow the reciprocal of a value to be used instead of dividing by
9568 the value if this enables optimizations. For example @code{x / y}
9569 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9570 is subject to common subexpression elimination. Note that this loses
9571 precision and increases the number of flops operating on the value.
9573 The default is @option{-fno-reciprocal-math}.
9575 @item -ffinite-math-only
9576 @opindex ffinite-math-only
9577 Allow optimizations for floating-point arithmetic that assume
9578 that arguments and results are not NaNs or +-Infs.
9580 This option is not turned on by any @option{-O} option since
9581 it can result in incorrect output for programs that depend on
9582 an exact implementation of IEEE or ISO rules/specifications for
9583 math functions. It may, however, yield faster code for programs
9584 that do not require the guarantees of these specifications.
9586 The default is @option{-fno-finite-math-only}.
9588 @item -fno-signed-zeros
9589 @opindex fno-signed-zeros
9590 Allow optimizations for floating-point arithmetic that ignore the
9591 signedness of zero. IEEE arithmetic specifies the behavior of
9592 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9593 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9594 This option implies that the sign of a zero result isn't significant.
9596 The default is @option{-fsigned-zeros}.
9598 @item -fno-trapping-math
9599 @opindex fno-trapping-math
9600 Compile code assuming that floating-point operations cannot generate
9601 user-visible traps. These traps include division by zero, overflow,
9602 underflow, inexact result and invalid operation. This option requires
9603 that @option{-fno-signaling-nans} be in effect. Setting this option may
9604 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9606 This option should never be turned on by any @option{-O} option since
9607 it can result in incorrect output for programs that depend on
9608 an exact implementation of IEEE or ISO rules/specifications for
9611 The default is @option{-ftrapping-math}.
9613 @item -frounding-math
9614 @opindex frounding-math
9615 Disable transformations and optimizations that assume default floating-point
9616 rounding behavior. This is round-to-zero for all floating point
9617 to integer conversions, and round-to-nearest for all other arithmetic
9618 truncations. This option should be specified for programs that change
9619 the FP rounding mode dynamically, or that may be executed with a
9620 non-default rounding mode. This option disables constant folding of
9621 floating-point expressions at compile time (which may be affected by
9622 rounding mode) and arithmetic transformations that are unsafe in the
9623 presence of sign-dependent rounding modes.
9625 The default is @option{-fno-rounding-math}.
9627 This option is experimental and does not currently guarantee to
9628 disable all GCC optimizations that are affected by rounding mode.
9629 Future versions of GCC may provide finer control of this setting
9630 using C99's @code{FENV_ACCESS} pragma. This command-line option
9631 will be used to specify the default state for @code{FENV_ACCESS}.
9633 @item -fsignaling-nans
9634 @opindex fsignaling-nans
9635 Compile code assuming that IEEE signaling NaNs may generate user-visible
9636 traps during floating-point operations. Setting this option disables
9637 optimizations that may change the number of exceptions visible with
9638 signaling NaNs. This option implies @option{-ftrapping-math}.
9640 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9643 The default is @option{-fno-signaling-nans}.
9645 This option is experimental and does not currently guarantee to
9646 disable all GCC optimizations that affect signaling NaN behavior.
9648 @item -fsingle-precision-constant
9649 @opindex fsingle-precision-constant
9650 Treat floating-point constants as single precision instead of
9651 implicitly converting them to double-precision constants.
9653 @item -fcx-limited-range
9654 @opindex fcx-limited-range
9655 When enabled, this option states that a range reduction step is not
9656 needed when performing complex division. Also, there is no checking
9657 whether the result of a complex multiplication or division is @code{NaN
9658 + I*NaN}, with an attempt to rescue the situation in that case. The
9659 default is @option{-fno-cx-limited-range}, but is enabled by
9660 @option{-ffast-math}.
9662 This option controls the default setting of the ISO C99
9663 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9666 @item -fcx-fortran-rules
9667 @opindex fcx-fortran-rules
9668 Complex multiplication and division follow Fortran rules. Range
9669 reduction is done as part of complex division, but 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.
9673 The default is @option{-fno-cx-fortran-rules}.
9677 The following options control optimizations that may improve
9678 performance, but are not enabled by any @option{-O} options. This
9679 section includes experimental options that may produce broken code.
9682 @item -fbranch-probabilities
9683 @opindex fbranch-probabilities
9684 After running a program compiled with @option{-fprofile-arcs}
9685 (@pxref{Debugging Options,, Options for Debugging Your Program or
9686 @command{gcc}}), you can compile it a second time using
9687 @option{-fbranch-probabilities}, to improve optimizations based on
9688 the number of times each branch was taken. When a program
9689 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9690 counts to a file called @file{@var{sourcename}.gcda} for each source
9691 file. The information in this data file is very dependent on the
9692 structure of the generated code, so you must use the same source code
9693 and the same optimization options for both compilations.
9695 With @option{-fbranch-probabilities}, GCC puts a
9696 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9697 These can be used to improve optimization. Currently, they are only
9698 used in one place: in @file{reorg.c}, instead of guessing which path a
9699 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9700 exactly determine which path is taken more often.
9702 @item -fprofile-values
9703 @opindex fprofile-values
9704 If combined with @option{-fprofile-arcs}, it adds code so that some
9705 data about values of expressions in the program is gathered.
9707 With @option{-fbranch-probabilities}, it reads back the data gathered
9708 from profiling values of expressions for usage in optimizations.
9710 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9712 @item -fprofile-reorder-functions
9713 @opindex fprofile-reorder-functions
9714 Function reordering based on profile instrumentation collects
9715 first time of execution of a function and orders these functions
9718 Enabled with @option{-fprofile-use}.
9722 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9723 to add code to gather information about values of expressions.
9725 With @option{-fbranch-probabilities}, it reads back the data gathered
9726 and actually performs the optimizations based on them.
9727 Currently the optimizations include specialization of division operations
9728 using the knowledge about the value of the denominator.
9730 @item -frename-registers
9731 @opindex frename-registers
9732 Attempt to avoid false dependencies in scheduled code by making use
9733 of registers left over after register allocation. This optimization
9734 most benefits processors with lots of registers. Depending on the
9735 debug information format adopted by the target, however, it can
9736 make debugging impossible, since variables no longer stay in
9737 a ``home register''.
9739 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9741 @item -fschedule-fusion
9742 @opindex fschedule-fusion
9743 Performs a target dependent pass over the instruction stream to schedule
9744 instructions of same type together because target machine can execute them
9745 more efficiently if they are adjacent to each other in the instruction flow.
9747 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9751 Perform tail duplication to enlarge superblock size. This transformation
9752 simplifies the control flow of the function allowing other optimizations to do
9755 Enabled with @option{-fprofile-use}.
9757 @item -funroll-loops
9758 @opindex funroll-loops
9759 Unroll loops whose number of iterations can be determined at compile time or
9760 upon entry to the loop. @option{-funroll-loops} implies
9761 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9762 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9763 a small constant number of iterations). This option makes code larger, and may
9764 or may not make it run faster.
9766 Enabled with @option{-fprofile-use}.
9768 @item -funroll-all-loops
9769 @opindex funroll-all-loops
9770 Unroll all loops, even if their number of iterations is uncertain when
9771 the loop is entered. This usually makes programs run more slowly.
9772 @option{-funroll-all-loops} implies the same options as
9773 @option{-funroll-loops}.
9776 @opindex fpeel-loops
9777 Peels loops for which there is enough information that they do not
9778 roll much (from profile feedback). It also turns on complete loop peeling
9779 (i.e.@: complete removal of loops with small constant number of iterations).
9781 Enabled with @option{-fprofile-use}.
9783 @item -fmove-loop-invariants
9784 @opindex fmove-loop-invariants
9785 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9786 at level @option{-O1}
9788 @item -funswitch-loops
9789 @opindex funswitch-loops
9790 Move branches with loop invariant conditions out of the loop, with duplicates
9791 of the loop on both branches (modified according to result of the condition).
9793 @item -ffunction-sections
9794 @itemx -fdata-sections
9795 @opindex ffunction-sections
9796 @opindex fdata-sections
9797 Place each function or data item into its own section in the output
9798 file if the target supports arbitrary sections. The name of the
9799 function or the name of the data item determines the section's name
9802 Use these options on systems where the linker can perform optimizations
9803 to improve locality of reference in the instruction space. Most systems
9804 using the ELF object format and SPARC processors running Solaris 2 have
9805 linkers with such optimizations. AIX may have these optimizations in
9808 Only use these options when there are significant benefits from doing
9809 so. When you specify these options, the assembler and linker
9810 create larger object and executable files and are also slower.
9811 You cannot use @command{gprof} on all systems if you
9812 specify this option, and you may have problems with debugging if
9813 you specify both this option and @option{-g}.
9815 @item -fbranch-target-load-optimize
9816 @opindex fbranch-target-load-optimize
9817 Perform branch target register load optimization before prologue / epilogue
9819 The use of target registers can typically be exposed only during reload,
9820 thus hoisting loads out of loops and doing inter-block scheduling needs
9821 a separate optimization pass.
9823 @item -fbranch-target-load-optimize2
9824 @opindex fbranch-target-load-optimize2
9825 Perform branch target register load optimization after prologue / epilogue
9828 @item -fbtr-bb-exclusive
9829 @opindex fbtr-bb-exclusive
9830 When performing branch target register load optimization, don't reuse
9831 branch target registers within any basic block.
9833 @item -fstack-protector
9834 @opindex fstack-protector
9835 Emit extra code to check for buffer overflows, such as stack smashing
9836 attacks. This is done by adding a guard variable to functions with
9837 vulnerable objects. This includes functions that call @code{alloca}, and
9838 functions with buffers larger than 8 bytes. The guards are initialized
9839 when a function is entered and then checked when the function exits.
9840 If a guard check fails, an error message is printed and the program exits.
9842 @item -fstack-protector-all
9843 @opindex fstack-protector-all
9844 Like @option{-fstack-protector} except that all functions are protected.
9846 @item -fstack-protector-strong
9847 @opindex fstack-protector-strong
9848 Like @option{-fstack-protector} but includes additional functions to
9849 be protected --- those that have local array definitions, or have
9850 references to local frame addresses.
9852 @item -fstack-protector-explicit
9853 @opindex fstack-protector-explicit
9854 Like @option{-fstack-protector} but only protects those functions which
9855 have the @code{stack_protect} attribute
9857 @item -fsection-anchors
9858 @opindex fsection-anchors
9859 Try to reduce the number of symbolic address calculations by using
9860 shared ``anchor'' symbols to address nearby objects. This transformation
9861 can help to reduce the number of GOT entries and GOT accesses on some
9864 For example, the implementation of the following function @code{foo}:
9868 int foo (void) @{ return a + b + c; @}
9872 usually calculates the addresses of all three variables, but if you
9873 compile it with @option{-fsection-anchors}, it accesses the variables
9874 from a common anchor point instead. The effect is similar to the
9875 following pseudocode (which isn't valid C):
9880 register int *xr = &x;
9881 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9885 Not all targets support this option.
9887 @item --param @var{name}=@var{value}
9889 In some places, GCC uses various constants to control the amount of
9890 optimization that is done. For example, GCC does not inline functions
9891 that contain more than a certain number of instructions. You can
9892 control some of these constants on the command line using the
9893 @option{--param} option.
9895 The names of specific parameters, and the meaning of the values, are
9896 tied to the internals of the compiler, and are subject to change
9897 without notice in future releases.
9899 In each case, the @var{value} is an integer. The allowable choices for
9903 @item predictable-branch-outcome
9904 When branch is predicted to be taken with probability lower than this threshold
9905 (in percent), then it is considered well predictable. The default is 10.
9907 @item max-crossjump-edges
9908 The maximum number of incoming edges to consider for cross-jumping.
9909 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9910 the number of edges incoming to each block. Increasing values mean
9911 more aggressive optimization, making the compilation time increase with
9912 probably small improvement in executable size.
9914 @item min-crossjump-insns
9915 The minimum number of instructions that must be matched at the end
9916 of two blocks before cross-jumping is performed on them. This
9917 value is ignored in the case where all instructions in the block being
9918 cross-jumped from are matched. The default value is 5.
9920 @item max-grow-copy-bb-insns
9921 The maximum code size expansion factor when copying basic blocks
9922 instead of jumping. The expansion is relative to a jump instruction.
9923 The default value is 8.
9925 @item max-goto-duplication-insns
9926 The maximum number of instructions to duplicate to a block that jumps
9927 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9928 passes, GCC factors computed gotos early in the compilation process,
9929 and unfactors them as late as possible. Only computed jumps at the
9930 end of a basic blocks with no more than max-goto-duplication-insns are
9931 unfactored. The default value is 8.
9933 @item max-delay-slot-insn-search
9934 The maximum number of instructions to consider when looking for an
9935 instruction to fill a delay slot. If more than this arbitrary number of
9936 instructions are searched, the time savings from filling the delay slot
9937 are minimal, so stop searching. Increasing values mean more
9938 aggressive optimization, making the compilation time increase with probably
9939 small improvement in execution time.
9941 @item max-delay-slot-live-search
9942 When trying to fill delay slots, the maximum number of instructions to
9943 consider when searching for a block with valid live register
9944 information. Increasing this arbitrarily chosen value means more
9945 aggressive optimization, increasing the compilation time. This parameter
9946 should be removed when the delay slot code is rewritten to maintain the
9949 @item max-gcse-memory
9950 The approximate maximum amount of memory that can be allocated in
9951 order to perform the global common subexpression elimination
9952 optimization. If more memory than specified is required, the
9953 optimization is not done.
9955 @item max-gcse-insertion-ratio
9956 If the ratio of expression insertions to deletions is larger than this value
9957 for any expression, then RTL PRE inserts or removes the expression and thus
9958 leaves partially redundant computations in the instruction stream. The default value is 20.
9960 @item max-pending-list-length
9961 The maximum number of pending dependencies scheduling allows
9962 before flushing the current state and starting over. Large functions
9963 with few branches or calls can create excessively large lists which
9964 needlessly consume memory and resources.
9966 @item max-modulo-backtrack-attempts
9967 The maximum number of backtrack attempts the scheduler should make
9968 when modulo scheduling a loop. Larger values can exponentially increase
9971 @item max-inline-insns-single
9972 Several parameters control the tree inliner used in GCC@.
9973 This number sets the maximum number of instructions (counted in GCC's
9974 internal representation) in a single function that the tree inliner
9975 considers for inlining. This only affects functions declared
9976 inline and methods implemented in a class declaration (C++).
9977 The default value is 400.
9979 @item max-inline-insns-auto
9980 When you use @option{-finline-functions} (included in @option{-O3}),
9981 a lot of functions that would otherwise not be considered for inlining
9982 by the compiler are investigated. To those functions, a different
9983 (more restrictive) limit compared to functions declared inline can
9985 The default value is 40.
9987 @item inline-min-speedup
9988 When estimated performance improvement of caller + callee runtime exceeds this
9989 threshold (in precent), the function can be inlined regardless the limit on
9990 @option{--param max-inline-insns-single} and @option{--param
9991 max-inline-insns-auto}.
9993 @item large-function-insns
9994 The limit specifying really large functions. For functions larger than this
9995 limit after inlining, inlining is constrained by
9996 @option{--param large-function-growth}. This parameter is useful primarily
9997 to avoid extreme compilation time caused by non-linear algorithms used by the
9999 The default value is 2700.
10001 @item large-function-growth
10002 Specifies maximal growth of large function caused by inlining in percents.
10003 The default value is 100 which limits large function growth to 2.0 times
10006 @item large-unit-insns
10007 The limit specifying large translation unit. Growth caused by inlining of
10008 units larger than this limit is limited by @option{--param inline-unit-growth}.
10009 For small units this might be too tight.
10010 For example, consider a unit consisting of function A
10011 that is inline and B that just calls A three times. If B is small relative to
10012 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10013 large units consisting of small inlineable functions, however, the overall unit
10014 growth limit is needed to avoid exponential explosion of code size. Thus for
10015 smaller units, the size is increased to @option{--param large-unit-insns}
10016 before applying @option{--param inline-unit-growth}. The default is 10000.
10018 @item inline-unit-growth
10019 Specifies maximal overall growth of the compilation unit caused by inlining.
10020 The default value is 15 which limits unit growth to 1.15 times the original
10021 size. Cold functions (either marked cold via an attribute or by profile
10022 feedback) are not accounted into the unit size.
10024 @item ipcp-unit-growth
10025 Specifies maximal overall growth of the compilation unit caused by
10026 interprocedural constant propagation. The default value is 10 which limits
10027 unit growth to 1.1 times the original size.
10029 @item large-stack-frame
10030 The limit specifying large stack frames. While inlining the algorithm is trying
10031 to not grow past this limit too much. The default value is 256 bytes.
10033 @item large-stack-frame-growth
10034 Specifies maximal growth of large stack frames caused by inlining in percents.
10035 The default value is 1000 which limits large stack frame growth to 11 times
10038 @item max-inline-insns-recursive
10039 @itemx max-inline-insns-recursive-auto
10040 Specifies the maximum number of instructions an out-of-line copy of a
10041 self-recursive inline
10042 function can grow into by performing recursive inlining.
10044 @option{--param max-inline-insns-recursive} applies to functions
10046 For functions not declared inline, recursive inlining
10047 happens only when @option{-finline-functions} (included in @option{-O3}) is
10048 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10049 default value is 450.
10051 @item max-inline-recursive-depth
10052 @itemx max-inline-recursive-depth-auto
10053 Specifies the maximum recursion depth used for recursive inlining.
10055 @option{--param max-inline-recursive-depth} applies to functions
10056 declared inline. For functions not declared inline, recursive inlining
10057 happens only when @option{-finline-functions} (included in @option{-O3}) is
10058 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10059 default value is 8.
10061 @item min-inline-recursive-probability
10062 Recursive inlining is profitable only for function having deep recursion
10063 in average and can hurt for function having little recursion depth by
10064 increasing the prologue size or complexity of function body to other
10067 When profile feedback is available (see @option{-fprofile-generate}) the actual
10068 recursion depth can be guessed from probability that function recurses via a
10069 given call expression. This parameter limits inlining only to call expressions
10070 whose probability exceeds the given threshold (in percents).
10071 The default value is 10.
10073 @item early-inlining-insns
10074 Specify growth that the early inliner can make. In effect it increases
10075 the amount of inlining for code having a large abstraction penalty.
10076 The default value is 14.
10078 @item max-early-inliner-iterations
10079 Limit of iterations of the early inliner. This basically bounds
10080 the number of nested indirect calls the early inliner can resolve.
10081 Deeper chains are still handled by late inlining.
10083 @item comdat-sharing-probability
10084 Probability (in percent) that C++ inline function with comdat visibility
10085 are shared across multiple compilation units. The default value is 20.
10087 @item profile-func-internal-id
10088 A parameter to control whether to use function internal id in profile
10089 database lookup. If the value is 0, the compiler uses an id that
10090 is based on function assembler name and filename, which makes old profile
10091 data more tolerant to source changes such as function reordering etc.
10092 The default value is 0.
10094 @item min-vect-loop-bound
10095 The minimum number of iterations under which loops are not vectorized
10096 when @option{-ftree-vectorize} is used. The number of iterations after
10097 vectorization needs to be greater than the value specified by this option
10098 to allow vectorization. The default value is 0.
10100 @item gcse-cost-distance-ratio
10101 Scaling factor in calculation of maximum distance an expression
10102 can be moved by GCSE optimizations. This is currently supported only in the
10103 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10104 is with simple expressions, i.e., the expressions that have cost
10105 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10106 hoisting of simple expressions. The default value is 10.
10108 @item gcse-unrestricted-cost
10109 Cost, roughly measured as the cost of a single typical machine
10110 instruction, at which GCSE optimizations do not constrain
10111 the distance an expression can travel. This is currently
10112 supported only in the code hoisting pass. The lesser the cost,
10113 the more aggressive code hoisting is. Specifying 0
10114 allows all expressions to travel unrestricted distances.
10115 The default value is 3.
10117 @item max-hoist-depth
10118 The depth of search in the dominator tree for expressions to hoist.
10119 This is used to avoid quadratic behavior in hoisting algorithm.
10120 The value of 0 does not limit on the search, but may slow down compilation
10121 of huge functions. The default value is 30.
10123 @item max-tail-merge-comparisons
10124 The maximum amount of similar bbs to compare a bb with. This is used to
10125 avoid quadratic behavior in tree tail merging. The default value is 10.
10127 @item max-tail-merge-iterations
10128 The maximum amount of iterations of the pass over the function. This is used to
10129 limit compilation time in tree tail merging. The default value is 2.
10131 @item max-unrolled-insns
10132 The maximum number of instructions that a loop may have to be unrolled.
10133 If a loop is unrolled, this parameter also determines how many times
10134 the loop code is unrolled.
10136 @item max-average-unrolled-insns
10137 The maximum number of instructions biased by probabilities of their execution
10138 that a loop may have to be unrolled. If a loop is unrolled,
10139 this parameter also determines how many times the loop code is unrolled.
10141 @item max-unroll-times
10142 The maximum number of unrollings of a single loop.
10144 @item max-peeled-insns
10145 The maximum number of instructions that a loop may have to be peeled.
10146 If a loop is peeled, this parameter also determines how many times
10147 the loop code is peeled.
10149 @item max-peel-times
10150 The maximum number of peelings of a single loop.
10152 @item max-peel-branches
10153 The maximum number of branches on the hot path through the peeled sequence.
10155 @item max-completely-peeled-insns
10156 The maximum number of insns of a completely peeled loop.
10158 @item max-completely-peel-times
10159 The maximum number of iterations of a loop to be suitable for complete peeling.
10161 @item max-completely-peel-loop-nest-depth
10162 The maximum depth of a loop nest suitable for complete peeling.
10164 @item max-unswitch-insns
10165 The maximum number of insns of an unswitched loop.
10167 @item max-unswitch-level
10168 The maximum number of branches unswitched in a single loop.
10170 @item lim-expensive
10171 The minimum cost of an expensive expression in the loop invariant motion.
10173 @item iv-consider-all-candidates-bound
10174 Bound on number of candidates for induction variables, below which
10175 all candidates are considered for each use in induction variable
10176 optimizations. If there are more candidates than this,
10177 only the most relevant ones are considered to avoid quadratic time complexity.
10179 @item iv-max-considered-uses
10180 The induction variable optimizations give up on loops that contain more
10181 induction variable uses.
10183 @item iv-always-prune-cand-set-bound
10184 If the number of candidates in the set is smaller than this value,
10185 always try to remove unnecessary ivs from the set
10186 when adding a new one.
10188 @item scev-max-expr-size
10189 Bound on size of expressions used in the scalar evolutions analyzer.
10190 Large expressions slow the analyzer.
10192 @item scev-max-expr-complexity
10193 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10194 Complex expressions slow the analyzer.
10196 @item omega-max-vars
10197 The maximum number of variables in an Omega constraint system.
10198 The default value is 128.
10200 @item omega-max-geqs
10201 The maximum number of inequalities in an Omega constraint system.
10202 The default value is 256.
10204 @item omega-max-eqs
10205 The maximum number of equalities in an Omega constraint system.
10206 The default value is 128.
10208 @item omega-max-wild-cards
10209 The maximum number of wildcard variables that the Omega solver is
10210 able to insert. The default value is 18.
10212 @item omega-hash-table-size
10213 The size of the hash table in the Omega solver. The default value is
10216 @item omega-max-keys
10217 The maximal number of keys used by the Omega solver. The default
10220 @item omega-eliminate-redundant-constraints
10221 When set to 1, use expensive methods to eliminate all redundant
10222 constraints. The default value is 0.
10224 @item vect-max-version-for-alignment-checks
10225 The maximum number of run-time checks that can be performed when
10226 doing loop versioning for alignment in the vectorizer.
10228 @item vect-max-version-for-alias-checks
10229 The maximum number of run-time checks that can be performed when
10230 doing loop versioning for alias in the vectorizer.
10232 @item vect-max-peeling-for-alignment
10233 The maximum number of loop peels to enhance access alignment
10234 for vectorizer. Value -1 means 'no limit'.
10236 @item max-iterations-to-track
10237 The maximum number of iterations of a loop the brute-force algorithm
10238 for analysis of the number of iterations of the loop tries to evaluate.
10240 @item hot-bb-count-ws-permille
10241 A basic block profile count is considered hot if it contributes to
10242 the given permillage (i.e. 0...1000) of the entire profiled execution.
10244 @item hot-bb-frequency-fraction
10245 Select fraction of the entry block frequency of executions of basic block in
10246 function given basic block needs to have to be considered hot.
10248 @item max-predicted-iterations
10249 The maximum number of loop iterations we predict statically. This is useful
10250 in cases where a function contains a single loop with known bound and
10251 another loop with unknown bound.
10252 The known number of iterations is predicted correctly, while
10253 the unknown number of iterations average to roughly 10. This means that the
10254 loop without bounds appears artificially cold relative to the other one.
10256 @item builtin-expect-probability
10257 Control the probability of the expression having the specified value. This
10258 parameter takes a percentage (i.e. 0 ... 100) as input.
10259 The default probability of 90 is obtained empirically.
10261 @item align-threshold
10263 Select fraction of the maximal frequency of executions of a basic block in
10264 a function to align the basic block.
10266 @item align-loop-iterations
10268 A loop expected to iterate at least the selected number of iterations is
10271 @item tracer-dynamic-coverage
10272 @itemx tracer-dynamic-coverage-feedback
10274 This value is used to limit superblock formation once the given percentage of
10275 executed instructions is covered. This limits unnecessary code size
10278 The @option{tracer-dynamic-coverage-feedback} is used only when profile
10279 feedback is available. The real profiles (as opposed to statically estimated
10280 ones) are much less balanced allowing the threshold to be larger value.
10282 @item tracer-max-code-growth
10283 Stop tail duplication once code growth has reached given percentage. This is
10284 a rather artificial limit, as most of the duplicates are eliminated later in
10285 cross jumping, so it may be set to much higher values than is the desired code
10288 @item tracer-min-branch-ratio
10290 Stop reverse growth when the reverse probability of best edge is less than this
10291 threshold (in percent).
10293 @item tracer-min-branch-ratio
10294 @itemx tracer-min-branch-ratio-feedback
10296 Stop forward growth if the best edge has probability lower than this
10299 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10300 compilation for profile feedback and one for compilation without. The value
10301 for compilation with profile feedback needs to be more conservative (higher) in
10302 order to make tracer effective.
10304 @item max-cse-path-length
10306 The maximum number of basic blocks on path that CSE considers.
10309 @item max-cse-insns
10310 The maximum number of instructions CSE processes before flushing.
10311 The default is 1000.
10313 @item ggc-min-expand
10315 GCC uses a garbage collector to manage its own memory allocation. This
10316 parameter specifies the minimum percentage by which the garbage
10317 collector's heap should be allowed to expand between collections.
10318 Tuning this may improve compilation speed; it has no effect on code
10321 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10322 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10323 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10324 GCC is not able to calculate RAM on a particular platform, the lower
10325 bound of 30% is used. Setting this parameter and
10326 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10327 every opportunity. This is extremely slow, but can be useful for
10330 @item ggc-min-heapsize
10332 Minimum size of the garbage collector's heap before it begins bothering
10333 to collect garbage. The first collection occurs after the heap expands
10334 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10335 tuning this may improve compilation speed, and has no effect on code
10338 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10339 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10340 with a lower bound of 4096 (four megabytes) and an upper bound of
10341 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10342 particular platform, the lower bound is used. Setting this parameter
10343 very large effectively disables garbage collection. Setting this
10344 parameter and @option{ggc-min-expand} to zero causes a full collection
10345 to occur at every opportunity.
10347 @item max-reload-search-insns
10348 The maximum number of instruction reload should look backward for equivalent
10349 register. Increasing values mean more aggressive optimization, making the
10350 compilation time increase with probably slightly better performance.
10351 The default value is 100.
10353 @item max-cselib-memory-locations
10354 The maximum number of memory locations cselib should take into account.
10355 Increasing values mean more aggressive optimization, making the compilation time
10356 increase with probably slightly better performance. The default value is 500.
10358 @item reorder-blocks-duplicate
10359 @itemx reorder-blocks-duplicate-feedback
10361 Used by the basic block reordering pass to decide whether to use unconditional
10362 branch or duplicate the code on its destination. Code is duplicated when its
10363 estimated size is smaller than this value multiplied by the estimated size of
10364 unconditional jump in the hot spots of the program.
10366 The @option{reorder-block-duplicate-feedback} is used only when profile
10367 feedback is available. It may be set to higher values than
10368 @option{reorder-block-duplicate} since information about the hot spots is more
10371 @item max-sched-ready-insns
10372 The maximum number of instructions ready to be issued the scheduler should
10373 consider at any given time during the first scheduling pass. Increasing
10374 values mean more thorough searches, making the compilation time increase
10375 with probably little benefit. The default value is 100.
10377 @item max-sched-region-blocks
10378 The maximum number of blocks in a region to be considered for
10379 interblock scheduling. The default value is 10.
10381 @item max-pipeline-region-blocks
10382 The maximum number of blocks in a region to be considered for
10383 pipelining in the selective scheduler. The default value is 15.
10385 @item max-sched-region-insns
10386 The maximum number of insns in a region to be considered for
10387 interblock scheduling. The default value is 100.
10389 @item max-pipeline-region-insns
10390 The maximum number of insns in a region to be considered for
10391 pipelining in the selective scheduler. The default value is 200.
10393 @item min-spec-prob
10394 The minimum probability (in percents) of reaching a source block
10395 for interblock speculative scheduling. The default value is 40.
10397 @item max-sched-extend-regions-iters
10398 The maximum number of iterations through CFG to extend regions.
10399 A value of 0 (the default) disables region extensions.
10401 @item max-sched-insn-conflict-delay
10402 The maximum conflict delay for an insn to be considered for speculative motion.
10403 The default value is 3.
10405 @item sched-spec-prob-cutoff
10406 The minimal probability of speculation success (in percents), so that
10407 speculative insns are scheduled.
10408 The default value is 40.
10410 @item sched-spec-state-edge-prob-cutoff
10411 The minimum probability an edge must have for the scheduler to save its
10413 The default value is 10.
10415 @item sched-mem-true-dep-cost
10416 Minimal distance (in CPU cycles) between store and load targeting same
10417 memory locations. The default value is 1.
10419 @item selsched-max-lookahead
10420 The maximum size of the lookahead window of selective scheduling. It is a
10421 depth of search for available instructions.
10422 The default value is 50.
10424 @item selsched-max-sched-times
10425 The maximum number of times that an instruction is scheduled during
10426 selective scheduling. This is the limit on the number of iterations
10427 through which the instruction may be pipelined. The default value is 2.
10429 @item selsched-max-insns-to-rename
10430 The maximum number of best instructions in the ready list that are considered
10431 for renaming in the selective scheduler. The default value is 2.
10434 The minimum value of stage count that swing modulo scheduler
10435 generates. The default value is 2.
10437 @item max-last-value-rtl
10438 The maximum size measured as number of RTLs that can be recorded in an expression
10439 in combiner for a pseudo register as last known value of that register. The default
10442 @item max-combine-insns
10443 The maximum number of instructions the RTL combiner tries to combine.
10444 The default value is 2 at @option{-Og} and 4 otherwise.
10446 @item integer-share-limit
10447 Small integer constants can use a shared data structure, reducing the
10448 compiler's memory usage and increasing its speed. This sets the maximum
10449 value of a shared integer constant. The default value is 256.
10451 @item ssp-buffer-size
10452 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10453 protection when @option{-fstack-protection} is used.
10455 @item min-size-for-stack-sharing
10456 The minimum size of variables taking part in stack slot sharing when not
10457 optimizing. The default value is 32.
10459 @item max-jump-thread-duplication-stmts
10460 Maximum number of statements allowed in a block that needs to be
10461 duplicated when threading jumps.
10463 @item max-fields-for-field-sensitive
10464 Maximum number of fields in a structure treated in
10465 a field sensitive manner during pointer analysis. The default is zero
10466 for @option{-O0} and @option{-O1},
10467 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10469 @item prefetch-latency
10470 Estimate on average number of instructions that are executed before
10471 prefetch finishes. The distance prefetched ahead is proportional
10472 to this constant. Increasing this number may also lead to less
10473 streams being prefetched (see @option{simultaneous-prefetches}).
10475 @item simultaneous-prefetches
10476 Maximum number of prefetches that can run at the same time.
10478 @item l1-cache-line-size
10479 The size of cache line in L1 cache, in bytes.
10481 @item l1-cache-size
10482 The size of L1 cache, in kilobytes.
10484 @item l2-cache-size
10485 The size of L2 cache, in kilobytes.
10487 @item min-insn-to-prefetch-ratio
10488 The minimum ratio between the number of instructions and the
10489 number of prefetches to enable prefetching in a loop.
10491 @item prefetch-min-insn-to-mem-ratio
10492 The minimum ratio between the number of instructions and the
10493 number of memory references to enable prefetching in a loop.
10495 @item use-canonical-types
10496 Whether the compiler should use the ``canonical'' type system. By
10497 default, this should always be 1, which uses a more efficient internal
10498 mechanism for comparing types in C++ and Objective-C++. However, if
10499 bugs in the canonical type system are causing compilation failures,
10500 set this value to 0 to disable canonical types.
10502 @item switch-conversion-max-branch-ratio
10503 Switch initialization conversion refuses to create arrays that are
10504 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10505 branches in the switch.
10507 @item max-partial-antic-length
10508 Maximum length of the partial antic set computed during the tree
10509 partial redundancy elimination optimization (@option{-ftree-pre}) when
10510 optimizing at @option{-O3} and above. For some sorts of source code
10511 the enhanced partial redundancy elimination optimization can run away,
10512 consuming all of the memory available on the host machine. This
10513 parameter sets a limit on the length of the sets that are computed,
10514 which prevents the runaway behavior. Setting a value of 0 for
10515 this parameter allows an unlimited set length.
10517 @item sccvn-max-scc-size
10518 Maximum size of a strongly connected component (SCC) during SCCVN
10519 processing. If this limit is hit, SCCVN processing for the whole
10520 function is not done and optimizations depending on it are
10521 disabled. The default maximum SCC size is 10000.
10523 @item sccvn-max-alias-queries-per-access
10524 Maximum number of alias-oracle queries we perform when looking for
10525 redundancies for loads and stores. If this limit is hit the search
10526 is aborted and the load or store is not considered redundant. The
10527 number of queries is algorithmically limited to the number of
10528 stores on all paths from the load to the function entry.
10529 The default maxmimum number of queries is 1000.
10531 @item ira-max-loops-num
10532 IRA uses regional register allocation by default. If a function
10533 contains more loops than the number given by this parameter, only at most
10534 the given number of the most frequently-executed loops form regions
10535 for regional register allocation. The default value of the
10538 @item ira-max-conflict-table-size
10539 Although IRA uses a sophisticated algorithm to compress the conflict
10540 table, the table can still require excessive amounts of memory for
10541 huge functions. If the conflict table for a function could be more
10542 than the size in MB given by this parameter, the register allocator
10543 instead uses a faster, simpler, and lower-quality
10544 algorithm that does not require building a pseudo-register conflict table.
10545 The default value of the parameter is 2000.
10547 @item ira-loop-reserved-regs
10548 IRA can be used to evaluate more accurate register pressure in loops
10549 for decisions to move loop invariants (see @option{-O3}). The number
10550 of available registers reserved for some other purposes is given
10551 by this parameter. The default value of the parameter is 2, which is
10552 the minimal number of registers needed by typical instructions.
10553 This value is the best found from numerous experiments.
10555 @item loop-invariant-max-bbs-in-loop
10556 Loop invariant motion can be very expensive, both in compilation time and
10557 in amount of needed compile-time memory, with very large loops. Loops
10558 with more basic blocks than this parameter won't have loop invariant
10559 motion optimization performed on them. The default value of the
10560 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10562 @item loop-max-datarefs-for-datadeps
10563 Building data dapendencies is expensive for very large loops. This
10564 parameter limits the number of data references in loops that are
10565 considered for data dependence analysis. These large loops are no
10566 handled by the optimizations using loop data dependencies.
10567 The default value is 1000.
10569 @item max-vartrack-size
10570 Sets a maximum number of hash table slots to use during variable
10571 tracking dataflow analysis of any function. If this limit is exceeded
10572 with variable tracking at assignments enabled, analysis for that
10573 function is retried without it, after removing all debug insns from
10574 the function. If the limit is exceeded even without debug insns, var
10575 tracking analysis is completely disabled for the function. Setting
10576 the parameter to zero makes it unlimited.
10578 @item max-vartrack-expr-depth
10579 Sets a maximum number of recursion levels when attempting to map
10580 variable names or debug temporaries to value expressions. This trades
10581 compilation time for more complete debug information. If this is set too
10582 low, value expressions that are available and could be represented in
10583 debug information may end up not being used; setting this higher may
10584 enable the compiler to find more complex debug expressions, but compile
10585 time and memory use may grow. The default is 12.
10587 @item min-nondebug-insn-uid
10588 Use uids starting at this parameter for nondebug insns. The range below
10589 the parameter is reserved exclusively for debug insns created by
10590 @option{-fvar-tracking-assignments}, but debug insns may get
10591 (non-overlapping) uids above it if the reserved range is exhausted.
10593 @item ipa-sra-ptr-growth-factor
10594 IPA-SRA replaces a pointer to an aggregate with one or more new
10595 parameters only when their cumulative size is less or equal to
10596 @option{ipa-sra-ptr-growth-factor} times the size of the original
10599 @item sra-max-scalarization-size-Ospeed
10600 @item sra-max-scalarization-size-Osize
10601 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10602 replace scalar parts of aggregates with uses of independent scalar
10603 variables. These parameters control the maximum size, in storage units,
10604 of aggregate which is considered for replacement when compiling for
10606 (@option{sra-max-scalarization-size-Ospeed}) or size
10607 (@option{sra-max-scalarization-size-Osize}) respectively.
10609 @item tm-max-aggregate-size
10610 When making copies of thread-local variables in a transaction, this
10611 parameter specifies the size in bytes after which variables are
10612 saved with the logging functions as opposed to save/restore code
10613 sequence pairs. This option only applies when using
10616 @item graphite-max-nb-scop-params
10617 To avoid exponential effects in the Graphite loop transforms, the
10618 number of parameters in a Static Control Part (SCoP) is bounded. The
10619 default value is 10 parameters. A variable whose value is unknown at
10620 compilation time and defined outside a SCoP is a parameter of the SCoP.
10622 @item graphite-max-bbs-per-function
10623 To avoid exponential effects in the detection of SCoPs, the size of
10624 the functions analyzed by Graphite is bounded. The default value is
10627 @item loop-block-tile-size
10628 Loop blocking or strip mining transforms, enabled with
10629 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10630 loop in the loop nest by a given number of iterations. The strip
10631 length can be changed using the @option{loop-block-tile-size}
10632 parameter. The default value is 51 iterations.
10634 @item loop-unroll-jam-size
10635 Specify the unroll factor for the @option{-floop-unroll-and-jam}. The
10636 default value is 4.
10638 @item loop-unroll-jam-depth
10639 Specify the dimension to be unrolled (counting from the most inner loop)
10640 for the @option{-floop-unroll-and-jam}. The default value is 2.
10642 @item ipa-cp-value-list-size
10643 IPA-CP attempts to track all possible values and types passed to a function's
10644 parameter in order to propagate them and perform devirtualization.
10645 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10646 stores per one formal parameter of a function.
10648 @item ipa-cp-eval-threshold
10649 IPA-CP calculates its own score of cloning profitability heuristics
10650 and performs those cloning opportunities with scores that exceed
10651 @option{ipa-cp-eval-threshold}.
10653 @item ipa-max-agg-items
10654 IPA-CP is also capable to propagate a number of scalar values passed
10655 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10656 number of such values per one parameter.
10658 @item ipa-cp-loop-hint-bonus
10659 When IPA-CP determines that a cloning candidate would make the number
10660 of iterations of a loop known, it adds a bonus of
10661 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10664 @item ipa-cp-array-index-hint-bonus
10665 When IPA-CP determines that a cloning candidate would make the index of
10666 an array access known, it adds a bonus of
10667 @option{ipa-cp-array-index-hint-bonus} to the profitability
10668 score of the candidate.
10670 @item ipa-max-aa-steps
10671 During its analysis of function bodies, IPA-CP employs alias analysis
10672 in order to track values pointed to by function parameters. In order
10673 not spend too much time analyzing huge functions, it gives up and
10674 consider all memory clobbered after examining
10675 @option{ipa-max-aa-steps} statements modifying memory.
10677 @item lto-partitions
10678 Specify desired number of partitions produced during WHOPR compilation.
10679 The number of partitions should exceed the number of CPUs used for compilation.
10680 The default value is 32.
10682 @item lto-minpartition
10683 Size of minimal partition for WHOPR (in estimated instructions).
10684 This prevents expenses of splitting very small programs into too many
10687 @item cxx-max-namespaces-for-diagnostic-help
10688 The maximum number of namespaces to consult for suggestions when C++
10689 name lookup fails for an identifier. The default is 1000.
10691 @item sink-frequency-threshold
10692 The maximum relative execution frequency (in percents) of the target block
10693 relative to a statement's original block to allow statement sinking of a
10694 statement. Larger numbers result in more aggressive statement sinking.
10695 The default value is 75. A small positive adjustment is applied for
10696 statements with memory operands as those are even more profitable so sink.
10698 @item max-stores-to-sink
10699 The maximum number of conditional stores paires that can be sunk. Set to 0
10700 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10701 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10703 @item allow-store-data-races
10704 Allow optimizers to introduce new data races on stores.
10705 Set to 1 to allow, otherwise to 0. This option is enabled by default
10706 at optimization level @option{-Ofast}.
10708 @item case-values-threshold
10709 The smallest number of different values for which it is best to use a
10710 jump-table instead of a tree of conditional branches. If the value is
10711 0, use the default for the machine. The default is 0.
10713 @item tree-reassoc-width
10714 Set the maximum number of instructions executed in parallel in
10715 reassociated tree. This parameter overrides target dependent
10716 heuristics used by default if has non zero value.
10718 @item sched-pressure-algorithm
10719 Choose between the two available implementations of
10720 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10721 and is the more likely to prevent instructions from being reordered.
10722 Algorithm 2 was designed to be a compromise between the relatively
10723 conservative approach taken by algorithm 1 and the rather aggressive
10724 approach taken by the default scheduler. It relies more heavily on
10725 having a regular register file and accurate register pressure classes.
10726 See @file{haifa-sched.c} in the GCC sources for more details.
10728 The default choice depends on the target.
10730 @item max-slsr-cand-scan
10731 Set the maximum number of existing candidates that are considered when
10732 seeking a basis for a new straight-line strength reduction candidate.
10735 Enable buffer overflow detection for global objects. This kind
10736 of protection is enabled by default if you are using
10737 @option{-fsanitize=address} option.
10738 To disable global objects protection use @option{--param asan-globals=0}.
10741 Enable buffer overflow detection for stack objects. This kind of
10742 protection is enabled by default when using@option{-fsanitize=address}.
10743 To disable stack protection use @option{--param asan-stack=0} option.
10745 @item asan-instrument-reads
10746 Enable buffer overflow detection for memory reads. This kind of
10747 protection is enabled by default when using @option{-fsanitize=address}.
10748 To disable memory reads protection use
10749 @option{--param asan-instrument-reads=0}.
10751 @item asan-instrument-writes
10752 Enable buffer overflow detection for memory writes. This kind of
10753 protection is enabled by default when using @option{-fsanitize=address}.
10754 To disable memory writes protection use
10755 @option{--param asan-instrument-writes=0} option.
10757 @item asan-memintrin
10758 Enable detection for built-in functions. This kind of protection
10759 is enabled by default when using @option{-fsanitize=address}.
10760 To disable built-in functions protection use
10761 @option{--param asan-memintrin=0}.
10763 @item asan-use-after-return
10764 Enable detection of use-after-return. This kind of protection
10765 is enabled by default when using @option{-fsanitize=address} option.
10766 To disable use-after-return detection use
10767 @option{--param asan-use-after-return=0}.
10769 @item asan-instrumentation-with-call-threshold
10770 If number of memory accesses in function being instrumented
10771 is greater or equal to this number, use callbacks instead of inline checks.
10772 E.g. to disable inline code use
10773 @option{--param asan-instrumentation-with-call-threshold=0}.
10775 @item chkp-max-ctor-size
10776 Static constructors generated by Pointer Bounds Checker may become very
10777 large and significantly increase compile time at optimization level
10778 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10779 in a single generated constructor. Default value is 5000.
10781 @item max-fsm-thread-path-insns
10782 Maximum number of instructions to copy when duplicating blocks on a
10783 finite state automaton jump thread path. The default is 100.
10785 @item max-fsm-thread-length
10786 Maximum number of basic blocks on a finite state automaton jump thread
10787 path. The default is 10.
10789 @item max-fsm-thread-paths
10790 Maximum number of new jump thread paths to create for a finite state
10791 automaton. The default is 50.
10796 @node Preprocessor Options
10797 @section Options Controlling the Preprocessor
10798 @cindex preprocessor options
10799 @cindex options, preprocessor
10801 These options control the C preprocessor, which is run on each C source
10802 file before actual compilation.
10804 If you use the @option{-E} option, nothing is done except preprocessing.
10805 Some of these options make sense only together with @option{-E} because
10806 they cause the preprocessor output to be unsuitable for actual
10810 @item -Wp,@var{option}
10812 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10813 and pass @var{option} directly through to the preprocessor. If
10814 @var{option} contains commas, it is split into multiple options at the
10815 commas. However, many options are modified, translated or interpreted
10816 by the compiler driver before being passed to the preprocessor, and
10817 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10818 interface is undocumented and subject to change, so whenever possible
10819 you should avoid using @option{-Wp} and let the driver handle the
10822 @item -Xpreprocessor @var{option}
10823 @opindex Xpreprocessor
10824 Pass @var{option} as an option to the preprocessor. You can use this to
10825 supply system-specific preprocessor options that GCC does not
10828 If you want to pass an option that takes an argument, you must use
10829 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10831 @item -no-integrated-cpp
10832 @opindex no-integrated-cpp
10833 Perform preprocessing as a separate pass before compilation.
10834 By default, GCC performs preprocessing as an integrated part of
10835 input tokenization and parsing.
10836 If this option is provided, the appropriate language front end
10837 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10838 and Objective-C, respectively) is instead invoked twice,
10839 once for preprocessing only and once for actual compilation
10840 of the preprocessed input.
10841 This option may be useful in conjunction with the @option{-B} or
10842 @option{-wrapper} options to specify an alternate preprocessor or
10843 perform additional processing of the program source between
10844 normal preprocessing and compilation.
10847 @include cppopts.texi
10849 @node Assembler Options
10850 @section Passing Options to the Assembler
10852 @c prevent bad page break with this line
10853 You can pass options to the assembler.
10856 @item -Wa,@var{option}
10858 Pass @var{option} as an option to the assembler. If @var{option}
10859 contains commas, it is split into multiple options at the commas.
10861 @item -Xassembler @var{option}
10862 @opindex Xassembler
10863 Pass @var{option} as an option to the assembler. You can use this to
10864 supply system-specific assembler options that GCC does not
10867 If you want to pass an option that takes an argument, you must use
10868 @option{-Xassembler} twice, once for the option and once for the argument.
10873 @section Options for Linking
10874 @cindex link options
10875 @cindex options, linking
10877 These options come into play when the compiler links object files into
10878 an executable output file. They are meaningless if the compiler is
10879 not doing a link step.
10883 @item @var{object-file-name}
10884 A file name that does not end in a special recognized suffix is
10885 considered to name an object file or library. (Object files are
10886 distinguished from libraries by the linker according to the file
10887 contents.) If linking is done, these object files are used as input
10896 If any of these options is used, then the linker is not run, and
10897 object file names should not be used as arguments. @xref{Overall
10901 @opindex fuse-ld=bfd
10902 Use the @command{bfd} linker instead of the default linker.
10904 @item -fuse-ld=gold
10905 @opindex fuse-ld=gold
10906 Use the @command{gold} linker instead of the default linker.
10909 @item -l@var{library}
10910 @itemx -l @var{library}
10912 Search the library named @var{library} when linking. (The second
10913 alternative with the library as a separate argument is only for
10914 POSIX compliance and is not recommended.)
10916 It makes a difference where in the command you write this option; the
10917 linker searches and processes libraries and object files in the order they
10918 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10919 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10920 to functions in @samp{z}, those functions may not be loaded.
10922 The linker searches a standard list of directories for the library,
10923 which is actually a file named @file{lib@var{library}.a}. The linker
10924 then uses this file as if it had been specified precisely by name.
10926 The directories searched include several standard system directories
10927 plus any that you specify with @option{-L}.
10929 Normally the files found this way are library files---archive files
10930 whose members are object files. The linker handles an archive file by
10931 scanning through it for members which define symbols that have so far
10932 been referenced but not defined. But if the file that is found is an
10933 ordinary object file, it is linked in the usual fashion. The only
10934 difference between using an @option{-l} option and specifying a file name
10935 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10936 and searches several directories.
10940 You need this special case of the @option{-l} option in order to
10941 link an Objective-C or Objective-C++ program.
10943 @item -nostartfiles
10944 @opindex nostartfiles
10945 Do not use the standard system startup files when linking.
10946 The standard system libraries are used normally, unless @option{-nostdlib}
10947 or @option{-nodefaultlibs} is used.
10949 @item -nodefaultlibs
10950 @opindex nodefaultlibs
10951 Do not use the standard system libraries when linking.
10952 Only the libraries you specify are passed to the linker, and options
10953 specifying linkage of the system libraries, such as @option{-static-libgcc}
10954 or @option{-shared-libgcc}, are ignored.
10955 The standard startup files are used normally, unless @option{-nostartfiles}
10958 The compiler may generate calls to @code{memcmp},
10959 @code{memset}, @code{memcpy} and @code{memmove}.
10960 These entries are usually resolved by entries in
10961 libc. These entry points should be supplied through some other
10962 mechanism when this option is specified.
10966 Do not use the standard system startup files or libraries when linking.
10967 No startup files and only the libraries you specify are passed to
10968 the linker, and options specifying linkage of the system libraries, such as
10969 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
10971 The compiler may generate calls to @code{memcmp}, @code{memset},
10972 @code{memcpy} and @code{memmove}.
10973 These entries are usually resolved by entries in
10974 libc. These entry points should be supplied through some other
10975 mechanism when this option is specified.
10977 @cindex @option{-lgcc}, use with @option{-nostdlib}
10978 @cindex @option{-nostdlib} and unresolved references
10979 @cindex unresolved references and @option{-nostdlib}
10980 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10981 @cindex @option{-nodefaultlibs} and unresolved references
10982 @cindex unresolved references and @option{-nodefaultlibs}
10983 One of the standard libraries bypassed by @option{-nostdlib} and
10984 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10985 which GCC uses to overcome shortcomings of particular machines, or special
10986 needs for some languages.
10987 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10988 Collection (GCC) Internals},
10989 for more discussion of @file{libgcc.a}.)
10990 In most cases, you need @file{libgcc.a} even when you want to avoid
10991 other standard libraries. In other words, when you specify @option{-nostdlib}
10992 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10993 This ensures that you have no unresolved references to internal GCC
10994 library subroutines.
10995 (An example of such an internal subroutine is @code{__main}, used to ensure C++
10996 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10997 GNU Compiler Collection (GCC) Internals}.)
11001 Produce a position independent executable on targets that support it.
11002 For predictable results, you must also specify the same set of options
11003 used for compilation (@option{-fpie}, @option{-fPIE},
11004 or model suboptions) when you specify this linker option.
11008 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11009 that support it. This instructs the linker to add all symbols, not
11010 only used ones, to the dynamic symbol table. This option is needed
11011 for some uses of @code{dlopen} or to allow obtaining backtraces
11012 from within a program.
11016 Remove all symbol table and relocation information from the executable.
11020 On systems that support dynamic linking, this prevents linking with the shared
11021 libraries. On other systems, this option has no effect.
11025 Produce a shared object which can then be linked with other objects to
11026 form an executable. Not all systems support this option. For predictable
11027 results, you must also specify the same set of options used for compilation
11028 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11029 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11030 needs to build supplementary stub code for constructors to work. On
11031 multi-libbed systems, @samp{gcc -shared} must select the correct support
11032 libraries to link against. Failing to supply the correct flags may lead
11033 to subtle defects. Supplying them in cases where they are not necessary
11036 @item -shared-libgcc
11037 @itemx -static-libgcc
11038 @opindex shared-libgcc
11039 @opindex static-libgcc
11040 On systems that provide @file{libgcc} as a shared library, these options
11041 force the use of either the shared or static version, respectively.
11042 If no shared version of @file{libgcc} was built when the compiler was
11043 configured, these options have no effect.
11045 There are several situations in which an application should use the
11046 shared @file{libgcc} instead of the static version. The most common
11047 of these is when the application wishes to throw and catch exceptions
11048 across different shared libraries. In that case, each of the libraries
11049 as well as the application itself should use the shared @file{libgcc}.
11051 Therefore, the G++ and GCJ drivers automatically add
11052 @option{-shared-libgcc} whenever you build a shared library or a main
11053 executable, because C++ and Java programs typically use exceptions, so
11054 this is the right thing to do.
11056 If, instead, you use the GCC driver to create shared libraries, you may
11057 find that they are not always linked with the shared @file{libgcc}.
11058 If GCC finds, at its configuration time, that you have a non-GNU linker
11059 or a GNU linker that does not support option @option{--eh-frame-hdr},
11060 it links the shared version of @file{libgcc} into shared libraries
11061 by default. Otherwise, it takes advantage of the linker and optimizes
11062 away the linking with the shared version of @file{libgcc}, linking with
11063 the static version of libgcc by default. This allows exceptions to
11064 propagate through such shared libraries, without incurring relocation
11065 costs at library load time.
11067 However, if a library or main executable is supposed to throw or catch
11068 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11069 for the languages used in the program, or using the option
11070 @option{-shared-libgcc}, such that it is linked with the shared
11073 @item -static-libasan
11074 @opindex static-libasan
11075 When the @option{-fsanitize=address} option is used to link a program,
11076 the GCC driver automatically links against @option{libasan}. If
11077 @file{libasan} is available as a shared library, and the @option{-static}
11078 option is not used, then this links against the shared version of
11079 @file{libasan}. The @option{-static-libasan} option directs the GCC
11080 driver to link @file{libasan} statically, without necessarily linking
11081 other libraries statically.
11083 @item -static-libtsan
11084 @opindex static-libtsan
11085 When the @option{-fsanitize=thread} option is used to link a program,
11086 the GCC driver automatically links against @option{libtsan}. If
11087 @file{libtsan} is available as a shared library, and the @option{-static}
11088 option is not used, then this links against the shared version of
11089 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11090 driver to link @file{libtsan} statically, without necessarily linking
11091 other libraries statically.
11093 @item -static-liblsan
11094 @opindex static-liblsan
11095 When the @option{-fsanitize=leak} option is used to link a program,
11096 the GCC driver automatically links against @option{liblsan}. If
11097 @file{liblsan} is available as a shared library, and the @option{-static}
11098 option is not used, then this links against the shared version of
11099 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11100 driver to link @file{liblsan} statically, without necessarily linking
11101 other libraries statically.
11103 @item -static-libubsan
11104 @opindex static-libubsan
11105 When the @option{-fsanitize=undefined} option is used to link a program,
11106 the GCC driver automatically links against @option{libubsan}. If
11107 @file{libubsan} is available as a shared library, and the @option{-static}
11108 option is not used, then this links against the shared version of
11109 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11110 driver to link @file{libubsan} statically, without necessarily linking
11111 other libraries statically.
11113 @item -static-libstdc++
11114 @opindex static-libstdc++
11115 When the @command{g++} program is used to link a C++ program, it
11116 normally automatically links against @option{libstdc++}. If
11117 @file{libstdc++} is available as a shared library, and the
11118 @option{-static} option is not used, then this links against the
11119 shared version of @file{libstdc++}. That is normally fine. However, it
11120 is sometimes useful to freeze the version of @file{libstdc++} used by
11121 the program without going all the way to a fully static link. The
11122 @option{-static-libstdc++} option directs the @command{g++} driver to
11123 link @file{libstdc++} statically, without necessarily linking other
11124 libraries statically.
11128 Bind references to global symbols when building a shared object. Warn
11129 about any unresolved references (unless overridden by the link editor
11130 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11133 @item -T @var{script}
11135 @cindex linker script
11136 Use @var{script} as the linker script. This option is supported by most
11137 systems using the GNU linker. On some targets, such as bare-board
11138 targets without an operating system, the @option{-T} option may be required
11139 when linking to avoid references to undefined symbols.
11141 @item -Xlinker @var{option}
11143 Pass @var{option} as an option to the linker. You can use this to
11144 supply system-specific linker options that GCC does not recognize.
11146 If you want to pass an option that takes a separate argument, you must use
11147 @option{-Xlinker} twice, once for the option and once for the argument.
11148 For example, to pass @option{-assert definitions}, you must write
11149 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11150 @option{-Xlinker "-assert definitions"}, because this passes the entire
11151 string as a single argument, which is not what the linker expects.
11153 When using the GNU linker, it is usually more convenient to pass
11154 arguments to linker options using the @option{@var{option}=@var{value}}
11155 syntax than as separate arguments. For example, you can specify
11156 @option{-Xlinker -Map=output.map} rather than
11157 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11158 this syntax for command-line options.
11160 @item -Wl,@var{option}
11162 Pass @var{option} as an option to the linker. If @var{option} contains
11163 commas, it is split into multiple options at the commas. You can use this
11164 syntax to pass an argument to the option.
11165 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11166 linker. When using the GNU linker, you can also get the same effect with
11167 @option{-Wl,-Map=output.map}.
11169 @item -u @var{symbol}
11171 Pretend the symbol @var{symbol} is undefined, to force linking of
11172 library modules to define it. You can use @option{-u} multiple times with
11173 different symbols to force loading of additional library modules.
11175 @item -z @var{keyword}
11177 @option{-z} is passed directly on to the linker along with the keyword
11178 @var{keyword}. See the section in the documentation of your linker for
11179 permitted values and their meanings.
11182 @node Directory Options
11183 @section Options for Directory Search
11184 @cindex directory options
11185 @cindex options, directory search
11186 @cindex search path
11188 These options specify directories to search for header files, for
11189 libraries and for parts of the compiler:
11194 Add the directory @var{dir} to the head of the list of directories to be
11195 searched for header files. This can be used to override a system header
11196 file, substituting your own version, since these directories are
11197 searched before the system header file directories. However, you should
11198 not use this option to add directories that contain vendor-supplied
11199 system header files (use @option{-isystem} for that). If you use more than
11200 one @option{-I} option, the directories are scanned in left-to-right
11201 order; the standard system directories come after.
11203 If a standard system include directory, or a directory specified with
11204 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11205 option is ignored. The directory is still searched but as a
11206 system directory at its normal position in the system include chain.
11207 This is to ensure that GCC's procedure to fix buggy system headers and
11208 the ordering for the @code{include_next} directive are not inadvertently changed.
11209 If you really need to change the search order for system directories,
11210 use the @option{-nostdinc} and/or @option{-isystem} options.
11212 @item -iplugindir=@var{dir}
11213 @opindex iplugindir=
11214 Set the directory to search for plugins that are passed
11215 by @option{-fplugin=@var{name}} instead of
11216 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11217 to be used by the user, but only passed by the driver.
11219 @item -iquote@var{dir}
11221 Add the directory @var{dir} to the head of the list of directories to
11222 be searched for header files only for the case of @code{#include
11223 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11224 otherwise just like @option{-I}.
11228 Add directory @var{dir} to the list of directories to be searched
11231 @item -B@var{prefix}
11233 This option specifies where to find the executables, libraries,
11234 include files, and data files of the compiler itself.
11236 The compiler driver program runs one or more of the subprograms
11237 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11238 @var{prefix} as a prefix for each program it tries to run, both with and
11239 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11241 For each subprogram to be run, the compiler driver first tries the
11242 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11243 is not specified, the driver tries two standard prefixes,
11244 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11245 those results in a file name that is found, the unmodified program
11246 name is searched for using the directories specified in your
11247 @env{PATH} environment variable.
11249 The compiler checks to see if the path provided by the @option{-B}
11250 refers to a directory, and if necessary it adds a directory
11251 separator character at the end of the path.
11253 @option{-B} prefixes that effectively specify directory names also apply
11254 to libraries in the linker, because the compiler translates these
11255 options into @option{-L} options for the linker. They also apply to
11256 include files in the preprocessor, because the compiler translates these
11257 options into @option{-isystem} options for the preprocessor. In this case,
11258 the compiler appends @samp{include} to the prefix.
11260 The runtime support file @file{libgcc.a} can also be searched for using
11261 the @option{-B} prefix, if needed. If it is not found there, the two
11262 standard prefixes above are tried, and that is all. The file is left
11263 out of the link if it is not found by those means.
11265 Another way to specify a prefix much like the @option{-B} prefix is to use
11266 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11269 As a special kludge, if the path provided by @option{-B} is
11270 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11271 9, then it is replaced by @file{[dir/]include}. This is to help
11272 with boot-strapping the compiler.
11274 @item -specs=@var{file}
11276 Process @var{file} after the compiler reads in the standard @file{specs}
11277 file, in order to override the defaults which the @command{gcc} driver
11278 program uses when determining what switches to pass to @command{cc1},
11279 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11280 @option{-specs=@var{file}} can be specified on the command line, and they
11281 are processed in order, from left to right.
11283 @item --sysroot=@var{dir}
11285 Use @var{dir} as the logical root directory for headers and libraries.
11286 For example, if the compiler normally searches for headers in
11287 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11288 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11290 If you use both this option and the @option{-isysroot} option, then
11291 the @option{--sysroot} option applies to libraries, but the
11292 @option{-isysroot} option applies to header files.
11294 The GNU linker (beginning with version 2.16) has the necessary support
11295 for this option. If your linker does not support this option, the
11296 header file aspect of @option{--sysroot} still works, but the
11297 library aspect does not.
11299 @item --no-sysroot-suffix
11300 @opindex no-sysroot-suffix
11301 For some targets, a suffix is added to the root directory specified
11302 with @option{--sysroot}, depending on the other options used, so that
11303 headers may for example be found in
11304 @file{@var{dir}/@var{suffix}/usr/include} instead of
11305 @file{@var{dir}/usr/include}. This option disables the addition of
11310 This option has been deprecated. Please use @option{-iquote} instead for
11311 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
11312 Any directories you specify with @option{-I} options before the @option{-I-}
11313 option are searched only for the case of @code{#include "@var{file}"};
11314 they are not searched for @code{#include <@var{file}>}.
11316 If additional directories are specified with @option{-I} options after
11317 the @option{-I-}, these directories are searched for all @code{#include}
11318 directives. (Ordinarily @emph{all} @option{-I} directories are used
11321 In addition, the @option{-I-} option inhibits the use of the current
11322 directory (where the current input file came from) as the first search
11323 directory for @code{#include "@var{file}"}. There is no way to
11324 override this effect of @option{-I-}. With @option{-I.} you can specify
11325 searching the directory that is current when the compiler is
11326 invoked. That is not exactly the same as what the preprocessor does
11327 by default, but it is often satisfactory.
11329 @option{-I-} does not inhibit the use of the standard system directories
11330 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11337 @section Specifying subprocesses and the switches to pass to them
11340 @command{gcc} is a driver program. It performs its job by invoking a
11341 sequence of other programs to do the work of compiling, assembling and
11342 linking. GCC interprets its command-line parameters and uses these to
11343 deduce which programs it should invoke, and which command-line options
11344 it ought to place on their command lines. This behavior is controlled
11345 by @dfn{spec strings}. In most cases there is one spec string for each
11346 program that GCC can invoke, but a few programs have multiple spec
11347 strings to control their behavior. The spec strings built into GCC can
11348 be overridden by using the @option{-specs=} command-line switch to specify
11351 @dfn{Spec files} are plaintext files that are used to construct spec
11352 strings. They consist of a sequence of directives separated by blank
11353 lines. The type of directive is determined by the first non-whitespace
11354 character on the line, which can be one of the following:
11357 @item %@var{command}
11358 Issues a @var{command} to the spec file processor. The commands that can
11362 @item %include <@var{file}>
11363 @cindex @code{%include}
11364 Search for @var{file} and insert its text at the current point in the
11367 @item %include_noerr <@var{file}>
11368 @cindex @code{%include_noerr}
11369 Just like @samp{%include}, but do not generate an error message if the include
11370 file cannot be found.
11372 @item %rename @var{old_name} @var{new_name}
11373 @cindex @code{%rename}
11374 Rename the spec string @var{old_name} to @var{new_name}.
11378 @item *[@var{spec_name}]:
11379 This tells the compiler to create, override or delete the named spec
11380 string. All lines after this directive up to the next directive or
11381 blank line are considered to be the text for the spec string. If this
11382 results in an empty string then the spec is deleted. (Or, if the
11383 spec did not exist, then nothing happens.) Otherwise, if the spec
11384 does not currently exist a new spec is created. If the spec does
11385 exist then its contents are overridden by the text of this
11386 directive, unless the first character of that text is the @samp{+}
11387 character, in which case the text is appended to the spec.
11389 @item [@var{suffix}]:
11390 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11391 and up to the next directive or blank line are considered to make up the
11392 spec string for the indicated suffix. When the compiler encounters an
11393 input file with the named suffix, it processes the spec string in
11394 order to work out how to compile that file. For example:
11398 z-compile -input %i
11401 This says that any input file whose name ends in @samp{.ZZ} should be
11402 passed to the program @samp{z-compile}, which should be invoked with the
11403 command-line switch @option{-input} and with the result of performing the
11404 @samp{%i} substitution. (See below.)
11406 As an alternative to providing a spec string, the text following a
11407 suffix directive can be one of the following:
11410 @item @@@var{language}
11411 This says that the suffix is an alias for a known @var{language}. This is
11412 similar to using the @option{-x} command-line switch to GCC to specify a
11413 language explicitly. For example:
11420 Says that .ZZ files are, in fact, C++ source files.
11423 This causes an error messages saying:
11426 @var{name} compiler not installed on this system.
11430 GCC already has an extensive list of suffixes built into it.
11431 This directive adds an entry to the end of the list of suffixes, but
11432 since the list is searched from the end backwards, it is effectively
11433 possible to override earlier entries using this technique.
11437 GCC has the following spec strings built into it. Spec files can
11438 override these strings or create their own. Note that individual
11439 targets can also add their own spec strings to this list.
11442 asm Options to pass to the assembler
11443 asm_final Options to pass to the assembler post-processor
11444 cpp Options to pass to the C preprocessor
11445 cc1 Options to pass to the C compiler
11446 cc1plus Options to pass to the C++ compiler
11447 endfile Object files to include at the end of the link
11448 link Options to pass to the linker
11449 lib Libraries to include on the command line to the linker
11450 libgcc Decides which GCC support library to pass to the linker
11451 linker Sets the name of the linker
11452 predefines Defines to be passed to the C preprocessor
11453 signed_char Defines to pass to CPP to say whether @code{char} is signed
11455 startfile Object files to include at the start of the link
11458 Here is a small example of a spec file:
11461 %rename lib old_lib
11464 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11467 This example renames the spec called @samp{lib} to @samp{old_lib} and
11468 then overrides the previous definition of @samp{lib} with a new one.
11469 The new definition adds in some extra command-line options before
11470 including the text of the old definition.
11472 @dfn{Spec strings} are a list of command-line options to be passed to their
11473 corresponding program. In addition, the spec strings can contain
11474 @samp{%}-prefixed sequences to substitute variable text or to
11475 conditionally insert text into the command line. Using these constructs
11476 it is possible to generate quite complex command lines.
11478 Here is a table of all defined @samp{%}-sequences for spec
11479 strings. Note that spaces are not generated automatically around the
11480 results of expanding these sequences. Therefore you can concatenate them
11481 together or combine them with constant text in a single argument.
11485 Substitute one @samp{%} into the program name or argument.
11488 Substitute the name of the input file being processed.
11491 Substitute the basename of the input file being processed.
11492 This is the substring up to (and not including) the last period
11493 and not including the directory.
11496 This is the same as @samp{%b}, but include the file suffix (text after
11500 Marks the argument containing or following the @samp{%d} as a
11501 temporary file name, so that that file is deleted if GCC exits
11502 successfully. Unlike @samp{%g}, this contributes no text to the
11505 @item %g@var{suffix}
11506 Substitute a file name that has suffix @var{suffix} and is chosen
11507 once per compilation, and mark the argument in the same way as
11508 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11509 name is now chosen in a way that is hard to predict even when previously
11510 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11511 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11512 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11513 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11514 was simply substituted with a file name chosen once per compilation,
11515 without regard to any appended suffix (which was therefore treated
11516 just like ordinary text), making such attacks more likely to succeed.
11518 @item %u@var{suffix}
11519 Like @samp{%g}, but generates a new temporary file name
11520 each time it appears instead of once per compilation.
11522 @item %U@var{suffix}
11523 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11524 new one if there is no such last file name. In the absence of any
11525 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11526 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11527 involves the generation of two distinct file names, one
11528 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11529 simply substituted with a file name chosen for the previous @samp{%u},
11530 without regard to any appended suffix.
11532 @item %j@var{suffix}
11533 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11534 writable, and if @option{-save-temps} is not used;
11535 otherwise, substitute the name
11536 of a temporary file, just like @samp{%u}. This temporary file is not
11537 meant for communication between processes, but rather as a junk
11538 disposal mechanism.
11540 @item %|@var{suffix}
11541 @itemx %m@var{suffix}
11542 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11543 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11544 all. These are the two most common ways to instruct a program that it
11545 should read from standard input or write to standard output. If you
11546 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11547 construct: see for example @file{f/lang-specs.h}.
11549 @item %.@var{SUFFIX}
11550 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11551 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11552 terminated by the next space or %.
11555 Marks the argument containing or following the @samp{%w} as the
11556 designated output file of this compilation. This puts the argument
11557 into the sequence of arguments that @samp{%o} substitutes.
11560 Substitutes the names of all the output files, with spaces
11561 automatically placed around them. You should write spaces
11562 around the @samp{%o} as well or the results are undefined.
11563 @samp{%o} is for use in the specs for running the linker.
11564 Input files whose names have no recognized suffix are not compiled
11565 at all, but they are included among the output files, so they are
11569 Substitutes the suffix for object files. Note that this is
11570 handled specially when it immediately follows @samp{%g, %u, or %U},
11571 because of the need for those to form complete file names. The
11572 handling is such that @samp{%O} is treated exactly as if it had already
11573 been substituted, except that @samp{%g, %u, and %U} do not currently
11574 support additional @var{suffix} characters following @samp{%O} as they do
11575 following, for example, @samp{.o}.
11578 Substitutes the standard macro predefinitions for the
11579 current target machine. Use this when running @command{cpp}.
11582 Like @samp{%p}, but puts @samp{__} before and after the name of each
11583 predefined macro, except for macros that start with @samp{__} or with
11584 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11588 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11589 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11590 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11591 and @option{-imultilib} as necessary.
11594 Current argument is the name of a library or startup file of some sort.
11595 Search for that file in a standard list of directories and substitute
11596 the full name found. The current working directory is included in the
11597 list of directories scanned.
11600 Current argument is the name of a linker script. Search for that file
11601 in the current list of directories to scan for libraries. If the file
11602 is located insert a @option{--script} option into the command line
11603 followed by the full path name found. If the file is not found then
11604 generate an error message. Note: the current working directory is not
11608 Print @var{str} as an error message. @var{str} is terminated by a newline.
11609 Use this when inconsistent options are detected.
11611 @item %(@var{name})
11612 Substitute the contents of spec string @var{name} at this point.
11614 @item %x@{@var{option}@}
11615 Accumulate an option for @samp{%X}.
11618 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11622 Output the accumulated assembler options specified by @option{-Wa}.
11625 Output the accumulated preprocessor options specified by @option{-Wp}.
11628 Process the @code{asm} spec. This is used to compute the
11629 switches to be passed to the assembler.
11632 Process the @code{asm_final} spec. This is a spec string for
11633 passing switches to an assembler post-processor, if such a program is
11637 Process the @code{link} spec. This is the spec for computing the
11638 command line passed to the linker. Typically it makes use of the
11639 @samp{%L %G %S %D and %E} sequences.
11642 Dump out a @option{-L} option for each directory that GCC believes might
11643 contain startup files. If the target supports multilibs then the
11644 current multilib directory is prepended to each of these paths.
11647 Process the @code{lib} spec. This is a spec string for deciding which
11648 libraries are included on the command line to the linker.
11651 Process the @code{libgcc} spec. This is a spec string for deciding
11652 which GCC support library is included on the command line to the linker.
11655 Process the @code{startfile} spec. This is a spec for deciding which
11656 object files are the first ones passed to the linker. Typically
11657 this might be a file named @file{crt0.o}.
11660 Process the @code{endfile} spec. This is a spec string that specifies
11661 the last object files that are passed to the linker.
11664 Process the @code{cpp} spec. This is used to construct the arguments
11665 to be passed to the C preprocessor.
11668 Process the @code{cc1} spec. This is used to construct the options to be
11669 passed to the actual C compiler (@command{cc1}).
11672 Process the @code{cc1plus} spec. This is used to construct the options to be
11673 passed to the actual C++ compiler (@command{cc1plus}).
11676 Substitute the variable part of a matched option. See below.
11677 Note that each comma in the substituted string is replaced by
11681 Remove all occurrences of @code{-S} from the command line. Note---this
11682 command is position dependent. @samp{%} commands in the spec string
11683 before this one see @code{-S}, @samp{%} commands in the spec string
11684 after this one do not.
11686 @item %:@var{function}(@var{args})
11687 Call the named function @var{function}, passing it @var{args}.
11688 @var{args} is first processed as a nested spec string, then split
11689 into an argument vector in the usual fashion. The function returns
11690 a string which is processed as if it had appeared literally as part
11691 of the current spec.
11693 The following built-in spec functions are provided:
11696 @item @code{getenv}
11697 The @code{getenv} spec function takes two arguments: an environment
11698 variable name and a string. If the environment variable is not
11699 defined, a fatal error is issued. Otherwise, the return value is the
11700 value of the environment variable concatenated with the string. For
11701 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11704 %:getenv(TOPDIR /include)
11707 expands to @file{/path/to/top/include}.
11709 @item @code{if-exists}
11710 The @code{if-exists} spec function takes one argument, an absolute
11711 pathname to a file. If the file exists, @code{if-exists} returns the
11712 pathname. Here is a small example of its usage:
11716 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11719 @item @code{if-exists-else}
11720 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11721 spec function, except that it takes two arguments. The first argument is
11722 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11723 returns the pathname. If it does not exist, it returns the second argument.
11724 This way, @code{if-exists-else} can be used to select one file or another,
11725 based on the existence of the first. Here is a small example of its usage:
11729 crt0%O%s %:if-exists(crti%O%s) \
11730 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11733 @item @code{replace-outfile}
11734 The @code{replace-outfile} spec function takes two arguments. It looks for the
11735 first argument in the outfiles array and replaces it with the second argument. Here
11736 is a small example of its usage:
11739 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11742 @item @code{remove-outfile}
11743 The @code{remove-outfile} spec function takes one argument. It looks for the
11744 first argument in the outfiles array and removes it. Here is a small example
11748 %:remove-outfile(-lm)
11751 @item @code{pass-through-libs}
11752 The @code{pass-through-libs} spec function takes any number of arguments. It
11753 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11754 assumes are the names of linker input library archive files) and returns a
11755 result containing all the found arguments each prepended by
11756 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11757 intended to be passed to the LTO linker plugin.
11760 %:pass-through-libs(%G %L %G)
11763 @item @code{print-asm-header}
11764 The @code{print-asm-header} function takes no arguments and simply
11765 prints a banner like:
11771 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11774 It is used to separate compiler options from assembler options
11775 in the @option{--target-help} output.
11778 @item %@{@code{S}@}
11779 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11780 If that switch is not specified, this substitutes nothing. Note that
11781 the leading dash is omitted when specifying this option, and it is
11782 automatically inserted if the substitution is performed. Thus the spec
11783 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11784 and outputs the command-line option @option{-foo}.
11786 @item %W@{@code{S}@}
11787 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11788 deleted on failure.
11790 @item %@{@code{S}*@}
11791 Substitutes all the switches specified to GCC whose names start
11792 with @code{-S}, but which also take an argument. This is used for
11793 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11794 GCC considers @option{-o foo} as being
11795 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11796 text, including the space. Thus two arguments are generated.
11798 @item %@{@code{S}*&@code{T}*@}
11799 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11800 (the order of @code{S} and @code{T} in the spec is not significant).
11801 There can be any number of ampersand-separated variables; for each the
11802 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11804 @item %@{@code{S}:@code{X}@}
11805 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11807 @item %@{!@code{S}:@code{X}@}
11808 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11810 @item %@{@code{S}*:@code{X}@}
11811 Substitutes @code{X} if one or more switches whose names start with
11812 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11813 once, no matter how many such switches appeared. However, if @code{%*}
11814 appears somewhere in @code{X}, then @code{X} is substituted once
11815 for each matching switch, with the @code{%*} replaced by the part of
11816 that switch matching the @code{*}.
11818 If @code{%*} appears as the last part of a spec sequence then a space
11819 is added after the end of the last substitution. If there is more
11820 text in the sequence, however, then a space is not generated. This
11821 allows the @code{%*} substitution to be used as part of a larger
11822 string. For example, a spec string like this:
11825 %@{mcu=*:--script=%*/memory.ld@}
11829 when matching an option like @option{-mcu=newchip} produces:
11832 --script=newchip/memory.ld
11835 @item %@{.@code{S}:@code{X}@}
11836 Substitutes @code{X}, if processing a file with suffix @code{S}.
11838 @item %@{!.@code{S}:@code{X}@}
11839 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
11841 @item %@{,@code{S}:@code{X}@}
11842 Substitutes @code{X}, if processing a file for language @code{S}.
11844 @item %@{!,@code{S}:@code{X}@}
11845 Substitutes @code{X}, if not processing a file for language @code{S}.
11847 @item %@{@code{S}|@code{P}:@code{X}@}
11848 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
11849 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
11850 @code{*} sequences as well, although they have a stronger binding than
11851 the @samp{|}. If @code{%*} appears in @code{X}, all of the
11852 alternatives must be starred, and only the first matching alternative
11855 For example, a spec string like this:
11858 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
11862 outputs the following command-line options from the following input
11863 command-line options:
11868 -d fred.c -foo -baz -boggle
11869 -d jim.d -bar -baz -boggle
11872 @item %@{S:X; T:Y; :D@}
11874 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
11875 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
11876 be as many clauses as you need. This may be combined with @code{.},
11877 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
11882 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
11883 construct may contain other nested @samp{%} constructs or spaces, or
11884 even newlines. They are processed as usual, as described above.
11885 Trailing white space in @code{X} is ignored. White space may also
11886 appear anywhere on the left side of the colon in these constructs,
11887 except between @code{.} or @code{*} and the corresponding word.
11889 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
11890 handled specifically in these constructs. If another value of
11891 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
11892 @option{-W} switch is found later in the command line, the earlier
11893 switch value is ignored, except with @{@code{S}*@} where @code{S} is
11894 just one letter, which passes all matching options.
11896 The character @samp{|} at the beginning of the predicate text is used to
11897 indicate that a command should be piped to the following command, but
11898 only if @option{-pipe} is specified.
11900 It is built into GCC which switches take arguments and which do not.
11901 (You might think it would be useful to generalize this to allow each
11902 compiler's spec to say which switches take arguments. But this cannot
11903 be done in a consistent fashion. GCC cannot even decide which input
11904 files have been specified without knowing which switches take arguments,
11905 and it must know which input files to compile in order to tell which
11908 GCC also knows implicitly that arguments starting in @option{-l} are to be
11909 treated as compiler output files, and passed to the linker in their
11910 proper position among the other output files.
11912 @c man begin OPTIONS
11914 @node Target Options
11915 @section Specifying Target Machine and Compiler Version
11916 @cindex target options
11917 @cindex cross compiling
11918 @cindex specifying machine version
11919 @cindex specifying compiler version and target machine
11920 @cindex compiler version, specifying
11921 @cindex target machine, specifying
11923 The usual way to run GCC is to run the executable called @command{gcc}, or
11924 @command{@var{machine}-gcc} when cross-compiling, or
11925 @command{@var{machine}-gcc-@var{version}} to run a version other than the
11926 one that was installed last.
11928 @node Submodel Options
11929 @section Hardware Models and Configurations
11930 @cindex submodel options
11931 @cindex specifying hardware config
11932 @cindex hardware models and configurations, specifying
11933 @cindex machine dependent options
11935 Each target machine types can have its own
11936 special options, starting with @samp{-m}, to choose among various
11937 hardware models or configurations---for example, 68010 vs 68020,
11938 floating coprocessor or none. A single installed version of the
11939 compiler can compile for any model or configuration, according to the
11942 Some configurations of the compiler also support additional special
11943 options, usually for compatibility with other compilers on the same
11946 @c This list is ordered alphanumerically by subsection name.
11947 @c It should be the same order and spelling as these options are listed
11948 @c in Machine Dependent Options
11951 * AArch64 Options::
11952 * Adapteva Epiphany Options::
11956 * Blackfin Options::
11961 * DEC Alpha Options::
11964 * GNU/Linux Options::
11974 * MicroBlaze Options::
11977 * MN10300 Options::
11981 * Nios II Options::
11983 * picoChip Options::
11984 * PowerPC Options::
11986 * RS/6000 and PowerPC Options::
11988 * S/390 and zSeries Options::
11991 * Solaris 2 Options::
11994 * System V Options::
11995 * TILE-Gx Options::
11996 * TILEPro Options::
12001 * VxWorks Options::
12003 * x86 Windows Options::
12004 * Xstormy16 Options::
12006 * zSeries Options::
12009 @node AArch64 Options
12010 @subsection AArch64 Options
12011 @cindex AArch64 Options
12013 These options are defined for AArch64 implementations:
12017 @item -mabi=@var{name}
12019 Generate code for the specified data model. Permissible values
12020 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12021 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12022 but long int and pointer are 64-bit.
12024 The default depends on the specific target configuration. Note that
12025 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12026 entire program with the same ABI, and link with a compatible set of libraries.
12029 @opindex mbig-endian
12030 Generate big-endian code. This is the default when GCC is configured for an
12031 @samp{aarch64_be-*-*} target.
12033 @item -mgeneral-regs-only
12034 @opindex mgeneral-regs-only
12035 Generate code which uses only the general registers.
12037 @item -mlittle-endian
12038 @opindex mlittle-endian
12039 Generate little-endian code. This is the default when GCC is configured for an
12040 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12042 @item -mcmodel=tiny
12043 @opindex mcmodel=tiny
12044 Generate code for the tiny code model. The program and its statically defined
12045 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12046 be statically or dynamically linked. This model is not fully implemented and
12047 mostly treated as @samp{small}.
12049 @item -mcmodel=small
12050 @opindex mcmodel=small
12051 Generate code for the small code model. The program and its statically defined
12052 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12053 be statically or dynamically linked. This is the default code model.
12055 @item -mcmodel=large
12056 @opindex mcmodel=large
12057 Generate code for the large code model. This makes no assumptions about
12058 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12059 statically linked only.
12061 @item -mstrict-align
12062 @opindex mstrict-align
12063 Do not assume that unaligned memory references are handled by the system.
12065 @item -momit-leaf-frame-pointer
12066 @itemx -mno-omit-leaf-frame-pointer
12067 @opindex momit-leaf-frame-pointer
12068 @opindex mno-omit-leaf-frame-pointer
12069 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12072 @item -mtls-dialect=desc
12073 @opindex mtls-dialect=desc
12074 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12075 of TLS variables. This is the default.
12077 @item -mtls-dialect=traditional
12078 @opindex mtls-dialect=traditional
12079 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12082 @item -mfix-cortex-a53-835769
12083 @itemx -mno-fix-cortex-a53-835769
12084 @opindex mfix-cortex-a53-835769
12085 @opindex mno-fix-cortex-a53-835769
12086 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12087 This involves inserting a NOP instruction between memory instructions and
12088 64-bit integer multiply-accumulate instructions.
12090 @item -march=@var{name}
12092 Specify the name of the target architecture, optionally suffixed by one or
12093 more feature modifiers. This option has the form
12094 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12095 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
12096 values for @var{feature} are documented in the sub-section below.
12098 Where conflicting feature modifiers are specified, the right-most feature is
12101 GCC uses this name to determine what kind of instructions it can emit when
12102 generating assembly code.
12104 Where @option{-march} is specified without either of @option{-mtune}
12105 or @option{-mcpu} also being specified, the code is tuned to perform
12106 well across a range of target processors implementing the target
12109 @item -mtune=@var{name}
12111 Specify the name of the target processor for which GCC should tune the
12112 performance of the code. Permissible values for this option are:
12113 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57},
12114 @samp{cortex-a72}, @samp{thunderx}, @samp{xgene1}.
12116 Additionally, this option can specify that GCC should tune the performance
12117 of the code for a big.LITTLE system. Permissible values for this
12118 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12120 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12121 are specified, the code is tuned to perform well across a range
12122 of target processors.
12124 This option cannot be suffixed by feature modifiers.
12126 @item -mcpu=@var{name}
12128 Specify the name of the target processor, optionally suffixed by one or more
12129 feature modifiers. This option has the form
12130 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12131 permissible values for @var{cpu} are the same as those available for
12134 The permissible values for @var{feature} are documented in the sub-section
12137 Where conflicting feature modifiers are specified, the right-most feature is
12140 GCC uses this name to determine what kind of instructions it can emit when
12141 generating assembly code (as if by @option{-march}) and to determine
12142 the target processor for which to tune for performance (as if
12143 by @option{-mtune}). Where this option is used in conjunction
12144 with @option{-march} or @option{-mtune}, those options take precedence
12145 over the appropriate part of this option.
12148 @subsubsection @option{-march} and @option{-mcpu} feature modifiers
12149 @cindex @option{-march} feature modifiers
12150 @cindex @option{-mcpu} feature modifiers
12151 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
12156 Enable CRC extension.
12158 Enable Crypto extension. This implies Advanced SIMD is enabled.
12160 Enable floating-point instructions.
12162 Enable Advanced SIMD instructions. This implies floating-point instructions
12163 are enabled. This is the default for all current possible values for options
12164 @option{-march} and @option{-mcpu=}.
12167 @node Adapteva Epiphany Options
12168 @subsection Adapteva Epiphany Options
12170 These @samp{-m} options are defined for Adapteva Epiphany:
12173 @item -mhalf-reg-file
12174 @opindex mhalf-reg-file
12175 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12176 That allows code to run on hardware variants that lack these registers.
12178 @item -mprefer-short-insn-regs
12179 @opindex mprefer-short-insn-regs
12180 Preferrentially allocate registers that allow short instruction generation.
12181 This can result in increased instruction count, so this may either reduce or
12182 increase overall code size.
12184 @item -mbranch-cost=@var{num}
12185 @opindex mbranch-cost
12186 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12187 This cost is only a heuristic and is not guaranteed to produce
12188 consistent results across releases.
12192 Enable the generation of conditional moves.
12194 @item -mnops=@var{num}
12196 Emit @var{num} NOPs before every other generated instruction.
12198 @item -mno-soft-cmpsf
12199 @opindex mno-soft-cmpsf
12200 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12201 and test the flags. This is faster than a software comparison, but can
12202 get incorrect results in the presence of NaNs, or when two different small
12203 numbers are compared such that their difference is calculated as zero.
12204 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12205 software comparisons.
12207 @item -mstack-offset=@var{num}
12208 @opindex mstack-offset
12209 Set the offset between the top of the stack and the stack pointer.
12210 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12211 can be used by leaf functions without stack allocation.
12212 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12213 Note also that this option changes the ABI; compiling a program with a
12214 different stack offset than the libraries have been compiled with
12215 generally does not work.
12216 This option can be useful if you want to evaluate if a different stack
12217 offset would give you better code, but to actually use a different stack
12218 offset to build working programs, it is recommended to configure the
12219 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12221 @item -mno-round-nearest
12222 @opindex mno-round-nearest
12223 Make the scheduler assume that the rounding mode has been set to
12224 truncating. The default is @option{-mround-nearest}.
12227 @opindex mlong-calls
12228 If not otherwise specified by an attribute, assume all calls might be beyond
12229 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12230 function address into a register before performing a (otherwise direct) call.
12231 This is the default.
12233 @item -mshort-calls
12234 @opindex short-calls
12235 If not otherwise specified by an attribute, assume all direct calls are
12236 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12237 for direct calls. The default is @option{-mlong-calls}.
12241 Assume addresses can be loaded as 16-bit unsigned values. This does not
12242 apply to function addresses for which @option{-mlong-calls} semantics
12245 @item -mfp-mode=@var{mode}
12247 Set the prevailing mode of the floating-point unit.
12248 This determines the floating-point mode that is provided and expected
12249 at function call and return time. Making this mode match the mode you
12250 predominantly need at function start can make your programs smaller and
12251 faster by avoiding unnecessary mode switches.
12253 @var{mode} can be set to one the following values:
12257 Any mode at function entry is valid, and retained or restored when
12258 the function returns, and when it calls other functions.
12259 This mode is useful for compiling libraries or other compilation units
12260 you might want to incorporate into different programs with different
12261 prevailing FPU modes, and the convenience of being able to use a single
12262 object file outweighs the size and speed overhead for any extra
12263 mode switching that might be needed, compared with what would be needed
12264 with a more specific choice of prevailing FPU mode.
12267 This is the mode used for floating-point calculations with
12268 truncating (i.e.@: round towards zero) rounding mode. That includes
12269 conversion from floating point to integer.
12271 @item round-nearest
12272 This is the mode used for floating-point calculations with
12273 round-to-nearest-or-even rounding mode.
12276 This is the mode used to perform integer calculations in the FPU, e.g.@:
12277 integer multiply, or integer multiply-and-accumulate.
12280 The default is @option{-mfp-mode=caller}
12282 @item -mnosplit-lohi
12283 @itemx -mno-postinc
12284 @itemx -mno-postmodify
12285 @opindex mnosplit-lohi
12286 @opindex mno-postinc
12287 @opindex mno-postmodify
12288 Code generation tweaks that disable, respectively, splitting of 32-bit
12289 loads, generation of post-increment addresses, and generation of
12290 post-modify addresses. The defaults are @option{msplit-lohi},
12291 @option{-mpost-inc}, and @option{-mpost-modify}.
12293 @item -mnovect-double
12294 @opindex mno-vect-double
12295 Change the preferred SIMD mode to SImode. The default is
12296 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12298 @item -max-vect-align=@var{num}
12299 @opindex max-vect-align
12300 The maximum alignment for SIMD vector mode types.
12301 @var{num} may be 4 or 8. The default is 8.
12302 Note that this is an ABI change, even though many library function
12303 interfaces are unaffected if they don't use SIMD vector modes
12304 in places that affect size and/or alignment of relevant types.
12306 @item -msplit-vecmove-early
12307 @opindex msplit-vecmove-early
12308 Split vector moves into single word moves before reload. In theory this
12309 can give better register allocation, but so far the reverse seems to be
12310 generally the case.
12312 @item -m1reg-@var{reg}
12314 Specify a register to hold the constant @minus{}1, which makes loading small negative
12315 constants and certain bitmasks faster.
12316 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12317 which specify use of that register as a fixed register,
12318 and @samp{none}, which means that no register is used for this
12319 purpose. The default is @option{-m1reg-none}.
12324 @subsection ARC Options
12325 @cindex ARC options
12327 The following options control the architecture variant for which code
12330 @c architecture variants
12333 @item -mbarrel-shifter
12334 @opindex mbarrel-shifter
12335 Generate instructions supported by barrel shifter. This is the default
12336 unless @option{-mcpu=ARC601} is in effect.
12338 @item -mcpu=@var{cpu}
12340 Set architecture type, register usage, and instruction scheduling
12341 parameters for @var{cpu}. There are also shortcut alias options
12342 available for backward compatibility and convenience. Supported
12343 values for @var{cpu} are
12349 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12353 Compile for ARC601. Alias: @option{-mARC601}.
12358 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12359 This is the default when configured with @option{--with-cpu=arc700}@.
12364 @itemx -mdpfp-compact
12365 @opindex mdpfp-compact
12366 FPX: Generate Double Precision FPX instructions, tuned for the compact
12370 @opindex mdpfp-fast
12371 FPX: Generate Double Precision FPX instructions, tuned for the fast
12374 @item -mno-dpfp-lrsr
12375 @opindex mno-dpfp-lrsr
12376 Disable LR and SR instructions from using FPX extension aux registers.
12380 Generate Extended arithmetic instructions. Currently only
12381 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12382 supported. This is always enabled for @option{-mcpu=ARC700}.
12386 Do not generate mpy instructions for ARC700.
12390 Generate 32x16 bit multiply and mac instructions.
12394 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12398 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12403 @itemx -mspfp-compact
12404 @opindex mspfp-compact
12405 FPX: Generate Single Precision FPX instructions, tuned for the compact
12409 @opindex mspfp-fast
12410 FPX: Generate Single Precision FPX instructions, tuned for the fast
12415 Enable generation of ARC SIMD instructions via target-specific
12416 builtins. Only valid for @option{-mcpu=ARC700}.
12419 @opindex msoft-float
12420 This option ignored; it is provided for compatibility purposes only.
12421 Software floating point code is emitted by default, and this default
12422 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12423 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12424 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12428 Generate swap instructions.
12432 The following options are passed through to the assembler, and also
12433 define preprocessor macro symbols.
12435 @c Flags used by the assembler, but for which we define preprocessor
12436 @c macro symbols as well.
12439 @opindex mdsp-packa
12440 Passed down to the assembler to enable the DSP Pack A extensions.
12441 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12445 Passed down to the assembler to enable the dual viterbi butterfly
12446 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12448 @c ARC700 4.10 extension instruction
12451 Passed down to the assembler to enable the Locked Load/Store
12452 Conditional extension. Also sets the preprocessor symbol
12457 Passed down to the assembler. Also sets the preprocessor symbol
12458 @code{__Xxmac_d16}.
12462 Passed down to the assembler. Also sets the preprocessor symbol
12465 @c ARC700 4.10 extension instruction
12468 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12469 extension instruction. Also sets the preprocessor symbol
12472 @c ARC700 4.10 extension instruction
12475 Passed down to the assembler to enable the swap byte ordering
12476 extension instruction. Also sets the preprocessor symbol
12480 @opindex mtelephony
12481 Passed down to the assembler to enable dual and single operand
12482 instructions for telephony. Also sets the preprocessor symbol
12483 @code{__Xtelephony}.
12487 Passed down to the assembler to enable the XY Memory extension. Also
12488 sets the preprocessor symbol @code{__Xxy}.
12492 The following options control how the assembly code is annotated:
12494 @c Assembly annotation options
12498 Annotate assembler instructions with estimated addresses.
12500 @item -mannotate-align
12501 @opindex mannotate-align
12502 Explain what alignment considerations lead to the decision to make an
12503 instruction short or long.
12507 The following options are passed through to the linker:
12509 @c options passed through to the linker
12513 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12514 This option is enabled by default in tool chains built for
12515 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12516 when profiling is not requested.
12518 @item -marclinux_prof
12519 @opindex marclinux_prof
12520 Passed through to the linker, to specify use of the
12521 @code{arclinux_prof} emulation. This option is enabled by default in
12522 tool chains built for @w{@code{arc-linux-uclibc}} and
12523 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12527 The following options control the semantics of generated code:
12529 @c semantically relevant code generation options
12531 @item -mepilogue-cfi
12532 @opindex mepilogue-cfi
12533 Enable generation of call frame information for epilogues.
12535 @item -mno-epilogue-cfi
12536 @opindex mno-epilogue-cfi
12537 Disable generation of call frame information for epilogues.
12540 @opindex mlong-calls
12541 Generate call insns as register indirect calls, thus providing access
12542 to the full 32-bit address range.
12544 @item -mmedium-calls
12545 @opindex mmedium-calls
12546 Don't use less than 25 bit addressing range for calls, which is the
12547 offset available for an unconditional branch-and-link
12548 instruction. Conditional execution of function calls is suppressed, to
12549 allow use of the 25-bit range, rather than the 21-bit range with
12550 conditional branch-and-link. This is the default for tool chains built
12551 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12555 Do not generate sdata references. This is the default for tool chains
12556 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12560 @opindex mucb-mcount
12561 Instrument with mcount calls as used in UCB code. I.e. do the
12562 counting in the callee, not the caller. By default ARC instrumentation
12563 counts in the caller.
12565 @item -mvolatile-cache
12566 @opindex mvolatile-cache
12567 Use ordinarily cached memory accesses for volatile references. This is the
12570 @item -mno-volatile-cache
12571 @opindex mno-volatile-cache
12572 Enable cache bypass for volatile references.
12576 The following options fine tune code generation:
12577 @c code generation tuning options
12580 @opindex malign-call
12581 Do alignment optimizations for call instructions.
12583 @item -mauto-modify-reg
12584 @opindex mauto-modify-reg
12585 Enable the use of pre/post modify with register displacement.
12587 @item -mbbit-peephole
12588 @opindex mbbit-peephole
12589 Enable bbit peephole2.
12593 This option disables a target-specific pass in @file{arc_reorg} to
12594 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12595 generation driven by the combiner pass.
12597 @item -mcase-vector-pcrel
12598 @opindex mcase-vector-pcrel
12599 Use pc-relative switch case tables - this enables case table shortening.
12600 This is the default for @option{-Os}.
12602 @item -mcompact-casesi
12603 @opindex mcompact-casesi
12604 Enable compact casesi pattern.
12605 This is the default for @option{-Os}.
12607 @item -mno-cond-exec
12608 @opindex mno-cond-exec
12609 Disable ARCompact specific pass to generate conditional execution instructions.
12610 Due to delay slot scheduling and interactions between operand numbers,
12611 literal sizes, instruction lengths, and the support for conditional execution,
12612 the target-independent pass to generate conditional execution is often lacking,
12613 so the ARC port has kept a special pass around that tries to find more
12614 conditional execution generating opportunities after register allocation,
12615 branch shortening, and delay slot scheduling have been done. This pass
12616 generally, but not always, improves performance and code size, at the cost of
12617 extra compilation time, which is why there is an option to switch it off.
12618 If you have a problem with call instructions exceeding their allowable
12619 offset range because they are conditionalized, you should consider using
12620 @option{-mmedium-calls} instead.
12622 @item -mearly-cbranchsi
12623 @opindex mearly-cbranchsi
12624 Enable pre-reload use of the cbranchsi pattern.
12626 @item -mexpand-adddi
12627 @opindex mexpand-adddi
12628 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12629 @code{add.f}, @code{adc} etc.
12631 @item -mindexed-loads
12632 @opindex mindexed-loads
12633 Enable the use of indexed loads. This can be problematic because some
12634 optimizers then assume that indexed stores exist, which is not
12639 Enable Local Register Allocation. This is still experimental for ARC,
12640 so by default the compiler uses standard reload
12641 (i.e. @option{-mno-lra}).
12643 @item -mlra-priority-none
12644 @opindex mlra-priority-none
12645 Don't indicate any priority for target registers.
12647 @item -mlra-priority-compact
12648 @opindex mlra-priority-compact
12649 Indicate target register priority for r0..r3 / r12..r15.
12651 @item -mlra-priority-noncompact
12652 @opindex mlra-priority-noncompact
12653 Reduce target regsiter priority for r0..r3 / r12..r15.
12655 @item -mno-millicode
12656 @opindex mno-millicode
12657 When optimizing for size (using @option{-Os}), prologues and epilogues
12658 that have to save or restore a large number of registers are often
12659 shortened by using call to a special function in libgcc; this is
12660 referred to as a @emph{millicode} call. As these calls can pose
12661 performance issues, and/or cause linking issues when linking in a
12662 nonstandard way, this option is provided to turn off millicode call
12666 @opindex mmixed-code
12667 Tweak register allocation to help 16-bit instruction generation.
12668 This generally has the effect of decreasing the average instruction size
12669 while increasing the instruction count.
12673 Enable 'q' instruction alternatives.
12674 This is the default for @option{-Os}.
12678 Enable Rcq constraint handling - most short code generation depends on this.
12679 This is the default.
12683 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12684 This is the default.
12686 @item -msize-level=@var{level}
12687 @opindex msize-level
12688 Fine-tune size optimization with regards to instruction lengths and alignment.
12689 The recognized values for @var{level} are:
12692 No size optimization. This level is deprecated and treated like @samp{1}.
12695 Short instructions are used opportunistically.
12698 In addition, alignment of loops and of code after barriers are dropped.
12701 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12705 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12706 the behavior when this is not set is equivalent to level @samp{1}.
12708 @item -mtune=@var{cpu}
12710 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12711 by @option{-mcpu=}.
12713 Supported values for @var{cpu} are
12717 Tune for ARC600 cpu.
12720 Tune for ARC601 cpu.
12723 Tune for ARC700 cpu with standard multiplier block.
12726 Tune for ARC700 cpu with XMAC block.
12729 Tune for ARC725D cpu.
12732 Tune for ARC750D cpu.
12736 @item -mmultcost=@var{num}
12738 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12739 normal instruction.
12741 @item -munalign-prob-threshold=@var{probability}
12742 @opindex munalign-prob-threshold
12743 Set probability threshold for unaligning branches.
12744 When tuning for @samp{ARC700} and optimizing for speed, branches without
12745 filled delay slot are preferably emitted unaligned and long, unless
12746 profiling indicates that the probability for the branch to be taken
12747 is below @var{probability}. @xref{Cross-profiling}.
12748 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12752 The following options are maintained for backward compatibility, but
12753 are now deprecated and will be removed in a future release:
12755 @c Deprecated options
12763 @opindex mbig-endian
12766 Compile code for big endian targets. Use of these options is now
12767 deprecated. Users wanting big-endian code, should use the
12768 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12769 building the tool chain, for which big-endian is the default.
12771 @item -mlittle-endian
12772 @opindex mlittle-endian
12775 Compile code for little endian targets. Use of these options is now
12776 deprecated. Users wanting little-endian code should use the
12777 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12778 building the tool chain, for which little-endian is the default.
12780 @item -mbarrel_shifter
12781 @opindex mbarrel_shifter
12782 Replaced by @option{-mbarrel-shifter}.
12784 @item -mdpfp_compact
12785 @opindex mdpfp_compact
12786 Replaced by @option{-mdpfp-compact}.
12789 @opindex mdpfp_fast
12790 Replaced by @option{-mdpfp-fast}.
12793 @opindex mdsp_packa
12794 Replaced by @option{-mdsp-packa}.
12798 Replaced by @option{-mea}.
12802 Replaced by @option{-mmac-24}.
12806 Replaced by @option{-mmac-d16}.
12808 @item -mspfp_compact
12809 @opindex mspfp_compact
12810 Replaced by @option{-mspfp-compact}.
12813 @opindex mspfp_fast
12814 Replaced by @option{-mspfp-fast}.
12816 @item -mtune=@var{cpu}
12818 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
12819 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
12820 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
12822 @item -multcost=@var{num}
12824 Replaced by @option{-mmultcost}.
12829 @subsection ARM Options
12830 @cindex ARM options
12832 These @samp{-m} options are defined for the ARM port:
12835 @item -mabi=@var{name}
12837 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
12838 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
12841 @opindex mapcs-frame
12842 Generate a stack frame that is compliant with the ARM Procedure Call
12843 Standard for all functions, even if this is not strictly necessary for
12844 correct execution of the code. Specifying @option{-fomit-frame-pointer}
12845 with this option causes the stack frames not to be generated for
12846 leaf functions. The default is @option{-mno-apcs-frame}.
12847 This option is deprecated.
12851 This is a synonym for @option{-mapcs-frame} and is deprecated.
12854 @c not currently implemented
12855 @item -mapcs-stack-check
12856 @opindex mapcs-stack-check
12857 Generate code to check the amount of stack space available upon entry to
12858 every function (that actually uses some stack space). If there is
12859 insufficient space available then either the function
12860 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
12861 called, depending upon the amount of stack space required. The runtime
12862 system is required to provide these functions. The default is
12863 @option{-mno-apcs-stack-check}, since this produces smaller code.
12865 @c not currently implemented
12867 @opindex mapcs-float
12868 Pass floating-point arguments using the floating-point registers. This is
12869 one of the variants of the APCS@. This option is recommended if the
12870 target hardware has a floating-point unit or if a lot of floating-point
12871 arithmetic is going to be performed by the code. The default is
12872 @option{-mno-apcs-float}, since the size of integer-only code is
12873 slightly increased if @option{-mapcs-float} is used.
12875 @c not currently implemented
12876 @item -mapcs-reentrant
12877 @opindex mapcs-reentrant
12878 Generate reentrant, position-independent code. The default is
12879 @option{-mno-apcs-reentrant}.
12882 @item -mthumb-interwork
12883 @opindex mthumb-interwork
12884 Generate code that supports calling between the ARM and Thumb
12885 instruction sets. Without this option, on pre-v5 architectures, the
12886 two instruction sets cannot be reliably used inside one program. The
12887 default is @option{-mno-thumb-interwork}, since slightly larger code
12888 is generated when @option{-mthumb-interwork} is specified. In AAPCS
12889 configurations this option is meaningless.
12891 @item -mno-sched-prolog
12892 @opindex mno-sched-prolog
12893 Prevent the reordering of instructions in the function prologue, or the
12894 merging of those instruction with the instructions in the function's
12895 body. This means that all functions start with a recognizable set
12896 of instructions (or in fact one of a choice from a small set of
12897 different function prologues), and this information can be used to
12898 locate the start of functions inside an executable piece of code. The
12899 default is @option{-msched-prolog}.
12901 @item -mfloat-abi=@var{name}
12902 @opindex mfloat-abi
12903 Specifies which floating-point ABI to use. Permissible values
12904 are: @samp{soft}, @samp{softfp} and @samp{hard}.
12906 Specifying @samp{soft} causes GCC to generate output containing
12907 library calls for floating-point operations.
12908 @samp{softfp} allows the generation of code using hardware floating-point
12909 instructions, but still uses the soft-float calling conventions.
12910 @samp{hard} allows generation of floating-point instructions
12911 and uses FPU-specific calling conventions.
12913 The default depends on the specific target configuration. Note that
12914 the hard-float and soft-float ABIs are not link-compatible; you must
12915 compile your entire program with the same ABI, and link with a
12916 compatible set of libraries.
12918 @item -mlittle-endian
12919 @opindex mlittle-endian
12920 Generate code for a processor running in little-endian mode. This is
12921 the default for all standard configurations.
12924 @opindex mbig-endian
12925 Generate code for a processor running in big-endian mode; the default is
12926 to compile code for a little-endian processor.
12928 @item -march=@var{name}
12930 This specifies the name of the target ARM architecture. GCC uses this
12931 name to determine what kind of instructions it can emit when generating
12932 assembly code. This option can be used in conjunction with or instead
12933 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
12934 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
12935 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
12936 @samp{armv6}, @samp{armv6j},
12937 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
12938 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
12939 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
12940 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
12942 @option{-march=armv7ve} is the armv7-a architecture with virtualization
12945 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
12946 architecture together with the optional CRC32 extensions.
12948 @option{-march=native} causes the compiler to auto-detect the architecture
12949 of the build computer. At present, this feature is only supported on
12950 GNU/Linux, and not all architectures are recognized. If the auto-detect
12951 is unsuccessful the option has no effect.
12953 @item -mtune=@var{name}
12955 This option specifies the name of the target ARM processor for
12956 which GCC should tune the performance of the code.
12957 For some ARM implementations better performance can be obtained by using
12959 Permissible names are: @samp{arm2}, @samp{arm250},
12960 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
12961 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
12962 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
12963 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
12965 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
12966 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
12967 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
12968 @samp{strongarm1110},
12969 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
12970 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
12971 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
12972 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
12973 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
12974 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
12975 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
12976 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
12977 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53},
12978 @samp{cortex-a57}, @samp{cortex-a72},
12980 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
12985 @samp{cortex-m0plus},
12986 @samp{cortex-m1.small-multiply},
12987 @samp{cortex-m0.small-multiply},
12988 @samp{cortex-m0plus.small-multiply},
12989 @samp{marvell-pj4},
12990 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
12991 @samp{fa526}, @samp{fa626},
12992 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
12995 Additionally, this option can specify that GCC should tune the performance
12996 of the code for a big.LITTLE system. Permissible names are:
12997 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53},
12998 @samp{cortex-a72.cortex-a53}.
13000 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13001 performance for a blend of processors within architecture @var{arch}.
13002 The aim is to generate code that run well on the current most popular
13003 processors, balancing between optimizations that benefit some CPUs in the
13004 range, and avoiding performance pitfalls of other CPUs. The effects of
13005 this option may change in future GCC versions as CPU models come and go.
13007 @option{-mtune=native} causes the compiler to auto-detect the CPU
13008 of the build computer. At present, this feature is only supported on
13009 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13010 unsuccessful the option has no effect.
13012 @item -mcpu=@var{name}
13014 This specifies the name of the target ARM processor. GCC uses this name
13015 to derive the name of the target ARM architecture (as if specified
13016 by @option{-march}) and the ARM processor type for which to tune for
13017 performance (as if specified by @option{-mtune}). Where this option
13018 is used in conjunction with @option{-march} or @option{-mtune},
13019 those options take precedence over the appropriate part of this option.
13021 Permissible names for this option are the same as those for
13024 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13025 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13026 See @option{-mtune} for more information.
13028 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13029 of the build computer. At present, this feature is only supported on
13030 GNU/Linux, and not all architectures are recognized. If the auto-detect
13031 is unsuccessful the option has no effect.
13033 @item -mfpu=@var{name}
13035 This specifies what floating-point hardware (or hardware emulation) is
13036 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13037 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13038 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13039 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13040 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13041 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13043 If @option{-msoft-float} is specified this specifies the format of
13044 floating-point values.
13046 If the selected floating-point hardware includes the NEON extension
13047 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13048 operations are not generated by GCC's auto-vectorization pass unless
13049 @option{-funsafe-math-optimizations} is also specified. This is
13050 because NEON hardware does not fully implement the IEEE 754 standard for
13051 floating-point arithmetic (in particular denormal values are treated as
13052 zero), so the use of NEON instructions may lead to a loss of precision.
13054 @item -mfp16-format=@var{name}
13055 @opindex mfp16-format
13056 Specify the format of the @code{__fp16} half-precision floating-point type.
13057 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13058 the default is @samp{none}, in which case the @code{__fp16} type is not
13059 defined. @xref{Half-Precision}, for more information.
13061 @item -mstructure-size-boundary=@var{n}
13062 @opindex mstructure-size-boundary
13063 The sizes of all structures and unions are rounded up to a multiple
13064 of the number of bits set by this option. Permissible values are 8, 32
13065 and 64. The default value varies for different toolchains. For the COFF
13066 targeted toolchain the default value is 8. A value of 64 is only allowed
13067 if the underlying ABI supports it.
13069 Specifying a larger number can produce faster, more efficient code, but
13070 can also increase the size of the program. Different values are potentially
13071 incompatible. Code compiled with one value cannot necessarily expect to
13072 work with code or libraries compiled with another value, if they exchange
13073 information using structures or unions.
13075 @item -mabort-on-noreturn
13076 @opindex mabort-on-noreturn
13077 Generate a call to the function @code{abort} at the end of a
13078 @code{noreturn} function. It is executed if the function tries to
13082 @itemx -mno-long-calls
13083 @opindex mlong-calls
13084 @opindex mno-long-calls
13085 Tells the compiler to perform function calls by first loading the
13086 address of the function into a register and then performing a subroutine
13087 call on this register. This switch is needed if the target function
13088 lies outside of the 64-megabyte addressing range of the offset-based
13089 version of subroutine call instruction.
13091 Even if this switch is enabled, not all function calls are turned
13092 into long calls. The heuristic is that static functions, functions
13093 that have the @code{short_call} attribute, functions that are inside
13094 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13095 definitions have already been compiled within the current compilation
13096 unit are not turned into long calls. The exceptions to this rule are
13097 that weak function definitions, functions with the @code{long_call}
13098 attribute or the @code{section} attribute, and functions that are within
13099 the scope of a @code{#pragma long_calls} directive are always
13100 turned into long calls.
13102 This feature is not enabled by default. Specifying
13103 @option{-mno-long-calls} restores the default behavior, as does
13104 placing the function calls within the scope of a @code{#pragma
13105 long_calls_off} directive. Note these switches have no effect on how
13106 the compiler generates code to handle function calls via function
13109 @item -msingle-pic-base
13110 @opindex msingle-pic-base
13111 Treat the register used for PIC addressing as read-only, rather than
13112 loading it in the prologue for each function. The runtime system is
13113 responsible for initializing this register with an appropriate value
13114 before execution begins.
13116 @item -mpic-register=@var{reg}
13117 @opindex mpic-register
13118 Specify the register to be used for PIC addressing.
13119 For standard PIC base case, the default is any suitable register
13120 determined by compiler. For single PIC base case, the default is
13121 @samp{R9} if target is EABI based or stack-checking is enabled,
13122 otherwise the default is @samp{R10}.
13124 @item -mpic-data-is-text-relative
13125 @opindex mpic-data-is-text-relative
13126 Assume that each data segments are relative to text segment at load time.
13127 Therefore, it permits addressing data using PC-relative operations.
13128 This option is on by default for targets other than VxWorks RTP.
13130 @item -mpoke-function-name
13131 @opindex mpoke-function-name
13132 Write the name of each function into the text section, directly
13133 preceding the function prologue. The generated code is similar to this:
13137 .ascii "arm_poke_function_name", 0
13140 .word 0xff000000 + (t1 - t0)
13141 arm_poke_function_name
13143 stmfd sp!, @{fp, ip, lr, pc@}
13147 When performing a stack backtrace, code can inspect the value of
13148 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13149 location @code{pc - 12} and the top 8 bits are set, then we know that
13150 there is a function name embedded immediately preceding this location
13151 and has length @code{((pc[-3]) & 0xff000000)}.
13158 Select between generating code that executes in ARM and Thumb
13159 states. The default for most configurations is to generate code
13160 that executes in ARM state, but the default can be changed by
13161 configuring GCC with the @option{--with-mode=}@var{state}
13165 @opindex mtpcs-frame
13166 Generate a stack frame that is compliant with the Thumb Procedure Call
13167 Standard for all non-leaf functions. (A leaf function is one that does
13168 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13170 @item -mtpcs-leaf-frame
13171 @opindex mtpcs-leaf-frame
13172 Generate a stack frame that is compliant with the Thumb Procedure Call
13173 Standard for all leaf functions. (A leaf function is one that does
13174 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13176 @item -mcallee-super-interworking
13177 @opindex mcallee-super-interworking
13178 Gives all externally visible functions in the file being compiled an ARM
13179 instruction set header which switches to Thumb mode before executing the
13180 rest of the function. This allows these functions to be called from
13181 non-interworking code. This option is not valid in AAPCS configurations
13182 because interworking is enabled by default.
13184 @item -mcaller-super-interworking
13185 @opindex mcaller-super-interworking
13186 Allows calls via function pointers (including virtual functions) to
13187 execute correctly regardless of whether the target code has been
13188 compiled for interworking or not. There is a small overhead in the cost
13189 of executing a function pointer if this option is enabled. This option
13190 is not valid in AAPCS configurations because interworking is enabled
13193 @item -mtp=@var{name}
13195 Specify the access model for the thread local storage pointer. The valid
13196 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13197 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13198 (supported in the arm6k architecture), and @samp{auto}, which uses the
13199 best available method for the selected processor. The default setting is
13202 @item -mtls-dialect=@var{dialect}
13203 @opindex mtls-dialect
13204 Specify the dialect to use for accessing thread local storage. Two
13205 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13206 @samp{gnu} dialect selects the original GNU scheme for supporting
13207 local and global dynamic TLS models. The @samp{gnu2} dialect
13208 selects the GNU descriptor scheme, which provides better performance
13209 for shared libraries. The GNU descriptor scheme is compatible with
13210 the original scheme, but does require new assembler, linker and
13211 library support. Initial and local exec TLS models are unaffected by
13212 this option and always use the original scheme.
13214 @item -mword-relocations
13215 @opindex mword-relocations
13216 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13217 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13218 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13221 @item -mfix-cortex-m3-ldrd
13222 @opindex mfix-cortex-m3-ldrd
13223 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13224 with overlapping destination and base registers are used. This option avoids
13225 generating these instructions. This option is enabled by default when
13226 @option{-mcpu=cortex-m3} is specified.
13228 @item -munaligned-access
13229 @itemx -mno-unaligned-access
13230 @opindex munaligned-access
13231 @opindex mno-unaligned-access
13232 Enables (or disables) reading and writing of 16- and 32- bit values
13233 from addresses that are not 16- or 32- bit aligned. By default
13234 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13235 architectures, and enabled for all other architectures. If unaligned
13236 access is not enabled then words in packed data structures are
13237 accessed a byte at a time.
13239 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13240 generated object file to either true or false, depending upon the
13241 setting of this option. If unaligned access is enabled then the
13242 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13245 @item -mneon-for-64bits
13246 @opindex mneon-for-64bits
13247 Enables using Neon to handle scalar 64-bits operations. This is
13248 disabled by default since the cost of moving data from core registers
13251 @item -mslow-flash-data
13252 @opindex mslow-flash-data
13253 Assume loading data from flash is slower than fetching instruction.
13254 Therefore literal load is minimized for better performance.
13255 This option is only supported when compiling for ARMv7 M-profile and
13258 @item -masm-syntax-unified
13259 @opindex masm-syntax-unified
13260 Assume inline assembler is using unified asm syntax. The default is
13261 currently off which implies divided syntax. Currently this option is
13262 available only for Thumb1 and has no effect on ARM state and Thumb2.
13263 However, this may change in future releases of GCC. Divided syntax
13264 should be considered deprecated.
13266 @item -mrestrict-it
13267 @opindex mrestrict-it
13268 Restricts generation of IT blocks to conform to the rules of ARMv8.
13269 IT blocks can only contain a single 16-bit instruction from a select
13270 set of instructions. This option is on by default for ARMv8 Thumb mode.
13274 @subsection AVR Options
13275 @cindex AVR Options
13277 These options are defined for AVR implementations:
13280 @item -mmcu=@var{mcu}
13282 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13284 The default for this option is@tie{}@samp{avr2}.
13286 GCC supports the following AVR devices and ISAs:
13288 @include avr-mmcu.texi
13290 @item -maccumulate-args
13291 @opindex maccumulate-args
13292 Accumulate outgoing function arguments and acquire/release the needed
13293 stack space for outgoing function arguments once in function
13294 prologue/epilogue. Without this option, outgoing arguments are pushed
13295 before calling a function and popped afterwards.
13297 Popping the arguments after the function call can be expensive on
13298 AVR so that accumulating the stack space might lead to smaller
13299 executables because arguments need not to be removed from the
13300 stack after such a function call.
13302 This option can lead to reduced code size for functions that perform
13303 several calls to functions that get their arguments on the stack like
13304 calls to printf-like functions.
13306 @item -mbranch-cost=@var{cost}
13307 @opindex mbranch-cost
13308 Set the branch costs for conditional branch instructions to
13309 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13310 integers. The default branch cost is 0.
13312 @item -mcall-prologues
13313 @opindex mcall-prologues
13314 Functions prologues/epilogues are expanded as calls to appropriate
13315 subroutines. Code size is smaller.
13319 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13320 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13321 and @code{long long} is 4 bytes. Please note that this option does not
13322 conform to the C standards, but it results in smaller code
13325 @item -mno-interrupts
13326 @opindex mno-interrupts
13327 Generated code is not compatible with hardware interrupts.
13328 Code size is smaller.
13332 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13333 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13334 Setting @option{-mrelax} just adds the @option{--relax} option to the
13335 linker command line when the linker is called.
13337 Jump relaxing is performed by the linker because jump offsets are not
13338 known before code is located. Therefore, the assembler code generated by the
13339 compiler is the same, but the instructions in the executable may
13340 differ from instructions in the assembler code.
13342 Relaxing must be turned on if linker stubs are needed, see the
13343 section on @code{EIND} and linker stubs below.
13347 Treat the stack pointer register as an 8-bit register,
13348 i.e.@: assume the high byte of the stack pointer is zero.
13349 In general, you don't need to set this option by hand.
13351 This option is used internally by the compiler to select and
13352 build multilibs for architectures @code{avr2} and @code{avr25}.
13353 These architectures mix devices with and without @code{SPH}.
13354 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13355 the compiler driver adds or removes this option from the compiler
13356 proper's command line, because the compiler then knows if the device
13357 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13362 Use address register @code{X} in a way proposed by the hardware. This means
13363 that @code{X} is only used in indirect, post-increment or
13364 pre-decrement addressing.
13366 Without this option, the @code{X} register may be used in the same way
13367 as @code{Y} or @code{Z} which then is emulated by additional
13369 For example, loading a value with @code{X+const} addressing with a
13370 small non-negative @code{const < 64} to a register @var{Rn} is
13374 adiw r26, const ; X += const
13375 ld @var{Rn}, X ; @var{Rn} = *X
13376 sbiw r26, const ; X -= const
13380 @opindex mtiny-stack
13381 Only change the lower 8@tie{}bits of the stack pointer.
13383 @item -Waddr-space-convert
13384 @opindex Waddr-space-convert
13385 Warn about conversions between address spaces in the case where the
13386 resulting address space is not contained in the incoming address space.
13389 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
13390 @cindex @code{EIND}
13391 Pointers in the implementation are 16@tie{}bits wide.
13392 The address of a function or label is represented as word address so
13393 that indirect jumps and calls can target any code address in the
13394 range of 64@tie{}Ki words.
13396 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13397 bytes of program memory space, there is a special function register called
13398 @code{EIND} that serves as most significant part of the target address
13399 when @code{EICALL} or @code{EIJMP} instructions are used.
13401 Indirect jumps and calls on these devices are handled as follows by
13402 the compiler and are subject to some limitations:
13407 The compiler never sets @code{EIND}.
13410 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13411 instructions or might read @code{EIND} directly in order to emulate an
13412 indirect call/jump by means of a @code{RET} instruction.
13415 The compiler assumes that @code{EIND} never changes during the startup
13416 code or during the application. In particular, @code{EIND} is not
13417 saved/restored in function or interrupt service routine
13421 For indirect calls to functions and computed goto, the linker
13422 generates @emph{stubs}. Stubs are jump pads sometimes also called
13423 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13424 The stub contains a direct jump to the desired address.
13427 Linker relaxation must be turned on so that the linker generates
13428 the stubs correctly in all situations. See the compiler option
13429 @option{-mrelax} and the linker option @option{--relax}.
13430 There are corner cases where the linker is supposed to generate stubs
13431 but aborts without relaxation and without a helpful error message.
13434 The default linker script is arranged for code with @code{EIND = 0}.
13435 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13436 linker script has to be used in order to place the sections whose
13437 name start with @code{.trampolines} into the segment where @code{EIND}
13441 The startup code from libgcc never sets @code{EIND}.
13442 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13443 For the impact of AVR-LibC on @code{EIND}, see the
13444 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13447 It is legitimate for user-specific startup code to set up @code{EIND}
13448 early, for example by means of initialization code located in
13449 section @code{.init3}. Such code runs prior to general startup code
13450 that initializes RAM and calls constructors, but after the bit
13451 of startup code from AVR-LibC that sets @code{EIND} to the segment
13452 where the vector table is located.
13454 #include <avr/io.h>
13457 __attribute__((section(".init3"),naked,used,no_instrument_function))
13458 init3_set_eind (void)
13460 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13461 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13466 The @code{__trampolines_start} symbol is defined in the linker script.
13469 Stubs are generated automatically by the linker if
13470 the following two conditions are met:
13473 @item The address of a label is taken by means of the @code{gs} modifier
13474 (short for @emph{generate stubs}) like so:
13476 LDI r24, lo8(gs(@var{func}))
13477 LDI r25, hi8(gs(@var{func}))
13479 @item The final location of that label is in a code segment
13480 @emph{outside} the segment where the stubs are located.
13484 The compiler emits such @code{gs} modifiers for code labels in the
13485 following situations:
13487 @item Taking address of a function or code label.
13488 @item Computed goto.
13489 @item If prologue-save function is used, see @option{-mcall-prologues}
13490 command-line option.
13491 @item Switch/case dispatch tables. If you do not want such dispatch
13492 tables you can specify the @option{-fno-jump-tables} command-line option.
13493 @item C and C++ constructors/destructors called during startup/shutdown.
13494 @item If the tools hit a @code{gs()} modifier explained above.
13498 Jumping to non-symbolic addresses like so is @emph{not} supported:
13503 /* Call function at word address 0x2 */
13504 return ((int(*)(void)) 0x2)();
13508 Instead, a stub has to be set up, i.e.@: the function has to be called
13509 through a symbol (@code{func_4} in the example):
13514 extern int func_4 (void);
13516 /* Call function at byte address 0x4 */
13521 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13522 Alternatively, @code{func_4} can be defined in the linker script.
13525 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13526 @cindex @code{RAMPD}
13527 @cindex @code{RAMPX}
13528 @cindex @code{RAMPY}
13529 @cindex @code{RAMPZ}
13530 Some AVR devices support memories larger than the 64@tie{}KiB range
13531 that can be accessed with 16-bit pointers. To access memory locations
13532 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13533 register is used as high part of the address:
13534 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13535 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13536 register, respectively, to get a wide address. Similarly,
13537 @code{RAMPD} is used together with direct addressing.
13541 The startup code initializes the @code{RAMP} special function
13542 registers with zero.
13545 If a @ref{AVR Named Address Spaces,named address space} other than
13546 generic or @code{__flash} is used, then @code{RAMPZ} is set
13547 as needed before the operation.
13550 If the device supports RAM larger than 64@tie{}KiB and the compiler
13551 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13552 is reset to zero after the operation.
13555 If the device comes with a specific @code{RAMP} register, the ISR
13556 prologue/epilogue saves/restores that SFR and initializes it with
13557 zero in case the ISR code might (implicitly) use it.
13560 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13561 If you use inline assembler to read from locations outside the
13562 16-bit address range and change one of the @code{RAMP} registers,
13563 you must reset it to zero after the access.
13567 @subsubsection AVR Built-in Macros
13569 GCC defines several built-in macros so that the user code can test
13570 for the presence or absence of features. Almost any of the following
13571 built-in macros are deduced from device capabilities and thus
13572 triggered by the @option{-mmcu=} command-line option.
13574 For even more AVR-specific built-in macros see
13575 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13580 Build-in macro that resolves to a decimal number that identifies the
13581 architecture and depends on the @option{-mmcu=@var{mcu}} option.
13582 Possible values are:
13584 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13585 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
13586 @code{105}, @code{106}, @code{107}
13588 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
13589 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
13590 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
13591 @code{avrxmega6}, @code{avrxmega7}, respectively.
13592 If @var{mcu} specifies a device, this built-in macro is set
13593 accordingly. For example, with @option{-mmcu=atmega8} the macro is
13594 defined to @code{4}.
13596 @item __AVR_@var{Device}__
13597 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
13598 the device's name. For example, @option{-mmcu=atmega8} defines the
13599 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
13600 @code{__AVR_ATtiny261A__}, etc.
13602 The built-in macros' names follow
13603 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13604 the device name as from the AVR user manual. The difference between
13605 @var{Device} in the built-in macro and @var{device} in
13606 @option{-mmcu=@var{device}} is that the latter is always lowercase.
13608 If @var{device} is not a device but only a core architecture like
13609 @samp{avr51}, this macro is not defined.
13611 @item __AVR_DEVICE_NAME__
13612 Setting @option{-mmcu=@var{device}} defines this built-in macro to
13613 the device's name. For example, with @option{-mmcu=atmega8} the macro
13614 is defined to @code{atmega8}.
13616 If @var{device} is not a device but only a core architecture like
13617 @samp{avr51}, this macro is not defined.
13619 @item __AVR_XMEGA__
13620 The device / architecture belongs to the XMEGA family of devices.
13622 @item __AVR_HAVE_ELPM__
13623 The device has the the @code{ELPM} instruction.
13625 @item __AVR_HAVE_ELPMX__
13626 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13627 R@var{n},Z+} instructions.
13629 @item __AVR_HAVE_MOVW__
13630 The device has the @code{MOVW} instruction to perform 16-bit
13631 register-register moves.
13633 @item __AVR_HAVE_LPMX__
13634 The device has the @code{LPM R@var{n},Z} and
13635 @code{LPM R@var{n},Z+} instructions.
13637 @item __AVR_HAVE_MUL__
13638 The device has a hardware multiplier.
13640 @item __AVR_HAVE_JMP_CALL__
13641 The device has the @code{JMP} and @code{CALL} instructions.
13642 This is the case for devices with at least 16@tie{}KiB of program
13645 @item __AVR_HAVE_EIJMP_EICALL__
13646 @itemx __AVR_3_BYTE_PC__
13647 The device has the @code{EIJMP} and @code{EICALL} instructions.
13648 This is the case for devices with more than 128@tie{}KiB of program memory.
13649 This also means that the program counter
13650 (PC) is 3@tie{}bytes wide.
13652 @item __AVR_2_BYTE_PC__
13653 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13654 with up to 128@tie{}KiB of program memory.
13656 @item __AVR_HAVE_8BIT_SP__
13657 @itemx __AVR_HAVE_16BIT_SP__
13658 The stack pointer (SP) register is treated as 8-bit respectively
13659 16-bit register by the compiler.
13660 The definition of these macros is affected by @option{-mtiny-stack}.
13662 @item __AVR_HAVE_SPH__
13664 The device has the SPH (high part of stack pointer) special function
13665 register or has an 8-bit stack pointer, respectively.
13666 The definition of these macros is affected by @option{-mmcu=} and
13667 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
13670 @item __AVR_HAVE_RAMPD__
13671 @itemx __AVR_HAVE_RAMPX__
13672 @itemx __AVR_HAVE_RAMPY__
13673 @itemx __AVR_HAVE_RAMPZ__
13674 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13675 @code{RAMPZ} special function register, respectively.
13677 @item __NO_INTERRUPTS__
13678 This macro reflects the @option{-mno-interrupts} command line option.
13680 @item __AVR_ERRATA_SKIP__
13681 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13682 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13683 instructions because of a hardware erratum. Skip instructions are
13684 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13685 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13688 @item __AVR_ISA_RMW__
13689 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13691 @item __AVR_SFR_OFFSET__=@var{offset}
13692 Instructions that can address I/O special function registers directly
13693 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13694 address as if addressed by an instruction to access RAM like @code{LD}
13695 or @code{STS}. This offset depends on the device architecture and has
13696 to be subtracted from the RAM address in order to get the
13697 respective I/O@tie{}address.
13699 @item __WITH_AVRLIBC__
13700 The compiler is configured to be used together with AVR-Libc.
13701 See the @option{--with-avrlibc} configure option.
13705 @node Blackfin Options
13706 @subsection Blackfin Options
13707 @cindex Blackfin Options
13710 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13712 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13713 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13714 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13715 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13716 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13717 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13718 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13719 @samp{bf561}, @samp{bf592}.
13721 The optional @var{sirevision} specifies the silicon revision of the target
13722 Blackfin processor. Any workarounds available for the targeted silicon revision
13723 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13724 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13725 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13726 hexadecimal digits representing the major and minor numbers in the silicon
13727 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13728 is not defined. If @var{sirevision} is @samp{any}, the
13729 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13730 If this optional @var{sirevision} is not used, GCC assumes the latest known
13731 silicon revision of the targeted Blackfin processor.
13733 GCC defines a preprocessor macro for the specified @var{cpu}.
13734 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13735 provided by libgloss to be linked in if @option{-msim} is not given.
13737 Without this option, @samp{bf532} is used as the processor by default.
13739 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13740 only the preprocessor macro is defined.
13744 Specifies that the program will be run on the simulator. This causes
13745 the simulator BSP provided by libgloss to be linked in. This option
13746 has effect only for @samp{bfin-elf} toolchain.
13747 Certain other options, such as @option{-mid-shared-library} and
13748 @option{-mfdpic}, imply @option{-msim}.
13750 @item -momit-leaf-frame-pointer
13751 @opindex momit-leaf-frame-pointer
13752 Don't keep the frame pointer in a register for leaf functions. This
13753 avoids the instructions to save, set up and restore frame pointers and
13754 makes an extra register available in leaf functions. The option
13755 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13756 which might make debugging harder.
13758 @item -mspecld-anomaly
13759 @opindex mspecld-anomaly
13760 When enabled, the compiler ensures that the generated code does not
13761 contain speculative loads after jump instructions. If this option is used,
13762 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13764 @item -mno-specld-anomaly
13765 @opindex mno-specld-anomaly
13766 Don't generate extra code to prevent speculative loads from occurring.
13768 @item -mcsync-anomaly
13769 @opindex mcsync-anomaly
13770 When enabled, the compiler ensures that the generated code does not
13771 contain CSYNC or SSYNC instructions too soon after conditional branches.
13772 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13774 @item -mno-csync-anomaly
13775 @opindex mno-csync-anomaly
13776 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13777 occurring too soon after a conditional branch.
13781 When enabled, the compiler is free to take advantage of the knowledge that
13782 the entire program fits into the low 64k of memory.
13785 @opindex mno-low-64k
13786 Assume that the program is arbitrarily large. This is the default.
13788 @item -mstack-check-l1
13789 @opindex mstack-check-l1
13790 Do stack checking using information placed into L1 scratchpad memory by the
13793 @item -mid-shared-library
13794 @opindex mid-shared-library
13795 Generate code that supports shared libraries via the library ID method.
13796 This allows for execute in place and shared libraries in an environment
13797 without virtual memory management. This option implies @option{-fPIC}.
13798 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13800 @item -mno-id-shared-library
13801 @opindex mno-id-shared-library
13802 Generate code that doesn't assume ID-based shared libraries are being used.
13803 This is the default.
13805 @item -mleaf-id-shared-library
13806 @opindex mleaf-id-shared-library
13807 Generate code that supports shared libraries via the library ID method,
13808 but assumes that this library or executable won't link against any other
13809 ID shared libraries. That allows the compiler to use faster code for jumps
13812 @item -mno-leaf-id-shared-library
13813 @opindex mno-leaf-id-shared-library
13814 Do not assume that the code being compiled won't link against any ID shared
13815 libraries. Slower code is generated for jump and call insns.
13817 @item -mshared-library-id=n
13818 @opindex mshared-library-id
13819 Specifies the identification number of the ID-based shared library being
13820 compiled. Specifying a value of 0 generates more compact code; specifying
13821 other values forces the allocation of that number to the current
13822 library but is no more space- or time-efficient than omitting this option.
13826 Generate code that allows the data segment to be located in a different
13827 area of memory from the text segment. This allows for execute in place in
13828 an environment without virtual memory management by eliminating relocations
13829 against the text section.
13831 @item -mno-sep-data
13832 @opindex mno-sep-data
13833 Generate code that assumes that the data segment follows the text segment.
13834 This is the default.
13837 @itemx -mno-long-calls
13838 @opindex mlong-calls
13839 @opindex mno-long-calls
13840 Tells the compiler to perform function calls by first loading the
13841 address of the function into a register and then performing a subroutine
13842 call on this register. This switch is needed if the target function
13843 lies outside of the 24-bit addressing range of the offset-based
13844 version of subroutine call instruction.
13846 This feature is not enabled by default. Specifying
13847 @option{-mno-long-calls} restores the default behavior. Note these
13848 switches have no effect on how the compiler generates code to handle
13849 function calls via function pointers.
13853 Link with the fast floating-point library. This library relaxes some of
13854 the IEEE floating-point standard's rules for checking inputs against
13855 Not-a-Number (NAN), in the interest of performance.
13858 @opindex minline-plt
13859 Enable inlining of PLT entries in function calls to functions that are
13860 not known to bind locally. It has no effect without @option{-mfdpic}.
13863 @opindex mmulticore
13864 Build a standalone application for multicore Blackfin processors.
13865 This option causes proper start files and link scripts supporting
13866 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
13867 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
13869 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
13870 selects the one-application-per-core programming model. Without
13871 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
13872 programming model is used. In this model, the main function of Core B
13873 should be named as @code{coreb_main}.
13875 If this option is not used, the single-core application programming
13880 Build a standalone application for Core A of BF561 when using
13881 the one-application-per-core programming model. Proper start files
13882 and link scripts are used to support Core A, and the macro
13883 @code{__BFIN_COREA} is defined.
13884 This option can only be used in conjunction with @option{-mmulticore}.
13888 Build a standalone application for Core B of BF561 when using
13889 the one-application-per-core programming model. Proper start files
13890 and link scripts are used to support Core B, and the macro
13891 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
13892 should be used instead of @code{main}.
13893 This option can only be used in conjunction with @option{-mmulticore}.
13897 Build a standalone application for SDRAM. Proper start files and
13898 link scripts are used to put the application into SDRAM, and the macro
13899 @code{__BFIN_SDRAM} is defined.
13900 The loader should initialize SDRAM before loading the application.
13904 Assume that ICPLBs are enabled at run time. This has an effect on certain
13905 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
13906 are enabled; for standalone applications the default is off.
13910 @subsection C6X Options
13911 @cindex C6X Options
13914 @item -march=@var{name}
13916 This specifies the name of the target architecture. GCC uses this
13917 name to determine what kind of instructions it can emit when generating
13918 assembly code. Permissible names are: @samp{c62x},
13919 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
13922 @opindex mbig-endian
13923 Generate code for a big-endian target.
13925 @item -mlittle-endian
13926 @opindex mlittle-endian
13927 Generate code for a little-endian target. This is the default.
13931 Choose startup files and linker script suitable for the simulator.
13933 @item -msdata=default
13934 @opindex msdata=default
13935 Put small global and static data in the @code{.neardata} section,
13936 which is pointed to by register @code{B14}. Put small uninitialized
13937 global and static data in the @code{.bss} section, which is adjacent
13938 to the @code{.neardata} section. Put small read-only data into the
13939 @code{.rodata} section. The corresponding sections used for large
13940 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
13943 @opindex msdata=all
13944 Put all data, not just small objects, into the sections reserved for
13945 small data, and use addressing relative to the @code{B14} register to
13949 @opindex msdata=none
13950 Make no use of the sections reserved for small data, and use absolute
13951 addresses to access all data. Put all initialized global and static
13952 data in the @code{.fardata} section, and all uninitialized data in the
13953 @code{.far} section. Put all constant data into the @code{.const}
13958 @subsection CRIS Options
13959 @cindex CRIS Options
13961 These options are defined specifically for the CRIS ports.
13964 @item -march=@var{architecture-type}
13965 @itemx -mcpu=@var{architecture-type}
13968 Generate code for the specified architecture. The choices for
13969 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
13970 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
13971 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
13974 @item -mtune=@var{architecture-type}
13976 Tune to @var{architecture-type} everything applicable about the generated
13977 code, except for the ABI and the set of available instructions. The
13978 choices for @var{architecture-type} are the same as for
13979 @option{-march=@var{architecture-type}}.
13981 @item -mmax-stack-frame=@var{n}
13982 @opindex mmax-stack-frame
13983 Warn when the stack frame of a function exceeds @var{n} bytes.
13989 The options @option{-metrax4} and @option{-metrax100} are synonyms for
13990 @option{-march=v3} and @option{-march=v8} respectively.
13992 @item -mmul-bug-workaround
13993 @itemx -mno-mul-bug-workaround
13994 @opindex mmul-bug-workaround
13995 @opindex mno-mul-bug-workaround
13996 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
13997 models where it applies. This option is active by default.
14001 Enable CRIS-specific verbose debug-related information in the assembly
14002 code. This option also has the effect of turning off the @samp{#NO_APP}
14003 formatted-code indicator to the assembler at the beginning of the
14008 Do not use condition-code results from previous instruction; always emit
14009 compare and test instructions before use of condition codes.
14011 @item -mno-side-effects
14012 @opindex mno-side-effects
14013 Do not emit instructions with side effects in addressing modes other than
14016 @item -mstack-align
14017 @itemx -mno-stack-align
14018 @itemx -mdata-align
14019 @itemx -mno-data-align
14020 @itemx -mconst-align
14021 @itemx -mno-const-align
14022 @opindex mstack-align
14023 @opindex mno-stack-align
14024 @opindex mdata-align
14025 @opindex mno-data-align
14026 @opindex mconst-align
14027 @opindex mno-const-align
14028 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14029 stack frame, individual data and constants to be aligned for the maximum
14030 single data access size for the chosen CPU model. The default is to
14031 arrange for 32-bit alignment. ABI details such as structure layout are
14032 not affected by these options.
14040 Similar to the stack- data- and const-align options above, these options
14041 arrange for stack frame, writable data and constants to all be 32-bit,
14042 16-bit or 8-bit aligned. The default is 32-bit alignment.
14044 @item -mno-prologue-epilogue
14045 @itemx -mprologue-epilogue
14046 @opindex mno-prologue-epilogue
14047 @opindex mprologue-epilogue
14048 With @option{-mno-prologue-epilogue}, the normal function prologue and
14049 epilogue which set up the stack frame are omitted and no return
14050 instructions or return sequences are generated in the code. Use this
14051 option only together with visual inspection of the compiled code: no
14052 warnings or errors are generated when call-saved registers must be saved,
14053 or storage for local variables needs to be allocated.
14057 @opindex mno-gotplt
14059 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14060 instruction sequences that load addresses for functions from the PLT part
14061 of the GOT rather than (traditional on other architectures) calls to the
14062 PLT@. The default is @option{-mgotplt}.
14066 Legacy no-op option only recognized with the cris-axis-elf and
14067 cris-axis-linux-gnu targets.
14071 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14075 This option, recognized for the cris-axis-elf, arranges
14076 to link with input-output functions from a simulator library. Code,
14077 initialized data and zero-initialized data are allocated consecutively.
14081 Like @option{-sim}, but pass linker options to locate initialized data at
14082 0x40000000 and zero-initialized data at 0x80000000.
14086 @subsection CR16 Options
14087 @cindex CR16 Options
14089 These options are defined specifically for the CR16 ports.
14095 Enable the use of multiply-accumulate instructions. Disabled by default.
14099 @opindex mcr16cplus
14101 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14106 Links the library libsim.a which is in compatible with simulator. Applicable
14107 to ELF compiler only.
14111 Choose integer type as 32-bit wide.
14115 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14117 @item -mdata-model=@var{model}
14118 @opindex mdata-model
14119 Choose a data model. The choices for @var{model} are @samp{near},
14120 @samp{far} or @samp{medium}. @samp{medium} is default.
14121 However, @samp{far} is not valid with @option{-mcr16c}, as the
14122 CR16C architecture does not support the far data model.
14125 @node Darwin Options
14126 @subsection Darwin Options
14127 @cindex Darwin options
14129 These options are defined for all architectures running the Darwin operating
14132 FSF GCC on Darwin does not create ``fat'' object files; it creates
14133 an object file for the single architecture that GCC was built to
14134 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14135 @option{-arch} options are used; it does so by running the compiler or
14136 linker multiple times and joining the results together with
14139 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14140 @samp{i686}) is determined by the flags that specify the ISA
14141 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14142 @option{-force_cpusubtype_ALL} option can be used to override this.
14144 The Darwin tools vary in their behavior when presented with an ISA
14145 mismatch. The assembler, @file{as}, only permits instructions to
14146 be used that are valid for the subtype of the file it is generating,
14147 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14148 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14149 and prints an error if asked to create a shared library with a less
14150 restrictive subtype than its input files (for instance, trying to put
14151 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14152 for executables, @command{ld}, quietly gives the executable the most
14153 restrictive subtype of any of its input files.
14158 Add the framework directory @var{dir} to the head of the list of
14159 directories to be searched for header files. These directories are
14160 interleaved with those specified by @option{-I} options and are
14161 scanned in a left-to-right order.
14163 A framework directory is a directory with frameworks in it. A
14164 framework is a directory with a @file{Headers} and/or
14165 @file{PrivateHeaders} directory contained directly in it that ends
14166 in @file{.framework}. The name of a framework is the name of this
14167 directory excluding the @file{.framework}. Headers associated with
14168 the framework are found in one of those two directories, with
14169 @file{Headers} being searched first. A subframework is a framework
14170 directory that is in a framework's @file{Frameworks} directory.
14171 Includes of subframework headers can only appear in a header of a
14172 framework that contains the subframework, or in a sibling subframework
14173 header. Two subframeworks are siblings if they occur in the same
14174 framework. A subframework should not have the same name as a
14175 framework; a warning is issued if this is violated. Currently a
14176 subframework cannot have subframeworks; in the future, the mechanism
14177 may be extended to support this. The standard frameworks can be found
14178 in @file{/System/Library/Frameworks} and
14179 @file{/Library/Frameworks}. An example include looks like
14180 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14181 the name of the framework and @file{header.h} is found in the
14182 @file{PrivateHeaders} or @file{Headers} directory.
14184 @item -iframework@var{dir}
14185 @opindex iframework
14186 Like @option{-F} except the directory is a treated as a system
14187 directory. The main difference between this @option{-iframework} and
14188 @option{-F} is that with @option{-iframework} the compiler does not
14189 warn about constructs contained within header files found via
14190 @var{dir}. This option is valid only for the C family of languages.
14194 Emit debugging information for symbols that are used. For stabs
14195 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14196 This is by default ON@.
14200 Emit debugging information for all symbols and types.
14202 @item -mmacosx-version-min=@var{version}
14203 The earliest version of MacOS X that this executable will run on
14204 is @var{version}. Typical values of @var{version} include @code{10.1},
14205 @code{10.2}, and @code{10.3.9}.
14207 If the compiler was built to use the system's headers by default,
14208 then the default for this option is the system version on which the
14209 compiler is running, otherwise the default is to make choices that
14210 are compatible with as many systems and code bases as possible.
14214 Enable kernel development mode. The @option{-mkernel} option sets
14215 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14216 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14217 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14218 applicable. This mode also sets @option{-mno-altivec},
14219 @option{-msoft-float}, @option{-fno-builtin} and
14220 @option{-mlong-branch} for PowerPC targets.
14222 @item -mone-byte-bool
14223 @opindex mone-byte-bool
14224 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14225 By default @code{sizeof(bool)} is @code{4} when compiling for
14226 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14227 option has no effect on x86.
14229 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14230 to generate code that is not binary compatible with code generated
14231 without that switch. Using this switch may require recompiling all
14232 other modules in a program, including system libraries. Use this
14233 switch to conform to a non-default data model.
14235 @item -mfix-and-continue
14236 @itemx -ffix-and-continue
14237 @itemx -findirect-data
14238 @opindex mfix-and-continue
14239 @opindex ffix-and-continue
14240 @opindex findirect-data
14241 Generate code suitable for fast turnaround development, such as to
14242 allow GDB to dynamically load @file{.o} files into already-running
14243 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14244 are provided for backwards compatibility.
14248 Loads all members of static archive libraries.
14249 See man ld(1) for more information.
14251 @item -arch_errors_fatal
14252 @opindex arch_errors_fatal
14253 Cause the errors having to do with files that have the wrong architecture
14256 @item -bind_at_load
14257 @opindex bind_at_load
14258 Causes the output file to be marked such that the dynamic linker will
14259 bind all undefined references when the file is loaded or launched.
14263 Produce a Mach-o bundle format file.
14264 See man ld(1) for more information.
14266 @item -bundle_loader @var{executable}
14267 @opindex bundle_loader
14268 This option specifies the @var{executable} that will load the build
14269 output file being linked. See man ld(1) for more information.
14272 @opindex dynamiclib
14273 When passed this option, GCC produces a dynamic library instead of
14274 an executable when linking, using the Darwin @file{libtool} command.
14276 @item -force_cpusubtype_ALL
14277 @opindex force_cpusubtype_ALL
14278 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14279 one controlled by the @option{-mcpu} or @option{-march} option.
14281 @item -allowable_client @var{client_name}
14282 @itemx -client_name
14283 @itemx -compatibility_version
14284 @itemx -current_version
14286 @itemx -dependency-file
14288 @itemx -dylinker_install_name
14290 @itemx -exported_symbols_list
14293 @itemx -flat_namespace
14294 @itemx -force_flat_namespace
14295 @itemx -headerpad_max_install_names
14298 @itemx -install_name
14299 @itemx -keep_private_externs
14300 @itemx -multi_module
14301 @itemx -multiply_defined
14302 @itemx -multiply_defined_unused
14305 @itemx -no_dead_strip_inits_and_terms
14306 @itemx -nofixprebinding
14307 @itemx -nomultidefs
14309 @itemx -noseglinkedit
14310 @itemx -pagezero_size
14312 @itemx -prebind_all_twolevel_modules
14313 @itemx -private_bundle
14315 @itemx -read_only_relocs
14317 @itemx -sectobjectsymbols
14321 @itemx -sectobjectsymbols
14324 @itemx -segs_read_only_addr
14326 @itemx -segs_read_write_addr
14327 @itemx -seg_addr_table
14328 @itemx -seg_addr_table_filename
14329 @itemx -seglinkedit
14331 @itemx -segs_read_only_addr
14332 @itemx -segs_read_write_addr
14333 @itemx -single_module
14335 @itemx -sub_library
14337 @itemx -sub_umbrella
14338 @itemx -twolevel_namespace
14341 @itemx -unexported_symbols_list
14342 @itemx -weak_reference_mismatches
14343 @itemx -whatsloaded
14344 @opindex allowable_client
14345 @opindex client_name
14346 @opindex compatibility_version
14347 @opindex current_version
14348 @opindex dead_strip
14349 @opindex dependency-file
14350 @opindex dylib_file
14351 @opindex dylinker_install_name
14353 @opindex exported_symbols_list
14355 @opindex flat_namespace
14356 @opindex force_flat_namespace
14357 @opindex headerpad_max_install_names
14358 @opindex image_base
14360 @opindex install_name
14361 @opindex keep_private_externs
14362 @opindex multi_module
14363 @opindex multiply_defined
14364 @opindex multiply_defined_unused
14365 @opindex noall_load
14366 @opindex no_dead_strip_inits_and_terms
14367 @opindex nofixprebinding
14368 @opindex nomultidefs
14370 @opindex noseglinkedit
14371 @opindex pagezero_size
14373 @opindex prebind_all_twolevel_modules
14374 @opindex private_bundle
14375 @opindex read_only_relocs
14377 @opindex sectobjectsymbols
14380 @opindex sectcreate
14381 @opindex sectobjectsymbols
14384 @opindex segs_read_only_addr
14385 @opindex segs_read_write_addr
14386 @opindex seg_addr_table
14387 @opindex seg_addr_table_filename
14388 @opindex seglinkedit
14390 @opindex segs_read_only_addr
14391 @opindex segs_read_write_addr
14392 @opindex single_module
14394 @opindex sub_library
14395 @opindex sub_umbrella
14396 @opindex twolevel_namespace
14399 @opindex unexported_symbols_list
14400 @opindex weak_reference_mismatches
14401 @opindex whatsloaded
14402 These options are passed to the Darwin linker. The Darwin linker man page
14403 describes them in detail.
14406 @node DEC Alpha Options
14407 @subsection DEC Alpha Options
14409 These @samp{-m} options are defined for the DEC Alpha implementations:
14412 @item -mno-soft-float
14413 @itemx -msoft-float
14414 @opindex mno-soft-float
14415 @opindex msoft-float
14416 Use (do not use) the hardware floating-point instructions for
14417 floating-point operations. When @option{-msoft-float} is specified,
14418 functions in @file{libgcc.a} are used to perform floating-point
14419 operations. Unless they are replaced by routines that emulate the
14420 floating-point operations, or compiled in such a way as to call such
14421 emulations routines, these routines issue floating-point
14422 operations. If you are compiling for an Alpha without floating-point
14423 operations, you must ensure that the library is built so as not to call
14426 Note that Alpha implementations without floating-point operations are
14427 required to have floating-point registers.
14430 @itemx -mno-fp-regs
14432 @opindex mno-fp-regs
14433 Generate code that uses (does not use) the floating-point register set.
14434 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14435 register set is not used, floating-point operands are passed in integer
14436 registers as if they were integers and floating-point results are passed
14437 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14438 so any function with a floating-point argument or return value called by code
14439 compiled with @option{-mno-fp-regs} must also be compiled with that
14442 A typical use of this option is building a kernel that does not use,
14443 and hence need not save and restore, any floating-point registers.
14447 The Alpha architecture implements floating-point hardware optimized for
14448 maximum performance. It is mostly compliant with the IEEE floating-point
14449 standard. However, for full compliance, software assistance is
14450 required. This option generates code fully IEEE-compliant code
14451 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14452 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14453 defined during compilation. The resulting code is less efficient but is
14454 able to correctly support denormalized numbers and exceptional IEEE
14455 values such as not-a-number and plus/minus infinity. Other Alpha
14456 compilers call this option @option{-ieee_with_no_inexact}.
14458 @item -mieee-with-inexact
14459 @opindex mieee-with-inexact
14460 This is like @option{-mieee} except the generated code also maintains
14461 the IEEE @var{inexact-flag}. Turning on this option causes the
14462 generated code to implement fully-compliant IEEE math. In addition to
14463 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14464 macro. On some Alpha implementations the resulting code may execute
14465 significantly slower than the code generated by default. Since there is
14466 very little code that depends on the @var{inexact-flag}, you should
14467 normally not specify this option. Other Alpha compilers call this
14468 option @option{-ieee_with_inexact}.
14470 @item -mfp-trap-mode=@var{trap-mode}
14471 @opindex mfp-trap-mode
14472 This option controls what floating-point related traps are enabled.
14473 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14474 The trap mode can be set to one of four values:
14478 This is the default (normal) setting. The only traps that are enabled
14479 are the ones that cannot be disabled in software (e.g., division by zero
14483 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14487 Like @samp{u}, but the instructions are marked to be safe for software
14488 completion (see Alpha architecture manual for details).
14491 Like @samp{su}, but inexact traps are enabled as well.
14494 @item -mfp-rounding-mode=@var{rounding-mode}
14495 @opindex mfp-rounding-mode
14496 Selects the IEEE rounding mode. Other Alpha compilers call this option
14497 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14502 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14503 the nearest machine number or towards the even machine number in case
14507 Round towards minus infinity.
14510 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14513 Dynamic rounding mode. A field in the floating-point control register
14514 (@var{fpcr}, see Alpha architecture reference manual) controls the
14515 rounding mode in effect. The C library initializes this register for
14516 rounding towards plus infinity. Thus, unless your program modifies the
14517 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14520 @item -mtrap-precision=@var{trap-precision}
14521 @opindex mtrap-precision
14522 In the Alpha architecture, floating-point traps are imprecise. This
14523 means without software assistance it is impossible to recover from a
14524 floating trap and program execution normally needs to be terminated.
14525 GCC can generate code that can assist operating system trap handlers
14526 in determining the exact location that caused a floating-point trap.
14527 Depending on the requirements of an application, different levels of
14528 precisions can be selected:
14532 Program precision. This option is the default and means a trap handler
14533 can only identify which program caused a floating-point exception.
14536 Function precision. The trap handler can determine the function that
14537 caused a floating-point exception.
14540 Instruction precision. The trap handler can determine the exact
14541 instruction that caused a floating-point exception.
14544 Other Alpha compilers provide the equivalent options called
14545 @option{-scope_safe} and @option{-resumption_safe}.
14547 @item -mieee-conformant
14548 @opindex mieee-conformant
14549 This option marks the generated code as IEEE conformant. You must not
14550 use this option unless you also specify @option{-mtrap-precision=i} and either
14551 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14552 is to emit the line @samp{.eflag 48} in the function prologue of the
14553 generated assembly file.
14555 @item -mbuild-constants
14556 @opindex mbuild-constants
14557 Normally GCC examines a 32- or 64-bit integer constant to
14558 see if it can construct it from smaller constants in two or three
14559 instructions. If it cannot, it outputs the constant as a literal and
14560 generates code to load it from the data segment at run time.
14562 Use this option to require GCC to construct @emph{all} integer constants
14563 using code, even if it takes more instructions (the maximum is six).
14565 You typically use this option to build a shared library dynamic
14566 loader. Itself a shared library, it must relocate itself in memory
14567 before it can find the variables and constants in its own data segment.
14585 Indicate whether GCC should generate code to use the optional BWX,
14586 CIX, FIX and MAX instruction sets. The default is to use the instruction
14587 sets supported by the CPU type specified via @option{-mcpu=} option or that
14588 of the CPU on which GCC was built if none is specified.
14591 @itemx -mfloat-ieee
14592 @opindex mfloat-vax
14593 @opindex mfloat-ieee
14594 Generate code that uses (does not use) VAX F and G floating-point
14595 arithmetic instead of IEEE single and double precision.
14597 @item -mexplicit-relocs
14598 @itemx -mno-explicit-relocs
14599 @opindex mexplicit-relocs
14600 @opindex mno-explicit-relocs
14601 Older Alpha assemblers provided no way to generate symbol relocations
14602 except via assembler macros. Use of these macros does not allow
14603 optimal instruction scheduling. GNU binutils as of version 2.12
14604 supports a new syntax that allows the compiler to explicitly mark
14605 which relocations should apply to which instructions. This option
14606 is mostly useful for debugging, as GCC detects the capabilities of
14607 the assembler when it is built and sets the default accordingly.
14610 @itemx -mlarge-data
14611 @opindex msmall-data
14612 @opindex mlarge-data
14613 When @option{-mexplicit-relocs} is in effect, static data is
14614 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14615 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14616 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14617 16-bit relocations off of the @code{$gp} register. This limits the
14618 size of the small data area to 64KB, but allows the variables to be
14619 directly accessed via a single instruction.
14621 The default is @option{-mlarge-data}. With this option the data area
14622 is limited to just below 2GB@. Programs that require more than 2GB of
14623 data must use @code{malloc} or @code{mmap} to allocate the data in the
14624 heap instead of in the program's data segment.
14626 When generating code for shared libraries, @option{-fpic} implies
14627 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14630 @itemx -mlarge-text
14631 @opindex msmall-text
14632 @opindex mlarge-text
14633 When @option{-msmall-text} is used, the compiler assumes that the
14634 code of the entire program (or shared library) fits in 4MB, and is
14635 thus reachable with a branch instruction. When @option{-msmall-data}
14636 is used, the compiler can assume that all local symbols share the
14637 same @code{$gp} value, and thus reduce the number of instructions
14638 required for a function call from 4 to 1.
14640 The default is @option{-mlarge-text}.
14642 @item -mcpu=@var{cpu_type}
14644 Set the instruction set and instruction scheduling parameters for
14645 machine type @var{cpu_type}. You can specify either the @samp{EV}
14646 style name or the corresponding chip number. GCC supports scheduling
14647 parameters for the EV4, EV5 and EV6 family of processors and
14648 chooses the default values for the instruction set from the processor
14649 you specify. If you do not specify a processor type, GCC defaults
14650 to the processor on which the compiler was built.
14652 Supported values for @var{cpu_type} are
14658 Schedules as an EV4 and has no instruction set extensions.
14662 Schedules as an EV5 and has no instruction set extensions.
14666 Schedules as an EV5 and supports the BWX extension.
14671 Schedules as an EV5 and supports the BWX and MAX extensions.
14675 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14679 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14682 Native toolchains also support the value @samp{native},
14683 which selects the best architecture option for the host processor.
14684 @option{-mcpu=native} has no effect if GCC does not recognize
14687 @item -mtune=@var{cpu_type}
14689 Set only the instruction scheduling parameters for machine type
14690 @var{cpu_type}. The instruction set is not changed.
14692 Native toolchains also support the value @samp{native},
14693 which selects the best architecture option for the host processor.
14694 @option{-mtune=native} has no effect if GCC does not recognize
14697 @item -mmemory-latency=@var{time}
14698 @opindex mmemory-latency
14699 Sets the latency the scheduler should assume for typical memory
14700 references as seen by the application. This number is highly
14701 dependent on the memory access patterns used by the application
14702 and the size of the external cache on the machine.
14704 Valid options for @var{time} are
14708 A decimal number representing clock cycles.
14714 The compiler contains estimates of the number of clock cycles for
14715 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14716 (also called Dcache, Scache, and Bcache), as well as to main memory.
14717 Note that L3 is only valid for EV5.
14723 @subsection FR30 Options
14724 @cindex FR30 Options
14726 These options are defined specifically for the FR30 port.
14730 @item -msmall-model
14731 @opindex msmall-model
14732 Use the small address space model. This can produce smaller code, but
14733 it does assume that all symbolic values and addresses fit into a
14738 Assume that runtime support has been provided and so there is no need
14739 to include the simulator library (@file{libsim.a}) on the linker
14745 @subsection FRV Options
14746 @cindex FRV Options
14752 Only use the first 32 general-purpose registers.
14757 Use all 64 general-purpose registers.
14762 Use only the first 32 floating-point registers.
14767 Use all 64 floating-point registers.
14770 @opindex mhard-float
14772 Use hardware instructions for floating-point operations.
14775 @opindex msoft-float
14777 Use library routines for floating-point operations.
14782 Dynamically allocate condition code registers.
14787 Do not try to dynamically allocate condition code registers, only
14788 use @code{icc0} and @code{fcc0}.
14793 Change ABI to use double word insns.
14798 Do not use double word instructions.
14803 Use floating-point double instructions.
14806 @opindex mno-double
14808 Do not use floating-point double instructions.
14813 Use media instructions.
14818 Do not use media instructions.
14823 Use multiply and add/subtract instructions.
14826 @opindex mno-muladd
14828 Do not use multiply and add/subtract instructions.
14833 Select the FDPIC ABI, which uses function descriptors to represent
14834 pointers to functions. Without any PIC/PIE-related options, it
14835 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
14836 assumes GOT entries and small data are within a 12-bit range from the
14837 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
14838 are computed with 32 bits.
14839 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14842 @opindex minline-plt
14844 Enable inlining of PLT entries in function calls to functions that are
14845 not known to bind locally. It has no effect without @option{-mfdpic}.
14846 It's enabled by default if optimizing for speed and compiling for
14847 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
14848 optimization option such as @option{-O3} or above is present in the
14854 Assume a large TLS segment when generating thread-local code.
14859 Do not assume a large TLS segment when generating thread-local code.
14864 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
14865 that is known to be in read-only sections. It's enabled by default,
14866 except for @option{-fpic} or @option{-fpie}: even though it may help
14867 make the global offset table smaller, it trades 1 instruction for 4.
14868 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
14869 one of which may be shared by multiple symbols, and it avoids the need
14870 for a GOT entry for the referenced symbol, so it's more likely to be a
14871 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
14873 @item -multilib-library-pic
14874 @opindex multilib-library-pic
14876 Link with the (library, not FD) pic libraries. It's implied by
14877 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
14878 @option{-fpic} without @option{-mfdpic}. You should never have to use
14882 @opindex mlinked-fp
14884 Follow the EABI requirement of always creating a frame pointer whenever
14885 a stack frame is allocated. This option is enabled by default and can
14886 be disabled with @option{-mno-linked-fp}.
14889 @opindex mlong-calls
14891 Use indirect addressing to call functions outside the current
14892 compilation unit. This allows the functions to be placed anywhere
14893 within the 32-bit address space.
14895 @item -malign-labels
14896 @opindex malign-labels
14898 Try to align labels to an 8-byte boundary by inserting NOPs into the
14899 previous packet. This option only has an effect when VLIW packing
14900 is enabled. It doesn't create new packets; it merely adds NOPs to
14903 @item -mlibrary-pic
14904 @opindex mlibrary-pic
14906 Generate position-independent EABI code.
14911 Use only the first four media accumulator registers.
14916 Use all eight media accumulator registers.
14921 Pack VLIW instructions.
14926 Do not pack VLIW instructions.
14929 @opindex mno-eflags
14931 Do not mark ABI switches in e_flags.
14934 @opindex mcond-move
14936 Enable the use of conditional-move instructions (default).
14938 This switch is mainly for debugging the compiler and will likely be removed
14939 in a future version.
14941 @item -mno-cond-move
14942 @opindex mno-cond-move
14944 Disable the use of conditional-move instructions.
14946 This switch is mainly for debugging the compiler and will likely be removed
14947 in a future version.
14952 Enable the use of conditional set instructions (default).
14954 This switch is mainly for debugging the compiler and will likely be removed
14955 in a future version.
14960 Disable the use of conditional set instructions.
14962 This switch is mainly for debugging the compiler and will likely be removed
14963 in a future version.
14966 @opindex mcond-exec
14968 Enable the use of conditional execution (default).
14970 This switch is mainly for debugging the compiler and will likely be removed
14971 in a future version.
14973 @item -mno-cond-exec
14974 @opindex mno-cond-exec
14976 Disable the use of conditional execution.
14978 This switch is mainly for debugging the compiler and will likely be removed
14979 in a future version.
14981 @item -mvliw-branch
14982 @opindex mvliw-branch
14984 Run a pass to pack branches into VLIW instructions (default).
14986 This switch is mainly for debugging the compiler and will likely be removed
14987 in a future version.
14989 @item -mno-vliw-branch
14990 @opindex mno-vliw-branch
14992 Do not run a pass to pack branches into VLIW instructions.
14994 This switch is mainly for debugging the compiler and will likely be removed
14995 in a future version.
14997 @item -mmulti-cond-exec
14998 @opindex mmulti-cond-exec
15000 Enable optimization of @code{&&} and @code{||} in conditional execution
15003 This switch is mainly for debugging the compiler and will likely be removed
15004 in a future version.
15006 @item -mno-multi-cond-exec
15007 @opindex mno-multi-cond-exec
15009 Disable optimization of @code{&&} and @code{||} in conditional execution.
15011 This switch is mainly for debugging the compiler and will likely be removed
15012 in a future version.
15014 @item -mnested-cond-exec
15015 @opindex mnested-cond-exec
15017 Enable nested conditional execution optimizations (default).
15019 This switch is mainly for debugging the compiler and will likely be removed
15020 in a future version.
15022 @item -mno-nested-cond-exec
15023 @opindex mno-nested-cond-exec
15025 Disable nested conditional execution optimizations.
15027 This switch is mainly for debugging the compiler and will likely be removed
15028 in a future version.
15030 @item -moptimize-membar
15031 @opindex moptimize-membar
15033 This switch removes redundant @code{membar} instructions from the
15034 compiler-generated code. It is enabled by default.
15036 @item -mno-optimize-membar
15037 @opindex mno-optimize-membar
15039 This switch disables the automatic removal of redundant @code{membar}
15040 instructions from the generated code.
15042 @item -mtomcat-stats
15043 @opindex mtomcat-stats
15045 Cause gas to print out tomcat statistics.
15047 @item -mcpu=@var{cpu}
15050 Select the processor type for which to generate code. Possible values are
15051 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15052 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15056 @node GNU/Linux Options
15057 @subsection GNU/Linux Options
15059 These @samp{-m} options are defined for GNU/Linux targets:
15064 Use the GNU C library. This is the default except
15065 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
15069 Use uClibc C library. This is the default on
15070 @samp{*-*-linux-*uclibc*} targets.
15074 Use Bionic C library. This is the default on
15075 @samp{*-*-linux-*android*} targets.
15079 Compile code compatible with Android platform. This is the default on
15080 @samp{*-*-linux-*android*} targets.
15082 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15083 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15084 this option makes the GCC driver pass Android-specific options to the linker.
15085 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15088 @item -tno-android-cc
15089 @opindex tno-android-cc
15090 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15091 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15092 @option{-fno-rtti} by default.
15094 @item -tno-android-ld
15095 @opindex tno-android-ld
15096 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15097 linking options to the linker.
15101 @node H8/300 Options
15102 @subsection H8/300 Options
15104 These @samp{-m} options are defined for the H8/300 implementations:
15109 Shorten some address references at link time, when possible; uses the
15110 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15111 ld, Using ld}, for a fuller description.
15115 Generate code for the H8/300H@.
15119 Generate code for the H8S@.
15123 Generate code for the H8S and H8/300H in the normal mode. This switch
15124 must be used either with @option{-mh} or @option{-ms}.
15128 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15132 Extended registers are stored on stack before execution of function
15133 with monitor attribute. Default option is @option{-mexr}.
15134 This option is valid only for H8S targets.
15138 Extended registers are not stored on stack before execution of function
15139 with monitor attribute. Default option is @option{-mno-exr}.
15140 This option is valid only for H8S targets.
15144 Make @code{int} data 32 bits by default.
15147 @opindex malign-300
15148 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15149 The default for the H8/300H and H8S is to align longs and floats on
15151 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15152 This option has no effect on the H8/300.
15156 @subsection HPPA Options
15157 @cindex HPPA Options
15159 These @samp{-m} options are defined for the HPPA family of computers:
15162 @item -march=@var{architecture-type}
15164 Generate code for the specified architecture. The choices for
15165 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15166 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15167 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15168 architecture option for your machine. Code compiled for lower numbered
15169 architectures runs on higher numbered architectures, but not the
15172 @item -mpa-risc-1-0
15173 @itemx -mpa-risc-1-1
15174 @itemx -mpa-risc-2-0
15175 @opindex mpa-risc-1-0
15176 @opindex mpa-risc-1-1
15177 @opindex mpa-risc-2-0
15178 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15180 @item -mjump-in-delay
15181 @opindex mjump-in-delay
15182 This option is ignored and provided for compatibility purposes only.
15184 @item -mdisable-fpregs
15185 @opindex mdisable-fpregs
15186 Prevent floating-point registers from being used in any manner. This is
15187 necessary for compiling kernels that perform lazy context switching of
15188 floating-point registers. If you use this option and attempt to perform
15189 floating-point operations, the compiler aborts.
15191 @item -mdisable-indexing
15192 @opindex mdisable-indexing
15193 Prevent the compiler from using indexing address modes. This avoids some
15194 rather obscure problems when compiling MIG generated code under MACH@.
15196 @item -mno-space-regs
15197 @opindex mno-space-regs
15198 Generate code that assumes the target has no space registers. This allows
15199 GCC to generate faster indirect calls and use unscaled index address modes.
15201 Such code is suitable for level 0 PA systems and kernels.
15203 @item -mfast-indirect-calls
15204 @opindex mfast-indirect-calls
15205 Generate code that assumes calls never cross space boundaries. This
15206 allows GCC to emit code that performs faster indirect calls.
15208 This option does not work in the presence of shared libraries or nested
15211 @item -mfixed-range=@var{register-range}
15212 @opindex mfixed-range
15213 Generate code treating the given register range as fixed registers.
15214 A fixed register is one that the register allocator cannot use. This is
15215 useful when compiling kernel code. A register range is specified as
15216 two registers separated by a dash. Multiple register ranges can be
15217 specified separated by a comma.
15219 @item -mlong-load-store
15220 @opindex mlong-load-store
15221 Generate 3-instruction load and store sequences as sometimes required by
15222 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15225 @item -mportable-runtime
15226 @opindex mportable-runtime
15227 Use the portable calling conventions proposed by HP for ELF systems.
15231 Enable the use of assembler directives only GAS understands.
15233 @item -mschedule=@var{cpu-type}
15235 Schedule code according to the constraints for the machine type
15236 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15237 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15238 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15239 proper scheduling option for your machine. The default scheduling is
15243 @opindex mlinker-opt
15244 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15245 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15246 linkers in which they give bogus error messages when linking some programs.
15249 @opindex msoft-float
15250 Generate output containing library calls for floating point.
15251 @strong{Warning:} the requisite libraries are not available for all HPPA
15252 targets. Normally the facilities of the machine's usual C compiler are
15253 used, but this cannot be done directly in cross-compilation. You must make
15254 your own arrangements to provide suitable library functions for
15257 @option{-msoft-float} changes the calling convention in the output file;
15258 therefore, it is only useful if you compile @emph{all} of a program with
15259 this option. In particular, you need to compile @file{libgcc.a}, the
15260 library that comes with GCC, with @option{-msoft-float} in order for
15265 Generate the predefine, @code{_SIO}, for server IO@. The default is
15266 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15267 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15268 options are available under HP-UX and HI-UX@.
15272 Use options specific to GNU @command{ld}.
15273 This passes @option{-shared} to @command{ld} when
15274 building a shared library. It is the default when GCC is configured,
15275 explicitly or implicitly, with the GNU linker. This option does not
15276 affect which @command{ld} is called; it only changes what parameters
15277 are passed to that @command{ld}.
15278 The @command{ld} that is called is determined by the
15279 @option{--with-ld} configure option, GCC's program search path, and
15280 finally by the user's @env{PATH}. The linker used by GCC can be printed
15281 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15282 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15286 Use options specific to HP @command{ld}.
15287 This passes @option{-b} to @command{ld} when building
15288 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15289 links. It is the default when GCC is configured, explicitly or
15290 implicitly, with the HP linker. This option does not affect
15291 which @command{ld} is called; it only changes what parameters are passed to that
15293 The @command{ld} that is called is determined by the @option{--with-ld}
15294 configure option, GCC's program search path, and finally by the user's
15295 @env{PATH}. The linker used by GCC can be printed using @samp{which
15296 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15297 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15300 @opindex mno-long-calls
15301 Generate code that uses long call sequences. This ensures that a call
15302 is always able to reach linker generated stubs. The default is to generate
15303 long calls only when the distance from the call site to the beginning
15304 of the function or translation unit, as the case may be, exceeds a
15305 predefined limit set by the branch type being used. The limits for
15306 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15307 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15310 Distances are measured from the beginning of functions when using the
15311 @option{-ffunction-sections} option, or when using the @option{-mgas}
15312 and @option{-mno-portable-runtime} options together under HP-UX with
15315 It is normally not desirable to use this option as it degrades
15316 performance. However, it may be useful in large applications,
15317 particularly when partial linking is used to build the application.
15319 The types of long calls used depends on the capabilities of the
15320 assembler and linker, and the type of code being generated. The
15321 impact on systems that support long absolute calls, and long pic
15322 symbol-difference or pc-relative calls should be relatively small.
15323 However, an indirect call is used on 32-bit ELF systems in pic code
15324 and it is quite long.
15326 @item -munix=@var{unix-std}
15328 Generate compiler predefines and select a startfile for the specified
15329 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15330 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15331 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15332 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15333 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15336 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15337 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15338 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15339 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15340 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15341 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15343 It is @emph{important} to note that this option changes the interfaces
15344 for various library routines. It also affects the operational behavior
15345 of the C library. Thus, @emph{extreme} care is needed in using this
15348 Library code that is intended to operate with more than one UNIX
15349 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15350 as appropriate. Most GNU software doesn't provide this capability.
15354 Suppress the generation of link options to search libdld.sl when the
15355 @option{-static} option is specified on HP-UX 10 and later.
15359 The HP-UX implementation of setlocale in libc has a dependency on
15360 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15361 when the @option{-static} option is specified, special link options
15362 are needed to resolve this dependency.
15364 On HP-UX 10 and later, the GCC driver adds the necessary options to
15365 link with libdld.sl when the @option{-static} option is specified.
15366 This causes the resulting binary to be dynamic. On the 64-bit port,
15367 the linkers generate dynamic binaries by default in any case. The
15368 @option{-nolibdld} option can be used to prevent the GCC driver from
15369 adding these link options.
15373 Add support for multithreading with the @dfn{dce thread} library
15374 under HP-UX@. This option sets flags for both the preprocessor and
15378 @node IA-64 Options
15379 @subsection IA-64 Options
15380 @cindex IA-64 Options
15382 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15386 @opindex mbig-endian
15387 Generate code for a big-endian target. This is the default for HP-UX@.
15389 @item -mlittle-endian
15390 @opindex mlittle-endian
15391 Generate code for a little-endian target. This is the default for AIX5
15397 @opindex mno-gnu-as
15398 Generate (or don't) code for the GNU assembler. This is the default.
15399 @c Also, this is the default if the configure option @option{--with-gnu-as}
15405 @opindex mno-gnu-ld
15406 Generate (or don't) code for the GNU linker. This is the default.
15407 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15412 Generate code that does not use a global pointer register. The result
15413 is not position independent code, and violates the IA-64 ABI@.
15415 @item -mvolatile-asm-stop
15416 @itemx -mno-volatile-asm-stop
15417 @opindex mvolatile-asm-stop
15418 @opindex mno-volatile-asm-stop
15419 Generate (or don't) a stop bit immediately before and after volatile asm
15422 @item -mregister-names
15423 @itemx -mno-register-names
15424 @opindex mregister-names
15425 @opindex mno-register-names
15426 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15427 the stacked registers. This may make assembler output more readable.
15433 Disable (or enable) optimizations that use the small data section. This may
15434 be useful for working around optimizer bugs.
15436 @item -mconstant-gp
15437 @opindex mconstant-gp
15438 Generate code that uses a single constant global pointer value. This is
15439 useful when compiling kernel code.
15443 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15444 This is useful when compiling firmware code.
15446 @item -minline-float-divide-min-latency
15447 @opindex minline-float-divide-min-latency
15448 Generate code for inline divides of floating-point values
15449 using the minimum latency algorithm.
15451 @item -minline-float-divide-max-throughput
15452 @opindex minline-float-divide-max-throughput
15453 Generate code for inline divides of floating-point values
15454 using the maximum throughput algorithm.
15456 @item -mno-inline-float-divide
15457 @opindex mno-inline-float-divide
15458 Do not generate inline code for divides of floating-point values.
15460 @item -minline-int-divide-min-latency
15461 @opindex minline-int-divide-min-latency
15462 Generate code for inline divides of integer values
15463 using the minimum latency algorithm.
15465 @item -minline-int-divide-max-throughput
15466 @opindex minline-int-divide-max-throughput
15467 Generate code for inline divides of integer values
15468 using the maximum throughput algorithm.
15470 @item -mno-inline-int-divide
15471 @opindex mno-inline-int-divide
15472 Do not generate inline code for divides of integer values.
15474 @item -minline-sqrt-min-latency
15475 @opindex minline-sqrt-min-latency
15476 Generate code for inline square roots
15477 using the minimum latency algorithm.
15479 @item -minline-sqrt-max-throughput
15480 @opindex minline-sqrt-max-throughput
15481 Generate code for inline square roots
15482 using the maximum throughput algorithm.
15484 @item -mno-inline-sqrt
15485 @opindex mno-inline-sqrt
15486 Do not generate inline code for @code{sqrt}.
15489 @itemx -mno-fused-madd
15490 @opindex mfused-madd
15491 @opindex mno-fused-madd
15492 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
15493 instructions. The default is to use these instructions.
15495 @item -mno-dwarf2-asm
15496 @itemx -mdwarf2-asm
15497 @opindex mno-dwarf2-asm
15498 @opindex mdwarf2-asm
15499 Don't (or do) generate assembler code for the DWARF 2 line number debugging
15500 info. This may be useful when not using the GNU assembler.
15502 @item -mearly-stop-bits
15503 @itemx -mno-early-stop-bits
15504 @opindex mearly-stop-bits
15505 @opindex mno-early-stop-bits
15506 Allow stop bits to be placed earlier than immediately preceding the
15507 instruction that triggered the stop bit. This can improve instruction
15508 scheduling, but does not always do so.
15510 @item -mfixed-range=@var{register-range}
15511 @opindex mfixed-range
15512 Generate code treating the given register range as fixed registers.
15513 A fixed register is one that the register allocator cannot use. This is
15514 useful when compiling kernel code. A register range is specified as
15515 two registers separated by a dash. Multiple register ranges can be
15516 specified separated by a comma.
15518 @item -mtls-size=@var{tls-size}
15520 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
15523 @item -mtune=@var{cpu-type}
15525 Tune the instruction scheduling for a particular CPU, Valid values are
15526 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
15527 and @samp{mckinley}.
15533 Generate code for a 32-bit or 64-bit environment.
15534 The 32-bit environment sets int, long and pointer to 32 bits.
15535 The 64-bit environment sets int to 32 bits and long and pointer
15536 to 64 bits. These are HP-UX specific flags.
15538 @item -mno-sched-br-data-spec
15539 @itemx -msched-br-data-spec
15540 @opindex mno-sched-br-data-spec
15541 @opindex msched-br-data-spec
15542 (Dis/En)able data speculative scheduling before reload.
15543 This results in generation of @code{ld.a} instructions and
15544 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15545 The default is 'disable'.
15547 @item -msched-ar-data-spec
15548 @itemx -mno-sched-ar-data-spec
15549 @opindex msched-ar-data-spec
15550 @opindex mno-sched-ar-data-spec
15551 (En/Dis)able data speculative scheduling after reload.
15552 This results in generation of @code{ld.a} instructions and
15553 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15554 The default is 'enable'.
15556 @item -mno-sched-control-spec
15557 @itemx -msched-control-spec
15558 @opindex mno-sched-control-spec
15559 @opindex msched-control-spec
15560 (Dis/En)able control speculative scheduling. This feature is
15561 available only during region scheduling (i.e.@: before reload).
15562 This results in generation of the @code{ld.s} instructions and
15563 the corresponding check instructions @code{chk.s}.
15564 The default is 'disable'.
15566 @item -msched-br-in-data-spec
15567 @itemx -mno-sched-br-in-data-spec
15568 @opindex msched-br-in-data-spec
15569 @opindex mno-sched-br-in-data-spec
15570 (En/Dis)able speculative scheduling of the instructions that
15571 are dependent on the data speculative loads before reload.
15572 This is effective only with @option{-msched-br-data-spec} enabled.
15573 The default is 'enable'.
15575 @item -msched-ar-in-data-spec
15576 @itemx -mno-sched-ar-in-data-spec
15577 @opindex msched-ar-in-data-spec
15578 @opindex mno-sched-ar-in-data-spec
15579 (En/Dis)able speculative scheduling of the instructions that
15580 are dependent on the data speculative loads after reload.
15581 This is effective only with @option{-msched-ar-data-spec} enabled.
15582 The default is 'enable'.
15584 @item -msched-in-control-spec
15585 @itemx -mno-sched-in-control-spec
15586 @opindex msched-in-control-spec
15587 @opindex mno-sched-in-control-spec
15588 (En/Dis)able speculative scheduling of the instructions that
15589 are dependent on the control speculative loads.
15590 This is effective only with @option{-msched-control-spec} enabled.
15591 The default is 'enable'.
15593 @item -mno-sched-prefer-non-data-spec-insns
15594 @itemx -msched-prefer-non-data-spec-insns
15595 @opindex mno-sched-prefer-non-data-spec-insns
15596 @opindex msched-prefer-non-data-spec-insns
15597 If enabled, data-speculative instructions are chosen for schedule
15598 only if there are no other choices at the moment. This makes
15599 the use of the data speculation much more conservative.
15600 The default is 'disable'.
15602 @item -mno-sched-prefer-non-control-spec-insns
15603 @itemx -msched-prefer-non-control-spec-insns
15604 @opindex mno-sched-prefer-non-control-spec-insns
15605 @opindex msched-prefer-non-control-spec-insns
15606 If enabled, control-speculative instructions are chosen for schedule
15607 only if there are no other choices at the moment. This makes
15608 the use of the control speculation much more conservative.
15609 The default is 'disable'.
15611 @item -mno-sched-count-spec-in-critical-path
15612 @itemx -msched-count-spec-in-critical-path
15613 @opindex mno-sched-count-spec-in-critical-path
15614 @opindex msched-count-spec-in-critical-path
15615 If enabled, speculative dependencies are considered during
15616 computation of the instructions priorities. This makes the use of the
15617 speculation a bit more conservative.
15618 The default is 'disable'.
15620 @item -msched-spec-ldc
15621 @opindex msched-spec-ldc
15622 Use a simple data speculation check. This option is on by default.
15624 @item -msched-control-spec-ldc
15625 @opindex msched-spec-ldc
15626 Use a simple check for control speculation. This option is on by default.
15628 @item -msched-stop-bits-after-every-cycle
15629 @opindex msched-stop-bits-after-every-cycle
15630 Place a stop bit after every cycle when scheduling. This option is on
15633 @item -msched-fp-mem-deps-zero-cost
15634 @opindex msched-fp-mem-deps-zero-cost
15635 Assume that floating-point stores and loads are not likely to cause a conflict
15636 when placed into the same instruction group. This option is disabled by
15639 @item -msel-sched-dont-check-control-spec
15640 @opindex msel-sched-dont-check-control-spec
15641 Generate checks for control speculation in selective scheduling.
15642 This flag is disabled by default.
15644 @item -msched-max-memory-insns=@var{max-insns}
15645 @opindex msched-max-memory-insns
15646 Limit on the number of memory insns per instruction group, giving lower
15647 priority to subsequent memory insns attempting to schedule in the same
15648 instruction group. Frequently useful to prevent cache bank conflicts.
15649 The default value is 1.
15651 @item -msched-max-memory-insns-hard-limit
15652 @opindex msched-max-memory-insns-hard-limit
15653 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
15654 disallowing more than that number in an instruction group.
15655 Otherwise, the limit is ``soft'', meaning that non-memory operations
15656 are preferred when the limit is reached, but memory operations may still
15662 @subsection LM32 Options
15663 @cindex LM32 options
15665 These @option{-m} options are defined for the LatticeMico32 architecture:
15668 @item -mbarrel-shift-enabled
15669 @opindex mbarrel-shift-enabled
15670 Enable barrel-shift instructions.
15672 @item -mdivide-enabled
15673 @opindex mdivide-enabled
15674 Enable divide and modulus instructions.
15676 @item -mmultiply-enabled
15677 @opindex multiply-enabled
15678 Enable multiply instructions.
15680 @item -msign-extend-enabled
15681 @opindex msign-extend-enabled
15682 Enable sign extend instructions.
15684 @item -muser-enabled
15685 @opindex muser-enabled
15686 Enable user-defined instructions.
15691 @subsection M32C Options
15692 @cindex M32C options
15695 @item -mcpu=@var{name}
15697 Select the CPU for which code is generated. @var{name} may be one of
15698 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
15699 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
15700 the M32C/80 series.
15704 Specifies that the program will be run on the simulator. This causes
15705 an alternate runtime library to be linked in which supports, for
15706 example, file I/O@. You must not use this option when generating
15707 programs that will run on real hardware; you must provide your own
15708 runtime library for whatever I/O functions are needed.
15710 @item -memregs=@var{number}
15712 Specifies the number of memory-based pseudo-registers GCC uses
15713 during code generation. These pseudo-registers are used like real
15714 registers, so there is a tradeoff between GCC's ability to fit the
15715 code into available registers, and the performance penalty of using
15716 memory instead of registers. Note that all modules in a program must
15717 be compiled with the same value for this option. Because of that, you
15718 must not use this option with GCC's default runtime libraries.
15722 @node M32R/D Options
15723 @subsection M32R/D Options
15724 @cindex M32R/D options
15726 These @option{-m} options are defined for Renesas M32R/D architectures:
15731 Generate code for the M32R/2@.
15735 Generate code for the M32R/X@.
15739 Generate code for the M32R@. This is the default.
15741 @item -mmodel=small
15742 @opindex mmodel=small
15743 Assume all objects live in the lower 16MB of memory (so that their addresses
15744 can be loaded with the @code{ld24} instruction), and assume all subroutines
15745 are reachable with the @code{bl} instruction.
15746 This is the default.
15748 The addressability of a particular object can be set with the
15749 @code{model} attribute.
15751 @item -mmodel=medium
15752 @opindex mmodel=medium
15753 Assume objects may be anywhere in the 32-bit address space (the compiler
15754 generates @code{seth/add3} instructions to load their addresses), and
15755 assume all subroutines are reachable with the @code{bl} instruction.
15757 @item -mmodel=large
15758 @opindex mmodel=large
15759 Assume objects may be anywhere in the 32-bit address space (the compiler
15760 generates @code{seth/add3} instructions to load their addresses), and
15761 assume subroutines may not be reachable with the @code{bl} instruction
15762 (the compiler generates the much slower @code{seth/add3/jl}
15763 instruction sequence).
15766 @opindex msdata=none
15767 Disable use of the small data area. Variables are put into
15768 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
15769 @code{section} attribute has been specified).
15770 This is the default.
15772 The small data area consists of sections @code{.sdata} and @code{.sbss}.
15773 Objects may be explicitly put in the small data area with the
15774 @code{section} attribute using one of these sections.
15776 @item -msdata=sdata
15777 @opindex msdata=sdata
15778 Put small global and static data in the small data area, but do not
15779 generate special code to reference them.
15782 @opindex msdata=use
15783 Put small global and static data in the small data area, and generate
15784 special instructions to reference them.
15788 @cindex smaller data references
15789 Put global and static objects less than or equal to @var{num} bytes
15790 into the small data or BSS sections instead of the normal data or BSS
15791 sections. The default value of @var{num} is 8.
15792 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
15793 for this option to have any effect.
15795 All modules should be compiled with the same @option{-G @var{num}} value.
15796 Compiling with different values of @var{num} may or may not work; if it
15797 doesn't the linker gives an error message---incorrect code is not
15802 Makes the M32R-specific code in the compiler display some statistics
15803 that might help in debugging programs.
15805 @item -malign-loops
15806 @opindex malign-loops
15807 Align all loops to a 32-byte boundary.
15809 @item -mno-align-loops
15810 @opindex mno-align-loops
15811 Do not enforce a 32-byte alignment for loops. This is the default.
15813 @item -missue-rate=@var{number}
15814 @opindex missue-rate=@var{number}
15815 Issue @var{number} instructions per cycle. @var{number} can only be 1
15818 @item -mbranch-cost=@var{number}
15819 @opindex mbranch-cost=@var{number}
15820 @var{number} can only be 1 or 2. If it is 1 then branches are
15821 preferred over conditional code, if it is 2, then the opposite applies.
15823 @item -mflush-trap=@var{number}
15824 @opindex mflush-trap=@var{number}
15825 Specifies the trap number to use to flush the cache. The default is
15826 12. Valid numbers are between 0 and 15 inclusive.
15828 @item -mno-flush-trap
15829 @opindex mno-flush-trap
15830 Specifies that the cache cannot be flushed by using a trap.
15832 @item -mflush-func=@var{name}
15833 @opindex mflush-func=@var{name}
15834 Specifies the name of the operating system function to call to flush
15835 the cache. The default is @samp{_flush_cache}, but a function call
15836 is only used if a trap is not available.
15838 @item -mno-flush-func
15839 @opindex mno-flush-func
15840 Indicates that there is no OS function for flushing the cache.
15844 @node M680x0 Options
15845 @subsection M680x0 Options
15846 @cindex M680x0 options
15848 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
15849 The default settings depend on which architecture was selected when
15850 the compiler was configured; the defaults for the most common choices
15854 @item -march=@var{arch}
15856 Generate code for a specific M680x0 or ColdFire instruction set
15857 architecture. Permissible values of @var{arch} for M680x0
15858 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
15859 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
15860 architectures are selected according to Freescale's ISA classification
15861 and the permissible values are: @samp{isaa}, @samp{isaaplus},
15862 @samp{isab} and @samp{isac}.
15864 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
15865 code for a ColdFire target. The @var{arch} in this macro is one of the
15866 @option{-march} arguments given above.
15868 When used together, @option{-march} and @option{-mtune} select code
15869 that runs on a family of similar processors but that is optimized
15870 for a particular microarchitecture.
15872 @item -mcpu=@var{cpu}
15874 Generate code for a specific M680x0 or ColdFire processor.
15875 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
15876 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
15877 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
15878 below, which also classifies the CPUs into families:
15880 @multitable @columnfractions 0.20 0.80
15881 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
15882 @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}
15883 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
15884 @item @samp{5206e} @tab @samp{5206e}
15885 @item @samp{5208} @tab @samp{5207} @samp{5208}
15886 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
15887 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
15888 @item @samp{5216} @tab @samp{5214} @samp{5216}
15889 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
15890 @item @samp{5225} @tab @samp{5224} @samp{5225}
15891 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
15892 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
15893 @item @samp{5249} @tab @samp{5249}
15894 @item @samp{5250} @tab @samp{5250}
15895 @item @samp{5271} @tab @samp{5270} @samp{5271}
15896 @item @samp{5272} @tab @samp{5272}
15897 @item @samp{5275} @tab @samp{5274} @samp{5275}
15898 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
15899 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
15900 @item @samp{5307} @tab @samp{5307}
15901 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
15902 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
15903 @item @samp{5407} @tab @samp{5407}
15904 @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}
15907 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
15908 @var{arch} is compatible with @var{cpu}. Other combinations of
15909 @option{-mcpu} and @option{-march} are rejected.
15911 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
15912 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
15913 where the value of @var{family} is given by the table above.
15915 @item -mtune=@var{tune}
15917 Tune the code for a particular microarchitecture within the
15918 constraints set by @option{-march} and @option{-mcpu}.
15919 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
15920 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
15921 and @samp{cpu32}. The ColdFire microarchitectures
15922 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
15924 You can also use @option{-mtune=68020-40} for code that needs
15925 to run relatively well on 68020, 68030 and 68040 targets.
15926 @option{-mtune=68020-60} is similar but includes 68060 targets
15927 as well. These two options select the same tuning decisions as
15928 @option{-m68020-40} and @option{-m68020-60} respectively.
15930 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
15931 when tuning for 680x0 architecture @var{arch}. It also defines
15932 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
15933 option is used. If GCC is tuning for a range of architectures,
15934 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
15935 it defines the macros for every architecture in the range.
15937 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
15938 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
15939 of the arguments given above.
15945 Generate output for a 68000. This is the default
15946 when the compiler is configured for 68000-based systems.
15947 It is equivalent to @option{-march=68000}.
15949 Use this option for microcontrollers with a 68000 or EC000 core,
15950 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
15954 Generate output for a 68010. This is the default
15955 when the compiler is configured for 68010-based systems.
15956 It is equivalent to @option{-march=68010}.
15962 Generate output for a 68020. This is the default
15963 when the compiler is configured for 68020-based systems.
15964 It is equivalent to @option{-march=68020}.
15968 Generate output for a 68030. This is the default when the compiler is
15969 configured for 68030-based systems. It is equivalent to
15970 @option{-march=68030}.
15974 Generate output for a 68040. This is the default when the compiler is
15975 configured for 68040-based systems. It is equivalent to
15976 @option{-march=68040}.
15978 This option inhibits the use of 68881/68882 instructions that have to be
15979 emulated by software on the 68040. Use this option if your 68040 does not
15980 have code to emulate those instructions.
15984 Generate output for a 68060. This is the default when the compiler is
15985 configured for 68060-based systems. It is equivalent to
15986 @option{-march=68060}.
15988 This option inhibits the use of 68020 and 68881/68882 instructions that
15989 have to be emulated by software on the 68060. Use this option if your 68060
15990 does not have code to emulate those instructions.
15994 Generate output for a CPU32. This is the default
15995 when the compiler is configured for CPU32-based systems.
15996 It is equivalent to @option{-march=cpu32}.
15998 Use this option for microcontrollers with a
15999 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16000 68336, 68340, 68341, 68349 and 68360.
16004 Generate output for a 520X ColdFire CPU@. This is the default
16005 when the compiler is configured for 520X-based systems.
16006 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16007 in favor of that option.
16009 Use this option for microcontroller with a 5200 core, including
16010 the MCF5202, MCF5203, MCF5204 and MCF5206.
16014 Generate output for a 5206e ColdFire CPU@. The option is now
16015 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16019 Generate output for a member of the ColdFire 528X family.
16020 The option is now deprecated in favor of the equivalent
16021 @option{-mcpu=528x}.
16025 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16026 in favor of the equivalent @option{-mcpu=5307}.
16030 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16031 in favor of the equivalent @option{-mcpu=5407}.
16035 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16036 This includes use of hardware floating-point instructions.
16037 The option is equivalent to @option{-mcpu=547x}, and is now
16038 deprecated in favor of that option.
16042 Generate output for a 68040, without using any of the new instructions.
16043 This results in code that can run relatively efficiently on either a
16044 68020/68881 or a 68030 or a 68040. The generated code does use the
16045 68881 instructions that are emulated on the 68040.
16047 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16051 Generate output for a 68060, without using any of the new instructions.
16052 This results in code that can run relatively efficiently on either a
16053 68020/68881 or a 68030 or a 68040. The generated code does use the
16054 68881 instructions that are emulated on the 68060.
16056 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16060 @opindex mhard-float
16062 Generate floating-point instructions. This is the default for 68020
16063 and above, and for ColdFire devices that have an FPU@. It defines the
16064 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16065 on ColdFire targets.
16068 @opindex msoft-float
16069 Do not generate floating-point instructions; use library calls instead.
16070 This is the default for 68000, 68010, and 68832 targets. It is also
16071 the default for ColdFire devices that have no FPU.
16077 Generate (do not generate) ColdFire hardware divide and remainder
16078 instructions. If @option{-march} is used without @option{-mcpu},
16079 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16080 architectures. Otherwise, the default is taken from the target CPU
16081 (either the default CPU, or the one specified by @option{-mcpu}). For
16082 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16083 @option{-mcpu=5206e}.
16085 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16089 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16090 Additionally, parameters passed on the stack are also aligned to a
16091 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16095 Do not consider type @code{int} to be 16 bits wide. This is the default.
16098 @itemx -mno-bitfield
16099 @opindex mnobitfield
16100 @opindex mno-bitfield
16101 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16102 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16106 Do use the bit-field instructions. The @option{-m68020} option implies
16107 @option{-mbitfield}. This is the default if you use a configuration
16108 designed for a 68020.
16112 Use a different function-calling convention, in which functions
16113 that take a fixed number of arguments return with the @code{rtd}
16114 instruction, which pops their arguments while returning. This
16115 saves one instruction in the caller since there is no need to pop
16116 the arguments there.
16118 This calling convention is incompatible with the one normally
16119 used on Unix, so you cannot use it if you need to call libraries
16120 compiled with the Unix compiler.
16122 Also, you must provide function prototypes for all functions that
16123 take variable numbers of arguments (including @code{printf});
16124 otherwise incorrect code is generated for calls to those
16127 In addition, seriously incorrect code results if you call a
16128 function with too many arguments. (Normally, extra arguments are
16129 harmlessly ignored.)
16131 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16132 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16136 Do not use the calling conventions selected by @option{-mrtd}.
16137 This is the default.
16140 @itemx -mno-align-int
16141 @opindex malign-int
16142 @opindex mno-align-int
16143 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16144 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16145 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16146 Aligning variables on 32-bit boundaries produces code that runs somewhat
16147 faster on processors with 32-bit busses at the expense of more memory.
16149 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16150 aligns structures containing the above types differently than
16151 most published application binary interface specifications for the m68k.
16155 Use the pc-relative addressing mode of the 68000 directly, instead of
16156 using a global offset table. At present, this option implies @option{-fpic},
16157 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16158 not presently supported with @option{-mpcrel}, though this could be supported for
16159 68020 and higher processors.
16161 @item -mno-strict-align
16162 @itemx -mstrict-align
16163 @opindex mno-strict-align
16164 @opindex mstrict-align
16165 Do not (do) assume that unaligned memory references are handled by
16169 Generate code that allows the data segment to be located in a different
16170 area of memory from the text segment. This allows for execute-in-place in
16171 an environment without virtual memory management. This option implies
16174 @item -mno-sep-data
16175 Generate code that assumes that the data segment follows the text segment.
16176 This is the default.
16178 @item -mid-shared-library
16179 Generate code that supports shared libraries via the library ID method.
16180 This allows for execute-in-place and shared libraries in an environment
16181 without virtual memory management. This option implies @option{-fPIC}.
16183 @item -mno-id-shared-library
16184 Generate code that doesn't assume ID-based shared libraries are being used.
16185 This is the default.
16187 @item -mshared-library-id=n
16188 Specifies the identification number of the ID-based shared library being
16189 compiled. Specifying a value of 0 generates more compact code; specifying
16190 other values forces the allocation of that number to the current
16191 library, but is no more space- or time-efficient than omitting this option.
16197 When generating position-independent code for ColdFire, generate code
16198 that works if the GOT has more than 8192 entries. This code is
16199 larger and slower than code generated without this option. On M680x0
16200 processors, this option is not needed; @option{-fPIC} suffices.
16202 GCC normally uses a single instruction to load values from the GOT@.
16203 While this is relatively efficient, it only works if the GOT
16204 is smaller than about 64k. Anything larger causes the linker
16205 to report an error such as:
16207 @cindex relocation truncated to fit (ColdFire)
16209 relocation truncated to fit: R_68K_GOT16O foobar
16212 If this happens, you should recompile your code with @option{-mxgot}.
16213 It should then work with very large GOTs. However, code generated with
16214 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16215 the value of a global symbol.
16217 Note that some linkers, including newer versions of the GNU linker,
16218 can create multiple GOTs and sort GOT entries. If you have such a linker,
16219 you should only need to use @option{-mxgot} when compiling a single
16220 object file that accesses more than 8192 GOT entries. Very few do.
16222 These options have no effect unless GCC is generating
16223 position-independent code.
16227 @node MCore Options
16228 @subsection MCore Options
16229 @cindex MCore options
16231 These are the @samp{-m} options defined for the Motorola M*Core
16237 @itemx -mno-hardlit
16239 @opindex mno-hardlit
16240 Inline constants into the code stream if it can be done in two
16241 instructions or less.
16247 Use the divide instruction. (Enabled by default).
16249 @item -mrelax-immediate
16250 @itemx -mno-relax-immediate
16251 @opindex mrelax-immediate
16252 @opindex mno-relax-immediate
16253 Allow arbitrary-sized immediates in bit operations.
16255 @item -mwide-bitfields
16256 @itemx -mno-wide-bitfields
16257 @opindex mwide-bitfields
16258 @opindex mno-wide-bitfields
16259 Always treat bit-fields as @code{int}-sized.
16261 @item -m4byte-functions
16262 @itemx -mno-4byte-functions
16263 @opindex m4byte-functions
16264 @opindex mno-4byte-functions
16265 Force all functions to be aligned to a 4-byte boundary.
16267 @item -mcallgraph-data
16268 @itemx -mno-callgraph-data
16269 @opindex mcallgraph-data
16270 @opindex mno-callgraph-data
16271 Emit callgraph information.
16274 @itemx -mno-slow-bytes
16275 @opindex mslow-bytes
16276 @opindex mno-slow-bytes
16277 Prefer word access when reading byte quantities.
16279 @item -mlittle-endian
16280 @itemx -mbig-endian
16281 @opindex mlittle-endian
16282 @opindex mbig-endian
16283 Generate code for a little-endian target.
16289 Generate code for the 210 processor.
16293 Assume that runtime support has been provided and so omit the
16294 simulator library (@file{libsim.a)} from the linker command line.
16296 @item -mstack-increment=@var{size}
16297 @opindex mstack-increment
16298 Set the maximum amount for a single stack increment operation. Large
16299 values can increase the speed of programs that contain functions
16300 that need a large amount of stack space, but they can also trigger a
16301 segmentation fault if the stack is extended too much. The default
16307 @subsection MeP Options
16308 @cindex MeP options
16314 Enables the @code{abs} instruction, which is the absolute difference
16315 between two registers.
16319 Enables all the optional instructions---average, multiply, divide, bit
16320 operations, leading zero, absolute difference, min/max, clip, and
16326 Enables the @code{ave} instruction, which computes the average of two
16329 @item -mbased=@var{n}
16331 Variables of size @var{n} bytes or smaller are placed in the
16332 @code{.based} section by default. Based variables use the @code{$tp}
16333 register as a base register, and there is a 128-byte limit to the
16334 @code{.based} section.
16338 Enables the bit operation instructions---bit test (@code{btstm}), set
16339 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16340 test-and-set (@code{tas}).
16342 @item -mc=@var{name}
16344 Selects which section constant data is placed in. @var{name} may
16345 be @samp{tiny}, @samp{near}, or @samp{far}.
16349 Enables the @code{clip} instruction. Note that @option{-mclip} is not
16350 useful unless you also provide @option{-mminmax}.
16352 @item -mconfig=@var{name}
16354 Selects one of the built-in core configurations. Each MeP chip has
16355 one or more modules in it; each module has a core CPU and a variety of
16356 coprocessors, optional instructions, and peripherals. The
16357 @code{MeP-Integrator} tool, not part of GCC, provides these
16358 configurations through this option; using this option is the same as
16359 using all the corresponding command-line options. The default
16360 configuration is @samp{default}.
16364 Enables the coprocessor instructions. By default, this is a 32-bit
16365 coprocessor. Note that the coprocessor is normally enabled via the
16366 @option{-mconfig=} option.
16370 Enables the 32-bit coprocessor's instructions.
16374 Enables the 64-bit coprocessor's instructions.
16378 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16382 Causes constant variables to be placed in the @code{.near} section.
16386 Enables the @code{div} and @code{divu} instructions.
16390 Generate big-endian code.
16394 Generate little-endian code.
16396 @item -mio-volatile
16397 @opindex mio-volatile
16398 Tells the compiler that any variable marked with the @code{io}
16399 attribute is to be considered volatile.
16403 Causes variables to be assigned to the @code{.far} section by default.
16407 Enables the @code{leadz} (leading zero) instruction.
16411 Causes variables to be assigned to the @code{.near} section by default.
16415 Enables the @code{min} and @code{max} instructions.
16419 Enables the multiplication and multiply-accumulate instructions.
16423 Disables all the optional instructions enabled by @option{-mall-opts}.
16427 Enables the @code{repeat} and @code{erepeat} instructions, used for
16428 low-overhead looping.
16432 Causes all variables to default to the @code{.tiny} section. Note
16433 that there is a 65536-byte limit to this section. Accesses to these
16434 variables use the @code{%gp} base register.
16438 Enables the saturation instructions. Note that the compiler does not
16439 currently generate these itself, but this option is included for
16440 compatibility with other tools, like @code{as}.
16444 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16448 Link the simulator run-time libraries.
16452 Link the simulator runtime libraries, excluding built-in support
16453 for reset and exception vectors and tables.
16457 Causes all functions to default to the @code{.far} section. Without
16458 this option, functions default to the @code{.near} section.
16460 @item -mtiny=@var{n}
16462 Variables that are @var{n} bytes or smaller are allocated to the
16463 @code{.tiny} section. These variables use the @code{$gp} base
16464 register. The default for this option is 4, but note that there's a
16465 65536-byte limit to the @code{.tiny} section.
16469 @node MicroBlaze Options
16470 @subsection MicroBlaze Options
16471 @cindex MicroBlaze Options
16476 @opindex msoft-float
16477 Use software emulation for floating point (default).
16480 @opindex mhard-float
16481 Use hardware floating-point instructions.
16485 Do not optimize block moves, use @code{memcpy}.
16487 @item -mno-clearbss
16488 @opindex mno-clearbss
16489 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
16491 @item -mcpu=@var{cpu-type}
16493 Use features of, and schedule code for, the given CPU.
16494 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
16495 where @var{X} is a major version, @var{YY} is the minor version, and
16496 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
16497 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
16499 @item -mxl-soft-mul
16500 @opindex mxl-soft-mul
16501 Use software multiply emulation (default).
16503 @item -mxl-soft-div
16504 @opindex mxl-soft-div
16505 Use software emulation for divides (default).
16507 @item -mxl-barrel-shift
16508 @opindex mxl-barrel-shift
16509 Use the hardware barrel shifter.
16511 @item -mxl-pattern-compare
16512 @opindex mxl-pattern-compare
16513 Use pattern compare instructions.
16515 @item -msmall-divides
16516 @opindex msmall-divides
16517 Use table lookup optimization for small signed integer divisions.
16519 @item -mxl-stack-check
16520 @opindex mxl-stack-check
16521 This option is deprecated. Use @option{-fstack-check} instead.
16524 @opindex mxl-gp-opt
16525 Use GP-relative @code{.sdata}/@code{.sbss} sections.
16527 @item -mxl-multiply-high
16528 @opindex mxl-multiply-high
16529 Use multiply high instructions for high part of 32x32 multiply.
16531 @item -mxl-float-convert
16532 @opindex mxl-float-convert
16533 Use hardware floating-point conversion instructions.
16535 @item -mxl-float-sqrt
16536 @opindex mxl-float-sqrt
16537 Use hardware floating-point square root instruction.
16540 @opindex mbig-endian
16541 Generate code for a big-endian target.
16543 @item -mlittle-endian
16544 @opindex mlittle-endian
16545 Generate code for a little-endian target.
16548 @opindex mxl-reorder
16549 Use reorder instructions (swap and byte reversed load/store).
16551 @item -mxl-mode-@var{app-model}
16552 Select application model @var{app-model}. Valid models are
16555 normal executable (default), uses startup code @file{crt0.o}.
16558 for use with Xilinx Microprocessor Debugger (XMD) based
16559 software intrusive debug agent called xmdstub. This uses startup file
16560 @file{crt1.o} and sets the start address of the program to 0x800.
16563 for applications that are loaded using a bootloader.
16564 This model uses startup file @file{crt2.o} which does not contain a processor
16565 reset vector handler. This is suitable for transferring control on a
16566 processor reset to the bootloader rather than the application.
16569 for applications that do not require any of the
16570 MicroBlaze vectors. This option may be useful for applications running
16571 within a monitoring application. This model uses @file{crt3.o} as a startup file.
16574 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
16575 @option{-mxl-mode-@var{app-model}}.
16580 @subsection MIPS Options
16581 @cindex MIPS options
16587 Generate big-endian code.
16591 Generate little-endian code. This is the default for @samp{mips*el-*-*}
16594 @item -march=@var{arch}
16596 Generate code that runs on @var{arch}, which can be the name of a
16597 generic MIPS ISA, or the name of a particular processor.
16599 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
16600 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
16601 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
16602 @samp{mips64r5} and @samp{mips64r6}.
16603 The processor names are:
16604 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
16605 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
16606 @samp{5kc}, @samp{5kf},
16608 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
16609 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
16610 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
16611 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
16612 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
16613 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
16615 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
16616 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
16619 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
16620 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
16621 @samp{rm7000}, @samp{rm9000},
16622 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
16625 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
16626 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
16627 @samp{xlr} and @samp{xlp}.
16628 The special value @samp{from-abi} selects the
16629 most compatible architecture for the selected ABI (that is,
16630 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
16632 The native Linux/GNU toolchain also supports the value @samp{native},
16633 which selects the best architecture option for the host processor.
16634 @option{-march=native} has no effect if GCC does not recognize
16637 In processor names, a final @samp{000} can be abbreviated as @samp{k}
16638 (for example, @option{-march=r2k}). Prefixes are optional, and
16639 @samp{vr} may be written @samp{r}.
16641 Names of the form @samp{@var{n}f2_1} refer to processors with
16642 FPUs clocked at half the rate of the core, names of the form
16643 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
16644 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
16645 processors with FPUs clocked a ratio of 3:2 with respect to the core.
16646 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
16647 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
16648 accepted as synonyms for @samp{@var{n}f1_1}.
16650 GCC defines two macros based on the value of this option. The first
16651 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
16652 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
16653 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
16654 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
16655 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
16657 Note that the @code{_MIPS_ARCH} macro uses the processor names given
16658 above. In other words, it has the full prefix and does not
16659 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
16660 the macro names the resolved architecture (either @code{"mips1"} or
16661 @code{"mips3"}). It names the default architecture when no
16662 @option{-march} option is given.
16664 @item -mtune=@var{arch}
16666 Optimize for @var{arch}. Among other things, this option controls
16667 the way instructions are scheduled, and the perceived cost of arithmetic
16668 operations. The list of @var{arch} values is the same as for
16671 When this option is not used, GCC optimizes for the processor
16672 specified by @option{-march}. By using @option{-march} and
16673 @option{-mtune} together, it is possible to generate code that
16674 runs on a family of processors, but optimize the code for one
16675 particular member of that family.
16677 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
16678 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
16679 @option{-march} ones described above.
16683 Equivalent to @option{-march=mips1}.
16687 Equivalent to @option{-march=mips2}.
16691 Equivalent to @option{-march=mips3}.
16695 Equivalent to @option{-march=mips4}.
16699 Equivalent to @option{-march=mips32}.
16703 Equivalent to @option{-march=mips32r3}.
16707 Equivalent to @option{-march=mips32r5}.
16711 Equivalent to @option{-march=mips32r6}.
16715 Equivalent to @option{-march=mips64}.
16719 Equivalent to @option{-march=mips64r2}.
16723 Equivalent to @option{-march=mips64r3}.
16727 Equivalent to @option{-march=mips64r5}.
16731 Equivalent to @option{-march=mips64r6}.
16736 @opindex mno-mips16
16737 Generate (do not generate) MIPS16 code. If GCC is targeting a
16738 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
16740 MIPS16 code generation can also be controlled on a per-function basis
16741 by means of @code{mips16} and @code{nomips16} attributes.
16742 @xref{Function Attributes}, for more information.
16744 @item -mflip-mips16
16745 @opindex mflip-mips16
16746 Generate MIPS16 code on alternating functions. This option is provided
16747 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
16748 not intended for ordinary use in compiling user code.
16750 @item -minterlink-compressed
16751 @item -mno-interlink-compressed
16752 @opindex minterlink-compressed
16753 @opindex mno-interlink-compressed
16754 Require (do not require) that code using the standard (uncompressed) MIPS ISA
16755 be link-compatible with MIPS16 and microMIPS code, and vice versa.
16757 For example, code using the standard ISA encoding cannot jump directly
16758 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
16759 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
16760 knows that the target of the jump is not compressed.
16762 @item -minterlink-mips16
16763 @itemx -mno-interlink-mips16
16764 @opindex minterlink-mips16
16765 @opindex mno-interlink-mips16
16766 Aliases of @option{-minterlink-compressed} and
16767 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
16768 and are retained for backwards compatibility.
16780 Generate code for the given ABI@.
16782 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
16783 generates 64-bit code when you select a 64-bit architecture, but you
16784 can use @option{-mgp32} to get 32-bit code instead.
16786 For information about the O64 ABI, see
16787 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
16789 GCC supports a variant of the o32 ABI in which floating-point registers
16790 are 64 rather than 32 bits wide. You can select this combination with
16791 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
16792 and @code{mfhc1} instructions and is therefore only supported for
16793 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
16795 The register assignments for arguments and return values remain the
16796 same, but each scalar value is passed in a single 64-bit register
16797 rather than a pair of 32-bit registers. For example, scalar
16798 floating-point values are returned in @samp{$f0} only, not a
16799 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
16800 remains the same in that the even-numbered double-precision registers
16803 Two additional variants of the o32 ABI are supported to enable
16804 a transition from 32-bit to 64-bit registers. These are FPXX
16805 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
16806 The FPXX extension mandates that all code must execute correctly
16807 when run using 32-bit or 64-bit registers. The code can be interlinked
16808 with either FP32 or FP64, but not both.
16809 The FP64A extension is similar to the FP64 extension but forbids the
16810 use of odd-numbered single-precision registers. This can be used
16811 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
16812 processors and allows both FP32 and FP64A code to interlink and
16813 run in the same process without changing FPU modes.
16816 @itemx -mno-abicalls
16818 @opindex mno-abicalls
16819 Generate (do not generate) code that is suitable for SVR4-style
16820 dynamic objects. @option{-mabicalls} is the default for SVR4-based
16825 Generate (do not generate) code that is fully position-independent,
16826 and that can therefore be linked into shared libraries. This option
16827 only affects @option{-mabicalls}.
16829 All @option{-mabicalls} code has traditionally been position-independent,
16830 regardless of options like @option{-fPIC} and @option{-fpic}. However,
16831 as an extension, the GNU toolchain allows executables to use absolute
16832 accesses for locally-binding symbols. It can also use shorter GP
16833 initialization sequences and generate direct calls to locally-defined
16834 functions. This mode is selected by @option{-mno-shared}.
16836 @option{-mno-shared} depends on binutils 2.16 or higher and generates
16837 objects that can only be linked by the GNU linker. However, the option
16838 does not affect the ABI of the final executable; it only affects the ABI
16839 of relocatable objects. Using @option{-mno-shared} generally makes
16840 executables both smaller and quicker.
16842 @option{-mshared} is the default.
16848 Assume (do not assume) that the static and dynamic linkers
16849 support PLTs and copy relocations. This option only affects
16850 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
16851 has no effect without @option{-msym32}.
16853 You can make @option{-mplt} the default by configuring
16854 GCC with @option{--with-mips-plt}. The default is
16855 @option{-mno-plt} otherwise.
16861 Lift (do not lift) the usual restrictions on the size of the global
16864 GCC normally uses a single instruction to load values from the GOT@.
16865 While this is relatively efficient, it only works if the GOT
16866 is smaller than about 64k. Anything larger causes the linker
16867 to report an error such as:
16869 @cindex relocation truncated to fit (MIPS)
16871 relocation truncated to fit: R_MIPS_GOT16 foobar
16874 If this happens, you should recompile your code with @option{-mxgot}.
16875 This works with very large GOTs, although the code is also
16876 less efficient, since it takes three instructions to fetch the
16877 value of a global symbol.
16879 Note that some linkers can create multiple GOTs. If you have such a
16880 linker, you should only need to use @option{-mxgot} when a single object
16881 file accesses more than 64k's worth of GOT entries. Very few do.
16883 These options have no effect unless GCC is generating position
16888 Assume that general-purpose registers are 32 bits wide.
16892 Assume that general-purpose registers are 64 bits wide.
16896 Assume that floating-point registers are 32 bits wide.
16900 Assume that floating-point registers are 64 bits wide.
16904 Do not assume the width of floating-point registers.
16907 @opindex mhard-float
16908 Use floating-point coprocessor instructions.
16911 @opindex msoft-float
16912 Do not use floating-point coprocessor instructions. Implement
16913 floating-point calculations using library calls instead.
16917 Equivalent to @option{-msoft-float}, but additionally asserts that the
16918 program being compiled does not perform any floating-point operations.
16919 This option is presently supported only by some bare-metal MIPS
16920 configurations, where it may select a special set of libraries
16921 that lack all floating-point support (including, for example, the
16922 floating-point @code{printf} formats).
16923 If code compiled with @option{-mno-float} accidentally contains
16924 floating-point operations, it is likely to suffer a link-time
16925 or run-time failure.
16927 @item -msingle-float
16928 @opindex msingle-float
16929 Assume that the floating-point coprocessor only supports single-precision
16932 @item -mdouble-float
16933 @opindex mdouble-float
16934 Assume that the floating-point coprocessor supports double-precision
16935 operations. This is the default.
16938 @itemx -mno-odd-spreg
16939 @opindex modd-spreg
16940 @opindex mno-odd-spreg
16941 Enable the use of odd-numbered single-precision floating-point registers
16942 for the o32 ABI. This is the default for processors that are known to
16943 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
16947 @itemx -mabs=legacy
16949 @opindex mabs=legacy
16950 These options control the treatment of the special not-a-number (NaN)
16951 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
16952 @code{neg.@i{fmt}} machine instructions.
16954 By default or when the @option{-mabs=legacy} is used the legacy
16955 treatment is selected. In this case these instructions are considered
16956 arithmetic and avoided where correct operation is required and the
16957 input operand might be a NaN. A longer sequence of instructions that
16958 manipulate the sign bit of floating-point datum manually is used
16959 instead unless the @option{-ffinite-math-only} option has also been
16962 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
16963 this case these instructions are considered non-arithmetic and therefore
16964 operating correctly in all cases, including in particular where the
16965 input operand is a NaN. These instructions are therefore always used
16966 for the respective operations.
16969 @itemx -mnan=legacy
16971 @opindex mnan=legacy
16972 These options control the encoding of the special not-a-number (NaN)
16973 IEEE 754 floating-point data.
16975 The @option{-mnan=legacy} option selects the legacy encoding. In this
16976 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
16977 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
16978 by the first bit of their trailing significand field being 1.
16980 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
16981 this case qNaNs are denoted by the first bit of their trailing
16982 significand field being 1, whereas sNaNs are denoted by the first bit of
16983 their trailing significand field being 0.
16985 The default is @option{-mnan=legacy} unless GCC has been configured with
16986 @option{--with-nan=2008}.
16992 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
16993 implement atomic memory built-in functions. When neither option is
16994 specified, GCC uses the instructions if the target architecture
16997 @option{-mllsc} is useful if the runtime environment can emulate the
16998 instructions and @option{-mno-llsc} can be useful when compiling for
16999 nonstandard ISAs. You can make either option the default by
17000 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17001 respectively. @option{--with-llsc} is the default for some
17002 configurations; see the installation documentation for details.
17008 Use (do not use) revision 1 of the MIPS DSP ASE@.
17009 @xref{MIPS DSP Built-in Functions}. This option defines the
17010 preprocessor macro @code{__mips_dsp}. It also defines
17011 @code{__mips_dsp_rev} to 1.
17017 Use (do not use) revision 2 of the MIPS DSP ASE@.
17018 @xref{MIPS DSP Built-in Functions}. This option defines the
17019 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17020 It also defines @code{__mips_dsp_rev} to 2.
17023 @itemx -mno-smartmips
17024 @opindex msmartmips
17025 @opindex mno-smartmips
17026 Use (do not use) the MIPS SmartMIPS ASE.
17028 @item -mpaired-single
17029 @itemx -mno-paired-single
17030 @opindex mpaired-single
17031 @opindex mno-paired-single
17032 Use (do not use) paired-single floating-point instructions.
17033 @xref{MIPS Paired-Single Support}. This option requires
17034 hardware floating-point support to be enabled.
17040 Use (do not use) MIPS Digital Media Extension instructions.
17041 This option can only be used when generating 64-bit code and requires
17042 hardware floating-point support to be enabled.
17047 @opindex mno-mips3d
17048 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17049 The option @option{-mips3d} implies @option{-mpaired-single}.
17052 @itemx -mno-micromips
17053 @opindex mmicromips
17054 @opindex mno-mmicromips
17055 Generate (do not generate) microMIPS code.
17057 MicroMIPS code generation can also be controlled on a per-function basis
17058 by means of @code{micromips} and @code{nomicromips} attributes.
17059 @xref{Function Attributes}, for more information.
17065 Use (do not use) MT Multithreading instructions.
17071 Use (do not use) the MIPS MCU ASE instructions.
17077 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17083 Use (do not use) the MIPS Virtualization Application Specific instructions.
17089 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17093 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17094 an explanation of the default and the way that the pointer size is
17099 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17101 The default size of @code{int}s, @code{long}s and pointers depends on
17102 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17103 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17104 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17105 or the same size as integer registers, whichever is smaller.
17111 Assume (do not assume) that all symbols have 32-bit values, regardless
17112 of the selected ABI@. This option is useful in combination with
17113 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17114 to generate shorter and faster references to symbolic addresses.
17118 Put definitions of externally-visible data in a small data section
17119 if that data is no bigger than @var{num} bytes. GCC can then generate
17120 more efficient accesses to the data; see @option{-mgpopt} for details.
17122 The default @option{-G} option depends on the configuration.
17124 @item -mlocal-sdata
17125 @itemx -mno-local-sdata
17126 @opindex mlocal-sdata
17127 @opindex mno-local-sdata
17128 Extend (do not extend) the @option{-G} behavior to local data too,
17129 such as to static variables in C@. @option{-mlocal-sdata} is the
17130 default for all configurations.
17132 If the linker complains that an application is using too much small data,
17133 you might want to try rebuilding the less performance-critical parts with
17134 @option{-mno-local-sdata}. You might also want to build large
17135 libraries with @option{-mno-local-sdata}, so that the libraries leave
17136 more room for the main program.
17138 @item -mextern-sdata
17139 @itemx -mno-extern-sdata
17140 @opindex mextern-sdata
17141 @opindex mno-extern-sdata
17142 Assume (do not assume) that externally-defined data is in
17143 a small data section if the size of that data is within the @option{-G} limit.
17144 @option{-mextern-sdata} is the default for all configurations.
17146 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17147 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17148 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17149 is placed in a small data section. If @var{Var} is defined by another
17150 module, you must either compile that module with a high-enough
17151 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17152 definition. If @var{Var} is common, you must link the application
17153 with a high-enough @option{-G} setting.
17155 The easiest way of satisfying these restrictions is to compile
17156 and link every module with the same @option{-G} option. However,
17157 you may wish to build a library that supports several different
17158 small data limits. You can do this by compiling the library with
17159 the highest supported @option{-G} setting and additionally using
17160 @option{-mno-extern-sdata} to stop the library from making assumptions
17161 about externally-defined data.
17167 Use (do not use) GP-relative accesses for symbols that are known to be
17168 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17169 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17172 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17173 might not hold the value of @code{_gp}. For example, if the code is
17174 part of a library that might be used in a boot monitor, programs that
17175 call boot monitor routines pass an unknown value in @code{$gp}.
17176 (In such situations, the boot monitor itself is usually compiled
17177 with @option{-G0}.)
17179 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17180 @option{-mno-extern-sdata}.
17182 @item -membedded-data
17183 @itemx -mno-embedded-data
17184 @opindex membedded-data
17185 @opindex mno-embedded-data
17186 Allocate variables to the read-only data section first if possible, then
17187 next in the small data section if possible, otherwise in data. This gives
17188 slightly slower code than the default, but reduces the amount of RAM required
17189 when executing, and thus may be preferred for some embedded systems.
17191 @item -muninit-const-in-rodata
17192 @itemx -mno-uninit-const-in-rodata
17193 @opindex muninit-const-in-rodata
17194 @opindex mno-uninit-const-in-rodata
17195 Put uninitialized @code{const} variables in the read-only data section.
17196 This option is only meaningful in conjunction with @option{-membedded-data}.
17198 @item -mcode-readable=@var{setting}
17199 @opindex mcode-readable
17200 Specify whether GCC may generate code that reads from executable sections.
17201 There are three possible settings:
17204 @item -mcode-readable=yes
17205 Instructions may freely access executable sections. This is the
17208 @item -mcode-readable=pcrel
17209 MIPS16 PC-relative load instructions can access executable sections,
17210 but other instructions must not do so. This option is useful on 4KSc
17211 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17212 It is also useful on processors that can be configured to have a dual
17213 instruction/data SRAM interface and that, like the M4K, automatically
17214 redirect PC-relative loads to the instruction RAM.
17216 @item -mcode-readable=no
17217 Instructions must not access executable sections. This option can be
17218 useful on targets that are configured to have a dual instruction/data
17219 SRAM interface but that (unlike the M4K) do not automatically redirect
17220 PC-relative loads to the instruction RAM.
17223 @item -msplit-addresses
17224 @itemx -mno-split-addresses
17225 @opindex msplit-addresses
17226 @opindex mno-split-addresses
17227 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17228 relocation operators. This option has been superseded by
17229 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17231 @item -mexplicit-relocs
17232 @itemx -mno-explicit-relocs
17233 @opindex mexplicit-relocs
17234 @opindex mno-explicit-relocs
17235 Use (do not use) assembler relocation operators when dealing with symbolic
17236 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17237 is to use assembler macros instead.
17239 @option{-mexplicit-relocs} is the default if GCC was configured
17240 to use an assembler that supports relocation operators.
17242 @item -mcheck-zero-division
17243 @itemx -mno-check-zero-division
17244 @opindex mcheck-zero-division
17245 @opindex mno-check-zero-division
17246 Trap (do not trap) on integer division by zero.
17248 The default is @option{-mcheck-zero-division}.
17250 @item -mdivide-traps
17251 @itemx -mdivide-breaks
17252 @opindex mdivide-traps
17253 @opindex mdivide-breaks
17254 MIPS systems check for division by zero by generating either a
17255 conditional trap or a break instruction. Using traps results in
17256 smaller code, but is only supported on MIPS II and later. Also, some
17257 versions of the Linux kernel have a bug that prevents trap from
17258 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17259 allow conditional traps on architectures that support them and
17260 @option{-mdivide-breaks} to force the use of breaks.
17262 The default is usually @option{-mdivide-traps}, but this can be
17263 overridden at configure time using @option{--with-divide=breaks}.
17264 Divide-by-zero checks can be completely disabled using
17265 @option{-mno-check-zero-division}.
17270 @opindex mno-memcpy
17271 Force (do not force) the use of @code{memcpy} for non-trivial block
17272 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17273 most constant-sized copies.
17276 @itemx -mno-long-calls
17277 @opindex mlong-calls
17278 @opindex mno-long-calls
17279 Disable (do not disable) use of the @code{jal} instruction. Calling
17280 functions using @code{jal} is more efficient but requires the caller
17281 and callee to be in the same 256 megabyte segment.
17283 This option has no effect on abicalls code. The default is
17284 @option{-mno-long-calls}.
17290 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17291 instructions, as provided by the R4650 ISA@.
17297 Enable (disable) use of the @code{madd} and @code{msub} integer
17298 instructions. The default is @option{-mimadd} on architectures
17299 that support @code{madd} and @code{msub} except for the 74k
17300 architecture where it was found to generate slower code.
17303 @itemx -mno-fused-madd
17304 @opindex mfused-madd
17305 @opindex mno-fused-madd
17306 Enable (disable) use of the floating-point multiply-accumulate
17307 instructions, when they are available. The default is
17308 @option{-mfused-madd}.
17310 On the R8000 CPU when multiply-accumulate instructions are used,
17311 the intermediate product is calculated to infinite precision
17312 and is not subject to the FCSR Flush to Zero bit. This may be
17313 undesirable in some circumstances. On other processors the result
17314 is numerically identical to the equivalent computation using
17315 separate multiply, add, subtract and negate instructions.
17319 Tell the MIPS assembler to not run its preprocessor over user
17320 assembler files (with a @samp{.s} suffix) when assembling them.
17325 @opindex mno-fix-24k
17326 Work around the 24K E48 (lost data on stores during refill) errata.
17327 The workarounds are implemented by the assembler rather than by GCC@.
17330 @itemx -mno-fix-r4000
17331 @opindex mfix-r4000
17332 @opindex mno-fix-r4000
17333 Work around certain R4000 CPU errata:
17336 A double-word or a variable shift may give an incorrect result if executed
17337 immediately after starting an integer division.
17339 A double-word or a variable shift may give an incorrect result if executed
17340 while an integer multiplication is in progress.
17342 An integer division may give an incorrect result if started in a delay slot
17343 of a taken branch or a jump.
17347 @itemx -mno-fix-r4400
17348 @opindex mfix-r4400
17349 @opindex mno-fix-r4400
17350 Work around certain R4400 CPU errata:
17353 A double-word or a variable shift may give an incorrect result if executed
17354 immediately after starting an integer division.
17358 @itemx -mno-fix-r10000
17359 @opindex mfix-r10000
17360 @opindex mno-fix-r10000
17361 Work around certain R10000 errata:
17364 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17365 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17368 This option can only be used if the target architecture supports
17369 branch-likely instructions. @option{-mfix-r10000} is the default when
17370 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17374 @itemx -mno-fix-rm7000
17375 @opindex mfix-rm7000
17376 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17377 workarounds are implemented by the assembler rather than by GCC@.
17380 @itemx -mno-fix-vr4120
17381 @opindex mfix-vr4120
17382 Work around certain VR4120 errata:
17385 @code{dmultu} does not always produce the correct result.
17387 @code{div} and @code{ddiv} do not always produce the correct result if one
17388 of the operands is negative.
17390 The workarounds for the division errata rely on special functions in
17391 @file{libgcc.a}. At present, these functions are only provided by
17392 the @code{mips64vr*-elf} configurations.
17394 Other VR4120 errata require a NOP to be inserted between certain pairs of
17395 instructions. These errata are handled by the assembler, not by GCC itself.
17398 @opindex mfix-vr4130
17399 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17400 workarounds are implemented by the assembler rather than by GCC,
17401 although GCC avoids using @code{mflo} and @code{mfhi} if the
17402 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17403 instructions are available instead.
17406 @itemx -mno-fix-sb1
17408 Work around certain SB-1 CPU core errata.
17409 (This flag currently works around the SB-1 revision 2
17410 ``F1'' and ``F2'' floating-point errata.)
17412 @item -mr10k-cache-barrier=@var{setting}
17413 @opindex mr10k-cache-barrier
17414 Specify whether GCC should insert cache barriers to avoid the
17415 side-effects of speculation on R10K processors.
17417 In common with many processors, the R10K tries to predict the outcome
17418 of a conditional branch and speculatively executes instructions from
17419 the ``taken'' branch. It later aborts these instructions if the
17420 predicted outcome is wrong. However, on the R10K, even aborted
17421 instructions can have side effects.
17423 This problem only affects kernel stores and, depending on the system,
17424 kernel loads. As an example, a speculatively-executed store may load
17425 the target memory into cache and mark the cache line as dirty, even if
17426 the store itself is later aborted. If a DMA operation writes to the
17427 same area of memory before the ``dirty'' line is flushed, the cached
17428 data overwrites the DMA-ed data. See the R10K processor manual
17429 for a full description, including other potential problems.
17431 One workaround is to insert cache barrier instructions before every memory
17432 access that might be speculatively executed and that might have side
17433 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17434 controls GCC's implementation of this workaround. It assumes that
17435 aborted accesses to any byte in the following regions does not have
17440 the memory occupied by the current function's stack frame;
17443 the memory occupied by an incoming stack argument;
17446 the memory occupied by an object with a link-time-constant address.
17449 It is the kernel's responsibility to ensure that speculative
17450 accesses to these regions are indeed safe.
17452 If the input program contains a function declaration such as:
17458 then the implementation of @code{foo} must allow @code{j foo} and
17459 @code{jal foo} to be executed speculatively. GCC honors this
17460 restriction for functions it compiles itself. It expects non-GCC
17461 functions (such as hand-written assembly code) to do the same.
17463 The option has three forms:
17466 @item -mr10k-cache-barrier=load-store
17467 Insert a cache barrier before a load or store that might be
17468 speculatively executed and that might have side effects even
17471 @item -mr10k-cache-barrier=store
17472 Insert a cache barrier before a store that might be speculatively
17473 executed and that might have side effects even if aborted.
17475 @item -mr10k-cache-barrier=none
17476 Disable the insertion of cache barriers. This is the default setting.
17479 @item -mflush-func=@var{func}
17480 @itemx -mno-flush-func
17481 @opindex mflush-func
17482 Specifies the function to call to flush the I and D caches, or to not
17483 call any such function. If called, the function must take the same
17484 arguments as the common @code{_flush_func}, that is, the address of the
17485 memory range for which the cache is being flushed, the size of the
17486 memory range, and the number 3 (to flush both caches). The default
17487 depends on the target GCC was configured for, but commonly is either
17488 @code{_flush_func} or @code{__cpu_flush}.
17490 @item mbranch-cost=@var{num}
17491 @opindex mbranch-cost
17492 Set the cost of branches to roughly @var{num} ``simple'' instructions.
17493 This cost is only a heuristic and is not guaranteed to produce
17494 consistent results across releases. A zero cost redundantly selects
17495 the default, which is based on the @option{-mtune} setting.
17497 @item -mbranch-likely
17498 @itemx -mno-branch-likely
17499 @opindex mbranch-likely
17500 @opindex mno-branch-likely
17501 Enable or disable use of Branch Likely instructions, regardless of the
17502 default for the selected architecture. By default, Branch Likely
17503 instructions may be generated if they are supported by the selected
17504 architecture. An exception is for the MIPS32 and MIPS64 architectures
17505 and processors that implement those architectures; for those, Branch
17506 Likely instructions are not be generated by default because the MIPS32
17507 and MIPS64 architectures specifically deprecate their use.
17509 @item -mfp-exceptions
17510 @itemx -mno-fp-exceptions
17511 @opindex mfp-exceptions
17512 Specifies whether FP exceptions are enabled. This affects how
17513 FP instructions are scheduled for some processors.
17514 The default is that FP exceptions are
17517 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
17518 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
17521 @item -mvr4130-align
17522 @itemx -mno-vr4130-align
17523 @opindex mvr4130-align
17524 The VR4130 pipeline is two-way superscalar, but can only issue two
17525 instructions together if the first one is 8-byte aligned. When this
17526 option is enabled, GCC aligns pairs of instructions that it
17527 thinks should execute in parallel.
17529 This option only has an effect when optimizing for the VR4130.
17530 It normally makes code faster, but at the expense of making it bigger.
17531 It is enabled by default at optimization level @option{-O3}.
17536 Enable (disable) generation of @code{synci} instructions on
17537 architectures that support it. The @code{synci} instructions (if
17538 enabled) are generated when @code{__builtin___clear_cache} is
17541 This option defaults to @option{-mno-synci}, but the default can be
17542 overridden by configuring GCC with @option{--with-synci}.
17544 When compiling code for single processor systems, it is generally safe
17545 to use @code{synci}. However, on many multi-core (SMP) systems, it
17546 does not invalidate the instruction caches on all cores and may lead
17547 to undefined behavior.
17549 @item -mrelax-pic-calls
17550 @itemx -mno-relax-pic-calls
17551 @opindex mrelax-pic-calls
17552 Try to turn PIC calls that are normally dispatched via register
17553 @code{$25} into direct calls. This is only possible if the linker can
17554 resolve the destination at link-time and if the destination is within
17555 range for a direct call.
17557 @option{-mrelax-pic-calls} is the default if GCC was configured to use
17558 an assembler and a linker that support the @code{.reloc} assembly
17559 directive and @option{-mexplicit-relocs} is in effect. With
17560 @option{-mno-explicit-relocs}, this optimization can be performed by the
17561 assembler and the linker alone without help from the compiler.
17563 @item -mmcount-ra-address
17564 @itemx -mno-mcount-ra-address
17565 @opindex mmcount-ra-address
17566 @opindex mno-mcount-ra-address
17567 Emit (do not emit) code that allows @code{_mcount} to modify the
17568 calling function's return address. When enabled, this option extends
17569 the usual @code{_mcount} interface with a new @var{ra-address}
17570 parameter, which has type @code{intptr_t *} and is passed in register
17571 @code{$12}. @code{_mcount} can then modify the return address by
17572 doing both of the following:
17575 Returning the new address in register @code{$31}.
17577 Storing the new address in @code{*@var{ra-address}},
17578 if @var{ra-address} is nonnull.
17581 The default is @option{-mno-mcount-ra-address}.
17586 @subsection MMIX Options
17587 @cindex MMIX Options
17589 These options are defined for the MMIX:
17593 @itemx -mno-libfuncs
17595 @opindex mno-libfuncs
17596 Specify that intrinsic library functions are being compiled, passing all
17597 values in registers, no matter the size.
17600 @itemx -mno-epsilon
17602 @opindex mno-epsilon
17603 Generate floating-point comparison instructions that compare with respect
17604 to the @code{rE} epsilon register.
17606 @item -mabi=mmixware
17608 @opindex mabi=mmixware
17610 Generate code that passes function parameters and return values that (in
17611 the called function) are seen as registers @code{$0} and up, as opposed to
17612 the GNU ABI which uses global registers @code{$231} and up.
17614 @item -mzero-extend
17615 @itemx -mno-zero-extend
17616 @opindex mzero-extend
17617 @opindex mno-zero-extend
17618 When reading data from memory in sizes shorter than 64 bits, use (do not
17619 use) zero-extending load instructions by default, rather than
17620 sign-extending ones.
17623 @itemx -mno-knuthdiv
17625 @opindex mno-knuthdiv
17626 Make the result of a division yielding a remainder have the same sign as
17627 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
17628 remainder follows the sign of the dividend. Both methods are
17629 arithmetically valid, the latter being almost exclusively used.
17631 @item -mtoplevel-symbols
17632 @itemx -mno-toplevel-symbols
17633 @opindex mtoplevel-symbols
17634 @opindex mno-toplevel-symbols
17635 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
17636 code can be used with the @code{PREFIX} assembly directive.
17640 Generate an executable in the ELF format, rather than the default
17641 @samp{mmo} format used by the @command{mmix} simulator.
17643 @item -mbranch-predict
17644 @itemx -mno-branch-predict
17645 @opindex mbranch-predict
17646 @opindex mno-branch-predict
17647 Use (do not use) the probable-branch instructions, when static branch
17648 prediction indicates a probable branch.
17650 @item -mbase-addresses
17651 @itemx -mno-base-addresses
17652 @opindex mbase-addresses
17653 @opindex mno-base-addresses
17654 Generate (do not generate) code that uses @emph{base addresses}. Using a
17655 base address automatically generates a request (handled by the assembler
17656 and the linker) for a constant to be set up in a global register. The
17657 register is used for one or more base address requests within the range 0
17658 to 255 from the value held in the register. The generally leads to short
17659 and fast code, but the number of different data items that can be
17660 addressed is limited. This means that a program that uses lots of static
17661 data may require @option{-mno-base-addresses}.
17663 @item -msingle-exit
17664 @itemx -mno-single-exit
17665 @opindex msingle-exit
17666 @opindex mno-single-exit
17667 Force (do not force) generated code to have a single exit point in each
17671 @node MN10300 Options
17672 @subsection MN10300 Options
17673 @cindex MN10300 options
17675 These @option{-m} options are defined for Matsushita MN10300 architectures:
17680 Generate code to avoid bugs in the multiply instructions for the MN10300
17681 processors. This is the default.
17683 @item -mno-mult-bug
17684 @opindex mno-mult-bug
17685 Do not generate code to avoid bugs in the multiply instructions for the
17686 MN10300 processors.
17690 Generate code using features specific to the AM33 processor.
17694 Do not generate code using features specific to the AM33 processor. This
17699 Generate code using features specific to the AM33/2.0 processor.
17703 Generate code using features specific to the AM34 processor.
17705 @item -mtune=@var{cpu-type}
17707 Use the timing characteristics of the indicated CPU type when
17708 scheduling instructions. This does not change the targeted processor
17709 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
17710 @samp{am33-2} or @samp{am34}.
17712 @item -mreturn-pointer-on-d0
17713 @opindex mreturn-pointer-on-d0
17714 When generating a function that returns a pointer, return the pointer
17715 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
17716 only in @code{a0}, and attempts to call such functions without a prototype
17717 result in errors. Note that this option is on by default; use
17718 @option{-mno-return-pointer-on-d0} to disable it.
17722 Do not link in the C run-time initialization object file.
17726 Indicate to the linker that it should perform a relaxation optimization pass
17727 to shorten branches, calls and absolute memory addresses. This option only
17728 has an effect when used on the command line for the final link step.
17730 This option makes symbolic debugging impossible.
17734 Allow the compiler to generate @emph{Long Instruction Word}
17735 instructions if the target is the @samp{AM33} or later. This is the
17736 default. This option defines the preprocessor macro @code{__LIW__}.
17740 Do not allow the compiler to generate @emph{Long Instruction Word}
17741 instructions. This option defines the preprocessor macro
17746 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
17747 instructions if the target is the @samp{AM33} or later. This is the
17748 default. This option defines the preprocessor macro @code{__SETLB__}.
17752 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
17753 instructions. This option defines the preprocessor macro
17754 @code{__NO_SETLB__}.
17758 @node Moxie Options
17759 @subsection Moxie Options
17760 @cindex Moxie Options
17766 Generate big-endian code. This is the default for @samp{moxie-*-*}
17771 Generate little-endian code.
17775 Generate mul.x and umul.x instructions. This is the default for
17776 @samp{moxiebox-*-*} configurations.
17780 Do not link in the C run-time initialization object file.
17784 @node MSP430 Options
17785 @subsection MSP430 Options
17786 @cindex MSP430 Options
17788 These options are defined for the MSP430:
17794 Force assembly output to always use hex constants. Normally such
17795 constants are signed decimals, but this option is available for
17796 testsuite and/or aesthetic purposes.
17800 Select the MCU to target. This is used to create a C preprocessor
17801 symbol based upon the MCU name, converted to upper case and pre- and
17802 post-fixed with @samp{__}. This in turn is used by the
17803 @file{msp430.h} header file to select an MCU-specific supplementary
17806 The option also sets the ISA to use. If the MCU name is one that is
17807 known to only support the 430 ISA then that is selected, otherwise the
17808 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
17809 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
17810 name selects the 430X ISA.
17812 In addition an MCU-specific linker script is added to the linker
17813 command line. The script's name is the name of the MCU with
17814 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
17815 command line defines the C preprocessor symbol @code{__XXX__} and
17816 cause the linker to search for a script called @file{xxx.ld}.
17818 This option is also passed on to the assembler.
17822 Specifies the ISA to use. Accepted values are @samp{msp430},
17823 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
17824 @option{-mmcu=} option should be used to select the ISA.
17828 Link to the simulator runtime libraries and linker script. Overrides
17829 any scripts that would be selected by the @option{-mmcu=} option.
17833 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
17837 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
17841 This option is passed to the assembler and linker, and allows the
17842 linker to perform certain optimizations that cannot be done until
17847 Describes the type of hardware multiply supported by the target.
17848 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
17849 for the original 16-bit-only multiply supported by early MCUs.
17850 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
17851 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
17852 A value of @samp{auto} can also be given. This tells GCC to deduce
17853 the hardware multiply support based upon the MCU name provided by the
17854 @option{-mmcu} option. If no @option{-mmcu} option is specified then
17855 @samp{32bit} hardware multiply support is assumed. @samp{auto} is the
17858 Hardware multiplies are normally performed by calling a library
17859 routine. This saves space in the generated code. When compiling at
17860 @option{-O3} or higher however the hardware multiplier is invoked
17861 inline. This makes for bigger, but faster code.
17863 The hardware multiply routines disable interrupts whilst running and
17864 restore the previous interrupt state when they finish. This makes
17865 them safe to use inside interrupt handlers as well as in normal code.
17869 Enable the use of a minimum runtime environment - no static
17870 initializers or constructors. This is intended for memory-constrained
17871 devices. The compiler includes special symbols in some objects
17872 that tell the linker and runtime which code fragments are required.
17876 @node NDS32 Options
17877 @subsection NDS32 Options
17878 @cindex NDS32 Options
17880 These options are defined for NDS32 implementations:
17885 @opindex mbig-endian
17886 Generate code in big-endian mode.
17888 @item -mlittle-endian
17889 @opindex mlittle-endian
17890 Generate code in little-endian mode.
17892 @item -mreduced-regs
17893 @opindex mreduced-regs
17894 Use reduced-set registers for register allocation.
17897 @opindex mfull-regs
17898 Use full-set registers for register allocation.
17902 Generate conditional move instructions.
17906 Do not generate conditional move instructions.
17910 Generate performance extension instructions.
17912 @item -mno-perf-ext
17913 @opindex mno-perf-ext
17914 Do not generate performance extension instructions.
17918 Generate v3 push25/pop25 instructions.
17921 @opindex mno-v3push
17922 Do not generate v3 push25/pop25 instructions.
17926 Generate 16-bit instructions.
17929 @opindex mno-16-bit
17930 Do not generate 16-bit instructions.
17932 @item -misr-vector-size=@var{num}
17933 @opindex misr-vector-size
17934 Specify the size of each interrupt vector, which must be 4 or 16.
17936 @item -mcache-block-size=@var{num}
17937 @opindex mcache-block-size
17938 Specify the size of each cache block,
17939 which must be a power of 2 between 4 and 512.
17941 @item -march=@var{arch}
17943 Specify the name of the target architecture.
17945 @item -mcmodel=@var{code-model}
17947 Set the code model to one of
17950 All the data and read-only data segments must be within 512KB addressing space.
17951 The text segment must be within 16MB addressing space.
17952 @item @samp{medium}
17953 The data segment must be within 512KB while the read-only data segment can be
17954 within 4GB addressing space. The text segment should be still within 16MB
17957 All the text and data segments can be within 4GB addressing space.
17961 @opindex mctor-dtor
17962 Enable constructor/destructor feature.
17966 Guide linker to relax instructions.
17970 @node Nios II Options
17971 @subsection Nios II Options
17972 @cindex Nios II options
17973 @cindex Altera Nios II options
17975 These are the options defined for the Altera Nios II processor.
17981 @cindex smaller data references
17982 Put global and static objects less than or equal to @var{num} bytes
17983 into the small data or BSS sections instead of the normal data or BSS
17984 sections. The default value of @var{num} is 8.
17986 @item -mgpopt=@var{option}
17991 Generate (do not generate) GP-relative accesses. The following
17992 @var{option} names are recognized:
17997 Do not generate GP-relative accesses.
18000 Generate GP-relative accesses for small data objects that are not
18001 external or weak. Also use GP-relative addressing for objects that
18002 have been explicitly placed in a small data section via a @code{section}
18006 As for @samp{local}, but also generate GP-relative accesses for
18007 small data objects that are external or weak. If you use this option,
18008 you must ensure that all parts of your program (including libraries) are
18009 compiled with the same @option{-G} setting.
18012 Generate GP-relative accesses for all data objects in the program. If you
18013 use this option, the entire data and BSS segments
18014 of your program must fit in 64K of memory and you must use an appropriate
18015 linker script to allocate them within the addressible range of the
18019 Generate GP-relative addresses for function pointers as well as data
18020 pointers. If you use this option, the entire text, data, and BSS segments
18021 of your program must fit in 64K of memory and you must use an appropriate
18022 linker script to allocate them within the addressible range of the
18027 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18028 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18030 The default is @option{-mgpopt} except when @option{-fpic} or
18031 @option{-fPIC} is specified to generate position-independent code.
18032 Note that the Nios II ABI does not permit GP-relative accesses from
18035 You may need to specify @option{-mno-gpopt} explicitly when building
18036 programs that include large amounts of small data, including large
18037 GOT data sections. In this case, the 16-bit offset for GP-relative
18038 addressing may not be large enough to allow access to the entire
18039 small data section.
18045 Generate little-endian (default) or big-endian (experimental) code,
18048 @item -mbypass-cache
18049 @itemx -mno-bypass-cache
18050 @opindex mno-bypass-cache
18051 @opindex mbypass-cache
18052 Force all load and store instructions to always bypass cache by
18053 using I/O variants of the instructions. The default is not to
18056 @item -mno-cache-volatile
18057 @itemx -mcache-volatile
18058 @opindex mcache-volatile
18059 @opindex mno-cache-volatile
18060 Volatile memory access bypass the cache using the I/O variants of
18061 the load and store instructions. The default is not to bypass the cache.
18063 @item -mno-fast-sw-div
18064 @itemx -mfast-sw-div
18065 @opindex mno-fast-sw-div
18066 @opindex mfast-sw-div
18067 Do not use table-based fast divide for small numbers. The default
18068 is to use the fast divide at @option{-O3} and above.
18072 @itemx -mno-hw-mulx
18076 @opindex mno-hw-mul
18078 @opindex mno-hw-mulx
18080 @opindex mno-hw-div
18082 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18083 instructions by the compiler. The default is to emit @code{mul}
18084 and not emit @code{div} and @code{mulx}.
18086 @item -mcustom-@var{insn}=@var{N}
18087 @itemx -mno-custom-@var{insn}
18088 @opindex mcustom-@var{insn}
18089 @opindex mno-custom-@var{insn}
18090 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18091 custom instruction with encoding @var{N} when generating code that uses
18092 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18093 instruction 253 for single-precision floating-point add operations instead
18094 of the default behavior of using a library call.
18096 The following values of @var{insn} are supported. Except as otherwise
18097 noted, floating-point operations are expected to be implemented with
18098 normal IEEE 754 semantics and correspond directly to the C operators or the
18099 equivalent GCC built-in functions (@pxref{Other Builtins}).
18101 Single-precision floating point:
18104 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18105 Binary arithmetic operations.
18111 Unary absolute value.
18113 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18114 Comparison operations.
18116 @item @samp{fmins}, @samp{fmaxs}
18117 Floating-point minimum and maximum. These instructions are only
18118 generated if @option{-ffinite-math-only} is specified.
18120 @item @samp{fsqrts}
18121 Unary square root operation.
18123 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18124 Floating-point trigonometric and exponential functions. These instructions
18125 are only generated if @option{-funsafe-math-optimizations} is also specified.
18129 Double-precision floating point:
18132 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18133 Binary arithmetic operations.
18139 Unary absolute value.
18141 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18142 Comparison operations.
18144 @item @samp{fmind}, @samp{fmaxd}
18145 Double-precision minimum and maximum. These instructions are only
18146 generated if @option{-ffinite-math-only} is specified.
18148 @item @samp{fsqrtd}
18149 Unary square root operation.
18151 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18152 Double-precision trigonometric and exponential functions. These instructions
18153 are only generated if @option{-funsafe-math-optimizations} is also specified.
18159 @item @samp{fextsd}
18160 Conversion from single precision to double precision.
18162 @item @samp{ftruncds}
18163 Conversion from double precision to single precision.
18165 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18166 Conversion from floating point to signed or unsigned integer types, with
18167 truncation towards zero.
18170 Conversion from single-precision floating point to signed integer,
18171 rounding to the nearest integer and ties away from zero.
18172 This corresponds to the @code{__builtin_lroundf} function when
18173 @option{-fno-math-errno} is used.
18175 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18176 Conversion from signed or unsigned integer types to floating-point types.
18180 In addition, all of the following transfer instructions for internal
18181 registers X and Y must be provided to use any of the double-precision
18182 floating-point instructions. Custom instructions taking two
18183 double-precision source operands expect the first operand in the
18184 64-bit register X. The other operand (or only operand of a unary
18185 operation) is given to the custom arithmetic instruction with the
18186 least significant half in source register @var{src1} and the most
18187 significant half in @var{src2}. A custom instruction that returns a
18188 double-precision result returns the most significant 32 bits in the
18189 destination register and the other half in 32-bit register Y.
18190 GCC automatically generates the necessary code sequences to write
18191 register X and/or read register Y when double-precision floating-point
18192 instructions are used.
18197 Write @var{src1} into the least significant half of X and @var{src2} into
18198 the most significant half of X.
18201 Write @var{src1} into Y.
18203 @item @samp{frdxhi}, @samp{frdxlo}
18204 Read the most or least (respectively) significant half of X and store it in
18208 Read the value of Y and store it into @var{dest}.
18211 Note that you can gain more local control over generation of Nios II custom
18212 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18213 and @code{target("no-custom-@var{insn}")} function attributes
18214 (@pxref{Function Attributes})
18215 or pragmas (@pxref{Function Specific Option Pragmas}).
18217 @item -mcustom-fpu-cfg=@var{name}
18218 @opindex mcustom-fpu-cfg
18220 This option enables a predefined, named set of custom instruction encodings
18221 (see @option{-mcustom-@var{insn}} above).
18222 Currently, the following sets are defined:
18224 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18225 @gccoptlist{-mcustom-fmuls=252 @gol
18226 -mcustom-fadds=253 @gol
18227 -mcustom-fsubs=254 @gol
18228 -fsingle-precision-constant}
18230 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18231 @gccoptlist{-mcustom-fmuls=252 @gol
18232 -mcustom-fadds=253 @gol
18233 -mcustom-fsubs=254 @gol
18234 -mcustom-fdivs=255 @gol
18235 -fsingle-precision-constant}
18237 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18238 @gccoptlist{-mcustom-floatus=243 @gol
18239 -mcustom-fixsi=244 @gol
18240 -mcustom-floatis=245 @gol
18241 -mcustom-fcmpgts=246 @gol
18242 -mcustom-fcmples=249 @gol
18243 -mcustom-fcmpeqs=250 @gol
18244 -mcustom-fcmpnes=251 @gol
18245 -mcustom-fmuls=252 @gol
18246 -mcustom-fadds=253 @gol
18247 -mcustom-fsubs=254 @gol
18248 -mcustom-fdivs=255 @gol
18249 -fsingle-precision-constant}
18251 Custom instruction assignments given by individual
18252 @option{-mcustom-@var{insn}=} options override those given by
18253 @option{-mcustom-fpu-cfg=}, regardless of the
18254 order of the options on the command line.
18256 Note that you can gain more local control over selection of a FPU
18257 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18258 function attribute (@pxref{Function Attributes})
18259 or pragma (@pxref{Function Specific Option Pragmas}).
18263 These additional @samp{-m} options are available for the Altera Nios II
18264 ELF (bare-metal) target:
18270 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18271 startup and termination code, and is typically used in conjunction with
18272 @option{-msys-crt0=} to specify the location of the alternate startup code
18273 provided by the HAL BSP.
18277 Link with a limited version of the C library, @option{-lsmallc}, rather than
18280 @item -msys-crt0=@var{startfile}
18282 @var{startfile} is the file name of the startfile (crt0) to use
18283 when linking. This option is only useful in conjunction with @option{-mhal}.
18285 @item -msys-lib=@var{systemlib}
18287 @var{systemlib} is the library name of the library that provides
18288 low-level system calls required by the C library,
18289 e.g. @code{read} and @code{write}.
18290 This option is typically used to link with a library provided by a HAL BSP.
18294 @node PDP-11 Options
18295 @subsection PDP-11 Options
18296 @cindex PDP-11 Options
18298 These options are defined for the PDP-11:
18303 Use hardware FPP floating point. This is the default. (FIS floating
18304 point on the PDP-11/40 is not supported.)
18307 @opindex msoft-float
18308 Do not use hardware floating point.
18312 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18316 Return floating-point results in memory. This is the default.
18320 Generate code for a PDP-11/40.
18324 Generate code for a PDP-11/45. This is the default.
18328 Generate code for a PDP-11/10.
18330 @item -mbcopy-builtin
18331 @opindex mbcopy-builtin
18332 Use inline @code{movmemhi} patterns for copying memory. This is the
18337 Do not use inline @code{movmemhi} patterns for copying memory.
18343 Use 16-bit @code{int}. This is the default.
18349 Use 32-bit @code{int}.
18352 @itemx -mno-float32
18354 @opindex mno-float32
18355 Use 64-bit @code{float}. This is the default.
18358 @itemx -mno-float64
18360 @opindex mno-float64
18361 Use 32-bit @code{float}.
18365 Use @code{abshi2} pattern. This is the default.
18369 Do not use @code{abshi2} pattern.
18371 @item -mbranch-expensive
18372 @opindex mbranch-expensive
18373 Pretend that branches are expensive. This is for experimenting with
18374 code generation only.
18376 @item -mbranch-cheap
18377 @opindex mbranch-cheap
18378 Do not pretend that branches are expensive. This is the default.
18382 Use Unix assembler syntax. This is the default when configured for
18383 @samp{pdp11-*-bsd}.
18387 Use DEC assembler syntax. This is the default when configured for any
18388 PDP-11 target other than @samp{pdp11-*-bsd}.
18391 @node picoChip Options
18392 @subsection picoChip Options
18393 @cindex picoChip options
18395 These @samp{-m} options are defined for picoChip implementations:
18399 @item -mae=@var{ae_type}
18401 Set the instruction set, register set, and instruction scheduling
18402 parameters for array element type @var{ae_type}. Supported values
18403 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18405 @option{-mae=ANY} selects a completely generic AE type. Code
18406 generated with this option runs on any of the other AE types. The
18407 code is not as efficient as it would be if compiled for a specific
18408 AE type, and some types of operation (e.g., multiplication) do not
18409 work properly on all types of AE.
18411 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
18412 for compiled code, and is the default.
18414 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
18415 option may suffer from poor performance of byte (char) manipulation,
18416 since the DSP AE does not provide hardware support for byte load/stores.
18418 @item -msymbol-as-address
18419 Enable the compiler to directly use a symbol name as an address in a
18420 load/store instruction, without first loading it into a
18421 register. Typically, the use of this option generates larger
18422 programs, which run faster than when the option isn't used. However, the
18423 results vary from program to program, so it is left as a user option,
18424 rather than being permanently enabled.
18426 @item -mno-inefficient-warnings
18427 Disables warnings about the generation of inefficient code. These
18428 warnings can be generated, for example, when compiling code that
18429 performs byte-level memory operations on the MAC AE type. The MAC AE has
18430 no hardware support for byte-level memory operations, so all byte
18431 load/stores must be synthesized from word load/store operations. This is
18432 inefficient and a warning is generated to indicate
18433 that you should rewrite the code to avoid byte operations, or to target
18434 an AE type that has the necessary hardware support. This option disables
18439 @node PowerPC Options
18440 @subsection PowerPC Options
18441 @cindex PowerPC options
18443 These are listed under @xref{RS/6000 and PowerPC Options}.
18446 @subsection RL78 Options
18447 @cindex RL78 Options
18453 Links in additional target libraries to support operation within a
18460 Specifies the type of hardware multiplication support to be used. The
18461 default is @samp{none}, which uses software multiplication functions.
18462 The @samp{g13} option is for the hardware multiply/divide peripheral
18463 only on the RL78/G13 targets. The @samp{rl78} option is for the
18464 standard hardware multiplication defined in the RL78 software manual.
18466 @item -m64bit-doubles
18467 @itemx -m32bit-doubles
18468 @opindex m64bit-doubles
18469 @opindex m32bit-doubles
18470 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
18471 or 32 bits (@option{-m32bit-doubles}) in size. The default is
18472 @option{-m32bit-doubles}.
18476 @node RS/6000 and PowerPC Options
18477 @subsection IBM RS/6000 and PowerPC Options
18478 @cindex RS/6000 and PowerPC Options
18479 @cindex IBM RS/6000 and PowerPC Options
18481 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
18483 @item -mpowerpc-gpopt
18484 @itemx -mno-powerpc-gpopt
18485 @itemx -mpowerpc-gfxopt
18486 @itemx -mno-powerpc-gfxopt
18489 @itemx -mno-powerpc64
18493 @itemx -mno-popcntb
18495 @itemx -mno-popcntd
18504 @itemx -mno-hard-dfp
18505 @opindex mpowerpc-gpopt
18506 @opindex mno-powerpc-gpopt
18507 @opindex mpowerpc-gfxopt
18508 @opindex mno-powerpc-gfxopt
18509 @opindex mpowerpc64
18510 @opindex mno-powerpc64
18514 @opindex mno-popcntb
18516 @opindex mno-popcntd
18522 @opindex mno-mfpgpr
18524 @opindex mno-hard-dfp
18525 You use these options to specify which instructions are available on the
18526 processor you are using. The default value of these options is
18527 determined when configuring GCC@. Specifying the
18528 @option{-mcpu=@var{cpu_type}} overrides the specification of these
18529 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
18530 rather than the options listed above.
18532 Specifying @option{-mpowerpc-gpopt} allows
18533 GCC to use the optional PowerPC architecture instructions in the
18534 General Purpose group, including floating-point square root. Specifying
18535 @option{-mpowerpc-gfxopt} allows GCC to
18536 use the optional PowerPC architecture instructions in the Graphics
18537 group, including floating-point select.
18539 The @option{-mmfcrf} option allows GCC to generate the move from
18540 condition register field instruction implemented on the POWER4
18541 processor and other processors that support the PowerPC V2.01
18543 The @option{-mpopcntb} option allows GCC to generate the popcount and
18544 double-precision FP reciprocal estimate instruction implemented on the
18545 POWER5 processor and other processors that support the PowerPC V2.02
18547 The @option{-mpopcntd} option allows GCC to generate the popcount
18548 instruction implemented on the POWER7 processor and other processors
18549 that support the PowerPC V2.06 architecture.
18550 The @option{-mfprnd} option allows GCC to generate the FP round to
18551 integer instructions implemented on the POWER5+ processor and other
18552 processors that support the PowerPC V2.03 architecture.
18553 The @option{-mcmpb} option allows GCC to generate the compare bytes
18554 instruction implemented on the POWER6 processor and other processors
18555 that support the PowerPC V2.05 architecture.
18556 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
18557 general-purpose register instructions implemented on the POWER6X
18558 processor and other processors that support the extended PowerPC V2.05
18560 The @option{-mhard-dfp} option allows GCC to generate the decimal
18561 floating-point instructions implemented on some POWER processors.
18563 The @option{-mpowerpc64} option allows GCC to generate the additional
18564 64-bit instructions that are found in the full PowerPC64 architecture
18565 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
18566 @option{-mno-powerpc64}.
18568 @item -mcpu=@var{cpu_type}
18570 Set architecture type, register usage, and
18571 instruction scheduling parameters for machine type @var{cpu_type}.
18572 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
18573 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
18574 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
18575 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
18576 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
18577 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
18578 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
18579 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
18580 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
18581 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
18582 @samp{powerpc64}, and @samp{rs64}.
18584 @option{-mcpu=powerpc}, and @option{-mcpu=powerpc64} specify pure 32-bit
18585 PowerPC and 64-bit PowerPC architecture machine
18586 types, with an appropriate, generic processor model assumed for
18587 scheduling purposes.
18589 The other options specify a specific processor. Code generated under
18590 those options runs best on that processor, and may not run at all on
18593 The @option{-mcpu} options automatically enable or disable the
18596 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
18597 -mpopcntb -mpopcntd -mpowerpc64 @gol
18598 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
18599 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
18600 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
18601 -mquad-memory -mquad-memory-atomic}
18603 The particular options set for any particular CPU varies between
18604 compiler versions, depending on what setting seems to produce optimal
18605 code for that CPU; it doesn't necessarily reflect the actual hardware's
18606 capabilities. If you wish to set an individual option to a particular
18607 value, you may specify it after the @option{-mcpu} option, like
18608 @option{-mcpu=970 -mno-altivec}.
18610 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
18611 not enabled or disabled by the @option{-mcpu} option at present because
18612 AIX does not have full support for these options. You may still
18613 enable or disable them individually if you're sure it'll work in your
18616 @item -mtune=@var{cpu_type}
18618 Set the instruction scheduling parameters for machine type
18619 @var{cpu_type}, but do not set the architecture type or register usage,
18620 as @option{-mcpu=@var{cpu_type}} does. The same
18621 values for @var{cpu_type} are used for @option{-mtune} as for
18622 @option{-mcpu}. If both are specified, the code generated uses the
18623 architecture and registers set by @option{-mcpu}, but the
18624 scheduling parameters set by @option{-mtune}.
18626 @item -mcmodel=small
18627 @opindex mcmodel=small
18628 Generate PowerPC64 code for the small model: The TOC is limited to
18631 @item -mcmodel=medium
18632 @opindex mcmodel=medium
18633 Generate PowerPC64 code for the medium model: The TOC and other static
18634 data may be up to a total of 4G in size.
18636 @item -mcmodel=large
18637 @opindex mcmodel=large
18638 Generate PowerPC64 code for the large model: The TOC may be up to 4G
18639 in size. Other data and code is only limited by the 64-bit address
18643 @itemx -mno-altivec
18645 @opindex mno-altivec
18646 Generate code that uses (does not use) AltiVec instructions, and also
18647 enable the use of built-in functions that allow more direct access to
18648 the AltiVec instruction set. You may also need to set
18649 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
18652 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
18653 @option{-maltivec=be}, the element order for Altivec intrinsics such
18654 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
18655 match array element order corresponding to the endianness of the
18656 target. That is, element zero identifies the leftmost element in a
18657 vector register when targeting a big-endian platform, and identifies
18658 the rightmost element in a vector register when targeting a
18659 little-endian platform.
18662 @opindex maltivec=be
18663 Generate Altivec instructions using big-endian element order,
18664 regardless of whether the target is big- or little-endian. This is
18665 the default when targeting a big-endian platform.
18667 The element order is used to interpret element numbers in Altivec
18668 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18669 @code{vec_insert}. By default, these match array element order
18670 corresponding to the endianness for the target.
18673 @opindex maltivec=le
18674 Generate Altivec instructions using little-endian element order,
18675 regardless of whether the target is big- or little-endian. This is
18676 the default when targeting a little-endian platform. This option is
18677 currently ignored when targeting a big-endian platform.
18679 The element order is used to interpret element numbers in Altivec
18680 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18681 @code{vec_insert}. By default, these match array element order
18682 corresponding to the endianness for the target.
18687 @opindex mno-vrsave
18688 Generate VRSAVE instructions when generating AltiVec code.
18690 @item -mgen-cell-microcode
18691 @opindex mgen-cell-microcode
18692 Generate Cell microcode instructions.
18694 @item -mwarn-cell-microcode
18695 @opindex mwarn-cell-microcode
18696 Warn when a Cell microcode instruction is emitted. An example
18697 of a Cell microcode instruction is a variable shift.
18700 @opindex msecure-plt
18701 Generate code that allows @command{ld} and @command{ld.so}
18702 to build executables and shared
18703 libraries with non-executable @code{.plt} and @code{.got} sections.
18705 32-bit SYSV ABI option.
18709 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
18711 requires @code{.plt} and @code{.got}
18712 sections that are both writable and executable.
18713 This is a PowerPC 32-bit SYSV ABI option.
18719 This switch enables or disables the generation of ISEL instructions.
18721 @item -misel=@var{yes/no}
18722 This switch has been deprecated. Use @option{-misel} and
18723 @option{-mno-isel} instead.
18729 This switch enables or disables the generation of SPE simd
18735 @opindex mno-paired
18736 This switch enables or disables the generation of PAIRED simd
18739 @item -mspe=@var{yes/no}
18740 This option has been deprecated. Use @option{-mspe} and
18741 @option{-mno-spe} instead.
18747 Generate code that uses (does not use) vector/scalar (VSX)
18748 instructions, and also enable the use of built-in functions that allow
18749 more direct access to the VSX instruction set.
18754 @opindex mno-crypto
18755 Enable the use (disable) of the built-in functions that allow direct
18756 access to the cryptographic instructions that were added in version
18757 2.07 of the PowerPC ISA.
18759 @item -mdirect-move
18760 @itemx -mno-direct-move
18761 @opindex mdirect-move
18762 @opindex mno-direct-move
18763 Generate code that uses (does not use) the instructions to move data
18764 between the general purpose registers and the vector/scalar (VSX)
18765 registers that were added in version 2.07 of the PowerPC ISA.
18767 @item -mpower8-fusion
18768 @itemx -mno-power8-fusion
18769 @opindex mpower8-fusion
18770 @opindex mno-power8-fusion
18771 Generate code that keeps (does not keeps) some integer operations
18772 adjacent so that the instructions can be fused together on power8 and
18775 @item -mpower8-vector
18776 @itemx -mno-power8-vector
18777 @opindex mpower8-vector
18778 @opindex mno-power8-vector
18779 Generate code that uses (does not use) the vector and scalar
18780 instructions that were added in version 2.07 of the PowerPC ISA. Also
18781 enable the use of built-in functions that allow more direct access to
18782 the vector instructions.
18784 @item -mquad-memory
18785 @itemx -mno-quad-memory
18786 @opindex mquad-memory
18787 @opindex mno-quad-memory
18788 Generate code that uses (does not use) the non-atomic quad word memory
18789 instructions. The @option{-mquad-memory} option requires use of
18792 @item -mquad-memory-atomic
18793 @itemx -mno-quad-memory-atomic
18794 @opindex mquad-memory-atomic
18795 @opindex mno-quad-memory-atomic
18796 Generate code that uses (does not use) the atomic quad word memory
18797 instructions. The @option{-mquad-memory-atomic} option requires use of
18800 @item -mupper-regs-df
18801 @itemx -mno-upper-regs-df
18802 @opindex mupper-regs-df
18803 @opindex mno-upper-regs-df
18804 Generate code that uses (does not use) the scalar double precision
18805 instructions that target all 64 registers in the vector/scalar
18806 floating point register set that were added in version 2.06 of the
18807 PowerPC ISA. The @option{-mupper-regs-df} turned on by default if you
18808 use either of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
18809 @option{-mvsx} options.
18811 @item -mupper-regs-sf
18812 @itemx -mno-upper-regs-sf
18813 @opindex mupper-regs-sf
18814 @opindex mno-upper-regs-sf
18815 Generate code that uses (does not use) the scalar single precision
18816 instructions that target all 64 registers in the vector/scalar
18817 floating point register set that were added in version 2.07 of the
18818 PowerPC ISA. The @option{-mupper-regs-sf} turned on by default if you
18819 use either of the @option{-mcpu=power8}, or @option{-mpower8-vector}
18823 @itemx -mno-upper-regs
18824 @opindex mupper-regs
18825 @opindex mno-upper-regs
18826 Generate code that uses (does not use) the scalar
18827 instructions that target all 64 registers in the vector/scalar
18828 floating point register set, depending on the model of the machine.
18830 If the @option{-mno-upper-regs} option is used, it turns off both
18831 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
18833 @item -mfloat-gprs=@var{yes/single/double/no}
18834 @itemx -mfloat-gprs
18835 @opindex mfloat-gprs
18836 This switch enables or disables the generation of floating-point
18837 operations on the general-purpose registers for architectures that
18840 The argument @samp{yes} or @samp{single} enables the use of
18841 single-precision floating-point operations.
18843 The argument @samp{double} enables the use of single and
18844 double-precision floating-point operations.
18846 The argument @samp{no} disables floating-point operations on the
18847 general-purpose registers.
18849 This option is currently only available on the MPC854x.
18855 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
18856 targets (including GNU/Linux). The 32-bit environment sets int, long
18857 and pointer to 32 bits and generates code that runs on any PowerPC
18858 variant. The 64-bit environment sets int to 32 bits and long and
18859 pointer to 64 bits, and generates code for PowerPC64, as for
18860 @option{-mpowerpc64}.
18863 @itemx -mno-fp-in-toc
18864 @itemx -mno-sum-in-toc
18865 @itemx -mminimal-toc
18867 @opindex mno-fp-in-toc
18868 @opindex mno-sum-in-toc
18869 @opindex mminimal-toc
18870 Modify generation of the TOC (Table Of Contents), which is created for
18871 every executable file. The @option{-mfull-toc} option is selected by
18872 default. In that case, GCC allocates at least one TOC entry for
18873 each unique non-automatic variable reference in your program. GCC
18874 also places floating-point constants in the TOC@. However, only
18875 16,384 entries are available in the TOC@.
18877 If you receive a linker error message that saying you have overflowed
18878 the available TOC space, you can reduce the amount of TOC space used
18879 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
18880 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
18881 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
18882 generate code to calculate the sum of an address and a constant at
18883 run time instead of putting that sum into the TOC@. You may specify one
18884 or both of these options. Each causes GCC to produce very slightly
18885 slower and larger code at the expense of conserving TOC space.
18887 If you still run out of space in the TOC even when you specify both of
18888 these options, specify @option{-mminimal-toc} instead. This option causes
18889 GCC to make only one TOC entry for every file. When you specify this
18890 option, GCC produces code that is slower and larger but which
18891 uses extremely little TOC space. You may wish to use this option
18892 only on files that contain less frequently-executed code.
18898 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
18899 @code{long} type, and the infrastructure needed to support them.
18900 Specifying @option{-maix64} implies @option{-mpowerpc64},
18901 while @option{-maix32} disables the 64-bit ABI and
18902 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
18905 @itemx -mno-xl-compat
18906 @opindex mxl-compat
18907 @opindex mno-xl-compat
18908 Produce code that conforms more closely to IBM XL compiler semantics
18909 when using AIX-compatible ABI@. Pass floating-point arguments to
18910 prototyped functions beyond the register save area (RSA) on the stack
18911 in addition to argument FPRs. Do not assume that most significant
18912 double in 128-bit long double value is properly rounded when comparing
18913 values and converting to double. Use XL symbol names for long double
18916 The AIX calling convention was extended but not initially documented to
18917 handle an obscure K&R C case of calling a function that takes the
18918 address of its arguments with fewer arguments than declared. IBM XL
18919 compilers access floating-point arguments that do not fit in the
18920 RSA from the stack when a subroutine is compiled without
18921 optimization. Because always storing floating-point arguments on the
18922 stack is inefficient and rarely needed, this option is not enabled by
18923 default and only is necessary when calling subroutines compiled by IBM
18924 XL compilers without optimization.
18928 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
18929 application written to use message passing with special startup code to
18930 enable the application to run. The system must have PE installed in the
18931 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
18932 must be overridden with the @option{-specs=} option to specify the
18933 appropriate directory location. The Parallel Environment does not
18934 support threads, so the @option{-mpe} option and the @option{-pthread}
18935 option are incompatible.
18937 @item -malign-natural
18938 @itemx -malign-power
18939 @opindex malign-natural
18940 @opindex malign-power
18941 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
18942 @option{-malign-natural} overrides the ABI-defined alignment of larger
18943 types, such as floating-point doubles, on their natural size-based boundary.
18944 The option @option{-malign-power} instructs GCC to follow the ABI-specified
18945 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
18947 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
18951 @itemx -mhard-float
18952 @opindex msoft-float
18953 @opindex mhard-float
18954 Generate code that does not use (uses) the floating-point register set.
18955 Software floating-point emulation is provided if you use the
18956 @option{-msoft-float} option, and pass the option to GCC when linking.
18958 @item -msingle-float
18959 @itemx -mdouble-float
18960 @opindex msingle-float
18961 @opindex mdouble-float
18962 Generate code for single- or double-precision floating-point operations.
18963 @option{-mdouble-float} implies @option{-msingle-float}.
18966 @opindex msimple-fpu
18967 Do not generate @code{sqrt} and @code{div} instructions for hardware
18968 floating-point unit.
18970 @item -mfpu=@var{name}
18972 Specify type of floating-point unit. Valid values for @var{name} are
18973 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
18974 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
18975 @samp{sp_full} (equivalent to @option{-msingle-float}),
18976 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
18979 @opindex mxilinx-fpu
18980 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
18983 @itemx -mno-multiple
18985 @opindex mno-multiple
18986 Generate code that uses (does not use) the load multiple word
18987 instructions and the store multiple word instructions. These
18988 instructions are generated by default on POWER systems, and not
18989 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
18990 PowerPC systems, since those instructions do not work when the
18991 processor is in little-endian mode. The exceptions are PPC740 and
18992 PPC750 which permit these instructions in little-endian mode.
18997 @opindex mno-string
18998 Generate code that uses (does not use) the load string instructions
18999 and the store string word instructions to save multiple registers and
19000 do small block moves. These instructions are generated by default on
19001 POWER systems, and not generated on PowerPC systems. Do not use
19002 @option{-mstring} on little-endian PowerPC systems, since those
19003 instructions do not work when the processor is in little-endian mode.
19004 The exceptions are PPC740 and PPC750 which permit these instructions
19005 in little-endian mode.
19010 @opindex mno-update
19011 Generate code that uses (does not use) the load or store instructions
19012 that update the base register to the address of the calculated memory
19013 location. These instructions are generated by default. If you use
19014 @option{-mno-update}, there is a small window between the time that the
19015 stack pointer is updated and the address of the previous frame is
19016 stored, which means code that walks the stack frame across interrupts or
19017 signals may get corrupted data.
19019 @item -mavoid-indexed-addresses
19020 @itemx -mno-avoid-indexed-addresses
19021 @opindex mavoid-indexed-addresses
19022 @opindex mno-avoid-indexed-addresses
19023 Generate code that tries to avoid (not avoid) the use of indexed load
19024 or store instructions. These instructions can incur a performance
19025 penalty on Power6 processors in certain situations, such as when
19026 stepping through large arrays that cross a 16M boundary. This option
19027 is enabled by default when targeting Power6 and disabled otherwise.
19030 @itemx -mno-fused-madd
19031 @opindex mfused-madd
19032 @opindex mno-fused-madd
19033 Generate code that uses (does not use) the floating-point multiply and
19034 accumulate instructions. These instructions are generated by default
19035 if hardware floating point is used. The machine-dependent
19036 @option{-mfused-madd} option is now mapped to the machine-independent
19037 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19038 mapped to @option{-ffp-contract=off}.
19044 Generate code that uses (does not use) the half-word multiply and
19045 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19046 These instructions are generated by default when targeting those
19053 Generate code that uses (does not use) the string-search @samp{dlmzb}
19054 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19055 generated by default when targeting those processors.
19057 @item -mno-bit-align
19059 @opindex mno-bit-align
19060 @opindex mbit-align
19061 On System V.4 and embedded PowerPC systems do not (do) force structures
19062 and unions that contain bit-fields to be aligned to the base type of the
19065 For example, by default a structure containing nothing but 8
19066 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19067 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19068 the structure is aligned to a 1-byte boundary and is 1 byte in
19071 @item -mno-strict-align
19072 @itemx -mstrict-align
19073 @opindex mno-strict-align
19074 @opindex mstrict-align
19075 On System V.4 and embedded PowerPC systems do not (do) assume that
19076 unaligned memory references are handled by the system.
19078 @item -mrelocatable
19079 @itemx -mno-relocatable
19080 @opindex mrelocatable
19081 @opindex mno-relocatable
19082 Generate code that allows (does not allow) a static executable to be
19083 relocated to a different address at run time. A simple embedded
19084 PowerPC system loader should relocate the entire contents of
19085 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19086 a table of 32-bit addresses generated by this option. For this to
19087 work, all objects linked together must be compiled with
19088 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19089 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19091 @item -mrelocatable-lib
19092 @itemx -mno-relocatable-lib
19093 @opindex mrelocatable-lib
19094 @opindex mno-relocatable-lib
19095 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19096 @code{.fixup} section to allow static executables to be relocated at
19097 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19098 alignment of @option{-mrelocatable}. Objects compiled with
19099 @option{-mrelocatable-lib} may be linked with objects compiled with
19100 any combination of the @option{-mrelocatable} options.
19106 On System V.4 and embedded PowerPC systems do not (do) assume that
19107 register 2 contains a pointer to a global area pointing to the addresses
19108 used in the program.
19111 @itemx -mlittle-endian
19113 @opindex mlittle-endian
19114 On System V.4 and embedded PowerPC systems compile code for the
19115 processor in little-endian mode. The @option{-mlittle-endian} option is
19116 the same as @option{-mlittle}.
19119 @itemx -mbig-endian
19121 @opindex mbig-endian
19122 On System V.4 and embedded PowerPC systems compile code for the
19123 processor in big-endian mode. The @option{-mbig-endian} option is
19124 the same as @option{-mbig}.
19126 @item -mdynamic-no-pic
19127 @opindex mdynamic-no-pic
19128 On Darwin and Mac OS X systems, compile code so that it is not
19129 relocatable, but that its external references are relocatable. The
19130 resulting code is suitable for applications, but not shared
19133 @item -msingle-pic-base
19134 @opindex msingle-pic-base
19135 Treat the register used for PIC addressing as read-only, rather than
19136 loading it in the prologue for each function. The runtime system is
19137 responsible for initializing this register with an appropriate value
19138 before execution begins.
19140 @item -mprioritize-restricted-insns=@var{priority}
19141 @opindex mprioritize-restricted-insns
19142 This option controls the priority that is assigned to
19143 dispatch-slot restricted instructions during the second scheduling
19144 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19145 or @samp{2} to assign no, highest, or second-highest (respectively)
19146 priority to dispatch-slot restricted
19149 @item -msched-costly-dep=@var{dependence_type}
19150 @opindex msched-costly-dep
19151 This option controls which dependences are considered costly
19152 by the target during instruction scheduling. The argument
19153 @var{dependence_type} takes one of the following values:
19157 No dependence is costly.
19160 All dependences are costly.
19162 @item @samp{true_store_to_load}
19163 A true dependence from store to load is costly.
19165 @item @samp{store_to_load}
19166 Any dependence from store to load is costly.
19169 Any dependence for which the latency is greater than or equal to
19170 @var{number} is costly.
19173 @item -minsert-sched-nops=@var{scheme}
19174 @opindex minsert-sched-nops
19175 This option controls which NOP insertion scheme is used during
19176 the second scheduling pass. The argument @var{scheme} takes one of the
19184 Pad with NOPs any dispatch group that has vacant issue slots,
19185 according to the scheduler's grouping.
19187 @item @samp{regroup_exact}
19188 Insert NOPs to force costly dependent insns into
19189 separate groups. Insert exactly as many NOPs as needed to force an insn
19190 to a new group, according to the estimated processor grouping.
19193 Insert NOPs to force costly dependent insns into
19194 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19198 @opindex mcall-sysv
19199 On System V.4 and embedded PowerPC systems compile code using calling
19200 conventions that adhere to the March 1995 draft of the System V
19201 Application Binary Interface, PowerPC processor supplement. This is the
19202 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19204 @item -mcall-sysv-eabi
19206 @opindex mcall-sysv-eabi
19207 @opindex mcall-eabi
19208 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19210 @item -mcall-sysv-noeabi
19211 @opindex mcall-sysv-noeabi
19212 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19214 @item -mcall-aixdesc
19216 On System V.4 and embedded PowerPC systems compile code for the AIX
19220 @opindex mcall-linux
19221 On System V.4 and embedded PowerPC systems compile code for the
19222 Linux-based GNU system.
19224 @item -mcall-freebsd
19225 @opindex mcall-freebsd
19226 On System V.4 and embedded PowerPC systems compile code for the
19227 FreeBSD operating system.
19229 @item -mcall-netbsd
19230 @opindex mcall-netbsd
19231 On System V.4 and embedded PowerPC systems compile code for the
19232 NetBSD operating system.
19234 @item -mcall-openbsd
19235 @opindex mcall-netbsd
19236 On System V.4 and embedded PowerPC systems compile code for the
19237 OpenBSD operating system.
19239 @item -maix-struct-return
19240 @opindex maix-struct-return
19241 Return all structures in memory (as specified by the AIX ABI)@.
19243 @item -msvr4-struct-return
19244 @opindex msvr4-struct-return
19245 Return structures smaller than 8 bytes in registers (as specified by the
19248 @item -mabi=@var{abi-type}
19250 Extend the current ABI with a particular extension, or remove such extension.
19251 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
19252 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
19253 @samp{elfv1}, @samp{elfv2}@.
19257 Extend the current ABI with SPE ABI extensions. This does not change
19258 the default ABI, instead it adds the SPE ABI extensions to the current
19262 @opindex mabi=no-spe
19263 Disable Book-E SPE ABI extensions for the current ABI@.
19265 @item -mabi=ibmlongdouble
19266 @opindex mabi=ibmlongdouble
19267 Change the current ABI to use IBM extended-precision long double.
19268 This is a PowerPC 32-bit SYSV ABI option.
19270 @item -mabi=ieeelongdouble
19271 @opindex mabi=ieeelongdouble
19272 Change the current ABI to use IEEE extended-precision long double.
19273 This is a PowerPC 32-bit Linux ABI option.
19276 @opindex mabi=elfv1
19277 Change the current ABI to use the ELFv1 ABI.
19278 This is the default ABI for big-endian PowerPC 64-bit Linux.
19279 Overriding the default ABI requires special system support and is
19280 likely to fail in spectacular ways.
19283 @opindex mabi=elfv2
19284 Change the current ABI to use the ELFv2 ABI.
19285 This is the default ABI for little-endian PowerPC 64-bit Linux.
19286 Overriding the default ABI requires special system support and is
19287 likely to fail in spectacular ways.
19290 @itemx -mno-prototype
19291 @opindex mprototype
19292 @opindex mno-prototype
19293 On System V.4 and embedded PowerPC systems assume that all calls to
19294 variable argument functions are properly prototyped. Otherwise, the
19295 compiler must insert an instruction before every non-prototyped call to
19296 set or clear bit 6 of the condition code register (@code{CR}) to
19297 indicate whether floating-point values are passed in the floating-point
19298 registers in case the function takes variable arguments. With
19299 @option{-mprototype}, only calls to prototyped variable argument functions
19300 set or clear the bit.
19304 On embedded PowerPC systems, assume that the startup module is called
19305 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19306 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19311 On embedded PowerPC systems, assume that the startup module is called
19312 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19317 On embedded PowerPC systems, assume that the startup module is called
19318 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19321 @item -myellowknife
19322 @opindex myellowknife
19323 On embedded PowerPC systems, assume that the startup module is called
19324 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19329 On System V.4 and embedded PowerPC systems, specify that you are
19330 compiling for a VxWorks system.
19334 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
19335 header to indicate that @samp{eabi} extended relocations are used.
19341 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19342 Embedded Applications Binary Interface (EABI), which is a set of
19343 modifications to the System V.4 specifications. Selecting @option{-meabi}
19344 means that the stack is aligned to an 8-byte boundary, a function
19345 @code{__eabi} is called from @code{main} to set up the EABI
19346 environment, and the @option{-msdata} option can use both @code{r2} and
19347 @code{r13} to point to two separate small data areas. Selecting
19348 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19349 no EABI initialization function is called from @code{main}, and the
19350 @option{-msdata} option only uses @code{r13} to point to a single
19351 small data area. The @option{-meabi} option is on by default if you
19352 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19355 @opindex msdata=eabi
19356 On System V.4 and embedded PowerPC systems, put small initialized
19357 @code{const} global and static data in the @code{.sdata2} section, which
19358 is pointed to by register @code{r2}. Put small initialized
19359 non-@code{const} global and static data in the @code{.sdata} section,
19360 which is pointed to by register @code{r13}. Put small uninitialized
19361 global and static data in the @code{.sbss} section, which is adjacent to
19362 the @code{.sdata} section. The @option{-msdata=eabi} option is
19363 incompatible with the @option{-mrelocatable} option. The
19364 @option{-msdata=eabi} option also sets the @option{-memb} option.
19367 @opindex msdata=sysv
19368 On System V.4 and embedded PowerPC systems, put small global and static
19369 data in the @code{.sdata} section, which is pointed to by register
19370 @code{r13}. Put small uninitialized global and static data in the
19371 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
19372 The @option{-msdata=sysv} option is incompatible with the
19373 @option{-mrelocatable} option.
19375 @item -msdata=default
19377 @opindex msdata=default
19379 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19380 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19381 same as @option{-msdata=sysv}.
19384 @opindex msdata=data
19385 On System V.4 and embedded PowerPC systems, put small global
19386 data in the @code{.sdata} section. Put small uninitialized global
19387 data in the @code{.sbss} section. Do not use register @code{r13}
19388 to address small data however. This is the default behavior unless
19389 other @option{-msdata} options are used.
19393 @opindex msdata=none
19395 On embedded PowerPC systems, put all initialized global and static data
19396 in the @code{.data} section, and all uninitialized data in the
19397 @code{.bss} section.
19399 @item -mblock-move-inline-limit=@var{num}
19400 @opindex mblock-move-inline-limit
19401 Inline all block moves (such as calls to @code{memcpy} or structure
19402 copies) less than or equal to @var{num} bytes. The minimum value for
19403 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19404 targets. The default value is target-specific.
19408 @cindex smaller data references (PowerPC)
19409 @cindex .sdata/.sdata2 references (PowerPC)
19410 On embedded PowerPC systems, put global and static items less than or
19411 equal to @var{num} bytes into the small data or BSS sections instead of
19412 the normal data or BSS section. By default, @var{num} is 8. The
19413 @option{-G @var{num}} switch is also passed to the linker.
19414 All modules should be compiled with the same @option{-G @var{num}} value.
19417 @itemx -mno-regnames
19419 @opindex mno-regnames
19420 On System V.4 and embedded PowerPC systems do (do not) emit register
19421 names in the assembly language output using symbolic forms.
19424 @itemx -mno-longcall
19426 @opindex mno-longcall
19427 By default assume that all calls are far away so that a longer and more
19428 expensive calling sequence is required. This is required for calls
19429 farther than 32 megabytes (33,554,432 bytes) from the current location.
19430 A short call is generated if the compiler knows
19431 the call cannot be that far away. This setting can be overridden by
19432 the @code{shortcall} function attribute, or by @code{#pragma
19435 Some linkers are capable of detecting out-of-range calls and generating
19436 glue code on the fly. On these systems, long calls are unnecessary and
19437 generate slower code. As of this writing, the AIX linker can do this,
19438 as can the GNU linker for PowerPC/64. It is planned to add this feature
19439 to the GNU linker for 32-bit PowerPC systems as well.
19441 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
19442 callee, L42}, plus a @dfn{branch island} (glue code). The two target
19443 addresses represent the callee and the branch island. The
19444 Darwin/PPC linker prefers the first address and generates a @code{bl
19445 callee} if the PPC @code{bl} instruction reaches the callee directly;
19446 otherwise, the linker generates @code{bl L42} to call the branch
19447 island. The branch island is appended to the body of the
19448 calling function; it computes the full 32-bit address of the callee
19451 On Mach-O (Darwin) systems, this option directs the compiler emit to
19452 the glue for every direct call, and the Darwin linker decides whether
19453 to use or discard it.
19455 In the future, GCC may ignore all longcall specifications
19456 when the linker is known to generate glue.
19458 @item -mtls-markers
19459 @itemx -mno-tls-markers
19460 @opindex mtls-markers
19461 @opindex mno-tls-markers
19462 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
19463 specifying the function argument. The relocation allows the linker to
19464 reliably associate function call with argument setup instructions for
19465 TLS optimization, which in turn allows GCC to better schedule the
19470 Adds support for multithreading with the @dfn{pthreads} library.
19471 This option sets flags for both the preprocessor and linker.
19476 This option enables use of the reciprocal estimate and
19477 reciprocal square root estimate instructions with additional
19478 Newton-Raphson steps to increase precision instead of doing a divide or
19479 square root and divide for floating-point arguments. You should use
19480 the @option{-ffast-math} option when using @option{-mrecip} (or at
19481 least @option{-funsafe-math-optimizations},
19482 @option{-finite-math-only}, @option{-freciprocal-math} and
19483 @option{-fno-trapping-math}). Note that while the throughput of the
19484 sequence is generally higher than the throughput of the non-reciprocal
19485 instruction, the precision of the sequence can be decreased by up to 2
19486 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
19489 @item -mrecip=@var{opt}
19490 @opindex mrecip=opt
19491 This option controls which reciprocal estimate instructions
19492 may be used. @var{opt} is a comma-separated list of options, which may
19493 be preceded by a @code{!} to invert the option:
19498 Enable all estimate instructions.
19501 Enable the default instructions, equivalent to @option{-mrecip}.
19504 Disable all estimate instructions, equivalent to @option{-mno-recip}.
19507 Enable the reciprocal approximation instructions for both
19508 single and double precision.
19511 Enable the single-precision reciprocal approximation instructions.
19514 Enable the double-precision reciprocal approximation instructions.
19517 Enable the reciprocal square root approximation instructions for both
19518 single and double precision.
19521 Enable the single-precision reciprocal square root approximation instructions.
19524 Enable the double-precision reciprocal square root approximation instructions.
19528 So, for example, @option{-mrecip=all,!rsqrtd} enables
19529 all of the reciprocal estimate instructions, except for the
19530 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
19531 which handle the double-precision reciprocal square root calculations.
19533 @item -mrecip-precision
19534 @itemx -mno-recip-precision
19535 @opindex mrecip-precision
19536 Assume (do not assume) that the reciprocal estimate instructions
19537 provide higher-precision estimates than is mandated by the PowerPC
19538 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
19539 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
19540 The double-precision square root estimate instructions are not generated by
19541 default on low-precision machines, since they do not provide an
19542 estimate that converges after three steps.
19544 @item -mveclibabi=@var{type}
19545 @opindex mveclibabi
19546 Specifies the ABI type to use for vectorizing intrinsics using an
19547 external library. The only type supported at present is @samp{mass},
19548 which specifies to use IBM's Mathematical Acceleration Subsystem
19549 (MASS) libraries for vectorizing intrinsics using external libraries.
19550 GCC currently emits calls to @code{acosd2}, @code{acosf4},
19551 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
19552 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
19553 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
19554 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
19555 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
19556 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
19557 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
19558 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
19559 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
19560 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
19561 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
19562 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
19563 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
19564 for power7. Both @option{-ftree-vectorize} and
19565 @option{-funsafe-math-optimizations} must also be enabled. The MASS
19566 libraries must be specified at link time.
19571 Generate (do not generate) the @code{friz} instruction when the
19572 @option{-funsafe-math-optimizations} option is used to optimize
19573 rounding of floating-point values to 64-bit integer and back to floating
19574 point. The @code{friz} instruction does not return the same value if
19575 the floating-point number is too large to fit in an integer.
19577 @item -mpointers-to-nested-functions
19578 @itemx -mno-pointers-to-nested-functions
19579 @opindex mpointers-to-nested-functions
19580 Generate (do not generate) code to load up the static chain register
19581 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
19582 systems where a function pointer points to a 3-word descriptor giving
19583 the function address, TOC value to be loaded in register @code{r2}, and
19584 static chain value to be loaded in register @code{r11}. The
19585 @option{-mpointers-to-nested-functions} is on by default. You cannot
19586 call through pointers to nested functions or pointers
19587 to functions compiled in other languages that use the static chain if
19588 you use the @option{-mno-pointers-to-nested-functions}.
19590 @item -msave-toc-indirect
19591 @itemx -mno-save-toc-indirect
19592 @opindex msave-toc-indirect
19593 Generate (do not generate) code to save the TOC value in the reserved
19594 stack location in the function prologue if the function calls through
19595 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
19596 saved in the prologue, it is saved just before the call through the
19597 pointer. The @option{-mno-save-toc-indirect} option is the default.
19599 @item -mcompat-align-parm
19600 @itemx -mno-compat-align-parm
19601 @opindex mcompat-align-parm
19602 Generate (do not generate) code to pass structure parameters with a
19603 maximum alignment of 64 bits, for compatibility with older versions
19606 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
19607 structure parameter on a 128-bit boundary when that structure contained
19608 a member requiring 128-bit alignment. This is corrected in more
19609 recent versions of GCC. This option may be used to generate code
19610 that is compatible with functions compiled with older versions of
19613 The @option{-mno-compat-align-parm} option is the default.
19617 @subsection RX Options
19620 These command-line options are defined for RX targets:
19623 @item -m64bit-doubles
19624 @itemx -m32bit-doubles
19625 @opindex m64bit-doubles
19626 @opindex m32bit-doubles
19627 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19628 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19629 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
19630 works on 32-bit values, which is why the default is
19631 @option{-m32bit-doubles}.
19637 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
19638 floating-point hardware. The default is enabled for the RX600
19639 series and disabled for the RX200 series.
19641 Floating-point instructions are only generated for 32-bit floating-point
19642 values, however, so the FPU hardware is not used for doubles if the
19643 @option{-m64bit-doubles} option is used.
19645 @emph{Note} If the @option{-fpu} option is enabled then
19646 @option{-funsafe-math-optimizations} is also enabled automatically.
19647 This is because the RX FPU instructions are themselves unsafe.
19649 @item -mcpu=@var{name}
19651 Selects the type of RX CPU to be targeted. Currently three types are
19652 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
19653 the specific @samp{RX610} CPU. The default is @samp{RX600}.
19655 The only difference between @samp{RX600} and @samp{RX610} is that the
19656 @samp{RX610} does not support the @code{MVTIPL} instruction.
19658 The @samp{RX200} series does not have a hardware floating-point unit
19659 and so @option{-nofpu} is enabled by default when this type is
19662 @item -mbig-endian-data
19663 @itemx -mlittle-endian-data
19664 @opindex mbig-endian-data
19665 @opindex mlittle-endian-data
19666 Store data (but not code) in the big-endian format. The default is
19667 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
19670 @item -msmall-data-limit=@var{N}
19671 @opindex msmall-data-limit
19672 Specifies the maximum size in bytes of global and static variables
19673 which can be placed into the small data area. Using the small data
19674 area can lead to smaller and faster code, but the size of area is
19675 limited and it is up to the programmer to ensure that the area does
19676 not overflow. Also when the small data area is used one of the RX's
19677 registers (usually @code{r13}) is reserved for use pointing to this
19678 area, so it is no longer available for use by the compiler. This
19679 could result in slower and/or larger code if variables are pushed onto
19680 the stack instead of being held in this register.
19682 Note, common variables (variables that have not been initialized) and
19683 constants are not placed into the small data area as they are assigned
19684 to other sections in the output executable.
19686 The default value is zero, which disables this feature. Note, this
19687 feature is not enabled by default with higher optimization levels
19688 (@option{-O2} etc) because of the potentially detrimental effects of
19689 reserving a register. It is up to the programmer to experiment and
19690 discover whether this feature is of benefit to their program. See the
19691 description of the @option{-mpid} option for a description of how the
19692 actual register to hold the small data area pointer is chosen.
19698 Use the simulator runtime. The default is to use the libgloss
19699 board-specific runtime.
19701 @item -mas100-syntax
19702 @itemx -mno-as100-syntax
19703 @opindex mas100-syntax
19704 @opindex mno-as100-syntax
19705 When generating assembler output use a syntax that is compatible with
19706 Renesas's AS100 assembler. This syntax can also be handled by the GAS
19707 assembler, but it has some restrictions so it is not generated by default.
19709 @item -mmax-constant-size=@var{N}
19710 @opindex mmax-constant-size
19711 Specifies the maximum size, in bytes, of a constant that can be used as
19712 an operand in a RX instruction. Although the RX instruction set does
19713 allow constants of up to 4 bytes in length to be used in instructions,
19714 a longer value equates to a longer instruction. Thus in some
19715 circumstances it can be beneficial to restrict the size of constants
19716 that are used in instructions. Constants that are too big are instead
19717 placed into a constant pool and referenced via register indirection.
19719 The value @var{N} can be between 0 and 4. A value of 0 (the default)
19720 or 4 means that constants of any size are allowed.
19724 Enable linker relaxation. Linker relaxation is a process whereby the
19725 linker attempts to reduce the size of a program by finding shorter
19726 versions of various instructions. Disabled by default.
19728 @item -mint-register=@var{N}
19729 @opindex mint-register
19730 Specify the number of registers to reserve for fast interrupt handler
19731 functions. The value @var{N} can be between 0 and 4. A value of 1
19732 means that register @code{r13} is reserved for the exclusive use
19733 of fast interrupt handlers. A value of 2 reserves @code{r13} and
19734 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
19735 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
19736 A value of 0, the default, does not reserve any registers.
19738 @item -msave-acc-in-interrupts
19739 @opindex msave-acc-in-interrupts
19740 Specifies that interrupt handler functions should preserve the
19741 accumulator register. This is only necessary if normal code might use
19742 the accumulator register, for example because it performs 64-bit
19743 multiplications. The default is to ignore the accumulator as this
19744 makes the interrupt handlers faster.
19750 Enables the generation of position independent data. When enabled any
19751 access to constant data is done via an offset from a base address
19752 held in a register. This allows the location of constant data to be
19753 determined at run time without requiring the executable to be
19754 relocated, which is a benefit to embedded applications with tight
19755 memory constraints. Data that can be modified is not affected by this
19758 Note, using this feature reserves a register, usually @code{r13}, for
19759 the constant data base address. This can result in slower and/or
19760 larger code, especially in complicated functions.
19762 The actual register chosen to hold the constant data base address
19763 depends upon whether the @option{-msmall-data-limit} and/or the
19764 @option{-mint-register} command-line options are enabled. Starting
19765 with register @code{r13} and proceeding downwards, registers are
19766 allocated first to satisfy the requirements of @option{-mint-register},
19767 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
19768 is possible for the small data area register to be @code{r8} if both
19769 @option{-mint-register=4} and @option{-mpid} are specified on the
19772 By default this feature is not enabled. The default can be restored
19773 via the @option{-mno-pid} command-line option.
19775 @item -mno-warn-multiple-fast-interrupts
19776 @itemx -mwarn-multiple-fast-interrupts
19777 @opindex mno-warn-multiple-fast-interrupts
19778 @opindex mwarn-multiple-fast-interrupts
19779 Prevents GCC from issuing a warning message if it finds more than one
19780 fast interrupt handler when it is compiling a file. The default is to
19781 issue a warning for each extra fast interrupt handler found, as the RX
19782 only supports one such interrupt.
19786 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
19787 has special significance to the RX port when used with the
19788 @code{interrupt} function attribute. This attribute indicates a
19789 function intended to process fast interrupts. GCC ensures
19790 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
19791 and/or @code{r13} and only provided that the normal use of the
19792 corresponding registers have been restricted via the
19793 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
19796 @node S/390 and zSeries Options
19797 @subsection S/390 and zSeries Options
19798 @cindex S/390 and zSeries Options
19800 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
19804 @itemx -msoft-float
19805 @opindex mhard-float
19806 @opindex msoft-float
19807 Use (do not use) the hardware floating-point instructions and registers
19808 for floating-point operations. When @option{-msoft-float} is specified,
19809 functions in @file{libgcc.a} are used to perform floating-point
19810 operations. When @option{-mhard-float} is specified, the compiler
19811 generates IEEE floating-point instructions. This is the default.
19814 @itemx -mno-hard-dfp
19816 @opindex mno-hard-dfp
19817 Use (do not use) the hardware decimal-floating-point instructions for
19818 decimal-floating-point operations. When @option{-mno-hard-dfp} is
19819 specified, functions in @file{libgcc.a} are used to perform
19820 decimal-floating-point operations. When @option{-mhard-dfp} is
19821 specified, the compiler generates decimal-floating-point hardware
19822 instructions. This is the default for @option{-march=z9-ec} or higher.
19824 @item -mlong-double-64
19825 @itemx -mlong-double-128
19826 @opindex mlong-double-64
19827 @opindex mlong-double-128
19828 These switches control the size of @code{long double} type. A size
19829 of 64 bits makes the @code{long double} type equivalent to the @code{double}
19830 type. This is the default.
19833 @itemx -mno-backchain
19834 @opindex mbackchain
19835 @opindex mno-backchain
19836 Store (do not store) the address of the caller's frame as backchain pointer
19837 into the callee's stack frame.
19838 A backchain may be needed to allow debugging using tools that do not understand
19839 DWARF 2 call frame information.
19840 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
19841 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
19842 the backchain is placed into the topmost word of the 96/160 byte register
19845 In general, code compiled with @option{-mbackchain} is call-compatible with
19846 code compiled with @option{-mmo-backchain}; however, use of the backchain
19847 for debugging purposes usually requires that the whole binary is built with
19848 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
19849 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
19850 to build a linux kernel use @option{-msoft-float}.
19852 The default is to not maintain the backchain.
19854 @item -mpacked-stack
19855 @itemx -mno-packed-stack
19856 @opindex mpacked-stack
19857 @opindex mno-packed-stack
19858 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
19859 specified, the compiler uses the all fields of the 96/160 byte register save
19860 area only for their default purpose; unused fields still take up stack space.
19861 When @option{-mpacked-stack} is specified, register save slots are densely
19862 packed at the top of the register save area; unused space is reused for other
19863 purposes, allowing for more efficient use of the available stack space.
19864 However, when @option{-mbackchain} is also in effect, the topmost word of
19865 the save area is always used to store the backchain, and the return address
19866 register is always saved two words below the backchain.
19868 As long as the stack frame backchain is not used, code generated with
19869 @option{-mpacked-stack} is call-compatible with code generated with
19870 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
19871 S/390 or zSeries generated code that uses the stack frame backchain at run
19872 time, not just for debugging purposes. Such code is not call-compatible
19873 with code compiled with @option{-mpacked-stack}. Also, note that the
19874 combination of @option{-mbackchain},
19875 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
19876 to build a linux kernel use @option{-msoft-float}.
19878 The default is to not use the packed stack layout.
19881 @itemx -mno-small-exec
19882 @opindex msmall-exec
19883 @opindex mno-small-exec
19884 Generate (or do not generate) code using the @code{bras} instruction
19885 to do subroutine calls.
19886 This only works reliably if the total executable size does not
19887 exceed 64k. The default is to use the @code{basr} instruction instead,
19888 which does not have this limitation.
19894 When @option{-m31} is specified, generate code compliant to the
19895 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
19896 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
19897 particular to generate 64-bit instructions. For the @samp{s390}
19898 targets, the default is @option{-m31}, while the @samp{s390x}
19899 targets default to @option{-m64}.
19905 When @option{-mzarch} is specified, generate code using the
19906 instructions available on z/Architecture.
19907 When @option{-mesa} is specified, generate code using the
19908 instructions available on ESA/390. Note that @option{-mesa} is
19909 not possible with @option{-m64}.
19910 When generating code compliant to the GNU/Linux for S/390 ABI,
19911 the default is @option{-mesa}. When generating code compliant
19912 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
19918 Generate (or do not generate) code using the @code{mvcle} instruction
19919 to perform block moves. When @option{-mno-mvcle} is specified,
19920 use a @code{mvc} loop instead. This is the default unless optimizing for
19927 Print (or do not print) additional debug information when compiling.
19928 The default is to not print debug information.
19930 @item -march=@var{cpu-type}
19932 Generate code that runs on @var{cpu-type}, which is the name of a system
19933 representing a certain processor type. Possible values for
19934 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
19935 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
19936 When generating code using the instructions available on z/Architecture,
19937 the default is @option{-march=z900}. Otherwise, the default is
19938 @option{-march=g5}.
19940 @item -mtune=@var{cpu-type}
19942 Tune to @var{cpu-type} everything applicable about the generated code,
19943 except for the ABI and the set of available instructions.
19944 The list of @var{cpu-type} values is the same as for @option{-march}.
19945 The default is the value used for @option{-march}.
19948 @itemx -mno-tpf-trace
19949 @opindex mtpf-trace
19950 @opindex mno-tpf-trace
19951 Generate code that adds (does not add) in TPF OS specific branches to trace
19952 routines in the operating system. This option is off by default, even
19953 when compiling for the TPF OS@.
19956 @itemx -mno-fused-madd
19957 @opindex mfused-madd
19958 @opindex mno-fused-madd
19959 Generate code that uses (does not use) the floating-point multiply and
19960 accumulate instructions. These instructions are generated by default if
19961 hardware floating point is used.
19963 @item -mwarn-framesize=@var{framesize}
19964 @opindex mwarn-framesize
19965 Emit a warning if the current function exceeds the given frame size. Because
19966 this is a compile-time check it doesn't need to be a real problem when the program
19967 runs. It is intended to identify functions that most probably cause
19968 a stack overflow. It is useful to be used in an environment with limited stack
19969 size e.g.@: the linux kernel.
19971 @item -mwarn-dynamicstack
19972 @opindex mwarn-dynamicstack
19973 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
19974 arrays. This is generally a bad idea with a limited stack size.
19976 @item -mstack-guard=@var{stack-guard}
19977 @itemx -mstack-size=@var{stack-size}
19978 @opindex mstack-guard
19979 @opindex mstack-size
19980 If these options are provided the S/390 back end emits additional instructions in
19981 the function prologue that trigger a trap if the stack size is @var{stack-guard}
19982 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
19983 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
19984 the frame size of the compiled function is chosen.
19985 These options are intended to be used to help debugging stack overflow problems.
19986 The additionally emitted code causes only little overhead and hence can also be
19987 used in production-like systems without greater performance degradation. The given
19988 values have to be exact powers of 2 and @var{stack-size} has to be greater than
19989 @var{stack-guard} without exceeding 64k.
19990 In order to be efficient the extra code makes the assumption that the stack starts
19991 at an address aligned to the value given by @var{stack-size}.
19992 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
19994 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
19996 If the hotpatch option is enabled, a ``hot-patching'' function
19997 prologue is generated for all functions in the compilation unit.
19998 The funtion label is prepended with the given number of two-byte
19999 Nop instructions (@var{pre-halfwords}, maximum 1000000). After
20000 the label, 2 * @var{post-halfwords} bytes are appended, using the
20001 larges nop like instructions the architecture allows (maximum
20004 If both arguments are zero, hotpatching is disabled.
20006 This option can be overridden for individual functions with the
20007 @code{hotpatch} attribute.
20010 @node Score Options
20011 @subsection Score Options
20012 @cindex Score Options
20014 These options are defined for Score implementations:
20019 Compile code for big-endian mode. This is the default.
20023 Compile code for little-endian mode.
20027 Disable generation of @code{bcnz} instructions.
20031 Enable generation of unaligned load and store instructions.
20035 Enable the use of multiply-accumulate instructions. Disabled by default.
20039 Specify the SCORE5 as the target architecture.
20043 Specify the SCORE5U of the target architecture.
20047 Specify the SCORE7 as the target architecture. This is the default.
20051 Specify the SCORE7D as the target architecture.
20055 @subsection SH Options
20057 These @samp{-m} options are defined for the SH implementations:
20062 Generate code for the SH1.
20066 Generate code for the SH2.
20069 Generate code for the SH2e.
20073 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20074 that the floating-point unit is not used.
20076 @item -m2a-single-only
20077 @opindex m2a-single-only
20078 Generate code for the SH2a-FPU, in such a way that no double-precision
20079 floating-point operations are used.
20082 @opindex m2a-single
20083 Generate code for the SH2a-FPU assuming the floating-point unit is in
20084 single-precision mode by default.
20088 Generate code for the SH2a-FPU assuming the floating-point unit is in
20089 double-precision mode by default.
20093 Generate code for the SH3.
20097 Generate code for the SH3e.
20101 Generate code for the SH4 without a floating-point unit.
20103 @item -m4-single-only
20104 @opindex m4-single-only
20105 Generate code for the SH4 with a floating-point unit that only
20106 supports single-precision arithmetic.
20110 Generate code for the SH4 assuming the floating-point unit is in
20111 single-precision mode by default.
20115 Generate code for the SH4.
20119 Generate code for SH4-100.
20121 @item -m4-100-nofpu
20122 @opindex m4-100-nofpu
20123 Generate code for SH4-100 in such a way that the
20124 floating-point unit is not used.
20126 @item -m4-100-single
20127 @opindex m4-100-single
20128 Generate code for SH4-100 assuming the floating-point unit is in
20129 single-precision mode by default.
20131 @item -m4-100-single-only
20132 @opindex m4-100-single-only
20133 Generate code for SH4-100 in such a way that no double-precision
20134 floating-point operations are used.
20138 Generate code for SH4-200.
20140 @item -m4-200-nofpu
20141 @opindex m4-200-nofpu
20142 Generate code for SH4-200 without in such a way that the
20143 floating-point unit is not used.
20145 @item -m4-200-single
20146 @opindex m4-200-single
20147 Generate code for SH4-200 assuming the floating-point unit is in
20148 single-precision mode by default.
20150 @item -m4-200-single-only
20151 @opindex m4-200-single-only
20152 Generate code for SH4-200 in such a way that no double-precision
20153 floating-point operations are used.
20157 Generate code for SH4-300.
20159 @item -m4-300-nofpu
20160 @opindex m4-300-nofpu
20161 Generate code for SH4-300 without in such a way that the
20162 floating-point unit is not used.
20164 @item -m4-300-single
20165 @opindex m4-300-single
20166 Generate code for SH4-300 in such a way that no double-precision
20167 floating-point operations are used.
20169 @item -m4-300-single-only
20170 @opindex m4-300-single-only
20171 Generate code for SH4-300 in such a way that no double-precision
20172 floating-point operations are used.
20176 Generate code for SH4-340 (no MMU, no FPU).
20180 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
20185 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20186 floating-point unit is not used.
20188 @item -m4a-single-only
20189 @opindex m4a-single-only
20190 Generate code for the SH4a, in such a way that no double-precision
20191 floating-point operations are used.
20194 @opindex m4a-single
20195 Generate code for the SH4a assuming the floating-point unit is in
20196 single-precision mode by default.
20200 Generate code for the SH4a.
20204 Same as @option{-m4a-nofpu}, except that it implicitly passes
20205 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20206 instructions at the moment.
20209 @opindex m5-32media
20210 Generate 32-bit code for SHmedia.
20212 @item -m5-32media-nofpu
20213 @opindex m5-32media-nofpu
20214 Generate 32-bit code for SHmedia in such a way that the
20215 floating-point unit is not used.
20218 @opindex m5-64media
20219 Generate 64-bit code for SHmedia.
20221 @item -m5-64media-nofpu
20222 @opindex m5-64media-nofpu
20223 Generate 64-bit code for SHmedia in such a way that the
20224 floating-point unit is not used.
20227 @opindex m5-compact
20228 Generate code for SHcompact.
20230 @item -m5-compact-nofpu
20231 @opindex m5-compact-nofpu
20232 Generate code for SHcompact in such a way that the
20233 floating-point unit is not used.
20237 Compile code for the processor in big-endian mode.
20241 Compile code for the processor in little-endian mode.
20245 Align doubles at 64-bit boundaries. Note that this changes the calling
20246 conventions, and thus some functions from the standard C library do
20247 not work unless you recompile it first with @option{-mdalign}.
20251 Shorten some address references at link time, when possible; uses the
20252 linker option @option{-relax}.
20256 Use 32-bit offsets in @code{switch} tables. The default is to use
20261 Enable the use of bit manipulation instructions on SH2A.
20265 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20266 alignment constraints.
20270 Comply with the calling conventions defined by Renesas.
20273 @opindex mno-renesas
20274 Comply with the calling conventions defined for GCC before the Renesas
20275 conventions were available. This option is the default for all
20276 targets of the SH toolchain.
20279 @opindex mnomacsave
20280 Mark the @code{MAC} register as call-clobbered, even if
20281 @option{-mrenesas} is given.
20287 Control the IEEE compliance of floating-point comparisons, which affects the
20288 handling of cases where the result of a comparison is unordered. By default
20289 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20290 enabled @option{-mno-ieee} is implicitly set, which results in faster
20291 floating-point greater-equal and less-equal comparisons. The implcit settings
20292 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20294 @item -minline-ic_invalidate
20295 @opindex minline-ic_invalidate
20296 Inline code to invalidate instruction cache entries after setting up
20297 nested function trampolines.
20298 This option has no effect if @option{-musermode} is in effect and the selected
20299 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20301 If the selected code generation option does not allow the use of the @code{icbi}
20302 instruction, and @option{-musermode} is not in effect, the inlined code
20303 manipulates the instruction cache address array directly with an associative
20304 write. This not only requires privileged mode at run time, but it also
20305 fails if the cache line had been mapped via the TLB and has become unmapped.
20309 Dump instruction size and location in the assembly code.
20312 @opindex mpadstruct
20313 This option is deprecated. It pads structures to multiple of 4 bytes,
20314 which is incompatible with the SH ABI@.
20316 @item -matomic-model=@var{model}
20317 @opindex matomic-model=@var{model}
20318 Sets the model of atomic operations and additional parameters as a comma
20319 separated list. For details on the atomic built-in functions see
20320 @ref{__atomic Builtins}. The following models and parameters are supported:
20325 Disable compiler generated atomic sequences and emit library calls for atomic
20326 operations. This is the default if the target is not @code{sh*-*-linux*}.
20329 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20330 built-in functions. The generated atomic sequences require additional support
20331 from the interrupt/exception handling code of the system and are only suitable
20332 for SH3* and SH4* single-core systems. This option is enabled by default when
20333 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
20334 this option also partially utilizes the hardware atomic instructions
20335 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20336 @samp{strict} is specified.
20339 Generate software atomic sequences that use a variable in the thread control
20340 block. This is a variation of the gUSA sequences which can also be used on
20341 SH1* and SH2* targets. The generated atomic sequences require additional
20342 support from the interrupt/exception handling code of the system and are only
20343 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20344 parameter has to be specified as well.
20347 Generate software atomic sequences that temporarily disable interrupts by
20348 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20349 in privileged mode and is only suitable for single-core systems. Additional
20350 support from the interrupt/exception handling code of the system is not
20351 required. This model is enabled by default when the target is
20352 @code{sh*-*-linux*} and SH1* or SH2*.
20355 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20356 instructions only. This is only available on SH4A and is suitable for
20357 multi-core systems. Since the hardware instructions support only 32 bit atomic
20358 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20359 Code compiled with this option is also compatible with other software
20360 atomic model interrupt/exception handling systems if executed on an SH4A
20361 system. Additional support from the interrupt/exception handling code of the
20362 system is not required for this model.
20365 This parameter specifies the offset in bytes of the variable in the thread
20366 control block structure that should be used by the generated atomic sequences
20367 when the @samp{soft-tcb} model has been selected. For other models this
20368 parameter is ignored. The specified value must be an integer multiple of four
20369 and in the range 0-1020.
20372 This parameter prevents mixed usage of multiple atomic models, even if they
20373 are compatible, and makes the compiler generate atomic sequences of the
20374 specified model only.
20380 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20381 Notice that depending on the particular hardware and software configuration
20382 this can degrade overall performance due to the operand cache line flushes
20383 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20384 processors the @code{tas.b} instruction must be used with caution since it
20385 can result in data corruption for certain cache configurations.
20388 @opindex mprefergot
20389 When generating position-independent code, emit function calls using
20390 the Global Offset Table instead of the Procedure Linkage Table.
20393 @itemx -mno-usermode
20395 @opindex mno-usermode
20396 Don't allow (allow) the compiler generating privileged mode code. Specifying
20397 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
20398 inlined code would not work in user mode. @option{-musermode} is the default
20399 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
20400 @option{-musermode} has no effect, since there is no user mode.
20402 @item -multcost=@var{number}
20403 @opindex multcost=@var{number}
20404 Set the cost to assume for a multiply insn.
20406 @item -mdiv=@var{strategy}
20407 @opindex mdiv=@var{strategy}
20408 Set the division strategy to be used for integer division operations.
20409 For SHmedia @var{strategy} can be one of:
20414 Performs the operation in floating point. This has a very high latency,
20415 but needs only a few instructions, so it might be a good choice if
20416 your code has enough easily-exploitable ILP to allow the compiler to
20417 schedule the floating-point instructions together with other instructions.
20418 Division by zero causes a floating-point exception.
20421 Uses integer operations to calculate the inverse of the divisor,
20422 and then multiplies the dividend with the inverse. This strategy allows
20423 CSE and hoisting of the inverse calculation. Division by zero calculates
20424 an unspecified result, but does not trap.
20427 A variant of @samp{inv} where, if no CSE or hoisting opportunities
20428 have been found, or if the entire operation has been hoisted to the same
20429 place, the last stages of the inverse calculation are intertwined with the
20430 final multiply to reduce the overall latency, at the expense of using a few
20431 more instructions, and thus offering fewer scheduling opportunities with
20435 Calls a library function that usually implements the @samp{inv:minlat}
20437 This gives high code density for @code{m5-*media-nofpu} compilations.
20440 Uses a different entry point of the same library function, where it
20441 assumes that a pointer to a lookup table has already been set up, which
20442 exposes the pointer load to CSE and code hoisting optimizations.
20447 Use the @samp{inv} algorithm for initial
20448 code generation, but if the code stays unoptimized, revert to the @samp{call},
20449 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
20450 potentially-trapping side effect of division by zero is carried by a
20451 separate instruction, so it is possible that all the integer instructions
20452 are hoisted out, but the marker for the side effect stays where it is.
20453 A recombination to floating-point operations or a call is not possible
20458 Variants of the @samp{inv:minlat} strategy. In the case
20459 that the inverse calculation is not separated from the multiply, they speed
20460 up division where the dividend fits into 20 bits (plus sign where applicable)
20461 by inserting a test to skip a number of operations in this case; this test
20462 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
20463 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
20467 For targets other than SHmedia @var{strategy} can be one of:
20472 Calls a library function that uses the single-step division instruction
20473 @code{div1} to perform the operation. Division by zero calculates an
20474 unspecified result and does not trap. This is the default except for SH4,
20475 SH2A and SHcompact.
20478 Calls a library function that performs the operation in double precision
20479 floating point. Division by zero causes a floating-point exception. This is
20480 the default for SHcompact with FPU. Specifying this for targets that do not
20481 have a double precision FPU defaults to @code{call-div1}.
20484 Calls a library function that uses a lookup table for small divisors and
20485 the @code{div1} instruction with case distinction for larger divisors. Division
20486 by zero calculates an unspecified result and does not trap. This is the default
20487 for SH4. Specifying this for targets that do not have dynamic shift
20488 instructions defaults to @code{call-div1}.
20492 When a division strategy has not been specified the default strategy is
20493 selected based on the current target. For SH2A the default strategy is to
20494 use the @code{divs} and @code{divu} instructions instead of library function
20497 @item -maccumulate-outgoing-args
20498 @opindex maccumulate-outgoing-args
20499 Reserve space once for outgoing arguments in the function prologue rather
20500 than around each call. Generally beneficial for performance and size. Also
20501 needed for unwinding to avoid changing the stack frame around conditional code.
20503 @item -mdivsi3_libfunc=@var{name}
20504 @opindex mdivsi3_libfunc=@var{name}
20505 Set the name of the library function used for 32-bit signed division to
20507 This only affects the name used in the @samp{call} and @samp{inv:call}
20508 division strategies, and the compiler still expects the same
20509 sets of input/output/clobbered registers as if this option were not present.
20511 @item -mfixed-range=@var{register-range}
20512 @opindex mfixed-range
20513 Generate code treating the given register range as fixed registers.
20514 A fixed register is one that the register allocator can not use. This is
20515 useful when compiling kernel code. A register range is specified as
20516 two registers separated by a dash. Multiple register ranges can be
20517 specified separated by a comma.
20519 @item -mindexed-addressing
20520 @opindex mindexed-addressing
20521 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
20522 This is only safe if the hardware and/or OS implement 32-bit wrap-around
20523 semantics for the indexed addressing mode. The architecture allows the
20524 implementation of processors with 64-bit MMU, which the OS could use to
20525 get 32-bit addressing, but since no current hardware implementation supports
20526 this or any other way to make the indexed addressing mode safe to use in
20527 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
20529 @item -mgettrcost=@var{number}
20530 @opindex mgettrcost=@var{number}
20531 Set the cost assumed for the @code{gettr} instruction to @var{number}.
20532 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
20536 Assume @code{pt*} instructions won't trap. This generally generates
20537 better-scheduled code, but is unsafe on current hardware.
20538 The current architecture
20539 definition says that @code{ptabs} and @code{ptrel} trap when the target
20541 This has the unintentional effect of making it unsafe to schedule these
20542 instructions before a branch, or hoist them out of a loop. For example,
20543 @code{__do_global_ctors}, a part of @file{libgcc}
20544 that runs constructors at program
20545 startup, calls functions in a list which is delimited by @minus{}1. With the
20546 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
20547 That means that all the constructors run a bit more quickly, but when
20548 the loop comes to the end of the list, the program crashes because @code{ptabs}
20549 loads @minus{}1 into a target register.
20551 Since this option is unsafe for any
20552 hardware implementing the current architecture specification, the default
20553 is @option{-mno-pt-fixed}. Unless specified explicitly with
20554 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
20555 this deters register allocation from using target registers for storing
20558 @item -minvalid-symbols
20559 @opindex minvalid-symbols
20560 Assume symbols might be invalid. Ordinary function symbols generated by
20561 the compiler are always valid to load with
20562 @code{movi}/@code{shori}/@code{ptabs} or
20563 @code{movi}/@code{shori}/@code{ptrel},
20564 but with assembler and/or linker tricks it is possible
20565 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
20566 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
20567 It prevents cross-basic-block CSE, hoisting and most scheduling
20568 of symbol loads. The default is @option{-mno-invalid-symbols}.
20570 @item -mbranch-cost=@var{num}
20571 @opindex mbranch-cost=@var{num}
20572 Assume @var{num} to be the cost for a branch instruction. Higher numbers
20573 make the compiler try to generate more branch-free code if possible.
20574 If not specified the value is selected depending on the processor type that
20575 is being compiled for.
20578 @itemx -mno-zdcbranch
20579 @opindex mzdcbranch
20580 @opindex mno-zdcbranch
20581 Assume (do not assume) that zero displacement conditional branch instructions
20582 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
20583 compiler prefers zero displacement branch code sequences. This is
20584 enabled by default when generating code for SH4 and SH4A. It can be explicitly
20585 disabled by specifying @option{-mno-zdcbranch}.
20588 @itemx -mno-fused-madd
20589 @opindex mfused-madd
20590 @opindex mno-fused-madd
20591 Generate code that uses (does not use) the floating-point multiply and
20592 accumulate instructions. These instructions are generated by default
20593 if hardware floating point is used. The machine-dependent
20594 @option{-mfused-madd} option is now mapped to the machine-independent
20595 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20596 mapped to @option{-ffp-contract=off}.
20602 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
20603 and cosine approximations. The option @option{-mfsca} must be used in
20604 combination with @option{-funsafe-math-optimizations}. It is enabled by default
20605 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
20606 approximations even if @option{-funsafe-math-optimizations} is in effect.
20612 Allow or disallow the compiler to emit the @code{fsrra} instruction for
20613 reciprocal square root approximations. The option @option{-mfsrra} must be used
20614 in combination with @option{-funsafe-math-optimizations} and
20615 @option{-ffinite-math-only}. It is enabled by default when generating code for
20616 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
20617 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
20620 @item -mpretend-cmove
20621 @opindex mpretend-cmove
20622 Prefer zero-displacement conditional branches for conditional move instruction
20623 patterns. This can result in faster code on the SH4 processor.
20627 @node Solaris 2 Options
20628 @subsection Solaris 2 Options
20629 @cindex Solaris 2 options
20631 These @samp{-m} options are supported on Solaris 2:
20634 @item -mclear-hwcap
20635 @opindex mclear-hwcap
20636 @option{-mclear-hwcap} tells the compiler to remove the hardware
20637 capabilities generated by the Solaris assembler. This is only necessary
20638 when object files use ISA extensions not supported by the current
20639 machine, but check at runtime whether or not to use them.
20641 @item -mimpure-text
20642 @opindex mimpure-text
20643 @option{-mimpure-text}, used in addition to @option{-shared}, tells
20644 the compiler to not pass @option{-z text} to the linker when linking a
20645 shared object. Using this option, you can link position-dependent
20646 code into a shared object.
20648 @option{-mimpure-text} suppresses the ``relocations remain against
20649 allocatable but non-writable sections'' linker error message.
20650 However, the necessary relocations trigger copy-on-write, and the
20651 shared object is not actually shared across processes. Instead of
20652 using @option{-mimpure-text}, you should compile all source code with
20653 @option{-fpic} or @option{-fPIC}.
20657 These switches are supported in addition to the above on Solaris 2:
20662 Add support for multithreading using the POSIX threads library. This
20663 option sets flags for both the preprocessor and linker. This option does
20664 not affect the thread safety of object code produced by the compiler or
20665 that of libraries supplied with it.
20669 This is a synonym for @option{-pthreads}.
20672 @node SPARC Options
20673 @subsection SPARC Options
20674 @cindex SPARC options
20676 These @samp{-m} options are supported on the SPARC:
20679 @item -mno-app-regs
20681 @opindex mno-app-regs
20683 Specify @option{-mapp-regs} to generate output using the global registers
20684 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
20685 global register 1, each global register 2 through 4 is then treated as an
20686 allocable register that is clobbered by function calls. This is the default.
20688 To be fully SVR4 ABI-compliant at the cost of some performance loss,
20689 specify @option{-mno-app-regs}. You should compile libraries and system
20690 software with this option.
20696 With @option{-mflat}, the compiler does not generate save/restore instructions
20697 and uses a ``flat'' or single register window model. This model is compatible
20698 with the regular register window model. The local registers and the input
20699 registers (0--5) are still treated as ``call-saved'' registers and are
20700 saved on the stack as needed.
20702 With @option{-mno-flat} (the default), the compiler generates save/restore
20703 instructions (except for leaf functions). This is the normal operating mode.
20706 @itemx -mhard-float
20708 @opindex mhard-float
20709 Generate output containing floating-point instructions. This is the
20713 @itemx -msoft-float
20715 @opindex msoft-float
20716 Generate output containing library calls for floating point.
20717 @strong{Warning:} the requisite libraries are not available for all SPARC
20718 targets. Normally the facilities of the machine's usual C compiler are
20719 used, but this cannot be done directly in cross-compilation. You must make
20720 your own arrangements to provide suitable library functions for
20721 cross-compilation. The embedded targets @samp{sparc-*-aout} and
20722 @samp{sparclite-*-*} do provide software floating-point support.
20724 @option{-msoft-float} changes the calling convention in the output file;
20725 therefore, it is only useful if you compile @emph{all} of a program with
20726 this option. In particular, you need to compile @file{libgcc.a}, the
20727 library that comes with GCC, with @option{-msoft-float} in order for
20730 @item -mhard-quad-float
20731 @opindex mhard-quad-float
20732 Generate output containing quad-word (long double) floating-point
20735 @item -msoft-quad-float
20736 @opindex msoft-quad-float
20737 Generate output containing library calls for quad-word (long double)
20738 floating-point instructions. The functions called are those specified
20739 in the SPARC ABI@. This is the default.
20741 As of this writing, there are no SPARC implementations that have hardware
20742 support for the quad-word floating-point instructions. They all invoke
20743 a trap handler for one of these instructions, and then the trap handler
20744 emulates the effect of the instruction. Because of the trap handler overhead,
20745 this is much slower than calling the ABI library routines. Thus the
20746 @option{-msoft-quad-float} option is the default.
20748 @item -mno-unaligned-doubles
20749 @itemx -munaligned-doubles
20750 @opindex mno-unaligned-doubles
20751 @opindex munaligned-doubles
20752 Assume that doubles have 8-byte alignment. This is the default.
20754 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
20755 alignment only if they are contained in another type, or if they have an
20756 absolute address. Otherwise, it assumes they have 4-byte alignment.
20757 Specifying this option avoids some rare compatibility problems with code
20758 generated by other compilers. It is not the default because it results
20759 in a performance loss, especially for floating-point code.
20762 @itemx -mno-user-mode
20763 @opindex muser-mode
20764 @opindex mno-user-mode
20765 Do not generate code that can only run in supervisor mode. This is relevant
20766 only for the @code{casa} instruction emitted for the LEON3 processor. The
20767 default is @option{-mno-user-mode}.
20769 @item -mno-faster-structs
20770 @itemx -mfaster-structs
20771 @opindex mno-faster-structs
20772 @opindex mfaster-structs
20773 With @option{-mfaster-structs}, the compiler assumes that structures
20774 should have 8-byte alignment. This enables the use of pairs of
20775 @code{ldd} and @code{std} instructions for copies in structure
20776 assignment, in place of twice as many @code{ld} and @code{st} pairs.
20777 However, the use of this changed alignment directly violates the SPARC
20778 ABI@. Thus, it's intended only for use on targets where the developer
20779 acknowledges that their resulting code is not directly in line with
20780 the rules of the ABI@.
20782 @item -mcpu=@var{cpu_type}
20784 Set the instruction set, register set, and instruction scheduling parameters
20785 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
20786 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
20787 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
20788 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
20789 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
20790 @samp{niagara3} and @samp{niagara4}.
20792 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
20793 which selects the best architecture option for the host processor.
20794 @option{-mcpu=native} has no effect if GCC does not recognize
20797 Default instruction scheduling parameters are used for values that select
20798 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
20799 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
20801 Here is a list of each supported architecture and their supported
20809 supersparc, hypersparc, leon, leon3
20812 f930, f934, sparclite86x
20818 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
20821 By default (unless configured otherwise), GCC generates code for the V7
20822 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
20823 additionally optimizes it for the Cypress CY7C602 chip, as used in the
20824 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
20825 SPARCStation 1, 2, IPX etc.
20827 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
20828 architecture. The only difference from V7 code is that the compiler emits
20829 the integer multiply and integer divide instructions which exist in SPARC-V8
20830 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
20831 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
20834 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
20835 the SPARC architecture. This adds the integer multiply, integer divide step
20836 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
20837 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
20838 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
20839 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
20840 MB86934 chip, which is the more recent SPARClite with FPU@.
20842 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
20843 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
20844 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
20845 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
20846 optimizes it for the TEMIC SPARClet chip.
20848 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
20849 architecture. This adds 64-bit integer and floating-point move instructions,
20850 3 additional floating-point condition code registers and conditional move
20851 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
20852 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
20853 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
20854 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
20855 @option{-mcpu=niagara}, the compiler additionally optimizes it for
20856 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
20857 additionally optimizes it for Sun UltraSPARC T2 chips. With
20858 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
20859 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
20860 additionally optimizes it for Sun UltraSPARC T4 chips.
20862 @item -mtune=@var{cpu_type}
20864 Set the instruction scheduling parameters for machine type
20865 @var{cpu_type}, but do not set the instruction set or register set that the
20866 option @option{-mcpu=@var{cpu_type}} does.
20868 The same values for @option{-mcpu=@var{cpu_type}} can be used for
20869 @option{-mtune=@var{cpu_type}}, but the only useful values are those
20870 that select a particular CPU implementation. Those are @samp{cypress},
20871 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
20872 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
20873 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
20874 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
20875 toolchains, @samp{native} can also be used.
20880 @opindex mno-v8plus
20881 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
20882 difference from the V8 ABI is that the global and out registers are
20883 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
20884 mode for all SPARC-V9 processors.
20890 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
20891 Visual Instruction Set extensions. The default is @option{-mno-vis}.
20897 With @option{-mvis2}, GCC generates code that takes advantage of
20898 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
20899 default is @option{-mvis2} when targeting a cpu that supports such
20900 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
20901 also sets @option{-mvis}.
20907 With @option{-mvis3}, GCC generates code that takes advantage of
20908 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
20909 default is @option{-mvis3} when targeting a cpu that supports such
20910 instructions, such as niagara-3 and later. Setting @option{-mvis3}
20911 also sets @option{-mvis2} and @option{-mvis}.
20916 @opindex mno-cbcond
20917 With @option{-mcbcond}, GCC generates code that takes advantage of
20918 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
20919 The default is @option{-mcbcond} when targeting a cpu that supports such
20920 instructions, such as niagara-4 and later.
20926 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
20927 population count instruction. The default is @option{-mpopc}
20928 when targeting a cpu that supports such instructions, such as Niagara-2 and
20935 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
20936 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
20937 when targeting a cpu that supports such instructions, such as Niagara-3 and
20941 @opindex mfix-at697f
20942 Enable the documented workaround for the single erratum of the Atmel AT697F
20943 processor (which corresponds to erratum #13 of the AT697E processor).
20946 @opindex mfix-ut699
20947 Enable the documented workarounds for the floating-point errata and the data
20948 cache nullify errata of the UT699 processor.
20951 These @samp{-m} options are supported in addition to the above
20952 on SPARC-V9 processors in 64-bit environments:
20959 Generate code for a 32-bit or 64-bit environment.
20960 The 32-bit environment sets int, long and pointer to 32 bits.
20961 The 64-bit environment sets int to 32 bits and long and pointer
20964 @item -mcmodel=@var{which}
20966 Set the code model to one of
20970 The Medium/Low code model: 64-bit addresses, programs
20971 must be linked in the low 32 bits of memory. Programs can be statically
20972 or dynamically linked.
20975 The Medium/Middle code model: 64-bit addresses, programs
20976 must be linked in the low 44 bits of memory, the text and data segments must
20977 be less than 2GB in size and the data segment must be located within 2GB of
20981 The Medium/Anywhere code model: 64-bit addresses, programs
20982 may be linked anywhere in memory, the text and data segments must be less
20983 than 2GB in size and the data segment must be located within 2GB of the
20987 The Medium/Anywhere code model for embedded systems:
20988 64-bit addresses, the text and data segments must be less than 2GB in
20989 size, both starting anywhere in memory (determined at link time). The
20990 global register %g4 points to the base of the data segment. Programs
20991 are statically linked and PIC is not supported.
20994 @item -mmemory-model=@var{mem-model}
20995 @opindex mmemory-model
20996 Set the memory model in force on the processor to one of
21000 The default memory model for the processor and operating system.
21003 Relaxed Memory Order
21006 Partial Store Order
21012 Sequential Consistency
21015 These memory models are formally defined in Appendix D of the Sparc V9
21016 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21019 @itemx -mno-stack-bias
21020 @opindex mstack-bias
21021 @opindex mno-stack-bias
21022 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21023 frame pointer if present, are offset by @minus{}2047 which must be added back
21024 when making stack frame references. This is the default in 64-bit mode.
21025 Otherwise, assume no such offset is present.
21029 @subsection SPU Options
21030 @cindex SPU options
21032 These @samp{-m} options are supported on the SPU:
21036 @itemx -merror-reloc
21037 @opindex mwarn-reloc
21038 @opindex merror-reloc
21040 The loader for SPU does not handle dynamic relocations. By default, GCC
21041 gives an error when it generates code that requires a dynamic
21042 relocation. @option{-mno-error-reloc} disables the error,
21043 @option{-mwarn-reloc} generates a warning instead.
21046 @itemx -munsafe-dma
21048 @opindex munsafe-dma
21050 Instructions that initiate or test completion of DMA must not be
21051 reordered with respect to loads and stores of the memory that is being
21053 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21054 memory accesses, but that can lead to inefficient code in places where the
21055 memory is known to not change. Rather than mark the memory as volatile,
21056 you can use @option{-msafe-dma} to tell the compiler to treat
21057 the DMA instructions as potentially affecting all memory.
21059 @item -mbranch-hints
21060 @opindex mbranch-hints
21062 By default, GCC generates a branch hint instruction to avoid
21063 pipeline stalls for always-taken or probably-taken branches. A hint
21064 is not generated closer than 8 instructions away from its branch.
21065 There is little reason to disable them, except for debugging purposes,
21066 or to make an object a little bit smaller.
21070 @opindex msmall-mem
21071 @opindex mlarge-mem
21073 By default, GCC generates code assuming that addresses are never larger
21074 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21075 a full 32-bit address.
21080 By default, GCC links against startup code that assumes the SPU-style
21081 main function interface (which has an unconventional parameter list).
21082 With @option{-mstdmain}, GCC links your program against startup
21083 code that assumes a C99-style interface to @code{main}, including a
21084 local copy of @code{argv} strings.
21086 @item -mfixed-range=@var{register-range}
21087 @opindex mfixed-range
21088 Generate code treating the given register range as fixed registers.
21089 A fixed register is one that the register allocator cannot use. This is
21090 useful when compiling kernel code. A register range is specified as
21091 two registers separated by a dash. Multiple register ranges can be
21092 specified separated by a comma.
21098 Compile code assuming that pointers to the PPU address space accessed
21099 via the @code{__ea} named address space qualifier are either 32 or 64
21100 bits wide. The default is 32 bits. As this is an ABI-changing option,
21101 all object code in an executable must be compiled with the same setting.
21103 @item -maddress-space-conversion
21104 @itemx -mno-address-space-conversion
21105 @opindex maddress-space-conversion
21106 @opindex mno-address-space-conversion
21107 Allow/disallow treating the @code{__ea} address space as superset
21108 of the generic address space. This enables explicit type casts
21109 between @code{__ea} and generic pointer as well as implicit
21110 conversions of generic pointers to @code{__ea} pointers. The
21111 default is to allow address space pointer conversions.
21113 @item -mcache-size=@var{cache-size}
21114 @opindex mcache-size
21115 This option controls the version of libgcc that the compiler links to an
21116 executable and selects a software-managed cache for accessing variables
21117 in the @code{__ea} address space with a particular cache size. Possible
21118 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21119 and @samp{128}. The default cache size is 64KB.
21121 @item -matomic-updates
21122 @itemx -mno-atomic-updates
21123 @opindex matomic-updates
21124 @opindex mno-atomic-updates
21125 This option controls the version of libgcc that the compiler links to an
21126 executable and selects whether atomic updates to the software-managed
21127 cache of PPU-side variables are used. If you use atomic updates, changes
21128 to a PPU variable from SPU code using the @code{__ea} named address space
21129 qualifier do not interfere with changes to other PPU variables residing
21130 in the same cache line from PPU code. If you do not use atomic updates,
21131 such interference may occur; however, writing back cache lines is
21132 more efficient. The default behavior is to use atomic updates.
21135 @itemx -mdual-nops=@var{n}
21136 @opindex mdual-nops
21137 By default, GCC inserts nops to increase dual issue when it expects
21138 it to increase performance. @var{n} can be a value from 0 to 10. A
21139 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21140 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21142 @item -mhint-max-nops=@var{n}
21143 @opindex mhint-max-nops
21144 Maximum number of nops to insert for a branch hint. A branch hint must
21145 be at least 8 instructions away from the branch it is affecting. GCC
21146 inserts up to @var{n} nops to enforce this, otherwise it does not
21147 generate the branch hint.
21149 @item -mhint-max-distance=@var{n}
21150 @opindex mhint-max-distance
21151 The encoding of the branch hint instruction limits the hint to be within
21152 256 instructions of the branch it is affecting. By default, GCC makes
21153 sure it is within 125.
21156 @opindex msafe-hints
21157 Work around a hardware bug that causes the SPU to stall indefinitely.
21158 By default, GCC inserts the @code{hbrp} instruction to make sure
21159 this stall won't happen.
21163 @node System V Options
21164 @subsection Options for System V
21166 These additional options are available on System V Release 4 for
21167 compatibility with other compilers on those systems:
21172 Create a shared object.
21173 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21177 Identify the versions of each tool used by the compiler, in a
21178 @code{.ident} assembler directive in the output.
21182 Refrain from adding @code{.ident} directives to the output file (this is
21185 @item -YP,@var{dirs}
21187 Search the directories @var{dirs}, and no others, for libraries
21188 specified with @option{-l}.
21190 @item -Ym,@var{dir}
21192 Look in the directory @var{dir} to find the M4 preprocessor.
21193 The assembler uses this option.
21194 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21195 @c the generic assembler that comes with Solaris takes just -Ym.
21198 @node TILE-Gx Options
21199 @subsection TILE-Gx Options
21200 @cindex TILE-Gx options
21202 These @samp{-m} options are supported on the TILE-Gx:
21205 @item -mcmodel=small
21206 @opindex mcmodel=small
21207 Generate code for the small model. The distance for direct calls is
21208 limited to 500M in either direction. PC-relative addresses are 32
21209 bits. Absolute addresses support the full address range.
21211 @item -mcmodel=large
21212 @opindex mcmodel=large
21213 Generate code for the large model. There is no limitation on call
21214 distance, pc-relative addresses, or absolute addresses.
21216 @item -mcpu=@var{name}
21218 Selects the type of CPU to be targeted. Currently the only supported
21219 type is @samp{tilegx}.
21225 Generate code for a 32-bit or 64-bit environment. The 32-bit
21226 environment sets int, long, and pointer to 32 bits. The 64-bit
21227 environment sets int to 32 bits and long and pointer to 64 bits.
21230 @itemx -mlittle-endian
21231 @opindex mbig-endian
21232 @opindex mlittle-endian
21233 Generate code in big/little endian mode, respectively.
21236 @node TILEPro Options
21237 @subsection TILEPro Options
21238 @cindex TILEPro options
21240 These @samp{-m} options are supported on the TILEPro:
21243 @item -mcpu=@var{name}
21245 Selects the type of CPU to be targeted. Currently the only supported
21246 type is @samp{tilepro}.
21250 Generate code for a 32-bit environment, which sets int, long, and
21251 pointer to 32 bits. This is the only supported behavior so the flag
21252 is essentially ignored.
21256 @subsection V850 Options
21257 @cindex V850 Options
21259 These @samp{-m} options are defined for V850 implementations:
21263 @itemx -mno-long-calls
21264 @opindex mlong-calls
21265 @opindex mno-long-calls
21266 Treat all calls as being far away (near). If calls are assumed to be
21267 far away, the compiler always loads the function's address into a
21268 register, and calls indirect through the pointer.
21274 Do not optimize (do optimize) basic blocks that use the same index
21275 pointer 4 or more times to copy pointer into the @code{ep} register, and
21276 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21277 option is on by default if you optimize.
21279 @item -mno-prolog-function
21280 @itemx -mprolog-function
21281 @opindex mno-prolog-function
21282 @opindex mprolog-function
21283 Do not use (do use) external functions to save and restore registers
21284 at the prologue and epilogue of a function. The external functions
21285 are slower, but use less code space if more than one function saves
21286 the same number of registers. The @option{-mprolog-function} option
21287 is on by default if you optimize.
21291 Try to make the code as small as possible. At present, this just turns
21292 on the @option{-mep} and @option{-mprolog-function} options.
21294 @item -mtda=@var{n}
21296 Put static or global variables whose size is @var{n} bytes or less into
21297 the tiny data area that register @code{ep} points to. The tiny data
21298 area can hold up to 256 bytes in total (128 bytes for byte references).
21300 @item -msda=@var{n}
21302 Put static or global variables whose size is @var{n} bytes or less into
21303 the small data area that register @code{gp} points to. The small data
21304 area can hold up to 64 kilobytes.
21306 @item -mzda=@var{n}
21308 Put static or global variables whose size is @var{n} bytes or less into
21309 the first 32 kilobytes of memory.
21313 Specify that the target processor is the V850.
21317 Specify that the target processor is the V850E3V5. The preprocessor
21318 constant @code{__v850e3v5__} is defined if this option is used.
21322 Specify that the target processor is the V850E3V5. This is an alias for
21323 the @option{-mv850e3v5} option.
21327 Specify that the target processor is the V850E2V3. The preprocessor
21328 constant @code{__v850e2v3__} is defined if this option is used.
21332 Specify that the target processor is the V850E2. The preprocessor
21333 constant @code{__v850e2__} is defined if this option is used.
21337 Specify that the target processor is the V850E1. The preprocessor
21338 constants @code{__v850e1__} and @code{__v850e__} are defined if
21339 this option is used.
21343 Specify that the target processor is the V850ES. This is an alias for
21344 the @option{-mv850e1} option.
21348 Specify that the target processor is the V850E@. The preprocessor
21349 constant @code{__v850e__} is defined if this option is used.
21351 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21352 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21353 are defined then a default target processor is chosen and the
21354 relevant @samp{__v850*__} preprocessor constant is defined.
21356 The preprocessor constants @code{__v850} and @code{__v851__} are always
21357 defined, regardless of which processor variant is the target.
21359 @item -mdisable-callt
21360 @itemx -mno-disable-callt
21361 @opindex mdisable-callt
21362 @opindex mno-disable-callt
21363 This option suppresses generation of the @code{CALLT} instruction for the
21364 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21367 This option is enabled by default when the RH850 ABI is
21368 in use (see @option{-mrh850-abi}), and disabled by default when the
21369 GCC ABI is in use. If @code{CALLT} instructions are being generated
21370 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
21376 Pass on (or do not pass on) the @option{-mrelax} command line option
21380 @itemx -mno-long-jumps
21381 @opindex mlong-jumps
21382 @opindex mno-long-jumps
21383 Disable (or re-enable) the generation of PC-relative jump instructions.
21386 @itemx -mhard-float
21387 @opindex msoft-float
21388 @opindex mhard-float
21389 Disable (or re-enable) the generation of hardware floating point
21390 instructions. This option is only significant when the target
21391 architecture is @samp{V850E2V3} or higher. If hardware floating point
21392 instructions are being generated then the C preprocessor symbol
21393 @code{__FPU_OK__} is defined, otherwise the symbol
21394 @code{__NO_FPU__} is defined.
21398 Enables the use of the e3v5 LOOP instruction. The use of this
21399 instruction is not enabled by default when the e3v5 architecture is
21400 selected because its use is still experimental.
21404 @opindex mrh850-abi
21406 Enables support for the RH850 version of the V850 ABI. This is the
21407 default. With this version of the ABI the following rules apply:
21411 Integer sized structures and unions are returned via a memory pointer
21412 rather than a register.
21415 Large structures and unions (more than 8 bytes in size) are passed by
21419 Functions are aligned to 16-bit boundaries.
21422 The @option{-m8byte-align} command line option is supported.
21425 The @option{-mdisable-callt} command line option is enabled by
21426 default. The @option{-mno-disable-callt} command line option is not
21430 When this version of the ABI is enabled the C preprocessor symbol
21431 @code{__V850_RH850_ABI__} is defined.
21435 Enables support for the old GCC version of the V850 ABI. With this
21436 version of the ABI the following rules apply:
21440 Integer sized structures and unions are returned in register @code{r10}.
21443 Large structures and unions (more than 8 bytes in size) are passed by
21447 Functions are aligned to 32-bit boundaries, unless optimizing for
21451 The @option{-m8byte-align} command line option is not supported.
21454 The @option{-mdisable-callt} command line option is supported but not
21455 enabled by default.
21458 When this version of the ABI is enabled the C preprocessor symbol
21459 @code{__V850_GCC_ABI__} is defined.
21461 @item -m8byte-align
21462 @itemx -mno-8byte-align
21463 @opindex m8byte-align
21464 @opindex mno-8byte-align
21465 Enables support for @code{double} and @code{long long} types to be
21466 aligned on 8-byte boundaries. The default is to restrict the
21467 alignment of all objects to at most 4-bytes. When
21468 @option{-m8byte-align} is in effect the C preprocessor symbol
21469 @code{__V850_8BYTE_ALIGN__} is defined.
21472 @opindex mbig-switch
21473 Generate code suitable for big switch tables. Use this option only if
21474 the assembler/linker complain about out of range branches within a switch
21479 This option causes r2 and r5 to be used in the code generated by
21480 the compiler. This setting is the default.
21482 @item -mno-app-regs
21483 @opindex mno-app-regs
21484 This option causes r2 and r5 to be treated as fixed registers.
21489 @subsection VAX Options
21490 @cindex VAX options
21492 These @samp{-m} options are defined for the VAX:
21497 Do not output certain jump instructions (@code{aobleq} and so on)
21498 that the Unix assembler for the VAX cannot handle across long
21503 Do output those jump instructions, on the assumption that the
21504 GNU assembler is being used.
21508 Output code for G-format floating-point numbers instead of D-format.
21511 @node Visium Options
21512 @subsection Visium Options
21513 @cindex Visium options
21519 A program which performs file I/O and is destined to run on an MCM target
21520 should be linked with this option. It causes the libraries libc.a and
21521 libdebug.a to be linked. The program should be run on the target under
21522 the control of the GDB remote debugging stub.
21526 A program which performs file I/O and is destined to run on the simulator
21527 should be linked with option. This causes libraries libc.a and libsim.a to
21531 @itemx -mhard-float
21533 @opindex mhard-float
21534 Generate code containing floating-point instructions. This is the
21538 @itemx -msoft-float
21540 @opindex msoft-float
21541 Generate code containing library calls for floating-point.
21543 @option{-msoft-float} changes the calling convention in the output file;
21544 therefore, it is only useful if you compile @emph{all} of a program with
21545 this option. In particular, you need to compile @file{libgcc.a}, the
21546 library that comes with GCC, with @option{-msoft-float} in order for
21549 @item -mcpu=@var{cpu_type}
21551 Set the instruction set, register set, and instruction scheduling parameters
21552 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21553 @samp{mcm}, @samp{gr5} and @samp{gr6}.
21555 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
21557 By default (unless configured otherwise), GCC generates code for the GR5
21558 variant of the Visium architecture.
21560 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
21561 architecture. The only difference from GR5 code is that the compiler will
21562 generate block move instructions.
21564 @item -mtune=@var{cpu_type}
21566 Set the instruction scheduling parameters for machine type @var{cpu_type},
21567 but do not set the instruction set or register set that the option
21568 @option{-mcpu=@var{cpu_type}} would.
21572 Generate code for the supervisor mode, where there are no restrictions on
21573 the access to general registers. This is the default.
21576 @opindex muser-mode
21577 Generate code for the user mode, where the access to some general registers
21578 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
21579 mode; on the GR6, only registers r29 to r31 are affected.
21583 @subsection VMS Options
21585 These @samp{-m} options are defined for the VMS implementations:
21588 @item -mvms-return-codes
21589 @opindex mvms-return-codes
21590 Return VMS condition codes from @code{main}. The default is to return POSIX-style
21591 condition (e.g.@ error) codes.
21593 @item -mdebug-main=@var{prefix}
21594 @opindex mdebug-main=@var{prefix}
21595 Flag the first routine whose name starts with @var{prefix} as the main
21596 routine for the debugger.
21600 Default to 64-bit memory allocation routines.
21602 @item -mpointer-size=@var{size}
21603 @opindex mpointer-size=@var{size}
21604 Set the default size of pointers. Possible options for @var{size} are
21605 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
21606 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
21607 The later option disables @code{pragma pointer_size}.
21610 @node VxWorks Options
21611 @subsection VxWorks Options
21612 @cindex VxWorks Options
21614 The options in this section are defined for all VxWorks targets.
21615 Options specific to the target hardware are listed with the other
21616 options for that target.
21621 GCC can generate code for both VxWorks kernels and real time processes
21622 (RTPs). This option switches from the former to the latter. It also
21623 defines the preprocessor macro @code{__RTP__}.
21626 @opindex non-static
21627 Link an RTP executable against shared libraries rather than static
21628 libraries. The options @option{-static} and @option{-shared} can
21629 also be used for RTPs (@pxref{Link Options}); @option{-static}
21636 These options are passed down to the linker. They are defined for
21637 compatibility with Diab.
21640 @opindex Xbind-lazy
21641 Enable lazy binding of function calls. This option is equivalent to
21642 @option{-Wl,-z,now} and is defined for compatibility with Diab.
21646 Disable lazy binding of function calls. This option is the default and
21647 is defined for compatibility with Diab.
21651 @subsection x86 Options
21652 @cindex x86 Options
21654 These @samp{-m} options are defined for the x86 family of computers.
21658 @item -march=@var{cpu-type}
21660 Generate instructions for the machine type @var{cpu-type}. In contrast to
21661 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
21662 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
21663 to generate code that may not run at all on processors other than the one
21664 indicated. Specifying @option{-march=@var{cpu-type}} implies
21665 @option{-mtune=@var{cpu-type}}.
21667 The choices for @var{cpu-type} are:
21671 This selects the CPU to generate code for at compilation time by determining
21672 the processor type of the compiling machine. Using @option{-march=native}
21673 enables all instruction subsets supported by the local machine (hence
21674 the result might not run on different machines). Using @option{-mtune=native}
21675 produces code optimized for the local machine under the constraints
21676 of the selected instruction set.
21679 Original Intel i386 CPU@.
21682 Intel i486 CPU@. (No scheduling is implemented for this chip.)
21686 Intel Pentium CPU with no MMX support.
21689 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
21692 Intel Pentium Pro CPU@.
21695 When used with @option{-march}, the Pentium Pro
21696 instruction set is used, so the code runs on all i686 family chips.
21697 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
21700 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
21705 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
21709 Intel Pentium M; low-power version of Intel Pentium III CPU
21710 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
21714 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
21717 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
21721 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
21722 SSE2 and SSE3 instruction set support.
21725 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
21726 instruction set support.
21729 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21730 SSE4.1, SSE4.2 and POPCNT instruction set support.
21733 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21734 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
21737 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21738 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
21741 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21742 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
21743 instruction set support.
21746 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21747 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
21748 BMI, BMI2 and F16C instruction set support.
21751 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21752 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
21753 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
21756 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
21757 instruction set support.
21760 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21761 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
21764 AMD K6 CPU with MMX instruction set support.
21768 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
21771 @itemx athlon-tbird
21772 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
21778 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
21779 instruction set support.
21785 Processors based on the AMD K8 core with x86-64 instruction set support,
21786 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
21787 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
21788 instruction set extensions.)
21791 @itemx opteron-sse3
21792 @itemx athlon64-sse3
21793 Improved versions of AMD K8 cores with SSE3 instruction set support.
21797 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
21798 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
21799 instruction set extensions.)
21802 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
21803 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
21804 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
21806 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
21807 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
21808 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
21811 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
21812 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
21813 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
21814 64-bit instruction set extensions.
21816 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
21817 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
21818 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
21819 SSE4.2, ABM and 64-bit instruction set extensions.
21822 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
21823 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
21824 instruction set extensions.)
21827 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
21828 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
21829 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
21832 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
21836 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
21837 instruction set support.
21840 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
21841 implemented for this chip.)
21844 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
21846 implemented for this chip.)
21849 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
21852 @item -mtune=@var{cpu-type}
21854 Tune to @var{cpu-type} everything applicable about the generated code, except
21855 for the ABI and the set of available instructions.
21856 While picking a specific @var{cpu-type} schedules things appropriately
21857 for that particular chip, the compiler does not generate any code that
21858 cannot run on the default machine type unless you use a
21859 @option{-march=@var{cpu-type}} option.
21860 For example, if GCC is configured for i686-pc-linux-gnu
21861 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
21862 but still runs on i686 machines.
21864 The choices for @var{cpu-type} are the same as for @option{-march}.
21865 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
21869 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
21870 If you know the CPU on which your code will run, then you should use
21871 the corresponding @option{-mtune} or @option{-march} option instead of
21872 @option{-mtune=generic}. But, if you do not know exactly what CPU users
21873 of your application will have, then you should use this option.
21875 As new processors are deployed in the marketplace, the behavior of this
21876 option will change. Therefore, if you upgrade to a newer version of
21877 GCC, code generation controlled by this option will change to reflect
21879 that are most common at the time that version of GCC is released.
21881 There is no @option{-march=generic} option because @option{-march}
21882 indicates the instruction set the compiler can use, and there is no
21883 generic instruction set applicable to all processors. In contrast,
21884 @option{-mtune} indicates the processor (or, in this case, collection of
21885 processors) for which the code is optimized.
21888 Produce code optimized for the most current Intel processors, which are
21889 Haswell and Silvermont for this version of GCC. If you know the CPU
21890 on which your code will run, then you should use the corresponding
21891 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
21892 But, if you want your application performs better on both Haswell and
21893 Silvermont, then you should use this option.
21895 As new Intel processors are deployed in the marketplace, the behavior of
21896 this option will change. Therefore, if you upgrade to a newer version of
21897 GCC, code generation controlled by this option will change to reflect
21898 the most current Intel processors at the time that version of GCC is
21901 There is no @option{-march=intel} option because @option{-march} indicates
21902 the instruction set the compiler can use, and there is no common
21903 instruction set applicable to all processors. In contrast,
21904 @option{-mtune} indicates the processor (or, in this case, collection of
21905 processors) for which the code is optimized.
21908 @item -mcpu=@var{cpu-type}
21910 A deprecated synonym for @option{-mtune}.
21912 @item -mfpmath=@var{unit}
21914 Generate floating-point arithmetic for selected unit @var{unit}. The choices
21915 for @var{unit} are:
21919 Use the standard 387 floating-point coprocessor present on the majority of chips and
21920 emulated otherwise. Code compiled with this option runs almost everywhere.
21921 The temporary results are computed in 80-bit precision instead of the precision
21922 specified by the type, resulting in slightly different results compared to most
21923 of other chips. See @option{-ffloat-store} for more detailed description.
21925 This is the default choice for x86-32 targets.
21928 Use scalar floating-point instructions present in the SSE instruction set.
21929 This instruction set is supported by Pentium III and newer chips,
21930 and in the AMD line
21931 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
21932 instruction set supports only single-precision arithmetic, thus the double and
21933 extended-precision arithmetic are still done using 387. A later version, present
21934 only in Pentium 4 and AMD x86-64 chips, supports double-precision
21937 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
21938 or @option{-msse2} switches to enable SSE extensions and make this option
21939 effective. For the x86-64 compiler, these extensions are enabled by default.
21941 The resulting code should be considerably faster in the majority of cases and avoid
21942 the numerical instability problems of 387 code, but may break some existing
21943 code that expects temporaries to be 80 bits.
21945 This is the default choice for the x86-64 compiler.
21950 Attempt to utilize both instruction sets at once. This effectively doubles the
21951 amount of available registers, and on chips with separate execution units for
21952 387 and SSE the execution resources too. Use this option with care, as it is
21953 still experimental, because the GCC register allocator does not model separate
21954 functional units well, resulting in unstable performance.
21957 @item -masm=@var{dialect}
21958 @opindex masm=@var{dialect}
21959 Output assembly instructions using selected @var{dialect}. Also affects
21960 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
21961 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
21962 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
21963 not support @samp{intel}.
21966 @itemx -mno-ieee-fp
21968 @opindex mno-ieee-fp
21969 Control whether or not the compiler uses IEEE floating-point
21970 comparisons. These correctly handle the case where the result of a
21971 comparison is unordered.
21974 @opindex msoft-float
21975 Generate output containing library calls for floating point.
21977 @strong{Warning:} the requisite libraries are not part of GCC@.
21978 Normally the facilities of the machine's usual C compiler are used, but
21979 this can't be done directly in cross-compilation. You must make your
21980 own arrangements to provide suitable library functions for
21983 On machines where a function returns floating-point results in the 80387
21984 register stack, some floating-point opcodes may be emitted even if
21985 @option{-msoft-float} is used.
21987 @item -mno-fp-ret-in-387
21988 @opindex mno-fp-ret-in-387
21989 Do not use the FPU registers for return values of functions.
21991 The usual calling convention has functions return values of types
21992 @code{float} and @code{double} in an FPU register, even if there
21993 is no FPU@. The idea is that the operating system should emulate
21996 The option @option{-mno-fp-ret-in-387} causes such values to be returned
21997 in ordinary CPU registers instead.
21999 @item -mno-fancy-math-387
22000 @opindex mno-fancy-math-387
22001 Some 387 emulators do not support the @code{sin}, @code{cos} and
22002 @code{sqrt} instructions for the 387. Specify this option to avoid
22003 generating those instructions. This option is the default on FreeBSD,
22004 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22005 indicates that the target CPU always has an FPU and so the
22006 instruction does not need emulation. These
22007 instructions are not generated unless you also use the
22008 @option{-funsafe-math-optimizations} switch.
22010 @item -malign-double
22011 @itemx -mno-align-double
22012 @opindex malign-double
22013 @opindex mno-align-double
22014 Control whether GCC aligns @code{double}, @code{long double}, and
22015 @code{long long} variables on a two-word boundary or a one-word
22016 boundary. Aligning @code{double} variables on a two-word boundary
22017 produces code that runs somewhat faster on a Pentium at the
22018 expense of more memory.
22020 On x86-64, @option{-malign-double} is enabled by default.
22022 @strong{Warning:} if you use the @option{-malign-double} switch,
22023 structures containing the above types are aligned differently than
22024 the published application binary interface specifications for the x86-32
22025 and are not binary compatible with structures in code compiled
22026 without that switch.
22028 @item -m96bit-long-double
22029 @itemx -m128bit-long-double
22030 @opindex m96bit-long-double
22031 @opindex m128bit-long-double
22032 These switches control the size of @code{long double} type. The x86-32
22033 application binary interface specifies the size to be 96 bits,
22034 so @option{-m96bit-long-double} is the default in 32-bit mode.
22036 Modern architectures (Pentium and newer) prefer @code{long double}
22037 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22038 conforming to the ABI, this is not possible. So specifying
22039 @option{-m128bit-long-double} aligns @code{long double}
22040 to a 16-byte boundary by padding the @code{long double} with an additional
22043 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22044 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22046 Notice that neither of these options enable any extra precision over the x87
22047 standard of 80 bits for a @code{long double}.
22049 @strong{Warning:} if you override the default value for your target ABI, this
22050 changes the size of
22051 structures and arrays containing @code{long double} variables,
22052 as well as modifying the function calling convention for functions taking
22053 @code{long double}. Hence they are not binary-compatible
22054 with code compiled without that switch.
22056 @item -mlong-double-64
22057 @itemx -mlong-double-80
22058 @itemx -mlong-double-128
22059 @opindex mlong-double-64
22060 @opindex mlong-double-80
22061 @opindex mlong-double-128
22062 These switches control the size of @code{long double} type. A size
22063 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22064 type. This is the default for 32-bit Bionic C library. A size
22065 of 128 bits makes the @code{long double} type equivalent to the
22066 @code{__float128} type. This is the default for 64-bit Bionic C library.
22068 @strong{Warning:} if you override the default value for your target ABI, this
22069 changes the size of
22070 structures and arrays containing @code{long double} variables,
22071 as well as modifying the function calling convention for functions taking
22072 @code{long double}. Hence they are not binary-compatible
22073 with code compiled without that switch.
22075 @item -malign-data=@var{type}
22076 @opindex malign-data
22077 Control how GCC aligns variables. Supported values for @var{type} are
22078 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22079 and earlier, @samp{abi} uses alignment value as specified by the
22080 psABI, and @samp{cacheline} uses increased alignment value to match
22081 the cache line size. @samp{compat} is the default.
22083 @item -mlarge-data-threshold=@var{threshold}
22084 @opindex mlarge-data-threshold
22085 When @option{-mcmodel=medium} is specified, data objects larger than
22086 @var{threshold} are placed in the large data section. This value must be the
22087 same across all objects linked into the binary, and defaults to 65535.
22091 Use a different function-calling convention, in which functions that
22092 take a fixed number of arguments return with the @code{ret @var{num}}
22093 instruction, which pops their arguments while returning. This saves one
22094 instruction in the caller since there is no need to pop the arguments
22097 You can specify that an individual function is called with this calling
22098 sequence with the function attribute @code{stdcall}. You can also
22099 override the @option{-mrtd} option by using the function attribute
22100 @code{cdecl}. @xref{Function Attributes}.
22102 @strong{Warning:} this calling convention is incompatible with the one
22103 normally used on Unix, so you cannot use it if you need to call
22104 libraries compiled with the Unix compiler.
22106 Also, you must provide function prototypes for all functions that
22107 take variable numbers of arguments (including @code{printf});
22108 otherwise incorrect code is generated for calls to those
22111 In addition, seriously incorrect code results if you call a
22112 function with too many arguments. (Normally, extra arguments are
22113 harmlessly ignored.)
22115 @item -mregparm=@var{num}
22117 Control how many registers are used to pass integer arguments. By
22118 default, no registers are used to pass arguments, and at most 3
22119 registers can be used. You can control this behavior for a specific
22120 function by using the function attribute @code{regparm}.
22121 @xref{Function Attributes}.
22123 @strong{Warning:} if you use this switch, and
22124 @var{num} is nonzero, then you must build all modules with the same
22125 value, including any libraries. This includes the system libraries and
22129 @opindex msseregparm
22130 Use SSE register passing conventions for float and double arguments
22131 and return values. You can control this behavior for a specific
22132 function by using the function attribute @code{sseregparm}.
22133 @xref{Function Attributes}.
22135 @strong{Warning:} if you use this switch then you must build all
22136 modules with the same value, including any libraries. This includes
22137 the system libraries and startup modules.
22139 @item -mvect8-ret-in-mem
22140 @opindex mvect8-ret-in-mem
22141 Return 8-byte vectors in memory instead of MMX registers. This is the
22142 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
22143 Studio compilers until version 12. Later compiler versions (starting
22144 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
22145 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
22146 you need to remain compatible with existing code produced by those
22147 previous compiler versions or older versions of GCC@.
22156 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
22157 is specified, the significands of results of floating-point operations are
22158 rounded to 24 bits (single precision); @option{-mpc64} rounds the
22159 significands of results of floating-point operations to 53 bits (double
22160 precision) and @option{-mpc80} rounds the significands of results of
22161 floating-point operations to 64 bits (extended double precision), which is
22162 the default. When this option is used, floating-point operations in higher
22163 precisions are not available to the programmer without setting the FPU
22164 control word explicitly.
22166 Setting the rounding of floating-point operations to less than the default
22167 80 bits can speed some programs by 2% or more. Note that some mathematical
22168 libraries assume that extended-precision (80-bit) floating-point operations
22169 are enabled by default; routines in such libraries could suffer significant
22170 loss of accuracy, typically through so-called ``catastrophic cancellation'',
22171 when this option is used to set the precision to less than extended precision.
22173 @item -mstackrealign
22174 @opindex mstackrealign
22175 Realign the stack at entry. On the x86, the @option{-mstackrealign}
22176 option generates an alternate prologue and epilogue that realigns the
22177 run-time stack if necessary. This supports mixing legacy codes that keep
22178 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
22179 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
22180 applicable to individual functions.
22182 @item -mpreferred-stack-boundary=@var{num}
22183 @opindex mpreferred-stack-boundary
22184 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
22185 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
22186 the default is 4 (16 bytes or 128 bits).
22188 @strong{Warning:} When generating code for the x86-64 architecture with
22189 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
22190 used to keep the stack boundary aligned to 8 byte boundary. Since
22191 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
22192 intended to be used in controlled environment where stack space is
22193 important limitation. This option leads to wrong code when functions
22194 compiled with 16 byte stack alignment (such as functions from a standard
22195 library) are called with misaligned stack. In this case, SSE
22196 instructions may lead to misaligned memory access traps. In addition,
22197 variable arguments are handled incorrectly for 16 byte aligned
22198 objects (including x87 long double and __int128), leading to wrong
22199 results. You must build all modules with
22200 @option{-mpreferred-stack-boundary=3}, including any libraries. This
22201 includes the system libraries and startup modules.
22203 @item -mincoming-stack-boundary=@var{num}
22204 @opindex mincoming-stack-boundary
22205 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
22206 boundary. If @option{-mincoming-stack-boundary} is not specified,
22207 the one specified by @option{-mpreferred-stack-boundary} is used.
22209 On Pentium and Pentium Pro, @code{double} and @code{long double} values
22210 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
22211 suffer significant run time performance penalties. On Pentium III, the
22212 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
22213 properly if it is not 16-byte aligned.
22215 To ensure proper alignment of this values on the stack, the stack boundary
22216 must be as aligned as that required by any value stored on the stack.
22217 Further, every function must be generated such that it keeps the stack
22218 aligned. Thus calling a function compiled with a higher preferred
22219 stack boundary from a function compiled with a lower preferred stack
22220 boundary most likely misaligns the stack. It is recommended that
22221 libraries that use callbacks always use the default setting.
22223 This extra alignment does consume extra stack space, and generally
22224 increases code size. Code that is sensitive to stack space usage, such
22225 as embedded systems and operating system kernels, may want to reduce the
22226 preferred alignment to @option{-mpreferred-stack-boundary=2}.
22272 @opindex mclfushopt
22290 @itemx -mprefetchwt1
22291 @opindex mprefetchwt1
22339 These switches enable the use of instructions in the MMX, SSE,
22340 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
22341 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
22342 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX or 3DNow!@:
22343 extended instruction sets. Each has a corresponding @option{-mno-} option
22344 to disable use of these instructions.
22346 These extensions are also available as built-in functions: see
22347 @ref{x86 Built-in Functions}, for details of the functions enabled and
22348 disabled by these switches.
22350 To generate SSE/SSE2 instructions automatically from floating-point
22351 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
22353 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
22354 generates new AVX instructions or AVX equivalence for all SSEx instructions
22357 These options enable GCC to use these extended instructions in
22358 generated code, even without @option{-mfpmath=sse}. Applications that
22359 perform run-time CPU detection must compile separate files for each
22360 supported architecture, using the appropriate flags. In particular,
22361 the file containing the CPU detection code should be compiled without
22364 @item -mdump-tune-features
22365 @opindex mdump-tune-features
22366 This option instructs GCC to dump the names of the x86 performance
22367 tuning features and default settings. The names can be used in
22368 @option{-mtune-ctrl=@var{feature-list}}.
22370 @item -mtune-ctrl=@var{feature-list}
22371 @opindex mtune-ctrl=@var{feature-list}
22372 This option is used to do fine grain control of x86 code generation features.
22373 @var{feature-list} is a comma separated list of @var{feature} names. See also
22374 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
22375 on if it is not preceded with @samp{^}, otherwise, it is turned off.
22376 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
22377 developers. Using it may lead to code paths not covered by testing and can
22378 potentially result in compiler ICEs or runtime errors.
22381 @opindex mno-default
22382 This option instructs GCC to turn off all tunable features. See also
22383 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
22387 This option instructs GCC to emit a @code{cld} instruction in the prologue
22388 of functions that use string instructions. String instructions depend on
22389 the DF flag to select between autoincrement or autodecrement mode. While the
22390 ABI specifies the DF flag to be cleared on function entry, some operating
22391 systems violate this specification by not clearing the DF flag in their
22392 exception dispatchers. The exception handler can be invoked with the DF flag
22393 set, which leads to wrong direction mode when string instructions are used.
22394 This option can be enabled by default on 32-bit x86 targets by configuring
22395 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
22396 instructions can be suppressed with the @option{-mno-cld} compiler option
22400 @opindex mvzeroupper
22401 This option instructs GCC to emit a @code{vzeroupper} instruction
22402 before a transfer of control flow out of the function to minimize
22403 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
22406 @item -mprefer-avx128
22407 @opindex mprefer-avx128
22408 This option instructs GCC to use 128-bit AVX instructions instead of
22409 256-bit AVX instructions in the auto-vectorizer.
22413 This option enables GCC to generate @code{CMPXCHG16B} instructions.
22414 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
22415 (or oword) data types.
22416 This is useful for high-resolution counters that can be updated
22417 by multiple processors (or cores). This instruction is generated as part of
22418 atomic built-in functions: see @ref{__sync Builtins} or
22419 @ref{__atomic Builtins} for details.
22423 This option enables generation of @code{SAHF} instructions in 64-bit code.
22424 Early Intel Pentium 4 CPUs with Intel 64 support,
22425 prior to the introduction of Pentium 4 G1 step in December 2005,
22426 lacked the @code{LAHF} and @code{SAHF} instructions
22427 which are supported by AMD64.
22428 These are load and store instructions, respectively, for certain status flags.
22429 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
22430 @code{drem}, and @code{remainder} built-in functions;
22431 see @ref{Other Builtins} for details.
22435 This option enables use of the @code{movbe} instruction to implement
22436 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
22440 This option enables built-in functions @code{__builtin_ia32_crc32qi},
22441 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
22442 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
22446 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
22447 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
22448 with an additional Newton-Raphson step
22449 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
22450 (and their vectorized
22451 variants) for single-precision floating-point arguments. These instructions
22452 are generated only when @option{-funsafe-math-optimizations} is enabled
22453 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
22454 Note that while the throughput of the sequence is higher than the throughput
22455 of the non-reciprocal instruction, the precision of the sequence can be
22456 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
22458 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
22459 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
22460 combination), and doesn't need @option{-mrecip}.
22462 Also note that GCC emits the above sequence with additional Newton-Raphson step
22463 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
22464 already with @option{-ffast-math} (or the above option combination), and
22465 doesn't need @option{-mrecip}.
22467 @item -mrecip=@var{opt}
22468 @opindex mrecip=opt
22469 This option controls which reciprocal estimate instructions
22470 may be used. @var{opt} is a comma-separated list of options, which may
22471 be preceded by a @samp{!} to invert the option:
22475 Enable all estimate instructions.
22478 Enable the default instructions, equivalent to @option{-mrecip}.
22481 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22484 Enable the approximation for scalar division.
22487 Enable the approximation for vectorized division.
22490 Enable the approximation for scalar square root.
22493 Enable the approximation for vectorized square root.
22496 So, for example, @option{-mrecip=all,!sqrt} enables
22497 all of the reciprocal approximations, except for square root.
22499 @item -mveclibabi=@var{type}
22500 @opindex mveclibabi
22501 Specifies the ABI type to use for vectorizing intrinsics using an
22502 external library. Supported values for @var{type} are @samp{svml}
22503 for the Intel short
22504 vector math library and @samp{acml} for the AMD math core library.
22505 To use this option, both @option{-ftree-vectorize} and
22506 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
22507 ABI-compatible library must be specified at link time.
22509 GCC currently emits calls to @code{vmldExp2},
22510 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
22511 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
22512 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
22513 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
22514 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
22515 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
22516 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
22517 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
22518 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
22519 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
22520 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
22521 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
22522 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
22523 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
22524 when @option{-mveclibabi=acml} is used.
22526 @item -mabi=@var{name}
22528 Generate code for the specified calling convention. Permissible values
22529 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
22530 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
22531 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
22532 You can control this behavior for specific functions by
22533 using the function attributes @code{ms_abi} and @code{sysv_abi}.
22534 @xref{Function Attributes}.
22536 @item -mtls-dialect=@var{type}
22537 @opindex mtls-dialect
22538 Generate code to access thread-local storage using the @samp{gnu} or
22539 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
22540 @samp{gnu2} is more efficient, but it may add compile- and run-time
22541 requirements that cannot be satisfied on all systems.
22544 @itemx -mno-push-args
22545 @opindex mpush-args
22546 @opindex mno-push-args
22547 Use PUSH operations to store outgoing parameters. This method is shorter
22548 and usually equally fast as method using SUB/MOV operations and is enabled
22549 by default. In some cases disabling it may improve performance because of
22550 improved scheduling and reduced dependencies.
22552 @item -maccumulate-outgoing-args
22553 @opindex maccumulate-outgoing-args
22554 If enabled, the maximum amount of space required for outgoing arguments is
22555 computed in the function prologue. This is faster on most modern CPUs
22556 because of reduced dependencies, improved scheduling and reduced stack usage
22557 when the preferred stack boundary is not equal to 2. The drawback is a notable
22558 increase in code size. This switch implies @option{-mno-push-args}.
22562 Support thread-safe exception handling on MinGW. Programs that rely
22563 on thread-safe exception handling must compile and link all code with the
22564 @option{-mthreads} option. When compiling, @option{-mthreads} defines
22565 @option{-D_MT}; when linking, it links in a special thread helper library
22566 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
22568 @item -mno-align-stringops
22569 @opindex mno-align-stringops
22570 Do not align the destination of inlined string operations. This switch reduces
22571 code size and improves performance in case the destination is already aligned,
22572 but GCC doesn't know about it.
22574 @item -minline-all-stringops
22575 @opindex minline-all-stringops
22576 By default GCC inlines string operations only when the destination is
22577 known to be aligned to least a 4-byte boundary.
22578 This enables more inlining and increases code
22579 size, but may improve performance of code that depends on fast
22580 @code{memcpy}, @code{strlen},
22581 and @code{memset} for short lengths.
22583 @item -minline-stringops-dynamically
22584 @opindex minline-stringops-dynamically
22585 For string operations of unknown size, use run-time checks with
22586 inline code for small blocks and a library call for large blocks.
22588 @item -mstringop-strategy=@var{alg}
22589 @opindex mstringop-strategy=@var{alg}
22590 Override the internal decision heuristic for the particular algorithm to use
22591 for inlining string operations. The allowed values for @var{alg} are:
22597 Expand using i386 @code{rep} prefix of the specified size.
22601 @itemx unrolled_loop
22602 Expand into an inline loop.
22605 Always use a library call.
22608 @item -mmemcpy-strategy=@var{strategy}
22609 @opindex mmemcpy-strategy=@var{strategy}
22610 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
22611 should be inlined and what inline algorithm to use when the expected size
22612 of the copy operation is known. @var{strategy}
22613 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
22614 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
22615 the max byte size with which inline algorithm @var{alg} is allowed. For the last
22616 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
22617 in the list must be specified in increasing order. The minimal byte size for
22618 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
22621 @item -mmemset-strategy=@var{strategy}
22622 @opindex mmemset-strategy=@var{strategy}
22623 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
22624 @code{__builtin_memset} expansion.
22626 @item -momit-leaf-frame-pointer
22627 @opindex momit-leaf-frame-pointer
22628 Don't keep the frame pointer in a register for leaf functions. This
22629 avoids the instructions to save, set up, and restore frame pointers and
22630 makes an extra register available in leaf functions. The option
22631 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
22632 which might make debugging harder.
22634 @item -mtls-direct-seg-refs
22635 @itemx -mno-tls-direct-seg-refs
22636 @opindex mtls-direct-seg-refs
22637 Controls whether TLS variables may be accessed with offsets from the
22638 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
22639 or whether the thread base pointer must be added. Whether or not this
22640 is valid depends on the operating system, and whether it maps the
22641 segment to cover the entire TLS area.
22643 For systems that use the GNU C Library, the default is on.
22646 @itemx -mno-sse2avx
22648 Specify that the assembler should encode SSE instructions with VEX
22649 prefix. The option @option{-mavx} turns this on by default.
22654 If profiling is active (@option{-pg}), put the profiling
22655 counter call before the prologue.
22656 Note: On x86 architectures the attribute @code{ms_hook_prologue}
22657 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
22659 @item -mrecord-mcount
22660 @itemx -mno-record-mcount
22661 @opindex mrecord-mcount
22662 If profiling is active (@option{-pg}), generate a __mcount_loc section
22663 that contains pointers to each profiling call. This is useful for
22664 automatically patching and out calls.
22667 @itemx -mno-nop-mcount
22668 @opindex mnop-mcount
22669 If profiling is active (@option{-pg}), generate the calls to
22670 the profiling functions as nops. This is useful when they
22671 should be patched in later dynamically. This is likely only
22672 useful together with @option{-mrecord-mcount}.
22674 @item -mskip-rax-setup
22675 @itemx -mno-skip-rax-setup
22676 @opindex mskip-rax-setup
22677 When generating code for the x86-64 architecture with SSE extensions
22678 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
22679 register when there are no variable arguments passed in vector registers.
22681 @strong{Warning:} Since RAX register is used to avoid unnecessarily
22682 saving vector registers on stack when passing variable arguments, the
22683 impacts of this option are callees may waste some stack space,
22684 misbehave or jump to a random location. GCC 4.4 or newer don't have
22685 those issues, regardless the RAX register value.
22688 @itemx -mno-8bit-idiv
22689 @opindex m8bit-idiv
22690 On some processors, like Intel Atom, 8-bit unsigned integer divide is
22691 much faster than 32-bit/64-bit integer divide. This option generates a
22692 run-time check. If both dividend and divisor are within range of 0
22693 to 255, 8-bit unsigned integer divide is used instead of
22694 32-bit/64-bit integer divide.
22696 @item -mavx256-split-unaligned-load
22697 @itemx -mavx256-split-unaligned-store
22698 @opindex mavx256-split-unaligned-load
22699 @opindex mavx256-split-unaligned-store
22700 Split 32-byte AVX unaligned load and store.
22702 @item -mstack-protector-guard=@var{guard}
22703 @opindex mstack-protector-guard=@var{guard}
22704 Generate stack protection code using canary at @var{guard}. Supported
22705 locations are @samp{global} for global canary or @samp{tls} for per-thread
22706 canary in the TLS block (the default). This option has effect only when
22707 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
22711 These @samp{-m} switches are supported in addition to the above
22712 on x86-64 processors in 64-bit environments.
22723 Generate code for a 16-bit, 32-bit or 64-bit environment.
22724 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
22726 generates code that runs on any i386 system.
22728 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
22729 types to 64 bits, and generates code for the x86-64 architecture.
22730 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
22731 and @option{-mdynamic-no-pic} options.
22733 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
22735 generates code for the x86-64 architecture.
22737 The @option{-m16} option is the same as @option{-m32}, except for that
22738 it outputs the @code{.code16gcc} assembly directive at the beginning of
22739 the assembly output so that the binary can run in 16-bit mode.
22741 @item -mno-red-zone
22742 @opindex mno-red-zone
22743 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
22744 by the x86-64 ABI; it is a 128-byte area beyond the location of the
22745 stack pointer that is not modified by signal or interrupt handlers
22746 and therefore can be used for temporary data without adjusting the stack
22747 pointer. The flag @option{-mno-red-zone} disables this red zone.
22749 @item -mcmodel=small
22750 @opindex mcmodel=small
22751 Generate code for the small code model: the program and its symbols must
22752 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
22753 Programs can be statically or dynamically linked. This is the default
22756 @item -mcmodel=kernel
22757 @opindex mcmodel=kernel
22758 Generate code for the kernel code model. The kernel runs in the
22759 negative 2 GB of the address space.
22760 This model has to be used for Linux kernel code.
22762 @item -mcmodel=medium
22763 @opindex mcmodel=medium
22764 Generate code for the medium model: the program is linked in the lower 2
22765 GB of the address space. Small symbols are also placed there. Symbols
22766 with sizes larger than @option{-mlarge-data-threshold} are put into
22767 large data or BSS sections and can be located above 2GB. Programs can
22768 be statically or dynamically linked.
22770 @item -mcmodel=large
22771 @opindex mcmodel=large
22772 Generate code for the large model. This model makes no assumptions
22773 about addresses and sizes of sections.
22775 @item -maddress-mode=long
22776 @opindex maddress-mode=long
22777 Generate code for long address mode. This is only supported for 64-bit
22778 and x32 environments. It is the default address mode for 64-bit
22781 @item -maddress-mode=short
22782 @opindex maddress-mode=short
22783 Generate code for short address mode. This is only supported for 32-bit
22784 and x32 environments. It is the default address mode for 32-bit and
22788 @node x86 Windows Options
22789 @subsection x86 Windows Options
22790 @cindex x86 Windows Options
22791 @cindex Windows Options for x86
22793 These additional options are available for Microsoft Windows targets:
22799 specifies that a console application is to be generated, by
22800 instructing the linker to set the PE header subsystem type
22801 required for console applications.
22802 This option is available for Cygwin and MinGW targets and is
22803 enabled by default on those targets.
22807 This option is available for Cygwin and MinGW targets. It
22808 specifies that a DLL---a dynamic link library---is to be
22809 generated, enabling the selection of the required runtime
22810 startup object and entry point.
22812 @item -mnop-fun-dllimport
22813 @opindex mnop-fun-dllimport
22814 This option is available for Cygwin and MinGW targets. It
22815 specifies that the @code{dllimport} attribute should be ignored.
22819 This option is available for MinGW targets. It specifies
22820 that MinGW-specific thread support is to be used.
22824 This option is available for MinGW-w64 targets. It causes
22825 the @code{UNICODE} preprocessor macro to be predefined, and
22826 chooses Unicode-capable runtime startup code.
22830 This option is available for Cygwin and MinGW targets. It
22831 specifies that the typical Microsoft Windows predefined macros are to
22832 be set in the pre-processor, but does not influence the choice
22833 of runtime library/startup code.
22837 This option is available for Cygwin and MinGW targets. It
22838 specifies that a GUI application is to be generated by
22839 instructing the linker to set the PE header subsystem type
22842 @item -fno-set-stack-executable
22843 @opindex fno-set-stack-executable
22844 This option is available for MinGW targets. It specifies that
22845 the executable flag for the stack used by nested functions isn't
22846 set. This is necessary for binaries running in kernel mode of
22847 Microsoft Windows, as there the User32 API, which is used to set executable
22848 privileges, isn't available.
22850 @item -fwritable-relocated-rdata
22851 @opindex fno-writable-relocated-rdata
22852 This option is available for MinGW and Cygwin targets. It specifies
22853 that relocated-data in read-only section is put into .data
22854 section. This is a necessary for older runtimes not supporting
22855 modification of .rdata sections for pseudo-relocation.
22857 @item -mpe-aligned-commons
22858 @opindex mpe-aligned-commons
22859 This option is available for Cygwin and MinGW targets. It
22860 specifies that the GNU extension to the PE file format that
22861 permits the correct alignment of COMMON variables should be
22862 used when generating code. It is enabled by default if
22863 GCC detects that the target assembler found during configuration
22864 supports the feature.
22867 See also under @ref{x86 Options} for standard options.
22869 @node Xstormy16 Options
22870 @subsection Xstormy16 Options
22871 @cindex Xstormy16 Options
22873 These options are defined for Xstormy16:
22878 Choose startup files and linker script suitable for the simulator.
22881 @node Xtensa Options
22882 @subsection Xtensa Options
22883 @cindex Xtensa Options
22885 These options are supported for Xtensa targets:
22889 @itemx -mno-const16
22891 @opindex mno-const16
22892 Enable or disable use of @code{CONST16} instructions for loading
22893 constant values. The @code{CONST16} instruction is currently not a
22894 standard option from Tensilica. When enabled, @code{CONST16}
22895 instructions are always used in place of the standard @code{L32R}
22896 instructions. The use of @code{CONST16} is enabled by default only if
22897 the @code{L32R} instruction is not available.
22900 @itemx -mno-fused-madd
22901 @opindex mfused-madd
22902 @opindex mno-fused-madd
22903 Enable or disable use of fused multiply/add and multiply/subtract
22904 instructions in the floating-point option. This has no effect if the
22905 floating-point option is not also enabled. Disabling fused multiply/add
22906 and multiply/subtract instructions forces the compiler to use separate
22907 instructions for the multiply and add/subtract operations. This may be
22908 desirable in some cases where strict IEEE 754-compliant results are
22909 required: the fused multiply add/subtract instructions do not round the
22910 intermediate result, thereby producing results with @emph{more} bits of
22911 precision than specified by the IEEE standard. Disabling fused multiply
22912 add/subtract instructions also ensures that the program output is not
22913 sensitive to the compiler's ability to combine multiply and add/subtract
22916 @item -mserialize-volatile
22917 @itemx -mno-serialize-volatile
22918 @opindex mserialize-volatile
22919 @opindex mno-serialize-volatile
22920 When this option is enabled, GCC inserts @code{MEMW} instructions before
22921 @code{volatile} memory references to guarantee sequential consistency.
22922 The default is @option{-mserialize-volatile}. Use
22923 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
22925 @item -mforce-no-pic
22926 @opindex mforce-no-pic
22927 For targets, like GNU/Linux, where all user-mode Xtensa code must be
22928 position-independent code (PIC), this option disables PIC for compiling
22931 @item -mtext-section-literals
22932 @itemx -mno-text-section-literals
22933 @opindex mtext-section-literals
22934 @opindex mno-text-section-literals
22935 These options control the treatment of literal pools. The default is
22936 @option{-mno-text-section-literals}, which places literals in a separate
22937 section in the output file. This allows the literal pool to be placed
22938 in a data RAM/ROM, and it also allows the linker to combine literal
22939 pools from separate object files to remove redundant literals and
22940 improve code size. With @option{-mtext-section-literals}, the literals
22941 are interspersed in the text section in order to keep them as close as
22942 possible to their references. This may be necessary for large assembly
22945 @item -mtarget-align
22946 @itemx -mno-target-align
22947 @opindex mtarget-align
22948 @opindex mno-target-align
22949 When this option is enabled, GCC instructs the assembler to
22950 automatically align instructions to reduce branch penalties at the
22951 expense of some code density. The assembler attempts to widen density
22952 instructions to align branch targets and the instructions following call
22953 instructions. If there are not enough preceding safe density
22954 instructions to align a target, no widening is performed. The
22955 default is @option{-mtarget-align}. These options do not affect the
22956 treatment of auto-aligned instructions like @code{LOOP}, which the
22957 assembler always aligns, either by widening density instructions or
22958 by inserting NOP instructions.
22961 @itemx -mno-longcalls
22962 @opindex mlongcalls
22963 @opindex mno-longcalls
22964 When this option is enabled, GCC instructs the assembler to translate
22965 direct calls to indirect calls unless it can determine that the target
22966 of a direct call is in the range allowed by the call instruction. This
22967 translation typically occurs for calls to functions in other source
22968 files. Specifically, the assembler translates a direct @code{CALL}
22969 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
22970 The default is @option{-mno-longcalls}. This option should be used in
22971 programs where the call target can potentially be out of range. This
22972 option is implemented in the assembler, not the compiler, so the
22973 assembly code generated by GCC still shows direct call
22974 instructions---look at the disassembled object code to see the actual
22975 instructions. Note that the assembler uses an indirect call for
22976 every cross-file call, not just those that really are out of range.
22979 @node zSeries Options
22980 @subsection zSeries Options
22981 @cindex zSeries options
22983 These are listed under @xref{S/390 and zSeries Options}.
22985 @node Code Gen Options
22986 @section Options for Code Generation Conventions
22987 @cindex code generation conventions
22988 @cindex options, code generation
22989 @cindex run-time options
22991 These machine-independent options control the interface conventions
22992 used in code generation.
22994 Most of them have both positive and negative forms; the negative form
22995 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
22996 one of the forms is listed---the one that is not the default. You
22997 can figure out the other form by either removing @samp{no-} or adding
23001 @item -fbounds-check
23002 @opindex fbounds-check
23003 For front ends that support it, generate additional code to check that
23004 indices used to access arrays are within the declared range. This is
23005 currently only supported by the Java and Fortran front ends, where
23006 this option defaults to true and false respectively.
23008 @item -fstack-reuse=@var{reuse-level}
23009 @opindex fstack_reuse
23010 This option controls stack space reuse for user declared local/auto variables
23011 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
23012 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
23013 local variables and temporaries, @samp{named_vars} enables the reuse only for
23014 user defined local variables with names, and @samp{none} disables stack reuse
23015 completely. The default value is @samp{all}. The option is needed when the
23016 program extends the lifetime of a scoped local variable or a compiler generated
23017 temporary beyond the end point defined by the language. When a lifetime of
23018 a variable ends, and if the variable lives in memory, the optimizing compiler
23019 has the freedom to reuse its stack space with other temporaries or scoped
23020 local variables whose live range does not overlap with it. Legacy code extending
23021 local lifetime is likely to break with the stack reuse optimization.
23040 if (*p == 10) // out of scope use of local1
23051 A(int k) : i(k), j(k) @{ @}
23058 void foo(const A& ar)
23065 foo(A(10)); // temp object's lifetime ends when foo returns
23071 ap->i+= 10; // ap references out of scope temp whose space
23072 // is reused with a. What is the value of ap->i?
23077 The lifetime of a compiler generated temporary is well defined by the C++
23078 standard. When a lifetime of a temporary ends, and if the temporary lives
23079 in memory, the optimizing compiler has the freedom to reuse its stack
23080 space with other temporaries or scoped local variables whose live range
23081 does not overlap with it. However some of the legacy code relies on
23082 the behavior of older compilers in which temporaries' stack space is
23083 not reused, the aggressive stack reuse can lead to runtime errors. This
23084 option is used to control the temporary stack reuse optimization.
23088 This option generates traps for signed overflow on addition, subtraction,
23089 multiplication operations.
23093 This option instructs the compiler to assume that signed arithmetic
23094 overflow of addition, subtraction and multiplication wraps around
23095 using twos-complement representation. This flag enables some optimizations
23096 and disables others. This option is enabled by default for the Java
23097 front end, as required by the Java language specification.
23100 @opindex fexceptions
23101 Enable exception handling. Generates extra code needed to propagate
23102 exceptions. For some targets, this implies GCC generates frame
23103 unwind information for all functions, which can produce significant data
23104 size overhead, although it does not affect execution. If you do not
23105 specify this option, GCC enables it by default for languages like
23106 C++ that normally require exception handling, and disables it for
23107 languages like C that do not normally require it. However, you may need
23108 to enable this option when compiling C code that needs to interoperate
23109 properly with exception handlers written in C++. You may also wish to
23110 disable this option if you are compiling older C++ programs that don't
23111 use exception handling.
23113 @item -fnon-call-exceptions
23114 @opindex fnon-call-exceptions
23115 Generate code that allows trapping instructions to throw exceptions.
23116 Note that this requires platform-specific runtime support that does
23117 not exist everywhere. Moreover, it only allows @emph{trapping}
23118 instructions to throw exceptions, i.e.@: memory references or floating-point
23119 instructions. It does not allow exceptions to be thrown from
23120 arbitrary signal handlers such as @code{SIGALRM}.
23122 @item -fdelete-dead-exceptions
23123 @opindex fdelete-dead-exceptions
23124 Consider that instructions that may throw exceptions but don't otherwise
23125 contribute to the execution of the program can be optimized away.
23126 This option is enabled by default for the Ada front end, as permitted by
23127 the Ada language specification.
23128 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23130 @item -funwind-tables
23131 @opindex funwind-tables
23132 Similar to @option{-fexceptions}, except that it just generates any needed
23133 static data, but does not affect the generated code in any other way.
23134 You normally do not need to enable this option; instead, a language processor
23135 that needs this handling enables it on your behalf.
23137 @item -fasynchronous-unwind-tables
23138 @opindex fasynchronous-unwind-tables
23139 Generate unwind table in DWARF 2 format, if supported by target machine. The
23140 table is exact at each instruction boundary, so it can be used for stack
23141 unwinding from asynchronous events (such as debugger or garbage collector).
23143 @item -fno-gnu-unique
23144 @opindex fno-gnu-unique
23145 On systems with recent GNU assembler and C library, the C++ compiler
23146 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23147 of template static data members and static local variables in inline
23148 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23149 is necessary to avoid problems with a library used by two different
23150 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23151 therefore disagreeing with the other one about the binding of the
23152 symbol. But this causes @code{dlclose} to be ignored for affected
23153 DSOs; if your program relies on reinitialization of a DSO via
23154 @code{dlclose} and @code{dlopen}, you can use
23155 @option{-fno-gnu-unique}.
23157 @item -fpcc-struct-return
23158 @opindex fpcc-struct-return
23159 Return ``short'' @code{struct} and @code{union} values in memory like
23160 longer ones, rather than in registers. This convention is less
23161 efficient, but it has the advantage of allowing intercallability between
23162 GCC-compiled files and files compiled with other compilers, particularly
23163 the Portable C Compiler (pcc).
23165 The precise convention for returning structures in memory depends
23166 on the target configuration macros.
23168 Short structures and unions are those whose size and alignment match
23169 that of some integer type.
23171 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23172 switch is not binary compatible with code compiled with the
23173 @option{-freg-struct-return} switch.
23174 Use it to conform to a non-default application binary interface.
23176 @item -freg-struct-return
23177 @opindex freg-struct-return
23178 Return @code{struct} and @code{union} values in registers when possible.
23179 This is more efficient for small structures than
23180 @option{-fpcc-struct-return}.
23182 If you specify neither @option{-fpcc-struct-return} nor
23183 @option{-freg-struct-return}, GCC defaults to whichever convention is
23184 standard for the target. If there is no standard convention, GCC
23185 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23186 the principal compiler. In those cases, we can choose the standard, and
23187 we chose the more efficient register return alternative.
23189 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23190 switch is not binary compatible with code compiled with the
23191 @option{-fpcc-struct-return} switch.
23192 Use it to conform to a non-default application binary interface.
23194 @item -fshort-enums
23195 @opindex fshort-enums
23196 Allocate to an @code{enum} type only as many bytes as it needs for the
23197 declared range of possible values. Specifically, the @code{enum} type
23198 is equivalent to the smallest integer type that has enough room.
23200 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23201 code that is not binary compatible with code generated without that switch.
23202 Use it to conform to a non-default application binary interface.
23204 @item -fshort-double
23205 @opindex fshort-double
23206 Use the same size for @code{double} as for @code{float}.
23208 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23209 code that is not binary compatible with code generated without that switch.
23210 Use it to conform to a non-default application binary interface.
23212 @item -fshort-wchar
23213 @opindex fshort-wchar
23214 Override the underlying type for @code{wchar_t} to be @code{short
23215 unsigned int} instead of the default for the target. This option is
23216 useful for building programs to run under WINE@.
23218 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23219 code that is not binary compatible with code generated without that switch.
23220 Use it to conform to a non-default application binary interface.
23223 @opindex fno-common
23224 In C code, controls the placement of uninitialized global variables.
23225 Unix C compilers have traditionally permitted multiple definitions of
23226 such variables in different compilation units by placing the variables
23228 This is the behavior specified by @option{-fcommon}, and is the default
23229 for GCC on most targets.
23230 On the other hand, this behavior is not required by ISO C, and on some
23231 targets may carry a speed or code size penalty on variable references.
23232 The @option{-fno-common} option specifies that the compiler should place
23233 uninitialized global variables in the data section of the object file,
23234 rather than generating them as common blocks.
23235 This has the effect that if the same variable is declared
23236 (without @code{extern}) in two different compilations,
23237 you get a multiple-definition error when you link them.
23238 In this case, you must compile with @option{-fcommon} instead.
23239 Compiling with @option{-fno-common} is useful on targets for which
23240 it provides better performance, or if you wish to verify that the
23241 program will work on other systems that always treat uninitialized
23242 variable declarations this way.
23246 Ignore the @code{#ident} directive.
23248 @item -finhibit-size-directive
23249 @opindex finhibit-size-directive
23250 Don't output a @code{.size} assembler directive, or anything else that
23251 would cause trouble if the function is split in the middle, and the
23252 two halves are placed at locations far apart in memory. This option is
23253 used when compiling @file{crtstuff.c}; you should not need to use it
23256 @item -fverbose-asm
23257 @opindex fverbose-asm
23258 Put extra commentary information in the generated assembly code to
23259 make it more readable. This option is generally only of use to those
23260 who actually need to read the generated assembly code (perhaps while
23261 debugging the compiler itself).
23263 @option{-fno-verbose-asm}, the default, causes the
23264 extra information to be omitted and is useful when comparing two assembler
23267 @item -frecord-gcc-switches
23268 @opindex frecord-gcc-switches
23269 This switch causes the command line used to invoke the
23270 compiler to be recorded into the object file that is being created.
23271 This switch is only implemented on some targets and the exact format
23272 of the recording is target and binary file format dependent, but it
23273 usually takes the form of a section containing ASCII text. This
23274 switch is related to the @option{-fverbose-asm} switch, but that
23275 switch only records information in the assembler output file as
23276 comments, so it never reaches the object file.
23277 See also @option{-grecord-gcc-switches} for another
23278 way of storing compiler options into the object file.
23282 @cindex global offset table
23284 Generate position-independent code (PIC) suitable for use in a shared
23285 library, if supported for the target machine. Such code accesses all
23286 constant addresses through a global offset table (GOT)@. The dynamic
23287 loader resolves the GOT entries when the program starts (the dynamic
23288 loader is not part of GCC; it is part of the operating system). If
23289 the GOT size for the linked executable exceeds a machine-specific
23290 maximum size, you get an error message from the linker indicating that
23291 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23292 instead. (These maximums are 8k on the SPARC and 32k
23293 on the m68k and RS/6000. The x86 has no such limit.)
23295 Position-independent code requires special support, and therefore works
23296 only on certain machines. For the x86, GCC supports PIC for System V
23297 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23298 position-independent.
23300 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23305 If supported for the target machine, emit position-independent code,
23306 suitable for dynamic linking and avoiding any limit on the size of the
23307 global offset table. This option makes a difference on the m68k,
23308 PowerPC and SPARC@.
23310 Position-independent code requires special support, and therefore works
23311 only on certain machines.
23313 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23320 These options are similar to @option{-fpic} and @option{-fPIC}, but
23321 generated position independent code can be only linked into executables.
23322 Usually these options are used when @option{-pie} GCC option is
23323 used during linking.
23325 @option{-fpie} and @option{-fPIE} both define the macros
23326 @code{__pie__} and @code{__PIE__}. The macros have the value 1
23327 for @option{-fpie} and 2 for @option{-fPIE}.
23329 @item -fno-jump-tables
23330 @opindex fno-jump-tables
23331 Do not use jump tables for switch statements even where it would be
23332 more efficient than other code generation strategies. This option is
23333 of use in conjunction with @option{-fpic} or @option{-fPIC} for
23334 building code that forms part of a dynamic linker and cannot
23335 reference the address of a jump table. On some targets, jump tables
23336 do not require a GOT and this option is not needed.
23338 @item -ffixed-@var{reg}
23340 Treat the register named @var{reg} as a fixed register; generated code
23341 should never refer to it (except perhaps as a stack pointer, frame
23342 pointer or in some other fixed role).
23344 @var{reg} must be the name of a register. The register names accepted
23345 are machine-specific and are defined in the @code{REGISTER_NAMES}
23346 macro in the machine description macro file.
23348 This flag does not have a negative form, because it specifies a
23351 @item -fcall-used-@var{reg}
23352 @opindex fcall-used
23353 Treat the register named @var{reg} as an allocable register that is
23354 clobbered by function calls. It may be allocated for temporaries or
23355 variables that do not live across a call. Functions compiled this way
23356 do not save and restore the register @var{reg}.
23358 It is an error to use this flag with the frame pointer or stack pointer.
23359 Use of this flag for other registers that have fixed pervasive roles in
23360 the machine's execution model produces disastrous results.
23362 This flag does not have a negative form, because it specifies a
23365 @item -fcall-saved-@var{reg}
23366 @opindex fcall-saved
23367 Treat the register named @var{reg} as an allocable register saved by
23368 functions. It may be allocated even for temporaries or variables that
23369 live across a call. Functions compiled this way save and restore
23370 the register @var{reg} if they use it.
23372 It is an error to use this flag with the frame pointer or stack pointer.
23373 Use of this flag for other registers that have fixed pervasive roles in
23374 the machine's execution model produces disastrous results.
23376 A different sort of disaster results from the use of this flag for
23377 a register in which function values may be returned.
23379 This flag does not have a negative form, because it specifies a
23382 @item -fpack-struct[=@var{n}]
23383 @opindex fpack-struct
23384 Without a value specified, pack all structure members together without
23385 holes. When a value is specified (which must be a small power of two), pack
23386 structure members according to this value, representing the maximum
23387 alignment (that is, objects with default alignment requirements larger than
23388 this are output potentially unaligned at the next fitting location.
23390 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
23391 code that is not binary compatible with code generated without that switch.
23392 Additionally, it makes the code suboptimal.
23393 Use it to conform to a non-default application binary interface.
23395 @item -finstrument-functions
23396 @opindex finstrument-functions
23397 Generate instrumentation calls for entry and exit to functions. Just
23398 after function entry and just before function exit, the following
23399 profiling functions are called with the address of the current
23400 function and its call site. (On some platforms,
23401 @code{__builtin_return_address} does not work beyond the current
23402 function, so the call site information may not be available to the
23403 profiling functions otherwise.)
23406 void __cyg_profile_func_enter (void *this_fn,
23408 void __cyg_profile_func_exit (void *this_fn,
23412 The first argument is the address of the start of the current function,
23413 which may be looked up exactly in the symbol table.
23415 This instrumentation is also done for functions expanded inline in other
23416 functions. The profiling calls indicate where, conceptually, the
23417 inline function is entered and exited. This means that addressable
23418 versions of such functions must be available. If all your uses of a
23419 function are expanded inline, this may mean an additional expansion of
23420 code size. If you use @code{extern inline} in your C code, an
23421 addressable version of such functions must be provided. (This is
23422 normally the case anyway, but if you get lucky and the optimizer always
23423 expands the functions inline, you might have gotten away without
23424 providing static copies.)
23426 A function may be given the attribute @code{no_instrument_function}, in
23427 which case this instrumentation is not done. This can be used, for
23428 example, for the profiling functions listed above, high-priority
23429 interrupt routines, and any functions from which the profiling functions
23430 cannot safely be called (perhaps signal handlers, if the profiling
23431 routines generate output or allocate memory).
23433 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23434 @opindex finstrument-functions-exclude-file-list
23436 Set the list of functions that are excluded from instrumentation (see
23437 the description of @option{-finstrument-functions}). If the file that
23438 contains a function definition matches with one of @var{file}, then
23439 that function is not instrumented. The match is done on substrings:
23440 if the @var{file} parameter is a substring of the file name, it is
23441 considered to be a match.
23446 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
23450 excludes any inline function defined in files whose pathnames
23451 contain @file{/bits/stl} or @file{include/sys}.
23453 If, for some reason, you want to include letter @samp{,} in one of
23454 @var{sym}, write @samp{\,}. For example,
23455 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
23456 (note the single quote surrounding the option).
23458 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
23459 @opindex finstrument-functions-exclude-function-list
23461 This is similar to @option{-finstrument-functions-exclude-file-list},
23462 but this option sets the list of function names to be excluded from
23463 instrumentation. The function name to be matched is its user-visible
23464 name, such as @code{vector<int> blah(const vector<int> &)}, not the
23465 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
23466 match is done on substrings: if the @var{sym} parameter is a substring
23467 of the function name, it is considered to be a match. For C99 and C++
23468 extended identifiers, the function name must be given in UTF-8, not
23469 using universal character names.
23471 @item -fstack-check
23472 @opindex fstack-check
23473 Generate code to verify that you do not go beyond the boundary of the
23474 stack. You should specify this flag if you are running in an
23475 environment with multiple threads, but you only rarely need to specify it in
23476 a single-threaded environment since stack overflow is automatically
23477 detected on nearly all systems if there is only one stack.
23479 Note that this switch does not actually cause checking to be done; the
23480 operating system or the language runtime must do that. The switch causes
23481 generation of code to ensure that they see the stack being extended.
23483 You can additionally specify a string parameter: @samp{no} means no
23484 checking, @samp{generic} means force the use of old-style checking,
23485 @samp{specific} means use the best checking method and is equivalent
23486 to bare @option{-fstack-check}.
23488 Old-style checking is a generic mechanism that requires no specific
23489 target support in the compiler but comes with the following drawbacks:
23493 Modified allocation strategy for large objects: they are always
23494 allocated dynamically if their size exceeds a fixed threshold.
23497 Fixed limit on the size of the static frame of functions: when it is
23498 topped by a particular function, stack checking is not reliable and
23499 a warning is issued by the compiler.
23502 Inefficiency: because of both the modified allocation strategy and the
23503 generic implementation, code performance is hampered.
23506 Note that old-style stack checking is also the fallback method for
23507 @samp{specific} if no target support has been added in the compiler.
23509 @item -fstack-limit-register=@var{reg}
23510 @itemx -fstack-limit-symbol=@var{sym}
23511 @itemx -fno-stack-limit
23512 @opindex fstack-limit-register
23513 @opindex fstack-limit-symbol
23514 @opindex fno-stack-limit
23515 Generate code to ensure that the stack does not grow beyond a certain value,
23516 either the value of a register or the address of a symbol. If a larger
23517 stack is required, a signal is raised at run time. For most targets,
23518 the signal is raised before the stack overruns the boundary, so
23519 it is possible to catch the signal without taking special precautions.
23521 For instance, if the stack starts at absolute address @samp{0x80000000}
23522 and grows downwards, you can use the flags
23523 @option{-fstack-limit-symbol=__stack_limit} and
23524 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
23525 of 128KB@. Note that this may only work with the GNU linker.
23527 @item -fsplit-stack
23528 @opindex fsplit-stack
23529 Generate code to automatically split the stack before it overflows.
23530 The resulting program has a discontiguous stack which can only
23531 overflow if the program is unable to allocate any more memory. This
23532 is most useful when running threaded programs, as it is no longer
23533 necessary to calculate a good stack size to use for each thread. This
23534 is currently only implemented for the x86 targets running
23537 When code compiled with @option{-fsplit-stack} calls code compiled
23538 without @option{-fsplit-stack}, there may not be much stack space
23539 available for the latter code to run. If compiling all code,
23540 including library code, with @option{-fsplit-stack} is not an option,
23541 then the linker can fix up these calls so that the code compiled
23542 without @option{-fsplit-stack} always has a large stack. Support for
23543 this is implemented in the gold linker in GNU binutils release 2.21
23546 @item -fleading-underscore
23547 @opindex fleading-underscore
23548 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
23549 change the way C symbols are represented in the object file. One use
23550 is to help link with legacy assembly code.
23552 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
23553 generate code that is not binary compatible with code generated without that
23554 switch. Use it to conform to a non-default application binary interface.
23555 Not all targets provide complete support for this switch.
23557 @item -ftls-model=@var{model}
23558 @opindex ftls-model
23559 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
23560 The @var{model} argument should be one of @samp{global-dynamic},
23561 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
23562 Note that the choice is subject to optimization: the compiler may use
23563 a more efficient model for symbols not visible outside of the translation
23564 unit, or if @option{-fpic} is not given on the command line.
23566 The default without @option{-fpic} is @samp{initial-exec}; with
23567 @option{-fpic} the default is @samp{global-dynamic}.
23569 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
23570 @opindex fvisibility
23571 Set the default ELF image symbol visibility to the specified option---all
23572 symbols are marked with this unless overridden within the code.
23573 Using this feature can very substantially improve linking and
23574 load times of shared object libraries, produce more optimized
23575 code, provide near-perfect API export and prevent symbol clashes.
23576 It is @strong{strongly} recommended that you use this in any shared objects
23579 Despite the nomenclature, @samp{default} always means public; i.e.,
23580 available to be linked against from outside the shared object.
23581 @samp{protected} and @samp{internal} are pretty useless in real-world
23582 usage so the only other commonly used option is @samp{hidden}.
23583 The default if @option{-fvisibility} isn't specified is
23584 @samp{default}, i.e., make every
23585 symbol public---this causes the same behavior as previous versions of
23588 A good explanation of the benefits offered by ensuring ELF
23589 symbols have the correct visibility is given by ``How To Write
23590 Shared Libraries'' by Ulrich Drepper (which can be found at
23591 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
23592 solution made possible by this option to marking things hidden when
23593 the default is public is to make the default hidden and mark things
23594 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
23595 and @code{__attribute__ ((visibility("default")))} instead of
23596 @code{__declspec(dllexport)} you get almost identical semantics with
23597 identical syntax. This is a great boon to those working with
23598 cross-platform projects.
23600 For those adding visibility support to existing code, you may find
23601 @code{#pragma GCC visibility} of use. This works by you enclosing
23602 the declarations you wish to set visibility for with (for example)
23603 @code{#pragma GCC visibility push(hidden)} and
23604 @code{#pragma GCC visibility pop}.
23605 Bear in mind that symbol visibility should be viewed @strong{as
23606 part of the API interface contract} and thus all new code should
23607 always specify visibility when it is not the default; i.e., declarations
23608 only for use within the local DSO should @strong{always} be marked explicitly
23609 as hidden as so to avoid PLT indirection overheads---making this
23610 abundantly clear also aids readability and self-documentation of the code.
23611 Note that due to ISO C++ specification requirements, @code{operator new} and
23612 @code{operator delete} must always be of default visibility.
23614 Be aware that headers from outside your project, in particular system
23615 headers and headers from any other library you use, may not be
23616 expecting to be compiled with visibility other than the default. You
23617 may need to explicitly say @code{#pragma GCC visibility push(default)}
23618 before including any such headers.
23620 @code{extern} declarations are not affected by @option{-fvisibility}, so
23621 a lot of code can be recompiled with @option{-fvisibility=hidden} with
23622 no modifications. However, this means that calls to @code{extern}
23623 functions with no explicit visibility use the PLT, so it is more
23624 effective to use @code{__attribute ((visibility))} and/or
23625 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
23626 declarations should be treated as hidden.
23628 Note that @option{-fvisibility} does affect C++ vague linkage
23629 entities. This means that, for instance, an exception class that is
23630 be thrown between DSOs must be explicitly marked with default
23631 visibility so that the @samp{type_info} nodes are unified between
23634 An overview of these techniques, their benefits and how to use them
23635 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
23637 @item -fstrict-volatile-bitfields
23638 @opindex fstrict-volatile-bitfields
23639 This option should be used if accesses to volatile bit-fields (or other
23640 structure fields, although the compiler usually honors those types
23641 anyway) should use a single access of the width of the
23642 field's type, aligned to a natural alignment if possible. For
23643 example, targets with memory-mapped peripheral registers might require
23644 all such accesses to be 16 bits wide; with this flag you can
23645 declare all peripheral bit-fields as @code{unsigned short} (assuming short
23646 is 16 bits on these targets) to force GCC to use 16-bit accesses
23647 instead of, perhaps, a more efficient 32-bit access.
23649 If this option is disabled, the compiler uses the most efficient
23650 instruction. In the previous example, that might be a 32-bit load
23651 instruction, even though that accesses bytes that do not contain
23652 any portion of the bit-field, or memory-mapped registers unrelated to
23653 the one being updated.
23655 In some cases, such as when the @code{packed} attribute is applied to a
23656 structure field, it may not be possible to access the field with a single
23657 read or write that is correctly aligned for the target machine. In this
23658 case GCC falls back to generating multiple accesses rather than code that
23659 will fault or truncate the result at run time.
23661 Note: Due to restrictions of the C/C++11 memory model, write accesses are
23662 not allowed to touch non bit-field members. It is therefore recommended
23663 to define all bits of the field's type as bit-field members.
23665 The default value of this option is determined by the application binary
23666 interface for the target processor.
23668 @item -fsync-libcalls
23669 @opindex fsync-libcalls
23670 This option controls whether any out-of-line instance of the @code{__sync}
23671 family of functions may be used to implement the C++11 @code{__atomic}
23672 family of functions.
23674 The default value of this option is enabled, thus the only useful form
23675 of the option is @option{-fno-sync-libcalls}. This option is used in
23676 the implementation of the @file{libatomic} runtime library.
23682 @node Environment Variables
23683 @section Environment Variables Affecting GCC
23684 @cindex environment variables
23686 @c man begin ENVIRONMENT
23687 This section describes several environment variables that affect how GCC
23688 operates. Some of them work by specifying directories or prefixes to use
23689 when searching for various kinds of files. Some are used to specify other
23690 aspects of the compilation environment.
23692 Note that you can also specify places to search using options such as
23693 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
23694 take precedence over places specified using environment variables, which
23695 in turn take precedence over those specified by the configuration of GCC@.
23696 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
23697 GNU Compiler Collection (GCC) Internals}.
23702 @c @itemx LC_COLLATE
23704 @c @itemx LC_MONETARY
23705 @c @itemx LC_NUMERIC
23710 @c @findex LC_COLLATE
23711 @findex LC_MESSAGES
23712 @c @findex LC_MONETARY
23713 @c @findex LC_NUMERIC
23717 These environment variables control the way that GCC uses
23718 localization information which allows GCC to work with different
23719 national conventions. GCC inspects the locale categories
23720 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
23721 so. These locale categories can be set to any value supported by your
23722 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
23723 Kingdom encoded in UTF-8.
23725 The @env{LC_CTYPE} environment variable specifies character
23726 classification. GCC uses it to determine the character boundaries in
23727 a string; this is needed for some multibyte encodings that contain quote
23728 and escape characters that are otherwise interpreted as a string
23731 The @env{LC_MESSAGES} environment variable specifies the language to
23732 use in diagnostic messages.
23734 If the @env{LC_ALL} environment variable is set, it overrides the value
23735 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
23736 and @env{LC_MESSAGES} default to the value of the @env{LANG}
23737 environment variable. If none of these variables are set, GCC
23738 defaults to traditional C English behavior.
23742 If @env{TMPDIR} is set, it specifies the directory to use for temporary
23743 files. GCC uses temporary files to hold the output of one stage of
23744 compilation which is to be used as input to the next stage: for example,
23745 the output of the preprocessor, which is the input to the compiler
23748 @item GCC_COMPARE_DEBUG
23749 @findex GCC_COMPARE_DEBUG
23750 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
23751 @option{-fcompare-debug} to the compiler driver. See the documentation
23752 of this option for more details.
23754 @item GCC_EXEC_PREFIX
23755 @findex GCC_EXEC_PREFIX
23756 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
23757 names of the subprograms executed by the compiler. No slash is added
23758 when this prefix is combined with the name of a subprogram, but you can
23759 specify a prefix that ends with a slash if you wish.
23761 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
23762 an appropriate prefix to use based on the pathname it is invoked with.
23764 If GCC cannot find the subprogram using the specified prefix, it
23765 tries looking in the usual places for the subprogram.
23767 The default value of @env{GCC_EXEC_PREFIX} is
23768 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
23769 the installed compiler. In many cases @var{prefix} is the value
23770 of @code{prefix} when you ran the @file{configure} script.
23772 Other prefixes specified with @option{-B} take precedence over this prefix.
23774 This prefix is also used for finding files such as @file{crt0.o} that are
23777 In addition, the prefix is used in an unusual way in finding the
23778 directories to search for header files. For each of the standard
23779 directories whose name normally begins with @samp{/usr/local/lib/gcc}
23780 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
23781 replacing that beginning with the specified prefix to produce an
23782 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
23783 @file{foo/bar} just before it searches the standard directory
23784 @file{/usr/local/lib/bar}.
23785 If a standard directory begins with the configured
23786 @var{prefix} then the value of @var{prefix} is replaced by
23787 @env{GCC_EXEC_PREFIX} when looking for header files.
23789 @item COMPILER_PATH
23790 @findex COMPILER_PATH
23791 The value of @env{COMPILER_PATH} is a colon-separated list of
23792 directories, much like @env{PATH}. GCC tries the directories thus
23793 specified when searching for subprograms, if it can't find the
23794 subprograms using @env{GCC_EXEC_PREFIX}.
23797 @findex LIBRARY_PATH
23798 The value of @env{LIBRARY_PATH} is a colon-separated list of
23799 directories, much like @env{PATH}. When configured as a native compiler,
23800 GCC tries the directories thus specified when searching for special
23801 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
23802 using GCC also uses these directories when searching for ordinary
23803 libraries for the @option{-l} option (but directories specified with
23804 @option{-L} come first).
23808 @cindex locale definition
23809 This variable is used to pass locale information to the compiler. One way in
23810 which this information is used is to determine the character set to be used
23811 when character literals, string literals and comments are parsed in C and C++.
23812 When the compiler is configured to allow multibyte characters,
23813 the following values for @env{LANG} are recognized:
23817 Recognize JIS characters.
23819 Recognize SJIS characters.
23821 Recognize EUCJP characters.
23824 If @env{LANG} is not defined, or if it has some other value, then the
23825 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
23826 recognize and translate multibyte characters.
23830 Some additional environment variables affect the behavior of the
23833 @include cppenv.texi
23837 @node Precompiled Headers
23838 @section Using Precompiled Headers
23839 @cindex precompiled headers
23840 @cindex speed of compilation
23842 Often large projects have many header files that are included in every
23843 source file. The time the compiler takes to process these header files
23844 over and over again can account for nearly all of the time required to
23845 build the project. To make builds faster, GCC allows you to
23846 @dfn{precompile} a header file.
23848 To create a precompiled header file, simply compile it as you would any
23849 other file, if necessary using the @option{-x} option to make the driver
23850 treat it as a C or C++ header file. You may want to use a
23851 tool like @command{make} to keep the precompiled header up-to-date when
23852 the headers it contains change.
23854 A precompiled header file is searched for when @code{#include} is
23855 seen in the compilation. As it searches for the included file
23856 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
23857 compiler looks for a precompiled header in each directory just before it
23858 looks for the include file in that directory. The name searched for is
23859 the name specified in the @code{#include} with @samp{.gch} appended. If
23860 the precompiled header file can't be used, it is ignored.
23862 For instance, if you have @code{#include "all.h"}, and you have
23863 @file{all.h.gch} in the same directory as @file{all.h}, then the
23864 precompiled header file is used if possible, and the original
23865 header is used otherwise.
23867 Alternatively, you might decide to put the precompiled header file in a
23868 directory and use @option{-I} to ensure that directory is searched
23869 before (or instead of) the directory containing the original header.
23870 Then, if you want to check that the precompiled header file is always
23871 used, you can put a file of the same name as the original header in this
23872 directory containing an @code{#error} command.
23874 This also works with @option{-include}. So yet another way to use
23875 precompiled headers, good for projects not designed with precompiled
23876 header files in mind, is to simply take most of the header files used by
23877 a project, include them from another header file, precompile that header
23878 file, and @option{-include} the precompiled header. If the header files
23879 have guards against multiple inclusion, they are skipped because
23880 they've already been included (in the precompiled header).
23882 If you need to precompile the same header file for different
23883 languages, targets, or compiler options, you can instead make a
23884 @emph{directory} named like @file{all.h.gch}, and put each precompiled
23885 header in the directory, perhaps using @option{-o}. It doesn't matter
23886 what you call the files in the directory; every precompiled header in
23887 the directory is considered. The first precompiled header
23888 encountered in the directory that is valid for this compilation is
23889 used; they're searched in no particular order.
23891 There are many other possibilities, limited only by your imagination,
23892 good sense, and the constraints of your build system.
23894 A precompiled header file can be used only when these conditions apply:
23898 Only one precompiled header can be used in a particular compilation.
23901 A precompiled header can't be used once the first C token is seen. You
23902 can have preprocessor directives before a precompiled header; you cannot
23903 include a precompiled header from inside another header.
23906 The precompiled header file must be produced for the same language as
23907 the current compilation. You can't use a C precompiled header for a C++
23911 The precompiled header file must have been produced by the same compiler
23912 binary as the current compilation is using.
23915 Any macros defined before the precompiled header is included must
23916 either be defined in the same way as when the precompiled header was
23917 generated, or must not affect the precompiled header, which usually
23918 means that they don't appear in the precompiled header at all.
23920 The @option{-D} option is one way to define a macro before a
23921 precompiled header is included; using a @code{#define} can also do it.
23922 There are also some options that define macros implicitly, like
23923 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
23926 @item If debugging information is output when using the precompiled
23927 header, using @option{-g} or similar, the same kind of debugging information
23928 must have been output when building the precompiled header. However,
23929 a precompiled header built using @option{-g} can be used in a compilation
23930 when no debugging information is being output.
23932 @item The same @option{-m} options must generally be used when building
23933 and using the precompiled header. @xref{Submodel Options},
23934 for any cases where this rule is relaxed.
23936 @item Each of the following options must be the same when building and using
23937 the precompiled header:
23939 @gccoptlist{-fexceptions}
23942 Some other command-line options starting with @option{-f},
23943 @option{-p}, or @option{-O} must be defined in the same way as when
23944 the precompiled header was generated. At present, it's not clear
23945 which options are safe to change and which are not; the safest choice
23946 is to use exactly the same options when generating and using the
23947 precompiled header. The following are known to be safe:
23949 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
23950 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
23951 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
23956 For all of these except the last, the compiler automatically
23957 ignores the precompiled header if the conditions aren't met. If you
23958 find an option combination that doesn't work and doesn't cause the
23959 precompiled header to be ignored, please consider filing a bug report,
23962 If you do use differing options when generating and using the
23963 precompiled header, the actual behavior is a mixture of the
23964 behavior for the options. For instance, if you use @option{-g} to
23965 generate the precompiled header but not when using it, you may or may
23966 not get debugging information for routines in the precompiled header.