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} -Wabi-tag -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 @gol
266 -Wmisleading-indentation -Wmissing-braces @gol
267 -Wmissing-field-initializers -Wmissing-include-dirs @gol
268 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
269 -Wodr -Wno-overflow -Wopenmp-simd @gol
270 -Woverride-init-side-effects @gol
271 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
272 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
273 -Wpointer-arith -Wno-pointer-to-int-cast @gol
274 -Wredundant-decls -Wno-return-local-addr @gol
275 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
276 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
277 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
278 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
279 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
280 -Wstrict-aliasing=n @gol -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
281 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
282 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
283 -Wmissing-format-attribute @gol
284 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
285 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
286 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
287 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
288 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
289 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
290 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
291 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
292 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
293 -Wzero-as-null-pointer-constant}
295 @item C and Objective-C-only Warning Options
296 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
297 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
298 -Wold-style-declaration -Wold-style-definition @gol
299 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
300 -Wdeclaration-after-statement -Wpointer-sign}
302 @item Debugging Options
303 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
304 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
305 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
306 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1,s2,...} @gol
307 -fsanitize-undefined-trap-on-error @gol
308 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
309 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
310 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
311 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
312 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
313 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
314 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
315 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
316 -fchkp-use-wrappers @gol
317 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
318 -fdisable-ipa-@var{pass_name} @gol
319 -fdisable-rtl-@var{pass_name} @gol
320 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
321 -fdisable-tree-@var{pass_name} @gol
322 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
323 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
324 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
325 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
326 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
328 -fdump-statistics @gol
330 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
331 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
332 -fdump-tree-cfg -fdump-tree-alias @gol
334 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
335 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
336 -fdump-tree-gimple@r{[}-raw@r{]} @gol
337 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
338 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
339 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
340 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
341 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
342 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
343 -fdump-tree-nrv -fdump-tree-vect @gol
344 -fdump-tree-sink @gol
345 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
346 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
347 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
348 -fdump-tree-vtable-verify @gol
349 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
350 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
351 -fdump-final-insns=@var{file} @gol
352 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
353 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
354 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
355 -fenable-@var{kind}-@var{pass} @gol
356 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
357 -fdebug-types-section -fmem-report-wpa @gol
358 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
360 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
361 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
362 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
363 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
364 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
365 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
366 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
367 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
368 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
369 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
370 -fdebug-prefix-map=@var{old}=@var{new} @gol
371 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
372 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
373 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
374 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
375 -print-prog-name=@var{program} -print-search-dirs -Q @gol
376 -print-sysroot -print-sysroot-headers-suffix @gol
377 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
379 @item Optimization Options
380 @xref{Optimize Options,,Options that Control Optimization}.
381 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
382 -falign-jumps[=@var{n}] @gol
383 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
384 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
385 -fauto-inc-dec -fbranch-probabilities @gol
386 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
387 -fbtr-bb-exclusive -fcaller-saves @gol
388 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
389 -fcompare-elim -fcprop-registers -fcrossjumping @gol
390 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
391 -fcx-limited-range @gol
392 -fdata-sections -fdce -fdelayed-branch @gol
393 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
394 -fdevirtualize-at-ltrans -fdse @gol
395 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
396 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
397 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
398 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
399 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
400 -fif-conversion2 -findirect-inlining @gol
401 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
402 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
403 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
404 -fira-algorithm=@var{algorithm} @gol
405 -fira-region=@var{region} -fira-hoist-pressure @gol
406 -fira-loop-pressure -fno-ira-share-save-slots @gol
407 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
408 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
409 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
410 -flive-range-shrinkage @gol
411 -floop-block -floop-interchange -floop-strip-mine @gol
412 -floop-unroll-and-jam -floop-nest-optimize @gol
413 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
414 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
415 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
416 -fmove-loop-invariants -fno-branch-count-reg @gol
417 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
418 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
419 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
420 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
421 -fomit-frame-pointer -foptimize-sibling-calls @gol
422 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
423 -fprefetch-loop-arrays -fprofile-report @gol
424 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
425 -fprofile-generate=@var{path} @gol
426 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
427 -fprofile-reorder-functions @gol
428 -freciprocal-math -free -frename-registers -freorder-blocks @gol
429 -freorder-blocks-and-partition -freorder-functions @gol
430 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
431 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
432 -fsched-spec-load -fsched-spec-load-dangerous @gol
433 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
434 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
435 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
436 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
437 -fschedule-fusion @gol
438 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
439 -fselective-scheduling -fselective-scheduling2 @gol
440 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
441 -fsemantic-interposition @gol
442 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
443 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt @gol
444 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
445 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
446 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
447 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
448 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
449 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
450 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
451 -ftree-loop-if-convert-stores -ftree-loop-im @gol
452 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
453 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
454 -ftree-loop-vectorize @gol
455 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
456 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
457 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
458 -ftree-vectorize -ftree-vrp @gol
459 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
460 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
461 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
462 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
463 --param @var{name}=@var{value}
464 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
466 @item Preprocessor Options
467 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
468 @gccoptlist{-A@var{question}=@var{answer} @gol
469 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
470 -C -dD -dI -dM -dN @gol
471 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
472 -idirafter @var{dir} @gol
473 -include @var{file} -imacros @var{file} @gol
474 -iprefix @var{file} -iwithprefix @var{dir} @gol
475 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
476 -imultilib @var{dir} -isysroot @var{dir} @gol
477 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
478 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
479 -remap -trigraphs -undef -U@var{macro} @gol
480 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
482 @item Assembler Option
483 @xref{Assembler Options,,Passing Options to the Assembler}.
484 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
487 @xref{Link Options,,Options for Linking}.
488 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
489 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
490 -s -static -static-libgcc -static-libstdc++ @gol
491 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
492 -static-libmpx -static-libmpxwrappers @gol
493 -shared -shared-libgcc -symbolic @gol
494 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
495 -u @var{symbol} -z @var{keyword}}
497 @item Directory Options
498 @xref{Directory Options,,Options for Directory Search}.
499 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
500 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
501 --sysroot=@var{dir} --no-sysroot-suffix}
503 @item Machine Dependent Options
504 @xref{Submodel Options,,Hardware Models and Configurations}.
505 @c This list is ordered alphanumerically by subsection name.
506 @c Try and put the significant identifier (CPU or system) first,
507 @c so users have a clue at guessing where the ones they want will be.
509 @emph{AArch64 Options}
510 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
511 -mgeneral-regs-only @gol
512 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
514 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
515 -mtls-dialect=desc -mtls-dialect=traditional @gol
516 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
517 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
518 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
520 @emph{Adapteva Epiphany Options}
521 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
522 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
523 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
524 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
525 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
526 -msplit-vecmove-early -m1reg-@var{reg}}
529 @gccoptlist{-mbarrel-shifter @gol
530 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
531 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
532 -mea -mno-mpy -mmul32x16 -mmul64 @gol
533 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
534 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
535 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
536 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
537 -mucb-mcount -mvolatile-cache @gol
538 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
539 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
540 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
541 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
542 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
543 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
546 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
547 -mabi=@var{name} @gol
548 -mapcs-stack-check -mno-apcs-stack-check @gol
549 -mapcs-float -mno-apcs-float @gol
550 -mapcs-reentrant -mno-apcs-reentrant @gol
551 -msched-prolog -mno-sched-prolog @gol
552 -mlittle-endian -mbig-endian @gol
553 -mfloat-abi=@var{name} @gol
554 -mfp16-format=@var{name}
555 -mthumb-interwork -mno-thumb-interwork @gol
556 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
557 -mtune=@var{name} -mprint-tune-info @gol
558 -mstructure-size-boundary=@var{n} @gol
559 -mabort-on-noreturn @gol
560 -mlong-calls -mno-long-calls @gol
561 -msingle-pic-base -mno-single-pic-base @gol
562 -mpic-register=@var{reg} @gol
563 -mnop-fun-dllimport @gol
564 -mpoke-function-name @gol
566 -mtpcs-frame -mtpcs-leaf-frame @gol
567 -mcaller-super-interworking -mcallee-super-interworking @gol
568 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
569 -mword-relocations @gol
570 -mfix-cortex-m3-ldrd @gol
571 -munaligned-access @gol
572 -mneon-for-64bits @gol
573 -mslow-flash-data @gol
574 -masm-syntax-unified @gol
578 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
579 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
580 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
582 @emph{Blackfin Options}
583 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
584 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
585 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
586 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
587 -mno-id-shared-library -mshared-library-id=@var{n} @gol
588 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
589 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
590 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
594 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
595 -msim -msdata=@var{sdata-type}}
598 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
599 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
600 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
601 -mstack-align -mdata-align -mconst-align @gol
602 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
603 -melf -maout -melinux -mlinux -sim -sim2 @gol
604 -mmul-bug-workaround -mno-mul-bug-workaround}
607 @gccoptlist{-mmac @gol
608 -mcr16cplus -mcr16c @gol
609 -msim -mint32 -mbit-ops
610 -mdata-model=@var{model}}
612 @emph{Darwin Options}
613 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
614 -arch_only -bind_at_load -bundle -bundle_loader @gol
615 -client_name -compatibility_version -current_version @gol
617 -dependency-file -dylib_file -dylinker_install_name @gol
618 -dynamic -dynamiclib -exported_symbols_list @gol
619 -filelist -flat_namespace -force_cpusubtype_ALL @gol
620 -force_flat_namespace -headerpad_max_install_names @gol
622 -image_base -init -install_name -keep_private_externs @gol
623 -multi_module -multiply_defined -multiply_defined_unused @gol
624 -noall_load -no_dead_strip_inits_and_terms @gol
625 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
626 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
627 -private_bundle -read_only_relocs -sectalign @gol
628 -sectobjectsymbols -whyload -seg1addr @gol
629 -sectcreate -sectobjectsymbols -sectorder @gol
630 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
631 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
632 -segprot -segs_read_only_addr -segs_read_write_addr @gol
633 -single_module -static -sub_library -sub_umbrella @gol
634 -twolevel_namespace -umbrella -undefined @gol
635 -unexported_symbols_list -weak_reference_mismatches @gol
636 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
637 -mkernel -mone-byte-bool}
639 @emph{DEC Alpha Options}
640 @gccoptlist{-mno-fp-regs -msoft-float @gol
641 -mieee -mieee-with-inexact -mieee-conformant @gol
642 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
643 -mtrap-precision=@var{mode} -mbuild-constants @gol
644 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
645 -mbwx -mmax -mfix -mcix @gol
646 -mfloat-vax -mfloat-ieee @gol
647 -mexplicit-relocs -msmall-data -mlarge-data @gol
648 -msmall-text -mlarge-text @gol
649 -mmemory-latency=@var{time}}
652 @gccoptlist{-msmall-model -mno-lsim}
655 @gccoptlist{-msim -mlra}
658 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
659 -mhard-float -msoft-float @gol
660 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
661 -mdouble -mno-double @gol
662 -mmedia -mno-media -mmuladd -mno-muladd @gol
663 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
664 -mlinked-fp -mlong-calls -malign-labels @gol
665 -mlibrary-pic -macc-4 -macc-8 @gol
666 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
667 -moptimize-membar -mno-optimize-membar @gol
668 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
669 -mvliw-branch -mno-vliw-branch @gol
670 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
671 -mno-nested-cond-exec -mtomcat-stats @gol
675 @emph{GNU/Linux Options}
676 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
677 -tno-android-cc -tno-android-ld}
679 @emph{H8/300 Options}
680 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
683 @gccoptlist{-march=@var{architecture-type} @gol
684 -mdisable-fpregs -mdisable-indexing @gol
685 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
686 -mfixed-range=@var{register-range} @gol
687 -mjump-in-delay -mlinker-opt -mlong-calls @gol
688 -mlong-load-store -mno-disable-fpregs @gol
689 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
690 -mno-jump-in-delay -mno-long-load-store @gol
691 -mno-portable-runtime -mno-soft-float @gol
692 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
693 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
694 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
695 -munix=@var{unix-std} -nolibdld -static -threads}
698 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
699 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
700 -mconstant-gp -mauto-pic -mfused-madd @gol
701 -minline-float-divide-min-latency @gol
702 -minline-float-divide-max-throughput @gol
703 -mno-inline-float-divide @gol
704 -minline-int-divide-min-latency @gol
705 -minline-int-divide-max-throughput @gol
706 -mno-inline-int-divide @gol
707 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
708 -mno-inline-sqrt @gol
709 -mdwarf2-asm -mearly-stop-bits @gol
710 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
711 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
712 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
713 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
714 -msched-spec-ldc -msched-spec-control-ldc @gol
715 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
716 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
717 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
718 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
721 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
722 -msign-extend-enabled -muser-enabled}
724 @emph{M32R/D Options}
725 @gccoptlist{-m32r2 -m32rx -m32r @gol
727 -malign-loops -mno-align-loops @gol
728 -missue-rate=@var{number} @gol
729 -mbranch-cost=@var{number} @gol
730 -mmodel=@var{code-size-model-type} @gol
731 -msdata=@var{sdata-type} @gol
732 -mno-flush-func -mflush-func=@var{name} @gol
733 -mno-flush-trap -mflush-trap=@var{number} @gol
737 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
739 @emph{M680x0 Options}
740 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
741 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
742 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
743 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
744 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
745 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
746 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
747 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
751 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
752 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
753 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
754 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
755 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
758 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
759 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
760 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
761 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
764 @emph{MicroBlaze Options}
765 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
766 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
767 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
768 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
769 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
772 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
773 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
774 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
775 -mips16 -mno-mips16 -mflip-mips16 @gol
776 -minterlink-compressed -mno-interlink-compressed @gol
777 -minterlink-mips16 -mno-interlink-mips16 @gol
778 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
779 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
780 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
781 -mno-float -msingle-float -mdouble-float @gol
782 -modd-spreg -mno-odd-spreg @gol
783 -mabs=@var{mode} -mnan=@var{encoding} @gol
784 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
787 -mvirt -mno-virt @gol
789 -mmicromips -mno-micromips @gol
790 -mfpu=@var{fpu-type} @gol
791 -msmartmips -mno-smartmips @gol
792 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
793 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
794 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
795 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
796 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
797 -membedded-data -mno-embedded-data @gol
798 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
799 -mcode-readable=@var{setting} @gol
800 -msplit-addresses -mno-split-addresses @gol
801 -mexplicit-relocs -mno-explicit-relocs @gol
802 -mcheck-zero-division -mno-check-zero-division @gol
803 -mdivide-traps -mdivide-breaks @gol
804 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
805 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
806 -mfix-24k -mno-fix-24k @gol
807 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
808 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
809 -mfix-vr4120 -mno-fix-vr4120 @gol
810 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
811 -mflush-func=@var{func} -mno-flush-func @gol
812 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
813 -mfp-exceptions -mno-fp-exceptions @gol
814 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
815 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
818 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
819 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
820 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
821 -mno-base-addresses -msingle-exit -mno-single-exit}
823 @emph{MN10300 Options}
824 @gccoptlist{-mmult-bug -mno-mult-bug @gol
825 -mno-am33 -mam33 -mam33-2 -mam34 @gol
826 -mtune=@var{cpu-type} @gol
827 -mreturn-pointer-on-d0 @gol
828 -mno-crt0 -mrelax -mliw -msetlb}
831 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
833 @emph{MSP430 Options}
834 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
835 -mcode-region= -mdata-region= @gol
839 @gccoptlist{-mbig-endian -mlittle-endian @gol
840 -mreduced-regs -mfull-regs @gol
841 -mcmov -mno-cmov @gol
842 -mperf-ext -mno-perf-ext @gol
843 -mv3push -mno-v3push @gol
844 -m16bit -mno-16bit @gol
845 -misr-vector-size=@var{num} @gol
846 -mcache-block-size=@var{num} @gol
847 -march=@var{arch} @gol
848 -mcmodel=@var{code-model} @gol
851 @emph{Nios II Options}
852 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
854 -mno-bypass-cache -mbypass-cache @gol
855 -mno-cache-volatile -mcache-volatile @gol
856 -mno-fast-sw-div -mfast-sw-div @gol
857 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
858 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
859 -mcustom-fpu-cfg=@var{name} @gol
860 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
862 @emph{Nvidia PTX Options}
863 @gccoptlist{-m32 -m64 -mmainkernel}
865 @emph{PDP-11 Options}
866 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
867 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
868 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
869 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
870 -mbranch-expensive -mbranch-cheap @gol
871 -munix-asm -mdec-asm}
873 @emph{picoChip Options}
874 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
875 -msymbol-as-address -mno-inefficient-warnings}
877 @emph{PowerPC Options}
878 See RS/6000 and PowerPC Options.
881 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
882 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
883 -m64bit-doubles -m32bit-doubles}
885 @emph{RS/6000 and PowerPC Options}
886 @gccoptlist{-mcpu=@var{cpu-type} @gol
887 -mtune=@var{cpu-type} @gol
888 -mcmodel=@var{code-model} @gol
890 -maltivec -mno-altivec @gol
891 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
892 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
893 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
894 -mfprnd -mno-fprnd @gol
895 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
896 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
897 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
898 -malign-power -malign-natural @gol
899 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
900 -msingle-float -mdouble-float -msimple-fpu @gol
901 -mstring -mno-string -mupdate -mno-update @gol
902 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
903 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
904 -mstrict-align -mno-strict-align -mrelocatable @gol
905 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
906 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
907 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
908 -mprioritize-restricted-insns=@var{priority} @gol
909 -msched-costly-dep=@var{dependence_type} @gol
910 -minsert-sched-nops=@var{scheme} @gol
911 -mcall-sysv -mcall-netbsd @gol
912 -maix-struct-return -msvr4-struct-return @gol
913 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
914 -mblock-move-inline-limit=@var{num} @gol
915 -misel -mno-isel @gol
916 -misel=yes -misel=no @gol
918 -mspe=yes -mspe=no @gol
920 -mgen-cell-microcode -mwarn-cell-microcode @gol
921 -mvrsave -mno-vrsave @gol
922 -mmulhw -mno-mulhw @gol
923 -mdlmzb -mno-dlmzb @gol
924 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
925 -mprototype -mno-prototype @gol
926 -msim -mmvme -mads -myellowknife -memb -msdata @gol
927 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
928 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
929 -mno-recip-precision @gol
930 -mveclibabi=@var{type} -mfriz -mno-friz @gol
931 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
932 -msave-toc-indirect -mno-save-toc-indirect @gol
933 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
934 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
935 -mquad-memory -mno-quad-memory @gol
936 -mquad-memory-atomic -mno-quad-memory-atomic @gol
937 -mcompat-align-parm -mno-compat-align-parm @gol
938 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
939 -mupper-regs -mno-upper-regs}
942 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
944 -mbig-endian-data -mlittle-endian-data @gol
947 -mas100-syntax -mno-as100-syntax@gol
949 -mmax-constant-size=@gol
952 -mallow-string-insns -mno-allow-string-insns@gol
953 -mno-warn-multiple-fast-interrupts@gol
954 -msave-acc-in-interrupts}
956 @emph{S/390 and zSeries Options}
957 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
958 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
959 -mlong-double-64 -mlong-double-128 @gol
960 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
961 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
962 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
963 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
964 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
965 -mhotpatch=@var{halfwords},@var{halfwords}}
968 @gccoptlist{-meb -mel @gol
972 -mscore5 -mscore5u -mscore7 -mscore7d}
975 @gccoptlist{-m1 -m2 -m2e @gol
976 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
978 -m4-nofpu -m4-single-only -m4-single -m4 @gol
979 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
980 -m5-64media -m5-64media-nofpu @gol
981 -m5-32media -m5-32media-nofpu @gol
982 -m5-compact -m5-compact-nofpu @gol
983 -mb -ml -mdalign -mrelax @gol
984 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
985 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
986 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
987 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
988 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
989 -maccumulate-outgoing-args -minvalid-symbols @gol
990 -matomic-model=@var{atomic-model} @gol
991 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
992 -mcbranch-force-delay-slot @gol
993 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
994 -mpretend-cmove -mtas}
996 @emph{Solaris 2 Options}
997 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1000 @emph{SPARC Options}
1001 @gccoptlist{-mcpu=@var{cpu-type} @gol
1002 -mtune=@var{cpu-type} @gol
1003 -mcmodel=@var{code-model} @gol
1004 -mmemory-model=@var{mem-model} @gol
1005 -m32 -m64 -mapp-regs -mno-app-regs @gol
1006 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1007 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1008 -mhard-quad-float -msoft-quad-float @gol
1009 -mstack-bias -mno-stack-bias @gol
1010 -munaligned-doubles -mno-unaligned-doubles @gol
1011 -muser-mode -mno-user-mode @gol
1012 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1013 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1014 -mcbcond -mno-cbcond @gol
1015 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1016 -mfix-at697f -mfix-ut699}
1019 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1020 -msafe-dma -munsafe-dma @gol
1022 -msmall-mem -mlarge-mem -mstdmain @gol
1023 -mfixed-range=@var{register-range} @gol
1025 -maddress-space-conversion -mno-address-space-conversion @gol
1026 -mcache-size=@var{cache-size} @gol
1027 -matomic-updates -mno-atomic-updates}
1029 @emph{System V Options}
1030 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1032 @emph{TILE-Gx Options}
1033 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1034 -mcmodel=@var{code-model}}
1036 @emph{TILEPro Options}
1037 @gccoptlist{-mcpu=@var{cpu} -m32}
1040 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1041 -mprolog-function -mno-prolog-function -mspace @gol
1042 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1043 -mapp-regs -mno-app-regs @gol
1044 -mdisable-callt -mno-disable-callt @gol
1045 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1046 -mv850e -mv850 -mv850e3v5 @gol
1057 @gccoptlist{-mg -mgnu -munix}
1059 @emph{Visium Options}
1060 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1061 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1064 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1065 -mpointer-size=@var{size}}
1067 @emph{VxWorks Options}
1068 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1069 -Xbind-lazy -Xbind-now}
1072 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1073 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1074 -mfpmath=@var{unit} @gol
1075 -masm=@var{dialect} -mno-fancy-math-387 @gol
1076 -mno-fp-ret-in-387 -msoft-float @gol
1077 -mno-wide-multiply -mrtd -malign-double @gol
1078 -mpreferred-stack-boundary=@var{num} @gol
1079 -mincoming-stack-boundary=@var{num} @gol
1080 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1081 -mrecip -mrecip=@var{opt} @gol
1082 -mvzeroupper -mprefer-avx128 @gol
1083 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1084 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
1085 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
1086 -mclflushopt -mxsavec -mxsaves @gol
1087 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1088 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mthreads @gol
1089 -mno-align-stringops -minline-all-stringops @gol
1090 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1091 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1092 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1093 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1094 -mregparm=@var{num} -msseregparm @gol
1095 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1096 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1097 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1098 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1099 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
1100 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1101 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1102 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
1104 @emph{x86 Windows Options}
1105 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1106 -mnop-fun-dllimport -mthread @gol
1107 -municode -mwin32 -mwindows -fno-set-stack-executable}
1109 @emph{Xstormy16 Options}
1112 @emph{Xtensa Options}
1113 @gccoptlist{-mconst16 -mno-const16 @gol
1114 -mfused-madd -mno-fused-madd @gol
1116 -mserialize-volatile -mno-serialize-volatile @gol
1117 -mtext-section-literals -mno-text-section-literals @gol
1118 -mtarget-align -mno-target-align @gol
1119 -mlongcalls -mno-longcalls}
1121 @emph{zSeries Options}
1122 See S/390 and zSeries Options.
1124 @item Code Generation Options
1125 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1126 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1127 -ffixed-@var{reg} -fexceptions @gol
1128 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1129 -fasynchronous-unwind-tables @gol
1130 -fno-gnu-unique @gol
1131 -finhibit-size-directive -finstrument-functions @gol
1132 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1133 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1134 -fno-common -fno-ident @gol
1135 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
1136 -fno-jump-tables @gol
1137 -frecord-gcc-switches @gol
1138 -freg-struct-return -fshort-enums @gol
1139 -fshort-double -fshort-wchar @gol
1140 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1141 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1142 -fno-stack-limit -fsplit-stack @gol
1143 -fleading-underscore -ftls-model=@var{model} @gol
1144 -fstack-reuse=@var{reuse_level} @gol
1145 -ftrapv -fwrapv -fbounds-check @gol
1146 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1147 -fstrict-volatile-bitfields -fsync-libcalls}
1151 @node Overall Options
1152 @section Options Controlling the Kind of Output
1154 Compilation can involve up to four stages: preprocessing, compilation
1155 proper, assembly and linking, always in that order. GCC is capable of
1156 preprocessing and compiling several files either into several
1157 assembler input files, or into one assembler input file; then each
1158 assembler input file produces an object file, and linking combines all
1159 the object files (those newly compiled, and those specified as input)
1160 into an executable file.
1162 @cindex file name suffix
1163 For any given input file, the file name suffix determines what kind of
1164 compilation is done:
1168 C source code that must be preprocessed.
1171 C source code that should not be preprocessed.
1174 C++ source code that should not be preprocessed.
1177 Objective-C source code. Note that you must link with the @file{libobjc}
1178 library to make an Objective-C program work.
1181 Objective-C source code that should not be preprocessed.
1185 Objective-C++ source code. Note that you must link with the @file{libobjc}
1186 library to make an Objective-C++ program work. Note that @samp{.M} refers
1187 to a literal capital M@.
1189 @item @var{file}.mii
1190 Objective-C++ source code that should not be preprocessed.
1193 C, C++, Objective-C or Objective-C++ header file to be turned into a
1194 precompiled header (default), or C, C++ header file to be turned into an
1195 Ada spec (via the @option{-fdump-ada-spec} switch).
1198 @itemx @var{file}.cp
1199 @itemx @var{file}.cxx
1200 @itemx @var{file}.cpp
1201 @itemx @var{file}.CPP
1202 @itemx @var{file}.c++
1204 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1205 the last two letters must both be literally @samp{x}. Likewise,
1206 @samp{.C} refers to a literal capital C@.
1210 Objective-C++ source code that must be preprocessed.
1212 @item @var{file}.mii
1213 Objective-C++ source code that should not be preprocessed.
1217 @itemx @var{file}.hp
1218 @itemx @var{file}.hxx
1219 @itemx @var{file}.hpp
1220 @itemx @var{file}.HPP
1221 @itemx @var{file}.h++
1222 @itemx @var{file}.tcc
1223 C++ header file to be turned into a precompiled header or Ada spec.
1226 @itemx @var{file}.for
1227 @itemx @var{file}.ftn
1228 Fixed form Fortran source code that should not be preprocessed.
1231 @itemx @var{file}.FOR
1232 @itemx @var{file}.fpp
1233 @itemx @var{file}.FPP
1234 @itemx @var{file}.FTN
1235 Fixed form Fortran source code that must be preprocessed (with the traditional
1238 @item @var{file}.f90
1239 @itemx @var{file}.f95
1240 @itemx @var{file}.f03
1241 @itemx @var{file}.f08
1242 Free form Fortran source code that should not be preprocessed.
1244 @item @var{file}.F90
1245 @itemx @var{file}.F95
1246 @itemx @var{file}.F03
1247 @itemx @var{file}.F08
1248 Free form Fortran source code that must be preprocessed (with the
1249 traditional preprocessor).
1254 @c FIXME: Descriptions of Java file types.
1260 @item @var{file}.ads
1261 Ada source code file that contains a library unit declaration (a
1262 declaration of a package, subprogram, or generic, or a generic
1263 instantiation), or a library unit renaming declaration (a package,
1264 generic, or subprogram renaming declaration). Such files are also
1267 @item @var{file}.adb
1268 Ada source code file containing a library unit body (a subprogram or
1269 package body). Such files are also called @dfn{bodies}.
1271 @c GCC also knows about some suffixes for languages not yet included:
1282 @itemx @var{file}.sx
1283 Assembler code that must be preprocessed.
1286 An object file to be fed straight into linking.
1287 Any file name with no recognized suffix is treated this way.
1291 You can specify the input language explicitly with the @option{-x} option:
1294 @item -x @var{language}
1295 Specify explicitly the @var{language} for the following input files
1296 (rather than letting the compiler choose a default based on the file
1297 name suffix). This option applies to all following input files until
1298 the next @option{-x} option. Possible values for @var{language} are:
1300 c c-header cpp-output
1301 c++ c++-header c++-cpp-output
1302 objective-c objective-c-header objective-c-cpp-output
1303 objective-c++ objective-c++-header objective-c++-cpp-output
1304 assembler assembler-with-cpp
1306 f77 f77-cpp-input f95 f95-cpp-input
1312 Turn off any specification of a language, so that subsequent files are
1313 handled according to their file name suffixes (as they are if @option{-x}
1314 has not been used at all).
1316 @item -pass-exit-codes
1317 @opindex pass-exit-codes
1318 Normally the @command{gcc} program exits with the code of 1 if any
1319 phase of the compiler returns a non-success return code. If you specify
1320 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1321 the numerically highest error produced by any phase returning an error
1322 indication. The C, C++, and Fortran front ends return 4 if an internal
1323 compiler error is encountered.
1326 If you only want some of the stages of compilation, you can use
1327 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1328 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1329 @command{gcc} is to stop. Note that some combinations (for example,
1330 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1335 Compile or assemble the source files, but do not link. The linking
1336 stage simply is not done. The ultimate output is in the form of an
1337 object file for each source file.
1339 By default, the object file name for a source file is made by replacing
1340 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1342 Unrecognized input files, not requiring compilation or assembly, are
1347 Stop after the stage of compilation proper; do not assemble. The output
1348 is in the form of an assembler code file for each non-assembler input
1351 By default, the assembler file name for a source file is made by
1352 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1354 Input files that don't require compilation are ignored.
1358 Stop after the preprocessing stage; do not run the compiler proper. The
1359 output is in the form of preprocessed source code, which is sent to the
1362 Input files that don't require preprocessing are ignored.
1364 @cindex output file option
1367 Place output in file @var{file}. This applies to whatever
1368 sort of output is being produced, whether it be an executable file,
1369 an object file, an assembler file or preprocessed C code.
1371 If @option{-o} is not specified, the default is to put an executable
1372 file in @file{a.out}, the object file for
1373 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1374 assembler file in @file{@var{source}.s}, a precompiled header file in
1375 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1380 Print (on standard error output) the commands executed to run the stages
1381 of compilation. Also print the version number of the compiler driver
1382 program and of the preprocessor and the compiler proper.
1386 Like @option{-v} except the commands are not executed and arguments
1387 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1388 This is useful for shell scripts to capture the driver-generated command lines.
1392 Use pipes rather than temporary files for communication between the
1393 various stages of compilation. This fails to work on some systems where
1394 the assembler is unable to read from a pipe; but the GNU assembler has
1399 Print (on the standard output) a description of the command-line options
1400 understood by @command{gcc}. If the @option{-v} option is also specified
1401 then @option{--help} is also passed on to the various processes
1402 invoked by @command{gcc}, so that they can display the command-line options
1403 they accept. If the @option{-Wextra} option has also been specified
1404 (prior to the @option{--help} option), then command-line options that
1405 have no documentation associated with them are also displayed.
1408 @opindex target-help
1409 Print (on the standard output) a description of target-specific command-line
1410 options for each tool. For some targets extra target-specific
1411 information may also be printed.
1413 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1414 Print (on the standard output) a description of the command-line
1415 options understood by the compiler that fit into all specified classes
1416 and qualifiers. These are the supported classes:
1419 @item @samp{optimizers}
1420 Display all of the optimization options supported by the
1423 @item @samp{warnings}
1424 Display all of the options controlling warning messages
1425 produced by the compiler.
1428 Display target-specific options. Unlike the
1429 @option{--target-help} option however, target-specific options of the
1430 linker and assembler are not displayed. This is because those
1431 tools do not currently support the extended @option{--help=} syntax.
1434 Display the values recognized by the @option{--param}
1437 @item @var{language}
1438 Display the options supported for @var{language}, where
1439 @var{language} is the name of one of the languages supported in this
1443 Display the options that are common to all languages.
1446 These are the supported qualifiers:
1449 @item @samp{undocumented}
1450 Display only those options that are undocumented.
1453 Display options taking an argument that appears after an equal
1454 sign in the same continuous piece of text, such as:
1455 @samp{--help=target}.
1457 @item @samp{separate}
1458 Display options taking an argument that appears as a separate word
1459 following the original option, such as: @samp{-o output-file}.
1462 Thus for example to display all the undocumented target-specific
1463 switches supported by the compiler, use:
1466 --help=target,undocumented
1469 The sense of a qualifier can be inverted by prefixing it with the
1470 @samp{^} character, so for example to display all binary warning
1471 options (i.e., ones that are either on or off and that do not take an
1472 argument) that have a description, use:
1475 --help=warnings,^joined,^undocumented
1478 The argument to @option{--help=} should not consist solely of inverted
1481 Combining several classes is possible, although this usually
1482 restricts the output so much that there is nothing to display. One
1483 case where it does work, however, is when one of the classes is
1484 @var{target}. For example, to display all the target-specific
1485 optimization options, use:
1488 --help=target,optimizers
1491 The @option{--help=} option can be repeated on the command line. Each
1492 successive use displays its requested class of options, skipping
1493 those that have already been displayed.
1495 If the @option{-Q} option appears on the command line before the
1496 @option{--help=} option, then the descriptive text displayed by
1497 @option{--help=} is changed. Instead of describing the displayed
1498 options, an indication is given as to whether the option is enabled,
1499 disabled or set to a specific value (assuming that the compiler
1500 knows this at the point where the @option{--help=} option is used).
1502 Here is a truncated example from the ARM port of @command{gcc}:
1505 % gcc -Q -mabi=2 --help=target -c
1506 The following options are target specific:
1508 -mabort-on-noreturn [disabled]
1512 The output is sensitive to the effects of previous command-line
1513 options, so for example it is possible to find out which optimizations
1514 are enabled at @option{-O2} by using:
1517 -Q -O2 --help=optimizers
1520 Alternatively you can discover which binary optimizations are enabled
1521 by @option{-O3} by using:
1524 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1525 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1526 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1529 @item -no-canonical-prefixes
1530 @opindex no-canonical-prefixes
1531 Do not expand any symbolic links, resolve references to @samp{/../}
1532 or @samp{/./}, or make the path absolute when generating a relative
1537 Display the version number and copyrights of the invoked GCC@.
1541 Invoke all subcommands under a wrapper program. The name of the
1542 wrapper program and its parameters are passed as a comma separated
1546 gcc -c t.c -wrapper gdb,--args
1550 This invokes all subprograms of @command{gcc} under
1551 @samp{gdb --args}, thus the invocation of @command{cc1} is
1552 @samp{gdb --args cc1 @dots{}}.
1554 @item -fplugin=@var{name}.so
1556 Load the plugin code in file @var{name}.so, assumed to be a
1557 shared object to be dlopen'd by the compiler. The base name of
1558 the shared object file is used to identify the plugin for the
1559 purposes of argument parsing (See
1560 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1561 Each plugin should define the callback functions specified in the
1564 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1565 @opindex fplugin-arg
1566 Define an argument called @var{key} with a value of @var{value}
1567 for the plugin called @var{name}.
1569 @item -fdump-ada-spec@r{[}-slim@r{]}
1570 @opindex fdump-ada-spec
1571 For C and C++ source and include files, generate corresponding Ada specs.
1572 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1573 GNAT User's Guide}, which provides detailed documentation on this feature.
1575 @item -fada-spec-parent=@var{unit}
1576 @opindex fada-spec-parent
1577 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1578 Ada specs as child units of parent @var{unit}.
1580 @item -fdump-go-spec=@var{file}
1581 @opindex fdump-go-spec
1582 For input files in any language, generate corresponding Go
1583 declarations in @var{file}. This generates Go @code{const},
1584 @code{type}, @code{var}, and @code{func} declarations which may be a
1585 useful way to start writing a Go interface to code written in some
1588 @include @value{srcdir}/../libiberty/at-file.texi
1592 @section Compiling C++ Programs
1594 @cindex suffixes for C++ source
1595 @cindex C++ source file suffixes
1596 C++ source files conventionally use one of the suffixes @samp{.C},
1597 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1598 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1599 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1600 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1601 files with these names and compiles them as C++ programs even if you
1602 call the compiler the same way as for compiling C programs (usually
1603 with the name @command{gcc}).
1607 However, the use of @command{gcc} does not add the C++ library.
1608 @command{g++} is a program that calls GCC and automatically specifies linking
1609 against the C++ library. It treats @samp{.c},
1610 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1611 files unless @option{-x} is used. This program is also useful when
1612 precompiling a C header file with a @samp{.h} extension for use in C++
1613 compilations. On many systems, @command{g++} is also installed with
1614 the name @command{c++}.
1616 @cindex invoking @command{g++}
1617 When you compile C++ programs, you may specify many of the same
1618 command-line options that you use for compiling programs in any
1619 language; or command-line options meaningful for C and related
1620 languages; or options that are meaningful only for C++ programs.
1621 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1622 explanations of options for languages related to C@.
1623 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1624 explanations of options that are meaningful only for C++ programs.
1626 @node C Dialect Options
1627 @section Options Controlling C Dialect
1628 @cindex dialect options
1629 @cindex language dialect options
1630 @cindex options, dialect
1632 The following options control the dialect of C (or languages derived
1633 from C, such as C++, Objective-C and Objective-C++) that the compiler
1637 @cindex ANSI support
1641 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1642 equivalent to @option{-std=c++98}.
1644 This turns off certain features of GCC that are incompatible with ISO
1645 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1646 such as the @code{asm} and @code{typeof} keywords, and
1647 predefined macros such as @code{unix} and @code{vax} that identify the
1648 type of system you are using. It also enables the undesirable and
1649 rarely used ISO trigraph feature. For the C compiler,
1650 it disables recognition of C++ style @samp{//} comments as well as
1651 the @code{inline} keyword.
1653 The alternate keywords @code{__asm__}, @code{__extension__},
1654 @code{__inline__} and @code{__typeof__} continue to work despite
1655 @option{-ansi}. You would not want to use them in an ISO C program, of
1656 course, but it is useful to put them in header files that might be included
1657 in compilations done with @option{-ansi}. Alternate predefined macros
1658 such as @code{__unix__} and @code{__vax__} are also available, with or
1659 without @option{-ansi}.
1661 The @option{-ansi} option does not cause non-ISO programs to be
1662 rejected gratuitously. For that, @option{-Wpedantic} is required in
1663 addition to @option{-ansi}. @xref{Warning Options}.
1665 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1666 option is used. Some header files may notice this macro and refrain
1667 from declaring certain functions or defining certain macros that the
1668 ISO standard doesn't call for; this is to avoid interfering with any
1669 programs that might use these names for other things.
1671 Functions that are normally built in but do not have semantics
1672 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1673 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1674 built-in functions provided by GCC}, for details of the functions
1679 Determine the language standard. @xref{Standards,,Language Standards
1680 Supported by GCC}, for details of these standard versions. This option
1681 is currently only supported when compiling C or C++.
1683 The compiler can accept several base standards, such as @samp{c90} or
1684 @samp{c++98}, and GNU dialects of those standards, such as
1685 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1686 compiler accepts all programs following that standard plus those
1687 using GNU extensions that do not contradict it. For example,
1688 @option{-std=c90} turns off certain features of GCC that are
1689 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1690 keywords, but not other GNU extensions that do not have a meaning in
1691 ISO C90, such as omitting the middle term of a @code{?:}
1692 expression. On the other hand, when a GNU dialect of a standard is
1693 specified, all features supported by the compiler are enabled, even when
1694 those features change the meaning of the base standard. As a result, some
1695 strict-conforming programs may be rejected. The particular standard
1696 is used by @option{-Wpedantic} to identify which features are GNU
1697 extensions given that version of the standard. For example
1698 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1699 comments, while @option{-std=gnu99 -Wpedantic} does not.
1701 A value for this option must be provided; possible values are
1707 Support all ISO C90 programs (certain GNU extensions that conflict
1708 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1710 @item iso9899:199409
1711 ISO C90 as modified in amendment 1.
1717 ISO C99. This standard is substantially completely supported, modulo
1718 bugs and floating-point issues
1719 (mainly but not entirely relating to optional C99 features from
1720 Annexes F and G). See
1721 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1722 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1727 ISO C11, the 2011 revision of the ISO C standard. This standard is
1728 substantially completely supported, modulo bugs, floating-point issues
1729 (mainly but not entirely relating to optional C11 features from
1730 Annexes F and G) and the optional Annexes K (Bounds-checking
1731 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1735 GNU dialect of ISO C90 (including some C99 features).
1739 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1743 GNU dialect of ISO C11. This is the default for C code.
1744 The name @samp{gnu1x} is deprecated.
1748 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1749 additional defect reports. Same as @option{-ansi} for C++ code.
1753 GNU dialect of @option{-std=c++98}. This is the default for
1758 The 2011 ISO C++ standard plus amendments.
1759 The name @samp{c++0x} is deprecated.
1763 GNU dialect of @option{-std=c++11}.
1764 The name @samp{gnu++0x} is deprecated.
1768 The 2014 ISO C++ standard plus amendments.
1769 The name @samp{c++1y} is deprecated.
1773 GNU dialect of @option{-std=c++14}.
1774 The name @samp{gnu++1y} is deprecated.
1777 The next revision of the ISO C++ standard, tentatively planned for
1778 2017. Support is highly experimental, and will almost certainly
1779 change in incompatible ways in future releases.
1782 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1783 and will almost certainly change in incompatible ways in future
1787 @item -fgnu89-inline
1788 @opindex fgnu89-inline
1789 The option @option{-fgnu89-inline} tells GCC to use the traditional
1790 GNU semantics for @code{inline} functions when in C99 mode.
1791 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1792 Using this option is roughly equivalent to adding the
1793 @code{gnu_inline} function attribute to all inline functions
1794 (@pxref{Function Attributes}).
1796 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1797 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1798 specifies the default behavior).
1799 This option is not supported in @option{-std=c90} or
1800 @option{-std=gnu90} mode.
1802 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1803 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1804 in effect for @code{inline} functions. @xref{Common Predefined
1805 Macros,,,cpp,The C Preprocessor}.
1807 @item -aux-info @var{filename}
1809 Output to the given filename prototyped declarations for all functions
1810 declared and/or defined in a translation unit, including those in header
1811 files. This option is silently ignored in any language other than C@.
1813 Besides declarations, the file indicates, in comments, the origin of
1814 each declaration (source file and line), whether the declaration was
1815 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1816 @samp{O} for old, respectively, in the first character after the line
1817 number and the colon), and whether it came from a declaration or a
1818 definition (@samp{C} or @samp{F}, respectively, in the following
1819 character). In the case of function definitions, a K&R-style list of
1820 arguments followed by their declarations is also provided, inside
1821 comments, after the declaration.
1823 @item -fallow-parameterless-variadic-functions
1824 @opindex fallow-parameterless-variadic-functions
1825 Accept variadic functions without named parameters.
1827 Although it is possible to define such a function, this is not very
1828 useful as it is not possible to read the arguments. This is only
1829 supported for C as this construct is allowed by C++.
1833 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1834 keyword, so that code can use these words as identifiers. You can use
1835 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1836 instead. @option{-ansi} implies @option{-fno-asm}.
1838 In C++, this switch only affects the @code{typeof} keyword, since
1839 @code{asm} and @code{inline} are standard keywords. You may want to
1840 use the @option{-fno-gnu-keywords} flag instead, which has the same
1841 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1842 switch only affects the @code{asm} and @code{typeof} keywords, since
1843 @code{inline} is a standard keyword in ISO C99.
1846 @itemx -fno-builtin-@var{function}
1847 @opindex fno-builtin
1848 @cindex built-in functions
1849 Don't recognize built-in functions that do not begin with
1850 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1851 functions provided by GCC}, for details of the functions affected,
1852 including those which are not built-in functions when @option{-ansi} or
1853 @option{-std} options for strict ISO C conformance are used because they
1854 do not have an ISO standard meaning.
1856 GCC normally generates special code to handle certain built-in functions
1857 more efficiently; for instance, calls to @code{alloca} may become single
1858 instructions which adjust the stack directly, and calls to @code{memcpy}
1859 may become inline copy loops. The resulting code is often both smaller
1860 and faster, but since the function calls no longer appear as such, you
1861 cannot set a breakpoint on those calls, nor can you change the behavior
1862 of the functions by linking with a different library. In addition,
1863 when a function is recognized as a built-in function, GCC may use
1864 information about that function to warn about problems with calls to
1865 that function, or to generate more efficient code, even if the
1866 resulting code still contains calls to that function. For example,
1867 warnings are given with @option{-Wformat} for bad calls to
1868 @code{printf} when @code{printf} is built in and @code{strlen} is
1869 known not to modify global memory.
1871 With the @option{-fno-builtin-@var{function}} option
1872 only the built-in function @var{function} is
1873 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1874 function is named that is not built-in in this version of GCC, this
1875 option is ignored. There is no corresponding
1876 @option{-fbuiltin-@var{function}} option; if you wish to enable
1877 built-in functions selectively when using @option{-fno-builtin} or
1878 @option{-ffreestanding}, you may define macros such as:
1881 #define abs(n) __builtin_abs ((n))
1882 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1887 @cindex hosted environment
1889 Assert that compilation targets a hosted environment. This implies
1890 @option{-fbuiltin}. A hosted environment is one in which the
1891 entire standard library is available, and in which @code{main} has a return
1892 type of @code{int}. Examples are nearly everything except a kernel.
1893 This is equivalent to @option{-fno-freestanding}.
1895 @item -ffreestanding
1896 @opindex ffreestanding
1897 @cindex hosted environment
1899 Assert that compilation targets a freestanding environment. This
1900 implies @option{-fno-builtin}. A freestanding environment
1901 is one in which the standard library may not exist, and program startup may
1902 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1903 This is equivalent to @option{-fno-hosted}.
1905 @xref{Standards,,Language Standards Supported by GCC}, for details of
1906 freestanding and hosted environments.
1910 @cindex OpenACC accelerator programming
1911 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1912 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1913 compiler generates accelerated code according to the OpenACC Application
1914 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1915 implies @option{-pthread}, and thus is only supported on targets that
1916 have support for @option{-pthread}.
1918 Note that this is an experimental feature, incomplete, and subject to
1919 change in future versions of GCC. See
1920 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1924 @cindex OpenMP parallel
1925 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1926 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1927 compiler generates parallel code according to the OpenMP Application
1928 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1929 implies @option{-pthread}, and thus is only supported on targets that
1930 have support for @option{-pthread}. @option{-fopenmp} implies
1931 @option{-fopenmp-simd}.
1934 @opindex fopenmp-simd
1937 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1938 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1943 @cindex Enable Cilk Plus
1944 Enable the usage of Cilk Plus language extension features for C/C++.
1945 When the option @option{-fcilkplus} is specified, enable the usage of
1946 the Cilk Plus Language extension features for C/C++. The present
1947 implementation follows ABI version 1.2. This is an experimental
1948 feature that is only partially complete, and whose interface may
1949 change in future versions of GCC as the official specification
1950 changes. Currently, all features but @code{_Cilk_for} have been
1955 When the option @option{-fgnu-tm} is specified, the compiler
1956 generates code for the Linux variant of Intel's current Transactional
1957 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1958 an experimental feature whose interface may change in future versions
1959 of GCC, as the official specification changes. Please note that not
1960 all architectures are supported for this feature.
1962 For more information on GCC's support for transactional memory,
1963 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1964 Transactional Memory Library}.
1966 Note that the transactional memory feature is not supported with
1967 non-call exceptions (@option{-fnon-call-exceptions}).
1969 @item -fms-extensions
1970 @opindex fms-extensions
1971 Accept some non-standard constructs used in Microsoft header files.
1973 In C++ code, this allows member names in structures to be similar
1974 to previous types declarations.
1983 Some cases of unnamed fields in structures and unions are only
1984 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1985 fields within structs/unions}, for details.
1987 Note that this option is off for all targets but x86
1988 targets using ms-abi.
1990 @item -fplan9-extensions
1991 @opindex fplan9-extensions
1992 Accept some non-standard constructs used in Plan 9 code.
1994 This enables @option{-fms-extensions}, permits passing pointers to
1995 structures with anonymous fields to functions that expect pointers to
1996 elements of the type of the field, and permits referring to anonymous
1997 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1998 struct/union fields within structs/unions}, for details. This is only
1999 supported for C, not C++.
2003 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2004 options for strict ISO C conformance) implies @option{-trigraphs}.
2006 @cindex traditional C language
2007 @cindex C language, traditional
2009 @itemx -traditional-cpp
2010 @opindex traditional-cpp
2011 @opindex traditional
2012 Formerly, these options caused GCC to attempt to emulate a pre-standard
2013 C compiler. They are now only supported with the @option{-E} switch.
2014 The preprocessor continues to support a pre-standard mode. See the GNU
2015 CPP manual for details.
2017 @item -fcond-mismatch
2018 @opindex fcond-mismatch
2019 Allow conditional expressions with mismatched types in the second and
2020 third arguments. The value of such an expression is void. This option
2021 is not supported for C++.
2023 @item -flax-vector-conversions
2024 @opindex flax-vector-conversions
2025 Allow implicit conversions between vectors with differing numbers of
2026 elements and/or incompatible element types. This option should not be
2029 @item -funsigned-char
2030 @opindex funsigned-char
2031 Let the type @code{char} be unsigned, like @code{unsigned char}.
2033 Each kind of machine has a default for what @code{char} should
2034 be. It is either like @code{unsigned char} by default or like
2035 @code{signed char} by default.
2037 Ideally, a portable program should always use @code{signed char} or
2038 @code{unsigned char} when it depends on the signedness of an object.
2039 But many programs have been written to use plain @code{char} and
2040 expect it to be signed, or expect it to be unsigned, depending on the
2041 machines they were written for. This option, and its inverse, let you
2042 make such a program work with the opposite default.
2044 The type @code{char} is always a distinct type from each of
2045 @code{signed char} or @code{unsigned char}, even though its behavior
2046 is always just like one of those two.
2049 @opindex fsigned-char
2050 Let the type @code{char} be signed, like @code{signed char}.
2052 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2053 the negative form of @option{-funsigned-char}. Likewise, the option
2054 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2056 @item -fsigned-bitfields
2057 @itemx -funsigned-bitfields
2058 @itemx -fno-signed-bitfields
2059 @itemx -fno-unsigned-bitfields
2060 @opindex fsigned-bitfields
2061 @opindex funsigned-bitfields
2062 @opindex fno-signed-bitfields
2063 @opindex fno-unsigned-bitfields
2064 These options control whether a bit-field is signed or unsigned, when the
2065 declaration does not use either @code{signed} or @code{unsigned}. By
2066 default, such a bit-field is signed, because this is consistent: the
2067 basic integer types such as @code{int} are signed types.
2070 @node C++ Dialect Options
2071 @section Options Controlling C++ Dialect
2073 @cindex compiler options, C++
2074 @cindex C++ options, command-line
2075 @cindex options, C++
2076 This section describes the command-line options that are only meaningful
2077 for C++ programs. You can also use most of the GNU compiler options
2078 regardless of what language your program is in. For example, you
2079 might compile a file @file{firstClass.C} like this:
2082 g++ -g -frepo -O -c firstClass.C
2086 In this example, only @option{-frepo} is an option meant
2087 only for C++ programs; you can use the other options with any
2088 language supported by GCC@.
2090 Here is a list of options that are @emph{only} for compiling C++ programs:
2094 @item -fabi-version=@var{n}
2095 @opindex fabi-version
2096 Use version @var{n} of the C++ ABI@. The default is version 0.
2098 Version 0 refers to the version conforming most closely to
2099 the C++ ABI specification. Therefore, the ABI obtained using version 0
2100 will change in different versions of G++ as ABI bugs are fixed.
2102 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2104 Version 2 is the version of the C++ ABI that first appeared in G++
2105 3.4, and was the default through G++ 4.9.
2107 Version 3 corrects an error in mangling a constant address as a
2110 Version 4, which first appeared in G++ 4.5, implements a standard
2111 mangling for vector types.
2113 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2114 attribute const/volatile on function pointer types, decltype of a
2115 plain decl, and use of a function parameter in the declaration of
2118 Version 6, which first appeared in G++ 4.7, corrects the promotion
2119 behavior of C++11 scoped enums and the mangling of template argument
2120 packs, const/static_cast, prefix ++ and --, and a class scope function
2121 used as a template argument.
2123 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2124 builtin type and corrects the mangling of lambdas in default argument
2127 Version 8, which first appeared in G++ 4.9, corrects the substitution
2128 behavior of function types with function-cv-qualifiers.
2130 See also @option{-Wabi}.
2132 @item -fabi-compat-version=@var{n}
2133 @opindex fabi-compat-version
2134 On targets that support strong aliases, G++
2135 works around mangling changes by creating an alias with the correct
2136 mangled name when defining a symbol with an incorrect mangled name.
2137 This switch specifies which ABI version to use for the alias.
2139 With @option{-fabi-version=0} (the default), this defaults to 2. If
2140 another ABI version is explicitly selected, this defaults to 0.
2142 The compatibility version is also set by @option{-Wabi=@var{n}}.
2144 @item -fno-access-control
2145 @opindex fno-access-control
2146 Turn off all access checking. This switch is mainly useful for working
2147 around bugs in the access control code.
2151 Check that the pointer returned by @code{operator new} is non-null
2152 before attempting to modify the storage allocated. This check is
2153 normally unnecessary because the C++ standard specifies that
2154 @code{operator new} only returns @code{0} if it is declared
2155 @code{throw()}, in which case the compiler always checks the
2156 return value even without this option. In all other cases, when
2157 @code{operator new} has a non-empty exception specification, memory
2158 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2159 @samp{new (nothrow)}.
2161 @item -fconstexpr-depth=@var{n}
2162 @opindex fconstexpr-depth
2163 Set the maximum nested evaluation depth for C++11 constexpr functions
2164 to @var{n}. A limit is needed to detect endless recursion during
2165 constant expression evaluation. The minimum specified by the standard
2168 @item -fdeduce-init-list
2169 @opindex fdeduce-init-list
2170 Enable deduction of a template type parameter as
2171 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2174 template <class T> auto forward(T t) -> decltype (realfn (t))
2181 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2185 This deduction was implemented as a possible extension to the
2186 originally proposed semantics for the C++11 standard, but was not part
2187 of the final standard, so it is disabled by default. This option is
2188 deprecated, and may be removed in a future version of G++.
2190 @item -ffriend-injection
2191 @opindex ffriend-injection
2192 Inject friend functions into the enclosing namespace, so that they are
2193 visible outside the scope of the class in which they are declared.
2194 Friend functions were documented to work this way in the old Annotated
2195 C++ Reference Manual.
2196 However, in ISO C++ a friend function that is not declared
2197 in an enclosing scope can only be found using argument dependent
2198 lookup. GCC defaults to the standard behavior.
2200 This option is for compatibility, and may be removed in a future
2203 @item -fno-elide-constructors
2204 @opindex fno-elide-constructors
2205 The C++ standard allows an implementation to omit creating a temporary
2206 that is only used to initialize another object of the same type.
2207 Specifying this option disables that optimization, and forces G++ to
2208 call the copy constructor in all cases.
2210 @item -fno-enforce-eh-specs
2211 @opindex fno-enforce-eh-specs
2212 Don't generate code to check for violation of exception specifications
2213 at run time. This option violates the C++ standard, but may be useful
2214 for reducing code size in production builds, much like defining
2215 @code{NDEBUG}. This does not give user code permission to throw
2216 exceptions in violation of the exception specifications; the compiler
2217 still optimizes based on the specifications, so throwing an
2218 unexpected exception results in undefined behavior at run time.
2220 @item -fextern-tls-init
2221 @itemx -fno-extern-tls-init
2222 @opindex fextern-tls-init
2223 @opindex fno-extern-tls-init
2224 The C++11 and OpenMP standards allow @code{thread_local} and
2225 @code{threadprivate} variables to have dynamic (runtime)
2226 initialization. To support this, any use of such a variable goes
2227 through a wrapper function that performs any necessary initialization.
2228 When the use and definition of the variable are in the same
2229 translation unit, this overhead can be optimized away, but when the
2230 use is in a different translation unit there is significant overhead
2231 even if the variable doesn't actually need dynamic initialization. If
2232 the programmer can be sure that no use of the variable in a
2233 non-defining TU needs to trigger dynamic initialization (either
2234 because the variable is statically initialized, or a use of the
2235 variable in the defining TU will be executed before any uses in
2236 another TU), they can avoid this overhead with the
2237 @option{-fno-extern-tls-init} option.
2239 On targets that support symbol aliases, the default is
2240 @option{-fextern-tls-init}. On targets that do not support symbol
2241 aliases, the default is @option{-fno-extern-tls-init}.
2244 @itemx -fno-for-scope
2246 @opindex fno-for-scope
2247 If @option{-ffor-scope} is specified, the scope of variables declared in
2248 a @i{for-init-statement} is limited to the @code{for} loop itself,
2249 as specified by the C++ standard.
2250 If @option{-fno-for-scope} is specified, the scope of variables declared in
2251 a @i{for-init-statement} extends to the end of the enclosing scope,
2252 as was the case in old versions of G++, and other (traditional)
2253 implementations of C++.
2255 If neither flag is given, the default is to follow the standard,
2256 but to allow and give a warning for old-style code that would
2257 otherwise be invalid, or have different behavior.
2259 @item -fno-gnu-keywords
2260 @opindex fno-gnu-keywords
2261 Do not recognize @code{typeof} as a keyword, so that code can use this
2262 word as an identifier. You can use the keyword @code{__typeof__} instead.
2263 @option{-ansi} implies @option{-fno-gnu-keywords}.
2265 @item -fno-implicit-templates
2266 @opindex fno-implicit-templates
2267 Never emit code for non-inline templates that are instantiated
2268 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2269 @xref{Template Instantiation}, for more information.
2271 @item -fno-implicit-inline-templates
2272 @opindex fno-implicit-inline-templates
2273 Don't emit code for implicit instantiations of inline templates, either.
2274 The default is to handle inlines differently so that compiles with and
2275 without optimization need the same set of explicit instantiations.
2277 @item -fno-implement-inlines
2278 @opindex fno-implement-inlines
2279 To save space, do not emit out-of-line copies of inline functions
2280 controlled by @code{#pragma implementation}. This causes linker
2281 errors if these functions are not inlined everywhere they are called.
2283 @item -fms-extensions
2284 @opindex fms-extensions
2285 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2286 int and getting a pointer to member function via non-standard syntax.
2288 @item -fno-nonansi-builtins
2289 @opindex fno-nonansi-builtins
2290 Disable built-in declarations of functions that are not mandated by
2291 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2292 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2295 @opindex fnothrow-opt
2296 Treat a @code{throw()} exception specification as if it were a
2297 @code{noexcept} specification to reduce or eliminate the text size
2298 overhead relative to a function with no exception specification. If
2299 the function has local variables of types with non-trivial
2300 destructors, the exception specification actually makes the
2301 function smaller because the EH cleanups for those variables can be
2302 optimized away. The semantic effect is that an exception thrown out of
2303 a function with such an exception specification results in a call
2304 to @code{terminate} rather than @code{unexpected}.
2306 @item -fno-operator-names
2307 @opindex fno-operator-names
2308 Do not treat the operator name keywords @code{and}, @code{bitand},
2309 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2310 synonyms as keywords.
2312 @item -fno-optional-diags
2313 @opindex fno-optional-diags
2314 Disable diagnostics that the standard says a compiler does not need to
2315 issue. Currently, the only such diagnostic issued by G++ is the one for
2316 a name having multiple meanings within a class.
2319 @opindex fpermissive
2320 Downgrade some diagnostics about nonconformant code from errors to
2321 warnings. Thus, using @option{-fpermissive} allows some
2322 nonconforming code to compile.
2324 @item -fno-pretty-templates
2325 @opindex fno-pretty-templates
2326 When an error message refers to a specialization of a function
2327 template, the compiler normally prints the signature of the
2328 template followed by the template arguments and any typedefs or
2329 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2330 rather than @code{void f(int)}) so that it's clear which template is
2331 involved. When an error message refers to a specialization of a class
2332 template, the compiler omits any template arguments that match
2333 the default template arguments for that template. If either of these
2334 behaviors make it harder to understand the error message rather than
2335 easier, you can use @option{-fno-pretty-templates} to disable them.
2339 Enable automatic template instantiation at link time. This option also
2340 implies @option{-fno-implicit-templates}. @xref{Template
2341 Instantiation}, for more information.
2345 Disable generation of information about every class with virtual
2346 functions for use by the C++ run-time type identification features
2347 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2348 of the language, you can save some space by using this flag. Note that
2349 exception handling uses the same information, but G++ generates it as
2350 needed. The @code{dynamic_cast} operator can still be used for casts that
2351 do not require run-time type information, i.e.@: casts to @code{void *} or to
2352 unambiguous base classes.
2354 @item -fsized-deallocation
2355 @opindex fsized-deallocation
2356 Enable the built-in global declarations
2358 void operator delete (void *, std::size_t) noexcept;
2359 void operator delete[] (void *, std::size_t) noexcept;
2361 as introduced in C++14. This is useful for user-defined replacement
2362 deallocation functions that, for example, use the size of the object
2363 to make deallocation faster. Enabled by default under
2364 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2365 warns about places that might want to add a definition.
2369 Emit statistics about front-end processing at the end of the compilation.
2370 This information is generally only useful to the G++ development team.
2372 @item -fstrict-enums
2373 @opindex fstrict-enums
2374 Allow the compiler to optimize using the assumption that a value of
2375 enumerated type can only be one of the values of the enumeration (as
2376 defined in the C++ standard; basically, a value that can be
2377 represented in the minimum number of bits needed to represent all the
2378 enumerators). This assumption may not be valid if the program uses a
2379 cast to convert an arbitrary integer value to the enumerated type.
2381 @item -ftemplate-backtrace-limit=@var{n}
2382 @opindex ftemplate-backtrace-limit
2383 Set the maximum number of template instantiation notes for a single
2384 warning or error to @var{n}. The default value is 10.
2386 @item -ftemplate-depth=@var{n}
2387 @opindex ftemplate-depth
2388 Set the maximum instantiation depth for template classes to @var{n}.
2389 A limit on the template instantiation depth is needed to detect
2390 endless recursions during template class instantiation. ANSI/ISO C++
2391 conforming programs must not rely on a maximum depth greater than 17
2392 (changed to 1024 in C++11). The default value is 900, as the compiler
2393 can run out of stack space before hitting 1024 in some situations.
2395 @item -fno-threadsafe-statics
2396 @opindex fno-threadsafe-statics
2397 Do not emit the extra code to use the routines specified in the C++
2398 ABI for thread-safe initialization of local statics. You can use this
2399 option to reduce code size slightly in code that doesn't need to be
2402 @item -fuse-cxa-atexit
2403 @opindex fuse-cxa-atexit
2404 Register destructors for objects with static storage duration with the
2405 @code{__cxa_atexit} function rather than the @code{atexit} function.
2406 This option is required for fully standards-compliant handling of static
2407 destructors, but only works if your C library supports
2408 @code{__cxa_atexit}.
2410 @item -fno-use-cxa-get-exception-ptr
2411 @opindex fno-use-cxa-get-exception-ptr
2412 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2413 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2414 if the runtime routine is not available.
2416 @item -fvisibility-inlines-hidden
2417 @opindex fvisibility-inlines-hidden
2418 This switch declares that the user does not attempt to compare
2419 pointers to inline functions or methods where the addresses of the two functions
2420 are taken in different shared objects.
2422 The effect of this is that GCC may, effectively, mark inline methods with
2423 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2424 appear in the export table of a DSO and do not require a PLT indirection
2425 when used within the DSO@. Enabling this option can have a dramatic effect
2426 on load and link times of a DSO as it massively reduces the size of the
2427 dynamic export table when the library makes heavy use of templates.
2429 The behavior of this switch is not quite the same as marking the
2430 methods as hidden directly, because it does not affect static variables
2431 local to the function or cause the compiler to deduce that
2432 the function is defined in only one shared object.
2434 You may mark a method as having a visibility explicitly to negate the
2435 effect of the switch for that method. For example, if you do want to
2436 compare pointers to a particular inline method, you might mark it as
2437 having default visibility. Marking the enclosing class with explicit
2438 visibility has no effect.
2440 Explicitly instantiated inline methods are unaffected by this option
2441 as their linkage might otherwise cross a shared library boundary.
2442 @xref{Template Instantiation}.
2444 @item -fvisibility-ms-compat
2445 @opindex fvisibility-ms-compat
2446 This flag attempts to use visibility settings to make GCC's C++
2447 linkage model compatible with that of Microsoft Visual Studio.
2449 The flag makes these changes to GCC's linkage model:
2453 It sets the default visibility to @code{hidden}, like
2454 @option{-fvisibility=hidden}.
2457 Types, but not their members, are not hidden by default.
2460 The One Definition Rule is relaxed for types without explicit
2461 visibility specifications that are defined in more than one
2462 shared object: those declarations are permitted if they are
2463 permitted when this option is not used.
2466 In new code it is better to use @option{-fvisibility=hidden} and
2467 export those classes that are intended to be externally visible.
2468 Unfortunately it is possible for code to rely, perhaps accidentally,
2469 on the Visual Studio behavior.
2471 Among the consequences of these changes are that static data members
2472 of the same type with the same name but defined in different shared
2473 objects are different, so changing one does not change the other;
2474 and that pointers to function members defined in different shared
2475 objects may not compare equal. When this flag is given, it is a
2476 violation of the ODR to define types with the same name differently.
2478 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2479 @opindex fvtable-verify
2480 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2481 feature that verifies at run time, for every virtual call, that
2482 the vtable pointer through which the call is made is valid for the type of
2483 the object, and has not been corrupted or overwritten. If an invalid vtable
2484 pointer is detected at run time, an error is reported and execution of the
2485 program is immediately halted.
2487 This option causes run-time data structures to be built at program startup,
2488 which are used for verifying the vtable pointers.
2489 The options @samp{std} and @samp{preinit}
2490 control the timing of when these data structures are built. In both cases the
2491 data structures are built before execution reaches @code{main}. Using
2492 @option{-fvtable-verify=std} causes the data structures to be built after
2493 shared libraries have been loaded and initialized.
2494 @option{-fvtable-verify=preinit} causes them to be built before shared
2495 libraries have been loaded and initialized.
2497 If this option appears multiple times in the command line with different
2498 values specified, @samp{none} takes highest priority over both @samp{std} and
2499 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2503 When used in conjunction with @option{-fvtable-verify=std} or
2504 @option{-fvtable-verify=preinit}, causes debug versions of the
2505 runtime functions for the vtable verification feature to be called.
2506 This flag also causes the compiler to log information about which
2507 vtable pointers it finds for each class.
2508 This information is written to a file named @file{vtv_set_ptr_data.log}
2509 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2510 if that is defined or the current working directory otherwise.
2512 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2513 file, be sure to delete any existing one.
2516 @opindex fvtv-counts
2517 This is a debugging flag. When used in conjunction with
2518 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2519 causes the compiler to keep track of the total number of virtual calls
2520 it encounters and the number of verifications it inserts. It also
2521 counts the number of calls to certain run-time library functions
2522 that it inserts and logs this information for each compilation unit.
2523 The compiler writes this information to a file named
2524 @file{vtv_count_data.log} in the directory named by the environment
2525 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2526 directory otherwise. It also counts the size of the vtable pointer sets
2527 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2528 in the same directory.
2530 Note: This feature @emph{appends} data to the log files. To get fresh log
2531 files, be sure to delete any existing ones.
2535 Do not use weak symbol support, even if it is provided by the linker.
2536 By default, G++ uses weak symbols if they are available. This
2537 option exists only for testing, and should not be used by end-users;
2538 it results in inferior code and has no benefits. This option may
2539 be removed in a future release of G++.
2543 Do not search for header files in the standard directories specific to
2544 C++, but do still search the other standard directories. (This option
2545 is used when building the C++ library.)
2548 In addition, these optimization, warning, and code generation options
2549 have meanings only for C++ programs:
2552 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2555 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2556 G++ to warn when it generates code that is probably not compatible with the
2557 vendor-neutral C++ ABI@. Since G++ now defaults to
2558 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2559 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2560 or an older compatibility version is selected (with
2561 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2563 Although an effort has been made to warn about
2564 all such cases, there are probably some cases that are not warned about,
2565 even though G++ is generating incompatible code. There may also be
2566 cases where warnings are emitted even though the code that is generated
2569 You should rewrite your code to avoid these warnings if you are
2570 concerned about the fact that code generated by G++ may not be binary
2571 compatible with code generated by other compilers.
2573 @option{-Wabi} can also be used with an explicit version number to
2574 warn about compatibility with a particular @option{-fabi-version}
2575 level, e.g. @option{-Wabi=2} to warn about changes relative to
2576 @option{-fabi-version=2}. Specifying a version number also sets
2577 @option{-fabi-compat-version=@var{n}}.
2579 The known incompatibilities in @option{-fabi-version=2} (which was the
2580 default from GCC 3.4 to 4.9) include:
2585 A template with a non-type template parameter of reference type was
2586 mangled incorrectly:
2589 template <int &> struct S @{@};
2593 This was fixed in @option{-fabi-version=3}.
2596 SIMD vector types declared using @code{__attribute ((vector_size))} were
2597 mangled in a non-standard way that does not allow for overloading of
2598 functions taking vectors of different sizes.
2600 The mangling was changed in @option{-fabi-version=4}.
2603 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2604 qualifiers, and @code{decltype} of a plain declaration was folded away.
2606 These mangling issues were fixed in @option{-fabi-version=5}.
2609 Scoped enumerators passed as arguments to a variadic function are
2610 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2611 On most targets this does not actually affect the parameter passing
2612 ABI, as there is no way to pass an argument smaller than @code{int}.
2614 Also, the ABI changed the mangling of template argument packs,
2615 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2616 a class scope function used as a template argument.
2618 These issues were corrected in @option{-fabi-version=6}.
2621 Lambdas in default argument scope were mangled incorrectly, and the
2622 ABI changed the mangling of @code{nullptr_t}.
2624 These issues were corrected in @option{-fabi-version=7}.
2627 When mangling a function type with function-cv-qualifiers, the
2628 un-qualified function type was incorrectly treated as a substitution
2631 This was fixed in @option{-fabi-version=8}.
2634 It also warns about psABI-related changes. The known psABI changes at this
2640 For SysV/x86-64, unions with @code{long double} members are
2641 passed in memory as specified in psABI. For example:
2651 @code{union U} is always passed in memory.
2655 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2658 Warn when a type with an ABI tag is used in a context that does not
2659 have that ABI tag. See @ref{C++ Attributes} for more information
2662 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2663 @opindex Wctor-dtor-privacy
2664 @opindex Wno-ctor-dtor-privacy
2665 Warn when a class seems unusable because all the constructors or
2666 destructors in that class are private, and it has neither friends nor
2667 public static member functions. Also warn if there are no non-private
2668 methods, and there's at least one private member function that isn't
2669 a constructor or destructor.
2671 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2672 @opindex Wdelete-non-virtual-dtor
2673 @opindex Wno-delete-non-virtual-dtor
2674 Warn when @code{delete} is used to destroy an instance of a class that
2675 has virtual functions and non-virtual destructor. It is unsafe to delete
2676 an instance of a derived class through a pointer to a base class if the
2677 base class does not have a virtual destructor. This warning is enabled
2680 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2681 @opindex Wliteral-suffix
2682 @opindex Wno-literal-suffix
2683 Warn when a string or character literal is followed by a ud-suffix which does
2684 not begin with an underscore. As a conforming extension, GCC treats such
2685 suffixes as separate preprocessing tokens in order to maintain backwards
2686 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2690 #define __STDC_FORMAT_MACROS
2691 #include <inttypes.h>
2696 printf("My int64: %" PRId64"\n", i64);
2700 In this case, @code{PRId64} is treated as a separate preprocessing token.
2702 This warning is enabled by default.
2704 @item -Wlto-type-mismatch
2705 @opindex Wlto-type-mismatch
2706 @opindex Wno-lto-type-mistmach
2708 During the link-time optimization warn about type mismatches in between
2709 global declarations from different compilation units.
2710 Requires @option{-flto} to be enabled. Enabled by default.
2712 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2714 @opindex Wno-narrowing
2715 Warn when a narrowing conversion prohibited by C++11 occurs within
2719 int i = @{ 2.2 @}; // error: narrowing from double to int
2722 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2724 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2725 required by the standard. Note that this does not affect the meaning
2726 of well-formed code; narrowing conversions are still considered
2727 ill-formed in SFINAE context.
2729 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2731 @opindex Wno-noexcept
2732 Warn when a noexcept-expression evaluates to false because of a call
2733 to a function that does not have a non-throwing exception
2734 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2735 the compiler to never throw an exception.
2737 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2738 @opindex Wnon-virtual-dtor
2739 @opindex Wno-non-virtual-dtor
2740 Warn when a class has virtual functions and an accessible non-virtual
2741 destructor itself or in an accessible polymorphic base class, in which
2742 case it is possible but unsafe to delete an instance of a derived
2743 class through a pointer to the class itself or base class. This
2744 warning is automatically enabled if @option{-Weffc++} is specified.
2746 @item -Wreorder @r{(C++ and Objective-C++ only)}
2748 @opindex Wno-reorder
2749 @cindex reordering, warning
2750 @cindex warning for reordering of member initializers
2751 Warn when the order of member initializers given in the code does not
2752 match the order in which they must be executed. For instance:
2758 A(): j (0), i (1) @{ @}
2763 The compiler rearranges the member initializers for @code{i}
2764 and @code{j} to match the declaration order of the members, emitting
2765 a warning to that effect. This warning is enabled by @option{-Wall}.
2767 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2768 @opindex fext-numeric-literals
2769 @opindex fno-ext-numeric-literals
2770 Accept imaginary, fixed-point, or machine-defined
2771 literal number suffixes as GNU extensions.
2772 When this option is turned off these suffixes are treated
2773 as C++11 user-defined literal numeric suffixes.
2774 This is on by default for all pre-C++11 dialects and all GNU dialects:
2775 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2776 @option{-std=gnu++14}.
2777 This option is off by default
2778 for ISO C++11 onwards (@option{-std=c++11}, ...).
2781 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2784 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2787 Warn about violations of the following style guidelines from Scott Meyers'
2788 @cite{Effective C++} series of books:
2792 Define a copy constructor and an assignment operator for classes
2793 with dynamically-allocated memory.
2796 Prefer initialization to assignment in constructors.
2799 Have @code{operator=} return a reference to @code{*this}.
2802 Don't try to return a reference when you must return an object.
2805 Distinguish between prefix and postfix forms of increment and
2806 decrement operators.
2809 Never overload @code{&&}, @code{||}, or @code{,}.
2813 This option also enables @option{-Wnon-virtual-dtor}, which is also
2814 one of the effective C++ recommendations. However, the check is
2815 extended to warn about the lack of virtual destructor in accessible
2816 non-polymorphic bases classes too.
2818 When selecting this option, be aware that the standard library
2819 headers do not obey all of these guidelines; use @samp{grep -v}
2820 to filter out those warnings.
2822 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2823 @opindex Wstrict-null-sentinel
2824 @opindex Wno-strict-null-sentinel
2825 Warn about the use of an uncasted @code{NULL} as sentinel. When
2826 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2827 to @code{__null}. Although it is a null pointer constant rather than a
2828 null pointer, it is guaranteed to be of the same size as a pointer.
2829 But this use is not portable across different compilers.
2831 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2832 @opindex Wno-non-template-friend
2833 @opindex Wnon-template-friend
2834 Disable warnings when non-templatized friend functions are declared
2835 within a template. Since the advent of explicit template specification
2836 support in G++, if the name of the friend is an unqualified-id (i.e.,
2837 @samp{friend foo(int)}), the C++ language specification demands that the
2838 friend declare or define an ordinary, nontemplate function. (Section
2839 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2840 could be interpreted as a particular specialization of a templatized
2841 function. Because this non-conforming behavior is no longer the default
2842 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2843 check existing code for potential trouble spots and is on by default.
2844 This new compiler behavior can be turned off with
2845 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2846 but disables the helpful warning.
2848 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2849 @opindex Wold-style-cast
2850 @opindex Wno-old-style-cast
2851 Warn if an old-style (C-style) cast to a non-void type is used within
2852 a C++ program. The new-style casts (@code{dynamic_cast},
2853 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2854 less vulnerable to unintended effects and much easier to search for.
2856 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2857 @opindex Woverloaded-virtual
2858 @opindex Wno-overloaded-virtual
2859 @cindex overloaded virtual function, warning
2860 @cindex warning for overloaded virtual function
2861 Warn when a function declaration hides virtual functions from a
2862 base class. For example, in:
2869 struct B: public A @{
2874 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2885 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2886 @opindex Wno-pmf-conversions
2887 @opindex Wpmf-conversions
2888 Disable the diagnostic for converting a bound pointer to member function
2891 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2892 @opindex Wsign-promo
2893 @opindex Wno-sign-promo
2894 Warn when overload resolution chooses a promotion from unsigned or
2895 enumerated type to a signed type, over a conversion to an unsigned type of
2896 the same size. Previous versions of G++ tried to preserve
2897 unsignedness, but the standard mandates the current behavior.
2899 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
2901 @opindex Wno-terminate
2902 Disable the warning about a throw-expression that will immediately
2903 result in a call to @code{terminate}.
2906 @node Objective-C and Objective-C++ Dialect Options
2907 @section Options Controlling Objective-C and Objective-C++ Dialects
2909 @cindex compiler options, Objective-C and Objective-C++
2910 @cindex Objective-C and Objective-C++ options, command-line
2911 @cindex options, Objective-C and Objective-C++
2912 (NOTE: This manual does not describe the Objective-C and Objective-C++
2913 languages themselves. @xref{Standards,,Language Standards
2914 Supported by GCC}, for references.)
2916 This section describes the command-line options that are only meaningful
2917 for Objective-C and Objective-C++ programs. You can also use most of
2918 the language-independent GNU compiler options.
2919 For example, you might compile a file @file{some_class.m} like this:
2922 gcc -g -fgnu-runtime -O -c some_class.m
2926 In this example, @option{-fgnu-runtime} is an option meant only for
2927 Objective-C and Objective-C++ programs; you can use the other options with
2928 any language supported by GCC@.
2930 Note that since Objective-C is an extension of the C language, Objective-C
2931 compilations may also use options specific to the C front-end (e.g.,
2932 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2933 C++-specific options (e.g., @option{-Wabi}).
2935 Here is a list of options that are @emph{only} for compiling Objective-C
2936 and Objective-C++ programs:
2939 @item -fconstant-string-class=@var{class-name}
2940 @opindex fconstant-string-class
2941 Use @var{class-name} as the name of the class to instantiate for each
2942 literal string specified with the syntax @code{@@"@dots{}"}. The default
2943 class name is @code{NXConstantString} if the GNU runtime is being used, and
2944 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2945 @option{-fconstant-cfstrings} option, if also present, overrides the
2946 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2947 to be laid out as constant CoreFoundation strings.
2950 @opindex fgnu-runtime
2951 Generate object code compatible with the standard GNU Objective-C
2952 runtime. This is the default for most types of systems.
2954 @item -fnext-runtime
2955 @opindex fnext-runtime
2956 Generate output compatible with the NeXT runtime. This is the default
2957 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2958 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2961 @item -fno-nil-receivers
2962 @opindex fno-nil-receivers
2963 Assume that all Objective-C message dispatches (@code{[receiver
2964 message:arg]}) in this translation unit ensure that the receiver is
2965 not @code{nil}. This allows for more efficient entry points in the
2966 runtime to be used. This option is only available in conjunction with
2967 the NeXT runtime and ABI version 0 or 1.
2969 @item -fobjc-abi-version=@var{n}
2970 @opindex fobjc-abi-version
2971 Use version @var{n} of the Objective-C ABI for the selected runtime.
2972 This option is currently supported only for the NeXT runtime. In that
2973 case, Version 0 is the traditional (32-bit) ABI without support for
2974 properties and other Objective-C 2.0 additions. Version 1 is the
2975 traditional (32-bit) ABI with support for properties and other
2976 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2977 nothing is specified, the default is Version 0 on 32-bit target
2978 machines, and Version 2 on 64-bit target machines.
2980 @item -fobjc-call-cxx-cdtors
2981 @opindex fobjc-call-cxx-cdtors
2982 For each Objective-C class, check if any of its instance variables is a
2983 C++ object with a non-trivial default constructor. If so, synthesize a
2984 special @code{- (id) .cxx_construct} instance method which runs
2985 non-trivial default constructors on any such instance variables, in order,
2986 and then return @code{self}. Similarly, check if any instance variable
2987 is a C++ object with a non-trivial destructor, and if so, synthesize a
2988 special @code{- (void) .cxx_destruct} method which runs
2989 all such default destructors, in reverse order.
2991 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2992 methods thusly generated only operate on instance variables
2993 declared in the current Objective-C class, and not those inherited
2994 from superclasses. It is the responsibility of the Objective-C
2995 runtime to invoke all such methods in an object's inheritance
2996 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2997 by the runtime immediately after a new object instance is allocated;
2998 the @code{- (void) .cxx_destruct} methods are invoked immediately
2999 before the runtime deallocates an object instance.
3001 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3002 support for invoking the @code{- (id) .cxx_construct} and
3003 @code{- (void) .cxx_destruct} methods.
3005 @item -fobjc-direct-dispatch
3006 @opindex fobjc-direct-dispatch
3007 Allow fast jumps to the message dispatcher. On Darwin this is
3008 accomplished via the comm page.
3010 @item -fobjc-exceptions
3011 @opindex fobjc-exceptions
3012 Enable syntactic support for structured exception handling in
3013 Objective-C, similar to what is offered by C++ and Java. This option
3014 is required to use the Objective-C keywords @code{@@try},
3015 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3016 @code{@@synchronized}. This option is available with both the GNU
3017 runtime and the NeXT runtime (but not available in conjunction with
3018 the NeXT runtime on Mac OS X 10.2 and earlier).
3022 Enable garbage collection (GC) in Objective-C and Objective-C++
3023 programs. This option is only available with the NeXT runtime; the
3024 GNU runtime has a different garbage collection implementation that
3025 does not require special compiler flags.
3027 @item -fobjc-nilcheck
3028 @opindex fobjc-nilcheck
3029 For the NeXT runtime with version 2 of the ABI, check for a nil
3030 receiver in method invocations before doing the actual method call.
3031 This is the default and can be disabled using
3032 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3033 checked for nil in this way no matter what this flag is set to.
3034 Currently this flag does nothing when the GNU runtime, or an older
3035 version of the NeXT runtime ABI, is used.
3037 @item -fobjc-std=objc1
3039 Conform to the language syntax of Objective-C 1.0, the language
3040 recognized by GCC 4.0. This only affects the Objective-C additions to
3041 the C/C++ language; it does not affect conformance to C/C++ standards,
3042 which is controlled by the separate C/C++ dialect option flags. When
3043 this option is used with the Objective-C or Objective-C++ compiler,
3044 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3045 This is useful if you need to make sure that your Objective-C code can
3046 be compiled with older versions of GCC@.
3048 @item -freplace-objc-classes
3049 @opindex freplace-objc-classes
3050 Emit a special marker instructing @command{ld(1)} not to statically link in
3051 the resulting object file, and allow @command{dyld(1)} to load it in at
3052 run time instead. This is used in conjunction with the Fix-and-Continue
3053 debugging mode, where the object file in question may be recompiled and
3054 dynamically reloaded in the course of program execution, without the need
3055 to restart the program itself. Currently, Fix-and-Continue functionality
3056 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3061 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3062 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3063 compile time) with static class references that get initialized at load time,
3064 which improves run-time performance. Specifying the @option{-fzero-link} flag
3065 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3066 to be retained. This is useful in Zero-Link debugging mode, since it allows
3067 for individual class implementations to be modified during program execution.
3068 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3069 regardless of command-line options.
3071 @item -fno-local-ivars
3072 @opindex fno-local-ivars
3073 @opindex flocal-ivars
3074 By default instance variables in Objective-C can be accessed as if
3075 they were local variables from within the methods of the class they're
3076 declared in. This can lead to shadowing between instance variables
3077 and other variables declared either locally inside a class method or
3078 globally with the same name. Specifying the @option{-fno-local-ivars}
3079 flag disables this behavior thus avoiding variable shadowing issues.
3081 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3082 @opindex fivar-visibility
3083 Set the default instance variable visibility to the specified option
3084 so that instance variables declared outside the scope of any access
3085 modifier directives default to the specified visibility.
3089 Dump interface declarations for all classes seen in the source file to a
3090 file named @file{@var{sourcename}.decl}.
3092 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3093 @opindex Wassign-intercept
3094 @opindex Wno-assign-intercept
3095 Warn whenever an Objective-C assignment is being intercepted by the
3098 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3099 @opindex Wno-protocol
3101 If a class is declared to implement a protocol, a warning is issued for
3102 every method in the protocol that is not implemented by the class. The
3103 default behavior is to issue a warning for every method not explicitly
3104 implemented in the class, even if a method implementation is inherited
3105 from the superclass. If you use the @option{-Wno-protocol} option, then
3106 methods inherited from the superclass are considered to be implemented,
3107 and no warning is issued for them.
3109 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3111 @opindex Wno-selector
3112 Warn if multiple methods of different types for the same selector are
3113 found during compilation. The check is performed on the list of methods
3114 in the final stage of compilation. Additionally, a check is performed
3115 for each selector appearing in a @code{@@selector(@dots{})}
3116 expression, and a corresponding method for that selector has been found
3117 during compilation. Because these checks scan the method table only at
3118 the end of compilation, these warnings are not produced if the final
3119 stage of compilation is not reached, for example because an error is
3120 found during compilation, or because the @option{-fsyntax-only} option is
3123 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3124 @opindex Wstrict-selector-match
3125 @opindex Wno-strict-selector-match
3126 Warn if multiple methods with differing argument and/or return types are
3127 found for a given selector when attempting to send a message using this
3128 selector to a receiver of type @code{id} or @code{Class}. When this flag
3129 is off (which is the default behavior), the compiler omits such warnings
3130 if any differences found are confined to types that share the same size
3133 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3134 @opindex Wundeclared-selector
3135 @opindex Wno-undeclared-selector
3136 Warn if a @code{@@selector(@dots{})} expression referring to an
3137 undeclared selector is found. A selector is considered undeclared if no
3138 method with that name has been declared before the
3139 @code{@@selector(@dots{})} expression, either explicitly in an
3140 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3141 an @code{@@implementation} section. This option always performs its
3142 checks as soon as a @code{@@selector(@dots{})} expression is found,
3143 while @option{-Wselector} only performs its checks in the final stage of
3144 compilation. This also enforces the coding style convention
3145 that methods and selectors must be declared before being used.
3147 @item -print-objc-runtime-info
3148 @opindex print-objc-runtime-info
3149 Generate C header describing the largest structure that is passed by
3154 @node Language Independent Options
3155 @section Options to Control Diagnostic Messages Formatting
3156 @cindex options to control diagnostics formatting
3157 @cindex diagnostic messages
3158 @cindex message formatting
3160 Traditionally, diagnostic messages have been formatted irrespective of
3161 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3162 options described below
3163 to control the formatting algorithm for diagnostic messages,
3164 e.g.@: how many characters per line, how often source location
3165 information should be reported. Note that some language front ends may not
3166 honor these options.
3169 @item -fmessage-length=@var{n}
3170 @opindex fmessage-length
3171 Try to format error messages so that they fit on lines of about
3172 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3173 done; each error message appears on a single line. This is the
3174 default for all front ends.
3176 @item -fdiagnostics-show-location=once
3177 @opindex fdiagnostics-show-location
3178 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3179 reporter to emit source location information @emph{once}; that is, in
3180 case the message is too long to fit on a single physical line and has to
3181 be wrapped, the source location won't be emitted (as prefix) again,
3182 over and over, in subsequent continuation lines. This is the default
3185 @item -fdiagnostics-show-location=every-line
3186 Only meaningful in line-wrapping mode. Instructs the diagnostic
3187 messages reporter to emit the same source location information (as
3188 prefix) for physical lines that result from the process of breaking
3189 a message which is too long to fit on a single line.
3191 @item -fdiagnostics-color[=@var{WHEN}]
3192 @itemx -fno-diagnostics-color
3193 @opindex fdiagnostics-color
3194 @cindex highlight, color, colour
3195 @vindex GCC_COLORS @r{environment variable}
3196 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3197 or @samp{auto}. The default depends on how the compiler has been configured,
3198 it can be any of the above @var{WHEN} options or also @samp{never}
3199 if @env{GCC_COLORS} environment variable isn't present in the environment,
3200 and @samp{auto} otherwise.
3201 @samp{auto} means to use color only when the standard error is a terminal.
3202 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3203 aliases for @option{-fdiagnostics-color=always} and
3204 @option{-fdiagnostics-color=never}, respectively.
3206 The colors are defined by the environment variable @env{GCC_COLORS}.
3207 Its value is a colon-separated list of capabilities and Select Graphic
3208 Rendition (SGR) substrings. SGR commands are interpreted by the
3209 terminal or terminal emulator. (See the section in the documentation
3210 of your text terminal for permitted values and their meanings as
3211 character attributes.) These substring values are integers in decimal
3212 representation and can be concatenated with semicolons.
3213 Common values to concatenate include
3215 @samp{4} for underline,
3217 @samp{7} for inverse,
3218 @samp{39} for default foreground color,
3219 @samp{30} to @samp{37} for foreground colors,
3220 @samp{90} to @samp{97} for 16-color mode foreground colors,
3221 @samp{38;5;0} to @samp{38;5;255}
3222 for 88-color and 256-color modes foreground colors,
3223 @samp{49} for default background color,
3224 @samp{40} to @samp{47} for background colors,
3225 @samp{100} to @samp{107} for 16-color mode background colors,
3226 and @samp{48;5;0} to @samp{48;5;255}
3227 for 88-color and 256-color modes background colors.
3229 The default @env{GCC_COLORS} is
3231 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3234 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3235 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3236 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3237 string disables colors.
3238 Supported capabilities are as follows.
3242 @vindex error GCC_COLORS @r{capability}
3243 SGR substring for error: markers.
3246 @vindex warning GCC_COLORS @r{capability}
3247 SGR substring for warning: markers.
3250 @vindex note GCC_COLORS @r{capability}
3251 SGR substring for note: markers.
3254 @vindex caret GCC_COLORS @r{capability}
3255 SGR substring for caret line.
3258 @vindex locus GCC_COLORS @r{capability}
3259 SGR substring for location information, @samp{file:line} or
3260 @samp{file:line:column} etc.
3263 @vindex quote GCC_COLORS @r{capability}
3264 SGR substring for information printed within quotes.
3267 @item -fno-diagnostics-show-option
3268 @opindex fno-diagnostics-show-option
3269 @opindex fdiagnostics-show-option
3270 By default, each diagnostic emitted includes text indicating the
3271 command-line option that directly controls the diagnostic (if such an
3272 option is known to the diagnostic machinery). Specifying the
3273 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3275 @item -fno-diagnostics-show-caret
3276 @opindex fno-diagnostics-show-caret
3277 @opindex fdiagnostics-show-caret
3278 By default, each diagnostic emitted includes the original source line
3279 and a caret '^' indicating the column. This option suppresses this
3280 information. The source line is truncated to @var{n} characters, if
3281 the @option{-fmessage-length=n} option is given. When the output is done
3282 to the terminal, the width is limited to the width given by the
3283 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3287 @node Warning Options
3288 @section Options to Request or Suppress Warnings
3289 @cindex options to control warnings
3290 @cindex warning messages
3291 @cindex messages, warning
3292 @cindex suppressing warnings
3294 Warnings are diagnostic messages that report constructions that
3295 are not inherently erroneous but that are risky or suggest there
3296 may have been an error.
3298 The following language-independent options do not enable specific
3299 warnings but control the kinds of diagnostics produced by GCC@.
3302 @cindex syntax checking
3304 @opindex fsyntax-only
3305 Check the code for syntax errors, but don't do anything beyond that.
3307 @item -fmax-errors=@var{n}
3308 @opindex fmax-errors
3309 Limits the maximum number of error messages to @var{n}, at which point
3310 GCC bails out rather than attempting to continue processing the source
3311 code. If @var{n} is 0 (the default), there is no limit on the number
3312 of error messages produced. If @option{-Wfatal-errors} is also
3313 specified, then @option{-Wfatal-errors} takes precedence over this
3318 Inhibit all warning messages.
3323 Make all warnings into errors.
3328 Make the specified warning into an error. The specifier for a warning
3329 is appended; for example @option{-Werror=switch} turns the warnings
3330 controlled by @option{-Wswitch} into errors. This switch takes a
3331 negative form, to be used to negate @option{-Werror} for specific
3332 warnings; for example @option{-Wno-error=switch} makes
3333 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3336 The warning message for each controllable warning includes the
3337 option that controls the warning. That option can then be used with
3338 @option{-Werror=} and @option{-Wno-error=} as described above.
3339 (Printing of the option in the warning message can be disabled using the
3340 @option{-fno-diagnostics-show-option} flag.)
3342 Note that specifying @option{-Werror=}@var{foo} automatically implies
3343 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3346 @item -Wfatal-errors
3347 @opindex Wfatal-errors
3348 @opindex Wno-fatal-errors
3349 This option causes the compiler to abort compilation on the first error
3350 occurred rather than trying to keep going and printing further error
3355 You can request many specific warnings with options beginning with
3356 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3357 implicit declarations. Each of these specific warning options also
3358 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3359 example, @option{-Wno-implicit}. This manual lists only one of the
3360 two forms, whichever is not the default. For further
3361 language-specific options also refer to @ref{C++ Dialect Options} and
3362 @ref{Objective-C and Objective-C++ Dialect Options}.
3364 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3365 options, such as @option{-Wunused}, which may turn on further options,
3366 such as @option{-Wunused-value}. The combined effect of positive and
3367 negative forms is that more specific options have priority over less
3368 specific ones, independently of their position in the command-line. For
3369 options of the same specificity, the last one takes effect. Options
3370 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3371 as if they appeared at the end of the command-line.
3373 When an unrecognized warning option is requested (e.g.,
3374 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3375 that the option is not recognized. However, if the @option{-Wno-} form
3376 is used, the behavior is slightly different: no diagnostic is
3377 produced for @option{-Wno-unknown-warning} unless other diagnostics
3378 are being produced. This allows the use of new @option{-Wno-} options
3379 with old compilers, but if something goes wrong, the compiler
3380 warns that an unrecognized option is present.
3387 Issue all the warnings demanded by strict ISO C and ISO C++;
3388 reject all programs that use forbidden extensions, and some other
3389 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3390 version of the ISO C standard specified by any @option{-std} option used.
3392 Valid ISO C and ISO C++ programs should compile properly with or without
3393 this option (though a rare few require @option{-ansi} or a
3394 @option{-std} option specifying the required version of ISO C)@. However,
3395 without this option, certain GNU extensions and traditional C and C++
3396 features are supported as well. With this option, they are rejected.
3398 @option{-Wpedantic} does not cause warning messages for use of the
3399 alternate keywords whose names begin and end with @samp{__}. Pedantic
3400 warnings are also disabled in the expression that follows
3401 @code{__extension__}. However, only system header files should use
3402 these escape routes; application programs should avoid them.
3403 @xref{Alternate Keywords}.
3405 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3406 C conformance. They soon find that it does not do quite what they want:
3407 it finds some non-ISO practices, but not all---only those for which
3408 ISO C @emph{requires} a diagnostic, and some others for which
3409 diagnostics have been added.
3411 A feature to report any failure to conform to ISO C might be useful in
3412 some instances, but would require considerable additional work and would
3413 be quite different from @option{-Wpedantic}. We don't have plans to
3414 support such a feature in the near future.
3416 Where the standard specified with @option{-std} represents a GNU
3417 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3418 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3419 extended dialect is based. Warnings from @option{-Wpedantic} are given
3420 where they are required by the base standard. (It does not make sense
3421 for such warnings to be given only for features not in the specified GNU
3422 C dialect, since by definition the GNU dialects of C include all
3423 features the compiler supports with the given option, and there would be
3424 nothing to warn about.)
3426 @item -pedantic-errors
3427 @opindex pedantic-errors
3428 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3429 requires a diagnostic, in some cases where there is undefined behavior
3430 at compile-time and in some other cases that do not prevent compilation
3431 of programs that are valid according to the standard. This is not
3432 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3433 by this option and not enabled by the latter and vice versa.
3438 This enables all the warnings about constructions that some users
3439 consider questionable, and that are easy to avoid (or modify to
3440 prevent the warning), even in conjunction with macros. This also
3441 enables some language-specific warnings described in @ref{C++ Dialect
3442 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3444 @option{-Wall} turns on the following warning flags:
3446 @gccoptlist{-Waddress @gol
3447 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3448 -Wc++11-compat -Wc++14-compat@gol
3449 -Wchar-subscripts @gol
3450 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3451 -Wimplicit-int @r{(C and Objective-C only)} @gol
3452 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3455 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3456 -Wmaybe-uninitialized @gol
3457 -Wmissing-braces @r{(only for C/ObjC)} @gol
3464 -Wsequence-point @gol
3465 -Wsign-compare @r{(only in C++)} @gol
3466 -Wstrict-aliasing @gol
3467 -Wstrict-overflow=1 @gol
3470 -Wuninitialized @gol
3471 -Wunknown-pragmas @gol
3472 -Wunused-function @gol
3475 -Wunused-variable @gol
3476 -Wvolatile-register-var @gol
3479 Note that some warning flags are not implied by @option{-Wall}. Some of
3480 them warn about constructions that users generally do not consider
3481 questionable, but which occasionally you might wish to check for;
3482 others warn about constructions that are necessary or hard to avoid in
3483 some cases, and there is no simple way to modify the code to suppress
3484 the warning. Some of them are enabled by @option{-Wextra} but many of
3485 them must be enabled individually.
3491 This enables some extra warning flags that are not enabled by
3492 @option{-Wall}. (This option used to be called @option{-W}. The older
3493 name is still supported, but the newer name is more descriptive.)
3495 @gccoptlist{-Wclobbered @gol
3497 -Wignored-qualifiers @gol
3498 -Wmissing-field-initializers @gol
3499 -Wmissing-parameter-type @r{(C only)} @gol
3500 -Wold-style-declaration @r{(C only)} @gol
3501 -Woverride-init @gol
3504 -Wuninitialized @gol
3505 -Wshift-negative-value @gol
3506 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3507 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3510 The option @option{-Wextra} also prints warning messages for the
3516 A pointer is compared against integer zero with @code{<}, @code{<=},
3517 @code{>}, or @code{>=}.
3520 (C++ only) An enumerator and a non-enumerator both appear in a
3521 conditional expression.
3524 (C++ only) Ambiguous virtual bases.
3527 (C++ only) Subscripting an array that has been declared @code{register}.
3530 (C++ only) Taking the address of a variable that has been declared
3534 (C++ only) A base class is not initialized in a derived class's copy
3539 @item -Wchar-subscripts
3540 @opindex Wchar-subscripts
3541 @opindex Wno-char-subscripts
3542 Warn if an array subscript has type @code{char}. This is a common cause
3543 of error, as programmers often forget that this type is signed on some
3545 This warning is enabled by @option{-Wall}.
3549 @opindex Wno-comment
3550 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3551 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3552 This warning is enabled by @option{-Wall}.
3554 @item -Wno-coverage-mismatch
3555 @opindex Wno-coverage-mismatch
3556 Warn if feedback profiles do not match when using the
3557 @option{-fprofile-use} option.
3558 If a source file is changed between compiling with @option{-fprofile-gen} and
3559 with @option{-fprofile-use}, the files with the profile feedback can fail
3560 to match the source file and GCC cannot use the profile feedback
3561 information. By default, this warning is enabled and is treated as an
3562 error. @option{-Wno-coverage-mismatch} can be used to disable the
3563 warning or @option{-Wno-error=coverage-mismatch} can be used to
3564 disable the error. Disabling the error for this warning can result in
3565 poorly optimized code and is useful only in the
3566 case of very minor changes such as bug fixes to an existing code-base.
3567 Completely disabling the warning is not recommended.
3570 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3572 Suppress warning messages emitted by @code{#warning} directives.
3574 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3575 @opindex Wdouble-promotion
3576 @opindex Wno-double-promotion
3577 Give a warning when a value of type @code{float} is implicitly
3578 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3579 floating-point unit implement @code{float} in hardware, but emulate
3580 @code{double} in software. On such a machine, doing computations
3581 using @code{double} values is much more expensive because of the
3582 overhead required for software emulation.
3584 It is easy to accidentally do computations with @code{double} because
3585 floating-point literals are implicitly of type @code{double}. For
3589 float area(float radius)
3591 return 3.14159 * radius * radius;
3595 the compiler performs the entire computation with @code{double}
3596 because the floating-point literal is a @code{double}.
3599 @itemx -Wformat=@var{n}
3602 @opindex ffreestanding
3603 @opindex fno-builtin
3605 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3606 the arguments supplied have types appropriate to the format string
3607 specified, and that the conversions specified in the format string make
3608 sense. This includes standard functions, and others specified by format
3609 attributes (@pxref{Function Attributes}), in the @code{printf},
3610 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3611 not in the C standard) families (or other target-specific families).
3612 Which functions are checked without format attributes having been
3613 specified depends on the standard version selected, and such checks of
3614 functions without the attribute specified are disabled by
3615 @option{-ffreestanding} or @option{-fno-builtin}.
3617 The formats are checked against the format features supported by GNU
3618 libc version 2.2. These include all ISO C90 and C99 features, as well
3619 as features from the Single Unix Specification and some BSD and GNU
3620 extensions. Other library implementations may not support all these
3621 features; GCC does not support warning about features that go beyond a
3622 particular library's limitations. However, if @option{-Wpedantic} is used
3623 with @option{-Wformat}, warnings are given about format features not
3624 in the selected standard version (but not for @code{strfmon} formats,
3625 since those are not in any version of the C standard). @xref{C Dialect
3626 Options,,Options Controlling C Dialect}.
3633 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3634 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3635 @option{-Wformat} also checks for null format arguments for several
3636 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3637 aspects of this level of format checking can be disabled by the
3638 options: @option{-Wno-format-contains-nul},
3639 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3640 @option{-Wformat} is enabled by @option{-Wall}.
3642 @item -Wno-format-contains-nul
3643 @opindex Wno-format-contains-nul
3644 @opindex Wformat-contains-nul
3645 If @option{-Wformat} is specified, do not warn about format strings that
3648 @item -Wno-format-extra-args
3649 @opindex Wno-format-extra-args
3650 @opindex Wformat-extra-args
3651 If @option{-Wformat} is specified, do not warn about excess arguments to a
3652 @code{printf} or @code{scanf} format function. The C standard specifies
3653 that such arguments are ignored.
3655 Where the unused arguments lie between used arguments that are
3656 specified with @samp{$} operand number specifications, normally
3657 warnings are still given, since the implementation could not know what
3658 type to pass to @code{va_arg} to skip the unused arguments. However,
3659 in the case of @code{scanf} formats, this option suppresses the
3660 warning if the unused arguments are all pointers, since the Single
3661 Unix Specification says that such unused arguments are allowed.
3663 @item -Wno-format-zero-length
3664 @opindex Wno-format-zero-length
3665 @opindex Wformat-zero-length
3666 If @option{-Wformat} is specified, do not warn about zero-length formats.
3667 The C standard specifies that zero-length formats are allowed.
3672 Enable @option{-Wformat} plus additional format checks. Currently
3673 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3676 @item -Wformat-nonliteral
3677 @opindex Wformat-nonliteral
3678 @opindex Wno-format-nonliteral
3679 If @option{-Wformat} is specified, also warn if the format string is not a
3680 string literal and so cannot be checked, unless the format function
3681 takes its format arguments as a @code{va_list}.
3683 @item -Wformat-security
3684 @opindex Wformat-security
3685 @opindex Wno-format-security
3686 If @option{-Wformat} is specified, also warn about uses of format
3687 functions that represent possible security problems. At present, this
3688 warns about calls to @code{printf} and @code{scanf} functions where the
3689 format string is not a string literal and there are no format arguments,
3690 as in @code{printf (foo);}. This may be a security hole if the format
3691 string came from untrusted input and contains @samp{%n}. (This is
3692 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3693 in future warnings may be added to @option{-Wformat-security} that are not
3694 included in @option{-Wformat-nonliteral}.)
3696 @item -Wformat-signedness
3697 @opindex Wformat-signedness
3698 @opindex Wno-format-signedness
3699 If @option{-Wformat} is specified, also warn if the format string
3700 requires an unsigned argument and the argument is signed and vice versa.
3703 @opindex Wformat-y2k
3704 @opindex Wno-format-y2k
3705 If @option{-Wformat} is specified, also warn about @code{strftime}
3706 formats that may yield only a two-digit year.
3711 @opindex Wno-nonnull
3712 Warn about passing a null pointer for arguments marked as
3713 requiring a non-null value by the @code{nonnull} function attribute.
3715 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3716 can be disabled with the @option{-Wno-nonnull} option.
3718 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3720 @opindex Wno-init-self
3721 Warn about uninitialized variables that are initialized with themselves.
3722 Note this option can only be used with the @option{-Wuninitialized} option.
3724 For example, GCC warns about @code{i} being uninitialized in the
3725 following snippet only when @option{-Winit-self} has been specified:
3736 This warning is enabled by @option{-Wall} in C++.
3738 @item -Wimplicit-int @r{(C and Objective-C only)}
3739 @opindex Wimplicit-int
3740 @opindex Wno-implicit-int
3741 Warn when a declaration does not specify a type.
3742 This warning is enabled by @option{-Wall}.
3744 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3745 @opindex Wimplicit-function-declaration
3746 @opindex Wno-implicit-function-declaration
3747 Give a warning whenever a function is used before being declared. In
3748 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3749 enabled by default and it is made into an error by
3750 @option{-pedantic-errors}. This warning is also enabled by
3753 @item -Wimplicit @r{(C and Objective-C only)}
3755 @opindex Wno-implicit
3756 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3757 This warning is enabled by @option{-Wall}.
3759 @item -Wignored-qualifiers @r{(C and C++ only)}
3760 @opindex Wignored-qualifiers
3761 @opindex Wno-ignored-qualifiers
3762 Warn if the return type of a function has a type qualifier
3763 such as @code{const}. For ISO C such a type qualifier has no effect,
3764 since the value returned by a function is not an lvalue.
3765 For C++, the warning is only emitted for scalar types or @code{void}.
3766 ISO C prohibits qualified @code{void} return types on function
3767 definitions, so such return types always receive a warning
3768 even without this option.
3770 This warning is also enabled by @option{-Wextra}.
3775 Warn if the type of @code{main} is suspicious. @code{main} should be
3776 a function with external linkage, returning int, taking either zero
3777 arguments, two, or three arguments of appropriate types. This warning
3778 is enabled by default in C++ and is enabled by either @option{-Wall}
3779 or @option{-Wpedantic}.
3781 @item -Wmisleading-indentation @r{(C and C++ only)}
3782 @opindex Wmisleading-indentation
3783 @opindex Wno-misleading-indentation
3784 Warn when the indentation of the code does not reflect the block structure.
3785 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3786 @code{for} clauses with a guarded statement that does not use braces,
3787 followed by an unguarded statement with the same indentation.
3789 This warning is disabled by default.
3791 In the following example, the call to ``bar'' is misleadingly indented as
3792 if it were guarded by the ``if'' conditional.
3795 if (some_condition ())
3797 bar (); /* Gotcha: this is not guarded by the "if". */
3800 In the case of mixed tabs and spaces, the warning uses the
3801 @option{-ftabstop=} option to determine if the statements line up
3804 The warning is not issued for code involving multiline preprocessor logic
3805 such as the following example.
3810 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3816 The warning is not issued after a @code{#line} directive, since this
3817 typically indicates autogenerated code, and no assumptions can be made
3818 about the layout of the file that the directive references.
3820 @item -Wmissing-braces
3821 @opindex Wmissing-braces
3822 @opindex Wno-missing-braces
3823 Warn if an aggregate or union initializer is not fully bracketed. In
3824 the following example, the initializer for @code{a} is not fully
3825 bracketed, but that for @code{b} is fully bracketed. This warning is
3826 enabled by @option{-Wall} in C.
3829 int a[2][2] = @{ 0, 1, 2, 3 @};
3830 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3833 This warning is enabled by @option{-Wall}.
3835 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3836 @opindex Wmissing-include-dirs
3837 @opindex Wno-missing-include-dirs
3838 Warn if a user-supplied include directory does not exist.
3841 @opindex Wparentheses
3842 @opindex Wno-parentheses
3843 Warn if parentheses are omitted in certain contexts, such
3844 as when there is an assignment in a context where a truth value
3845 is expected, or when operators are nested whose precedence people
3846 often get confused about.
3848 Also warn if a comparison like @code{x<=y<=z} appears; this is
3849 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3850 interpretation from that of ordinary mathematical notation.
3852 Also warn about constructions where there may be confusion to which
3853 @code{if} statement an @code{else} branch belongs. Here is an example of
3868 In C/C++, every @code{else} branch belongs to the innermost possible
3869 @code{if} statement, which in this example is @code{if (b)}. This is
3870 often not what the programmer expected, as illustrated in the above
3871 example by indentation the programmer chose. When there is the
3872 potential for this confusion, GCC issues a warning when this flag
3873 is specified. To eliminate the warning, add explicit braces around
3874 the innermost @code{if} statement so there is no way the @code{else}
3875 can belong to the enclosing @code{if}. The resulting code
3892 Also warn for dangerous uses of the GNU extension to
3893 @code{?:} with omitted middle operand. When the condition
3894 in the @code{?}: operator is a boolean expression, the omitted value is
3895 always 1. Often programmers expect it to be a value computed
3896 inside the conditional expression instead.
3898 This warning is enabled by @option{-Wall}.
3900 @item -Wsequence-point
3901 @opindex Wsequence-point
3902 @opindex Wno-sequence-point
3903 Warn about code that may have undefined semantics because of violations
3904 of sequence point rules in the C and C++ standards.
3906 The C and C++ standards define the order in which expressions in a C/C++
3907 program are evaluated in terms of @dfn{sequence points}, which represent
3908 a partial ordering between the execution of parts of the program: those
3909 executed before the sequence point, and those executed after it. These
3910 occur after the evaluation of a full expression (one which is not part
3911 of a larger expression), after the evaluation of the first operand of a
3912 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3913 function is called (but after the evaluation of its arguments and the
3914 expression denoting the called function), and in certain other places.
3915 Other than as expressed by the sequence point rules, the order of
3916 evaluation of subexpressions of an expression is not specified. All
3917 these rules describe only a partial order rather than a total order,
3918 since, for example, if two functions are called within one expression
3919 with no sequence point between them, the order in which the functions
3920 are called is not specified. However, the standards committee have
3921 ruled that function calls do not overlap.
3923 It is not specified when between sequence points modifications to the
3924 values of objects take effect. Programs whose behavior depends on this
3925 have undefined behavior; the C and C++ standards specify that ``Between
3926 the previous and next sequence point an object shall have its stored
3927 value modified at most once by the evaluation of an expression.
3928 Furthermore, the prior value shall be read only to determine the value
3929 to be stored.''. If a program breaks these rules, the results on any
3930 particular implementation are entirely unpredictable.
3932 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3933 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3934 diagnosed by this option, and it may give an occasional false positive
3935 result, but in general it has been found fairly effective at detecting
3936 this sort of problem in programs.
3938 The standard is worded confusingly, therefore there is some debate
3939 over the precise meaning of the sequence point rules in subtle cases.
3940 Links to discussions of the problem, including proposed formal
3941 definitions, may be found on the GCC readings page, at
3942 @uref{http://gcc.gnu.org/@/readings.html}.
3944 This warning is enabled by @option{-Wall} for C and C++.
3946 @item -Wno-return-local-addr
3947 @opindex Wno-return-local-addr
3948 @opindex Wreturn-local-addr
3949 Do not warn about returning a pointer (or in C++, a reference) to a
3950 variable that goes out of scope after the function returns.
3953 @opindex Wreturn-type
3954 @opindex Wno-return-type
3955 Warn whenever a function is defined with a return type that defaults
3956 to @code{int}. Also warn about any @code{return} statement with no
3957 return value in a function whose return type is not @code{void}
3958 (falling off the end of the function body is considered returning
3959 without a value), and about a @code{return} statement with an
3960 expression in a function whose return type is @code{void}.
3962 For C++, a function without return type always produces a diagnostic
3963 message, even when @option{-Wno-return-type} is specified. The only
3964 exceptions are @code{main} and functions defined in system headers.
3966 This warning is enabled by @option{-Wall}.
3968 @item -Wshift-count-negative
3969 @opindex Wshift-count-negative
3970 @opindex Wno-shift-count-negative
3971 Warn if shift count is negative. This warning is enabled by default.
3973 @item -Wshift-count-overflow
3974 @opindex Wshift-count-overflow
3975 @opindex Wno-shift-count-overflow
3976 Warn if shift count >= width of type. This warning is enabled by default.
3978 @item -Wshift-negative-value
3979 @opindex Wshift-negative-value
3980 @opindex Wno-shift-negative-value
3981 Warn if left shifting a negative value. This warning is enabled by
3982 @option{-Wextra} in C99 and C++11 modes (and newer).
3987 Warn whenever a @code{switch} statement has an index of enumerated type
3988 and lacks a @code{case} for one or more of the named codes of that
3989 enumeration. (The presence of a @code{default} label prevents this
3990 warning.) @code{case} labels outside the enumeration range also
3991 provoke warnings when this option is used (even if there is a
3992 @code{default} label).
3993 This warning is enabled by @option{-Wall}.
3995 @item -Wswitch-default
3996 @opindex Wswitch-default
3997 @opindex Wno-switch-default
3998 Warn whenever a @code{switch} statement does not have a @code{default}
4002 @opindex Wswitch-enum
4003 @opindex Wno-switch-enum
4004 Warn whenever a @code{switch} statement has an index of enumerated type
4005 and lacks a @code{case} for one or more of the named codes of that
4006 enumeration. @code{case} labels outside the enumeration range also
4007 provoke warnings when this option is used. The only difference
4008 between @option{-Wswitch} and this option is that this option gives a
4009 warning about an omitted enumeration code even if there is a
4010 @code{default} label.
4013 @opindex Wswitch-bool
4014 @opindex Wno-switch-bool
4015 Warn whenever a @code{switch} statement has an index of boolean type.
4016 It is possible to suppress this warning by casting the controlling
4017 expression to a type other than @code{bool}. For example:
4020 switch ((int) (a == 4))
4026 This warning is enabled by default for C and C++ programs.
4028 @item -Wsync-nand @r{(C and C++ only)}
4030 @opindex Wno-sync-nand
4031 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4032 built-in functions are used. These functions changed semantics in GCC 4.4.
4036 @opindex Wno-trigraphs
4037 Warn if any trigraphs are encountered that might change the meaning of
4038 the program (trigraphs within comments are not warned about).
4039 This warning is enabled by @option{-Wall}.
4041 @item -Wunused-but-set-parameter
4042 @opindex Wunused-but-set-parameter
4043 @opindex Wno-unused-but-set-parameter
4044 Warn whenever a function parameter is assigned to, but otherwise unused
4045 (aside from its declaration).
4047 To suppress this warning use the @code{unused} attribute
4048 (@pxref{Variable Attributes}).
4050 This warning is also enabled by @option{-Wunused} together with
4053 @item -Wunused-but-set-variable
4054 @opindex Wunused-but-set-variable
4055 @opindex Wno-unused-but-set-variable
4056 Warn whenever a local variable is assigned to, but otherwise unused
4057 (aside from its declaration).
4058 This warning is enabled by @option{-Wall}.
4060 To suppress this warning use the @code{unused} attribute
4061 (@pxref{Variable Attributes}).
4063 This warning is also enabled by @option{-Wunused}, which is enabled
4066 @item -Wunused-function
4067 @opindex Wunused-function
4068 @opindex Wno-unused-function
4069 Warn whenever a static function is declared but not defined or a
4070 non-inline static function is unused.
4071 This warning is enabled by @option{-Wall}.
4073 @item -Wunused-label
4074 @opindex Wunused-label
4075 @opindex Wno-unused-label
4076 Warn whenever a label is declared but not used.
4077 This warning is enabled by @option{-Wall}.
4079 To suppress this warning use the @code{unused} attribute
4080 (@pxref{Variable Attributes}).
4082 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4083 @opindex Wunused-local-typedefs
4084 Warn when a typedef locally defined in a function is not used.
4085 This warning is enabled by @option{-Wall}.
4087 @item -Wunused-parameter
4088 @opindex Wunused-parameter
4089 @opindex Wno-unused-parameter
4090 Warn whenever a function parameter is unused aside from its declaration.
4092 To suppress this warning use the @code{unused} attribute
4093 (@pxref{Variable Attributes}).
4095 @item -Wno-unused-result
4096 @opindex Wunused-result
4097 @opindex Wno-unused-result
4098 Do not warn if a caller of a function marked with attribute
4099 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4100 its return value. The default is @option{-Wunused-result}.
4102 @item -Wunused-variable
4103 @opindex Wunused-variable
4104 @opindex Wno-unused-variable
4105 Warn whenever a local variable or non-constant static variable is unused
4106 aside from its declaration.
4107 This warning is enabled by @option{-Wall}.
4109 To suppress this warning use the @code{unused} attribute
4110 (@pxref{Variable Attributes}).
4112 @item -Wunused-value
4113 @opindex Wunused-value
4114 @opindex Wno-unused-value
4115 Warn whenever a statement computes a result that is explicitly not
4116 used. To suppress this warning cast the unused expression to
4117 @code{void}. This includes an expression-statement or the left-hand
4118 side of a comma expression that contains no side effects. For example,
4119 an expression such as @code{x[i,j]} causes a warning, while
4120 @code{x[(void)i,j]} does not.
4122 This warning is enabled by @option{-Wall}.
4127 All the above @option{-Wunused} options combined.
4129 In order to get a warning about an unused function parameter, you must
4130 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4131 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4133 @item -Wuninitialized
4134 @opindex Wuninitialized
4135 @opindex Wno-uninitialized
4136 Warn if an automatic variable is used without first being initialized
4137 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4138 warn if a non-static reference or non-static @code{const} member
4139 appears in a class without constructors.
4141 If you want to warn about code that uses the uninitialized value of the
4142 variable in its own initializer, use the @option{-Winit-self} option.
4144 These warnings occur for individual uninitialized or clobbered
4145 elements of structure, union or array variables as well as for
4146 variables that are uninitialized or clobbered as a whole. They do
4147 not occur for variables or elements declared @code{volatile}. Because
4148 these warnings depend on optimization, the exact variables or elements
4149 for which there are warnings depends on the precise optimization
4150 options and version of GCC used.
4152 Note that there may be no warning about a variable that is used only
4153 to compute a value that itself is never used, because such
4154 computations may be deleted by data flow analysis before the warnings
4157 @item -Wmaybe-uninitialized
4158 @opindex Wmaybe-uninitialized
4159 @opindex Wno-maybe-uninitialized
4160 For an automatic variable, if there exists a path from the function
4161 entry to a use of the variable that is initialized, but there exist
4162 some other paths for which the variable is not initialized, the compiler
4163 emits a warning if it cannot prove the uninitialized paths are not
4164 executed at run time. These warnings are made optional because GCC is
4165 not smart enough to see all the reasons why the code might be correct
4166 in spite of appearing to have an error. Here is one example of how
4187 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4188 always initialized, but GCC doesn't know this. To suppress the
4189 warning, you need to provide a default case with assert(0) or
4192 @cindex @code{longjmp} warnings
4193 This option also warns when a non-volatile automatic variable might be
4194 changed by a call to @code{longjmp}. These warnings as well are possible
4195 only in optimizing compilation.
4197 The compiler sees only the calls to @code{setjmp}. It cannot know
4198 where @code{longjmp} will be called; in fact, a signal handler could
4199 call it at any point in the code. As a result, you may get a warning
4200 even when there is in fact no problem because @code{longjmp} cannot
4201 in fact be called at the place that would cause a problem.
4203 Some spurious warnings can be avoided if you declare all the functions
4204 you use that never return as @code{noreturn}. @xref{Function
4207 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4209 @item -Wunknown-pragmas
4210 @opindex Wunknown-pragmas
4211 @opindex Wno-unknown-pragmas
4212 @cindex warning for unknown pragmas
4213 @cindex unknown pragmas, warning
4214 @cindex pragmas, warning of unknown
4215 Warn when a @code{#pragma} directive is encountered that is not understood by
4216 GCC@. If this command-line option is used, warnings are even issued
4217 for unknown pragmas in system header files. This is not the case if
4218 the warnings are only enabled by the @option{-Wall} command-line option.
4221 @opindex Wno-pragmas
4223 Do not warn about misuses of pragmas, such as incorrect parameters,
4224 invalid syntax, or conflicts between pragmas. See also
4225 @option{-Wunknown-pragmas}.
4227 @item -Wstrict-aliasing
4228 @opindex Wstrict-aliasing
4229 @opindex Wno-strict-aliasing
4230 This option is only active when @option{-fstrict-aliasing} is active.
4231 It warns about code that might break the strict aliasing rules that the
4232 compiler is using for optimization. The warning does not catch all
4233 cases, but does attempt to catch the more common pitfalls. It is
4234 included in @option{-Wall}.
4235 It is equivalent to @option{-Wstrict-aliasing=3}
4237 @item -Wstrict-aliasing=n
4238 @opindex Wstrict-aliasing=n
4239 This option is only active when @option{-fstrict-aliasing} is active.
4240 It warns about code that might break the strict aliasing rules that the
4241 compiler is using for optimization.
4242 Higher levels correspond to higher accuracy (fewer false positives).
4243 Higher levels also correspond to more effort, similar to the way @option{-O}
4245 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4247 Level 1: Most aggressive, quick, least accurate.
4248 Possibly useful when higher levels
4249 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4250 false negatives. However, it has many false positives.
4251 Warns for all pointer conversions between possibly incompatible types,
4252 even if never dereferenced. Runs in the front end only.
4254 Level 2: Aggressive, quick, not too precise.
4255 May still have many false positives (not as many as level 1 though),
4256 and few false negatives (but possibly more than level 1).
4257 Unlike level 1, it only warns when an address is taken. Warns about
4258 incomplete types. Runs in the front end only.
4260 Level 3 (default for @option{-Wstrict-aliasing}):
4261 Should have very few false positives and few false
4262 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4263 Takes care of the common pun+dereference pattern in the front end:
4264 @code{*(int*)&some_float}.
4265 If optimization is enabled, it also runs in the back end, where it deals
4266 with multiple statement cases using flow-sensitive points-to information.
4267 Only warns when the converted pointer is dereferenced.
4268 Does not warn about incomplete types.
4270 @item -Wstrict-overflow
4271 @itemx -Wstrict-overflow=@var{n}
4272 @opindex Wstrict-overflow
4273 @opindex Wno-strict-overflow
4274 This option is only active when @option{-fstrict-overflow} is active.
4275 It warns about cases where the compiler optimizes based on the
4276 assumption that signed overflow does not occur. Note that it does not
4277 warn about all cases where the code might overflow: it only warns
4278 about cases where the compiler implements some optimization. Thus
4279 this warning depends on the optimization level.
4281 An optimization that assumes that signed overflow does not occur is
4282 perfectly safe if the values of the variables involved are such that
4283 overflow never does, in fact, occur. Therefore this warning can
4284 easily give a false positive: a warning about code that is not
4285 actually a problem. To help focus on important issues, several
4286 warning levels are defined. No warnings are issued for the use of
4287 undefined signed overflow when estimating how many iterations a loop
4288 requires, in particular when determining whether a loop will be
4292 @item -Wstrict-overflow=1
4293 Warn about cases that are both questionable and easy to avoid. For
4294 example, with @option{-fstrict-overflow}, the compiler simplifies
4295 @code{x + 1 > x} to @code{1}. This level of
4296 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4297 are not, and must be explicitly requested.
4299 @item -Wstrict-overflow=2
4300 Also warn about other cases where a comparison is simplified to a
4301 constant. For example: @code{abs (x) >= 0}. This can only be
4302 simplified when @option{-fstrict-overflow} is in effect, because
4303 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4304 zero. @option{-Wstrict-overflow} (with no level) is the same as
4305 @option{-Wstrict-overflow=2}.
4307 @item -Wstrict-overflow=3
4308 Also warn about other cases where a comparison is simplified. For
4309 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4311 @item -Wstrict-overflow=4
4312 Also warn about other simplifications not covered by the above cases.
4313 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4315 @item -Wstrict-overflow=5
4316 Also warn about cases where the compiler reduces the magnitude of a
4317 constant involved in a comparison. For example: @code{x + 2 > y} is
4318 simplified to @code{x + 1 >= y}. This is reported only at the
4319 highest warning level because this simplification applies to many
4320 comparisons, so this warning level gives a very large number of
4324 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4325 @opindex Wsuggest-attribute=
4326 @opindex Wno-suggest-attribute=
4327 Warn for cases where adding an attribute may be beneficial. The
4328 attributes currently supported are listed below.
4331 @item -Wsuggest-attribute=pure
4332 @itemx -Wsuggest-attribute=const
4333 @itemx -Wsuggest-attribute=noreturn
4334 @opindex Wsuggest-attribute=pure
4335 @opindex Wno-suggest-attribute=pure
4336 @opindex Wsuggest-attribute=const
4337 @opindex Wno-suggest-attribute=const
4338 @opindex Wsuggest-attribute=noreturn
4339 @opindex Wno-suggest-attribute=noreturn
4341 Warn about functions that might be candidates for attributes
4342 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4343 functions visible in other compilation units or (in the case of @code{pure} and
4344 @code{const}) if it cannot prove that the function returns normally. A function
4345 returns normally if it doesn't contain an infinite loop or return abnormally
4346 by throwing, calling @code{abort} or trapping. This analysis requires option
4347 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4348 higher. Higher optimization levels improve the accuracy of the analysis.
4350 @item -Wsuggest-attribute=format
4351 @itemx -Wmissing-format-attribute
4352 @opindex Wsuggest-attribute=format
4353 @opindex Wmissing-format-attribute
4354 @opindex Wno-suggest-attribute=format
4355 @opindex Wno-missing-format-attribute
4359 Warn about function pointers that might be candidates for @code{format}
4360 attributes. Note these are only possible candidates, not absolute ones.
4361 GCC guesses that function pointers with @code{format} attributes that
4362 are used in assignment, initialization, parameter passing or return
4363 statements should have a corresponding @code{format} attribute in the
4364 resulting type. I.e.@: the left-hand side of the assignment or
4365 initialization, the type of the parameter variable, or the return type
4366 of the containing function respectively should also have a @code{format}
4367 attribute to avoid the warning.
4369 GCC also warns about function definitions that might be
4370 candidates for @code{format} attributes. Again, these are only
4371 possible candidates. GCC guesses that @code{format} attributes
4372 might be appropriate for any function that calls a function like
4373 @code{vprintf} or @code{vscanf}, but this might not always be the
4374 case, and some functions for which @code{format} attributes are
4375 appropriate may not be detected.
4378 @item -Wsuggest-final-types
4379 @opindex Wno-suggest-final-types
4380 @opindex Wsuggest-final-types
4381 Warn about types with virtual methods where code quality would be improved
4382 if the type were declared with the C++11 @code{final} specifier,
4384 declared in an anonymous namespace. This allows GCC to more aggressively
4385 devirtualize the polymorphic calls. This warning is more effective with link
4386 time optimization, where the information about the class hierarchy graph is
4389 @item -Wsuggest-final-methods
4390 @opindex Wno-suggest-final-methods
4391 @opindex Wsuggest-final-methods
4392 Warn about virtual methods where code quality would be improved if the method
4393 were declared with the C++11 @code{final} specifier,
4394 or, if possible, its type were
4395 declared in an anonymous namespace or with the @code{final} specifier.
4397 more effective with link time optimization, where the information about the
4398 class hierarchy graph is more complete. It is recommended to first consider
4399 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4402 @item -Wsuggest-override
4403 Warn about overriding virtual functions that are not marked with the override
4406 @item -Warray-bounds
4407 @itemx -Warray-bounds=@var{n}
4408 @opindex Wno-array-bounds
4409 @opindex Warray-bounds
4410 This option is only active when @option{-ftree-vrp} is active
4411 (default for @option{-O2} and above). It warns about subscripts to arrays
4412 that are always out of bounds. This warning is enabled by @option{-Wall}.
4415 @item -Warray-bounds=1
4416 This is the warning level of @option{-Warray-bounds} and is enabled
4417 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4419 @item -Warray-bounds=2
4420 This warning level also warns about out of bounds access for
4421 arrays at the end of a struct and for arrays accessed through
4422 pointers. This warning level may give a larger number of
4423 false positives and is deactivated by default.
4426 @item -Wbool-compare
4427 @opindex Wno-bool-compare
4428 @opindex Wbool-compare
4429 Warn about boolean expression compared with an integer value different from
4430 @code{true}/@code{false}. For instance, the following comparison is
4435 if ((n > 1) == 2) @{ @dots{} @}
4437 This warning is enabled by @option{-Wall}.
4439 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4440 @opindex Wno-discarded-qualifiers
4441 @opindex Wdiscarded-qualifiers
4442 Do not warn if type qualifiers on pointers are being discarded.
4443 Typically, the compiler warns if a @code{const char *} variable is
4444 passed to a function that takes a @code{char *} parameter. This option
4445 can be used to suppress such a warning.
4447 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4448 @opindex Wno-discarded-array-qualifiers
4449 @opindex Wdiscarded-array-qualifiers
4450 Do not warn if type qualifiers on arrays which are pointer targets
4451 are being discarded. Typically, the compiler warns if a
4452 @code{const int (*)[]} variable is passed to a function that
4453 takes a @code{int (*)[]} parameter. This option can be used to
4454 suppress such a warning.
4456 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4457 @opindex Wno-incompatible-pointer-types
4458 @opindex Wincompatible-pointer-types
4459 Do not warn when there is a conversion between pointers that have incompatible
4460 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4461 which warns for pointer argument passing or assignment with different
4464 @item -Wno-int-conversion @r{(C and Objective-C only)}
4465 @opindex Wno-int-conversion
4466 @opindex Wint-conversion
4467 Do not warn about incompatible integer to pointer and pointer to integer
4468 conversions. This warning is about implicit conversions; for explicit
4469 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4470 @option{-Wno-pointer-to-int-cast} may be used.
4472 @item -Wno-div-by-zero
4473 @opindex Wno-div-by-zero
4474 @opindex Wdiv-by-zero
4475 Do not warn about compile-time integer division by zero. Floating-point
4476 division by zero is not warned about, as it can be a legitimate way of
4477 obtaining infinities and NaNs.
4479 @item -Wsystem-headers
4480 @opindex Wsystem-headers
4481 @opindex Wno-system-headers
4482 @cindex warnings from system headers
4483 @cindex system headers, warnings from
4484 Print warning messages for constructs found in system header files.
4485 Warnings from system headers are normally suppressed, on the assumption
4486 that they usually do not indicate real problems and would only make the
4487 compiler output harder to read. Using this command-line option tells
4488 GCC to emit warnings from system headers as if they occurred in user
4489 code. However, note that using @option{-Wall} in conjunction with this
4490 option does @emph{not} warn about unknown pragmas in system
4491 headers---for that, @option{-Wunknown-pragmas} must also be used.
4494 @opindex Wtrampolines
4495 @opindex Wno-trampolines
4496 Warn about trampolines generated for pointers to nested functions.
4497 A trampoline is a small piece of data or code that is created at run
4498 time on the stack when the address of a nested function is taken, and is
4499 used to call the nested function indirectly. For some targets, it is
4500 made up of data only and thus requires no special treatment. But, for
4501 most targets, it is made up of code and thus requires the stack to be
4502 made executable in order for the program to work properly.
4505 @opindex Wfloat-equal
4506 @opindex Wno-float-equal
4507 Warn if floating-point values are used in equality comparisons.
4509 The idea behind this is that sometimes it is convenient (for the
4510 programmer) to consider floating-point values as approximations to
4511 infinitely precise real numbers. If you are doing this, then you need
4512 to compute (by analyzing the code, or in some other way) the maximum or
4513 likely maximum error that the computation introduces, and allow for it
4514 when performing comparisons (and when producing output, but that's a
4515 different problem). In particular, instead of testing for equality, you
4516 should check to see whether the two values have ranges that overlap; and
4517 this is done with the relational operators, so equality comparisons are
4520 @item -Wtraditional @r{(C and Objective-C only)}
4521 @opindex Wtraditional
4522 @opindex Wno-traditional
4523 Warn about certain constructs that behave differently in traditional and
4524 ISO C@. Also warn about ISO C constructs that have no traditional C
4525 equivalent, and/or problematic constructs that should be avoided.
4529 Macro parameters that appear within string literals in the macro body.
4530 In traditional C macro replacement takes place within string literals,
4531 but in ISO C it does not.
4534 In traditional C, some preprocessor directives did not exist.
4535 Traditional preprocessors only considered a line to be a directive
4536 if the @samp{#} appeared in column 1 on the line. Therefore
4537 @option{-Wtraditional} warns about directives that traditional C
4538 understands but ignores because the @samp{#} does not appear as the
4539 first character on the line. It also suggests you hide directives like
4540 @code{#pragma} not understood by traditional C by indenting them. Some
4541 traditional implementations do not recognize @code{#elif}, so this option
4542 suggests avoiding it altogether.
4545 A function-like macro that appears without arguments.
4548 The unary plus operator.
4551 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4552 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4553 constants.) Note, these suffixes appear in macros defined in the system
4554 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4555 Use of these macros in user code might normally lead to spurious
4556 warnings, however GCC's integrated preprocessor has enough context to
4557 avoid warning in these cases.
4560 A function declared external in one block and then used after the end of
4564 A @code{switch} statement has an operand of type @code{long}.
4567 A non-@code{static} function declaration follows a @code{static} one.
4568 This construct is not accepted by some traditional C compilers.
4571 The ISO type of an integer constant has a different width or
4572 signedness from its traditional type. This warning is only issued if
4573 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4574 typically represent bit patterns, are not warned about.
4577 Usage of ISO string concatenation is detected.
4580 Initialization of automatic aggregates.
4583 Identifier conflicts with labels. Traditional C lacks a separate
4584 namespace for labels.
4587 Initialization of unions. If the initializer is zero, the warning is
4588 omitted. This is done under the assumption that the zero initializer in
4589 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4590 initializer warnings and relies on default initialization to zero in the
4594 Conversions by prototypes between fixed/floating-point values and vice
4595 versa. The absence of these prototypes when compiling with traditional
4596 C causes serious problems. This is a subset of the possible
4597 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4600 Use of ISO C style function definitions. This warning intentionally is
4601 @emph{not} issued for prototype declarations or variadic functions
4602 because these ISO C features appear in your code when using
4603 libiberty's traditional C compatibility macros, @code{PARAMS} and
4604 @code{VPARAMS}. This warning is also bypassed for nested functions
4605 because that feature is already a GCC extension and thus not relevant to
4606 traditional C compatibility.
4609 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4610 @opindex Wtraditional-conversion
4611 @opindex Wno-traditional-conversion
4612 Warn if a prototype causes a type conversion that is different from what
4613 would happen to the same argument in the absence of a prototype. This
4614 includes conversions of fixed point to floating and vice versa, and
4615 conversions changing the width or signedness of a fixed-point argument
4616 except when the same as the default promotion.
4618 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4619 @opindex Wdeclaration-after-statement
4620 @opindex Wno-declaration-after-statement
4621 Warn when a declaration is found after a statement in a block. This
4622 construct, known from C++, was introduced with ISO C99 and is by default
4623 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4628 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4630 @item -Wno-endif-labels
4631 @opindex Wno-endif-labels
4632 @opindex Wendif-labels
4633 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4638 Warn whenever a local variable or type declaration shadows another
4639 variable, parameter, type, class member (in C++), or instance variable
4640 (in Objective-C) or whenever a built-in function is shadowed. Note
4641 that in C++, the compiler warns if a local variable shadows an
4642 explicit typedef, but not if it shadows a struct/class/enum.
4644 @item -Wno-shadow-ivar @r{(Objective-C only)}
4645 @opindex Wno-shadow-ivar
4646 @opindex Wshadow-ivar
4647 Do not warn whenever a local variable shadows an instance variable in an
4650 @item -Wlarger-than=@var{len}
4651 @opindex Wlarger-than=@var{len}
4652 @opindex Wlarger-than-@var{len}
4653 Warn whenever an object of larger than @var{len} bytes is defined.
4655 @item -Wframe-larger-than=@var{len}
4656 @opindex Wframe-larger-than
4657 Warn if the size of a function frame is larger than @var{len} bytes.
4658 The computation done to determine the stack frame size is approximate
4659 and not conservative.
4660 The actual requirements may be somewhat greater than @var{len}
4661 even if you do not get a warning. In addition, any space allocated
4662 via @code{alloca}, variable-length arrays, or related constructs
4663 is not included by the compiler when determining
4664 whether or not to issue a warning.
4666 @item -Wno-free-nonheap-object
4667 @opindex Wno-free-nonheap-object
4668 @opindex Wfree-nonheap-object
4669 Do not warn when attempting to free an object that was not allocated
4672 @item -Wstack-usage=@var{len}
4673 @opindex Wstack-usage
4674 Warn if the stack usage of a function might be larger than @var{len} bytes.
4675 The computation done to determine the stack usage is conservative.
4676 Any space allocated via @code{alloca}, variable-length arrays, or related
4677 constructs is included by the compiler when determining whether or not to
4680 The message is in keeping with the output of @option{-fstack-usage}.
4684 If the stack usage is fully static but exceeds the specified amount, it's:
4687 warning: stack usage is 1120 bytes
4690 If the stack usage is (partly) dynamic but bounded, it's:
4693 warning: stack usage might be 1648 bytes
4696 If the stack usage is (partly) dynamic and not bounded, it's:
4699 warning: stack usage might be unbounded
4703 @item -Wunsafe-loop-optimizations
4704 @opindex Wunsafe-loop-optimizations
4705 @opindex Wno-unsafe-loop-optimizations
4706 Warn if the loop cannot be optimized because the compiler cannot
4707 assume anything on the bounds of the loop indices. With
4708 @option{-funsafe-loop-optimizations} warn if the compiler makes
4711 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4712 @opindex Wno-pedantic-ms-format
4713 @opindex Wpedantic-ms-format
4714 When used in combination with @option{-Wformat}
4715 and @option{-pedantic} without GNU extensions, this option
4716 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4717 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4718 which depend on the MS runtime.
4720 @item -Wpointer-arith
4721 @opindex Wpointer-arith
4722 @opindex Wno-pointer-arith
4723 Warn about anything that depends on the ``size of'' a function type or
4724 of @code{void}. GNU C assigns these types a size of 1, for
4725 convenience in calculations with @code{void *} pointers and pointers
4726 to functions. In C++, warn also when an arithmetic operation involves
4727 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4730 @opindex Wtype-limits
4731 @opindex Wno-type-limits
4732 Warn if a comparison is always true or always false due to the limited
4733 range of the data type, but do not warn for constant expressions. For
4734 example, warn if an unsigned variable is compared against zero with
4735 @code{<} or @code{>=}. This warning is also enabled by
4738 @item -Wbad-function-cast @r{(C and Objective-C only)}
4739 @opindex Wbad-function-cast
4740 @opindex Wno-bad-function-cast
4741 Warn when a function call is cast to a non-matching type.
4742 For example, warn if a call to a function returning an integer type
4743 is cast to a pointer type.
4745 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4746 @opindex Wc90-c99-compat
4747 @opindex Wno-c90-c99-compat
4748 Warn about features not present in ISO C90, but present in ISO C99.
4749 For instance, warn about use of variable length arrays, @code{long long}
4750 type, @code{bool} type, compound literals, designated initializers, and so
4751 on. This option is independent of the standards mode. Warnings are disabled
4752 in the expression that follows @code{__extension__}.
4754 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4755 @opindex Wc99-c11-compat
4756 @opindex Wno-c99-c11-compat
4757 Warn about features not present in ISO C99, but present in ISO C11.
4758 For instance, warn about use of anonymous structures and unions,
4759 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4760 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4761 and so on. This option is independent of the standards mode. Warnings are
4762 disabled in the expression that follows @code{__extension__}.
4764 @item -Wc++-compat @r{(C and Objective-C only)}
4765 @opindex Wc++-compat
4766 Warn about ISO C constructs that are outside of the common subset of
4767 ISO C and ISO C++, e.g.@: request for implicit conversion from
4768 @code{void *} to a pointer to non-@code{void} type.
4770 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4771 @opindex Wc++11-compat
4772 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4773 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4774 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4775 enabled by @option{-Wall}.
4777 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4778 @opindex Wc++14-compat
4779 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4780 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4784 @opindex Wno-cast-qual
4785 Warn whenever a pointer is cast so as to remove a type qualifier from
4786 the target type. For example, warn if a @code{const char *} is cast
4787 to an ordinary @code{char *}.
4789 Also warn when making a cast that introduces a type qualifier in an
4790 unsafe way. For example, casting @code{char **} to @code{const char **}
4791 is unsafe, as in this example:
4794 /* p is char ** value. */
4795 const char **q = (const char **) p;
4796 /* Assignment of readonly string to const char * is OK. */
4798 /* Now char** pointer points to read-only memory. */
4803 @opindex Wcast-align
4804 @opindex Wno-cast-align
4805 Warn whenever a pointer is cast such that the required alignment of the
4806 target is increased. For example, warn if a @code{char *} is cast to
4807 an @code{int *} on machines where integers can only be accessed at
4808 two- or four-byte boundaries.
4810 @item -Wwrite-strings
4811 @opindex Wwrite-strings
4812 @opindex Wno-write-strings
4813 When compiling C, give string constants the type @code{const
4814 char[@var{length}]} so that copying the address of one into a
4815 non-@code{const} @code{char *} pointer produces a warning. These
4816 warnings help you find at compile time code that can try to write
4817 into a string constant, but only if you have been very careful about
4818 using @code{const} in declarations and prototypes. Otherwise, it is
4819 just a nuisance. This is why we did not make @option{-Wall} request
4822 When compiling C++, warn about the deprecated conversion from string
4823 literals to @code{char *}. This warning is enabled by default for C++
4828 @opindex Wno-clobbered
4829 Warn for variables that might be changed by @code{longjmp} or
4830 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4832 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4833 @opindex Wconditionally-supported
4834 @opindex Wno-conditionally-supported
4835 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4838 @opindex Wconversion
4839 @opindex Wno-conversion
4840 Warn for implicit conversions that may alter a value. This includes
4841 conversions between real and integer, like @code{abs (x)} when
4842 @code{x} is @code{double}; conversions between signed and unsigned,
4843 like @code{unsigned ui = -1}; and conversions to smaller types, like
4844 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4845 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4846 changed by the conversion like in @code{abs (2.0)}. Warnings about
4847 conversions between signed and unsigned integers can be disabled by
4848 using @option{-Wno-sign-conversion}.
4850 For C++, also warn for confusing overload resolution for user-defined
4851 conversions; and conversions that never use a type conversion
4852 operator: conversions to @code{void}, the same type, a base class or a
4853 reference to them. Warnings about conversions between signed and
4854 unsigned integers are disabled by default in C++ unless
4855 @option{-Wsign-conversion} is explicitly enabled.
4857 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4858 @opindex Wconversion-null
4859 @opindex Wno-conversion-null
4860 Do not warn for conversions between @code{NULL} and non-pointer
4861 types. @option{-Wconversion-null} is enabled by default.
4863 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4864 @opindex Wzero-as-null-pointer-constant
4865 @opindex Wno-zero-as-null-pointer-constant
4866 Warn when a literal '0' is used as null pointer constant. This can
4867 be useful to facilitate the conversion to @code{nullptr} in C++11.
4871 @opindex Wno-date-time
4872 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4873 are encountered as they might prevent bit-wise-identical reproducible
4876 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4877 @opindex Wdelete-incomplete
4878 @opindex Wno-delete-incomplete
4879 Warn when deleting a pointer to incomplete type, which may cause
4880 undefined behavior at runtime. This warning is enabled by default.
4882 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4883 @opindex Wuseless-cast
4884 @opindex Wno-useless-cast
4885 Warn when an expression is casted to its own type.
4888 @opindex Wempty-body
4889 @opindex Wno-empty-body
4890 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
4891 while} statement. This warning is also enabled by @option{-Wextra}.
4893 @item -Wenum-compare
4894 @opindex Wenum-compare
4895 @opindex Wno-enum-compare
4896 Warn about a comparison between values of different enumerated types.
4897 In C++ enumeral mismatches in conditional expressions are also
4898 diagnosed and the warning is enabled by default. In C this warning is
4899 enabled by @option{-Wall}.
4901 @item -Wjump-misses-init @r{(C, Objective-C only)}
4902 @opindex Wjump-misses-init
4903 @opindex Wno-jump-misses-init
4904 Warn if a @code{goto} statement or a @code{switch} statement jumps
4905 forward across the initialization of a variable, or jumps backward to a
4906 label after the variable has been initialized. This only warns about
4907 variables that are initialized when they are declared. This warning is
4908 only supported for C and Objective-C; in C++ this sort of branch is an
4911 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4912 can be disabled with the @option{-Wno-jump-misses-init} option.
4914 @item -Wsign-compare
4915 @opindex Wsign-compare
4916 @opindex Wno-sign-compare
4917 @cindex warning for comparison of signed and unsigned values
4918 @cindex comparison of signed and unsigned values, warning
4919 @cindex signed and unsigned values, comparison warning
4920 Warn when a comparison between signed and unsigned values could produce
4921 an incorrect result when the signed value is converted to unsigned.
4922 This warning is also enabled by @option{-Wextra}; to get the other warnings
4923 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4925 @item -Wsign-conversion
4926 @opindex Wsign-conversion
4927 @opindex Wno-sign-conversion
4928 Warn for implicit conversions that may change the sign of an integer
4929 value, like assigning a signed integer expression to an unsigned
4930 integer variable. An explicit cast silences the warning. In C, this
4931 option is enabled also by @option{-Wconversion}.
4933 @item -Wfloat-conversion
4934 @opindex Wfloat-conversion
4935 @opindex Wno-float-conversion
4936 Warn for implicit conversions that reduce the precision of a real value.
4937 This includes conversions from real to integer, and from higher precision
4938 real to lower precision real values. This option is also enabled by
4939 @option{-Wconversion}.
4941 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
4942 @opindex Wsized-deallocation
4943 @opindex Wno-sized-deallocation
4944 Warn about a definition of an unsized deallocation function
4946 void operator delete (void *) noexcept;
4947 void operator delete[] (void *) noexcept;
4949 without a definition of the corresponding sized deallocation function
4951 void operator delete (void *, std::size_t) noexcept;
4952 void operator delete[] (void *, std::size_t) noexcept;
4954 or vice versa. Enabled by @option{-Wextra} along with
4955 @option{-fsized-deallocation}.
4957 @item -Wsizeof-pointer-memaccess
4958 @opindex Wsizeof-pointer-memaccess
4959 @opindex Wno-sizeof-pointer-memaccess
4960 Warn for suspicious length parameters to certain string and memory built-in
4961 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4962 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4963 but a pointer, and suggests a possible fix, or about
4964 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4967 @item -Wsizeof-array-argument
4968 @opindex Wsizeof-array-argument
4969 @opindex Wno-sizeof-array-argument
4970 Warn when the @code{sizeof} operator is applied to a parameter that is
4971 declared as an array in a function definition. This warning is enabled by
4972 default for C and C++ programs.
4974 @item -Wmemset-transposed-args
4975 @opindex Wmemset-transposed-args
4976 @opindex Wno-memset-transposed-args
4977 Warn for suspicious calls to the @code{memset} built-in function, if the
4978 second argument is not zero and the third argument is zero. This warns e.g.@
4979 about @code{memset (buf, sizeof buf, 0)} where most probably
4980 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4981 is only emitted if the third argument is literal zero. If it is some
4982 expression that is folded to zero, a cast of zero to some type, etc.,
4983 it is far less likely that the user has mistakenly exchanged the arguments
4984 and no warning is emitted. This warning is enabled by @option{-Wall}.
4988 @opindex Wno-address
4989 Warn about suspicious uses of memory addresses. These include using
4990 the address of a function in a conditional expression, such as
4991 @code{void func(void); if (func)}, and comparisons against the memory
4992 address of a string literal, such as @code{if (x == "abc")}. Such
4993 uses typically indicate a programmer error: the address of a function
4994 always evaluates to true, so their use in a conditional usually
4995 indicate that the programmer forgot the parentheses in a function
4996 call; and comparisons against string literals result in unspecified
4997 behavior and are not portable in C, so they usually indicate that the
4998 programmer intended to use @code{strcmp}. This warning is enabled by
5002 @opindex Wlogical-op
5003 @opindex Wno-logical-op
5004 Warn about suspicious uses of logical operators in expressions.
5005 This includes using logical operators in contexts where a
5006 bit-wise operator is likely to be expected. Also warns when
5007 the operands of a logical operator are the same:
5010 if (a < 0 && a < 0) @{ @dots{} @}
5013 @item -Wlogical-not-parentheses
5014 @opindex Wlogical-not-parentheses
5015 @opindex Wno-logical-not-parentheses
5016 Warn about logical not used on the left hand side operand of a comparison.
5017 This option does not warn if the RHS operand is of a boolean type. Its
5018 purpose is to detect suspicious code like the following:
5022 if (!a > 1) @{ @dots{} @}
5025 It is possible to suppress the warning by wrapping the LHS into
5028 if ((!a) > 1) @{ @dots{} @}
5031 This warning is enabled by @option{-Wall}.
5033 @item -Waggregate-return
5034 @opindex Waggregate-return
5035 @opindex Wno-aggregate-return
5036 Warn if any functions that return structures or unions are defined or
5037 called. (In languages where you can return an array, this also elicits
5040 @item -Wno-aggressive-loop-optimizations
5041 @opindex Wno-aggressive-loop-optimizations
5042 @opindex Waggressive-loop-optimizations
5043 Warn if in a loop with constant number of iterations the compiler detects
5044 undefined behavior in some statement during one or more of the iterations.
5046 @item -Wno-attributes
5047 @opindex Wno-attributes
5048 @opindex Wattributes
5049 Do not warn if an unexpected @code{__attribute__} is used, such as
5050 unrecognized attributes, function attributes applied to variables,
5051 etc. This does not stop errors for incorrect use of supported
5054 @item -Wno-builtin-macro-redefined
5055 @opindex Wno-builtin-macro-redefined
5056 @opindex Wbuiltin-macro-redefined
5057 Do not warn if certain built-in macros are redefined. This suppresses
5058 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5059 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5061 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5062 @opindex Wstrict-prototypes
5063 @opindex Wno-strict-prototypes
5064 Warn if a function is declared or defined without specifying the
5065 argument types. (An old-style function definition is permitted without
5066 a warning if preceded by a declaration that specifies the argument
5069 @item -Wold-style-declaration @r{(C and Objective-C only)}
5070 @opindex Wold-style-declaration
5071 @opindex Wno-old-style-declaration
5072 Warn for obsolescent usages, according to the C Standard, in a
5073 declaration. For example, warn if storage-class specifiers like
5074 @code{static} are not the first things in a declaration. This warning
5075 is also enabled by @option{-Wextra}.
5077 @item -Wold-style-definition @r{(C and Objective-C only)}
5078 @opindex Wold-style-definition
5079 @opindex Wno-old-style-definition
5080 Warn if an old-style function definition is used. A warning is given
5081 even if there is a previous prototype.
5083 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5084 @opindex Wmissing-parameter-type
5085 @opindex Wno-missing-parameter-type
5086 A function parameter is declared without a type specifier in K&R-style
5093 This warning is also enabled by @option{-Wextra}.
5095 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5096 @opindex Wmissing-prototypes
5097 @opindex Wno-missing-prototypes
5098 Warn if a global function is defined without a previous prototype
5099 declaration. This warning is issued even if the definition itself
5100 provides a prototype. Use this option to detect global functions
5101 that do not have a matching prototype declaration in a header file.
5102 This option is not valid for C++ because all function declarations
5103 provide prototypes and a non-matching declaration declares an
5104 overload rather than conflict with an earlier declaration.
5105 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5107 @item -Wmissing-declarations
5108 @opindex Wmissing-declarations
5109 @opindex Wno-missing-declarations
5110 Warn if a global function is defined without a previous declaration.
5111 Do so even if the definition itself provides a prototype.
5112 Use this option to detect global functions that are not declared in
5113 header files. In C, no warnings are issued for functions with previous
5114 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5115 missing prototypes. In C++, no warnings are issued for function templates,
5116 or for inline functions, or for functions in anonymous namespaces.
5118 @item -Wmissing-field-initializers
5119 @opindex Wmissing-field-initializers
5120 @opindex Wno-missing-field-initializers
5124 Warn if a structure's initializer has some fields missing. For
5125 example, the following code causes such a warning, because
5126 @code{x.h} is implicitly zero:
5129 struct s @{ int f, g, h; @};
5130 struct s x = @{ 3, 4 @};
5133 This option does not warn about designated initializers, so the following
5134 modification does not trigger a warning:
5137 struct s @{ int f, g, h; @};
5138 struct s x = @{ .f = 3, .g = 4 @};
5141 In C++ this option does not warn either about the empty @{ @}
5142 initializer, for example:
5145 struct s @{ int f, g, h; @};
5149 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5150 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5152 @item -Wno-multichar
5153 @opindex Wno-multichar
5155 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5156 Usually they indicate a typo in the user's code, as they have
5157 implementation-defined values, and should not be used in portable code.
5159 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5160 @opindex Wnormalized=
5161 @opindex Wnormalized
5162 @opindex Wno-normalized
5165 @cindex character set, input normalization
5166 In ISO C and ISO C++, two identifiers are different if they are
5167 different sequences of characters. However, sometimes when characters
5168 outside the basic ASCII character set are used, you can have two
5169 different character sequences that look the same. To avoid confusion,
5170 the ISO 10646 standard sets out some @dfn{normalization rules} which
5171 when applied ensure that two sequences that look the same are turned into
5172 the same sequence. GCC can warn you if you are using identifiers that
5173 have not been normalized; this option controls that warning.
5175 There are four levels of warning supported by GCC@. The default is
5176 @option{-Wnormalized=nfc}, which warns about any identifier that is
5177 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5178 recommended form for most uses. It is equivalent to
5179 @option{-Wnormalized}.
5181 Unfortunately, there are some characters allowed in identifiers by
5182 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5183 identifiers. That is, there's no way to use these symbols in portable
5184 ISO C or C++ and have all your identifiers in NFC@.
5185 @option{-Wnormalized=id} suppresses the warning for these characters.
5186 It is hoped that future versions of the standards involved will correct
5187 this, which is why this option is not the default.
5189 You can switch the warning off for all characters by writing
5190 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5191 only do this if you are using some other normalization scheme (like
5192 ``D''), because otherwise you can easily create bugs that are
5193 literally impossible to see.
5195 Some characters in ISO 10646 have distinct meanings but look identical
5196 in some fonts or display methodologies, especially once formatting has
5197 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5198 LETTER N'', displays just like a regular @code{n} that has been
5199 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5200 normalization scheme to convert all these into a standard form as
5201 well, and GCC warns if your code is not in NFKC if you use
5202 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5203 about every identifier that contains the letter O because it might be
5204 confused with the digit 0, and so is not the default, but may be
5205 useful as a local coding convention if the programming environment
5206 cannot be fixed to display these characters distinctly.
5208 @item -Wno-deprecated
5209 @opindex Wno-deprecated
5210 @opindex Wdeprecated
5211 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5213 @item -Wno-deprecated-declarations
5214 @opindex Wno-deprecated-declarations
5215 @opindex Wdeprecated-declarations
5216 Do not warn about uses of functions (@pxref{Function Attributes}),
5217 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5218 Attributes}) marked as deprecated by using the @code{deprecated}
5222 @opindex Wno-overflow
5224 Do not warn about compile-time overflow in constant expressions.
5229 Warn about One Definition Rule violations during link-time optimization.
5230 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5233 @opindex Wopenm-simd
5234 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5235 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5236 option can be used to relax the cost model.
5238 @item -Woverride-init @r{(C and Objective-C only)}
5239 @opindex Woverride-init
5240 @opindex Wno-override-init
5244 Warn if an initialized field without side effects is overridden when
5245 using designated initializers (@pxref{Designated Inits, , Designated
5248 This warning is included in @option{-Wextra}. To get other
5249 @option{-Wextra} warnings without this one, use @option{-Wextra
5250 -Wno-override-init}.
5252 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5253 @opindex Woverride-init-side-effects
5254 @opindex Wno-override-init-side-effects
5255 Warn if an initialized field with side effects is overridden when
5256 using designated initializers (@pxref{Designated Inits, , Designated
5257 Initializers}). This warning is enabled by default.
5262 Warn if a structure is given the packed attribute, but the packed
5263 attribute has no effect on the layout or size of the structure.
5264 Such structures may be mis-aligned for little benefit. For
5265 instance, in this code, the variable @code{f.x} in @code{struct bar}
5266 is misaligned even though @code{struct bar} does not itself
5267 have the packed attribute:
5274 @} __attribute__((packed));
5282 @item -Wpacked-bitfield-compat
5283 @opindex Wpacked-bitfield-compat
5284 @opindex Wno-packed-bitfield-compat
5285 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5286 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5287 the change can lead to differences in the structure layout. GCC
5288 informs you when the offset of such a field has changed in GCC 4.4.
5289 For example there is no longer a 4-bit padding between field @code{a}
5290 and @code{b} in this structure:
5297 @} __attribute__ ((packed));
5300 This warning is enabled by default. Use
5301 @option{-Wno-packed-bitfield-compat} to disable this warning.
5306 Warn if padding is included in a structure, either to align an element
5307 of the structure or to align the whole structure. Sometimes when this
5308 happens it is possible to rearrange the fields of the structure to
5309 reduce the padding and so make the structure smaller.
5311 @item -Wredundant-decls
5312 @opindex Wredundant-decls
5313 @opindex Wno-redundant-decls
5314 Warn if anything is declared more than once in the same scope, even in
5315 cases where multiple declaration is valid and changes nothing.
5317 @item -Wnested-externs @r{(C and Objective-C only)}
5318 @opindex Wnested-externs
5319 @opindex Wno-nested-externs
5320 Warn if an @code{extern} declaration is encountered within a function.
5322 @item -Wno-inherited-variadic-ctor
5323 @opindex Winherited-variadic-ctor
5324 @opindex Wno-inherited-variadic-ctor
5325 Suppress warnings about use of C++11 inheriting constructors when the
5326 base class inherited from has a C variadic constructor; the warning is
5327 on by default because the ellipsis is not inherited.
5332 Warn if a function that is declared as inline cannot be inlined.
5333 Even with this option, the compiler does not warn about failures to
5334 inline functions declared in system headers.
5336 The compiler uses a variety of heuristics to determine whether or not
5337 to inline a function. For example, the compiler takes into account
5338 the size of the function being inlined and the amount of inlining
5339 that has already been done in the current function. Therefore,
5340 seemingly insignificant changes in the source program can cause the
5341 warnings produced by @option{-Winline} to appear or disappear.
5343 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5344 @opindex Wno-invalid-offsetof
5345 @opindex Winvalid-offsetof
5346 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5347 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5348 to a non-standard-layout type is undefined. In existing C++ implementations,
5349 however, @code{offsetof} typically gives meaningful results.
5350 This flag is for users who are aware that they are
5351 writing nonportable code and who have deliberately chosen to ignore the
5354 The restrictions on @code{offsetof} may be relaxed in a future version
5355 of the C++ standard.
5357 @item -Wno-int-to-pointer-cast
5358 @opindex Wno-int-to-pointer-cast
5359 @opindex Wint-to-pointer-cast
5360 Suppress warnings from casts to pointer type of an integer of a
5361 different size. In C++, casting to a pointer type of smaller size is
5362 an error. @option{Wint-to-pointer-cast} is enabled by default.
5365 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5366 @opindex Wno-pointer-to-int-cast
5367 @opindex Wpointer-to-int-cast
5368 Suppress warnings from casts from a pointer to an integer type of a
5372 @opindex Winvalid-pch
5373 @opindex Wno-invalid-pch
5374 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5375 the search path but can't be used.
5379 @opindex Wno-long-long
5380 Warn if @code{long long} type is used. This is enabled by either
5381 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5382 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5384 @item -Wvariadic-macros
5385 @opindex Wvariadic-macros
5386 @opindex Wno-variadic-macros
5387 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5388 alternate syntax is used in ISO C99 mode. This is enabled by either
5389 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5390 messages, use @option{-Wno-variadic-macros}.
5394 @opindex Wno-varargs
5395 Warn upon questionable usage of the macros used to handle variable
5396 arguments like @code{va_start}. This is default. To inhibit the
5397 warning messages, use @option{-Wno-varargs}.
5399 @item -Wvector-operation-performance
5400 @opindex Wvector-operation-performance
5401 @opindex Wno-vector-operation-performance
5402 Warn if vector operation is not implemented via SIMD capabilities of the
5403 architecture. Mainly useful for the performance tuning.
5404 Vector operation can be implemented @code{piecewise}, which means that the
5405 scalar operation is performed on every vector element;
5406 @code{in parallel}, which means that the vector operation is implemented
5407 using scalars of wider type, which normally is more performance efficient;
5408 and @code{as a single scalar}, which means that vector fits into a
5411 @item -Wno-virtual-move-assign
5412 @opindex Wvirtual-move-assign
5413 @opindex Wno-virtual-move-assign
5414 Suppress warnings about inheriting from a virtual base with a
5415 non-trivial C++11 move assignment operator. This is dangerous because
5416 if the virtual base is reachable along more than one path, it is
5417 moved multiple times, which can mean both objects end up in the
5418 moved-from state. If the move assignment operator is written to avoid
5419 moving from a moved-from object, this warning can be disabled.
5424 Warn if variable length array is used in the code.
5425 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5426 the variable length array.
5428 @item -Wvolatile-register-var
5429 @opindex Wvolatile-register-var
5430 @opindex Wno-volatile-register-var
5431 Warn if a register variable is declared volatile. The volatile
5432 modifier does not inhibit all optimizations that may eliminate reads
5433 and/or writes to register variables. This warning is enabled by
5436 @item -Wdisabled-optimization
5437 @opindex Wdisabled-optimization
5438 @opindex Wno-disabled-optimization
5439 Warn if a requested optimization pass is disabled. This warning does
5440 not generally indicate that there is anything wrong with your code; it
5441 merely indicates that GCC's optimizers are unable to handle the code
5442 effectively. Often, the problem is that your code is too big or too
5443 complex; GCC refuses to optimize programs when the optimization
5444 itself is likely to take inordinate amounts of time.
5446 @item -Wpointer-sign @r{(C and Objective-C only)}
5447 @opindex Wpointer-sign
5448 @opindex Wno-pointer-sign
5449 Warn for pointer argument passing or assignment with different signedness.
5450 This option is only supported for C and Objective-C@. It is implied by
5451 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5452 @option{-Wno-pointer-sign}.
5454 @item -Wstack-protector
5455 @opindex Wstack-protector
5456 @opindex Wno-stack-protector
5457 This option is only active when @option{-fstack-protector} is active. It
5458 warns about functions that are not protected against stack smashing.
5460 @item -Woverlength-strings
5461 @opindex Woverlength-strings
5462 @opindex Wno-overlength-strings
5463 Warn about string constants that are longer than the ``minimum
5464 maximum'' length specified in the C standard. Modern compilers
5465 generally allow string constants that are much longer than the
5466 standard's minimum limit, but very portable programs should avoid
5467 using longer strings.
5469 The limit applies @emph{after} string constant concatenation, and does
5470 not count the trailing NUL@. In C90, the limit was 509 characters; in
5471 C99, it was raised to 4095. C++98 does not specify a normative
5472 minimum maximum, so we do not diagnose overlength strings in C++@.
5474 This option is implied by @option{-Wpedantic}, and can be disabled with
5475 @option{-Wno-overlength-strings}.
5477 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5478 @opindex Wunsuffixed-float-constants
5480 Issue a warning for any floating constant that does not have
5481 a suffix. When used together with @option{-Wsystem-headers} it
5482 warns about such constants in system header files. This can be useful
5483 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5484 from the decimal floating-point extension to C99.
5486 @item -Wno-designated-init @r{(C and Objective-C only)}
5487 Suppress warnings when a positional initializer is used to initialize
5488 a structure that has been marked with the @code{designated_init}
5493 @node Debugging Options
5494 @section Options for Debugging Your Program or GCC
5495 @cindex options, debugging
5496 @cindex debugging information options
5498 GCC has various special options that are used for debugging
5499 either your program or GCC:
5504 Produce debugging information in the operating system's native format
5505 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5508 On most systems that use stabs format, @option{-g} enables use of extra
5509 debugging information that only GDB can use; this extra information
5510 makes debugging work better in GDB but probably makes other debuggers
5512 refuse to read the program. If you want to control for certain whether
5513 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5514 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5516 GCC allows you to use @option{-g} with
5517 @option{-O}. The shortcuts taken by optimized code may occasionally
5518 produce surprising results: some variables you declared may not exist
5519 at all; flow of control may briefly move where you did not expect it;
5520 some statements may not be executed because they compute constant
5521 results or their values are already at hand; some statements may
5522 execute in different places because they have been moved out of loops.
5524 Nevertheless it proves possible to debug optimized output. This makes
5525 it reasonable to use the optimizer for programs that might have bugs.
5527 The following options are useful when GCC is generated with the
5528 capability for more than one debugging format.
5531 @opindex gsplit-dwarf
5532 Separate as much dwarf debugging information as possible into a
5533 separate output file with the extension .dwo. This option allows
5534 the build system to avoid linking files with debug information. To
5535 be useful, this option requires a debugger capable of reading .dwo
5540 Produce debugging information for use by GDB@. This means to use the
5541 most expressive format available (DWARF 2, stabs, or the native format
5542 if neither of those are supported), including GDB extensions if at all
5547 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5549 @item -ggnu-pubnames
5550 @opindex ggnu-pubnames
5551 Generate .debug_pubnames and .debug_pubtypes sections in a format
5552 suitable for conversion into a GDB@ index. This option is only useful
5553 with a linker that can produce GDB@ index version 7.
5557 Produce debugging information in stabs format (if that is supported),
5558 without GDB extensions. This is the format used by DBX on most BSD
5559 systems. On MIPS, Alpha and System V Release 4 systems this option
5560 produces stabs debugging output that is not understood by DBX or SDB@.
5561 On System V Release 4 systems this option requires the GNU assembler.
5563 @item -feliminate-unused-debug-symbols
5564 @opindex feliminate-unused-debug-symbols
5565 Produce debugging information in stabs format (if that is supported),
5566 for only symbols that are actually used.
5568 @item -femit-class-debug-always
5569 @opindex femit-class-debug-always
5570 Instead of emitting debugging information for a C++ class in only one
5571 object file, emit it in all object files using the class. This option
5572 should be used only with debuggers that are unable to handle the way GCC
5573 normally emits debugging information for classes because using this
5574 option increases the size of debugging information by as much as a
5577 @item -fdebug-types-section
5578 @opindex fdebug-types-section
5579 @opindex fno-debug-types-section
5580 When using DWARF Version 4 or higher, type DIEs can be put into
5581 their own @code{.debug_types} section instead of making them part of the
5582 @code{.debug_info} section. It is more efficient to put them in a separate
5583 comdat sections since the linker can then remove duplicates.
5584 But not all DWARF consumers support @code{.debug_types} sections yet
5585 and on some objects @code{.debug_types} produces larger instead of smaller
5586 debugging information.
5590 Produce debugging information in stabs format (if that is supported),
5591 using GNU extensions understood only by the GNU debugger (GDB)@. The
5592 use of these extensions is likely to make other debuggers crash or
5593 refuse to read the program.
5597 Produce debugging information in COFF format (if that is supported).
5598 This is the format used by SDB on most System V systems prior to
5603 Produce debugging information in XCOFF format (if that is supported).
5604 This is the format used by the DBX debugger on IBM RS/6000 systems.
5608 Produce debugging information in XCOFF format (if that is supported),
5609 using GNU extensions understood only by the GNU debugger (GDB)@. The
5610 use of these extensions is likely to make other debuggers crash or
5611 refuse to read the program, and may cause assemblers other than the GNU
5612 assembler (GAS) to fail with an error.
5614 @item -gdwarf-@var{version}
5615 @opindex gdwarf-@var{version}
5616 Produce debugging information in DWARF format (if that is supported).
5617 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5618 for most targets is 4. DWARF Version 5 is only experimental.
5620 Note that with DWARF Version 2, some ports require and always
5621 use some non-conflicting DWARF 3 extensions in the unwind tables.
5623 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5624 for maximum benefit.
5626 @item -grecord-gcc-switches
5627 @opindex grecord-gcc-switches
5628 This switch causes the command-line options used to invoke the
5629 compiler that may affect code generation to be appended to the
5630 DW_AT_producer attribute in DWARF debugging information. The options
5631 are concatenated with spaces separating them from each other and from
5632 the compiler version. See also @option{-frecord-gcc-switches} for another
5633 way of storing compiler options into the object file. This is the default.
5635 @item -gno-record-gcc-switches
5636 @opindex gno-record-gcc-switches
5637 Disallow appending command-line options to the DW_AT_producer attribute
5638 in DWARF debugging information.
5640 @item -gstrict-dwarf
5641 @opindex gstrict-dwarf
5642 Disallow using extensions of later DWARF standard version than selected
5643 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5644 DWARF extensions from later standard versions is allowed.
5646 @item -gno-strict-dwarf
5647 @opindex gno-strict-dwarf
5648 Allow using extensions of later DWARF standard version than selected with
5649 @option{-gdwarf-@var{version}}.
5651 @item -gz@r{[}=@var{type}@r{]}
5653 Produce compressed debug sections in DWARF format, if that is supported.
5654 If @var{type} is not given, the default type depends on the capabilities
5655 of the assembler and linker used. @var{type} may be one of
5656 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5657 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5658 compression in traditional GNU format). If the linker doesn't support
5659 writing compressed debug sections, the option is rejected. Otherwise,
5660 if the assembler does not support them, @option{-gz} is silently ignored
5661 when producing object files.
5665 Produce debugging information in Alpha/VMS debug format (if that is
5666 supported). This is the format used by DEBUG on Alpha/VMS systems.
5669 @itemx -ggdb@var{level}
5670 @itemx -gstabs@var{level}
5671 @itemx -gcoff@var{level}
5672 @itemx -gxcoff@var{level}
5673 @itemx -gvms@var{level}
5674 Request debugging information and also use @var{level} to specify how
5675 much information. The default level is 2.
5677 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5680 Level 1 produces minimal information, enough for making backtraces in
5681 parts of the program that you don't plan to debug. This includes
5682 descriptions of functions and external variables, and line number
5683 tables, but no information about local variables.
5685 Level 3 includes extra information, such as all the macro definitions
5686 present in the program. Some debuggers support macro expansion when
5687 you use @option{-g3}.
5689 @option{-gdwarf-2} does not accept a concatenated debug level, because
5690 GCC used to support an option @option{-gdwarf} that meant to generate
5691 debug information in version 1 of the DWARF format (which is very
5692 different from version 2), and it would have been too confusing. That
5693 debug format is long obsolete, but the option cannot be changed now.
5694 Instead use an additional @option{-g@var{level}} option to change the
5695 debug level for DWARF.
5699 Turn off generation of debug info, if leaving out this option
5700 generates it, or turn it on at level 2 otherwise. The position of this
5701 argument in the command line does not matter; it takes effect after all
5702 other options are processed, and it does so only once, no matter how
5703 many times it is given. This is mainly intended to be used with
5704 @option{-fcompare-debug}.
5706 @item -fsanitize=address
5707 @opindex fsanitize=address
5708 Enable AddressSanitizer, a fast memory error detector.
5709 Memory access instructions are instrumented to detect
5710 out-of-bounds and use-after-free bugs.
5711 See @uref{http://code.google.com/p/address-sanitizer/} for
5712 more details. The run-time behavior can be influenced using the
5713 @env{ASAN_OPTIONS} environment variable; see
5714 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5715 a list of supported options.
5717 @item -fsanitize=kernel-address
5718 @opindex fsanitize=kernel-address
5719 Enable AddressSanitizer for Linux kernel.
5720 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5722 @item -fsanitize=thread
5723 @opindex fsanitize=thread
5724 Enable ThreadSanitizer, a fast data race detector.
5725 Memory access instructions are instrumented to detect
5726 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5727 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5728 environment variable; see
5729 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5732 @item -fsanitize=leak
5733 @opindex fsanitize=leak
5734 Enable LeakSanitizer, a memory leak detector.
5735 This option only matters for linking of executables and if neither
5736 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5737 case the executable is linked against a library that overrides @code{malloc}
5738 and other allocator functions. See
5739 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5740 details. The run-time behavior can be influenced using the
5741 @env{LSAN_OPTIONS} environment variable.
5743 @item -fsanitize=undefined
5744 @opindex fsanitize=undefined
5745 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5746 Various computations are instrumented to detect undefined behavior
5747 at runtime. Current suboptions are:
5751 @item -fsanitize=shift
5752 @opindex fsanitize=shift
5753 This option enables checking that the result of a shift operation is
5754 not undefined. Note that what exactly is considered undefined differs
5755 slightly between C and C++, as well as between ISO C90 and C99, etc.
5757 @item -fsanitize=integer-divide-by-zero
5758 @opindex fsanitize=integer-divide-by-zero
5759 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5761 @item -fsanitize=unreachable
5762 @opindex fsanitize=unreachable
5763 With this option, the compiler turns the @code{__builtin_unreachable}
5764 call into a diagnostics message call instead. When reaching the
5765 @code{__builtin_unreachable} call, the behavior is undefined.
5767 @item -fsanitize=vla-bound
5768 @opindex fsanitize=vla-bound
5769 This option instructs the compiler to check that the size of a variable
5770 length array is positive.
5772 @item -fsanitize=null
5773 @opindex fsanitize=null
5774 This option enables pointer checking. Particularly, the application
5775 built with this option turned on will issue an error message when it
5776 tries to dereference a NULL pointer, or if a reference (possibly an
5777 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5778 on an object pointed by a NULL pointer.
5780 @item -fsanitize=return
5781 @opindex fsanitize=return
5782 This option enables return statement checking. Programs
5783 built with this option turned on will issue an error message
5784 when the end of a non-void function is reached without actually
5785 returning a value. This option works in C++ only.
5787 @item -fsanitize=signed-integer-overflow
5788 @opindex fsanitize=signed-integer-overflow
5789 This option enables signed integer overflow checking. We check that
5790 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5791 does not overflow in the signed arithmetics. Note, integer promotion
5792 rules must be taken into account. That is, the following is not an
5795 signed char a = SCHAR_MAX;
5799 @item -fsanitize=bounds
5800 @opindex fsanitize=bounds
5801 This option enables instrumentation of array bounds. Various out of bounds
5802 accesses are detected. Flexible array members, flexible array member-like
5803 arrays, and initializers of variables with static storage are not instrumented.
5805 @item -fsanitize=bounds-strict
5806 @opindex fsanitize=bounds-strict
5807 This option enables strict instrumentation of array bounds. Most out of bounds
5808 accesses are detected, including flexible array members and flexible array
5809 member-like arrays. Initializers of variables with static storage are not
5812 @item -fsanitize=alignment
5813 @opindex fsanitize=alignment
5815 This option enables checking of alignment of pointers when they are
5816 dereferenced, or when a reference is bound to insufficiently aligned target,
5817 or when a method or constructor is invoked on insufficiently aligned object.
5819 @item -fsanitize=object-size
5820 @opindex fsanitize=object-size
5821 This option enables instrumentation of memory references using the
5822 @code{__builtin_object_size} function. Various out of bounds pointer
5823 accesses are detected.
5825 @item -fsanitize=float-divide-by-zero
5826 @opindex fsanitize=float-divide-by-zero
5827 Detect floating-point division by zero. Unlike other similar options,
5828 @option{-fsanitize=float-divide-by-zero} is not enabled by
5829 @option{-fsanitize=undefined}, since floating-point division by zero can
5830 be a legitimate way of obtaining infinities and NaNs.
5832 @item -fsanitize=float-cast-overflow
5833 @opindex fsanitize=float-cast-overflow
5834 This option enables floating-point type to integer conversion checking.
5835 We check that the result of the conversion does not overflow.
5836 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
5837 not enabled by @option{-fsanitize=undefined}.
5838 This option does not work well with @code{FE_INVALID} exceptions enabled.
5840 @item -fsanitize=nonnull-attribute
5841 @opindex fsanitize=nonnull-attribute
5843 This option enables instrumentation of calls, checking whether null values
5844 are not passed to arguments marked as requiring a non-null value by the
5845 @code{nonnull} function attribute.
5847 @item -fsanitize=returns-nonnull-attribute
5848 @opindex fsanitize=returns-nonnull-attribute
5850 This option enables instrumentation of return statements in functions
5851 marked with @code{returns_nonnull} function attribute, to detect returning
5852 of null values from such functions.
5854 @item -fsanitize=bool
5855 @opindex fsanitize=bool
5857 This option enables instrumentation of loads from bool. If a value other
5858 than 0/1 is loaded, a run-time error is issued.
5860 @item -fsanitize=enum
5861 @opindex fsanitize=enum
5863 This option enables instrumentation of loads from an enum type. If
5864 a value outside the range of values for the enum type is loaded,
5865 a run-time error is issued.
5867 @item -fsanitize=vptr
5868 @opindex fsanitize=vptr
5870 This option enables instrumentation of C++ member function calls, member
5871 accesses and some conversions between pointers to base and derived classes,
5872 to verify the referenced object has the correct dynamic type.
5876 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5877 @option{-fsanitize=undefined} gives a diagnostic message.
5878 This currently works only for the C family of languages.
5880 @item -fno-sanitize=all
5881 @opindex fno-sanitize=all
5883 This option disables all previously enabled sanitizers.
5884 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
5887 @item -fasan-shadow-offset=@var{number}
5888 @opindex fasan-shadow-offset
5889 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5890 It is useful for experimenting with different shadow memory layouts in
5891 Kernel AddressSanitizer.
5893 @item -fsanitize-sections=@var{s1,s2,...}
5894 @opindex fsanitize-sections
5895 Sanitize global variables in selected user-defined sections. @var{si} may
5898 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5899 @opindex fsanitize-recover
5900 @opindex fno-sanitize-recover
5901 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5902 mentioned in comma-separated list of @var{opts}. Enabling this option
5903 for a sanitizer component causes it to attempt to continue
5904 running the program as if no error happened. This means multiple
5905 runtime errors can be reported in a single program run, and the exit
5906 code of the program may indicate success even when errors
5907 have been reported. The @option{-fno-sanitize-recover=} option
5908 can be used to alter
5909 this behavior: only the first detected error is reported
5910 and program then exits with a non-zero exit code.
5912 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5913 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5914 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5915 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5916 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
5917 accepted, the former enables recovery for all sanitizers that support it,
5918 the latter disables recovery for all sanitizers that support it.
5920 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5922 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5925 Similarly @option{-fno-sanitize-recover} is equivalent to
5927 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5930 @item -fsanitize-undefined-trap-on-error
5931 @opindex fsanitize-undefined-trap-on-error
5932 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
5933 report undefined behavior using @code{__builtin_trap} rather than
5934 a @code{libubsan} library routine. The advantage of this is that the
5935 @code{libubsan} library is not needed and is not linked in, so this
5936 is usable even in freestanding environments.
5938 @item -fcheck-pointer-bounds
5939 @opindex fcheck-pointer-bounds
5940 @opindex fno-check-pointer-bounds
5941 @cindex Pointer Bounds Checker options
5942 Enable Pointer Bounds Checker instrumentation. Each memory reference
5943 is instrumented with checks of the pointer used for memory access against
5944 bounds associated with that pointer.
5947 is only an implementation for Intel MPX available, thus x86 target
5948 and @option{-mmpx} are required to enable this feature.
5949 MPX-based instrumentation requires
5950 a runtime library to enable MPX in hardware and handle bounds
5951 violation signals. By default when @option{-fcheck-pointer-bounds}
5952 and @option{-mmpx} options are used to link a program, the GCC driver
5953 links against the @file{libmpx} runtime library and @file{libmpxwrappers}
5954 library. It also passes '-z bndplt' to a linker in case it supports this
5955 option (which is checked on libmpx configuration). Note that old versions
5956 of linker may ignore option. Gold linker doesn't support '-z bndplt'
5957 option. With no '-z bndplt' support in linker all calls to dynamic libraries
5958 lose passed bounds reducing overall protection level. It's highly
5959 recommended to use linker with '-z bndplt' support. In case such linker
5960 is not available it is adviced to always use @option{-static-libmpxwrappers}
5961 for better protection level or use @option{-static} to completely avoid
5962 external calls to dynamic libraries. MPX-based instrumentation
5963 may be used for debugging and also may be included in production code
5964 to increase program security. Depending on usage, you may
5965 have different requirements for the runtime library. The current version
5966 of the MPX runtime library is more oriented for use as a debugging
5967 tool. MPX runtime library usage implies @option{-lpthread}. See
5968 also @option{-static-libmpx}. The runtime library behavior can be
5969 influenced using various @env{CHKP_RT_*} environment variables. See
5970 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
5973 Generated instrumentation may be controlled by various
5974 @option{-fchkp-*} options and by the @code{bnd_variable_size}
5975 structure field attribute (@pxref{Type Attributes}) and
5976 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
5977 (@pxref{Function Attributes}). GCC also provides a number of built-in
5978 functions for controlling the Pointer Bounds Checker. @xref{Pointer
5979 Bounds Checker builtins}, for more information.
5981 @item -fchkp-check-incomplete-type
5982 @opindex fchkp-check-incomplete-type
5983 @opindex fno-chkp-check-incomplete-type
5984 Generate pointer bounds checks for variables with incomplete type.
5987 @item -fchkp-narrow-bounds
5988 @opindex fchkp-narrow-bounds
5989 @opindex fno-chkp-narrow-bounds
5990 Controls bounds used by Pointer Bounds Checker for pointers to object
5991 fields. If narrowing is enabled then field bounds are used. Otherwise
5992 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
5993 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
5995 @item -fchkp-first-field-has-own-bounds
5996 @opindex fchkp-first-field-has-own-bounds
5997 @opindex fno-chkp-first-field-has-own-bounds
5998 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
5999 first field in the structure. By default a pointer to the first field has
6000 the same bounds as a pointer to the whole structure.
6002 @item -fchkp-narrow-to-innermost-array
6003 @opindex fchkp-narrow-to-innermost-array
6004 @opindex fno-chkp-narrow-to-innermost-array
6005 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
6006 case of nested static array access. By default this option is disabled and
6007 bounds of the outermost array are used.
6009 @item -fchkp-optimize
6010 @opindex fchkp-optimize
6011 @opindex fno-chkp-optimize
6012 Enables Pointer Bounds Checker optimizations. Enabled by default at
6013 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
6015 @item -fchkp-use-fast-string-functions
6016 @opindex fchkp-use-fast-string-functions
6017 @opindex fno-chkp-use-fast-string-functions
6018 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
6019 by Pointer Bounds Checker. Disabled by default.
6021 @item -fchkp-use-nochk-string-functions
6022 @opindex fchkp-use-nochk-string-functions
6023 @opindex fno-chkp-use-nochk-string-functions
6024 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
6025 by Pointer Bounds Checker. Disabled by default.
6027 @item -fchkp-use-static-bounds
6028 @opindex fchkp-use-static-bounds
6029 @opindex fno-chkp-use-static-bounds
6030 Allow Pointer Bounds Checker to generate static bounds holding
6031 bounds of static variables. Enabled by default.
6033 @item -fchkp-use-static-const-bounds
6034 @opindex fchkp-use-static-const-bounds
6035 @opindex fno-chkp-use-static-const-bounds
6036 Use statically-initialized bounds for constant bounds instead of
6037 generating them each time they are required. By default enabled when
6038 @option{-fchkp-use-static-bounds} is enabled.
6040 @item -fchkp-treat-zero-dynamic-size-as-infinite
6041 @opindex fchkp-treat-zero-dynamic-size-as-infinite
6042 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
6043 With this option, objects with incomplete type whose
6044 dynamically-obtained size is zero are treated as having infinite size
6045 instead by Pointer Bounds
6046 Checker. This option may be helpful if a program is linked with a library
6047 missing size information for some symbols. Disabled by default.
6049 @item -fchkp-check-read
6050 @opindex fchkp-check-read
6051 @opindex fno-chkp-check-read
6052 Instructs Pointer Bounds Checker to generate checks for all read
6053 accesses to memory. Enabled by default.
6055 @item -fchkp-check-write
6056 @opindex fchkp-check-write
6057 @opindex fno-chkp-check-write
6058 Instructs Pointer Bounds Checker to generate checks for all write
6059 accesses to memory. Enabled by default.
6061 @item -fchkp-store-bounds
6062 @opindex fchkp-store-bounds
6063 @opindex fno-chkp-store-bounds
6064 Instructs Pointer Bounds Checker to generate bounds stores for
6065 pointer writes. Enabled by default.
6067 @item -fchkp-instrument-calls
6068 @opindex fchkp-instrument-calls
6069 @opindex fno-chkp-instrument-calls
6070 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
6073 @item -fchkp-instrument-marked-only
6074 @opindex fchkp-instrument-marked-only
6075 @opindex fno-chkp-instrument-marked-only
6076 Instructs Pointer Bounds Checker to instrument only functions
6077 marked with the @code{bnd_instrument} attribute
6078 (@pxref{Function Attributes}). Disabled by default.
6080 @item -fchkp-use-wrappers
6081 @opindex fchkp-use-wrappers
6082 @opindex fno-chkp-use-wrappers
6083 Allows Pointer Bounds Checker to replace calls to built-in functions
6084 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
6085 is used to link a program, the GCC driver automatically links
6086 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
6089 @item -fdump-final-insns@r{[}=@var{file}@r{]}
6090 @opindex fdump-final-insns
6091 Dump the final internal representation (RTL) to @var{file}. If the
6092 optional argument is omitted (or if @var{file} is @code{.}), the name
6093 of the dump file is determined by appending @code{.gkd} to the
6094 compilation output file name.
6096 @item -fcompare-debug@r{[}=@var{opts}@r{]}
6097 @opindex fcompare-debug
6098 @opindex fno-compare-debug
6099 If no error occurs during compilation, run the compiler a second time,
6100 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
6101 passed to the second compilation. Dump the final internal
6102 representation in both compilations, and print an error if they differ.
6104 If the equal sign is omitted, the default @option{-gtoggle} is used.
6106 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
6107 and nonzero, implicitly enables @option{-fcompare-debug}. If
6108 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
6109 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
6112 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
6113 is equivalent to @option{-fno-compare-debug}, which disables the dumping
6114 of the final representation and the second compilation, preventing even
6115 @env{GCC_COMPARE_DEBUG} from taking effect.
6117 To verify full coverage during @option{-fcompare-debug} testing, set
6118 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
6119 which GCC rejects as an invalid option in any actual compilation
6120 (rather than preprocessing, assembly or linking). To get just a
6121 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
6122 not overridden} will do.
6124 @item -fcompare-debug-second
6125 @opindex fcompare-debug-second
6126 This option is implicitly passed to the compiler for the second
6127 compilation requested by @option{-fcompare-debug}, along with options to
6128 silence warnings, and omitting other options that would cause
6129 side-effect compiler outputs to files or to the standard output. Dump
6130 files and preserved temporary files are renamed so as to contain the
6131 @code{.gk} additional extension during the second compilation, to avoid
6132 overwriting those generated by the first.
6134 When this option is passed to the compiler driver, it causes the
6135 @emph{first} compilation to be skipped, which makes it useful for little
6136 other than debugging the compiler proper.
6138 @item -feliminate-dwarf2-dups
6139 @opindex feliminate-dwarf2-dups
6140 Compress DWARF 2 debugging information by eliminating duplicated
6141 information about each symbol. This option only makes sense when
6142 generating DWARF 2 debugging information with @option{-gdwarf-2}.
6144 @item -femit-struct-debug-baseonly
6145 @opindex femit-struct-debug-baseonly
6146 Emit debug information for struct-like types
6147 only when the base name of the compilation source file
6148 matches the base name of file in which the struct is defined.
6150 This option substantially reduces the size of debugging information,
6151 but at significant potential loss in type information to the debugger.
6152 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6153 See @option{-femit-struct-debug-detailed} for more detailed control.
6155 This option works only with DWARF 2.
6157 @item -femit-struct-debug-reduced
6158 @opindex femit-struct-debug-reduced
6159 Emit debug information for struct-like types
6160 only when the base name of the compilation source file
6161 matches the base name of file in which the type is defined,
6162 unless the struct is a template or defined in a system header.
6164 This option significantly reduces the size of debugging information,
6165 with some potential loss in type information to the debugger.
6166 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6167 See @option{-femit-struct-debug-detailed} for more detailed control.
6169 This option works only with DWARF 2.
6171 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6172 @opindex femit-struct-debug-detailed
6173 Specify the struct-like types
6174 for which the compiler generates debug information.
6175 The intent is to reduce duplicate struct debug information
6176 between different object files within the same program.
6178 This option is a detailed version of
6179 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6180 which serves for most needs.
6182 A specification has the syntax@*
6183 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6185 The optional first word limits the specification to
6186 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6187 A struct type is used directly when it is the type of a variable, member.
6188 Indirect uses arise through pointers to structs.
6189 That is, when use of an incomplete struct is valid, the use is indirect.
6191 @samp{struct one direct; struct two * indirect;}.
6193 The optional second word limits the specification to
6194 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6195 Generic structs are a bit complicated to explain.
6196 For C++, these are non-explicit specializations of template classes,
6197 or non-template classes within the above.
6198 Other programming languages have generics,
6199 but @option{-femit-struct-debug-detailed} does not yet implement them.
6201 The third word specifies the source files for those
6202 structs for which the compiler should emit debug information.
6203 The values @samp{none} and @samp{any} have the normal meaning.
6204 The value @samp{base} means that
6205 the base of name of the file in which the type declaration appears
6206 must match the base of the name of the main compilation file.
6207 In practice, this means that when compiling @file{foo.c}, debug information
6208 is generated for types declared in that file and @file{foo.h},
6209 but not other header files.
6210 The value @samp{sys} means those types satisfying @samp{base}
6211 or declared in system or compiler headers.
6213 You may need to experiment to determine the best settings for your application.
6215 The default is @option{-femit-struct-debug-detailed=all}.
6217 This option works only with DWARF 2.
6219 @item -fno-merge-debug-strings
6220 @opindex fmerge-debug-strings
6221 @opindex fno-merge-debug-strings
6222 Direct the linker to not merge together strings in the debugging
6223 information that are identical in different object files. Merging is
6224 not supported by all assemblers or linkers. Merging decreases the size
6225 of the debug information in the output file at the cost of increasing
6226 link processing time. Merging is enabled by default.
6228 @item -fdebug-prefix-map=@var{old}=@var{new}
6229 @opindex fdebug-prefix-map
6230 When compiling files in directory @file{@var{old}}, record debugging
6231 information describing them as in @file{@var{new}} instead.
6233 @item -fno-dwarf2-cfi-asm
6234 @opindex fdwarf2-cfi-asm
6235 @opindex fno-dwarf2-cfi-asm
6236 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
6237 instead of using GAS @code{.cfi_*} directives.
6239 @cindex @command{prof}
6242 Generate extra code to write profile information suitable for the
6243 analysis program @command{prof}. You must use this option when compiling
6244 the source files you want data about, and you must also use it when
6247 @cindex @command{gprof}
6250 Generate extra code to write profile information suitable for the
6251 analysis program @command{gprof}. You must use this option when compiling
6252 the source files you want data about, and you must also use it when
6257 Makes the compiler print out each function name as it is compiled, and
6258 print some statistics about each pass when it finishes.
6261 @opindex ftime-report
6262 Makes the compiler print some statistics about the time consumed by each
6263 pass when it finishes.
6266 @opindex fmem-report
6267 Makes the compiler print some statistics about permanent memory
6268 allocation when it finishes.
6270 @item -fmem-report-wpa
6271 @opindex fmem-report-wpa
6272 Makes the compiler print some statistics about permanent memory
6273 allocation for the WPA phase only.
6275 @item -fpre-ipa-mem-report
6276 @opindex fpre-ipa-mem-report
6277 @item -fpost-ipa-mem-report
6278 @opindex fpost-ipa-mem-report
6279 Makes the compiler print some statistics about permanent memory
6280 allocation before or after interprocedural optimization.
6282 @item -fprofile-report
6283 @opindex fprofile-report
6284 Makes the compiler print some statistics about consistency of the
6285 (estimated) profile and effect of individual passes.
6288 @opindex fstack-usage
6289 Makes the compiler output stack usage information for the program, on a
6290 per-function basis. The filename for the dump is made by appending
6291 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6292 the output file, if explicitly specified and it is not an executable,
6293 otherwise it is the basename of the source file. An entry is made up
6298 The name of the function.
6302 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6305 The qualifier @code{static} means that the function manipulates the stack
6306 statically: a fixed number of bytes are allocated for the frame on function
6307 entry and released on function exit; no stack adjustments are otherwise made
6308 in the function. The second field is this fixed number of bytes.
6310 The qualifier @code{dynamic} means that the function manipulates the stack
6311 dynamically: in addition to the static allocation described above, stack
6312 adjustments are made in the body of the function, for example to push/pop
6313 arguments around function calls. If the qualifier @code{bounded} is also
6314 present, the amount of these adjustments is bounded at compile time and
6315 the second field is an upper bound of the total amount of stack used by
6316 the function. If it is not present, the amount of these adjustments is
6317 not bounded at compile time and the second field only represents the
6320 @item -fprofile-arcs
6321 @opindex fprofile-arcs
6322 Add code so that program flow @dfn{arcs} are instrumented. During
6323 execution the program records how many times each branch and call is
6324 executed and how many times it is taken or returns. When the compiled
6325 program exits it saves this data to a file called
6326 @file{@var{auxname}.gcda} for each source file. The data may be used for
6327 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6328 test coverage analysis (@option{-ftest-coverage}). Each object file's
6329 @var{auxname} is generated from the name of the output file, if
6330 explicitly specified and it is not the final executable, otherwise it is
6331 the basename of the source file. In both cases any suffix is removed
6332 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6333 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6334 @xref{Cross-profiling}.
6336 @cindex @command{gcov}
6340 This option is used to compile and link code instrumented for coverage
6341 analysis. The option is a synonym for @option{-fprofile-arcs}
6342 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6343 linking). See the documentation for those options for more details.
6348 Compile the source files with @option{-fprofile-arcs} plus optimization
6349 and code generation options. For test coverage analysis, use the
6350 additional @option{-ftest-coverage} option. You do not need to profile
6351 every source file in a program.
6354 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6355 (the latter implies the former).
6358 Run the program on a representative workload to generate the arc profile
6359 information. This may be repeated any number of times. You can run
6360 concurrent instances of your program, and provided that the file system
6361 supports locking, the data files will be correctly updated. Also
6362 @code{fork} calls are detected and correctly handled (double counting
6366 For profile-directed optimizations, compile the source files again with
6367 the same optimization and code generation options plus
6368 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6369 Control Optimization}).
6372 For test coverage analysis, use @command{gcov} to produce human readable
6373 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6374 @command{gcov} documentation for further information.
6378 With @option{-fprofile-arcs}, for each function of your program GCC
6379 creates a program flow graph, then finds a spanning tree for the graph.
6380 Only arcs that are not on the spanning tree have to be instrumented: the
6381 compiler adds code to count the number of times that these arcs are
6382 executed. When an arc is the only exit or only entrance to a block, the
6383 instrumentation code can be added to the block; otherwise, a new basic
6384 block must be created to hold the instrumentation code.
6387 @item -ftest-coverage
6388 @opindex ftest-coverage
6389 Produce a notes file that the @command{gcov} code-coverage utility
6390 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6391 show program coverage. Each source file's note file is called
6392 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6393 above for a description of @var{auxname} and instructions on how to
6394 generate test coverage data. Coverage data matches the source files
6395 more closely if you do not optimize.
6397 @item -fdbg-cnt-list
6398 @opindex fdbg-cnt-list
6399 Print the name and the counter upper bound for all debug counters.
6402 @item -fdbg-cnt=@var{counter-value-list}
6404 Set the internal debug counter upper bound. @var{counter-value-list}
6405 is a comma-separated list of @var{name}:@var{value} pairs
6406 which sets the upper bound of each debug counter @var{name} to @var{value}.
6407 All debug counters have the initial upper bound of @code{UINT_MAX};
6408 thus @code{dbg_cnt} returns true always unless the upper bound
6409 is set by this option.
6410 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6411 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6413 @item -fenable-@var{kind}-@var{pass}
6414 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6418 This is a set of options that are used to explicitly disable/enable
6419 optimization passes. These options are intended for use for debugging GCC.
6420 Compiler users should use regular options for enabling/disabling
6425 @item -fdisable-ipa-@var{pass}
6426 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6427 statically invoked in the compiler multiple times, the pass name should be
6428 appended with a sequential number starting from 1.
6430 @item -fdisable-rtl-@var{pass}
6431 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6432 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6433 statically invoked in the compiler multiple times, the pass name should be
6434 appended with a sequential number starting from 1. @var{range-list} is a
6435 comma-separated list of function ranges or assembler names. Each range is a number
6436 pair separated by a colon. The range is inclusive in both ends. If the range
6437 is trivial, the number pair can be simplified as a single number. If the
6438 function's call graph node's @var{uid} falls within one of the specified ranges,
6439 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6440 function header of a dump file, and the pass names can be dumped by using
6441 option @option{-fdump-passes}.
6443 @item -fdisable-tree-@var{pass}
6444 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6445 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6448 @item -fenable-ipa-@var{pass}
6449 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6450 statically invoked in the compiler multiple times, the pass name should be
6451 appended with a sequential number starting from 1.
6453 @item -fenable-rtl-@var{pass}
6454 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6455 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6456 description and examples.
6458 @item -fenable-tree-@var{pass}
6459 @itemx -fenable-tree-@var{pass}=@var{range-list}
6460 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6461 of option arguments.
6465 Here are some examples showing uses of these options.
6469 # disable ccp1 for all functions
6471 # disable complete unroll for function whose cgraph node uid is 1
6472 -fenable-tree-cunroll=1
6473 # disable gcse2 for functions at the following ranges [1,1],
6474 # [300,400], and [400,1000]
6475 # disable gcse2 for functions foo and foo2
6476 -fdisable-rtl-gcse2=foo,foo2
6477 # disable early inlining
6478 -fdisable-tree-einline
6479 # disable ipa inlining
6480 -fdisable-ipa-inline
6481 # enable tree full unroll
6482 -fenable-tree-unroll
6486 @item -d@var{letters}
6487 @itemx -fdump-rtl-@var{pass}
6488 @itemx -fdump-rtl-@var{pass}=@var{filename}
6490 @opindex fdump-rtl-@var{pass}
6491 Says to make debugging dumps during compilation at times specified by
6492 @var{letters}. This is used for debugging the RTL-based passes of the
6493 compiler. The file names for most of the dumps are made by appending
6494 a pass number and a word to the @var{dumpname}, and the files are
6495 created in the directory of the output file. In case of
6496 @option{=@var{filename}} option, the dump is output on the given file
6497 instead of the pass numbered dump files. Note that the pass number is
6498 computed statically as passes get registered into the pass manager.
6499 Thus the numbering is not related to the dynamic order of execution of
6500 passes. In particular, a pass installed by a plugin could have a
6501 number over 200 even if it executed quite early. @var{dumpname} is
6502 generated from the name of the output file, if explicitly specified
6503 and it is not an executable, otherwise it is the basename of the
6504 source file. These switches may have different effects when
6505 @option{-E} is used for preprocessing.
6507 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6508 @option{-d} option @var{letters}. Here are the possible
6509 letters for use in @var{pass} and @var{letters}, and their meanings:
6513 @item -fdump-rtl-alignments
6514 @opindex fdump-rtl-alignments
6515 Dump after branch alignments have been computed.
6517 @item -fdump-rtl-asmcons
6518 @opindex fdump-rtl-asmcons
6519 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6521 @item -fdump-rtl-auto_inc_dec
6522 @opindex fdump-rtl-auto_inc_dec
6523 Dump after auto-inc-dec discovery. This pass is only run on
6524 architectures that have auto inc or auto dec instructions.
6526 @item -fdump-rtl-barriers
6527 @opindex fdump-rtl-barriers
6528 Dump after cleaning up the barrier instructions.
6530 @item -fdump-rtl-bbpart
6531 @opindex fdump-rtl-bbpart
6532 Dump after partitioning hot and cold basic blocks.
6534 @item -fdump-rtl-bbro
6535 @opindex fdump-rtl-bbro
6536 Dump after block reordering.
6538 @item -fdump-rtl-btl1
6539 @itemx -fdump-rtl-btl2
6540 @opindex fdump-rtl-btl2
6541 @opindex fdump-rtl-btl2
6542 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6543 after the two branch
6544 target load optimization passes.
6546 @item -fdump-rtl-bypass
6547 @opindex fdump-rtl-bypass
6548 Dump after jump bypassing and control flow optimizations.
6550 @item -fdump-rtl-combine
6551 @opindex fdump-rtl-combine
6552 Dump after the RTL instruction combination pass.
6554 @item -fdump-rtl-compgotos
6555 @opindex fdump-rtl-compgotos
6556 Dump after duplicating the computed gotos.
6558 @item -fdump-rtl-ce1
6559 @itemx -fdump-rtl-ce2
6560 @itemx -fdump-rtl-ce3
6561 @opindex fdump-rtl-ce1
6562 @opindex fdump-rtl-ce2
6563 @opindex fdump-rtl-ce3
6564 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6565 @option{-fdump-rtl-ce3} enable dumping after the three
6566 if conversion passes.
6568 @item -fdump-rtl-cprop_hardreg
6569 @opindex fdump-rtl-cprop_hardreg
6570 Dump after hard register copy propagation.
6572 @item -fdump-rtl-csa
6573 @opindex fdump-rtl-csa
6574 Dump after combining stack adjustments.
6576 @item -fdump-rtl-cse1
6577 @itemx -fdump-rtl-cse2
6578 @opindex fdump-rtl-cse1
6579 @opindex fdump-rtl-cse2
6580 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6581 the two common subexpression elimination passes.
6583 @item -fdump-rtl-dce
6584 @opindex fdump-rtl-dce
6585 Dump after the standalone dead code elimination passes.
6587 @item -fdump-rtl-dbr
6588 @opindex fdump-rtl-dbr
6589 Dump after delayed branch scheduling.
6591 @item -fdump-rtl-dce1
6592 @itemx -fdump-rtl-dce2
6593 @opindex fdump-rtl-dce1
6594 @opindex fdump-rtl-dce2
6595 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6596 the two dead store elimination passes.
6599 @opindex fdump-rtl-eh
6600 Dump after finalization of EH handling code.
6602 @item -fdump-rtl-eh_ranges
6603 @opindex fdump-rtl-eh_ranges
6604 Dump after conversion of EH handling range regions.
6606 @item -fdump-rtl-expand
6607 @opindex fdump-rtl-expand
6608 Dump after RTL generation.
6610 @item -fdump-rtl-fwprop1
6611 @itemx -fdump-rtl-fwprop2
6612 @opindex fdump-rtl-fwprop1
6613 @opindex fdump-rtl-fwprop2
6614 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6615 dumping after the two forward propagation passes.
6617 @item -fdump-rtl-gcse1
6618 @itemx -fdump-rtl-gcse2
6619 @opindex fdump-rtl-gcse1
6620 @opindex fdump-rtl-gcse2
6621 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6622 after global common subexpression elimination.
6624 @item -fdump-rtl-init-regs
6625 @opindex fdump-rtl-init-regs
6626 Dump after the initialization of the registers.
6628 @item -fdump-rtl-initvals
6629 @opindex fdump-rtl-initvals
6630 Dump after the computation of the initial value sets.
6632 @item -fdump-rtl-into_cfglayout
6633 @opindex fdump-rtl-into_cfglayout
6634 Dump after converting to cfglayout mode.
6636 @item -fdump-rtl-ira
6637 @opindex fdump-rtl-ira
6638 Dump after iterated register allocation.
6640 @item -fdump-rtl-jump
6641 @opindex fdump-rtl-jump
6642 Dump after the second jump optimization.
6644 @item -fdump-rtl-loop2
6645 @opindex fdump-rtl-loop2
6646 @option{-fdump-rtl-loop2} enables dumping after the rtl
6647 loop optimization passes.
6649 @item -fdump-rtl-mach
6650 @opindex fdump-rtl-mach
6651 Dump after performing the machine dependent reorganization pass, if that
6654 @item -fdump-rtl-mode_sw
6655 @opindex fdump-rtl-mode_sw
6656 Dump after removing redundant mode switches.
6658 @item -fdump-rtl-rnreg
6659 @opindex fdump-rtl-rnreg
6660 Dump after register renumbering.
6662 @item -fdump-rtl-outof_cfglayout
6663 @opindex fdump-rtl-outof_cfglayout
6664 Dump after converting from cfglayout mode.
6666 @item -fdump-rtl-peephole2
6667 @opindex fdump-rtl-peephole2
6668 Dump after the peephole pass.
6670 @item -fdump-rtl-postreload
6671 @opindex fdump-rtl-postreload
6672 Dump after post-reload optimizations.
6674 @item -fdump-rtl-pro_and_epilogue
6675 @opindex fdump-rtl-pro_and_epilogue
6676 Dump after generating the function prologues and epilogues.
6678 @item -fdump-rtl-sched1
6679 @itemx -fdump-rtl-sched2
6680 @opindex fdump-rtl-sched1
6681 @opindex fdump-rtl-sched2
6682 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6683 after the basic block scheduling passes.
6685 @item -fdump-rtl-ree
6686 @opindex fdump-rtl-ree
6687 Dump after sign/zero extension elimination.
6689 @item -fdump-rtl-seqabstr
6690 @opindex fdump-rtl-seqabstr
6691 Dump after common sequence discovery.
6693 @item -fdump-rtl-shorten
6694 @opindex fdump-rtl-shorten
6695 Dump after shortening branches.
6697 @item -fdump-rtl-sibling
6698 @opindex fdump-rtl-sibling
6699 Dump after sibling call optimizations.
6701 @item -fdump-rtl-split1
6702 @itemx -fdump-rtl-split2
6703 @itemx -fdump-rtl-split3
6704 @itemx -fdump-rtl-split4
6705 @itemx -fdump-rtl-split5
6706 @opindex fdump-rtl-split1
6707 @opindex fdump-rtl-split2
6708 @opindex fdump-rtl-split3
6709 @opindex fdump-rtl-split4
6710 @opindex fdump-rtl-split5
6711 These options enable dumping after five rounds of
6712 instruction splitting.
6714 @item -fdump-rtl-sms
6715 @opindex fdump-rtl-sms
6716 Dump after modulo scheduling. This pass is only run on some
6719 @item -fdump-rtl-stack
6720 @opindex fdump-rtl-stack
6721 Dump after conversion from GCC's ``flat register file'' registers to the
6722 x87's stack-like registers. This pass is only run on x86 variants.
6724 @item -fdump-rtl-subreg1
6725 @itemx -fdump-rtl-subreg2
6726 @opindex fdump-rtl-subreg1
6727 @opindex fdump-rtl-subreg2
6728 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6729 the two subreg expansion passes.
6731 @item -fdump-rtl-unshare
6732 @opindex fdump-rtl-unshare
6733 Dump after all rtl has been unshared.
6735 @item -fdump-rtl-vartrack
6736 @opindex fdump-rtl-vartrack
6737 Dump after variable tracking.
6739 @item -fdump-rtl-vregs
6740 @opindex fdump-rtl-vregs
6741 Dump after converting virtual registers to hard registers.
6743 @item -fdump-rtl-web
6744 @opindex fdump-rtl-web
6745 Dump after live range splitting.
6747 @item -fdump-rtl-regclass
6748 @itemx -fdump-rtl-subregs_of_mode_init
6749 @itemx -fdump-rtl-subregs_of_mode_finish
6750 @itemx -fdump-rtl-dfinit
6751 @itemx -fdump-rtl-dfinish
6752 @opindex fdump-rtl-regclass
6753 @opindex fdump-rtl-subregs_of_mode_init
6754 @opindex fdump-rtl-subregs_of_mode_finish
6755 @opindex fdump-rtl-dfinit
6756 @opindex fdump-rtl-dfinish
6757 These dumps are defined but always produce empty files.
6760 @itemx -fdump-rtl-all
6762 @opindex fdump-rtl-all
6763 Produce all the dumps listed above.
6767 Annotate the assembler output with miscellaneous debugging information.
6771 Dump all macro definitions, at the end of preprocessing, in addition to
6776 Produce a core dump whenever an error occurs.
6780 Annotate the assembler output with a comment indicating which
6781 pattern and alternative is used. The length of each instruction is
6786 Dump the RTL in the assembler output as a comment before each instruction.
6787 Also turns on @option{-dp} annotation.
6791 Just generate RTL for a function instead of compiling it. Usually used
6792 with @option{-fdump-rtl-expand}.
6796 @opindex fdump-noaddr
6797 When doing debugging dumps, suppress address output. This makes it more
6798 feasible to use diff on debugging dumps for compiler invocations with
6799 different compiler binaries and/or different
6800 text / bss / data / heap / stack / dso start locations.
6803 @opindex freport-bug
6804 Collect and dump debug information into temporary file if ICE in C/C++
6807 @item -fdump-unnumbered
6808 @opindex fdump-unnumbered
6809 When doing debugging dumps, suppress instruction numbers and address output.
6810 This makes it more feasible to use diff on debugging dumps for compiler
6811 invocations with different options, in particular with and without
6814 @item -fdump-unnumbered-links
6815 @opindex fdump-unnumbered-links
6816 When doing debugging dumps (see @option{-d} option above), suppress
6817 instruction numbers for the links to the previous and next instructions
6820 @item -fdump-translation-unit @r{(C++ only)}
6821 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6822 @opindex fdump-translation-unit
6823 Dump a representation of the tree structure for the entire translation
6824 unit to a file. The file name is made by appending @file{.tu} to the
6825 source file name, and the file is created in the same directory as the
6826 output file. If the @samp{-@var{options}} form is used, @var{options}
6827 controls the details of the dump as described for the
6828 @option{-fdump-tree} options.
6830 @item -fdump-class-hierarchy @r{(C++ only)}
6831 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6832 @opindex fdump-class-hierarchy
6833 Dump a representation of each class's hierarchy and virtual function
6834 table layout to a file. The file name is made by appending
6835 @file{.class} to the source file name, and the file is created in the
6836 same directory as the output file. If the @samp{-@var{options}} form
6837 is used, @var{options} controls the details of the dump as described
6838 for the @option{-fdump-tree} options.
6840 @item -fdump-ipa-@var{switch}
6842 Control the dumping at various stages of inter-procedural analysis
6843 language tree to a file. The file name is generated by appending a
6844 switch specific suffix to the source file name, and the file is created
6845 in the same directory as the output file. The following dumps are
6850 Enables all inter-procedural analysis dumps.
6853 Dumps information about call-graph optimization, unused function removal,
6854 and inlining decisions.
6857 Dump after function inlining.
6862 @opindex fdump-passes
6863 Dump the list of optimization passes that are turned on and off by
6864 the current command-line options.
6866 @item -fdump-statistics-@var{option}
6867 @opindex fdump-statistics
6868 Enable and control dumping of pass statistics in a separate file. The
6869 file name is generated by appending a suffix ending in
6870 @samp{.statistics} to the source file name, and the file is created in
6871 the same directory as the output file. If the @samp{-@var{option}}
6872 form is used, @samp{-stats} causes counters to be summed over the
6873 whole compilation unit while @samp{-details} dumps every event as
6874 the passes generate them. The default with no option is to sum
6875 counters for each function compiled.
6877 @item -fdump-tree-@var{switch}
6878 @itemx -fdump-tree-@var{switch}-@var{options}
6879 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6881 Control the dumping at various stages of processing the intermediate
6882 language tree to a file. The file name is generated by appending a
6883 switch-specific suffix to the source file name, and the file is
6884 created in the same directory as the output file. In case of
6885 @option{=@var{filename}} option, the dump is output on the given file
6886 instead of the auto named dump files. If the @samp{-@var{options}}
6887 form is used, @var{options} is a list of @samp{-} separated options
6888 which control the details of the dump. Not all options are applicable
6889 to all dumps; those that are not meaningful are ignored. The
6890 following options are available
6894 Print the address of each node. Usually this is not meaningful as it
6895 changes according to the environment and source file. Its primary use
6896 is for tying up a dump file with a debug environment.
6898 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6899 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6900 use working backward from mangled names in the assembly file.
6902 When dumping front-end intermediate representations, inhibit dumping
6903 of members of a scope or body of a function merely because that scope
6904 has been reached. Only dump such items when they are directly reachable
6907 When dumping pretty-printed trees, this option inhibits dumping the
6908 bodies of control structures.
6910 When dumping RTL, print the RTL in slim (condensed) form instead of
6911 the default LISP-like representation.
6913 Print a raw representation of the tree. By default, trees are
6914 pretty-printed into a C-like representation.
6916 Enable more detailed dumps (not honored by every dump option). Also
6917 include information from the optimization passes.
6919 Enable dumping various statistics about the pass (not honored by every dump
6922 Enable showing basic block boundaries (disabled in raw dumps).
6924 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6925 dump a representation of the control flow graph suitable for viewing with
6926 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6927 the file is pretty-printed as a subgraph, so that GraphViz can render them
6928 all in a single plot.
6930 This option currently only works for RTL dumps, and the RTL is always
6931 dumped in slim form.
6933 Enable showing virtual operands for every statement.
6935 Enable showing line numbers for statements.
6937 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6939 Enable showing the tree dump for each statement.
6941 Enable showing the EH region number holding each statement.
6943 Enable showing scalar evolution analysis details.
6945 Enable showing optimization information (only available in certain
6948 Enable showing missed optimization information (only available in certain
6951 Enable other detailed optimization information (only available in
6953 @item =@var{filename}
6954 Instead of an auto named dump file, output into the given file
6955 name. The file names @file{stdout} and @file{stderr} are treated
6956 specially and are considered already open standard streams. For
6960 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6961 -fdump-tree-pre=stderr file.c
6964 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6965 output on to @file{stderr}. If two conflicting dump filenames are
6966 given for the same pass, then the latter option overrides the earlier
6970 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6971 and @option{lineno}.
6974 Turn on all optimization options, i.e., @option{optimized},
6975 @option{missed}, and @option{note}.
6978 The following tree dumps are possible:
6982 @opindex fdump-tree-original
6983 Dump before any tree based optimization, to @file{@var{file}.original}.
6986 @opindex fdump-tree-optimized
6987 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6990 @opindex fdump-tree-gimple
6991 Dump each function before and after the gimplification pass to a file. The
6992 file name is made by appending @file{.gimple} to the source file name.
6995 @opindex fdump-tree-cfg
6996 Dump the control flow graph of each function to a file. The file name is
6997 made by appending @file{.cfg} to the source file name.
7000 @opindex fdump-tree-ch
7001 Dump each function after copying loop headers. The file name is made by
7002 appending @file{.ch} to the source file name.
7005 @opindex fdump-tree-ssa
7006 Dump SSA related information to a file. The file name is made by appending
7007 @file{.ssa} to the source file name.
7010 @opindex fdump-tree-alias
7011 Dump aliasing information for each function. The file name is made by
7012 appending @file{.alias} to the source file name.
7015 @opindex fdump-tree-ccp
7016 Dump each function after CCP@. The file name is made by appending
7017 @file{.ccp} to the source file name.
7020 @opindex fdump-tree-storeccp
7021 Dump each function after STORE-CCP@. The file name is made by appending
7022 @file{.storeccp} to the source file name.
7025 @opindex fdump-tree-pre
7026 Dump trees after partial redundancy elimination. The file name is made
7027 by appending @file{.pre} to the source file name.
7030 @opindex fdump-tree-fre
7031 Dump trees after full redundancy elimination. The file name is made
7032 by appending @file{.fre} to the source file name.
7035 @opindex fdump-tree-copyprop
7036 Dump trees after copy propagation. The file name is made
7037 by appending @file{.copyprop} to the source file name.
7039 @item store_copyprop
7040 @opindex fdump-tree-store_copyprop
7041 Dump trees after store copy-propagation. The file name is made
7042 by appending @file{.store_copyprop} to the source file name.
7045 @opindex fdump-tree-dce
7046 Dump each function after dead code elimination. The file name is made by
7047 appending @file{.dce} to the source file name.
7050 @opindex fdump-tree-sra
7051 Dump each function after performing scalar replacement of aggregates. The
7052 file name is made by appending @file{.sra} to the source file name.
7055 @opindex fdump-tree-sink
7056 Dump each function after performing code sinking. The file name is made
7057 by appending @file{.sink} to the source file name.
7060 @opindex fdump-tree-dom
7061 Dump each function after applying dominator tree optimizations. The file
7062 name is made by appending @file{.dom} to the source file name.
7065 @opindex fdump-tree-dse
7066 Dump each function after applying dead store elimination. The file
7067 name is made by appending @file{.dse} to the source file name.
7070 @opindex fdump-tree-phiopt
7071 Dump each function after optimizing PHI nodes into straightline code. The file
7072 name is made by appending @file{.phiopt} to the source file name.
7075 @opindex fdump-tree-forwprop
7076 Dump each function after forward propagating single use variables. The file
7077 name is made by appending @file{.forwprop} to the source file name.
7080 @opindex fdump-tree-copyrename
7081 Dump each function after applying the copy rename optimization. The file
7082 name is made by appending @file{.copyrename} to the source file name.
7085 @opindex fdump-tree-nrv
7086 Dump each function after applying the named return value optimization on
7087 generic trees. The file name is made by appending @file{.nrv} to the source
7091 @opindex fdump-tree-vect
7092 Dump each function after applying vectorization of loops. The file name is
7093 made by appending @file{.vect} to the source file name.
7096 @opindex fdump-tree-slp
7097 Dump each function after applying vectorization of basic blocks. The file name
7098 is made by appending @file{.slp} to the source file name.
7101 @opindex fdump-tree-vrp
7102 Dump each function after Value Range Propagation (VRP). The file name
7103 is made by appending @file{.vrp} to the source file name.
7106 @opindex fdump-tree-all
7107 Enable all the available tree dumps with the flags provided in this option.
7111 @itemx -fopt-info-@var{options}
7112 @itemx -fopt-info-@var{options}=@var{filename}
7114 Controls optimization dumps from various optimization passes. If the
7115 @samp{-@var{options}} form is used, @var{options} is a list of
7116 @samp{-} separated option keywords to select the dump details and
7119 The @var{options} can be divided into two groups: options describing the
7120 verbosity of the dump, and options describing which optimizations
7121 should be included. The options from both the groups can be freely
7122 mixed as they are non-overlapping. However, in case of any conflicts,
7123 the later options override the earlier options on the command
7126 The following options control the dump verbosity:
7130 Print information when an optimization is successfully applied. It is
7131 up to a pass to decide which information is relevant. For example, the
7132 vectorizer passes print the source location of loops which are
7133 successfully vectorized.
7135 Print information about missed optimizations. Individual passes
7136 control which information to include in the output.
7138 Print verbose information about optimizations, such as certain
7139 transformations, more detailed messages about decisions etc.
7141 Print detailed optimization information. This includes
7142 @samp{optimized}, @samp{missed}, and @samp{note}.
7145 One or more of the following option keywords can be used to describe a
7146 group of optimizations:
7150 Enable dumps from all interprocedural optimizations.
7152 Enable dumps from all loop optimizations.
7154 Enable dumps from all inlining optimizations.
7156 Enable dumps from all vectorization optimizations.
7158 Enable dumps from all optimizations. This is a superset of
7159 the optimization groups listed above.
7163 omitted, it defaults to @samp{optimized-optall}, which means to dump all
7164 info about successful optimizations from all the passes.
7166 If the @var{filename} is provided, then the dumps from all the
7167 applicable optimizations are concatenated into the @var{filename}.
7168 Otherwise the dump is output onto @file{stderr}. Though multiple
7169 @option{-fopt-info} options are accepted, only one of them can include
7170 a @var{filename}. If other filenames are provided then all but the
7171 first such option are ignored.
7173 Note that the output @var{filename} is overwritten
7174 in case of multiple translation units. If a combined output from
7175 multiple translation units is desired, @file{stderr} should be used
7178 In the following example, the optimization info is output to
7187 gcc -O3 -fopt-info-missed=missed.all
7191 outputs missed optimization report from all the passes into
7192 @file{missed.all}, and this one:
7195 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
7199 prints information about missed optimization opportunities from
7200 vectorization passes on @file{stderr}.
7201 Note that @option{-fopt-info-vec-missed} is equivalent to
7202 @option{-fopt-info-missed-vec}.
7206 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
7210 outputs information about missed optimizations as well as
7211 optimized locations from all the inlining passes into
7217 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
7221 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
7222 in conflict since only one output file is allowed. In this case, only
7223 the first option takes effect and the subsequent options are
7224 ignored. Thus only @file{vec.miss} is produced which contains
7225 dumps from the vectorizer about missed opportunities.
7227 @item -frandom-seed=@var{number}
7228 @opindex frandom-seed
7229 This option provides a seed that GCC uses in place of
7230 random numbers in generating certain symbol names
7231 that have to be different in every compiled file. It is also used to
7232 place unique stamps in coverage data files and the object files that
7233 produce them. You can use the @option{-frandom-seed} option to produce
7234 reproducibly identical object files.
7236 The @var{number} should be different for every file you compile.
7238 @item -fsched-verbose=@var{n}
7239 @opindex fsched-verbose
7240 On targets that use instruction scheduling, this option controls the
7241 amount of debugging output the scheduler prints. This information is
7242 written to standard error, unless @option{-fdump-rtl-sched1} or
7243 @option{-fdump-rtl-sched2} is specified, in which case it is output
7244 to the usual dump listing file, @file{.sched1} or @file{.sched2}
7245 respectively. However for @var{n} greater than nine, the output is
7246 always printed to standard error.
7248 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
7249 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
7250 For @var{n} greater than one, it also output basic block probabilities,
7251 detailed ready list information and unit/insn info. For @var{n} greater
7252 than two, it includes RTL at abort point, control-flow and regions info.
7253 And for @var{n} over four, @option{-fsched-verbose} also includes
7257 @itemx -save-temps=cwd
7259 Store the usual ``temporary'' intermediate files permanently; place them
7260 in the current directory and name them based on the source file. Thus,
7261 compiling @file{foo.c} with @option{-c -save-temps} produces files
7262 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7263 preprocessed @file{foo.i} output file even though the compiler now
7264 normally uses an integrated preprocessor.
7266 When used in combination with the @option{-x} command-line option,
7267 @option{-save-temps} is sensible enough to avoid over writing an
7268 input source file with the same extension as an intermediate file.
7269 The corresponding intermediate file may be obtained by renaming the
7270 source file before using @option{-save-temps}.
7272 If you invoke GCC in parallel, compiling several different source
7273 files that share a common base name in different subdirectories or the
7274 same source file compiled for multiple output destinations, it is
7275 likely that the different parallel compilers will interfere with each
7276 other, and overwrite the temporary files. For instance:
7279 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7280 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7283 may result in @file{foo.i} and @file{foo.o} being written to
7284 simultaneously by both compilers.
7286 @item -save-temps=obj
7287 @opindex save-temps=obj
7288 Store the usual ``temporary'' intermediate files permanently. If the
7289 @option{-o} option is used, the temporary files are based on the
7290 object file. If the @option{-o} option is not used, the
7291 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7296 gcc -save-temps=obj -c foo.c
7297 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7298 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7302 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7303 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7304 @file{dir2/yfoobar.o}.
7306 @item -time@r{[}=@var{file}@r{]}
7308 Report the CPU time taken by each subprocess in the compilation
7309 sequence. For C source files, this is the compiler proper and assembler
7310 (plus the linker if linking is done).
7312 Without the specification of an output file, the output looks like this:
7319 The first number on each line is the ``user time'', that is time spent
7320 executing the program itself. The second number is ``system time'',
7321 time spent executing operating system routines on behalf of the program.
7322 Both numbers are in seconds.
7324 With the specification of an output file, the output is appended to the
7325 named file, and it looks like this:
7328 0.12 0.01 cc1 @var{options}
7329 0.00 0.01 as @var{options}
7332 The ``user time'' and the ``system time'' are moved before the program
7333 name, and the options passed to the program are displayed, so that one
7334 can later tell what file was being compiled, and with which options.
7336 @item -fvar-tracking
7337 @opindex fvar-tracking
7338 Run variable tracking pass. It computes where variables are stored at each
7339 position in code. Better debugging information is then generated
7340 (if the debugging information format supports this information).
7342 It is enabled by default when compiling with optimization (@option{-Os},
7343 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7344 the debug info format supports it.
7346 @item -fvar-tracking-assignments
7347 @opindex fvar-tracking-assignments
7348 @opindex fno-var-tracking-assignments
7349 Annotate assignments to user variables early in the compilation and
7350 attempt to carry the annotations over throughout the compilation all the
7351 way to the end, in an attempt to improve debug information while
7352 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7354 It can be enabled even if var-tracking is disabled, in which case
7355 annotations are created and maintained, but discarded at the end.
7356 By default, this flag is enabled together with @option{-fvar-tracking},
7357 except when selective scheduling is enabled.
7359 @item -fvar-tracking-assignments-toggle
7360 @opindex fvar-tracking-assignments-toggle
7361 @opindex fno-var-tracking-assignments-toggle
7362 Toggle @option{-fvar-tracking-assignments}, in the same way that
7363 @option{-gtoggle} toggles @option{-g}.
7365 @item -print-file-name=@var{library}
7366 @opindex print-file-name
7367 Print the full absolute name of the library file @var{library} that
7368 would be used when linking---and don't do anything else. With this
7369 option, GCC does not compile or link anything; it just prints the
7372 @item -print-multi-directory
7373 @opindex print-multi-directory
7374 Print the directory name corresponding to the multilib selected by any
7375 other switches present in the command line. This directory is supposed
7376 to exist in @env{GCC_EXEC_PREFIX}.
7378 @item -print-multi-lib
7379 @opindex print-multi-lib
7380 Print the mapping from multilib directory names to compiler switches
7381 that enable them. The directory name is separated from the switches by
7382 @samp{;}, and each switch starts with an @samp{@@} instead of the
7383 @samp{-}, without spaces between multiple switches. This is supposed to
7384 ease shell processing.
7386 @item -print-multi-os-directory
7387 @opindex print-multi-os-directory
7388 Print the path to OS libraries for the selected
7389 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7390 present in the @file{lib} subdirectory and no multilibs are used, this is
7391 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7392 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7393 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7394 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7396 @item -print-multiarch
7397 @opindex print-multiarch
7398 Print the path to OS libraries for the selected multiarch,
7399 relative to some @file{lib} subdirectory.
7401 @item -print-prog-name=@var{program}
7402 @opindex print-prog-name
7403 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7405 @item -print-libgcc-file-name
7406 @opindex print-libgcc-file-name
7407 Same as @option{-print-file-name=libgcc.a}.
7409 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7410 but you do want to link with @file{libgcc.a}. You can do:
7413 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7416 @item -print-search-dirs
7417 @opindex print-search-dirs
7418 Print the name of the configured installation directory and a list of
7419 program and library directories @command{gcc} searches---and don't do anything else.
7421 This is useful when @command{gcc} prints the error message
7422 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7423 To resolve this you either need to put @file{cpp0} and the other compiler
7424 components where @command{gcc} expects to find them, or you can set the environment
7425 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7426 Don't forget the trailing @samp{/}.
7427 @xref{Environment Variables}.
7429 @item -print-sysroot
7430 @opindex print-sysroot
7431 Print the target sysroot directory that is used during
7432 compilation. This is the target sysroot specified either at configure
7433 time or using the @option{--sysroot} option, possibly with an extra
7434 suffix that depends on compilation options. If no target sysroot is
7435 specified, the option prints nothing.
7437 @item -print-sysroot-headers-suffix
7438 @opindex print-sysroot-headers-suffix
7439 Print the suffix added to the target sysroot when searching for
7440 headers, or give an error if the compiler is not configured with such
7441 a suffix---and don't do anything else.
7444 @opindex dumpmachine
7445 Print the compiler's target machine (for example,
7446 @samp{i686-pc-linux-gnu})---and don't do anything else.
7449 @opindex dumpversion
7450 Print the compiler version (for example, @code{3.0})---and don't do
7455 Print the compiler's built-in specs---and don't do anything else. (This
7456 is used when GCC itself is being built.) @xref{Spec Files}.
7458 @item -fno-eliminate-unused-debug-types
7459 @opindex feliminate-unused-debug-types
7460 @opindex fno-eliminate-unused-debug-types
7461 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7462 output for types that are nowhere used in the source file being compiled.
7463 Sometimes it is useful to have GCC emit debugging
7464 information for all types declared in a compilation
7465 unit, regardless of whether or not they are actually used
7466 in that compilation unit, for example
7467 if, in the debugger, you want to cast a value to a type that is
7468 not actually used in your program (but is declared). More often,
7469 however, this results in a significant amount of wasted space.
7472 @node Optimize Options
7473 @section Options That Control Optimization
7474 @cindex optimize options
7475 @cindex options, optimization
7477 These options control various sorts of optimizations.
7479 Without any optimization option, the compiler's goal is to reduce the
7480 cost of compilation and to make debugging produce the expected
7481 results. Statements are independent: if you stop the program with a
7482 breakpoint between statements, you can then assign a new value to any
7483 variable or change the program counter to any other statement in the
7484 function and get exactly the results you expect from the source
7487 Turning on optimization flags makes the compiler attempt to improve
7488 the performance and/or code size at the expense of compilation time
7489 and possibly the ability to debug the program.
7491 The compiler performs optimization based on the knowledge it has of the
7492 program. Compiling multiple files at once to a single output file mode allows
7493 the compiler to use information gained from all of the files when compiling
7496 Not all optimizations are controlled directly by a flag. Only
7497 optimizations that have a flag are listed in this section.
7499 Most optimizations are only enabled if an @option{-O} level is set on
7500 the command line. Otherwise they are disabled, even if individual
7501 optimization flags are specified.
7503 Depending on the target and how GCC was configured, a slightly different
7504 set of optimizations may be enabled at each @option{-O} level than
7505 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7506 to find out the exact set of optimizations that are enabled at each level.
7507 @xref{Overall Options}, for examples.
7514 Optimize. Optimizing compilation takes somewhat more time, and a lot
7515 more memory for a large function.
7517 With @option{-O}, the compiler tries to reduce code size and execution
7518 time, without performing any optimizations that take a great deal of
7521 @option{-O} turns on the following optimization flags:
7524 -fbranch-count-reg @gol
7525 -fcombine-stack-adjustments @gol
7527 -fcprop-registers @gol
7530 -fdelayed-branch @gol
7532 -fforward-propagate @gol
7533 -fguess-branch-probability @gol
7534 -fif-conversion2 @gol
7535 -fif-conversion @gol
7536 -finline-functions-called-once @gol
7537 -fipa-pure-const @gol
7539 -fipa-reference @gol
7540 -fmerge-constants @gol
7541 -fmove-loop-invariants @gol
7543 -fsplit-wide-types @gol
7548 -ftree-copy-prop @gol
7549 -ftree-copyrename @gol
7551 -ftree-dominator-opts @gol
7553 -ftree-forwprop @gol
7563 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7564 where doing so does not interfere with debugging.
7568 Optimize even more. GCC performs nearly all supported optimizations
7569 that do not involve a space-speed tradeoff.
7570 As compared to @option{-O}, this option increases both compilation time
7571 and the performance of the generated code.
7573 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7574 also turns on the following optimization flags:
7575 @gccoptlist{-fthread-jumps @gol
7576 -falign-functions -falign-jumps @gol
7577 -falign-loops -falign-labels @gol
7580 -fcse-follow-jumps -fcse-skip-blocks @gol
7581 -fdelete-null-pointer-checks @gol
7582 -fdevirtualize -fdevirtualize-speculatively @gol
7583 -fexpensive-optimizations @gol
7584 -fgcse -fgcse-lm @gol
7585 -fhoist-adjacent-loads @gol
7586 -finline-small-functions @gol
7587 -findirect-inlining @gol
7589 -fipa-cp-alignment @gol
7592 -fisolate-erroneous-paths-dereference @gol
7594 -foptimize-sibling-calls @gol
7595 -foptimize-strlen @gol
7596 -fpartial-inlining @gol
7598 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7599 -frerun-cse-after-loop @gol
7600 -fsched-interblock -fsched-spec @gol
7601 -fschedule-insns -fschedule-insns2 @gol
7602 -fstrict-aliasing -fstrict-overflow @gol
7603 -ftree-builtin-call-dce @gol
7604 -ftree-switch-conversion -ftree-tail-merge @gol
7609 Please note the warning under @option{-fgcse} about
7610 invoking @option{-O2} on programs that use computed gotos.
7614 Optimize yet more. @option{-O3} turns on all optimizations specified
7615 by @option{-O2} and also turns on the @option{-finline-functions},
7616 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7617 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7618 @option{-ftree-loop-distribute-patterns},
7619 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7620 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7624 Reduce compilation time and make debugging produce the expected
7625 results. This is the default.
7629 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7630 do not typically increase code size. It also performs further
7631 optimizations designed to reduce code size.
7633 @option{-Os} disables the following optimization flags:
7634 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7635 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7636 -fprefetch-loop-arrays}
7640 Disregard strict standards compliance. @option{-Ofast} enables all
7641 @option{-O3} optimizations. It also enables optimizations that are not
7642 valid for all standard-compliant programs.
7643 It turns on @option{-ffast-math} and the Fortran-specific
7644 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7648 Optimize debugging experience. @option{-Og} enables optimizations
7649 that do not interfere with debugging. It should be the optimization
7650 level of choice for the standard edit-compile-debug cycle, offering
7651 a reasonable level of optimization while maintaining fast compilation
7652 and a good debugging experience.
7654 If you use multiple @option{-O} options, with or without level numbers,
7655 the last such option is the one that is effective.
7658 Options of the form @option{-f@var{flag}} specify machine-independent
7659 flags. Most flags have both positive and negative forms; the negative
7660 form of @option{-ffoo} is @option{-fno-foo}. In the table
7661 below, only one of the forms is listed---the one you typically
7662 use. You can figure out the other form by either removing @samp{no-}
7665 The following options control specific optimizations. They are either
7666 activated by @option{-O} options or are related to ones that are. You
7667 can use the following flags in the rare cases when ``fine-tuning'' of
7668 optimizations to be performed is desired.
7671 @item -fno-defer-pop
7672 @opindex fno-defer-pop
7673 Always pop the arguments to each function call as soon as that function
7674 returns. For machines that must pop arguments after a function call,
7675 the compiler normally lets arguments accumulate on the stack for several
7676 function calls and pops them all at once.
7678 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7680 @item -fforward-propagate
7681 @opindex fforward-propagate
7682 Perform a forward propagation pass on RTL@. The pass tries to combine two
7683 instructions and checks if the result can be simplified. If loop unrolling
7684 is active, two passes are performed and the second is scheduled after
7687 This option is enabled by default at optimization levels @option{-O},
7688 @option{-O2}, @option{-O3}, @option{-Os}.
7690 @item -ffp-contract=@var{style}
7691 @opindex ffp-contract
7692 @option{-ffp-contract=off} disables floating-point expression contraction.
7693 @option{-ffp-contract=fast} enables floating-point expression contraction
7694 such as forming of fused multiply-add operations if the target has
7695 native support for them.
7696 @option{-ffp-contract=on} enables floating-point expression contraction
7697 if allowed by the language standard. This is currently not implemented
7698 and treated equal to @option{-ffp-contract=off}.
7700 The default is @option{-ffp-contract=fast}.
7702 @item -fomit-frame-pointer
7703 @opindex fomit-frame-pointer
7704 Don't keep the frame pointer in a register for functions that
7705 don't need one. This avoids the instructions to save, set up and
7706 restore frame pointers; it also makes an extra register available
7707 in many functions. @strong{It also makes debugging impossible on
7710 On some machines, such as the VAX, this flag has no effect, because
7711 the standard calling sequence automatically handles the frame pointer
7712 and nothing is saved by pretending it doesn't exist. The
7713 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7714 whether a target machine supports this flag. @xref{Registers,,Register
7715 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7717 The default setting (when not optimizing for
7718 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7719 @option{-fomit-frame-pointer}. You can configure GCC with the
7720 @option{--enable-frame-pointer} configure option to change the default.
7722 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7724 @item -foptimize-sibling-calls
7725 @opindex foptimize-sibling-calls
7726 Optimize sibling and tail recursive calls.
7728 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7730 @item -foptimize-strlen
7731 @opindex foptimize-strlen
7732 Optimize various standard C string functions (e.g. @code{strlen},
7733 @code{strchr} or @code{strcpy}) and
7734 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7736 Enabled at levels @option{-O2}, @option{-O3}.
7740 Do not expand any functions inline apart from those marked with
7741 the @code{always_inline} attribute. This is the default when not
7744 Single functions can be exempted from inlining by marking them
7745 with the @code{noinline} attribute.
7747 @item -finline-small-functions
7748 @opindex finline-small-functions
7749 Integrate functions into their callers when their body is smaller than expected
7750 function call code (so overall size of program gets smaller). The compiler
7751 heuristically decides which functions are simple enough to be worth integrating
7752 in this way. This inlining applies to all functions, even those not declared
7755 Enabled at level @option{-O2}.
7757 @item -findirect-inlining
7758 @opindex findirect-inlining
7759 Inline also indirect calls that are discovered to be known at compile
7760 time thanks to previous inlining. This option has any effect only
7761 when inlining itself is turned on by the @option{-finline-functions}
7762 or @option{-finline-small-functions} options.
7764 Enabled at level @option{-O2}.
7766 @item -finline-functions
7767 @opindex finline-functions
7768 Consider all functions for inlining, even if they are not declared inline.
7769 The compiler heuristically decides which functions are worth integrating
7772 If all calls to a given function are integrated, and the function is
7773 declared @code{static}, then the function is normally not output as
7774 assembler code in its own right.
7776 Enabled at level @option{-O3}.
7778 @item -finline-functions-called-once
7779 @opindex finline-functions-called-once
7780 Consider all @code{static} functions called once for inlining into their
7781 caller even if they are not marked @code{inline}. If a call to a given
7782 function is integrated, then the function is not output as assembler code
7785 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7787 @item -fearly-inlining
7788 @opindex fearly-inlining
7789 Inline functions marked by @code{always_inline} and functions whose body seems
7790 smaller than the function call overhead early before doing
7791 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7792 makes profiling significantly cheaper and usually inlining faster on programs
7793 having large chains of nested wrapper functions.
7799 Perform interprocedural scalar replacement of aggregates, removal of
7800 unused parameters and replacement of parameters passed by reference
7801 by parameters passed by value.
7803 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7805 @item -finline-limit=@var{n}
7806 @opindex finline-limit
7807 By default, GCC limits the size of functions that can be inlined. This flag
7808 allows coarse control of this limit. @var{n} is the size of functions that
7809 can be inlined in number of pseudo instructions.
7811 Inlining is actually controlled by a number of parameters, which may be
7812 specified individually by using @option{--param @var{name}=@var{value}}.
7813 The @option{-finline-limit=@var{n}} option sets some of these parameters
7817 @item max-inline-insns-single
7818 is set to @var{n}/2.
7819 @item max-inline-insns-auto
7820 is set to @var{n}/2.
7823 See below for a documentation of the individual
7824 parameters controlling inlining and for the defaults of these parameters.
7826 @emph{Note:} there may be no value to @option{-finline-limit} that results
7827 in default behavior.
7829 @emph{Note:} pseudo instruction represents, in this particular context, an
7830 abstract measurement of function's size. In no way does it represent a count
7831 of assembly instructions and as such its exact meaning might change from one
7832 release to an another.
7834 @item -fno-keep-inline-dllexport
7835 @opindex fno-keep-inline-dllexport
7836 This is a more fine-grained version of @option{-fkeep-inline-functions},
7837 which applies only to functions that are declared using the @code{dllexport}
7838 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7841 @item -fkeep-inline-functions
7842 @opindex fkeep-inline-functions
7843 In C, emit @code{static} functions that are declared @code{inline}
7844 into the object file, even if the function has been inlined into all
7845 of its callers. This switch does not affect functions using the
7846 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7847 inline functions into the object file.
7849 @item -fkeep-static-consts
7850 @opindex fkeep-static-consts
7851 Emit variables declared @code{static const} when optimization isn't turned
7852 on, even if the variables aren't referenced.
7854 GCC enables this option by default. If you want to force the compiler to
7855 check if a variable is referenced, regardless of whether or not
7856 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7858 @item -fmerge-constants
7859 @opindex fmerge-constants
7860 Attempt to merge identical constants (string constants and floating-point
7861 constants) across compilation units.
7863 This option is the default for optimized compilation if the assembler and
7864 linker support it. Use @option{-fno-merge-constants} to inhibit this
7867 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7869 @item -fmerge-all-constants
7870 @opindex fmerge-all-constants
7871 Attempt to merge identical constants and identical variables.
7873 This option implies @option{-fmerge-constants}. In addition to
7874 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7875 arrays or initialized constant variables with integral or floating-point
7876 types. Languages like C or C++ require each variable, including multiple
7877 instances of the same variable in recursive calls, to have distinct locations,
7878 so using this option results in non-conforming
7881 @item -fmodulo-sched
7882 @opindex fmodulo-sched
7883 Perform swing modulo scheduling immediately before the first scheduling
7884 pass. This pass looks at innermost loops and reorders their
7885 instructions by overlapping different iterations.
7887 @item -fmodulo-sched-allow-regmoves
7888 @opindex fmodulo-sched-allow-regmoves
7889 Perform more aggressive SMS-based modulo scheduling with register moves
7890 allowed. By setting this flag certain anti-dependences edges are
7891 deleted, which triggers the generation of reg-moves based on the
7892 life-range analysis. This option is effective only with
7893 @option{-fmodulo-sched} enabled.
7895 @item -fno-branch-count-reg
7896 @opindex fno-branch-count-reg
7897 Do not use ``decrement and branch'' instructions on a count register,
7898 but instead generate a sequence of instructions that decrement a
7899 register, compare it against zero, then branch based upon the result.
7900 This option is only meaningful on architectures that support such
7901 instructions, which include x86, PowerPC, IA-64 and S/390.
7903 Enabled by default at @option{-O1} and higher.
7905 The default is @option{-fbranch-count-reg}.
7907 @item -fno-function-cse
7908 @opindex fno-function-cse
7909 Do not put function addresses in registers; make each instruction that
7910 calls a constant function contain the function's address explicitly.
7912 This option results in less efficient code, but some strange hacks
7913 that alter the assembler output may be confused by the optimizations
7914 performed when this option is not used.
7916 The default is @option{-ffunction-cse}
7918 @item -fno-zero-initialized-in-bss
7919 @opindex fno-zero-initialized-in-bss
7920 If the target supports a BSS section, GCC by default puts variables that
7921 are initialized to zero into BSS@. This can save space in the resulting
7924 This option turns off this behavior because some programs explicitly
7925 rely on variables going to the data section---e.g., so that the
7926 resulting executable can find the beginning of that section and/or make
7927 assumptions based on that.
7929 The default is @option{-fzero-initialized-in-bss}.
7931 @item -fthread-jumps
7932 @opindex fthread-jumps
7933 Perform optimizations that check to see if a jump branches to a
7934 location where another comparison subsumed by the first is found. If
7935 so, the first branch is redirected to either the destination of the
7936 second branch or a point immediately following it, depending on whether
7937 the condition is known to be true or false.
7939 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7941 @item -fsplit-wide-types
7942 @opindex fsplit-wide-types
7943 When using a type that occupies multiple registers, such as @code{long
7944 long} on a 32-bit system, split the registers apart and allocate them
7945 independently. This normally generates better code for those types,
7946 but may make debugging more difficult.
7948 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7951 @item -fcse-follow-jumps
7952 @opindex fcse-follow-jumps
7953 In common subexpression elimination (CSE), scan through jump instructions
7954 when the target of the jump is not reached by any other path. For
7955 example, when CSE encounters an @code{if} statement with an
7956 @code{else} clause, CSE follows the jump when the condition
7959 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7961 @item -fcse-skip-blocks
7962 @opindex fcse-skip-blocks
7963 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7964 follow jumps that conditionally skip over blocks. When CSE
7965 encounters a simple @code{if} statement with no else clause,
7966 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7967 body of the @code{if}.
7969 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7971 @item -frerun-cse-after-loop
7972 @opindex frerun-cse-after-loop
7973 Re-run common subexpression elimination after loop optimizations are
7976 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7980 Perform a global common subexpression elimination pass.
7981 This pass also performs global constant and copy propagation.
7983 @emph{Note:} When compiling a program using computed gotos, a GCC
7984 extension, you may get better run-time performance if you disable
7985 the global common subexpression elimination pass by adding
7986 @option{-fno-gcse} to the command line.
7988 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7992 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7993 attempts to move loads that are only killed by stores into themselves. This
7994 allows a loop containing a load/store sequence to be changed to a load outside
7995 the loop, and a copy/store within the loop.
7997 Enabled by default when @option{-fgcse} is enabled.
8001 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8002 global common subexpression elimination. This pass attempts to move
8003 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8004 loops containing a load/store sequence can be changed to a load before
8005 the loop and a store after the loop.
8007 Not enabled at any optimization level.
8011 When @option{-fgcse-las} is enabled, the global common subexpression
8012 elimination pass eliminates redundant loads that come after stores to the
8013 same memory location (both partial and full redundancies).
8015 Not enabled at any optimization level.
8017 @item -fgcse-after-reload
8018 @opindex fgcse-after-reload
8019 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8020 pass is performed after reload. The purpose of this pass is to clean up
8023 @item -faggressive-loop-optimizations
8024 @opindex faggressive-loop-optimizations
8025 This option tells the loop optimizer to use language constraints to
8026 derive bounds for the number of iterations of a loop. This assumes that
8027 loop code does not invoke undefined behavior by for example causing signed
8028 integer overflows or out-of-bound array accesses. The bounds for the
8029 number of iterations of a loop are used to guide loop unrolling and peeling
8030 and loop exit test optimizations.
8031 This option is enabled by default.
8033 @item -funsafe-loop-optimizations
8034 @opindex funsafe-loop-optimizations
8035 This option tells the loop optimizer to assume that loop indices do not
8036 overflow, and that loops with nontrivial exit condition are not
8037 infinite. This enables a wider range of loop optimizations even if
8038 the loop optimizer itself cannot prove that these assumptions are valid.
8039 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
8040 if it finds this kind of loop.
8042 @item -fcrossjumping
8043 @opindex fcrossjumping
8044 Perform cross-jumping transformation.
8045 This transformation unifies equivalent code and saves code size. The
8046 resulting code may or may not perform better than without cross-jumping.
8048 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8050 @item -fauto-inc-dec
8051 @opindex fauto-inc-dec
8052 Combine increments or decrements of addresses with memory accesses.
8053 This pass is always skipped on architectures that do not have
8054 instructions to support this. Enabled by default at @option{-O} and
8055 higher on architectures that support this.
8059 Perform dead code elimination (DCE) on RTL@.
8060 Enabled by default at @option{-O} and higher.
8064 Perform dead store elimination (DSE) on RTL@.
8065 Enabled by default at @option{-O} and higher.
8067 @item -fif-conversion
8068 @opindex fif-conversion
8069 Attempt to transform conditional jumps into branch-less equivalents. This
8070 includes use of conditional moves, min, max, set flags and abs instructions, and
8071 some tricks doable by standard arithmetics. The use of conditional execution
8072 on chips where it is available is controlled by @option{-fif-conversion2}.
8074 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8076 @item -fif-conversion2
8077 @opindex fif-conversion2
8078 Use conditional execution (where available) to transform conditional jumps into
8079 branch-less equivalents.
8081 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8083 @item -fdeclone-ctor-dtor
8084 @opindex fdeclone-ctor-dtor
8085 The C++ ABI requires multiple entry points for constructors and
8086 destructors: one for a base subobject, one for a complete object, and
8087 one for a virtual destructor that calls operator delete afterwards.
8088 For a hierarchy with virtual bases, the base and complete variants are
8089 clones, which means two copies of the function. With this option, the
8090 base and complete variants are changed to be thunks that call a common
8093 Enabled by @option{-Os}.
8095 @item -fdelete-null-pointer-checks
8096 @opindex fdelete-null-pointer-checks
8097 Assume that programs cannot safely dereference null pointers, and that
8098 no code or data element resides at address zero.
8099 This option enables simple constant
8100 folding optimizations at all optimization levels. In addition, other
8101 optimization passes in GCC use this flag to control global dataflow
8102 analyses that eliminate useless checks for null pointers; these assume
8103 that a memory access to address zero always results in a trap, so
8104 that if a pointer is checked after it has already been dereferenced,
8107 Note however that in some environments this assumption is not true.
8108 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8109 for programs that depend on that behavior.
8111 This option is enabled by default on most targets. On Nios II ELF, it
8112 defaults to off. On AVR and CR16, this option is completely disabled.
8114 Passes that use the dataflow information
8115 are enabled independently at different optimization levels.
8117 @item -fdevirtualize
8118 @opindex fdevirtualize
8119 Attempt to convert calls to virtual functions to direct calls. This
8120 is done both within a procedure and interprocedurally as part of
8121 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8122 propagation (@option{-fipa-cp}).
8123 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8125 @item -fdevirtualize-speculatively
8126 @opindex fdevirtualize-speculatively
8127 Attempt to convert calls to virtual functions to speculative direct calls.
8128 Based on the analysis of the type inheritance graph, determine for a given call
8129 the set of likely targets. If the set is small, preferably of size 1, change
8130 the call into a conditional deciding between direct and indirect calls. The
8131 speculative calls enable more optimizations, such as inlining. When they seem
8132 useless after further optimization, they are converted back into original form.
8134 @item -fdevirtualize-at-ltrans
8135 @opindex fdevirtualize-at-ltrans
8136 Stream extra information needed for aggressive devirtualization when running
8137 the link-time optimizer in local transformation mode.
8138 This option enables more devirtualization but
8139 significantly increases the size of streamed data. For this reason it is
8140 disabled by default.
8142 @item -fexpensive-optimizations
8143 @opindex fexpensive-optimizations
8144 Perform a number of minor optimizations that are relatively expensive.
8146 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8150 Attempt to remove redundant extension instructions. This is especially
8151 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8152 registers after writing to their lower 32-bit half.
8154 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8155 @option{-O3}, @option{-Os}.
8157 @item -fno-lifetime-dse
8158 @opindex fno-lifetime-dse
8159 In C++ the value of an object is only affected by changes within its
8160 lifetime: when the constructor begins, the object has an indeterminate
8161 value, and any changes during the lifetime of the object are dead when
8162 the object is destroyed. Normally dead store elimination will take
8163 advantage of this; if your code relies on the value of the object
8164 storage persisting beyond the lifetime of the object, you can use this
8165 flag to disable this optimization.
8167 @item -flive-range-shrinkage
8168 @opindex flive-range-shrinkage
8169 Attempt to decrease register pressure through register live range
8170 shrinkage. This is helpful for fast processors with small or moderate
8173 @item -fira-algorithm=@var{algorithm}
8174 @opindex fira-algorithm
8175 Use the specified coloring algorithm for the integrated register
8176 allocator. The @var{algorithm} argument can be @samp{priority}, which
8177 specifies Chow's priority coloring, or @samp{CB}, which specifies
8178 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8179 for all architectures, but for those targets that do support it, it is
8180 the default because it generates better code.
8182 @item -fira-region=@var{region}
8183 @opindex fira-region
8184 Use specified regions for the integrated register allocator. The
8185 @var{region} argument should be one of the following:
8190 Use all loops as register allocation regions.
8191 This can give the best results for machines with a small and/or
8192 irregular register set.
8195 Use all loops except for loops with small register pressure
8196 as the regions. This value usually gives
8197 the best results in most cases and for most architectures,
8198 and is enabled by default when compiling with optimization for speed
8199 (@option{-O}, @option{-O2}, @dots{}).
8202 Use all functions as a single region.
8203 This typically results in the smallest code size, and is enabled by default for
8204 @option{-Os} or @option{-O0}.
8208 @item -fira-hoist-pressure
8209 @opindex fira-hoist-pressure
8210 Use IRA to evaluate register pressure in the code hoisting pass for
8211 decisions to hoist expressions. This option usually results in smaller
8212 code, but it can slow the compiler down.
8214 This option is enabled at level @option{-Os} for all targets.
8216 @item -fira-loop-pressure
8217 @opindex fira-loop-pressure
8218 Use IRA to evaluate register pressure in loops for decisions to move
8219 loop invariants. This option usually results in generation
8220 of faster and smaller code on machines with large register files (>= 32
8221 registers), but it can slow the compiler down.
8223 This option is enabled at level @option{-O3} for some targets.
8225 @item -fno-ira-share-save-slots
8226 @opindex fno-ira-share-save-slots
8227 Disable sharing of stack slots used for saving call-used hard
8228 registers living through a call. Each hard register gets a
8229 separate stack slot, and as a result function stack frames are
8232 @item -fno-ira-share-spill-slots
8233 @opindex fno-ira-share-spill-slots
8234 Disable sharing of stack slots allocated for pseudo-registers. Each
8235 pseudo-register that does not get a hard register gets a separate
8236 stack slot, and as a result function stack frames are larger.
8238 @item -fira-verbose=@var{n}
8239 @opindex fira-verbose
8240 Control the verbosity of the dump file for the integrated register allocator.
8241 The default value is 5. If the value @var{n} is greater or equal to 10,
8242 the dump output is sent to stderr using the same format as @var{n} minus 10.
8246 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8247 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8248 values if it is profitable.
8250 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8252 @item -fdelayed-branch
8253 @opindex fdelayed-branch
8254 If supported for the target machine, attempt to reorder instructions
8255 to exploit instruction slots available after delayed branch
8258 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8260 @item -fschedule-insns
8261 @opindex fschedule-insns
8262 If supported for the target machine, attempt to reorder instructions to
8263 eliminate execution stalls due to required data being unavailable. This
8264 helps machines that have slow floating point or memory load instructions
8265 by allowing other instructions to be issued until the result of the load
8266 or floating-point instruction is required.
8268 Enabled at levels @option{-O2}, @option{-O3}.
8270 @item -fschedule-insns2
8271 @opindex fschedule-insns2
8272 Similar to @option{-fschedule-insns}, but requests an additional pass of
8273 instruction scheduling after register allocation has been done. This is
8274 especially useful on machines with a relatively small number of
8275 registers and where memory load instructions take more than one cycle.
8277 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8279 @item -fno-sched-interblock
8280 @opindex fno-sched-interblock
8281 Don't schedule instructions across basic blocks. This is normally
8282 enabled by default when scheduling before register allocation, i.e.@:
8283 with @option{-fschedule-insns} or at @option{-O2} or higher.
8285 @item -fno-sched-spec
8286 @opindex fno-sched-spec
8287 Don't allow speculative motion of non-load instructions. This is normally
8288 enabled by default when scheduling before register allocation, i.e.@:
8289 with @option{-fschedule-insns} or at @option{-O2} or higher.
8291 @item -fsched-pressure
8292 @opindex fsched-pressure
8293 Enable register pressure sensitive insn scheduling before register
8294 allocation. This only makes sense when scheduling before register
8295 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8296 @option{-O2} or higher. Usage of this option can improve the
8297 generated code and decrease its size by preventing register pressure
8298 increase above the number of available hard registers and subsequent
8299 spills in register allocation.
8301 @item -fsched-spec-load
8302 @opindex fsched-spec-load
8303 Allow speculative motion of some load instructions. This only makes
8304 sense when scheduling before register allocation, i.e.@: with
8305 @option{-fschedule-insns} or at @option{-O2} or higher.
8307 @item -fsched-spec-load-dangerous
8308 @opindex fsched-spec-load-dangerous
8309 Allow speculative motion of more load instructions. This only makes
8310 sense when scheduling before register allocation, i.e.@: with
8311 @option{-fschedule-insns} or at @option{-O2} or higher.
8313 @item -fsched-stalled-insns
8314 @itemx -fsched-stalled-insns=@var{n}
8315 @opindex fsched-stalled-insns
8316 Define how many insns (if any) can be moved prematurely from the queue
8317 of stalled insns into the ready list during the second scheduling pass.
8318 @option{-fno-sched-stalled-insns} means that no insns are moved
8319 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8320 on how many queued insns can be moved prematurely.
8321 @option{-fsched-stalled-insns} without a value is equivalent to
8322 @option{-fsched-stalled-insns=1}.
8324 @item -fsched-stalled-insns-dep
8325 @itemx -fsched-stalled-insns-dep=@var{n}
8326 @opindex fsched-stalled-insns-dep
8327 Define how many insn groups (cycles) are examined for a dependency
8328 on a stalled insn that is a candidate for premature removal from the queue
8329 of stalled insns. This has an effect only during the second scheduling pass,
8330 and only if @option{-fsched-stalled-insns} is used.
8331 @option{-fno-sched-stalled-insns-dep} is equivalent to
8332 @option{-fsched-stalled-insns-dep=0}.
8333 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8334 @option{-fsched-stalled-insns-dep=1}.
8336 @item -fsched2-use-superblocks
8337 @opindex fsched2-use-superblocks
8338 When scheduling after register allocation, use superblock scheduling.
8339 This allows motion across basic block boundaries,
8340 resulting in faster schedules. This option is experimental, as not all machine
8341 descriptions used by GCC model the CPU closely enough to avoid unreliable
8342 results from the algorithm.
8344 This only makes sense when scheduling after register allocation, i.e.@: with
8345 @option{-fschedule-insns2} or at @option{-O2} or higher.
8347 @item -fsched-group-heuristic
8348 @opindex fsched-group-heuristic
8349 Enable the group heuristic in the scheduler. This heuristic favors
8350 the instruction that belongs to a schedule group. This is enabled
8351 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8352 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8354 @item -fsched-critical-path-heuristic
8355 @opindex fsched-critical-path-heuristic
8356 Enable the critical-path heuristic in the scheduler. This heuristic favors
8357 instructions on the critical path. This is enabled by default when
8358 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8359 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8361 @item -fsched-spec-insn-heuristic
8362 @opindex fsched-spec-insn-heuristic
8363 Enable the speculative instruction heuristic in the scheduler. This
8364 heuristic favors speculative instructions with greater dependency weakness.
8365 This is enabled by default when scheduling is enabled, i.e.@:
8366 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8367 or at @option{-O2} or higher.
8369 @item -fsched-rank-heuristic
8370 @opindex fsched-rank-heuristic
8371 Enable the rank heuristic in the scheduler. This heuristic favors
8372 the instruction belonging to a basic block with greater size or frequency.
8373 This is enabled by default when scheduling is enabled, i.e.@:
8374 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8375 at @option{-O2} or higher.
8377 @item -fsched-last-insn-heuristic
8378 @opindex fsched-last-insn-heuristic
8379 Enable the last-instruction heuristic in the scheduler. This heuristic
8380 favors the instruction that is less dependent on the last instruction
8381 scheduled. This is enabled by default when scheduling is enabled,
8382 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8383 at @option{-O2} or higher.
8385 @item -fsched-dep-count-heuristic
8386 @opindex fsched-dep-count-heuristic
8387 Enable the dependent-count heuristic in the scheduler. This heuristic
8388 favors the instruction that has more instructions depending on it.
8389 This is enabled by default when scheduling is enabled, i.e.@:
8390 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8391 at @option{-O2} or higher.
8393 @item -freschedule-modulo-scheduled-loops
8394 @opindex freschedule-modulo-scheduled-loops
8395 Modulo scheduling is performed before traditional scheduling. If a loop
8396 is modulo scheduled, later scheduling passes may change its schedule.
8397 Use this option to control that behavior.
8399 @item -fselective-scheduling
8400 @opindex fselective-scheduling
8401 Schedule instructions using selective scheduling algorithm. Selective
8402 scheduling runs instead of the first scheduler pass.
8404 @item -fselective-scheduling2
8405 @opindex fselective-scheduling2
8406 Schedule instructions using selective scheduling algorithm. Selective
8407 scheduling runs instead of the second scheduler pass.
8409 @item -fsel-sched-pipelining
8410 @opindex fsel-sched-pipelining
8411 Enable software pipelining of innermost loops during selective scheduling.
8412 This option has no effect unless one of @option{-fselective-scheduling} or
8413 @option{-fselective-scheduling2} is turned on.
8415 @item -fsel-sched-pipelining-outer-loops
8416 @opindex fsel-sched-pipelining-outer-loops
8417 When pipelining loops during selective scheduling, also pipeline outer loops.
8418 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8420 @item -fsemantic-interposition
8421 @opindex fsemantic-interposition
8422 Some object formats, like ELF, allow interposing of symbols by the
8424 This means that for symbols exported from the DSO, the compiler cannot perform
8425 interprocedural propagation, inlining and other optimizations in anticipation
8426 that the function or variable in question may change. While this feature is
8427 useful, for example, to rewrite memory allocation functions by a debugging
8428 implementation, it is expensive in the terms of code quality.
8429 With @option{-fno-semantic-interposition} the compiler assumes that
8430 if interposition happens for functions the overwriting function will have
8431 precisely the same semantics (and side effects).
8432 Similarly if interposition happens
8433 for variables, the constructor of the variable will be the same. The flag
8434 has no effect for functions explicitly declared inline
8435 (where it is never allowed for interposition to change semantics)
8436 and for symbols explicitly declared weak.
8439 @opindex fshrink-wrap
8440 Emit function prologues only before parts of the function that need it,
8441 rather than at the top of the function. This flag is enabled by default at
8442 @option{-O} and higher.
8444 @item -fcaller-saves
8445 @opindex fcaller-saves
8446 Enable allocation of values to registers that are clobbered by
8447 function calls, by emitting extra instructions to save and restore the
8448 registers around such calls. Such allocation is done only when it
8449 seems to result in better code.
8451 This option is always enabled by default on certain machines, usually
8452 those which have no call-preserved registers to use instead.
8454 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8456 @item -fcombine-stack-adjustments
8457 @opindex fcombine-stack-adjustments
8458 Tracks stack adjustments (pushes and pops) and stack memory references
8459 and then tries to find ways to combine them.
8461 Enabled by default at @option{-O1} and higher.
8465 Use caller save registers for allocation if those registers are not used by
8466 any called function. In that case it is not necessary to save and restore
8467 them around calls. This is only possible if called functions are part of
8468 same compilation unit as current function and they are compiled before it.
8470 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8472 @item -fconserve-stack
8473 @opindex fconserve-stack
8474 Attempt to minimize stack usage. The compiler attempts to use less
8475 stack space, even if that makes the program slower. This option
8476 implies setting the @option{large-stack-frame} parameter to 100
8477 and the @option{large-stack-frame-growth} parameter to 400.
8479 @item -ftree-reassoc
8480 @opindex ftree-reassoc
8481 Perform reassociation on trees. This flag is enabled by default
8482 at @option{-O} and higher.
8486 Perform partial redundancy elimination (PRE) on trees. This flag is
8487 enabled by default at @option{-O2} and @option{-O3}.
8489 @item -ftree-partial-pre
8490 @opindex ftree-partial-pre
8491 Make partial redundancy elimination (PRE) more aggressive. This flag is
8492 enabled by default at @option{-O3}.
8494 @item -ftree-forwprop
8495 @opindex ftree-forwprop
8496 Perform forward propagation on trees. This flag is enabled by default
8497 at @option{-O} and higher.
8501 Perform full redundancy elimination (FRE) on trees. The difference
8502 between FRE and PRE is that FRE only considers expressions
8503 that are computed on all paths leading to the redundant computation.
8504 This analysis is faster than PRE, though it exposes fewer redundancies.
8505 This flag is enabled by default at @option{-O} and higher.
8507 @item -ftree-phiprop
8508 @opindex ftree-phiprop
8509 Perform hoisting of loads from conditional pointers on trees. This
8510 pass is enabled by default at @option{-O} and higher.
8512 @item -fhoist-adjacent-loads
8513 @opindex fhoist-adjacent-loads
8514 Speculatively hoist loads from both branches of an if-then-else if the
8515 loads are from adjacent locations in the same structure and the target
8516 architecture has a conditional move instruction. This flag is enabled
8517 by default at @option{-O2} and higher.
8519 @item -ftree-copy-prop
8520 @opindex ftree-copy-prop
8521 Perform copy propagation on trees. This pass eliminates unnecessary
8522 copy operations. This flag is enabled by default at @option{-O} and
8525 @item -fipa-pure-const
8526 @opindex fipa-pure-const
8527 Discover which functions are pure or constant.
8528 Enabled by default at @option{-O} and higher.
8530 @item -fipa-reference
8531 @opindex fipa-reference
8532 Discover which static variables do not escape the
8534 Enabled by default at @option{-O} and higher.
8538 Perform interprocedural pointer analysis and interprocedural modification
8539 and reference analysis. This option can cause excessive memory and
8540 compile-time usage on large compilation units. It is not enabled by
8541 default at any optimization level.
8544 @opindex fipa-profile
8545 Perform interprocedural profile propagation. The functions called only from
8546 cold functions are marked as cold. Also functions executed once (such as
8547 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8548 functions and loop less parts of functions executed once are then optimized for
8550 Enabled by default at @option{-O} and higher.
8554 Perform interprocedural constant propagation.
8555 This optimization analyzes the program to determine when values passed
8556 to functions are constants and then optimizes accordingly.
8557 This optimization can substantially increase performance
8558 if the application has constants passed to functions.
8559 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8561 @item -fipa-cp-clone
8562 @opindex fipa-cp-clone
8563 Perform function cloning to make interprocedural constant propagation stronger.
8564 When enabled, interprocedural constant propagation performs function cloning
8565 when externally visible function can be called with constant arguments.
8566 Because this optimization can create multiple copies of functions,
8567 it may significantly increase code size
8568 (see @option{--param ipcp-unit-growth=@var{value}}).
8569 This flag is enabled by default at @option{-O3}.
8571 @item -fipa-cp-alignment
8572 @opindex -fipa-cp-alignment
8573 When enabled, this optimization propagates alignment of function
8574 parameters to support better vectorization and string operations.
8576 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8577 requires that @option{-fipa-cp} is enabled.
8581 Perform Identical Code Folding for functions and read-only variables.
8582 The optimization reduces code size and may disturb unwind stacks by replacing
8583 a function by equivalent one with a different name. The optimization works
8584 more effectively with link time optimization enabled.
8586 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8587 works on different levels and thus the optimizations are not same - there are
8588 equivalences that are found only by GCC and equivalences found only by Gold.
8590 This flag is enabled by default at @option{-O2} and @option{-Os}.
8592 @item -fisolate-erroneous-paths-dereference
8593 @opindex fisolate-erroneous-paths-dereference
8594 Detect paths that trigger erroneous or undefined behavior due to
8595 dereferencing a null pointer. Isolate those paths from the main control
8596 flow and turn the statement with erroneous or undefined behavior into a trap.
8597 This flag is enabled by default at @option{-O2} and higher and depends on
8598 @option{-fdelete-null-pointer-checks} also being enabled.
8600 @item -fisolate-erroneous-paths-attribute
8601 @opindex fisolate-erroneous-paths-attribute
8602 Detect paths that trigger erroneous or undefined behavior due a null value
8603 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8604 attribute. Isolate those paths from the main control flow and turn the
8605 statement with erroneous or undefined behavior into a trap. This is not
8606 currently enabled, but may be enabled by @option{-O2} in the future.
8610 Perform forward store motion on trees. This flag is
8611 enabled by default at @option{-O} and higher.
8613 @item -ftree-bit-ccp
8614 @opindex ftree-bit-ccp
8615 Perform sparse conditional bit constant propagation on trees and propagate
8616 pointer alignment information.
8617 This pass only operates on local scalar variables and is enabled by default
8618 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8622 Perform sparse conditional constant propagation (CCP) on trees. This
8623 pass only operates on local scalar variables and is enabled by default
8624 at @option{-O} and higher.
8627 @opindex fssa-phiopt
8628 Perform pattern matching on SSA PHI nodes to optimize conditional
8629 code. This pass is enabled by default at @option{-O} and higher.
8631 @item -ftree-switch-conversion
8632 @opindex ftree-switch-conversion
8633 Perform conversion of simple initializations in a switch to
8634 initializations from a scalar array. This flag is enabled by default
8635 at @option{-O2} and higher.
8637 @item -ftree-tail-merge
8638 @opindex ftree-tail-merge
8639 Look for identical code sequences. When found, replace one with a jump to the
8640 other. This optimization is known as tail merging or cross jumping. This flag
8641 is enabled by default at @option{-O2} and higher. The compilation time
8643 be limited using @option{max-tail-merge-comparisons} parameter and
8644 @option{max-tail-merge-iterations} parameter.
8648 Perform dead code elimination (DCE) on trees. This flag is enabled by
8649 default at @option{-O} and higher.
8651 @item -ftree-builtin-call-dce
8652 @opindex ftree-builtin-call-dce
8653 Perform conditional dead code elimination (DCE) for calls to built-in functions
8654 that may set @code{errno} but are otherwise side-effect free. This flag is
8655 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8658 @item -ftree-dominator-opts
8659 @opindex ftree-dominator-opts
8660 Perform a variety of simple scalar cleanups (constant/copy
8661 propagation, redundancy elimination, range propagation and expression
8662 simplification) based on a dominator tree traversal. This also
8663 performs jump threading (to reduce jumps to jumps). This flag is
8664 enabled by default at @option{-O} and higher.
8668 Perform dead store elimination (DSE) on trees. A dead store is a store into
8669 a memory location that is later overwritten by another store without
8670 any intervening loads. In this case the earlier store can be deleted. This
8671 flag is enabled by default at @option{-O} and higher.
8675 Perform loop header copying on trees. This is beneficial since it increases
8676 effectiveness of code motion optimizations. It also saves one jump. This flag
8677 is enabled by default at @option{-O} and higher. It is not enabled
8678 for @option{-Os}, since it usually increases code size.
8680 @item -ftree-loop-optimize
8681 @opindex ftree-loop-optimize
8682 Perform loop optimizations on trees. This flag is enabled by default
8683 at @option{-O} and higher.
8685 @item -ftree-loop-linear
8686 @opindex ftree-loop-linear
8687 Perform loop interchange transformations on tree. Same as
8688 @option{-floop-interchange}. To use this code transformation, GCC has
8689 to be configured with @option{--with-isl} to enable the Graphite loop
8690 transformation infrastructure.
8692 @item -floop-interchange
8693 @opindex floop-interchange
8694 Perform loop interchange transformations on loops. Interchanging two
8695 nested loops switches the inner and outer loops. For example, given a
8700 A(J, I) = A(J, I) * C
8705 loop interchange transforms the loop as if it were written:
8709 A(J, I) = A(J, I) * C
8713 which can be beneficial when @code{N} is larger than the caches,
8714 because in Fortran, the elements of an array are stored in memory
8715 contiguously by column, and the original loop iterates over rows,
8716 potentially creating at each access a cache miss. This optimization
8717 applies to all the languages supported by GCC and is not limited to
8718 Fortran. To use this code transformation, GCC has to be configured
8719 with @option{--with-isl} to enable the Graphite loop transformation
8722 @item -floop-strip-mine
8723 @opindex floop-strip-mine
8724 Perform loop strip mining transformations on loops. Strip mining
8725 splits a loop into two nested loops. The outer loop has strides
8726 equal to the strip size and the inner loop has strides of the
8727 original loop within a strip. The strip length can be changed
8728 using the @option{loop-block-tile-size} parameter. For example,
8736 loop strip mining transforms the loop as if it were written:
8739 DO I = II, min (II + 50, N)
8744 This optimization applies to all the languages supported by GCC and is
8745 not limited to Fortran. To use this code transformation, GCC has to
8746 be configured with @option{--with-isl} to enable the Graphite loop
8747 transformation infrastructure.
8750 @opindex floop-block
8751 Perform loop blocking transformations on loops. Blocking strip mines
8752 each loop in the loop nest such that the memory accesses of the
8753 element loops fit inside caches. The strip length can be changed
8754 using the @option{loop-block-tile-size} parameter. For example, given
8759 A(J, I) = B(I) + C(J)
8764 loop blocking transforms the loop as if it were written:
8768 DO I = II, min (II + 50, N)
8769 DO J = JJ, min (JJ + 50, M)
8770 A(J, I) = B(I) + C(J)
8776 which can be beneficial when @code{M} is larger than the caches,
8777 because the innermost loop iterates over a smaller amount of data
8778 which can be kept in the caches. This optimization applies to all the
8779 languages supported by GCC and is not limited to Fortran. To use this
8780 code transformation, GCC has to be configured with @option{--with-isl}
8781 to enable the Graphite loop transformation infrastructure.
8783 @item -fgraphite-identity
8784 @opindex fgraphite-identity
8785 Enable the identity transformation for graphite. For every SCoP we generate
8786 the polyhedral representation and transform it back to gimple. Using
8787 @option{-fgraphite-identity} we can check the costs or benefits of the
8788 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8789 are also performed by the code generator ISL, like index splitting and
8790 dead code elimination in loops.
8792 @item -floop-nest-optimize
8793 @opindex floop-nest-optimize
8794 Enable the ISL based loop nest optimizer. This is a generic loop nest
8795 optimizer based on the Pluto optimization algorithms. It calculates a loop
8796 structure optimized for data-locality and parallelism. This option
8799 @item -floop-unroll-and-jam
8800 @opindex floop-unroll-and-jam
8801 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8802 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8803 The unrolled dimension (counting from the most inner one) can be changed
8804 using the @option{loop-unroll-jam-depth} parameter. .
8806 @item -floop-parallelize-all
8807 @opindex floop-parallelize-all
8808 Use the Graphite data dependence analysis to identify loops that can
8809 be parallelized. Parallelize all the loops that can be analyzed to
8810 not contain loop carried dependences without checking that it is
8811 profitable to parallelize the loops.
8813 @item -fcheck-data-deps
8814 @opindex fcheck-data-deps
8815 Compare the results of several data dependence analyzers. This option
8816 is used for debugging the data dependence analyzers.
8818 @item -ftree-loop-if-convert
8819 @opindex ftree-loop-if-convert
8820 Attempt to transform conditional jumps in the innermost loops to
8821 branch-less equivalents. The intent is to remove control-flow from
8822 the innermost loops in order to improve the ability of the
8823 vectorization pass to handle these loops. This is enabled by default
8824 if vectorization is enabled.
8826 @item -ftree-loop-if-convert-stores
8827 @opindex ftree-loop-if-convert-stores
8828 Attempt to also if-convert conditional jumps containing memory writes.
8829 This transformation can be unsafe for multi-threaded programs as it
8830 transforms conditional memory writes into unconditional memory writes.
8833 for (i = 0; i < N; i++)
8839 for (i = 0; i < N; i++)
8840 A[i] = cond ? expr : A[i];
8842 potentially producing data races.
8844 @item -ftree-loop-distribution
8845 @opindex ftree-loop-distribution
8846 Perform loop distribution. This flag can improve cache performance on
8847 big loop bodies and allow further loop optimizations, like
8848 parallelization or vectorization, to take place. For example, the loop
8865 @item -ftree-loop-distribute-patterns
8866 @opindex ftree-loop-distribute-patterns
8867 Perform loop distribution of patterns that can be code generated with
8868 calls to a library. This flag is enabled by default at @option{-O3}.
8870 This pass distributes the initialization loops and generates a call to
8871 memset zero. For example, the loop
8887 and the initialization loop is transformed into a call to memset zero.
8889 @item -ftree-loop-im
8890 @opindex ftree-loop-im
8891 Perform loop invariant motion on trees. This pass moves only invariants that
8892 are hard to handle at RTL level (function calls, operations that expand to
8893 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8894 operands of conditions that are invariant out of the loop, so that we can use
8895 just trivial invariantness analysis in loop unswitching. The pass also includes
8898 @item -ftree-loop-ivcanon
8899 @opindex ftree-loop-ivcanon
8900 Create a canonical counter for number of iterations in loops for which
8901 determining number of iterations requires complicated analysis. Later
8902 optimizations then may determine the number easily. Useful especially
8903 in connection with unrolling.
8907 Perform induction variable optimizations (strength reduction, induction
8908 variable merging and induction variable elimination) on trees.
8910 @item -ftree-parallelize-loops=n
8911 @opindex ftree-parallelize-loops
8912 Parallelize loops, i.e., split their iteration space to run in n threads.
8913 This is only possible for loops whose iterations are independent
8914 and can be arbitrarily reordered. The optimization is only
8915 profitable on multiprocessor machines, for loops that are CPU-intensive,
8916 rather than constrained e.g.@: by memory bandwidth. This option
8917 implies @option{-pthread}, and thus is only supported on targets
8918 that have support for @option{-pthread}.
8922 Perform function-local points-to analysis on trees. This flag is
8923 enabled by default at @option{-O} and higher.
8927 Perform scalar replacement of aggregates. This pass replaces structure
8928 references with scalars to prevent committing structures to memory too
8929 early. This flag is enabled by default at @option{-O} and higher.
8931 @item -ftree-copyrename
8932 @opindex ftree-copyrename
8933 Perform copy renaming on trees. This pass attempts to rename compiler
8934 temporaries to other variables at copy locations, usually resulting in
8935 variable names which more closely resemble the original variables. This flag
8936 is enabled by default at @option{-O} and higher.
8938 @item -ftree-coalesce-inlined-vars
8939 @opindex ftree-coalesce-inlined-vars
8940 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8941 combine small user-defined variables too, but only if they are inlined
8942 from other functions. It is a more limited form of
8943 @option{-ftree-coalesce-vars}. This may harm debug information of such
8944 inlined variables, but it keeps variables of the inlined-into
8945 function apart from each other, such that they are more likely to
8946 contain the expected values in a debugging session.
8948 @item -ftree-coalesce-vars
8949 @opindex ftree-coalesce-vars
8950 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8951 combine small user-defined variables too, instead of just compiler
8952 temporaries. This may severely limit the ability to debug an optimized
8953 program compiled with @option{-fno-var-tracking-assignments}. In the
8954 negated form, this flag prevents SSA coalescing of user variables,
8955 including inlined ones. This option is enabled by default.
8959 Perform temporary expression replacement during the SSA->normal phase. Single
8960 use/single def temporaries are replaced at their use location with their
8961 defining expression. This results in non-GIMPLE code, but gives the expanders
8962 much more complex trees to work on resulting in better RTL generation. This is
8963 enabled by default at @option{-O} and higher.
8967 Perform straight-line strength reduction on trees. This recognizes related
8968 expressions involving multiplications and replaces them by less expensive
8969 calculations when possible. This is enabled by default at @option{-O} and
8972 @item -ftree-vectorize
8973 @opindex ftree-vectorize
8974 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8975 and @option{-ftree-slp-vectorize} if not explicitly specified.
8977 @item -ftree-loop-vectorize
8978 @opindex ftree-loop-vectorize
8979 Perform loop vectorization on trees. This flag is enabled by default at
8980 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8982 @item -ftree-slp-vectorize
8983 @opindex ftree-slp-vectorize
8984 Perform basic block vectorization on trees. This flag is enabled by default at
8985 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8987 @item -fvect-cost-model=@var{model}
8988 @opindex fvect-cost-model
8989 Alter the cost model used for vectorization. The @var{model} argument
8990 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8991 With the @samp{unlimited} model the vectorized code-path is assumed
8992 to be profitable while with the @samp{dynamic} model a runtime check
8993 guards the vectorized code-path to enable it only for iteration
8994 counts that will likely execute faster than when executing the original
8995 scalar loop. The @samp{cheap} model disables vectorization of
8996 loops where doing so would be cost prohibitive for example due to
8997 required runtime checks for data dependence or alignment but otherwise
8998 is equal to the @samp{dynamic} model.
8999 The default cost model depends on other optimization flags and is
9000 either @samp{dynamic} or @samp{cheap}.
9002 @item -fsimd-cost-model=@var{model}
9003 @opindex fsimd-cost-model
9004 Alter the cost model used for vectorization of loops marked with the OpenMP
9005 or Cilk Plus simd directive. The @var{model} argument should be one of
9006 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9007 have the same meaning as described in @option{-fvect-cost-model} and by
9008 default a cost model defined with @option{-fvect-cost-model} is used.
9012 Perform Value Range Propagation on trees. This is similar to the
9013 constant propagation pass, but instead of values, ranges of values are
9014 propagated. This allows the optimizers to remove unnecessary range
9015 checks like array bound checks and null pointer checks. This is
9016 enabled by default at @option{-O2} and higher. Null pointer check
9017 elimination is only done if @option{-fdelete-null-pointer-checks} is
9020 @item -fsplit-ivs-in-unroller
9021 @opindex fsplit-ivs-in-unroller
9022 Enables expression of values of induction variables in later iterations
9023 of the unrolled loop using the value in the first iteration. This breaks
9024 long dependency chains, thus improving efficiency of the scheduling passes.
9026 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9027 same effect. However, that is not reliable in cases where the loop body
9028 is more complicated than a single basic block. It also does not work at all
9029 on some architectures due to restrictions in the CSE pass.
9031 This optimization is enabled by default.
9033 @item -fvariable-expansion-in-unroller
9034 @opindex fvariable-expansion-in-unroller
9035 With this option, the compiler creates multiple copies of some
9036 local variables when unrolling a loop, which can result in superior code.
9038 @item -fpartial-inlining
9039 @opindex fpartial-inlining
9040 Inline parts of functions. This option has any effect only
9041 when inlining itself is turned on by the @option{-finline-functions}
9042 or @option{-finline-small-functions} options.
9044 Enabled at level @option{-O2}.
9046 @item -fpredictive-commoning
9047 @opindex fpredictive-commoning
9048 Perform predictive commoning optimization, i.e., reusing computations
9049 (especially memory loads and stores) performed in previous
9050 iterations of loops.
9052 This option is enabled at level @option{-O3}.
9054 @item -fprefetch-loop-arrays
9055 @opindex fprefetch-loop-arrays
9056 If supported by the target machine, generate instructions to prefetch
9057 memory to improve the performance of loops that access large arrays.
9059 This option may generate better or worse code; results are highly
9060 dependent on the structure of loops within the source code.
9062 Disabled at level @option{-Os}.
9065 @itemx -fno-peephole2
9066 @opindex fno-peephole
9067 @opindex fno-peephole2
9068 Disable any machine-specific peephole optimizations. The difference
9069 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9070 are implemented in the compiler; some targets use one, some use the
9071 other, a few use both.
9073 @option{-fpeephole} is enabled by default.
9074 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9076 @item -fno-guess-branch-probability
9077 @opindex fno-guess-branch-probability
9078 Do not guess branch probabilities using heuristics.
9080 GCC uses heuristics to guess branch probabilities if they are
9081 not provided by profiling feedback (@option{-fprofile-arcs}). These
9082 heuristics are based on the control flow graph. If some branch probabilities
9083 are specified by @code{__builtin_expect}, then the heuristics are
9084 used to guess branch probabilities for the rest of the control flow graph,
9085 taking the @code{__builtin_expect} info into account. The interactions
9086 between the heuristics and @code{__builtin_expect} can be complex, and in
9087 some cases, it may be useful to disable the heuristics so that the effects
9088 of @code{__builtin_expect} are easier to understand.
9090 The default is @option{-fguess-branch-probability} at levels
9091 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9093 @item -freorder-blocks
9094 @opindex freorder-blocks
9095 Reorder basic blocks in the compiled function in order to reduce number of
9096 taken branches and improve code locality.
9098 Enabled at levels @option{-O2}, @option{-O3}.
9100 @item -freorder-blocks-and-partition
9101 @opindex freorder-blocks-and-partition
9102 In addition to reordering basic blocks in the compiled function, in order
9103 to reduce number of taken branches, partitions hot and cold basic blocks
9104 into separate sections of the assembly and .o files, to improve
9105 paging and cache locality performance.
9107 This optimization is automatically turned off in the presence of
9108 exception handling, for linkonce sections, for functions with a user-defined
9109 section attribute and on any architecture that does not support named
9112 Enabled for x86 at levels @option{-O2}, @option{-O3}.
9114 @item -freorder-functions
9115 @opindex freorder-functions
9116 Reorder functions in the object file in order to
9117 improve code locality. This is implemented by using special
9118 subsections @code{.text.hot} for most frequently executed functions and
9119 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9120 the linker so object file format must support named sections and linker must
9121 place them in a reasonable way.
9123 Also profile feedback must be available to make this option effective. See
9124 @option{-fprofile-arcs} for details.
9126 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9128 @item -fstrict-aliasing
9129 @opindex fstrict-aliasing
9130 Allow the compiler to assume the strictest aliasing rules applicable to
9131 the language being compiled. For C (and C++), this activates
9132 optimizations based on the type of expressions. In particular, an
9133 object of one type is assumed never to reside at the same address as an
9134 object of a different type, unless the types are almost the same. For
9135 example, an @code{unsigned int} can alias an @code{int}, but not a
9136 @code{void*} or a @code{double}. A character type may alias any other
9139 @anchor{Type-punning}Pay special attention to code like this:
9152 The practice of reading from a different union member than the one most
9153 recently written to (called ``type-punning'') is common. Even with
9154 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9155 is accessed through the union type. So, the code above works as
9156 expected. @xref{Structures unions enumerations and bit-fields
9157 implementation}. However, this code might not:
9168 Similarly, access by taking the address, casting the resulting pointer
9169 and dereferencing the result has undefined behavior, even if the cast
9170 uses a union type, e.g.:
9174 return ((union a_union *) &d)->i;
9178 The @option{-fstrict-aliasing} option is enabled at levels
9179 @option{-O2}, @option{-O3}, @option{-Os}.
9181 @item -fstrict-overflow
9182 @opindex fstrict-overflow
9183 Allow the compiler to assume strict signed overflow rules, depending
9184 on the language being compiled. For C (and C++) this means that
9185 overflow when doing arithmetic with signed numbers is undefined, which
9186 means that the compiler may assume that it does not happen. This
9187 permits various optimizations. For example, the compiler assumes
9188 that an expression like @code{i + 10 > i} is always true for
9189 signed @code{i}. This assumption is only valid if signed overflow is
9190 undefined, as the expression is false if @code{i + 10} overflows when
9191 using twos complement arithmetic. When this option is in effect any
9192 attempt to determine whether an operation on signed numbers
9193 overflows must be written carefully to not actually involve overflow.
9195 This option also allows the compiler to assume strict pointer
9196 semantics: given a pointer to an object, if adding an offset to that
9197 pointer does not produce a pointer to the same object, the addition is
9198 undefined. This permits the compiler to conclude that @code{p + u >
9199 p} is always true for a pointer @code{p} and unsigned integer
9200 @code{u}. This assumption is only valid because pointer wraparound is
9201 undefined, as the expression is false if @code{p + u} overflows using
9202 twos complement arithmetic.
9204 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
9205 that integer signed overflow is fully defined: it wraps. When
9206 @option{-fwrapv} is used, there is no difference between
9207 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
9208 integers. With @option{-fwrapv} certain types of overflow are
9209 permitted. For example, if the compiler gets an overflow when doing
9210 arithmetic on constants, the overflowed value can still be used with
9211 @option{-fwrapv}, but not otherwise.
9213 The @option{-fstrict-overflow} option is enabled at levels
9214 @option{-O2}, @option{-O3}, @option{-Os}.
9216 @item -falign-functions
9217 @itemx -falign-functions=@var{n}
9218 @opindex falign-functions
9219 Align the start of functions to the next power-of-two greater than
9220 @var{n}, skipping up to @var{n} bytes. For instance,
9221 @option{-falign-functions=32} aligns functions to the next 32-byte
9222 boundary, but @option{-falign-functions=24} aligns to the next
9223 32-byte boundary only if this can be done by skipping 23 bytes or less.
9225 @option{-fno-align-functions} and @option{-falign-functions=1} are
9226 equivalent and mean that functions are not aligned.
9228 Some assemblers only support this flag when @var{n} is a power of two;
9229 in that case, it is rounded up.
9231 If @var{n} is not specified or is zero, use a machine-dependent default.
9233 Enabled at levels @option{-O2}, @option{-O3}.
9235 @item -falign-labels
9236 @itemx -falign-labels=@var{n}
9237 @opindex falign-labels
9238 Align all branch targets to a power-of-two boundary, skipping up to
9239 @var{n} bytes like @option{-falign-functions}. This option can easily
9240 make code slower, because it must insert dummy operations for when the
9241 branch target is reached in the usual flow of the code.
9243 @option{-fno-align-labels} and @option{-falign-labels=1} are
9244 equivalent and mean that labels are not aligned.
9246 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9247 are greater than this value, then their values are used instead.
9249 If @var{n} is not specified or is zero, use a machine-dependent default
9250 which is very likely to be @samp{1}, meaning no alignment.
9252 Enabled at levels @option{-O2}, @option{-O3}.
9255 @itemx -falign-loops=@var{n}
9256 @opindex falign-loops
9257 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9258 like @option{-falign-functions}. If the loops are
9259 executed many times, this makes up for any execution of the dummy
9262 @option{-fno-align-loops} and @option{-falign-loops=1} are
9263 equivalent and mean that loops are not aligned.
9265 If @var{n} is not specified or is zero, use a machine-dependent default.
9267 Enabled at levels @option{-O2}, @option{-O3}.
9270 @itemx -falign-jumps=@var{n}
9271 @opindex falign-jumps
9272 Align branch targets to a power-of-two boundary, for branch targets
9273 where the targets can only be reached by jumping, skipping up to @var{n}
9274 bytes like @option{-falign-functions}. In this case, no dummy operations
9277 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9278 equivalent and mean that loops are not aligned.
9280 If @var{n} is not specified or is zero, use a machine-dependent default.
9282 Enabled at levels @option{-O2}, @option{-O3}.
9284 @item -funit-at-a-time
9285 @opindex funit-at-a-time
9286 This option is left for compatibility reasons. @option{-funit-at-a-time}
9287 has no effect, while @option{-fno-unit-at-a-time} implies
9288 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9292 @item -fno-toplevel-reorder
9293 @opindex fno-toplevel-reorder
9294 Do not reorder top-level functions, variables, and @code{asm}
9295 statements. Output them in the same order that they appear in the
9296 input file. When this option is used, unreferenced static variables
9297 are not removed. This option is intended to support existing code
9298 that relies on a particular ordering. For new code, it is better to
9299 use attributes when possible.
9301 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9302 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9307 Constructs webs as commonly used for register allocation purposes and assign
9308 each web individual pseudo register. This allows the register allocation pass
9309 to operate on pseudos directly, but also strengthens several other optimization
9310 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9311 however, make debugging impossible, since variables no longer stay in a
9314 Enabled by default with @option{-funroll-loops}.
9316 @item -fwhole-program
9317 @opindex fwhole-program
9318 Assume that the current compilation unit represents the whole program being
9319 compiled. All public functions and variables with the exception of @code{main}
9320 and those merged by attribute @code{externally_visible} become static functions
9321 and in effect are optimized more aggressively by interprocedural optimizers.
9323 This option should not be used in combination with @option{-flto}.
9324 Instead relying on a linker plugin should provide safer and more precise
9327 @item -flto[=@var{n}]
9329 This option runs the standard link-time optimizer. When invoked
9330 with source code, it generates GIMPLE (one of GCC's internal
9331 representations) and writes it to special ELF sections in the object
9332 file. When the object files are linked together, all the function
9333 bodies are read from these ELF sections and instantiated as if they
9334 had been part of the same translation unit.
9336 To use the link-time optimizer, @option{-flto} and optimization
9337 options should be specified at compile time and during the final link.
9341 gcc -c -O2 -flto foo.c
9342 gcc -c -O2 -flto bar.c
9343 gcc -o myprog -flto -O2 foo.o bar.o
9346 The first two invocations to GCC save a bytecode representation
9347 of GIMPLE into special ELF sections inside @file{foo.o} and
9348 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9349 @file{foo.o} and @file{bar.o}, merges the two files into a single
9350 internal image, and compiles the result as usual. Since both
9351 @file{foo.o} and @file{bar.o} are merged into a single image, this
9352 causes all the interprocedural analyses and optimizations in GCC to
9353 work across the two files as if they were a single one. This means,
9354 for example, that the inliner is able to inline functions in
9355 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9357 Another (simpler) way to enable link-time optimization is:
9360 gcc -o myprog -flto -O2 foo.c bar.c
9363 The above generates bytecode for @file{foo.c} and @file{bar.c},
9364 merges them together into a single GIMPLE representation and optimizes
9365 them as usual to produce @file{myprog}.
9367 The only important thing to keep in mind is that to enable link-time
9368 optimizations you need to use the GCC driver to perform the link-step.
9369 GCC then automatically performs link-time optimization if any of the
9370 objects involved were compiled with the @option{-flto} command-line option.
9372 should specify the optimization options to be used for link-time
9373 optimization though GCC tries to be clever at guessing an
9374 optimization level to use from the options used at compile-time
9375 if you fail to specify one at link-time. You can always override
9376 the automatic decision to do link-time optimization at link-time
9377 by passing @option{-fno-lto} to the link command.
9379 To make whole program optimization effective, it is necessary to make
9380 certain whole program assumptions. The compiler needs to know
9381 what functions and variables can be accessed by libraries and runtime
9382 outside of the link-time optimized unit. When supported by the linker,
9383 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9384 to the compiler about used and externally visible symbols. When
9385 the linker plugin is not available, @option{-fwhole-program} should be
9386 used to allow the compiler to make these assumptions, which leads
9387 to more aggressive optimization decisions.
9389 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9390 compiled with @option{-flto}, the generated object file is larger than
9391 a regular object file because it contains GIMPLE bytecodes and the usual
9392 final code (see @option{-ffat-lto-objects}. This means that
9393 object files with LTO information can be linked as normal object
9394 files; if @option{-fno-lto} is passed to the linker, no
9395 interprocedural optimizations are applied. Note that when
9396 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9397 but you cannot perform a regular, non-LTO link on them.
9399 Additionally, the optimization flags used to compile individual files
9400 are not necessarily related to those used at link time. For instance,
9403 gcc -c -O0 -ffat-lto-objects -flto foo.c
9404 gcc -c -O0 -ffat-lto-objects -flto bar.c
9405 gcc -o myprog -O3 foo.o bar.o
9408 This produces individual object files with unoptimized assembler
9409 code, but the resulting binary @file{myprog} is optimized at
9410 @option{-O3}. If, instead, the final binary is generated with
9411 @option{-fno-lto}, then @file{myprog} is not optimized.
9413 When producing the final binary, GCC only
9414 applies link-time optimizations to those files that contain bytecode.
9415 Therefore, you can mix and match object files and libraries with
9416 GIMPLE bytecodes and final object code. GCC automatically selects
9417 which files to optimize in LTO mode and which files to link without
9420 There are some code generation flags preserved by GCC when
9421 generating bytecodes, as they need to be used during the final link
9422 stage. Generally options specified at link-time override those
9423 specified at compile-time.
9425 If you do not specify an optimization level option @option{-O} at
9426 link-time then GCC computes one based on the optimization levels
9427 used when compiling the object files. The highest optimization
9430 Currently, the following options and their setting are take from
9431 the first object file that explicitely specified it:
9432 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9433 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9434 and all the @option{-m} target flags.
9436 Certain ABI changing flags are required to match in all compilation-units
9437 and trying to override this at link-time with a conflicting value
9438 is ignored. This includes options such as @option{-freg-struct-return}
9439 and @option{-fpcc-struct-return}.
9441 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9442 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9443 are passed through to the link stage and merged conservatively for
9444 conflicting translation units. Specifically
9445 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9446 precedence and for example @option{-ffp-contract=off} takes precedence
9447 over @option{-ffp-contract=fast}. You can override them at linke-time.
9449 It is recommended that you compile all the files participating in the
9450 same link with the same options and also specify those options at
9453 If LTO encounters objects with C linkage declared with incompatible
9454 types in separate translation units to be linked together (undefined
9455 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9456 issued. The behavior is still undefined at run time. Similar
9457 diagnostics may be raised for other languages.
9459 Another feature of LTO is that it is possible to apply interprocedural
9460 optimizations on files written in different languages:
9465 gfortran -c -flto baz.f90
9466 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9469 Notice that the final link is done with @command{g++} to get the C++
9470 runtime libraries and @option{-lgfortran} is added to get the Fortran
9471 runtime libraries. In general, when mixing languages in LTO mode, you
9472 should use the same link command options as when mixing languages in a
9473 regular (non-LTO) compilation.
9475 If object files containing GIMPLE bytecode are stored in a library archive, say
9476 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9477 are using a linker with plugin support. To create static libraries suitable
9478 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9479 and @command{ranlib};
9480 to show the symbols of object files with GIMPLE bytecode, use
9481 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9482 and @command{nm} have been compiled with plugin support. At link time, use the the
9483 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9484 the LTO optimization process:
9487 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9490 With the linker plugin enabled, the linker extracts the needed
9491 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9492 to make them part of the aggregated GIMPLE image to be optimized.
9494 If you are not using a linker with plugin support and/or do not
9495 enable the linker plugin, then the objects inside @file{libfoo.a}
9496 are extracted and linked as usual, but they do not participate
9497 in the LTO optimization process. In order to make a static library suitable
9498 for both LTO optimization and usual linkage, compile its object files with
9499 @option{-flto} @option{-ffat-lto-objects}.
9501 Link-time optimizations do not require the presence of the whole program to
9502 operate. If the program does not require any symbols to be exported, it is
9503 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9504 the interprocedural optimizers to use more aggressive assumptions which may
9505 lead to improved optimization opportunities.
9506 Use of @option{-fwhole-program} is not needed when linker plugin is
9507 active (see @option{-fuse-linker-plugin}).
9509 The current implementation of LTO makes no
9510 attempt to generate bytecode that is portable between different
9511 types of hosts. The bytecode files are versioned and there is a
9512 strict version check, so bytecode files generated in one version of
9513 GCC do not work with an older or newer version of GCC.
9515 Link-time optimization does not work well with generation of debugging
9516 information. Combining @option{-flto} with
9517 @option{-g} is currently experimental and expected to produce unexpected
9520 If you specify the optional @var{n}, the optimization and code
9521 generation done at link time is executed in parallel using @var{n}
9522 parallel jobs by utilizing an installed @command{make} program. The
9523 environment variable @env{MAKE} may be used to override the program
9524 used. The default value for @var{n} is 1.
9526 You can also specify @option{-flto=jobserver} to use GNU make's
9527 job server mode to determine the number of parallel jobs. This
9528 is useful when the Makefile calling GCC is already executing in parallel.
9529 You must prepend a @samp{+} to the command recipe in the parent Makefile
9530 for this to work. This option likely only works if @env{MAKE} is
9533 @item -flto-partition=@var{alg}
9534 @opindex flto-partition
9535 Specify the partitioning algorithm used by the link-time optimizer.
9536 The value is either @samp{1to1} to specify a partitioning mirroring
9537 the original source files or @samp{balanced} to specify partitioning
9538 into equally sized chunks (whenever possible) or @samp{max} to create
9539 new partition for every symbol where possible. Specifying @samp{none}
9540 as an algorithm disables partitioning and streaming completely.
9541 The default value is @samp{balanced}. While @samp{1to1} can be used
9542 as an workaround for various code ordering issues, the @samp{max}
9543 partitioning is intended for internal testing only.
9544 The value @samp{one} specifies that exactly one partition should be
9545 used while the value @samp{none} bypasses partitioning and executes
9546 the link-time optimization step directly from the WPA phase.
9548 @item -flto-odr-type-merging
9549 @opindex flto-odr-type-merging
9550 Enable streaming of mangled types names of C++ types and their unification
9551 at linktime. This increases size of LTO object files, but enable
9552 diagnostics about One Definition Rule violations.
9554 @item -flto-compression-level=@var{n}
9555 @opindex flto-compression-level
9556 This option specifies the level of compression used for intermediate
9557 language written to LTO object files, and is only meaningful in
9558 conjunction with LTO mode (@option{-flto}). Valid
9559 values are 0 (no compression) to 9 (maximum compression). Values
9560 outside this range are clamped to either 0 or 9. If the option is not
9561 given, a default balanced compression setting is used.
9564 @opindex flto-report
9565 Prints a report with internal details on the workings of the link-time
9566 optimizer. The contents of this report vary from version to version.
9567 It is meant to be useful to GCC developers when processing object
9568 files in LTO mode (via @option{-flto}).
9570 Disabled by default.
9572 @item -flto-report-wpa
9573 @opindex flto-report-wpa
9574 Like @option{-flto-report}, but only print for the WPA phase of Link
9577 @item -fuse-linker-plugin
9578 @opindex fuse-linker-plugin
9579 Enables the use of a linker plugin during link-time optimization. This
9580 option relies on plugin support in the linker, which is available in gold
9581 or in GNU ld 2.21 or newer.
9583 This option enables the extraction of object files with GIMPLE bytecode out
9584 of library archives. This improves the quality of optimization by exposing
9585 more code to the link-time optimizer. This information specifies what
9586 symbols can be accessed externally (by non-LTO object or during dynamic
9587 linking). Resulting code quality improvements on binaries (and shared
9588 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9589 See @option{-flto} for a description of the effect of this flag and how to
9592 This option is enabled by default when LTO support in GCC is enabled
9593 and GCC was configured for use with
9594 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9596 @item -ffat-lto-objects
9597 @opindex ffat-lto-objects
9598 Fat LTO objects are object files that contain both the intermediate language
9599 and the object code. This makes them usable for both LTO linking and normal
9600 linking. This option is effective only when compiling with @option{-flto}
9601 and is ignored at link time.
9603 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9604 requires the complete toolchain to be aware of LTO. It requires a linker with
9605 linker plugin support for basic functionality. Additionally,
9606 @command{nm}, @command{ar} and @command{ranlib}
9607 need to support linker plugins to allow a full-featured build environment
9608 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9609 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9610 to these tools. With non fat LTO makefiles need to be modified to use them.
9612 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9615 @item -fcompare-elim
9616 @opindex fcompare-elim
9617 After register allocation and post-register allocation instruction splitting,
9618 identify arithmetic instructions that compute processor flags similar to a
9619 comparison operation based on that arithmetic. If possible, eliminate the
9620 explicit comparison operation.
9622 This pass only applies to certain targets that cannot explicitly represent
9623 the comparison operation before register allocation is complete.
9625 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9627 @item -fcprop-registers
9628 @opindex fcprop-registers
9629 After register allocation and post-register allocation instruction splitting,
9630 perform a copy-propagation pass to try to reduce scheduling dependencies
9631 and occasionally eliminate the copy.
9633 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9635 @item -fprofile-correction
9636 @opindex fprofile-correction
9637 Profiles collected using an instrumented binary for multi-threaded programs may
9638 be inconsistent due to missed counter updates. When this option is specified,
9639 GCC uses heuristics to correct or smooth out such inconsistencies. By
9640 default, GCC emits an error message when an inconsistent profile is detected.
9642 @item -fprofile-dir=@var{path}
9643 @opindex fprofile-dir
9645 Set the directory to search for the profile data files in to @var{path}.
9646 This option affects only the profile data generated by
9647 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9648 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9649 and its related options. Both absolute and relative paths can be used.
9650 By default, GCC uses the current directory as @var{path}, thus the
9651 profile data file appears in the same directory as the object file.
9653 @item -fprofile-generate
9654 @itemx -fprofile-generate=@var{path}
9655 @opindex fprofile-generate
9657 Enable options usually used for instrumenting application to produce
9658 profile useful for later recompilation with profile feedback based
9659 optimization. You must use @option{-fprofile-generate} both when
9660 compiling and when linking your program.
9662 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9664 If @var{path} is specified, GCC looks at the @var{path} to find
9665 the profile feedback data files. See @option{-fprofile-dir}.
9668 @itemx -fprofile-use=@var{path}
9669 @opindex fprofile-use
9670 Enable profile feedback-directed optimizations,
9671 and the following optimizations
9672 which are generally profitable only with profile feedback available:
9673 @option{-fbranch-probabilities}, @option{-fvpt},
9674 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9675 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9677 By default, GCC emits an error message if the feedback profiles do not
9678 match the source code. This error can be turned into a warning by using
9679 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9682 If @var{path} is specified, GCC looks at the @var{path} to find
9683 the profile feedback data files. See @option{-fprofile-dir}.
9685 @item -fauto-profile
9686 @itemx -fauto-profile=@var{path}
9687 @opindex fauto-profile
9688 Enable sampling-based feedback-directed optimizations,
9689 and the following optimizations
9690 which are generally profitable only with profile feedback available:
9691 @option{-fbranch-probabilities}, @option{-fvpt},
9692 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9693 @option{-ftree-vectorize},
9694 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9695 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9696 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9698 @var{path} is the name of a file containing AutoFDO profile information.
9699 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9701 Producing an AutoFDO profile data file requires running your program
9702 with the @command{perf} utility on a supported GNU/Linux target system.
9703 For more information, see @uref{https://perf.wiki.kernel.org/}.
9707 perf record -e br_inst_retired:near_taken -b -o perf.data \
9711 Then use the @command{create_gcov} tool to convert the raw profile data
9712 to a format that can be used by GCC.@ You must also supply the
9713 unstripped binary for your program to this tool.
9714 See @uref{https://github.com/google/autofdo}.
9718 create_gcov --binary=your_program.unstripped --profile=perf.data \
9723 The following options control compiler behavior regarding floating-point
9724 arithmetic. These options trade off between speed and
9725 correctness. All must be specifically enabled.
9729 @opindex ffloat-store
9730 Do not store floating-point variables in registers, and inhibit other
9731 options that might change whether a floating-point value is taken from a
9734 @cindex floating-point precision
9735 This option prevents undesirable excess precision on machines such as
9736 the 68000 where the floating registers (of the 68881) keep more
9737 precision than a @code{double} is supposed to have. Similarly for the
9738 x86 architecture. For most programs, the excess precision does only
9739 good, but a few programs rely on the precise definition of IEEE floating
9740 point. Use @option{-ffloat-store} for such programs, after modifying
9741 them to store all pertinent intermediate computations into variables.
9743 @item -fexcess-precision=@var{style}
9744 @opindex fexcess-precision
9745 This option allows further control over excess precision on machines
9746 where floating-point registers have more precision than the IEEE
9747 @code{float} and @code{double} types and the processor does not
9748 support operations rounding to those types. By default,
9749 @option{-fexcess-precision=fast} is in effect; this means that
9750 operations are carried out in the precision of the registers and that
9751 it is unpredictable when rounding to the types specified in the source
9752 code takes place. When compiling C, if
9753 @option{-fexcess-precision=standard} is specified then excess
9754 precision follows the rules specified in ISO C99; in particular,
9755 both casts and assignments cause values to be rounded to their
9756 semantic types (whereas @option{-ffloat-store} only affects
9757 assignments). This option is enabled by default for C if a strict
9758 conformance option such as @option{-std=c99} is used.
9761 @option{-fexcess-precision=standard} is not implemented for languages
9762 other than C, and has no effect if
9763 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9764 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9765 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9766 semantics apply without excess precision, and in the latter, rounding
9771 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9772 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9773 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9775 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9777 This option is not turned on by any @option{-O} option besides
9778 @option{-Ofast} since it can result in incorrect output for programs
9779 that depend on an exact implementation of IEEE or ISO rules/specifications
9780 for math functions. It may, however, yield faster code for programs
9781 that do not require the guarantees of these specifications.
9783 @item -fno-math-errno
9784 @opindex fno-math-errno
9785 Do not set @code{errno} after calling math functions that are executed
9786 with a single instruction, e.g., @code{sqrt}. A program that relies on
9787 IEEE exceptions for math error handling may want to use this flag
9788 for speed while maintaining IEEE arithmetic compatibility.
9790 This option is not turned on by any @option{-O} option since
9791 it can result in incorrect output for programs that depend on
9792 an exact implementation of IEEE or ISO rules/specifications for
9793 math functions. It may, however, yield faster code for programs
9794 that do not require the guarantees of these specifications.
9796 The default is @option{-fmath-errno}.
9798 On Darwin systems, the math library never sets @code{errno}. There is
9799 therefore no reason for the compiler to consider the possibility that
9800 it might, and @option{-fno-math-errno} is the default.
9802 @item -funsafe-math-optimizations
9803 @opindex funsafe-math-optimizations
9805 Allow optimizations for floating-point arithmetic that (a) assume
9806 that arguments and results are valid and (b) may violate IEEE or
9807 ANSI standards. When used at link-time, it may include libraries
9808 or startup files that change the default FPU control word or other
9809 similar optimizations.
9811 This option is not turned on by any @option{-O} option since
9812 it can result in incorrect output for programs that depend on
9813 an exact implementation of IEEE or ISO rules/specifications for
9814 math functions. It may, however, yield faster code for programs
9815 that do not require the guarantees of these specifications.
9816 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9817 @option{-fassociative-math} and @option{-freciprocal-math}.
9819 The default is @option{-fno-unsafe-math-optimizations}.
9821 @item -fassociative-math
9822 @opindex fassociative-math
9824 Allow re-association of operands in series of floating-point operations.
9825 This violates the ISO C and C++ language standard by possibly changing
9826 computation result. NOTE: re-ordering may change the sign of zero as
9827 well as ignore NaNs and inhibit or create underflow or overflow (and
9828 thus cannot be used on code that relies on rounding behavior like
9829 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9830 and thus may not be used when ordered comparisons are required.
9831 This option requires that both @option{-fno-signed-zeros} and
9832 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9833 much sense with @option{-frounding-math}. For Fortran the option
9834 is automatically enabled when both @option{-fno-signed-zeros} and
9835 @option{-fno-trapping-math} are in effect.
9837 The default is @option{-fno-associative-math}.
9839 @item -freciprocal-math
9840 @opindex freciprocal-math
9842 Allow the reciprocal of a value to be used instead of dividing by
9843 the value if this enables optimizations. For example @code{x / y}
9844 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9845 is subject to common subexpression elimination. Note that this loses
9846 precision and increases the number of flops operating on the value.
9848 The default is @option{-fno-reciprocal-math}.
9850 @item -ffinite-math-only
9851 @opindex ffinite-math-only
9852 Allow optimizations for floating-point arithmetic that assume
9853 that arguments and results are not NaNs or +-Infs.
9855 This option is not turned on by any @option{-O} option since
9856 it can result in incorrect output for programs that depend on
9857 an exact implementation of IEEE or ISO rules/specifications for
9858 math functions. It may, however, yield faster code for programs
9859 that do not require the guarantees of these specifications.
9861 The default is @option{-fno-finite-math-only}.
9863 @item -fno-signed-zeros
9864 @opindex fno-signed-zeros
9865 Allow optimizations for floating-point arithmetic that ignore the
9866 signedness of zero. IEEE arithmetic specifies the behavior of
9867 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9868 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9869 This option implies that the sign of a zero result isn't significant.
9871 The default is @option{-fsigned-zeros}.
9873 @item -fno-trapping-math
9874 @opindex fno-trapping-math
9875 Compile code assuming that floating-point operations cannot generate
9876 user-visible traps. These traps include division by zero, overflow,
9877 underflow, inexact result and invalid operation. This option requires
9878 that @option{-fno-signaling-nans} be in effect. Setting this option may
9879 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9881 This option should never be turned on by any @option{-O} option since
9882 it can result in incorrect output for programs that depend on
9883 an exact implementation of IEEE or ISO rules/specifications for
9886 The default is @option{-ftrapping-math}.
9888 @item -frounding-math
9889 @opindex frounding-math
9890 Disable transformations and optimizations that assume default floating-point
9891 rounding behavior. This is round-to-zero for all floating point
9892 to integer conversions, and round-to-nearest for all other arithmetic
9893 truncations. This option should be specified for programs that change
9894 the FP rounding mode dynamically, or that may be executed with a
9895 non-default rounding mode. This option disables constant folding of
9896 floating-point expressions at compile time (which may be affected by
9897 rounding mode) and arithmetic transformations that are unsafe in the
9898 presence of sign-dependent rounding modes.
9900 The default is @option{-fno-rounding-math}.
9902 This option is experimental and does not currently guarantee to
9903 disable all GCC optimizations that are affected by rounding mode.
9904 Future versions of GCC may provide finer control of this setting
9905 using C99's @code{FENV_ACCESS} pragma. This command-line option
9906 will be used to specify the default state for @code{FENV_ACCESS}.
9908 @item -fsignaling-nans
9909 @opindex fsignaling-nans
9910 Compile code assuming that IEEE signaling NaNs may generate user-visible
9911 traps during floating-point operations. Setting this option disables
9912 optimizations that may change the number of exceptions visible with
9913 signaling NaNs. This option implies @option{-ftrapping-math}.
9915 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9918 The default is @option{-fno-signaling-nans}.
9920 This option is experimental and does not currently guarantee to
9921 disable all GCC optimizations that affect signaling NaN behavior.
9923 @item -fsingle-precision-constant
9924 @opindex fsingle-precision-constant
9925 Treat floating-point constants as single precision instead of
9926 implicitly converting them to double-precision constants.
9928 @item -fcx-limited-range
9929 @opindex fcx-limited-range
9930 When enabled, this option states that a range reduction step is not
9931 needed when performing complex division. Also, there is no checking
9932 whether the result of a complex multiplication or division is @code{NaN
9933 + I*NaN}, with an attempt to rescue the situation in that case. The
9934 default is @option{-fno-cx-limited-range}, but is enabled by
9935 @option{-ffast-math}.
9937 This option controls the default setting of the ISO C99
9938 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9941 @item -fcx-fortran-rules
9942 @opindex fcx-fortran-rules
9943 Complex multiplication and division follow Fortran rules. Range
9944 reduction is done as part of complex division, but there is no checking
9945 whether the result of a complex multiplication or division is @code{NaN
9946 + I*NaN}, with an attempt to rescue the situation in that case.
9948 The default is @option{-fno-cx-fortran-rules}.
9952 The following options control optimizations that may improve
9953 performance, but are not enabled by any @option{-O} options. This
9954 section includes experimental options that may produce broken code.
9957 @item -fbranch-probabilities
9958 @opindex fbranch-probabilities
9959 After running a program compiled with @option{-fprofile-arcs}
9960 (@pxref{Debugging Options,, Options for Debugging Your Program or
9961 @command{gcc}}), you can compile it a second time using
9962 @option{-fbranch-probabilities}, to improve optimizations based on
9963 the number of times each branch was taken. When a program
9964 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9965 counts to a file called @file{@var{sourcename}.gcda} for each source
9966 file. The information in this data file is very dependent on the
9967 structure of the generated code, so you must use the same source code
9968 and the same optimization options for both compilations.
9970 With @option{-fbranch-probabilities}, GCC puts a
9971 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9972 These can be used to improve optimization. Currently, they are only
9973 used in one place: in @file{reorg.c}, instead of guessing which path a
9974 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9975 exactly determine which path is taken more often.
9977 @item -fprofile-values
9978 @opindex fprofile-values
9979 If combined with @option{-fprofile-arcs}, it adds code so that some
9980 data about values of expressions in the program is gathered.
9982 With @option{-fbranch-probabilities}, it reads back the data gathered
9983 from profiling values of expressions for usage in optimizations.
9985 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9987 @item -fprofile-reorder-functions
9988 @opindex fprofile-reorder-functions
9989 Function reordering based on profile instrumentation collects
9990 first time of execution of a function and orders these functions
9993 Enabled with @option{-fprofile-use}.
9997 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9998 to add code to gather information about values of expressions.
10000 With @option{-fbranch-probabilities}, it reads back the data gathered
10001 and actually performs the optimizations based on them.
10002 Currently the optimizations include specialization of division operations
10003 using the knowledge about the value of the denominator.
10005 @item -frename-registers
10006 @opindex frename-registers
10007 Attempt to avoid false dependencies in scheduled code by making use
10008 of registers left over after register allocation. This optimization
10009 most benefits processors with lots of registers. Depending on the
10010 debug information format adopted by the target, however, it can
10011 make debugging impossible, since variables no longer stay in
10012 a ``home register''.
10014 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
10016 @item -fschedule-fusion
10017 @opindex fschedule-fusion
10018 Performs a target dependent pass over the instruction stream to schedule
10019 instructions of same type together because target machine can execute them
10020 more efficiently if they are adjacent to each other in the instruction flow.
10022 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10026 Perform tail duplication to enlarge superblock size. This transformation
10027 simplifies the control flow of the function allowing other optimizations to do
10030 Enabled with @option{-fprofile-use}.
10032 @item -funroll-loops
10033 @opindex funroll-loops
10034 Unroll loops whose number of iterations can be determined at compile time or
10035 upon entry to the loop. @option{-funroll-loops} implies
10036 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10037 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10038 a small constant number of iterations). This option makes code larger, and may
10039 or may not make it run faster.
10041 Enabled with @option{-fprofile-use}.
10043 @item -funroll-all-loops
10044 @opindex funroll-all-loops
10045 Unroll all loops, even if their number of iterations is uncertain when
10046 the loop is entered. This usually makes programs run more slowly.
10047 @option{-funroll-all-loops} implies the same options as
10048 @option{-funroll-loops}.
10051 @opindex fpeel-loops
10052 Peels loops for which there is enough information that they do not
10053 roll much (from profile feedback). It also turns on complete loop peeling
10054 (i.e.@: complete removal of loops with small constant number of iterations).
10056 Enabled with @option{-fprofile-use}.
10058 @item -fmove-loop-invariants
10059 @opindex fmove-loop-invariants
10060 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10061 at level @option{-O1}
10063 @item -funswitch-loops
10064 @opindex funswitch-loops
10065 Move branches with loop invariant conditions out of the loop, with duplicates
10066 of the loop on both branches (modified according to result of the condition).
10068 @item -ffunction-sections
10069 @itemx -fdata-sections
10070 @opindex ffunction-sections
10071 @opindex fdata-sections
10072 Place each function or data item into its own section in the output
10073 file if the target supports arbitrary sections. The name of the
10074 function or the name of the data item determines the section's name
10075 in the output file.
10077 Use these options on systems where the linker can perform optimizations
10078 to improve locality of reference in the instruction space. Most systems
10079 using the ELF object format and SPARC processors running Solaris 2 have
10080 linkers with such optimizations. AIX may have these optimizations in
10083 Only use these options when there are significant benefits from doing
10084 so. When you specify these options, the assembler and linker
10085 create larger object and executable files and are also slower.
10086 You cannot use @command{gprof} on all systems if you
10087 specify this option, and you may have problems with debugging if
10088 you specify both this option and @option{-g}.
10090 @item -fbranch-target-load-optimize
10091 @opindex fbranch-target-load-optimize
10092 Perform branch target register load optimization before prologue / epilogue
10094 The use of target registers can typically be exposed only during reload,
10095 thus hoisting loads out of loops and doing inter-block scheduling needs
10096 a separate optimization pass.
10098 @item -fbranch-target-load-optimize2
10099 @opindex fbranch-target-load-optimize2
10100 Perform branch target register load optimization after prologue / epilogue
10103 @item -fbtr-bb-exclusive
10104 @opindex fbtr-bb-exclusive
10105 When performing branch target register load optimization, don't reuse
10106 branch target registers within any basic block.
10108 @item -fstack-protector
10109 @opindex fstack-protector
10110 Emit extra code to check for buffer overflows, such as stack smashing
10111 attacks. This is done by adding a guard variable to functions with
10112 vulnerable objects. This includes functions that call @code{alloca}, and
10113 functions with buffers larger than 8 bytes. The guards are initialized
10114 when a function is entered and then checked when the function exits.
10115 If a guard check fails, an error message is printed and the program exits.
10117 @item -fstack-protector-all
10118 @opindex fstack-protector-all
10119 Like @option{-fstack-protector} except that all functions are protected.
10121 @item -fstack-protector-strong
10122 @opindex fstack-protector-strong
10123 Like @option{-fstack-protector} but includes additional functions to
10124 be protected --- those that have local array definitions, or have
10125 references to local frame addresses.
10127 @item -fstack-protector-explicit
10128 @opindex fstack-protector-explicit
10129 Like @option{-fstack-protector} but only protects those functions which
10130 have the @code{stack_protect} attribute
10133 @opindex fstdarg-opt
10134 Optimize the prologue of variadic argument functions with respect to usage of
10137 @item -fsection-anchors
10138 @opindex fsection-anchors
10139 Try to reduce the number of symbolic address calculations by using
10140 shared ``anchor'' symbols to address nearby objects. This transformation
10141 can help to reduce the number of GOT entries and GOT accesses on some
10144 For example, the implementation of the following function @code{foo}:
10147 static int a, b, c;
10148 int foo (void) @{ return a + b + c; @}
10152 usually calculates the addresses of all three variables, but if you
10153 compile it with @option{-fsection-anchors}, it accesses the variables
10154 from a common anchor point instead. The effect is similar to the
10155 following pseudocode (which isn't valid C):
10160 register int *xr = &x;
10161 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10165 Not all targets support this option.
10167 @item --param @var{name}=@var{value}
10169 In some places, GCC uses various constants to control the amount of
10170 optimization that is done. For example, GCC does not inline functions
10171 that contain more than a certain number of instructions. You can
10172 control some of these constants on the command line using the
10173 @option{--param} option.
10175 The names of specific parameters, and the meaning of the values, are
10176 tied to the internals of the compiler, and are subject to change
10177 without notice in future releases.
10179 In each case, the @var{value} is an integer. The allowable choices for
10183 @item predictable-branch-outcome
10184 When branch is predicted to be taken with probability lower than this threshold
10185 (in percent), then it is considered well predictable. The default is 10.
10187 @item max-crossjump-edges
10188 The maximum number of incoming edges to consider for cross-jumping.
10189 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10190 the number of edges incoming to each block. Increasing values mean
10191 more aggressive optimization, making the compilation time increase with
10192 probably small improvement in executable size.
10194 @item min-crossjump-insns
10195 The minimum number of instructions that must be matched at the end
10196 of two blocks before cross-jumping is performed on them. This
10197 value is ignored in the case where all instructions in the block being
10198 cross-jumped from are matched. The default value is 5.
10200 @item max-grow-copy-bb-insns
10201 The maximum code size expansion factor when copying basic blocks
10202 instead of jumping. The expansion is relative to a jump instruction.
10203 The default value is 8.
10205 @item max-goto-duplication-insns
10206 The maximum number of instructions to duplicate to a block that jumps
10207 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10208 passes, GCC factors computed gotos early in the compilation process,
10209 and unfactors them as late as possible. Only computed jumps at the
10210 end of a basic blocks with no more than max-goto-duplication-insns are
10211 unfactored. The default value is 8.
10213 @item max-delay-slot-insn-search
10214 The maximum number of instructions to consider when looking for an
10215 instruction to fill a delay slot. If more than this arbitrary number of
10216 instructions are searched, the time savings from filling the delay slot
10217 are minimal, so stop searching. Increasing values mean more
10218 aggressive optimization, making the compilation time increase with probably
10219 small improvement in execution time.
10221 @item max-delay-slot-live-search
10222 When trying to fill delay slots, the maximum number of instructions to
10223 consider when searching for a block with valid live register
10224 information. Increasing this arbitrarily chosen value means more
10225 aggressive optimization, increasing the compilation time. This parameter
10226 should be removed when the delay slot code is rewritten to maintain the
10227 control-flow graph.
10229 @item max-gcse-memory
10230 The approximate maximum amount of memory that can be allocated in
10231 order to perform the global common subexpression elimination
10232 optimization. If more memory than specified is required, the
10233 optimization is not done.
10235 @item max-gcse-insertion-ratio
10236 If the ratio of expression insertions to deletions is larger than this value
10237 for any expression, then RTL PRE inserts or removes the expression and thus
10238 leaves partially redundant computations in the instruction stream. The default value is 20.
10240 @item max-pending-list-length
10241 The maximum number of pending dependencies scheduling allows
10242 before flushing the current state and starting over. Large functions
10243 with few branches or calls can create excessively large lists which
10244 needlessly consume memory and resources.
10246 @item max-modulo-backtrack-attempts
10247 The maximum number of backtrack attempts the scheduler should make
10248 when modulo scheduling a loop. Larger values can exponentially increase
10251 @item max-inline-insns-single
10252 Several parameters control the tree inliner used in GCC@.
10253 This number sets the maximum number of instructions (counted in GCC's
10254 internal representation) in a single function that the tree inliner
10255 considers for inlining. This only affects functions declared
10256 inline and methods implemented in a class declaration (C++).
10257 The default value is 400.
10259 @item max-inline-insns-auto
10260 When you use @option{-finline-functions} (included in @option{-O3}),
10261 a lot of functions that would otherwise not be considered for inlining
10262 by the compiler are investigated. To those functions, a different
10263 (more restrictive) limit compared to functions declared inline can
10265 The default value is 40.
10267 @item inline-min-speedup
10268 When estimated performance improvement of caller + callee runtime exceeds this
10269 threshold (in precent), the function can be inlined regardless the limit on
10270 @option{--param max-inline-insns-single} and @option{--param
10271 max-inline-insns-auto}.
10273 @item large-function-insns
10274 The limit specifying really large functions. For functions larger than this
10275 limit after inlining, inlining is constrained by
10276 @option{--param large-function-growth}. This parameter is useful primarily
10277 to avoid extreme compilation time caused by non-linear algorithms used by the
10279 The default value is 2700.
10281 @item large-function-growth
10282 Specifies maximal growth of large function caused by inlining in percents.
10283 The default value is 100 which limits large function growth to 2.0 times
10286 @item large-unit-insns
10287 The limit specifying large translation unit. Growth caused by inlining of
10288 units larger than this limit is limited by @option{--param inline-unit-growth}.
10289 For small units this might be too tight.
10290 For example, consider a unit consisting of function A
10291 that is inline and B that just calls A three times. If B is small relative to
10292 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10293 large units consisting of small inlineable functions, however, the overall unit
10294 growth limit is needed to avoid exponential explosion of code size. Thus for
10295 smaller units, the size is increased to @option{--param large-unit-insns}
10296 before applying @option{--param inline-unit-growth}. The default is 10000.
10298 @item inline-unit-growth
10299 Specifies maximal overall growth of the compilation unit caused by inlining.
10300 The default value is 20 which limits unit growth to 1.2 times the original
10301 size. Cold functions (either marked cold via an attribute or by profile
10302 feedback) are not accounted into the unit size.
10304 @item ipcp-unit-growth
10305 Specifies maximal overall growth of the compilation unit caused by
10306 interprocedural constant propagation. The default value is 10 which limits
10307 unit growth to 1.1 times the original size.
10309 @item large-stack-frame
10310 The limit specifying large stack frames. While inlining the algorithm is trying
10311 to not grow past this limit too much. The default value is 256 bytes.
10313 @item large-stack-frame-growth
10314 Specifies maximal growth of large stack frames caused by inlining in percents.
10315 The default value is 1000 which limits large stack frame growth to 11 times
10318 @item max-inline-insns-recursive
10319 @itemx max-inline-insns-recursive-auto
10320 Specifies the maximum number of instructions an out-of-line copy of a
10321 self-recursive inline
10322 function can grow into by performing recursive inlining.
10324 @option{--param max-inline-insns-recursive} applies to functions
10326 For functions not declared inline, recursive inlining
10327 happens only when @option{-finline-functions} (included in @option{-O3}) is
10328 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10329 default value is 450.
10331 @item max-inline-recursive-depth
10332 @itemx max-inline-recursive-depth-auto
10333 Specifies the maximum recursion depth used for recursive inlining.
10335 @option{--param max-inline-recursive-depth} applies to functions
10336 declared inline. For functions not declared inline, recursive inlining
10337 happens only when @option{-finline-functions} (included in @option{-O3}) is
10338 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10339 default value is 8.
10341 @item min-inline-recursive-probability
10342 Recursive inlining is profitable only for function having deep recursion
10343 in average and can hurt for function having little recursion depth by
10344 increasing the prologue size or complexity of function body to other
10347 When profile feedback is available (see @option{-fprofile-generate}) the actual
10348 recursion depth can be guessed from probability that function recurses via a
10349 given call expression. This parameter limits inlining only to call expressions
10350 whose probability exceeds the given threshold (in percents).
10351 The default value is 10.
10353 @item early-inlining-insns
10354 Specify growth that the early inliner can make. In effect it increases
10355 the amount of inlining for code having a large abstraction penalty.
10356 The default value is 14.
10358 @item max-early-inliner-iterations
10359 Limit of iterations of the early inliner. This basically bounds
10360 the number of nested indirect calls the early inliner can resolve.
10361 Deeper chains are still handled by late inlining.
10363 @item comdat-sharing-probability
10364 Probability (in percent) that C++ inline function with comdat visibility
10365 are shared across multiple compilation units. The default value is 20.
10367 @item profile-func-internal-id
10368 A parameter to control whether to use function internal id in profile
10369 database lookup. If the value is 0, the compiler uses an id that
10370 is based on function assembler name and filename, which makes old profile
10371 data more tolerant to source changes such as function reordering etc.
10372 The default value is 0.
10374 @item min-vect-loop-bound
10375 The minimum number of iterations under which loops are not vectorized
10376 when @option{-ftree-vectorize} is used. The number of iterations after
10377 vectorization needs to be greater than the value specified by this option
10378 to allow vectorization. The default value is 0.
10380 @item gcse-cost-distance-ratio
10381 Scaling factor in calculation of maximum distance an expression
10382 can be moved by GCSE optimizations. This is currently supported only in the
10383 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10384 is with simple expressions, i.e., the expressions that have cost
10385 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10386 hoisting of simple expressions. The default value is 10.
10388 @item gcse-unrestricted-cost
10389 Cost, roughly measured as the cost of a single typical machine
10390 instruction, at which GCSE optimizations do not constrain
10391 the distance an expression can travel. This is currently
10392 supported only in the code hoisting pass. The lesser the cost,
10393 the more aggressive code hoisting is. Specifying 0
10394 allows all expressions to travel unrestricted distances.
10395 The default value is 3.
10397 @item max-hoist-depth
10398 The depth of search in the dominator tree for expressions to hoist.
10399 This is used to avoid quadratic behavior in hoisting algorithm.
10400 The value of 0 does not limit on the search, but may slow down compilation
10401 of huge functions. The default value is 30.
10403 @item max-tail-merge-comparisons
10404 The maximum amount of similar bbs to compare a bb with. This is used to
10405 avoid quadratic behavior in tree tail merging. The default value is 10.
10407 @item max-tail-merge-iterations
10408 The maximum amount of iterations of the pass over the function. This is used to
10409 limit compilation time in tree tail merging. The default value is 2.
10411 @item max-unrolled-insns
10412 The maximum number of instructions that a loop may have to be unrolled.
10413 If a loop is unrolled, this parameter also determines how many times
10414 the loop code is unrolled.
10416 @item max-average-unrolled-insns
10417 The maximum number of instructions biased by probabilities of their execution
10418 that a loop may have to be unrolled. If a loop is unrolled,
10419 this parameter also determines how many times the loop code is unrolled.
10421 @item max-unroll-times
10422 The maximum number of unrollings of a single loop.
10424 @item max-peeled-insns
10425 The maximum number of instructions that a loop may have to be peeled.
10426 If a loop is peeled, this parameter also determines how many times
10427 the loop code is peeled.
10429 @item max-peel-times
10430 The maximum number of peelings of a single loop.
10432 @item max-peel-branches
10433 The maximum number of branches on the hot path through the peeled sequence.
10435 @item max-completely-peeled-insns
10436 The maximum number of insns of a completely peeled loop.
10438 @item max-completely-peel-times
10439 The maximum number of iterations of a loop to be suitable for complete peeling.
10441 @item max-completely-peel-loop-nest-depth
10442 The maximum depth of a loop nest suitable for complete peeling.
10444 @item max-unswitch-insns
10445 The maximum number of insns of an unswitched loop.
10447 @item max-unswitch-level
10448 The maximum number of branches unswitched in a single loop.
10450 @item lim-expensive
10451 The minimum cost of an expensive expression in the loop invariant motion.
10453 @item iv-consider-all-candidates-bound
10454 Bound on number of candidates for induction variables, below which
10455 all candidates are considered for each use in induction variable
10456 optimizations. If there are more candidates than this,
10457 only the most relevant ones are considered to avoid quadratic time complexity.
10459 @item iv-max-considered-uses
10460 The induction variable optimizations give up on loops that contain more
10461 induction variable uses.
10463 @item iv-always-prune-cand-set-bound
10464 If the number of candidates in the set is smaller than this value,
10465 always try to remove unnecessary ivs from the set
10466 when adding a new one.
10468 @item scev-max-expr-size
10469 Bound on size of expressions used in the scalar evolutions analyzer.
10470 Large expressions slow the analyzer.
10472 @item scev-max-expr-complexity
10473 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10474 Complex expressions slow the analyzer.
10476 @item omega-max-vars
10477 The maximum number of variables in an Omega constraint system.
10478 The default value is 128.
10480 @item omega-max-geqs
10481 The maximum number of inequalities in an Omega constraint system.
10482 The default value is 256.
10484 @item omega-max-eqs
10485 The maximum number of equalities in an Omega constraint system.
10486 The default value is 128.
10488 @item omega-max-wild-cards
10489 The maximum number of wildcard variables that the Omega solver is
10490 able to insert. The default value is 18.
10492 @item omega-hash-table-size
10493 The size of the hash table in the Omega solver. The default value is
10496 @item omega-max-keys
10497 The maximal number of keys used by the Omega solver. The default
10500 @item omega-eliminate-redundant-constraints
10501 When set to 1, use expensive methods to eliminate all redundant
10502 constraints. The default value is 0.
10504 @item vect-max-version-for-alignment-checks
10505 The maximum number of run-time checks that can be performed when
10506 doing loop versioning for alignment in the vectorizer.
10508 @item vect-max-version-for-alias-checks
10509 The maximum number of run-time checks that can be performed when
10510 doing loop versioning for alias in the vectorizer.
10512 @item vect-max-peeling-for-alignment
10513 The maximum number of loop peels to enhance access alignment
10514 for vectorizer. Value -1 means 'no limit'.
10516 @item max-iterations-to-track
10517 The maximum number of iterations of a loop the brute-force algorithm
10518 for analysis of the number of iterations of the loop tries to evaluate.
10520 @item hot-bb-count-ws-permille
10521 A basic block profile count is considered hot if it contributes to
10522 the given permillage (i.e. 0...1000) of the entire profiled execution.
10524 @item hot-bb-frequency-fraction
10525 Select fraction of the entry block frequency of executions of basic block in
10526 function given basic block needs to have to be considered hot.
10528 @item max-predicted-iterations
10529 The maximum number of loop iterations we predict statically. This is useful
10530 in cases where a function contains a single loop with known bound and
10531 another loop with unknown bound.
10532 The known number of iterations is predicted correctly, while
10533 the unknown number of iterations average to roughly 10. This means that the
10534 loop without bounds appears artificially cold relative to the other one.
10536 @item builtin-expect-probability
10537 Control the probability of the expression having the specified value. This
10538 parameter takes a percentage (i.e. 0 ... 100) as input.
10539 The default probability of 90 is obtained empirically.
10541 @item align-threshold
10543 Select fraction of the maximal frequency of executions of a basic block in
10544 a function to align the basic block.
10546 @item align-loop-iterations
10548 A loop expected to iterate at least the selected number of iterations is
10551 @item tracer-dynamic-coverage
10552 @itemx tracer-dynamic-coverage-feedback
10554 This value is used to limit superblock formation once the given percentage of
10555 executed instructions is covered. This limits unnecessary code size
10558 The @option{tracer-dynamic-coverage-feedback} parameter
10559 is used only when profile
10560 feedback is available. The real profiles (as opposed to statically estimated
10561 ones) are much less balanced allowing the threshold to be larger value.
10563 @item tracer-max-code-growth
10564 Stop tail duplication once code growth has reached given percentage. This is
10565 a rather artificial limit, as most of the duplicates are eliminated later in
10566 cross jumping, so it may be set to much higher values than is the desired code
10569 @item tracer-min-branch-ratio
10571 Stop reverse growth when the reverse probability of best edge is less than this
10572 threshold (in percent).
10574 @item tracer-min-branch-ratio
10575 @itemx tracer-min-branch-ratio-feedback
10577 Stop forward growth if the best edge has probability lower than this
10580 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10581 compilation for profile feedback and one for compilation without. The value
10582 for compilation with profile feedback needs to be more conservative (higher) in
10583 order to make tracer effective.
10585 @item max-cse-path-length
10587 The maximum number of basic blocks on path that CSE considers.
10590 @item max-cse-insns
10591 The maximum number of instructions CSE processes before flushing.
10592 The default is 1000.
10594 @item ggc-min-expand
10596 GCC uses a garbage collector to manage its own memory allocation. This
10597 parameter specifies the minimum percentage by which the garbage
10598 collector's heap should be allowed to expand between collections.
10599 Tuning this may improve compilation speed; it has no effect on code
10602 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10603 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10604 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10605 GCC is not able to calculate RAM on a particular platform, the lower
10606 bound of 30% is used. Setting this parameter and
10607 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10608 every opportunity. This is extremely slow, but can be useful for
10611 @item ggc-min-heapsize
10613 Minimum size of the garbage collector's heap before it begins bothering
10614 to collect garbage. The first collection occurs after the heap expands
10615 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10616 tuning this may improve compilation speed, and has no effect on code
10619 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10620 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10621 with a lower bound of 4096 (four megabytes) and an upper bound of
10622 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10623 particular platform, the lower bound is used. Setting this parameter
10624 very large effectively disables garbage collection. Setting this
10625 parameter and @option{ggc-min-expand} to zero causes a full collection
10626 to occur at every opportunity.
10628 @item max-reload-search-insns
10629 The maximum number of instruction reload should look backward for equivalent
10630 register. Increasing values mean more aggressive optimization, making the
10631 compilation time increase with probably slightly better performance.
10632 The default value is 100.
10634 @item max-cselib-memory-locations
10635 The maximum number of memory locations cselib should take into account.
10636 Increasing values mean more aggressive optimization, making the compilation time
10637 increase with probably slightly better performance. The default value is 500.
10639 @item reorder-blocks-duplicate
10640 @itemx reorder-blocks-duplicate-feedback
10642 Used by the basic block reordering pass to decide whether to use unconditional
10643 branch or duplicate the code on its destination. Code is duplicated when its
10644 estimated size is smaller than this value multiplied by the estimated size of
10645 unconditional jump in the hot spots of the program.
10647 The @option{reorder-block-duplicate-feedback} parameter
10648 is used only when profile
10649 feedback is available. It may be set to higher values than
10650 @option{reorder-block-duplicate} since information about the hot spots is more
10653 @item max-sched-ready-insns
10654 The maximum number of instructions ready to be issued the scheduler should
10655 consider at any given time during the first scheduling pass. Increasing
10656 values mean more thorough searches, making the compilation time increase
10657 with probably little benefit. The default value is 100.
10659 @item max-sched-region-blocks
10660 The maximum number of blocks in a region to be considered for
10661 interblock scheduling. The default value is 10.
10663 @item max-pipeline-region-blocks
10664 The maximum number of blocks in a region to be considered for
10665 pipelining in the selective scheduler. The default value is 15.
10667 @item max-sched-region-insns
10668 The maximum number of insns in a region to be considered for
10669 interblock scheduling. The default value is 100.
10671 @item max-pipeline-region-insns
10672 The maximum number of insns in a region to be considered for
10673 pipelining in the selective scheduler. The default value is 200.
10675 @item min-spec-prob
10676 The minimum probability (in percents) of reaching a source block
10677 for interblock speculative scheduling. The default value is 40.
10679 @item max-sched-extend-regions-iters
10680 The maximum number of iterations through CFG to extend regions.
10681 A value of 0 (the default) disables region extensions.
10683 @item max-sched-insn-conflict-delay
10684 The maximum conflict delay for an insn to be considered for speculative motion.
10685 The default value is 3.
10687 @item sched-spec-prob-cutoff
10688 The minimal probability of speculation success (in percents), so that
10689 speculative insns are scheduled.
10690 The default value is 40.
10692 @item sched-spec-state-edge-prob-cutoff
10693 The minimum probability an edge must have for the scheduler to save its
10695 The default value is 10.
10697 @item sched-mem-true-dep-cost
10698 Minimal distance (in CPU cycles) between store and load targeting same
10699 memory locations. The default value is 1.
10701 @item selsched-max-lookahead
10702 The maximum size of the lookahead window of selective scheduling. It is a
10703 depth of search for available instructions.
10704 The default value is 50.
10706 @item selsched-max-sched-times
10707 The maximum number of times that an instruction is scheduled during
10708 selective scheduling. This is the limit on the number of iterations
10709 through which the instruction may be pipelined. The default value is 2.
10711 @item selsched-max-insns-to-rename
10712 The maximum number of best instructions in the ready list that are considered
10713 for renaming in the selective scheduler. The default value is 2.
10716 The minimum value of stage count that swing modulo scheduler
10717 generates. The default value is 2.
10719 @item max-last-value-rtl
10720 The maximum size measured as number of RTLs that can be recorded in an expression
10721 in combiner for a pseudo register as last known value of that register. The default
10724 @item max-combine-insns
10725 The maximum number of instructions the RTL combiner tries to combine.
10726 The default value is 2 at @option{-Og} and 4 otherwise.
10728 @item integer-share-limit
10729 Small integer constants can use a shared data structure, reducing the
10730 compiler's memory usage and increasing its speed. This sets the maximum
10731 value of a shared integer constant. The default value is 256.
10733 @item ssp-buffer-size
10734 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10735 protection when @option{-fstack-protection} is used.
10737 @item min-size-for-stack-sharing
10738 The minimum size of variables taking part in stack slot sharing when not
10739 optimizing. The default value is 32.
10741 @item max-jump-thread-duplication-stmts
10742 Maximum number of statements allowed in a block that needs to be
10743 duplicated when threading jumps.
10745 @item max-fields-for-field-sensitive
10746 Maximum number of fields in a structure treated in
10747 a field sensitive manner during pointer analysis. The default is zero
10748 for @option{-O0} and @option{-O1},
10749 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10751 @item prefetch-latency
10752 Estimate on average number of instructions that are executed before
10753 prefetch finishes. The distance prefetched ahead is proportional
10754 to this constant. Increasing this number may also lead to less
10755 streams being prefetched (see @option{simultaneous-prefetches}).
10757 @item simultaneous-prefetches
10758 Maximum number of prefetches that can run at the same time.
10760 @item l1-cache-line-size
10761 The size of cache line in L1 cache, in bytes.
10763 @item l1-cache-size
10764 The size of L1 cache, in kilobytes.
10766 @item l2-cache-size
10767 The size of L2 cache, in kilobytes.
10769 @item min-insn-to-prefetch-ratio
10770 The minimum ratio between the number of instructions and the
10771 number of prefetches to enable prefetching in a loop.
10773 @item prefetch-min-insn-to-mem-ratio
10774 The minimum ratio between the number of instructions and the
10775 number of memory references to enable prefetching in a loop.
10777 @item use-canonical-types
10778 Whether the compiler should use the ``canonical'' type system. By
10779 default, this should always be 1, which uses a more efficient internal
10780 mechanism for comparing types in C++ and Objective-C++. However, if
10781 bugs in the canonical type system are causing compilation failures,
10782 set this value to 0 to disable canonical types.
10784 @item switch-conversion-max-branch-ratio
10785 Switch initialization conversion refuses to create arrays that are
10786 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10787 branches in the switch.
10789 @item max-partial-antic-length
10790 Maximum length of the partial antic set computed during the tree
10791 partial redundancy elimination optimization (@option{-ftree-pre}) when
10792 optimizing at @option{-O3} and above. For some sorts of source code
10793 the enhanced partial redundancy elimination optimization can run away,
10794 consuming all of the memory available on the host machine. This
10795 parameter sets a limit on the length of the sets that are computed,
10796 which prevents the runaway behavior. Setting a value of 0 for
10797 this parameter allows an unlimited set length.
10799 @item sccvn-max-scc-size
10800 Maximum size of a strongly connected component (SCC) during SCCVN
10801 processing. If this limit is hit, SCCVN processing for the whole
10802 function is not done and optimizations depending on it are
10803 disabled. The default maximum SCC size is 10000.
10805 @item sccvn-max-alias-queries-per-access
10806 Maximum number of alias-oracle queries we perform when looking for
10807 redundancies for loads and stores. If this limit is hit the search
10808 is aborted and the load or store is not considered redundant. The
10809 number of queries is algorithmically limited to the number of
10810 stores on all paths from the load to the function entry.
10811 The default maxmimum number of queries is 1000.
10813 @item ira-max-loops-num
10814 IRA uses regional register allocation by default. If a function
10815 contains more loops than the number given by this parameter, only at most
10816 the given number of the most frequently-executed loops form regions
10817 for regional register allocation. The default value of the
10820 @item ira-max-conflict-table-size
10821 Although IRA uses a sophisticated algorithm to compress the conflict
10822 table, the table can still require excessive amounts of memory for
10823 huge functions. If the conflict table for a function could be more
10824 than the size in MB given by this parameter, the register allocator
10825 instead uses a faster, simpler, and lower-quality
10826 algorithm that does not require building a pseudo-register conflict table.
10827 The default value of the parameter is 2000.
10829 @item ira-loop-reserved-regs
10830 IRA can be used to evaluate more accurate register pressure in loops
10831 for decisions to move loop invariants (see @option{-O3}). The number
10832 of available registers reserved for some other purposes is given
10833 by this parameter. The default value of the parameter is 2, which is
10834 the minimal number of registers needed by typical instructions.
10835 This value is the best found from numerous experiments.
10837 @item lra-inheritance-ebb-probability-cutoff
10838 LRA tries to reuse values reloaded in registers in subsequent insns.
10839 This optimization is called inheritance. EBB is used as a region to
10840 do this optimization. The parameter defines a minimal fall-through
10841 edge probability in percentage used to add BB to inheritance EBB in
10842 LRA. The default value of the parameter is 40. The value was chosen
10843 from numerous runs of SPEC2000 on x86-64.
10845 @item loop-invariant-max-bbs-in-loop
10846 Loop invariant motion can be very expensive, both in compilation time and
10847 in amount of needed compile-time memory, with very large loops. Loops
10848 with more basic blocks than this parameter won't have loop invariant
10849 motion optimization performed on them. The default value of the
10850 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10852 @item loop-max-datarefs-for-datadeps
10853 Building data dapendencies is expensive for very large loops. This
10854 parameter limits the number of data references in loops that are
10855 considered for data dependence analysis. These large loops are no
10856 handled by the optimizations using loop data dependencies.
10857 The default value is 1000.
10859 @item max-vartrack-size
10860 Sets a maximum number of hash table slots to use during variable
10861 tracking dataflow analysis of any function. If this limit is exceeded
10862 with variable tracking at assignments enabled, analysis for that
10863 function is retried without it, after removing all debug insns from
10864 the function. If the limit is exceeded even without debug insns, var
10865 tracking analysis is completely disabled for the function. Setting
10866 the parameter to zero makes it unlimited.
10868 @item max-vartrack-expr-depth
10869 Sets a maximum number of recursion levels when attempting to map
10870 variable names or debug temporaries to value expressions. This trades
10871 compilation time for more complete debug information. If this is set too
10872 low, value expressions that are available and could be represented in
10873 debug information may end up not being used; setting this higher may
10874 enable the compiler to find more complex debug expressions, but compile
10875 time and memory use may grow. The default is 12.
10877 @item min-nondebug-insn-uid
10878 Use uids starting at this parameter for nondebug insns. The range below
10879 the parameter is reserved exclusively for debug insns created by
10880 @option{-fvar-tracking-assignments}, but debug insns may get
10881 (non-overlapping) uids above it if the reserved range is exhausted.
10883 @item ipa-sra-ptr-growth-factor
10884 IPA-SRA replaces a pointer to an aggregate with one or more new
10885 parameters only when their cumulative size is less or equal to
10886 @option{ipa-sra-ptr-growth-factor} times the size of the original
10889 @item sra-max-scalarization-size-Ospeed
10890 @item sra-max-scalarization-size-Osize
10891 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10892 replace scalar parts of aggregates with uses of independent scalar
10893 variables. These parameters control the maximum size, in storage units,
10894 of aggregate which is considered for replacement when compiling for
10896 (@option{sra-max-scalarization-size-Ospeed}) or size
10897 (@option{sra-max-scalarization-size-Osize}) respectively.
10899 @item tm-max-aggregate-size
10900 When making copies of thread-local variables in a transaction, this
10901 parameter specifies the size in bytes after which variables are
10902 saved with the logging functions as opposed to save/restore code
10903 sequence pairs. This option only applies when using
10906 @item graphite-max-nb-scop-params
10907 To avoid exponential effects in the Graphite loop transforms, the
10908 number of parameters in a Static Control Part (SCoP) is bounded. The
10909 default value is 10 parameters. A variable whose value is unknown at
10910 compilation time and defined outside a SCoP is a parameter of the SCoP.
10912 @item graphite-max-bbs-per-function
10913 To avoid exponential effects in the detection of SCoPs, the size of
10914 the functions analyzed by Graphite is bounded. The default value is
10917 @item loop-block-tile-size
10918 Loop blocking or strip mining transforms, enabled with
10919 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10920 loop in the loop nest by a given number of iterations. The strip
10921 length can be changed using the @option{loop-block-tile-size}
10922 parameter. The default value is 51 iterations.
10924 @item loop-unroll-jam-size
10925 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10926 default value is 4.
10928 @item loop-unroll-jam-depth
10929 Specify the dimension to be unrolled (counting from the most inner loop)
10930 for the @option{-floop-unroll-and-jam}. The default value is 2.
10932 @item ipa-cp-value-list-size
10933 IPA-CP attempts to track all possible values and types passed to a function's
10934 parameter in order to propagate them and perform devirtualization.
10935 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10936 stores per one formal parameter of a function.
10938 @item ipa-cp-eval-threshold
10939 IPA-CP calculates its own score of cloning profitability heuristics
10940 and performs those cloning opportunities with scores that exceed
10941 @option{ipa-cp-eval-threshold}.
10943 @item ipa-cp-recursion-penalty
10944 Percentage penalty the recursive functions will receive when they
10945 are evaluated for cloning.
10947 @item ipa-cp-single-call-penalty
10948 Percentage penalty functions containg a single call to another
10949 function will receive when they are evaluated for cloning.
10952 @item ipa-max-agg-items
10953 IPA-CP is also capable to propagate a number of scalar values passed
10954 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10955 number of such values per one parameter.
10957 @item ipa-cp-loop-hint-bonus
10958 When IPA-CP determines that a cloning candidate would make the number
10959 of iterations of a loop known, it adds a bonus of
10960 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10963 @item ipa-cp-array-index-hint-bonus
10964 When IPA-CP determines that a cloning candidate would make the index of
10965 an array access known, it adds a bonus of
10966 @option{ipa-cp-array-index-hint-bonus} to the profitability
10967 score of the candidate.
10969 @item ipa-max-aa-steps
10970 During its analysis of function bodies, IPA-CP employs alias analysis
10971 in order to track values pointed to by function parameters. In order
10972 not spend too much time analyzing huge functions, it gives up and
10973 consider all memory clobbered after examining
10974 @option{ipa-max-aa-steps} statements modifying memory.
10976 @item lto-partitions
10977 Specify desired number of partitions produced during WHOPR compilation.
10978 The number of partitions should exceed the number of CPUs used for compilation.
10979 The default value is 32.
10981 @item lto-minpartition
10982 Size of minimal partition for WHOPR (in estimated instructions).
10983 This prevents expenses of splitting very small programs into too many
10986 @item cxx-max-namespaces-for-diagnostic-help
10987 The maximum number of namespaces to consult for suggestions when C++
10988 name lookup fails for an identifier. The default is 1000.
10990 @item sink-frequency-threshold
10991 The maximum relative execution frequency (in percents) of the target block
10992 relative to a statement's original block to allow statement sinking of a
10993 statement. Larger numbers result in more aggressive statement sinking.
10994 The default value is 75. A small positive adjustment is applied for
10995 statements with memory operands as those are even more profitable so sink.
10997 @item max-stores-to-sink
10998 The maximum number of conditional stores paires that can be sunk. Set to 0
10999 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11000 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
11002 @item allow-store-data-races
11003 Allow optimizers to introduce new data races on stores.
11004 Set to 1 to allow, otherwise to 0. This option is enabled by default
11005 at optimization level @option{-Ofast}.
11007 @item case-values-threshold
11008 The smallest number of different values for which it is best to use a
11009 jump-table instead of a tree of conditional branches. If the value is
11010 0, use the default for the machine. The default is 0.
11012 @item tree-reassoc-width
11013 Set the maximum number of instructions executed in parallel in
11014 reassociated tree. This parameter overrides target dependent
11015 heuristics used by default if has non zero value.
11017 @item sched-pressure-algorithm
11018 Choose between the two available implementations of
11019 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11020 and is the more likely to prevent instructions from being reordered.
11021 Algorithm 2 was designed to be a compromise between the relatively
11022 conservative approach taken by algorithm 1 and the rather aggressive
11023 approach taken by the default scheduler. It relies more heavily on
11024 having a regular register file and accurate register pressure classes.
11025 See @file{haifa-sched.c} in the GCC sources for more details.
11027 The default choice depends on the target.
11029 @item max-slsr-cand-scan
11030 Set the maximum number of existing candidates that are considered when
11031 seeking a basis for a new straight-line strength reduction candidate.
11034 Enable buffer overflow detection for global objects. This kind
11035 of protection is enabled by default if you are using
11036 @option{-fsanitize=address} option.
11037 To disable global objects protection use @option{--param asan-globals=0}.
11040 Enable buffer overflow detection for stack objects. This kind of
11041 protection is enabled by default when using@option{-fsanitize=address}.
11042 To disable stack protection use @option{--param asan-stack=0} option.
11044 @item asan-instrument-reads
11045 Enable buffer overflow detection for memory reads. This kind of
11046 protection is enabled by default when using @option{-fsanitize=address}.
11047 To disable memory reads protection use
11048 @option{--param asan-instrument-reads=0}.
11050 @item asan-instrument-writes
11051 Enable buffer overflow detection for memory writes. This kind of
11052 protection is enabled by default when using @option{-fsanitize=address}.
11053 To disable memory writes protection use
11054 @option{--param asan-instrument-writes=0} option.
11056 @item asan-memintrin
11057 Enable detection for built-in functions. This kind of protection
11058 is enabled by default when using @option{-fsanitize=address}.
11059 To disable built-in functions protection use
11060 @option{--param asan-memintrin=0}.
11062 @item asan-use-after-return
11063 Enable detection of use-after-return. This kind of protection
11064 is enabled by default when using @option{-fsanitize=address} option.
11065 To disable use-after-return detection use
11066 @option{--param asan-use-after-return=0}.
11068 @item asan-instrumentation-with-call-threshold
11069 If number of memory accesses in function being instrumented
11070 is greater or equal to this number, use callbacks instead of inline checks.
11071 E.g. to disable inline code use
11072 @option{--param asan-instrumentation-with-call-threshold=0}.
11074 @item chkp-max-ctor-size
11075 Static constructors generated by Pointer Bounds Checker may become very
11076 large and significantly increase compile time at optimization level
11077 @option{-O1} and higher. This parameter is a maximum nubmer of statements
11078 in a single generated constructor. Default value is 5000.
11080 @item max-fsm-thread-path-insns
11081 Maximum number of instructions to copy when duplicating blocks on a
11082 finite state automaton jump thread path. The default is 100.
11084 @item max-fsm-thread-length
11085 Maximum number of basic blocks on a finite state automaton jump thread
11086 path. The default is 10.
11088 @item max-fsm-thread-paths
11089 Maximum number of new jump thread paths to create for a finite state
11090 automaton. The default is 50.
11095 @node Preprocessor Options
11096 @section Options Controlling the Preprocessor
11097 @cindex preprocessor options
11098 @cindex options, preprocessor
11100 These options control the C preprocessor, which is run on each C source
11101 file before actual compilation.
11103 If you use the @option{-E} option, nothing is done except preprocessing.
11104 Some of these options make sense only together with @option{-E} because
11105 they cause the preprocessor output to be unsuitable for actual
11109 @item -Wp,@var{option}
11111 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11112 and pass @var{option} directly through to the preprocessor. If
11113 @var{option} contains commas, it is split into multiple options at the
11114 commas. However, many options are modified, translated or interpreted
11115 by the compiler driver before being passed to the preprocessor, and
11116 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11117 interface is undocumented and subject to change, so whenever possible
11118 you should avoid using @option{-Wp} and let the driver handle the
11121 @item -Xpreprocessor @var{option}
11122 @opindex Xpreprocessor
11123 Pass @var{option} as an option to the preprocessor. You can use this to
11124 supply system-specific preprocessor options that GCC does not
11127 If you want to pass an option that takes an argument, you must use
11128 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11130 @item -no-integrated-cpp
11131 @opindex no-integrated-cpp
11132 Perform preprocessing as a separate pass before compilation.
11133 By default, GCC performs preprocessing as an integrated part of
11134 input tokenization and parsing.
11135 If this option is provided, the appropriate language front end
11136 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11137 and Objective-C, respectively) is instead invoked twice,
11138 once for preprocessing only and once for actual compilation
11139 of the preprocessed input.
11140 This option may be useful in conjunction with the @option{-B} or
11141 @option{-wrapper} options to specify an alternate preprocessor or
11142 perform additional processing of the program source between
11143 normal preprocessing and compilation.
11146 @include cppopts.texi
11148 @node Assembler Options
11149 @section Passing Options to the Assembler
11151 @c prevent bad page break with this line
11152 You can pass options to the assembler.
11155 @item -Wa,@var{option}
11157 Pass @var{option} as an option to the assembler. If @var{option}
11158 contains commas, it is split into multiple options at the commas.
11160 @item -Xassembler @var{option}
11161 @opindex Xassembler
11162 Pass @var{option} as an option to the assembler. You can use this to
11163 supply system-specific assembler options that GCC does not
11166 If you want to pass an option that takes an argument, you must use
11167 @option{-Xassembler} twice, once for the option and once for the argument.
11172 @section Options for Linking
11173 @cindex link options
11174 @cindex options, linking
11176 These options come into play when the compiler links object files into
11177 an executable output file. They are meaningless if the compiler is
11178 not doing a link step.
11182 @item @var{object-file-name}
11183 A file name that does not end in a special recognized suffix is
11184 considered to name an object file or library. (Object files are
11185 distinguished from libraries by the linker according to the file
11186 contents.) If linking is done, these object files are used as input
11195 If any of these options is used, then the linker is not run, and
11196 object file names should not be used as arguments. @xref{Overall
11200 @opindex fuse-ld=bfd
11201 Use the @command{bfd} linker instead of the default linker.
11203 @item -fuse-ld=gold
11204 @opindex fuse-ld=gold
11205 Use the @command{gold} linker instead of the default linker.
11208 @item -l@var{library}
11209 @itemx -l @var{library}
11211 Search the library named @var{library} when linking. (The second
11212 alternative with the library as a separate argument is only for
11213 POSIX compliance and is not recommended.)
11215 It makes a difference where in the command you write this option; the
11216 linker searches and processes libraries and object files in the order they
11217 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11218 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11219 to functions in @samp{z}, those functions may not be loaded.
11221 The linker searches a standard list of directories for the library,
11222 which is actually a file named @file{lib@var{library}.a}. The linker
11223 then uses this file as if it had been specified precisely by name.
11225 The directories searched include several standard system directories
11226 plus any that you specify with @option{-L}.
11228 Normally the files found this way are library files---archive files
11229 whose members are object files. The linker handles an archive file by
11230 scanning through it for members which define symbols that have so far
11231 been referenced but not defined. But if the file that is found is an
11232 ordinary object file, it is linked in the usual fashion. The only
11233 difference between using an @option{-l} option and specifying a file name
11234 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11235 and searches several directories.
11239 You need this special case of the @option{-l} option in order to
11240 link an Objective-C or Objective-C++ program.
11242 @item -nostartfiles
11243 @opindex nostartfiles
11244 Do not use the standard system startup files when linking.
11245 The standard system libraries are used normally, unless @option{-nostdlib}
11246 or @option{-nodefaultlibs} is used.
11248 @item -nodefaultlibs
11249 @opindex nodefaultlibs
11250 Do not use the standard system libraries when linking.
11251 Only the libraries you specify are passed to the linker, and options
11252 specifying linkage of the system libraries, such as @option{-static-libgcc}
11253 or @option{-shared-libgcc}, are ignored.
11254 The standard startup files are used normally, unless @option{-nostartfiles}
11257 The compiler may generate calls to @code{memcmp},
11258 @code{memset}, @code{memcpy} and @code{memmove}.
11259 These entries are usually resolved by entries in
11260 libc. These entry points should be supplied through some other
11261 mechanism when this option is specified.
11265 Do not use the standard system startup files or libraries when linking.
11266 No startup files and only the libraries you specify are passed to
11267 the linker, and options specifying linkage of the system libraries, such as
11268 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11270 The compiler may generate calls to @code{memcmp}, @code{memset},
11271 @code{memcpy} and @code{memmove}.
11272 These entries are usually resolved by entries in
11273 libc. These entry points should be supplied through some other
11274 mechanism when this option is specified.
11276 @cindex @option{-lgcc}, use with @option{-nostdlib}
11277 @cindex @option{-nostdlib} and unresolved references
11278 @cindex unresolved references and @option{-nostdlib}
11279 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11280 @cindex @option{-nodefaultlibs} and unresolved references
11281 @cindex unresolved references and @option{-nodefaultlibs}
11282 One of the standard libraries bypassed by @option{-nostdlib} and
11283 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11284 which GCC uses to overcome shortcomings of particular machines, or special
11285 needs for some languages.
11286 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11287 Collection (GCC) Internals},
11288 for more discussion of @file{libgcc.a}.)
11289 In most cases, you need @file{libgcc.a} even when you want to avoid
11290 other standard libraries. In other words, when you specify @option{-nostdlib}
11291 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11292 This ensures that you have no unresolved references to internal GCC
11293 library subroutines.
11294 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11295 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11296 GNU Compiler Collection (GCC) Internals}.)
11300 Produce a position independent executable on targets that support it.
11301 For predictable results, you must also specify the same set of options
11302 used for compilation (@option{-fpie}, @option{-fPIE},
11303 or model suboptions) when you specify this linker option.
11307 Don't produce a position independent executable.
11311 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11312 that support it. This instructs the linker to add all symbols, not
11313 only used ones, to the dynamic symbol table. This option is needed
11314 for some uses of @code{dlopen} or to allow obtaining backtraces
11315 from within a program.
11319 Remove all symbol table and relocation information from the executable.
11323 On systems that support dynamic linking, this prevents linking with the shared
11324 libraries. On other systems, this option has no effect.
11328 Produce a shared object which can then be linked with other objects to
11329 form an executable. Not all systems support this option. For predictable
11330 results, you must also specify the same set of options used for compilation
11331 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11332 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11333 needs to build supplementary stub code for constructors to work. On
11334 multi-libbed systems, @samp{gcc -shared} must select the correct support
11335 libraries to link against. Failing to supply the correct flags may lead
11336 to subtle defects. Supplying them in cases where they are not necessary
11339 @item -shared-libgcc
11340 @itemx -static-libgcc
11341 @opindex shared-libgcc
11342 @opindex static-libgcc
11343 On systems that provide @file{libgcc} as a shared library, these options
11344 force the use of either the shared or static version, respectively.
11345 If no shared version of @file{libgcc} was built when the compiler was
11346 configured, these options have no effect.
11348 There are several situations in which an application should use the
11349 shared @file{libgcc} instead of the static version. The most common
11350 of these is when the application wishes to throw and catch exceptions
11351 across different shared libraries. In that case, each of the libraries
11352 as well as the application itself should use the shared @file{libgcc}.
11354 Therefore, the G++ and GCJ drivers automatically add
11355 @option{-shared-libgcc} whenever you build a shared library or a main
11356 executable, because C++ and Java programs typically use exceptions, so
11357 this is the right thing to do.
11359 If, instead, you use the GCC driver to create shared libraries, you may
11360 find that they are not always linked with the shared @file{libgcc}.
11361 If GCC finds, at its configuration time, that you have a non-GNU linker
11362 or a GNU linker that does not support option @option{--eh-frame-hdr},
11363 it links the shared version of @file{libgcc} into shared libraries
11364 by default. Otherwise, it takes advantage of the linker and optimizes
11365 away the linking with the shared version of @file{libgcc}, linking with
11366 the static version of libgcc by default. This allows exceptions to
11367 propagate through such shared libraries, without incurring relocation
11368 costs at library load time.
11370 However, if a library or main executable is supposed to throw or catch
11371 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11372 for the languages used in the program, or using the option
11373 @option{-shared-libgcc}, such that it is linked with the shared
11376 @item -static-libasan
11377 @opindex static-libasan
11378 When the @option{-fsanitize=address} option is used to link a program,
11379 the GCC driver automatically links against @option{libasan}. If
11380 @file{libasan} is available as a shared library, and the @option{-static}
11381 option is not used, then this links against the shared version of
11382 @file{libasan}. The @option{-static-libasan} option directs the GCC
11383 driver to link @file{libasan} statically, without necessarily linking
11384 other libraries statically.
11386 @item -static-libtsan
11387 @opindex static-libtsan
11388 When the @option{-fsanitize=thread} option is used to link a program,
11389 the GCC driver automatically links against @option{libtsan}. If
11390 @file{libtsan} is available as a shared library, and the @option{-static}
11391 option is not used, then this links against the shared version of
11392 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11393 driver to link @file{libtsan} statically, without necessarily linking
11394 other libraries statically.
11396 @item -static-liblsan
11397 @opindex static-liblsan
11398 When the @option{-fsanitize=leak} option is used to link a program,
11399 the GCC driver automatically links against @option{liblsan}. If
11400 @file{liblsan} is available as a shared library, and the @option{-static}
11401 option is not used, then this links against the shared version of
11402 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11403 driver to link @file{liblsan} statically, without necessarily linking
11404 other libraries statically.
11406 @item -static-libubsan
11407 @opindex static-libubsan
11408 When the @option{-fsanitize=undefined} option is used to link a program,
11409 the GCC driver automatically links against @option{libubsan}. If
11410 @file{libubsan} is available as a shared library, and the @option{-static}
11411 option is not used, then this links against the shared version of
11412 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11413 driver to link @file{libubsan} statically, without necessarily linking
11414 other libraries statically.
11416 @item -static-libmpx
11417 @opindex static-libmpx
11418 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11419 used to link a program, the GCC driver automatically links against
11420 @file{libmpx}. If @file{libmpx} is available as a shared library,
11421 and the @option{-static} option is not used, then this links against
11422 the shared version of @file{libmpx}. The @option{-static-libmpx}
11423 option directs the GCC driver to link @file{libmpx} statically,
11424 without necessarily linking other libraries statically.
11426 @item -static-libmpxwrappers
11427 @opindex static-libmpxwrappers
11428 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11429 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11430 GCC driver automatically links against @file{libmpxwrappers}. If
11431 @file{libmpxwrappers} is available as a shared library, and the
11432 @option{-static} option is not used, then this links against the shared
11433 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11434 option directs the GCC driver to link @file{libmpxwrappers} statically,
11435 without necessarily linking other libraries statically.
11437 @item -static-libstdc++
11438 @opindex static-libstdc++
11439 When the @command{g++} program is used to link a C++ program, it
11440 normally automatically links against @option{libstdc++}. If
11441 @file{libstdc++} is available as a shared library, and the
11442 @option{-static} option is not used, then this links against the
11443 shared version of @file{libstdc++}. That is normally fine. However, it
11444 is sometimes useful to freeze the version of @file{libstdc++} used by
11445 the program without going all the way to a fully static link. The
11446 @option{-static-libstdc++} option directs the @command{g++} driver to
11447 link @file{libstdc++} statically, without necessarily linking other
11448 libraries statically.
11452 Bind references to global symbols when building a shared object. Warn
11453 about any unresolved references (unless overridden by the link editor
11454 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11457 @item -T @var{script}
11459 @cindex linker script
11460 Use @var{script} as the linker script. This option is supported by most
11461 systems using the GNU linker. On some targets, such as bare-board
11462 targets without an operating system, the @option{-T} option may be required
11463 when linking to avoid references to undefined symbols.
11465 @item -Xlinker @var{option}
11467 Pass @var{option} as an option to the linker. You can use this to
11468 supply system-specific linker options that GCC does not recognize.
11470 If you want to pass an option that takes a separate argument, you must use
11471 @option{-Xlinker} twice, once for the option and once for the argument.
11472 For example, to pass @option{-assert definitions}, you must write
11473 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11474 @option{-Xlinker "-assert definitions"}, because this passes the entire
11475 string as a single argument, which is not what the linker expects.
11477 When using the GNU linker, it is usually more convenient to pass
11478 arguments to linker options using the @option{@var{option}=@var{value}}
11479 syntax than as separate arguments. For example, you can specify
11480 @option{-Xlinker -Map=output.map} rather than
11481 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11482 this syntax for command-line options.
11484 @item -Wl,@var{option}
11486 Pass @var{option} as an option to the linker. If @var{option} contains
11487 commas, it is split into multiple options at the commas. You can use this
11488 syntax to pass an argument to the option.
11489 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11490 linker. When using the GNU linker, you can also get the same effect with
11491 @option{-Wl,-Map=output.map}.
11493 @item -u @var{symbol}
11495 Pretend the symbol @var{symbol} is undefined, to force linking of
11496 library modules to define it. You can use @option{-u} multiple times with
11497 different symbols to force loading of additional library modules.
11499 @item -z @var{keyword}
11501 @option{-z} is passed directly on to the linker along with the keyword
11502 @var{keyword}. See the section in the documentation of your linker for
11503 permitted values and their meanings.
11506 @node Directory Options
11507 @section Options for Directory Search
11508 @cindex directory options
11509 @cindex options, directory search
11510 @cindex search path
11512 These options specify directories to search for header files, for
11513 libraries and for parts of the compiler:
11518 Add the directory @var{dir} to the head of the list of directories to be
11519 searched for header files. This can be used to override a system header
11520 file, substituting your own version, since these directories are
11521 searched before the system header file directories. However, you should
11522 not use this option to add directories that contain vendor-supplied
11523 system header files (use @option{-isystem} for that). If you use more than
11524 one @option{-I} option, the directories are scanned in left-to-right
11525 order; the standard system directories come after.
11527 If a standard system include directory, or a directory specified with
11528 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11529 option is ignored. The directory is still searched but as a
11530 system directory at its normal position in the system include chain.
11531 This is to ensure that GCC's procedure to fix buggy system headers and
11532 the ordering for the @code{include_next} directive are not inadvertently changed.
11533 If you really need to change the search order for system directories,
11534 use the @option{-nostdinc} and/or @option{-isystem} options.
11536 @item -iplugindir=@var{dir}
11537 @opindex iplugindir=
11538 Set the directory to search for plugins that are passed
11539 by @option{-fplugin=@var{name}} instead of
11540 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11541 to be used by the user, but only passed by the driver.
11543 @item -iquote@var{dir}
11545 Add the directory @var{dir} to the head of the list of directories to
11546 be searched for header files only for the case of @code{#include
11547 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11548 otherwise just like @option{-I}.
11552 Add directory @var{dir} to the list of directories to be searched
11555 @item -B@var{prefix}
11557 This option specifies where to find the executables, libraries,
11558 include files, and data files of the compiler itself.
11560 The compiler driver program runs one or more of the subprograms
11561 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11562 @var{prefix} as a prefix for each program it tries to run, both with and
11563 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11565 For each subprogram to be run, the compiler driver first tries the
11566 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11567 is not specified, the driver tries two standard prefixes,
11568 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11569 those results in a file name that is found, the unmodified program
11570 name is searched for using the directories specified in your
11571 @env{PATH} environment variable.
11573 The compiler checks to see if the path provided by @option{-B}
11574 refers to a directory, and if necessary it adds a directory
11575 separator character at the end of the path.
11577 @option{-B} prefixes that effectively specify directory names also apply
11578 to libraries in the linker, because the compiler translates these
11579 options into @option{-L} options for the linker. They also apply to
11580 include files in the preprocessor, because the compiler translates these
11581 options into @option{-isystem} options for the preprocessor. In this case,
11582 the compiler appends @samp{include} to the prefix.
11584 The runtime support file @file{libgcc.a} can also be searched for using
11585 the @option{-B} prefix, if needed. If it is not found there, the two
11586 standard prefixes above are tried, and that is all. The file is left
11587 out of the link if it is not found by those means.
11589 Another way to specify a prefix much like the @option{-B} prefix is to use
11590 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11593 As a special kludge, if the path provided by @option{-B} is
11594 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11595 9, then it is replaced by @file{[dir/]include}. This is to help
11596 with boot-strapping the compiler.
11598 @item -specs=@var{file}
11600 Process @var{file} after the compiler reads in the standard @file{specs}
11601 file, in order to override the defaults which the @command{gcc} driver
11602 program uses when determining what switches to pass to @command{cc1},
11603 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11604 @option{-specs=@var{file}} can be specified on the command line, and they
11605 are processed in order, from left to right.
11607 @item --sysroot=@var{dir}
11609 Use @var{dir} as the logical root directory for headers and libraries.
11610 For example, if the compiler normally searches for headers in
11611 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11612 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11614 If you use both this option and the @option{-isysroot} option, then
11615 the @option{--sysroot} option applies to libraries, but the
11616 @option{-isysroot} option applies to header files.
11618 The GNU linker (beginning with version 2.16) has the necessary support
11619 for this option. If your linker does not support this option, the
11620 header file aspect of @option{--sysroot} still works, but the
11621 library aspect does not.
11623 @item --no-sysroot-suffix
11624 @opindex no-sysroot-suffix
11625 For some targets, a suffix is added to the root directory specified
11626 with @option{--sysroot}, depending on the other options used, so that
11627 headers may for example be found in
11628 @file{@var{dir}/@var{suffix}/usr/include} instead of
11629 @file{@var{dir}/usr/include}. This option disables the addition of
11634 This option has been deprecated. Please use @option{-iquote} instead for
11635 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11637 Any directories you specify with @option{-I} options before the @option{-I-}
11638 option are searched only for the case of @code{#include "@var{file}"};
11639 they are not searched for @code{#include <@var{file}>}.
11641 If additional directories are specified with @option{-I} options after
11642 the @option{-I-} option, these directories are searched for all @code{#include}
11643 directives. (Ordinarily @emph{all} @option{-I} directories are used
11646 In addition, the @option{-I-} option inhibits the use of the current
11647 directory (where the current input file came from) as the first search
11648 directory for @code{#include "@var{file}"}. There is no way to
11649 override this effect of @option{-I-}. With @option{-I.} you can specify
11650 searching the directory that is current when the compiler is
11651 invoked. That is not exactly the same as what the preprocessor does
11652 by default, but it is often satisfactory.
11654 @option{-I-} does not inhibit the use of the standard system directories
11655 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11662 @section Specifying Subprocesses and the Switches to Pass to Them
11665 @command{gcc} is a driver program. It performs its job by invoking a
11666 sequence of other programs to do the work of compiling, assembling and
11667 linking. GCC interprets its command-line parameters and uses these to
11668 deduce which programs it should invoke, and which command-line options
11669 it ought to place on their command lines. This behavior is controlled
11670 by @dfn{spec strings}. In most cases there is one spec string for each
11671 program that GCC can invoke, but a few programs have multiple spec
11672 strings to control their behavior. The spec strings built into GCC can
11673 be overridden by using the @option{-specs=} command-line switch to specify
11676 @dfn{Spec files} are plaintext files that are used to construct spec
11677 strings. They consist of a sequence of directives separated by blank
11678 lines. The type of directive is determined by the first non-whitespace
11679 character on the line, which can be one of the following:
11682 @item %@var{command}
11683 Issues a @var{command} to the spec file processor. The commands that can
11687 @item %include <@var{file}>
11688 @cindex @code{%include}
11689 Search for @var{file} and insert its text at the current point in the
11692 @item %include_noerr <@var{file}>
11693 @cindex @code{%include_noerr}
11694 Just like @samp{%include}, but do not generate an error message if the include
11695 file cannot be found.
11697 @item %rename @var{old_name} @var{new_name}
11698 @cindex @code{%rename}
11699 Rename the spec string @var{old_name} to @var{new_name}.
11703 @item *[@var{spec_name}]:
11704 This tells the compiler to create, override or delete the named spec
11705 string. All lines after this directive up to the next directive or
11706 blank line are considered to be the text for the spec string. If this
11707 results in an empty string then the spec is deleted. (Or, if the
11708 spec did not exist, then nothing happens.) Otherwise, if the spec
11709 does not currently exist a new spec is created. If the spec does
11710 exist then its contents are overridden by the text of this
11711 directive, unless the first character of that text is the @samp{+}
11712 character, in which case the text is appended to the spec.
11714 @item [@var{suffix}]:
11715 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11716 and up to the next directive or blank line are considered to make up the
11717 spec string for the indicated suffix. When the compiler encounters an
11718 input file with the named suffix, it processes the spec string in
11719 order to work out how to compile that file. For example:
11723 z-compile -input %i
11726 This says that any input file whose name ends in @samp{.ZZ} should be
11727 passed to the program @samp{z-compile}, which should be invoked with the
11728 command-line switch @option{-input} and with the result of performing the
11729 @samp{%i} substitution. (See below.)
11731 As an alternative to providing a spec string, the text following a
11732 suffix directive can be one of the following:
11735 @item @@@var{language}
11736 This says that the suffix is an alias for a known @var{language}. This is
11737 similar to using the @option{-x} command-line switch to GCC to specify a
11738 language explicitly. For example:
11745 Says that .ZZ files are, in fact, C++ source files.
11748 This causes an error messages saying:
11751 @var{name} compiler not installed on this system.
11755 GCC already has an extensive list of suffixes built into it.
11756 This directive adds an entry to the end of the list of suffixes, but
11757 since the list is searched from the end backwards, it is effectively
11758 possible to override earlier entries using this technique.
11762 GCC has the following spec strings built into it. Spec files can
11763 override these strings or create their own. Note that individual
11764 targets can also add their own spec strings to this list.
11767 asm Options to pass to the assembler
11768 asm_final Options to pass to the assembler post-processor
11769 cpp Options to pass to the C preprocessor
11770 cc1 Options to pass to the C compiler
11771 cc1plus Options to pass to the C++ compiler
11772 endfile Object files to include at the end of the link
11773 link Options to pass to the linker
11774 lib Libraries to include on the command line to the linker
11775 libgcc Decides which GCC support library to pass to the linker
11776 linker Sets the name of the linker
11777 predefines Defines to be passed to the C preprocessor
11778 signed_char Defines to pass to CPP to say whether @code{char} is signed
11780 startfile Object files to include at the start of the link
11783 Here is a small example of a spec file:
11786 %rename lib old_lib
11789 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11792 This example renames the spec called @samp{lib} to @samp{old_lib} and
11793 then overrides the previous definition of @samp{lib} with a new one.
11794 The new definition adds in some extra command-line options before
11795 including the text of the old definition.
11797 @dfn{Spec strings} are a list of command-line options to be passed to their
11798 corresponding program. In addition, the spec strings can contain
11799 @samp{%}-prefixed sequences to substitute variable text or to
11800 conditionally insert text into the command line. Using these constructs
11801 it is possible to generate quite complex command lines.
11803 Here is a table of all defined @samp{%}-sequences for spec
11804 strings. Note that spaces are not generated automatically around the
11805 results of expanding these sequences. Therefore you can concatenate them
11806 together or combine them with constant text in a single argument.
11810 Substitute one @samp{%} into the program name or argument.
11813 Substitute the name of the input file being processed.
11816 Substitute the basename of the input file being processed.
11817 This is the substring up to (and not including) the last period
11818 and not including the directory.
11821 This is the same as @samp{%b}, but include the file suffix (text after
11825 Marks the argument containing or following the @samp{%d} as a
11826 temporary file name, so that that file is deleted if GCC exits
11827 successfully. Unlike @samp{%g}, this contributes no text to the
11830 @item %g@var{suffix}
11831 Substitute a file name that has suffix @var{suffix} and is chosen
11832 once per compilation, and mark the argument in the same way as
11833 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11834 name is now chosen in a way that is hard to predict even when previously
11835 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11836 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11837 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11838 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11839 was simply substituted with a file name chosen once per compilation,
11840 without regard to any appended suffix (which was therefore treated
11841 just like ordinary text), making such attacks more likely to succeed.
11843 @item %u@var{suffix}
11844 Like @samp{%g}, but generates a new temporary file name
11845 each time it appears instead of once per compilation.
11847 @item %U@var{suffix}
11848 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11849 new one if there is no such last file name. In the absence of any
11850 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11851 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11852 involves the generation of two distinct file names, one
11853 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11854 simply substituted with a file name chosen for the previous @samp{%u},
11855 without regard to any appended suffix.
11857 @item %j@var{suffix}
11858 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11859 writable, and if @option{-save-temps} is not used;
11860 otherwise, substitute the name
11861 of a temporary file, just like @samp{%u}. This temporary file is not
11862 meant for communication between processes, but rather as a junk
11863 disposal mechanism.
11865 @item %|@var{suffix}
11866 @itemx %m@var{suffix}
11867 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11868 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11869 all. These are the two most common ways to instruct a program that it
11870 should read from standard input or write to standard output. If you
11871 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11872 construct: see for example @file{f/lang-specs.h}.
11874 @item %.@var{SUFFIX}
11875 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11876 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11877 terminated by the next space or %.
11880 Marks the argument containing or following the @samp{%w} as the
11881 designated output file of this compilation. This puts the argument
11882 into the sequence of arguments that @samp{%o} substitutes.
11885 Substitutes the names of all the output files, with spaces
11886 automatically placed around them. You should write spaces
11887 around the @samp{%o} as well or the results are undefined.
11888 @samp{%o} is for use in the specs for running the linker.
11889 Input files whose names have no recognized suffix are not compiled
11890 at all, but they are included among the output files, so they are
11894 Substitutes the suffix for object files. Note that this is
11895 handled specially when it immediately follows @samp{%g, %u, or %U},
11896 because of the need for those to form complete file names. The
11897 handling is such that @samp{%O} is treated exactly as if it had already
11898 been substituted, except that @samp{%g, %u, and %U} do not currently
11899 support additional @var{suffix} characters following @samp{%O} as they do
11900 following, for example, @samp{.o}.
11903 Substitutes the standard macro predefinitions for the
11904 current target machine. Use this when running @command{cpp}.
11907 Like @samp{%p}, but puts @samp{__} before and after the name of each
11908 predefined macro, except for macros that start with @samp{__} or with
11909 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11913 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11914 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11915 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11916 and @option{-imultilib} as necessary.
11919 Current argument is the name of a library or startup file of some sort.
11920 Search for that file in a standard list of directories and substitute
11921 the full name found. The current working directory is included in the
11922 list of directories scanned.
11925 Current argument is the name of a linker script. Search for that file
11926 in the current list of directories to scan for libraries. If the file
11927 is located insert a @option{--script} option into the command line
11928 followed by the full path name found. If the file is not found then
11929 generate an error message. Note: the current working directory is not
11933 Print @var{str} as an error message. @var{str} is terminated by a newline.
11934 Use this when inconsistent options are detected.
11936 @item %(@var{name})
11937 Substitute the contents of spec string @var{name} at this point.
11939 @item %x@{@var{option}@}
11940 Accumulate an option for @samp{%X}.
11943 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11947 Output the accumulated assembler options specified by @option{-Wa}.
11950 Output the accumulated preprocessor options specified by @option{-Wp}.
11953 Process the @code{asm} spec. This is used to compute the
11954 switches to be passed to the assembler.
11957 Process the @code{asm_final} spec. This is a spec string for
11958 passing switches to an assembler post-processor, if such a program is
11962 Process the @code{link} spec. This is the spec for computing the
11963 command line passed to the linker. Typically it makes use of the
11964 @samp{%L %G %S %D and %E} sequences.
11967 Dump out a @option{-L} option for each directory that GCC believes might
11968 contain startup files. If the target supports multilibs then the
11969 current multilib directory is prepended to each of these paths.
11972 Process the @code{lib} spec. This is a spec string for deciding which
11973 libraries are included on the command line to the linker.
11976 Process the @code{libgcc} spec. This is a spec string for deciding
11977 which GCC support library is included on the command line to the linker.
11980 Process the @code{startfile} spec. This is a spec for deciding which
11981 object files are the first ones passed to the linker. Typically
11982 this might be a file named @file{crt0.o}.
11985 Process the @code{endfile} spec. This is a spec string that specifies
11986 the last object files that are passed to the linker.
11989 Process the @code{cpp} spec. This is used to construct the arguments
11990 to be passed to the C preprocessor.
11993 Process the @code{cc1} spec. This is used to construct the options to be
11994 passed to the actual C compiler (@command{cc1}).
11997 Process the @code{cc1plus} spec. This is used to construct the options to be
11998 passed to the actual C++ compiler (@command{cc1plus}).
12001 Substitute the variable part of a matched option. See below.
12002 Note that each comma in the substituted string is replaced by
12006 Remove all occurrences of @code{-S} from the command line. Note---this
12007 command is position dependent. @samp{%} commands in the spec string
12008 before this one see @code{-S}, @samp{%} commands in the spec string
12009 after this one do not.
12011 @item %:@var{function}(@var{args})
12012 Call the named function @var{function}, passing it @var{args}.
12013 @var{args} is first processed as a nested spec string, then split
12014 into an argument vector in the usual fashion. The function returns
12015 a string which is processed as if it had appeared literally as part
12016 of the current spec.
12018 The following built-in spec functions are provided:
12021 @item @code{getenv}
12022 The @code{getenv} spec function takes two arguments: an environment
12023 variable name and a string. If the environment variable is not
12024 defined, a fatal error is issued. Otherwise, the return value is the
12025 value of the environment variable concatenated with the string. For
12026 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
12029 %:getenv(TOPDIR /include)
12032 expands to @file{/path/to/top/include}.
12034 @item @code{if-exists}
12035 The @code{if-exists} spec function takes one argument, an absolute
12036 pathname to a file. If the file exists, @code{if-exists} returns the
12037 pathname. Here is a small example of its usage:
12041 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
12044 @item @code{if-exists-else}
12045 The @code{if-exists-else} spec function is similar to the @code{if-exists}
12046 spec function, except that it takes two arguments. The first argument is
12047 an absolute pathname to a file. If the file exists, @code{if-exists-else}
12048 returns the pathname. If it does not exist, it returns the second argument.
12049 This way, @code{if-exists-else} can be used to select one file or another,
12050 based on the existence of the first. Here is a small example of its usage:
12054 crt0%O%s %:if-exists(crti%O%s) \
12055 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
12058 @item @code{replace-outfile}
12059 The @code{replace-outfile} spec function takes two arguments. It looks for the
12060 first argument in the outfiles array and replaces it with the second argument. Here
12061 is a small example of its usage:
12064 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
12067 @item @code{remove-outfile}
12068 The @code{remove-outfile} spec function takes one argument. It looks for the
12069 first argument in the outfiles array and removes it. Here is a small example
12073 %:remove-outfile(-lm)
12076 @item @code{pass-through-libs}
12077 The @code{pass-through-libs} spec function takes any number of arguments. It
12078 finds any @option{-l} options and any non-options ending in @file{.a} (which it
12079 assumes are the names of linker input library archive files) and returns a
12080 result containing all the found arguments each prepended by
12081 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
12082 intended to be passed to the LTO linker plugin.
12085 %:pass-through-libs(%G %L %G)
12088 @item @code{print-asm-header}
12089 The @code{print-asm-header} function takes no arguments and simply
12090 prints a banner like:
12096 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
12099 It is used to separate compiler options from assembler options
12100 in the @option{--target-help} output.
12103 @item %@{@code{S}@}
12104 Substitutes the @code{-S} switch, if that switch is given to GCC@.
12105 If that switch is not specified, this substitutes nothing. Note that
12106 the leading dash is omitted when specifying this option, and it is
12107 automatically inserted if the substitution is performed. Thus the spec
12108 string @samp{%@{foo@}} matches the command-line option @option{-foo}
12109 and outputs the command-line option @option{-foo}.
12111 @item %W@{@code{S}@}
12112 Like %@{@code{S}@} but mark last argument supplied within as a file to be
12113 deleted on failure.
12115 @item %@{@code{S}*@}
12116 Substitutes all the switches specified to GCC whose names start
12117 with @code{-S}, but which also take an argument. This is used for
12118 switches like @option{-o}, @option{-D}, @option{-I}, etc.
12119 GCC considers @option{-o foo} as being
12120 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
12121 text, including the space. Thus two arguments are generated.
12123 @item %@{@code{S}*&@code{T}*@}
12124 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
12125 (the order of @code{S} and @code{T} in the spec is not significant).
12126 There can be any number of ampersand-separated variables; for each the
12127 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
12129 @item %@{@code{S}:@code{X}@}
12130 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
12132 @item %@{!@code{S}:@code{X}@}
12133 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
12135 @item %@{@code{S}*:@code{X}@}
12136 Substitutes @code{X} if one or more switches whose names start with
12137 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
12138 once, no matter how many such switches appeared. However, if @code{%*}
12139 appears somewhere in @code{X}, then @code{X} is substituted once
12140 for each matching switch, with the @code{%*} replaced by the part of
12141 that switch matching the @code{*}.
12143 If @code{%*} appears as the last part of a spec sequence then a space
12144 is added after the end of the last substitution. If there is more
12145 text in the sequence, however, then a space is not generated. This
12146 allows the @code{%*} substitution to be used as part of a larger
12147 string. For example, a spec string like this:
12150 %@{mcu=*:--script=%*/memory.ld@}
12154 when matching an option like @option{-mcu=newchip} produces:
12157 --script=newchip/memory.ld
12160 @item %@{.@code{S}:@code{X}@}
12161 Substitutes @code{X}, if processing a file with suffix @code{S}.
12163 @item %@{!.@code{S}:@code{X}@}
12164 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
12166 @item %@{,@code{S}:@code{X}@}
12167 Substitutes @code{X}, if processing a file for language @code{S}.
12169 @item %@{!,@code{S}:@code{X}@}
12170 Substitutes @code{X}, if not processing a file for language @code{S}.
12172 @item %@{@code{S}|@code{P}:@code{X}@}
12173 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
12174 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
12175 @code{*} sequences as well, although they have a stronger binding than
12176 the @samp{|}. If @code{%*} appears in @code{X}, all of the
12177 alternatives must be starred, and only the first matching alternative
12180 For example, a spec string like this:
12183 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
12187 outputs the following command-line options from the following input
12188 command-line options:
12193 -d fred.c -foo -baz -boggle
12194 -d jim.d -bar -baz -boggle
12197 @item %@{S:X; T:Y; :D@}
12199 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
12200 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
12201 be as many clauses as you need. This may be combined with @code{.},
12202 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
12207 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
12208 construct may contain other nested @samp{%} constructs or spaces, or
12209 even newlines. They are processed as usual, as described above.
12210 Trailing white space in @code{X} is ignored. White space may also
12211 appear anywhere on the left side of the colon in these constructs,
12212 except between @code{.} or @code{*} and the corresponding word.
12214 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
12215 handled specifically in these constructs. If another value of
12216 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
12217 @option{-W} switch is found later in the command line, the earlier
12218 switch value is ignored, except with @{@code{S}*@} where @code{S} is
12219 just one letter, which passes all matching options.
12221 The character @samp{|} at the beginning of the predicate text is used to
12222 indicate that a command should be piped to the following command, but
12223 only if @option{-pipe} is specified.
12225 It is built into GCC which switches take arguments and which do not.
12226 (You might think it would be useful to generalize this to allow each
12227 compiler's spec to say which switches take arguments. But this cannot
12228 be done in a consistent fashion. GCC cannot even decide which input
12229 files have been specified without knowing which switches take arguments,
12230 and it must know which input files to compile in order to tell which
12233 GCC also knows implicitly that arguments starting in @option{-l} are to be
12234 treated as compiler output files, and passed to the linker in their
12235 proper position among the other output files.
12237 @c man begin OPTIONS
12239 @node Target Options
12240 @section Specifying Target Machine and Compiler Version
12241 @cindex target options
12242 @cindex cross compiling
12243 @cindex specifying machine version
12244 @cindex specifying compiler version and target machine
12245 @cindex compiler version, specifying
12246 @cindex target machine, specifying
12248 The usual way to run GCC is to run the executable called @command{gcc}, or
12249 @command{@var{machine}-gcc} when cross-compiling, or
12250 @command{@var{machine}-gcc-@var{version}} to run a version other than the
12251 one that was installed last.
12253 @node Submodel Options
12254 @section Hardware Models and Configurations
12255 @cindex submodel options
12256 @cindex specifying hardware config
12257 @cindex hardware models and configurations, specifying
12258 @cindex machine dependent options
12260 Each target machine types can have its own
12261 special options, starting with @samp{-m}, to choose among various
12262 hardware models or configurations---for example, 68010 vs 68020,
12263 floating coprocessor or none. A single installed version of the
12264 compiler can compile for any model or configuration, according to the
12267 Some configurations of the compiler also support additional special
12268 options, usually for compatibility with other compilers on the same
12271 @c This list is ordered alphanumerically by subsection name.
12272 @c It should be the same order and spelling as these options are listed
12273 @c in Machine Dependent Options
12276 * AArch64 Options::
12277 * Adapteva Epiphany Options::
12281 * Blackfin Options::
12286 * DEC Alpha Options::
12290 * GNU/Linux Options::
12300 * MicroBlaze Options::
12303 * MN10300 Options::
12307 * Nios II Options::
12308 * Nvidia PTX Options::
12310 * picoChip Options::
12311 * PowerPC Options::
12313 * RS/6000 and PowerPC Options::
12315 * S/390 and zSeries Options::
12318 * Solaris 2 Options::
12321 * System V Options::
12322 * TILE-Gx Options::
12323 * TILEPro Options::
12328 * VxWorks Options::
12330 * x86 Windows Options::
12331 * Xstormy16 Options::
12333 * zSeries Options::
12336 @node AArch64 Options
12337 @subsection AArch64 Options
12338 @cindex AArch64 Options
12340 These options are defined for AArch64 implementations:
12344 @item -mabi=@var{name}
12346 Generate code for the specified data model. Permissible values
12347 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12348 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12349 but long int and pointer are 64-bit.
12351 The default depends on the specific target configuration. Note that
12352 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12353 entire program with the same ABI, and link with a compatible set of libraries.
12356 @opindex mbig-endian
12357 Generate big-endian code. This is the default when GCC is configured for an
12358 @samp{aarch64_be-*-*} target.
12360 @item -mgeneral-regs-only
12361 @opindex mgeneral-regs-only
12362 Generate code which uses only the general registers.
12364 @item -mlittle-endian
12365 @opindex mlittle-endian
12366 Generate little-endian code. This is the default when GCC is configured for an
12367 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12369 @item -mcmodel=tiny
12370 @opindex mcmodel=tiny
12371 Generate code for the tiny code model. The program and its statically defined
12372 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12373 be statically or dynamically linked. This model is not fully implemented and
12374 mostly treated as @samp{small}.
12376 @item -mcmodel=small
12377 @opindex mcmodel=small
12378 Generate code for the small code model. The program and its statically defined
12379 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12380 be statically or dynamically linked. This is the default code model.
12382 @item -mcmodel=large
12383 @opindex mcmodel=large
12384 Generate code for the large code model. This makes no assumptions about
12385 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12386 statically linked only.
12388 @item -mstrict-align
12389 @opindex mstrict-align
12390 Do not assume that unaligned memory references are handled by the system.
12392 @item -momit-leaf-frame-pointer
12393 @itemx -mno-omit-leaf-frame-pointer
12394 @opindex momit-leaf-frame-pointer
12395 @opindex mno-omit-leaf-frame-pointer
12396 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12399 @item -mtls-dialect=desc
12400 @opindex mtls-dialect=desc
12401 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12402 of TLS variables. This is the default.
12404 @item -mtls-dialect=traditional
12405 @opindex mtls-dialect=traditional
12406 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12409 @item -mfix-cortex-a53-835769
12410 @itemx -mno-fix-cortex-a53-835769
12411 @opindex mfix-cortex-a53-835769
12412 @opindex mno-fix-cortex-a53-835769
12413 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12414 This involves inserting a NOP instruction between memory instructions and
12415 64-bit integer multiply-accumulate instructions.
12417 @item -mfix-cortex-a53-843419
12418 @itemx -mno-fix-cortex-a53-843419
12419 @opindex mfix-cortex-a53-843419
12420 @opindex mno-fix-cortex-a53-843419
12421 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
12422 This erratum workaround is made at link time and this will only pass the
12423 corresponding flag to the linker.
12425 @item -march=@var{name}
12427 Specify the name of the target architecture, optionally suffixed by one or
12428 more feature modifiers. This option has the form
12429 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12430 only permissible value for @var{arch} is @samp{armv8-a}.
12431 The permissible values for @var{feature} are documented in the sub-section
12432 below. Additionally on native AArch64 GNU/Linux systems the value
12433 @samp{native} is available. This option causes the compiler to pick the
12434 architecture of the host system. If the compiler is unable to recognize the
12435 architecture of the host system this option has no effect.
12437 Where conflicting feature modifiers are specified, the right-most feature is
12440 GCC uses this name to determine what kind of instructions it can emit when
12441 generating assembly code.
12443 Where @option{-march} is specified without either of @option{-mtune}
12444 or @option{-mcpu} also being specified, the code is tuned to perform
12445 well across a range of target processors implementing the target
12448 @item -mtune=@var{name}
12450 Specify the name of the target processor for which GCC should tune the
12451 performance of the code. Permissible values for this option are:
12452 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
12453 @samp{exynos-m1}, @samp{thunderx}, @samp{xgene1}.
12455 Additionally, this option can specify that GCC should tune the performance
12456 of the code for a big.LITTLE system. Permissible values for this
12457 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12459 Additionally on native AArch64 GNU/Linux systems the value @samp{native}
12461 This option causes the compiler to pick the architecture of and tune the
12462 performance of the code for the processor of the host system.
12463 If the compiler is unable to recognize the processor of the host system
12464 this option has no effect.
12466 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12467 are specified, the code is tuned to perform well across a range
12468 of target processors.
12470 This option cannot be suffixed by feature modifiers.
12472 @item -mcpu=@var{name}
12474 Specify the name of the target processor, optionally suffixed by one or more
12475 feature modifiers. This option has the form
12476 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12477 permissible values for @var{cpu} are the same as those available for
12478 @option{-mtune}. Additionally on native AArch64 GNU/Linux systems the
12479 value @samp{native} is available.
12480 This option causes the compiler to tune the performance of the code for the
12481 processor of the host system. If the compiler is unable to recognize the
12482 processor of the host system this option has no effect.
12484 The permissible values for @var{feature} are documented in the sub-section
12487 Where conflicting feature modifiers are specified, the right-most feature is
12490 GCC uses this name to determine what kind of instructions it can emit when
12491 generating assembly code (as if by @option{-march}) and to determine
12492 the target processor for which to tune for performance (as if
12493 by @option{-mtune}). Where this option is used in conjunction
12494 with @option{-march} or @option{-mtune}, those options take precedence
12495 over the appropriate part of this option.
12498 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12499 @cindex @option{-march} feature modifiers
12500 @cindex @option{-mcpu} feature modifiers
12501 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
12506 Enable CRC extension.
12508 Enable Crypto extension. This implies Advanced SIMD is enabled.
12510 Enable floating-point instructions.
12512 Enable Advanced SIMD instructions. This implies floating-point instructions
12513 are enabled. This is the default for all current possible values for options
12514 @option{-march} and @option{-mcpu=}.
12517 @node Adapteva Epiphany Options
12518 @subsection Adapteva Epiphany Options
12520 These @samp{-m} options are defined for Adapteva Epiphany:
12523 @item -mhalf-reg-file
12524 @opindex mhalf-reg-file
12525 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12526 That allows code to run on hardware variants that lack these registers.
12528 @item -mprefer-short-insn-regs
12529 @opindex mprefer-short-insn-regs
12530 Preferrentially allocate registers that allow short instruction generation.
12531 This can result in increased instruction count, so this may either reduce or
12532 increase overall code size.
12534 @item -mbranch-cost=@var{num}
12535 @opindex mbranch-cost
12536 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12537 This cost is only a heuristic and is not guaranteed to produce
12538 consistent results across releases.
12542 Enable the generation of conditional moves.
12544 @item -mnops=@var{num}
12546 Emit @var{num} NOPs before every other generated instruction.
12548 @item -mno-soft-cmpsf
12549 @opindex mno-soft-cmpsf
12550 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12551 and test the flags. This is faster than a software comparison, but can
12552 get incorrect results in the presence of NaNs, or when two different small
12553 numbers are compared such that their difference is calculated as zero.
12554 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12555 software comparisons.
12557 @item -mstack-offset=@var{num}
12558 @opindex mstack-offset
12559 Set the offset between the top of the stack and the stack pointer.
12560 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12561 can be used by leaf functions without stack allocation.
12562 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12563 Note also that this option changes the ABI; compiling a program with a
12564 different stack offset than the libraries have been compiled with
12565 generally does not work.
12566 This option can be useful if you want to evaluate if a different stack
12567 offset would give you better code, but to actually use a different stack
12568 offset to build working programs, it is recommended to configure the
12569 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12571 @item -mno-round-nearest
12572 @opindex mno-round-nearest
12573 Make the scheduler assume that the rounding mode has been set to
12574 truncating. The default is @option{-mround-nearest}.
12577 @opindex mlong-calls
12578 If not otherwise specified by an attribute, assume all calls might be beyond
12579 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12580 function address into a register before performing a (otherwise direct) call.
12581 This is the default.
12583 @item -mshort-calls
12584 @opindex short-calls
12585 If not otherwise specified by an attribute, assume all direct calls are
12586 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12587 for direct calls. The default is @option{-mlong-calls}.
12591 Assume addresses can be loaded as 16-bit unsigned values. This does not
12592 apply to function addresses for which @option{-mlong-calls} semantics
12595 @item -mfp-mode=@var{mode}
12597 Set the prevailing mode of the floating-point unit.
12598 This determines the floating-point mode that is provided and expected
12599 at function call and return time. Making this mode match the mode you
12600 predominantly need at function start can make your programs smaller and
12601 faster by avoiding unnecessary mode switches.
12603 @var{mode} can be set to one the following values:
12607 Any mode at function entry is valid, and retained or restored when
12608 the function returns, and when it calls other functions.
12609 This mode is useful for compiling libraries or other compilation units
12610 you might want to incorporate into different programs with different
12611 prevailing FPU modes, and the convenience of being able to use a single
12612 object file outweighs the size and speed overhead for any extra
12613 mode switching that might be needed, compared with what would be needed
12614 with a more specific choice of prevailing FPU mode.
12617 This is the mode used for floating-point calculations with
12618 truncating (i.e.@: round towards zero) rounding mode. That includes
12619 conversion from floating point to integer.
12621 @item round-nearest
12622 This is the mode used for floating-point calculations with
12623 round-to-nearest-or-even rounding mode.
12626 This is the mode used to perform integer calculations in the FPU, e.g.@:
12627 integer multiply, or integer multiply-and-accumulate.
12630 The default is @option{-mfp-mode=caller}
12632 @item -mnosplit-lohi
12633 @itemx -mno-postinc
12634 @itemx -mno-postmodify
12635 @opindex mnosplit-lohi
12636 @opindex mno-postinc
12637 @opindex mno-postmodify
12638 Code generation tweaks that disable, respectively, splitting of 32-bit
12639 loads, generation of post-increment addresses, and generation of
12640 post-modify addresses. The defaults are @option{msplit-lohi},
12641 @option{-mpost-inc}, and @option{-mpost-modify}.
12643 @item -mnovect-double
12644 @opindex mno-vect-double
12645 Change the preferred SIMD mode to SImode. The default is
12646 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12648 @item -max-vect-align=@var{num}
12649 @opindex max-vect-align
12650 The maximum alignment for SIMD vector mode types.
12651 @var{num} may be 4 or 8. The default is 8.
12652 Note that this is an ABI change, even though many library function
12653 interfaces are unaffected if they don't use SIMD vector modes
12654 in places that affect size and/or alignment of relevant types.
12656 @item -msplit-vecmove-early
12657 @opindex msplit-vecmove-early
12658 Split vector moves into single word moves before reload. In theory this
12659 can give better register allocation, but so far the reverse seems to be
12660 generally the case.
12662 @item -m1reg-@var{reg}
12664 Specify a register to hold the constant @minus{}1, which makes loading small negative
12665 constants and certain bitmasks faster.
12666 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12667 which specify use of that register as a fixed register,
12668 and @samp{none}, which means that no register is used for this
12669 purpose. The default is @option{-m1reg-none}.
12674 @subsection ARC Options
12675 @cindex ARC options
12677 The following options control the architecture variant for which code
12680 @c architecture variants
12683 @item -mbarrel-shifter
12684 @opindex mbarrel-shifter
12685 Generate instructions supported by barrel shifter. This is the default
12686 unless @option{-mcpu=ARC601} is in effect.
12688 @item -mcpu=@var{cpu}
12690 Set architecture type, register usage, and instruction scheduling
12691 parameters for @var{cpu}. There are also shortcut alias options
12692 available for backward compatibility and convenience. Supported
12693 values for @var{cpu} are
12699 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12703 Compile for ARC601. Alias: @option{-mARC601}.
12708 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12709 This is the default when configured with @option{--with-cpu=arc700}@.
12714 @itemx -mdpfp-compact
12715 @opindex mdpfp-compact
12716 FPX: Generate Double Precision FPX instructions, tuned for the compact
12720 @opindex mdpfp-fast
12721 FPX: Generate Double Precision FPX instructions, tuned for the fast
12724 @item -mno-dpfp-lrsr
12725 @opindex mno-dpfp-lrsr
12726 Disable LR and SR instructions from using FPX extension aux registers.
12730 Generate Extended arithmetic instructions. Currently only
12731 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12732 supported. This is always enabled for @option{-mcpu=ARC700}.
12736 Do not generate mpy instructions for ARC700.
12740 Generate 32x16 bit multiply and mac instructions.
12744 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12748 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12753 @itemx -mspfp-compact
12754 @opindex mspfp-compact
12755 FPX: Generate Single Precision FPX instructions, tuned for the compact
12759 @opindex mspfp-fast
12760 FPX: Generate Single Precision FPX instructions, tuned for the fast
12765 Enable generation of ARC SIMD instructions via target-specific
12766 builtins. Only valid for @option{-mcpu=ARC700}.
12769 @opindex msoft-float
12770 This option ignored; it is provided for compatibility purposes only.
12771 Software floating point code is emitted by default, and this default
12772 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12773 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12774 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12778 Generate swap instructions.
12782 The following options are passed through to the assembler, and also
12783 define preprocessor macro symbols.
12785 @c Flags used by the assembler, but for which we define preprocessor
12786 @c macro symbols as well.
12789 @opindex mdsp-packa
12790 Passed down to the assembler to enable the DSP Pack A extensions.
12791 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12795 Passed down to the assembler to enable the dual viterbi butterfly
12796 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12798 @c ARC700 4.10 extension instruction
12801 Passed down to the assembler to enable the Locked Load/Store
12802 Conditional extension. Also sets the preprocessor symbol
12807 Passed down to the assembler. Also sets the preprocessor symbol
12808 @code{__Xxmac_d16}.
12812 Passed down to the assembler. Also sets the preprocessor symbol
12815 @c ARC700 4.10 extension instruction
12818 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12819 extension instruction. Also sets the preprocessor symbol
12822 @c ARC700 4.10 extension instruction
12825 Passed down to the assembler to enable the swap byte ordering
12826 extension instruction. Also sets the preprocessor symbol
12830 @opindex mtelephony
12831 Passed down to the assembler to enable dual and single operand
12832 instructions for telephony. Also sets the preprocessor symbol
12833 @code{__Xtelephony}.
12837 Passed down to the assembler to enable the XY Memory extension. Also
12838 sets the preprocessor symbol @code{__Xxy}.
12842 The following options control how the assembly code is annotated:
12844 @c Assembly annotation options
12848 Annotate assembler instructions with estimated addresses.
12850 @item -mannotate-align
12851 @opindex mannotate-align
12852 Explain what alignment considerations lead to the decision to make an
12853 instruction short or long.
12857 The following options are passed through to the linker:
12859 @c options passed through to the linker
12863 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12864 This option is enabled by default in tool chains built for
12865 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12866 when profiling is not requested.
12868 @item -marclinux_prof
12869 @opindex marclinux_prof
12870 Passed through to the linker, to specify use of the
12871 @code{arclinux_prof} emulation. This option is enabled by default in
12872 tool chains built for @w{@code{arc-linux-uclibc}} and
12873 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12877 The following options control the semantics of generated code:
12879 @c semantically relevant code generation options
12881 @item -mepilogue-cfi
12882 @opindex mepilogue-cfi
12883 Enable generation of call frame information for epilogues.
12885 @item -mno-epilogue-cfi
12886 @opindex mno-epilogue-cfi
12887 Disable generation of call frame information for epilogues.
12890 @opindex mlong-calls
12891 Generate call insns as register indirect calls, thus providing access
12892 to the full 32-bit address range.
12894 @item -mmedium-calls
12895 @opindex mmedium-calls
12896 Don't use less than 25 bit addressing range for calls, which is the
12897 offset available for an unconditional branch-and-link
12898 instruction. Conditional execution of function calls is suppressed, to
12899 allow use of the 25-bit range, rather than the 21-bit range with
12900 conditional branch-and-link. This is the default for tool chains built
12901 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12905 Do not generate sdata references. This is the default for tool chains
12906 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12910 @opindex mucb-mcount
12911 Instrument with mcount calls as used in UCB code. I.e. do the
12912 counting in the callee, not the caller. By default ARC instrumentation
12913 counts in the caller.
12915 @item -mvolatile-cache
12916 @opindex mvolatile-cache
12917 Use ordinarily cached memory accesses for volatile references. This is the
12920 @item -mno-volatile-cache
12921 @opindex mno-volatile-cache
12922 Enable cache bypass for volatile references.
12926 The following options fine tune code generation:
12927 @c code generation tuning options
12930 @opindex malign-call
12931 Do alignment optimizations for call instructions.
12933 @item -mauto-modify-reg
12934 @opindex mauto-modify-reg
12935 Enable the use of pre/post modify with register displacement.
12937 @item -mbbit-peephole
12938 @opindex mbbit-peephole
12939 Enable bbit peephole2.
12943 This option disables a target-specific pass in @file{arc_reorg} to
12944 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12945 generation driven by the combiner pass.
12947 @item -mcase-vector-pcrel
12948 @opindex mcase-vector-pcrel
12949 Use pc-relative switch case tables - this enables case table shortening.
12950 This is the default for @option{-Os}.
12952 @item -mcompact-casesi
12953 @opindex mcompact-casesi
12954 Enable compact casesi pattern.
12955 This is the default for @option{-Os}.
12957 @item -mno-cond-exec
12958 @opindex mno-cond-exec
12959 Disable ARCompact specific pass to generate conditional execution instructions.
12960 Due to delay slot scheduling and interactions between operand numbers,
12961 literal sizes, instruction lengths, and the support for conditional execution,
12962 the target-independent pass to generate conditional execution is often lacking,
12963 so the ARC port has kept a special pass around that tries to find more
12964 conditional execution generating opportunities after register allocation,
12965 branch shortening, and delay slot scheduling have been done. This pass
12966 generally, but not always, improves performance and code size, at the cost of
12967 extra compilation time, which is why there is an option to switch it off.
12968 If you have a problem with call instructions exceeding their allowable
12969 offset range because they are conditionalized, you should consider using
12970 @option{-mmedium-calls} instead.
12972 @item -mearly-cbranchsi
12973 @opindex mearly-cbranchsi
12974 Enable pre-reload use of the cbranchsi pattern.
12976 @item -mexpand-adddi
12977 @opindex mexpand-adddi
12978 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12979 @code{add.f}, @code{adc} etc.
12981 @item -mindexed-loads
12982 @opindex mindexed-loads
12983 Enable the use of indexed loads. This can be problematic because some
12984 optimizers then assume that indexed stores exist, which is not
12989 Enable Local Register Allocation. This is still experimental for ARC,
12990 so by default the compiler uses standard reload
12991 (i.e. @option{-mno-lra}).
12993 @item -mlra-priority-none
12994 @opindex mlra-priority-none
12995 Don't indicate any priority for target registers.
12997 @item -mlra-priority-compact
12998 @opindex mlra-priority-compact
12999 Indicate target register priority for r0..r3 / r12..r15.
13001 @item -mlra-priority-noncompact
13002 @opindex mlra-priority-noncompact
13003 Reduce target regsiter priority for r0..r3 / r12..r15.
13005 @item -mno-millicode
13006 @opindex mno-millicode
13007 When optimizing for size (using @option{-Os}), prologues and epilogues
13008 that have to save or restore a large number of registers are often
13009 shortened by using call to a special function in libgcc; this is
13010 referred to as a @emph{millicode} call. As these calls can pose
13011 performance issues, and/or cause linking issues when linking in a
13012 nonstandard way, this option is provided to turn off millicode call
13016 @opindex mmixed-code
13017 Tweak register allocation to help 16-bit instruction generation.
13018 This generally has the effect of decreasing the average instruction size
13019 while increasing the instruction count.
13023 Enable 'q' instruction alternatives.
13024 This is the default for @option{-Os}.
13028 Enable Rcq constraint handling - most short code generation depends on this.
13029 This is the default.
13033 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13034 This is the default.
13036 @item -msize-level=@var{level}
13037 @opindex msize-level
13038 Fine-tune size optimization with regards to instruction lengths and alignment.
13039 The recognized values for @var{level} are:
13042 No size optimization. This level is deprecated and treated like @samp{1}.
13045 Short instructions are used opportunistically.
13048 In addition, alignment of loops and of code after barriers are dropped.
13051 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13055 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13056 the behavior when this is not set is equivalent to level @samp{1}.
13058 @item -mtune=@var{cpu}
13060 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13061 by @option{-mcpu=}.
13063 Supported values for @var{cpu} are
13067 Tune for ARC600 cpu.
13070 Tune for ARC601 cpu.
13073 Tune for ARC700 cpu with standard multiplier block.
13076 Tune for ARC700 cpu with XMAC block.
13079 Tune for ARC725D cpu.
13082 Tune for ARC750D cpu.
13086 @item -mmultcost=@var{num}
13088 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13089 normal instruction.
13091 @item -munalign-prob-threshold=@var{probability}
13092 @opindex munalign-prob-threshold
13093 Set probability threshold for unaligning branches.
13094 When tuning for @samp{ARC700} and optimizing for speed, branches without
13095 filled delay slot are preferably emitted unaligned and long, unless
13096 profiling indicates that the probability for the branch to be taken
13097 is below @var{probability}. @xref{Cross-profiling}.
13098 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13102 The following options are maintained for backward compatibility, but
13103 are now deprecated and will be removed in a future release:
13105 @c Deprecated options
13113 @opindex mbig-endian
13116 Compile code for big endian targets. Use of these options is now
13117 deprecated. Users wanting big-endian code, should use the
13118 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13119 building the tool chain, for which big-endian is the default.
13121 @item -mlittle-endian
13122 @opindex mlittle-endian
13125 Compile code for little endian targets. Use of these options is now
13126 deprecated. Users wanting little-endian code should use the
13127 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13128 building the tool chain, for which little-endian is the default.
13130 @item -mbarrel_shifter
13131 @opindex mbarrel_shifter
13132 Replaced by @option{-mbarrel-shifter}.
13134 @item -mdpfp_compact
13135 @opindex mdpfp_compact
13136 Replaced by @option{-mdpfp-compact}.
13139 @opindex mdpfp_fast
13140 Replaced by @option{-mdpfp-fast}.
13143 @opindex mdsp_packa
13144 Replaced by @option{-mdsp-packa}.
13148 Replaced by @option{-mea}.
13152 Replaced by @option{-mmac-24}.
13156 Replaced by @option{-mmac-d16}.
13158 @item -mspfp_compact
13159 @opindex mspfp_compact
13160 Replaced by @option{-mspfp-compact}.
13163 @opindex mspfp_fast
13164 Replaced by @option{-mspfp-fast}.
13166 @item -mtune=@var{cpu}
13168 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13169 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13170 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13172 @item -multcost=@var{num}
13174 Replaced by @option{-mmultcost}.
13179 @subsection ARM Options
13180 @cindex ARM options
13182 These @samp{-m} options are defined for the ARM port:
13185 @item -mabi=@var{name}
13187 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13188 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13191 @opindex mapcs-frame
13192 Generate a stack frame that is compliant with the ARM Procedure Call
13193 Standard for all functions, even if this is not strictly necessary for
13194 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13195 with this option causes the stack frames not to be generated for
13196 leaf functions. The default is @option{-mno-apcs-frame}.
13197 This option is deprecated.
13201 This is a synonym for @option{-mapcs-frame} and is deprecated.
13204 @c not currently implemented
13205 @item -mapcs-stack-check
13206 @opindex mapcs-stack-check
13207 Generate code to check the amount of stack space available upon entry to
13208 every function (that actually uses some stack space). If there is
13209 insufficient space available then either the function
13210 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13211 called, depending upon the amount of stack space required. The runtime
13212 system is required to provide these functions. The default is
13213 @option{-mno-apcs-stack-check}, since this produces smaller code.
13215 @c not currently implemented
13217 @opindex mapcs-float
13218 Pass floating-point arguments using the floating-point registers. This is
13219 one of the variants of the APCS@. This option is recommended if the
13220 target hardware has a floating-point unit or if a lot of floating-point
13221 arithmetic is going to be performed by the code. The default is
13222 @option{-mno-apcs-float}, since the size of integer-only code is
13223 slightly increased if @option{-mapcs-float} is used.
13225 @c not currently implemented
13226 @item -mapcs-reentrant
13227 @opindex mapcs-reentrant
13228 Generate reentrant, position-independent code. The default is
13229 @option{-mno-apcs-reentrant}.
13232 @item -mthumb-interwork
13233 @opindex mthumb-interwork
13234 Generate code that supports calling between the ARM and Thumb
13235 instruction sets. Without this option, on pre-v5 architectures, the
13236 two instruction sets cannot be reliably used inside one program. The
13237 default is @option{-mno-thumb-interwork}, since slightly larger code
13238 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13239 configurations this option is meaningless.
13241 @item -mno-sched-prolog
13242 @opindex mno-sched-prolog
13243 Prevent the reordering of instructions in the function prologue, or the
13244 merging of those instruction with the instructions in the function's
13245 body. This means that all functions start with a recognizable set
13246 of instructions (or in fact one of a choice from a small set of
13247 different function prologues), and this information can be used to
13248 locate the start of functions inside an executable piece of code. The
13249 default is @option{-msched-prolog}.
13251 @item -mfloat-abi=@var{name}
13252 @opindex mfloat-abi
13253 Specifies which floating-point ABI to use. Permissible values
13254 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13256 Specifying @samp{soft} causes GCC to generate output containing
13257 library calls for floating-point operations.
13258 @samp{softfp} allows the generation of code using hardware floating-point
13259 instructions, but still uses the soft-float calling conventions.
13260 @samp{hard} allows generation of floating-point instructions
13261 and uses FPU-specific calling conventions.
13263 The default depends on the specific target configuration. Note that
13264 the hard-float and soft-float ABIs are not link-compatible; you must
13265 compile your entire program with the same ABI, and link with a
13266 compatible set of libraries.
13268 @item -mlittle-endian
13269 @opindex mlittle-endian
13270 Generate code for a processor running in little-endian mode. This is
13271 the default for all standard configurations.
13274 @opindex mbig-endian
13275 Generate code for a processor running in big-endian mode; the default is
13276 to compile code for a little-endian processor.
13278 @item -march=@var{name}
13280 This specifies the name of the target ARM architecture. GCC uses this
13281 name to determine what kind of instructions it can emit when generating
13282 assembly code. This option can be used in conjunction with or instead
13283 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13284 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13285 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13286 @samp{armv6}, @samp{armv6j},
13287 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
13288 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13289 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
13290 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13292 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13295 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13296 architecture together with the optional CRC32 extensions.
13298 @option{-march=native} causes the compiler to auto-detect the architecture
13299 of the build computer. At present, this feature is only supported on
13300 GNU/Linux, and not all architectures are recognized. If the auto-detect
13301 is unsuccessful the option has no effect.
13303 @item -mtune=@var{name}
13305 This option specifies the name of the target ARM processor for
13306 which GCC should tune the performance of the code.
13307 For some ARM implementations better performance can be obtained by using
13309 Permissible names are: @samp{arm2}, @samp{arm250},
13310 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
13311 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
13312 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
13313 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
13315 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
13316 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
13317 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
13318 @samp{strongarm1110},
13319 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
13320 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
13321 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
13322 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13323 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13324 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13325 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13326 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
13327 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
13328 @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
13330 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13335 @samp{cortex-m0plus},
13336 @samp{cortex-m1.small-multiply},
13337 @samp{cortex-m0.small-multiply},
13338 @samp{cortex-m0plus.small-multiply},
13340 @samp{marvell-pj4},
13341 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13342 @samp{fa526}, @samp{fa626},
13343 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13346 Additionally, this option can specify that GCC should tune the performance
13347 of the code for a big.LITTLE system. Permissible names are:
13348 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
13349 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
13351 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13352 performance for a blend of processors within architecture @var{arch}.
13353 The aim is to generate code that run well on the current most popular
13354 processors, balancing between optimizations that benefit some CPUs in the
13355 range, and avoiding performance pitfalls of other CPUs. The effects of
13356 this option may change in future GCC versions as CPU models come and go.
13358 @option{-mtune=native} causes the compiler to auto-detect the CPU
13359 of the build computer. At present, this feature is only supported on
13360 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13361 unsuccessful the option has no effect.
13363 @item -mcpu=@var{name}
13365 This specifies the name of the target ARM processor. GCC uses this name
13366 to derive the name of the target ARM architecture (as if specified
13367 by @option{-march}) and the ARM processor type for which to tune for
13368 performance (as if specified by @option{-mtune}). Where this option
13369 is used in conjunction with @option{-march} or @option{-mtune},
13370 those options take precedence over the appropriate part of this option.
13372 Permissible names for this option are the same as those for
13375 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13376 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13377 See @option{-mtune} for more information.
13379 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13380 of the build computer. At present, this feature is only supported on
13381 GNU/Linux, and not all architectures are recognized. If the auto-detect
13382 is unsuccessful the option has no effect.
13384 @item -mfpu=@var{name}
13386 This specifies what floating-point hardware (or hardware emulation) is
13387 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13388 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13389 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13390 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13391 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13392 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13394 If @option{-msoft-float} is specified this specifies the format of
13395 floating-point values.
13397 If the selected floating-point hardware includes the NEON extension
13398 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13399 operations are not generated by GCC's auto-vectorization pass unless
13400 @option{-funsafe-math-optimizations} is also specified. This is
13401 because NEON hardware does not fully implement the IEEE 754 standard for
13402 floating-point arithmetic (in particular denormal values are treated as
13403 zero), so the use of NEON instructions may lead to a loss of precision.
13405 @item -mfp16-format=@var{name}
13406 @opindex mfp16-format
13407 Specify the format of the @code{__fp16} half-precision floating-point type.
13408 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13409 the default is @samp{none}, in which case the @code{__fp16} type is not
13410 defined. @xref{Half-Precision}, for more information.
13412 @item -mstructure-size-boundary=@var{n}
13413 @opindex mstructure-size-boundary
13414 The sizes of all structures and unions are rounded up to a multiple
13415 of the number of bits set by this option. Permissible values are 8, 32
13416 and 64. The default value varies for different toolchains. For the COFF
13417 targeted toolchain the default value is 8. A value of 64 is only allowed
13418 if the underlying ABI supports it.
13420 Specifying a larger number can produce faster, more efficient code, but
13421 can also increase the size of the program. Different values are potentially
13422 incompatible. Code compiled with one value cannot necessarily expect to
13423 work with code or libraries compiled with another value, if they exchange
13424 information using structures or unions.
13426 @item -mabort-on-noreturn
13427 @opindex mabort-on-noreturn
13428 Generate a call to the function @code{abort} at the end of a
13429 @code{noreturn} function. It is executed if the function tries to
13433 @itemx -mno-long-calls
13434 @opindex mlong-calls
13435 @opindex mno-long-calls
13436 Tells the compiler to perform function calls by first loading the
13437 address of the function into a register and then performing a subroutine
13438 call on this register. This switch is needed if the target function
13439 lies outside of the 64-megabyte addressing range of the offset-based
13440 version of subroutine call instruction.
13442 Even if this switch is enabled, not all function calls are turned
13443 into long calls. The heuristic is that static functions, functions
13444 that have the @code{short_call} attribute, functions that are inside
13445 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13446 definitions have already been compiled within the current compilation
13447 unit are not turned into long calls. The exceptions to this rule are
13448 that weak function definitions, functions with the @code{long_call}
13449 attribute or the @code{section} attribute, and functions that are within
13450 the scope of a @code{#pragma long_calls} directive are always
13451 turned into long calls.
13453 This feature is not enabled by default. Specifying
13454 @option{-mno-long-calls} restores the default behavior, as does
13455 placing the function calls within the scope of a @code{#pragma
13456 long_calls_off} directive. Note these switches have no effect on how
13457 the compiler generates code to handle function calls via function
13460 @item -msingle-pic-base
13461 @opindex msingle-pic-base
13462 Treat the register used for PIC addressing as read-only, rather than
13463 loading it in the prologue for each function. The runtime system is
13464 responsible for initializing this register with an appropriate value
13465 before execution begins.
13467 @item -mpic-register=@var{reg}
13468 @opindex mpic-register
13469 Specify the register to be used for PIC addressing.
13470 For standard PIC base case, the default is any suitable register
13471 determined by compiler. For single PIC base case, the default is
13472 @samp{R9} if target is EABI based or stack-checking is enabled,
13473 otherwise the default is @samp{R10}.
13475 @item -mpic-data-is-text-relative
13476 @opindex mpic-data-is-text-relative
13477 Assume that each data segments are relative to text segment at load time.
13478 Therefore, it permits addressing data using PC-relative operations.
13479 This option is on by default for targets other than VxWorks RTP.
13481 @item -mpoke-function-name
13482 @opindex mpoke-function-name
13483 Write the name of each function into the text section, directly
13484 preceding the function prologue. The generated code is similar to this:
13488 .ascii "arm_poke_function_name", 0
13491 .word 0xff000000 + (t1 - t0)
13492 arm_poke_function_name
13494 stmfd sp!, @{fp, ip, lr, pc@}
13498 When performing a stack backtrace, code can inspect the value of
13499 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13500 location @code{pc - 12} and the top 8 bits are set, then we know that
13501 there is a function name embedded immediately preceding this location
13502 and has length @code{((pc[-3]) & 0xff000000)}.
13509 Select between generating code that executes in ARM and Thumb
13510 states. The default for most configurations is to generate code
13511 that executes in ARM state, but the default can be changed by
13512 configuring GCC with the @option{--with-mode=}@var{state}
13515 You can also override the ARM and Thumb mode for each function
13516 by using the @code{target("thumb")} and @code{target("arm")} function attributes
13517 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
13520 @opindex mtpcs-frame
13521 Generate a stack frame that is compliant with the Thumb Procedure Call
13522 Standard for all non-leaf functions. (A leaf function is one that does
13523 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13525 @item -mtpcs-leaf-frame
13526 @opindex mtpcs-leaf-frame
13527 Generate a stack frame that is compliant with the Thumb Procedure Call
13528 Standard for all leaf functions. (A leaf function is one that does
13529 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13531 @item -mcallee-super-interworking
13532 @opindex mcallee-super-interworking
13533 Gives all externally visible functions in the file being compiled an ARM
13534 instruction set header which switches to Thumb mode before executing the
13535 rest of the function. This allows these functions to be called from
13536 non-interworking code. This option is not valid in AAPCS configurations
13537 because interworking is enabled by default.
13539 @item -mcaller-super-interworking
13540 @opindex mcaller-super-interworking
13541 Allows calls via function pointers (including virtual functions) to
13542 execute correctly regardless of whether the target code has been
13543 compiled for interworking or not. There is a small overhead in the cost
13544 of executing a function pointer if this option is enabled. This option
13545 is not valid in AAPCS configurations because interworking is enabled
13548 @item -mtp=@var{name}
13550 Specify the access model for the thread local storage pointer. The valid
13551 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13552 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13553 (supported in the arm6k architecture), and @samp{auto}, which uses the
13554 best available method for the selected processor. The default setting is
13557 @item -mtls-dialect=@var{dialect}
13558 @opindex mtls-dialect
13559 Specify the dialect to use for accessing thread local storage. Two
13560 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13561 @samp{gnu} dialect selects the original GNU scheme for supporting
13562 local and global dynamic TLS models. The @samp{gnu2} dialect
13563 selects the GNU descriptor scheme, which provides better performance
13564 for shared libraries. The GNU descriptor scheme is compatible with
13565 the original scheme, but does require new assembler, linker and
13566 library support. Initial and local exec TLS models are unaffected by
13567 this option and always use the original scheme.
13569 @item -mword-relocations
13570 @opindex mword-relocations
13571 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13572 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13573 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13576 @item -mfix-cortex-m3-ldrd
13577 @opindex mfix-cortex-m3-ldrd
13578 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13579 with overlapping destination and base registers are used. This option avoids
13580 generating these instructions. This option is enabled by default when
13581 @option{-mcpu=cortex-m3} is specified.
13583 @item -munaligned-access
13584 @itemx -mno-unaligned-access
13585 @opindex munaligned-access
13586 @opindex mno-unaligned-access
13587 Enables (or disables) reading and writing of 16- and 32- bit values
13588 from addresses that are not 16- or 32- bit aligned. By default
13589 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13590 architectures, and enabled for all other architectures. If unaligned
13591 access is not enabled then words in packed data structures are
13592 accessed a byte at a time.
13594 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13595 generated object file to either true or false, depending upon the
13596 setting of this option. If unaligned access is enabled then the
13597 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13600 @item -mneon-for-64bits
13601 @opindex mneon-for-64bits
13602 Enables using Neon to handle scalar 64-bits operations. This is
13603 disabled by default since the cost of moving data from core registers
13606 @item -mslow-flash-data
13607 @opindex mslow-flash-data
13608 Assume loading data from flash is slower than fetching instruction.
13609 Therefore literal load is minimized for better performance.
13610 This option is only supported when compiling for ARMv7 M-profile and
13613 @item -masm-syntax-unified
13614 @opindex masm-syntax-unified
13615 Assume inline assembler is using unified asm syntax. The default is
13616 currently off which implies divided syntax. Currently this option is
13617 available only for Thumb1 and has no effect on ARM state and Thumb2.
13618 However, this may change in future releases of GCC. Divided syntax
13619 should be considered deprecated.
13621 @item -mrestrict-it
13622 @opindex mrestrict-it
13623 Restricts generation of IT blocks to conform to the rules of ARMv8.
13624 IT blocks can only contain a single 16-bit instruction from a select
13625 set of instructions. This option is on by default for ARMv8 Thumb mode.
13627 @item -mprint-tune-info
13628 @opindex mprint-tune-info
13629 Print CPU tuning information as comment in assembler file. This is
13630 an option used only for regression testing of the compiler and not
13631 intended for ordinary use in compiling code. This option is disabled
13636 @subsection AVR Options
13637 @cindex AVR Options
13639 These options are defined for AVR implementations:
13642 @item -mmcu=@var{mcu}
13644 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13646 The default for this option is@tie{}@samp{avr2}.
13648 GCC supports the following AVR devices and ISAs:
13650 @include avr-mmcu.texi
13652 @item -maccumulate-args
13653 @opindex maccumulate-args
13654 Accumulate outgoing function arguments and acquire/release the needed
13655 stack space for outgoing function arguments once in function
13656 prologue/epilogue. Without this option, outgoing arguments are pushed
13657 before calling a function and popped afterwards.
13659 Popping the arguments after the function call can be expensive on
13660 AVR so that accumulating the stack space might lead to smaller
13661 executables because arguments need not to be removed from the
13662 stack after such a function call.
13664 This option can lead to reduced code size for functions that perform
13665 several calls to functions that get their arguments on the stack like
13666 calls to printf-like functions.
13668 @item -mbranch-cost=@var{cost}
13669 @opindex mbranch-cost
13670 Set the branch costs for conditional branch instructions to
13671 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13672 integers. The default branch cost is 0.
13674 @item -mcall-prologues
13675 @opindex mcall-prologues
13676 Functions prologues/epilogues are expanded as calls to appropriate
13677 subroutines. Code size is smaller.
13681 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13682 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13683 and @code{long long} is 4 bytes. Please note that this option does not
13684 conform to the C standards, but it results in smaller code
13687 @item -mn-flash=@var{num}
13689 Assume that the flash memory has a size of
13690 @var{num} times 64@tie{}KiB.
13692 @item -mno-interrupts
13693 @opindex mno-interrupts
13694 Generated code is not compatible with hardware interrupts.
13695 Code size is smaller.
13699 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13700 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13701 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
13702 the assembler's command line and the @option{--relax} option to the
13703 linker's command line.
13705 Jump relaxing is performed by the linker because jump offsets are not
13706 known before code is located. Therefore, the assembler code generated by the
13707 compiler is the same, but the instructions in the executable may
13708 differ from instructions in the assembler code.
13710 Relaxing must be turned on if linker stubs are needed, see the
13711 section on @code{EIND} and linker stubs below.
13715 Assume that the device supports the Read-Modify-Write
13716 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
13720 Treat the stack pointer register as an 8-bit register,
13721 i.e.@: assume the high byte of the stack pointer is zero.
13722 In general, you don't need to set this option by hand.
13724 This option is used internally by the compiler to select and
13725 build multilibs for architectures @code{avr2} and @code{avr25}.
13726 These architectures mix devices with and without @code{SPH}.
13727 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13728 the compiler driver adds or removes this option from the compiler
13729 proper's command line, because the compiler then knows if the device
13730 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13735 Use address register @code{X} in a way proposed by the hardware. This means
13736 that @code{X} is only used in indirect, post-increment or
13737 pre-decrement addressing.
13739 Without this option, the @code{X} register may be used in the same way
13740 as @code{Y} or @code{Z} which then is emulated by additional
13742 For example, loading a value with @code{X+const} addressing with a
13743 small non-negative @code{const < 64} to a register @var{Rn} is
13747 adiw r26, const ; X += const
13748 ld @var{Rn}, X ; @var{Rn} = *X
13749 sbiw r26, const ; X -= const
13753 @opindex mtiny-stack
13754 Only change the lower 8@tie{}bits of the stack pointer.
13757 @opindex nodevicelib
13758 Don't link against AVR-LibC's device specific library @code{libdev.a}.
13760 @item -Waddr-space-convert
13761 @opindex Waddr-space-convert
13762 Warn about conversions between address spaces in the case where the
13763 resulting address space is not contained in the incoming address space.
13766 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
13767 @cindex @code{EIND}
13768 Pointers in the implementation are 16@tie{}bits wide.
13769 The address of a function or label is represented as word address so
13770 that indirect jumps and calls can target any code address in the
13771 range of 64@tie{}Ki words.
13773 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13774 bytes of program memory space, there is a special function register called
13775 @code{EIND} that serves as most significant part of the target address
13776 when @code{EICALL} or @code{EIJMP} instructions are used.
13778 Indirect jumps and calls on these devices are handled as follows by
13779 the compiler and are subject to some limitations:
13784 The compiler never sets @code{EIND}.
13787 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13788 instructions or might read @code{EIND} directly in order to emulate an
13789 indirect call/jump by means of a @code{RET} instruction.
13792 The compiler assumes that @code{EIND} never changes during the startup
13793 code or during the application. In particular, @code{EIND} is not
13794 saved/restored in function or interrupt service routine
13798 For indirect calls to functions and computed goto, the linker
13799 generates @emph{stubs}. Stubs are jump pads sometimes also called
13800 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13801 The stub contains a direct jump to the desired address.
13804 Linker relaxation must be turned on so that the linker generates
13805 the stubs correctly in all situations. See the compiler option
13806 @option{-mrelax} and the linker option @option{--relax}.
13807 There are corner cases where the linker is supposed to generate stubs
13808 but aborts without relaxation and without a helpful error message.
13811 The default linker script is arranged for code with @code{EIND = 0}.
13812 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13813 linker script has to be used in order to place the sections whose
13814 name start with @code{.trampolines} into the segment where @code{EIND}
13818 The startup code from libgcc never sets @code{EIND}.
13819 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13820 For the impact of AVR-LibC on @code{EIND}, see the
13821 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13824 It is legitimate for user-specific startup code to set up @code{EIND}
13825 early, for example by means of initialization code located in
13826 section @code{.init3}. Such code runs prior to general startup code
13827 that initializes RAM and calls constructors, but after the bit
13828 of startup code from AVR-LibC that sets @code{EIND} to the segment
13829 where the vector table is located.
13831 #include <avr/io.h>
13834 __attribute__((section(".init3"),naked,used,no_instrument_function))
13835 init3_set_eind (void)
13837 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13838 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13843 The @code{__trampolines_start} symbol is defined in the linker script.
13846 Stubs are generated automatically by the linker if
13847 the following two conditions are met:
13850 @item The address of a label is taken by means of the @code{gs} modifier
13851 (short for @emph{generate stubs}) like so:
13853 LDI r24, lo8(gs(@var{func}))
13854 LDI r25, hi8(gs(@var{func}))
13856 @item The final location of that label is in a code segment
13857 @emph{outside} the segment where the stubs are located.
13861 The compiler emits such @code{gs} modifiers for code labels in the
13862 following situations:
13864 @item Taking address of a function or code label.
13865 @item Computed goto.
13866 @item If prologue-save function is used, see @option{-mcall-prologues}
13867 command-line option.
13868 @item Switch/case dispatch tables. If you do not want such dispatch
13869 tables you can specify the @option{-fno-jump-tables} command-line option.
13870 @item C and C++ constructors/destructors called during startup/shutdown.
13871 @item If the tools hit a @code{gs()} modifier explained above.
13875 Jumping to non-symbolic addresses like so is @emph{not} supported:
13880 /* Call function at word address 0x2 */
13881 return ((int(*)(void)) 0x2)();
13885 Instead, a stub has to be set up, i.e.@: the function has to be called
13886 through a symbol (@code{func_4} in the example):
13891 extern int func_4 (void);
13893 /* Call function at byte address 0x4 */
13898 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13899 Alternatively, @code{func_4} can be defined in the linker script.
13902 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13903 @cindex @code{RAMPD}
13904 @cindex @code{RAMPX}
13905 @cindex @code{RAMPY}
13906 @cindex @code{RAMPZ}
13907 Some AVR devices support memories larger than the 64@tie{}KiB range
13908 that can be accessed with 16-bit pointers. To access memory locations
13909 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13910 register is used as high part of the address:
13911 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13912 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13913 register, respectively, to get a wide address. Similarly,
13914 @code{RAMPD} is used together with direct addressing.
13918 The startup code initializes the @code{RAMP} special function
13919 registers with zero.
13922 If a @ref{AVR Named Address Spaces,named address space} other than
13923 generic or @code{__flash} is used, then @code{RAMPZ} is set
13924 as needed before the operation.
13927 If the device supports RAM larger than 64@tie{}KiB and the compiler
13928 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13929 is reset to zero after the operation.
13932 If the device comes with a specific @code{RAMP} register, the ISR
13933 prologue/epilogue saves/restores that SFR and initializes it with
13934 zero in case the ISR code might (implicitly) use it.
13937 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13938 If you use inline assembler to read from locations outside the
13939 16-bit address range and change one of the @code{RAMP} registers,
13940 you must reset it to zero after the access.
13944 @subsubsection AVR Built-in Macros
13946 GCC defines several built-in macros so that the user code can test
13947 for the presence or absence of features. Almost any of the following
13948 built-in macros are deduced from device capabilities and thus
13949 triggered by the @option{-mmcu=} command-line option.
13951 For even more AVR-specific built-in macros see
13952 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13957 Build-in macro that resolves to a decimal number that identifies the
13958 architecture and depends on the @option{-mmcu=@var{mcu}} option.
13959 Possible values are:
13961 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13962 @code{4}, @code{5}, @code{51}, @code{6}
13964 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
13965 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
13969 @code{100}, @code{102}, @code{104},
13970 @code{105}, @code{106}, @code{107}
13972 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
13973 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
13974 If @var{mcu} specifies a device, this built-in macro is set
13975 accordingly. For example, with @option{-mmcu=atmega8} the macro is
13976 defined to @code{4}.
13978 @item __AVR_@var{Device}__
13979 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
13980 the device's name. For example, @option{-mmcu=atmega8} defines the
13981 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
13982 @code{__AVR_ATtiny261A__}, etc.
13984 The built-in macros' names follow
13985 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13986 the device name as from the AVR user manual. The difference between
13987 @var{Device} in the built-in macro and @var{device} in
13988 @option{-mmcu=@var{device}} is that the latter is always lowercase.
13990 If @var{device} is not a device but only a core architecture like
13991 @samp{avr51}, this macro is not defined.
13993 @item __AVR_DEVICE_NAME__
13994 Setting @option{-mmcu=@var{device}} defines this built-in macro to
13995 the device's name. For example, with @option{-mmcu=atmega8} the macro
13996 is defined to @code{atmega8}.
13998 If @var{device} is not a device but only a core architecture like
13999 @samp{avr51}, this macro is not defined.
14001 @item __AVR_XMEGA__
14002 The device / architecture belongs to the XMEGA family of devices.
14004 @item __AVR_HAVE_ELPM__
14005 The device has the the @code{ELPM} instruction.
14007 @item __AVR_HAVE_ELPMX__
14008 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14009 R@var{n},Z+} instructions.
14011 @item __AVR_HAVE_MOVW__
14012 The device has the @code{MOVW} instruction to perform 16-bit
14013 register-register moves.
14015 @item __AVR_HAVE_LPMX__
14016 The device has the @code{LPM R@var{n},Z} and
14017 @code{LPM R@var{n},Z+} instructions.
14019 @item __AVR_HAVE_MUL__
14020 The device has a hardware multiplier.
14022 @item __AVR_HAVE_JMP_CALL__
14023 The device has the @code{JMP} and @code{CALL} instructions.
14024 This is the case for devices with at least 16@tie{}KiB of program
14027 @item __AVR_HAVE_EIJMP_EICALL__
14028 @itemx __AVR_3_BYTE_PC__
14029 The device has the @code{EIJMP} and @code{EICALL} instructions.
14030 This is the case for devices with more than 128@tie{}KiB of program memory.
14031 This also means that the program counter
14032 (PC) is 3@tie{}bytes wide.
14034 @item __AVR_2_BYTE_PC__
14035 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14036 with up to 128@tie{}KiB of program memory.
14038 @item __AVR_HAVE_8BIT_SP__
14039 @itemx __AVR_HAVE_16BIT_SP__
14040 The stack pointer (SP) register is treated as 8-bit respectively
14041 16-bit register by the compiler.
14042 The definition of these macros is affected by @option{-mtiny-stack}.
14044 @item __AVR_HAVE_SPH__
14046 The device has the SPH (high part of stack pointer) special function
14047 register or has an 8-bit stack pointer, respectively.
14048 The definition of these macros is affected by @option{-mmcu=} and
14049 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14052 @item __AVR_HAVE_RAMPD__
14053 @itemx __AVR_HAVE_RAMPX__
14054 @itemx __AVR_HAVE_RAMPY__
14055 @itemx __AVR_HAVE_RAMPZ__
14056 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14057 @code{RAMPZ} special function register, respectively.
14059 @item __NO_INTERRUPTS__
14060 This macro reflects the @option{-mno-interrupts} command-line option.
14062 @item __AVR_ERRATA_SKIP__
14063 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14064 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14065 instructions because of a hardware erratum. Skip instructions are
14066 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14067 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14070 @item __AVR_ISA_RMW__
14071 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14073 @item __AVR_SFR_OFFSET__=@var{offset}
14074 Instructions that can address I/O special function registers directly
14075 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14076 address as if addressed by an instruction to access RAM like @code{LD}
14077 or @code{STS}. This offset depends on the device architecture and has
14078 to be subtracted from the RAM address in order to get the
14079 respective I/O@tie{}address.
14081 @item __WITH_AVRLIBC__
14082 The compiler is configured to be used together with AVR-Libc.
14083 See the @option{--with-avrlibc} configure option.
14087 @node Blackfin Options
14088 @subsection Blackfin Options
14089 @cindex Blackfin Options
14092 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14094 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14095 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14096 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14097 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14098 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14099 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14100 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14101 @samp{bf561}, @samp{bf592}.
14103 The optional @var{sirevision} specifies the silicon revision of the target
14104 Blackfin processor. Any workarounds available for the targeted silicon revision
14105 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14106 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14107 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14108 hexadecimal digits representing the major and minor numbers in the silicon
14109 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14110 is not defined. If @var{sirevision} is @samp{any}, the
14111 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14112 If this optional @var{sirevision} is not used, GCC assumes the latest known
14113 silicon revision of the targeted Blackfin processor.
14115 GCC defines a preprocessor macro for the specified @var{cpu}.
14116 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14117 provided by libgloss to be linked in if @option{-msim} is not given.
14119 Without this option, @samp{bf532} is used as the processor by default.
14121 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14122 only the preprocessor macro is defined.
14126 Specifies that the program will be run on the simulator. This causes
14127 the simulator BSP provided by libgloss to be linked in. This option
14128 has effect only for @samp{bfin-elf} toolchain.
14129 Certain other options, such as @option{-mid-shared-library} and
14130 @option{-mfdpic}, imply @option{-msim}.
14132 @item -momit-leaf-frame-pointer
14133 @opindex momit-leaf-frame-pointer
14134 Don't keep the frame pointer in a register for leaf functions. This
14135 avoids the instructions to save, set up and restore frame pointers and
14136 makes an extra register available in leaf functions. The option
14137 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14138 which might make debugging harder.
14140 @item -mspecld-anomaly
14141 @opindex mspecld-anomaly
14142 When enabled, the compiler ensures that the generated code does not
14143 contain speculative loads after jump instructions. If this option is used,
14144 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14146 @item -mno-specld-anomaly
14147 @opindex mno-specld-anomaly
14148 Don't generate extra code to prevent speculative loads from occurring.
14150 @item -mcsync-anomaly
14151 @opindex mcsync-anomaly
14152 When enabled, the compiler ensures that the generated code does not
14153 contain CSYNC or SSYNC instructions too soon after conditional branches.
14154 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14156 @item -mno-csync-anomaly
14157 @opindex mno-csync-anomaly
14158 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14159 occurring too soon after a conditional branch.
14163 When enabled, the compiler is free to take advantage of the knowledge that
14164 the entire program fits into the low 64k of memory.
14167 @opindex mno-low-64k
14168 Assume that the program is arbitrarily large. This is the default.
14170 @item -mstack-check-l1
14171 @opindex mstack-check-l1
14172 Do stack checking using information placed into L1 scratchpad memory by the
14175 @item -mid-shared-library
14176 @opindex mid-shared-library
14177 Generate code that supports shared libraries via the library ID method.
14178 This allows for execute in place and shared libraries in an environment
14179 without virtual memory management. This option implies @option{-fPIC}.
14180 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14182 @item -mno-id-shared-library
14183 @opindex mno-id-shared-library
14184 Generate code that doesn't assume ID-based shared libraries are being used.
14185 This is the default.
14187 @item -mleaf-id-shared-library
14188 @opindex mleaf-id-shared-library
14189 Generate code that supports shared libraries via the library ID method,
14190 but assumes that this library or executable won't link against any other
14191 ID shared libraries. That allows the compiler to use faster code for jumps
14194 @item -mno-leaf-id-shared-library
14195 @opindex mno-leaf-id-shared-library
14196 Do not assume that the code being compiled won't link against any ID shared
14197 libraries. Slower code is generated for jump and call insns.
14199 @item -mshared-library-id=n
14200 @opindex mshared-library-id
14201 Specifies the identification number of the ID-based shared library being
14202 compiled. Specifying a value of 0 generates more compact code; specifying
14203 other values forces the allocation of that number to the current
14204 library but is no more space- or time-efficient than omitting this option.
14208 Generate code that allows the data segment to be located in a different
14209 area of memory from the text segment. This allows for execute in place in
14210 an environment without virtual memory management by eliminating relocations
14211 against the text section.
14213 @item -mno-sep-data
14214 @opindex mno-sep-data
14215 Generate code that assumes that the data segment follows the text segment.
14216 This is the default.
14219 @itemx -mno-long-calls
14220 @opindex mlong-calls
14221 @opindex mno-long-calls
14222 Tells the compiler to perform function calls by first loading the
14223 address of the function into a register and then performing a subroutine
14224 call on this register. This switch is needed if the target function
14225 lies outside of the 24-bit addressing range of the offset-based
14226 version of subroutine call instruction.
14228 This feature is not enabled by default. Specifying
14229 @option{-mno-long-calls} restores the default behavior. Note these
14230 switches have no effect on how the compiler generates code to handle
14231 function calls via function pointers.
14235 Link with the fast floating-point library. This library relaxes some of
14236 the IEEE floating-point standard's rules for checking inputs against
14237 Not-a-Number (NAN), in the interest of performance.
14240 @opindex minline-plt
14241 Enable inlining of PLT entries in function calls to functions that are
14242 not known to bind locally. It has no effect without @option{-mfdpic}.
14245 @opindex mmulticore
14246 Build a standalone application for multicore Blackfin processors.
14247 This option causes proper start files and link scripts supporting
14248 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14249 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14251 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14252 selects the one-application-per-core programming model. Without
14253 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14254 programming model is used. In this model, the main function of Core B
14255 should be named as @code{coreb_main}.
14257 If this option is not used, the single-core application programming
14262 Build a standalone application for Core A of BF561 when using
14263 the one-application-per-core programming model. Proper start files
14264 and link scripts are used to support Core A, and the macro
14265 @code{__BFIN_COREA} is defined.
14266 This option can only be used in conjunction with @option{-mmulticore}.
14270 Build a standalone application for Core B of BF561 when using
14271 the one-application-per-core programming model. Proper start files
14272 and link scripts are used to support Core B, and the macro
14273 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14274 should be used instead of @code{main}.
14275 This option can only be used in conjunction with @option{-mmulticore}.
14279 Build a standalone application for SDRAM. Proper start files and
14280 link scripts are used to put the application into SDRAM, and the macro
14281 @code{__BFIN_SDRAM} is defined.
14282 The loader should initialize SDRAM before loading the application.
14286 Assume that ICPLBs are enabled at run time. This has an effect on certain
14287 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14288 are enabled; for standalone applications the default is off.
14292 @subsection C6X Options
14293 @cindex C6X Options
14296 @item -march=@var{name}
14298 This specifies the name of the target architecture. GCC uses this
14299 name to determine what kind of instructions it can emit when generating
14300 assembly code. Permissible names are: @samp{c62x},
14301 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14304 @opindex mbig-endian
14305 Generate code for a big-endian target.
14307 @item -mlittle-endian
14308 @opindex mlittle-endian
14309 Generate code for a little-endian target. This is the default.
14313 Choose startup files and linker script suitable for the simulator.
14315 @item -msdata=default
14316 @opindex msdata=default
14317 Put small global and static data in the @code{.neardata} section,
14318 which is pointed to by register @code{B14}. Put small uninitialized
14319 global and static data in the @code{.bss} section, which is adjacent
14320 to the @code{.neardata} section. Put small read-only data into the
14321 @code{.rodata} section. The corresponding sections used for large
14322 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
14325 @opindex msdata=all
14326 Put all data, not just small objects, into the sections reserved for
14327 small data, and use addressing relative to the @code{B14} register to
14331 @opindex msdata=none
14332 Make no use of the sections reserved for small data, and use absolute
14333 addresses to access all data. Put all initialized global and static
14334 data in the @code{.fardata} section, and all uninitialized data in the
14335 @code{.far} section. Put all constant data into the @code{.const}
14340 @subsection CRIS Options
14341 @cindex CRIS Options
14343 These options are defined specifically for the CRIS ports.
14346 @item -march=@var{architecture-type}
14347 @itemx -mcpu=@var{architecture-type}
14350 Generate code for the specified architecture. The choices for
14351 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
14352 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
14353 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14356 @item -mtune=@var{architecture-type}
14358 Tune to @var{architecture-type} everything applicable about the generated
14359 code, except for the ABI and the set of available instructions. The
14360 choices for @var{architecture-type} are the same as for
14361 @option{-march=@var{architecture-type}}.
14363 @item -mmax-stack-frame=@var{n}
14364 @opindex mmax-stack-frame
14365 Warn when the stack frame of a function exceeds @var{n} bytes.
14371 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14372 @option{-march=v3} and @option{-march=v8} respectively.
14374 @item -mmul-bug-workaround
14375 @itemx -mno-mul-bug-workaround
14376 @opindex mmul-bug-workaround
14377 @opindex mno-mul-bug-workaround
14378 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14379 models where it applies. This option is active by default.
14383 Enable CRIS-specific verbose debug-related information in the assembly
14384 code. This option also has the effect of turning off the @samp{#NO_APP}
14385 formatted-code indicator to the assembler at the beginning of the
14390 Do not use condition-code results from previous instruction; always emit
14391 compare and test instructions before use of condition codes.
14393 @item -mno-side-effects
14394 @opindex mno-side-effects
14395 Do not emit instructions with side effects in addressing modes other than
14398 @item -mstack-align
14399 @itemx -mno-stack-align
14400 @itemx -mdata-align
14401 @itemx -mno-data-align
14402 @itemx -mconst-align
14403 @itemx -mno-const-align
14404 @opindex mstack-align
14405 @opindex mno-stack-align
14406 @opindex mdata-align
14407 @opindex mno-data-align
14408 @opindex mconst-align
14409 @opindex mno-const-align
14410 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14411 stack frame, individual data and constants to be aligned for the maximum
14412 single data access size for the chosen CPU model. The default is to
14413 arrange for 32-bit alignment. ABI details such as structure layout are
14414 not affected by these options.
14422 Similar to the stack- data- and const-align options above, these options
14423 arrange for stack frame, writable data and constants to all be 32-bit,
14424 16-bit or 8-bit aligned. The default is 32-bit alignment.
14426 @item -mno-prologue-epilogue
14427 @itemx -mprologue-epilogue
14428 @opindex mno-prologue-epilogue
14429 @opindex mprologue-epilogue
14430 With @option{-mno-prologue-epilogue}, the normal function prologue and
14431 epilogue which set up the stack frame are omitted and no return
14432 instructions or return sequences are generated in the code. Use this
14433 option only together with visual inspection of the compiled code: no
14434 warnings or errors are generated when call-saved registers must be saved,
14435 or storage for local variables needs to be allocated.
14439 @opindex mno-gotplt
14441 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14442 instruction sequences that load addresses for functions from the PLT part
14443 of the GOT rather than (traditional on other architectures) calls to the
14444 PLT@. The default is @option{-mgotplt}.
14448 Legacy no-op option only recognized with the cris-axis-elf and
14449 cris-axis-linux-gnu targets.
14453 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14457 This option, recognized for the cris-axis-elf, arranges
14458 to link with input-output functions from a simulator library. Code,
14459 initialized data and zero-initialized data are allocated consecutively.
14463 Like @option{-sim}, but pass linker options to locate initialized data at
14464 0x40000000 and zero-initialized data at 0x80000000.
14468 @subsection CR16 Options
14469 @cindex CR16 Options
14471 These options are defined specifically for the CR16 ports.
14477 Enable the use of multiply-accumulate instructions. Disabled by default.
14481 @opindex mcr16cplus
14483 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14488 Links the library libsim.a which is in compatible with simulator. Applicable
14489 to ELF compiler only.
14493 Choose integer type as 32-bit wide.
14497 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14499 @item -mdata-model=@var{model}
14500 @opindex mdata-model
14501 Choose a data model. The choices for @var{model} are @samp{near},
14502 @samp{far} or @samp{medium}. @samp{medium} is default.
14503 However, @samp{far} is not valid with @option{-mcr16c}, as the
14504 CR16C architecture does not support the far data model.
14507 @node Darwin Options
14508 @subsection Darwin Options
14509 @cindex Darwin options
14511 These options are defined for all architectures running the Darwin operating
14514 FSF GCC on Darwin does not create ``fat'' object files; it creates
14515 an object file for the single architecture that GCC was built to
14516 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14517 @option{-arch} options are used; it does so by running the compiler or
14518 linker multiple times and joining the results together with
14521 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14522 @samp{i686}) is determined by the flags that specify the ISA
14523 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14524 @option{-force_cpusubtype_ALL} option can be used to override this.
14526 The Darwin tools vary in their behavior when presented with an ISA
14527 mismatch. The assembler, @file{as}, only permits instructions to
14528 be used that are valid for the subtype of the file it is generating,
14529 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14530 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14531 and prints an error if asked to create a shared library with a less
14532 restrictive subtype than its input files (for instance, trying to put
14533 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14534 for executables, @command{ld}, quietly gives the executable the most
14535 restrictive subtype of any of its input files.
14540 Add the framework directory @var{dir} to the head of the list of
14541 directories to be searched for header files. These directories are
14542 interleaved with those specified by @option{-I} options and are
14543 scanned in a left-to-right order.
14545 A framework directory is a directory with frameworks in it. A
14546 framework is a directory with a @file{Headers} and/or
14547 @file{PrivateHeaders} directory contained directly in it that ends
14548 in @file{.framework}. The name of a framework is the name of this
14549 directory excluding the @file{.framework}. Headers associated with
14550 the framework are found in one of those two directories, with
14551 @file{Headers} being searched first. A subframework is a framework
14552 directory that is in a framework's @file{Frameworks} directory.
14553 Includes of subframework headers can only appear in a header of a
14554 framework that contains the subframework, or in a sibling subframework
14555 header. Two subframeworks are siblings if they occur in the same
14556 framework. A subframework should not have the same name as a
14557 framework; a warning is issued if this is violated. Currently a
14558 subframework cannot have subframeworks; in the future, the mechanism
14559 may be extended to support this. The standard frameworks can be found
14560 in @file{/System/Library/Frameworks} and
14561 @file{/Library/Frameworks}. An example include looks like
14562 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14563 the name of the framework and @file{header.h} is found in the
14564 @file{PrivateHeaders} or @file{Headers} directory.
14566 @item -iframework@var{dir}
14567 @opindex iframework
14568 Like @option{-F} except the directory is a treated as a system
14569 directory. The main difference between this @option{-iframework} and
14570 @option{-F} is that with @option{-iframework} the compiler does not
14571 warn about constructs contained within header files found via
14572 @var{dir}. This option is valid only for the C family of languages.
14576 Emit debugging information for symbols that are used. For stabs
14577 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14578 This is by default ON@.
14582 Emit debugging information for all symbols and types.
14584 @item -mmacosx-version-min=@var{version}
14585 The earliest version of MacOS X that this executable will run on
14586 is @var{version}. Typical values of @var{version} include @code{10.1},
14587 @code{10.2}, and @code{10.3.9}.
14589 If the compiler was built to use the system's headers by default,
14590 then the default for this option is the system version on which the
14591 compiler is running, otherwise the default is to make choices that
14592 are compatible with as many systems and code bases as possible.
14596 Enable kernel development mode. The @option{-mkernel} option sets
14597 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14598 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14599 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14600 applicable. This mode also sets @option{-mno-altivec},
14601 @option{-msoft-float}, @option{-fno-builtin} and
14602 @option{-mlong-branch} for PowerPC targets.
14604 @item -mone-byte-bool
14605 @opindex mone-byte-bool
14606 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14607 By default @code{sizeof(bool)} is @code{4} when compiling for
14608 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14609 option has no effect on x86.
14611 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14612 to generate code that is not binary compatible with code generated
14613 without that switch. Using this switch may require recompiling all
14614 other modules in a program, including system libraries. Use this
14615 switch to conform to a non-default data model.
14617 @item -mfix-and-continue
14618 @itemx -ffix-and-continue
14619 @itemx -findirect-data
14620 @opindex mfix-and-continue
14621 @opindex ffix-and-continue
14622 @opindex findirect-data
14623 Generate code suitable for fast turnaround development, such as to
14624 allow GDB to dynamically load @file{.o} files into already-running
14625 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14626 are provided for backwards compatibility.
14630 Loads all members of static archive libraries.
14631 See man ld(1) for more information.
14633 @item -arch_errors_fatal
14634 @opindex arch_errors_fatal
14635 Cause the errors having to do with files that have the wrong architecture
14638 @item -bind_at_load
14639 @opindex bind_at_load
14640 Causes the output file to be marked such that the dynamic linker will
14641 bind all undefined references when the file is loaded or launched.
14645 Produce a Mach-o bundle format file.
14646 See man ld(1) for more information.
14648 @item -bundle_loader @var{executable}
14649 @opindex bundle_loader
14650 This option specifies the @var{executable} that will load the build
14651 output file being linked. See man ld(1) for more information.
14654 @opindex dynamiclib
14655 When passed this option, GCC produces a dynamic library instead of
14656 an executable when linking, using the Darwin @file{libtool} command.
14658 @item -force_cpusubtype_ALL
14659 @opindex force_cpusubtype_ALL
14660 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14661 one controlled by the @option{-mcpu} or @option{-march} option.
14663 @item -allowable_client @var{client_name}
14664 @itemx -client_name
14665 @itemx -compatibility_version
14666 @itemx -current_version
14668 @itemx -dependency-file
14670 @itemx -dylinker_install_name
14672 @itemx -exported_symbols_list
14675 @itemx -flat_namespace
14676 @itemx -force_flat_namespace
14677 @itemx -headerpad_max_install_names
14680 @itemx -install_name
14681 @itemx -keep_private_externs
14682 @itemx -multi_module
14683 @itemx -multiply_defined
14684 @itemx -multiply_defined_unused
14687 @itemx -no_dead_strip_inits_and_terms
14688 @itemx -nofixprebinding
14689 @itemx -nomultidefs
14691 @itemx -noseglinkedit
14692 @itemx -pagezero_size
14694 @itemx -prebind_all_twolevel_modules
14695 @itemx -private_bundle
14697 @itemx -read_only_relocs
14699 @itemx -sectobjectsymbols
14703 @itemx -sectobjectsymbols
14706 @itemx -segs_read_only_addr
14708 @itemx -segs_read_write_addr
14709 @itemx -seg_addr_table
14710 @itemx -seg_addr_table_filename
14711 @itemx -seglinkedit
14713 @itemx -segs_read_only_addr
14714 @itemx -segs_read_write_addr
14715 @itemx -single_module
14717 @itemx -sub_library
14719 @itemx -sub_umbrella
14720 @itemx -twolevel_namespace
14723 @itemx -unexported_symbols_list
14724 @itemx -weak_reference_mismatches
14725 @itemx -whatsloaded
14726 @opindex allowable_client
14727 @opindex client_name
14728 @opindex compatibility_version
14729 @opindex current_version
14730 @opindex dead_strip
14731 @opindex dependency-file
14732 @opindex dylib_file
14733 @opindex dylinker_install_name
14735 @opindex exported_symbols_list
14737 @opindex flat_namespace
14738 @opindex force_flat_namespace
14739 @opindex headerpad_max_install_names
14740 @opindex image_base
14742 @opindex install_name
14743 @opindex keep_private_externs
14744 @opindex multi_module
14745 @opindex multiply_defined
14746 @opindex multiply_defined_unused
14747 @opindex noall_load
14748 @opindex no_dead_strip_inits_and_terms
14749 @opindex nofixprebinding
14750 @opindex nomultidefs
14752 @opindex noseglinkedit
14753 @opindex pagezero_size
14755 @opindex prebind_all_twolevel_modules
14756 @opindex private_bundle
14757 @opindex read_only_relocs
14759 @opindex sectobjectsymbols
14762 @opindex sectcreate
14763 @opindex sectobjectsymbols
14766 @opindex segs_read_only_addr
14767 @opindex segs_read_write_addr
14768 @opindex seg_addr_table
14769 @opindex seg_addr_table_filename
14770 @opindex seglinkedit
14772 @opindex segs_read_only_addr
14773 @opindex segs_read_write_addr
14774 @opindex single_module
14776 @opindex sub_library
14777 @opindex sub_umbrella
14778 @opindex twolevel_namespace
14781 @opindex unexported_symbols_list
14782 @opindex weak_reference_mismatches
14783 @opindex whatsloaded
14784 These options are passed to the Darwin linker. The Darwin linker man page
14785 describes them in detail.
14788 @node DEC Alpha Options
14789 @subsection DEC Alpha Options
14791 These @samp{-m} options are defined for the DEC Alpha implementations:
14794 @item -mno-soft-float
14795 @itemx -msoft-float
14796 @opindex mno-soft-float
14797 @opindex msoft-float
14798 Use (do not use) the hardware floating-point instructions for
14799 floating-point operations. When @option{-msoft-float} is specified,
14800 functions in @file{libgcc.a} are used to perform floating-point
14801 operations. Unless they are replaced by routines that emulate the
14802 floating-point operations, or compiled in such a way as to call such
14803 emulations routines, these routines issue floating-point
14804 operations. If you are compiling for an Alpha without floating-point
14805 operations, you must ensure that the library is built so as not to call
14808 Note that Alpha implementations without floating-point operations are
14809 required to have floating-point registers.
14812 @itemx -mno-fp-regs
14814 @opindex mno-fp-regs
14815 Generate code that uses (does not use) the floating-point register set.
14816 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14817 register set is not used, floating-point operands are passed in integer
14818 registers as if they were integers and floating-point results are passed
14819 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14820 so any function with a floating-point argument or return value called by code
14821 compiled with @option{-mno-fp-regs} must also be compiled with that
14824 A typical use of this option is building a kernel that does not use,
14825 and hence need not save and restore, any floating-point registers.
14829 The Alpha architecture implements floating-point hardware optimized for
14830 maximum performance. It is mostly compliant with the IEEE floating-point
14831 standard. However, for full compliance, software assistance is
14832 required. This option generates code fully IEEE-compliant code
14833 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14834 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14835 defined during compilation. The resulting code is less efficient but is
14836 able to correctly support denormalized numbers and exceptional IEEE
14837 values such as not-a-number and plus/minus infinity. Other Alpha
14838 compilers call this option @option{-ieee_with_no_inexact}.
14840 @item -mieee-with-inexact
14841 @opindex mieee-with-inexact
14842 This is like @option{-mieee} except the generated code also maintains
14843 the IEEE @var{inexact-flag}. Turning on this option causes the
14844 generated code to implement fully-compliant IEEE math. In addition to
14845 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14846 macro. On some Alpha implementations the resulting code may execute
14847 significantly slower than the code generated by default. Since there is
14848 very little code that depends on the @var{inexact-flag}, you should
14849 normally not specify this option. Other Alpha compilers call this
14850 option @option{-ieee_with_inexact}.
14852 @item -mfp-trap-mode=@var{trap-mode}
14853 @opindex mfp-trap-mode
14854 This option controls what floating-point related traps are enabled.
14855 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14856 The trap mode can be set to one of four values:
14860 This is the default (normal) setting. The only traps that are enabled
14861 are the ones that cannot be disabled in software (e.g., division by zero
14865 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14869 Like @samp{u}, but the instructions are marked to be safe for software
14870 completion (see Alpha architecture manual for details).
14873 Like @samp{su}, but inexact traps are enabled as well.
14876 @item -mfp-rounding-mode=@var{rounding-mode}
14877 @opindex mfp-rounding-mode
14878 Selects the IEEE rounding mode. Other Alpha compilers call this option
14879 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14884 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14885 the nearest machine number or towards the even machine number in case
14889 Round towards minus infinity.
14892 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14895 Dynamic rounding mode. A field in the floating-point control register
14896 (@var{fpcr}, see Alpha architecture reference manual) controls the
14897 rounding mode in effect. The C library initializes this register for
14898 rounding towards plus infinity. Thus, unless your program modifies the
14899 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14902 @item -mtrap-precision=@var{trap-precision}
14903 @opindex mtrap-precision
14904 In the Alpha architecture, floating-point traps are imprecise. This
14905 means without software assistance it is impossible to recover from a
14906 floating trap and program execution normally needs to be terminated.
14907 GCC can generate code that can assist operating system trap handlers
14908 in determining the exact location that caused a floating-point trap.
14909 Depending on the requirements of an application, different levels of
14910 precisions can be selected:
14914 Program precision. This option is the default and means a trap handler
14915 can only identify which program caused a floating-point exception.
14918 Function precision. The trap handler can determine the function that
14919 caused a floating-point exception.
14922 Instruction precision. The trap handler can determine the exact
14923 instruction that caused a floating-point exception.
14926 Other Alpha compilers provide the equivalent options called
14927 @option{-scope_safe} and @option{-resumption_safe}.
14929 @item -mieee-conformant
14930 @opindex mieee-conformant
14931 This option marks the generated code as IEEE conformant. You must not
14932 use this option unless you also specify @option{-mtrap-precision=i} and either
14933 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14934 is to emit the line @samp{.eflag 48} in the function prologue of the
14935 generated assembly file.
14937 @item -mbuild-constants
14938 @opindex mbuild-constants
14939 Normally GCC examines a 32- or 64-bit integer constant to
14940 see if it can construct it from smaller constants in two or three
14941 instructions. If it cannot, it outputs the constant as a literal and
14942 generates code to load it from the data segment at run time.
14944 Use this option to require GCC to construct @emph{all} integer constants
14945 using code, even if it takes more instructions (the maximum is six).
14947 You typically use this option to build a shared library dynamic
14948 loader. Itself a shared library, it must relocate itself in memory
14949 before it can find the variables and constants in its own data segment.
14967 Indicate whether GCC should generate code to use the optional BWX,
14968 CIX, FIX and MAX instruction sets. The default is to use the instruction
14969 sets supported by the CPU type specified via @option{-mcpu=} option or that
14970 of the CPU on which GCC was built if none is specified.
14973 @itemx -mfloat-ieee
14974 @opindex mfloat-vax
14975 @opindex mfloat-ieee
14976 Generate code that uses (does not use) VAX F and G floating-point
14977 arithmetic instead of IEEE single and double precision.
14979 @item -mexplicit-relocs
14980 @itemx -mno-explicit-relocs
14981 @opindex mexplicit-relocs
14982 @opindex mno-explicit-relocs
14983 Older Alpha assemblers provided no way to generate symbol relocations
14984 except via assembler macros. Use of these macros does not allow
14985 optimal instruction scheduling. GNU binutils as of version 2.12
14986 supports a new syntax that allows the compiler to explicitly mark
14987 which relocations should apply to which instructions. This option
14988 is mostly useful for debugging, as GCC detects the capabilities of
14989 the assembler when it is built and sets the default accordingly.
14992 @itemx -mlarge-data
14993 @opindex msmall-data
14994 @opindex mlarge-data
14995 When @option{-mexplicit-relocs} is in effect, static data is
14996 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14997 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14998 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14999 16-bit relocations off of the @code{$gp} register. This limits the
15000 size of the small data area to 64KB, but allows the variables to be
15001 directly accessed via a single instruction.
15003 The default is @option{-mlarge-data}. With this option the data area
15004 is limited to just below 2GB@. Programs that require more than 2GB of
15005 data must use @code{malloc} or @code{mmap} to allocate the data in the
15006 heap instead of in the program's data segment.
15008 When generating code for shared libraries, @option{-fpic} implies
15009 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15012 @itemx -mlarge-text
15013 @opindex msmall-text
15014 @opindex mlarge-text
15015 When @option{-msmall-text} is used, the compiler assumes that the
15016 code of the entire program (or shared library) fits in 4MB, and is
15017 thus reachable with a branch instruction. When @option{-msmall-data}
15018 is used, the compiler can assume that all local symbols share the
15019 same @code{$gp} value, and thus reduce the number of instructions
15020 required for a function call from 4 to 1.
15022 The default is @option{-mlarge-text}.
15024 @item -mcpu=@var{cpu_type}
15026 Set the instruction set and instruction scheduling parameters for
15027 machine type @var{cpu_type}. You can specify either the @samp{EV}
15028 style name or the corresponding chip number. GCC supports scheduling
15029 parameters for the EV4, EV5 and EV6 family of processors and
15030 chooses the default values for the instruction set from the processor
15031 you specify. If you do not specify a processor type, GCC defaults
15032 to the processor on which the compiler was built.
15034 Supported values for @var{cpu_type} are
15040 Schedules as an EV4 and has no instruction set extensions.
15044 Schedules as an EV5 and has no instruction set extensions.
15048 Schedules as an EV5 and supports the BWX extension.
15053 Schedules as an EV5 and supports the BWX and MAX extensions.
15057 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15061 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15064 Native toolchains also support the value @samp{native},
15065 which selects the best architecture option for the host processor.
15066 @option{-mcpu=native} has no effect if GCC does not recognize
15069 @item -mtune=@var{cpu_type}
15071 Set only the instruction scheduling parameters for machine type
15072 @var{cpu_type}. The instruction set is not changed.
15074 Native toolchains also support the value @samp{native},
15075 which selects the best architecture option for the host processor.
15076 @option{-mtune=native} has no effect if GCC does not recognize
15079 @item -mmemory-latency=@var{time}
15080 @opindex mmemory-latency
15081 Sets the latency the scheduler should assume for typical memory
15082 references as seen by the application. This number is highly
15083 dependent on the memory access patterns used by the application
15084 and the size of the external cache on the machine.
15086 Valid options for @var{time} are
15090 A decimal number representing clock cycles.
15096 The compiler contains estimates of the number of clock cycles for
15097 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15098 (also called Dcache, Scache, and Bcache), as well as to main memory.
15099 Note that L3 is only valid for EV5.
15105 @subsection FR30 Options
15106 @cindex FR30 Options
15108 These options are defined specifically for the FR30 port.
15112 @item -msmall-model
15113 @opindex msmall-model
15114 Use the small address space model. This can produce smaller code, but
15115 it does assume that all symbolic values and addresses fit into a
15120 Assume that runtime support has been provided and so there is no need
15121 to include the simulator library (@file{libsim.a}) on the linker
15127 @subsection FT32 Options
15128 @cindex FT32 Options
15130 These options are defined specifically for the FT32 port.
15136 Specifies that the program will be run on the simulator. This causes
15137 an alternate runtime startup and library to be linked.
15138 You must not use this option when generating programs that will run on
15139 real hardware; you must provide your own runtime library for whatever
15140 I/O functions are needed.
15144 Enable Local Register Allocation. This is still experimental for FT32,
15145 so by default the compiler uses standard reload.
15150 @subsection FRV Options
15151 @cindex FRV Options
15157 Only use the first 32 general-purpose registers.
15162 Use all 64 general-purpose registers.
15167 Use only the first 32 floating-point registers.
15172 Use all 64 floating-point registers.
15175 @opindex mhard-float
15177 Use hardware instructions for floating-point operations.
15180 @opindex msoft-float
15182 Use library routines for floating-point operations.
15187 Dynamically allocate condition code registers.
15192 Do not try to dynamically allocate condition code registers, only
15193 use @code{icc0} and @code{fcc0}.
15198 Change ABI to use double word insns.
15203 Do not use double word instructions.
15208 Use floating-point double instructions.
15211 @opindex mno-double
15213 Do not use floating-point double instructions.
15218 Use media instructions.
15223 Do not use media instructions.
15228 Use multiply and add/subtract instructions.
15231 @opindex mno-muladd
15233 Do not use multiply and add/subtract instructions.
15238 Select the FDPIC ABI, which uses function descriptors to represent
15239 pointers to functions. Without any PIC/PIE-related options, it
15240 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15241 assumes GOT entries and small data are within a 12-bit range from the
15242 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15243 are computed with 32 bits.
15244 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15247 @opindex minline-plt
15249 Enable inlining of PLT entries in function calls to functions that are
15250 not known to bind locally. It has no effect without @option{-mfdpic}.
15251 It's enabled by default if optimizing for speed and compiling for
15252 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15253 optimization option such as @option{-O3} or above is present in the
15259 Assume a large TLS segment when generating thread-local code.
15264 Do not assume a large TLS segment when generating thread-local code.
15269 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15270 that is known to be in read-only sections. It's enabled by default,
15271 except for @option{-fpic} or @option{-fpie}: even though it may help
15272 make the global offset table smaller, it trades 1 instruction for 4.
15273 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15274 one of which may be shared by multiple symbols, and it avoids the need
15275 for a GOT entry for the referenced symbol, so it's more likely to be a
15276 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15278 @item -multilib-library-pic
15279 @opindex multilib-library-pic
15281 Link with the (library, not FD) pic libraries. It's implied by
15282 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15283 @option{-fpic} without @option{-mfdpic}. You should never have to use
15287 @opindex mlinked-fp
15289 Follow the EABI requirement of always creating a frame pointer whenever
15290 a stack frame is allocated. This option is enabled by default and can
15291 be disabled with @option{-mno-linked-fp}.
15294 @opindex mlong-calls
15296 Use indirect addressing to call functions outside the current
15297 compilation unit. This allows the functions to be placed anywhere
15298 within the 32-bit address space.
15300 @item -malign-labels
15301 @opindex malign-labels
15303 Try to align labels to an 8-byte boundary by inserting NOPs into the
15304 previous packet. This option only has an effect when VLIW packing
15305 is enabled. It doesn't create new packets; it merely adds NOPs to
15308 @item -mlibrary-pic
15309 @opindex mlibrary-pic
15311 Generate position-independent EABI code.
15316 Use only the first four media accumulator registers.
15321 Use all eight media accumulator registers.
15326 Pack VLIW instructions.
15331 Do not pack VLIW instructions.
15334 @opindex mno-eflags
15336 Do not mark ABI switches in e_flags.
15339 @opindex mcond-move
15341 Enable the use of conditional-move instructions (default).
15343 This switch is mainly for debugging the compiler and will likely be removed
15344 in a future version.
15346 @item -mno-cond-move
15347 @opindex mno-cond-move
15349 Disable the use of conditional-move instructions.
15351 This switch is mainly for debugging the compiler and will likely be removed
15352 in a future version.
15357 Enable the use of conditional set instructions (default).
15359 This switch is mainly for debugging the compiler and will likely be removed
15360 in a future version.
15365 Disable the use of conditional set instructions.
15367 This switch is mainly for debugging the compiler and will likely be removed
15368 in a future version.
15371 @opindex mcond-exec
15373 Enable the use of conditional execution (default).
15375 This switch is mainly for debugging the compiler and will likely be removed
15376 in a future version.
15378 @item -mno-cond-exec
15379 @opindex mno-cond-exec
15381 Disable the use of conditional execution.
15383 This switch is mainly for debugging the compiler and will likely be removed
15384 in a future version.
15386 @item -mvliw-branch
15387 @opindex mvliw-branch
15389 Run a pass to pack branches into VLIW instructions (default).
15391 This switch is mainly for debugging the compiler and will likely be removed
15392 in a future version.
15394 @item -mno-vliw-branch
15395 @opindex mno-vliw-branch
15397 Do not run a pass to pack branches into VLIW instructions.
15399 This switch is mainly for debugging the compiler and will likely be removed
15400 in a future version.
15402 @item -mmulti-cond-exec
15403 @opindex mmulti-cond-exec
15405 Enable optimization of @code{&&} and @code{||} in conditional execution
15408 This switch is mainly for debugging the compiler and will likely be removed
15409 in a future version.
15411 @item -mno-multi-cond-exec
15412 @opindex mno-multi-cond-exec
15414 Disable optimization of @code{&&} and @code{||} in conditional execution.
15416 This switch is mainly for debugging the compiler and will likely be removed
15417 in a future version.
15419 @item -mnested-cond-exec
15420 @opindex mnested-cond-exec
15422 Enable nested conditional execution optimizations (default).
15424 This switch is mainly for debugging the compiler and will likely be removed
15425 in a future version.
15427 @item -mno-nested-cond-exec
15428 @opindex mno-nested-cond-exec
15430 Disable nested conditional execution optimizations.
15432 This switch is mainly for debugging the compiler and will likely be removed
15433 in a future version.
15435 @item -moptimize-membar
15436 @opindex moptimize-membar
15438 This switch removes redundant @code{membar} instructions from the
15439 compiler-generated code. It is enabled by default.
15441 @item -mno-optimize-membar
15442 @opindex mno-optimize-membar
15444 This switch disables the automatic removal of redundant @code{membar}
15445 instructions from the generated code.
15447 @item -mtomcat-stats
15448 @opindex mtomcat-stats
15450 Cause gas to print out tomcat statistics.
15452 @item -mcpu=@var{cpu}
15455 Select the processor type for which to generate code. Possible values are
15456 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15457 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15461 @node GNU/Linux Options
15462 @subsection GNU/Linux Options
15464 These @samp{-m} options are defined for GNU/Linux targets:
15469 Use the GNU C library. This is the default except
15470 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
15471 @samp{*-*-linux-*android*} targets.
15475 Use uClibc C library. This is the default on
15476 @samp{*-*-linux-*uclibc*} targets.
15480 Use the musl C library. This is the default on
15481 @samp{*-*-linux-*musl*} targets.
15485 Use Bionic C library. This is the default on
15486 @samp{*-*-linux-*android*} targets.
15490 Compile code compatible with Android platform. This is the default on
15491 @samp{*-*-linux-*android*} targets.
15493 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15494 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15495 this option makes the GCC driver pass Android-specific options to the linker.
15496 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15499 @item -tno-android-cc
15500 @opindex tno-android-cc
15501 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15502 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15503 @option{-fno-rtti} by default.
15505 @item -tno-android-ld
15506 @opindex tno-android-ld
15507 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15508 linking options to the linker.
15512 @node H8/300 Options
15513 @subsection H8/300 Options
15515 These @samp{-m} options are defined for the H8/300 implementations:
15520 Shorten some address references at link time, when possible; uses the
15521 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15522 ld, Using ld}, for a fuller description.
15526 Generate code for the H8/300H@.
15530 Generate code for the H8S@.
15534 Generate code for the H8S and H8/300H in the normal mode. This switch
15535 must be used either with @option{-mh} or @option{-ms}.
15539 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15543 Extended registers are stored on stack before execution of function
15544 with monitor attribute. Default option is @option{-mexr}.
15545 This option is valid only for H8S targets.
15549 Extended registers are not stored on stack before execution of function
15550 with monitor attribute. Default option is @option{-mno-exr}.
15551 This option is valid only for H8S targets.
15555 Make @code{int} data 32 bits by default.
15558 @opindex malign-300
15559 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15560 The default for the H8/300H and H8S is to align longs and floats on
15562 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15563 This option has no effect on the H8/300.
15567 @subsection HPPA Options
15568 @cindex HPPA Options
15570 These @samp{-m} options are defined for the HPPA family of computers:
15573 @item -march=@var{architecture-type}
15575 Generate code for the specified architecture. The choices for
15576 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15577 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15578 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15579 architecture option for your machine. Code compiled for lower numbered
15580 architectures runs on higher numbered architectures, but not the
15583 @item -mpa-risc-1-0
15584 @itemx -mpa-risc-1-1
15585 @itemx -mpa-risc-2-0
15586 @opindex mpa-risc-1-0
15587 @opindex mpa-risc-1-1
15588 @opindex mpa-risc-2-0
15589 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15591 @item -mjump-in-delay
15592 @opindex mjump-in-delay
15593 This option is ignored and provided for compatibility purposes only.
15595 @item -mdisable-fpregs
15596 @opindex mdisable-fpregs
15597 Prevent floating-point registers from being used in any manner. This is
15598 necessary for compiling kernels that perform lazy context switching of
15599 floating-point registers. If you use this option and attempt to perform
15600 floating-point operations, the compiler aborts.
15602 @item -mdisable-indexing
15603 @opindex mdisable-indexing
15604 Prevent the compiler from using indexing address modes. This avoids some
15605 rather obscure problems when compiling MIG generated code under MACH@.
15607 @item -mno-space-regs
15608 @opindex mno-space-regs
15609 Generate code that assumes the target has no space registers. This allows
15610 GCC to generate faster indirect calls and use unscaled index address modes.
15612 Such code is suitable for level 0 PA systems and kernels.
15614 @item -mfast-indirect-calls
15615 @opindex mfast-indirect-calls
15616 Generate code that assumes calls never cross space boundaries. This
15617 allows GCC to emit code that performs faster indirect calls.
15619 This option does not work in the presence of shared libraries or nested
15622 @item -mfixed-range=@var{register-range}
15623 @opindex mfixed-range
15624 Generate code treating the given register range as fixed registers.
15625 A fixed register is one that the register allocator cannot use. This is
15626 useful when compiling kernel code. A register range is specified as
15627 two registers separated by a dash. Multiple register ranges can be
15628 specified separated by a comma.
15630 @item -mlong-load-store
15631 @opindex mlong-load-store
15632 Generate 3-instruction load and store sequences as sometimes required by
15633 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15636 @item -mportable-runtime
15637 @opindex mportable-runtime
15638 Use the portable calling conventions proposed by HP for ELF systems.
15642 Enable the use of assembler directives only GAS understands.
15644 @item -mschedule=@var{cpu-type}
15646 Schedule code according to the constraints for the machine type
15647 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15648 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15649 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15650 proper scheduling option for your machine. The default scheduling is
15654 @opindex mlinker-opt
15655 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15656 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15657 linkers in which they give bogus error messages when linking some programs.
15660 @opindex msoft-float
15661 Generate output containing library calls for floating point.
15662 @strong{Warning:} the requisite libraries are not available for all HPPA
15663 targets. Normally the facilities of the machine's usual C compiler are
15664 used, but this cannot be done directly in cross-compilation. You must make
15665 your own arrangements to provide suitable library functions for
15668 @option{-msoft-float} changes the calling convention in the output file;
15669 therefore, it is only useful if you compile @emph{all} of a program with
15670 this option. In particular, you need to compile @file{libgcc.a}, the
15671 library that comes with GCC, with @option{-msoft-float} in order for
15676 Generate the predefine, @code{_SIO}, for server IO@. The default is
15677 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15678 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15679 options are available under HP-UX and HI-UX@.
15683 Use options specific to GNU @command{ld}.
15684 This passes @option{-shared} to @command{ld} when
15685 building a shared library. It is the default when GCC is configured,
15686 explicitly or implicitly, with the GNU linker. This option does not
15687 affect which @command{ld} is called; it only changes what parameters
15688 are passed to that @command{ld}.
15689 The @command{ld} that is called is determined by the
15690 @option{--with-ld} configure option, GCC's program search path, and
15691 finally by the user's @env{PATH}. The linker used by GCC can be printed
15692 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15693 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15697 Use options specific to HP @command{ld}.
15698 This passes @option{-b} to @command{ld} when building
15699 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15700 links. It is the default when GCC is configured, explicitly or
15701 implicitly, with the HP linker. This option does not affect
15702 which @command{ld} is called; it only changes what parameters are passed to that
15704 The @command{ld} that is called is determined by the @option{--with-ld}
15705 configure option, GCC's program search path, and finally by the user's
15706 @env{PATH}. The linker used by GCC can be printed using @samp{which
15707 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15708 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15711 @opindex mno-long-calls
15712 Generate code that uses long call sequences. This ensures that a call
15713 is always able to reach linker generated stubs. The default is to generate
15714 long calls only when the distance from the call site to the beginning
15715 of the function or translation unit, as the case may be, exceeds a
15716 predefined limit set by the branch type being used. The limits for
15717 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15718 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15721 Distances are measured from the beginning of functions when using the
15722 @option{-ffunction-sections} option, or when using the @option{-mgas}
15723 and @option{-mno-portable-runtime} options together under HP-UX with
15726 It is normally not desirable to use this option as it degrades
15727 performance. However, it may be useful in large applications,
15728 particularly when partial linking is used to build the application.
15730 The types of long calls used depends on the capabilities of the
15731 assembler and linker, and the type of code being generated. The
15732 impact on systems that support long absolute calls, and long pic
15733 symbol-difference or pc-relative calls should be relatively small.
15734 However, an indirect call is used on 32-bit ELF systems in pic code
15735 and it is quite long.
15737 @item -munix=@var{unix-std}
15739 Generate compiler predefines and select a startfile for the specified
15740 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15741 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15742 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15743 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15744 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15747 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15748 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15749 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15750 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15751 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15752 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15754 It is @emph{important} to note that this option changes the interfaces
15755 for various library routines. It also affects the operational behavior
15756 of the C library. Thus, @emph{extreme} care is needed in using this
15759 Library code that is intended to operate with more than one UNIX
15760 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15761 as appropriate. Most GNU software doesn't provide this capability.
15765 Suppress the generation of link options to search libdld.sl when the
15766 @option{-static} option is specified on HP-UX 10 and later.
15770 The HP-UX implementation of setlocale in libc has a dependency on
15771 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15772 when the @option{-static} option is specified, special link options
15773 are needed to resolve this dependency.
15775 On HP-UX 10 and later, the GCC driver adds the necessary options to
15776 link with libdld.sl when the @option{-static} option is specified.
15777 This causes the resulting binary to be dynamic. On the 64-bit port,
15778 the linkers generate dynamic binaries by default in any case. The
15779 @option{-nolibdld} option can be used to prevent the GCC driver from
15780 adding these link options.
15784 Add support for multithreading with the @dfn{dce thread} library
15785 under HP-UX@. This option sets flags for both the preprocessor and
15789 @node IA-64 Options
15790 @subsection IA-64 Options
15791 @cindex IA-64 Options
15793 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15797 @opindex mbig-endian
15798 Generate code for a big-endian target. This is the default for HP-UX@.
15800 @item -mlittle-endian
15801 @opindex mlittle-endian
15802 Generate code for a little-endian target. This is the default for AIX5
15808 @opindex mno-gnu-as
15809 Generate (or don't) code for the GNU assembler. This is the default.
15810 @c Also, this is the default if the configure option @option{--with-gnu-as}
15816 @opindex mno-gnu-ld
15817 Generate (or don't) code for the GNU linker. This is the default.
15818 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15823 Generate code that does not use a global pointer register. The result
15824 is not position independent code, and violates the IA-64 ABI@.
15826 @item -mvolatile-asm-stop
15827 @itemx -mno-volatile-asm-stop
15828 @opindex mvolatile-asm-stop
15829 @opindex mno-volatile-asm-stop
15830 Generate (or don't) a stop bit immediately before and after volatile asm
15833 @item -mregister-names
15834 @itemx -mno-register-names
15835 @opindex mregister-names
15836 @opindex mno-register-names
15837 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15838 the stacked registers. This may make assembler output more readable.
15844 Disable (or enable) optimizations that use the small data section. This may
15845 be useful for working around optimizer bugs.
15847 @item -mconstant-gp
15848 @opindex mconstant-gp
15849 Generate code that uses a single constant global pointer value. This is
15850 useful when compiling kernel code.
15854 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15855 This is useful when compiling firmware code.
15857 @item -minline-float-divide-min-latency
15858 @opindex minline-float-divide-min-latency
15859 Generate code for inline divides of floating-point values
15860 using the minimum latency algorithm.
15862 @item -minline-float-divide-max-throughput
15863 @opindex minline-float-divide-max-throughput
15864 Generate code for inline divides of floating-point values
15865 using the maximum throughput algorithm.
15867 @item -mno-inline-float-divide
15868 @opindex mno-inline-float-divide
15869 Do not generate inline code for divides of floating-point values.
15871 @item -minline-int-divide-min-latency
15872 @opindex minline-int-divide-min-latency
15873 Generate code for inline divides of integer values
15874 using the minimum latency algorithm.
15876 @item -minline-int-divide-max-throughput
15877 @opindex minline-int-divide-max-throughput
15878 Generate code for inline divides of integer values
15879 using the maximum throughput algorithm.
15881 @item -mno-inline-int-divide
15882 @opindex mno-inline-int-divide
15883 Do not generate inline code for divides of integer values.
15885 @item -minline-sqrt-min-latency
15886 @opindex minline-sqrt-min-latency
15887 Generate code for inline square roots
15888 using the minimum latency algorithm.
15890 @item -minline-sqrt-max-throughput
15891 @opindex minline-sqrt-max-throughput
15892 Generate code for inline square roots
15893 using the maximum throughput algorithm.
15895 @item -mno-inline-sqrt
15896 @opindex mno-inline-sqrt
15897 Do not generate inline code for @code{sqrt}.
15900 @itemx -mno-fused-madd
15901 @opindex mfused-madd
15902 @opindex mno-fused-madd
15903 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
15904 instructions. The default is to use these instructions.
15906 @item -mno-dwarf2-asm
15907 @itemx -mdwarf2-asm
15908 @opindex mno-dwarf2-asm
15909 @opindex mdwarf2-asm
15910 Don't (or do) generate assembler code for the DWARF 2 line number debugging
15911 info. This may be useful when not using the GNU assembler.
15913 @item -mearly-stop-bits
15914 @itemx -mno-early-stop-bits
15915 @opindex mearly-stop-bits
15916 @opindex mno-early-stop-bits
15917 Allow stop bits to be placed earlier than immediately preceding the
15918 instruction that triggered the stop bit. This can improve instruction
15919 scheduling, but does not always do so.
15921 @item -mfixed-range=@var{register-range}
15922 @opindex mfixed-range
15923 Generate code treating the given register range as fixed registers.
15924 A fixed register is one that the register allocator cannot use. This is
15925 useful when compiling kernel code. A register range is specified as
15926 two registers separated by a dash. Multiple register ranges can be
15927 specified separated by a comma.
15929 @item -mtls-size=@var{tls-size}
15931 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
15934 @item -mtune=@var{cpu-type}
15936 Tune the instruction scheduling for a particular CPU, Valid values are
15937 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
15938 and @samp{mckinley}.
15944 Generate code for a 32-bit or 64-bit environment.
15945 The 32-bit environment sets int, long and pointer to 32 bits.
15946 The 64-bit environment sets int to 32 bits and long and pointer
15947 to 64 bits. These are HP-UX specific flags.
15949 @item -mno-sched-br-data-spec
15950 @itemx -msched-br-data-spec
15951 @opindex mno-sched-br-data-spec
15952 @opindex msched-br-data-spec
15953 (Dis/En)able data speculative scheduling before reload.
15954 This results in generation of @code{ld.a} instructions and
15955 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15956 The default is 'disable'.
15958 @item -msched-ar-data-spec
15959 @itemx -mno-sched-ar-data-spec
15960 @opindex msched-ar-data-spec
15961 @opindex mno-sched-ar-data-spec
15962 (En/Dis)able data speculative scheduling after reload.
15963 This results in generation of @code{ld.a} instructions and
15964 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15965 The default is 'enable'.
15967 @item -mno-sched-control-spec
15968 @itemx -msched-control-spec
15969 @opindex mno-sched-control-spec
15970 @opindex msched-control-spec
15971 (Dis/En)able control speculative scheduling. This feature is
15972 available only during region scheduling (i.e.@: before reload).
15973 This results in generation of the @code{ld.s} instructions and
15974 the corresponding check instructions @code{chk.s}.
15975 The default is 'disable'.
15977 @item -msched-br-in-data-spec
15978 @itemx -mno-sched-br-in-data-spec
15979 @opindex msched-br-in-data-spec
15980 @opindex mno-sched-br-in-data-spec
15981 (En/Dis)able speculative scheduling of the instructions that
15982 are dependent on the data speculative loads before reload.
15983 This is effective only with @option{-msched-br-data-spec} enabled.
15984 The default is 'enable'.
15986 @item -msched-ar-in-data-spec
15987 @itemx -mno-sched-ar-in-data-spec
15988 @opindex msched-ar-in-data-spec
15989 @opindex mno-sched-ar-in-data-spec
15990 (En/Dis)able speculative scheduling of the instructions that
15991 are dependent on the data speculative loads after reload.
15992 This is effective only with @option{-msched-ar-data-spec} enabled.
15993 The default is 'enable'.
15995 @item -msched-in-control-spec
15996 @itemx -mno-sched-in-control-spec
15997 @opindex msched-in-control-spec
15998 @opindex mno-sched-in-control-spec
15999 (En/Dis)able speculative scheduling of the instructions that
16000 are dependent on the control speculative loads.
16001 This is effective only with @option{-msched-control-spec} enabled.
16002 The default is 'enable'.
16004 @item -mno-sched-prefer-non-data-spec-insns
16005 @itemx -msched-prefer-non-data-spec-insns
16006 @opindex mno-sched-prefer-non-data-spec-insns
16007 @opindex msched-prefer-non-data-spec-insns
16008 If enabled, data-speculative instructions are chosen for schedule
16009 only if there are no other choices at the moment. This makes
16010 the use of the data speculation much more conservative.
16011 The default is 'disable'.
16013 @item -mno-sched-prefer-non-control-spec-insns
16014 @itemx -msched-prefer-non-control-spec-insns
16015 @opindex mno-sched-prefer-non-control-spec-insns
16016 @opindex msched-prefer-non-control-spec-insns
16017 If enabled, control-speculative instructions are chosen for schedule
16018 only if there are no other choices at the moment. This makes
16019 the use of the control speculation much more conservative.
16020 The default is 'disable'.
16022 @item -mno-sched-count-spec-in-critical-path
16023 @itemx -msched-count-spec-in-critical-path
16024 @opindex mno-sched-count-spec-in-critical-path
16025 @opindex msched-count-spec-in-critical-path
16026 If enabled, speculative dependencies are considered during
16027 computation of the instructions priorities. This makes the use of the
16028 speculation a bit more conservative.
16029 The default is 'disable'.
16031 @item -msched-spec-ldc
16032 @opindex msched-spec-ldc
16033 Use a simple data speculation check. This option is on by default.
16035 @item -msched-control-spec-ldc
16036 @opindex msched-spec-ldc
16037 Use a simple check for control speculation. This option is on by default.
16039 @item -msched-stop-bits-after-every-cycle
16040 @opindex msched-stop-bits-after-every-cycle
16041 Place a stop bit after every cycle when scheduling. This option is on
16044 @item -msched-fp-mem-deps-zero-cost
16045 @opindex msched-fp-mem-deps-zero-cost
16046 Assume that floating-point stores and loads are not likely to cause a conflict
16047 when placed into the same instruction group. This option is disabled by
16050 @item -msel-sched-dont-check-control-spec
16051 @opindex msel-sched-dont-check-control-spec
16052 Generate checks for control speculation in selective scheduling.
16053 This flag is disabled by default.
16055 @item -msched-max-memory-insns=@var{max-insns}
16056 @opindex msched-max-memory-insns
16057 Limit on the number of memory insns per instruction group, giving lower
16058 priority to subsequent memory insns attempting to schedule in the same
16059 instruction group. Frequently useful to prevent cache bank conflicts.
16060 The default value is 1.
16062 @item -msched-max-memory-insns-hard-limit
16063 @opindex msched-max-memory-insns-hard-limit
16064 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16065 disallowing more than that number in an instruction group.
16066 Otherwise, the limit is ``soft'', meaning that non-memory operations
16067 are preferred when the limit is reached, but memory operations may still
16073 @subsection LM32 Options
16074 @cindex LM32 options
16076 These @option{-m} options are defined for the LatticeMico32 architecture:
16079 @item -mbarrel-shift-enabled
16080 @opindex mbarrel-shift-enabled
16081 Enable barrel-shift instructions.
16083 @item -mdivide-enabled
16084 @opindex mdivide-enabled
16085 Enable divide and modulus instructions.
16087 @item -mmultiply-enabled
16088 @opindex multiply-enabled
16089 Enable multiply instructions.
16091 @item -msign-extend-enabled
16092 @opindex msign-extend-enabled
16093 Enable sign extend instructions.
16095 @item -muser-enabled
16096 @opindex muser-enabled
16097 Enable user-defined instructions.
16102 @subsection M32C Options
16103 @cindex M32C options
16106 @item -mcpu=@var{name}
16108 Select the CPU for which code is generated. @var{name} may be one of
16109 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16110 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16111 the M32C/80 series.
16115 Specifies that the program will be run on the simulator. This causes
16116 an alternate runtime library to be linked in which supports, for
16117 example, file I/O@. You must not use this option when generating
16118 programs that will run on real hardware; you must provide your own
16119 runtime library for whatever I/O functions are needed.
16121 @item -memregs=@var{number}
16123 Specifies the number of memory-based pseudo-registers GCC uses
16124 during code generation. These pseudo-registers are used like real
16125 registers, so there is a tradeoff between GCC's ability to fit the
16126 code into available registers, and the performance penalty of using
16127 memory instead of registers. Note that all modules in a program must
16128 be compiled with the same value for this option. Because of that, you
16129 must not use this option with GCC's default runtime libraries.
16133 @node M32R/D Options
16134 @subsection M32R/D Options
16135 @cindex M32R/D options
16137 These @option{-m} options are defined for Renesas M32R/D architectures:
16142 Generate code for the M32R/2@.
16146 Generate code for the M32R/X@.
16150 Generate code for the M32R@. This is the default.
16152 @item -mmodel=small
16153 @opindex mmodel=small
16154 Assume all objects live in the lower 16MB of memory (so that their addresses
16155 can be loaded with the @code{ld24} instruction), and assume all subroutines
16156 are reachable with the @code{bl} instruction.
16157 This is the default.
16159 The addressability of a particular object can be set with the
16160 @code{model} attribute.
16162 @item -mmodel=medium
16163 @opindex mmodel=medium
16164 Assume objects may be anywhere in the 32-bit address space (the compiler
16165 generates @code{seth/add3} instructions to load their addresses), and
16166 assume all subroutines are reachable with the @code{bl} instruction.
16168 @item -mmodel=large
16169 @opindex mmodel=large
16170 Assume objects may be anywhere in the 32-bit address space (the compiler
16171 generates @code{seth/add3} instructions to load their addresses), and
16172 assume subroutines may not be reachable with the @code{bl} instruction
16173 (the compiler generates the much slower @code{seth/add3/jl}
16174 instruction sequence).
16177 @opindex msdata=none
16178 Disable use of the small data area. Variables are put into
16179 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16180 @code{section} attribute has been specified).
16181 This is the default.
16183 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16184 Objects may be explicitly put in the small data area with the
16185 @code{section} attribute using one of these sections.
16187 @item -msdata=sdata
16188 @opindex msdata=sdata
16189 Put small global and static data in the small data area, but do not
16190 generate special code to reference them.
16193 @opindex msdata=use
16194 Put small global and static data in the small data area, and generate
16195 special instructions to reference them.
16199 @cindex smaller data references
16200 Put global and static objects less than or equal to @var{num} bytes
16201 into the small data or BSS sections instead of the normal data or BSS
16202 sections. The default value of @var{num} is 8.
16203 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16204 for this option to have any effect.
16206 All modules should be compiled with the same @option{-G @var{num}} value.
16207 Compiling with different values of @var{num} may or may not work; if it
16208 doesn't the linker gives an error message---incorrect code is not
16213 Makes the M32R-specific code in the compiler display some statistics
16214 that might help in debugging programs.
16216 @item -malign-loops
16217 @opindex malign-loops
16218 Align all loops to a 32-byte boundary.
16220 @item -mno-align-loops
16221 @opindex mno-align-loops
16222 Do not enforce a 32-byte alignment for loops. This is the default.
16224 @item -missue-rate=@var{number}
16225 @opindex missue-rate=@var{number}
16226 Issue @var{number} instructions per cycle. @var{number} can only be 1
16229 @item -mbranch-cost=@var{number}
16230 @opindex mbranch-cost=@var{number}
16231 @var{number} can only be 1 or 2. If it is 1 then branches are
16232 preferred over conditional code, if it is 2, then the opposite applies.
16234 @item -mflush-trap=@var{number}
16235 @opindex mflush-trap=@var{number}
16236 Specifies the trap number to use to flush the cache. The default is
16237 12. Valid numbers are between 0 and 15 inclusive.
16239 @item -mno-flush-trap
16240 @opindex mno-flush-trap
16241 Specifies that the cache cannot be flushed by using a trap.
16243 @item -mflush-func=@var{name}
16244 @opindex mflush-func=@var{name}
16245 Specifies the name of the operating system function to call to flush
16246 the cache. The default is @samp{_flush_cache}, but a function call
16247 is only used if a trap is not available.
16249 @item -mno-flush-func
16250 @opindex mno-flush-func
16251 Indicates that there is no OS function for flushing the cache.
16255 @node M680x0 Options
16256 @subsection M680x0 Options
16257 @cindex M680x0 options
16259 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16260 The default settings depend on which architecture was selected when
16261 the compiler was configured; the defaults for the most common choices
16265 @item -march=@var{arch}
16267 Generate code for a specific M680x0 or ColdFire instruction set
16268 architecture. Permissible values of @var{arch} for M680x0
16269 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16270 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16271 architectures are selected according to Freescale's ISA classification
16272 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16273 @samp{isab} and @samp{isac}.
16275 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16276 code for a ColdFire target. The @var{arch} in this macro is one of the
16277 @option{-march} arguments given above.
16279 When used together, @option{-march} and @option{-mtune} select code
16280 that runs on a family of similar processors but that is optimized
16281 for a particular microarchitecture.
16283 @item -mcpu=@var{cpu}
16285 Generate code for a specific M680x0 or ColdFire processor.
16286 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16287 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16288 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16289 below, which also classifies the CPUs into families:
16291 @multitable @columnfractions 0.20 0.80
16292 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16293 @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}
16294 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16295 @item @samp{5206e} @tab @samp{5206e}
16296 @item @samp{5208} @tab @samp{5207} @samp{5208}
16297 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16298 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16299 @item @samp{5216} @tab @samp{5214} @samp{5216}
16300 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16301 @item @samp{5225} @tab @samp{5224} @samp{5225}
16302 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16303 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16304 @item @samp{5249} @tab @samp{5249}
16305 @item @samp{5250} @tab @samp{5250}
16306 @item @samp{5271} @tab @samp{5270} @samp{5271}
16307 @item @samp{5272} @tab @samp{5272}
16308 @item @samp{5275} @tab @samp{5274} @samp{5275}
16309 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16310 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16311 @item @samp{5307} @tab @samp{5307}
16312 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16313 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16314 @item @samp{5407} @tab @samp{5407}
16315 @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}
16318 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16319 @var{arch} is compatible with @var{cpu}. Other combinations of
16320 @option{-mcpu} and @option{-march} are rejected.
16322 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
16323 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
16324 where the value of @var{family} is given by the table above.
16326 @item -mtune=@var{tune}
16328 Tune the code for a particular microarchitecture within the
16329 constraints set by @option{-march} and @option{-mcpu}.
16330 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16331 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16332 and @samp{cpu32}. The ColdFire microarchitectures
16333 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16335 You can also use @option{-mtune=68020-40} for code that needs
16336 to run relatively well on 68020, 68030 and 68040 targets.
16337 @option{-mtune=68020-60} is similar but includes 68060 targets
16338 as well. These two options select the same tuning decisions as
16339 @option{-m68020-40} and @option{-m68020-60} respectively.
16341 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
16342 when tuning for 680x0 architecture @var{arch}. It also defines
16343 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16344 option is used. If GCC is tuning for a range of architectures,
16345 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16346 it defines the macros for every architecture in the range.
16348 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
16349 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16350 of the arguments given above.
16356 Generate output for a 68000. This is the default
16357 when the compiler is configured for 68000-based systems.
16358 It is equivalent to @option{-march=68000}.
16360 Use this option for microcontrollers with a 68000 or EC000 core,
16361 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16365 Generate output for a 68010. This is the default
16366 when the compiler is configured for 68010-based systems.
16367 It is equivalent to @option{-march=68010}.
16373 Generate output for a 68020. This is the default
16374 when the compiler is configured for 68020-based systems.
16375 It is equivalent to @option{-march=68020}.
16379 Generate output for a 68030. This is the default when the compiler is
16380 configured for 68030-based systems. It is equivalent to
16381 @option{-march=68030}.
16385 Generate output for a 68040. This is the default when the compiler is
16386 configured for 68040-based systems. It is equivalent to
16387 @option{-march=68040}.
16389 This option inhibits the use of 68881/68882 instructions that have to be
16390 emulated by software on the 68040. Use this option if your 68040 does not
16391 have code to emulate those instructions.
16395 Generate output for a 68060. This is the default when the compiler is
16396 configured for 68060-based systems. It is equivalent to
16397 @option{-march=68060}.
16399 This option inhibits the use of 68020 and 68881/68882 instructions that
16400 have to be emulated by software on the 68060. Use this option if your 68060
16401 does not have code to emulate those instructions.
16405 Generate output for a CPU32. This is the default
16406 when the compiler is configured for CPU32-based systems.
16407 It is equivalent to @option{-march=cpu32}.
16409 Use this option for microcontrollers with a
16410 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16411 68336, 68340, 68341, 68349 and 68360.
16415 Generate output for a 520X ColdFire CPU@. This is the default
16416 when the compiler is configured for 520X-based systems.
16417 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16418 in favor of that option.
16420 Use this option for microcontroller with a 5200 core, including
16421 the MCF5202, MCF5203, MCF5204 and MCF5206.
16425 Generate output for a 5206e ColdFire CPU@. The option is now
16426 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16430 Generate output for a member of the ColdFire 528X family.
16431 The option is now deprecated in favor of the equivalent
16432 @option{-mcpu=528x}.
16436 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16437 in favor of the equivalent @option{-mcpu=5307}.
16441 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16442 in favor of the equivalent @option{-mcpu=5407}.
16446 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16447 This includes use of hardware floating-point instructions.
16448 The option is equivalent to @option{-mcpu=547x}, and is now
16449 deprecated in favor of that option.
16453 Generate output for a 68040, without using any of the new instructions.
16454 This results in code that can run relatively efficiently on either a
16455 68020/68881 or a 68030 or a 68040. The generated code does use the
16456 68881 instructions that are emulated on the 68040.
16458 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16462 Generate output for a 68060, without using any of the new instructions.
16463 This results in code that can run relatively efficiently on either a
16464 68020/68881 or a 68030 or a 68040. The generated code does use the
16465 68881 instructions that are emulated on the 68060.
16467 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16471 @opindex mhard-float
16473 Generate floating-point instructions. This is the default for 68020
16474 and above, and for ColdFire devices that have an FPU@. It defines the
16475 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16476 on ColdFire targets.
16479 @opindex msoft-float
16480 Do not generate floating-point instructions; use library calls instead.
16481 This is the default for 68000, 68010, and 68832 targets. It is also
16482 the default for ColdFire devices that have no FPU.
16488 Generate (do not generate) ColdFire hardware divide and remainder
16489 instructions. If @option{-march} is used without @option{-mcpu},
16490 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16491 architectures. Otherwise, the default is taken from the target CPU
16492 (either the default CPU, or the one specified by @option{-mcpu}). For
16493 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16494 @option{-mcpu=5206e}.
16496 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16500 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16501 Additionally, parameters passed on the stack are also aligned to a
16502 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16506 Do not consider type @code{int} to be 16 bits wide. This is the default.
16509 @itemx -mno-bitfield
16510 @opindex mnobitfield
16511 @opindex mno-bitfield
16512 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16513 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16517 Do use the bit-field instructions. The @option{-m68020} option implies
16518 @option{-mbitfield}. This is the default if you use a configuration
16519 designed for a 68020.
16523 Use a different function-calling convention, in which functions
16524 that take a fixed number of arguments return with the @code{rtd}
16525 instruction, which pops their arguments while returning. This
16526 saves one instruction in the caller since there is no need to pop
16527 the arguments there.
16529 This calling convention is incompatible with the one normally
16530 used on Unix, so you cannot use it if you need to call libraries
16531 compiled with the Unix compiler.
16533 Also, you must provide function prototypes for all functions that
16534 take variable numbers of arguments (including @code{printf});
16535 otherwise incorrect code is generated for calls to those
16538 In addition, seriously incorrect code results if you call a
16539 function with too many arguments. (Normally, extra arguments are
16540 harmlessly ignored.)
16542 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16543 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16547 Do not use the calling conventions selected by @option{-mrtd}.
16548 This is the default.
16551 @itemx -mno-align-int
16552 @opindex malign-int
16553 @opindex mno-align-int
16554 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16555 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16556 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16557 Aligning variables on 32-bit boundaries produces code that runs somewhat
16558 faster on processors with 32-bit busses at the expense of more memory.
16560 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16561 aligns structures containing the above types differently than
16562 most published application binary interface specifications for the m68k.
16566 Use the pc-relative addressing mode of the 68000 directly, instead of
16567 using a global offset table. At present, this option implies @option{-fpic},
16568 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16569 not presently supported with @option{-mpcrel}, though this could be supported for
16570 68020 and higher processors.
16572 @item -mno-strict-align
16573 @itemx -mstrict-align
16574 @opindex mno-strict-align
16575 @opindex mstrict-align
16576 Do not (do) assume that unaligned memory references are handled by
16580 Generate code that allows the data segment to be located in a different
16581 area of memory from the text segment. This allows for execute-in-place in
16582 an environment without virtual memory management. This option implies
16585 @item -mno-sep-data
16586 Generate code that assumes that the data segment follows the text segment.
16587 This is the default.
16589 @item -mid-shared-library
16590 Generate code that supports shared libraries via the library ID method.
16591 This allows for execute-in-place and shared libraries in an environment
16592 without virtual memory management. This option implies @option{-fPIC}.
16594 @item -mno-id-shared-library
16595 Generate code that doesn't assume ID-based shared libraries are being used.
16596 This is the default.
16598 @item -mshared-library-id=n
16599 Specifies the identification number of the ID-based shared library being
16600 compiled. Specifying a value of 0 generates more compact code; specifying
16601 other values forces the allocation of that number to the current
16602 library, but is no more space- or time-efficient than omitting this option.
16608 When generating position-independent code for ColdFire, generate code
16609 that works if the GOT has more than 8192 entries. This code is
16610 larger and slower than code generated without this option. On M680x0
16611 processors, this option is not needed; @option{-fPIC} suffices.
16613 GCC normally uses a single instruction to load values from the GOT@.
16614 While this is relatively efficient, it only works if the GOT
16615 is smaller than about 64k. Anything larger causes the linker
16616 to report an error such as:
16618 @cindex relocation truncated to fit (ColdFire)
16620 relocation truncated to fit: R_68K_GOT16O foobar
16623 If this happens, you should recompile your code with @option{-mxgot}.
16624 It should then work with very large GOTs. However, code generated with
16625 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16626 the value of a global symbol.
16628 Note that some linkers, including newer versions of the GNU linker,
16629 can create multiple GOTs and sort GOT entries. If you have such a linker,
16630 you should only need to use @option{-mxgot} when compiling a single
16631 object file that accesses more than 8192 GOT entries. Very few do.
16633 These options have no effect unless GCC is generating
16634 position-independent code.
16638 @node MCore Options
16639 @subsection MCore Options
16640 @cindex MCore options
16642 These are the @samp{-m} options defined for the Motorola M*Core
16648 @itemx -mno-hardlit
16650 @opindex mno-hardlit
16651 Inline constants into the code stream if it can be done in two
16652 instructions or less.
16658 Use the divide instruction. (Enabled by default).
16660 @item -mrelax-immediate
16661 @itemx -mno-relax-immediate
16662 @opindex mrelax-immediate
16663 @opindex mno-relax-immediate
16664 Allow arbitrary-sized immediates in bit operations.
16666 @item -mwide-bitfields
16667 @itemx -mno-wide-bitfields
16668 @opindex mwide-bitfields
16669 @opindex mno-wide-bitfields
16670 Always treat bit-fields as @code{int}-sized.
16672 @item -m4byte-functions
16673 @itemx -mno-4byte-functions
16674 @opindex m4byte-functions
16675 @opindex mno-4byte-functions
16676 Force all functions to be aligned to a 4-byte boundary.
16678 @item -mcallgraph-data
16679 @itemx -mno-callgraph-data
16680 @opindex mcallgraph-data
16681 @opindex mno-callgraph-data
16682 Emit callgraph information.
16685 @itemx -mno-slow-bytes
16686 @opindex mslow-bytes
16687 @opindex mno-slow-bytes
16688 Prefer word access when reading byte quantities.
16690 @item -mlittle-endian
16691 @itemx -mbig-endian
16692 @opindex mlittle-endian
16693 @opindex mbig-endian
16694 Generate code for a little-endian target.
16700 Generate code for the 210 processor.
16704 Assume that runtime support has been provided and so omit the
16705 simulator library (@file{libsim.a)} from the linker command line.
16707 @item -mstack-increment=@var{size}
16708 @opindex mstack-increment
16709 Set the maximum amount for a single stack increment operation. Large
16710 values can increase the speed of programs that contain functions
16711 that need a large amount of stack space, but they can also trigger a
16712 segmentation fault if the stack is extended too much. The default
16718 @subsection MeP Options
16719 @cindex MeP options
16725 Enables the @code{abs} instruction, which is the absolute difference
16726 between two registers.
16730 Enables all the optional instructions---average, multiply, divide, bit
16731 operations, leading zero, absolute difference, min/max, clip, and
16737 Enables the @code{ave} instruction, which computes the average of two
16740 @item -mbased=@var{n}
16742 Variables of size @var{n} bytes or smaller are placed in the
16743 @code{.based} section by default. Based variables use the @code{$tp}
16744 register as a base register, and there is a 128-byte limit to the
16745 @code{.based} section.
16749 Enables the bit operation instructions---bit test (@code{btstm}), set
16750 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16751 test-and-set (@code{tas}).
16753 @item -mc=@var{name}
16755 Selects which section constant data is placed in. @var{name} may
16756 be @samp{tiny}, @samp{near}, or @samp{far}.
16760 Enables the @code{clip} instruction. Note that @option{-mclip} is not
16761 useful unless you also provide @option{-mminmax}.
16763 @item -mconfig=@var{name}
16765 Selects one of the built-in core configurations. Each MeP chip has
16766 one or more modules in it; each module has a core CPU and a variety of
16767 coprocessors, optional instructions, and peripherals. The
16768 @code{MeP-Integrator} tool, not part of GCC, provides these
16769 configurations through this option; using this option is the same as
16770 using all the corresponding command-line options. The default
16771 configuration is @samp{default}.
16775 Enables the coprocessor instructions. By default, this is a 32-bit
16776 coprocessor. Note that the coprocessor is normally enabled via the
16777 @option{-mconfig=} option.
16781 Enables the 32-bit coprocessor's instructions.
16785 Enables the 64-bit coprocessor's instructions.
16789 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16793 Causes constant variables to be placed in the @code{.near} section.
16797 Enables the @code{div} and @code{divu} instructions.
16801 Generate big-endian code.
16805 Generate little-endian code.
16807 @item -mio-volatile
16808 @opindex mio-volatile
16809 Tells the compiler that any variable marked with the @code{io}
16810 attribute is to be considered volatile.
16814 Causes variables to be assigned to the @code{.far} section by default.
16818 Enables the @code{leadz} (leading zero) instruction.
16822 Causes variables to be assigned to the @code{.near} section by default.
16826 Enables the @code{min} and @code{max} instructions.
16830 Enables the multiplication and multiply-accumulate instructions.
16834 Disables all the optional instructions enabled by @option{-mall-opts}.
16838 Enables the @code{repeat} and @code{erepeat} instructions, used for
16839 low-overhead looping.
16843 Causes all variables to default to the @code{.tiny} section. Note
16844 that there is a 65536-byte limit to this section. Accesses to these
16845 variables use the @code{%gp} base register.
16849 Enables the saturation instructions. Note that the compiler does not
16850 currently generate these itself, but this option is included for
16851 compatibility with other tools, like @code{as}.
16855 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16859 Link the simulator run-time libraries.
16863 Link the simulator runtime libraries, excluding built-in support
16864 for reset and exception vectors and tables.
16868 Causes all functions to default to the @code{.far} section. Without
16869 this option, functions default to the @code{.near} section.
16871 @item -mtiny=@var{n}
16873 Variables that are @var{n} bytes or smaller are allocated to the
16874 @code{.tiny} section. These variables use the @code{$gp} base
16875 register. The default for this option is 4, but note that there's a
16876 65536-byte limit to the @code{.tiny} section.
16880 @node MicroBlaze Options
16881 @subsection MicroBlaze Options
16882 @cindex MicroBlaze Options
16887 @opindex msoft-float
16888 Use software emulation for floating point (default).
16891 @opindex mhard-float
16892 Use hardware floating-point instructions.
16896 Do not optimize block moves, use @code{memcpy}.
16898 @item -mno-clearbss
16899 @opindex mno-clearbss
16900 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
16902 @item -mcpu=@var{cpu-type}
16904 Use features of, and schedule code for, the given CPU.
16905 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
16906 where @var{X} is a major version, @var{YY} is the minor version, and
16907 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
16908 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
16910 @item -mxl-soft-mul
16911 @opindex mxl-soft-mul
16912 Use software multiply emulation (default).
16914 @item -mxl-soft-div
16915 @opindex mxl-soft-div
16916 Use software emulation for divides (default).
16918 @item -mxl-barrel-shift
16919 @opindex mxl-barrel-shift
16920 Use the hardware barrel shifter.
16922 @item -mxl-pattern-compare
16923 @opindex mxl-pattern-compare
16924 Use pattern compare instructions.
16926 @item -msmall-divides
16927 @opindex msmall-divides
16928 Use table lookup optimization for small signed integer divisions.
16930 @item -mxl-stack-check
16931 @opindex mxl-stack-check
16932 This option is deprecated. Use @option{-fstack-check} instead.
16935 @opindex mxl-gp-opt
16936 Use GP-relative @code{.sdata}/@code{.sbss} sections.
16938 @item -mxl-multiply-high
16939 @opindex mxl-multiply-high
16940 Use multiply high instructions for high part of 32x32 multiply.
16942 @item -mxl-float-convert
16943 @opindex mxl-float-convert
16944 Use hardware floating-point conversion instructions.
16946 @item -mxl-float-sqrt
16947 @opindex mxl-float-sqrt
16948 Use hardware floating-point square root instruction.
16951 @opindex mbig-endian
16952 Generate code for a big-endian target.
16954 @item -mlittle-endian
16955 @opindex mlittle-endian
16956 Generate code for a little-endian target.
16959 @opindex mxl-reorder
16960 Use reorder instructions (swap and byte reversed load/store).
16962 @item -mxl-mode-@var{app-model}
16963 Select application model @var{app-model}. Valid models are
16966 normal executable (default), uses startup code @file{crt0.o}.
16969 for use with Xilinx Microprocessor Debugger (XMD) based
16970 software intrusive debug agent called xmdstub. This uses startup file
16971 @file{crt1.o} and sets the start address of the program to 0x800.
16974 for applications that are loaded using a bootloader.
16975 This model uses startup file @file{crt2.o} which does not contain a processor
16976 reset vector handler. This is suitable for transferring control on a
16977 processor reset to the bootloader rather than the application.
16980 for applications that do not require any of the
16981 MicroBlaze vectors. This option may be useful for applications running
16982 within a monitoring application. This model uses @file{crt3.o} as a startup file.
16985 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
16986 @option{-mxl-mode-@var{app-model}}.
16991 @subsection MIPS Options
16992 @cindex MIPS options
16998 Generate big-endian code.
17002 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17005 @item -march=@var{arch}
17007 Generate code that runs on @var{arch}, which can be the name of a
17008 generic MIPS ISA, or the name of a particular processor.
17010 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17011 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17012 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17013 @samp{mips64r5} and @samp{mips64r6}.
17014 The processor names are:
17015 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17016 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17017 @samp{5kc}, @samp{5kf},
17019 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17020 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17021 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17022 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17023 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17024 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17026 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17027 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17030 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17031 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17032 @samp{rm7000}, @samp{rm9000},
17033 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17036 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17037 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17038 @samp{xlr} and @samp{xlp}.
17039 The special value @samp{from-abi} selects the
17040 most compatible architecture for the selected ABI (that is,
17041 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17043 The native Linux/GNU toolchain also supports the value @samp{native},
17044 which selects the best architecture option for the host processor.
17045 @option{-march=native} has no effect if GCC does not recognize
17048 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17049 (for example, @option{-march=r2k}). Prefixes are optional, and
17050 @samp{vr} may be written @samp{r}.
17052 Names of the form @samp{@var{n}f2_1} refer to processors with
17053 FPUs clocked at half the rate of the core, names of the form
17054 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17055 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17056 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17057 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17058 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17059 accepted as synonyms for @samp{@var{n}f1_1}.
17061 GCC defines two macros based on the value of this option. The first
17062 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17063 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17064 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17065 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17066 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17068 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17069 above. In other words, it has the full prefix and does not
17070 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17071 the macro names the resolved architecture (either @code{"mips1"} or
17072 @code{"mips3"}). It names the default architecture when no
17073 @option{-march} option is given.
17075 @item -mtune=@var{arch}
17077 Optimize for @var{arch}. Among other things, this option controls
17078 the way instructions are scheduled, and the perceived cost of arithmetic
17079 operations. The list of @var{arch} values is the same as for
17082 When this option is not used, GCC optimizes for the processor
17083 specified by @option{-march}. By using @option{-march} and
17084 @option{-mtune} together, it is possible to generate code that
17085 runs on a family of processors, but optimize the code for one
17086 particular member of that family.
17088 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17089 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17090 @option{-march} ones described above.
17094 Equivalent to @option{-march=mips1}.
17098 Equivalent to @option{-march=mips2}.
17102 Equivalent to @option{-march=mips3}.
17106 Equivalent to @option{-march=mips4}.
17110 Equivalent to @option{-march=mips32}.
17114 Equivalent to @option{-march=mips32r3}.
17118 Equivalent to @option{-march=mips32r5}.
17122 Equivalent to @option{-march=mips32r6}.
17126 Equivalent to @option{-march=mips64}.
17130 Equivalent to @option{-march=mips64r2}.
17134 Equivalent to @option{-march=mips64r3}.
17138 Equivalent to @option{-march=mips64r5}.
17142 Equivalent to @option{-march=mips64r6}.
17147 @opindex mno-mips16
17148 Generate (do not generate) MIPS16 code. If GCC is targeting a
17149 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17151 MIPS16 code generation can also be controlled on a per-function basis
17152 by means of @code{mips16} and @code{nomips16} attributes.
17153 @xref{Function Attributes}, for more information.
17155 @item -mflip-mips16
17156 @opindex mflip-mips16
17157 Generate MIPS16 code on alternating functions. This option is provided
17158 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17159 not intended for ordinary use in compiling user code.
17161 @item -minterlink-compressed
17162 @item -mno-interlink-compressed
17163 @opindex minterlink-compressed
17164 @opindex mno-interlink-compressed
17165 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17166 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17168 For example, code using the standard ISA encoding cannot jump directly
17169 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17170 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17171 knows that the target of the jump is not compressed.
17173 @item -minterlink-mips16
17174 @itemx -mno-interlink-mips16
17175 @opindex minterlink-mips16
17176 @opindex mno-interlink-mips16
17177 Aliases of @option{-minterlink-compressed} and
17178 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17179 and are retained for backwards compatibility.
17191 Generate code for the given ABI@.
17193 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17194 generates 64-bit code when you select a 64-bit architecture, but you
17195 can use @option{-mgp32} to get 32-bit code instead.
17197 For information about the O64 ABI, see
17198 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17200 GCC supports a variant of the o32 ABI in which floating-point registers
17201 are 64 rather than 32 bits wide. You can select this combination with
17202 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17203 and @code{mfhc1} instructions and is therefore only supported for
17204 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17206 The register assignments for arguments and return values remain the
17207 same, but each scalar value is passed in a single 64-bit register
17208 rather than a pair of 32-bit registers. For example, scalar
17209 floating-point values are returned in @samp{$f0} only, not a
17210 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17211 remains the same in that the even-numbered double-precision registers
17214 Two additional variants of the o32 ABI are supported to enable
17215 a transition from 32-bit to 64-bit registers. These are FPXX
17216 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17217 The FPXX extension mandates that all code must execute correctly
17218 when run using 32-bit or 64-bit registers. The code can be interlinked
17219 with either FP32 or FP64, but not both.
17220 The FP64A extension is similar to the FP64 extension but forbids the
17221 use of odd-numbered single-precision registers. This can be used
17222 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17223 processors and allows both FP32 and FP64A code to interlink and
17224 run in the same process without changing FPU modes.
17227 @itemx -mno-abicalls
17229 @opindex mno-abicalls
17230 Generate (do not generate) code that is suitable for SVR4-style
17231 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17236 Generate (do not generate) code that is fully position-independent,
17237 and that can therefore be linked into shared libraries. This option
17238 only affects @option{-mabicalls}.
17240 All @option{-mabicalls} code has traditionally been position-independent,
17241 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17242 as an extension, the GNU toolchain allows executables to use absolute
17243 accesses for locally-binding symbols. It can also use shorter GP
17244 initialization sequences and generate direct calls to locally-defined
17245 functions. This mode is selected by @option{-mno-shared}.
17247 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17248 objects that can only be linked by the GNU linker. However, the option
17249 does not affect the ABI of the final executable; it only affects the ABI
17250 of relocatable objects. Using @option{-mno-shared} generally makes
17251 executables both smaller and quicker.
17253 @option{-mshared} is the default.
17259 Assume (do not assume) that the static and dynamic linkers
17260 support PLTs and copy relocations. This option only affects
17261 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17262 has no effect without @option{-msym32}.
17264 You can make @option{-mplt} the default by configuring
17265 GCC with @option{--with-mips-plt}. The default is
17266 @option{-mno-plt} otherwise.
17272 Lift (do not lift) the usual restrictions on the size of the global
17275 GCC normally uses a single instruction to load values from the GOT@.
17276 While this is relatively efficient, it only works if the GOT
17277 is smaller than about 64k. Anything larger causes the linker
17278 to report an error such as:
17280 @cindex relocation truncated to fit (MIPS)
17282 relocation truncated to fit: R_MIPS_GOT16 foobar
17285 If this happens, you should recompile your code with @option{-mxgot}.
17286 This works with very large GOTs, although the code is also
17287 less efficient, since it takes three instructions to fetch the
17288 value of a global symbol.
17290 Note that some linkers can create multiple GOTs. If you have such a
17291 linker, you should only need to use @option{-mxgot} when a single object
17292 file accesses more than 64k's worth of GOT entries. Very few do.
17294 These options have no effect unless GCC is generating position
17299 Assume that general-purpose registers are 32 bits wide.
17303 Assume that general-purpose registers are 64 bits wide.
17307 Assume that floating-point registers are 32 bits wide.
17311 Assume that floating-point registers are 64 bits wide.
17315 Do not assume the width of floating-point registers.
17318 @opindex mhard-float
17319 Use floating-point coprocessor instructions.
17322 @opindex msoft-float
17323 Do not use floating-point coprocessor instructions. Implement
17324 floating-point calculations using library calls instead.
17328 Equivalent to @option{-msoft-float}, but additionally asserts that the
17329 program being compiled does not perform any floating-point operations.
17330 This option is presently supported only by some bare-metal MIPS
17331 configurations, where it may select a special set of libraries
17332 that lack all floating-point support (including, for example, the
17333 floating-point @code{printf} formats).
17334 If code compiled with @option{-mno-float} accidentally contains
17335 floating-point operations, it is likely to suffer a link-time
17336 or run-time failure.
17338 @item -msingle-float
17339 @opindex msingle-float
17340 Assume that the floating-point coprocessor only supports single-precision
17343 @item -mdouble-float
17344 @opindex mdouble-float
17345 Assume that the floating-point coprocessor supports double-precision
17346 operations. This is the default.
17349 @itemx -mno-odd-spreg
17350 @opindex modd-spreg
17351 @opindex mno-odd-spreg
17352 Enable the use of odd-numbered single-precision floating-point registers
17353 for the o32 ABI. This is the default for processors that are known to
17354 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
17358 @itemx -mabs=legacy
17360 @opindex mabs=legacy
17361 These options control the treatment of the special not-a-number (NaN)
17362 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17363 @code{neg.@i{fmt}} machine instructions.
17365 By default or when @option{-mabs=legacy} is used the legacy
17366 treatment is selected. In this case these instructions are considered
17367 arithmetic and avoided where correct operation is required and the
17368 input operand might be a NaN. A longer sequence of instructions that
17369 manipulate the sign bit of floating-point datum manually is used
17370 instead unless the @option{-ffinite-math-only} option has also been
17373 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17374 this case these instructions are considered non-arithmetic and therefore
17375 operating correctly in all cases, including in particular where the
17376 input operand is a NaN. These instructions are therefore always used
17377 for the respective operations.
17380 @itemx -mnan=legacy
17382 @opindex mnan=legacy
17383 These options control the encoding of the special not-a-number (NaN)
17384 IEEE 754 floating-point data.
17386 The @option{-mnan=legacy} option selects the legacy encoding. In this
17387 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17388 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17389 by the first bit of their trailing significand field being 1.
17391 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17392 this case qNaNs are denoted by the first bit of their trailing
17393 significand field being 1, whereas sNaNs are denoted by the first bit of
17394 their trailing significand field being 0.
17396 The default is @option{-mnan=legacy} unless GCC has been configured with
17397 @option{--with-nan=2008}.
17403 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17404 implement atomic memory built-in functions. When neither option is
17405 specified, GCC uses the instructions if the target architecture
17408 @option{-mllsc} is useful if the runtime environment can emulate the
17409 instructions and @option{-mno-llsc} can be useful when compiling for
17410 nonstandard ISAs. You can make either option the default by
17411 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17412 respectively. @option{--with-llsc} is the default for some
17413 configurations; see the installation documentation for details.
17419 Use (do not use) revision 1 of the MIPS DSP ASE@.
17420 @xref{MIPS DSP Built-in Functions}. This option defines the
17421 preprocessor macro @code{__mips_dsp}. It also defines
17422 @code{__mips_dsp_rev} to 1.
17428 Use (do not use) revision 2 of the MIPS DSP ASE@.
17429 @xref{MIPS DSP Built-in Functions}. This option defines the
17430 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17431 It also defines @code{__mips_dsp_rev} to 2.
17434 @itemx -mno-smartmips
17435 @opindex msmartmips
17436 @opindex mno-smartmips
17437 Use (do not use) the MIPS SmartMIPS ASE.
17439 @item -mpaired-single
17440 @itemx -mno-paired-single
17441 @opindex mpaired-single
17442 @opindex mno-paired-single
17443 Use (do not use) paired-single floating-point instructions.
17444 @xref{MIPS Paired-Single Support}. This option requires
17445 hardware floating-point support to be enabled.
17451 Use (do not use) MIPS Digital Media Extension instructions.
17452 This option can only be used when generating 64-bit code and requires
17453 hardware floating-point support to be enabled.
17458 @opindex mno-mips3d
17459 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17460 The option @option{-mips3d} implies @option{-mpaired-single}.
17463 @itemx -mno-micromips
17464 @opindex mmicromips
17465 @opindex mno-mmicromips
17466 Generate (do not generate) microMIPS code.
17468 MicroMIPS code generation can also be controlled on a per-function basis
17469 by means of @code{micromips} and @code{nomicromips} attributes.
17470 @xref{Function Attributes}, for more information.
17476 Use (do not use) MT Multithreading instructions.
17482 Use (do not use) the MIPS MCU ASE instructions.
17488 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17494 Use (do not use) the MIPS Virtualization Application Specific instructions.
17500 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17504 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17505 an explanation of the default and the way that the pointer size is
17510 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17512 The default size of @code{int}s, @code{long}s and pointers depends on
17513 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17514 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17515 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17516 or the same size as integer registers, whichever is smaller.
17522 Assume (do not assume) that all symbols have 32-bit values, regardless
17523 of the selected ABI@. This option is useful in combination with
17524 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17525 to generate shorter and faster references to symbolic addresses.
17529 Put definitions of externally-visible data in a small data section
17530 if that data is no bigger than @var{num} bytes. GCC can then generate
17531 more efficient accesses to the data; see @option{-mgpopt} for details.
17533 The default @option{-G} option depends on the configuration.
17535 @item -mlocal-sdata
17536 @itemx -mno-local-sdata
17537 @opindex mlocal-sdata
17538 @opindex mno-local-sdata
17539 Extend (do not extend) the @option{-G} behavior to local data too,
17540 such as to static variables in C@. @option{-mlocal-sdata} is the
17541 default for all configurations.
17543 If the linker complains that an application is using too much small data,
17544 you might want to try rebuilding the less performance-critical parts with
17545 @option{-mno-local-sdata}. You might also want to build large
17546 libraries with @option{-mno-local-sdata}, so that the libraries leave
17547 more room for the main program.
17549 @item -mextern-sdata
17550 @itemx -mno-extern-sdata
17551 @opindex mextern-sdata
17552 @opindex mno-extern-sdata
17553 Assume (do not assume) that externally-defined data is in
17554 a small data section if the size of that data is within the @option{-G} limit.
17555 @option{-mextern-sdata} is the default for all configurations.
17557 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17558 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17559 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17560 is placed in a small data section. If @var{Var} is defined by another
17561 module, you must either compile that module with a high-enough
17562 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17563 definition. If @var{Var} is common, you must link the application
17564 with a high-enough @option{-G} setting.
17566 The easiest way of satisfying these restrictions is to compile
17567 and link every module with the same @option{-G} option. However,
17568 you may wish to build a library that supports several different
17569 small data limits. You can do this by compiling the library with
17570 the highest supported @option{-G} setting and additionally using
17571 @option{-mno-extern-sdata} to stop the library from making assumptions
17572 about externally-defined data.
17578 Use (do not use) GP-relative accesses for symbols that are known to be
17579 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17580 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17583 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17584 might not hold the value of @code{_gp}. For example, if the code is
17585 part of a library that might be used in a boot monitor, programs that
17586 call boot monitor routines pass an unknown value in @code{$gp}.
17587 (In such situations, the boot monitor itself is usually compiled
17588 with @option{-G0}.)
17590 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17591 @option{-mno-extern-sdata}.
17593 @item -membedded-data
17594 @itemx -mno-embedded-data
17595 @opindex membedded-data
17596 @opindex mno-embedded-data
17597 Allocate variables to the read-only data section first if possible, then
17598 next in the small data section if possible, otherwise in data. This gives
17599 slightly slower code than the default, but reduces the amount of RAM required
17600 when executing, and thus may be preferred for some embedded systems.
17602 @item -muninit-const-in-rodata
17603 @itemx -mno-uninit-const-in-rodata
17604 @opindex muninit-const-in-rodata
17605 @opindex mno-uninit-const-in-rodata
17606 Put uninitialized @code{const} variables in the read-only data section.
17607 This option is only meaningful in conjunction with @option{-membedded-data}.
17609 @item -mcode-readable=@var{setting}
17610 @opindex mcode-readable
17611 Specify whether GCC may generate code that reads from executable sections.
17612 There are three possible settings:
17615 @item -mcode-readable=yes
17616 Instructions may freely access executable sections. This is the
17619 @item -mcode-readable=pcrel
17620 MIPS16 PC-relative load instructions can access executable sections,
17621 but other instructions must not do so. This option is useful on 4KSc
17622 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17623 It is also useful on processors that can be configured to have a dual
17624 instruction/data SRAM interface and that, like the M4K, automatically
17625 redirect PC-relative loads to the instruction RAM.
17627 @item -mcode-readable=no
17628 Instructions must not access executable sections. This option can be
17629 useful on targets that are configured to have a dual instruction/data
17630 SRAM interface but that (unlike the M4K) do not automatically redirect
17631 PC-relative loads to the instruction RAM.
17634 @item -msplit-addresses
17635 @itemx -mno-split-addresses
17636 @opindex msplit-addresses
17637 @opindex mno-split-addresses
17638 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17639 relocation operators. This option has been superseded by
17640 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17642 @item -mexplicit-relocs
17643 @itemx -mno-explicit-relocs
17644 @opindex mexplicit-relocs
17645 @opindex mno-explicit-relocs
17646 Use (do not use) assembler relocation operators when dealing with symbolic
17647 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17648 is to use assembler macros instead.
17650 @option{-mexplicit-relocs} is the default if GCC was configured
17651 to use an assembler that supports relocation operators.
17653 @item -mcheck-zero-division
17654 @itemx -mno-check-zero-division
17655 @opindex mcheck-zero-division
17656 @opindex mno-check-zero-division
17657 Trap (do not trap) on integer division by zero.
17659 The default is @option{-mcheck-zero-division}.
17661 @item -mdivide-traps
17662 @itemx -mdivide-breaks
17663 @opindex mdivide-traps
17664 @opindex mdivide-breaks
17665 MIPS systems check for division by zero by generating either a
17666 conditional trap or a break instruction. Using traps results in
17667 smaller code, but is only supported on MIPS II and later. Also, some
17668 versions of the Linux kernel have a bug that prevents trap from
17669 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17670 allow conditional traps on architectures that support them and
17671 @option{-mdivide-breaks} to force the use of breaks.
17673 The default is usually @option{-mdivide-traps}, but this can be
17674 overridden at configure time using @option{--with-divide=breaks}.
17675 Divide-by-zero checks can be completely disabled using
17676 @option{-mno-check-zero-division}.
17681 @opindex mno-memcpy
17682 Force (do not force) the use of @code{memcpy} for non-trivial block
17683 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17684 most constant-sized copies.
17687 @itemx -mno-long-calls
17688 @opindex mlong-calls
17689 @opindex mno-long-calls
17690 Disable (do not disable) use of the @code{jal} instruction. Calling
17691 functions using @code{jal} is more efficient but requires the caller
17692 and callee to be in the same 256 megabyte segment.
17694 This option has no effect on abicalls code. The default is
17695 @option{-mno-long-calls}.
17701 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17702 instructions, as provided by the R4650 ISA@.
17708 Enable (disable) use of the @code{madd} and @code{msub} integer
17709 instructions. The default is @option{-mimadd} on architectures
17710 that support @code{madd} and @code{msub} except for the 74k
17711 architecture where it was found to generate slower code.
17714 @itemx -mno-fused-madd
17715 @opindex mfused-madd
17716 @opindex mno-fused-madd
17717 Enable (disable) use of the floating-point multiply-accumulate
17718 instructions, when they are available. The default is
17719 @option{-mfused-madd}.
17721 On the R8000 CPU when multiply-accumulate instructions are used,
17722 the intermediate product is calculated to infinite precision
17723 and is not subject to the FCSR Flush to Zero bit. This may be
17724 undesirable in some circumstances. On other processors the result
17725 is numerically identical to the equivalent computation using
17726 separate multiply, add, subtract and negate instructions.
17730 Tell the MIPS assembler to not run its preprocessor over user
17731 assembler files (with a @samp{.s} suffix) when assembling them.
17736 @opindex mno-fix-24k
17737 Work around the 24K E48 (lost data on stores during refill) errata.
17738 The workarounds are implemented by the assembler rather than by GCC@.
17741 @itemx -mno-fix-r4000
17742 @opindex mfix-r4000
17743 @opindex mno-fix-r4000
17744 Work around certain R4000 CPU errata:
17747 A double-word or a variable shift may give an incorrect result if executed
17748 immediately after starting an integer division.
17750 A double-word or a variable shift may give an incorrect result if executed
17751 while an integer multiplication is in progress.
17753 An integer division may give an incorrect result if started in a delay slot
17754 of a taken branch or a jump.
17758 @itemx -mno-fix-r4400
17759 @opindex mfix-r4400
17760 @opindex mno-fix-r4400
17761 Work around certain R4400 CPU errata:
17764 A double-word or a variable shift may give an incorrect result if executed
17765 immediately after starting an integer division.
17769 @itemx -mno-fix-r10000
17770 @opindex mfix-r10000
17771 @opindex mno-fix-r10000
17772 Work around certain R10000 errata:
17775 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17776 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17779 This option can only be used if the target architecture supports
17780 branch-likely instructions. @option{-mfix-r10000} is the default when
17781 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17785 @itemx -mno-fix-rm7000
17786 @opindex mfix-rm7000
17787 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17788 workarounds are implemented by the assembler rather than by GCC@.
17791 @itemx -mno-fix-vr4120
17792 @opindex mfix-vr4120
17793 Work around certain VR4120 errata:
17796 @code{dmultu} does not always produce the correct result.
17798 @code{div} and @code{ddiv} do not always produce the correct result if one
17799 of the operands is negative.
17801 The workarounds for the division errata rely on special functions in
17802 @file{libgcc.a}. At present, these functions are only provided by
17803 the @code{mips64vr*-elf} configurations.
17805 Other VR4120 errata require a NOP to be inserted between certain pairs of
17806 instructions. These errata are handled by the assembler, not by GCC itself.
17809 @opindex mfix-vr4130
17810 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17811 workarounds are implemented by the assembler rather than by GCC,
17812 although GCC avoids using @code{mflo} and @code{mfhi} if the
17813 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17814 instructions are available instead.
17817 @itemx -mno-fix-sb1
17819 Work around certain SB-1 CPU core errata.
17820 (This flag currently works around the SB-1 revision 2
17821 ``F1'' and ``F2'' floating-point errata.)
17823 @item -mr10k-cache-barrier=@var{setting}
17824 @opindex mr10k-cache-barrier
17825 Specify whether GCC should insert cache barriers to avoid the
17826 side-effects of speculation on R10K processors.
17828 In common with many processors, the R10K tries to predict the outcome
17829 of a conditional branch and speculatively executes instructions from
17830 the ``taken'' branch. It later aborts these instructions if the
17831 predicted outcome is wrong. However, on the R10K, even aborted
17832 instructions can have side effects.
17834 This problem only affects kernel stores and, depending on the system,
17835 kernel loads. As an example, a speculatively-executed store may load
17836 the target memory into cache and mark the cache line as dirty, even if
17837 the store itself is later aborted. If a DMA operation writes to the
17838 same area of memory before the ``dirty'' line is flushed, the cached
17839 data overwrites the DMA-ed data. See the R10K processor manual
17840 for a full description, including other potential problems.
17842 One workaround is to insert cache barrier instructions before every memory
17843 access that might be speculatively executed and that might have side
17844 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17845 controls GCC's implementation of this workaround. It assumes that
17846 aborted accesses to any byte in the following regions does not have
17851 the memory occupied by the current function's stack frame;
17854 the memory occupied by an incoming stack argument;
17857 the memory occupied by an object with a link-time-constant address.
17860 It is the kernel's responsibility to ensure that speculative
17861 accesses to these regions are indeed safe.
17863 If the input program contains a function declaration such as:
17869 then the implementation of @code{foo} must allow @code{j foo} and
17870 @code{jal foo} to be executed speculatively. GCC honors this
17871 restriction for functions it compiles itself. It expects non-GCC
17872 functions (such as hand-written assembly code) to do the same.
17874 The option has three forms:
17877 @item -mr10k-cache-barrier=load-store
17878 Insert a cache barrier before a load or store that might be
17879 speculatively executed and that might have side effects even
17882 @item -mr10k-cache-barrier=store
17883 Insert a cache barrier before a store that might be speculatively
17884 executed and that might have side effects even if aborted.
17886 @item -mr10k-cache-barrier=none
17887 Disable the insertion of cache barriers. This is the default setting.
17890 @item -mflush-func=@var{func}
17891 @itemx -mno-flush-func
17892 @opindex mflush-func
17893 Specifies the function to call to flush the I and D caches, or to not
17894 call any such function. If called, the function must take the same
17895 arguments as the common @code{_flush_func}, that is, the address of the
17896 memory range for which the cache is being flushed, the size of the
17897 memory range, and the number 3 (to flush both caches). The default
17898 depends on the target GCC was configured for, but commonly is either
17899 @code{_flush_func} or @code{__cpu_flush}.
17901 @item mbranch-cost=@var{num}
17902 @opindex mbranch-cost
17903 Set the cost of branches to roughly @var{num} ``simple'' instructions.
17904 This cost is only a heuristic and is not guaranteed to produce
17905 consistent results across releases. A zero cost redundantly selects
17906 the default, which is based on the @option{-mtune} setting.
17908 @item -mbranch-likely
17909 @itemx -mno-branch-likely
17910 @opindex mbranch-likely
17911 @opindex mno-branch-likely
17912 Enable or disable use of Branch Likely instructions, regardless of the
17913 default for the selected architecture. By default, Branch Likely
17914 instructions may be generated if they are supported by the selected
17915 architecture. An exception is for the MIPS32 and MIPS64 architectures
17916 and processors that implement those architectures; for those, Branch
17917 Likely instructions are not be generated by default because the MIPS32
17918 and MIPS64 architectures specifically deprecate their use.
17920 @item -mfp-exceptions
17921 @itemx -mno-fp-exceptions
17922 @opindex mfp-exceptions
17923 Specifies whether FP exceptions are enabled. This affects how
17924 FP instructions are scheduled for some processors.
17925 The default is that FP exceptions are
17928 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
17929 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
17932 @item -mvr4130-align
17933 @itemx -mno-vr4130-align
17934 @opindex mvr4130-align
17935 The VR4130 pipeline is two-way superscalar, but can only issue two
17936 instructions together if the first one is 8-byte aligned. When this
17937 option is enabled, GCC aligns pairs of instructions that it
17938 thinks should execute in parallel.
17940 This option only has an effect when optimizing for the VR4130.
17941 It normally makes code faster, but at the expense of making it bigger.
17942 It is enabled by default at optimization level @option{-O3}.
17947 Enable (disable) generation of @code{synci} instructions on
17948 architectures that support it. The @code{synci} instructions (if
17949 enabled) are generated when @code{__builtin___clear_cache} is
17952 This option defaults to @option{-mno-synci}, but the default can be
17953 overridden by configuring GCC with @option{--with-synci}.
17955 When compiling code for single processor systems, it is generally safe
17956 to use @code{synci}. However, on many multi-core (SMP) systems, it
17957 does not invalidate the instruction caches on all cores and may lead
17958 to undefined behavior.
17960 @item -mrelax-pic-calls
17961 @itemx -mno-relax-pic-calls
17962 @opindex mrelax-pic-calls
17963 Try to turn PIC calls that are normally dispatched via register
17964 @code{$25} into direct calls. This is only possible if the linker can
17965 resolve the destination at link-time and if the destination is within
17966 range for a direct call.
17968 @option{-mrelax-pic-calls} is the default if GCC was configured to use
17969 an assembler and a linker that support the @code{.reloc} assembly
17970 directive and @option{-mexplicit-relocs} is in effect. With
17971 @option{-mno-explicit-relocs}, this optimization can be performed by the
17972 assembler and the linker alone without help from the compiler.
17974 @item -mmcount-ra-address
17975 @itemx -mno-mcount-ra-address
17976 @opindex mmcount-ra-address
17977 @opindex mno-mcount-ra-address
17978 Emit (do not emit) code that allows @code{_mcount} to modify the
17979 calling function's return address. When enabled, this option extends
17980 the usual @code{_mcount} interface with a new @var{ra-address}
17981 parameter, which has type @code{intptr_t *} and is passed in register
17982 @code{$12}. @code{_mcount} can then modify the return address by
17983 doing both of the following:
17986 Returning the new address in register @code{$31}.
17988 Storing the new address in @code{*@var{ra-address}},
17989 if @var{ra-address} is nonnull.
17992 The default is @option{-mno-mcount-ra-address}.
17997 @subsection MMIX Options
17998 @cindex MMIX Options
18000 These options are defined for the MMIX:
18004 @itemx -mno-libfuncs
18006 @opindex mno-libfuncs
18007 Specify that intrinsic library functions are being compiled, passing all
18008 values in registers, no matter the size.
18011 @itemx -mno-epsilon
18013 @opindex mno-epsilon
18014 Generate floating-point comparison instructions that compare with respect
18015 to the @code{rE} epsilon register.
18017 @item -mabi=mmixware
18019 @opindex mabi=mmixware
18021 Generate code that passes function parameters and return values that (in
18022 the called function) are seen as registers @code{$0} and up, as opposed to
18023 the GNU ABI which uses global registers @code{$231} and up.
18025 @item -mzero-extend
18026 @itemx -mno-zero-extend
18027 @opindex mzero-extend
18028 @opindex mno-zero-extend
18029 When reading data from memory in sizes shorter than 64 bits, use (do not
18030 use) zero-extending load instructions by default, rather than
18031 sign-extending ones.
18034 @itemx -mno-knuthdiv
18036 @opindex mno-knuthdiv
18037 Make the result of a division yielding a remainder have the same sign as
18038 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18039 remainder follows the sign of the dividend. Both methods are
18040 arithmetically valid, the latter being almost exclusively used.
18042 @item -mtoplevel-symbols
18043 @itemx -mno-toplevel-symbols
18044 @opindex mtoplevel-symbols
18045 @opindex mno-toplevel-symbols
18046 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18047 code can be used with the @code{PREFIX} assembly directive.
18051 Generate an executable in the ELF format, rather than the default
18052 @samp{mmo} format used by the @command{mmix} simulator.
18054 @item -mbranch-predict
18055 @itemx -mno-branch-predict
18056 @opindex mbranch-predict
18057 @opindex mno-branch-predict
18058 Use (do not use) the probable-branch instructions, when static branch
18059 prediction indicates a probable branch.
18061 @item -mbase-addresses
18062 @itemx -mno-base-addresses
18063 @opindex mbase-addresses
18064 @opindex mno-base-addresses
18065 Generate (do not generate) code that uses @emph{base addresses}. Using a
18066 base address automatically generates a request (handled by the assembler
18067 and the linker) for a constant to be set up in a global register. The
18068 register is used for one or more base address requests within the range 0
18069 to 255 from the value held in the register. The generally leads to short
18070 and fast code, but the number of different data items that can be
18071 addressed is limited. This means that a program that uses lots of static
18072 data may require @option{-mno-base-addresses}.
18074 @item -msingle-exit
18075 @itemx -mno-single-exit
18076 @opindex msingle-exit
18077 @opindex mno-single-exit
18078 Force (do not force) generated code to have a single exit point in each
18082 @node MN10300 Options
18083 @subsection MN10300 Options
18084 @cindex MN10300 options
18086 These @option{-m} options are defined for Matsushita MN10300 architectures:
18091 Generate code to avoid bugs in the multiply instructions for the MN10300
18092 processors. This is the default.
18094 @item -mno-mult-bug
18095 @opindex mno-mult-bug
18096 Do not generate code to avoid bugs in the multiply instructions for the
18097 MN10300 processors.
18101 Generate code using features specific to the AM33 processor.
18105 Do not generate code using features specific to the AM33 processor. This
18110 Generate code using features specific to the AM33/2.0 processor.
18114 Generate code using features specific to the AM34 processor.
18116 @item -mtune=@var{cpu-type}
18118 Use the timing characteristics of the indicated CPU type when
18119 scheduling instructions. This does not change the targeted processor
18120 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18121 @samp{am33-2} or @samp{am34}.
18123 @item -mreturn-pointer-on-d0
18124 @opindex mreturn-pointer-on-d0
18125 When generating a function that returns a pointer, return the pointer
18126 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18127 only in @code{a0}, and attempts to call such functions without a prototype
18128 result in errors. Note that this option is on by default; use
18129 @option{-mno-return-pointer-on-d0} to disable it.
18133 Do not link in the C run-time initialization object file.
18137 Indicate to the linker that it should perform a relaxation optimization pass
18138 to shorten branches, calls and absolute memory addresses. This option only
18139 has an effect when used on the command line for the final link step.
18141 This option makes symbolic debugging impossible.
18145 Allow the compiler to generate @emph{Long Instruction Word}
18146 instructions if the target is the @samp{AM33} or later. This is the
18147 default. This option defines the preprocessor macro @code{__LIW__}.
18151 Do not allow the compiler to generate @emph{Long Instruction Word}
18152 instructions. This option defines the preprocessor macro
18157 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18158 instructions if the target is the @samp{AM33} or later. This is the
18159 default. This option defines the preprocessor macro @code{__SETLB__}.
18163 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18164 instructions. This option defines the preprocessor macro
18165 @code{__NO_SETLB__}.
18169 @node Moxie Options
18170 @subsection Moxie Options
18171 @cindex Moxie Options
18177 Generate big-endian code. This is the default for @samp{moxie-*-*}
18182 Generate little-endian code.
18186 Generate mul.x and umul.x instructions. This is the default for
18187 @samp{moxiebox-*-*} configurations.
18191 Do not link in the C run-time initialization object file.
18195 @node MSP430 Options
18196 @subsection MSP430 Options
18197 @cindex MSP430 Options
18199 These options are defined for the MSP430:
18205 Force assembly output to always use hex constants. Normally such
18206 constants are signed decimals, but this option is available for
18207 testsuite and/or aesthetic purposes.
18211 Select the MCU to target. This is used to create a C preprocessor
18212 symbol based upon the MCU name, converted to upper case and pre- and
18213 post-fixed with @samp{__}. This in turn is used by the
18214 @file{msp430.h} header file to select an MCU-specific supplementary
18217 The option also sets the ISA to use. If the MCU name is one that is
18218 known to only support the 430 ISA then that is selected, otherwise the
18219 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18220 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18221 name selects the 430X ISA.
18223 In addition an MCU-specific linker script is added to the linker
18224 command line. The script's name is the name of the MCU with
18225 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18226 command line defines the C preprocessor symbol @code{__XXX__} and
18227 cause the linker to search for a script called @file{xxx.ld}.
18229 This option is also passed on to the assembler.
18233 Specifies the ISA to use. Accepted values are @samp{msp430},
18234 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18235 @option{-mmcu=} option should be used to select the ISA.
18239 Link to the simulator runtime libraries and linker script. Overrides
18240 any scripts that would be selected by the @option{-mmcu=} option.
18244 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18248 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18252 This option is passed to the assembler and linker, and allows the
18253 linker to perform certain optimizations that cannot be done until
18258 Describes the type of hardware multiply supported by the target.
18259 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
18260 for the original 16-bit-only multiply supported by early MCUs.
18261 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
18262 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18263 A value of @samp{auto} can also be given. This tells GCC to deduce
18264 the hardware multiply support based upon the MCU name provided by the
18265 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18266 @samp{32bit} hardware multiply support is assumed. @samp{auto} is the
18269 Hardware multiplies are normally performed by calling a library
18270 routine. This saves space in the generated code. When compiling at
18271 @option{-O3} or higher however the hardware multiplier is invoked
18272 inline. This makes for bigger, but faster code.
18274 The hardware multiply routines disable interrupts whilst running and
18275 restore the previous interrupt state when they finish. This makes
18276 them safe to use inside interrupt handlers as well as in normal code.
18280 Enable the use of a minimum runtime environment - no static
18281 initializers or constructors. This is intended for memory-constrained
18282 devices. The compiler includes special symbols in some objects
18283 that tell the linker and runtime which code fragments are required.
18285 @item -mcode-region=
18286 @itemx -mdata-region=
18287 @opindex mcode-region
18288 @opindex mdata-region
18289 These options tell the compiler where to place functions and data that
18290 do not have one of the @code{lower}, @code{upper}, @code{either} or
18291 @code{section} attributes. Possible values are @code{lower},
18292 @code{upper}, @code{either} or @code{any}. The first three behave
18293 like the corresponding attribute. The fourth possible value -
18294 @code{any} - is the default. It leaves placement entirely up to the
18295 linker script and how it assigns the standard sections (.text, .data
18296 etc) to the memory regions.
18300 @node NDS32 Options
18301 @subsection NDS32 Options
18302 @cindex NDS32 Options
18304 These options are defined for NDS32 implementations:
18309 @opindex mbig-endian
18310 Generate code in big-endian mode.
18312 @item -mlittle-endian
18313 @opindex mlittle-endian
18314 Generate code in little-endian mode.
18316 @item -mreduced-regs
18317 @opindex mreduced-regs
18318 Use reduced-set registers for register allocation.
18321 @opindex mfull-regs
18322 Use full-set registers for register allocation.
18326 Generate conditional move instructions.
18330 Do not generate conditional move instructions.
18334 Generate performance extension instructions.
18336 @item -mno-perf-ext
18337 @opindex mno-perf-ext
18338 Do not generate performance extension instructions.
18342 Generate v3 push25/pop25 instructions.
18345 @opindex mno-v3push
18346 Do not generate v3 push25/pop25 instructions.
18350 Generate 16-bit instructions.
18353 @opindex mno-16-bit
18354 Do not generate 16-bit instructions.
18356 @item -misr-vector-size=@var{num}
18357 @opindex misr-vector-size
18358 Specify the size of each interrupt vector, which must be 4 or 16.
18360 @item -mcache-block-size=@var{num}
18361 @opindex mcache-block-size
18362 Specify the size of each cache block,
18363 which must be a power of 2 between 4 and 512.
18365 @item -march=@var{arch}
18367 Specify the name of the target architecture.
18369 @item -mcmodel=@var{code-model}
18371 Set the code model to one of
18374 All the data and read-only data segments must be within 512KB addressing space.
18375 The text segment must be within 16MB addressing space.
18376 @item @samp{medium}
18377 The data segment must be within 512KB while the read-only data segment can be
18378 within 4GB addressing space. The text segment should be still within 16MB
18381 All the text and data segments can be within 4GB addressing space.
18385 @opindex mctor-dtor
18386 Enable constructor/destructor feature.
18390 Guide linker to relax instructions.
18394 @node Nios II Options
18395 @subsection Nios II Options
18396 @cindex Nios II options
18397 @cindex Altera Nios II options
18399 These are the options defined for the Altera Nios II processor.
18405 @cindex smaller data references
18406 Put global and static objects less than or equal to @var{num} bytes
18407 into the small data or BSS sections instead of the normal data or BSS
18408 sections. The default value of @var{num} is 8.
18410 @item -mgpopt=@var{option}
18415 Generate (do not generate) GP-relative accesses. The following
18416 @var{option} names are recognized:
18421 Do not generate GP-relative accesses.
18424 Generate GP-relative accesses for small data objects that are not
18425 external or weak. Also use GP-relative addressing for objects that
18426 have been explicitly placed in a small data section via a @code{section}
18430 As for @samp{local}, but also generate GP-relative accesses for
18431 small data objects that are external or weak. If you use this option,
18432 you must ensure that all parts of your program (including libraries) are
18433 compiled with the same @option{-G} setting.
18436 Generate GP-relative accesses for all data objects in the program. If you
18437 use this option, the entire data and BSS segments
18438 of your program must fit in 64K of memory and you must use an appropriate
18439 linker script to allocate them within the addressible range of the
18443 Generate GP-relative addresses for function pointers as well as data
18444 pointers. If you use this option, the entire text, data, and BSS segments
18445 of your program must fit in 64K of memory and you must use an appropriate
18446 linker script to allocate them within the addressible range of the
18451 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18452 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18454 The default is @option{-mgpopt} except when @option{-fpic} or
18455 @option{-fPIC} is specified to generate position-independent code.
18456 Note that the Nios II ABI does not permit GP-relative accesses from
18459 You may need to specify @option{-mno-gpopt} explicitly when building
18460 programs that include large amounts of small data, including large
18461 GOT data sections. In this case, the 16-bit offset for GP-relative
18462 addressing may not be large enough to allow access to the entire
18463 small data section.
18469 Generate little-endian (default) or big-endian (experimental) code,
18472 @item -mbypass-cache
18473 @itemx -mno-bypass-cache
18474 @opindex mno-bypass-cache
18475 @opindex mbypass-cache
18476 Force all load and store instructions to always bypass cache by
18477 using I/O variants of the instructions. The default is not to
18480 @item -mno-cache-volatile
18481 @itemx -mcache-volatile
18482 @opindex mcache-volatile
18483 @opindex mno-cache-volatile
18484 Volatile memory access bypass the cache using the I/O variants of
18485 the load and store instructions. The default is not to bypass the cache.
18487 @item -mno-fast-sw-div
18488 @itemx -mfast-sw-div
18489 @opindex mno-fast-sw-div
18490 @opindex mfast-sw-div
18491 Do not use table-based fast divide for small numbers. The default
18492 is to use the fast divide at @option{-O3} and above.
18496 @itemx -mno-hw-mulx
18500 @opindex mno-hw-mul
18502 @opindex mno-hw-mulx
18504 @opindex mno-hw-div
18506 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18507 instructions by the compiler. The default is to emit @code{mul}
18508 and not emit @code{div} and @code{mulx}.
18510 @item -mcustom-@var{insn}=@var{N}
18511 @itemx -mno-custom-@var{insn}
18512 @opindex mcustom-@var{insn}
18513 @opindex mno-custom-@var{insn}
18514 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18515 custom instruction with encoding @var{N} when generating code that uses
18516 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18517 instruction 253 for single-precision floating-point add operations instead
18518 of the default behavior of using a library call.
18520 The following values of @var{insn} are supported. Except as otherwise
18521 noted, floating-point operations are expected to be implemented with
18522 normal IEEE 754 semantics and correspond directly to the C operators or the
18523 equivalent GCC built-in functions (@pxref{Other Builtins}).
18525 Single-precision floating point:
18528 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18529 Binary arithmetic operations.
18535 Unary absolute value.
18537 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18538 Comparison operations.
18540 @item @samp{fmins}, @samp{fmaxs}
18541 Floating-point minimum and maximum. These instructions are only
18542 generated if @option{-ffinite-math-only} is specified.
18544 @item @samp{fsqrts}
18545 Unary square root operation.
18547 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18548 Floating-point trigonometric and exponential functions. These instructions
18549 are only generated if @option{-funsafe-math-optimizations} is also specified.
18553 Double-precision floating point:
18556 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18557 Binary arithmetic operations.
18563 Unary absolute value.
18565 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18566 Comparison operations.
18568 @item @samp{fmind}, @samp{fmaxd}
18569 Double-precision minimum and maximum. These instructions are only
18570 generated if @option{-ffinite-math-only} is specified.
18572 @item @samp{fsqrtd}
18573 Unary square root operation.
18575 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18576 Double-precision trigonometric and exponential functions. These instructions
18577 are only generated if @option{-funsafe-math-optimizations} is also specified.
18583 @item @samp{fextsd}
18584 Conversion from single precision to double precision.
18586 @item @samp{ftruncds}
18587 Conversion from double precision to single precision.
18589 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18590 Conversion from floating point to signed or unsigned integer types, with
18591 truncation towards zero.
18594 Conversion from single-precision floating point to signed integer,
18595 rounding to the nearest integer and ties away from zero.
18596 This corresponds to the @code{__builtin_lroundf} function when
18597 @option{-fno-math-errno} is used.
18599 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18600 Conversion from signed or unsigned integer types to floating-point types.
18604 In addition, all of the following transfer instructions for internal
18605 registers X and Y must be provided to use any of the double-precision
18606 floating-point instructions. Custom instructions taking two
18607 double-precision source operands expect the first operand in the
18608 64-bit register X. The other operand (or only operand of a unary
18609 operation) is given to the custom arithmetic instruction with the
18610 least significant half in source register @var{src1} and the most
18611 significant half in @var{src2}. A custom instruction that returns a
18612 double-precision result returns the most significant 32 bits in the
18613 destination register and the other half in 32-bit register Y.
18614 GCC automatically generates the necessary code sequences to write
18615 register X and/or read register Y when double-precision floating-point
18616 instructions are used.
18621 Write @var{src1} into the least significant half of X and @var{src2} into
18622 the most significant half of X.
18625 Write @var{src1} into Y.
18627 @item @samp{frdxhi}, @samp{frdxlo}
18628 Read the most or least (respectively) significant half of X and store it in
18632 Read the value of Y and store it into @var{dest}.
18635 Note that you can gain more local control over generation of Nios II custom
18636 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18637 and @code{target("no-custom-@var{insn}")} function attributes
18638 (@pxref{Function Attributes})
18639 or pragmas (@pxref{Function Specific Option Pragmas}).
18641 @item -mcustom-fpu-cfg=@var{name}
18642 @opindex mcustom-fpu-cfg
18644 This option enables a predefined, named set of custom instruction encodings
18645 (see @option{-mcustom-@var{insn}} above).
18646 Currently, the following sets are defined:
18648 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18649 @gccoptlist{-mcustom-fmuls=252 @gol
18650 -mcustom-fadds=253 @gol
18651 -mcustom-fsubs=254 @gol
18652 -fsingle-precision-constant}
18654 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18655 @gccoptlist{-mcustom-fmuls=252 @gol
18656 -mcustom-fadds=253 @gol
18657 -mcustom-fsubs=254 @gol
18658 -mcustom-fdivs=255 @gol
18659 -fsingle-precision-constant}
18661 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18662 @gccoptlist{-mcustom-floatus=243 @gol
18663 -mcustom-fixsi=244 @gol
18664 -mcustom-floatis=245 @gol
18665 -mcustom-fcmpgts=246 @gol
18666 -mcustom-fcmples=249 @gol
18667 -mcustom-fcmpeqs=250 @gol
18668 -mcustom-fcmpnes=251 @gol
18669 -mcustom-fmuls=252 @gol
18670 -mcustom-fadds=253 @gol
18671 -mcustom-fsubs=254 @gol
18672 -mcustom-fdivs=255 @gol
18673 -fsingle-precision-constant}
18675 Custom instruction assignments given by individual
18676 @option{-mcustom-@var{insn}=} options override those given by
18677 @option{-mcustom-fpu-cfg=}, regardless of the
18678 order of the options on the command line.
18680 Note that you can gain more local control over selection of a FPU
18681 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18682 function attribute (@pxref{Function Attributes})
18683 or pragma (@pxref{Function Specific Option Pragmas}).
18687 These additional @samp{-m} options are available for the Altera Nios II
18688 ELF (bare-metal) target:
18694 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18695 startup and termination code, and is typically used in conjunction with
18696 @option{-msys-crt0=} to specify the location of the alternate startup code
18697 provided by the HAL BSP.
18701 Link with a limited version of the C library, @option{-lsmallc}, rather than
18704 @item -msys-crt0=@var{startfile}
18706 @var{startfile} is the file name of the startfile (crt0) to use
18707 when linking. This option is only useful in conjunction with @option{-mhal}.
18709 @item -msys-lib=@var{systemlib}
18711 @var{systemlib} is the library name of the library that provides
18712 low-level system calls required by the C library,
18713 e.g. @code{read} and @code{write}.
18714 This option is typically used to link with a library provided by a HAL BSP.
18718 @node Nvidia PTX Options
18719 @subsection Nvidia PTX Options
18720 @cindex Nvidia PTX options
18721 @cindex nvptx options
18723 These options are defined for Nvidia PTX:
18731 Generate code for 32-bit or 64-bit ABI.
18734 @opindex mmainkernel
18735 Link in code for a __main kernel. This is for stand-alone instead of
18736 offloading execution.
18740 @node PDP-11 Options
18741 @subsection PDP-11 Options
18742 @cindex PDP-11 Options
18744 These options are defined for the PDP-11:
18749 Use hardware FPP floating point. This is the default. (FIS floating
18750 point on the PDP-11/40 is not supported.)
18753 @opindex msoft-float
18754 Do not use hardware floating point.
18758 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18762 Return floating-point results in memory. This is the default.
18766 Generate code for a PDP-11/40.
18770 Generate code for a PDP-11/45. This is the default.
18774 Generate code for a PDP-11/10.
18776 @item -mbcopy-builtin
18777 @opindex mbcopy-builtin
18778 Use inline @code{movmemhi} patterns for copying memory. This is the
18783 Do not use inline @code{movmemhi} patterns for copying memory.
18789 Use 16-bit @code{int}. This is the default.
18795 Use 32-bit @code{int}.
18798 @itemx -mno-float32
18800 @opindex mno-float32
18801 Use 64-bit @code{float}. This is the default.
18804 @itemx -mno-float64
18806 @opindex mno-float64
18807 Use 32-bit @code{float}.
18811 Use @code{abshi2} pattern. This is the default.
18815 Do not use @code{abshi2} pattern.
18817 @item -mbranch-expensive
18818 @opindex mbranch-expensive
18819 Pretend that branches are expensive. This is for experimenting with
18820 code generation only.
18822 @item -mbranch-cheap
18823 @opindex mbranch-cheap
18824 Do not pretend that branches are expensive. This is the default.
18828 Use Unix assembler syntax. This is the default when configured for
18829 @samp{pdp11-*-bsd}.
18833 Use DEC assembler syntax. This is the default when configured for any
18834 PDP-11 target other than @samp{pdp11-*-bsd}.
18837 @node picoChip Options
18838 @subsection picoChip Options
18839 @cindex picoChip options
18841 These @samp{-m} options are defined for picoChip implementations:
18845 @item -mae=@var{ae_type}
18847 Set the instruction set, register set, and instruction scheduling
18848 parameters for array element type @var{ae_type}. Supported values
18849 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18851 @option{-mae=ANY} selects a completely generic AE type. Code
18852 generated with this option runs on any of the other AE types. The
18853 code is not as efficient as it would be if compiled for a specific
18854 AE type, and some types of operation (e.g., multiplication) do not
18855 work properly on all types of AE.
18857 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
18858 for compiled code, and is the default.
18860 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
18861 option may suffer from poor performance of byte (char) manipulation,
18862 since the DSP AE does not provide hardware support for byte load/stores.
18864 @item -msymbol-as-address
18865 Enable the compiler to directly use a symbol name as an address in a
18866 load/store instruction, without first loading it into a
18867 register. Typically, the use of this option generates larger
18868 programs, which run faster than when the option isn't used. However, the
18869 results vary from program to program, so it is left as a user option,
18870 rather than being permanently enabled.
18872 @item -mno-inefficient-warnings
18873 Disables warnings about the generation of inefficient code. These
18874 warnings can be generated, for example, when compiling code that
18875 performs byte-level memory operations on the MAC AE type. The MAC AE has
18876 no hardware support for byte-level memory operations, so all byte
18877 load/stores must be synthesized from word load/store operations. This is
18878 inefficient and a warning is generated to indicate
18879 that you should rewrite the code to avoid byte operations, or to target
18880 an AE type that has the necessary hardware support. This option disables
18885 @node PowerPC Options
18886 @subsection PowerPC Options
18887 @cindex PowerPC options
18889 These are listed under @xref{RS/6000 and PowerPC Options}.
18892 @subsection RL78 Options
18893 @cindex RL78 Options
18899 Links in additional target libraries to support operation within a
18908 Specifies the type of hardware multiplication and division support to
18909 be used. The simplest is @code{none}, which uses software for both
18910 multiplication and division. This is the default. The @code{g13}
18911 value is for the hardware multiply/divide peripheral found on the
18912 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
18913 the multiplication and division instructions supported by the RL78/G14
18914 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
18915 the value @code{mg10} is an alias for @code{none}.
18917 In addition a C preprocessor macro is defined, based upon the setting
18918 of this option. Possible values are: @code{__RL78_MUL_NONE__},
18919 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
18926 Specifies the RL78 core to target. The default is the G14 core, also
18927 known as an S3 core or just RL78. The G13 or S2 core does not have
18928 multiply or divide instructions, instead it uses a hardware peripheral
18929 for these operations. The G10 or S1 core does not have register
18930 banks, so it uses a different calling convention.
18932 If this option is set it also selects the type of hardware multiply
18933 support to use, unless this is overridden by an explicit
18934 @option{-mmul=none} option on the command line. Thus specifying
18935 @option{-mcpu=g13} enables the use of the G13 hardware multiply
18936 peripheral and specifying @option{-mcpu=g10} disables the use of
18937 hardware multipications altogether.
18939 Note, although the RL78/G14 core is the default target, specifying
18940 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
18941 change the behaviour of the toolchain since it also enables G14
18942 hardware multiply support. If these options are not specified on the
18943 command line then software multiplication routines will be used even
18944 though the code targets the RL78 core. This is for backwards
18945 compatibility with older toolchains which did not have hardware
18946 multiply and divide support.
18948 In addition a C preprocessor macro is defined, based upon the setting
18949 of this option. Possible values are: @code{__RL78_G10__},
18950 @code{__RL78_G13__} or @code{__RL78_G14__}.
18960 These are aliases for the corresponding @option{-mcpu=} option. They
18961 are provided for backwards compatibility.
18965 Allow the compiler to use all of the available registers. By default
18966 registers @code{r24..r31} are reserved for use in interrupt handlers.
18967 With this option enabled these registers can be used in ordinary
18970 @item -m64bit-doubles
18971 @itemx -m32bit-doubles
18972 @opindex m64bit-doubles
18973 @opindex m32bit-doubles
18974 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
18975 or 32 bits (@option{-m32bit-doubles}) in size. The default is
18976 @option{-m32bit-doubles}.
18980 @node RS/6000 and PowerPC Options
18981 @subsection IBM RS/6000 and PowerPC Options
18982 @cindex RS/6000 and PowerPC Options
18983 @cindex IBM RS/6000 and PowerPC Options
18985 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
18987 @item -mpowerpc-gpopt
18988 @itemx -mno-powerpc-gpopt
18989 @itemx -mpowerpc-gfxopt
18990 @itemx -mno-powerpc-gfxopt
18993 @itemx -mno-powerpc64
18997 @itemx -mno-popcntb
18999 @itemx -mno-popcntd
19008 @itemx -mno-hard-dfp
19009 @opindex mpowerpc-gpopt
19010 @opindex mno-powerpc-gpopt
19011 @opindex mpowerpc-gfxopt
19012 @opindex mno-powerpc-gfxopt
19013 @opindex mpowerpc64
19014 @opindex mno-powerpc64
19018 @opindex mno-popcntb
19020 @opindex mno-popcntd
19026 @opindex mno-mfpgpr
19028 @opindex mno-hard-dfp
19029 You use these options to specify which instructions are available on the
19030 processor you are using. The default value of these options is
19031 determined when configuring GCC@. Specifying the
19032 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19033 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19034 rather than the options listed above.
19036 Specifying @option{-mpowerpc-gpopt} allows
19037 GCC to use the optional PowerPC architecture instructions in the
19038 General Purpose group, including floating-point square root. Specifying
19039 @option{-mpowerpc-gfxopt} allows GCC to
19040 use the optional PowerPC architecture instructions in the Graphics
19041 group, including floating-point select.
19043 The @option{-mmfcrf} option allows GCC to generate the move from
19044 condition register field instruction implemented on the POWER4
19045 processor and other processors that support the PowerPC V2.01
19047 The @option{-mpopcntb} option allows GCC to generate the popcount and
19048 double-precision FP reciprocal estimate instruction implemented on the
19049 POWER5 processor and other processors that support the PowerPC V2.02
19051 The @option{-mpopcntd} option allows GCC to generate the popcount
19052 instruction implemented on the POWER7 processor and other processors
19053 that support the PowerPC V2.06 architecture.
19054 The @option{-mfprnd} option allows GCC to generate the FP round to
19055 integer instructions implemented on the POWER5+ processor and other
19056 processors that support the PowerPC V2.03 architecture.
19057 The @option{-mcmpb} option allows GCC to generate the compare bytes
19058 instruction implemented on the POWER6 processor and other processors
19059 that support the PowerPC V2.05 architecture.
19060 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19061 general-purpose register instructions implemented on the POWER6X
19062 processor and other processors that support the extended PowerPC V2.05
19064 The @option{-mhard-dfp} option allows GCC to generate the decimal
19065 floating-point instructions implemented on some POWER processors.
19067 The @option{-mpowerpc64} option allows GCC to generate the additional
19068 64-bit instructions that are found in the full PowerPC64 architecture
19069 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19070 @option{-mno-powerpc64}.
19072 @item -mcpu=@var{cpu_type}
19074 Set architecture type, register usage, and
19075 instruction scheduling parameters for machine type @var{cpu_type}.
19076 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19077 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19078 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19079 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19080 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19081 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19082 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19083 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19084 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19085 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
19086 @samp{powerpc64}, @samp{powerpc64le}, and @samp{rs64}.
19088 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
19089 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
19090 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
19091 architecture machine types, with an appropriate, generic processor
19092 model assumed for scheduling purposes.
19094 The other options specify a specific processor. Code generated under
19095 those options runs best on that processor, and may not run at all on
19098 The @option{-mcpu} options automatically enable or disable the
19101 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19102 -mpopcntb -mpopcntd -mpowerpc64 @gol
19103 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19104 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19105 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19106 -mquad-memory -mquad-memory-atomic}
19108 The particular options set for any particular CPU varies between
19109 compiler versions, depending on what setting seems to produce optimal
19110 code for that CPU; it doesn't necessarily reflect the actual hardware's
19111 capabilities. If you wish to set an individual option to a particular
19112 value, you may specify it after the @option{-mcpu} option, like
19113 @option{-mcpu=970 -mno-altivec}.
19115 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19116 not enabled or disabled by the @option{-mcpu} option at present because
19117 AIX does not have full support for these options. You may still
19118 enable or disable them individually if you're sure it'll work in your
19121 @item -mtune=@var{cpu_type}
19123 Set the instruction scheduling parameters for machine type
19124 @var{cpu_type}, but do not set the architecture type or register usage,
19125 as @option{-mcpu=@var{cpu_type}} does. The same
19126 values for @var{cpu_type} are used for @option{-mtune} as for
19127 @option{-mcpu}. If both are specified, the code generated uses the
19128 architecture and registers set by @option{-mcpu}, but the
19129 scheduling parameters set by @option{-mtune}.
19131 @item -mcmodel=small
19132 @opindex mcmodel=small
19133 Generate PowerPC64 code for the small model: The TOC is limited to
19136 @item -mcmodel=medium
19137 @opindex mcmodel=medium
19138 Generate PowerPC64 code for the medium model: The TOC and other static
19139 data may be up to a total of 4G in size.
19141 @item -mcmodel=large
19142 @opindex mcmodel=large
19143 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19144 in size. Other data and code is only limited by the 64-bit address
19148 @itemx -mno-altivec
19150 @opindex mno-altivec
19151 Generate code that uses (does not use) AltiVec instructions, and also
19152 enable the use of built-in functions that allow more direct access to
19153 the AltiVec instruction set. You may also need to set
19154 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19157 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19158 @option{-maltivec=be}, the element order for Altivec intrinsics such
19159 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19160 match array element order corresponding to the endianness of the
19161 target. That is, element zero identifies the leftmost element in a
19162 vector register when targeting a big-endian platform, and identifies
19163 the rightmost element in a vector register when targeting a
19164 little-endian platform.
19167 @opindex maltivec=be
19168 Generate Altivec instructions using big-endian element order,
19169 regardless of whether the target is big- or little-endian. This is
19170 the default when targeting a big-endian platform.
19172 The element order is used to interpret element numbers in Altivec
19173 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19174 @code{vec_insert}. By default, these match array element order
19175 corresponding to the endianness for the target.
19178 @opindex maltivec=le
19179 Generate Altivec instructions using little-endian element order,
19180 regardless of whether the target is big- or little-endian. This is
19181 the default when targeting a little-endian platform. This option is
19182 currently ignored when targeting a big-endian platform.
19184 The element order is used to interpret element numbers in Altivec
19185 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19186 @code{vec_insert}. By default, these match array element order
19187 corresponding to the endianness for the target.
19192 @opindex mno-vrsave
19193 Generate VRSAVE instructions when generating AltiVec code.
19195 @item -mgen-cell-microcode
19196 @opindex mgen-cell-microcode
19197 Generate Cell microcode instructions.
19199 @item -mwarn-cell-microcode
19200 @opindex mwarn-cell-microcode
19201 Warn when a Cell microcode instruction is emitted. An example
19202 of a Cell microcode instruction is a variable shift.
19205 @opindex msecure-plt
19206 Generate code that allows @command{ld} and @command{ld.so}
19207 to build executables and shared
19208 libraries with non-executable @code{.plt} and @code{.got} sections.
19210 32-bit SYSV ABI option.
19214 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19216 requires @code{.plt} and @code{.got}
19217 sections that are both writable and executable.
19218 This is a PowerPC 32-bit SYSV ABI option.
19224 This switch enables or disables the generation of ISEL instructions.
19226 @item -misel=@var{yes/no}
19227 This switch has been deprecated. Use @option{-misel} and
19228 @option{-mno-isel} instead.
19234 This switch enables or disables the generation of SPE simd
19240 @opindex mno-paired
19241 This switch enables or disables the generation of PAIRED simd
19244 @item -mspe=@var{yes/no}
19245 This option has been deprecated. Use @option{-mspe} and
19246 @option{-mno-spe} instead.
19252 Generate code that uses (does not use) vector/scalar (VSX)
19253 instructions, and also enable the use of built-in functions that allow
19254 more direct access to the VSX instruction set.
19259 @opindex mno-crypto
19260 Enable the use (disable) of the built-in functions that allow direct
19261 access to the cryptographic instructions that were added in version
19262 2.07 of the PowerPC ISA.
19264 @item -mdirect-move
19265 @itemx -mno-direct-move
19266 @opindex mdirect-move
19267 @opindex mno-direct-move
19268 Generate code that uses (does not use) the instructions to move data
19269 between the general purpose registers and the vector/scalar (VSX)
19270 registers that were added in version 2.07 of the PowerPC ISA.
19272 @item -mpower8-fusion
19273 @itemx -mno-power8-fusion
19274 @opindex mpower8-fusion
19275 @opindex mno-power8-fusion
19276 Generate code that keeps (does not keeps) some integer operations
19277 adjacent so that the instructions can be fused together on power8 and
19280 @item -mpower8-vector
19281 @itemx -mno-power8-vector
19282 @opindex mpower8-vector
19283 @opindex mno-power8-vector
19284 Generate code that uses (does not use) the vector and scalar
19285 instructions that were added in version 2.07 of the PowerPC ISA. Also
19286 enable the use of built-in functions that allow more direct access to
19287 the vector instructions.
19289 @item -mquad-memory
19290 @itemx -mno-quad-memory
19291 @opindex mquad-memory
19292 @opindex mno-quad-memory
19293 Generate code that uses (does not use) the non-atomic quad word memory
19294 instructions. The @option{-mquad-memory} option requires use of
19297 @item -mquad-memory-atomic
19298 @itemx -mno-quad-memory-atomic
19299 @opindex mquad-memory-atomic
19300 @opindex mno-quad-memory-atomic
19301 Generate code that uses (does not use) the atomic quad word memory
19302 instructions. The @option{-mquad-memory-atomic} option requires use of
19305 @item -mupper-regs-df
19306 @itemx -mno-upper-regs-df
19307 @opindex mupper-regs-df
19308 @opindex mno-upper-regs-df
19309 Generate code that uses (does not use) the scalar double precision
19310 instructions that target all 64 registers in the vector/scalar
19311 floating point register set that were added in version 2.06 of the
19312 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
19313 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19314 @option{-mvsx} options.
19316 @item -mupper-regs-sf
19317 @itemx -mno-upper-regs-sf
19318 @opindex mupper-regs-sf
19319 @opindex mno-upper-regs-sf
19320 Generate code that uses (does not use) the scalar single precision
19321 instructions that target all 64 registers in the vector/scalar
19322 floating point register set that were added in version 2.07 of the
19323 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
19324 use either of the @option{-mcpu=power8} or @option{-mpower8-vector}
19328 @itemx -mno-upper-regs
19329 @opindex mupper-regs
19330 @opindex mno-upper-regs
19331 Generate code that uses (does not use) the scalar
19332 instructions that target all 64 registers in the vector/scalar
19333 floating point register set, depending on the model of the machine.
19335 If the @option{-mno-upper-regs} option is used, it turns off both
19336 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19338 @item -mfloat-gprs=@var{yes/single/double/no}
19339 @itemx -mfloat-gprs
19340 @opindex mfloat-gprs
19341 This switch enables or disables the generation of floating-point
19342 operations on the general-purpose registers for architectures that
19345 The argument @samp{yes} or @samp{single} enables the use of
19346 single-precision floating-point operations.
19348 The argument @samp{double} enables the use of single and
19349 double-precision floating-point operations.
19351 The argument @samp{no} disables floating-point operations on the
19352 general-purpose registers.
19354 This option is currently only available on the MPC854x.
19360 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19361 targets (including GNU/Linux). The 32-bit environment sets int, long
19362 and pointer to 32 bits and generates code that runs on any PowerPC
19363 variant. The 64-bit environment sets int to 32 bits and long and
19364 pointer to 64 bits, and generates code for PowerPC64, as for
19365 @option{-mpowerpc64}.
19368 @itemx -mno-fp-in-toc
19369 @itemx -mno-sum-in-toc
19370 @itemx -mminimal-toc
19372 @opindex mno-fp-in-toc
19373 @opindex mno-sum-in-toc
19374 @opindex mminimal-toc
19375 Modify generation of the TOC (Table Of Contents), which is created for
19376 every executable file. The @option{-mfull-toc} option is selected by
19377 default. In that case, GCC allocates at least one TOC entry for
19378 each unique non-automatic variable reference in your program. GCC
19379 also places floating-point constants in the TOC@. However, only
19380 16,384 entries are available in the TOC@.
19382 If you receive a linker error message that saying you have overflowed
19383 the available TOC space, you can reduce the amount of TOC space used
19384 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19385 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19386 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19387 generate code to calculate the sum of an address and a constant at
19388 run time instead of putting that sum into the TOC@. You may specify one
19389 or both of these options. Each causes GCC to produce very slightly
19390 slower and larger code at the expense of conserving TOC space.
19392 If you still run out of space in the TOC even when you specify both of
19393 these options, specify @option{-mminimal-toc} instead. This option causes
19394 GCC to make only one TOC entry for every file. When you specify this
19395 option, GCC produces code that is slower and larger but which
19396 uses extremely little TOC space. You may wish to use this option
19397 only on files that contain less frequently-executed code.
19403 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19404 @code{long} type, and the infrastructure needed to support them.
19405 Specifying @option{-maix64} implies @option{-mpowerpc64},
19406 while @option{-maix32} disables the 64-bit ABI and
19407 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19410 @itemx -mno-xl-compat
19411 @opindex mxl-compat
19412 @opindex mno-xl-compat
19413 Produce code that conforms more closely to IBM XL compiler semantics
19414 when using AIX-compatible ABI@. Pass floating-point arguments to
19415 prototyped functions beyond the register save area (RSA) on the stack
19416 in addition to argument FPRs. Do not assume that most significant
19417 double in 128-bit long double value is properly rounded when comparing
19418 values and converting to double. Use XL symbol names for long double
19421 The AIX calling convention was extended but not initially documented to
19422 handle an obscure K&R C case of calling a function that takes the
19423 address of its arguments with fewer arguments than declared. IBM XL
19424 compilers access floating-point arguments that do not fit in the
19425 RSA from the stack when a subroutine is compiled without
19426 optimization. Because always storing floating-point arguments on the
19427 stack is inefficient and rarely needed, this option is not enabled by
19428 default and only is necessary when calling subroutines compiled by IBM
19429 XL compilers without optimization.
19433 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19434 application written to use message passing with special startup code to
19435 enable the application to run. The system must have PE installed in the
19436 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19437 must be overridden with the @option{-specs=} option to specify the
19438 appropriate directory location. The Parallel Environment does not
19439 support threads, so the @option{-mpe} option and the @option{-pthread}
19440 option are incompatible.
19442 @item -malign-natural
19443 @itemx -malign-power
19444 @opindex malign-natural
19445 @opindex malign-power
19446 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19447 @option{-malign-natural} overrides the ABI-defined alignment of larger
19448 types, such as floating-point doubles, on their natural size-based boundary.
19449 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19450 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19452 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19456 @itemx -mhard-float
19457 @opindex msoft-float
19458 @opindex mhard-float
19459 Generate code that does not use (uses) the floating-point register set.
19460 Software floating-point emulation is provided if you use the
19461 @option{-msoft-float} option, and pass the option to GCC when linking.
19463 @item -msingle-float
19464 @itemx -mdouble-float
19465 @opindex msingle-float
19466 @opindex mdouble-float
19467 Generate code for single- or double-precision floating-point operations.
19468 @option{-mdouble-float} implies @option{-msingle-float}.
19471 @opindex msimple-fpu
19472 Do not generate @code{sqrt} and @code{div} instructions for hardware
19473 floating-point unit.
19475 @item -mfpu=@var{name}
19477 Specify type of floating-point unit. Valid values for @var{name} are
19478 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19479 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19480 @samp{sp_full} (equivalent to @option{-msingle-float}),
19481 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19484 @opindex mxilinx-fpu
19485 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19488 @itemx -mno-multiple
19490 @opindex mno-multiple
19491 Generate code that uses (does not use) the load multiple word
19492 instructions and the store multiple word instructions. These
19493 instructions are generated by default on POWER systems, and not
19494 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19495 PowerPC systems, since those instructions do not work when the
19496 processor is in little-endian mode. The exceptions are PPC740 and
19497 PPC750 which permit these instructions in little-endian mode.
19502 @opindex mno-string
19503 Generate code that uses (does not use) the load string instructions
19504 and the store string word instructions to save multiple registers and
19505 do small block moves. These instructions are generated by default on
19506 POWER systems, and not generated on PowerPC systems. Do not use
19507 @option{-mstring} on little-endian PowerPC systems, since those
19508 instructions do not work when the processor is in little-endian mode.
19509 The exceptions are PPC740 and PPC750 which permit these instructions
19510 in little-endian mode.
19515 @opindex mno-update
19516 Generate code that uses (does not use) the load or store instructions
19517 that update the base register to the address of the calculated memory
19518 location. These instructions are generated by default. If you use
19519 @option{-mno-update}, there is a small window between the time that the
19520 stack pointer is updated and the address of the previous frame is
19521 stored, which means code that walks the stack frame across interrupts or
19522 signals may get corrupted data.
19524 @item -mavoid-indexed-addresses
19525 @itemx -mno-avoid-indexed-addresses
19526 @opindex mavoid-indexed-addresses
19527 @opindex mno-avoid-indexed-addresses
19528 Generate code that tries to avoid (not avoid) the use of indexed load
19529 or store instructions. These instructions can incur a performance
19530 penalty on Power6 processors in certain situations, such as when
19531 stepping through large arrays that cross a 16M boundary. This option
19532 is enabled by default when targeting Power6 and disabled otherwise.
19535 @itemx -mno-fused-madd
19536 @opindex mfused-madd
19537 @opindex mno-fused-madd
19538 Generate code that uses (does not use) the floating-point multiply and
19539 accumulate instructions. These instructions are generated by default
19540 if hardware floating point is used. The machine-dependent
19541 @option{-mfused-madd} option is now mapped to the machine-independent
19542 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19543 mapped to @option{-ffp-contract=off}.
19549 Generate code that uses (does not use) the half-word multiply and
19550 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19551 These instructions are generated by default when targeting those
19558 Generate code that uses (does not use) the string-search @samp{dlmzb}
19559 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19560 generated by default when targeting those processors.
19562 @item -mno-bit-align
19564 @opindex mno-bit-align
19565 @opindex mbit-align
19566 On System V.4 and embedded PowerPC systems do not (do) force structures
19567 and unions that contain bit-fields to be aligned to the base type of the
19570 For example, by default a structure containing nothing but 8
19571 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19572 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19573 the structure is aligned to a 1-byte boundary and is 1 byte in
19576 @item -mno-strict-align
19577 @itemx -mstrict-align
19578 @opindex mno-strict-align
19579 @opindex mstrict-align
19580 On System V.4 and embedded PowerPC systems do not (do) assume that
19581 unaligned memory references are handled by the system.
19583 @item -mrelocatable
19584 @itemx -mno-relocatable
19585 @opindex mrelocatable
19586 @opindex mno-relocatable
19587 Generate code that allows (does not allow) a static executable to be
19588 relocated to a different address at run time. A simple embedded
19589 PowerPC system loader should relocate the entire contents of
19590 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19591 a table of 32-bit addresses generated by this option. For this to
19592 work, all objects linked together must be compiled with
19593 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19594 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19596 @item -mrelocatable-lib
19597 @itemx -mno-relocatable-lib
19598 @opindex mrelocatable-lib
19599 @opindex mno-relocatable-lib
19600 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19601 @code{.fixup} section to allow static executables to be relocated at
19602 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19603 alignment of @option{-mrelocatable}. Objects compiled with
19604 @option{-mrelocatable-lib} may be linked with objects compiled with
19605 any combination of the @option{-mrelocatable} options.
19611 On System V.4 and embedded PowerPC systems do not (do) assume that
19612 register 2 contains a pointer to a global area pointing to the addresses
19613 used in the program.
19616 @itemx -mlittle-endian
19618 @opindex mlittle-endian
19619 On System V.4 and embedded PowerPC systems compile code for the
19620 processor in little-endian mode. The @option{-mlittle-endian} option is
19621 the same as @option{-mlittle}.
19624 @itemx -mbig-endian
19626 @opindex mbig-endian
19627 On System V.4 and embedded PowerPC systems compile code for the
19628 processor in big-endian mode. The @option{-mbig-endian} option is
19629 the same as @option{-mbig}.
19631 @item -mdynamic-no-pic
19632 @opindex mdynamic-no-pic
19633 On Darwin and Mac OS X systems, compile code so that it is not
19634 relocatable, but that its external references are relocatable. The
19635 resulting code is suitable for applications, but not shared
19638 @item -msingle-pic-base
19639 @opindex msingle-pic-base
19640 Treat the register used for PIC addressing as read-only, rather than
19641 loading it in the prologue for each function. The runtime system is
19642 responsible for initializing this register with an appropriate value
19643 before execution begins.
19645 @item -mprioritize-restricted-insns=@var{priority}
19646 @opindex mprioritize-restricted-insns
19647 This option controls the priority that is assigned to
19648 dispatch-slot restricted instructions during the second scheduling
19649 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19650 or @samp{2} to assign no, highest, or second-highest (respectively)
19651 priority to dispatch-slot restricted
19654 @item -msched-costly-dep=@var{dependence_type}
19655 @opindex msched-costly-dep
19656 This option controls which dependences are considered costly
19657 by the target during instruction scheduling. The argument
19658 @var{dependence_type} takes one of the following values:
19662 No dependence is costly.
19665 All dependences are costly.
19667 @item @samp{true_store_to_load}
19668 A true dependence from store to load is costly.
19670 @item @samp{store_to_load}
19671 Any dependence from store to load is costly.
19674 Any dependence for which the latency is greater than or equal to
19675 @var{number} is costly.
19678 @item -minsert-sched-nops=@var{scheme}
19679 @opindex minsert-sched-nops
19680 This option controls which NOP insertion scheme is used during
19681 the second scheduling pass. The argument @var{scheme} takes one of the
19689 Pad with NOPs any dispatch group that has vacant issue slots,
19690 according to the scheduler's grouping.
19692 @item @samp{regroup_exact}
19693 Insert NOPs to force costly dependent insns into
19694 separate groups. Insert exactly as many NOPs as needed to force an insn
19695 to a new group, according to the estimated processor grouping.
19698 Insert NOPs to force costly dependent insns into
19699 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19703 @opindex mcall-sysv
19704 On System V.4 and embedded PowerPC systems compile code using calling
19705 conventions that adhere to the March 1995 draft of the System V
19706 Application Binary Interface, PowerPC processor supplement. This is the
19707 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19709 @item -mcall-sysv-eabi
19711 @opindex mcall-sysv-eabi
19712 @opindex mcall-eabi
19713 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19715 @item -mcall-sysv-noeabi
19716 @opindex mcall-sysv-noeabi
19717 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19719 @item -mcall-aixdesc
19721 On System V.4 and embedded PowerPC systems compile code for the AIX
19725 @opindex mcall-linux
19726 On System V.4 and embedded PowerPC systems compile code for the
19727 Linux-based GNU system.
19729 @item -mcall-freebsd
19730 @opindex mcall-freebsd
19731 On System V.4 and embedded PowerPC systems compile code for the
19732 FreeBSD operating system.
19734 @item -mcall-netbsd
19735 @opindex mcall-netbsd
19736 On System V.4 and embedded PowerPC systems compile code for the
19737 NetBSD operating system.
19739 @item -mcall-openbsd
19740 @opindex mcall-netbsd
19741 On System V.4 and embedded PowerPC systems compile code for the
19742 OpenBSD operating system.
19744 @item -maix-struct-return
19745 @opindex maix-struct-return
19746 Return all structures in memory (as specified by the AIX ABI)@.
19748 @item -msvr4-struct-return
19749 @opindex msvr4-struct-return
19750 Return structures smaller than 8 bytes in registers (as specified by the
19753 @item -mabi=@var{abi-type}
19755 Extend the current ABI with a particular extension, or remove such extension.
19756 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
19757 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
19758 @samp{elfv1}, @samp{elfv2}@.
19762 Extend the current ABI with SPE ABI extensions. This does not change
19763 the default ABI, instead it adds the SPE ABI extensions to the current
19767 @opindex mabi=no-spe
19768 Disable Book-E SPE ABI extensions for the current ABI@.
19770 @item -mabi=ibmlongdouble
19771 @opindex mabi=ibmlongdouble
19772 Change the current ABI to use IBM extended-precision long double.
19773 This is a PowerPC 32-bit SYSV ABI option.
19775 @item -mabi=ieeelongdouble
19776 @opindex mabi=ieeelongdouble
19777 Change the current ABI to use IEEE extended-precision long double.
19778 This is a PowerPC 32-bit Linux ABI option.
19781 @opindex mabi=elfv1
19782 Change the current ABI to use the ELFv1 ABI.
19783 This is the default ABI for big-endian PowerPC 64-bit Linux.
19784 Overriding the default ABI requires special system support and is
19785 likely to fail in spectacular ways.
19788 @opindex mabi=elfv2
19789 Change the current ABI to use the ELFv2 ABI.
19790 This is the default ABI for little-endian PowerPC 64-bit Linux.
19791 Overriding the default ABI requires special system support and is
19792 likely to fail in spectacular ways.
19795 @itemx -mno-prototype
19796 @opindex mprototype
19797 @opindex mno-prototype
19798 On System V.4 and embedded PowerPC systems assume that all calls to
19799 variable argument functions are properly prototyped. Otherwise, the
19800 compiler must insert an instruction before every non-prototyped call to
19801 set or clear bit 6 of the condition code register (@code{CR}) to
19802 indicate whether floating-point values are passed in the floating-point
19803 registers in case the function takes variable arguments. With
19804 @option{-mprototype}, only calls to prototyped variable argument functions
19805 set or clear the bit.
19809 On embedded PowerPC systems, assume that the startup module is called
19810 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19811 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19816 On embedded PowerPC systems, assume that the startup module is called
19817 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19822 On embedded PowerPC systems, assume that the startup module is called
19823 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19826 @item -myellowknife
19827 @opindex myellowknife
19828 On embedded PowerPC systems, assume that the startup module is called
19829 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19834 On System V.4 and embedded PowerPC systems, specify that you are
19835 compiling for a VxWorks system.
19839 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
19840 header to indicate that @samp{eabi} extended relocations are used.
19846 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19847 Embedded Applications Binary Interface (EABI), which is a set of
19848 modifications to the System V.4 specifications. Selecting @option{-meabi}
19849 means that the stack is aligned to an 8-byte boundary, a function
19850 @code{__eabi} is called from @code{main} to set up the EABI
19851 environment, and the @option{-msdata} option can use both @code{r2} and
19852 @code{r13} to point to two separate small data areas. Selecting
19853 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19854 no EABI initialization function is called from @code{main}, and the
19855 @option{-msdata} option only uses @code{r13} to point to a single
19856 small data area. The @option{-meabi} option is on by default if you
19857 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19860 @opindex msdata=eabi
19861 On System V.4 and embedded PowerPC systems, put small initialized
19862 @code{const} global and static data in the @code{.sdata2} section, which
19863 is pointed to by register @code{r2}. Put small initialized
19864 non-@code{const} global and static data in the @code{.sdata} section,
19865 which is pointed to by register @code{r13}. Put small uninitialized
19866 global and static data in the @code{.sbss} section, which is adjacent to
19867 the @code{.sdata} section. The @option{-msdata=eabi} option is
19868 incompatible with the @option{-mrelocatable} option. The
19869 @option{-msdata=eabi} option also sets the @option{-memb} option.
19872 @opindex msdata=sysv
19873 On System V.4 and embedded PowerPC systems, put small global and static
19874 data in the @code{.sdata} section, which is pointed to by register
19875 @code{r13}. Put small uninitialized global and static data in the
19876 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
19877 The @option{-msdata=sysv} option is incompatible with the
19878 @option{-mrelocatable} option.
19880 @item -msdata=default
19882 @opindex msdata=default
19884 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19885 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19886 same as @option{-msdata=sysv}.
19889 @opindex msdata=data
19890 On System V.4 and embedded PowerPC systems, put small global
19891 data in the @code{.sdata} section. Put small uninitialized global
19892 data in the @code{.sbss} section. Do not use register @code{r13}
19893 to address small data however. This is the default behavior unless
19894 other @option{-msdata} options are used.
19898 @opindex msdata=none
19900 On embedded PowerPC systems, put all initialized global and static data
19901 in the @code{.data} section, and all uninitialized data in the
19902 @code{.bss} section.
19904 @item -mblock-move-inline-limit=@var{num}
19905 @opindex mblock-move-inline-limit
19906 Inline all block moves (such as calls to @code{memcpy} or structure
19907 copies) less than or equal to @var{num} bytes. The minimum value for
19908 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19909 targets. The default value is target-specific.
19913 @cindex smaller data references (PowerPC)
19914 @cindex .sdata/.sdata2 references (PowerPC)
19915 On embedded PowerPC systems, put global and static items less than or
19916 equal to @var{num} bytes into the small data or BSS sections instead of
19917 the normal data or BSS section. By default, @var{num} is 8. The
19918 @option{-G @var{num}} switch is also passed to the linker.
19919 All modules should be compiled with the same @option{-G @var{num}} value.
19922 @itemx -mno-regnames
19924 @opindex mno-regnames
19925 On System V.4 and embedded PowerPC systems do (do not) emit register
19926 names in the assembly language output using symbolic forms.
19929 @itemx -mno-longcall
19931 @opindex mno-longcall
19932 By default assume that all calls are far away so that a longer and more
19933 expensive calling sequence is required. This is required for calls
19934 farther than 32 megabytes (33,554,432 bytes) from the current location.
19935 A short call is generated if the compiler knows
19936 the call cannot be that far away. This setting can be overridden by
19937 the @code{shortcall} function attribute, or by @code{#pragma
19940 Some linkers are capable of detecting out-of-range calls and generating
19941 glue code on the fly. On these systems, long calls are unnecessary and
19942 generate slower code. As of this writing, the AIX linker can do this,
19943 as can the GNU linker for PowerPC/64. It is planned to add this feature
19944 to the GNU linker for 32-bit PowerPC systems as well.
19946 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
19947 callee, L42}, plus a @dfn{branch island} (glue code). The two target
19948 addresses represent the callee and the branch island. The
19949 Darwin/PPC linker prefers the first address and generates a @code{bl
19950 callee} if the PPC @code{bl} instruction reaches the callee directly;
19951 otherwise, the linker generates @code{bl L42} to call the branch
19952 island. The branch island is appended to the body of the
19953 calling function; it computes the full 32-bit address of the callee
19956 On Mach-O (Darwin) systems, this option directs the compiler emit to
19957 the glue for every direct call, and the Darwin linker decides whether
19958 to use or discard it.
19960 In the future, GCC may ignore all longcall specifications
19961 when the linker is known to generate glue.
19963 @item -mtls-markers
19964 @itemx -mno-tls-markers
19965 @opindex mtls-markers
19966 @opindex mno-tls-markers
19967 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
19968 specifying the function argument. The relocation allows the linker to
19969 reliably associate function call with argument setup instructions for
19970 TLS optimization, which in turn allows GCC to better schedule the
19975 Adds support for multithreading with the @dfn{pthreads} library.
19976 This option sets flags for both the preprocessor and linker.
19981 This option enables use of the reciprocal estimate and
19982 reciprocal square root estimate instructions with additional
19983 Newton-Raphson steps to increase precision instead of doing a divide or
19984 square root and divide for floating-point arguments. You should use
19985 the @option{-ffast-math} option when using @option{-mrecip} (or at
19986 least @option{-funsafe-math-optimizations},
19987 @option{-finite-math-only}, @option{-freciprocal-math} and
19988 @option{-fno-trapping-math}). Note that while the throughput of the
19989 sequence is generally higher than the throughput of the non-reciprocal
19990 instruction, the precision of the sequence can be decreased by up to 2
19991 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
19994 @item -mrecip=@var{opt}
19995 @opindex mrecip=opt
19996 This option controls which reciprocal estimate instructions
19997 may be used. @var{opt} is a comma-separated list of options, which may
19998 be preceded by a @code{!} to invert the option:
20003 Enable all estimate instructions.
20006 Enable the default instructions, equivalent to @option{-mrecip}.
20009 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20012 Enable the reciprocal approximation instructions for both
20013 single and double precision.
20016 Enable the single-precision reciprocal approximation instructions.
20019 Enable the double-precision reciprocal approximation instructions.
20022 Enable the reciprocal square root approximation instructions for both
20023 single and double precision.
20026 Enable the single-precision reciprocal square root approximation instructions.
20029 Enable the double-precision reciprocal square root approximation instructions.
20033 So, for example, @option{-mrecip=all,!rsqrtd} enables
20034 all of the reciprocal estimate instructions, except for the
20035 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20036 which handle the double-precision reciprocal square root calculations.
20038 @item -mrecip-precision
20039 @itemx -mno-recip-precision
20040 @opindex mrecip-precision
20041 Assume (do not assume) that the reciprocal estimate instructions
20042 provide higher-precision estimates than is mandated by the PowerPC
20043 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20044 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20045 The double-precision square root estimate instructions are not generated by
20046 default on low-precision machines, since they do not provide an
20047 estimate that converges after three steps.
20049 @item -mveclibabi=@var{type}
20050 @opindex mveclibabi
20051 Specifies the ABI type to use for vectorizing intrinsics using an
20052 external library. The only type supported at present is @samp{mass},
20053 which specifies to use IBM's Mathematical Acceleration Subsystem
20054 (MASS) libraries for vectorizing intrinsics using external libraries.
20055 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20056 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20057 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20058 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20059 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20060 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20061 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20062 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20063 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20064 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20065 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20066 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20067 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20068 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20069 for power7. Both @option{-ftree-vectorize} and
20070 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20071 libraries must be specified at link time.
20076 Generate (do not generate) the @code{friz} instruction when the
20077 @option{-funsafe-math-optimizations} option is used to optimize
20078 rounding of floating-point values to 64-bit integer and back to floating
20079 point. The @code{friz} instruction does not return the same value if
20080 the floating-point number is too large to fit in an integer.
20082 @item -mpointers-to-nested-functions
20083 @itemx -mno-pointers-to-nested-functions
20084 @opindex mpointers-to-nested-functions
20085 Generate (do not generate) code to load up the static chain register
20086 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20087 systems where a function pointer points to a 3-word descriptor giving
20088 the function address, TOC value to be loaded in register @code{r2}, and
20089 static chain value to be loaded in register @code{r11}. The
20090 @option{-mpointers-to-nested-functions} is on by default. You cannot
20091 call through pointers to nested functions or pointers
20092 to functions compiled in other languages that use the static chain if
20093 you use @option{-mno-pointers-to-nested-functions}.
20095 @item -msave-toc-indirect
20096 @itemx -mno-save-toc-indirect
20097 @opindex msave-toc-indirect
20098 Generate (do not generate) code to save the TOC value in the reserved
20099 stack location in the function prologue if the function calls through
20100 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20101 saved in the prologue, it is saved just before the call through the
20102 pointer. The @option{-mno-save-toc-indirect} option is the default.
20104 @item -mcompat-align-parm
20105 @itemx -mno-compat-align-parm
20106 @opindex mcompat-align-parm
20107 Generate (do not generate) code to pass structure parameters with a
20108 maximum alignment of 64 bits, for compatibility with older versions
20111 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20112 structure parameter on a 128-bit boundary when that structure contained
20113 a member requiring 128-bit alignment. This is corrected in more
20114 recent versions of GCC. This option may be used to generate code
20115 that is compatible with functions compiled with older versions of
20118 The @option{-mno-compat-align-parm} option is the default.
20122 @subsection RX Options
20125 These command-line options are defined for RX targets:
20128 @item -m64bit-doubles
20129 @itemx -m32bit-doubles
20130 @opindex m64bit-doubles
20131 @opindex m32bit-doubles
20132 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20133 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20134 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20135 works on 32-bit values, which is why the default is
20136 @option{-m32bit-doubles}.
20142 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20143 floating-point hardware. The default is enabled for the RX600
20144 series and disabled for the RX200 series.
20146 Floating-point instructions are only generated for 32-bit floating-point
20147 values, however, so the FPU hardware is not used for doubles if the
20148 @option{-m64bit-doubles} option is used.
20150 @emph{Note} If the @option{-fpu} option is enabled then
20151 @option{-funsafe-math-optimizations} is also enabled automatically.
20152 This is because the RX FPU instructions are themselves unsafe.
20154 @item -mcpu=@var{name}
20156 Selects the type of RX CPU to be targeted. Currently three types are
20157 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20158 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20160 The only difference between @samp{RX600} and @samp{RX610} is that the
20161 @samp{RX610} does not support the @code{MVTIPL} instruction.
20163 The @samp{RX200} series does not have a hardware floating-point unit
20164 and so @option{-nofpu} is enabled by default when this type is
20167 @item -mbig-endian-data
20168 @itemx -mlittle-endian-data
20169 @opindex mbig-endian-data
20170 @opindex mlittle-endian-data
20171 Store data (but not code) in the big-endian format. The default is
20172 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20175 @item -msmall-data-limit=@var{N}
20176 @opindex msmall-data-limit
20177 Specifies the maximum size in bytes of global and static variables
20178 which can be placed into the small data area. Using the small data
20179 area can lead to smaller and faster code, but the size of area is
20180 limited and it is up to the programmer to ensure that the area does
20181 not overflow. Also when the small data area is used one of the RX's
20182 registers (usually @code{r13}) is reserved for use pointing to this
20183 area, so it is no longer available for use by the compiler. This
20184 could result in slower and/or larger code if variables are pushed onto
20185 the stack instead of being held in this register.
20187 Note, common variables (variables that have not been initialized) and
20188 constants are not placed into the small data area as they are assigned
20189 to other sections in the output executable.
20191 The default value is zero, which disables this feature. Note, this
20192 feature is not enabled by default with higher optimization levels
20193 (@option{-O2} etc) because of the potentially detrimental effects of
20194 reserving a register. It is up to the programmer to experiment and
20195 discover whether this feature is of benefit to their program. See the
20196 description of the @option{-mpid} option for a description of how the
20197 actual register to hold the small data area pointer is chosen.
20203 Use the simulator runtime. The default is to use the libgloss
20204 board-specific runtime.
20206 @item -mas100-syntax
20207 @itemx -mno-as100-syntax
20208 @opindex mas100-syntax
20209 @opindex mno-as100-syntax
20210 When generating assembler output use a syntax that is compatible with
20211 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20212 assembler, but it has some restrictions so it is not generated by default.
20214 @item -mmax-constant-size=@var{N}
20215 @opindex mmax-constant-size
20216 Specifies the maximum size, in bytes, of a constant that can be used as
20217 an operand in a RX instruction. Although the RX instruction set does
20218 allow constants of up to 4 bytes in length to be used in instructions,
20219 a longer value equates to a longer instruction. Thus in some
20220 circumstances it can be beneficial to restrict the size of constants
20221 that are used in instructions. Constants that are too big are instead
20222 placed into a constant pool and referenced via register indirection.
20224 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20225 or 4 means that constants of any size are allowed.
20229 Enable linker relaxation. Linker relaxation is a process whereby the
20230 linker attempts to reduce the size of a program by finding shorter
20231 versions of various instructions. Disabled by default.
20233 @item -mint-register=@var{N}
20234 @opindex mint-register
20235 Specify the number of registers to reserve for fast interrupt handler
20236 functions. The value @var{N} can be between 0 and 4. A value of 1
20237 means that register @code{r13} is reserved for the exclusive use
20238 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20239 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20240 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20241 A value of 0, the default, does not reserve any registers.
20243 @item -msave-acc-in-interrupts
20244 @opindex msave-acc-in-interrupts
20245 Specifies that interrupt handler functions should preserve the
20246 accumulator register. This is only necessary if normal code might use
20247 the accumulator register, for example because it performs 64-bit
20248 multiplications. The default is to ignore the accumulator as this
20249 makes the interrupt handlers faster.
20255 Enables the generation of position independent data. When enabled any
20256 access to constant data is done via an offset from a base address
20257 held in a register. This allows the location of constant data to be
20258 determined at run time without requiring the executable to be
20259 relocated, which is a benefit to embedded applications with tight
20260 memory constraints. Data that can be modified is not affected by this
20263 Note, using this feature reserves a register, usually @code{r13}, for
20264 the constant data base address. This can result in slower and/or
20265 larger code, especially in complicated functions.
20267 The actual register chosen to hold the constant data base address
20268 depends upon whether the @option{-msmall-data-limit} and/or the
20269 @option{-mint-register} command-line options are enabled. Starting
20270 with register @code{r13} and proceeding downwards, registers are
20271 allocated first to satisfy the requirements of @option{-mint-register},
20272 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20273 is possible for the small data area register to be @code{r8} if both
20274 @option{-mint-register=4} and @option{-mpid} are specified on the
20277 By default this feature is not enabled. The default can be restored
20278 via the @option{-mno-pid} command-line option.
20280 @item -mno-warn-multiple-fast-interrupts
20281 @itemx -mwarn-multiple-fast-interrupts
20282 @opindex mno-warn-multiple-fast-interrupts
20283 @opindex mwarn-multiple-fast-interrupts
20284 Prevents GCC from issuing a warning message if it finds more than one
20285 fast interrupt handler when it is compiling a file. The default is to
20286 issue a warning for each extra fast interrupt handler found, as the RX
20287 only supports one such interrupt.
20289 @item -mallow-string-insns
20290 @itemx -mno-allow-string-insns
20291 @opindex mallow-string-insns
20292 @opindex mno-allow-string-insns
20293 Enables or disables the use of the string manipulation instructions
20294 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
20295 @code{SWHILE} and also the @code{RMPA} instruction. These
20296 instructions may prefetch data, which is not safe to do if accessing
20297 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
20298 for more information).
20300 The default is to allow these instructions, but it is not possible for
20301 GCC to reliably detect all circumstances where a string instruction
20302 might be used to access an I/O register, so their use cannot be
20303 disabled automatically. Instead it is reliant upon the programmer to
20304 use the @option{-mno-allow-string-insns} option if their program
20305 accesses I/O space.
20307 When the instructions are enabled GCC defines the C preprocessor
20308 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
20309 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
20312 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20313 has special significance to the RX port when used with the
20314 @code{interrupt} function attribute. This attribute indicates a
20315 function intended to process fast interrupts. GCC ensures
20316 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20317 and/or @code{r13} and only provided that the normal use of the
20318 corresponding registers have been restricted via the
20319 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20322 @node S/390 and zSeries Options
20323 @subsection S/390 and zSeries Options
20324 @cindex S/390 and zSeries Options
20326 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20330 @itemx -msoft-float
20331 @opindex mhard-float
20332 @opindex msoft-float
20333 Use (do not use) the hardware floating-point instructions and registers
20334 for floating-point operations. When @option{-msoft-float} is specified,
20335 functions in @file{libgcc.a} are used to perform floating-point
20336 operations. When @option{-mhard-float} is specified, the compiler
20337 generates IEEE floating-point instructions. This is the default.
20340 @itemx -mno-hard-dfp
20342 @opindex mno-hard-dfp
20343 Use (do not use) the hardware decimal-floating-point instructions for
20344 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20345 specified, functions in @file{libgcc.a} are used to perform
20346 decimal-floating-point operations. When @option{-mhard-dfp} is
20347 specified, the compiler generates decimal-floating-point hardware
20348 instructions. This is the default for @option{-march=z9-ec} or higher.
20350 @item -mlong-double-64
20351 @itemx -mlong-double-128
20352 @opindex mlong-double-64
20353 @opindex mlong-double-128
20354 These switches control the size of @code{long double} type. A size
20355 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20356 type. This is the default.
20359 @itemx -mno-backchain
20360 @opindex mbackchain
20361 @opindex mno-backchain
20362 Store (do not store) the address of the caller's frame as backchain pointer
20363 into the callee's stack frame.
20364 A backchain may be needed to allow debugging using tools that do not understand
20365 DWARF 2 call frame information.
20366 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20367 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20368 the backchain is placed into the topmost word of the 96/160 byte register
20371 In general, code compiled with @option{-mbackchain} is call-compatible with
20372 code compiled with @option{-mmo-backchain}; however, use of the backchain
20373 for debugging purposes usually requires that the whole binary is built with
20374 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20375 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20376 to build a linux kernel use @option{-msoft-float}.
20378 The default is to not maintain the backchain.
20380 @item -mpacked-stack
20381 @itemx -mno-packed-stack
20382 @opindex mpacked-stack
20383 @opindex mno-packed-stack
20384 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20385 specified, the compiler uses the all fields of the 96/160 byte register save
20386 area only for their default purpose; unused fields still take up stack space.
20387 When @option{-mpacked-stack} is specified, register save slots are densely
20388 packed at the top of the register save area; unused space is reused for other
20389 purposes, allowing for more efficient use of the available stack space.
20390 However, when @option{-mbackchain} is also in effect, the topmost word of
20391 the save area is always used to store the backchain, and the return address
20392 register is always saved two words below the backchain.
20394 As long as the stack frame backchain is not used, code generated with
20395 @option{-mpacked-stack} is call-compatible with code generated with
20396 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20397 S/390 or zSeries generated code that uses the stack frame backchain at run
20398 time, not just for debugging purposes. Such code is not call-compatible
20399 with code compiled with @option{-mpacked-stack}. Also, note that the
20400 combination of @option{-mbackchain},
20401 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20402 to build a linux kernel use @option{-msoft-float}.
20404 The default is to not use the packed stack layout.
20407 @itemx -mno-small-exec
20408 @opindex msmall-exec
20409 @opindex mno-small-exec
20410 Generate (or do not generate) code using the @code{bras} instruction
20411 to do subroutine calls.
20412 This only works reliably if the total executable size does not
20413 exceed 64k. The default is to use the @code{basr} instruction instead,
20414 which does not have this limitation.
20420 When @option{-m31} is specified, generate code compliant to the
20421 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20422 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20423 particular to generate 64-bit instructions. For the @samp{s390}
20424 targets, the default is @option{-m31}, while the @samp{s390x}
20425 targets default to @option{-m64}.
20431 When @option{-mzarch} is specified, generate code using the
20432 instructions available on z/Architecture.
20433 When @option{-mesa} is specified, generate code using the
20434 instructions available on ESA/390. Note that @option{-mesa} is
20435 not possible with @option{-m64}.
20436 When generating code compliant to the GNU/Linux for S/390 ABI,
20437 the default is @option{-mesa}. When generating code compliant
20438 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20444 Generate (or do not generate) code using the @code{mvcle} instruction
20445 to perform block moves. When @option{-mno-mvcle} is specified,
20446 use a @code{mvc} loop instead. This is the default unless optimizing for
20453 Print (or do not print) additional debug information when compiling.
20454 The default is to not print debug information.
20456 @item -march=@var{cpu-type}
20458 Generate code that runs on @var{cpu-type}, which is the name of a system
20459 representing a certain processor type. Possible values for
20460 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20461 @samp{z9-109}, @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12},
20463 When generating code using the instructions available on z/Architecture,
20464 the default is @option{-march=z900}. Otherwise, the default is
20465 @option{-march=g5}.
20467 @item -mtune=@var{cpu-type}
20469 Tune to @var{cpu-type} everything applicable about the generated code,
20470 except for the ABI and the set of available instructions.
20471 The list of @var{cpu-type} values is the same as for @option{-march}.
20472 The default is the value used for @option{-march}.
20475 @itemx -mno-tpf-trace
20476 @opindex mtpf-trace
20477 @opindex mno-tpf-trace
20478 Generate code that adds (does not add) in TPF OS specific branches to trace
20479 routines in the operating system. This option is off by default, even
20480 when compiling for the TPF OS@.
20483 @itemx -mno-fused-madd
20484 @opindex mfused-madd
20485 @opindex mno-fused-madd
20486 Generate code that uses (does not use) the floating-point multiply and
20487 accumulate instructions. These instructions are generated by default if
20488 hardware floating point is used.
20490 @item -mwarn-framesize=@var{framesize}
20491 @opindex mwarn-framesize
20492 Emit a warning if the current function exceeds the given frame size. Because
20493 this is a compile-time check it doesn't need to be a real problem when the program
20494 runs. It is intended to identify functions that most probably cause
20495 a stack overflow. It is useful to be used in an environment with limited stack
20496 size e.g.@: the linux kernel.
20498 @item -mwarn-dynamicstack
20499 @opindex mwarn-dynamicstack
20500 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20501 arrays. This is generally a bad idea with a limited stack size.
20503 @item -mstack-guard=@var{stack-guard}
20504 @itemx -mstack-size=@var{stack-size}
20505 @opindex mstack-guard
20506 @opindex mstack-size
20507 If these options are provided the S/390 back end emits additional instructions in
20508 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20509 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20510 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20511 the frame size of the compiled function is chosen.
20512 These options are intended to be used to help debugging stack overflow problems.
20513 The additionally emitted code causes only little overhead and hence can also be
20514 used in production-like systems without greater performance degradation. The given
20515 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20516 @var{stack-guard} without exceeding 64k.
20517 In order to be efficient the extra code makes the assumption that the stack starts
20518 at an address aligned to the value given by @var{stack-size}.
20519 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20521 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20523 If the hotpatch option is enabled, a ``hot-patching'' function
20524 prologue is generated for all functions in the compilation unit.
20525 The funtion label is prepended with the given number of two-byte
20526 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
20527 the label, 2 * @var{post-halfwords} bytes are appended, using the
20528 largest NOP like instructions the architecture allows (maximum
20531 If both arguments are zero, hotpatching is disabled.
20533 This option can be overridden for individual functions with the
20534 @code{hotpatch} attribute.
20537 @node Score Options
20538 @subsection Score Options
20539 @cindex Score Options
20541 These options are defined for Score implementations:
20546 Compile code for big-endian mode. This is the default.
20550 Compile code for little-endian mode.
20554 Disable generation of @code{bcnz} instructions.
20558 Enable generation of unaligned load and store instructions.
20562 Enable the use of multiply-accumulate instructions. Disabled by default.
20566 Specify the SCORE5 as the target architecture.
20570 Specify the SCORE5U of the target architecture.
20574 Specify the SCORE7 as the target architecture. This is the default.
20578 Specify the SCORE7D as the target architecture.
20582 @subsection SH Options
20584 These @samp{-m} options are defined for the SH implementations:
20589 Generate code for the SH1.
20593 Generate code for the SH2.
20596 Generate code for the SH2e.
20600 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20601 that the floating-point unit is not used.
20603 @item -m2a-single-only
20604 @opindex m2a-single-only
20605 Generate code for the SH2a-FPU, in such a way that no double-precision
20606 floating-point operations are used.
20609 @opindex m2a-single
20610 Generate code for the SH2a-FPU assuming the floating-point unit is in
20611 single-precision mode by default.
20615 Generate code for the SH2a-FPU assuming the floating-point unit is in
20616 double-precision mode by default.
20620 Generate code for the SH3.
20624 Generate code for the SH3e.
20628 Generate code for the SH4 without a floating-point unit.
20630 @item -m4-single-only
20631 @opindex m4-single-only
20632 Generate code for the SH4 with a floating-point unit that only
20633 supports single-precision arithmetic.
20637 Generate code for the SH4 assuming the floating-point unit is in
20638 single-precision mode by default.
20642 Generate code for the SH4.
20646 Generate code for SH4-100.
20648 @item -m4-100-nofpu
20649 @opindex m4-100-nofpu
20650 Generate code for SH4-100 in such a way that the
20651 floating-point unit is not used.
20653 @item -m4-100-single
20654 @opindex m4-100-single
20655 Generate code for SH4-100 assuming the floating-point unit is in
20656 single-precision mode by default.
20658 @item -m4-100-single-only
20659 @opindex m4-100-single-only
20660 Generate code for SH4-100 in such a way that no double-precision
20661 floating-point operations are used.
20665 Generate code for SH4-200.
20667 @item -m4-200-nofpu
20668 @opindex m4-200-nofpu
20669 Generate code for SH4-200 without in such a way that the
20670 floating-point unit is not used.
20672 @item -m4-200-single
20673 @opindex m4-200-single
20674 Generate code for SH4-200 assuming the floating-point unit is in
20675 single-precision mode by default.
20677 @item -m4-200-single-only
20678 @opindex m4-200-single-only
20679 Generate code for SH4-200 in such a way that no double-precision
20680 floating-point operations are used.
20684 Generate code for SH4-300.
20686 @item -m4-300-nofpu
20687 @opindex m4-300-nofpu
20688 Generate code for SH4-300 without in such a way that the
20689 floating-point unit is not used.
20691 @item -m4-300-single
20692 @opindex m4-300-single
20693 Generate code for SH4-300 in such a way that no double-precision
20694 floating-point operations are used.
20696 @item -m4-300-single-only
20697 @opindex m4-300-single-only
20698 Generate code for SH4-300 in such a way that no double-precision
20699 floating-point operations are used.
20703 Generate code for SH4-340 (no MMU, no FPU).
20707 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
20712 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20713 floating-point unit is not used.
20715 @item -m4a-single-only
20716 @opindex m4a-single-only
20717 Generate code for the SH4a, in such a way that no double-precision
20718 floating-point operations are used.
20721 @opindex m4a-single
20722 Generate code for the SH4a assuming the floating-point unit is in
20723 single-precision mode by default.
20727 Generate code for the SH4a.
20731 Same as @option{-m4a-nofpu}, except that it implicitly passes
20732 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20733 instructions at the moment.
20736 @opindex m5-32media
20737 Generate 32-bit code for SHmedia.
20739 @item -m5-32media-nofpu
20740 @opindex m5-32media-nofpu
20741 Generate 32-bit code for SHmedia in such a way that the
20742 floating-point unit is not used.
20745 @opindex m5-64media
20746 Generate 64-bit code for SHmedia.
20748 @item -m5-64media-nofpu
20749 @opindex m5-64media-nofpu
20750 Generate 64-bit code for SHmedia in such a way that the
20751 floating-point unit is not used.
20754 @opindex m5-compact
20755 Generate code for SHcompact.
20757 @item -m5-compact-nofpu
20758 @opindex m5-compact-nofpu
20759 Generate code for SHcompact in such a way that the
20760 floating-point unit is not used.
20764 Compile code for the processor in big-endian mode.
20768 Compile code for the processor in little-endian mode.
20772 Align doubles at 64-bit boundaries. Note that this changes the calling
20773 conventions, and thus some functions from the standard C library do
20774 not work unless you recompile it first with @option{-mdalign}.
20778 Shorten some address references at link time, when possible; uses the
20779 linker option @option{-relax}.
20783 Use 32-bit offsets in @code{switch} tables. The default is to use
20788 Enable the use of bit manipulation instructions on SH2A.
20792 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20793 alignment constraints.
20797 Comply with the calling conventions defined by Renesas.
20800 @opindex mno-renesas
20801 Comply with the calling conventions defined for GCC before the Renesas
20802 conventions were available. This option is the default for all
20803 targets of the SH toolchain.
20806 @opindex mnomacsave
20807 Mark the @code{MAC} register as call-clobbered, even if
20808 @option{-mrenesas} is given.
20814 Control the IEEE compliance of floating-point comparisons, which affects the
20815 handling of cases where the result of a comparison is unordered. By default
20816 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20817 enabled @option{-mno-ieee} is implicitly set, which results in faster
20818 floating-point greater-equal and less-equal comparisons. The implcit settings
20819 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20821 @item -minline-ic_invalidate
20822 @opindex minline-ic_invalidate
20823 Inline code to invalidate instruction cache entries after setting up
20824 nested function trampolines.
20825 This option has no effect if @option{-musermode} is in effect and the selected
20826 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20828 If the selected code generation option does not allow the use of the @code{icbi}
20829 instruction, and @option{-musermode} is not in effect, the inlined code
20830 manipulates the instruction cache address array directly with an associative
20831 write. This not only requires privileged mode at run time, but it also
20832 fails if the cache line had been mapped via the TLB and has become unmapped.
20836 Dump instruction size and location in the assembly code.
20839 @opindex mpadstruct
20840 This option is deprecated. It pads structures to multiple of 4 bytes,
20841 which is incompatible with the SH ABI@.
20843 @item -matomic-model=@var{model}
20844 @opindex matomic-model=@var{model}
20845 Sets the model of atomic operations and additional parameters as a comma
20846 separated list. For details on the atomic built-in functions see
20847 @ref{__atomic Builtins}. The following models and parameters are supported:
20852 Disable compiler generated atomic sequences and emit library calls for atomic
20853 operations. This is the default if the target is not @code{sh*-*-linux*}.
20856 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20857 built-in functions. The generated atomic sequences require additional support
20858 from the interrupt/exception handling code of the system and are only suitable
20859 for SH3* and SH4* single-core systems. This option is enabled by default when
20860 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
20861 this option also partially utilizes the hardware atomic instructions
20862 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20863 @samp{strict} is specified.
20866 Generate software atomic sequences that use a variable in the thread control
20867 block. This is a variation of the gUSA sequences which can also be used on
20868 SH1* and SH2* targets. The generated atomic sequences require additional
20869 support from the interrupt/exception handling code of the system and are only
20870 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20871 parameter has to be specified as well.
20874 Generate software atomic sequences that temporarily disable interrupts by
20875 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20876 in privileged mode and is only suitable for single-core systems. Additional
20877 support from the interrupt/exception handling code of the system is not
20878 required. This model is enabled by default when the target is
20879 @code{sh*-*-linux*} and SH1* or SH2*.
20882 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20883 instructions only. This is only available on SH4A and is suitable for
20884 multi-core systems. Since the hardware instructions support only 32 bit atomic
20885 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20886 Code compiled with this option is also compatible with other software
20887 atomic model interrupt/exception handling systems if executed on an SH4A
20888 system. Additional support from the interrupt/exception handling code of the
20889 system is not required for this model.
20892 This parameter specifies the offset in bytes of the variable in the thread
20893 control block structure that should be used by the generated atomic sequences
20894 when the @samp{soft-tcb} model has been selected. For other models this
20895 parameter is ignored. The specified value must be an integer multiple of four
20896 and in the range 0-1020.
20899 This parameter prevents mixed usage of multiple atomic models, even if they
20900 are compatible, and makes the compiler generate atomic sequences of the
20901 specified model only.
20907 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20908 Notice that depending on the particular hardware and software configuration
20909 this can degrade overall performance due to the operand cache line flushes
20910 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20911 processors the @code{tas.b} instruction must be used with caution since it
20912 can result in data corruption for certain cache configurations.
20915 @opindex mprefergot
20916 When generating position-independent code, emit function calls using
20917 the Global Offset Table instead of the Procedure Linkage Table.
20920 @itemx -mno-usermode
20922 @opindex mno-usermode
20923 Don't allow (allow) the compiler generating privileged mode code. Specifying
20924 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
20925 inlined code would not work in user mode. @option{-musermode} is the default
20926 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
20927 @option{-musermode} has no effect, since there is no user mode.
20929 @item -multcost=@var{number}
20930 @opindex multcost=@var{number}
20931 Set the cost to assume for a multiply insn.
20933 @item -mdiv=@var{strategy}
20934 @opindex mdiv=@var{strategy}
20935 Set the division strategy to be used for integer division operations.
20936 For SHmedia @var{strategy} can be one of:
20941 Performs the operation in floating point. This has a very high latency,
20942 but needs only a few instructions, so it might be a good choice if
20943 your code has enough easily-exploitable ILP to allow the compiler to
20944 schedule the floating-point instructions together with other instructions.
20945 Division by zero causes a floating-point exception.
20948 Uses integer operations to calculate the inverse of the divisor,
20949 and then multiplies the dividend with the inverse. This strategy allows
20950 CSE and hoisting of the inverse calculation. Division by zero calculates
20951 an unspecified result, but does not trap.
20954 A variant of @samp{inv} where, if no CSE or hoisting opportunities
20955 have been found, or if the entire operation has been hoisted to the same
20956 place, the last stages of the inverse calculation are intertwined with the
20957 final multiply to reduce the overall latency, at the expense of using a few
20958 more instructions, and thus offering fewer scheduling opportunities with
20962 Calls a library function that usually implements the @samp{inv:minlat}
20964 This gives high code density for @code{m5-*media-nofpu} compilations.
20967 Uses a different entry point of the same library function, where it
20968 assumes that a pointer to a lookup table has already been set up, which
20969 exposes the pointer load to CSE and code hoisting optimizations.
20974 Use the @samp{inv} algorithm for initial
20975 code generation, but if the code stays unoptimized, revert to the @samp{call},
20976 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
20977 potentially-trapping side effect of division by zero is carried by a
20978 separate instruction, so it is possible that all the integer instructions
20979 are hoisted out, but the marker for the side effect stays where it is.
20980 A recombination to floating-point operations or a call is not possible
20985 Variants of the @samp{inv:minlat} strategy. In the case
20986 that the inverse calculation is not separated from the multiply, they speed
20987 up division where the dividend fits into 20 bits (plus sign where applicable)
20988 by inserting a test to skip a number of operations in this case; this test
20989 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
20990 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
20994 For targets other than SHmedia @var{strategy} can be one of:
20999 Calls a library function that uses the single-step division instruction
21000 @code{div1} to perform the operation. Division by zero calculates an
21001 unspecified result and does not trap. This is the default except for SH4,
21002 SH2A and SHcompact.
21005 Calls a library function that performs the operation in double precision
21006 floating point. Division by zero causes a floating-point exception. This is
21007 the default for SHcompact with FPU. Specifying this for targets that do not
21008 have a double precision FPU defaults to @code{call-div1}.
21011 Calls a library function that uses a lookup table for small divisors and
21012 the @code{div1} instruction with case distinction for larger divisors. Division
21013 by zero calculates an unspecified result and does not trap. This is the default
21014 for SH4. Specifying this for targets that do not have dynamic shift
21015 instructions defaults to @code{call-div1}.
21019 When a division strategy has not been specified the default strategy is
21020 selected based on the current target. For SH2A the default strategy is to
21021 use the @code{divs} and @code{divu} instructions instead of library function
21024 @item -maccumulate-outgoing-args
21025 @opindex maccumulate-outgoing-args
21026 Reserve space once for outgoing arguments in the function prologue rather
21027 than around each call. Generally beneficial for performance and size. Also
21028 needed for unwinding to avoid changing the stack frame around conditional code.
21030 @item -mdivsi3_libfunc=@var{name}
21031 @opindex mdivsi3_libfunc=@var{name}
21032 Set the name of the library function used for 32-bit signed division to
21034 This only affects the name used in the @samp{call} and @samp{inv:call}
21035 division strategies, and the compiler still expects the same
21036 sets of input/output/clobbered registers as if this option were not present.
21038 @item -mfixed-range=@var{register-range}
21039 @opindex mfixed-range
21040 Generate code treating the given register range as fixed registers.
21041 A fixed register is one that the register allocator can not use. This is
21042 useful when compiling kernel code. A register range is specified as
21043 two registers separated by a dash. Multiple register ranges can be
21044 specified separated by a comma.
21046 @item -mindexed-addressing
21047 @opindex mindexed-addressing
21048 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
21049 This is only safe if the hardware and/or OS implement 32-bit wrap-around
21050 semantics for the indexed addressing mode. The architecture allows the
21051 implementation of processors with 64-bit MMU, which the OS could use to
21052 get 32-bit addressing, but since no current hardware implementation supports
21053 this or any other way to make the indexed addressing mode safe to use in
21054 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
21056 @item -mgettrcost=@var{number}
21057 @opindex mgettrcost=@var{number}
21058 Set the cost assumed for the @code{gettr} instruction to @var{number}.
21059 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
21063 Assume @code{pt*} instructions won't trap. This generally generates
21064 better-scheduled code, but is unsafe on current hardware.
21065 The current architecture
21066 definition says that @code{ptabs} and @code{ptrel} trap when the target
21068 This has the unintentional effect of making it unsafe to schedule these
21069 instructions before a branch, or hoist them out of a loop. For example,
21070 @code{__do_global_ctors}, a part of @file{libgcc}
21071 that runs constructors at program
21072 startup, calls functions in a list which is delimited by @minus{}1. With the
21073 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
21074 That means that all the constructors run a bit more quickly, but when
21075 the loop comes to the end of the list, the program crashes because @code{ptabs}
21076 loads @minus{}1 into a target register.
21078 Since this option is unsafe for any
21079 hardware implementing the current architecture specification, the default
21080 is @option{-mno-pt-fixed}. Unless specified explicitly with
21081 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
21082 this deters register allocation from using target registers for storing
21085 @item -minvalid-symbols
21086 @opindex minvalid-symbols
21087 Assume symbols might be invalid. Ordinary function symbols generated by
21088 the compiler are always valid to load with
21089 @code{movi}/@code{shori}/@code{ptabs} or
21090 @code{movi}/@code{shori}/@code{ptrel},
21091 but with assembler and/or linker tricks it is possible
21092 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
21093 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
21094 It prevents cross-basic-block CSE, hoisting and most scheduling
21095 of symbol loads. The default is @option{-mno-invalid-symbols}.
21097 @item -mbranch-cost=@var{num}
21098 @opindex mbranch-cost=@var{num}
21099 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21100 make the compiler try to generate more branch-free code if possible.
21101 If not specified the value is selected depending on the processor type that
21102 is being compiled for.
21105 @itemx -mno-zdcbranch
21106 @opindex mzdcbranch
21107 @opindex mno-zdcbranch
21108 Assume (do not assume) that zero displacement conditional branch instructions
21109 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21110 compiler prefers zero displacement branch code sequences. This is
21111 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21112 disabled by specifying @option{-mno-zdcbranch}.
21114 @item -mcbranch-force-delay-slot
21115 @opindex mcbranch-force-delay-slot
21116 Force the usage of delay slots for conditional branches, which stuffs the delay
21117 slot with a @code{nop} if a suitable instruction can't be found. By default
21118 this option is disabled. It can be enabled to work around hardware bugs as
21119 found in the original SH7055.
21122 @itemx -mno-fused-madd
21123 @opindex mfused-madd
21124 @opindex mno-fused-madd
21125 Generate code that uses (does not use) the floating-point multiply and
21126 accumulate instructions. These instructions are generated by default
21127 if hardware floating point is used. The machine-dependent
21128 @option{-mfused-madd} option is now mapped to the machine-independent
21129 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21130 mapped to @option{-ffp-contract=off}.
21136 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21137 and cosine approximations. The option @option{-mfsca} must be used in
21138 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21139 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21140 approximations even if @option{-funsafe-math-optimizations} is in effect.
21146 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21147 reciprocal square root approximations. The option @option{-mfsrra} must be used
21148 in combination with @option{-funsafe-math-optimizations} and
21149 @option{-ffinite-math-only}. It is enabled by default when generating code for
21150 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21151 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21154 @item -mpretend-cmove
21155 @opindex mpretend-cmove
21156 Prefer zero-displacement conditional branches for conditional move instruction
21157 patterns. This can result in faster code on the SH4 processor.
21161 @node Solaris 2 Options
21162 @subsection Solaris 2 Options
21163 @cindex Solaris 2 options
21165 These @samp{-m} options are supported on Solaris 2:
21168 @item -mclear-hwcap
21169 @opindex mclear-hwcap
21170 @option{-mclear-hwcap} tells the compiler to remove the hardware
21171 capabilities generated by the Solaris assembler. This is only necessary
21172 when object files use ISA extensions not supported by the current
21173 machine, but check at runtime whether or not to use them.
21175 @item -mimpure-text
21176 @opindex mimpure-text
21177 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21178 the compiler to not pass @option{-z text} to the linker when linking a
21179 shared object. Using this option, you can link position-dependent
21180 code into a shared object.
21182 @option{-mimpure-text} suppresses the ``relocations remain against
21183 allocatable but non-writable sections'' linker error message.
21184 However, the necessary relocations trigger copy-on-write, and the
21185 shared object is not actually shared across processes. Instead of
21186 using @option{-mimpure-text}, you should compile all source code with
21187 @option{-fpic} or @option{-fPIC}.
21191 These switches are supported in addition to the above on Solaris 2:
21196 Add support for multithreading using the POSIX threads library. This
21197 option sets flags for both the preprocessor and linker. This option does
21198 not affect the thread safety of object code produced by the compiler or
21199 that of libraries supplied with it.
21203 This is a synonym for @option{-pthreads}.
21206 @node SPARC Options
21207 @subsection SPARC Options
21208 @cindex SPARC options
21210 These @samp{-m} options are supported on the SPARC:
21213 @item -mno-app-regs
21215 @opindex mno-app-regs
21217 Specify @option{-mapp-regs} to generate output using the global registers
21218 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21219 global register 1, each global register 2 through 4 is then treated as an
21220 allocable register that is clobbered by function calls. This is the default.
21222 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21223 specify @option{-mno-app-regs}. You should compile libraries and system
21224 software with this option.
21230 With @option{-mflat}, the compiler does not generate save/restore instructions
21231 and uses a ``flat'' or single register window model. This model is compatible
21232 with the regular register window model. The local registers and the input
21233 registers (0--5) are still treated as ``call-saved'' registers and are
21234 saved on the stack as needed.
21236 With @option{-mno-flat} (the default), the compiler generates save/restore
21237 instructions (except for leaf functions). This is the normal operating mode.
21240 @itemx -mhard-float
21242 @opindex mhard-float
21243 Generate output containing floating-point instructions. This is the
21247 @itemx -msoft-float
21249 @opindex msoft-float
21250 Generate output containing library calls for floating point.
21251 @strong{Warning:} the requisite libraries are not available for all SPARC
21252 targets. Normally the facilities of the machine's usual C compiler are
21253 used, but this cannot be done directly in cross-compilation. You must make
21254 your own arrangements to provide suitable library functions for
21255 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21256 @samp{sparclite-*-*} do provide software floating-point support.
21258 @option{-msoft-float} changes the calling convention in the output file;
21259 therefore, it is only useful if you compile @emph{all} of a program with
21260 this option. In particular, you need to compile @file{libgcc.a}, the
21261 library that comes with GCC, with @option{-msoft-float} in order for
21264 @item -mhard-quad-float
21265 @opindex mhard-quad-float
21266 Generate output containing quad-word (long double) floating-point
21269 @item -msoft-quad-float
21270 @opindex msoft-quad-float
21271 Generate output containing library calls for quad-word (long double)
21272 floating-point instructions. The functions called are those specified
21273 in the SPARC ABI@. This is the default.
21275 As of this writing, there are no SPARC implementations that have hardware
21276 support for the quad-word floating-point instructions. They all invoke
21277 a trap handler for one of these instructions, and then the trap handler
21278 emulates the effect of the instruction. Because of the trap handler overhead,
21279 this is much slower than calling the ABI library routines. Thus the
21280 @option{-msoft-quad-float} option is the default.
21282 @item -mno-unaligned-doubles
21283 @itemx -munaligned-doubles
21284 @opindex mno-unaligned-doubles
21285 @opindex munaligned-doubles
21286 Assume that doubles have 8-byte alignment. This is the default.
21288 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21289 alignment only if they are contained in another type, or if they have an
21290 absolute address. Otherwise, it assumes they have 4-byte alignment.
21291 Specifying this option avoids some rare compatibility problems with code
21292 generated by other compilers. It is not the default because it results
21293 in a performance loss, especially for floating-point code.
21296 @itemx -mno-user-mode
21297 @opindex muser-mode
21298 @opindex mno-user-mode
21299 Do not generate code that can only run in supervisor mode. This is relevant
21300 only for the @code{casa} instruction emitted for the LEON3 processor. The
21301 default is @option{-mno-user-mode}.
21303 @item -mno-faster-structs
21304 @itemx -mfaster-structs
21305 @opindex mno-faster-structs
21306 @opindex mfaster-structs
21307 With @option{-mfaster-structs}, the compiler assumes that structures
21308 should have 8-byte alignment. This enables the use of pairs of
21309 @code{ldd} and @code{std} instructions for copies in structure
21310 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21311 However, the use of this changed alignment directly violates the SPARC
21312 ABI@. Thus, it's intended only for use on targets where the developer
21313 acknowledges that their resulting code is not directly in line with
21314 the rules of the ABI@.
21316 @item -mcpu=@var{cpu_type}
21318 Set the instruction set, register set, and instruction scheduling parameters
21319 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21320 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21321 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21322 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21323 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21324 @samp{niagara3} and @samp{niagara4}.
21326 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21327 which selects the best architecture option for the host processor.
21328 @option{-mcpu=native} has no effect if GCC does not recognize
21331 Default instruction scheduling parameters are used for values that select
21332 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21333 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21335 Here is a list of each supported architecture and their supported
21343 supersparc, hypersparc, leon, leon3
21346 f930, f934, sparclite86x
21352 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21355 By default (unless configured otherwise), GCC generates code for the V7
21356 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21357 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21358 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21359 SPARCStation 1, 2, IPX etc.
21361 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21362 architecture. The only difference from V7 code is that the compiler emits
21363 the integer multiply and integer divide instructions which exist in SPARC-V8
21364 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21365 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21368 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21369 the SPARC architecture. This adds the integer multiply, integer divide step
21370 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21371 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21372 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21373 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21374 MB86934 chip, which is the more recent SPARClite with FPU@.
21376 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21377 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21378 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21379 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21380 optimizes it for the TEMIC SPARClet chip.
21382 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21383 architecture. This adds 64-bit integer and floating-point move instructions,
21384 3 additional floating-point condition code registers and conditional move
21385 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21386 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21387 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21388 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21389 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21390 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21391 additionally optimizes it for Sun UltraSPARC T2 chips. With
21392 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21393 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21394 additionally optimizes it for Sun UltraSPARC T4 chips.
21396 @item -mtune=@var{cpu_type}
21398 Set the instruction scheduling parameters for machine type
21399 @var{cpu_type}, but do not set the instruction set or register set that the
21400 option @option{-mcpu=@var{cpu_type}} does.
21402 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21403 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21404 that select a particular CPU implementation. Those are @samp{cypress},
21405 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21406 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21407 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21408 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21409 toolchains, @samp{native} can also be used.
21414 @opindex mno-v8plus
21415 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21416 difference from the V8 ABI is that the global and out registers are
21417 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21418 mode for all SPARC-V9 processors.
21424 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21425 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21431 With @option{-mvis2}, GCC generates code that takes advantage of
21432 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21433 default is @option{-mvis2} when targeting a cpu that supports such
21434 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21435 also sets @option{-mvis}.
21441 With @option{-mvis3}, GCC generates code that takes advantage of
21442 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21443 default is @option{-mvis3} when targeting a cpu that supports such
21444 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21445 also sets @option{-mvis2} and @option{-mvis}.
21450 @opindex mno-cbcond
21451 With @option{-mcbcond}, GCC generates code that takes advantage of
21452 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21453 The default is @option{-mcbcond} when targeting a cpu that supports such
21454 instructions, such as niagara-4 and later.
21460 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21461 population count instruction. The default is @option{-mpopc}
21462 when targeting a cpu that supports such instructions, such as Niagara-2 and
21469 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21470 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21471 when targeting a cpu that supports such instructions, such as Niagara-3 and
21475 @opindex mfix-at697f
21476 Enable the documented workaround for the single erratum of the Atmel AT697F
21477 processor (which corresponds to erratum #13 of the AT697E processor).
21480 @opindex mfix-ut699
21481 Enable the documented workarounds for the floating-point errata and the data
21482 cache nullify errata of the UT699 processor.
21485 These @samp{-m} options are supported in addition to the above
21486 on SPARC-V9 processors in 64-bit environments:
21493 Generate code for a 32-bit or 64-bit environment.
21494 The 32-bit environment sets int, long and pointer to 32 bits.
21495 The 64-bit environment sets int to 32 bits and long and pointer
21498 @item -mcmodel=@var{which}
21500 Set the code model to one of
21504 The Medium/Low code model: 64-bit addresses, programs
21505 must be linked in the low 32 bits of memory. Programs can be statically
21506 or dynamically linked.
21509 The Medium/Middle code model: 64-bit addresses, programs
21510 must be linked in the low 44 bits of memory, the text and data segments must
21511 be less than 2GB in size and the data segment must be located within 2GB of
21515 The Medium/Anywhere code model: 64-bit addresses, programs
21516 may be linked anywhere in memory, the text and data segments must be less
21517 than 2GB in size and the data segment must be located within 2GB of the
21521 The Medium/Anywhere code model for embedded systems:
21522 64-bit addresses, the text and data segments must be less than 2GB in
21523 size, both starting anywhere in memory (determined at link time). The
21524 global register %g4 points to the base of the data segment. Programs
21525 are statically linked and PIC is not supported.
21528 @item -mmemory-model=@var{mem-model}
21529 @opindex mmemory-model
21530 Set the memory model in force on the processor to one of
21534 The default memory model for the processor and operating system.
21537 Relaxed Memory Order
21540 Partial Store Order
21546 Sequential Consistency
21549 These memory models are formally defined in Appendix D of the Sparc V9
21550 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21553 @itemx -mno-stack-bias
21554 @opindex mstack-bias
21555 @opindex mno-stack-bias
21556 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21557 frame pointer if present, are offset by @minus{}2047 which must be added back
21558 when making stack frame references. This is the default in 64-bit mode.
21559 Otherwise, assume no such offset is present.
21563 @subsection SPU Options
21564 @cindex SPU options
21566 These @samp{-m} options are supported on the SPU:
21570 @itemx -merror-reloc
21571 @opindex mwarn-reloc
21572 @opindex merror-reloc
21574 The loader for SPU does not handle dynamic relocations. By default, GCC
21575 gives an error when it generates code that requires a dynamic
21576 relocation. @option{-mno-error-reloc} disables the error,
21577 @option{-mwarn-reloc} generates a warning instead.
21580 @itemx -munsafe-dma
21582 @opindex munsafe-dma
21584 Instructions that initiate or test completion of DMA must not be
21585 reordered with respect to loads and stores of the memory that is being
21587 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21588 memory accesses, but that can lead to inefficient code in places where the
21589 memory is known to not change. Rather than mark the memory as volatile,
21590 you can use @option{-msafe-dma} to tell the compiler to treat
21591 the DMA instructions as potentially affecting all memory.
21593 @item -mbranch-hints
21594 @opindex mbranch-hints
21596 By default, GCC generates a branch hint instruction to avoid
21597 pipeline stalls for always-taken or probably-taken branches. A hint
21598 is not generated closer than 8 instructions away from its branch.
21599 There is little reason to disable them, except for debugging purposes,
21600 or to make an object a little bit smaller.
21604 @opindex msmall-mem
21605 @opindex mlarge-mem
21607 By default, GCC generates code assuming that addresses are never larger
21608 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21609 a full 32-bit address.
21614 By default, GCC links against startup code that assumes the SPU-style
21615 main function interface (which has an unconventional parameter list).
21616 With @option{-mstdmain}, GCC links your program against startup
21617 code that assumes a C99-style interface to @code{main}, including a
21618 local copy of @code{argv} strings.
21620 @item -mfixed-range=@var{register-range}
21621 @opindex mfixed-range
21622 Generate code treating the given register range as fixed registers.
21623 A fixed register is one that the register allocator cannot use. This is
21624 useful when compiling kernel code. A register range is specified as
21625 two registers separated by a dash. Multiple register ranges can be
21626 specified separated by a comma.
21632 Compile code assuming that pointers to the PPU address space accessed
21633 via the @code{__ea} named address space qualifier are either 32 or 64
21634 bits wide. The default is 32 bits. As this is an ABI-changing option,
21635 all object code in an executable must be compiled with the same setting.
21637 @item -maddress-space-conversion
21638 @itemx -mno-address-space-conversion
21639 @opindex maddress-space-conversion
21640 @opindex mno-address-space-conversion
21641 Allow/disallow treating the @code{__ea} address space as superset
21642 of the generic address space. This enables explicit type casts
21643 between @code{__ea} and generic pointer as well as implicit
21644 conversions of generic pointers to @code{__ea} pointers. The
21645 default is to allow address space pointer conversions.
21647 @item -mcache-size=@var{cache-size}
21648 @opindex mcache-size
21649 This option controls the version of libgcc that the compiler links to an
21650 executable and selects a software-managed cache for accessing variables
21651 in the @code{__ea} address space with a particular cache size. Possible
21652 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21653 and @samp{128}. The default cache size is 64KB.
21655 @item -matomic-updates
21656 @itemx -mno-atomic-updates
21657 @opindex matomic-updates
21658 @opindex mno-atomic-updates
21659 This option controls the version of libgcc that the compiler links to an
21660 executable and selects whether atomic updates to the software-managed
21661 cache of PPU-side variables are used. If you use atomic updates, changes
21662 to a PPU variable from SPU code using the @code{__ea} named address space
21663 qualifier do not interfere with changes to other PPU variables residing
21664 in the same cache line from PPU code. If you do not use atomic updates,
21665 such interference may occur; however, writing back cache lines is
21666 more efficient. The default behavior is to use atomic updates.
21669 @itemx -mdual-nops=@var{n}
21670 @opindex mdual-nops
21671 By default, GCC inserts nops to increase dual issue when it expects
21672 it to increase performance. @var{n} can be a value from 0 to 10. A
21673 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21674 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21676 @item -mhint-max-nops=@var{n}
21677 @opindex mhint-max-nops
21678 Maximum number of nops to insert for a branch hint. A branch hint must
21679 be at least 8 instructions away from the branch it is affecting. GCC
21680 inserts up to @var{n} nops to enforce this, otherwise it does not
21681 generate the branch hint.
21683 @item -mhint-max-distance=@var{n}
21684 @opindex mhint-max-distance
21685 The encoding of the branch hint instruction limits the hint to be within
21686 256 instructions of the branch it is affecting. By default, GCC makes
21687 sure it is within 125.
21690 @opindex msafe-hints
21691 Work around a hardware bug that causes the SPU to stall indefinitely.
21692 By default, GCC inserts the @code{hbrp} instruction to make sure
21693 this stall won't happen.
21697 @node System V Options
21698 @subsection Options for System V
21700 These additional options are available on System V Release 4 for
21701 compatibility with other compilers on those systems:
21706 Create a shared object.
21707 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21711 Identify the versions of each tool used by the compiler, in a
21712 @code{.ident} assembler directive in the output.
21716 Refrain from adding @code{.ident} directives to the output file (this is
21719 @item -YP,@var{dirs}
21721 Search the directories @var{dirs}, and no others, for libraries
21722 specified with @option{-l}.
21724 @item -Ym,@var{dir}
21726 Look in the directory @var{dir} to find the M4 preprocessor.
21727 The assembler uses this option.
21728 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21729 @c the generic assembler that comes with Solaris takes just -Ym.
21732 @node TILE-Gx Options
21733 @subsection TILE-Gx Options
21734 @cindex TILE-Gx options
21736 These @samp{-m} options are supported on the TILE-Gx:
21739 @item -mcmodel=small
21740 @opindex mcmodel=small
21741 Generate code for the small model. The distance for direct calls is
21742 limited to 500M in either direction. PC-relative addresses are 32
21743 bits. Absolute addresses support the full address range.
21745 @item -mcmodel=large
21746 @opindex mcmodel=large
21747 Generate code for the large model. There is no limitation on call
21748 distance, pc-relative addresses, or absolute addresses.
21750 @item -mcpu=@var{name}
21752 Selects the type of CPU to be targeted. Currently the only supported
21753 type is @samp{tilegx}.
21759 Generate code for a 32-bit or 64-bit environment. The 32-bit
21760 environment sets int, long, and pointer to 32 bits. The 64-bit
21761 environment sets int to 32 bits and long and pointer to 64 bits.
21764 @itemx -mlittle-endian
21765 @opindex mbig-endian
21766 @opindex mlittle-endian
21767 Generate code in big/little endian mode, respectively.
21770 @node TILEPro Options
21771 @subsection TILEPro Options
21772 @cindex TILEPro options
21774 These @samp{-m} options are supported on the TILEPro:
21777 @item -mcpu=@var{name}
21779 Selects the type of CPU to be targeted. Currently the only supported
21780 type is @samp{tilepro}.
21784 Generate code for a 32-bit environment, which sets int, long, and
21785 pointer to 32 bits. This is the only supported behavior so the flag
21786 is essentially ignored.
21790 @subsection V850 Options
21791 @cindex V850 Options
21793 These @samp{-m} options are defined for V850 implementations:
21797 @itemx -mno-long-calls
21798 @opindex mlong-calls
21799 @opindex mno-long-calls
21800 Treat all calls as being far away (near). If calls are assumed to be
21801 far away, the compiler always loads the function's address into a
21802 register, and calls indirect through the pointer.
21808 Do not optimize (do optimize) basic blocks that use the same index
21809 pointer 4 or more times to copy pointer into the @code{ep} register, and
21810 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21811 option is on by default if you optimize.
21813 @item -mno-prolog-function
21814 @itemx -mprolog-function
21815 @opindex mno-prolog-function
21816 @opindex mprolog-function
21817 Do not use (do use) external functions to save and restore registers
21818 at the prologue and epilogue of a function. The external functions
21819 are slower, but use less code space if more than one function saves
21820 the same number of registers. The @option{-mprolog-function} option
21821 is on by default if you optimize.
21825 Try to make the code as small as possible. At present, this just turns
21826 on the @option{-mep} and @option{-mprolog-function} options.
21828 @item -mtda=@var{n}
21830 Put static or global variables whose size is @var{n} bytes or less into
21831 the tiny data area that register @code{ep} points to. The tiny data
21832 area can hold up to 256 bytes in total (128 bytes for byte references).
21834 @item -msda=@var{n}
21836 Put static or global variables whose size is @var{n} bytes or less into
21837 the small data area that register @code{gp} points to. The small data
21838 area can hold up to 64 kilobytes.
21840 @item -mzda=@var{n}
21842 Put static or global variables whose size is @var{n} bytes or less into
21843 the first 32 kilobytes of memory.
21847 Specify that the target processor is the V850.
21851 Specify that the target processor is the V850E3V5. The preprocessor
21852 constant @code{__v850e3v5__} is defined if this option is used.
21856 Specify that the target processor is the V850E3V5. This is an alias for
21857 the @option{-mv850e3v5} option.
21861 Specify that the target processor is the V850E2V3. The preprocessor
21862 constant @code{__v850e2v3__} is defined if this option is used.
21866 Specify that the target processor is the V850E2. The preprocessor
21867 constant @code{__v850e2__} is defined if this option is used.
21871 Specify that the target processor is the V850E1. The preprocessor
21872 constants @code{__v850e1__} and @code{__v850e__} are defined if
21873 this option is used.
21877 Specify that the target processor is the V850ES. This is an alias for
21878 the @option{-mv850e1} option.
21882 Specify that the target processor is the V850E@. The preprocessor
21883 constant @code{__v850e__} is defined if this option is used.
21885 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21886 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21887 are defined then a default target processor is chosen and the
21888 relevant @samp{__v850*__} preprocessor constant is defined.
21890 The preprocessor constants @code{__v850} and @code{__v851__} are always
21891 defined, regardless of which processor variant is the target.
21893 @item -mdisable-callt
21894 @itemx -mno-disable-callt
21895 @opindex mdisable-callt
21896 @opindex mno-disable-callt
21897 This option suppresses generation of the @code{CALLT} instruction for the
21898 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21901 This option is enabled by default when the RH850 ABI is
21902 in use (see @option{-mrh850-abi}), and disabled by default when the
21903 GCC ABI is in use. If @code{CALLT} instructions are being generated
21904 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
21910 Pass on (or do not pass on) the @option{-mrelax} command-line option
21914 @itemx -mno-long-jumps
21915 @opindex mlong-jumps
21916 @opindex mno-long-jumps
21917 Disable (or re-enable) the generation of PC-relative jump instructions.
21920 @itemx -mhard-float
21921 @opindex msoft-float
21922 @opindex mhard-float
21923 Disable (or re-enable) the generation of hardware floating point
21924 instructions. This option is only significant when the target
21925 architecture is @samp{V850E2V3} or higher. If hardware floating point
21926 instructions are being generated then the C preprocessor symbol
21927 @code{__FPU_OK__} is defined, otherwise the symbol
21928 @code{__NO_FPU__} is defined.
21932 Enables the use of the e3v5 LOOP instruction. The use of this
21933 instruction is not enabled by default when the e3v5 architecture is
21934 selected because its use is still experimental.
21938 @opindex mrh850-abi
21940 Enables support for the RH850 version of the V850 ABI. This is the
21941 default. With this version of the ABI the following rules apply:
21945 Integer sized structures and unions are returned via a memory pointer
21946 rather than a register.
21949 Large structures and unions (more than 8 bytes in size) are passed by
21953 Functions are aligned to 16-bit boundaries.
21956 The @option{-m8byte-align} command-line option is supported.
21959 The @option{-mdisable-callt} command-line option is enabled by
21960 default. The @option{-mno-disable-callt} command-line option is not
21964 When this version of the ABI is enabled the C preprocessor symbol
21965 @code{__V850_RH850_ABI__} is defined.
21969 Enables support for the old GCC version of the V850 ABI. With this
21970 version of the ABI the following rules apply:
21974 Integer sized structures and unions are returned in register @code{r10}.
21977 Large structures and unions (more than 8 bytes in size) are passed by
21981 Functions are aligned to 32-bit boundaries, unless optimizing for
21985 The @option{-m8byte-align} command-line option is not supported.
21988 The @option{-mdisable-callt} command-line option is supported but not
21989 enabled by default.
21992 When this version of the ABI is enabled the C preprocessor symbol
21993 @code{__V850_GCC_ABI__} is defined.
21995 @item -m8byte-align
21996 @itemx -mno-8byte-align
21997 @opindex m8byte-align
21998 @opindex mno-8byte-align
21999 Enables support for @code{double} and @code{long long} types to be
22000 aligned on 8-byte boundaries. The default is to restrict the
22001 alignment of all objects to at most 4-bytes. When
22002 @option{-m8byte-align} is in effect the C preprocessor symbol
22003 @code{__V850_8BYTE_ALIGN__} is defined.
22006 @opindex mbig-switch
22007 Generate code suitable for big switch tables. Use this option only if
22008 the assembler/linker complain about out of range branches within a switch
22013 This option causes r2 and r5 to be used in the code generated by
22014 the compiler. This setting is the default.
22016 @item -mno-app-regs
22017 @opindex mno-app-regs
22018 This option causes r2 and r5 to be treated as fixed registers.
22023 @subsection VAX Options
22024 @cindex VAX options
22026 These @samp{-m} options are defined for the VAX:
22031 Do not output certain jump instructions (@code{aobleq} and so on)
22032 that the Unix assembler for the VAX cannot handle across long
22037 Do output those jump instructions, on the assumption that the
22038 GNU assembler is being used.
22042 Output code for G-format floating-point numbers instead of D-format.
22045 @node Visium Options
22046 @subsection Visium Options
22047 @cindex Visium options
22053 A program which performs file I/O and is destined to run on an MCM target
22054 should be linked with this option. It causes the libraries libc.a and
22055 libdebug.a to be linked. The program should be run on the target under
22056 the control of the GDB remote debugging stub.
22060 A program which performs file I/O and is destined to run on the simulator
22061 should be linked with option. This causes libraries libc.a and libsim.a to
22065 @itemx -mhard-float
22067 @opindex mhard-float
22068 Generate code containing floating-point instructions. This is the
22072 @itemx -msoft-float
22074 @opindex msoft-float
22075 Generate code containing library calls for floating-point.
22077 @option{-msoft-float} changes the calling convention in the output file;
22078 therefore, it is only useful if you compile @emph{all} of a program with
22079 this option. In particular, you need to compile @file{libgcc.a}, the
22080 library that comes with GCC, with @option{-msoft-float} in order for
22083 @item -mcpu=@var{cpu_type}
22085 Set the instruction set, register set, and instruction scheduling parameters
22086 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22087 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22089 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22091 By default (unless configured otherwise), GCC generates code for the GR5
22092 variant of the Visium architecture.
22094 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22095 architecture. The only difference from GR5 code is that the compiler will
22096 generate block move instructions.
22098 @item -mtune=@var{cpu_type}
22100 Set the instruction scheduling parameters for machine type @var{cpu_type},
22101 but do not set the instruction set or register set that the option
22102 @option{-mcpu=@var{cpu_type}} would.
22106 Generate code for the supervisor mode, where there are no restrictions on
22107 the access to general registers. This is the default.
22110 @opindex muser-mode
22111 Generate code for the user mode, where the access to some general registers
22112 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22113 mode; on the GR6, only registers r29 to r31 are affected.
22117 @subsection VMS Options
22119 These @samp{-m} options are defined for the VMS implementations:
22122 @item -mvms-return-codes
22123 @opindex mvms-return-codes
22124 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22125 condition (e.g.@ error) codes.
22127 @item -mdebug-main=@var{prefix}
22128 @opindex mdebug-main=@var{prefix}
22129 Flag the first routine whose name starts with @var{prefix} as the main
22130 routine for the debugger.
22134 Default to 64-bit memory allocation routines.
22136 @item -mpointer-size=@var{size}
22137 @opindex mpointer-size=@var{size}
22138 Set the default size of pointers. Possible options for @var{size} are
22139 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22140 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22141 The later option disables @code{pragma pointer_size}.
22144 @node VxWorks Options
22145 @subsection VxWorks Options
22146 @cindex VxWorks Options
22148 The options in this section are defined for all VxWorks targets.
22149 Options specific to the target hardware are listed with the other
22150 options for that target.
22155 GCC can generate code for both VxWorks kernels and real time processes
22156 (RTPs). This option switches from the former to the latter. It also
22157 defines the preprocessor macro @code{__RTP__}.
22160 @opindex non-static
22161 Link an RTP executable against shared libraries rather than static
22162 libraries. The options @option{-static} and @option{-shared} can
22163 also be used for RTPs (@pxref{Link Options}); @option{-static}
22170 These options are passed down to the linker. They are defined for
22171 compatibility with Diab.
22174 @opindex Xbind-lazy
22175 Enable lazy binding of function calls. This option is equivalent to
22176 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22180 Disable lazy binding of function calls. This option is the default and
22181 is defined for compatibility with Diab.
22185 @subsection x86 Options
22186 @cindex x86 Options
22188 These @samp{-m} options are defined for the x86 family of computers.
22192 @item -march=@var{cpu-type}
22194 Generate instructions for the machine type @var{cpu-type}. In contrast to
22195 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
22196 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
22197 to generate code that may not run at all on processors other than the one
22198 indicated. Specifying @option{-march=@var{cpu-type}} implies
22199 @option{-mtune=@var{cpu-type}}.
22201 The choices for @var{cpu-type} are:
22205 This selects the CPU to generate code for at compilation time by determining
22206 the processor type of the compiling machine. Using @option{-march=native}
22207 enables all instruction subsets supported by the local machine (hence
22208 the result might not run on different machines). Using @option{-mtune=native}
22209 produces code optimized for the local machine under the constraints
22210 of the selected instruction set.
22213 Original Intel i386 CPU@.
22216 Intel i486 CPU@. (No scheduling is implemented for this chip.)
22220 Intel Pentium CPU with no MMX support.
22223 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
22226 Intel Pentium Pro CPU@.
22229 When used with @option{-march}, the Pentium Pro
22230 instruction set is used, so the code runs on all i686 family chips.
22231 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
22234 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
22239 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
22243 Intel Pentium M; low-power version of Intel Pentium III CPU
22244 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
22248 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
22251 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
22255 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
22256 SSE2 and SSE3 instruction set support.
22259 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
22260 instruction set support.
22263 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22264 SSE4.1, SSE4.2 and POPCNT instruction set support.
22267 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22268 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
22271 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22272 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
22275 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22276 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
22277 instruction set support.
22280 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22281 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22282 BMI, BMI2 and F16C instruction set support.
22285 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22286 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22287 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
22290 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
22291 instruction set support.
22294 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22295 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
22298 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22299 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22300 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
22301 AVX512CD instruction set support.
22304 AMD K6 CPU with MMX instruction set support.
22308 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
22311 @itemx athlon-tbird
22312 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
22318 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
22319 instruction set support.
22325 Processors based on the AMD K8 core with x86-64 instruction set support,
22326 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
22327 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
22328 instruction set extensions.)
22331 @itemx opteron-sse3
22332 @itemx athlon64-sse3
22333 Improved versions of AMD K8 cores with SSE3 instruction set support.
22337 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
22338 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
22339 instruction set extensions.)
22342 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
22343 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
22344 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
22346 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22347 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
22348 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
22351 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22352 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
22353 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
22354 64-bit instruction set extensions.
22356 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22357 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
22358 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
22359 SSE4.2, ABM and 64-bit instruction set extensions.
22362 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
22363 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
22364 instruction set extensions.)
22367 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
22368 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
22369 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
22372 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
22376 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
22377 instruction set support.
22380 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
22381 implemented for this chip.)
22384 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
22386 implemented for this chip.)
22389 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
22392 @item -mtune=@var{cpu-type}
22394 Tune to @var{cpu-type} everything applicable about the generated code, except
22395 for the ABI and the set of available instructions.
22396 While picking a specific @var{cpu-type} schedules things appropriately
22397 for that particular chip, the compiler does not generate any code that
22398 cannot run on the default machine type unless you use a
22399 @option{-march=@var{cpu-type}} option.
22400 For example, if GCC is configured for i686-pc-linux-gnu
22401 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
22402 but still runs on i686 machines.
22404 The choices for @var{cpu-type} are the same as for @option{-march}.
22405 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
22409 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
22410 If you know the CPU on which your code will run, then you should use
22411 the corresponding @option{-mtune} or @option{-march} option instead of
22412 @option{-mtune=generic}. But, if you do not know exactly what CPU users
22413 of your application will have, then you should use this option.
22415 As new processors are deployed in the marketplace, the behavior of this
22416 option will change. Therefore, if you upgrade to a newer version of
22417 GCC, code generation controlled by this option will change to reflect
22419 that are most common at the time that version of GCC is released.
22421 There is no @option{-march=generic} option because @option{-march}
22422 indicates the instruction set the compiler can use, and there is no
22423 generic instruction set applicable to all processors. In contrast,
22424 @option{-mtune} indicates the processor (or, in this case, collection of
22425 processors) for which the code is optimized.
22428 Produce code optimized for the most current Intel processors, which are
22429 Haswell and Silvermont for this version of GCC. If you know the CPU
22430 on which your code will run, then you should use the corresponding
22431 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
22432 But, if you want your application performs better on both Haswell and
22433 Silvermont, then you should use this option.
22435 As new Intel processors are deployed in the marketplace, the behavior of
22436 this option will change. Therefore, if you upgrade to a newer version of
22437 GCC, code generation controlled by this option will change to reflect
22438 the most current Intel processors at the time that version of GCC is
22441 There is no @option{-march=intel} option because @option{-march} indicates
22442 the instruction set the compiler can use, and there is no common
22443 instruction set applicable to all processors. In contrast,
22444 @option{-mtune} indicates the processor (or, in this case, collection of
22445 processors) for which the code is optimized.
22448 @item -mcpu=@var{cpu-type}
22450 A deprecated synonym for @option{-mtune}.
22452 @item -mfpmath=@var{unit}
22454 Generate floating-point arithmetic for selected unit @var{unit}. The choices
22455 for @var{unit} are:
22459 Use the standard 387 floating-point coprocessor present on the majority of chips and
22460 emulated otherwise. Code compiled with this option runs almost everywhere.
22461 The temporary results are computed in 80-bit precision instead of the precision
22462 specified by the type, resulting in slightly different results compared to most
22463 of other chips. See @option{-ffloat-store} for more detailed description.
22465 This is the default choice for x86-32 targets.
22468 Use scalar floating-point instructions present in the SSE instruction set.
22469 This instruction set is supported by Pentium III and newer chips,
22470 and in the AMD line
22471 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
22472 instruction set supports only single-precision arithmetic, thus the double and
22473 extended-precision arithmetic are still done using 387. A later version, present
22474 only in Pentium 4 and AMD x86-64 chips, supports double-precision
22477 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
22478 or @option{-msse2} switches to enable SSE extensions and make this option
22479 effective. For the x86-64 compiler, these extensions are enabled by default.
22481 The resulting code should be considerably faster in the majority of cases and avoid
22482 the numerical instability problems of 387 code, but may break some existing
22483 code that expects temporaries to be 80 bits.
22485 This is the default choice for the x86-64 compiler.
22490 Attempt to utilize both instruction sets at once. This effectively doubles the
22491 amount of available registers, and on chips with separate execution units for
22492 387 and SSE the execution resources too. Use this option with care, as it is
22493 still experimental, because the GCC register allocator does not model separate
22494 functional units well, resulting in unstable performance.
22497 @item -masm=@var{dialect}
22498 @opindex masm=@var{dialect}
22499 Output assembly instructions using selected @var{dialect}. Also affects
22500 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22501 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22502 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22503 not support @samp{intel}.
22506 @itemx -mno-ieee-fp
22508 @opindex mno-ieee-fp
22509 Control whether or not the compiler uses IEEE floating-point
22510 comparisons. These correctly handle the case where the result of a
22511 comparison is unordered.
22514 @opindex msoft-float
22515 Generate output containing library calls for floating point.
22517 @strong{Warning:} the requisite libraries are not part of GCC@.
22518 Normally the facilities of the machine's usual C compiler are used, but
22519 this can't be done directly in cross-compilation. You must make your
22520 own arrangements to provide suitable library functions for
22523 On machines where a function returns floating-point results in the 80387
22524 register stack, some floating-point opcodes may be emitted even if
22525 @option{-msoft-float} is used.
22527 @item -mno-fp-ret-in-387
22528 @opindex mno-fp-ret-in-387
22529 Do not use the FPU registers for return values of functions.
22531 The usual calling convention has functions return values of types
22532 @code{float} and @code{double} in an FPU register, even if there
22533 is no FPU@. The idea is that the operating system should emulate
22536 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22537 in ordinary CPU registers instead.
22539 @item -mno-fancy-math-387
22540 @opindex mno-fancy-math-387
22541 Some 387 emulators do not support the @code{sin}, @code{cos} and
22542 @code{sqrt} instructions for the 387. Specify this option to avoid
22543 generating those instructions. This option is the default on FreeBSD,
22544 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22545 indicates that the target CPU always has an FPU and so the
22546 instruction does not need emulation. These
22547 instructions are not generated unless you also use the
22548 @option{-funsafe-math-optimizations} switch.
22550 @item -malign-double
22551 @itemx -mno-align-double
22552 @opindex malign-double
22553 @opindex mno-align-double
22554 Control whether GCC aligns @code{double}, @code{long double}, and
22555 @code{long long} variables on a two-word boundary or a one-word
22556 boundary. Aligning @code{double} variables on a two-word boundary
22557 produces code that runs somewhat faster on a Pentium at the
22558 expense of more memory.
22560 On x86-64, @option{-malign-double} is enabled by default.
22562 @strong{Warning:} if you use the @option{-malign-double} switch,
22563 structures containing the above types are aligned differently than
22564 the published application binary interface specifications for the x86-32
22565 and are not binary compatible with structures in code compiled
22566 without that switch.
22568 @item -m96bit-long-double
22569 @itemx -m128bit-long-double
22570 @opindex m96bit-long-double
22571 @opindex m128bit-long-double
22572 These switches control the size of @code{long double} type. The x86-32
22573 application binary interface specifies the size to be 96 bits,
22574 so @option{-m96bit-long-double} is the default in 32-bit mode.
22576 Modern architectures (Pentium and newer) prefer @code{long double}
22577 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22578 conforming to the ABI, this is not possible. So specifying
22579 @option{-m128bit-long-double} aligns @code{long double}
22580 to a 16-byte boundary by padding the @code{long double} with an additional
22583 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22584 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22586 Notice that neither of these options enable any extra precision over the x87
22587 standard of 80 bits for a @code{long double}.
22589 @strong{Warning:} if you override the default value for your target ABI, this
22590 changes the size of
22591 structures and arrays containing @code{long double} variables,
22592 as well as modifying the function calling convention for functions taking
22593 @code{long double}. Hence they are not binary-compatible
22594 with code compiled without that switch.
22596 @item -mlong-double-64
22597 @itemx -mlong-double-80
22598 @itemx -mlong-double-128
22599 @opindex mlong-double-64
22600 @opindex mlong-double-80
22601 @opindex mlong-double-128
22602 These switches control the size of @code{long double} type. A size
22603 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22604 type. This is the default for 32-bit Bionic C library. A size
22605 of 128 bits makes the @code{long double} type equivalent to the
22606 @code{__float128} type. This is the default for 64-bit Bionic C library.
22608 @strong{Warning:} if you override the default value for your target ABI, this
22609 changes the size of
22610 structures and arrays containing @code{long double} variables,
22611 as well as modifying the function calling convention for functions taking
22612 @code{long double}. Hence they are not binary-compatible
22613 with code compiled without that switch.
22615 @item -malign-data=@var{type}
22616 @opindex malign-data
22617 Control how GCC aligns variables. Supported values for @var{type} are
22618 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22619 and earlier, @samp{abi} uses alignment value as specified by the
22620 psABI, and @samp{cacheline} uses increased alignment value to match
22621 the cache line size. @samp{compat} is the default.
22623 @item -mlarge-data-threshold=@var{threshold}
22624 @opindex mlarge-data-threshold
22625 When @option{-mcmodel=medium} is specified, data objects larger than
22626 @var{threshold} are placed in the large data section. This value must be the
22627 same across all objects linked into the binary, and defaults to 65535.
22631 Use a different function-calling convention, in which functions that
22632 take a fixed number of arguments return with the @code{ret @var{num}}
22633 instruction, which pops their arguments while returning. This saves one
22634 instruction in the caller since there is no need to pop the arguments
22637 You can specify that an individual function is called with this calling
22638 sequence with the function attribute @code{stdcall}. You can also
22639 override the @option{-mrtd} option by using the function attribute
22640 @code{cdecl}. @xref{Function Attributes}.
22642 @strong{Warning:} this calling convention is incompatible with the one
22643 normally used on Unix, so you cannot use it if you need to call
22644 libraries compiled with the Unix compiler.
22646 Also, you must provide function prototypes for all functions that
22647 take variable numbers of arguments (including @code{printf});
22648 otherwise incorrect code is generated for calls to those
22651 In addition, seriously incorrect code results if you call a
22652 function with too many arguments. (Normally, extra arguments are
22653 harmlessly ignored.)
22655 @item -mregparm=@var{num}
22657 Control how many registers are used to pass integer arguments. By
22658 default, no registers are used to pass arguments, and at most 3
22659 registers can be used. You can control this behavior for a specific
22660 function by using the function attribute @code{regparm}.
22661 @xref{Function Attributes}.
22663 @strong{Warning:} if you use this switch, and
22664 @var{num} is nonzero, then you must build all modules with the same
22665 value, including any libraries. This includes the system libraries and
22669 @opindex msseregparm
22670 Use SSE register passing conventions for float and double arguments
22671 and return values. You can control this behavior for a specific
22672 function by using the function attribute @code{sseregparm}.
22673 @xref{Function Attributes}.
22675 @strong{Warning:} if you use this switch then you must build all
22676 modules with the same value, including any libraries. This includes
22677 the system libraries and startup modules.
22679 @item -mvect8-ret-in-mem
22680 @opindex mvect8-ret-in-mem
22681 Return 8-byte vectors in memory instead of MMX registers. This is the
22682 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
22683 Studio compilers until version 12. Later compiler versions (starting
22684 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
22685 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
22686 you need to remain compatible with existing code produced by those
22687 previous compiler versions or older versions of GCC@.
22696 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
22697 is specified, the significands of results of floating-point operations are
22698 rounded to 24 bits (single precision); @option{-mpc64} rounds the
22699 significands of results of floating-point operations to 53 bits (double
22700 precision) and @option{-mpc80} rounds the significands of results of
22701 floating-point operations to 64 bits (extended double precision), which is
22702 the default. When this option is used, floating-point operations in higher
22703 precisions are not available to the programmer without setting the FPU
22704 control word explicitly.
22706 Setting the rounding of floating-point operations to less than the default
22707 80 bits can speed some programs by 2% or more. Note that some mathematical
22708 libraries assume that extended-precision (80-bit) floating-point operations
22709 are enabled by default; routines in such libraries could suffer significant
22710 loss of accuracy, typically through so-called ``catastrophic cancellation'',
22711 when this option is used to set the precision to less than extended precision.
22713 @item -mstackrealign
22714 @opindex mstackrealign
22715 Realign the stack at entry. On the x86, the @option{-mstackrealign}
22716 option generates an alternate prologue and epilogue that realigns the
22717 run-time stack if necessary. This supports mixing legacy codes that keep
22718 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
22719 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
22720 applicable to individual functions.
22722 @item -mpreferred-stack-boundary=@var{num}
22723 @opindex mpreferred-stack-boundary
22724 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
22725 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
22726 the default is 4 (16 bytes or 128 bits).
22728 @strong{Warning:} When generating code for the x86-64 architecture with
22729 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
22730 used to keep the stack boundary aligned to 8 byte boundary. Since
22731 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
22732 intended to be used in controlled environment where stack space is
22733 important limitation. This option leads to wrong code when functions
22734 compiled with 16 byte stack alignment (such as functions from a standard
22735 library) are called with misaligned stack. In this case, SSE
22736 instructions may lead to misaligned memory access traps. In addition,
22737 variable arguments are handled incorrectly for 16 byte aligned
22738 objects (including x87 long double and __int128), leading to wrong
22739 results. You must build all modules with
22740 @option{-mpreferred-stack-boundary=3}, including any libraries. This
22741 includes the system libraries and startup modules.
22743 @item -mincoming-stack-boundary=@var{num}
22744 @opindex mincoming-stack-boundary
22745 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
22746 boundary. If @option{-mincoming-stack-boundary} is not specified,
22747 the one specified by @option{-mpreferred-stack-boundary} is used.
22749 On Pentium and Pentium Pro, @code{double} and @code{long double} values
22750 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
22751 suffer significant run time performance penalties. On Pentium III, the
22752 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
22753 properly if it is not 16-byte aligned.
22755 To ensure proper alignment of this values on the stack, the stack boundary
22756 must be as aligned as that required by any value stored on the stack.
22757 Further, every function must be generated such that it keeps the stack
22758 aligned. Thus calling a function compiled with a higher preferred
22759 stack boundary from a function compiled with a lower preferred stack
22760 boundary most likely misaligns the stack. It is recommended that
22761 libraries that use callbacks always use the default setting.
22763 This extra alignment does consume extra stack space, and generally
22764 increases code size. Code that is sensitive to stack space usage, such
22765 as embedded systems and operating system kernels, may want to reduce the
22766 preferred alignment to @option{-mpreferred-stack-boundary=2}.
22812 @opindex mclfushopt
22830 @itemx -mprefetchwt1
22831 @opindex mprefetchwt1
22882 These switches enable the use of instructions in the MMX, SSE,
22883 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
22884 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
22885 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX or 3DNow!@:
22886 extended instruction sets. Each has a corresponding @option{-mno-} option
22887 to disable use of these instructions.
22889 These extensions are also available as built-in functions: see
22890 @ref{x86 Built-in Functions}, for details of the functions enabled and
22891 disabled by these switches.
22893 To generate SSE/SSE2 instructions automatically from floating-point
22894 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
22896 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
22897 generates new AVX instructions or AVX equivalence for all SSEx instructions
22900 These options enable GCC to use these extended instructions in
22901 generated code, even without @option{-mfpmath=sse}. Applications that
22902 perform run-time CPU detection must compile separate files for each
22903 supported architecture, using the appropriate flags. In particular,
22904 the file containing the CPU detection code should be compiled without
22907 @item -mdump-tune-features
22908 @opindex mdump-tune-features
22909 This option instructs GCC to dump the names of the x86 performance
22910 tuning features and default settings. The names can be used in
22911 @option{-mtune-ctrl=@var{feature-list}}.
22913 @item -mtune-ctrl=@var{feature-list}
22914 @opindex mtune-ctrl=@var{feature-list}
22915 This option is used to do fine grain control of x86 code generation features.
22916 @var{feature-list} is a comma separated list of @var{feature} names. See also
22917 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
22918 on if it is not preceded with @samp{^}, otherwise, it is turned off.
22919 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
22920 developers. Using it may lead to code paths not covered by testing and can
22921 potentially result in compiler ICEs or runtime errors.
22924 @opindex mno-default
22925 This option instructs GCC to turn off all tunable features. See also
22926 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
22930 This option instructs GCC to emit a @code{cld} instruction in the prologue
22931 of functions that use string instructions. String instructions depend on
22932 the DF flag to select between autoincrement or autodecrement mode. While the
22933 ABI specifies the DF flag to be cleared on function entry, some operating
22934 systems violate this specification by not clearing the DF flag in their
22935 exception dispatchers. The exception handler can be invoked with the DF flag
22936 set, which leads to wrong direction mode when string instructions are used.
22937 This option can be enabled by default on 32-bit x86 targets by configuring
22938 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
22939 instructions can be suppressed with the @option{-mno-cld} compiler option
22943 @opindex mvzeroupper
22944 This option instructs GCC to emit a @code{vzeroupper} instruction
22945 before a transfer of control flow out of the function to minimize
22946 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
22949 @item -mprefer-avx128
22950 @opindex mprefer-avx128
22951 This option instructs GCC to use 128-bit AVX instructions instead of
22952 256-bit AVX instructions in the auto-vectorizer.
22956 This option enables GCC to generate @code{CMPXCHG16B} instructions.
22957 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
22958 (or oword) data types.
22959 This is useful for high-resolution counters that can be updated
22960 by multiple processors (or cores). This instruction is generated as part of
22961 atomic built-in functions: see @ref{__sync Builtins} or
22962 @ref{__atomic Builtins} for details.
22966 This option enables generation of @code{SAHF} instructions in 64-bit code.
22967 Early Intel Pentium 4 CPUs with Intel 64 support,
22968 prior to the introduction of Pentium 4 G1 step in December 2005,
22969 lacked the @code{LAHF} and @code{SAHF} instructions
22970 which are supported by AMD64.
22971 These are load and store instructions, respectively, for certain status flags.
22972 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
22973 @code{drem}, and @code{remainder} built-in functions;
22974 see @ref{Other Builtins} for details.
22978 This option enables use of the @code{movbe} instruction to implement
22979 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
22983 This option enables built-in functions @code{__builtin_ia32_crc32qi},
22984 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
22985 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
22989 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
22990 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
22991 with an additional Newton-Raphson step
22992 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
22993 (and their vectorized
22994 variants) for single-precision floating-point arguments. These instructions
22995 are generated only when @option{-funsafe-math-optimizations} is enabled
22996 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
22997 Note that while the throughput of the sequence is higher than the throughput
22998 of the non-reciprocal instruction, the precision of the sequence can be
22999 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
23001 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
23002 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
23003 combination), and doesn't need @option{-mrecip}.
23005 Also note that GCC emits the above sequence with additional Newton-Raphson step
23006 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
23007 already with @option{-ffast-math} (or the above option combination), and
23008 doesn't need @option{-mrecip}.
23010 @item -mrecip=@var{opt}
23011 @opindex mrecip=opt
23012 This option controls which reciprocal estimate instructions
23013 may be used. @var{opt} is a comma-separated list of options, which may
23014 be preceded by a @samp{!} to invert the option:
23018 Enable all estimate instructions.
23021 Enable the default instructions, equivalent to @option{-mrecip}.
23024 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23027 Enable the approximation for scalar division.
23030 Enable the approximation for vectorized division.
23033 Enable the approximation for scalar square root.
23036 Enable the approximation for vectorized square root.
23039 So, for example, @option{-mrecip=all,!sqrt} enables
23040 all of the reciprocal approximations, except for square root.
23042 @item -mveclibabi=@var{type}
23043 @opindex mveclibabi
23044 Specifies the ABI type to use for vectorizing intrinsics using an
23045 external library. Supported values for @var{type} are @samp{svml}
23046 for the Intel short
23047 vector math library and @samp{acml} for the AMD math core library.
23048 To use this option, both @option{-ftree-vectorize} and
23049 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
23050 ABI-compatible library must be specified at link time.
23052 GCC currently emits calls to @code{vmldExp2},
23053 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
23054 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
23055 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
23056 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
23057 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
23058 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
23059 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
23060 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
23061 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
23062 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
23063 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
23064 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
23065 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
23066 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
23067 when @option{-mveclibabi=acml} is used.
23069 @item -mabi=@var{name}
23071 Generate code for the specified calling convention. Permissible values
23072 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
23073 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
23074 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
23075 You can control this behavior for specific functions by
23076 using the function attributes @code{ms_abi} and @code{sysv_abi}.
23077 @xref{Function Attributes}.
23079 @item -mtls-dialect=@var{type}
23080 @opindex mtls-dialect
23081 Generate code to access thread-local storage using the @samp{gnu} or
23082 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
23083 @samp{gnu2} is more efficient, but it may add compile- and run-time
23084 requirements that cannot be satisfied on all systems.
23087 @itemx -mno-push-args
23088 @opindex mpush-args
23089 @opindex mno-push-args
23090 Use PUSH operations to store outgoing parameters. This method is shorter
23091 and usually equally fast as method using SUB/MOV operations and is enabled
23092 by default. In some cases disabling it may improve performance because of
23093 improved scheduling and reduced dependencies.
23095 @item -maccumulate-outgoing-args
23096 @opindex maccumulate-outgoing-args
23097 If enabled, the maximum amount of space required for outgoing arguments is
23098 computed in the function prologue. This is faster on most modern CPUs
23099 because of reduced dependencies, improved scheduling and reduced stack usage
23100 when the preferred stack boundary is not equal to 2. The drawback is a notable
23101 increase in code size. This switch implies @option{-mno-push-args}.
23105 Support thread-safe exception handling on MinGW. Programs that rely
23106 on thread-safe exception handling must compile and link all code with the
23107 @option{-mthreads} option. When compiling, @option{-mthreads} defines
23108 @option{-D_MT}; when linking, it links in a special thread helper library
23109 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
23111 @item -mno-align-stringops
23112 @opindex mno-align-stringops
23113 Do not align the destination of inlined string operations. This switch reduces
23114 code size and improves performance in case the destination is already aligned,
23115 but GCC doesn't know about it.
23117 @item -minline-all-stringops
23118 @opindex minline-all-stringops
23119 By default GCC inlines string operations only when the destination is
23120 known to be aligned to least a 4-byte boundary.
23121 This enables more inlining and increases code
23122 size, but may improve performance of code that depends on fast
23123 @code{memcpy}, @code{strlen},
23124 and @code{memset} for short lengths.
23126 @item -minline-stringops-dynamically
23127 @opindex minline-stringops-dynamically
23128 For string operations of unknown size, use run-time checks with
23129 inline code for small blocks and a library call for large blocks.
23131 @item -mstringop-strategy=@var{alg}
23132 @opindex mstringop-strategy=@var{alg}
23133 Override the internal decision heuristic for the particular algorithm to use
23134 for inlining string operations. The allowed values for @var{alg} are:
23140 Expand using i386 @code{rep} prefix of the specified size.
23144 @itemx unrolled_loop
23145 Expand into an inline loop.
23148 Always use a library call.
23151 @item -mmemcpy-strategy=@var{strategy}
23152 @opindex mmemcpy-strategy=@var{strategy}
23153 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
23154 should be inlined and what inline algorithm to use when the expected size
23155 of the copy operation is known. @var{strategy}
23156 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
23157 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
23158 the max byte size with which inline algorithm @var{alg} is allowed. For the last
23159 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
23160 in the list must be specified in increasing order. The minimal byte size for
23161 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
23164 @item -mmemset-strategy=@var{strategy}
23165 @opindex mmemset-strategy=@var{strategy}
23166 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
23167 @code{__builtin_memset} expansion.
23169 @item -momit-leaf-frame-pointer
23170 @opindex momit-leaf-frame-pointer
23171 Don't keep the frame pointer in a register for leaf functions. This
23172 avoids the instructions to save, set up, and restore frame pointers and
23173 makes an extra register available in leaf functions. The option
23174 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
23175 which might make debugging harder.
23177 @item -mtls-direct-seg-refs
23178 @itemx -mno-tls-direct-seg-refs
23179 @opindex mtls-direct-seg-refs
23180 Controls whether TLS variables may be accessed with offsets from the
23181 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
23182 or whether the thread base pointer must be added. Whether or not this
23183 is valid depends on the operating system, and whether it maps the
23184 segment to cover the entire TLS area.
23186 For systems that use the GNU C Library, the default is on.
23189 @itemx -mno-sse2avx
23191 Specify that the assembler should encode SSE instructions with VEX
23192 prefix. The option @option{-mavx} turns this on by default.
23197 If profiling is active (@option{-pg}), put the profiling
23198 counter call before the prologue.
23199 Note: On x86 architectures the attribute @code{ms_hook_prologue}
23200 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
23202 @item -mrecord-mcount
23203 @itemx -mno-record-mcount
23204 @opindex mrecord-mcount
23205 If profiling is active (@option{-pg}), generate a __mcount_loc section
23206 that contains pointers to each profiling call. This is useful for
23207 automatically patching and out calls.
23210 @itemx -mno-nop-mcount
23211 @opindex mnop-mcount
23212 If profiling is active (@option{-pg}), generate the calls to
23213 the profiling functions as nops. This is useful when they
23214 should be patched in later dynamically. This is likely only
23215 useful together with @option{-mrecord-mcount}.
23217 @item -mskip-rax-setup
23218 @itemx -mno-skip-rax-setup
23219 @opindex mskip-rax-setup
23220 When generating code for the x86-64 architecture with SSE extensions
23221 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
23222 register when there are no variable arguments passed in vector registers.
23224 @strong{Warning:} Since RAX register is used to avoid unnecessarily
23225 saving vector registers on stack when passing variable arguments, the
23226 impacts of this option are callees may waste some stack space,
23227 misbehave or jump to a random location. GCC 4.4 or newer don't have
23228 those issues, regardless the RAX register value.
23231 @itemx -mno-8bit-idiv
23232 @opindex m8bit-idiv
23233 On some processors, like Intel Atom, 8-bit unsigned integer divide is
23234 much faster than 32-bit/64-bit integer divide. This option generates a
23235 run-time check. If both dividend and divisor are within range of 0
23236 to 255, 8-bit unsigned integer divide is used instead of
23237 32-bit/64-bit integer divide.
23239 @item -mavx256-split-unaligned-load
23240 @itemx -mavx256-split-unaligned-store
23241 @opindex mavx256-split-unaligned-load
23242 @opindex mavx256-split-unaligned-store
23243 Split 32-byte AVX unaligned load and store.
23245 @item -mstack-protector-guard=@var{guard}
23246 @opindex mstack-protector-guard=@var{guard}
23247 Generate stack protection code using canary at @var{guard}. Supported
23248 locations are @samp{global} for global canary or @samp{tls} for per-thread
23249 canary in the TLS block (the default). This option has effect only when
23250 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
23254 These @samp{-m} switches are supported in addition to the above
23255 on x86-64 processors in 64-bit environments.
23266 Generate code for a 16-bit, 32-bit or 64-bit environment.
23267 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
23269 generates code that runs on any i386 system.
23271 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
23272 types to 64 bits, and generates code for the x86-64 architecture.
23273 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
23274 and @option{-mdynamic-no-pic} options.
23276 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
23278 generates code for the x86-64 architecture.
23280 The @option{-m16} option is the same as @option{-m32}, except for that
23281 it outputs the @code{.code16gcc} assembly directive at the beginning of
23282 the assembly output so that the binary can run in 16-bit mode.
23284 @item -mno-red-zone
23285 @opindex mno-red-zone
23286 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
23287 by the x86-64 ABI; it is a 128-byte area beyond the location of the
23288 stack pointer that is not modified by signal or interrupt handlers
23289 and therefore can be used for temporary data without adjusting the stack
23290 pointer. The flag @option{-mno-red-zone} disables this red zone.
23292 @item -mcmodel=small
23293 @opindex mcmodel=small
23294 Generate code for the small code model: the program and its symbols must
23295 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
23296 Programs can be statically or dynamically linked. This is the default
23299 @item -mcmodel=kernel
23300 @opindex mcmodel=kernel
23301 Generate code for the kernel code model. The kernel runs in the
23302 negative 2 GB of the address space.
23303 This model has to be used for Linux kernel code.
23305 @item -mcmodel=medium
23306 @opindex mcmodel=medium
23307 Generate code for the medium model: the program is linked in the lower 2
23308 GB of the address space. Small symbols are also placed there. Symbols
23309 with sizes larger than @option{-mlarge-data-threshold} are put into
23310 large data or BSS sections and can be located above 2GB. Programs can
23311 be statically or dynamically linked.
23313 @item -mcmodel=large
23314 @opindex mcmodel=large
23315 Generate code for the large model. This model makes no assumptions
23316 about addresses and sizes of sections.
23318 @item -maddress-mode=long
23319 @opindex maddress-mode=long
23320 Generate code for long address mode. This is only supported for 64-bit
23321 and x32 environments. It is the default address mode for 64-bit
23324 @item -maddress-mode=short
23325 @opindex maddress-mode=short
23326 Generate code for short address mode. This is only supported for 32-bit
23327 and x32 environments. It is the default address mode for 32-bit and
23331 @node x86 Windows Options
23332 @subsection x86 Windows Options
23333 @cindex x86 Windows Options
23334 @cindex Windows Options for x86
23336 These additional options are available for Microsoft Windows targets:
23342 specifies that a console application is to be generated, by
23343 instructing the linker to set the PE header subsystem type
23344 required for console applications.
23345 This option is available for Cygwin and MinGW targets and is
23346 enabled by default on those targets.
23350 This option is available for Cygwin and MinGW targets. It
23351 specifies that a DLL---a dynamic link library---is to be
23352 generated, enabling the selection of the required runtime
23353 startup object and entry point.
23355 @item -mnop-fun-dllimport
23356 @opindex mnop-fun-dllimport
23357 This option is available for Cygwin and MinGW targets. It
23358 specifies that the @code{dllimport} attribute should be ignored.
23362 This option is available for MinGW targets. It specifies
23363 that MinGW-specific thread support is to be used.
23367 This option is available for MinGW-w64 targets. It causes
23368 the @code{UNICODE} preprocessor macro to be predefined, and
23369 chooses Unicode-capable runtime startup code.
23373 This option is available for Cygwin and MinGW targets. It
23374 specifies that the typical Microsoft Windows predefined macros are to
23375 be set in the pre-processor, but does not influence the choice
23376 of runtime library/startup code.
23380 This option is available for Cygwin and MinGW targets. It
23381 specifies that a GUI application is to be generated by
23382 instructing the linker to set the PE header subsystem type
23385 @item -fno-set-stack-executable
23386 @opindex fno-set-stack-executable
23387 This option is available for MinGW targets. It specifies that
23388 the executable flag for the stack used by nested functions isn't
23389 set. This is necessary for binaries running in kernel mode of
23390 Microsoft Windows, as there the User32 API, which is used to set executable
23391 privileges, isn't available.
23393 @item -fwritable-relocated-rdata
23394 @opindex fno-writable-relocated-rdata
23395 This option is available for MinGW and Cygwin targets. It specifies
23396 that relocated-data in read-only section is put into .data
23397 section. This is a necessary for older runtimes not supporting
23398 modification of .rdata sections for pseudo-relocation.
23400 @item -mpe-aligned-commons
23401 @opindex mpe-aligned-commons
23402 This option is available for Cygwin and MinGW targets. It
23403 specifies that the GNU extension to the PE file format that
23404 permits the correct alignment of COMMON variables should be
23405 used when generating code. It is enabled by default if
23406 GCC detects that the target assembler found during configuration
23407 supports the feature.
23410 See also under @ref{x86 Options} for standard options.
23412 @node Xstormy16 Options
23413 @subsection Xstormy16 Options
23414 @cindex Xstormy16 Options
23416 These options are defined for Xstormy16:
23421 Choose startup files and linker script suitable for the simulator.
23424 @node Xtensa Options
23425 @subsection Xtensa Options
23426 @cindex Xtensa Options
23428 These options are supported for Xtensa targets:
23432 @itemx -mno-const16
23434 @opindex mno-const16
23435 Enable or disable use of @code{CONST16} instructions for loading
23436 constant values. The @code{CONST16} instruction is currently not a
23437 standard option from Tensilica. When enabled, @code{CONST16}
23438 instructions are always used in place of the standard @code{L32R}
23439 instructions. The use of @code{CONST16} is enabled by default only if
23440 the @code{L32R} instruction is not available.
23443 @itemx -mno-fused-madd
23444 @opindex mfused-madd
23445 @opindex mno-fused-madd
23446 Enable or disable use of fused multiply/add and multiply/subtract
23447 instructions in the floating-point option. This has no effect if the
23448 floating-point option is not also enabled. Disabling fused multiply/add
23449 and multiply/subtract instructions forces the compiler to use separate
23450 instructions for the multiply and add/subtract operations. This may be
23451 desirable in some cases where strict IEEE 754-compliant results are
23452 required: the fused multiply add/subtract instructions do not round the
23453 intermediate result, thereby producing results with @emph{more} bits of
23454 precision than specified by the IEEE standard. Disabling fused multiply
23455 add/subtract instructions also ensures that the program output is not
23456 sensitive to the compiler's ability to combine multiply and add/subtract
23459 @item -mserialize-volatile
23460 @itemx -mno-serialize-volatile
23461 @opindex mserialize-volatile
23462 @opindex mno-serialize-volatile
23463 When this option is enabled, GCC inserts @code{MEMW} instructions before
23464 @code{volatile} memory references to guarantee sequential consistency.
23465 The default is @option{-mserialize-volatile}. Use
23466 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
23468 @item -mforce-no-pic
23469 @opindex mforce-no-pic
23470 For targets, like GNU/Linux, where all user-mode Xtensa code must be
23471 position-independent code (PIC), this option disables PIC for compiling
23474 @item -mtext-section-literals
23475 @itemx -mno-text-section-literals
23476 @opindex mtext-section-literals
23477 @opindex mno-text-section-literals
23478 These options control the treatment of literal pools. The default is
23479 @option{-mno-text-section-literals}, which places literals in a separate
23480 section in the output file. This allows the literal pool to be placed
23481 in a data RAM/ROM, and it also allows the linker to combine literal
23482 pools from separate object files to remove redundant literals and
23483 improve code size. With @option{-mtext-section-literals}, the literals
23484 are interspersed in the text section in order to keep them as close as
23485 possible to their references. This may be necessary for large assembly
23488 @item -mtarget-align
23489 @itemx -mno-target-align
23490 @opindex mtarget-align
23491 @opindex mno-target-align
23492 When this option is enabled, GCC instructs the assembler to
23493 automatically align instructions to reduce branch penalties at the
23494 expense of some code density. The assembler attempts to widen density
23495 instructions to align branch targets and the instructions following call
23496 instructions. If there are not enough preceding safe density
23497 instructions to align a target, no widening is performed. The
23498 default is @option{-mtarget-align}. These options do not affect the
23499 treatment of auto-aligned instructions like @code{LOOP}, which the
23500 assembler always aligns, either by widening density instructions or
23501 by inserting NOP instructions.
23504 @itemx -mno-longcalls
23505 @opindex mlongcalls
23506 @opindex mno-longcalls
23507 When this option is enabled, GCC instructs the assembler to translate
23508 direct calls to indirect calls unless it can determine that the target
23509 of a direct call is in the range allowed by the call instruction. This
23510 translation typically occurs for calls to functions in other source
23511 files. Specifically, the assembler translates a direct @code{CALL}
23512 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
23513 The default is @option{-mno-longcalls}. This option should be used in
23514 programs where the call target can potentially be out of range. This
23515 option is implemented in the assembler, not the compiler, so the
23516 assembly code generated by GCC still shows direct call
23517 instructions---look at the disassembled object code to see the actual
23518 instructions. Note that the assembler uses an indirect call for
23519 every cross-file call, not just those that really are out of range.
23522 @node zSeries Options
23523 @subsection zSeries Options
23524 @cindex zSeries options
23526 These are listed under @xref{S/390 and zSeries Options}.
23528 @node Code Gen Options
23529 @section Options for Code Generation Conventions
23530 @cindex code generation conventions
23531 @cindex options, code generation
23532 @cindex run-time options
23534 These machine-independent options control the interface conventions
23535 used in code generation.
23537 Most of them have both positive and negative forms; the negative form
23538 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
23539 one of the forms is listed---the one that is not the default. You
23540 can figure out the other form by either removing @samp{no-} or adding
23544 @item -fbounds-check
23545 @opindex fbounds-check
23546 For front ends that support it, generate additional code to check that
23547 indices used to access arrays are within the declared range. This is
23548 currently only supported by the Java and Fortran front ends, where
23549 this option defaults to true and false respectively.
23551 @item -fstack-reuse=@var{reuse-level}
23552 @opindex fstack_reuse
23553 This option controls stack space reuse for user declared local/auto variables
23554 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
23555 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
23556 local variables and temporaries, @samp{named_vars} enables the reuse only for
23557 user defined local variables with names, and @samp{none} disables stack reuse
23558 completely. The default value is @samp{all}. The option is needed when the
23559 program extends the lifetime of a scoped local variable or a compiler generated
23560 temporary beyond the end point defined by the language. When a lifetime of
23561 a variable ends, and if the variable lives in memory, the optimizing compiler
23562 has the freedom to reuse its stack space with other temporaries or scoped
23563 local variables whose live range does not overlap with it. Legacy code extending
23564 local lifetime is likely to break with the stack reuse optimization.
23583 if (*p == 10) // out of scope use of local1
23594 A(int k) : i(k), j(k) @{ @}
23601 void foo(const A& ar)
23608 foo(A(10)); // temp object's lifetime ends when foo returns
23614 ap->i+= 10; // ap references out of scope temp whose space
23615 // is reused with a. What is the value of ap->i?
23620 The lifetime of a compiler generated temporary is well defined by the C++
23621 standard. When a lifetime of a temporary ends, and if the temporary lives
23622 in memory, the optimizing compiler has the freedom to reuse its stack
23623 space with other temporaries or scoped local variables whose live range
23624 does not overlap with it. However some of the legacy code relies on
23625 the behavior of older compilers in which temporaries' stack space is
23626 not reused, the aggressive stack reuse can lead to runtime errors. This
23627 option is used to control the temporary stack reuse optimization.
23631 This option generates traps for signed overflow on addition, subtraction,
23632 multiplication operations.
23636 This option instructs the compiler to assume that signed arithmetic
23637 overflow of addition, subtraction and multiplication wraps around
23638 using twos-complement representation. This flag enables some optimizations
23639 and disables others. This option is enabled by default for the Java
23640 front end, as required by the Java language specification.
23643 @opindex fexceptions
23644 Enable exception handling. Generates extra code needed to propagate
23645 exceptions. For some targets, this implies GCC generates frame
23646 unwind information for all functions, which can produce significant data
23647 size overhead, although it does not affect execution. If you do not
23648 specify this option, GCC enables it by default for languages like
23649 C++ that normally require exception handling, and disables it for
23650 languages like C that do not normally require it. However, you may need
23651 to enable this option when compiling C code that needs to interoperate
23652 properly with exception handlers written in C++. You may also wish to
23653 disable this option if you are compiling older C++ programs that don't
23654 use exception handling.
23656 @item -fnon-call-exceptions
23657 @opindex fnon-call-exceptions
23658 Generate code that allows trapping instructions to throw exceptions.
23659 Note that this requires platform-specific runtime support that does
23660 not exist everywhere. Moreover, it only allows @emph{trapping}
23661 instructions to throw exceptions, i.e.@: memory references or floating-point
23662 instructions. It does not allow exceptions to be thrown from
23663 arbitrary signal handlers such as @code{SIGALRM}.
23665 @item -fdelete-dead-exceptions
23666 @opindex fdelete-dead-exceptions
23667 Consider that instructions that may throw exceptions but don't otherwise
23668 contribute to the execution of the program can be optimized away.
23669 This option is enabled by default for the Ada front end, as permitted by
23670 the Ada language specification.
23671 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23673 @item -funwind-tables
23674 @opindex funwind-tables
23675 Similar to @option{-fexceptions}, except that it just generates any needed
23676 static data, but does not affect the generated code in any other way.
23677 You normally do not need to enable this option; instead, a language processor
23678 that needs this handling enables it on your behalf.
23680 @item -fasynchronous-unwind-tables
23681 @opindex fasynchronous-unwind-tables
23682 Generate unwind table in DWARF 2 format, if supported by target machine. The
23683 table is exact at each instruction boundary, so it can be used for stack
23684 unwinding from asynchronous events (such as debugger or garbage collector).
23686 @item -fno-gnu-unique
23687 @opindex fno-gnu-unique
23688 On systems with recent GNU assembler and C library, the C++ compiler
23689 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23690 of template static data members and static local variables in inline
23691 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23692 is necessary to avoid problems with a library used by two different
23693 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23694 therefore disagreeing with the other one about the binding of the
23695 symbol. But this causes @code{dlclose} to be ignored for affected
23696 DSOs; if your program relies on reinitialization of a DSO via
23697 @code{dlclose} and @code{dlopen}, you can use
23698 @option{-fno-gnu-unique}.
23700 @item -fpcc-struct-return
23701 @opindex fpcc-struct-return
23702 Return ``short'' @code{struct} and @code{union} values in memory like
23703 longer ones, rather than in registers. This convention is less
23704 efficient, but it has the advantage of allowing intercallability between
23705 GCC-compiled files and files compiled with other compilers, particularly
23706 the Portable C Compiler (pcc).
23708 The precise convention for returning structures in memory depends
23709 on the target configuration macros.
23711 Short structures and unions are those whose size and alignment match
23712 that of some integer type.
23714 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23715 switch is not binary compatible with code compiled with the
23716 @option{-freg-struct-return} switch.
23717 Use it to conform to a non-default application binary interface.
23719 @item -freg-struct-return
23720 @opindex freg-struct-return
23721 Return @code{struct} and @code{union} values in registers when possible.
23722 This is more efficient for small structures than
23723 @option{-fpcc-struct-return}.
23725 If you specify neither @option{-fpcc-struct-return} nor
23726 @option{-freg-struct-return}, GCC defaults to whichever convention is
23727 standard for the target. If there is no standard convention, GCC
23728 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23729 the principal compiler. In those cases, we can choose the standard, and
23730 we chose the more efficient register return alternative.
23732 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23733 switch is not binary compatible with code compiled with the
23734 @option{-fpcc-struct-return} switch.
23735 Use it to conform to a non-default application binary interface.
23737 @item -fshort-enums
23738 @opindex fshort-enums
23739 Allocate to an @code{enum} type only as many bytes as it needs for the
23740 declared range of possible values. Specifically, the @code{enum} type
23741 is equivalent to the smallest integer type that has enough room.
23743 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23744 code that is not binary compatible with code generated without that switch.
23745 Use it to conform to a non-default application binary interface.
23747 @item -fshort-double
23748 @opindex fshort-double
23749 Use the same size for @code{double} as for @code{float}.
23751 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23752 code that is not binary compatible with code generated without that switch.
23753 Use it to conform to a non-default application binary interface.
23755 @item -fshort-wchar
23756 @opindex fshort-wchar
23757 Override the underlying type for @code{wchar_t} to be @code{short
23758 unsigned int} instead of the default for the target. This option is
23759 useful for building programs to run under WINE@.
23761 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23762 code that is not binary compatible with code generated without that switch.
23763 Use it to conform to a non-default application binary interface.
23766 @opindex fno-common
23767 In C code, controls the placement of uninitialized global variables.
23768 Unix C compilers have traditionally permitted multiple definitions of
23769 such variables in different compilation units by placing the variables
23771 This is the behavior specified by @option{-fcommon}, and is the default
23772 for GCC on most targets.
23773 On the other hand, this behavior is not required by ISO C, and on some
23774 targets may carry a speed or code size penalty on variable references.
23775 The @option{-fno-common} option specifies that the compiler should place
23776 uninitialized global variables in the data section of the object file,
23777 rather than generating them as common blocks.
23778 This has the effect that if the same variable is declared
23779 (without @code{extern}) in two different compilations,
23780 you get a multiple-definition error when you link them.
23781 In this case, you must compile with @option{-fcommon} instead.
23782 Compiling with @option{-fno-common} is useful on targets for which
23783 it provides better performance, or if you wish to verify that the
23784 program will work on other systems that always treat uninitialized
23785 variable declarations this way.
23789 Ignore the @code{#ident} directive.
23791 @item -finhibit-size-directive
23792 @opindex finhibit-size-directive
23793 Don't output a @code{.size} assembler directive, or anything else that
23794 would cause trouble if the function is split in the middle, and the
23795 two halves are placed at locations far apart in memory. This option is
23796 used when compiling @file{crtstuff.c}; you should not need to use it
23799 @item -fverbose-asm
23800 @opindex fverbose-asm
23801 Put extra commentary information in the generated assembly code to
23802 make it more readable. This option is generally only of use to those
23803 who actually need to read the generated assembly code (perhaps while
23804 debugging the compiler itself).
23806 @option{-fno-verbose-asm}, the default, causes the
23807 extra information to be omitted and is useful when comparing two assembler
23810 @item -frecord-gcc-switches
23811 @opindex frecord-gcc-switches
23812 This switch causes the command line used to invoke the
23813 compiler to be recorded into the object file that is being created.
23814 This switch is only implemented on some targets and the exact format
23815 of the recording is target and binary file format dependent, but it
23816 usually takes the form of a section containing ASCII text. This
23817 switch is related to the @option{-fverbose-asm} switch, but that
23818 switch only records information in the assembler output file as
23819 comments, so it never reaches the object file.
23820 See also @option{-grecord-gcc-switches} for another
23821 way of storing compiler options into the object file.
23825 @cindex global offset table
23827 Generate position-independent code (PIC) suitable for use in a shared
23828 library, if supported for the target machine. Such code accesses all
23829 constant addresses through a global offset table (GOT)@. The dynamic
23830 loader resolves the GOT entries when the program starts (the dynamic
23831 loader is not part of GCC; it is part of the operating system). If
23832 the GOT size for the linked executable exceeds a machine-specific
23833 maximum size, you get an error message from the linker indicating that
23834 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23835 instead. (These maximums are 8k on the SPARC and 32k
23836 on the m68k and RS/6000. The x86 has no such limit.)
23838 Position-independent code requires special support, and therefore works
23839 only on certain machines. For the x86, GCC supports PIC for System V
23840 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23841 position-independent.
23843 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23848 If supported for the target machine, emit position-independent code,
23849 suitable for dynamic linking and avoiding any limit on the size of the
23850 global offset table. This option makes a difference on the m68k,
23851 PowerPC and SPARC@.
23853 Position-independent code requires special support, and therefore works
23854 only on certain machines.
23856 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23863 These options are similar to @option{-fpic} and @option{-fPIC}, but
23864 generated position independent code can be only linked into executables.
23865 Usually these options are used when @option{-pie} GCC option is
23866 used during linking.
23868 @option{-fpie} and @option{-fPIE} both define the macros
23869 @code{__pie__} and @code{__PIE__}. The macros have the value 1
23870 for @option{-fpie} and 2 for @option{-fPIE}.
23874 Do not use PLT for external function calls in position-independent code.
23875 Instead, load callee address at call site from GOT and branch to it.
23876 This leads to more efficient code by eliminating PLT stubs and exposing
23877 GOT load to optimizations. On architectures such as 32-bit x86 where
23878 PLT stubs expect GOT pointer in a specific register, this gives more
23879 register allocation freedom to the compiler. Lazy binding requires PLT:
23880 with @option{-fno-plt} all external symbols are resolved at load time.
23882 Alternatively, function attribute @code{noplt} can be used to avoid PLT
23883 for calls to specific external functions by marking those functions with
23886 Additionally, a few targets also convert calls to those functions that are
23887 marked to not use the PLT to use the GOT instead for non-position independent
23890 @item -fno-jump-tables
23891 @opindex fno-jump-tables
23892 Do not use jump tables for switch statements even where it would be
23893 more efficient than other code generation strategies. This option is
23894 of use in conjunction with @option{-fpic} or @option{-fPIC} for
23895 building code that forms part of a dynamic linker and cannot
23896 reference the address of a jump table. On some targets, jump tables
23897 do not require a GOT and this option is not needed.
23899 @item -ffixed-@var{reg}
23901 Treat the register named @var{reg} as a fixed register; generated code
23902 should never refer to it (except perhaps as a stack pointer, frame
23903 pointer or in some other fixed role).
23905 @var{reg} must be the name of a register. The register names accepted
23906 are machine-specific and are defined in the @code{REGISTER_NAMES}
23907 macro in the machine description macro file.
23909 This flag does not have a negative form, because it specifies a
23912 @item -fcall-used-@var{reg}
23913 @opindex fcall-used
23914 Treat the register named @var{reg} as an allocable register that is
23915 clobbered by function calls. It may be allocated for temporaries or
23916 variables that do not live across a call. Functions compiled this way
23917 do not save and restore the register @var{reg}.
23919 It is an error to use this flag with the frame pointer or stack pointer.
23920 Use of this flag for other registers that have fixed pervasive roles in
23921 the machine's execution model produces disastrous results.
23923 This flag does not have a negative form, because it specifies a
23926 @item -fcall-saved-@var{reg}
23927 @opindex fcall-saved
23928 Treat the register named @var{reg} as an allocable register saved by
23929 functions. It may be allocated even for temporaries or variables that
23930 live across a call. Functions compiled this way save and restore
23931 the register @var{reg} if they use it.
23933 It is an error to use this flag with the frame pointer or stack pointer.
23934 Use of this flag for other registers that have fixed pervasive roles in
23935 the machine's execution model produces disastrous results.
23937 A different sort of disaster results from the use of this flag for
23938 a register in which function values may be returned.
23940 This flag does not have a negative form, because it specifies a
23943 @item -fpack-struct[=@var{n}]
23944 @opindex fpack-struct
23945 Without a value specified, pack all structure members together without
23946 holes. When a value is specified (which must be a small power of two), pack
23947 structure members according to this value, representing the maximum
23948 alignment (that is, objects with default alignment requirements larger than
23949 this are output potentially unaligned at the next fitting location.
23951 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
23952 code that is not binary compatible with code generated without that switch.
23953 Additionally, it makes the code suboptimal.
23954 Use it to conform to a non-default application binary interface.
23956 @item -finstrument-functions
23957 @opindex finstrument-functions
23958 Generate instrumentation calls for entry and exit to functions. Just
23959 after function entry and just before function exit, the following
23960 profiling functions are called with the address of the current
23961 function and its call site. (On some platforms,
23962 @code{__builtin_return_address} does not work beyond the current
23963 function, so the call site information may not be available to the
23964 profiling functions otherwise.)
23967 void __cyg_profile_func_enter (void *this_fn,
23969 void __cyg_profile_func_exit (void *this_fn,
23973 The first argument is the address of the start of the current function,
23974 which may be looked up exactly in the symbol table.
23976 This instrumentation is also done for functions expanded inline in other
23977 functions. The profiling calls indicate where, conceptually, the
23978 inline function is entered and exited. This means that addressable
23979 versions of such functions must be available. If all your uses of a
23980 function are expanded inline, this may mean an additional expansion of
23981 code size. If you use @code{extern inline} in your C code, an
23982 addressable version of such functions must be provided. (This is
23983 normally the case anyway, but if you get lucky and the optimizer always
23984 expands the functions inline, you might have gotten away without
23985 providing static copies.)
23987 A function may be given the attribute @code{no_instrument_function}, in
23988 which case this instrumentation is not done. This can be used, for
23989 example, for the profiling functions listed above, high-priority
23990 interrupt routines, and any functions from which the profiling functions
23991 cannot safely be called (perhaps signal handlers, if the profiling
23992 routines generate output or allocate memory).
23994 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23995 @opindex finstrument-functions-exclude-file-list
23997 Set the list of functions that are excluded from instrumentation (see
23998 the description of @option{-finstrument-functions}). If the file that
23999 contains a function definition matches with one of @var{file}, then
24000 that function is not instrumented. The match is done on substrings:
24001 if the @var{file} parameter is a substring of the file name, it is
24002 considered to be a match.
24007 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
24011 excludes any inline function defined in files whose pathnames
24012 contain @file{/bits/stl} or @file{include/sys}.
24014 If, for some reason, you want to include letter @samp{,} in one of
24015 @var{sym}, write @samp{\,}. For example,
24016 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
24017 (note the single quote surrounding the option).
24019 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
24020 @opindex finstrument-functions-exclude-function-list
24022 This is similar to @option{-finstrument-functions-exclude-file-list},
24023 but this option sets the list of function names to be excluded from
24024 instrumentation. The function name to be matched is its user-visible
24025 name, such as @code{vector<int> blah(const vector<int> &)}, not the
24026 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
24027 match is done on substrings: if the @var{sym} parameter is a substring
24028 of the function name, it is considered to be a match. For C99 and C++
24029 extended identifiers, the function name must be given in UTF-8, not
24030 using universal character names.
24032 @item -fstack-check
24033 @opindex fstack-check
24034 Generate code to verify that you do not go beyond the boundary of the
24035 stack. You should specify this flag if you are running in an
24036 environment with multiple threads, but you only rarely need to specify it in
24037 a single-threaded environment since stack overflow is automatically
24038 detected on nearly all systems if there is only one stack.
24040 Note that this switch does not actually cause checking to be done; the
24041 operating system or the language runtime must do that. The switch causes
24042 generation of code to ensure that they see the stack being extended.
24044 You can additionally specify a string parameter: @samp{no} means no
24045 checking, @samp{generic} means force the use of old-style checking,
24046 @samp{specific} means use the best checking method and is equivalent
24047 to bare @option{-fstack-check}.
24049 Old-style checking is a generic mechanism that requires no specific
24050 target support in the compiler but comes with the following drawbacks:
24054 Modified allocation strategy for large objects: they are always
24055 allocated dynamically if their size exceeds a fixed threshold.
24058 Fixed limit on the size of the static frame of functions: when it is
24059 topped by a particular function, stack checking is not reliable and
24060 a warning is issued by the compiler.
24063 Inefficiency: because of both the modified allocation strategy and the
24064 generic implementation, code performance is hampered.
24067 Note that old-style stack checking is also the fallback method for
24068 @samp{specific} if no target support has been added in the compiler.
24070 @item -fstack-limit-register=@var{reg}
24071 @itemx -fstack-limit-symbol=@var{sym}
24072 @itemx -fno-stack-limit
24073 @opindex fstack-limit-register
24074 @opindex fstack-limit-symbol
24075 @opindex fno-stack-limit
24076 Generate code to ensure that the stack does not grow beyond a certain value,
24077 either the value of a register or the address of a symbol. If a larger
24078 stack is required, a signal is raised at run time. For most targets,
24079 the signal is raised before the stack overruns the boundary, so
24080 it is possible to catch the signal without taking special precautions.
24082 For instance, if the stack starts at absolute address @samp{0x80000000}
24083 and grows downwards, you can use the flags
24084 @option{-fstack-limit-symbol=__stack_limit} and
24085 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
24086 of 128KB@. Note that this may only work with the GNU linker.
24088 @item -fsplit-stack
24089 @opindex fsplit-stack
24090 Generate code to automatically split the stack before it overflows.
24091 The resulting program has a discontiguous stack which can only
24092 overflow if the program is unable to allocate any more memory. This
24093 is most useful when running threaded programs, as it is no longer
24094 necessary to calculate a good stack size to use for each thread. This
24095 is currently only implemented for the x86 targets running
24098 When code compiled with @option{-fsplit-stack} calls code compiled
24099 without @option{-fsplit-stack}, there may not be much stack space
24100 available for the latter code to run. If compiling all code,
24101 including library code, with @option{-fsplit-stack} is not an option,
24102 then the linker can fix up these calls so that the code compiled
24103 without @option{-fsplit-stack} always has a large stack. Support for
24104 this is implemented in the gold linker in GNU binutils release 2.21
24107 @item -fleading-underscore
24108 @opindex fleading-underscore
24109 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
24110 change the way C symbols are represented in the object file. One use
24111 is to help link with legacy assembly code.
24113 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
24114 generate code that is not binary compatible with code generated without that
24115 switch. Use it to conform to a non-default application binary interface.
24116 Not all targets provide complete support for this switch.
24118 @item -ftls-model=@var{model}
24119 @opindex ftls-model
24120 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
24121 The @var{model} argument should be one of @samp{global-dynamic},
24122 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
24123 Note that the choice is subject to optimization: the compiler may use
24124 a more efficient model for symbols not visible outside of the translation
24125 unit, or if @option{-fpic} is not given on the command line.
24127 The default without @option{-fpic} is @samp{initial-exec}; with
24128 @option{-fpic} the default is @samp{global-dynamic}.
24130 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
24131 @opindex fvisibility
24132 Set the default ELF image symbol visibility to the specified option---all
24133 symbols are marked with this unless overridden within the code.
24134 Using this feature can very substantially improve linking and
24135 load times of shared object libraries, produce more optimized
24136 code, provide near-perfect API export and prevent symbol clashes.
24137 It is @strong{strongly} recommended that you use this in any shared objects
24140 Despite the nomenclature, @samp{default} always means public; i.e.,
24141 available to be linked against from outside the shared object.
24142 @samp{protected} and @samp{internal} are pretty useless in real-world
24143 usage so the only other commonly used option is @samp{hidden}.
24144 The default if @option{-fvisibility} isn't specified is
24145 @samp{default}, i.e., make every symbol public.
24147 A good explanation of the benefits offered by ensuring ELF
24148 symbols have the correct visibility is given by ``How To Write
24149 Shared Libraries'' by Ulrich Drepper (which can be found at
24150 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
24151 solution made possible by this option to marking things hidden when
24152 the default is public is to make the default hidden and mark things
24153 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
24154 and @code{__attribute__ ((visibility("default")))} instead of
24155 @code{__declspec(dllexport)} you get almost identical semantics with
24156 identical syntax. This is a great boon to those working with
24157 cross-platform projects.
24159 For those adding visibility support to existing code, you may find
24160 @code{#pragma GCC visibility} of use. This works by you enclosing
24161 the declarations you wish to set visibility for with (for example)
24162 @code{#pragma GCC visibility push(hidden)} and
24163 @code{#pragma GCC visibility pop}.
24164 Bear in mind that symbol visibility should be viewed @strong{as
24165 part of the API interface contract} and thus all new code should
24166 always specify visibility when it is not the default; i.e., declarations
24167 only for use within the local DSO should @strong{always} be marked explicitly
24168 as hidden as so to avoid PLT indirection overheads---making this
24169 abundantly clear also aids readability and self-documentation of the code.
24170 Note that due to ISO C++ specification requirements, @code{operator new} and
24171 @code{operator delete} must always be of default visibility.
24173 Be aware that headers from outside your project, in particular system
24174 headers and headers from any other library you use, may not be
24175 expecting to be compiled with visibility other than the default. You
24176 may need to explicitly say @code{#pragma GCC visibility push(default)}
24177 before including any such headers.
24179 @code{extern} declarations are not affected by @option{-fvisibility}, so
24180 a lot of code can be recompiled with @option{-fvisibility=hidden} with
24181 no modifications. However, this means that calls to @code{extern}
24182 functions with no explicit visibility use the PLT, so it is more
24183 effective to use @code{__attribute ((visibility))} and/or
24184 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
24185 declarations should be treated as hidden.
24187 Note that @option{-fvisibility} does affect C++ vague linkage
24188 entities. This means that, for instance, an exception class that is
24189 be thrown between DSOs must be explicitly marked with default
24190 visibility so that the @samp{type_info} nodes are unified between
24193 An overview of these techniques, their benefits and how to use them
24194 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
24196 @item -fstrict-volatile-bitfields
24197 @opindex fstrict-volatile-bitfields
24198 This option should be used if accesses to volatile bit-fields (or other
24199 structure fields, although the compiler usually honors those types
24200 anyway) should use a single access of the width of the
24201 field's type, aligned to a natural alignment if possible. For
24202 example, targets with memory-mapped peripheral registers might require
24203 all such accesses to be 16 bits wide; with this flag you can
24204 declare all peripheral bit-fields as @code{unsigned short} (assuming short
24205 is 16 bits on these targets) to force GCC to use 16-bit accesses
24206 instead of, perhaps, a more efficient 32-bit access.
24208 If this option is disabled, the compiler uses the most efficient
24209 instruction. In the previous example, that might be a 32-bit load
24210 instruction, even though that accesses bytes that do not contain
24211 any portion of the bit-field, or memory-mapped registers unrelated to
24212 the one being updated.
24214 In some cases, such as when the @code{packed} attribute is applied to a
24215 structure field, it may not be possible to access the field with a single
24216 read or write that is correctly aligned for the target machine. In this
24217 case GCC falls back to generating multiple accesses rather than code that
24218 will fault or truncate the result at run time.
24220 Note: Due to restrictions of the C/C++11 memory model, write accesses are
24221 not allowed to touch non bit-field members. It is therefore recommended
24222 to define all bits of the field's type as bit-field members.
24224 The default value of this option is determined by the application binary
24225 interface for the target processor.
24227 @item -fsync-libcalls
24228 @opindex fsync-libcalls
24229 This option controls whether any out-of-line instance of the @code{__sync}
24230 family of functions may be used to implement the C++11 @code{__atomic}
24231 family of functions.
24233 The default value of this option is enabled, thus the only useful form
24234 of the option is @option{-fno-sync-libcalls}. This option is used in
24235 the implementation of the @file{libatomic} runtime library.
24241 @node Environment Variables
24242 @section Environment Variables Affecting GCC
24243 @cindex environment variables
24245 @c man begin ENVIRONMENT
24246 This section describes several environment variables that affect how GCC
24247 operates. Some of them work by specifying directories or prefixes to use
24248 when searching for various kinds of files. Some are used to specify other
24249 aspects of the compilation environment.
24251 Note that you can also specify places to search using options such as
24252 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
24253 take precedence over places specified using environment variables, which
24254 in turn take precedence over those specified by the configuration of GCC@.
24255 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
24256 GNU Compiler Collection (GCC) Internals}.
24261 @c @itemx LC_COLLATE
24263 @c @itemx LC_MONETARY
24264 @c @itemx LC_NUMERIC
24269 @c @findex LC_COLLATE
24270 @findex LC_MESSAGES
24271 @c @findex LC_MONETARY
24272 @c @findex LC_NUMERIC
24276 These environment variables control the way that GCC uses
24277 localization information which allows GCC to work with different
24278 national conventions. GCC inspects the locale categories
24279 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
24280 so. These locale categories can be set to any value supported by your
24281 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
24282 Kingdom encoded in UTF-8.
24284 The @env{LC_CTYPE} environment variable specifies character
24285 classification. GCC uses it to determine the character boundaries in
24286 a string; this is needed for some multibyte encodings that contain quote
24287 and escape characters that are otherwise interpreted as a string
24290 The @env{LC_MESSAGES} environment variable specifies the language to
24291 use in diagnostic messages.
24293 If the @env{LC_ALL} environment variable is set, it overrides the value
24294 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
24295 and @env{LC_MESSAGES} default to the value of the @env{LANG}
24296 environment variable. If none of these variables are set, GCC
24297 defaults to traditional C English behavior.
24301 If @env{TMPDIR} is set, it specifies the directory to use for temporary
24302 files. GCC uses temporary files to hold the output of one stage of
24303 compilation which is to be used as input to the next stage: for example,
24304 the output of the preprocessor, which is the input to the compiler
24307 @item GCC_COMPARE_DEBUG
24308 @findex GCC_COMPARE_DEBUG
24309 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
24310 @option{-fcompare-debug} to the compiler driver. See the documentation
24311 of this option for more details.
24313 @item GCC_EXEC_PREFIX
24314 @findex GCC_EXEC_PREFIX
24315 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
24316 names of the subprograms executed by the compiler. No slash is added
24317 when this prefix is combined with the name of a subprogram, but you can
24318 specify a prefix that ends with a slash if you wish.
24320 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
24321 an appropriate prefix to use based on the pathname it is invoked with.
24323 If GCC cannot find the subprogram using the specified prefix, it
24324 tries looking in the usual places for the subprogram.
24326 The default value of @env{GCC_EXEC_PREFIX} is
24327 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
24328 the installed compiler. In many cases @var{prefix} is the value
24329 of @code{prefix} when you ran the @file{configure} script.
24331 Other prefixes specified with @option{-B} take precedence over this prefix.
24333 This prefix is also used for finding files such as @file{crt0.o} that are
24336 In addition, the prefix is used in an unusual way in finding the
24337 directories to search for header files. For each of the standard
24338 directories whose name normally begins with @samp{/usr/local/lib/gcc}
24339 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
24340 replacing that beginning with the specified prefix to produce an
24341 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
24342 @file{foo/bar} just before it searches the standard directory
24343 @file{/usr/local/lib/bar}.
24344 If a standard directory begins with the configured
24345 @var{prefix} then the value of @var{prefix} is replaced by
24346 @env{GCC_EXEC_PREFIX} when looking for header files.
24348 @item COMPILER_PATH
24349 @findex COMPILER_PATH
24350 The value of @env{COMPILER_PATH} is a colon-separated list of
24351 directories, much like @env{PATH}. GCC tries the directories thus
24352 specified when searching for subprograms, if it can't find the
24353 subprograms using @env{GCC_EXEC_PREFIX}.
24356 @findex LIBRARY_PATH
24357 The value of @env{LIBRARY_PATH} is a colon-separated list of
24358 directories, much like @env{PATH}. When configured as a native compiler,
24359 GCC tries the directories thus specified when searching for special
24360 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
24361 using GCC also uses these directories when searching for ordinary
24362 libraries for the @option{-l} option (but directories specified with
24363 @option{-L} come first).
24367 @cindex locale definition
24368 This variable is used to pass locale information to the compiler. One way in
24369 which this information is used is to determine the character set to be used
24370 when character literals, string literals and comments are parsed in C and C++.
24371 When the compiler is configured to allow multibyte characters,
24372 the following values for @env{LANG} are recognized:
24376 Recognize JIS characters.
24378 Recognize SJIS characters.
24380 Recognize EUCJP characters.
24383 If @env{LANG} is not defined, or if it has some other value, then the
24384 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
24385 recognize and translate multibyte characters.
24389 Some additional environment variables affect the behavior of the
24392 @include cppenv.texi
24396 @node Precompiled Headers
24397 @section Using Precompiled Headers
24398 @cindex precompiled headers
24399 @cindex speed of compilation
24401 Often large projects have many header files that are included in every
24402 source file. The time the compiler takes to process these header files
24403 over and over again can account for nearly all of the time required to
24404 build the project. To make builds faster, GCC allows you to
24405 @dfn{precompile} a header file.
24407 To create a precompiled header file, simply compile it as you would any
24408 other file, if necessary using the @option{-x} option to make the driver
24409 treat it as a C or C++ header file. You may want to use a
24410 tool like @command{make} to keep the precompiled header up-to-date when
24411 the headers it contains change.
24413 A precompiled header file is searched for when @code{#include} is
24414 seen in the compilation. As it searches for the included file
24415 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
24416 compiler looks for a precompiled header in each directory just before it
24417 looks for the include file in that directory. The name searched for is
24418 the name specified in the @code{#include} with @samp{.gch} appended. If
24419 the precompiled header file can't be used, it is ignored.
24421 For instance, if you have @code{#include "all.h"}, and you have
24422 @file{all.h.gch} in the same directory as @file{all.h}, then the
24423 precompiled header file is used if possible, and the original
24424 header is used otherwise.
24426 Alternatively, you might decide to put the precompiled header file in a
24427 directory and use @option{-I} to ensure that directory is searched
24428 before (or instead of) the directory containing the original header.
24429 Then, if you want to check that the precompiled header file is always
24430 used, you can put a file of the same name as the original header in this
24431 directory containing an @code{#error} command.
24433 This also works with @option{-include}. So yet another way to use
24434 precompiled headers, good for projects not designed with precompiled
24435 header files in mind, is to simply take most of the header files used by
24436 a project, include them from another header file, precompile that header
24437 file, and @option{-include} the precompiled header. If the header files
24438 have guards against multiple inclusion, they are skipped because
24439 they've already been included (in the precompiled header).
24441 If you need to precompile the same header file for different
24442 languages, targets, or compiler options, you can instead make a
24443 @emph{directory} named like @file{all.h.gch}, and put each precompiled
24444 header in the directory, perhaps using @option{-o}. It doesn't matter
24445 what you call the files in the directory; every precompiled header in
24446 the directory is considered. The first precompiled header
24447 encountered in the directory that is valid for this compilation is
24448 used; they're searched in no particular order.
24450 There are many other possibilities, limited only by your imagination,
24451 good sense, and the constraints of your build system.
24453 A precompiled header file can be used only when these conditions apply:
24457 Only one precompiled header can be used in a particular compilation.
24460 A precompiled header can't be used once the first C token is seen. You
24461 can have preprocessor directives before a precompiled header; you cannot
24462 include a precompiled header from inside another header.
24465 The precompiled header file must be produced for the same language as
24466 the current compilation. You can't use a C precompiled header for a C++
24470 The precompiled header file must have been produced by the same compiler
24471 binary as the current compilation is using.
24474 Any macros defined before the precompiled header is included must
24475 either be defined in the same way as when the precompiled header was
24476 generated, or must not affect the precompiled header, which usually
24477 means that they don't appear in the precompiled header at all.
24479 The @option{-D} option is one way to define a macro before a
24480 precompiled header is included; using a @code{#define} can also do it.
24481 There are also some options that define macros implicitly, like
24482 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
24485 @item If debugging information is output when using the precompiled
24486 header, using @option{-g} or similar, the same kind of debugging information
24487 must have been output when building the precompiled header. However,
24488 a precompiled header built using @option{-g} can be used in a compilation
24489 when no debugging information is being output.
24491 @item The same @option{-m} options must generally be used when building
24492 and using the precompiled header. @xref{Submodel Options},
24493 for any cases where this rule is relaxed.
24495 @item Each of the following options must be the same when building and using
24496 the precompiled header:
24498 @gccoptlist{-fexceptions}
24501 Some other command-line options starting with @option{-f},
24502 @option{-p}, or @option{-O} must be defined in the same way as when
24503 the precompiled header was generated. At present, it's not clear
24504 which options are safe to change and which are not; the safest choice
24505 is to use exactly the same options when generating and using the
24506 precompiled header. The following are known to be safe:
24508 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
24509 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
24510 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
24515 For all of these except the last, the compiler automatically
24516 ignores the precompiled header if the conditions aren't met. If you
24517 find an option combination that doesn't work and doesn't cause the
24518 precompiled header to be ignored, please consider filing a bug report,
24521 If you do use differing options when generating and using the
24522 precompiled header, the actual behavior is a mixture of the
24523 behavior for the options. For instance, if you use @option{-g} to
24524 generate the precompiled header but not when using it, you may or may
24525 not get debugging information for routines in the precompiled header.