1 @c Copyright (C) 1988-2018 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{} 1988-2018 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), dbx(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.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
126 @xref{Option Index}, for an index to GCC's options.
129 * Option Summary:: Brief list of all options, without explanations.
130 * Overall Options:: Controlling the kind of output:
131 an executable, object files, assembler files,
132 or preprocessed source.
133 * Invoking G++:: Compiling C++ programs.
134 * C Dialect Options:: Controlling the variant of C language compiled.
135 * C++ Dialect Options:: Variations on C++.
136 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
138 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
140 * Warning Options:: How picky should the compiler be?
141 * Debugging Options:: Producing debuggable code.
142 * Optimize Options:: How much optimization?
143 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
144 * Preprocessor Options:: Controlling header files and macro definitions.
145 Also, getting dependency information for Make.
146 * Assembler Options:: Passing options to the assembler.
147 * Link Options:: Specifying libraries and so on.
148 * Directory Options:: Where to find header files and libraries.
149 Where to find the compiler executable files.
150 * Code Gen Options:: Specifying conventions for function calls, data layout
152 * Developer Options:: Printing GCC configuration info, statistics, and
154 * Submodel Options:: Target-specific options, such as compiling for a
155 specific processor variant.
156 * Spec Files:: How to pass switches to sub-processes.
157 * Environment Variables:: Env vars that affect GCC.
158 * Precompiled Headers:: Compiling a header once, and using it many times.
164 @section Option Summary
166 Here is a summary of all the options, grouped by type. Explanations are
167 in the following sections.
170 @item Overall Options
171 @xref{Overall Options,,Options Controlling the Kind of Output}.
172 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
173 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
174 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
175 @@@var{file} -ffile-prefix-map=@var{old}=@var{new} @gol
176 -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
177 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
179 @item C Language Options
180 @xref{C Dialect Options,,Options Controlling C Dialect}.
181 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
182 -fpermitted-flt-eval-methods=@var{standard} @gol
183 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
184 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
185 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
186 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
187 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
188 -fsigned-bitfields -fsigned-char @gol
189 -funsigned-bitfields -funsigned-char}
191 @item C++ Language Options
192 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
193 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
194 -faligned-new=@var{n} -fargs-in-order=@var{n} -fcheck-new @gol
195 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
196 -ffriend-injection @gol
197 -fno-elide-constructors @gol
198 -fno-enforce-eh-specs @gol
199 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
200 -fno-implicit-templates @gol
201 -fno-implicit-inline-templates @gol
202 -fno-implement-inlines -fms-extensions @gol
203 -fnew-inheriting-ctors @gol
204 -fnew-ttp-matching @gol
205 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
206 -fno-optional-diags -fpermissive @gol
207 -fno-pretty-templates @gol
208 -frepo -fno-rtti -fsized-deallocation @gol
209 -ftemplate-backtrace-limit=@var{n} @gol
210 -ftemplate-depth=@var{n} @gol
211 -fno-threadsafe-statics -fuse-cxa-atexit @gol
212 -fno-weak -nostdinc++ @gol
213 -fvisibility-inlines-hidden @gol
214 -fvisibility-ms-compat @gol
215 -fext-numeric-literals @gol
216 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
217 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
218 -Wnamespaces -Wnarrowing @gol
219 -Wnoexcept -Wnoexcept-type -Wclass-memaccess @gol
220 -Wnon-virtual-dtor -Wreorder -Wregister @gol
221 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
222 -Wno-non-template-friend -Wold-style-cast @gol
223 -Woverloaded-virtual -Wno-pmf-conversions @gol
224 -Wsign-promo -Wvirtual-inheritance}
226 @item Objective-C and Objective-C++ Language Options
227 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
228 Objective-C and Objective-C++ Dialects}.
229 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
230 -fgnu-runtime -fnext-runtime @gol
231 -fno-nil-receivers @gol
232 -fobjc-abi-version=@var{n} @gol
233 -fobjc-call-cxx-cdtors @gol
234 -fobjc-direct-dispatch @gol
235 -fobjc-exceptions @gol
238 -fobjc-std=objc1 @gol
239 -fno-local-ivars @gol
240 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
241 -freplace-objc-classes @gol
244 -Wassign-intercept @gol
245 -Wno-protocol -Wselector @gol
246 -Wstrict-selector-match @gol
247 -Wundeclared-selector}
249 @item Diagnostic Message Formatting Options
250 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
251 @gccoptlist{-fmessage-length=@var{n} @gol
252 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
253 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
254 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
255 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
256 -fdiagnostics-show-template-tree -fno-elide-type @gol
259 @item Warning Options
260 @xref{Warning Options,,Options to Request or Suppress Warnings}.
261 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
262 -pedantic-errors @gol
263 -w -Wextra -Wall -Waddress -Waggregate-return @gol
264 -Walloc-zero -Walloc-size-larger-than=@var{n}
265 -Walloca -Walloca-larger-than=@var{n} @gol
266 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
267 -Wno-attributes -Wbool-compare -Wbool-operation @gol
268 -Wno-builtin-declaration-mismatch @gol
269 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
270 -Wc++-compat -Wc++11-compat -Wc++14-compat @gol
271 -Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual @gol
272 -Wchar-subscripts -Wchkp -Wcatch-value -Wcatch-value=@var{n} @gol
273 -Wclobbered -Wcomment -Wconditionally-supported @gol
274 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
275 -Wdelete-incomplete @gol
276 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
277 -Wdisabled-optimization @gol
278 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
279 -Wno-div-by-zero -Wdouble-promotion @gol
280 -Wduplicated-branches -Wduplicated-cond @gol
281 -Wempty-body -Wenum-compare -Wno-endif-labels -Wexpansion-to-defined @gol
282 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
283 -Wfloat-equal -Wformat -Wformat=2 @gol
284 -Wno-format-contains-nul -Wno-format-extra-args @gol
285 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
286 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
287 -Wformat-y2k -Wframe-address @gol
288 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
289 -Wif-not-aligned @gol
290 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
291 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
292 -Wimplicit-function-declaration -Wimplicit-int @gol
293 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
294 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
295 -Winvalid-pch -Wlarger-than=@var{len} @gol
296 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
297 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
298 -Wmisleading-indentation -Wmissing-attributes -Wmissing-braces @gol
299 -Wmissing-field-initializers -Wmissing-include-dirs @gol
300 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol
301 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
302 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
303 -Woverride-init-side-effects -Woverlength-strings @gol
304 -Wpacked -Wpacked-bitfield-compat -Wpacked-not-aligned -Wpadded @gol
305 -Wparentheses -Wno-pedantic-ms-format @gol
306 -Wplacement-new -Wplacement-new=@var{n} @gol
307 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
308 -Wno-pragmas -Wredundant-decls -Wrestrict -Wno-return-local-addr @gol
309 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
310 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
311 -Wshift-overflow -Wshift-overflow=@var{n} @gol
312 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
313 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
314 -Wno-scalar-storage-order -Wsizeof-pointer-div @gol
315 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
316 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
317 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
318 -Wstringop-overflow=@var{n} -Wstringop-truncation @gol
319 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@r{]} @gol
320 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
321 -Wmissing-format-attribute -Wsubobject-linkage @gol
322 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
323 -Wswitch-unreachable -Wsync-nand @gol
324 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
325 -Wtype-limits -Wundef @gol
326 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
327 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
328 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
329 -Wunused-parameter -Wno-unused-result @gol
330 -Wunused-value -Wunused-variable @gol
331 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
332 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
333 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
334 -Wvla -Wvla-larger-than=@var{n} -Wvolatile-register-var -Wwrite-strings @gol
335 -Wzero-as-null-pointer-constant -Whsa}
337 @item C and Objective-C-only Warning Options
338 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
339 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
340 -Wold-style-declaration -Wold-style-definition @gol
341 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
342 -Wdeclaration-after-statement -Wpointer-sign}
344 @item Debugging Options
345 @xref{Debugging Options,,Options for Debugging Your Program}.
346 @gccoptlist{-g -g@var{level} -gdwarf -gdwarf-@var{version} @gol
347 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
348 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
349 -gas-loc-support -gno-as-loc-support @gol
350 -gas-locview-support -gno-as-locview-support @gol
351 -gcolumn-info -gno-column-info @gol
352 -gstatement-frontiers -gno-statement-frontiers @gol
353 -gvariable-location-views -gno-variable-location-views @gol
354 -ginternal-reset-location-views -gno-internal-reset-location-views @gol
355 -ginline-points -gno-inline-points @gol
356 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
357 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
358 -fno-eliminate-unused-debug-types @gol
359 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
360 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
361 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
362 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
363 -fvar-tracking -fvar-tracking-assignments}
365 @item Optimization Options
366 @xref{Optimize Options,,Options that Control Optimization}.
367 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
368 -falign-jumps[=@var{n}] @gol
369 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
370 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
371 -fauto-inc-dec -fbranch-probabilities @gol
372 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
373 -fbtr-bb-exclusive -fcaller-saves @gol
374 -fcombine-stack-adjustments -fconserve-stack @gol
375 -fcompare-elim -fcprop-registers -fcrossjumping @gol
376 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
377 -fcx-limited-range @gol
378 -fdata-sections -fdce -fdelayed-branch @gol
379 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
380 -fdevirtualize-at-ltrans -fdse @gol
381 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
382 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
383 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
384 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
385 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
386 -fif-conversion2 -findirect-inlining @gol
387 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
388 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
389 -fipa-bit-cp -fipa-vrp @gol
390 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
391 -fira-algorithm=@var{algorithm} @gol
392 -fira-region=@var{region} -fira-hoist-pressure @gol
393 -fira-loop-pressure -fno-ira-share-save-slots @gol
394 -fno-ira-share-spill-slots @gol
395 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
396 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
397 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
398 -floop-block -floop-interchange -floop-strip-mine @gol
399 -floop-unroll-and-jam -floop-nest-optimize @gol
400 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
401 -flto-partition=@var{alg} -fmerge-all-constants @gol
402 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
403 -fmove-loop-invariants -fno-branch-count-reg @gol
404 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
405 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
406 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
407 -fno-sched-spec -fno-signed-zeros @gol
408 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
409 -fomit-frame-pointer -foptimize-sibling-calls @gol
410 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
411 -fprefetch-loop-arrays @gol
412 -fprofile-correction @gol
413 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
414 -fprofile-reorder-functions @gol
415 -freciprocal-math -free -frename-registers -freorder-blocks @gol
416 -freorder-blocks-algorithm=@var{algorithm} @gol
417 -freorder-blocks-and-partition -freorder-functions @gol
418 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
419 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
420 -fsched-spec-load -fsched-spec-load-dangerous @gol
421 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
422 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
423 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
424 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
425 -fschedule-fusion @gol
426 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
427 -fselective-scheduling -fselective-scheduling2 @gol
428 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
429 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
430 -fsignaling-nans @gol
431 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
433 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
434 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
435 -fthread-jumps -ftracer -ftree-bit-ccp @gol
436 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
437 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
438 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
439 -ftree-loop-if-convert -ftree-loop-im @gol
440 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
441 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
442 -ftree-loop-vectorize @gol
443 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
444 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
445 -ftree-switch-conversion -ftree-tail-merge @gol
446 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
447 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
448 -funsafe-math-optimizations -funswitch-loops @gol
449 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
450 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
451 --param @var{name}=@var{value}
452 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
454 @item Program Instrumentation Options
455 @xref{Instrumentation Options,,Program Instrumentation Options}.
456 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
457 -fprofile-abs-path @gol
458 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
459 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
460 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
461 -fsanitize-undefined-trap-on-error -fbounds-check @gol
462 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
463 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
464 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
465 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
466 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
467 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
468 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
469 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
470 -fchkp-use-wrappers -fchkp-flexible-struct-trailing-arrays@gol
471 -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]} @gol
472 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
473 -fstack-protector-explicit -fstack-check @gol
474 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
475 -fno-stack-limit -fsplit-stack @gol
476 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
477 -fvtv-counts -fvtv-debug @gol
478 -finstrument-functions @gol
479 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
480 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
482 @item Preprocessor Options
483 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
484 @gccoptlist{-A@var{question}=@var{answer} @gol
485 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
486 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
487 -dD -dI -dM -dN -dU @gol
488 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
489 -fexec-charset=@var{charset} -fextended-identifiers @gol
490 -finput-charset=@var{charset} -fmacro-prefix-map=@var{old}=@var{new} @gol
491 -fno-canonical-system-headers @gol -fpch-deps -fpch-preprocess @gol
492 -fpreprocessed -ftabstop=@var{width} -ftrack-macro-expansion @gol
493 -fwide-exec-charset=@var{charset} -fworking-directory @gol
494 -H -imacros @var{file} -include @var{file} @gol
495 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
496 -no-integrated-cpp -P -pthread -remap @gol
497 -traditional -traditional-cpp -trigraphs @gol
498 -U@var{macro} -undef @gol
499 -Wp,@var{option} -Xpreprocessor @var{option}}
501 @item Assembler Options
502 @xref{Assembler Options,,Passing Options to the Assembler}.
503 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
506 @xref{Link Options,,Options for Linking}.
507 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
508 -nostartfiles -nodefaultlibs -nostdlib -pie -pthread -rdynamic @gol
509 -s -static -static-pie -static-libgcc -static-libstdc++ @gol
510 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
511 -static-libmpx -static-libmpxwrappers @gol
512 -shared -shared-libgcc -symbolic @gol
513 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
514 -u @var{symbol} -z @var{keyword}}
516 @item Directory Options
517 @xref{Directory Options,,Options for Directory Search}.
518 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
519 -idirafter @var{dir} @gol
520 -imacros @var{file} -imultilib @var{dir} @gol
521 -iplugindir=@var{dir} -iprefix @var{file} @gol
522 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
523 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
524 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
525 -nostdinc -nostdinc++ --sysroot=@var{dir}}
527 @item Code Generation Options
528 @xref{Code Gen Options,,Options for Code Generation Conventions}.
529 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
530 -ffixed-@var{reg} -fexceptions @gol
531 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
532 -fasynchronous-unwind-tables @gol
534 -finhibit-size-directive -fno-common -fno-ident @gol
535 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
536 -fno-jump-tables @gol
537 -frecord-gcc-switches @gol
538 -freg-struct-return -fshort-enums -fshort-wchar @gol
539 -fverbose-asm -fpack-struct[=@var{n}] @gol
540 -fleading-underscore -ftls-model=@var{model} @gol
541 -fstack-reuse=@var{reuse_level} @gol
542 -ftrampolines -ftrapv -fwrapv @gol
543 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
544 -fstrict-volatile-bitfields -fsync-libcalls}
546 @item Developer Options
547 @xref{Developer Options,,GCC Developer Options}.
548 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
549 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
550 -fdbg-cnt=@var{counter-value-list} @gol
551 -fdisable-ipa-@var{pass_name} @gol
552 -fdisable-rtl-@var{pass_name} @gol
553 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
554 -fdisable-tree-@var{pass_name} @gol
555 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
556 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
557 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
558 -fdump-final-insns@r{[}=@var{file}@r{]} @gol
559 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
561 -fdump-lang-@var{switch} @gol
562 -fdump-lang-@var{switch}-@var{options} @gol
563 -fdump-lang-@var{switch}-@var{options}=@var{filename} @gol
565 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
566 -fdump-statistics @gol
568 -fdump-tree-@var{switch} @gol
569 -fdump-tree-@var{switch}-@var{options} @gol
570 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
571 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
572 -fenable-@var{kind}-@var{pass} @gol
573 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
574 -fira-verbose=@var{n} @gol
575 -flto-report -flto-report-wpa -fmem-report-wpa @gol
576 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
577 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
578 -fprofile-report @gol
579 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
580 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
581 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
582 -fvar-tracking-assignments-toggle -gtoggle @gol
583 -print-file-name=@var{library} -print-libgcc-file-name @gol
584 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
585 -print-prog-name=@var{program} -print-search-dirs -Q @gol
586 -print-sysroot -print-sysroot-headers-suffix @gol
587 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
589 @item Machine-Dependent Options
590 @xref{Submodel Options,,Machine-Dependent Options}.
591 @c This list is ordered alphanumerically by subsection name.
592 @c Try and put the significant identifier (CPU or system) first,
593 @c so users have a clue at guessing where the ones they want will be.
595 @emph{AArch64 Options}
596 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
597 -mgeneral-regs-only @gol
598 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
600 -momit-leaf-frame-pointer @gol
601 -mtls-dialect=desc -mtls-dialect=traditional @gol
602 -mtls-size=@var{size} @gol
603 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol
604 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol
605 -mpc-relative-literal-loads @gol
606 -msign-return-address=@var{scope} @gol
607 -march=@var{name} -mcpu=@var{name} -mtune=@var{name} @gol
608 -moverride=@var{string} -mverbose-cost-dump}
610 @emph{Adapteva Epiphany Options}
611 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
612 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
613 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
614 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
615 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
616 -msplit-vecmove-early -m1reg-@var{reg}}
619 @gccoptlist{-mbarrel-shifter -mjli-always @gol
620 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
621 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
622 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
623 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
624 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
625 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
626 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol
627 -mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol
628 -mvolatile-cache -mtp-regno=@var{regno} @gol
629 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
630 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
631 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
632 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
633 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
634 -mtune=@var{cpu} -mmultcost=@var{num} @gol
635 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
636 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu} -mrf16}
639 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
640 -mabi=@var{name} @gol
641 -mapcs-stack-check -mno-apcs-stack-check @gol
642 -mapcs-reentrant -mno-apcs-reentrant @gol
643 -msched-prolog -mno-sched-prolog @gol
644 -mlittle-endian -mbig-endian @gol
646 -mfloat-abi=@var{name} @gol
647 -mfp16-format=@var{name}
648 -mthumb-interwork -mno-thumb-interwork @gol
649 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
650 -mtune=@var{name} -mprint-tune-info @gol
651 -mstructure-size-boundary=@var{n} @gol
652 -mabort-on-noreturn @gol
653 -mlong-calls -mno-long-calls @gol
654 -msingle-pic-base -mno-single-pic-base @gol
655 -mpic-register=@var{reg} @gol
656 -mnop-fun-dllimport @gol
657 -mpoke-function-name @gol
658 -mthumb -marm -mflip-thumb @gol
659 -mtpcs-frame -mtpcs-leaf-frame @gol
660 -mcaller-super-interworking -mcallee-super-interworking @gol
661 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
662 -mword-relocations @gol
663 -mfix-cortex-m3-ldrd @gol
664 -munaligned-access @gol
665 -mneon-for-64bits @gol
666 -mslow-flash-data @gol
667 -masm-syntax-unified @gol
669 -mverbose-cost-dump @gol
674 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
675 -mbranch-cost=@var{cost} @gol
676 -mcall-prologues -mgas-isr-prologues -mint8 @gol
677 -mn_flash=@var{size} -mno-interrupts @gol
678 -mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack @gol
679 -mfract-convert-truncate @gol
680 -mshort-calls -nodevicelib @gol
681 -Waddr-space-convert -Wmisspelled-isr}
683 @emph{Blackfin Options}
684 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
685 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
686 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
687 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
688 -mno-id-shared-library -mshared-library-id=@var{n} @gol
689 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
690 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
691 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
695 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
696 -msim -msdata=@var{sdata-type}}
699 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
700 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
701 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
702 -mstack-align -mdata-align -mconst-align @gol
703 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
704 -melf -maout -melinux -mlinux -sim -sim2 @gol
705 -mmul-bug-workaround -mno-mul-bug-workaround}
708 @gccoptlist{-mmac @gol
709 -mcr16cplus -mcr16c @gol
710 -msim -mint32 -mbit-ops
711 -mdata-model=@var{model}}
713 @emph{Darwin Options}
714 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
715 -arch_only -bind_at_load -bundle -bundle_loader @gol
716 -client_name -compatibility_version -current_version @gol
718 -dependency-file -dylib_file -dylinker_install_name @gol
719 -dynamic -dynamiclib -exported_symbols_list @gol
720 -filelist -flat_namespace -force_cpusubtype_ALL @gol
721 -force_flat_namespace -headerpad_max_install_names @gol
723 -image_base -init -install_name -keep_private_externs @gol
724 -multi_module -multiply_defined -multiply_defined_unused @gol
725 -noall_load -no_dead_strip_inits_and_terms @gol
726 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
727 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
728 -private_bundle -read_only_relocs -sectalign @gol
729 -sectobjectsymbols -whyload -seg1addr @gol
730 -sectcreate -sectobjectsymbols -sectorder @gol
731 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
732 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
733 -segprot -segs_read_only_addr -segs_read_write_addr @gol
734 -single_module -static -sub_library -sub_umbrella @gol
735 -twolevel_namespace -umbrella -undefined @gol
736 -unexported_symbols_list -weak_reference_mismatches @gol
737 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
738 -mkernel -mone-byte-bool}
740 @emph{DEC Alpha Options}
741 @gccoptlist{-mno-fp-regs -msoft-float @gol
742 -mieee -mieee-with-inexact -mieee-conformant @gol
743 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
744 -mtrap-precision=@var{mode} -mbuild-constants @gol
745 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
746 -mbwx -mmax -mfix -mcix @gol
747 -mfloat-vax -mfloat-ieee @gol
748 -mexplicit-relocs -msmall-data -mlarge-data @gol
749 -msmall-text -mlarge-text @gol
750 -mmemory-latency=@var{time}}
753 @gccoptlist{-msmall-model -mno-lsim}
756 @gccoptlist{-msim -mlra -mnodiv -mft32b -mcompress -mnopm}
759 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
760 -mhard-float -msoft-float @gol
761 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
762 -mdouble -mno-double @gol
763 -mmedia -mno-media -mmuladd -mno-muladd @gol
764 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
765 -mlinked-fp -mlong-calls -malign-labels @gol
766 -mlibrary-pic -macc-4 -macc-8 @gol
767 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
768 -moptimize-membar -mno-optimize-membar @gol
769 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
770 -mvliw-branch -mno-vliw-branch @gol
771 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
772 -mno-nested-cond-exec -mtomcat-stats @gol
776 @emph{GNU/Linux Options}
777 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
778 -tno-android-cc -tno-android-ld}
780 @emph{H8/300 Options}
781 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
784 @gccoptlist{-march=@var{architecture-type} @gol
785 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
786 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
787 -mfixed-range=@var{register-range} @gol
788 -mjump-in-delay -mlinker-opt -mlong-calls @gol
789 -mlong-load-store -mno-disable-fpregs @gol
790 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
791 -mno-jump-in-delay -mno-long-load-store @gol
792 -mno-portable-runtime -mno-soft-float @gol
793 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
794 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
795 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
796 -munix=@var{unix-std} -nolibdld -static -threads}
799 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
800 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
801 -mconstant-gp -mauto-pic -mfused-madd @gol
802 -minline-float-divide-min-latency @gol
803 -minline-float-divide-max-throughput @gol
804 -mno-inline-float-divide @gol
805 -minline-int-divide-min-latency @gol
806 -minline-int-divide-max-throughput @gol
807 -mno-inline-int-divide @gol
808 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
809 -mno-inline-sqrt @gol
810 -mdwarf2-asm -mearly-stop-bits @gol
811 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
812 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
813 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
814 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
815 -msched-spec-ldc -msched-spec-control-ldc @gol
816 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
817 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
818 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
819 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
822 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
823 -msign-extend-enabled -muser-enabled}
825 @emph{M32R/D Options}
826 @gccoptlist{-m32r2 -m32rx -m32r @gol
828 -malign-loops -mno-align-loops @gol
829 -missue-rate=@var{number} @gol
830 -mbranch-cost=@var{number} @gol
831 -mmodel=@var{code-size-model-type} @gol
832 -msdata=@var{sdata-type} @gol
833 -mno-flush-func -mflush-func=@var{name} @gol
834 -mno-flush-trap -mflush-trap=@var{number} @gol
838 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
840 @emph{M680x0 Options}
841 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
842 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
843 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
844 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
845 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
846 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
847 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
848 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
849 -mxgot -mno-xgot -mlong-jump-table-offsets}
852 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
853 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
854 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
855 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
856 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
859 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
860 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
861 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
862 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
865 @emph{MicroBlaze Options}
866 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
867 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
868 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
869 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
870 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
873 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
874 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
875 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
876 -mips16 -mno-mips16 -mflip-mips16 @gol
877 -minterlink-compressed -mno-interlink-compressed @gol
878 -minterlink-mips16 -mno-interlink-mips16 @gol
879 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
880 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
881 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
882 -mno-float -msingle-float -mdouble-float @gol
883 -modd-spreg -mno-odd-spreg @gol
884 -mabs=@var{mode} -mnan=@var{encoding} @gol
885 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
888 -mvirt -mno-virt @gol
890 -mmicromips -mno-micromips @gol
892 -mfpu=@var{fpu-type} @gol
893 -msmartmips -mno-smartmips @gol
894 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
895 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
896 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
897 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
898 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
899 -membedded-data -mno-embedded-data @gol
900 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
901 -mcode-readable=@var{setting} @gol
902 -msplit-addresses -mno-split-addresses @gol
903 -mexplicit-relocs -mno-explicit-relocs @gol
904 -mcheck-zero-division -mno-check-zero-division @gol
905 -mdivide-traps -mdivide-breaks @gol
906 -mload-store-pairs -mno-load-store-pairs @gol
907 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
908 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
909 -mfix-24k -mno-fix-24k @gol
910 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
911 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
912 -mfix-vr4120 -mno-fix-vr4120 @gol
913 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
914 -mflush-func=@var{func} -mno-flush-func @gol
915 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
916 -mcompact-branches=@var{policy} @gol
917 -mfp-exceptions -mno-fp-exceptions @gol
918 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
919 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
920 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
921 -mframe-header-opt -mno-frame-header-opt}
924 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
925 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
926 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
927 -mno-base-addresses -msingle-exit -mno-single-exit}
929 @emph{MN10300 Options}
930 @gccoptlist{-mmult-bug -mno-mult-bug @gol
931 -mno-am33 -mam33 -mam33-2 -mam34 @gol
932 -mtune=@var{cpu-type} @gol
933 -mreturn-pointer-on-d0 @gol
934 -mno-crt0 -mrelax -mliw -msetlb}
937 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
939 @emph{MSP430 Options}
940 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
942 -mcode-region= -mdata-region= @gol
943 -msilicon-errata= -msilicon-errata-warn= @gol
947 @gccoptlist{-mbig-endian -mlittle-endian @gol
948 -mreduced-regs -mfull-regs @gol
949 -mcmov -mno-cmov @gol
950 -mext-perf -mno-ext-perf @gol
951 -mext-perf2 -mno-ext-perf2 @gol
952 -mext-string -mno-ext-string @gol
953 -mv3push -mno-v3push @gol
954 -m16bit -mno-16bit @gol
955 -misr-vector-size=@var{num} @gol
956 -mcache-block-size=@var{num} @gol
957 -march=@var{arch} @gol
958 -mcmodel=@var{code-model} @gol
961 @emph{Nios II Options}
962 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
963 -mgprel-sec=@var{regexp} -mr0rel-sec=@var{regexp} @gol
965 -mno-bypass-cache -mbypass-cache @gol
966 -mno-cache-volatile -mcache-volatile @gol
967 -mno-fast-sw-div -mfast-sw-div @gol
968 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
969 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
970 -mcustom-fpu-cfg=@var{name} @gol
971 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
972 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
974 @emph{Nvidia PTX Options}
975 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
977 @emph{PDP-11 Options}
978 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
979 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
980 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
981 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
982 -mbranch-expensive -mbranch-cheap @gol
983 -munix-asm -mdec-asm}
985 @emph{picoChip Options}
986 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
987 -msymbol-as-address -mno-inefficient-warnings}
989 @emph{PowerPC Options}
990 See RS/6000 and PowerPC Options.
992 @emph{PowerPC SPE Options}
993 @gccoptlist{-mcpu=@var{cpu-type} @gol
994 -mtune=@var{cpu-type} @gol
995 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb @gol
996 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
997 -m32 -mxl-compat -mno-xl-compat @gol
998 -malign-power -malign-natural @gol
999 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1000 -msingle-float -mdouble-float @gol
1001 -mupdate -mno-update @gol
1002 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1003 -mstrict-align -mno-strict-align -mrelocatable @gol
1004 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1005 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1006 -msingle-pic-base @gol
1007 -mprioritize-restricted-insns=@var{priority} @gol
1008 -msched-costly-dep=@var{dependence_type} @gol
1009 -minsert-sched-nops=@var{scheme} @gol
1010 -mcall-sysv -mcall-netbsd @gol
1011 -maix-struct-return -msvr4-struct-return @gol
1012 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1013 -mblock-move-inline-limit=@var{num} @gol
1014 -misel -mno-isel @gol
1015 -misel=yes -misel=no @gol
1017 -mspe=yes -mspe=no @gol
1018 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1019 -mprototype -mno-prototype @gol
1020 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1021 -msdata=@var{opt} -mvxworks -G @var{num} @gol
1022 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1023 -mno-recip-precision @gol
1024 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1025 -msave-toc-indirect -mno-save-toc-indirect @gol
1026 -mcompat-align-parm -mno-compat-align-parm @gol
1027 -mfloat128 -mno-float128 @gol
1028 -mgnu-attribute -mno-gnu-attribute @gol
1029 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1030 -mstack-protector-guard-offset=@var{offset}}
1032 @emph{RISC-V Options}
1033 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
1035 -mabi=@var{ABI-string} @gol
1036 -mfdiv -mno-fdiv @gol
1038 -march=@var{ISA-string} @gol
1039 -mtune=@var{processor-string} @gol
1040 -mpreferred-stack-boundary=@var{num} @gol
1041 -msmall-data-limit=@var{N-bytes} @gol
1042 -msave-restore -mno-save-restore @gol
1043 -mstrict-align -mno-strict-align @gol
1044 -mcmodel=medlow -mcmodel=medany @gol
1045 -mexplicit-relocs -mno-explicit-relocs @gol
1046 -mrelax -mno-relax @gol}
1049 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
1050 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
1051 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
1053 @emph{RS/6000 and PowerPC Options}
1054 @gccoptlist{-mcpu=@var{cpu-type} @gol
1055 -mtune=@var{cpu-type} @gol
1056 -mcmodel=@var{code-model} @gol
1058 -maltivec -mno-altivec @gol
1059 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1060 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1061 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1062 -mfprnd -mno-fprnd @gol
1063 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
1064 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1065 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1066 -malign-power -malign-natural @gol
1067 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1068 -msingle-float -mdouble-float -msimple-fpu @gol
1069 -mupdate -mno-update @gol
1070 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1071 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1072 -mstrict-align -mno-strict-align -mrelocatable @gol
1073 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1074 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1075 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
1076 -mprioritize-restricted-insns=@var{priority} @gol
1077 -msched-costly-dep=@var{dependence_type} @gol
1078 -minsert-sched-nops=@var{scheme} @gol
1079 -mcall-aixdesc -mcall-eabi -mcall-freebsd @gol
1080 -mcall-linux -mcall-netbsd -mcall-openbsd @gol
1081 -mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi @gol
1082 -mtraceback=@var{traceback_type} @gol
1083 -maix-struct-return -msvr4-struct-return @gol
1084 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1085 -mblock-move-inline-limit=@var{num} @gol
1086 -mblock-compare-inline-limit=@var{num} @gol
1087 -mblock-compare-inline-loop-limit=@var{num} @gol
1088 -mstring-compare-inline-limit=@var{num} @gol
1089 -misel -mno-isel @gol
1090 -misel=yes -misel=no @gol
1092 -mvrsave -mno-vrsave @gol
1093 -mmulhw -mno-mulhw @gol
1094 -mdlmzb -mno-dlmzb @gol
1095 -mprototype -mno-prototype @gol
1096 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1097 -msdata=@var{opt} -mreadonly-in-sdata -mvxworks -G @var{num} @gol
1098 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1099 -mno-recip-precision @gol
1100 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1101 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1102 -msave-toc-indirect -mno-save-toc-indirect @gol
1103 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1104 -mcrypto -mno-crypto -mhtm -mno-htm @gol
1105 -mquad-memory -mno-quad-memory @gol
1106 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1107 -mcompat-align-parm -mno-compat-align-parm @gol
1108 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1109 -mgnu-attribute -mno-gnu-attribute @gol
1110 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1111 -mstack-protector-guard-offset=@var{offset}}
1114 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1116 -mbig-endian-data -mlittle-endian-data @gol
1119 -mas100-syntax -mno-as100-syntax@gol
1121 -mmax-constant-size=@gol
1124 -mallow-string-insns -mno-allow-string-insns@gol
1126 -mno-warn-multiple-fast-interrupts@gol
1127 -msave-acc-in-interrupts}
1129 @emph{S/390 and zSeries Options}
1130 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1131 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1132 -mlong-double-64 -mlong-double-128 @gol
1133 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1134 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1135 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1136 -mhtm -mvx -mzvector @gol
1137 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1138 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1139 -mhotpatch=@var{halfwords},@var{halfwords}}
1141 @emph{Score Options}
1142 @gccoptlist{-meb -mel @gol
1146 -mscore5 -mscore5u -mscore7 -mscore7d}
1149 @gccoptlist{-m1 -m2 -m2e @gol
1150 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1152 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1153 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1154 -mb -ml -mdalign -mrelax @gol
1155 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1156 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1157 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1158 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1159 -maccumulate-outgoing-args @gol
1160 -matomic-model=@var{atomic-model} @gol
1161 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1162 -mcbranch-force-delay-slot @gol
1163 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1164 -mpretend-cmove -mtas}
1166 @emph{Solaris 2 Options}
1167 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1170 @emph{SPARC Options}
1171 @gccoptlist{-mcpu=@var{cpu-type} @gol
1172 -mtune=@var{cpu-type} @gol
1173 -mcmodel=@var{code-model} @gol
1174 -mmemory-model=@var{mem-model} @gol
1175 -m32 -m64 -mapp-regs -mno-app-regs @gol
1176 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1177 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1178 -mhard-quad-float -msoft-quad-float @gol
1179 -mstack-bias -mno-stack-bias @gol
1180 -mstd-struct-return -mno-std-struct-return @gol
1181 -munaligned-doubles -mno-unaligned-doubles @gol
1182 -muser-mode -mno-user-mode @gol
1183 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1184 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1185 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1186 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1187 -mpopc -mno-popc -msubxc -mno-subxc @gol
1188 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1192 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1193 -msafe-dma -munsafe-dma @gol
1195 -msmall-mem -mlarge-mem -mstdmain @gol
1196 -mfixed-range=@var{register-range} @gol
1198 -maddress-space-conversion -mno-address-space-conversion @gol
1199 -mcache-size=@var{cache-size} @gol
1200 -matomic-updates -mno-atomic-updates}
1202 @emph{System V Options}
1203 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1205 @emph{TILE-Gx Options}
1206 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1207 -mcmodel=@var{code-model}}
1209 @emph{TILEPro Options}
1210 @gccoptlist{-mcpu=@var{cpu} -m32}
1213 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1214 -mprolog-function -mno-prolog-function -mspace @gol
1215 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1216 -mapp-regs -mno-app-regs @gol
1217 -mdisable-callt -mno-disable-callt @gol
1218 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1219 -mv850e -mv850 -mv850e3v5 @gol
1230 @gccoptlist{-mg -mgnu -munix}
1232 @emph{Visium Options}
1233 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1234 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1237 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1238 -mpointer-size=@var{size}}
1240 @emph{VxWorks Options}
1241 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1242 -Xbind-lazy -Xbind-now}
1245 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1246 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1247 -mfpmath=@var{unit} @gol
1248 -masm=@var{dialect} -mno-fancy-math-387 @gol
1249 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1250 -mno-wide-multiply -mrtd -malign-double @gol
1251 -mpreferred-stack-boundary=@var{num} @gol
1252 -mincoming-stack-boundary=@var{num} @gol
1253 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1254 -mrecip -mrecip=@var{opt} @gol
1255 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol
1256 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1257 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1258 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1259 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd @gol
1260 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1261 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1262 -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx @gol
1263 -mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes @gol
1264 -mcet -mibt -mshstk -mforce-indirect-call -mavx512vbmi2 @gol
1265 -mvpclmulqdq -mavx512bitalg -mavx512vpopcntdq @gol
1266 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1267 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1268 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1269 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1270 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1271 -mregparm=@var{num} -msseregparm @gol
1272 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1273 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1274 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1275 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1276 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1277 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1278 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1279 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1280 -mstack-protector-guard-reg=@var{reg} @gol
1281 -mstack-protector-guard-offset=@var{offset} @gol
1282 -mstack-protector-guard-symbol=@var{symbol} -mmitigate-rop @gol
1283 -mgeneral-regs-only -mcall-ms2sysv-xlogues @gol
1284 -mindirect-branch=@var{choice} -mfunction-return=@var{choice} @gol
1285 -mindirect-branch-register}
1287 @emph{x86 Windows Options}
1288 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1289 -mnop-fun-dllimport -mthread @gol
1290 -municode -mwin32 -mwindows -fno-set-stack-executable}
1292 @emph{Xstormy16 Options}
1295 @emph{Xtensa Options}
1296 @gccoptlist{-mconst16 -mno-const16 @gol
1297 -mfused-madd -mno-fused-madd @gol
1299 -mserialize-volatile -mno-serialize-volatile @gol
1300 -mtext-section-literals -mno-text-section-literals @gol
1301 -mauto-litpools -mno-auto-litpools @gol
1302 -mtarget-align -mno-target-align @gol
1303 -mlongcalls -mno-longcalls}
1305 @emph{zSeries Options}
1306 See S/390 and zSeries Options.
1310 @node Overall Options
1311 @section Options Controlling the Kind of Output
1313 Compilation can involve up to four stages: preprocessing, compilation
1314 proper, assembly and linking, always in that order. GCC is capable of
1315 preprocessing and compiling several files either into several
1316 assembler input files, or into one assembler input file; then each
1317 assembler input file produces an object file, and linking combines all
1318 the object files (those newly compiled, and those specified as input)
1319 into an executable file.
1321 @cindex file name suffix
1322 For any given input file, the file name suffix determines what kind of
1323 compilation is done:
1327 C source code that must be preprocessed.
1330 C source code that should not be preprocessed.
1333 C++ source code that should not be preprocessed.
1336 Objective-C source code. Note that you must link with the @file{libobjc}
1337 library to make an Objective-C program work.
1340 Objective-C source code that should not be preprocessed.
1344 Objective-C++ source code. Note that you must link with the @file{libobjc}
1345 library to make an Objective-C++ program work. Note that @samp{.M} refers
1346 to a literal capital M@.
1348 @item @var{file}.mii
1349 Objective-C++ source code that should not be preprocessed.
1352 C, C++, Objective-C or Objective-C++ header file to be turned into a
1353 precompiled header (default), or C, C++ header file to be turned into an
1354 Ada spec (via the @option{-fdump-ada-spec} switch).
1357 @itemx @var{file}.cp
1358 @itemx @var{file}.cxx
1359 @itemx @var{file}.cpp
1360 @itemx @var{file}.CPP
1361 @itemx @var{file}.c++
1363 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1364 the last two letters must both be literally @samp{x}. Likewise,
1365 @samp{.C} refers to a literal capital C@.
1369 Objective-C++ source code that must be preprocessed.
1371 @item @var{file}.mii
1372 Objective-C++ source code that should not be preprocessed.
1376 @itemx @var{file}.hp
1377 @itemx @var{file}.hxx
1378 @itemx @var{file}.hpp
1379 @itemx @var{file}.HPP
1380 @itemx @var{file}.h++
1381 @itemx @var{file}.tcc
1382 C++ header file to be turned into a precompiled header or Ada spec.
1385 @itemx @var{file}.for
1386 @itemx @var{file}.ftn
1387 Fixed form Fortran source code that should not be preprocessed.
1390 @itemx @var{file}.FOR
1391 @itemx @var{file}.fpp
1392 @itemx @var{file}.FPP
1393 @itemx @var{file}.FTN
1394 Fixed form Fortran source code that must be preprocessed (with the traditional
1397 @item @var{file}.f90
1398 @itemx @var{file}.f95
1399 @itemx @var{file}.f03
1400 @itemx @var{file}.f08
1401 Free form Fortran source code that should not be preprocessed.
1403 @item @var{file}.F90
1404 @itemx @var{file}.F95
1405 @itemx @var{file}.F03
1406 @itemx @var{file}.F08
1407 Free form Fortran source code that must be preprocessed (with the
1408 traditional preprocessor).
1413 @item @var{file}.brig
1414 BRIG files (binary representation of HSAIL).
1416 @item @var{file}.ads
1417 Ada source code file that contains a library unit declaration (a
1418 declaration of a package, subprogram, or generic, or a generic
1419 instantiation), or a library unit renaming declaration (a package,
1420 generic, or subprogram renaming declaration). Such files are also
1423 @item @var{file}.adb
1424 Ada source code file containing a library unit body (a subprogram or
1425 package body). Such files are also called @dfn{bodies}.
1427 @c GCC also knows about some suffixes for languages not yet included:
1438 @itemx @var{file}.sx
1439 Assembler code that must be preprocessed.
1442 An object file to be fed straight into linking.
1443 Any file name with no recognized suffix is treated this way.
1447 You can specify the input language explicitly with the @option{-x} option:
1450 @item -x @var{language}
1451 Specify explicitly the @var{language} for the following input files
1452 (rather than letting the compiler choose a default based on the file
1453 name suffix). This option applies to all following input files until
1454 the next @option{-x} option. Possible values for @var{language} are:
1456 c c-header cpp-output
1457 c++ c++-header c++-cpp-output
1458 objective-c objective-c-header objective-c-cpp-output
1459 objective-c++ objective-c++-header objective-c++-cpp-output
1460 assembler assembler-with-cpp
1462 f77 f77-cpp-input f95 f95-cpp-input
1468 Turn off any specification of a language, so that subsequent files are
1469 handled according to their file name suffixes (as they are if @option{-x}
1470 has not been used at all).
1473 If you only want some of the stages of compilation, you can use
1474 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1475 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1476 @command{gcc} is to stop. Note that some combinations (for example,
1477 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1482 Compile or assemble the source files, but do not link. The linking
1483 stage simply is not done. The ultimate output is in the form of an
1484 object file for each source file.
1486 By default, the object file name for a source file is made by replacing
1487 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1489 Unrecognized input files, not requiring compilation or assembly, are
1494 Stop after the stage of compilation proper; do not assemble. The output
1495 is in the form of an assembler code file for each non-assembler input
1498 By default, the assembler file name for a source file is made by
1499 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1501 Input files that don't require compilation are ignored.
1505 Stop after the preprocessing stage; do not run the compiler proper. The
1506 output is in the form of preprocessed source code, which is sent to the
1509 Input files that don't require preprocessing are ignored.
1511 @cindex output file option
1514 Place output in file @var{file}. This applies to whatever
1515 sort of output is being produced, whether it be an executable file,
1516 an object file, an assembler file or preprocessed C code.
1518 If @option{-o} is not specified, the default is to put an executable
1519 file in @file{a.out}, the object file for
1520 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1521 assembler file in @file{@var{source}.s}, a precompiled header file in
1522 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1527 Print (on standard error output) the commands executed to run the stages
1528 of compilation. Also print the version number of the compiler driver
1529 program and of the preprocessor and the compiler proper.
1533 Like @option{-v} except the commands are not executed and arguments
1534 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1535 This is useful for shell scripts to capture the driver-generated command lines.
1539 Print (on the standard output) a description of the command-line options
1540 understood by @command{gcc}. If the @option{-v} option is also specified
1541 then @option{--help} is also passed on to the various processes
1542 invoked by @command{gcc}, so that they can display the command-line options
1543 they accept. If the @option{-Wextra} option has also been specified
1544 (prior to the @option{--help} option), then command-line options that
1545 have no documentation associated with them are also displayed.
1548 @opindex target-help
1549 Print (on the standard output) a description of target-specific command-line
1550 options for each tool. For some targets extra target-specific
1551 information may also be printed.
1553 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1554 Print (on the standard output) a description of the command-line
1555 options understood by the compiler that fit into all specified classes
1556 and qualifiers. These are the supported classes:
1559 @item @samp{optimizers}
1560 Display all of the optimization options supported by the
1563 @item @samp{warnings}
1564 Display all of the options controlling warning messages
1565 produced by the compiler.
1568 Display target-specific options. Unlike the
1569 @option{--target-help} option however, target-specific options of the
1570 linker and assembler are not displayed. This is because those
1571 tools do not currently support the extended @option{--help=} syntax.
1574 Display the values recognized by the @option{--param}
1577 @item @var{language}
1578 Display the options supported for @var{language}, where
1579 @var{language} is the name of one of the languages supported in this
1583 Display the options that are common to all languages.
1586 These are the supported qualifiers:
1589 @item @samp{undocumented}
1590 Display only those options that are undocumented.
1593 Display options taking an argument that appears after an equal
1594 sign in the same continuous piece of text, such as:
1595 @samp{--help=target}.
1597 @item @samp{separate}
1598 Display options taking an argument that appears as a separate word
1599 following the original option, such as: @samp{-o output-file}.
1602 Thus for example to display all the undocumented target-specific
1603 switches supported by the compiler, use:
1606 --help=target,undocumented
1609 The sense of a qualifier can be inverted by prefixing it with the
1610 @samp{^} character, so for example to display all binary warning
1611 options (i.e., ones that are either on or off and that do not take an
1612 argument) that have a description, use:
1615 --help=warnings,^joined,^undocumented
1618 The argument to @option{--help=} should not consist solely of inverted
1621 Combining several classes is possible, although this usually
1622 restricts the output so much that there is nothing to display. One
1623 case where it does work, however, is when one of the classes is
1624 @var{target}. For example, to display all the target-specific
1625 optimization options, use:
1628 --help=target,optimizers
1631 The @option{--help=} option can be repeated on the command line. Each
1632 successive use displays its requested class of options, skipping
1633 those that have already been displayed.
1635 If the @option{-Q} option appears on the command line before the
1636 @option{--help=} option, then the descriptive text displayed by
1637 @option{--help=} is changed. Instead of describing the displayed
1638 options, an indication is given as to whether the option is enabled,
1639 disabled or set to a specific value (assuming that the compiler
1640 knows this at the point where the @option{--help=} option is used).
1642 Here is a truncated example from the ARM port of @command{gcc}:
1645 % gcc -Q -mabi=2 --help=target -c
1646 The following options are target specific:
1648 -mabort-on-noreturn [disabled]
1652 The output is sensitive to the effects of previous command-line
1653 options, so for example it is possible to find out which optimizations
1654 are enabled at @option{-O2} by using:
1657 -Q -O2 --help=optimizers
1660 Alternatively you can discover which binary optimizations are enabled
1661 by @option{-O3} by using:
1664 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1665 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1666 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1671 Display the version number and copyrights of the invoked GCC@.
1673 @item -pass-exit-codes
1674 @opindex pass-exit-codes
1675 Normally the @command{gcc} program exits with the code of 1 if any
1676 phase of the compiler returns a non-success return code. If you specify
1677 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1678 the numerically highest error produced by any phase returning an error
1679 indication. The C, C++, and Fortran front ends return 4 if an internal
1680 compiler error is encountered.
1684 Use pipes rather than temporary files for communication between the
1685 various stages of compilation. This fails to work on some systems where
1686 the assembler is unable to read from a pipe; but the GNU assembler has
1689 @item -specs=@var{file}
1691 Process @var{file} after the compiler reads in the standard @file{specs}
1692 file, in order to override the defaults which the @command{gcc} driver
1693 program uses when determining what switches to pass to @command{cc1},
1694 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1695 @option{-specs=@var{file}} can be specified on the command line, and they
1696 are processed in order, from left to right. @xref{Spec Files}, for
1697 information about the format of the @var{file}.
1701 Invoke all subcommands under a wrapper program. The name of the
1702 wrapper program and its parameters are passed as a comma separated
1706 gcc -c t.c -wrapper gdb,--args
1710 This invokes all subprograms of @command{gcc} under
1711 @samp{gdb --args}, thus the invocation of @command{cc1} is
1712 @samp{gdb --args cc1 @dots{}}.
1714 @item -ffile-prefix-map=@var{old}=@var{new}
1715 @opindex ffile-prefix-map
1716 When compiling files residing in directory @file{@var{old}}, record
1717 any references to them in the result of the compilation as if the
1718 files resided in directory @file{@var{new}} instead. Specifying this
1719 option is equivalent to specifying all the individual
1720 @option{-f*-prefix-map} options. This can be used to make reproducible
1721 builds that are location independent. See also
1722 @option{-fmacro-prefix-map} and @option{-fdebug-prefix-map}.
1724 @item -fplugin=@var{name}.so
1726 Load the plugin code in file @var{name}.so, assumed to be a
1727 shared object to be dlopen'd by the compiler. The base name of
1728 the shared object file is used to identify the plugin for the
1729 purposes of argument parsing (See
1730 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1731 Each plugin should define the callback functions specified in the
1734 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1735 @opindex fplugin-arg
1736 Define an argument called @var{key} with a value of @var{value}
1737 for the plugin called @var{name}.
1739 @item -fdump-ada-spec@r{[}-slim@r{]}
1740 @opindex fdump-ada-spec
1741 For C and C++ source and include files, generate corresponding Ada specs.
1742 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1743 GNAT User's Guide}, which provides detailed documentation on this feature.
1745 @item -fada-spec-parent=@var{unit}
1746 @opindex fada-spec-parent
1747 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1748 Ada specs as child units of parent @var{unit}.
1750 @item -fdump-go-spec=@var{file}
1751 @opindex fdump-go-spec
1752 For input files in any language, generate corresponding Go
1753 declarations in @var{file}. This generates Go @code{const},
1754 @code{type}, @code{var}, and @code{func} declarations which may be a
1755 useful way to start writing a Go interface to code written in some
1758 @include @value{srcdir}/../libiberty/at-file.texi
1762 @section Compiling C++ Programs
1764 @cindex suffixes for C++ source
1765 @cindex C++ source file suffixes
1766 C++ source files conventionally use one of the suffixes @samp{.C},
1767 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1768 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1769 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1770 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1771 files with these names and compiles them as C++ programs even if you
1772 call the compiler the same way as for compiling C programs (usually
1773 with the name @command{gcc}).
1777 However, the use of @command{gcc} does not add the C++ library.
1778 @command{g++} is a program that calls GCC and automatically specifies linking
1779 against the C++ library. It treats @samp{.c},
1780 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1781 files unless @option{-x} is used. This program is also useful when
1782 precompiling a C header file with a @samp{.h} extension for use in C++
1783 compilations. On many systems, @command{g++} is also installed with
1784 the name @command{c++}.
1786 @cindex invoking @command{g++}
1787 When you compile C++ programs, you may specify many of the same
1788 command-line options that you use for compiling programs in any
1789 language; or command-line options meaningful for C and related
1790 languages; or options that are meaningful only for C++ programs.
1791 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1792 explanations of options for languages related to C@.
1793 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1794 explanations of options that are meaningful only for C++ programs.
1796 @node C Dialect Options
1797 @section Options Controlling C Dialect
1798 @cindex dialect options
1799 @cindex language dialect options
1800 @cindex options, dialect
1802 The following options control the dialect of C (or languages derived
1803 from C, such as C++, Objective-C and Objective-C++) that the compiler
1807 @cindex ANSI support
1811 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1812 equivalent to @option{-std=c++98}.
1814 This turns off certain features of GCC that are incompatible with ISO
1815 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1816 such as the @code{asm} and @code{typeof} keywords, and
1817 predefined macros such as @code{unix} and @code{vax} that identify the
1818 type of system you are using. It also enables the undesirable and
1819 rarely used ISO trigraph feature. For the C compiler,
1820 it disables recognition of C++ style @samp{//} comments as well as
1821 the @code{inline} keyword.
1823 The alternate keywords @code{__asm__}, @code{__extension__},
1824 @code{__inline__} and @code{__typeof__} continue to work despite
1825 @option{-ansi}. You would not want to use them in an ISO C program, of
1826 course, but it is useful to put them in header files that might be included
1827 in compilations done with @option{-ansi}. Alternate predefined macros
1828 such as @code{__unix__} and @code{__vax__} are also available, with or
1829 without @option{-ansi}.
1831 The @option{-ansi} option does not cause non-ISO programs to be
1832 rejected gratuitously. For that, @option{-Wpedantic} is required in
1833 addition to @option{-ansi}. @xref{Warning Options}.
1835 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1836 option is used. Some header files may notice this macro and refrain
1837 from declaring certain functions or defining certain macros that the
1838 ISO standard doesn't call for; this is to avoid interfering with any
1839 programs that might use these names for other things.
1841 Functions that are normally built in but do not have semantics
1842 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1843 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1844 built-in functions provided by GCC}, for details of the functions
1849 Determine the language standard. @xref{Standards,,Language Standards
1850 Supported by GCC}, for details of these standard versions. This option
1851 is currently only supported when compiling C or C++.
1853 The compiler can accept several base standards, such as @samp{c90} or
1854 @samp{c++98}, and GNU dialects of those standards, such as
1855 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1856 compiler accepts all programs following that standard plus those
1857 using GNU extensions that do not contradict it. For example,
1858 @option{-std=c90} turns off certain features of GCC that are
1859 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1860 keywords, but not other GNU extensions that do not have a meaning in
1861 ISO C90, such as omitting the middle term of a @code{?:}
1862 expression. On the other hand, when a GNU dialect of a standard is
1863 specified, all features supported by the compiler are enabled, even when
1864 those features change the meaning of the base standard. As a result, some
1865 strict-conforming programs may be rejected. The particular standard
1866 is used by @option{-Wpedantic} to identify which features are GNU
1867 extensions given that version of the standard. For example
1868 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1869 comments, while @option{-std=gnu99 -Wpedantic} does not.
1871 A value for this option must be provided; possible values are
1877 Support all ISO C90 programs (certain GNU extensions that conflict
1878 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1880 @item iso9899:199409
1881 ISO C90 as modified in amendment 1.
1887 ISO C99. This standard is substantially completely supported, modulo
1888 bugs and floating-point issues
1889 (mainly but not entirely relating to optional C99 features from
1890 Annexes F and G). See
1891 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1892 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1897 ISO C11, the 2011 revision of the ISO C standard. This standard is
1898 substantially completely supported, modulo bugs, floating-point issues
1899 (mainly but not entirely relating to optional C11 features from
1900 Annexes F and G) and the optional Annexes K (Bounds-checking
1901 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1907 ISO C17, the 2017 revision of the ISO C standard (expected to be
1908 published in 2018). This standard is
1909 same as C11 except for corrections of defects (all of which are also
1910 applied with @option{-std=c11}) and a new value of
1911 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1915 GNU dialect of ISO C90 (including some C99 features).
1919 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1923 GNU dialect of ISO C11.
1924 The name @samp{gnu1x} is deprecated.
1928 GNU dialect of ISO C17. This is the default for C code.
1932 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1933 additional defect reports. Same as @option{-ansi} for C++ code.
1937 GNU dialect of @option{-std=c++98}.
1941 The 2011 ISO C++ standard plus amendments.
1942 The name @samp{c++0x} is deprecated.
1946 GNU dialect of @option{-std=c++11}.
1947 The name @samp{gnu++0x} is deprecated.
1951 The 2014 ISO C++ standard plus amendments.
1952 The name @samp{c++1y} is deprecated.
1956 GNU dialect of @option{-std=c++14}.
1957 This is the default for C++ code.
1958 The name @samp{gnu++1y} is deprecated.
1962 The 2017 ISO C++ standard plus amendments.
1963 The name @samp{c++1z} is deprecated.
1967 GNU dialect of @option{-std=c++17}.
1968 The name @samp{gnu++1z} is deprecated.
1971 The next revision of the ISO C++ standard, tentatively planned for
1972 2020. Support is highly experimental, and will almost certainly
1973 change in incompatible ways in future releases.
1976 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
1977 and will almost certainly change in incompatible ways in future
1981 @item -fgnu89-inline
1982 @opindex fgnu89-inline
1983 The option @option{-fgnu89-inline} tells GCC to use the traditional
1984 GNU semantics for @code{inline} functions when in C99 mode.
1985 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1986 Using this option is roughly equivalent to adding the
1987 @code{gnu_inline} function attribute to all inline functions
1988 (@pxref{Function Attributes}).
1990 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1991 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1992 specifies the default behavior).
1993 This option is not supported in @option{-std=c90} or
1994 @option{-std=gnu90} mode.
1996 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1997 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1998 in effect for @code{inline} functions. @xref{Common Predefined
1999 Macros,,,cpp,The C Preprocessor}.
2001 @item -fpermitted-flt-eval-methods=@var{style}
2002 @opindex fpermitted-flt-eval-methods
2003 @opindex fpermitted-flt-eval-methods=c11
2004 @opindex fpermitted-flt-eval-methods=ts-18661-3
2005 ISO/IEC TS 18661-3 defines new permissible values for
2006 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
2007 a semantic type that is an interchange or extended format should be
2008 evaluated to the precision and range of that type. These new values are
2009 a superset of those permitted under C99/C11, which does not specify the
2010 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
2011 conforming to C11 may not have been written expecting the possibility of
2014 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
2015 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
2016 or the extended set of values specified in ISO/IEC TS 18661-3.
2018 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
2020 The default when in a standards compliant mode (@option{-std=c11} or similar)
2021 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
2022 dialect (@option{-std=gnu11} or similar) is
2023 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
2025 @item -aux-info @var{filename}
2027 Output to the given filename prototyped declarations for all functions
2028 declared and/or defined in a translation unit, including those in header
2029 files. This option is silently ignored in any language other than C@.
2031 Besides declarations, the file indicates, in comments, the origin of
2032 each declaration (source file and line), whether the declaration was
2033 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
2034 @samp{O} for old, respectively, in the first character after the line
2035 number and the colon), and whether it came from a declaration or a
2036 definition (@samp{C} or @samp{F}, respectively, in the following
2037 character). In the case of function definitions, a K&R-style list of
2038 arguments followed by their declarations is also provided, inside
2039 comments, after the declaration.
2041 @item -fallow-parameterless-variadic-functions
2042 @opindex fallow-parameterless-variadic-functions
2043 Accept variadic functions without named parameters.
2045 Although it is possible to define such a function, this is not very
2046 useful as it is not possible to read the arguments. This is only
2047 supported for C as this construct is allowed by C++.
2051 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
2052 keyword, so that code can use these words as identifiers. You can use
2053 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
2054 instead. @option{-ansi} implies @option{-fno-asm}.
2056 In C++, this switch only affects the @code{typeof} keyword, since
2057 @code{asm} and @code{inline} are standard keywords. You may want to
2058 use the @option{-fno-gnu-keywords} flag instead, which has the same
2059 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
2060 switch only affects the @code{asm} and @code{typeof} keywords, since
2061 @code{inline} is a standard keyword in ISO C99.
2064 @itemx -fno-builtin-@var{function}
2065 @opindex fno-builtin
2066 @cindex built-in functions
2067 Don't recognize built-in functions that do not begin with
2068 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
2069 functions provided by GCC}, for details of the functions affected,
2070 including those which are not built-in functions when @option{-ansi} or
2071 @option{-std} options for strict ISO C conformance are used because they
2072 do not have an ISO standard meaning.
2074 GCC normally generates special code to handle certain built-in functions
2075 more efficiently; for instance, calls to @code{alloca} may become single
2076 instructions which adjust the stack directly, and calls to @code{memcpy}
2077 may become inline copy loops. The resulting code is often both smaller
2078 and faster, but since the function calls no longer appear as such, you
2079 cannot set a breakpoint on those calls, nor can you change the behavior
2080 of the functions by linking with a different library. In addition,
2081 when a function is recognized as a built-in function, GCC may use
2082 information about that function to warn about problems with calls to
2083 that function, or to generate more efficient code, even if the
2084 resulting code still contains calls to that function. For example,
2085 warnings are given with @option{-Wformat} for bad calls to
2086 @code{printf} when @code{printf} is built in and @code{strlen} is
2087 known not to modify global memory.
2089 With the @option{-fno-builtin-@var{function}} option
2090 only the built-in function @var{function} is
2091 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2092 function is named that is not built-in in this version of GCC, this
2093 option is ignored. There is no corresponding
2094 @option{-fbuiltin-@var{function}} option; if you wish to enable
2095 built-in functions selectively when using @option{-fno-builtin} or
2096 @option{-ffreestanding}, you may define macros such as:
2099 #define abs(n) __builtin_abs ((n))
2100 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2106 Enable parsing of function definitions marked with @code{__GIMPLE}.
2107 This is an experimental feature that allows unit testing of GIMPLE
2112 @cindex hosted environment
2114 Assert that compilation targets a hosted environment. This implies
2115 @option{-fbuiltin}. A hosted environment is one in which the
2116 entire standard library is available, and in which @code{main} has a return
2117 type of @code{int}. Examples are nearly everything except a kernel.
2118 This is equivalent to @option{-fno-freestanding}.
2120 @item -ffreestanding
2121 @opindex ffreestanding
2122 @cindex hosted environment
2124 Assert that compilation targets a freestanding environment. This
2125 implies @option{-fno-builtin}. A freestanding environment
2126 is one in which the standard library may not exist, and program startup may
2127 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2128 This is equivalent to @option{-fno-hosted}.
2130 @xref{Standards,,Language Standards Supported by GCC}, for details of
2131 freestanding and hosted environments.
2135 @cindex OpenACC accelerator programming
2136 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2137 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2138 compiler generates accelerated code according to the OpenACC Application
2139 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2140 implies @option{-pthread}, and thus is only supported on targets that
2141 have support for @option{-pthread}.
2143 @item -fopenacc-dim=@var{geom}
2144 @opindex fopenacc-dim
2145 @cindex OpenACC accelerator programming
2146 Specify default compute dimensions for parallel offload regions that do
2147 not explicitly specify. The @var{geom} value is a triple of
2148 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2149 can be omitted, to use a target-specific default value.
2153 @cindex OpenMP parallel
2154 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2155 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2156 compiler generates parallel code according to the OpenMP Application
2157 Program Interface v4.5 @w{@uref{http://www.openmp.org/}}. This option
2158 implies @option{-pthread}, and thus is only supported on targets that
2159 have support for @option{-pthread}. @option{-fopenmp} implies
2160 @option{-fopenmp-simd}.
2163 @opindex fopenmp-simd
2166 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2167 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2172 When the option @option{-fgnu-tm} is specified, the compiler
2173 generates code for the Linux variant of Intel's current Transactional
2174 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2175 an experimental feature whose interface may change in future versions
2176 of GCC, as the official specification changes. Please note that not
2177 all architectures are supported for this feature.
2179 For more information on GCC's support for transactional memory,
2180 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2181 Transactional Memory Library}.
2183 Note that the transactional memory feature is not supported with
2184 non-call exceptions (@option{-fnon-call-exceptions}).
2186 @item -fms-extensions
2187 @opindex fms-extensions
2188 Accept some non-standard constructs used in Microsoft header files.
2190 In C++ code, this allows member names in structures to be similar
2191 to previous types declarations.
2200 Some cases of unnamed fields in structures and unions are only
2201 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2202 fields within structs/unions}, for details.
2204 Note that this option is off for all targets but x86
2205 targets using ms-abi.
2207 @item -fplan9-extensions
2208 @opindex fplan9-extensions
2209 Accept some non-standard constructs used in Plan 9 code.
2211 This enables @option{-fms-extensions}, permits passing pointers to
2212 structures with anonymous fields to functions that expect pointers to
2213 elements of the type of the field, and permits referring to anonymous
2214 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2215 struct/union fields within structs/unions}, for details. This is only
2216 supported for C, not C++.
2218 @item -fcond-mismatch
2219 @opindex fcond-mismatch
2220 Allow conditional expressions with mismatched types in the second and
2221 third arguments. The value of such an expression is void. This option
2222 is not supported for C++.
2224 @item -flax-vector-conversions
2225 @opindex flax-vector-conversions
2226 Allow implicit conversions between vectors with differing numbers of
2227 elements and/or incompatible element types. This option should not be
2230 @item -funsigned-char
2231 @opindex funsigned-char
2232 Let the type @code{char} be unsigned, like @code{unsigned char}.
2234 Each kind of machine has a default for what @code{char} should
2235 be. It is either like @code{unsigned char} by default or like
2236 @code{signed char} by default.
2238 Ideally, a portable program should always use @code{signed char} or
2239 @code{unsigned char} when it depends on the signedness of an object.
2240 But many programs have been written to use plain @code{char} and
2241 expect it to be signed, or expect it to be unsigned, depending on the
2242 machines they were written for. This option, and its inverse, let you
2243 make such a program work with the opposite default.
2245 The type @code{char} is always a distinct type from each of
2246 @code{signed char} or @code{unsigned char}, even though its behavior
2247 is always just like one of those two.
2250 @opindex fsigned-char
2251 Let the type @code{char} be signed, like @code{signed char}.
2253 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2254 the negative form of @option{-funsigned-char}. Likewise, the option
2255 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2257 @item -fsigned-bitfields
2258 @itemx -funsigned-bitfields
2259 @itemx -fno-signed-bitfields
2260 @itemx -fno-unsigned-bitfields
2261 @opindex fsigned-bitfields
2262 @opindex funsigned-bitfields
2263 @opindex fno-signed-bitfields
2264 @opindex fno-unsigned-bitfields
2265 These options control whether a bit-field is signed or unsigned, when the
2266 declaration does not use either @code{signed} or @code{unsigned}. By
2267 default, such a bit-field is signed, because this is consistent: the
2268 basic integer types such as @code{int} are signed types.
2270 @item -fsso-struct=@var{endianness}
2271 @opindex fsso-struct
2272 Set the default scalar storage order of structures and unions to the
2273 specified endianness. The accepted values are @samp{big-endian},
2274 @samp{little-endian} and @samp{native} for the native endianness of
2275 the target (the default). This option is not supported for C++.
2277 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2278 code that is not binary compatible with code generated without it if the
2279 specified endianness is not the native endianness of the target.
2282 @node C++ Dialect Options
2283 @section Options Controlling C++ Dialect
2285 @cindex compiler options, C++
2286 @cindex C++ options, command-line
2287 @cindex options, C++
2288 This section describes the command-line options that are only meaningful
2289 for C++ programs. You can also use most of the GNU compiler options
2290 regardless of what language your program is in. For example, you
2291 might compile a file @file{firstClass.C} like this:
2294 g++ -g -fstrict-enums -O -c firstClass.C
2298 In this example, only @option{-fstrict-enums} is an option meant
2299 only for C++ programs; you can use the other options with any
2300 language supported by GCC@.
2302 Some options for compiling C programs, such as @option{-std}, are also
2303 relevant for C++ programs.
2304 @xref{C Dialect Options,,Options Controlling C Dialect}.
2306 Here is a list of options that are @emph{only} for compiling C++ programs:
2310 @item -fabi-version=@var{n}
2311 @opindex fabi-version
2312 Use version @var{n} of the C++ ABI@. The default is version 0.
2314 Version 0 refers to the version conforming most closely to
2315 the C++ ABI specification. Therefore, the ABI obtained using version 0
2316 will change in different versions of G++ as ABI bugs are fixed.
2318 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2320 Version 2 is the version of the C++ ABI that first appeared in G++
2321 3.4, and was the default through G++ 4.9.
2323 Version 3 corrects an error in mangling a constant address as a
2326 Version 4, which first appeared in G++ 4.5, implements a standard
2327 mangling for vector types.
2329 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2330 attribute const/volatile on function pointer types, decltype of a
2331 plain decl, and use of a function parameter in the declaration of
2334 Version 6, which first appeared in G++ 4.7, corrects the promotion
2335 behavior of C++11 scoped enums and the mangling of template argument
2336 packs, const/static_cast, prefix ++ and --, and a class scope function
2337 used as a template argument.
2339 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2340 builtin type and corrects the mangling of lambdas in default argument
2343 Version 8, which first appeared in G++ 4.9, corrects the substitution
2344 behavior of function types with function-cv-qualifiers.
2346 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2349 Version 10, which first appeared in G++ 6.1, adds mangling of
2350 attributes that affect type identity, such as ia32 calling convention
2351 attributes (e.g. @samp{stdcall}).
2353 Version 11, which first appeared in G++ 7, corrects the mangling of
2354 sizeof... expressions and operator names. For multiple entities with
2355 the same name within a function, that are declared in different scopes,
2356 the mangling now changes starting with the twelfth occurrence. It also
2357 implies @option{-fnew-inheriting-ctors}.
2359 See also @option{-Wabi}.
2361 @item -fabi-compat-version=@var{n}
2362 @opindex fabi-compat-version
2363 On targets that support strong aliases, G++
2364 works around mangling changes by creating an alias with the correct
2365 mangled name when defining a symbol with an incorrect mangled name.
2366 This switch specifies which ABI version to use for the alias.
2368 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2369 compatibility). If another ABI version is explicitly selected, this
2370 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2371 use @option{-fabi-compat-version=2}.
2373 If this option is not provided but @option{-Wabi=@var{n}} is, that
2374 version is used for compatibility aliases. If this option is provided
2375 along with @option{-Wabi} (without the version), the version from this
2376 option is used for the warning.
2378 @item -fno-access-control
2379 @opindex fno-access-control
2380 Turn off all access checking. This switch is mainly useful for working
2381 around bugs in the access control code.
2384 @opindex faligned-new
2385 Enable support for C++17 @code{new} of types that require more
2386 alignment than @code{void* ::operator new(std::size_t)} provides. A
2387 numeric argument such as @code{-faligned-new=32} can be used to
2388 specify how much alignment (in bytes) is provided by that function,
2389 but few users will need to override the default of
2390 @code{alignof(std::max_align_t)}.
2392 This flag is enabled by default for @option{-std=c++17}.
2396 Check that the pointer returned by @code{operator new} is non-null
2397 before attempting to modify the storage allocated. This check is
2398 normally unnecessary because the C++ standard specifies that
2399 @code{operator new} only returns @code{0} if it is declared
2400 @code{throw()}, in which case the compiler always checks the
2401 return value even without this option. In all other cases, when
2402 @code{operator new} has a non-empty exception specification, memory
2403 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2404 @samp{new (nothrow)}.
2408 Enable support for the C++ Extensions for Concepts Technical
2409 Specification, ISO 19217 (2015), which allows code like
2412 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2413 template <Addable T> T add (T a, T b) @{ return a + b; @}
2416 @item -fconstexpr-depth=@var{n}
2417 @opindex fconstexpr-depth
2418 Set the maximum nested evaluation depth for C++11 constexpr functions
2419 to @var{n}. A limit is needed to detect endless recursion during
2420 constant expression evaluation. The minimum specified by the standard
2423 @item -fconstexpr-loop-limit=@var{n}
2424 @opindex fconstexpr-loop-limit
2425 Set the maximum number of iterations for a loop in C++14 constexpr functions
2426 to @var{n}. A limit is needed to detect infinite loops during
2427 constant expression evaluation. The default is 262144 (1<<18).
2429 @item -fdeduce-init-list
2430 @opindex fdeduce-init-list
2431 Enable deduction of a template type parameter as
2432 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2435 template <class T> auto forward(T t) -> decltype (realfn (t))
2442 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2446 This deduction was implemented as a possible extension to the
2447 originally proposed semantics for the C++11 standard, but was not part
2448 of the final standard, so it is disabled by default. This option is
2449 deprecated, and may be removed in a future version of G++.
2451 @item -ffriend-injection
2452 @opindex ffriend-injection
2453 Inject friend functions into the enclosing namespace, so that they are
2454 visible outside the scope of the class in which they are declared.
2455 Friend functions were documented to work this way in the old Annotated
2456 C++ Reference Manual.
2457 However, in ISO C++ a friend function that is not declared
2458 in an enclosing scope can only be found using argument dependent
2459 lookup. GCC defaults to the standard behavior.
2461 This option is deprecated and will be removed.
2463 @item -fno-elide-constructors
2464 @opindex fno-elide-constructors
2465 The C++ standard allows an implementation to omit creating a temporary
2466 that is only used to initialize another object of the same type.
2467 Specifying this option disables that optimization, and forces G++ to
2468 call the copy constructor in all cases. This option also causes G++
2469 to call trivial member functions which otherwise would be expanded inline.
2471 In C++17, the compiler is required to omit these temporaries, but this
2472 option still affects trivial member functions.
2474 @item -fno-enforce-eh-specs
2475 @opindex fno-enforce-eh-specs
2476 Don't generate code to check for violation of exception specifications
2477 at run time. This option violates the C++ standard, but may be useful
2478 for reducing code size in production builds, much like defining
2479 @code{NDEBUG}. This does not give user code permission to throw
2480 exceptions in violation of the exception specifications; the compiler
2481 still optimizes based on the specifications, so throwing an
2482 unexpected exception results in undefined behavior at run time.
2484 @item -fextern-tls-init
2485 @itemx -fno-extern-tls-init
2486 @opindex fextern-tls-init
2487 @opindex fno-extern-tls-init
2488 The C++11 and OpenMP standards allow @code{thread_local} and
2489 @code{threadprivate} variables to have dynamic (runtime)
2490 initialization. To support this, any use of such a variable goes
2491 through a wrapper function that performs any necessary initialization.
2492 When the use and definition of the variable are in the same
2493 translation unit, this overhead can be optimized away, but when the
2494 use is in a different translation unit there is significant overhead
2495 even if the variable doesn't actually need dynamic initialization. If
2496 the programmer can be sure that no use of the variable in a
2497 non-defining TU needs to trigger dynamic initialization (either
2498 because the variable is statically initialized, or a use of the
2499 variable in the defining TU will be executed before any uses in
2500 another TU), they can avoid this overhead with the
2501 @option{-fno-extern-tls-init} option.
2503 On targets that support symbol aliases, the default is
2504 @option{-fextern-tls-init}. On targets that do not support symbol
2505 aliases, the default is @option{-fno-extern-tls-init}.
2508 @itemx -fno-for-scope
2510 @opindex fno-for-scope
2511 If @option{-ffor-scope} is specified, the scope of variables declared in
2512 a @i{for-init-statement} is limited to the @code{for} loop itself,
2513 as specified by the C++ standard.
2514 If @option{-fno-for-scope} is specified, the scope of variables declared in
2515 a @i{for-init-statement} extends to the end of the enclosing scope,
2516 as was the case in old versions of G++, and other (traditional)
2517 implementations of C++.
2519 This option is deprecated and the associated non-standard
2520 functionality will be removed.
2522 @item -fno-gnu-keywords
2523 @opindex fno-gnu-keywords
2524 Do not recognize @code{typeof} as a keyword, so that code can use this
2525 word as an identifier. You can use the keyword @code{__typeof__} instead.
2526 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2527 @option{-std=c++98}, @option{-std=c++11}, etc.
2529 @item -fno-implicit-templates
2530 @opindex fno-implicit-templates
2531 Never emit code for non-inline templates that are instantiated
2532 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2533 @xref{Template Instantiation}, for more information.
2535 @item -fno-implicit-inline-templates
2536 @opindex fno-implicit-inline-templates
2537 Don't emit code for implicit instantiations of inline templates, either.
2538 The default is to handle inlines differently so that compiles with and
2539 without optimization need the same set of explicit instantiations.
2541 @item -fno-implement-inlines
2542 @opindex fno-implement-inlines
2543 To save space, do not emit out-of-line copies of inline functions
2544 controlled by @code{#pragma implementation}. This causes linker
2545 errors if these functions are not inlined everywhere they are called.
2547 @item -fms-extensions
2548 @opindex fms-extensions
2549 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2550 int and getting a pointer to member function via non-standard syntax.
2552 @item -fnew-inheriting-ctors
2553 @opindex fnew-inheriting-ctors
2554 Enable the P0136 adjustment to the semantics of C++11 constructor
2555 inheritance. This is part of C++17 but also considered to be a Defect
2556 Report against C++11 and C++14. This flag is enabled by default
2557 unless @option{-fabi-version=10} or lower is specified.
2559 @item -fnew-ttp-matching
2560 @opindex fnew-ttp-matching
2561 Enable the P0522 resolution to Core issue 150, template template
2562 parameters and default arguments: this allows a template with default
2563 template arguments as an argument for a template template parameter
2564 with fewer template parameters. This flag is enabled by default for
2565 @option{-std=c++17}.
2567 @item -fno-nonansi-builtins
2568 @opindex fno-nonansi-builtins
2569 Disable built-in declarations of functions that are not mandated by
2570 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2571 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2574 @opindex fnothrow-opt
2575 Treat a @code{throw()} exception specification as if it were a
2576 @code{noexcept} specification to reduce or eliminate the text size
2577 overhead relative to a function with no exception specification. If
2578 the function has local variables of types with non-trivial
2579 destructors, the exception specification actually makes the
2580 function smaller because the EH cleanups for those variables can be
2581 optimized away. The semantic effect is that an exception thrown out of
2582 a function with such an exception specification results in a call
2583 to @code{terminate} rather than @code{unexpected}.
2585 @item -fno-operator-names
2586 @opindex fno-operator-names
2587 Do not treat the operator name keywords @code{and}, @code{bitand},
2588 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2589 synonyms as keywords.
2591 @item -fno-optional-diags
2592 @opindex fno-optional-diags
2593 Disable diagnostics that the standard says a compiler does not need to
2594 issue. Currently, the only such diagnostic issued by G++ is the one for
2595 a name having multiple meanings within a class.
2598 @opindex fpermissive
2599 Downgrade some diagnostics about nonconformant code from errors to
2600 warnings. Thus, using @option{-fpermissive} allows some
2601 nonconforming code to compile.
2603 @item -fno-pretty-templates
2604 @opindex fno-pretty-templates
2605 When an error message refers to a specialization of a function
2606 template, the compiler normally prints the signature of the
2607 template followed by the template arguments and any typedefs or
2608 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2609 rather than @code{void f(int)}) so that it's clear which template is
2610 involved. When an error message refers to a specialization of a class
2611 template, the compiler omits any template arguments that match
2612 the default template arguments for that template. If either of these
2613 behaviors make it harder to understand the error message rather than
2614 easier, you can use @option{-fno-pretty-templates} to disable them.
2618 Enable automatic template instantiation at link time. This option also
2619 implies @option{-fno-implicit-templates}. @xref{Template
2620 Instantiation}, for more information.
2624 Disable generation of information about every class with virtual
2625 functions for use by the C++ run-time type identification features
2626 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2627 of the language, you can save some space by using this flag. Note that
2628 exception handling uses the same information, but G++ generates it as
2629 needed. The @code{dynamic_cast} operator can still be used for casts that
2630 do not require run-time type information, i.e.@: casts to @code{void *} or to
2631 unambiguous base classes.
2633 @item -fsized-deallocation
2634 @opindex fsized-deallocation
2635 Enable the built-in global declarations
2637 void operator delete (void *, std::size_t) noexcept;
2638 void operator delete[] (void *, std::size_t) noexcept;
2640 as introduced in C++14. This is useful for user-defined replacement
2641 deallocation functions that, for example, use the size of the object
2642 to make deallocation faster. Enabled by default under
2643 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2644 warns about places that might want to add a definition.
2646 @item -fstrict-enums
2647 @opindex fstrict-enums
2648 Allow the compiler to optimize using the assumption that a value of
2649 enumerated type can only be one of the values of the enumeration (as
2650 defined in the C++ standard; basically, a value that can be
2651 represented in the minimum number of bits needed to represent all the
2652 enumerators). This assumption may not be valid if the program uses a
2653 cast to convert an arbitrary integer value to the enumerated type.
2655 @item -fstrong-eval-order
2656 @opindex fstrong-eval-order
2657 Evaluate member access, array subscripting, and shift expressions in
2658 left-to-right order, and evaluate assignment in right-to-left order,
2659 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2660 @option{-fstrong-eval-order=some} enables just the ordering of member
2661 access and shift expressions, and is the default without
2662 @option{-std=c++17}.
2664 @item -ftemplate-backtrace-limit=@var{n}
2665 @opindex ftemplate-backtrace-limit
2666 Set the maximum number of template instantiation notes for a single
2667 warning or error to @var{n}. The default value is 10.
2669 @item -ftemplate-depth=@var{n}
2670 @opindex ftemplate-depth
2671 Set the maximum instantiation depth for template classes to @var{n}.
2672 A limit on the template instantiation depth is needed to detect
2673 endless recursions during template class instantiation. ANSI/ISO C++
2674 conforming programs must not rely on a maximum depth greater than 17
2675 (changed to 1024 in C++11). The default value is 900, as the compiler
2676 can run out of stack space before hitting 1024 in some situations.
2678 @item -fno-threadsafe-statics
2679 @opindex fno-threadsafe-statics
2680 Do not emit the extra code to use the routines specified in the C++
2681 ABI for thread-safe initialization of local statics. You can use this
2682 option to reduce code size slightly in code that doesn't need to be
2685 @item -fuse-cxa-atexit
2686 @opindex fuse-cxa-atexit
2687 Register destructors for objects with static storage duration with the
2688 @code{__cxa_atexit} function rather than the @code{atexit} function.
2689 This option is required for fully standards-compliant handling of static
2690 destructors, but only works if your C library supports
2691 @code{__cxa_atexit}.
2693 @item -fno-use-cxa-get-exception-ptr
2694 @opindex fno-use-cxa-get-exception-ptr
2695 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2696 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2697 if the runtime routine is not available.
2699 @item -fvisibility-inlines-hidden
2700 @opindex fvisibility-inlines-hidden
2701 This switch declares that the user does not attempt to compare
2702 pointers to inline functions or methods where the addresses of the two functions
2703 are taken in different shared objects.
2705 The effect of this is that GCC may, effectively, mark inline methods with
2706 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2707 appear in the export table of a DSO and do not require a PLT indirection
2708 when used within the DSO@. Enabling this option can have a dramatic effect
2709 on load and link times of a DSO as it massively reduces the size of the
2710 dynamic export table when the library makes heavy use of templates.
2712 The behavior of this switch is not quite the same as marking the
2713 methods as hidden directly, because it does not affect static variables
2714 local to the function or cause the compiler to deduce that
2715 the function is defined in only one shared object.
2717 You may mark a method as having a visibility explicitly to negate the
2718 effect of the switch for that method. For example, if you do want to
2719 compare pointers to a particular inline method, you might mark it as
2720 having default visibility. Marking the enclosing class with explicit
2721 visibility has no effect.
2723 Explicitly instantiated inline methods are unaffected by this option
2724 as their linkage might otherwise cross a shared library boundary.
2725 @xref{Template Instantiation}.
2727 @item -fvisibility-ms-compat
2728 @opindex fvisibility-ms-compat
2729 This flag attempts to use visibility settings to make GCC's C++
2730 linkage model compatible with that of Microsoft Visual Studio.
2732 The flag makes these changes to GCC's linkage model:
2736 It sets the default visibility to @code{hidden}, like
2737 @option{-fvisibility=hidden}.
2740 Types, but not their members, are not hidden by default.
2743 The One Definition Rule is relaxed for types without explicit
2744 visibility specifications that are defined in more than one
2745 shared object: those declarations are permitted if they are
2746 permitted when this option is not used.
2749 In new code it is better to use @option{-fvisibility=hidden} and
2750 export those classes that are intended to be externally visible.
2751 Unfortunately it is possible for code to rely, perhaps accidentally,
2752 on the Visual Studio behavior.
2754 Among the consequences of these changes are that static data members
2755 of the same type with the same name but defined in different shared
2756 objects are different, so changing one does not change the other;
2757 and that pointers to function members defined in different shared
2758 objects may not compare equal. When this flag is given, it is a
2759 violation of the ODR to define types with the same name differently.
2763 Do not use weak symbol support, even if it is provided by the linker.
2764 By default, G++ uses weak symbols if they are available. This
2765 option exists only for testing, and should not be used by end-users;
2766 it results in inferior code and has no benefits. This option may
2767 be removed in a future release of G++.
2771 Do not search for header files in the standard directories specific to
2772 C++, but do still search the other standard directories. (This option
2773 is used when building the C++ library.)
2776 In addition, these optimization, warning, and code generation options
2777 have meanings only for C++ programs:
2780 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2783 Warn when G++ it generates code that is probably not compatible with
2784 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2785 ABI with each major release, normally @option{-Wabi} will warn only if
2786 there is a check added later in a release series for an ABI issue
2787 discovered since the initial release. @option{-Wabi} will warn about
2788 more things if an older ABI version is selected (with
2789 @option{-fabi-version=@var{n}}).
2791 @option{-Wabi} can also be used with an explicit version number to
2792 warn about compatibility with a particular @option{-fabi-version}
2793 level, e.g. @option{-Wabi=2} to warn about changes relative to
2794 @option{-fabi-version=2}.
2796 If an explicit version number is provided and
2797 @option{-fabi-compat-version} is not specified, the version number
2798 from this option is used for compatibility aliases. If no explicit
2799 version number is provided with this option, but
2800 @option{-fabi-compat-version} is specified, that version number is
2801 used for ABI warnings.
2803 Although an effort has been made to warn about
2804 all such cases, there are probably some cases that are not warned about,
2805 even though G++ is generating incompatible code. There may also be
2806 cases where warnings are emitted even though the code that is generated
2809 You should rewrite your code to avoid these warnings if you are
2810 concerned about the fact that code generated by G++ may not be binary
2811 compatible with code generated by other compilers.
2813 Known incompatibilities in @option{-fabi-version=2} (which was the
2814 default from GCC 3.4 to 4.9) include:
2819 A template with a non-type template parameter of reference type was
2820 mangled incorrectly:
2823 template <int &> struct S @{@};
2827 This was fixed in @option{-fabi-version=3}.
2830 SIMD vector types declared using @code{__attribute ((vector_size))} were
2831 mangled in a non-standard way that does not allow for overloading of
2832 functions taking vectors of different sizes.
2834 The mangling was changed in @option{-fabi-version=4}.
2837 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2838 qualifiers, and @code{decltype} of a plain declaration was folded away.
2840 These mangling issues were fixed in @option{-fabi-version=5}.
2843 Scoped enumerators passed as arguments to a variadic function are
2844 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2845 On most targets this does not actually affect the parameter passing
2846 ABI, as there is no way to pass an argument smaller than @code{int}.
2848 Also, the ABI changed the mangling of template argument packs,
2849 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2850 a class scope function used as a template argument.
2852 These issues were corrected in @option{-fabi-version=6}.
2855 Lambdas in default argument scope were mangled incorrectly, and the
2856 ABI changed the mangling of @code{nullptr_t}.
2858 These issues were corrected in @option{-fabi-version=7}.
2861 When mangling a function type with function-cv-qualifiers, the
2862 un-qualified function type was incorrectly treated as a substitution
2865 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2868 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2869 unaligned accesses. Note that this did not affect the ABI of a
2870 function with a @code{nullptr_t} parameter, as parameters have a
2873 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2876 Target-specific attributes that affect the identity of a type, such as
2877 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2878 did not affect the mangled name, leading to name collisions when
2879 function pointers were used as template arguments.
2881 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2885 It also warns about psABI-related changes. The known psABI changes at this
2891 For SysV/x86-64, unions with @code{long double} members are
2892 passed in memory as specified in psABI. For example:
2902 @code{union U} is always passed in memory.
2906 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2909 Warn when a type with an ABI tag is used in a context that does not
2910 have that ABI tag. See @ref{C++ Attributes} for more information
2913 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2914 @opindex Wctor-dtor-privacy
2915 @opindex Wno-ctor-dtor-privacy
2916 Warn when a class seems unusable because all the constructors or
2917 destructors in that class are private, and it has neither friends nor
2918 public static member functions. Also warn if there are no non-private
2919 methods, and there's at least one private member function that isn't
2920 a constructor or destructor.
2922 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2923 @opindex Wdelete-non-virtual-dtor
2924 @opindex Wno-delete-non-virtual-dtor
2925 Warn when @code{delete} is used to destroy an instance of a class that
2926 has virtual functions and non-virtual destructor. It is unsafe to delete
2927 an instance of a derived class through a pointer to a base class if the
2928 base class does not have a virtual destructor. This warning is enabled
2931 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2932 @opindex Wliteral-suffix
2933 @opindex Wno-literal-suffix
2934 Warn when a string or character literal is followed by a ud-suffix which does
2935 not begin with an underscore. As a conforming extension, GCC treats such
2936 suffixes as separate preprocessing tokens in order to maintain backwards
2937 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2941 #define __STDC_FORMAT_MACROS
2942 #include <inttypes.h>
2947 printf("My int64: %" PRId64"\n", i64);
2951 In this case, @code{PRId64} is treated as a separate preprocessing token.
2953 Additionally, warn when a user-defined literal operator is declared with
2954 a literal suffix identifier that doesn't begin with an underscore. Literal
2955 suffix identifiers that don't begin with an underscore are reserved for
2956 future standardization.
2958 This warning is enabled by default.
2960 @item -Wlto-type-mismatch
2961 @opindex Wlto-type-mismatch
2962 @opindex Wno-lto-type-mismatch
2964 During the link-time optimization warn about type mismatches in
2965 global declarations from different compilation units.
2966 Requires @option{-flto} to be enabled. Enabled by default.
2968 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2970 @opindex Wno-narrowing
2971 For C++11 and later standards, narrowing conversions are diagnosed by default,
2972 as required by the standard. A narrowing conversion from a constant produces
2973 an error, and a narrowing conversion from a non-constant produces a warning,
2974 but @option{-Wno-narrowing} suppresses the diagnostic.
2975 Note that this does not affect the meaning of well-formed code;
2976 narrowing conversions are still considered ill-formed in SFINAE contexts.
2978 With @option{-Wnarrowing} in C++98, warn when a narrowing
2979 conversion prohibited by C++11 occurs within
2983 int i = @{ 2.2 @}; // error: narrowing from double to int
2986 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2988 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2990 @opindex Wno-noexcept
2991 Warn when a noexcept-expression evaluates to false because of a call
2992 to a function that does not have a non-throwing exception
2993 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2994 the compiler to never throw an exception.
2996 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
2997 @opindex Wnoexcept-type
2998 @opindex Wno-noexcept-type
2999 Warn if the C++17 feature making @code{noexcept} part of a function
3000 type changes the mangled name of a symbol relative to C++14. Enabled
3001 by @option{-Wabi} and @option{-Wc++17-compat}.
3006 template <class T> void f(T t) @{ t(); @};
3008 void h() @{ f(g); @}
3012 In C++14, @code{f} calls calls @code{f<void(*)()>}, but in
3013 C++17 it calls @code{f<void(*)()noexcept>}.
3015 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
3016 @opindex Wclass-memaccess
3017 Warn when the destination of a call to a raw memory function such as
3018 @code{memset} or @code{memcpy} is an object of class type, and when writing
3019 into such an object might bypass the class non-trivial or deleted constructor
3020 or copy assignment, violate const-correctness or encapsulation, or corrupt
3021 virtual table pointers. Modifying the representation of such objects may
3022 violate invariants maintained by member functions of the class. For example,
3023 the call to @code{memset} below is undefined because it modifies a non-trivial
3024 class object and is, therefore, diagnosed. The safe way to either initialize
3025 or clear the storage of objects of such types is by using the appropriate
3026 constructor or assignment operator, if one is available.
3028 std::string str = "abc";
3029 memset (&str, 0, sizeof str);
3031 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
3032 Explicitly casting the pointer to the class object to @code{void *} or
3033 to a type that can be safely accessed by the raw memory function suppresses
3036 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
3037 @opindex Wnon-virtual-dtor
3038 @opindex Wno-non-virtual-dtor
3039 Warn when a class has virtual functions and an accessible non-virtual
3040 destructor itself or in an accessible polymorphic base class, in which
3041 case it is possible but unsafe to delete an instance of a derived
3042 class through a pointer to the class itself or base class. This
3043 warning is automatically enabled if @option{-Weffc++} is specified.
3045 @item -Wregister @r{(C++ and Objective-C++ only)}
3047 @opindex Wno-register
3048 Warn on uses of the @code{register} storage class specifier, except
3049 when it is part of the GNU @ref{Explicit Register Variables} extension.
3050 The use of the @code{register} keyword as storage class specifier has
3051 been deprecated in C++11 and removed in C++17.
3052 Enabled by default with @option{-std=c++17}.
3054 @item -Wreorder @r{(C++ and Objective-C++ only)}
3056 @opindex Wno-reorder
3057 @cindex reordering, warning
3058 @cindex warning for reordering of member initializers
3059 Warn when the order of member initializers given in the code does not
3060 match the order in which they must be executed. For instance:
3066 A(): j (0), i (1) @{ @}
3071 The compiler rearranges the member initializers for @code{i}
3072 and @code{j} to match the declaration order of the members, emitting
3073 a warning to that effect. This warning is enabled by @option{-Wall}.
3075 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
3076 @opindex fext-numeric-literals
3077 @opindex fno-ext-numeric-literals
3078 Accept imaginary, fixed-point, or machine-defined
3079 literal number suffixes as GNU extensions.
3080 When this option is turned off these suffixes are treated
3081 as C++11 user-defined literal numeric suffixes.
3082 This is on by default for all pre-C++11 dialects and all GNU dialects:
3083 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3084 @option{-std=gnu++14}.
3085 This option is off by default
3086 for ISO C++11 onwards (@option{-std=c++11}, ...).
3089 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3092 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3095 Warn about violations of the following style guidelines from Scott Meyers'
3096 @cite{Effective C++} series of books:
3100 Define a copy constructor and an assignment operator for classes
3101 with dynamically-allocated memory.
3104 Prefer initialization to assignment in constructors.
3107 Have @code{operator=} return a reference to @code{*this}.
3110 Don't try to return a reference when you must return an object.
3113 Distinguish between prefix and postfix forms of increment and
3114 decrement operators.
3117 Never overload @code{&&}, @code{||}, or @code{,}.
3121 This option also enables @option{-Wnon-virtual-dtor}, which is also
3122 one of the effective C++ recommendations. However, the check is
3123 extended to warn about the lack of virtual destructor in accessible
3124 non-polymorphic bases classes too.
3126 When selecting this option, be aware that the standard library
3127 headers do not obey all of these guidelines; use @samp{grep -v}
3128 to filter out those warnings.
3130 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3131 @opindex Wstrict-null-sentinel
3132 @opindex Wno-strict-null-sentinel
3133 Warn about the use of an uncasted @code{NULL} as sentinel. When
3134 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3135 to @code{__null}. Although it is a null pointer constant rather than a
3136 null pointer, it is guaranteed to be of the same size as a pointer.
3137 But this use is not portable across different compilers.
3139 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3140 @opindex Wno-non-template-friend
3141 @opindex Wnon-template-friend
3142 Disable warnings when non-template friend functions are declared
3143 within a template. In very old versions of GCC that predate implementation
3144 of the ISO standard, declarations such as
3145 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3146 could be interpreted as a particular specialization of a template
3147 function; the warning exists to diagnose compatibility problems,
3148 and is enabled by default.
3150 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3151 @opindex Wold-style-cast
3152 @opindex Wno-old-style-cast
3153 Warn if an old-style (C-style) cast to a non-void type is used within
3154 a C++ program. The new-style casts (@code{dynamic_cast},
3155 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3156 less vulnerable to unintended effects and much easier to search for.
3158 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3159 @opindex Woverloaded-virtual
3160 @opindex Wno-overloaded-virtual
3161 @cindex overloaded virtual function, warning
3162 @cindex warning for overloaded virtual function
3163 Warn when a function declaration hides virtual functions from a
3164 base class. For example, in:
3171 struct B: public A @{
3176 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3187 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3188 @opindex Wno-pmf-conversions
3189 @opindex Wpmf-conversions
3190 Disable the diagnostic for converting a bound pointer to member function
3193 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3194 @opindex Wsign-promo
3195 @opindex Wno-sign-promo
3196 Warn when overload resolution chooses a promotion from unsigned or
3197 enumerated type to a signed type, over a conversion to an unsigned type of
3198 the same size. Previous versions of G++ tried to preserve
3199 unsignedness, but the standard mandates the current behavior.
3201 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3203 Warn when a primary template declaration is encountered. Some coding
3204 rules disallow templates, and this may be used to enforce that rule.
3205 The warning is inactive inside a system header file, such as the STL, so
3206 one can still use the STL. One may also instantiate or specialize
3209 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3210 @opindex Wmultiple-inheritance
3211 Warn when a class is defined with multiple direct base classes. Some
3212 coding rules disallow multiple inheritance, and this may be used to
3213 enforce that rule. The warning is inactive inside a system header file,
3214 such as the STL, so one can still use the STL. One may also define
3215 classes that indirectly use multiple inheritance.
3217 @item -Wvirtual-inheritance
3218 @opindex Wvirtual-inheritance
3219 Warn when a class is defined with a virtual direct base class. Some
3220 coding rules disallow multiple inheritance, and this may be used to
3221 enforce that rule. The warning is inactive inside a system header file,
3222 such as the STL, so one can still use the STL. One may also define
3223 classes that indirectly use virtual inheritance.
3226 @opindex Wnamespaces
3227 Warn when a namespace definition is opened. Some coding rules disallow
3228 namespaces, and this may be used to enforce that rule. The warning is
3229 inactive inside a system header file, such as the STL, so one can still
3230 use the STL. One may also use using directives and qualified names.
3232 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3234 @opindex Wno-terminate
3235 Disable the warning about a throw-expression that will immediately
3236 result in a call to @code{terminate}.
3239 @node Objective-C and Objective-C++ Dialect Options
3240 @section Options Controlling Objective-C and Objective-C++ Dialects
3242 @cindex compiler options, Objective-C and Objective-C++
3243 @cindex Objective-C and Objective-C++ options, command-line
3244 @cindex options, Objective-C and Objective-C++
3245 (NOTE: This manual does not describe the Objective-C and Objective-C++
3246 languages themselves. @xref{Standards,,Language Standards
3247 Supported by GCC}, for references.)
3249 This section describes the command-line options that are only meaningful
3250 for Objective-C and Objective-C++ programs. You can also use most of
3251 the language-independent GNU compiler options.
3252 For example, you might compile a file @file{some_class.m} like this:
3255 gcc -g -fgnu-runtime -O -c some_class.m
3259 In this example, @option{-fgnu-runtime} is an option meant only for
3260 Objective-C and Objective-C++ programs; you can use the other options with
3261 any language supported by GCC@.
3263 Note that since Objective-C is an extension of the C language, Objective-C
3264 compilations may also use options specific to the C front-end (e.g.,
3265 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3266 C++-specific options (e.g., @option{-Wabi}).
3268 Here is a list of options that are @emph{only} for compiling Objective-C
3269 and Objective-C++ programs:
3272 @item -fconstant-string-class=@var{class-name}
3273 @opindex fconstant-string-class
3274 Use @var{class-name} as the name of the class to instantiate for each
3275 literal string specified with the syntax @code{@@"@dots{}"}. The default
3276 class name is @code{NXConstantString} if the GNU runtime is being used, and
3277 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3278 @option{-fconstant-cfstrings} option, if also present, overrides the
3279 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3280 to be laid out as constant CoreFoundation strings.
3283 @opindex fgnu-runtime
3284 Generate object code compatible with the standard GNU Objective-C
3285 runtime. This is the default for most types of systems.
3287 @item -fnext-runtime
3288 @opindex fnext-runtime
3289 Generate output compatible with the NeXT runtime. This is the default
3290 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3291 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3294 @item -fno-nil-receivers
3295 @opindex fno-nil-receivers
3296 Assume that all Objective-C message dispatches (@code{[receiver
3297 message:arg]}) in this translation unit ensure that the receiver is
3298 not @code{nil}. This allows for more efficient entry points in the
3299 runtime to be used. This option is only available in conjunction with
3300 the NeXT runtime and ABI version 0 or 1.
3302 @item -fobjc-abi-version=@var{n}
3303 @opindex fobjc-abi-version
3304 Use version @var{n} of the Objective-C ABI for the selected runtime.
3305 This option is currently supported only for the NeXT runtime. In that
3306 case, Version 0 is the traditional (32-bit) ABI without support for
3307 properties and other Objective-C 2.0 additions. Version 1 is the
3308 traditional (32-bit) ABI with support for properties and other
3309 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3310 nothing is specified, the default is Version 0 on 32-bit target
3311 machines, and Version 2 on 64-bit target machines.
3313 @item -fobjc-call-cxx-cdtors
3314 @opindex fobjc-call-cxx-cdtors
3315 For each Objective-C class, check if any of its instance variables is a
3316 C++ object with a non-trivial default constructor. If so, synthesize a
3317 special @code{- (id) .cxx_construct} instance method which runs
3318 non-trivial default constructors on any such instance variables, in order,
3319 and then return @code{self}. Similarly, check if any instance variable
3320 is a C++ object with a non-trivial destructor, and if so, synthesize a
3321 special @code{- (void) .cxx_destruct} method which runs
3322 all such default destructors, in reverse order.
3324 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3325 methods thusly generated only operate on instance variables
3326 declared in the current Objective-C class, and not those inherited
3327 from superclasses. It is the responsibility of the Objective-C
3328 runtime to invoke all such methods in an object's inheritance
3329 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3330 by the runtime immediately after a new object instance is allocated;
3331 the @code{- (void) .cxx_destruct} methods are invoked immediately
3332 before the runtime deallocates an object instance.
3334 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3335 support for invoking the @code{- (id) .cxx_construct} and
3336 @code{- (void) .cxx_destruct} methods.
3338 @item -fobjc-direct-dispatch
3339 @opindex fobjc-direct-dispatch
3340 Allow fast jumps to the message dispatcher. On Darwin this is
3341 accomplished via the comm page.
3343 @item -fobjc-exceptions
3344 @opindex fobjc-exceptions
3345 Enable syntactic support for structured exception handling in
3346 Objective-C, similar to what is offered by C++. This option
3347 is required to use the Objective-C keywords @code{@@try},
3348 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3349 @code{@@synchronized}. This option is available with both the GNU
3350 runtime and the NeXT runtime (but not available in conjunction with
3351 the NeXT runtime on Mac OS X 10.2 and earlier).
3355 Enable garbage collection (GC) in Objective-C and Objective-C++
3356 programs. This option is only available with the NeXT runtime; the
3357 GNU runtime has a different garbage collection implementation that
3358 does not require special compiler flags.
3360 @item -fobjc-nilcheck
3361 @opindex fobjc-nilcheck
3362 For the NeXT runtime with version 2 of the ABI, check for a nil
3363 receiver in method invocations before doing the actual method call.
3364 This is the default and can be disabled using
3365 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3366 checked for nil in this way no matter what this flag is set to.
3367 Currently this flag does nothing when the GNU runtime, or an older
3368 version of the NeXT runtime ABI, is used.
3370 @item -fobjc-std=objc1
3372 Conform to the language syntax of Objective-C 1.0, the language
3373 recognized by GCC 4.0. This only affects the Objective-C additions to
3374 the C/C++ language; it does not affect conformance to C/C++ standards,
3375 which is controlled by the separate C/C++ dialect option flags. When
3376 this option is used with the Objective-C or Objective-C++ compiler,
3377 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3378 This is useful if you need to make sure that your Objective-C code can
3379 be compiled with older versions of GCC@.
3381 @item -freplace-objc-classes
3382 @opindex freplace-objc-classes
3383 Emit a special marker instructing @command{ld(1)} not to statically link in
3384 the resulting object file, and allow @command{dyld(1)} to load it in at
3385 run time instead. This is used in conjunction with the Fix-and-Continue
3386 debugging mode, where the object file in question may be recompiled and
3387 dynamically reloaded in the course of program execution, without the need
3388 to restart the program itself. Currently, Fix-and-Continue functionality
3389 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3394 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3395 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3396 compile time) with static class references that get initialized at load time,
3397 which improves run-time performance. Specifying the @option{-fzero-link} flag
3398 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3399 to be retained. This is useful in Zero-Link debugging mode, since it allows
3400 for individual class implementations to be modified during program execution.
3401 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3402 regardless of command-line options.
3404 @item -fno-local-ivars
3405 @opindex fno-local-ivars
3406 @opindex flocal-ivars
3407 By default instance variables in Objective-C can be accessed as if
3408 they were local variables from within the methods of the class they're
3409 declared in. This can lead to shadowing between instance variables
3410 and other variables declared either locally inside a class method or
3411 globally with the same name. Specifying the @option{-fno-local-ivars}
3412 flag disables this behavior thus avoiding variable shadowing issues.
3414 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3415 @opindex fivar-visibility
3416 Set the default instance variable visibility to the specified option
3417 so that instance variables declared outside the scope of any access
3418 modifier directives default to the specified visibility.
3422 Dump interface declarations for all classes seen in the source file to a
3423 file named @file{@var{sourcename}.decl}.
3425 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3426 @opindex Wassign-intercept
3427 @opindex Wno-assign-intercept
3428 Warn whenever an Objective-C assignment is being intercepted by the
3431 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3432 @opindex Wno-protocol
3434 If a class is declared to implement a protocol, a warning is issued for
3435 every method in the protocol that is not implemented by the class. The
3436 default behavior is to issue a warning for every method not explicitly
3437 implemented in the class, even if a method implementation is inherited
3438 from the superclass. If you use the @option{-Wno-protocol} option, then
3439 methods inherited from the superclass are considered to be implemented,
3440 and no warning is issued for them.
3442 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3444 @opindex Wno-selector
3445 Warn if multiple methods of different types for the same selector are
3446 found during compilation. The check is performed on the list of methods
3447 in the final stage of compilation. Additionally, a check is performed
3448 for each selector appearing in a @code{@@selector(@dots{})}
3449 expression, and a corresponding method for that selector has been found
3450 during compilation. Because these checks scan the method table only at
3451 the end of compilation, these warnings are not produced if the final
3452 stage of compilation is not reached, for example because an error is
3453 found during compilation, or because the @option{-fsyntax-only} option is
3456 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3457 @opindex Wstrict-selector-match
3458 @opindex Wno-strict-selector-match
3459 Warn if multiple methods with differing argument and/or return types are
3460 found for a given selector when attempting to send a message using this
3461 selector to a receiver of type @code{id} or @code{Class}. When this flag
3462 is off (which is the default behavior), the compiler omits such warnings
3463 if any differences found are confined to types that share the same size
3466 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3467 @opindex Wundeclared-selector
3468 @opindex Wno-undeclared-selector
3469 Warn if a @code{@@selector(@dots{})} expression referring to an
3470 undeclared selector is found. A selector is considered undeclared if no
3471 method with that name has been declared before the
3472 @code{@@selector(@dots{})} expression, either explicitly in an
3473 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3474 an @code{@@implementation} section. This option always performs its
3475 checks as soon as a @code{@@selector(@dots{})} expression is found,
3476 while @option{-Wselector} only performs its checks in the final stage of
3477 compilation. This also enforces the coding style convention
3478 that methods and selectors must be declared before being used.
3480 @item -print-objc-runtime-info
3481 @opindex print-objc-runtime-info
3482 Generate C header describing the largest structure that is passed by
3487 @node Diagnostic Message Formatting Options
3488 @section Options to Control Diagnostic Messages Formatting
3489 @cindex options to control diagnostics formatting
3490 @cindex diagnostic messages
3491 @cindex message formatting
3493 Traditionally, diagnostic messages have been formatted irrespective of
3494 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3495 options described below
3496 to control the formatting algorithm for diagnostic messages,
3497 e.g.@: how many characters per line, how often source location
3498 information should be reported. Note that some language front ends may not
3499 honor these options.
3502 @item -fmessage-length=@var{n}
3503 @opindex fmessage-length
3504 Try to format error messages so that they fit on lines of about
3505 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3506 done; each error message appears on a single line. This is the
3507 default for all front ends.
3509 @item -fdiagnostics-show-location=once
3510 @opindex fdiagnostics-show-location
3511 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3512 reporter to emit source location information @emph{once}; that is, in
3513 case the message is too long to fit on a single physical line and has to
3514 be wrapped, the source location won't be emitted (as prefix) again,
3515 over and over, in subsequent continuation lines. This is the default
3518 @item -fdiagnostics-show-location=every-line
3519 Only meaningful in line-wrapping mode. Instructs the diagnostic
3520 messages reporter to emit the same source location information (as
3521 prefix) for physical lines that result from the process of breaking
3522 a message which is too long to fit on a single line.
3524 @item -fdiagnostics-color[=@var{WHEN}]
3525 @itemx -fno-diagnostics-color
3526 @opindex fdiagnostics-color
3527 @cindex highlight, color
3528 @vindex GCC_COLORS @r{environment variable}
3529 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3530 or @samp{auto}. The default depends on how the compiler has been configured,
3531 it can be any of the above @var{WHEN} options or also @samp{never}
3532 if @env{GCC_COLORS} environment variable isn't present in the environment,
3533 and @samp{auto} otherwise.
3534 @samp{auto} means to use color only when the standard error is a terminal.
3535 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3536 aliases for @option{-fdiagnostics-color=always} and
3537 @option{-fdiagnostics-color=never}, respectively.
3539 The colors are defined by the environment variable @env{GCC_COLORS}.
3540 Its value is a colon-separated list of capabilities and Select Graphic
3541 Rendition (SGR) substrings. SGR commands are interpreted by the
3542 terminal or terminal emulator. (See the section in the documentation
3543 of your text terminal for permitted values and their meanings as
3544 character attributes.) These substring values are integers in decimal
3545 representation and can be concatenated with semicolons.
3546 Common values to concatenate include
3548 @samp{4} for underline,
3550 @samp{7} for inverse,
3551 @samp{39} for default foreground color,
3552 @samp{30} to @samp{37} for foreground colors,
3553 @samp{90} to @samp{97} for 16-color mode foreground colors,
3554 @samp{38;5;0} to @samp{38;5;255}
3555 for 88-color and 256-color modes foreground colors,
3556 @samp{49} for default background color,
3557 @samp{40} to @samp{47} for background colors,
3558 @samp{100} to @samp{107} for 16-color mode background colors,
3559 and @samp{48;5;0} to @samp{48;5;255}
3560 for 88-color and 256-color modes background colors.
3562 The default @env{GCC_COLORS} is
3564 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3565 quote=01:fixit-insert=32:fixit-delete=31:\
3566 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3570 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3571 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3572 @samp{01} is bold, and @samp{31} is red.
3573 Setting @env{GCC_COLORS} to the empty string disables colors.
3574 Supported capabilities are as follows.
3578 @vindex error GCC_COLORS @r{capability}
3579 SGR substring for error: markers.
3582 @vindex warning GCC_COLORS @r{capability}
3583 SGR substring for warning: markers.
3586 @vindex note GCC_COLORS @r{capability}
3587 SGR substring for note: markers.
3590 @vindex range1 GCC_COLORS @r{capability}
3591 SGR substring for first additional range.
3594 @vindex range2 GCC_COLORS @r{capability}
3595 SGR substring for second additional range.
3598 @vindex locus GCC_COLORS @r{capability}
3599 SGR substring for location information, @samp{file:line} or
3600 @samp{file:line:column} etc.
3603 @vindex quote GCC_COLORS @r{capability}
3604 SGR substring for information printed within quotes.
3607 @vindex fixit-insert GCC_COLORS @r{capability}
3608 SGR substring for fix-it hints suggesting text to
3609 be inserted or replaced.
3612 @vindex fixit-delete GCC_COLORS @r{capability}
3613 SGR substring for fix-it hints suggesting text to
3616 @item diff-filename=
3617 @vindex diff-filename GCC_COLORS @r{capability}
3618 SGR substring for filename headers within generated patches.
3621 @vindex diff-hunk GCC_COLORS @r{capability}
3622 SGR substring for the starts of hunks within generated patches.
3625 @vindex diff-delete GCC_COLORS @r{capability}
3626 SGR substring for deleted lines within generated patches.
3629 @vindex diff-insert GCC_COLORS @r{capability}
3630 SGR substring for inserted lines within generated patches.
3633 @vindex type-diff GCC_COLORS @r{capability}
3634 SGR substring for highlighting mismatching types within template
3635 arguments in the C++ frontend.
3638 @item -fno-diagnostics-show-option
3639 @opindex fno-diagnostics-show-option
3640 @opindex fdiagnostics-show-option
3641 By default, each diagnostic emitted includes text indicating the
3642 command-line option that directly controls the diagnostic (if such an
3643 option is known to the diagnostic machinery). Specifying the
3644 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3646 @item -fno-diagnostics-show-caret
3647 @opindex fno-diagnostics-show-caret
3648 @opindex fdiagnostics-show-caret
3649 By default, each diagnostic emitted includes the original source line
3650 and a caret @samp{^} indicating the column. This option suppresses this
3651 information. The source line is truncated to @var{n} characters, if
3652 the @option{-fmessage-length=n} option is given. When the output is done
3653 to the terminal, the width is limited to the width given by the
3654 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3656 @item -fdiagnostics-parseable-fixits
3657 @opindex fdiagnostics-parseable-fixits
3658 Emit fix-it hints in a machine-parseable format, suitable for consumption
3659 by IDEs. For each fix-it, a line will be printed after the relevant
3660 diagnostic, starting with the string ``fix-it:''. For example:
3663 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3666 The location is expressed as a half-open range, expressed as a count of
3667 bytes, starting at byte 1 for the initial column. In the above example,
3668 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3672 00000000011111111112222222222
3673 12345678901234567890123456789
3674 gtk_widget_showall (dlg);
3679 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3680 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3681 (e.g. vertical tab as ``\013'').
3683 An empty replacement string indicates that the given range is to be removed.
3684 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3685 be inserted at the given position.
3687 @item -fdiagnostics-generate-patch
3688 @opindex fdiagnostics-generate-patch
3689 Print fix-it hints to stderr in unified diff format, after any diagnostics
3690 are printed. For example:
3697 void show_cb(GtkDialog *dlg)
3699 - gtk_widget_showall(dlg);
3700 + gtk_widget_show_all(dlg);
3705 The diff may or may not be colorized, following the same rules
3706 as for diagnostics (see @option{-fdiagnostics-color}).
3708 @item -fdiagnostics-show-template-tree
3709 @opindex fdiagnostics-show-template-tree
3711 In the C++ frontend, when printing diagnostics showing mismatching
3712 template types, such as:
3715 could not convert 'std::map<int, std::vector<double> >()'
3716 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3719 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3720 tree-like structure showing the common and differing parts of the types,
3730 The parts that differ are highlighted with color (``double'' and
3731 ``float'' in this case).
3733 @item -fno-elide-type
3734 @opindex fno-elide-type
3735 @opindex felide-type
3736 By default when the C++ frontend prints diagnostics showing mismatching
3737 template types, common parts of the types are printed as ``[...]'' to
3738 simplify the error message. For example:
3741 could not convert 'std::map<int, std::vector<double> >()'
3742 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3745 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
3746 This flag also affects the output of the
3747 @option{-fdiagnostics-show-template-tree} flag.
3749 @item -fno-show-column
3750 @opindex fno-show-column
3751 Do not print column numbers in diagnostics. This may be necessary if
3752 diagnostics are being scanned by a program that does not understand the
3753 column numbers, such as @command{dejagnu}.
3757 @node Warning Options
3758 @section Options to Request or Suppress Warnings
3759 @cindex options to control warnings
3760 @cindex warning messages
3761 @cindex messages, warning
3762 @cindex suppressing warnings
3764 Warnings are diagnostic messages that report constructions that
3765 are not inherently erroneous but that are risky or suggest there
3766 may have been an error.
3768 The following language-independent options do not enable specific
3769 warnings but control the kinds of diagnostics produced by GCC@.
3772 @cindex syntax checking
3774 @opindex fsyntax-only
3775 Check the code for syntax errors, but don't do anything beyond that.
3777 @item -fmax-errors=@var{n}
3778 @opindex fmax-errors
3779 Limits the maximum number of error messages to @var{n}, at which point
3780 GCC bails out rather than attempting to continue processing the source
3781 code. If @var{n} is 0 (the default), there is no limit on the number
3782 of error messages produced. If @option{-Wfatal-errors} is also
3783 specified, then @option{-Wfatal-errors} takes precedence over this
3788 Inhibit all warning messages.
3793 Make all warnings into errors.
3798 Make the specified warning into an error. The specifier for a warning
3799 is appended; for example @option{-Werror=switch} turns the warnings
3800 controlled by @option{-Wswitch} into errors. This switch takes a
3801 negative form, to be used to negate @option{-Werror} for specific
3802 warnings; for example @option{-Wno-error=switch} makes
3803 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3806 The warning message for each controllable warning includes the
3807 option that controls the warning. That option can then be used with
3808 @option{-Werror=} and @option{-Wno-error=} as described above.
3809 (Printing of the option in the warning message can be disabled using the
3810 @option{-fno-diagnostics-show-option} flag.)
3812 Note that specifying @option{-Werror=}@var{foo} automatically implies
3813 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3816 @item -Wfatal-errors
3817 @opindex Wfatal-errors
3818 @opindex Wno-fatal-errors
3819 This option causes the compiler to abort compilation on the first error
3820 occurred rather than trying to keep going and printing further error
3825 You can request many specific warnings with options beginning with
3826 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3827 implicit declarations. Each of these specific warning options also
3828 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3829 example, @option{-Wno-implicit}. This manual lists only one of the
3830 two forms, whichever is not the default. For further
3831 language-specific options also refer to @ref{C++ Dialect Options} and
3832 @ref{Objective-C and Objective-C++ Dialect Options}.
3834 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3835 options, such as @option{-Wunused}, which may turn on further options,
3836 such as @option{-Wunused-value}. The combined effect of positive and
3837 negative forms is that more specific options have priority over less
3838 specific ones, independently of their position in the command-line. For
3839 options of the same specificity, the last one takes effect. Options
3840 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3841 as if they appeared at the end of the command-line.
3843 When an unrecognized warning option is requested (e.g.,
3844 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3845 that the option is not recognized. However, if the @option{-Wno-} form
3846 is used, the behavior is slightly different: no diagnostic is
3847 produced for @option{-Wno-unknown-warning} unless other diagnostics
3848 are being produced. This allows the use of new @option{-Wno-} options
3849 with old compilers, but if something goes wrong, the compiler
3850 warns that an unrecognized option is present.
3857 Issue all the warnings demanded by strict ISO C and ISO C++;
3858 reject all programs that use forbidden extensions, and some other
3859 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3860 version of the ISO C standard specified by any @option{-std} option used.
3862 Valid ISO C and ISO C++ programs should compile properly with or without
3863 this option (though a rare few require @option{-ansi} or a
3864 @option{-std} option specifying the required version of ISO C)@. However,
3865 without this option, certain GNU extensions and traditional C and C++
3866 features are supported as well. With this option, they are rejected.
3868 @option{-Wpedantic} does not cause warning messages for use of the
3869 alternate keywords whose names begin and end with @samp{__}. Pedantic
3870 warnings are also disabled in the expression that follows
3871 @code{__extension__}. However, only system header files should use
3872 these escape routes; application programs should avoid them.
3873 @xref{Alternate Keywords}.
3875 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3876 C conformance. They soon find that it does not do quite what they want:
3877 it finds some non-ISO practices, but not all---only those for which
3878 ISO C @emph{requires} a diagnostic, and some others for which
3879 diagnostics have been added.
3881 A feature to report any failure to conform to ISO C might be useful in
3882 some instances, but would require considerable additional work and would
3883 be quite different from @option{-Wpedantic}. We don't have plans to
3884 support such a feature in the near future.
3886 Where the standard specified with @option{-std} represents a GNU
3887 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3888 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3889 extended dialect is based. Warnings from @option{-Wpedantic} are given
3890 where they are required by the base standard. (It does not make sense
3891 for such warnings to be given only for features not in the specified GNU
3892 C dialect, since by definition the GNU dialects of C include all
3893 features the compiler supports with the given option, and there would be
3894 nothing to warn about.)
3896 @item -pedantic-errors
3897 @opindex pedantic-errors
3898 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3899 requires a diagnostic, in some cases where there is undefined behavior
3900 at compile-time and in some other cases that do not prevent compilation
3901 of programs that are valid according to the standard. This is not
3902 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3903 by this option and not enabled by the latter and vice versa.
3908 This enables all the warnings about constructions that some users
3909 consider questionable, and that are easy to avoid (or modify to
3910 prevent the warning), even in conjunction with macros. This also
3911 enables some language-specific warnings described in @ref{C++ Dialect
3912 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3914 @option{-Wall} turns on the following warning flags:
3916 @gccoptlist{-Waddress @gol
3917 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3919 -Wbool-operation @gol
3920 -Wc++11-compat -Wc++14-compat @gol
3921 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
3922 -Wchar-subscripts @gol
3924 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3925 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3927 -Wint-in-bool-context @gol
3928 -Wimplicit @r{(C and Objective-C only)} @gol
3929 -Wimplicit-int @r{(C and Objective-C only)} @gol
3930 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3931 -Winit-self @r{(only for C++)} @gol
3932 -Wlogical-not-parentheses @gol
3933 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3934 -Wmaybe-uninitialized @gol
3935 -Wmemset-elt-size @gol
3936 -Wmemset-transposed-args @gol
3937 -Wmisleading-indentation @r{(only for C/C++)} @gol
3938 -Wmissing-attributes @gol
3939 -Wmissing-braces @r{(only for C/ObjC)} @gol
3940 -Wmultistatement-macros @gol
3941 -Wnarrowing @r{(only for C++)} @gol
3943 -Wnonnull-compare @gol
3950 -Wsequence-point @gol
3951 -Wsign-compare @r{(only in C++)} @gol
3952 -Wsizeof-pointer-div @gol
3953 -Wsizeof-pointer-memaccess @gol
3954 -Wstrict-aliasing @gol
3955 -Wstrict-overflow=1 @gol
3957 -Wtautological-compare @gol
3959 -Wuninitialized @gol
3960 -Wunknown-pragmas @gol
3961 -Wunused-function @gol
3964 -Wunused-variable @gol
3965 -Wvolatile-register-var @gol
3968 Note that some warning flags are not implied by @option{-Wall}. Some of
3969 them warn about constructions that users generally do not consider
3970 questionable, but which occasionally you might wish to check for;
3971 others warn about constructions that are necessary or hard to avoid in
3972 some cases, and there is no simple way to modify the code to suppress
3973 the warning. Some of them are enabled by @option{-Wextra} but many of
3974 them must be enabled individually.
3980 This enables some extra warning flags that are not enabled by
3981 @option{-Wall}. (This option used to be called @option{-W}. The older
3982 name is still supported, but the newer name is more descriptive.)
3984 @gccoptlist{-Wclobbered @gol
3985 -Wcast-function-type @gol
3987 -Wignored-qualifiers @gol
3988 -Wimplicit-fallthrough=3 @gol
3989 -Wmissing-field-initializers @gol
3990 -Wmissing-parameter-type @r{(C only)} @gol
3991 -Wold-style-declaration @r{(C only)} @gol
3992 -Woverride-init @gol
3993 -Wsign-compare @r{(C only)} @gol
3995 -Wuninitialized @gol
3996 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3997 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3998 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
4001 The option @option{-Wextra} also prints warning messages for the
4007 A pointer is compared against integer zero with @code{<}, @code{<=},
4008 @code{>}, or @code{>=}.
4011 (C++ only) An enumerator and a non-enumerator both appear in a
4012 conditional expression.
4015 (C++ only) Ambiguous virtual bases.
4018 (C++ only) Subscripting an array that has been declared @code{register}.
4021 (C++ only) Taking the address of a variable that has been declared
4025 (C++ only) A base class is not initialized in the copy constructor
4030 @item -Wchar-subscripts
4031 @opindex Wchar-subscripts
4032 @opindex Wno-char-subscripts
4033 Warn if an array subscript has type @code{char}. This is a common cause
4034 of error, as programmers often forget that this type is signed on some
4036 This warning is enabled by @option{-Wall}.
4040 Warn about an invalid memory access that is found by Pointer Bounds Checker
4041 (@option{-fcheck-pointer-bounds}).
4043 @item -Wno-coverage-mismatch
4044 @opindex Wno-coverage-mismatch
4045 Warn if feedback profiles do not match when using the
4046 @option{-fprofile-use} option.
4047 If a source file is changed between compiling with @option{-fprofile-gen} and
4048 with @option{-fprofile-use}, the files with the profile feedback can fail
4049 to match the source file and GCC cannot use the profile feedback
4050 information. By default, this warning is enabled and is treated as an
4051 error. @option{-Wno-coverage-mismatch} can be used to disable the
4052 warning or @option{-Wno-error=coverage-mismatch} can be used to
4053 disable the error. Disabling the error for this warning can result in
4054 poorly optimized code and is useful only in the
4055 case of very minor changes such as bug fixes to an existing code-base.
4056 Completely disabling the warning is not recommended.
4059 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
4061 Suppress warning messages emitted by @code{#warning} directives.
4063 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
4064 @opindex Wdouble-promotion
4065 @opindex Wno-double-promotion
4066 Give a warning when a value of type @code{float} is implicitly
4067 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
4068 floating-point unit implement @code{float} in hardware, but emulate
4069 @code{double} in software. On such a machine, doing computations
4070 using @code{double} values is much more expensive because of the
4071 overhead required for software emulation.
4073 It is easy to accidentally do computations with @code{double} because
4074 floating-point literals are implicitly of type @code{double}. For
4078 float area(float radius)
4080 return 3.14159 * radius * radius;
4084 the compiler performs the entire computation with @code{double}
4085 because the floating-point literal is a @code{double}.
4087 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4088 @opindex Wduplicate-decl-specifier
4089 @opindex Wno-duplicate-decl-specifier
4090 Warn if a declaration has duplicate @code{const}, @code{volatile},
4091 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4095 @itemx -Wformat=@var{n}
4098 @opindex ffreestanding
4099 @opindex fno-builtin
4101 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4102 the arguments supplied have types appropriate to the format string
4103 specified, and that the conversions specified in the format string make
4104 sense. This includes standard functions, and others specified by format
4105 attributes (@pxref{Function Attributes}), in the @code{printf},
4106 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4107 not in the C standard) families (or other target-specific families).
4108 Which functions are checked without format attributes having been
4109 specified depends on the standard version selected, and such checks of
4110 functions without the attribute specified are disabled by
4111 @option{-ffreestanding} or @option{-fno-builtin}.
4113 The formats are checked against the format features supported by GNU
4114 libc version 2.2. These include all ISO C90 and C99 features, as well
4115 as features from the Single Unix Specification and some BSD and GNU
4116 extensions. Other library implementations may not support all these
4117 features; GCC does not support warning about features that go beyond a
4118 particular library's limitations. However, if @option{-Wpedantic} is used
4119 with @option{-Wformat}, warnings are given about format features not
4120 in the selected standard version (but not for @code{strfmon} formats,
4121 since those are not in any version of the C standard). @xref{C Dialect
4122 Options,,Options Controlling C Dialect}.
4129 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4130 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4131 @option{-Wformat} also checks for null format arguments for several
4132 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4133 aspects of this level of format checking can be disabled by the
4134 options: @option{-Wno-format-contains-nul},
4135 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4136 @option{-Wformat} is enabled by @option{-Wall}.
4138 @item -Wno-format-contains-nul
4139 @opindex Wno-format-contains-nul
4140 @opindex Wformat-contains-nul
4141 If @option{-Wformat} is specified, do not warn about format strings that
4144 @item -Wno-format-extra-args
4145 @opindex Wno-format-extra-args
4146 @opindex Wformat-extra-args
4147 If @option{-Wformat} is specified, do not warn about excess arguments to a
4148 @code{printf} or @code{scanf} format function. The C standard specifies
4149 that such arguments are ignored.
4151 Where the unused arguments lie between used arguments that are
4152 specified with @samp{$} operand number specifications, normally
4153 warnings are still given, since the implementation could not know what
4154 type to pass to @code{va_arg} to skip the unused arguments. However,
4155 in the case of @code{scanf} formats, this option suppresses the
4156 warning if the unused arguments are all pointers, since the Single
4157 Unix Specification says that such unused arguments are allowed.
4159 @item -Wformat-overflow
4160 @itemx -Wformat-overflow=@var{level}
4161 @opindex Wformat-overflow
4162 @opindex Wno-format-overflow
4163 Warn about calls to formatted input/output functions such as @code{sprintf}
4164 and @code{vsprintf} that might overflow the destination buffer. When the
4165 exact number of bytes written by a format directive cannot be determined
4166 at compile-time it is estimated based on heuristics that depend on the
4167 @var{level} argument and on optimization. While enabling optimization
4168 will in most cases improve the accuracy of the warning, it may also
4169 result in false positives.
4172 @item -Wformat-overflow
4173 @itemx -Wformat-overflow=1
4174 @opindex Wformat-overflow
4175 @opindex Wno-format-overflow
4176 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4177 employs a conservative approach that warns only about calls that most
4178 likely overflow the buffer. At this level, numeric arguments to format
4179 directives with unknown values are assumed to have the value of one, and
4180 strings of unknown length to be empty. Numeric arguments that are known
4181 to be bounded to a subrange of their type, or string arguments whose output
4182 is bounded either by their directive's precision or by a finite set of
4183 string literals, are assumed to take on the value within the range that
4184 results in the most bytes on output. For example, the call to @code{sprintf}
4185 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4186 the terminating NUL character (@code{'\0'}) appended by the function
4187 to the destination buffer will be written past its end. Increasing
4188 the size of the buffer by a single byte is sufficient to avoid the
4189 warning, though it may not be sufficient to avoid the overflow.
4192 void f (int a, int b)
4195 sprintf (buf, "a = %i, b = %i\n", a, b);
4199 @item -Wformat-overflow=2
4200 Level @var{2} warns also about calls that might overflow the destination
4201 buffer given an argument of sufficient length or magnitude. At level
4202 @var{2}, unknown numeric arguments are assumed to have the minimum
4203 representable value for signed types with a precision greater than 1, and
4204 the maximum representable value otherwise. Unknown string arguments whose
4205 length cannot be assumed to be bounded either by the directive's precision,
4206 or by a finite set of string literals they may evaluate to, or the character
4207 array they may point to, are assumed to be 1 character long.
4209 At level @var{2}, the call in the example above is again diagnosed, but
4210 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4211 @code{%i} directive will write some of its digits beyond the end of
4212 the destination buffer. To make the call safe regardless of the values
4213 of the two variables, the size of the destination buffer must be increased
4214 to at least 34 bytes. GCC includes the minimum size of the buffer in
4215 an informational note following the warning.
4217 An alternative to increasing the size of the destination buffer is to
4218 constrain the range of formatted values. The maximum length of string
4219 arguments can be bounded by specifying the precision in the format
4220 directive. When numeric arguments of format directives can be assumed
4221 to be bounded by less than the precision of their type, choosing
4222 an appropriate length modifier to the format specifier will reduce
4223 the required buffer size. For example, if @var{a} and @var{b} in the
4224 example above can be assumed to be within the precision of
4225 the @code{short int} type then using either the @code{%hi} format
4226 directive or casting the argument to @code{short} reduces the maximum
4227 required size of the buffer to 24 bytes.
4230 void f (int a, int b)
4233 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4238 @item -Wno-format-zero-length
4239 @opindex Wno-format-zero-length
4240 @opindex Wformat-zero-length
4241 If @option{-Wformat} is specified, do not warn about zero-length formats.
4242 The C standard specifies that zero-length formats are allowed.
4247 Enable @option{-Wformat} plus additional format checks. Currently
4248 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4251 @item -Wformat-nonliteral
4252 @opindex Wformat-nonliteral
4253 @opindex Wno-format-nonliteral
4254 If @option{-Wformat} is specified, also warn if the format string is not a
4255 string literal and so cannot be checked, unless the format function
4256 takes its format arguments as a @code{va_list}.
4258 @item -Wformat-security
4259 @opindex Wformat-security
4260 @opindex Wno-format-security
4261 If @option{-Wformat} is specified, also warn about uses of format
4262 functions that represent possible security problems. At present, this
4263 warns about calls to @code{printf} and @code{scanf} functions where the
4264 format string is not a string literal and there are no format arguments,
4265 as in @code{printf (foo);}. This may be a security hole if the format
4266 string came from untrusted input and contains @samp{%n}. (This is
4267 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4268 in future warnings may be added to @option{-Wformat-security} that are not
4269 included in @option{-Wformat-nonliteral}.)
4271 @item -Wformat-signedness
4272 @opindex Wformat-signedness
4273 @opindex Wno-format-signedness
4274 If @option{-Wformat} is specified, also warn if the format string
4275 requires an unsigned argument and the argument is signed and vice versa.
4277 @item -Wformat-truncation
4278 @itemx -Wformat-truncation=@var{level}
4279 @opindex Wformat-truncation
4280 @opindex Wno-format-truncation
4281 Warn about calls to formatted input/output functions such as @code{snprintf}
4282 and @code{vsnprintf} that might result in output truncation. When the exact
4283 number of bytes written by a format directive cannot be determined at
4284 compile-time it is estimated based on heuristics that depend on
4285 the @var{level} argument and on optimization. While enabling optimization
4286 will in most cases improve the accuracy of the warning, it may also result
4287 in false positives. Except as noted otherwise, the option uses the same
4288 logic @option{-Wformat-overflow}.
4291 @item -Wformat-truncation
4292 @itemx -Wformat-truncation=1
4293 @opindex Wformat-truncation
4294 @opindex Wno-format-overflow
4295 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4296 employs a conservative approach that warns only about calls to bounded
4297 functions whose return value is unused and that will most likely result
4298 in output truncation.
4300 @item -Wformat-truncation=2
4301 Level @var{2} warns also about calls to bounded functions whose return
4302 value is used and that might result in truncation given an argument of
4303 sufficient length or magnitude.
4307 @opindex Wformat-y2k
4308 @opindex Wno-format-y2k
4309 If @option{-Wformat} is specified, also warn about @code{strftime}
4310 formats that may yield only a two-digit year.
4315 @opindex Wno-nonnull
4316 Warn about passing a null pointer for arguments marked as
4317 requiring a non-null value by the @code{nonnull} function attribute.
4319 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4320 can be disabled with the @option{-Wno-nonnull} option.
4322 @item -Wnonnull-compare
4323 @opindex Wnonnull-compare
4324 @opindex Wno-nonnull-compare
4325 Warn when comparing an argument marked with the @code{nonnull}
4326 function attribute against null inside the function.
4328 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4329 can be disabled with the @option{-Wno-nonnull-compare} option.
4331 @item -Wnull-dereference
4332 @opindex Wnull-dereference
4333 @opindex Wno-null-dereference
4334 Warn if the compiler detects paths that trigger erroneous or
4335 undefined behavior due to dereferencing a null pointer. This option
4336 is only active when @option{-fdelete-null-pointer-checks} is active,
4337 which is enabled by optimizations in most targets. The precision of
4338 the warnings depends on the optimization options used.
4340 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4342 @opindex Wno-init-self
4343 Warn about uninitialized variables that are initialized with themselves.
4344 Note this option can only be used with the @option{-Wuninitialized} option.
4346 For example, GCC warns about @code{i} being uninitialized in the
4347 following snippet only when @option{-Winit-self} has been specified:
4358 This warning is enabled by @option{-Wall} in C++.
4360 @item -Wimplicit-int @r{(C and Objective-C only)}
4361 @opindex Wimplicit-int
4362 @opindex Wno-implicit-int
4363 Warn when a declaration does not specify a type.
4364 This warning is enabled by @option{-Wall}.
4366 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4367 @opindex Wimplicit-function-declaration
4368 @opindex Wno-implicit-function-declaration
4369 Give a warning whenever a function is used before being declared. In
4370 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4371 enabled by default and it is made into an error by
4372 @option{-pedantic-errors}. This warning is also enabled by
4375 @item -Wimplicit @r{(C and Objective-C only)}
4377 @opindex Wno-implicit
4378 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4379 This warning is enabled by @option{-Wall}.
4381 @item -Wimplicit-fallthrough
4382 @opindex Wimplicit-fallthrough
4383 @opindex Wno-implicit-fallthrough
4384 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4385 and @option{-Wno-implicit-fallthrough} is the same as
4386 @option{-Wimplicit-fallthrough=0}.
4388 @item -Wimplicit-fallthrough=@var{n}
4389 @opindex Wimplicit-fallthrough=
4390 Warn when a switch case falls through. For example:
4408 This warning does not warn when the last statement of a case cannot
4409 fall through, e.g. when there is a return statement or a call to function
4410 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4411 also takes into account control flow statements, such as ifs, and only
4412 warns when appropriate. E.g.@:
4422 @} else if (i < 1) @{
4432 Since there are occasions where a switch case fall through is desirable,
4433 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4434 to be used along with a null statement to suppress this warning that
4435 would normally occur:
4443 __attribute__ ((fallthrough));
4450 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4451 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4452 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4453 Instead of these attributes, it is also possible to add a fallthrough comment
4454 to silence the warning. The whole body of the C or C++ style comment should
4455 match the given regular expressions listed below. The option argument @var{n}
4456 specifies what kind of comments are accepted:
4460 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4462 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4463 expression, any comment is used as fallthrough comment.
4465 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4466 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4468 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4469 following regular expressions:
4473 @item @code{-fallthrough}
4475 @item @code{@@fallthrough@@}
4477 @item @code{lint -fallthrough[ \t]*}
4479 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4481 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4483 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4487 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4488 following regular expressions:
4492 @item @code{-fallthrough}
4494 @item @code{@@fallthrough@@}
4496 @item @code{lint -fallthrough[ \t]*}
4498 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4502 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4503 fallthrough comments, only attributes disable the warning.
4507 The comment needs to be followed after optional whitespace and other comments
4508 by @code{case} or @code{default} keywords or by a user label that precedes some
4509 @code{case} or @code{default} label.
4524 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4526 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4527 @opindex Wif-not-aligned
4528 @opindex Wno-if-not-aligned
4529 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4530 should be issued. This is is enabled by default.
4531 Use @option{-Wno-if-not-aligned} to disable it.
4533 @item -Wignored-qualifiers @r{(C and C++ only)}
4534 @opindex Wignored-qualifiers
4535 @opindex Wno-ignored-qualifiers
4536 Warn if the return type of a function has a type qualifier
4537 such as @code{const}. For ISO C such a type qualifier has no effect,
4538 since the value returned by a function is not an lvalue.
4539 For C++, the warning is only emitted for scalar types or @code{void}.
4540 ISO C prohibits qualified @code{void} return types on function
4541 definitions, so such return types always receive a warning
4542 even without this option.
4544 This warning is also enabled by @option{-Wextra}.
4546 @item -Wignored-attributes @r{(C and C++ only)}
4547 @opindex Wignored-attributes
4548 @opindex Wno-ignored-attributes
4549 Warn when an attribute is ignored. This is different from the
4550 @option{-Wattributes} option in that it warns whenever the compiler decides
4551 to drop an attribute, not that the attribute is either unknown, used in a
4552 wrong place, etc. This warning is enabled by default.
4557 Warn if the type of @code{main} is suspicious. @code{main} should be
4558 a function with external linkage, returning int, taking either zero
4559 arguments, two, or three arguments of appropriate types. This warning
4560 is enabled by default in C++ and is enabled by either @option{-Wall}
4561 or @option{-Wpedantic}.
4563 @item -Wmisleading-indentation @r{(C and C++ only)}
4564 @opindex Wmisleading-indentation
4565 @opindex Wno-misleading-indentation
4566 Warn when the indentation of the code does not reflect the block structure.
4567 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4568 @code{for} clauses with a guarded statement that does not use braces,
4569 followed by an unguarded statement with the same indentation.
4571 In the following example, the call to ``bar'' is misleadingly indented as
4572 if it were guarded by the ``if'' conditional.
4575 if (some_condition ())
4577 bar (); /* Gotcha: this is not guarded by the "if". */
4580 In the case of mixed tabs and spaces, the warning uses the
4581 @option{-ftabstop=} option to determine if the statements line up
4584 The warning is not issued for code involving multiline preprocessor logic
4585 such as the following example.
4590 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4596 The warning is not issued after a @code{#line} directive, since this
4597 typically indicates autogenerated code, and no assumptions can be made
4598 about the layout of the file that the directive references.
4600 This warning is enabled by @option{-Wall} in C and C++.
4602 @item -Wmissing-attributes
4603 @opindex Wmissing-attributes
4604 @opindex Wno-missing-attributes
4605 Warn when a declaration of a function is missing one or more attributes
4606 that a related function is declared with and whose absence may adversely
4607 affect the correctness or efficiency of generated code. For example, in
4608 C++, the warning is issued when an explicit specialization of a primary
4609 template declared with attribute @code{alloc_align}, @code{alloc_size},
4610 @code{assume_aligned}, @code{format}, @code{format_arg}, @code{malloc},
4611 or @code{nonnull} is declared without it. Attributes @code{deprecated},
4612 @code{error}, and @code{warning} suppress the warning.
4613 (@pxref{Function Attributes}).
4615 @option{-Wmissing-attributes} is enabled by @option{-Wall}.
4617 For example, since the declaration of the primary function template
4618 below makes use of both attribute @code{malloc} and @code{alloc_size}
4619 the declaration of the explicit specialization of the template is
4620 diagnosed because it is missing one of the attributes.
4624 T* __attribute__ ((malloc, alloc_size (1)))
4628 void* __attribute__ ((malloc)) // missing alloc_size
4629 allocate<void> (size_t);
4632 @item -Wmissing-braces
4633 @opindex Wmissing-braces
4634 @opindex Wno-missing-braces
4635 Warn if an aggregate or union initializer is not fully bracketed. In
4636 the following example, the initializer for @code{a} is not fully
4637 bracketed, but that for @code{b} is fully bracketed. This warning is
4638 enabled by @option{-Wall} in C.
4641 int a[2][2] = @{ 0, 1, 2, 3 @};
4642 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4645 This warning is enabled by @option{-Wall}.
4647 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4648 @opindex Wmissing-include-dirs
4649 @opindex Wno-missing-include-dirs
4650 Warn if a user-supplied include directory does not exist.
4652 @item -Wmultistatement-macros
4653 @opindex Wmultistatement-macros
4654 @opindex Wno-multistatement-macros
4655 Warn about unsafe multiple statement macros that appear to be guarded
4656 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
4657 @code{while}, in which only the first statement is actually guarded after
4658 the macro is expanded.
4663 #define DOIT x++; y++
4668 will increment @code{y} unconditionally, not just when @code{c} holds.
4669 The can usually be fixed by wrapping the macro in a do-while loop:
4671 #define DOIT do @{ x++; y++; @} while (0)
4676 This warning is enabled by @option{-Wall} in C and C++.
4679 @opindex Wparentheses
4680 @opindex Wno-parentheses
4681 Warn if parentheses are omitted in certain contexts, such
4682 as when there is an assignment in a context where a truth value
4683 is expected, or when operators are nested whose precedence people
4684 often get confused about.
4686 Also warn if a comparison like @code{x<=y<=z} appears; this is
4687 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4688 interpretation from that of ordinary mathematical notation.
4690 Also warn for dangerous uses of the GNU extension to
4691 @code{?:} with omitted middle operand. When the condition
4692 in the @code{?}: operator is a boolean expression, the omitted value is
4693 always 1. Often programmers expect it to be a value computed
4694 inside the conditional expression instead.
4696 For C++ this also warns for some cases of unnecessary parentheses in
4697 declarations, which can indicate an attempt at a function call instead
4701 // Declares a local variable called mymutex.
4702 std::unique_lock<std::mutex> (mymutex);
4703 // User meant std::unique_lock<std::mutex> lock (mymutex);
4707 This warning is enabled by @option{-Wall}.
4709 @item -Wsequence-point
4710 @opindex Wsequence-point
4711 @opindex Wno-sequence-point
4712 Warn about code that may have undefined semantics because of violations
4713 of sequence point rules in the C and C++ standards.
4715 The C and C++ standards define the order in which expressions in a C/C++
4716 program are evaluated in terms of @dfn{sequence points}, which represent
4717 a partial ordering between the execution of parts of the program: those
4718 executed before the sequence point, and those executed after it. These
4719 occur after the evaluation of a full expression (one which is not part
4720 of a larger expression), after the evaluation of the first operand of a
4721 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4722 function is called (but after the evaluation of its arguments and the
4723 expression denoting the called function), and in certain other places.
4724 Other than as expressed by the sequence point rules, the order of
4725 evaluation of subexpressions of an expression is not specified. All
4726 these rules describe only a partial order rather than a total order,
4727 since, for example, if two functions are called within one expression
4728 with no sequence point between them, the order in which the functions
4729 are called is not specified. However, the standards committee have
4730 ruled that function calls do not overlap.
4732 It is not specified when between sequence points modifications to the
4733 values of objects take effect. Programs whose behavior depends on this
4734 have undefined behavior; the C and C++ standards specify that ``Between
4735 the previous and next sequence point an object shall have its stored
4736 value modified at most once by the evaluation of an expression.
4737 Furthermore, the prior value shall be read only to determine the value
4738 to be stored.''. If a program breaks these rules, the results on any
4739 particular implementation are entirely unpredictable.
4741 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4742 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4743 diagnosed by this option, and it may give an occasional false positive
4744 result, but in general it has been found fairly effective at detecting
4745 this sort of problem in programs.
4747 The C++17 standard will define the order of evaluation of operands in
4748 more cases: in particular it requires that the right-hand side of an
4749 assignment be evaluated before the left-hand side, so the above
4750 examples are no longer undefined. But this warning will still warn
4751 about them, to help people avoid writing code that is undefined in C
4752 and earlier revisions of C++.
4754 The standard is worded confusingly, therefore there is some debate
4755 over the precise meaning of the sequence point rules in subtle cases.
4756 Links to discussions of the problem, including proposed formal
4757 definitions, may be found on the GCC readings page, at
4758 @uref{http://gcc.gnu.org/@/readings.html}.
4760 This warning is enabled by @option{-Wall} for C and C++.
4762 @item -Wno-return-local-addr
4763 @opindex Wno-return-local-addr
4764 @opindex Wreturn-local-addr
4765 Do not warn about returning a pointer (or in C++, a reference) to a
4766 variable that goes out of scope after the function returns.
4769 @opindex Wreturn-type
4770 @opindex Wno-return-type
4771 Warn whenever a function is defined with a return type that defaults
4772 to @code{int}. Also warn about any @code{return} statement with no
4773 return value in a function whose return type is not @code{void}
4774 (falling off the end of the function body is considered returning
4777 For C only, warn about a @code{return} statement with an expression in a
4778 function whose return type is @code{void}, unless the expression type is
4779 also @code{void}. As a GNU extension, the latter case is accepted
4780 without a warning unless @option{-Wpedantic} is used.
4782 For C++, a function without return type always produces a diagnostic
4783 message, even when @option{-Wno-return-type} is specified. The only
4784 exceptions are @code{main} and functions defined in system headers.
4786 This warning is enabled by @option{-Wall}.
4788 @item -Wshift-count-negative
4789 @opindex Wshift-count-negative
4790 @opindex Wno-shift-count-negative
4791 Warn if shift count is negative. This warning is enabled by default.
4793 @item -Wshift-count-overflow
4794 @opindex Wshift-count-overflow
4795 @opindex Wno-shift-count-overflow
4796 Warn if shift count >= width of type. This warning is enabled by default.
4798 @item -Wshift-negative-value
4799 @opindex Wshift-negative-value
4800 @opindex Wno-shift-negative-value
4801 Warn if left shifting a negative value. This warning is enabled by
4802 @option{-Wextra} in C99 and C++11 modes (and newer).
4804 @item -Wshift-overflow
4805 @itemx -Wshift-overflow=@var{n}
4806 @opindex Wshift-overflow
4807 @opindex Wno-shift-overflow
4808 Warn about left shift overflows. This warning is enabled by
4809 default in C99 and C++11 modes (and newer).
4812 @item -Wshift-overflow=1
4813 This is the warning level of @option{-Wshift-overflow} and is enabled
4814 by default in C99 and C++11 modes (and newer). This warning level does
4815 not warn about left-shifting 1 into the sign bit. (However, in C, such
4816 an overflow is still rejected in contexts where an integer constant expression
4819 @item -Wshift-overflow=2
4820 This warning level also warns about left-shifting 1 into the sign bit,
4821 unless C++14 mode is active.
4827 Warn whenever a @code{switch} statement has an index of enumerated type
4828 and lacks a @code{case} for one or more of the named codes of that
4829 enumeration. (The presence of a @code{default} label prevents this
4830 warning.) @code{case} labels outside the enumeration range also
4831 provoke warnings when this option is used (even if there is a
4832 @code{default} label).
4833 This warning is enabled by @option{-Wall}.
4835 @item -Wswitch-default
4836 @opindex Wswitch-default
4837 @opindex Wno-switch-default
4838 Warn whenever a @code{switch} statement does not have a @code{default}
4842 @opindex Wswitch-enum
4843 @opindex Wno-switch-enum
4844 Warn whenever a @code{switch} statement has an index of enumerated type
4845 and lacks a @code{case} for one or more of the named codes of that
4846 enumeration. @code{case} labels outside the enumeration range also
4847 provoke warnings when this option is used. The only difference
4848 between @option{-Wswitch} and this option is that this option gives a
4849 warning about an omitted enumeration code even if there is a
4850 @code{default} label.
4853 @opindex Wswitch-bool
4854 @opindex Wno-switch-bool
4855 Warn whenever a @code{switch} statement has an index of boolean type
4856 and the case values are outside the range of a boolean type.
4857 It is possible to suppress this warning by casting the controlling
4858 expression to a type other than @code{bool}. For example:
4861 switch ((int) (a == 4))
4867 This warning is enabled by default for C and C++ programs.
4869 @item -Wswitch-unreachable
4870 @opindex Wswitch-unreachable
4871 @opindex Wno-switch-unreachable
4872 Warn whenever a @code{switch} statement contains statements between the
4873 controlling expression and the first case label, which will never be
4874 executed. For example:
4886 @option{-Wswitch-unreachable} does not warn if the statement between the
4887 controlling expression and the first case label is just a declaration:
4900 This warning is enabled by default for C and C++ programs.
4902 @item -Wsync-nand @r{(C and C++ only)}
4904 @opindex Wno-sync-nand
4905 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4906 built-in functions are used. These functions changed semantics in GCC 4.4.
4908 @item -Wunused-but-set-parameter
4909 @opindex Wunused-but-set-parameter
4910 @opindex Wno-unused-but-set-parameter
4911 Warn whenever a function parameter is assigned to, but otherwise unused
4912 (aside from its declaration).
4914 To suppress this warning use the @code{unused} attribute
4915 (@pxref{Variable Attributes}).
4917 This warning is also enabled by @option{-Wunused} together with
4920 @item -Wunused-but-set-variable
4921 @opindex Wunused-but-set-variable
4922 @opindex Wno-unused-but-set-variable
4923 Warn whenever a local variable is assigned to, but otherwise unused
4924 (aside from its declaration).
4925 This warning is enabled by @option{-Wall}.
4927 To suppress this warning use the @code{unused} attribute
4928 (@pxref{Variable Attributes}).
4930 This warning is also enabled by @option{-Wunused}, which is enabled
4933 @item -Wunused-function
4934 @opindex Wunused-function
4935 @opindex Wno-unused-function
4936 Warn whenever a static function is declared but not defined or a
4937 non-inline static function is unused.
4938 This warning is enabled by @option{-Wall}.
4940 @item -Wunused-label
4941 @opindex Wunused-label
4942 @opindex Wno-unused-label
4943 Warn whenever a label is declared but not used.
4944 This warning is enabled by @option{-Wall}.
4946 To suppress this warning use the @code{unused} attribute
4947 (@pxref{Variable Attributes}).
4949 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4950 @opindex Wunused-local-typedefs
4951 Warn when a typedef locally defined in a function is not used.
4952 This warning is enabled by @option{-Wall}.
4954 @item -Wunused-parameter
4955 @opindex Wunused-parameter
4956 @opindex Wno-unused-parameter
4957 Warn whenever a function parameter is unused aside from its declaration.
4959 To suppress this warning use the @code{unused} attribute
4960 (@pxref{Variable Attributes}).
4962 @item -Wno-unused-result
4963 @opindex Wunused-result
4964 @opindex Wno-unused-result
4965 Do not warn if a caller of a function marked with attribute
4966 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4967 its return value. The default is @option{-Wunused-result}.
4969 @item -Wunused-variable
4970 @opindex Wunused-variable
4971 @opindex Wno-unused-variable
4972 Warn whenever a local or static variable is unused aside from its
4973 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4974 but not for C++. This warning is enabled by @option{-Wall}.
4976 To suppress this warning use the @code{unused} attribute
4977 (@pxref{Variable Attributes}).
4979 @item -Wunused-const-variable
4980 @itemx -Wunused-const-variable=@var{n}
4981 @opindex Wunused-const-variable
4982 @opindex Wno-unused-const-variable
4983 Warn whenever a constant static variable is unused aside from its declaration.
4984 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4985 for C, but not for C++. In C this declares variable storage, but in C++ this
4986 is not an error since const variables take the place of @code{#define}s.
4988 To suppress this warning use the @code{unused} attribute
4989 (@pxref{Variable Attributes}).
4992 @item -Wunused-const-variable=1
4993 This is the warning level that is enabled by @option{-Wunused-variable} for
4994 C. It warns only about unused static const variables defined in the main
4995 compilation unit, but not about static const variables declared in any
4998 @item -Wunused-const-variable=2
4999 This warning level also warns for unused constant static variables in
5000 headers (excluding system headers). This is the warning level of
5001 @option{-Wunused-const-variable} and must be explicitly requested since
5002 in C++ this isn't an error and in C it might be harder to clean up all
5006 @item -Wunused-value
5007 @opindex Wunused-value
5008 @opindex Wno-unused-value
5009 Warn whenever a statement computes a result that is explicitly not
5010 used. To suppress this warning cast the unused expression to
5011 @code{void}. This includes an expression-statement or the left-hand
5012 side of a comma expression that contains no side effects. For example,
5013 an expression such as @code{x[i,j]} causes a warning, while
5014 @code{x[(void)i,j]} does not.
5016 This warning is enabled by @option{-Wall}.
5021 All the above @option{-Wunused} options combined.
5023 In order to get a warning about an unused function parameter, you must
5024 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
5025 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
5027 @item -Wuninitialized
5028 @opindex Wuninitialized
5029 @opindex Wno-uninitialized
5030 Warn if an automatic variable is used without first being initialized
5031 or if a variable may be clobbered by a @code{setjmp} call. In C++,
5032 warn if a non-static reference or non-static @code{const} member
5033 appears in a class without constructors.
5035 If you want to warn about code that uses the uninitialized value of the
5036 variable in its own initializer, use the @option{-Winit-self} option.
5038 These warnings occur for individual uninitialized or clobbered
5039 elements of structure, union or array variables as well as for
5040 variables that are uninitialized or clobbered as a whole. They do
5041 not occur for variables or elements declared @code{volatile}. Because
5042 these warnings depend on optimization, the exact variables or elements
5043 for which there are warnings depends on the precise optimization
5044 options and version of GCC used.
5046 Note that there may be no warning about a variable that is used only
5047 to compute a value that itself is never used, because such
5048 computations may be deleted by data flow analysis before the warnings
5051 @item -Winvalid-memory-model
5052 @opindex Winvalid-memory-model
5053 @opindex Wno-invalid-memory-model
5054 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
5055 and the C11 atomic generic functions with a memory consistency argument
5056 that is either invalid for the operation or outside the range of values
5057 of the @code{memory_order} enumeration. For example, since the
5058 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
5059 defined for the relaxed, release, and sequentially consistent memory
5060 orders the following code is diagnosed:
5065 __atomic_store_n (i, 0, memory_order_consume);
5069 @option{-Winvalid-memory-model} is enabled by default.
5071 @item -Wmaybe-uninitialized
5072 @opindex Wmaybe-uninitialized
5073 @opindex Wno-maybe-uninitialized
5074 For an automatic (i.e.@ local) variable, if there exists a path from the
5075 function entry to a use of the variable that is initialized, but there exist
5076 some other paths for which the variable is not initialized, the compiler
5077 emits a warning if it cannot prove the uninitialized paths are not
5078 executed at run time.
5080 These warnings are only possible in optimizing compilation, because otherwise
5081 GCC does not keep track of the state of variables.
5083 These warnings are made optional because GCC may not be able to determine when
5084 the code is correct in spite of appearing to have an error. Here is one
5085 example of how this can happen:
5105 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
5106 always initialized, but GCC doesn't know this. To suppress the
5107 warning, you need to provide a default case with assert(0) or
5110 @cindex @code{longjmp} warnings
5111 This option also warns when a non-volatile automatic variable might be
5112 changed by a call to @code{longjmp}.
5113 The compiler sees only the calls to @code{setjmp}. It cannot know
5114 where @code{longjmp} will be called; in fact, a signal handler could
5115 call it at any point in the code. As a result, you may get a warning
5116 even when there is in fact no problem because @code{longjmp} cannot
5117 in fact be called at the place that would cause a problem.
5119 Some spurious warnings can be avoided if you declare all the functions
5120 you use that never return as @code{noreturn}. @xref{Function
5123 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5125 @item -Wunknown-pragmas
5126 @opindex Wunknown-pragmas
5127 @opindex Wno-unknown-pragmas
5128 @cindex warning for unknown pragmas
5129 @cindex unknown pragmas, warning
5130 @cindex pragmas, warning of unknown
5131 Warn when a @code{#pragma} directive is encountered that is not understood by
5132 GCC@. If this command-line option is used, warnings are even issued
5133 for unknown pragmas in system header files. This is not the case if
5134 the warnings are only enabled by the @option{-Wall} command-line option.
5137 @opindex Wno-pragmas
5139 Do not warn about misuses of pragmas, such as incorrect parameters,
5140 invalid syntax, or conflicts between pragmas. See also
5141 @option{-Wunknown-pragmas}.
5143 @item -Wstrict-aliasing
5144 @opindex Wstrict-aliasing
5145 @opindex Wno-strict-aliasing
5146 This option is only active when @option{-fstrict-aliasing} is active.
5147 It warns about code that might break the strict aliasing rules that the
5148 compiler is using for optimization. The warning does not catch all
5149 cases, but does attempt to catch the more common pitfalls. It is
5150 included in @option{-Wall}.
5151 It is equivalent to @option{-Wstrict-aliasing=3}
5153 @item -Wstrict-aliasing=n
5154 @opindex Wstrict-aliasing=n
5155 This option is only active when @option{-fstrict-aliasing} is active.
5156 It warns about code that might break the strict aliasing rules that the
5157 compiler is using for optimization.
5158 Higher levels correspond to higher accuracy (fewer false positives).
5159 Higher levels also correspond to more effort, similar to the way @option{-O}
5161 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5163 Level 1: Most aggressive, quick, least accurate.
5164 Possibly useful when higher levels
5165 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5166 false negatives. However, it has many false positives.
5167 Warns for all pointer conversions between possibly incompatible types,
5168 even if never dereferenced. Runs in the front end only.
5170 Level 2: Aggressive, quick, not too precise.
5171 May still have many false positives (not as many as level 1 though),
5172 and few false negatives (but possibly more than level 1).
5173 Unlike level 1, it only warns when an address is taken. Warns about
5174 incomplete types. Runs in the front end only.
5176 Level 3 (default for @option{-Wstrict-aliasing}):
5177 Should have very few false positives and few false
5178 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5179 Takes care of the common pun+dereference pattern in the front end:
5180 @code{*(int*)&some_float}.
5181 If optimization is enabled, it also runs in the back end, where it deals
5182 with multiple statement cases using flow-sensitive points-to information.
5183 Only warns when the converted pointer is dereferenced.
5184 Does not warn about incomplete types.
5186 @item -Wstrict-overflow
5187 @itemx -Wstrict-overflow=@var{n}
5188 @opindex Wstrict-overflow
5189 @opindex Wno-strict-overflow
5190 This option is only active when signed overflow is undefined.
5191 It warns about cases where the compiler optimizes based on the
5192 assumption that signed overflow does not occur. Note that it does not
5193 warn about all cases where the code might overflow: it only warns
5194 about cases where the compiler implements some optimization. Thus
5195 this warning depends on the optimization level.
5197 An optimization that assumes that signed overflow does not occur is
5198 perfectly safe if the values of the variables involved are such that
5199 overflow never does, in fact, occur. Therefore this warning can
5200 easily give a false positive: a warning about code that is not
5201 actually a problem. To help focus on important issues, several
5202 warning levels are defined. No warnings are issued for the use of
5203 undefined signed overflow when estimating how many iterations a loop
5204 requires, in particular when determining whether a loop will be
5208 @item -Wstrict-overflow=1
5209 Warn about cases that are both questionable and easy to avoid. For
5210 example the compiler simplifies
5211 @code{x + 1 > x} to @code{1}. This level of
5212 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5213 are not, and must be explicitly requested.
5215 @item -Wstrict-overflow=2
5216 Also warn about other cases where a comparison is simplified to a
5217 constant. For example: @code{abs (x) >= 0}. This can only be
5218 simplified when signed integer overflow is undefined, because
5219 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5220 zero. @option{-Wstrict-overflow} (with no level) is the same as
5221 @option{-Wstrict-overflow=2}.
5223 @item -Wstrict-overflow=3
5224 Also warn about other cases where a comparison is simplified. For
5225 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5227 @item -Wstrict-overflow=4
5228 Also warn about other simplifications not covered by the above cases.
5229 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5231 @item -Wstrict-overflow=5
5232 Also warn about cases where the compiler reduces the magnitude of a
5233 constant involved in a comparison. For example: @code{x + 2 > y} is
5234 simplified to @code{x + 1 >= y}. This is reported only at the
5235 highest warning level because this simplification applies to many
5236 comparisons, so this warning level gives a very large number of
5240 @item -Wstringop-overflow
5241 @itemx -Wstringop-overflow=@var{type}
5242 @opindex Wstringop-overflow
5243 @opindex Wno-stringop-overflow
5244 Warn for calls to string manipulation functions such as @code{memcpy} and
5245 @code{strcpy} that are determined to overflow the destination buffer. The
5246 optional argument is one greater than the type of Object Size Checking to
5247 perform to determine the size of the destination. @xref{Object Size Checking}.
5248 The argument is meaningful only for functions that operate on character arrays
5249 but not for raw memory functions like @code{memcpy} which always make use
5250 of Object Size type-0. The option also warns for calls that specify a size
5251 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5252 The option produces the best results with optimization enabled but can detect
5253 a small subset of simple buffer overflows even without optimization in
5254 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5255 correspond to the standard functions. In any case, the option warns about
5256 just a subset of buffer overflows detected by the corresponding overflow
5257 checking built-ins. For example, the option will issue a warning for
5258 the @code{strcpy} call below because it copies at least 5 characters
5259 (the string @code{"blue"} including the terminating NUL) into the buffer
5263 enum Color @{ blue, purple, yellow @};
5264 const char* f (enum Color clr)
5266 static char buf [4];
5270 case blue: str = "blue"; break;
5271 case purple: str = "purple"; break;
5272 case yellow: str = "yellow"; break;
5275 return strcpy (buf, str); // warning here
5279 Option @option{-Wstringop-overflow=2} is enabled by default.
5282 @item -Wstringop-overflow
5283 @itemx -Wstringop-overflow=1
5284 @opindex Wstringop-overflow
5285 @opindex Wno-stringop-overflow
5286 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5287 to determine the sizes of destination objects. This is the default setting
5288 of the option. At this setting the option will not warn for writes past
5289 the end of subobjects of larger objects accessed by pointers unless the
5290 size of the largest surrounding object is known. When the destination may
5291 be one of several objects it is assumed to be the largest one of them. On
5292 Linux systems, when optimization is enabled at this setting the option warns
5293 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5296 @item -Wstringop-overflow=2
5297 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5298 to determine the sizes of destination objects. At this setting the option
5299 will warn about overflows when writing to members of the largest complete
5300 objects whose exact size is known. It will, however, not warn for excessive
5301 writes to the same members of unknown objects referenced by pointers since
5302 they may point to arrays containing unknown numbers of elements.
5304 @item -Wstringop-overflow=3
5305 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5306 to determine the sizes of destination objects. At this setting the option
5307 warns about overflowing the smallest object or data member. This is the
5308 most restrictive setting of the option that may result in warnings for safe
5311 @item -Wstringop-overflow=4
5312 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5313 to determine the sizes of destination objects. At this setting the option
5314 will warn about overflowing any data members, and when the destination is
5315 one of several objects it uses the size of the largest of them to decide
5316 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5317 setting of the option may result in warnings for benign code.
5320 @item -Wstringop-truncation
5321 @opindex Wstringop-truncation
5322 @opindex Wno-stringop-truncation
5323 Warn for calls to bounded string manipulation functions such as @code{strncat},
5324 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5325 or leave the destination unchanged.
5327 In the following example, the call to @code{strncat} specifies a bound that
5328 is less than the length of the source string. As a result, the copy of
5329 the source will be truncated and so the call is diagnosed. To avoid the
5330 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5333 void append (char *buf, size_t bufsize)
5335 strncat (buf, ".txt", 3);
5339 As another example, the following call to @code{strncpy} results in copying
5340 to @code{d} just the characters preceding the terminating NUL, without
5341 appending the NUL to the end. Assuming the result of @code{strncpy} is
5342 necessarily a NUL-terminated string is a common mistake, and so the call
5343 is diagnosed. To avoid the warning when the result is not expected to be
5344 NUL-terminated, call @code{memcpy} instead.
5347 void copy (char *d, const char *s)
5349 strncpy (d, s, strlen (s));
5353 In the following example, the call to @code{strncpy} specifies the size
5354 of the destination buffer as the bound. If the length of the source
5355 string is equal to or greater than this size the result of the copy will
5356 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5357 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5358 element of the buffer to @code{NUL}.
5361 void copy (const char *s)
5364 strncpy (buf, s, sizeof buf);
5369 In situations where a character array is intended to store a sequence
5370 of bytes with no terminating @code{NUL} such an array may be annotated
5371 with attribute @code{nonstring} to avoid this warning. Such arrays,
5372 however, are not suitable arguments to functions that expect
5373 @code{NUL}-terminated strings. To help detect accidental misuses of
5374 such arrays GCC issues warnings unless it can prove that the use is
5375 safe. @xref{Common Variable Attributes}.
5377 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5378 @opindex Wsuggest-attribute=
5379 @opindex Wno-suggest-attribute=
5380 Warn for cases where adding an attribute may be beneficial. The
5381 attributes currently supported are listed below.
5384 @item -Wsuggest-attribute=pure
5385 @itemx -Wsuggest-attribute=const
5386 @itemx -Wsuggest-attribute=noreturn
5387 @itemx -Wsuggest-attribute=malloc
5388 @opindex Wsuggest-attribute=pure
5389 @opindex Wno-suggest-attribute=pure
5390 @opindex Wsuggest-attribute=const
5391 @opindex Wno-suggest-attribute=const
5392 @opindex Wsuggest-attribute=noreturn
5393 @opindex Wno-suggest-attribute=noreturn
5394 @opindex Wsuggest-attribute=malloc
5395 @opindex Wno-suggest-attribute=malloc
5397 Warn about functions that might be candidates for attributes
5398 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5399 only warns for functions visible in other compilation units or (in the case of
5400 @code{pure} and @code{const}) if it cannot prove that the function returns
5401 normally. A function returns normally if it doesn't contain an infinite loop or
5402 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5403 requires option @option{-fipa-pure-const}, which is enabled by default at
5404 @option{-O} and higher. Higher optimization levels improve the accuracy
5407 @item -Wsuggest-attribute=format
5408 @itemx -Wmissing-format-attribute
5409 @opindex Wsuggest-attribute=format
5410 @opindex Wmissing-format-attribute
5411 @opindex Wno-suggest-attribute=format
5412 @opindex Wno-missing-format-attribute
5416 Warn about function pointers that might be candidates for @code{format}
5417 attributes. Note these are only possible candidates, not absolute ones.
5418 GCC guesses that function pointers with @code{format} attributes that
5419 are used in assignment, initialization, parameter passing or return
5420 statements should have a corresponding @code{format} attribute in the
5421 resulting type. I.e.@: the left-hand side of the assignment or
5422 initialization, the type of the parameter variable, or the return type
5423 of the containing function respectively should also have a @code{format}
5424 attribute to avoid the warning.
5426 GCC also warns about function definitions that might be
5427 candidates for @code{format} attributes. Again, these are only
5428 possible candidates. GCC guesses that @code{format} attributes
5429 might be appropriate for any function that calls a function like
5430 @code{vprintf} or @code{vscanf}, but this might not always be the
5431 case, and some functions for which @code{format} attributes are
5432 appropriate may not be detected.
5434 @item -Wsuggest-attribute=cold
5435 @opindex Wsuggest-attribute=cold
5436 @opindex Wno-suggest-attribute=cold
5438 Warn about functions that might be candidates for @code{cold} attribute. This
5439 is based on static detection and generally will only warn about functions which
5440 always leads to a call to another @code{cold} function such as wrappers of
5441 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5444 @item -Wsuggest-final-types
5445 @opindex Wno-suggest-final-types
5446 @opindex Wsuggest-final-types
5447 Warn about types with virtual methods where code quality would be improved
5448 if the type were declared with the C++11 @code{final} specifier,
5450 declared in an anonymous namespace. This allows GCC to more aggressively
5451 devirtualize the polymorphic calls. This warning is more effective with link
5452 time optimization, where the information about the class hierarchy graph is
5455 @item -Wsuggest-final-methods
5456 @opindex Wno-suggest-final-methods
5457 @opindex Wsuggest-final-methods
5458 Warn about virtual methods where code quality would be improved if the method
5459 were declared with the C++11 @code{final} specifier,
5460 or, if possible, its type were
5461 declared in an anonymous namespace or with the @code{final} specifier.
5463 more effective with link-time optimization, where the information about the
5464 class hierarchy graph is more complete. It is recommended to first consider
5465 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5468 @item -Wsuggest-override
5469 Warn about overriding virtual functions that are not marked with the override
5473 @opindex Wno-alloc-zero
5474 @opindex Walloc-zero
5475 Warn about calls to allocation functions decorated with attribute
5476 @code{alloc_size} that specify zero bytes, including those to the built-in
5477 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5478 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5479 when called with a zero size differs among implementations (and in the case
5480 of @code{realloc} has been deprecated) relying on it may result in subtle
5481 portability bugs and should be avoided.
5483 @item -Walloc-size-larger-than=@var{n}
5484 Warn about calls to functions decorated with attribute @code{alloc_size}
5485 that attempt to allocate objects larger than the specified number of bytes,
5486 or where the result of the size computation in an integer type with infinite
5487 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5488 may end in one of the standard suffixes designating a multiple of bytes
5489 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5490 @code{MB} and @code{MiB} for megabyte and mebibyte, and so on.
5491 @xref{Function Attributes}.
5496 This option warns on all uses of @code{alloca} in the source.
5498 @item -Walloca-larger-than=@var{n}
5499 This option warns on calls to @code{alloca} that are not bounded by a
5500 controlling predicate limiting its argument of integer type to at most
5501 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5502 Arguments of non-integer types are considered unbounded even if they
5503 appear to be constrained to the expected range.
5505 For example, a bounded case of @code{alloca} could be:
5508 void func (size_t n)
5519 In the above example, passing @code{-Walloca-larger-than=1000} would not
5520 issue a warning because the call to @code{alloca} is known to be at most
5521 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5522 the compiler would emit a warning.
5524 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5525 controlling predicate constraining its integer argument. For example:
5530 void *p = alloca (n);
5535 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5536 a warning, but this time because of the lack of bounds checking.
5538 Note, that even seemingly correct code involving signed integers could
5542 void func (signed int n)
5552 In the above example, @var{n} could be negative, causing a larger than
5553 expected argument to be implicitly cast into the @code{alloca} call.
5555 This option also warns when @code{alloca} is used in a loop.
5557 This warning is not enabled by @option{-Wall}, and is only active when
5558 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5560 See also @option{-Wvla-larger-than=@var{n}}.
5562 @item -Warray-bounds
5563 @itemx -Warray-bounds=@var{n}
5564 @opindex Wno-array-bounds
5565 @opindex Warray-bounds
5566 This option is only active when @option{-ftree-vrp} is active
5567 (default for @option{-O2} and above). It warns about subscripts to arrays
5568 that are always out of bounds. This warning is enabled by @option{-Wall}.
5571 @item -Warray-bounds=1
5572 This is the warning level of @option{-Warray-bounds} and is enabled
5573 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5575 @item -Warray-bounds=2
5576 This warning level also warns about out of bounds access for
5577 arrays at the end of a struct and for arrays accessed through
5578 pointers. This warning level may give a larger number of
5579 false positives and is deactivated by default.
5582 @item -Wattribute-alias
5583 Warn about declarations using the @code{alias} and similar attributes whose
5584 target is incompatible with the type of the alias. @xref{Function Attributes,
5585 ,Declaring Attributes of Functions}.
5587 @item -Wbool-compare
5588 @opindex Wno-bool-compare
5589 @opindex Wbool-compare
5590 Warn about boolean expression compared with an integer value different from
5591 @code{true}/@code{false}. For instance, the following comparison is
5596 if ((n > 1) == 2) @{ @dots{} @}
5598 This warning is enabled by @option{-Wall}.
5600 @item -Wbool-operation
5601 @opindex Wno-bool-operation
5602 @opindex Wbool-operation
5603 Warn about suspicious operations on expressions of a boolean type. For
5604 instance, bitwise negation of a boolean is very likely a bug in the program.
5605 For C, this warning also warns about incrementing or decrementing a boolean,
5606 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5607 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
5609 This warning is enabled by @option{-Wall}.
5611 @item -Wduplicated-branches
5612 @opindex Wno-duplicated-branches
5613 @opindex Wduplicated-branches
5614 Warn when an if-else has identical branches. This warning detects cases like
5621 It doesn't warn when both branches contain just a null statement. This warning
5622 also warn for conditional operators:
5624 int i = x ? *p : *p;
5627 @item -Wduplicated-cond
5628 @opindex Wno-duplicated-cond
5629 @opindex Wduplicated-cond
5630 Warn about duplicated conditions in an if-else-if chain. For instance,
5631 warn for the following code:
5633 if (p->q != NULL) @{ @dots{} @}
5634 else if (p->q != NULL) @{ @dots{} @}
5637 @item -Wframe-address
5638 @opindex Wno-frame-address
5639 @opindex Wframe-address
5640 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5641 is called with an argument greater than 0. Such calls may return indeterminate
5642 values or crash the program. The warning is included in @option{-Wall}.
5644 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5645 @opindex Wno-discarded-qualifiers
5646 @opindex Wdiscarded-qualifiers
5647 Do not warn if type qualifiers on pointers are being discarded.
5648 Typically, the compiler warns if a @code{const char *} variable is
5649 passed to a function that takes a @code{char *} parameter. This option
5650 can be used to suppress such a warning.
5652 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5653 @opindex Wno-discarded-array-qualifiers
5654 @opindex Wdiscarded-array-qualifiers
5655 Do not warn if type qualifiers on arrays which are pointer targets
5656 are being discarded. Typically, the compiler warns if a
5657 @code{const int (*)[]} variable is passed to a function that
5658 takes a @code{int (*)[]} parameter. This option can be used to
5659 suppress such a warning.
5661 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5662 @opindex Wno-incompatible-pointer-types
5663 @opindex Wincompatible-pointer-types
5664 Do not warn when there is a conversion between pointers that have incompatible
5665 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5666 which warns for pointer argument passing or assignment with different
5669 @item -Wno-int-conversion @r{(C and Objective-C only)}
5670 @opindex Wno-int-conversion
5671 @opindex Wint-conversion
5672 Do not warn about incompatible integer to pointer and pointer to integer
5673 conversions. This warning is about implicit conversions; for explicit
5674 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5675 @option{-Wno-pointer-to-int-cast} may be used.
5677 @item -Wno-div-by-zero
5678 @opindex Wno-div-by-zero
5679 @opindex Wdiv-by-zero
5680 Do not warn about compile-time integer division by zero. Floating-point
5681 division by zero is not warned about, as it can be a legitimate way of
5682 obtaining infinities and NaNs.
5684 @item -Wsystem-headers
5685 @opindex Wsystem-headers
5686 @opindex Wno-system-headers
5687 @cindex warnings from system headers
5688 @cindex system headers, warnings from
5689 Print warning messages for constructs found in system header files.
5690 Warnings from system headers are normally suppressed, on the assumption
5691 that they usually do not indicate real problems and would only make the
5692 compiler output harder to read. Using this command-line option tells
5693 GCC to emit warnings from system headers as if they occurred in user
5694 code. However, note that using @option{-Wall} in conjunction with this
5695 option does @emph{not} warn about unknown pragmas in system
5696 headers---for that, @option{-Wunknown-pragmas} must also be used.
5698 @item -Wtautological-compare
5699 @opindex Wtautological-compare
5700 @opindex Wno-tautological-compare
5701 Warn if a self-comparison always evaluates to true or false. This
5702 warning detects various mistakes such as:
5706 if (i > i) @{ @dots{} @}
5709 This warning also warns about bitwise comparisons that always evaluate
5710 to true or false, for instance:
5712 if ((a & 16) == 10) @{ @dots{} @}
5714 will always be false.
5716 This warning is enabled by @option{-Wall}.
5719 @opindex Wtrampolines
5720 @opindex Wno-trampolines
5721 Warn about trampolines generated for pointers to nested functions.
5722 A trampoline is a small piece of data or code that is created at run
5723 time on the stack when the address of a nested function is taken, and is
5724 used to call the nested function indirectly. For some targets, it is
5725 made up of data only and thus requires no special treatment. But, for
5726 most targets, it is made up of code and thus requires the stack to be
5727 made executable in order for the program to work properly.
5730 @opindex Wfloat-equal
5731 @opindex Wno-float-equal
5732 Warn if floating-point values are used in equality comparisons.
5734 The idea behind this is that sometimes it is convenient (for the
5735 programmer) to consider floating-point values as approximations to
5736 infinitely precise real numbers. If you are doing this, then you need
5737 to compute (by analyzing the code, or in some other way) the maximum or
5738 likely maximum error that the computation introduces, and allow for it
5739 when performing comparisons (and when producing output, but that's a
5740 different problem). In particular, instead of testing for equality, you
5741 should check to see whether the two values have ranges that overlap; and
5742 this is done with the relational operators, so equality comparisons are
5745 @item -Wtraditional @r{(C and Objective-C only)}
5746 @opindex Wtraditional
5747 @opindex Wno-traditional
5748 Warn about certain constructs that behave differently in traditional and
5749 ISO C@. Also warn about ISO C constructs that have no traditional C
5750 equivalent, and/or problematic constructs that should be avoided.
5754 Macro parameters that appear within string literals in the macro body.
5755 In traditional C macro replacement takes place within string literals,
5756 but in ISO C it does not.
5759 In traditional C, some preprocessor directives did not exist.
5760 Traditional preprocessors only considered a line to be a directive
5761 if the @samp{#} appeared in column 1 on the line. Therefore
5762 @option{-Wtraditional} warns about directives that traditional C
5763 understands but ignores because the @samp{#} does not appear as the
5764 first character on the line. It also suggests you hide directives like
5765 @code{#pragma} not understood by traditional C by indenting them. Some
5766 traditional implementations do not recognize @code{#elif}, so this option
5767 suggests avoiding it altogether.
5770 A function-like macro that appears without arguments.
5773 The unary plus operator.
5776 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5777 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5778 constants.) Note, these suffixes appear in macros defined in the system
5779 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5780 Use of these macros in user code might normally lead to spurious
5781 warnings, however GCC's integrated preprocessor has enough context to
5782 avoid warning in these cases.
5785 A function declared external in one block and then used after the end of
5789 A @code{switch} statement has an operand of type @code{long}.
5792 A non-@code{static} function declaration follows a @code{static} one.
5793 This construct is not accepted by some traditional C compilers.
5796 The ISO type of an integer constant has a different width or
5797 signedness from its traditional type. This warning is only issued if
5798 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5799 typically represent bit patterns, are not warned about.
5802 Usage of ISO string concatenation is detected.
5805 Initialization of automatic aggregates.
5808 Identifier conflicts with labels. Traditional C lacks a separate
5809 namespace for labels.
5812 Initialization of unions. If the initializer is zero, the warning is
5813 omitted. This is done under the assumption that the zero initializer in
5814 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5815 initializer warnings and relies on default initialization to zero in the
5819 Conversions by prototypes between fixed/floating-point values and vice
5820 versa. The absence of these prototypes when compiling with traditional
5821 C causes serious problems. This is a subset of the possible
5822 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5825 Use of ISO C style function definitions. This warning intentionally is
5826 @emph{not} issued for prototype declarations or variadic functions
5827 because these ISO C features appear in your code when using
5828 libiberty's traditional C compatibility macros, @code{PARAMS} and
5829 @code{VPARAMS}. This warning is also bypassed for nested functions
5830 because that feature is already a GCC extension and thus not relevant to
5831 traditional C compatibility.
5834 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5835 @opindex Wtraditional-conversion
5836 @opindex Wno-traditional-conversion
5837 Warn if a prototype causes a type conversion that is different from what
5838 would happen to the same argument in the absence of a prototype. This
5839 includes conversions of fixed point to floating and vice versa, and
5840 conversions changing the width or signedness of a fixed-point argument
5841 except when the same as the default promotion.
5843 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5844 @opindex Wdeclaration-after-statement
5845 @opindex Wno-declaration-after-statement
5846 Warn when a declaration is found after a statement in a block. This
5847 construct, known from C++, was introduced with ISO C99 and is by default
5848 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5853 Warn whenever a local variable or type declaration shadows another
5854 variable, parameter, type, class member (in C++), or instance variable
5855 (in Objective-C) or whenever a built-in function is shadowed. Note
5856 that in C++, the compiler warns if a local variable shadows an
5857 explicit typedef, but not if it shadows a struct/class/enum.
5858 Same as @option{-Wshadow=global}.
5860 @item -Wno-shadow-ivar @r{(Objective-C only)}
5861 @opindex Wno-shadow-ivar
5862 @opindex Wshadow-ivar
5863 Do not warn whenever a local variable shadows an instance variable in an
5866 @item -Wshadow=global
5867 @opindex Wshadow=local
5868 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5870 @item -Wshadow=local
5871 @opindex Wshadow=local
5872 Warn when a local variable shadows another local variable or parameter.
5873 This warning is enabled by @option{-Wshadow=global}.
5875 @item -Wshadow=compatible-local
5876 @opindex Wshadow=compatible-local
5877 Warn when a local variable shadows another local variable or parameter
5878 whose type is compatible with that of the shadowing variable. In C++,
5879 type compatibility here means the type of the shadowing variable can be
5880 converted to that of the shadowed variable. The creation of this flag
5881 (in addition to @option{-Wshadow=local}) is based on the idea that when
5882 a local variable shadows another one of incompatible type, it is most
5883 likely intentional, not a bug or typo, as shown in the following example:
5887 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5889 for (int i = 0; i < N; ++i)
5898 Since the two variable @code{i} in the example above have incompatible types,
5899 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5900 Because their types are incompatible, if a programmer accidentally uses one
5901 in place of the other, type checking will catch that and emit an error or
5902 warning. So not warning (about shadowing) in this case will not lead to
5903 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5904 possibly reduce the number of warnings triggered by intentional shadowing.
5906 This warning is enabled by @option{-Wshadow=local}.
5908 @item -Wlarger-than=@var{len}
5909 @opindex Wlarger-than=@var{len}
5910 @opindex Wlarger-than-@var{len}
5911 Warn whenever an object of larger than @var{len} bytes is defined.
5913 @item -Wframe-larger-than=@var{len}
5914 @opindex Wframe-larger-than
5915 Warn if the size of a function frame is larger than @var{len} bytes.
5916 The computation done to determine the stack frame size is approximate
5917 and not conservative.
5918 The actual requirements may be somewhat greater than @var{len}
5919 even if you do not get a warning. In addition, any space allocated
5920 via @code{alloca}, variable-length arrays, or related constructs
5921 is not included by the compiler when determining
5922 whether or not to issue a warning.
5924 @item -Wno-free-nonheap-object
5925 @opindex Wno-free-nonheap-object
5926 @opindex Wfree-nonheap-object
5927 Do not warn when attempting to free an object that was not allocated
5930 @item -Wstack-usage=@var{len}
5931 @opindex Wstack-usage
5932 Warn if the stack usage of a function might be larger than @var{len} bytes.
5933 The computation done to determine the stack usage is conservative.
5934 Any space allocated via @code{alloca}, variable-length arrays, or related
5935 constructs is included by the compiler when determining whether or not to
5938 The message is in keeping with the output of @option{-fstack-usage}.
5942 If the stack usage is fully static but exceeds the specified amount, it's:
5945 warning: stack usage is 1120 bytes
5948 If the stack usage is (partly) dynamic but bounded, it's:
5951 warning: stack usage might be 1648 bytes
5954 If the stack usage is (partly) dynamic and not bounded, it's:
5957 warning: stack usage might be unbounded
5961 @item -Wunsafe-loop-optimizations
5962 @opindex Wunsafe-loop-optimizations
5963 @opindex Wno-unsafe-loop-optimizations
5964 Warn if the loop cannot be optimized because the compiler cannot
5965 assume anything on the bounds of the loop indices. With
5966 @option{-funsafe-loop-optimizations} warn if the compiler makes
5969 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5970 @opindex Wno-pedantic-ms-format
5971 @opindex Wpedantic-ms-format
5972 When used in combination with @option{-Wformat}
5973 and @option{-pedantic} without GNU extensions, this option
5974 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5975 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5976 which depend on the MS runtime.
5979 @opindex Waligned-new
5980 @opindex Wno-aligned-new
5981 Warn about a new-expression of a type that requires greater alignment
5982 than the @code{alignof(std::max_align_t)} but uses an allocation
5983 function without an explicit alignment parameter. This option is
5984 enabled by @option{-Wall}.
5986 Normally this only warns about global allocation functions, but
5987 @option{-Waligned-new=all} also warns about class member allocation
5990 @item -Wplacement-new
5991 @itemx -Wplacement-new=@var{n}
5992 @opindex Wplacement-new
5993 @opindex Wno-placement-new
5994 Warn about placement new expressions with undefined behavior, such as
5995 constructing an object in a buffer that is smaller than the type of
5996 the object. For example, the placement new expression below is diagnosed
5997 because it attempts to construct an array of 64 integers in a buffer only
6003 This warning is enabled by default.
6006 @item -Wplacement-new=1
6007 This is the default warning level of @option{-Wplacement-new}. At this
6008 level the warning is not issued for some strictly undefined constructs that
6009 GCC allows as extensions for compatibility with legacy code. For example,
6010 the following @code{new} expression is not diagnosed at this level even
6011 though it has undefined behavior according to the C++ standard because
6012 it writes past the end of the one-element array.
6014 struct S @{ int n, a[1]; @};
6015 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
6016 new (s->a)int [32]();
6019 @item -Wplacement-new=2
6020 At this level, in addition to diagnosing all the same constructs as at level
6021 1, a diagnostic is also issued for placement new expressions that construct
6022 an object in the last member of structure whose type is an array of a single
6023 element and whose size is less than the size of the object being constructed.
6024 While the previous example would be diagnosed, the following construct makes
6025 use of the flexible member array extension to avoid the warning at level 2.
6027 struct S @{ int n, a[]; @};
6028 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
6029 new (s->a)int [32]();
6034 @item -Wpointer-arith
6035 @opindex Wpointer-arith
6036 @opindex Wno-pointer-arith
6037 Warn about anything that depends on the ``size of'' a function type or
6038 of @code{void}. GNU C assigns these types a size of 1, for
6039 convenience in calculations with @code{void *} pointers and pointers
6040 to functions. In C++, warn also when an arithmetic operation involves
6041 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
6043 @item -Wpointer-compare
6044 @opindex Wpointer-compare
6045 @opindex Wno-pointer-compare
6046 Warn if a pointer is compared with a zero character constant. This usually
6047 means that the pointer was meant to be dereferenced. For example:
6050 const char *p = foo ();
6055 Note that the code above is invalid in C++11.
6057 This warning is enabled by default.
6060 @opindex Wtype-limits
6061 @opindex Wno-type-limits
6062 Warn if a comparison is always true or always false due to the limited
6063 range of the data type, but do not warn for constant expressions. For
6064 example, warn if an unsigned variable is compared against zero with
6065 @code{<} or @code{>=}. This warning is also enabled by
6068 @include cppwarnopts.texi
6070 @item -Wbad-function-cast @r{(C and Objective-C only)}
6071 @opindex Wbad-function-cast
6072 @opindex Wno-bad-function-cast
6073 Warn when a function call is cast to a non-matching type.
6074 For example, warn if a call to a function returning an integer type
6075 is cast to a pointer type.
6077 @item -Wc90-c99-compat @r{(C and Objective-C only)}
6078 @opindex Wc90-c99-compat
6079 @opindex Wno-c90-c99-compat
6080 Warn about features not present in ISO C90, but present in ISO C99.
6081 For instance, warn about use of variable length arrays, @code{long long}
6082 type, @code{bool} type, compound literals, designated initializers, and so
6083 on. This option is independent of the standards mode. Warnings are disabled
6084 in the expression that follows @code{__extension__}.
6086 @item -Wc99-c11-compat @r{(C and Objective-C only)}
6087 @opindex Wc99-c11-compat
6088 @opindex Wno-c99-c11-compat
6089 Warn about features not present in ISO C99, but present in ISO C11.
6090 For instance, warn about use of anonymous structures and unions,
6091 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
6092 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
6093 and so on. This option is independent of the standards mode. Warnings are
6094 disabled in the expression that follows @code{__extension__}.
6096 @item -Wc++-compat @r{(C and Objective-C only)}
6097 @opindex Wc++-compat
6098 Warn about ISO C constructs that are outside of the common subset of
6099 ISO C and ISO C++, e.g.@: request for implicit conversion from
6100 @code{void *} to a pointer to non-@code{void} type.
6102 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
6103 @opindex Wc++11-compat
6104 Warn about C++ constructs whose meaning differs between ISO C++ 1998
6105 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
6106 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
6107 enabled by @option{-Wall}.
6109 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
6110 @opindex Wc++14-compat
6111 Warn about C++ constructs whose meaning differs between ISO C++ 2011
6112 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
6114 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
6115 @opindex Wc++17-compat
6116 Warn about C++ constructs whose meaning differs between ISO C++ 2014
6117 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6121 @opindex Wno-cast-qual
6122 Warn whenever a pointer is cast so as to remove a type qualifier from
6123 the target type. For example, warn if a @code{const char *} is cast
6124 to an ordinary @code{char *}.
6126 Also warn when making a cast that introduces a type qualifier in an
6127 unsafe way. For example, casting @code{char **} to @code{const char **}
6128 is unsafe, as in this example:
6131 /* p is char ** value. */
6132 const char **q = (const char **) p;
6133 /* Assignment of readonly string to const char * is OK. */
6135 /* Now char** pointer points to read-only memory. */
6140 @opindex Wcast-align
6141 @opindex Wno-cast-align
6142 Warn whenever a pointer is cast such that the required alignment of the
6143 target is increased. For example, warn if a @code{char *} is cast to
6144 an @code{int *} on machines where integers can only be accessed at
6145 two- or four-byte boundaries.
6147 @item -Wcast-align=strict
6148 @opindex Wcast-align=strict
6149 Warn whenever a pointer is cast such that the required alignment of the
6150 target is increased. For example, warn if a @code{char *} is cast to
6151 an @code{int *} regardless of the target machine.
6153 @item -Wcast-function-type
6154 @opindex Wcast-function-type
6155 @opindex Wno-cast-function-type
6156 Warn when a function pointer is cast to an incompatible function pointer.
6157 In a cast involving function types with a variable argument list only
6158 the types of initial arguments that are provided are considered.
6159 Any parameter of pointer-type matches any other pointer-type. Any benign
6160 differences in integral types are ignored, like @code{int} vs. @code{long}
6161 on ILP32 targets. Likewise type qualifiers are ignored. The function
6162 type @code{void (*) (void)} is special and matches everything, which can
6163 be used to suppress this warning.
6164 In a cast involving pointer to member types this warning warns whenever
6165 the type cast is changing the pointer to member type.
6166 This warning is enabled by @option{-Wextra}.
6168 @item -Wwrite-strings
6169 @opindex Wwrite-strings
6170 @opindex Wno-write-strings
6171 When compiling C, give string constants the type @code{const
6172 char[@var{length}]} so that copying the address of one into a
6173 non-@code{const} @code{char *} pointer produces a warning. These
6174 warnings help you find at compile time code that can try to write
6175 into a string constant, but only if you have been very careful about
6176 using @code{const} in declarations and prototypes. Otherwise, it is
6177 just a nuisance. This is why we did not make @option{-Wall} request
6180 When compiling C++, warn about the deprecated conversion from string
6181 literals to @code{char *}. This warning is enabled by default for C++
6185 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6186 @opindex Wcatch-value
6187 @opindex Wno-catch-value
6188 Warn about catch handlers that do not catch via reference.
6189 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6190 warn about polymorphic class types that are caught by value.
6191 With @option{-Wcatch-value=2} warn about all class types that are caught
6192 by value. With @option{-Wcatch-value=3} warn about all types that are
6193 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6197 @opindex Wno-clobbered
6198 Warn for variables that might be changed by @code{longjmp} or
6199 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6201 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6202 @opindex Wconditionally-supported
6203 @opindex Wno-conditionally-supported
6204 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6207 @opindex Wconversion
6208 @opindex Wno-conversion
6209 Warn for implicit conversions that may alter a value. This includes
6210 conversions between real and integer, like @code{abs (x)} when
6211 @code{x} is @code{double}; conversions between signed and unsigned,
6212 like @code{unsigned ui = -1}; and conversions to smaller types, like
6213 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6214 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6215 changed by the conversion like in @code{abs (2.0)}. Warnings about
6216 conversions between signed and unsigned integers can be disabled by
6217 using @option{-Wno-sign-conversion}.
6219 For C++, also warn for confusing overload resolution for user-defined
6220 conversions; and conversions that never use a type conversion
6221 operator: conversions to @code{void}, the same type, a base class or a
6222 reference to them. Warnings about conversions between signed and
6223 unsigned integers are disabled by default in C++ unless
6224 @option{-Wsign-conversion} is explicitly enabled.
6226 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6227 @opindex Wconversion-null
6228 @opindex Wno-conversion-null
6229 Do not warn for conversions between @code{NULL} and non-pointer
6230 types. @option{-Wconversion-null} is enabled by default.
6232 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6233 @opindex Wzero-as-null-pointer-constant
6234 @opindex Wno-zero-as-null-pointer-constant
6235 Warn when a literal @samp{0} is used as null pointer constant. This can
6236 be useful to facilitate the conversion to @code{nullptr} in C++11.
6238 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6239 @opindex Wsubobject-linkage
6240 @opindex Wno-subobject-linkage
6241 Warn if a class type has a base or a field whose type uses the anonymous
6242 namespace or depends on a type with no linkage. If a type A depends on
6243 a type B with no or internal linkage, defining it in multiple
6244 translation units would be an ODR violation because the meaning of B
6245 is different in each translation unit. If A only appears in a single
6246 translation unit, the best way to silence the warning is to give it
6247 internal linkage by putting it in an anonymous namespace as well. The
6248 compiler doesn't give this warning for types defined in the main .C
6249 file, as those are unlikely to have multiple definitions.
6250 @option{-Wsubobject-linkage} is enabled by default.
6252 @item -Wdangling-else
6253 @opindex Wdangling-else
6254 @opindex Wno-dangling-else
6255 Warn about constructions where there may be confusion to which
6256 @code{if} statement an @code{else} branch belongs. Here is an example of
6271 In C/C++, every @code{else} branch belongs to the innermost possible
6272 @code{if} statement, which in this example is @code{if (b)}. This is
6273 often not what the programmer expected, as illustrated in the above
6274 example by indentation the programmer chose. When there is the
6275 potential for this confusion, GCC issues a warning when this flag
6276 is specified. To eliminate the warning, add explicit braces around
6277 the innermost @code{if} statement so there is no way the @code{else}
6278 can belong to the enclosing @code{if}. The resulting code
6295 This warning is enabled by @option{-Wparentheses}.
6299 @opindex Wno-date-time
6300 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6301 are encountered as they might prevent bit-wise-identical reproducible
6304 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6305 @opindex Wdelete-incomplete
6306 @opindex Wno-delete-incomplete
6307 Warn when deleting a pointer to incomplete type, which may cause
6308 undefined behavior at runtime. This warning is enabled by default.
6310 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6311 @opindex Wuseless-cast
6312 @opindex Wno-useless-cast
6313 Warn when an expression is casted to its own type.
6316 @opindex Wempty-body
6317 @opindex Wno-empty-body
6318 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6319 while} statement. This warning is also enabled by @option{-Wextra}.
6321 @item -Wenum-compare
6322 @opindex Wenum-compare
6323 @opindex Wno-enum-compare
6324 Warn about a comparison between values of different enumerated types.
6325 In C++ enumerated type mismatches in conditional expressions are also
6326 diagnosed and the warning is enabled by default. In C this warning is
6327 enabled by @option{-Wall}.
6329 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6330 @opindex Wextra-semi
6331 @opindex Wno-extra-semi
6332 Warn about redundant semicolon after in-class function definition.
6334 @item -Wjump-misses-init @r{(C, Objective-C only)}
6335 @opindex Wjump-misses-init
6336 @opindex Wno-jump-misses-init
6337 Warn if a @code{goto} statement or a @code{switch} statement jumps
6338 forward across the initialization of a variable, or jumps backward to a
6339 label after the variable has been initialized. This only warns about
6340 variables that are initialized when they are declared. This warning is
6341 only supported for C and Objective-C; in C++ this sort of branch is an
6344 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6345 can be disabled with the @option{-Wno-jump-misses-init} option.
6347 @item -Wsign-compare
6348 @opindex Wsign-compare
6349 @opindex Wno-sign-compare
6350 @cindex warning for comparison of signed and unsigned values
6351 @cindex comparison of signed and unsigned values, warning
6352 @cindex signed and unsigned values, comparison warning
6353 Warn when a comparison between signed and unsigned values could produce
6354 an incorrect result when the signed value is converted to unsigned.
6355 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6356 also enabled by @option{-Wextra}.
6358 @item -Wsign-conversion
6359 @opindex Wsign-conversion
6360 @opindex Wno-sign-conversion
6361 Warn for implicit conversions that may change the sign of an integer
6362 value, like assigning a signed integer expression to an unsigned
6363 integer variable. An explicit cast silences the warning. In C, this
6364 option is enabled also by @option{-Wconversion}.
6366 @item -Wfloat-conversion
6367 @opindex Wfloat-conversion
6368 @opindex Wno-float-conversion
6369 Warn for implicit conversions that reduce the precision of a real value.
6370 This includes conversions from real to integer, and from higher precision
6371 real to lower precision real values. This option is also enabled by
6372 @option{-Wconversion}.
6374 @item -Wno-scalar-storage-order
6375 @opindex -Wno-scalar-storage-order
6376 @opindex -Wscalar-storage-order
6377 Do not warn on suspicious constructs involving reverse scalar storage order.
6379 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6380 @opindex Wsized-deallocation
6381 @opindex Wno-sized-deallocation
6382 Warn about a definition of an unsized deallocation function
6384 void operator delete (void *) noexcept;
6385 void operator delete[] (void *) noexcept;
6387 without a definition of the corresponding sized deallocation function
6389 void operator delete (void *, std::size_t) noexcept;
6390 void operator delete[] (void *, std::size_t) noexcept;
6392 or vice versa. Enabled by @option{-Wextra} along with
6393 @option{-fsized-deallocation}.
6395 @item -Wsizeof-pointer-div
6396 @opindex Wsizeof-pointer-div
6397 @opindex Wno-sizeof-pointer-div
6398 Warn for suspicious divisions of two sizeof expressions that divide
6399 the pointer size by the element size, which is the usual way to compute
6400 the array size but won't work out correctly with pointers. This warning
6401 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
6402 not an array, but a pointer. This warning is enabled by @option{-Wall}.
6404 @item -Wsizeof-pointer-memaccess
6405 @opindex Wsizeof-pointer-memaccess
6406 @opindex Wno-sizeof-pointer-memaccess
6407 Warn for suspicious length parameters to certain string and memory built-in
6408 functions if the argument uses @code{sizeof}. This warning triggers for
6409 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
6410 an array, but a pointer, and suggests a possible fix, or about
6411 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
6412 also warns about calls to bounded string copy functions like @code{strncat}
6413 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
6414 the source array. For example, in the following function the call to
6415 @code{strncat} specifies the size of the source string as the bound. That
6416 is almost certainly a mistake and so the call is diagnosed.
6418 void make_file (const char *name)
6420 char path[PATH_MAX];
6421 strncpy (path, name, sizeof path - 1);
6422 strncat (path, ".text", sizeof ".text");
6427 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
6429 @item -Wsizeof-array-argument
6430 @opindex Wsizeof-array-argument
6431 @opindex Wno-sizeof-array-argument
6432 Warn when the @code{sizeof} operator is applied to a parameter that is
6433 declared as an array in a function definition. This warning is enabled by
6434 default for C and C++ programs.
6436 @item -Wmemset-elt-size
6437 @opindex Wmemset-elt-size
6438 @opindex Wno-memset-elt-size
6439 Warn for suspicious calls to the @code{memset} built-in function, if the
6440 first argument references an array, and the third argument is a number
6441 equal to the number of elements, but not equal to the size of the array
6442 in memory. This indicates that the user has omitted a multiplication by
6443 the element size. This warning is enabled by @option{-Wall}.
6445 @item -Wmemset-transposed-args
6446 @opindex Wmemset-transposed-args
6447 @opindex Wno-memset-transposed-args
6448 Warn for suspicious calls to the @code{memset} built-in function, if the
6449 second argument is not zero and the third argument is zero. This warns e.g.@
6450 about @code{memset (buf, sizeof buf, 0)} where most probably
6451 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6452 is only emitted if the third argument is literal zero. If it is some
6453 expression that is folded to zero, a cast of zero to some type, etc.,
6454 it is far less likely that the user has mistakenly exchanged the arguments
6455 and no warning is emitted. This warning is enabled by @option{-Wall}.
6459 @opindex Wno-address
6460 Warn about suspicious uses of memory addresses. These include using
6461 the address of a function in a conditional expression, such as
6462 @code{void func(void); if (func)}, and comparisons against the memory
6463 address of a string literal, such as @code{if (x == "abc")}. Such
6464 uses typically indicate a programmer error: the address of a function
6465 always evaluates to true, so their use in a conditional usually
6466 indicate that the programmer forgot the parentheses in a function
6467 call; and comparisons against string literals result in unspecified
6468 behavior and are not portable in C, so they usually indicate that the
6469 programmer intended to use @code{strcmp}. This warning is enabled by
6473 @opindex Wlogical-op
6474 @opindex Wno-logical-op
6475 Warn about suspicious uses of logical operators in expressions.
6476 This includes using logical operators in contexts where a
6477 bit-wise operator is likely to be expected. Also warns when
6478 the operands of a logical operator are the same:
6481 if (a < 0 && a < 0) @{ @dots{} @}
6484 @item -Wlogical-not-parentheses
6485 @opindex Wlogical-not-parentheses
6486 @opindex Wno-logical-not-parentheses
6487 Warn about logical not used on the left hand side operand of a comparison.
6488 This option does not warn if the right operand is considered to be a boolean
6489 expression. Its purpose is to detect suspicious code like the following:
6493 if (!a > 1) @{ @dots{} @}
6496 It is possible to suppress the warning by wrapping the LHS into
6499 if ((!a) > 1) @{ @dots{} @}
6502 This warning is enabled by @option{-Wall}.
6504 @item -Waggregate-return
6505 @opindex Waggregate-return
6506 @opindex Wno-aggregate-return
6507 Warn if any functions that return structures or unions are defined or
6508 called. (In languages where you can return an array, this also elicits
6511 @item -Wno-aggressive-loop-optimizations
6512 @opindex Wno-aggressive-loop-optimizations
6513 @opindex Waggressive-loop-optimizations
6514 Warn if in a loop with constant number of iterations the compiler detects
6515 undefined behavior in some statement during one or more of the iterations.
6517 @item -Wno-attributes
6518 @opindex Wno-attributes
6519 @opindex Wattributes
6520 Do not warn if an unexpected @code{__attribute__} is used, such as
6521 unrecognized attributes, function attributes applied to variables,
6522 etc. This does not stop errors for incorrect use of supported
6525 @item -Wno-builtin-declaration-mismatch
6526 @opindex Wno-builtin-declaration-mismatch
6527 @opindex Wbuiltin-declaration-mismatch
6528 Warn if a built-in function is declared with the wrong signature or
6530 This warning is enabled by default.
6532 @item -Wno-builtin-macro-redefined
6533 @opindex Wno-builtin-macro-redefined
6534 @opindex Wbuiltin-macro-redefined
6535 Do not warn if certain built-in macros are redefined. This suppresses
6536 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6537 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6539 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6540 @opindex Wstrict-prototypes
6541 @opindex Wno-strict-prototypes
6542 Warn if a function is declared or defined without specifying the
6543 argument types. (An old-style function definition is permitted without
6544 a warning if preceded by a declaration that specifies the argument
6547 @item -Wold-style-declaration @r{(C and Objective-C only)}
6548 @opindex Wold-style-declaration
6549 @opindex Wno-old-style-declaration
6550 Warn for obsolescent usages, according to the C Standard, in a
6551 declaration. For example, warn if storage-class specifiers like
6552 @code{static} are not the first things in a declaration. This warning
6553 is also enabled by @option{-Wextra}.
6555 @item -Wold-style-definition @r{(C and Objective-C only)}
6556 @opindex Wold-style-definition
6557 @opindex Wno-old-style-definition
6558 Warn if an old-style function definition is used. A warning is given
6559 even if there is a previous prototype.
6561 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6562 @opindex Wmissing-parameter-type
6563 @opindex Wno-missing-parameter-type
6564 A function parameter is declared without a type specifier in K&R-style
6571 This warning is also enabled by @option{-Wextra}.
6573 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6574 @opindex Wmissing-prototypes
6575 @opindex Wno-missing-prototypes
6576 Warn if a global function is defined without a previous prototype
6577 declaration. This warning is issued even if the definition itself
6578 provides a prototype. Use this option to detect global functions
6579 that do not have a matching prototype declaration in a header file.
6580 This option is not valid for C++ because all function declarations
6581 provide prototypes and a non-matching declaration declares an
6582 overload rather than conflict with an earlier declaration.
6583 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6585 @item -Wmissing-declarations
6586 @opindex Wmissing-declarations
6587 @opindex Wno-missing-declarations
6588 Warn if a global function is defined without a previous declaration.
6589 Do so even if the definition itself provides a prototype.
6590 Use this option to detect global functions that are not declared in
6591 header files. In C, no warnings are issued for functions with previous
6592 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6593 missing prototypes. In C++, no warnings are issued for function templates,
6594 or for inline functions, or for functions in anonymous namespaces.
6596 @item -Wmissing-field-initializers
6597 @opindex Wmissing-field-initializers
6598 @opindex Wno-missing-field-initializers
6602 Warn if a structure's initializer has some fields missing. For
6603 example, the following code causes such a warning, because
6604 @code{x.h} is implicitly zero:
6607 struct s @{ int f, g, h; @};
6608 struct s x = @{ 3, 4 @};
6611 This option does not warn about designated initializers, so the following
6612 modification does not trigger a warning:
6615 struct s @{ int f, g, h; @};
6616 struct s x = @{ .f = 3, .g = 4 @};
6619 In C this option does not warn about the universal zero initializer
6623 struct s @{ int f, g, h; @};
6624 struct s x = @{ 0 @};
6627 Likewise, in C++ this option does not warn about the empty @{ @}
6628 initializer, for example:
6631 struct s @{ int f, g, h; @};
6635 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6636 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6638 @item -Wno-multichar
6639 @opindex Wno-multichar
6641 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6642 Usually they indicate a typo in the user's code, as they have
6643 implementation-defined values, and should not be used in portable code.
6645 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
6646 @opindex Wnormalized=
6647 @opindex Wnormalized
6648 @opindex Wno-normalized
6651 @cindex character set, input normalization
6652 In ISO C and ISO C++, two identifiers are different if they are
6653 different sequences of characters. However, sometimes when characters
6654 outside the basic ASCII character set are used, you can have two
6655 different character sequences that look the same. To avoid confusion,
6656 the ISO 10646 standard sets out some @dfn{normalization rules} which
6657 when applied ensure that two sequences that look the same are turned into
6658 the same sequence. GCC can warn you if you are using identifiers that
6659 have not been normalized; this option controls that warning.
6661 There are four levels of warning supported by GCC@. The default is
6662 @option{-Wnormalized=nfc}, which warns about any identifier that is
6663 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6664 recommended form for most uses. It is equivalent to
6665 @option{-Wnormalized}.
6667 Unfortunately, there are some characters allowed in identifiers by
6668 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6669 identifiers. That is, there's no way to use these symbols in portable
6670 ISO C or C++ and have all your identifiers in NFC@.
6671 @option{-Wnormalized=id} suppresses the warning for these characters.
6672 It is hoped that future versions of the standards involved will correct
6673 this, which is why this option is not the default.
6675 You can switch the warning off for all characters by writing
6676 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6677 only do this if you are using some other normalization scheme (like
6678 ``D''), because otherwise you can easily create bugs that are
6679 literally impossible to see.
6681 Some characters in ISO 10646 have distinct meanings but look identical
6682 in some fonts or display methodologies, especially once formatting has
6683 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6684 LETTER N'', displays just like a regular @code{n} that has been
6685 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6686 normalization scheme to convert all these into a standard form as
6687 well, and GCC warns if your code is not in NFKC if you use
6688 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6689 about every identifier that contains the letter O because it might be
6690 confused with the digit 0, and so is not the default, but may be
6691 useful as a local coding convention if the programming environment
6692 cannot be fixed to display these characters distinctly.
6694 @item -Wno-deprecated
6695 @opindex Wno-deprecated
6696 @opindex Wdeprecated
6697 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6699 @item -Wno-deprecated-declarations
6700 @opindex Wno-deprecated-declarations
6701 @opindex Wdeprecated-declarations
6702 Do not warn about uses of functions (@pxref{Function Attributes}),
6703 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6704 Attributes}) marked as deprecated by using the @code{deprecated}
6708 @opindex Wno-overflow
6710 Do not warn about compile-time overflow in constant expressions.
6715 Warn about One Definition Rule violations during link-time optimization.
6716 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6719 @opindex Wopenm-simd
6720 Warn if the vectorizer cost model overrides the OpenMP
6721 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6722 option can be used to relax the cost model.
6724 @item -Woverride-init @r{(C and Objective-C only)}
6725 @opindex Woverride-init
6726 @opindex Wno-override-init
6730 Warn if an initialized field without side effects is overridden when
6731 using designated initializers (@pxref{Designated Inits, , Designated
6734 This warning is included in @option{-Wextra}. To get other
6735 @option{-Wextra} warnings without this one, use @option{-Wextra
6736 -Wno-override-init}.
6738 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6739 @opindex Woverride-init-side-effects
6740 @opindex Wno-override-init-side-effects
6741 Warn if an initialized field with side effects is overridden when
6742 using designated initializers (@pxref{Designated Inits, , Designated
6743 Initializers}). This warning is enabled by default.
6748 Warn if a structure is given the packed attribute, but the packed
6749 attribute has no effect on the layout or size of the structure.
6750 Such structures may be mis-aligned for little benefit. For
6751 instance, in this code, the variable @code{f.x} in @code{struct bar}
6752 is misaligned even though @code{struct bar} does not itself
6753 have the packed attribute:
6760 @} __attribute__((packed));
6768 @item -Wpacked-bitfield-compat
6769 @opindex Wpacked-bitfield-compat
6770 @opindex Wno-packed-bitfield-compat
6771 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6772 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6773 the change can lead to differences in the structure layout. GCC
6774 informs you when the offset of such a field has changed in GCC 4.4.
6775 For example there is no longer a 4-bit padding between field @code{a}
6776 and @code{b} in this structure:
6783 @} __attribute__ ((packed));
6786 This warning is enabled by default. Use
6787 @option{-Wno-packed-bitfield-compat} to disable this warning.
6789 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
6790 @opindex Wpacked-not-aligned
6791 @opindex Wno-packed-not-aligned
6792 Warn if a structure field with explicitly specified alignment in a
6793 packed struct or union is misaligned. For example, a warning will
6794 be issued on @code{struct S}, like, @code{warning: alignment 1 of
6795 'struct S' is less than 8}, in this code:
6799 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
6800 struct __attribute__ ((packed)) S @{
6806 This warning is enabled by @option{-Wall}.
6811 Warn if padding is included in a structure, either to align an element
6812 of the structure or to align the whole structure. Sometimes when this
6813 happens it is possible to rearrange the fields of the structure to
6814 reduce the padding and so make the structure smaller.
6816 @item -Wredundant-decls
6817 @opindex Wredundant-decls
6818 @opindex Wno-redundant-decls
6819 Warn if anything is declared more than once in the same scope, even in
6820 cases where multiple declaration is valid and changes nothing.
6824 @opindex Wno-restrict
6825 Warn when an object referenced by a @code{restrict}-qualified parameter
6826 (or, in C++, a @code{__restrict}-qualified parameter) is aliased by another
6827 argument, or when copies between such objects overlap. For example,
6828 the call to the @code{strcpy} function below attempts to truncate the string
6829 by replacing its initial characters with the last four. However, because
6830 the call writes the terminating NUL into @code{a[4]}, the copies overlap and
6831 the call is diagnosed.
6836 char a[] = "abcd1234";
6841 The @option{-Wrestrict} option detects some instances of simple overlap
6842 even without optimization but works best at @option{-O2} and above. It
6843 is included in @option{-Wall}.
6845 @item -Wnested-externs @r{(C and Objective-C only)}
6846 @opindex Wnested-externs
6847 @opindex Wno-nested-externs
6848 Warn if an @code{extern} declaration is encountered within a function.
6850 @item -Wno-inherited-variadic-ctor
6851 @opindex Winherited-variadic-ctor
6852 @opindex Wno-inherited-variadic-ctor
6853 Suppress warnings about use of C++11 inheriting constructors when the
6854 base class inherited from has a C variadic constructor; the warning is
6855 on by default because the ellipsis is not inherited.
6860 Warn if a function that is declared as inline cannot be inlined.
6861 Even with this option, the compiler does not warn about failures to
6862 inline functions declared in system headers.
6864 The compiler uses a variety of heuristics to determine whether or not
6865 to inline a function. For example, the compiler takes into account
6866 the size of the function being inlined and the amount of inlining
6867 that has already been done in the current function. Therefore,
6868 seemingly insignificant changes in the source program can cause the
6869 warnings produced by @option{-Winline} to appear or disappear.
6871 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6872 @opindex Wno-invalid-offsetof
6873 @opindex Winvalid-offsetof
6874 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6875 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6876 to a non-standard-layout type is undefined. In existing C++ implementations,
6877 however, @code{offsetof} typically gives meaningful results.
6878 This flag is for users who are aware that they are
6879 writing nonportable code and who have deliberately chosen to ignore the
6882 The restrictions on @code{offsetof} may be relaxed in a future version
6883 of the C++ standard.
6885 @item -Wint-in-bool-context
6886 @opindex Wint-in-bool-context
6887 @opindex Wno-int-in-bool-context
6888 Warn for suspicious use of integer values where boolean values are expected,
6889 such as conditional expressions (?:) using non-boolean integer constants in
6890 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6891 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6892 for all kinds of multiplications regardless of the data type.
6893 This warning is enabled by @option{-Wall}.
6895 @item -Wno-int-to-pointer-cast
6896 @opindex Wno-int-to-pointer-cast
6897 @opindex Wint-to-pointer-cast
6898 Suppress warnings from casts to pointer type of an integer of a
6899 different size. In C++, casting to a pointer type of smaller size is
6900 an error. @option{Wint-to-pointer-cast} is enabled by default.
6903 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6904 @opindex Wno-pointer-to-int-cast
6905 @opindex Wpointer-to-int-cast
6906 Suppress warnings from casts from a pointer to an integer type of a
6910 @opindex Winvalid-pch
6911 @opindex Wno-invalid-pch
6912 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6913 the search path but cannot be used.
6917 @opindex Wno-long-long
6918 Warn if @code{long long} type is used. This is enabled by either
6919 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6920 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6922 @item -Wvariadic-macros
6923 @opindex Wvariadic-macros
6924 @opindex Wno-variadic-macros
6925 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6926 alternate syntax is used in ISO C99 mode. This is enabled by either
6927 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6928 messages, use @option{-Wno-variadic-macros}.
6932 @opindex Wno-varargs
6933 Warn upon questionable usage of the macros used to handle variable
6934 arguments like @code{va_start}. This is default. To inhibit the
6935 warning messages, use @option{-Wno-varargs}.
6937 @item -Wvector-operation-performance
6938 @opindex Wvector-operation-performance
6939 @opindex Wno-vector-operation-performance
6940 Warn if vector operation is not implemented via SIMD capabilities of the
6941 architecture. Mainly useful for the performance tuning.
6942 Vector operation can be implemented @code{piecewise}, which means that the
6943 scalar operation is performed on every vector element;
6944 @code{in parallel}, which means that the vector operation is implemented
6945 using scalars of wider type, which normally is more performance efficient;
6946 and @code{as a single scalar}, which means that vector fits into a
6949 @item -Wno-virtual-move-assign
6950 @opindex Wvirtual-move-assign
6951 @opindex Wno-virtual-move-assign
6952 Suppress warnings about inheriting from a virtual base with a
6953 non-trivial C++11 move assignment operator. This is dangerous because
6954 if the virtual base is reachable along more than one path, it is
6955 moved multiple times, which can mean both objects end up in the
6956 moved-from state. If the move assignment operator is written to avoid
6957 moving from a moved-from object, this warning can be disabled.
6962 Warn if a variable-length array is used in the code.
6963 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6964 the variable-length array.
6966 @item -Wvla-larger-than=@var{n}
6967 If this option is used, the compiler will warn on uses of
6968 variable-length arrays where the size is either unbounded, or bounded
6969 by an argument that can be larger than @var{n} bytes. This is similar
6970 to how @option{-Walloca-larger-than=@var{n}} works, but with
6971 variable-length arrays.
6973 Note that GCC may optimize small variable-length arrays of a known
6974 value into plain arrays, so this warning may not get triggered for
6977 This warning is not enabled by @option{-Wall}, and is only active when
6978 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6980 See also @option{-Walloca-larger-than=@var{n}}.
6982 @item -Wvolatile-register-var
6983 @opindex Wvolatile-register-var
6984 @opindex Wno-volatile-register-var
6985 Warn if a register variable is declared volatile. The volatile
6986 modifier does not inhibit all optimizations that may eliminate reads
6987 and/or writes to register variables. This warning is enabled by
6990 @item -Wdisabled-optimization
6991 @opindex Wdisabled-optimization
6992 @opindex Wno-disabled-optimization
6993 Warn if a requested optimization pass is disabled. This warning does
6994 not generally indicate that there is anything wrong with your code; it
6995 merely indicates that GCC's optimizers are unable to handle the code
6996 effectively. Often, the problem is that your code is too big or too
6997 complex; GCC refuses to optimize programs when the optimization
6998 itself is likely to take inordinate amounts of time.
7000 @item -Wpointer-sign @r{(C and Objective-C only)}
7001 @opindex Wpointer-sign
7002 @opindex Wno-pointer-sign
7003 Warn for pointer argument passing or assignment with different signedness.
7004 This option is only supported for C and Objective-C@. It is implied by
7005 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
7006 @option{-Wno-pointer-sign}.
7008 @item -Wstack-protector
7009 @opindex Wstack-protector
7010 @opindex Wno-stack-protector
7011 This option is only active when @option{-fstack-protector} is active. It
7012 warns about functions that are not protected against stack smashing.
7014 @item -Woverlength-strings
7015 @opindex Woverlength-strings
7016 @opindex Wno-overlength-strings
7017 Warn about string constants that are longer than the ``minimum
7018 maximum'' length specified in the C standard. Modern compilers
7019 generally allow string constants that are much longer than the
7020 standard's minimum limit, but very portable programs should avoid
7021 using longer strings.
7023 The limit applies @emph{after} string constant concatenation, and does
7024 not count the trailing NUL@. In C90, the limit was 509 characters; in
7025 C99, it was raised to 4095. C++98 does not specify a normative
7026 minimum maximum, so we do not diagnose overlength strings in C++@.
7028 This option is implied by @option{-Wpedantic}, and can be disabled with
7029 @option{-Wno-overlength-strings}.
7031 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
7032 @opindex Wunsuffixed-float-constants
7034 Issue a warning for any floating constant that does not have
7035 a suffix. When used together with @option{-Wsystem-headers} it
7036 warns about such constants in system header files. This can be useful
7037 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
7038 from the decimal floating-point extension to C99.
7040 @item -Wno-designated-init @r{(C and Objective-C only)}
7041 Suppress warnings when a positional initializer is used to initialize
7042 a structure that has been marked with the @code{designated_init}
7046 Issue a warning when HSAIL cannot be emitted for the compiled function or
7051 @node Debugging Options
7052 @section Options for Debugging Your Program
7053 @cindex options, debugging
7054 @cindex debugging information options
7056 To tell GCC to emit extra information for use by a debugger, in almost
7057 all cases you need only to add @option{-g} to your other options.
7059 GCC allows you to use @option{-g} with
7060 @option{-O}. The shortcuts taken by optimized code may occasionally
7061 be surprising: some variables you declared may not exist
7062 at all; flow of control may briefly move where you did not expect it;
7063 some statements may not be executed because they compute constant
7064 results or their values are already at hand; some statements may
7065 execute in different places because they have been moved out of loops.
7066 Nevertheless it is possible to debug optimized output. This makes
7067 it reasonable to use the optimizer for programs that might have bugs.
7069 If you are not using some other optimization option, consider
7070 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
7071 With no @option{-O} option at all, some compiler passes that collect
7072 information useful for debugging do not run at all, so that
7073 @option{-Og} may result in a better debugging experience.
7078 Produce debugging information in the operating system's native format
7079 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
7082 On most systems that use stabs format, @option{-g} enables use of extra
7083 debugging information that only GDB can use; this extra information
7084 makes debugging work better in GDB but probably makes other debuggers
7086 refuse to read the program. If you want to control for certain whether
7087 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
7088 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
7092 Produce debugging information for use by GDB@. This means to use the
7093 most expressive format available (DWARF, stabs, or the native format
7094 if neither of those are supported), including GDB extensions if at all
7098 @itemx -gdwarf-@var{version}
7100 Produce debugging information in DWARF format (if that is supported).
7101 The value of @var{version} may be either 2, 3, 4 or 5; the default version
7102 for most targets is 4. DWARF Version 5 is only experimental.
7104 Note that with DWARF Version 2, some ports require and always
7105 use some non-conflicting DWARF 3 extensions in the unwind tables.
7107 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
7108 for maximum benefit.
7110 GCC no longer supports DWARF Version 1, which is substantially
7111 different than Version 2 and later. For historical reasons, some
7112 other DWARF-related options such as
7113 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
7114 in their names, but apply to all currently-supported versions of DWARF.
7118 Produce debugging information in stabs format (if that is supported),
7119 without GDB extensions. This is the format used by DBX on most BSD
7120 systems. On MIPS, Alpha and System V Release 4 systems this option
7121 produces stabs debugging output that is not understood by DBX@.
7122 On System V Release 4 systems this option requires the GNU assembler.
7126 Produce debugging information in stabs format (if that is supported),
7127 using GNU extensions understood only by the GNU debugger (GDB)@. The
7128 use of these extensions is likely to make other debuggers crash or
7129 refuse to read the program.
7133 Produce debugging information in XCOFF format (if that is supported).
7134 This is the format used by the DBX debugger on IBM RS/6000 systems.
7138 Produce debugging information in XCOFF format (if that is supported),
7139 using GNU extensions understood only by the GNU debugger (GDB)@. The
7140 use of these extensions is likely to make other debuggers crash or
7141 refuse to read the program, and may cause assemblers other than the GNU
7142 assembler (GAS) to fail with an error.
7146 Produce debugging information in Alpha/VMS debug format (if that is
7147 supported). This is the format used by DEBUG on Alpha/VMS systems.
7150 @itemx -ggdb@var{level}
7151 @itemx -gstabs@var{level}
7152 @itemx -gxcoff@var{level}
7153 @itemx -gvms@var{level}
7154 Request debugging information and also use @var{level} to specify how
7155 much information. The default level is 2.
7157 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7160 Level 1 produces minimal information, enough for making backtraces in
7161 parts of the program that you don't plan to debug. This includes
7162 descriptions of functions and external variables, and line number
7163 tables, but no information about local variables.
7165 Level 3 includes extra information, such as all the macro definitions
7166 present in the program. Some debuggers support macro expansion when
7167 you use @option{-g3}.
7169 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7170 confusion with @option{-gdwarf-@var{level}}.
7171 Instead use an additional @option{-g@var{level}} option to change the
7172 debug level for DWARF.
7174 @item -feliminate-unused-debug-symbols
7175 @opindex feliminate-unused-debug-symbols
7176 Produce debugging information in stabs format (if that is supported),
7177 for only symbols that are actually used.
7179 @item -femit-class-debug-always
7180 @opindex femit-class-debug-always
7181 Instead of emitting debugging information for a C++ class in only one
7182 object file, emit it in all object files using the class. This option
7183 should be used only with debuggers that are unable to handle the way GCC
7184 normally emits debugging information for classes because using this
7185 option increases the size of debugging information by as much as a
7188 @item -fno-merge-debug-strings
7189 @opindex fmerge-debug-strings
7190 @opindex fno-merge-debug-strings
7191 Direct the linker to not merge together strings in the debugging
7192 information that are identical in different object files. Merging is
7193 not supported by all assemblers or linkers. Merging decreases the size
7194 of the debug information in the output file at the cost of increasing
7195 link processing time. Merging is enabled by default.
7197 @item -fdebug-prefix-map=@var{old}=@var{new}
7198 @opindex fdebug-prefix-map
7199 When compiling files residing in directory @file{@var{old}}, record
7200 debugging information describing them as if the files resided in
7201 directory @file{@var{new}} instead. This can be used to replace a
7202 build-time path with an install-time path in the debug info. It can
7203 also be used to change an absolute path to a relative path by using
7204 @file{.} for @var{new}. This can give more reproducible builds, which
7205 are location independent, but may require an extra command to tell GDB
7206 where to find the source files. See also @option{-ffile-prefix-map}.
7208 @item -fvar-tracking
7209 @opindex fvar-tracking
7210 Run variable tracking pass. It computes where variables are stored at each
7211 position in code. Better debugging information is then generated
7212 (if the debugging information format supports this information).
7214 It is enabled by default when compiling with optimization (@option{-Os},
7215 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7216 the debug info format supports it.
7218 @item -fvar-tracking-assignments
7219 @opindex fvar-tracking-assignments
7220 @opindex fno-var-tracking-assignments
7221 Annotate assignments to user variables early in the compilation and
7222 attempt to carry the annotations over throughout the compilation all the
7223 way to the end, in an attempt to improve debug information while
7224 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7226 It can be enabled even if var-tracking is disabled, in which case
7227 annotations are created and maintained, but discarded at the end.
7228 By default, this flag is enabled together with @option{-fvar-tracking},
7229 except when selective scheduling is enabled.
7232 @opindex gsplit-dwarf
7233 Separate as much DWARF debugging information as possible into a
7234 separate output file with the extension @file{.dwo}. This option allows
7235 the build system to avoid linking files with debug information. To
7236 be useful, this option requires a debugger capable of reading @file{.dwo}
7241 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7243 @item -ggnu-pubnames
7244 @opindex ggnu-pubnames
7245 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7246 suitable for conversion into a GDB@ index. This option is only useful
7247 with a linker that can produce GDB@ index version 7.
7249 @item -fdebug-types-section
7250 @opindex fdebug-types-section
7251 @opindex fno-debug-types-section
7252 When using DWARF Version 4 or higher, type DIEs can be put into
7253 their own @code{.debug_types} section instead of making them part of the
7254 @code{.debug_info} section. It is more efficient to put them in a separate
7255 comdat sections since the linker can then remove duplicates.
7256 But not all DWARF consumers support @code{.debug_types} sections yet
7257 and on some objects @code{.debug_types} produces larger instead of smaller
7258 debugging information.
7260 @item -grecord-gcc-switches
7261 @itemx -gno-record-gcc-switches
7262 @opindex grecord-gcc-switches
7263 @opindex gno-record-gcc-switches
7264 This switch causes the command-line options used to invoke the
7265 compiler that may affect code generation to be appended to the
7266 DW_AT_producer attribute in DWARF debugging information. The options
7267 are concatenated with spaces separating them from each other and from
7268 the compiler version.
7269 It is enabled by default.
7270 See also @option{-frecord-gcc-switches} for another
7271 way of storing compiler options into the object file.
7273 @item -gstrict-dwarf
7274 @opindex gstrict-dwarf
7275 Disallow using extensions of later DWARF standard version than selected
7276 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7277 DWARF extensions from later standard versions is allowed.
7279 @item -gno-strict-dwarf
7280 @opindex gno-strict-dwarf
7281 Allow using extensions of later DWARF standard version than selected with
7282 @option{-gdwarf-@var{version}}.
7284 @item -gas-loc-support
7285 @opindex gas-loc-support
7286 Inform the compiler that the assembler supports @code{.loc} directives.
7287 It may then use them for the assembler to generate DWARF2+ line number
7290 This is generally desirable, because assembler-generated line-number
7291 tables are a lot more compact than those the compiler can generate
7294 This option will be enabled by default if, at GCC configure time, the
7295 assembler was found to support such directives.
7297 @item -gno-as-loc-support
7298 @opindex gno-as-loc-support
7299 Force GCC to generate DWARF2+ line number tables internally, if DWARF2+
7300 line number tables are to be generated.
7302 @item gas-locview-support
7303 @opindex gas-locview-support
7304 Inform the compiler that the assembler supports @code{view} assignment
7305 and reset assertion checking in @code{.loc} directives.
7307 This option will be enabled by default if, at GCC configure time, the
7308 assembler was found to support them.
7310 @item gno-as-locview-support
7311 Force GCC to assign view numbers internally, if
7312 @option{-gvariable-location-views} are explicitly requested.
7315 @itemx -gno-column-info
7316 @opindex gcolumn-info
7317 @opindex gno-column-info
7318 Emit location column information into DWARF debugging information, rather
7319 than just file and line.
7320 This option is enabled by default.
7322 @item -gstatement-frontiers
7323 @itemx -gno-statement-frontiers
7324 @opindex gstatement-frontiers
7325 @opindex gno-statement-frontiers
7326 This option causes GCC to create markers in the internal representation
7327 at the beginning of statements, and to keep them roughly in place
7328 throughout compilation, using them to guide the output of @code{is_stmt}
7329 markers in the line number table. This is enabled by default when
7330 compiling with optimization (@option{-Os}, @option{-O}, @option{-O2},
7331 @dots{}), and outputting DWARF 2 debug information at the normal level.
7333 @item -gvariable-location-views
7334 @itemx -gvariable-location-views=incompat5
7335 @itemx -gno-variable-location-views
7336 @opindex gvariable-location-views
7337 @opindex gvariable-location-views=incompat5
7338 @opindex gno-variable-location-views
7339 Augment variable location lists with progressive view numbers implied
7340 from the line number table. This enables debug information consumers to
7341 inspect state at certain points of the program, even if no instructions
7342 associated with the corresponding source locations are present at that
7343 point. If the assembler lacks support for view numbers in line number
7344 tables, this will cause the compiler to emit the line number table,
7345 which generally makes them somewhat less compact. The augmented line
7346 number tables and location lists are fully backward-compatible, so they
7347 can be consumed by debug information consumers that are not aware of
7348 these augmentations, but they won't derive any benefit from them either.
7350 This is enabled by default when outputting DWARF 2 debug information at
7351 the normal level, as long as there is assembler support,
7352 @option{-fvar-tracking-assignments} is enabled and
7353 @option{-gstrict-dwarf} is not. When assembler support is not
7354 available, this may still be enabled, but it will force GCC to output
7355 internal line number tables, and if
7356 @option{-ginternal-reset-location-views} is not enabled, that will most
7357 certainly lead to silently mismatching location views.
7359 There is a proposed representation for view numbers that is not backward
7360 compatible with the location list format introduced in DWARF 5, that can
7361 be enabled with @option{-gvariable-location-views=incompat5}. This
7362 option may be removed in the future, is only provided as a reference
7363 implementation of the proposed representation. Debug information
7364 consumers are not expected to support this extended format, and they
7365 would be rendered unable to decode location lists using it.
7367 @item -ginternal-reset-location-views
7368 @itemx -gnointernal-reset-location-views
7369 @opindex ginternal-reset-location-views
7370 @opindex gno-internal-reset-location-views
7371 Attempt to determine location views that can be omitted from location
7372 view lists. This requires the compiler to have very accurate insn
7373 length estimates, which isn't always the case, and it may cause
7374 incorrect view lists to be generated silently when using an assembler
7375 that does not support location view lists. The GNU assembler will flag
7376 any such error as a @code{view number mismatch}. This is only enabled
7377 on ports that define a reliable estimation function.
7379 @item -ginline-points
7380 @itemx -gno-inline-points
7381 @opindex ginline-points
7382 @opindex gno-inline-points
7383 Generate extended debug information for inlined functions. Location
7384 view tracking markers are inserted at inlined entry points, so that
7385 address and view numbers can be computed and output in debug
7386 information. This can be enabled independently of location views, in
7387 which case the view numbers won't be output, but it can only be enabled
7388 along with statement frontiers, and it is only enabled by default if
7389 location views are enabled.
7391 @item -gz@r{[}=@var{type}@r{]}
7393 Produce compressed debug sections in DWARF format, if that is supported.
7394 If @var{type} is not given, the default type depends on the capabilities
7395 of the assembler and linker used. @var{type} may be one of
7396 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
7397 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
7398 compression in traditional GNU format). If the linker doesn't support
7399 writing compressed debug sections, the option is rejected. Otherwise,
7400 if the assembler does not support them, @option{-gz} is silently ignored
7401 when producing object files.
7403 @item -femit-struct-debug-baseonly
7404 @opindex femit-struct-debug-baseonly
7405 Emit debug information for struct-like types
7406 only when the base name of the compilation source file
7407 matches the base name of file in which the struct is defined.
7409 This option substantially reduces the size of debugging information,
7410 but at significant potential loss in type information to the debugger.
7411 See @option{-femit-struct-debug-reduced} for a less aggressive option.
7412 See @option{-femit-struct-debug-detailed} for more detailed control.
7414 This option works only with DWARF debug output.
7416 @item -femit-struct-debug-reduced
7417 @opindex femit-struct-debug-reduced
7418 Emit debug information for struct-like types
7419 only when the base name of the compilation source file
7420 matches the base name of file in which the type is defined,
7421 unless the struct is a template or defined in a system header.
7423 This option significantly reduces the size of debugging information,
7424 with some potential loss in type information to the debugger.
7425 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
7426 See @option{-femit-struct-debug-detailed} for more detailed control.
7428 This option works only with DWARF debug output.
7430 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
7431 @opindex femit-struct-debug-detailed
7432 Specify the struct-like types
7433 for which the compiler generates debug information.
7434 The intent is to reduce duplicate struct debug information
7435 between different object files within the same program.
7437 This option is a detailed version of
7438 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
7439 which serves for most needs.
7441 A specification has the syntax@*
7442 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
7444 The optional first word limits the specification to
7445 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
7446 A struct type is used directly when it is the type of a variable, member.
7447 Indirect uses arise through pointers to structs.
7448 That is, when use of an incomplete struct is valid, the use is indirect.
7450 @samp{struct one direct; struct two * indirect;}.
7452 The optional second word limits the specification to
7453 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
7454 Generic structs are a bit complicated to explain.
7455 For C++, these are non-explicit specializations of template classes,
7456 or non-template classes within the above.
7457 Other programming languages have generics,
7458 but @option{-femit-struct-debug-detailed} does not yet implement them.
7460 The third word specifies the source files for those
7461 structs for which the compiler should emit debug information.
7462 The values @samp{none} and @samp{any} have the normal meaning.
7463 The value @samp{base} means that
7464 the base of name of the file in which the type declaration appears
7465 must match the base of the name of the main compilation file.
7466 In practice, this means that when compiling @file{foo.c}, debug information
7467 is generated for types declared in that file and @file{foo.h},
7468 but not other header files.
7469 The value @samp{sys} means those types satisfying @samp{base}
7470 or declared in system or compiler headers.
7472 You may need to experiment to determine the best settings for your application.
7474 The default is @option{-femit-struct-debug-detailed=all}.
7476 This option works only with DWARF debug output.
7478 @item -fno-dwarf2-cfi-asm
7479 @opindex fdwarf2-cfi-asm
7480 @opindex fno-dwarf2-cfi-asm
7481 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
7482 instead of using GAS @code{.cfi_*} directives.
7484 @item -fno-eliminate-unused-debug-types
7485 @opindex feliminate-unused-debug-types
7486 @opindex fno-eliminate-unused-debug-types
7487 Normally, when producing DWARF output, GCC avoids producing debug symbol
7488 output for types that are nowhere used in the source file being compiled.
7489 Sometimes it is useful to have GCC emit debugging
7490 information for all types declared in a compilation
7491 unit, regardless of whether or not they are actually used
7492 in that compilation unit, for example
7493 if, in the debugger, you want to cast a value to a type that is
7494 not actually used in your program (but is declared). More often,
7495 however, this results in a significant amount of wasted space.
7498 @node Optimize Options
7499 @section Options That Control Optimization
7500 @cindex optimize options
7501 @cindex options, optimization
7503 These options control various sorts of optimizations.
7505 Without any optimization option, the compiler's goal is to reduce the
7506 cost of compilation and to make debugging produce the expected
7507 results. Statements are independent: if you stop the program with a
7508 breakpoint between statements, you can then assign a new value to any
7509 variable or change the program counter to any other statement in the
7510 function and get exactly the results you expect from the source
7513 Turning on optimization flags makes the compiler attempt to improve
7514 the performance and/or code size at the expense of compilation time
7515 and possibly the ability to debug the program.
7517 The compiler performs optimization based on the knowledge it has of the
7518 program. Compiling multiple files at once to a single output file mode allows
7519 the compiler to use information gained from all of the files when compiling
7522 Not all optimizations are controlled directly by a flag. Only
7523 optimizations that have a flag are listed in this section.
7525 Most optimizations are only enabled if an @option{-O} level is set on
7526 the command line. Otherwise they are disabled, even if individual
7527 optimization flags are specified.
7529 Depending on the target and how GCC was configured, a slightly different
7530 set of optimizations may be enabled at each @option{-O} level than
7531 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7532 to find out the exact set of optimizations that are enabled at each level.
7533 @xref{Overall Options}, for examples.
7540 Optimize. Optimizing compilation takes somewhat more time, and a lot
7541 more memory for a large function.
7543 With @option{-O}, the compiler tries to reduce code size and execution
7544 time, without performing any optimizations that take a great deal of
7547 @option{-O} turns on the following optimization flags:
7550 -fbranch-count-reg @gol
7551 -fcombine-stack-adjustments @gol
7553 -fcprop-registers @gol
7556 -fdelayed-branch @gol
7558 -fforward-propagate @gol
7559 -fguess-branch-probability @gol
7560 -fif-conversion2 @gol
7561 -fif-conversion @gol
7562 -finline-functions-called-once @gol
7563 -fipa-pure-const @gol
7565 -fipa-reference @gol
7566 -fmerge-constants @gol
7567 -fmove-loop-invariants @gol
7568 -fomit-frame-pointer @gol
7569 -freorder-blocks @gol
7571 -fshrink-wrap-separate @gol
7572 -fsplit-wide-types @gol
7578 -ftree-coalesce-vars @gol
7579 -ftree-copy-prop @gol
7581 -ftree-dominator-opts @gol
7583 -ftree-forwprop @gol
7595 Optimize even more. GCC performs nearly all supported optimizations
7596 that do not involve a space-speed tradeoff.
7597 As compared to @option{-O}, this option increases both compilation time
7598 and the performance of the generated code.
7600 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7601 also turns on the following optimization flags:
7602 @gccoptlist{-fthread-jumps @gol
7603 -falign-functions -falign-jumps @gol
7604 -falign-loops -falign-labels @gol
7607 -fcse-follow-jumps -fcse-skip-blocks @gol
7608 -fdelete-null-pointer-checks @gol
7609 -fdevirtualize -fdevirtualize-speculatively @gol
7610 -fexpensive-optimizations @gol
7611 -fgcse -fgcse-lm @gol
7612 -fhoist-adjacent-loads @gol
7613 -finline-small-functions @gol
7614 -findirect-inlining @gol
7620 -fisolate-erroneous-paths-dereference @gol
7622 -foptimize-sibling-calls @gol
7623 -foptimize-strlen @gol
7624 -fpartial-inlining @gol
7626 -freorder-blocks-algorithm=stc @gol
7627 -freorder-blocks-and-partition -freorder-functions @gol
7628 -frerun-cse-after-loop @gol
7629 -fsched-interblock -fsched-spec @gol
7630 -fschedule-insns -fschedule-insns2 @gol
7631 -fstore-merging @gol
7632 -fstrict-aliasing @gol
7633 -ftree-builtin-call-dce @gol
7634 -ftree-switch-conversion -ftree-tail-merge @gol
7635 -fcode-hoisting @gol
7640 Please note the warning under @option{-fgcse} about
7641 invoking @option{-O2} on programs that use computed gotos.
7645 Optimize yet more. @option{-O3} turns on all optimizations specified
7646 by @option{-O2} and also turns on the following optimization flags:
7647 @gccoptlist{-finline-functions @gol
7648 -funswitch-loops @gol
7649 -fpredictive-commoning @gol
7650 -fgcse-after-reload @gol
7651 -ftree-loop-vectorize @gol
7652 -ftree-loop-distribution @gol
7653 -ftree-loop-distribute-patterns @gol
7654 -floop-interchange @gol
7656 -ftree-slp-vectorize @gol
7657 -fvect-cost-model @gol
7658 -ftree-partial-pre @gol
7664 Reduce compilation time and make debugging produce the expected
7665 results. This is the default.
7669 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7670 do not typically increase code size. It also performs further
7671 optimizations designed to reduce code size.
7673 @option{-Os} disables the following optimization flags:
7674 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7675 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7676 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7680 Disregard strict standards compliance. @option{-Ofast} enables all
7681 @option{-O3} optimizations. It also enables optimizations that are not
7682 valid for all standard-compliant programs.
7683 It turns on @option{-ffast-math} and the Fortran-specific
7684 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
7685 specified, and @option{-fno-protect-parens}.
7689 Optimize debugging experience. @option{-Og} enables optimizations
7690 that do not interfere with debugging. It should be the optimization
7691 level of choice for the standard edit-compile-debug cycle, offering
7692 a reasonable level of optimization while maintaining fast compilation
7693 and a good debugging experience.
7696 If you use multiple @option{-O} options, with or without level numbers,
7697 the last such option is the one that is effective.
7699 Options of the form @option{-f@var{flag}} specify machine-independent
7700 flags. Most flags have both positive and negative forms; the negative
7701 form of @option{-ffoo} is @option{-fno-foo}. In the table
7702 below, only one of the forms is listed---the one you typically
7703 use. You can figure out the other form by either removing @samp{no-}
7706 The following options control specific optimizations. They are either
7707 activated by @option{-O} options or are related to ones that are. You
7708 can use the following flags in the rare cases when ``fine-tuning'' of
7709 optimizations to be performed is desired.
7712 @item -fno-defer-pop
7713 @opindex fno-defer-pop
7714 Always pop the arguments to each function call as soon as that function
7715 returns. For machines that must pop arguments after a function call,
7716 the compiler normally lets arguments accumulate on the stack for several
7717 function calls and pops them all at once.
7719 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7721 @item -fforward-propagate
7722 @opindex fforward-propagate
7723 Perform a forward propagation pass on RTL@. The pass tries to combine two
7724 instructions and checks if the result can be simplified. If loop unrolling
7725 is active, two passes are performed and the second is scheduled after
7728 This option is enabled by default at optimization levels @option{-O},
7729 @option{-O2}, @option{-O3}, @option{-Os}.
7731 @item -ffp-contract=@var{style}
7732 @opindex ffp-contract
7733 @option{-ffp-contract=off} disables floating-point expression contraction.
7734 @option{-ffp-contract=fast} enables floating-point expression contraction
7735 such as forming of fused multiply-add operations if the target has
7736 native support for them.
7737 @option{-ffp-contract=on} enables floating-point expression contraction
7738 if allowed by the language standard. This is currently not implemented
7739 and treated equal to @option{-ffp-contract=off}.
7741 The default is @option{-ffp-contract=fast}.
7743 @item -fomit-frame-pointer
7744 @opindex fomit-frame-pointer
7745 Omit the frame pointer in functions that don't need one. This avoids the
7746 instructions to save, set up and restore the frame pointer; on many targets
7747 it also makes an extra register available.
7749 On some targets this flag has no effect because the standard calling sequence
7750 always uses a frame pointer, so it cannot be omitted.
7752 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
7753 is used in all functions. Several targets always omit the frame pointer in
7756 Enabled by default at @option{-O} and higher.
7758 @item -foptimize-sibling-calls
7759 @opindex foptimize-sibling-calls
7760 Optimize sibling and tail recursive calls.
7762 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7764 @item -foptimize-strlen
7765 @opindex foptimize-strlen
7766 Optimize various standard C string functions (e.g. @code{strlen},
7767 @code{strchr} or @code{strcpy}) and
7768 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7770 Enabled at levels @option{-O2}, @option{-O3}.
7774 Do not expand any functions inline apart from those marked with
7775 the @code{always_inline} attribute. This is the default when not
7778 Single functions can be exempted from inlining by marking them
7779 with the @code{noinline} attribute.
7781 @item -finline-small-functions
7782 @opindex finline-small-functions
7783 Integrate functions into their callers when their body is smaller than expected
7784 function call code (so overall size of program gets smaller). The compiler
7785 heuristically decides which functions are simple enough to be worth integrating
7786 in this way. This inlining applies to all functions, even those not declared
7789 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7791 @item -findirect-inlining
7792 @opindex findirect-inlining
7793 Inline also indirect calls that are discovered to be known at compile
7794 time thanks to previous inlining. This option has any effect only
7795 when inlining itself is turned on by the @option{-finline-functions}
7796 or @option{-finline-small-functions} options.
7798 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7800 @item -finline-functions
7801 @opindex finline-functions
7802 Consider all functions for inlining, even if they are not declared inline.
7803 The compiler heuristically decides which functions are worth integrating
7806 If all calls to a given function are integrated, and the function is
7807 declared @code{static}, then the function is normally not output as
7808 assembler code in its own right.
7810 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7812 @item -finline-functions-called-once
7813 @opindex finline-functions-called-once
7814 Consider all @code{static} functions called once for inlining into their
7815 caller even if they are not marked @code{inline}. If a call to a given
7816 function is integrated, then the function is not output as assembler code
7819 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7821 @item -fearly-inlining
7822 @opindex fearly-inlining
7823 Inline functions marked by @code{always_inline} and functions whose body seems
7824 smaller than the function call overhead early before doing
7825 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7826 makes profiling significantly cheaper and usually inlining faster on programs
7827 having large chains of nested wrapper functions.
7833 Perform interprocedural scalar replacement of aggregates, removal of
7834 unused parameters and replacement of parameters passed by reference
7835 by parameters passed by value.
7837 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7839 @item -finline-limit=@var{n}
7840 @opindex finline-limit
7841 By default, GCC limits the size of functions that can be inlined. This flag
7842 allows coarse control of this limit. @var{n} is the size of functions that
7843 can be inlined in number of pseudo instructions.
7845 Inlining is actually controlled by a number of parameters, which may be
7846 specified individually by using @option{--param @var{name}=@var{value}}.
7847 The @option{-finline-limit=@var{n}} option sets some of these parameters
7851 @item max-inline-insns-single
7852 is set to @var{n}/2.
7853 @item max-inline-insns-auto
7854 is set to @var{n}/2.
7857 See below for a documentation of the individual
7858 parameters controlling inlining and for the defaults of these parameters.
7860 @emph{Note:} there may be no value to @option{-finline-limit} that results
7861 in default behavior.
7863 @emph{Note:} pseudo instruction represents, in this particular context, an
7864 abstract measurement of function's size. In no way does it represent a count
7865 of assembly instructions and as such its exact meaning might change from one
7866 release to an another.
7868 @item -fno-keep-inline-dllexport
7869 @opindex fno-keep-inline-dllexport
7870 This is a more fine-grained version of @option{-fkeep-inline-functions},
7871 which applies only to functions that are declared using the @code{dllexport}
7872 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
7875 @item -fkeep-inline-functions
7876 @opindex fkeep-inline-functions
7877 In C, emit @code{static} functions that are declared @code{inline}
7878 into the object file, even if the function has been inlined into all
7879 of its callers. This switch does not affect functions using the
7880 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7881 inline functions into the object file.
7883 @item -fkeep-static-functions
7884 @opindex fkeep-static-functions
7885 Emit @code{static} functions into the object file, even if the function
7888 @item -fkeep-static-consts
7889 @opindex fkeep-static-consts
7890 Emit variables declared @code{static const} when optimization isn't turned
7891 on, even if the variables aren't referenced.
7893 GCC enables this option by default. If you want to force the compiler to
7894 check if a variable is referenced, regardless of whether or not
7895 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7897 @item -fmerge-constants
7898 @opindex fmerge-constants
7899 Attempt to merge identical constants (string constants and floating-point
7900 constants) across compilation units.
7902 This option is the default for optimized compilation if the assembler and
7903 linker support it. Use @option{-fno-merge-constants} to inhibit this
7906 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7908 @item -fmerge-all-constants
7909 @opindex fmerge-all-constants
7910 Attempt to merge identical constants and identical variables.
7912 This option implies @option{-fmerge-constants}. In addition to
7913 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7914 arrays or initialized constant variables with integral or floating-point
7915 types. Languages like C or C++ require each variable, including multiple
7916 instances of the same variable in recursive calls, to have distinct locations,
7917 so using this option results in non-conforming
7920 @item -fmodulo-sched
7921 @opindex fmodulo-sched
7922 Perform swing modulo scheduling immediately before the first scheduling
7923 pass. This pass looks at innermost loops and reorders their
7924 instructions by overlapping different iterations.
7926 @item -fmodulo-sched-allow-regmoves
7927 @opindex fmodulo-sched-allow-regmoves
7928 Perform more aggressive SMS-based modulo scheduling with register moves
7929 allowed. By setting this flag certain anti-dependences edges are
7930 deleted, which triggers the generation of reg-moves based on the
7931 life-range analysis. This option is effective only with
7932 @option{-fmodulo-sched} enabled.
7934 @item -fno-branch-count-reg
7935 @opindex fno-branch-count-reg
7936 Avoid running a pass scanning for opportunities to use ``decrement and
7937 branch'' instructions on a count register instead of generating sequences
7938 of instructions that decrement a register, compare it against zero, and
7939 then branch based upon the result. This option is only meaningful on
7940 architectures that support such instructions, which include x86, PowerPC,
7941 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7942 doesn't remove the decrement and branch instructions from the generated
7943 instruction stream introduced by other optimization passes.
7945 Enabled by default at @option{-O1} and higher.
7947 The default is @option{-fbranch-count-reg}.
7949 @item -fno-function-cse
7950 @opindex fno-function-cse
7951 Do not put function addresses in registers; make each instruction that
7952 calls a constant function contain the function's address explicitly.
7954 This option results in less efficient code, but some strange hacks
7955 that alter the assembler output may be confused by the optimizations
7956 performed when this option is not used.
7958 The default is @option{-ffunction-cse}
7960 @item -fno-zero-initialized-in-bss
7961 @opindex fno-zero-initialized-in-bss
7962 If the target supports a BSS section, GCC by default puts variables that
7963 are initialized to zero into BSS@. This can save space in the resulting
7966 This option turns off this behavior because some programs explicitly
7967 rely on variables going to the data section---e.g., so that the
7968 resulting executable can find the beginning of that section and/or make
7969 assumptions based on that.
7971 The default is @option{-fzero-initialized-in-bss}.
7973 @item -fthread-jumps
7974 @opindex fthread-jumps
7975 Perform optimizations that check to see if a jump branches to a
7976 location where another comparison subsumed by the first is found. If
7977 so, the first branch is redirected to either the destination of the
7978 second branch or a point immediately following it, depending on whether
7979 the condition is known to be true or false.
7981 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7983 @item -fsplit-wide-types
7984 @opindex fsplit-wide-types
7985 When using a type that occupies multiple registers, such as @code{long
7986 long} on a 32-bit system, split the registers apart and allocate them
7987 independently. This normally generates better code for those types,
7988 but may make debugging more difficult.
7990 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7993 @item -fcse-follow-jumps
7994 @opindex fcse-follow-jumps
7995 In common subexpression elimination (CSE), scan through jump instructions
7996 when the target of the jump is not reached by any other path. For
7997 example, when CSE encounters an @code{if} statement with an
7998 @code{else} clause, CSE follows the jump when the condition
8001 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8003 @item -fcse-skip-blocks
8004 @opindex fcse-skip-blocks
8005 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8006 follow jumps that conditionally skip over blocks. When CSE
8007 encounters a simple @code{if} statement with no else clause,
8008 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8009 body of the @code{if}.
8011 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8013 @item -frerun-cse-after-loop
8014 @opindex frerun-cse-after-loop
8015 Re-run common subexpression elimination after loop optimizations are
8018 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8022 Perform a global common subexpression elimination pass.
8023 This pass also performs global constant and copy propagation.
8025 @emph{Note:} When compiling a program using computed gotos, a GCC
8026 extension, you may get better run-time performance if you disable
8027 the global common subexpression elimination pass by adding
8028 @option{-fno-gcse} to the command line.
8030 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8034 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8035 attempts to move loads that are only killed by stores into themselves. This
8036 allows a loop containing a load/store sequence to be changed to a load outside
8037 the loop, and a copy/store within the loop.
8039 Enabled by default when @option{-fgcse} is enabled.
8043 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8044 global common subexpression elimination. This pass attempts to move
8045 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8046 loops containing a load/store sequence can be changed to a load before
8047 the loop and a store after the loop.
8049 Not enabled at any optimization level.
8053 When @option{-fgcse-las} is enabled, the global common subexpression
8054 elimination pass eliminates redundant loads that come after stores to the
8055 same memory location (both partial and full redundancies).
8057 Not enabled at any optimization level.
8059 @item -fgcse-after-reload
8060 @opindex fgcse-after-reload
8061 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8062 pass is performed after reload. The purpose of this pass is to clean up
8065 @item -faggressive-loop-optimizations
8066 @opindex faggressive-loop-optimizations
8067 This option tells the loop optimizer to use language constraints to
8068 derive bounds for the number of iterations of a loop. This assumes that
8069 loop code does not invoke undefined behavior by for example causing signed
8070 integer overflows or out-of-bound array accesses. The bounds for the
8071 number of iterations of a loop are used to guide loop unrolling and peeling
8072 and loop exit test optimizations.
8073 This option is enabled by default.
8075 @item -funconstrained-commons
8076 @opindex funconstrained-commons
8077 This option tells the compiler that variables declared in common blocks
8078 (e.g. Fortran) may later be overridden with longer trailing arrays. This
8079 prevents certain optimizations that depend on knowing the array bounds.
8081 @item -fcrossjumping
8082 @opindex fcrossjumping
8083 Perform cross-jumping transformation.
8084 This transformation unifies equivalent code and saves code size. The
8085 resulting code may or may not perform better than without cross-jumping.
8087 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8089 @item -fauto-inc-dec
8090 @opindex fauto-inc-dec
8091 Combine increments or decrements of addresses with memory accesses.
8092 This pass is always skipped on architectures that do not have
8093 instructions to support this. Enabled by default at @option{-O} and
8094 higher on architectures that support this.
8098 Perform dead code elimination (DCE) on RTL@.
8099 Enabled by default at @option{-O} and higher.
8103 Perform dead store elimination (DSE) on RTL@.
8104 Enabled by default at @option{-O} and higher.
8106 @item -fif-conversion
8107 @opindex fif-conversion
8108 Attempt to transform conditional jumps into branch-less equivalents. This
8109 includes use of conditional moves, min, max, set flags and abs instructions, and
8110 some tricks doable by standard arithmetics. The use of conditional execution
8111 on chips where it is available is controlled by @option{-fif-conversion2}.
8113 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8115 @item -fif-conversion2
8116 @opindex fif-conversion2
8117 Use conditional execution (where available) to transform conditional jumps into
8118 branch-less equivalents.
8120 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8122 @item -fdeclone-ctor-dtor
8123 @opindex fdeclone-ctor-dtor
8124 The C++ ABI requires multiple entry points for constructors and
8125 destructors: one for a base subobject, one for a complete object, and
8126 one for a virtual destructor that calls operator delete afterwards.
8127 For a hierarchy with virtual bases, the base and complete variants are
8128 clones, which means two copies of the function. With this option, the
8129 base and complete variants are changed to be thunks that call a common
8132 Enabled by @option{-Os}.
8134 @item -fdelete-null-pointer-checks
8135 @opindex fdelete-null-pointer-checks
8136 Assume that programs cannot safely dereference null pointers, and that
8137 no code or data element resides at address zero.
8138 This option enables simple constant
8139 folding optimizations at all optimization levels. In addition, other
8140 optimization passes in GCC use this flag to control global dataflow
8141 analyses that eliminate useless checks for null pointers; these assume
8142 that a memory access to address zero always results in a trap, so
8143 that if a pointer is checked after it has already been dereferenced,
8146 Note however that in some environments this assumption is not true.
8147 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8148 for programs that depend on that behavior.
8150 This option is enabled by default on most targets. On Nios II ELF, it
8151 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
8153 Passes that use the dataflow information
8154 are enabled independently at different optimization levels.
8156 @item -fdevirtualize
8157 @opindex fdevirtualize
8158 Attempt to convert calls to virtual functions to direct calls. This
8159 is done both within a procedure and interprocedurally as part of
8160 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8161 propagation (@option{-fipa-cp}).
8162 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8164 @item -fdevirtualize-speculatively
8165 @opindex fdevirtualize-speculatively
8166 Attempt to convert calls to virtual functions to speculative direct calls.
8167 Based on the analysis of the type inheritance graph, determine for a given call
8168 the set of likely targets. If the set is small, preferably of size 1, change
8169 the call into a conditional deciding between direct and indirect calls. The
8170 speculative calls enable more optimizations, such as inlining. When they seem
8171 useless after further optimization, they are converted back into original form.
8173 @item -fdevirtualize-at-ltrans
8174 @opindex fdevirtualize-at-ltrans
8175 Stream extra information needed for aggressive devirtualization when running
8176 the link-time optimizer in local transformation mode.
8177 This option enables more devirtualization but
8178 significantly increases the size of streamed data. For this reason it is
8179 disabled by default.
8181 @item -fexpensive-optimizations
8182 @opindex fexpensive-optimizations
8183 Perform a number of minor optimizations that are relatively expensive.
8185 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8189 Attempt to remove redundant extension instructions. This is especially
8190 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8191 registers after writing to their lower 32-bit half.
8193 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8194 @option{-O3}, @option{-Os}.
8196 @item -fno-lifetime-dse
8197 @opindex fno-lifetime-dse
8198 In C++ the value of an object is only affected by changes within its
8199 lifetime: when the constructor begins, the object has an indeterminate
8200 value, and any changes during the lifetime of the object are dead when
8201 the object is destroyed. Normally dead store elimination will take
8202 advantage of this; if your code relies on the value of the object
8203 storage persisting beyond the lifetime of the object, you can use this
8204 flag to disable this optimization. To preserve stores before the
8205 constructor starts (e.g. because your operator new clears the object
8206 storage) but still treat the object as dead after the destructor you,
8207 can use @option{-flifetime-dse=1}. The default behavior can be
8208 explicitly selected with @option{-flifetime-dse=2}.
8209 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
8211 @item -flive-range-shrinkage
8212 @opindex flive-range-shrinkage
8213 Attempt to decrease register pressure through register live range
8214 shrinkage. This is helpful for fast processors with small or moderate
8217 @item -fira-algorithm=@var{algorithm}
8218 @opindex fira-algorithm
8219 Use the specified coloring algorithm for the integrated register
8220 allocator. The @var{algorithm} argument can be @samp{priority}, which
8221 specifies Chow's priority coloring, or @samp{CB}, which specifies
8222 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8223 for all architectures, but for those targets that do support it, it is
8224 the default because it generates better code.
8226 @item -fira-region=@var{region}
8227 @opindex fira-region
8228 Use specified regions for the integrated register allocator. The
8229 @var{region} argument should be one of the following:
8234 Use all loops as register allocation regions.
8235 This can give the best results for machines with a small and/or
8236 irregular register set.
8239 Use all loops except for loops with small register pressure
8240 as the regions. This value usually gives
8241 the best results in most cases and for most architectures,
8242 and is enabled by default when compiling with optimization for speed
8243 (@option{-O}, @option{-O2}, @dots{}).
8246 Use all functions as a single region.
8247 This typically results in the smallest code size, and is enabled by default for
8248 @option{-Os} or @option{-O0}.
8252 @item -fira-hoist-pressure
8253 @opindex fira-hoist-pressure
8254 Use IRA to evaluate register pressure in the code hoisting pass for
8255 decisions to hoist expressions. This option usually results in smaller
8256 code, but it can slow the compiler down.
8258 This option is enabled at level @option{-Os} for all targets.
8260 @item -fira-loop-pressure
8261 @opindex fira-loop-pressure
8262 Use IRA to evaluate register pressure in loops for decisions to move
8263 loop invariants. This option usually results in generation
8264 of faster and smaller code on machines with large register files (>= 32
8265 registers), but it can slow the compiler down.
8267 This option is enabled at level @option{-O3} for some targets.
8269 @item -fno-ira-share-save-slots
8270 @opindex fno-ira-share-save-slots
8271 Disable sharing of stack slots used for saving call-used hard
8272 registers living through a call. Each hard register gets a
8273 separate stack slot, and as a result function stack frames are
8276 @item -fno-ira-share-spill-slots
8277 @opindex fno-ira-share-spill-slots
8278 Disable sharing of stack slots allocated for pseudo-registers. Each
8279 pseudo-register that does not get a hard register gets a separate
8280 stack slot, and as a result function stack frames are larger.
8284 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8285 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8286 values if it is profitable.
8288 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8290 @item -fdelayed-branch
8291 @opindex fdelayed-branch
8292 If supported for the target machine, attempt to reorder instructions
8293 to exploit instruction slots available after delayed branch
8296 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8298 @item -fschedule-insns
8299 @opindex fschedule-insns
8300 If supported for the target machine, attempt to reorder instructions to
8301 eliminate execution stalls due to required data being unavailable. This
8302 helps machines that have slow floating point or memory load instructions
8303 by allowing other instructions to be issued until the result of the load
8304 or floating-point instruction is required.
8306 Enabled at levels @option{-O2}, @option{-O3}.
8308 @item -fschedule-insns2
8309 @opindex fschedule-insns2
8310 Similar to @option{-fschedule-insns}, but requests an additional pass of
8311 instruction scheduling after register allocation has been done. This is
8312 especially useful on machines with a relatively small number of
8313 registers and where memory load instructions take more than one cycle.
8315 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8317 @item -fno-sched-interblock
8318 @opindex fno-sched-interblock
8319 Don't schedule instructions across basic blocks. This is normally
8320 enabled by default when scheduling before register allocation, i.e.@:
8321 with @option{-fschedule-insns} or at @option{-O2} or higher.
8323 @item -fno-sched-spec
8324 @opindex fno-sched-spec
8325 Don't allow speculative motion of non-load instructions. This is normally
8326 enabled by default when scheduling before register allocation, i.e.@:
8327 with @option{-fschedule-insns} or at @option{-O2} or higher.
8329 @item -fsched-pressure
8330 @opindex fsched-pressure
8331 Enable register pressure sensitive insn scheduling before register
8332 allocation. This only makes sense when scheduling before register
8333 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8334 @option{-O2} or higher. Usage of this option can improve the
8335 generated code and decrease its size by preventing register pressure
8336 increase above the number of available hard registers and subsequent
8337 spills in register allocation.
8339 @item -fsched-spec-load
8340 @opindex fsched-spec-load
8341 Allow speculative motion of some load instructions. This only makes
8342 sense when scheduling before register allocation, i.e.@: with
8343 @option{-fschedule-insns} or at @option{-O2} or higher.
8345 @item -fsched-spec-load-dangerous
8346 @opindex fsched-spec-load-dangerous
8347 Allow speculative motion of more load instructions. This only makes
8348 sense when scheduling before register allocation, i.e.@: with
8349 @option{-fschedule-insns} or at @option{-O2} or higher.
8351 @item -fsched-stalled-insns
8352 @itemx -fsched-stalled-insns=@var{n}
8353 @opindex fsched-stalled-insns
8354 Define how many insns (if any) can be moved prematurely from the queue
8355 of stalled insns into the ready list during the second scheduling pass.
8356 @option{-fno-sched-stalled-insns} means that no insns are moved
8357 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8358 on how many queued insns can be moved prematurely.
8359 @option{-fsched-stalled-insns} without a value is equivalent to
8360 @option{-fsched-stalled-insns=1}.
8362 @item -fsched-stalled-insns-dep
8363 @itemx -fsched-stalled-insns-dep=@var{n}
8364 @opindex fsched-stalled-insns-dep
8365 Define how many insn groups (cycles) are examined for a dependency
8366 on a stalled insn that is a candidate for premature removal from the queue
8367 of stalled insns. This has an effect only during the second scheduling pass,
8368 and only if @option{-fsched-stalled-insns} is used.
8369 @option{-fno-sched-stalled-insns-dep} is equivalent to
8370 @option{-fsched-stalled-insns-dep=0}.
8371 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8372 @option{-fsched-stalled-insns-dep=1}.
8374 @item -fsched2-use-superblocks
8375 @opindex fsched2-use-superblocks
8376 When scheduling after register allocation, use superblock scheduling.
8377 This allows motion across basic block boundaries,
8378 resulting in faster schedules. This option is experimental, as not all machine
8379 descriptions used by GCC model the CPU closely enough to avoid unreliable
8380 results from the algorithm.
8382 This only makes sense when scheduling after register allocation, i.e.@: with
8383 @option{-fschedule-insns2} or at @option{-O2} or higher.
8385 @item -fsched-group-heuristic
8386 @opindex fsched-group-heuristic
8387 Enable the group heuristic in the scheduler. This heuristic favors
8388 the instruction that belongs to a schedule group. This is enabled
8389 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8390 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8392 @item -fsched-critical-path-heuristic
8393 @opindex fsched-critical-path-heuristic
8394 Enable the critical-path heuristic in the scheduler. This heuristic favors
8395 instructions on the critical path. This is enabled by default when
8396 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8397 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8399 @item -fsched-spec-insn-heuristic
8400 @opindex fsched-spec-insn-heuristic
8401 Enable the speculative instruction heuristic in the scheduler. This
8402 heuristic favors speculative instructions with greater dependency weakness.
8403 This is enabled by default when scheduling is enabled, i.e.@:
8404 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8405 or at @option{-O2} or higher.
8407 @item -fsched-rank-heuristic
8408 @opindex fsched-rank-heuristic
8409 Enable the rank heuristic in the scheduler. This heuristic favors
8410 the instruction belonging to a basic block with greater size or frequency.
8411 This is enabled by default when scheduling is enabled, i.e.@:
8412 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8413 at @option{-O2} or higher.
8415 @item -fsched-last-insn-heuristic
8416 @opindex fsched-last-insn-heuristic
8417 Enable the last-instruction heuristic in the scheduler. This heuristic
8418 favors the instruction that is less dependent on the last instruction
8419 scheduled. This is enabled by default when scheduling is enabled,
8420 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8421 at @option{-O2} or higher.
8423 @item -fsched-dep-count-heuristic
8424 @opindex fsched-dep-count-heuristic
8425 Enable the dependent-count heuristic in the scheduler. This heuristic
8426 favors the instruction that has more instructions depending on it.
8427 This is enabled by default when scheduling is enabled, i.e.@:
8428 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8429 at @option{-O2} or higher.
8431 @item -freschedule-modulo-scheduled-loops
8432 @opindex freschedule-modulo-scheduled-loops
8433 Modulo scheduling is performed before traditional scheduling. If a loop
8434 is modulo scheduled, later scheduling passes may change its schedule.
8435 Use this option to control that behavior.
8437 @item -fselective-scheduling
8438 @opindex fselective-scheduling
8439 Schedule instructions using selective scheduling algorithm. Selective
8440 scheduling runs instead of the first scheduler pass.
8442 @item -fselective-scheduling2
8443 @opindex fselective-scheduling2
8444 Schedule instructions using selective scheduling algorithm. Selective
8445 scheduling runs instead of the second scheduler pass.
8447 @item -fsel-sched-pipelining
8448 @opindex fsel-sched-pipelining
8449 Enable software pipelining of innermost loops during selective scheduling.
8450 This option has no effect unless one of @option{-fselective-scheduling} or
8451 @option{-fselective-scheduling2} is turned on.
8453 @item -fsel-sched-pipelining-outer-loops
8454 @opindex fsel-sched-pipelining-outer-loops
8455 When pipelining loops during selective scheduling, also pipeline outer loops.
8456 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8458 @item -fsemantic-interposition
8459 @opindex fsemantic-interposition
8460 Some object formats, like ELF, allow interposing of symbols by the
8462 This means that for symbols exported from the DSO, the compiler cannot perform
8463 interprocedural propagation, inlining and other optimizations in anticipation
8464 that the function or variable in question may change. While this feature is
8465 useful, for example, to rewrite memory allocation functions by a debugging
8466 implementation, it is expensive in the terms of code quality.
8467 With @option{-fno-semantic-interposition} the compiler assumes that
8468 if interposition happens for functions the overwriting function will have
8469 precisely the same semantics (and side effects).
8470 Similarly if interposition happens
8471 for variables, the constructor of the variable will be the same. The flag
8472 has no effect for functions explicitly declared inline
8473 (where it is never allowed for interposition to change semantics)
8474 and for symbols explicitly declared weak.
8477 @opindex fshrink-wrap
8478 Emit function prologues only before parts of the function that need it,
8479 rather than at the top of the function. This flag is enabled by default at
8480 @option{-O} and higher.
8482 @item -fshrink-wrap-separate
8483 @opindex fshrink-wrap-separate
8484 Shrink-wrap separate parts of the prologue and epilogue separately, so that
8485 those parts are only executed when needed.
8486 This option is on by default, but has no effect unless @option{-fshrink-wrap}
8487 is also turned on and the target supports this.
8489 @item -fcaller-saves
8490 @opindex fcaller-saves
8491 Enable allocation of values to registers that are clobbered by
8492 function calls, by emitting extra instructions to save and restore the
8493 registers around such calls. Such allocation is done only when it
8494 seems to result in better code.
8496 This option is always enabled by default on certain machines, usually
8497 those which have no call-preserved registers to use instead.
8499 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8501 @item -fcombine-stack-adjustments
8502 @opindex fcombine-stack-adjustments
8503 Tracks stack adjustments (pushes and pops) and stack memory references
8504 and then tries to find ways to combine them.
8506 Enabled by default at @option{-O1} and higher.
8510 Use caller save registers for allocation if those registers are not used by
8511 any called function. In that case it is not necessary to save and restore
8512 them around calls. This is only possible if called functions are part of
8513 same compilation unit as current function and they are compiled before it.
8515 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
8516 is disabled if generated code will be instrumented for profiling
8517 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
8518 exactly (this happens on targets that do not expose prologues
8519 and epilogues in RTL).
8521 @item -fconserve-stack
8522 @opindex fconserve-stack
8523 Attempt to minimize stack usage. The compiler attempts to use less
8524 stack space, even if that makes the program slower. This option
8525 implies setting the @option{large-stack-frame} parameter to 100
8526 and the @option{large-stack-frame-growth} parameter to 400.
8528 @item -ftree-reassoc
8529 @opindex ftree-reassoc
8530 Perform reassociation on trees. This flag is enabled by default
8531 at @option{-O} and higher.
8533 @item -fcode-hoisting
8534 @opindex fcode-hoisting
8535 Perform code hoisting. Code hoisting tries to move the
8536 evaluation of expressions executed on all paths to the function exit
8537 as early as possible. This is especially useful as a code size
8538 optimization, but it often helps for code speed as well.
8539 This flag is enabled by default at @option{-O2} and higher.
8543 Perform partial redundancy elimination (PRE) on trees. This flag is
8544 enabled by default at @option{-O2} and @option{-O3}.
8546 @item -ftree-partial-pre
8547 @opindex ftree-partial-pre
8548 Make partial redundancy elimination (PRE) more aggressive. This flag is
8549 enabled by default at @option{-O3}.
8551 @item -ftree-forwprop
8552 @opindex ftree-forwprop
8553 Perform forward propagation on trees. This flag is enabled by default
8554 at @option{-O} and higher.
8558 Perform full redundancy elimination (FRE) on trees. The difference
8559 between FRE and PRE is that FRE only considers expressions
8560 that are computed on all paths leading to the redundant computation.
8561 This analysis is faster than PRE, though it exposes fewer redundancies.
8562 This flag is enabled by default at @option{-O} and higher.
8564 @item -ftree-phiprop
8565 @opindex ftree-phiprop
8566 Perform hoisting of loads from conditional pointers on trees. This
8567 pass is enabled by default at @option{-O} and higher.
8569 @item -fhoist-adjacent-loads
8570 @opindex fhoist-adjacent-loads
8571 Speculatively hoist loads from both branches of an if-then-else if the
8572 loads are from adjacent locations in the same structure and the target
8573 architecture has a conditional move instruction. This flag is enabled
8574 by default at @option{-O2} and higher.
8576 @item -ftree-copy-prop
8577 @opindex ftree-copy-prop
8578 Perform copy propagation on trees. This pass eliminates unnecessary
8579 copy operations. This flag is enabled by default at @option{-O} and
8582 @item -fipa-pure-const
8583 @opindex fipa-pure-const
8584 Discover which functions are pure or constant.
8585 Enabled by default at @option{-O} and higher.
8587 @item -fipa-reference
8588 @opindex fipa-reference
8589 Discover which static variables do not escape the
8591 Enabled by default at @option{-O} and higher.
8595 Perform interprocedural pointer analysis and interprocedural modification
8596 and reference analysis. This option can cause excessive memory and
8597 compile-time usage on large compilation units. It is not enabled by
8598 default at any optimization level.
8601 @opindex fipa-profile
8602 Perform interprocedural profile propagation. The functions called only from
8603 cold functions are marked as cold. Also functions executed once (such as
8604 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8605 functions and loop less parts of functions executed once are then optimized for
8607 Enabled by default at @option{-O} and higher.
8611 Perform interprocedural constant propagation.
8612 This optimization analyzes the program to determine when values passed
8613 to functions are constants and then optimizes accordingly.
8614 This optimization can substantially increase performance
8615 if the application has constants passed to functions.
8616 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8618 @item -fipa-cp-clone
8619 @opindex fipa-cp-clone
8620 Perform function cloning to make interprocedural constant propagation stronger.
8621 When enabled, interprocedural constant propagation performs function cloning
8622 when externally visible function can be called with constant arguments.
8623 Because this optimization can create multiple copies of functions,
8624 it may significantly increase code size
8625 (see @option{--param ipcp-unit-growth=@var{value}}).
8626 This flag is enabled by default at @option{-O3}.
8629 @opindex -fipa-bit-cp
8630 When enabled, perform interprocedural bitwise constant
8631 propagation. This flag is enabled by default at @option{-O2}. It
8632 requires that @option{-fipa-cp} is enabled.
8636 When enabled, perform interprocedural propagation of value
8637 ranges. This flag is enabled by default at @option{-O2}. It requires
8638 that @option{-fipa-cp} is enabled.
8642 Perform Identical Code Folding for functions and read-only variables.
8643 The optimization reduces code size and may disturb unwind stacks by replacing
8644 a function by equivalent one with a different name. The optimization works
8645 more effectively with link-time optimization enabled.
8647 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8648 works on different levels and thus the optimizations are not same - there are
8649 equivalences that are found only by GCC and equivalences found only by Gold.
8651 This flag is enabled by default at @option{-O2} and @option{-Os}.
8653 @item -fisolate-erroneous-paths-dereference
8654 @opindex fisolate-erroneous-paths-dereference
8655 Detect paths that trigger erroneous or undefined behavior due to
8656 dereferencing a null pointer. Isolate those paths from the main control
8657 flow and turn the statement with erroneous or undefined behavior into a trap.
8658 This flag is enabled by default at @option{-O2} and higher and depends on
8659 @option{-fdelete-null-pointer-checks} also being enabled.
8661 @item -fisolate-erroneous-paths-attribute
8662 @opindex fisolate-erroneous-paths-attribute
8663 Detect paths that trigger erroneous or undefined behavior due to a null value
8664 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8665 attribute. Isolate those paths from the main control flow and turn the
8666 statement with erroneous or undefined behavior into a trap. This is not
8667 currently enabled, but may be enabled by @option{-O2} in the future.
8671 Perform forward store motion on trees. This flag is
8672 enabled by default at @option{-O} and higher.
8674 @item -ftree-bit-ccp
8675 @opindex ftree-bit-ccp
8676 Perform sparse conditional bit constant propagation on trees and propagate
8677 pointer alignment information.
8678 This pass only operates on local scalar variables and is enabled by default
8679 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8683 Perform sparse conditional constant propagation (CCP) on trees. This
8684 pass only operates on local scalar variables and is enabled by default
8685 at @option{-O} and higher.
8687 @item -fssa-backprop
8688 @opindex fssa-backprop
8689 Propagate information about uses of a value up the definition chain
8690 in order to simplify the definitions. For example, this pass strips
8691 sign operations if the sign of a value never matters. The flag is
8692 enabled by default at @option{-O} and higher.
8695 @opindex fssa-phiopt
8696 Perform pattern matching on SSA PHI nodes to optimize conditional
8697 code. This pass is enabled by default at @option{-O} and higher.
8699 @item -ftree-switch-conversion
8700 @opindex ftree-switch-conversion
8701 Perform conversion of simple initializations in a switch to
8702 initializations from a scalar array. This flag is enabled by default
8703 at @option{-O2} and higher.
8705 @item -ftree-tail-merge
8706 @opindex ftree-tail-merge
8707 Look for identical code sequences. When found, replace one with a jump to the
8708 other. This optimization is known as tail merging or cross jumping. This flag
8709 is enabled by default at @option{-O2} and higher. The compilation time
8711 be limited using @option{max-tail-merge-comparisons} parameter and
8712 @option{max-tail-merge-iterations} parameter.
8716 Perform dead code elimination (DCE) on trees. This flag is enabled by
8717 default at @option{-O} and higher.
8719 @item -ftree-builtin-call-dce
8720 @opindex ftree-builtin-call-dce
8721 Perform conditional dead code elimination (DCE) for calls to built-in functions
8722 that may set @code{errno} but are otherwise free of side effects. This flag is
8723 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8726 @item -ftree-dominator-opts
8727 @opindex ftree-dominator-opts
8728 Perform a variety of simple scalar cleanups (constant/copy
8729 propagation, redundancy elimination, range propagation and expression
8730 simplification) based on a dominator tree traversal. This also
8731 performs jump threading (to reduce jumps to jumps). This flag is
8732 enabled by default at @option{-O} and higher.
8736 Perform dead store elimination (DSE) on trees. A dead store is a store into
8737 a memory location that is later overwritten by another store without
8738 any intervening loads. In this case the earlier store can be deleted. This
8739 flag is enabled by default at @option{-O} and higher.
8743 Perform loop header copying on trees. This is beneficial since it increases
8744 effectiveness of code motion optimizations. It also saves one jump. This flag
8745 is enabled by default at @option{-O} and higher. It is not enabled
8746 for @option{-Os}, since it usually increases code size.
8748 @item -ftree-loop-optimize
8749 @opindex ftree-loop-optimize
8750 Perform loop optimizations on trees. This flag is enabled by default
8751 at @option{-O} and higher.
8753 @item -ftree-loop-linear
8754 @itemx -floop-strip-mine
8756 @opindex ftree-loop-linear
8757 @opindex floop-strip-mine
8758 @opindex floop-block
8759 Perform loop nest optimizations. Same as
8760 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8761 to be configured with @option{--with-isl} to enable the Graphite loop
8762 transformation infrastructure.
8764 @item -fgraphite-identity
8765 @opindex fgraphite-identity
8766 Enable the identity transformation for graphite. For every SCoP we generate
8767 the polyhedral representation and transform it back to gimple. Using
8768 @option{-fgraphite-identity} we can check the costs or benefits of the
8769 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8770 are also performed by the code generator isl, like index splitting and
8771 dead code elimination in loops.
8773 @item -floop-nest-optimize
8774 @opindex floop-nest-optimize
8775 Enable the isl based loop nest optimizer. This is a generic loop nest
8776 optimizer based on the Pluto optimization algorithms. It calculates a loop
8777 structure optimized for data-locality and parallelism. This option
8780 @item -floop-parallelize-all
8781 @opindex floop-parallelize-all
8782 Use the Graphite data dependence analysis to identify loops that can
8783 be parallelized. Parallelize all the loops that can be analyzed to
8784 not contain loop carried dependences without checking that it is
8785 profitable to parallelize the loops.
8787 @item -ftree-coalesce-vars
8788 @opindex ftree-coalesce-vars
8789 While transforming the program out of the SSA representation, attempt to
8790 reduce copying by coalescing versions of different user-defined
8791 variables, instead of just compiler temporaries. This may severely
8792 limit the ability to debug an optimized program compiled with
8793 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8794 prevents SSA coalescing of user variables. This option is enabled by
8795 default if optimization is enabled, and it does very little otherwise.
8797 @item -ftree-loop-if-convert
8798 @opindex ftree-loop-if-convert
8799 Attempt to transform conditional jumps in the innermost loops to
8800 branch-less equivalents. The intent is to remove control-flow from
8801 the innermost loops in order to improve the ability of the
8802 vectorization pass to handle these loops. This is enabled by default
8803 if vectorization is enabled.
8805 @item -ftree-loop-distribution
8806 @opindex ftree-loop-distribution
8807 Perform loop distribution. This flag can improve cache performance on
8808 big loop bodies and allow further loop optimizations, like
8809 parallelization or vectorization, to take place. For example, the loop
8826 @item -ftree-loop-distribute-patterns
8827 @opindex ftree-loop-distribute-patterns
8828 Perform loop distribution of patterns that can be code generated with
8829 calls to a library. This flag is enabled by default at @option{-O3}.
8831 This pass distributes the initialization loops and generates a call to
8832 memset zero. For example, the loop
8848 and the initialization loop is transformed into a call to memset zero.
8850 @item -floop-interchange
8851 @opindex floop-interchange
8852 Perform loop interchange outside of graphite. This flag can improve cache
8853 performance on loop nest and allow further loop optimizations, like
8854 vectorization, to take place. For example, the loop
8856 for (int i = 0; i < N; i++)
8857 for (int j = 0; j < N; j++)
8858 for (int k = 0; k < N; k++)
8859 c[i][j] = c[i][j] + a[i][k]*b[k][j];
8863 for (int i = 0; i < N; i++)
8864 for (int k = 0; k < N; k++)
8865 for (int j = 0; j < N; j++)
8866 c[i][j] = c[i][j] + a[i][k]*b[k][j];
8869 @item -ftree-loop-im
8870 @opindex ftree-loop-im
8871 Perform loop invariant motion on trees. This pass moves only invariants that
8872 are hard to handle at RTL level (function calls, operations that expand to
8873 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8874 operands of conditions that are invariant out of the loop, so that we can use
8875 just trivial invariantness analysis in loop unswitching. The pass also includes
8878 @item -ftree-loop-ivcanon
8879 @opindex ftree-loop-ivcanon
8880 Create a canonical counter for number of iterations in loops for which
8881 determining number of iterations requires complicated analysis. Later
8882 optimizations then may determine the number easily. Useful especially
8883 in connection with unrolling.
8887 Perform induction variable optimizations (strength reduction, induction
8888 variable merging and induction variable elimination) on trees.
8890 @item -ftree-parallelize-loops=n
8891 @opindex ftree-parallelize-loops
8892 Parallelize loops, i.e., split their iteration space to run in n threads.
8893 This is only possible for loops whose iterations are independent
8894 and can be arbitrarily reordered. The optimization is only
8895 profitable on multiprocessor machines, for loops that are CPU-intensive,
8896 rather than constrained e.g.@: by memory bandwidth. This option
8897 implies @option{-pthread}, and thus is only supported on targets
8898 that have support for @option{-pthread}.
8902 Perform function-local points-to analysis on trees. This flag is
8903 enabled by default at @option{-O} and higher.
8907 Perform scalar replacement of aggregates. This pass replaces structure
8908 references with scalars to prevent committing structures to memory too
8909 early. This flag is enabled by default at @option{-O} and higher.
8911 @item -fstore-merging
8912 @opindex fstore-merging
8913 Perform merging of narrow stores to consecutive memory addresses. This pass
8914 merges contiguous stores of immediate values narrower than a word into fewer
8915 wider stores to reduce the number of instructions. This is enabled by default
8916 at @option{-O2} and higher as well as @option{-Os}.
8920 Perform temporary expression replacement during the SSA->normal phase. Single
8921 use/single def temporaries are replaced at their use location with their
8922 defining expression. This results in non-GIMPLE code, but gives the expanders
8923 much more complex trees to work on resulting in better RTL generation. This is
8924 enabled by default at @option{-O} and higher.
8928 Perform straight-line strength reduction on trees. This recognizes related
8929 expressions involving multiplications and replaces them by less expensive
8930 calculations when possible. This is enabled by default at @option{-O} and
8933 @item -ftree-vectorize
8934 @opindex ftree-vectorize
8935 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8936 and @option{-ftree-slp-vectorize} if not explicitly specified.
8938 @item -ftree-loop-vectorize
8939 @opindex ftree-loop-vectorize
8940 Perform loop vectorization on trees. This flag is enabled by default at
8941 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8943 @item -ftree-slp-vectorize
8944 @opindex ftree-slp-vectorize
8945 Perform basic block vectorization on trees. This flag is enabled by default at
8946 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8948 @item -fvect-cost-model=@var{model}
8949 @opindex fvect-cost-model
8950 Alter the cost model used for vectorization. The @var{model} argument
8951 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8952 With the @samp{unlimited} model the vectorized code-path is assumed
8953 to be profitable while with the @samp{dynamic} model a runtime check
8954 guards the vectorized code-path to enable it only for iteration
8955 counts that will likely execute faster than when executing the original
8956 scalar loop. The @samp{cheap} model disables vectorization of
8957 loops where doing so would be cost prohibitive for example due to
8958 required runtime checks for data dependence or alignment but otherwise
8959 is equal to the @samp{dynamic} model.
8960 The default cost model depends on other optimization flags and is
8961 either @samp{dynamic} or @samp{cheap}.
8963 @item -fsimd-cost-model=@var{model}
8964 @opindex fsimd-cost-model
8965 Alter the cost model used for vectorization of loops marked with the OpenMP
8966 simd directive. The @var{model} argument should be one of
8967 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8968 have the same meaning as described in @option{-fvect-cost-model} and by
8969 default a cost model defined with @option{-fvect-cost-model} is used.
8973 Perform Value Range Propagation on trees. This is similar to the
8974 constant propagation pass, but instead of values, ranges of values are
8975 propagated. This allows the optimizers to remove unnecessary range
8976 checks like array bound checks and null pointer checks. This is
8977 enabled by default at @option{-O2} and higher. Null pointer check
8978 elimination is only done if @option{-fdelete-null-pointer-checks} is
8982 @opindex fsplit-paths
8983 Split paths leading to loop backedges. This can improve dead code
8984 elimination and common subexpression elimination. This is enabled by
8985 default at @option{-O2} and above.
8987 @item -fsplit-ivs-in-unroller
8988 @opindex fsplit-ivs-in-unroller
8989 Enables expression of values of induction variables in later iterations
8990 of the unrolled loop using the value in the first iteration. This breaks
8991 long dependency chains, thus improving efficiency of the scheduling passes.
8993 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8994 same effect. However, that is not reliable in cases where the loop body
8995 is more complicated than a single basic block. It also does not work at all
8996 on some architectures due to restrictions in the CSE pass.
8998 This optimization is enabled by default.
9000 @item -fvariable-expansion-in-unroller
9001 @opindex fvariable-expansion-in-unroller
9002 With this option, the compiler creates multiple copies of some
9003 local variables when unrolling a loop, which can result in superior code.
9005 @item -fpartial-inlining
9006 @opindex fpartial-inlining
9007 Inline parts of functions. This option has any effect only
9008 when inlining itself is turned on by the @option{-finline-functions}
9009 or @option{-finline-small-functions} options.
9011 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9013 @item -fpredictive-commoning
9014 @opindex fpredictive-commoning
9015 Perform predictive commoning optimization, i.e., reusing computations
9016 (especially memory loads and stores) performed in previous
9017 iterations of loops.
9019 This option is enabled at level @option{-O3}.
9021 @item -fprefetch-loop-arrays
9022 @opindex fprefetch-loop-arrays
9023 If supported by the target machine, generate instructions to prefetch
9024 memory to improve the performance of loops that access large arrays.
9026 This option may generate better or worse code; results are highly
9027 dependent on the structure of loops within the source code.
9029 Disabled at level @option{-Os}.
9031 @item -fno-printf-return-value
9032 @opindex fno-printf-return-value
9033 Do not substitute constants for known return value of formatted output
9034 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
9035 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
9036 transformation allows GCC to optimize or even eliminate branches based
9037 on the known return value of these functions called with arguments that
9038 are either constant, or whose values are known to be in a range that
9039 makes determining the exact return value possible. For example, when
9040 @option{-fprintf-return-value} is in effect, both the branch and the
9041 body of the @code{if} statement (but not the call to @code{snprint})
9042 can be optimized away when @code{i} is a 32-bit or smaller integer
9043 because the return value is guaranteed to be at most 8.
9047 if (snprintf (buf, "%08x", i) >= sizeof buf)
9051 The @option{-fprintf-return-value} option relies on other optimizations
9052 and yields best results with @option{-O2} and above. It works in tandem
9053 with the @option{-Wformat-overflow} and @option{-Wformat-truncation}
9054 options. The @option{-fprintf-return-value} option is enabled by default.
9057 @itemx -fno-peephole2
9058 @opindex fno-peephole
9059 @opindex fno-peephole2
9060 Disable any machine-specific peephole optimizations. The difference
9061 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9062 are implemented in the compiler; some targets use one, some use the
9063 other, a few use both.
9065 @option{-fpeephole} is enabled by default.
9066 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9068 @item -fno-guess-branch-probability
9069 @opindex fno-guess-branch-probability
9070 Do not guess branch probabilities using heuristics.
9072 GCC uses heuristics to guess branch probabilities if they are
9073 not provided by profiling feedback (@option{-fprofile-arcs}). These
9074 heuristics are based on the control flow graph. If some branch probabilities
9075 are specified by @code{__builtin_expect}, then the heuristics are
9076 used to guess branch probabilities for the rest of the control flow graph,
9077 taking the @code{__builtin_expect} info into account. The interactions
9078 between the heuristics and @code{__builtin_expect} can be complex, and in
9079 some cases, it may be useful to disable the heuristics so that the effects
9080 of @code{__builtin_expect} are easier to understand.
9082 The default is @option{-fguess-branch-probability} at levels
9083 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9085 @item -freorder-blocks
9086 @opindex freorder-blocks
9087 Reorder basic blocks in the compiled function in order to reduce number of
9088 taken branches and improve code locality.
9090 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9092 @item -freorder-blocks-algorithm=@var{algorithm}
9093 @opindex freorder-blocks-algorithm
9094 Use the specified algorithm for basic block reordering. The
9095 @var{algorithm} argument can be @samp{simple}, which does not increase
9096 code size (except sometimes due to secondary effects like alignment),
9097 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9098 put all often executed code together, minimizing the number of branches
9099 executed by making extra copies of code.
9101 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9102 @samp{stc} at levels @option{-O2}, @option{-O3}.
9104 @item -freorder-blocks-and-partition
9105 @opindex freorder-blocks-and-partition
9106 In addition to reordering basic blocks in the compiled function, in order
9107 to reduce number of taken branches, partitions hot and cold basic blocks
9108 into separate sections of the assembly and @file{.o} files, to improve
9109 paging and cache locality performance.
9111 This optimization is automatically turned off in the presence of
9112 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
9113 section attribute and on any architecture that does not support named
9114 sections. When @option{-fsplit-stack} is used this option is not
9115 enabled by default (to avoid linker errors), but may be enabled
9116 explicitly (if using a working linker).
9118 Enabled for x86 at levels @option{-O2}, @option{-O3}, @option{-Os}.
9120 @item -freorder-functions
9121 @opindex freorder-functions
9122 Reorder functions in the object file in order to
9123 improve code locality. This is implemented by using special
9124 subsections @code{.text.hot} for most frequently executed functions and
9125 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9126 the linker so object file format must support named sections and linker must
9127 place them in a reasonable way.
9129 Also profile feedback must be available to make this option effective. See
9130 @option{-fprofile-arcs} for details.
9132 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9134 @item -fstrict-aliasing
9135 @opindex fstrict-aliasing
9136 Allow the compiler to assume the strictest aliasing rules applicable to
9137 the language being compiled. For C (and C++), this activates
9138 optimizations based on the type of expressions. In particular, an
9139 object of one type is assumed never to reside at the same address as an
9140 object of a different type, unless the types are almost the same. For
9141 example, an @code{unsigned int} can alias an @code{int}, but not a
9142 @code{void*} or a @code{double}. A character type may alias any other
9145 @anchor{Type-punning}Pay special attention to code like this:
9158 The practice of reading from a different union member than the one most
9159 recently written to (called ``type-punning'') is common. Even with
9160 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9161 is accessed through the union type. So, the code above works as
9162 expected. @xref{Structures unions enumerations and bit-fields
9163 implementation}. However, this code might not:
9174 Similarly, access by taking the address, casting the resulting pointer
9175 and dereferencing the result has undefined behavior, even if the cast
9176 uses a union type, e.g.:
9180 return ((union a_union *) &d)->i;
9184 The @option{-fstrict-aliasing} option is enabled at levels
9185 @option{-O2}, @option{-O3}, @option{-Os}.
9187 @item -falign-functions
9188 @itemx -falign-functions=@var{n}
9189 @opindex falign-functions
9190 Align the start of functions to the next power-of-two greater than
9191 @var{n}, skipping up to @var{n} bytes. For instance,
9192 @option{-falign-functions=32} aligns functions to the next 32-byte
9193 boundary, but @option{-falign-functions=24} aligns to the next
9194 32-byte boundary only if this can be done by skipping 23 bytes or less.
9196 @option{-fno-align-functions} and @option{-falign-functions=1} are
9197 equivalent and mean that functions are not aligned.
9199 Some assemblers only support this flag when @var{n} is a power of two;
9200 in that case, it is rounded up.
9202 If @var{n} is not specified or is zero, use a machine-dependent default.
9203 The maximum allowed @var{n} option value is 65536.
9205 Enabled at levels @option{-O2}, @option{-O3}.
9207 @item -flimit-function-alignment
9208 If this option is enabled, the compiler tries to avoid unnecessarily
9209 overaligning functions. It attempts to instruct the assembler to align
9210 by the amount specified by @option{-falign-functions}, but not to
9211 skip more bytes than the size of the function.
9213 @item -falign-labels
9214 @itemx -falign-labels=@var{n}
9215 @opindex falign-labels
9216 Align all branch targets to a power-of-two boundary, skipping up to
9217 @var{n} bytes like @option{-falign-functions}. This option can easily
9218 make code slower, because it must insert dummy operations for when the
9219 branch target is reached in the usual flow of the code.
9221 @option{-fno-align-labels} and @option{-falign-labels=1} are
9222 equivalent and mean that labels are not aligned.
9224 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9225 are greater than this value, then their values are used instead.
9227 If @var{n} is not specified or is zero, use a machine-dependent default
9228 which is very likely to be @samp{1}, meaning no alignment.
9229 The maximum allowed @var{n} option value is 65536.
9231 Enabled at levels @option{-O2}, @option{-O3}.
9234 @itemx -falign-loops=@var{n}
9235 @opindex falign-loops
9236 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9237 like @option{-falign-functions}. If the loops are
9238 executed many times, this makes up for any execution of the dummy
9241 @option{-fno-align-loops} and @option{-falign-loops=1} are
9242 equivalent and mean that loops are not aligned.
9243 The maximum allowed @var{n} option value is 65536.
9245 If @var{n} is not specified or is zero, use a machine-dependent default.
9247 Enabled at levels @option{-O2}, @option{-O3}.
9250 @itemx -falign-jumps=@var{n}
9251 @opindex falign-jumps
9252 Align branch targets to a power-of-two boundary, for branch targets
9253 where the targets can only be reached by jumping, skipping up to @var{n}
9254 bytes like @option{-falign-functions}. In this case, no dummy operations
9257 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9258 equivalent and mean that loops are not aligned.
9260 If @var{n} is not specified or is zero, use a machine-dependent default.
9261 The maximum allowed @var{n} option value is 65536.
9263 Enabled at levels @option{-O2}, @option{-O3}.
9265 @item -funit-at-a-time
9266 @opindex funit-at-a-time
9267 This option is left for compatibility reasons. @option{-funit-at-a-time}
9268 has no effect, while @option{-fno-unit-at-a-time} implies
9269 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9273 @item -fno-toplevel-reorder
9274 @opindex fno-toplevel-reorder
9275 Do not reorder top-level functions, variables, and @code{asm}
9276 statements. Output them in the same order that they appear in the
9277 input file. When this option is used, unreferenced static variables
9278 are not removed. This option is intended to support existing code
9279 that relies on a particular ordering. For new code, it is better to
9280 use attributes when possible.
9282 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9283 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9288 Constructs webs as commonly used for register allocation purposes and assign
9289 each web individual pseudo register. This allows the register allocation pass
9290 to operate on pseudos directly, but also strengthens several other optimization
9291 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9292 however, make debugging impossible, since variables no longer stay in a
9295 Enabled by default with @option{-funroll-loops}.
9297 @item -fwhole-program
9298 @opindex fwhole-program
9299 Assume that the current compilation unit represents the whole program being
9300 compiled. All public functions and variables with the exception of @code{main}
9301 and those merged by attribute @code{externally_visible} become static functions
9302 and in effect are optimized more aggressively by interprocedural optimizers.
9304 This option should not be used in combination with @option{-flto}.
9305 Instead relying on a linker plugin should provide safer and more precise
9308 @item -flto[=@var{n}]
9310 This option runs the standard link-time optimizer. When invoked
9311 with source code, it generates GIMPLE (one of GCC's internal
9312 representations) and writes it to special ELF sections in the object
9313 file. When the object files are linked together, all the function
9314 bodies are read from these ELF sections and instantiated as if they
9315 had been part of the same translation unit.
9317 To use the link-time optimizer, @option{-flto} and optimization
9318 options should be specified at compile time and during the final link.
9319 It is recommended that you compile all the files participating in the
9320 same link with the same options and also specify those options at
9325 gcc -c -O2 -flto foo.c
9326 gcc -c -O2 -flto bar.c
9327 gcc -o myprog -flto -O2 foo.o bar.o
9330 The first two invocations to GCC save a bytecode representation
9331 of GIMPLE into special ELF sections inside @file{foo.o} and
9332 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9333 @file{foo.o} and @file{bar.o}, merges the two files into a single
9334 internal image, and compiles the result as usual. Since both
9335 @file{foo.o} and @file{bar.o} are merged into a single image, this
9336 causes all the interprocedural analyses and optimizations in GCC to
9337 work across the two files as if they were a single one. This means,
9338 for example, that the inliner is able to inline functions in
9339 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9341 Another (simpler) way to enable link-time optimization is:
9344 gcc -o myprog -flto -O2 foo.c bar.c
9347 The above generates bytecode for @file{foo.c} and @file{bar.c},
9348 merges them together into a single GIMPLE representation and optimizes
9349 them as usual to produce @file{myprog}.
9351 The only important thing to keep in mind is that to enable link-time
9352 optimizations you need to use the GCC driver to perform the link step.
9353 GCC then automatically performs link-time optimization if any of the
9354 objects involved were compiled with the @option{-flto} command-line option.
9356 should specify the optimization options to be used for link-time
9357 optimization though GCC tries to be clever at guessing an
9358 optimization level to use from the options used at compile time
9359 if you fail to specify one at link time. You can always override
9360 the automatic decision to do link-time optimization
9361 by passing @option{-fno-lto} to the link command.
9363 To make whole program optimization effective, it is necessary to make
9364 certain whole program assumptions. The compiler needs to know
9365 what functions and variables can be accessed by libraries and runtime
9366 outside of the link-time optimized unit. When supported by the linker,
9367 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9368 to the compiler about used and externally visible symbols. When
9369 the linker plugin is not available, @option{-fwhole-program} should be
9370 used to allow the compiler to make these assumptions, which leads
9371 to more aggressive optimization decisions.
9373 When @option{-fuse-linker-plugin} is not enabled, when a file is
9374 compiled with @option{-flto}, the generated object file is larger than
9375 a regular object file because it contains GIMPLE bytecodes and the usual
9376 final code (see @option{-ffat-lto-objects}. This means that
9377 object files with LTO information can be linked as normal object
9378 files; if @option{-fno-lto} is passed to the linker, no
9379 interprocedural optimizations are applied. Note that when
9380 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
9381 but you cannot perform a regular, non-LTO link on them.
9383 Additionally, the optimization flags used to compile individual files
9384 are not necessarily related to those used at link time. For instance,
9387 gcc -c -O0 -ffat-lto-objects -flto foo.c
9388 gcc -c -O0 -ffat-lto-objects -flto bar.c
9389 gcc -o myprog -O3 foo.o bar.o
9392 This produces individual object files with unoptimized assembler
9393 code, but the resulting binary @file{myprog} is optimized at
9394 @option{-O3}. If, instead, the final binary is generated with
9395 @option{-fno-lto}, then @file{myprog} is not optimized.
9397 When producing the final binary, GCC only
9398 applies link-time optimizations to those files that contain bytecode.
9399 Therefore, you can mix and match object files and libraries with
9400 GIMPLE bytecodes and final object code. GCC automatically selects
9401 which files to optimize in LTO mode and which files to link without
9404 There are some code generation flags preserved by GCC when
9405 generating bytecodes, as they need to be used during the final link
9406 stage. Generally options specified at link time override those
9407 specified at compile time.
9409 If you do not specify an optimization level option @option{-O} at
9410 link time, then GCC uses the highest optimization level
9411 used when compiling the object files.
9413 Currently, the following options and their settings are taken from
9414 the first object file that explicitly specifies them:
9415 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9416 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9417 and all the @option{-m} target flags.
9419 Certain ABI-changing flags are required to match in all compilation units,
9420 and trying to override this at link time with a conflicting value
9421 is ignored. This includes options such as @option{-freg-struct-return}
9422 and @option{-fpcc-struct-return}.
9424 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9425 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9426 are passed through to the link stage and merged conservatively for
9427 conflicting translation units. Specifically
9428 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9429 precedence; and for example @option{-ffp-contract=off} takes precedence
9430 over @option{-ffp-contract=fast}. You can override them at link time.
9432 If LTO encounters objects with C linkage declared with incompatible
9433 types in separate translation units to be linked together (undefined
9434 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9435 issued. The behavior is still undefined at run time. Similar
9436 diagnostics may be raised for other languages.
9438 Another feature of LTO is that it is possible to apply interprocedural
9439 optimizations on files written in different languages:
9444 gfortran -c -flto baz.f90
9445 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9448 Notice that the final link is done with @command{g++} to get the C++
9449 runtime libraries and @option{-lgfortran} is added to get the Fortran
9450 runtime libraries. In general, when mixing languages in LTO mode, you
9451 should use the same link command options as when mixing languages in a
9452 regular (non-LTO) compilation.
9454 If object files containing GIMPLE bytecode are stored in a library archive, say
9455 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9456 are using a linker with plugin support. To create static libraries suitable
9457 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9458 and @command{ranlib};
9459 to show the symbols of object files with GIMPLE bytecode, use
9460 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9461 and @command{nm} have been compiled with plugin support. At link time, use the the
9462 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9463 the LTO optimization process:
9466 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9469 With the linker plugin enabled, the linker extracts the needed
9470 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9471 to make them part of the aggregated GIMPLE image to be optimized.
9473 If you are not using a linker with plugin support and/or do not
9474 enable the linker plugin, then the objects inside @file{libfoo.a}
9475 are extracted and linked as usual, but they do not participate
9476 in the LTO optimization process. In order to make a static library suitable
9477 for both LTO optimization and usual linkage, compile its object files with
9478 @option{-flto} @option{-ffat-lto-objects}.
9480 Link-time optimizations do not require the presence of the whole program to
9481 operate. If the program does not require any symbols to be exported, it is
9482 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9483 the interprocedural optimizers to use more aggressive assumptions which may
9484 lead to improved optimization opportunities.
9485 Use of @option{-fwhole-program} is not needed when linker plugin is
9486 active (see @option{-fuse-linker-plugin}).
9488 The current implementation of LTO makes no
9489 attempt to generate bytecode that is portable between different
9490 types of hosts. The bytecode files are versioned and there is a
9491 strict version check, so bytecode files generated in one version of
9492 GCC do not work with an older or newer version of GCC.
9494 Link-time optimization does not work well with generation of debugging
9495 information on systems other than those using a combination of ELF and
9498 If you specify the optional @var{n}, the optimization and code
9499 generation done at link time is executed in parallel using @var{n}
9500 parallel jobs by utilizing an installed @command{make} program. The
9501 environment variable @env{MAKE} may be used to override the program
9502 used. The default value for @var{n} is 1.
9504 You can also specify @option{-flto=jobserver} to use GNU make's
9505 job server mode to determine the number of parallel jobs. This
9506 is useful when the Makefile calling GCC is already executing in parallel.
9507 You must prepend a @samp{+} to the command recipe in the parent Makefile
9508 for this to work. This option likely only works if @env{MAKE} is
9511 @item -flto-partition=@var{alg}
9512 @opindex flto-partition
9513 Specify the partitioning algorithm used by the link-time optimizer.
9514 The value is either @samp{1to1} to specify a partitioning mirroring
9515 the original source files or @samp{balanced} to specify partitioning
9516 into equally sized chunks (whenever possible) or @samp{max} to create
9517 new partition for every symbol where possible. Specifying @samp{none}
9518 as an algorithm disables partitioning and streaming completely.
9519 The default value is @samp{balanced}. While @samp{1to1} can be used
9520 as an workaround for various code ordering issues, the @samp{max}
9521 partitioning is intended for internal testing only.
9522 The value @samp{one} specifies that exactly one partition should be
9523 used while the value @samp{none} bypasses partitioning and executes
9524 the link-time optimization step directly from the WPA phase.
9526 @item -flto-odr-type-merging
9527 @opindex flto-odr-type-merging
9528 Enable streaming of mangled types names of C++ types and their unification
9529 at link time. This increases size of LTO object files, but enables
9530 diagnostics about One Definition Rule violations.
9532 @item -flto-compression-level=@var{n}
9533 @opindex flto-compression-level
9534 This option specifies the level of compression used for intermediate
9535 language written to LTO object files, and is only meaningful in
9536 conjunction with LTO mode (@option{-flto}). Valid
9537 values are 0 (no compression) to 9 (maximum compression). Values
9538 outside this range are clamped to either 0 or 9. If the option is not
9539 given, a default balanced compression setting is used.
9541 @item -fuse-linker-plugin
9542 @opindex fuse-linker-plugin
9543 Enables the use of a linker plugin during link-time optimization. This
9544 option relies on plugin support in the linker, which is available in gold
9545 or in GNU ld 2.21 or newer.
9547 This option enables the extraction of object files with GIMPLE bytecode out
9548 of library archives. This improves the quality of optimization by exposing
9549 more code to the link-time optimizer. This information specifies what
9550 symbols can be accessed externally (by non-LTO object or during dynamic
9551 linking). Resulting code quality improvements on binaries (and shared
9552 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9553 See @option{-flto} for a description of the effect of this flag and how to
9556 This option is enabled by default when LTO support in GCC is enabled
9557 and GCC was configured for use with
9558 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9560 @item -ffat-lto-objects
9561 @opindex ffat-lto-objects
9562 Fat LTO objects are object files that contain both the intermediate language
9563 and the object code. This makes them usable for both LTO linking and normal
9564 linking. This option is effective only when compiling with @option{-flto}
9565 and is ignored at link time.
9567 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9568 requires the complete toolchain to be aware of LTO. It requires a linker with
9569 linker plugin support for basic functionality. Additionally,
9570 @command{nm}, @command{ar} and @command{ranlib}
9571 need to support linker plugins to allow a full-featured build environment
9572 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9573 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9574 to these tools. With non fat LTO makefiles need to be modified to use them.
9576 Note that modern binutils provide plugin auto-load mechanism.
9577 Installing the linker plugin into @file{$libdir/bfd-plugins} has the same
9578 effect as usage of the command wrappers (@command{gcc-ar}, @command{gcc-nm} and
9579 @command{gcc-ranlib}).
9581 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9584 @item -fcompare-elim
9585 @opindex fcompare-elim
9586 After register allocation and post-register allocation instruction splitting,
9587 identify arithmetic instructions that compute processor flags similar to a
9588 comparison operation based on that arithmetic. If possible, eliminate the
9589 explicit comparison operation.
9591 This pass only applies to certain targets that cannot explicitly represent
9592 the comparison operation before register allocation is complete.
9594 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9596 @item -fcprop-registers
9597 @opindex fcprop-registers
9598 After register allocation and post-register allocation instruction splitting,
9599 perform a copy-propagation pass to try to reduce scheduling dependencies
9600 and occasionally eliminate the copy.
9602 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9604 @item -fprofile-correction
9605 @opindex fprofile-correction
9606 Profiles collected using an instrumented binary for multi-threaded programs may
9607 be inconsistent due to missed counter updates. When this option is specified,
9608 GCC uses heuristics to correct or smooth out such inconsistencies. By
9609 default, GCC emits an error message when an inconsistent profile is detected.
9612 @itemx -fprofile-use=@var{path}
9613 @opindex fprofile-use
9614 Enable profile feedback-directed optimizations,
9615 and the following optimizations
9616 which are generally profitable only with profile feedback available:
9617 @option{-fbranch-probabilities}, @option{-fvpt},
9618 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9619 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9621 Before you can use this option, you must first generate profiling information.
9622 @xref{Instrumentation Options}, for information about the
9623 @option{-fprofile-generate} option.
9625 By default, GCC emits an error message if the feedback profiles do not
9626 match the source code. This error can be turned into a warning by using
9627 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9630 If @var{path} is specified, GCC looks at the @var{path} to find
9631 the profile feedback data files. See @option{-fprofile-dir}.
9633 @item -fauto-profile
9634 @itemx -fauto-profile=@var{path}
9635 @opindex fauto-profile
9636 Enable sampling-based feedback-directed optimizations,
9637 and the following optimizations
9638 which are generally profitable only with profile feedback available:
9639 @option{-fbranch-probabilities}, @option{-fvpt},
9640 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9641 @option{-ftree-vectorize},
9642 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9643 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9644 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9646 @var{path} is the name of a file containing AutoFDO profile information.
9647 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9649 Producing an AutoFDO profile data file requires running your program
9650 with the @command{perf} utility on a supported GNU/Linux target system.
9651 For more information, see @uref{https://perf.wiki.kernel.org/}.
9655 perf record -e br_inst_retired:near_taken -b -o perf.data \
9659 Then use the @command{create_gcov} tool to convert the raw profile data
9660 to a format that can be used by GCC.@ You must also supply the
9661 unstripped binary for your program to this tool.
9662 See @uref{https://github.com/google/autofdo}.
9666 create_gcov --binary=your_program.unstripped --profile=perf.data \
9671 The following options control compiler behavior regarding floating-point
9672 arithmetic. These options trade off between speed and
9673 correctness. All must be specifically enabled.
9677 @opindex ffloat-store
9678 Do not store floating-point variables in registers, and inhibit other
9679 options that might change whether a floating-point value is taken from a
9682 @cindex floating-point precision
9683 This option prevents undesirable excess precision on machines such as
9684 the 68000 where the floating registers (of the 68881) keep more
9685 precision than a @code{double} is supposed to have. Similarly for the
9686 x86 architecture. For most programs, the excess precision does only
9687 good, but a few programs rely on the precise definition of IEEE floating
9688 point. Use @option{-ffloat-store} for such programs, after modifying
9689 them to store all pertinent intermediate computations into variables.
9691 @item -fexcess-precision=@var{style}
9692 @opindex fexcess-precision
9693 This option allows further control over excess precision on machines
9694 where floating-point operations occur in a format with more precision or
9695 range than the IEEE standard and interchange floating-point types. By
9696 default, @option{-fexcess-precision=fast} is in effect; this means that
9697 operations may be carried out in a wider precision than the types specified
9698 in the source if that would result in faster code, and it is unpredictable
9699 when rounding to the types specified in the source code takes place.
9700 When compiling C, if @option{-fexcess-precision=standard} is specified then
9701 excess precision follows the rules specified in ISO C99; in particular,
9702 both casts and assignments cause values to be rounded to their
9703 semantic types (whereas @option{-ffloat-store} only affects
9704 assignments). This option is enabled by default for C if a strict
9705 conformance option such as @option{-std=c99} is used.
9706 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
9707 regardless of whether a strict conformance option is used.
9710 @option{-fexcess-precision=standard} is not implemented for languages
9711 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
9712 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9713 semantics apply without excess precision, and in the latter, rounding
9718 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9719 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9720 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
9721 @option{-fexcess-precision=fast}.
9723 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9725 This option is not turned on by any @option{-O} option besides
9726 @option{-Ofast} since it can result in incorrect output for programs
9727 that depend on an exact implementation of IEEE or ISO rules/specifications
9728 for math functions. It may, however, yield faster code for programs
9729 that do not require the guarantees of these specifications.
9731 @item -fno-math-errno
9732 @opindex fno-math-errno
9733 Do not set @code{errno} after calling math functions that are executed
9734 with a single instruction, e.g., @code{sqrt}. A program that relies on
9735 IEEE exceptions for math error handling may want to use this flag
9736 for speed while maintaining IEEE arithmetic compatibility.
9738 This option is not turned on by any @option{-O} option since
9739 it can result in incorrect output for programs that depend on
9740 an exact implementation of IEEE or ISO rules/specifications for
9741 math functions. It may, however, yield faster code for programs
9742 that do not require the guarantees of these specifications.
9744 The default is @option{-fmath-errno}.
9746 On Darwin systems, the math library never sets @code{errno}. There is
9747 therefore no reason for the compiler to consider the possibility that
9748 it might, and @option{-fno-math-errno} is the default.
9750 @item -funsafe-math-optimizations
9751 @opindex funsafe-math-optimizations
9753 Allow optimizations for floating-point arithmetic that (a) assume
9754 that arguments and results are valid and (b) may violate IEEE or
9755 ANSI standards. When used at link time, it may include libraries
9756 or startup files that change the default FPU control word or other
9757 similar optimizations.
9759 This option is not turned on by any @option{-O} option since
9760 it can result in incorrect output for programs that depend on
9761 an exact implementation of IEEE or ISO rules/specifications for
9762 math functions. It may, however, yield faster code for programs
9763 that do not require the guarantees of these specifications.
9764 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9765 @option{-fassociative-math} and @option{-freciprocal-math}.
9767 The default is @option{-fno-unsafe-math-optimizations}.
9769 @item -fassociative-math
9770 @opindex fassociative-math
9772 Allow re-association of operands in series of floating-point operations.
9773 This violates the ISO C and C++ language standard by possibly changing
9774 computation result. NOTE: re-ordering may change the sign of zero as
9775 well as ignore NaNs and inhibit or create underflow or overflow (and
9776 thus cannot be used on code that relies on rounding behavior like
9777 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9778 and thus may not be used when ordered comparisons are required.
9779 This option requires that both @option{-fno-signed-zeros} and
9780 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9781 much sense with @option{-frounding-math}. For Fortran the option
9782 is automatically enabled when both @option{-fno-signed-zeros} and
9783 @option{-fno-trapping-math} are in effect.
9785 The default is @option{-fno-associative-math}.
9787 @item -freciprocal-math
9788 @opindex freciprocal-math
9790 Allow the reciprocal of a value to be used instead of dividing by
9791 the value if this enables optimizations. For example @code{x / y}
9792 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9793 is subject to common subexpression elimination. Note that this loses
9794 precision and increases the number of flops operating on the value.
9796 The default is @option{-fno-reciprocal-math}.
9798 @item -ffinite-math-only
9799 @opindex ffinite-math-only
9800 Allow optimizations for floating-point arithmetic that assume
9801 that arguments and results are not NaNs or +-Infs.
9803 This option is not turned on by any @option{-O} option since
9804 it can result in incorrect output for programs that depend on
9805 an exact implementation of IEEE or ISO rules/specifications for
9806 math functions. It may, however, yield faster code for programs
9807 that do not require the guarantees of these specifications.
9809 The default is @option{-fno-finite-math-only}.
9811 @item -fno-signed-zeros
9812 @opindex fno-signed-zeros
9813 Allow optimizations for floating-point arithmetic that ignore the
9814 signedness of zero. IEEE arithmetic specifies the behavior of
9815 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9816 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9817 This option implies that the sign of a zero result isn't significant.
9819 The default is @option{-fsigned-zeros}.
9821 @item -fno-trapping-math
9822 @opindex fno-trapping-math
9823 Compile code assuming that floating-point operations cannot generate
9824 user-visible traps. These traps include division by zero, overflow,
9825 underflow, inexact result and invalid operation. This option requires
9826 that @option{-fno-signaling-nans} be in effect. Setting this option may
9827 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9829 This option should never be turned on by any @option{-O} option since
9830 it can result in incorrect output for programs that depend on
9831 an exact implementation of IEEE or ISO rules/specifications for
9834 The default is @option{-ftrapping-math}.
9836 @item -frounding-math
9837 @opindex frounding-math
9838 Disable transformations and optimizations that assume default floating-point
9839 rounding behavior. This is round-to-zero for all floating point
9840 to integer conversions, and round-to-nearest for all other arithmetic
9841 truncations. This option should be specified for programs that change
9842 the FP rounding mode dynamically, or that may be executed with a
9843 non-default rounding mode. This option disables constant folding of
9844 floating-point expressions at compile time (which may be affected by
9845 rounding mode) and arithmetic transformations that are unsafe in the
9846 presence of sign-dependent rounding modes.
9848 The default is @option{-fno-rounding-math}.
9850 This option is experimental and does not currently guarantee to
9851 disable all GCC optimizations that are affected by rounding mode.
9852 Future versions of GCC may provide finer control of this setting
9853 using C99's @code{FENV_ACCESS} pragma. This command-line option
9854 will be used to specify the default state for @code{FENV_ACCESS}.
9856 @item -fsignaling-nans
9857 @opindex fsignaling-nans
9858 Compile code assuming that IEEE signaling NaNs may generate user-visible
9859 traps during floating-point operations. Setting this option disables
9860 optimizations that may change the number of exceptions visible with
9861 signaling NaNs. This option implies @option{-ftrapping-math}.
9863 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9866 The default is @option{-fno-signaling-nans}.
9868 This option is experimental and does not currently guarantee to
9869 disable all GCC optimizations that affect signaling NaN behavior.
9871 @item -fno-fp-int-builtin-inexact
9872 @opindex fno-fp-int-builtin-inexact
9873 Do not allow the built-in functions @code{ceil}, @code{floor},
9874 @code{round} and @code{trunc}, and their @code{float} and @code{long
9875 double} variants, to generate code that raises the ``inexact''
9876 floating-point exception for noninteger arguments. ISO C99 and C11
9877 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9878 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9881 The default is @option{-ffp-int-builtin-inexact}, allowing the
9882 exception to be raised. This option does nothing unless
9883 @option{-ftrapping-math} is in effect.
9885 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9886 generate a call to a library function then the ``inexact'' exception
9887 may be raised if the library implementation does not follow TS 18661.
9889 @item -fsingle-precision-constant
9890 @opindex fsingle-precision-constant
9891 Treat floating-point constants as single precision instead of
9892 implicitly converting them to double-precision constants.
9894 @item -fcx-limited-range
9895 @opindex fcx-limited-range
9896 When enabled, this option states that a range reduction step is not
9897 needed when performing complex division. Also, there is no checking
9898 whether the result of a complex multiplication or division is @code{NaN
9899 + I*NaN}, with an attempt to rescue the situation in that case. The
9900 default is @option{-fno-cx-limited-range}, but is enabled by
9901 @option{-ffast-math}.
9903 This option controls the default setting of the ISO C99
9904 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9907 @item -fcx-fortran-rules
9908 @opindex fcx-fortran-rules
9909 Complex multiplication and division follow Fortran rules. Range
9910 reduction is done as part of complex division, but there is no checking
9911 whether the result of a complex multiplication or division is @code{NaN
9912 + I*NaN}, with an attempt to rescue the situation in that case.
9914 The default is @option{-fno-cx-fortran-rules}.
9918 The following options control optimizations that may improve
9919 performance, but are not enabled by any @option{-O} options. This
9920 section includes experimental options that may produce broken code.
9923 @item -fbranch-probabilities
9924 @opindex fbranch-probabilities
9925 After running a program compiled with @option{-fprofile-arcs}
9926 (@pxref{Instrumentation Options}),
9927 you can compile it a second time using
9928 @option{-fbranch-probabilities}, to improve optimizations based on
9929 the number of times each branch was taken. When a program
9930 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9931 counts to a file called @file{@var{sourcename}.gcda} for each source
9932 file. The information in this data file is very dependent on the
9933 structure of the generated code, so you must use the same source code
9934 and the same optimization options for both compilations.
9936 With @option{-fbranch-probabilities}, GCC puts a
9937 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9938 These can be used to improve optimization. Currently, they are only
9939 used in one place: in @file{reorg.c}, instead of guessing which path a
9940 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9941 exactly determine which path is taken more often.
9943 @item -fprofile-values
9944 @opindex fprofile-values
9945 If combined with @option{-fprofile-arcs}, it adds code so that some
9946 data about values of expressions in the program is gathered.
9948 With @option{-fbranch-probabilities}, it reads back the data gathered
9949 from profiling values of expressions for usage in optimizations.
9951 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9953 @item -fprofile-reorder-functions
9954 @opindex fprofile-reorder-functions
9955 Function reordering based on profile instrumentation collects
9956 first time of execution of a function and orders these functions
9959 Enabled with @option{-fprofile-use}.
9963 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9964 to add code to gather information about values of expressions.
9966 With @option{-fbranch-probabilities}, it reads back the data gathered
9967 and actually performs the optimizations based on them.
9968 Currently the optimizations include specialization of division operations
9969 using the knowledge about the value of the denominator.
9971 @item -frename-registers
9972 @opindex frename-registers
9973 Attempt to avoid false dependencies in scheduled code by making use
9974 of registers left over after register allocation. This optimization
9975 most benefits processors with lots of registers. Depending on the
9976 debug information format adopted by the target, however, it can
9977 make debugging impossible, since variables no longer stay in
9978 a ``home register''.
9980 Enabled by default with @option{-funroll-loops}.
9982 @item -fschedule-fusion
9983 @opindex fschedule-fusion
9984 Performs a target dependent pass over the instruction stream to schedule
9985 instructions of same type together because target machine can execute them
9986 more efficiently if they are adjacent to each other in the instruction flow.
9988 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9992 Perform tail duplication to enlarge superblock size. This transformation
9993 simplifies the control flow of the function allowing other optimizations to do
9996 Enabled with @option{-fprofile-use}.
9998 @item -funroll-loops
9999 @opindex funroll-loops
10000 Unroll loops whose number of iterations can be determined at compile time or
10001 upon entry to the loop. @option{-funroll-loops} implies
10002 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10003 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10004 a small constant number of iterations). This option makes code larger, and may
10005 or may not make it run faster.
10007 Enabled with @option{-fprofile-use}.
10009 @item -funroll-all-loops
10010 @opindex funroll-all-loops
10011 Unroll all loops, even if their number of iterations is uncertain when
10012 the loop is entered. This usually makes programs run more slowly.
10013 @option{-funroll-all-loops} implies the same options as
10014 @option{-funroll-loops}.
10017 @opindex fpeel-loops
10018 Peels loops for which there is enough information that they do not
10019 roll much (from profile feedback or static analysis). It also turns on
10020 complete loop peeling (i.e.@: complete removal of loops with small constant
10021 number of iterations).
10023 Enabled with @option{-O3} and/or @option{-fprofile-use}.
10025 @item -fmove-loop-invariants
10026 @opindex fmove-loop-invariants
10027 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10028 at level @option{-O1}
10030 @item -fsplit-loops
10031 @opindex fsplit-loops
10032 Split a loop into two if it contains a condition that's always true
10033 for one side of the iteration space and false for the other.
10035 @item -funswitch-loops
10036 @opindex funswitch-loops
10037 Move branches with loop invariant conditions out of the loop, with duplicates
10038 of the loop on both branches (modified according to result of the condition).
10040 @item -floop-unroll-and-jam
10041 @opindex floop-unroll-and-jam
10042 Apply unroll and jam transformations on feasible loops. In a loop
10043 nest this unrolls the outer loop by some factor and fuses the resulting
10044 multiple inner loops.
10046 @item -ffunction-sections
10047 @itemx -fdata-sections
10048 @opindex ffunction-sections
10049 @opindex fdata-sections
10050 Place each function or data item into its own section in the output
10051 file if the target supports arbitrary sections. The name of the
10052 function or the name of the data item determines the section's name
10053 in the output file.
10055 Use these options on systems where the linker can perform optimizations to
10056 improve locality of reference in the instruction space. Most systems using the
10057 ELF object format have linkers with such optimizations. On AIX, the linker
10058 rearranges sections (CSECTs) based on the call graph. The performance impact
10061 Together with a linker garbage collection (linker @option{--gc-sections}
10062 option) these options may lead to smaller statically-linked executables (after
10065 On ELF/DWARF systems these options do not degenerate the quality of the debug
10066 information. There could be issues with other object files/debug info formats.
10068 Only use these options when there are significant benefits from doing so. When
10069 you specify these options, the assembler and linker create larger object and
10070 executable files and are also slower. These options affect code generation.
10071 They prevent optimizations by the compiler and assembler using relative
10072 locations inside a translation unit since the locations are unknown until
10073 link time. An example of such an optimization is relaxing calls to short call
10076 @item -fbranch-target-load-optimize
10077 @opindex fbranch-target-load-optimize
10078 Perform branch target register load optimization before prologue / epilogue
10080 The use of target registers can typically be exposed only during reload,
10081 thus hoisting loads out of loops and doing inter-block scheduling needs
10082 a separate optimization pass.
10084 @item -fbranch-target-load-optimize2
10085 @opindex fbranch-target-load-optimize2
10086 Perform branch target register load optimization after prologue / epilogue
10089 @item -fbtr-bb-exclusive
10090 @opindex fbtr-bb-exclusive
10091 When performing branch target register load optimization, don't reuse
10092 branch target registers within any basic block.
10095 @opindex fstdarg-opt
10096 Optimize the prologue of variadic argument functions with respect to usage of
10099 @item -fsection-anchors
10100 @opindex fsection-anchors
10101 Try to reduce the number of symbolic address calculations by using
10102 shared ``anchor'' symbols to address nearby objects. This transformation
10103 can help to reduce the number of GOT entries and GOT accesses on some
10106 For example, the implementation of the following function @code{foo}:
10109 static int a, b, c;
10110 int foo (void) @{ return a + b + c; @}
10114 usually calculates the addresses of all three variables, but if you
10115 compile it with @option{-fsection-anchors}, it accesses the variables
10116 from a common anchor point instead. The effect is similar to the
10117 following pseudocode (which isn't valid C):
10122 register int *xr = &x;
10123 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10127 Not all targets support this option.
10129 @item --param @var{name}=@var{value}
10131 In some places, GCC uses various constants to control the amount of
10132 optimization that is done. For example, GCC does not inline functions
10133 that contain more than a certain number of instructions. You can
10134 control some of these constants on the command line using the
10135 @option{--param} option.
10137 The names of specific parameters, and the meaning of the values, are
10138 tied to the internals of the compiler, and are subject to change
10139 without notice in future releases.
10141 In each case, the @var{value} is an integer. The allowable choices for
10145 @item predictable-branch-outcome
10146 When branch is predicted to be taken with probability lower than this threshold
10147 (in percent), then it is considered well predictable. The default is 10.
10149 @item max-rtl-if-conversion-insns
10150 RTL if-conversion tries to remove conditional branches around a block and
10151 replace them with conditionally executed instructions. This parameter
10152 gives the maximum number of instructions in a block which should be
10153 considered for if-conversion. The default is 10, though the compiler will
10154 also use other heuristics to decide whether if-conversion is likely to be
10157 @item max-rtl-if-conversion-predictable-cost
10158 @itemx max-rtl-if-conversion-unpredictable-cost
10159 RTL if-conversion will try to remove conditional branches around a block
10160 and replace them with conditionally executed instructions. These parameters
10161 give the maximum permissible cost for the sequence that would be generated
10162 by if-conversion depending on whether the branch is statically determined
10163 to be predictable or not. The units for this parameter are the same as
10164 those for the GCC internal seq_cost metric. The compiler will try to
10165 provide a reasonable default for this parameter using the BRANCH_COST
10168 @item max-crossjump-edges
10169 The maximum number of incoming edges to consider for cross-jumping.
10170 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10171 the number of edges incoming to each block. Increasing values mean
10172 more aggressive optimization, making the compilation time increase with
10173 probably small improvement in executable size.
10175 @item min-crossjump-insns
10176 The minimum number of instructions that must be matched at the end
10177 of two blocks before cross-jumping is performed on them. This
10178 value is ignored in the case where all instructions in the block being
10179 cross-jumped from are matched. The default value is 5.
10181 @item max-grow-copy-bb-insns
10182 The maximum code size expansion factor when copying basic blocks
10183 instead of jumping. The expansion is relative to a jump instruction.
10184 The default value is 8.
10186 @item max-goto-duplication-insns
10187 The maximum number of instructions to duplicate to a block that jumps
10188 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10189 passes, GCC factors computed gotos early in the compilation process,
10190 and unfactors them as late as possible. Only computed jumps at the
10191 end of a basic blocks with no more than max-goto-duplication-insns are
10192 unfactored. The default value is 8.
10194 @item max-delay-slot-insn-search
10195 The maximum number of instructions to consider when looking for an
10196 instruction to fill a delay slot. If more than this arbitrary number of
10197 instructions are searched, the time savings from filling the delay slot
10198 are minimal, so stop searching. Increasing values mean more
10199 aggressive optimization, making the compilation time increase with probably
10200 small improvement in execution time.
10202 @item max-delay-slot-live-search
10203 When trying to fill delay slots, the maximum number of instructions to
10204 consider when searching for a block with valid live register
10205 information. Increasing this arbitrarily chosen value means more
10206 aggressive optimization, increasing the compilation time. This parameter
10207 should be removed when the delay slot code is rewritten to maintain the
10208 control-flow graph.
10210 @item max-gcse-memory
10211 The approximate maximum amount of memory that can be allocated in
10212 order to perform the global common subexpression elimination
10213 optimization. If more memory than specified is required, the
10214 optimization is not done.
10216 @item max-gcse-insertion-ratio
10217 If the ratio of expression insertions to deletions is larger than this value
10218 for any expression, then RTL PRE inserts or removes the expression and thus
10219 leaves partially redundant computations in the instruction stream. The default value is 20.
10221 @item max-pending-list-length
10222 The maximum number of pending dependencies scheduling allows
10223 before flushing the current state and starting over. Large functions
10224 with few branches or calls can create excessively large lists which
10225 needlessly consume memory and resources.
10227 @item max-modulo-backtrack-attempts
10228 The maximum number of backtrack attempts the scheduler should make
10229 when modulo scheduling a loop. Larger values can exponentially increase
10232 @item max-inline-insns-single
10233 Several parameters control the tree inliner used in GCC@.
10234 This number sets the maximum number of instructions (counted in GCC's
10235 internal representation) in a single function that the tree inliner
10236 considers for inlining. This only affects functions declared
10237 inline and methods implemented in a class declaration (C++).
10238 The default value is 400.
10240 @item max-inline-insns-auto
10241 When you use @option{-finline-functions} (included in @option{-O3}),
10242 a lot of functions that would otherwise not be considered for inlining
10243 by the compiler are investigated. To those functions, a different
10244 (more restrictive) limit compared to functions declared inline can
10246 The default value is 30.
10248 @item inline-min-speedup
10249 When estimated performance improvement of caller + callee runtime exceeds this
10250 threshold (in percent), the function can be inlined regardless of the limit on
10251 @option{--param max-inline-insns-single} and @option{--param
10252 max-inline-insns-auto}.
10253 The default value is 15.
10255 @item large-function-insns
10256 The limit specifying really large functions. For functions larger than this
10257 limit after inlining, inlining is constrained by
10258 @option{--param large-function-growth}. This parameter is useful primarily
10259 to avoid extreme compilation time caused by non-linear algorithms used by the
10261 The default value is 2700.
10263 @item large-function-growth
10264 Specifies maximal growth of large function caused by inlining in percents.
10265 The default value is 100 which limits large function growth to 2.0 times
10268 @item large-unit-insns
10269 The limit specifying large translation unit. Growth caused by inlining of
10270 units larger than this limit is limited by @option{--param inline-unit-growth}.
10271 For small units this might be too tight.
10272 For example, consider a unit consisting of function A
10273 that is inline and B that just calls A three times. If B is small relative to
10274 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10275 large units consisting of small inlineable functions, however, the overall unit
10276 growth limit is needed to avoid exponential explosion of code size. Thus for
10277 smaller units, the size is increased to @option{--param large-unit-insns}
10278 before applying @option{--param inline-unit-growth}. The default is 10000.
10280 @item inline-unit-growth
10281 Specifies maximal overall growth of the compilation unit caused by inlining.
10282 The default value is 20 which limits unit growth to 1.2 times the original
10283 size. Cold functions (either marked cold via an attribute or by profile
10284 feedback) are not accounted into the unit size.
10286 @item ipcp-unit-growth
10287 Specifies maximal overall growth of the compilation unit caused by
10288 interprocedural constant propagation. The default value is 10 which limits
10289 unit growth to 1.1 times the original size.
10291 @item large-stack-frame
10292 The limit specifying large stack frames. While inlining the algorithm is trying
10293 to not grow past this limit too much. The default value is 256 bytes.
10295 @item large-stack-frame-growth
10296 Specifies maximal growth of large stack frames caused by inlining in percents.
10297 The default value is 1000 which limits large stack frame growth to 11 times
10300 @item max-inline-insns-recursive
10301 @itemx max-inline-insns-recursive-auto
10302 Specifies the maximum number of instructions an out-of-line copy of a
10303 self-recursive inline
10304 function can grow into by performing recursive inlining.
10306 @option{--param max-inline-insns-recursive} applies to functions
10308 For functions not declared inline, recursive inlining
10309 happens only when @option{-finline-functions} (included in @option{-O3}) is
10310 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10311 default value is 450.
10313 @item max-inline-recursive-depth
10314 @itemx max-inline-recursive-depth-auto
10315 Specifies the maximum recursion depth used for recursive inlining.
10317 @option{--param max-inline-recursive-depth} applies to functions
10318 declared inline. For functions not declared inline, recursive inlining
10319 happens only when @option{-finline-functions} (included in @option{-O3}) is
10320 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10321 default value is 8.
10323 @item min-inline-recursive-probability
10324 Recursive inlining is profitable only for function having deep recursion
10325 in average and can hurt for function having little recursion depth by
10326 increasing the prologue size or complexity of function body to other
10329 When profile feedback is available (see @option{-fprofile-generate}) the actual
10330 recursion depth can be guessed from the probability that function recurses
10331 via a given call expression. This parameter limits inlining only to call
10332 expressions whose probability exceeds the given threshold (in percents).
10333 The default value is 10.
10335 @item early-inlining-insns
10336 Specify growth that the early inliner can make. In effect it increases
10337 the amount of inlining for code having a large abstraction penalty.
10338 The default value is 14.
10340 @item max-early-inliner-iterations
10341 Limit of iterations of the early inliner. This basically bounds
10342 the number of nested indirect calls the early inliner can resolve.
10343 Deeper chains are still handled by late inlining.
10345 @item comdat-sharing-probability
10346 Probability (in percent) that C++ inline function with comdat visibility
10347 are shared across multiple compilation units. The default value is 20.
10349 @item profile-func-internal-id
10350 A parameter to control whether to use function internal id in profile
10351 database lookup. If the value is 0, the compiler uses an id that
10352 is based on function assembler name and filename, which makes old profile
10353 data more tolerant to source changes such as function reordering etc.
10354 The default value is 0.
10356 @item min-vect-loop-bound
10357 The minimum number of iterations under which loops are not vectorized
10358 when @option{-ftree-vectorize} is used. The number of iterations after
10359 vectorization needs to be greater than the value specified by this option
10360 to allow vectorization. The default value is 0.
10362 @item gcse-cost-distance-ratio
10363 Scaling factor in calculation of maximum distance an expression
10364 can be moved by GCSE optimizations. This is currently supported only in the
10365 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10366 is with simple expressions, i.e., the expressions that have cost
10367 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10368 hoisting of simple expressions. The default value is 10.
10370 @item gcse-unrestricted-cost
10371 Cost, roughly measured as the cost of a single typical machine
10372 instruction, at which GCSE optimizations do not constrain
10373 the distance an expression can travel. This is currently
10374 supported only in the code hoisting pass. The lesser the cost,
10375 the more aggressive code hoisting is. Specifying 0
10376 allows all expressions to travel unrestricted distances.
10377 The default value is 3.
10379 @item max-hoist-depth
10380 The depth of search in the dominator tree for expressions to hoist.
10381 This is used to avoid quadratic behavior in hoisting algorithm.
10382 The value of 0 does not limit on the search, but may slow down compilation
10383 of huge functions. The default value is 30.
10385 @item max-tail-merge-comparisons
10386 The maximum amount of similar bbs to compare a bb with. This is used to
10387 avoid quadratic behavior in tree tail merging. The default value is 10.
10389 @item max-tail-merge-iterations
10390 The maximum amount of iterations of the pass over the function. This is used to
10391 limit compilation time in tree tail merging. The default value is 2.
10393 @item store-merging-allow-unaligned
10394 Allow the store merging pass to introduce unaligned stores if it is legal to
10395 do so. The default value is 1.
10397 @item max-stores-to-merge
10398 The maximum number of stores to attempt to merge into wider stores in the store
10399 merging pass. The minimum value is 2 and the default is 64.
10401 @item max-unrolled-insns
10402 The maximum number of instructions that a loop may have to be unrolled.
10403 If a loop is unrolled, this parameter also determines how many times
10404 the loop code is unrolled.
10406 @item max-average-unrolled-insns
10407 The maximum number of instructions biased by probabilities of their execution
10408 that a loop may have to be unrolled. If a loop is unrolled,
10409 this parameter also determines how many times the loop code is unrolled.
10411 @item max-unroll-times
10412 The maximum number of unrollings of a single loop.
10414 @item max-peeled-insns
10415 The maximum number of instructions that a loop may have to be peeled.
10416 If a loop is peeled, this parameter also determines how many times
10417 the loop code is peeled.
10419 @item max-peel-times
10420 The maximum number of peelings of a single loop.
10422 @item max-peel-branches
10423 The maximum number of branches on the hot path through the peeled sequence.
10425 @item max-completely-peeled-insns
10426 The maximum number of insns of a completely peeled loop.
10428 @item max-completely-peel-times
10429 The maximum number of iterations of a loop to be suitable for complete peeling.
10431 @item max-completely-peel-loop-nest-depth
10432 The maximum depth of a loop nest suitable for complete peeling.
10434 @item max-unswitch-insns
10435 The maximum number of insns of an unswitched loop.
10437 @item max-unswitch-level
10438 The maximum number of branches unswitched in a single loop.
10440 @item max-loop-headers-insns
10441 The maximum number of insns in loop header duplicated by the copy loop headers
10444 @item lim-expensive
10445 The minimum cost of an expensive expression in the loop invariant motion.
10447 @item iv-consider-all-candidates-bound
10448 Bound on number of candidates for induction variables, below which
10449 all candidates are considered for each use in induction variable
10450 optimizations. If there are more candidates than this,
10451 only the most relevant ones are considered to avoid quadratic time complexity.
10453 @item iv-max-considered-uses
10454 The induction variable optimizations give up on loops that contain more
10455 induction variable uses.
10457 @item iv-always-prune-cand-set-bound
10458 If the number of candidates in the set is smaller than this value,
10459 always try to remove unnecessary ivs from the set
10460 when adding a new one.
10462 @item avg-loop-niter
10463 Average number of iterations of a loop.
10465 @item dse-max-object-size
10466 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
10467 Larger values may result in larger compilation times.
10469 @item scev-max-expr-size
10470 Bound on size of expressions used in the scalar evolutions analyzer.
10471 Large expressions slow the analyzer.
10473 @item scev-max-expr-complexity
10474 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10475 Complex expressions slow the analyzer.
10477 @item max-tree-if-conversion-phi-args
10478 Maximum number of arguments in a PHI supported by TREE if conversion
10479 unless the loop is marked with simd pragma.
10481 @item vect-max-version-for-alignment-checks
10482 The maximum number of run-time checks that can be performed when
10483 doing loop versioning for alignment in the vectorizer.
10485 @item vect-max-version-for-alias-checks
10486 The maximum number of run-time checks that can be performed when
10487 doing loop versioning for alias in the vectorizer.
10489 @item vect-max-peeling-for-alignment
10490 The maximum number of loop peels to enhance access alignment
10491 for vectorizer. Value -1 means no limit.
10493 @item max-iterations-to-track
10494 The maximum number of iterations of a loop the brute-force algorithm
10495 for analysis of the number of iterations of the loop tries to evaluate.
10497 @item hot-bb-count-ws-permille
10498 A basic block profile count is considered hot if it contributes to
10499 the given permillage (i.e. 0...1000) of the entire profiled execution.
10501 @item hot-bb-frequency-fraction
10502 Select fraction of the entry block frequency of executions of basic block in
10503 function given basic block needs to have to be considered hot.
10505 @item max-predicted-iterations
10506 The maximum number of loop iterations we predict statically. This is useful
10507 in cases where a function contains a single loop with known bound and
10508 another loop with unknown bound.
10509 The known number of iterations is predicted correctly, while
10510 the unknown number of iterations average to roughly 10. This means that the
10511 loop without bounds appears artificially cold relative to the other one.
10513 @item builtin-expect-probability
10514 Control the probability of the expression having the specified value. This
10515 parameter takes a percentage (i.e. 0 ... 100) as input.
10516 The default probability of 90 is obtained empirically.
10518 @item align-threshold
10520 Select fraction of the maximal frequency of executions of a basic block in
10521 a function to align the basic block.
10523 @item align-loop-iterations
10525 A loop expected to iterate at least the selected number of iterations is
10528 @item tracer-dynamic-coverage
10529 @itemx tracer-dynamic-coverage-feedback
10531 This value is used to limit superblock formation once the given percentage of
10532 executed instructions is covered. This limits unnecessary code size
10535 The @option{tracer-dynamic-coverage-feedback} parameter
10536 is used only when profile
10537 feedback is available. The real profiles (as opposed to statically estimated
10538 ones) are much less balanced allowing the threshold to be larger value.
10540 @item tracer-max-code-growth
10541 Stop tail duplication once code growth has reached given percentage. This is
10542 a rather artificial limit, as most of the duplicates are eliminated later in
10543 cross jumping, so it may be set to much higher values than is the desired code
10546 @item tracer-min-branch-ratio
10548 Stop reverse growth when the reverse probability of best edge is less than this
10549 threshold (in percent).
10551 @item tracer-min-branch-probability
10552 @itemx tracer-min-branch-probability-feedback
10554 Stop forward growth if the best edge has probability lower than this
10557 Similarly to @option{tracer-dynamic-coverage} two parameters are
10558 provided. @option{tracer-min-branch-probability-feedback} is used for
10559 compilation with profile feedback and @option{tracer-min-branch-probability}
10560 compilation without. The value for compilation with profile feedback
10561 needs to be more conservative (higher) in order to make tracer
10564 @item stack-clash-protection-guard-size
10565 Specify the size of the operating system provided stack guard as
10566 2 raised to @var{num} bytes. The default value is 12 (4096 bytes).
10567 Acceptable values are between 12 and 30. Higher values may reduce the
10568 number of explicit probes, but a value larger than the operating system
10569 provided guard will leave code vulnerable to stack clash style attacks.
10571 @item stack-clash-protection-probe-interval
10572 Stack clash protection involves probing stack space as it is allocated. This
10573 param controls the maximum distance between probes into the stack as 2 raised
10574 to @var{num} bytes. Acceptable values are between 10 and 16 and defaults to
10575 12. Higher values may reduce the number of explicit probes, but a value
10576 larger than the operating system provided guard will leave code vulnerable to
10577 stack clash style attacks.
10579 @item max-cse-path-length
10581 The maximum number of basic blocks on path that CSE considers.
10584 @item max-cse-insns
10585 The maximum number of instructions CSE processes before flushing.
10586 The default is 1000.
10588 @item ggc-min-expand
10590 GCC uses a garbage collector to manage its own memory allocation. This
10591 parameter specifies the minimum percentage by which the garbage
10592 collector's heap should be allowed to expand between collections.
10593 Tuning this may improve compilation speed; it has no effect on code
10596 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10597 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10598 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10599 GCC is not able to calculate RAM on a particular platform, the lower
10600 bound of 30% is used. Setting this parameter and
10601 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10602 every opportunity. This is extremely slow, but can be useful for
10605 @item ggc-min-heapsize
10607 Minimum size of the garbage collector's heap before it begins bothering
10608 to collect garbage. The first collection occurs after the heap expands
10609 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10610 tuning this may improve compilation speed, and has no effect on code
10613 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10614 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10615 with a lower bound of 4096 (four megabytes) and an upper bound of
10616 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10617 particular platform, the lower bound is used. Setting this parameter
10618 very large effectively disables garbage collection. Setting this
10619 parameter and @option{ggc-min-expand} to zero causes a full collection
10620 to occur at every opportunity.
10622 @item max-reload-search-insns
10623 The maximum number of instruction reload should look backward for equivalent
10624 register. Increasing values mean more aggressive optimization, making the
10625 compilation time increase with probably slightly better performance.
10626 The default value is 100.
10628 @item max-cselib-memory-locations
10629 The maximum number of memory locations cselib should take into account.
10630 Increasing values mean more aggressive optimization, making the compilation time
10631 increase with probably slightly better performance. The default value is 500.
10633 @item max-sched-ready-insns
10634 The maximum number of instructions ready to be issued the scheduler should
10635 consider at any given time during the first scheduling pass. Increasing
10636 values mean more thorough searches, making the compilation time increase
10637 with probably little benefit. The default value is 100.
10639 @item max-sched-region-blocks
10640 The maximum number of blocks in a region to be considered for
10641 interblock scheduling. The default value is 10.
10643 @item max-pipeline-region-blocks
10644 The maximum number of blocks in a region to be considered for
10645 pipelining in the selective scheduler. The default value is 15.
10647 @item max-sched-region-insns
10648 The maximum number of insns in a region to be considered for
10649 interblock scheduling. The default value is 100.
10651 @item max-pipeline-region-insns
10652 The maximum number of insns in a region to be considered for
10653 pipelining in the selective scheduler. The default value is 200.
10655 @item min-spec-prob
10656 The minimum probability (in percents) of reaching a source block
10657 for interblock speculative scheduling. The default value is 40.
10659 @item max-sched-extend-regions-iters
10660 The maximum number of iterations through CFG to extend regions.
10661 A value of 0 (the default) disables region extensions.
10663 @item max-sched-insn-conflict-delay
10664 The maximum conflict delay for an insn to be considered for speculative motion.
10665 The default value is 3.
10667 @item sched-spec-prob-cutoff
10668 The minimal probability of speculation success (in percents), so that
10669 speculative insns are scheduled.
10670 The default value is 40.
10672 @item sched-state-edge-prob-cutoff
10673 The minimum probability an edge must have for the scheduler to save its
10675 The default value is 10.
10677 @item sched-mem-true-dep-cost
10678 Minimal distance (in CPU cycles) between store and load targeting same
10679 memory locations. The default value is 1.
10681 @item selsched-max-lookahead
10682 The maximum size of the lookahead window of selective scheduling. It is a
10683 depth of search for available instructions.
10684 The default value is 50.
10686 @item selsched-max-sched-times
10687 The maximum number of times that an instruction is scheduled during
10688 selective scheduling. This is the limit on the number of iterations
10689 through which the instruction may be pipelined. The default value is 2.
10691 @item selsched-insns-to-rename
10692 The maximum number of best instructions in the ready list that are considered
10693 for renaming in the selective scheduler. The default value is 2.
10696 The minimum value of stage count that swing modulo scheduler
10697 generates. The default value is 2.
10699 @item max-last-value-rtl
10700 The maximum size measured as number of RTLs that can be recorded in an expression
10701 in combiner for a pseudo register as last known value of that register. The default
10704 @item max-combine-insns
10705 The maximum number of instructions the RTL combiner tries to combine.
10706 The default value is 2 at @option{-Og} and 4 otherwise.
10708 @item integer-share-limit
10709 Small integer constants can use a shared data structure, reducing the
10710 compiler's memory usage and increasing its speed. This sets the maximum
10711 value of a shared integer constant. The default value is 256.
10713 @item ssp-buffer-size
10714 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10715 protection when @option{-fstack-protection} is used.
10717 @item min-size-for-stack-sharing
10718 The minimum size of variables taking part in stack slot sharing when not
10719 optimizing. The default value is 32.
10721 @item max-jump-thread-duplication-stmts
10722 Maximum number of statements allowed in a block that needs to be
10723 duplicated when threading jumps.
10725 @item max-fields-for-field-sensitive
10726 Maximum number of fields in a structure treated in
10727 a field sensitive manner during pointer analysis. The default is zero
10728 for @option{-O0} and @option{-O1},
10729 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10731 @item prefetch-latency
10732 Estimate on average number of instructions that are executed before
10733 prefetch finishes. The distance prefetched ahead is proportional
10734 to this constant. Increasing this number may also lead to less
10735 streams being prefetched (see @option{simultaneous-prefetches}).
10737 @item simultaneous-prefetches
10738 Maximum number of prefetches that can run at the same time.
10740 @item l1-cache-line-size
10741 The size of cache line in L1 cache, in bytes.
10743 @item l1-cache-size
10744 The size of L1 cache, in kilobytes.
10746 @item l2-cache-size
10747 The size of L2 cache, in kilobytes.
10749 @item loop-interchange-max-num-stmts
10750 The maximum number of stmts in a loop to be interchanged.
10752 @item loop-interchange-stride-ratio
10753 The minimum ratio between stride of two loops for interchange to be profitable.
10755 @item min-insn-to-prefetch-ratio
10756 The minimum ratio between the number of instructions and the
10757 number of prefetches to enable prefetching in a loop.
10759 @item prefetch-min-insn-to-mem-ratio
10760 The minimum ratio between the number of instructions and the
10761 number of memory references to enable prefetching in a loop.
10763 @item use-canonical-types
10764 Whether the compiler should use the ``canonical'' type system. By
10765 default, this should always be 1, which uses a more efficient internal
10766 mechanism for comparing types in C++ and Objective-C++. However, if
10767 bugs in the canonical type system are causing compilation failures,
10768 set this value to 0 to disable canonical types.
10770 @item switch-conversion-max-branch-ratio
10771 Switch initialization conversion refuses to create arrays that are
10772 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10773 branches in the switch.
10775 @item max-partial-antic-length
10776 Maximum length of the partial antic set computed during the tree
10777 partial redundancy elimination optimization (@option{-ftree-pre}) when
10778 optimizing at @option{-O3} and above. For some sorts of source code
10779 the enhanced partial redundancy elimination optimization can run away,
10780 consuming all of the memory available on the host machine. This
10781 parameter sets a limit on the length of the sets that are computed,
10782 which prevents the runaway behavior. Setting a value of 0 for
10783 this parameter allows an unlimited set length.
10785 @item sccvn-max-scc-size
10786 Maximum size of a strongly connected component (SCC) during SCCVN
10787 processing. If this limit is hit, SCCVN processing for the whole
10788 function is not done and optimizations depending on it are
10789 disabled. The default maximum SCC size is 10000.
10791 @item sccvn-max-alias-queries-per-access
10792 Maximum number of alias-oracle queries we perform when looking for
10793 redundancies for loads and stores. If this limit is hit the search
10794 is aborted and the load or store is not considered redundant. The
10795 number of queries is algorithmically limited to the number of
10796 stores on all paths from the load to the function entry.
10797 The default maximum number of queries is 1000.
10799 @item ira-max-loops-num
10800 IRA uses regional register allocation by default. If a function
10801 contains more loops than the number given by this parameter, only at most
10802 the given number of the most frequently-executed loops form regions
10803 for regional register allocation. The default value of the
10806 @item ira-max-conflict-table-size
10807 Although IRA uses a sophisticated algorithm to compress the conflict
10808 table, the table can still require excessive amounts of memory for
10809 huge functions. If the conflict table for a function could be more
10810 than the size in MB given by this parameter, the register allocator
10811 instead uses a faster, simpler, and lower-quality
10812 algorithm that does not require building a pseudo-register conflict table.
10813 The default value of the parameter is 2000.
10815 @item ira-loop-reserved-regs
10816 IRA can be used to evaluate more accurate register pressure in loops
10817 for decisions to move loop invariants (see @option{-O3}). The number
10818 of available registers reserved for some other purposes is given
10819 by this parameter. The default value of the parameter is 2, which is
10820 the minimal number of registers needed by typical instructions.
10821 This value is the best found from numerous experiments.
10823 @item lra-inheritance-ebb-probability-cutoff
10824 LRA tries to reuse values reloaded in registers in subsequent insns.
10825 This optimization is called inheritance. EBB is used as a region to
10826 do this optimization. The parameter defines a minimal fall-through
10827 edge probability in percentage used to add BB to inheritance EBB in
10828 LRA. The default value of the parameter is 40. The value was chosen
10829 from numerous runs of SPEC2000 on x86-64.
10831 @item loop-invariant-max-bbs-in-loop
10832 Loop invariant motion can be very expensive, both in compilation time and
10833 in amount of needed compile-time memory, with very large loops. Loops
10834 with more basic blocks than this parameter won't have loop invariant
10835 motion optimization performed on them. The default value of the
10836 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10838 @item loop-max-datarefs-for-datadeps
10839 Building data dependencies is expensive for very large loops. This
10840 parameter limits the number of data references in loops that are
10841 considered for data dependence analysis. These large loops are no
10842 handled by the optimizations using loop data dependencies.
10843 The default value is 1000.
10845 @item max-vartrack-size
10846 Sets a maximum number of hash table slots to use during variable
10847 tracking dataflow analysis of any function. If this limit is exceeded
10848 with variable tracking at assignments enabled, analysis for that
10849 function is retried without it, after removing all debug insns from
10850 the function. If the limit is exceeded even without debug insns, var
10851 tracking analysis is completely disabled for the function. Setting
10852 the parameter to zero makes it unlimited.
10854 @item max-vartrack-expr-depth
10855 Sets a maximum number of recursion levels when attempting to map
10856 variable names or debug temporaries to value expressions. This trades
10857 compilation time for more complete debug information. If this is set too
10858 low, value expressions that are available and could be represented in
10859 debug information may end up not being used; setting this higher may
10860 enable the compiler to find more complex debug expressions, but compile
10861 time and memory use may grow. The default is 12.
10863 @item max-debug-marker-count
10864 Sets a threshold on the number of debug markers (e.g. begin stmt
10865 markers) to avoid complexity explosion at inlining or expanding to RTL.
10866 If a function has more such gimple stmts than the set limit, such stmts
10867 will be dropped from the inlined copy of a function, and from its RTL
10868 expansion. The default is 100000.
10870 @item min-nondebug-insn-uid
10871 Use uids starting at this parameter for nondebug insns. The range below
10872 the parameter is reserved exclusively for debug insns created by
10873 @option{-fvar-tracking-assignments}, but debug insns may get
10874 (non-overlapping) uids above it if the reserved range is exhausted.
10876 @item ipa-sra-ptr-growth-factor
10877 IPA-SRA replaces a pointer to an aggregate with one or more new
10878 parameters only when their cumulative size is less or equal to
10879 @option{ipa-sra-ptr-growth-factor} times the size of the original
10882 @item sra-max-scalarization-size-Ospeed
10883 @itemx sra-max-scalarization-size-Osize
10884 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10885 replace scalar parts of aggregates with uses of independent scalar
10886 variables. These parameters control the maximum size, in storage units,
10887 of aggregate which is considered for replacement when compiling for
10889 (@option{sra-max-scalarization-size-Ospeed}) or size
10890 (@option{sra-max-scalarization-size-Osize}) respectively.
10892 @item tm-max-aggregate-size
10893 When making copies of thread-local variables in a transaction, this
10894 parameter specifies the size in bytes after which variables are
10895 saved with the logging functions as opposed to save/restore code
10896 sequence pairs. This option only applies when using
10899 @item graphite-max-nb-scop-params
10900 To avoid exponential effects in the Graphite loop transforms, the
10901 number of parameters in a Static Control Part (SCoP) is bounded. The
10902 default value is 10 parameters, a value of zero can be used to lift
10903 the bound. A variable whose value is unknown at compilation time and
10904 defined outside a SCoP is a parameter of the SCoP.
10906 @item loop-block-tile-size
10907 Loop blocking or strip mining transforms, enabled with
10908 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10909 loop in the loop nest by a given number of iterations. The strip
10910 length can be changed using the @option{loop-block-tile-size}
10911 parameter. The default value is 51 iterations.
10913 @item loop-unroll-jam-size
10914 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10915 default value is 4.
10917 @item loop-unroll-jam-depth
10918 Specify the dimension to be unrolled (counting from the most inner loop)
10919 for the @option{-floop-unroll-and-jam}. The default value is 2.
10921 @item ipa-cp-value-list-size
10922 IPA-CP attempts to track all possible values and types passed to a function's
10923 parameter in order to propagate them and perform devirtualization.
10924 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10925 stores per one formal parameter of a function.
10927 @item ipa-cp-eval-threshold
10928 IPA-CP calculates its own score of cloning profitability heuristics
10929 and performs those cloning opportunities with scores that exceed
10930 @option{ipa-cp-eval-threshold}.
10932 @item ipa-cp-recursion-penalty
10933 Percentage penalty the recursive functions will receive when they
10934 are evaluated for cloning.
10936 @item ipa-cp-single-call-penalty
10937 Percentage penalty functions containing a single call to another
10938 function will receive when they are evaluated for cloning.
10941 @item ipa-max-agg-items
10942 IPA-CP is also capable to propagate a number of scalar values passed
10943 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10944 number of such values per one parameter.
10946 @item ipa-cp-loop-hint-bonus
10947 When IPA-CP determines that a cloning candidate would make the number
10948 of iterations of a loop known, it adds a bonus of
10949 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10952 @item ipa-cp-array-index-hint-bonus
10953 When IPA-CP determines that a cloning candidate would make the index of
10954 an array access known, it adds a bonus of
10955 @option{ipa-cp-array-index-hint-bonus} to the profitability
10956 score of the candidate.
10958 @item ipa-max-aa-steps
10959 During its analysis of function bodies, IPA-CP employs alias analysis
10960 in order to track values pointed to by function parameters. In order
10961 not spend too much time analyzing huge functions, it gives up and
10962 consider all memory clobbered after examining
10963 @option{ipa-max-aa-steps} statements modifying memory.
10965 @item lto-partitions
10966 Specify desired number of partitions produced during WHOPR compilation.
10967 The number of partitions should exceed the number of CPUs used for compilation.
10968 The default value is 32.
10970 @item lto-min-partition
10971 Size of minimal partition for WHOPR (in estimated instructions).
10972 This prevents expenses of splitting very small programs into too many
10975 @item lto-max-partition
10976 Size of max partition for WHOPR (in estimated instructions).
10977 to provide an upper bound for individual size of partition.
10978 Meant to be used only with balanced partitioning.
10980 @item cxx-max-namespaces-for-diagnostic-help
10981 The maximum number of namespaces to consult for suggestions when C++
10982 name lookup fails for an identifier. The default is 1000.
10984 @item sink-frequency-threshold
10985 The maximum relative execution frequency (in percents) of the target block
10986 relative to a statement's original block to allow statement sinking of a
10987 statement. Larger numbers result in more aggressive statement sinking.
10988 The default value is 75. A small positive adjustment is applied for
10989 statements with memory operands as those are even more profitable so sink.
10991 @item max-stores-to-sink
10992 The maximum number of conditional store pairs that can be sunk. Set to 0
10993 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10994 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10996 @item allow-store-data-races
10997 Allow optimizers to introduce new data races on stores.
10998 Set to 1 to allow, otherwise to 0. This option is enabled by default
10999 at optimization level @option{-Ofast}.
11001 @item case-values-threshold
11002 The smallest number of different values for which it is best to use a
11003 jump-table instead of a tree of conditional branches. If the value is
11004 0, use the default for the machine. The default is 0.
11006 @item tree-reassoc-width
11007 Set the maximum number of instructions executed in parallel in
11008 reassociated tree. This parameter overrides target dependent
11009 heuristics used by default if has non zero value.
11011 @item sched-pressure-algorithm
11012 Choose between the two available implementations of
11013 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11014 and is the more likely to prevent instructions from being reordered.
11015 Algorithm 2 was designed to be a compromise between the relatively
11016 conservative approach taken by algorithm 1 and the rather aggressive
11017 approach taken by the default scheduler. It relies more heavily on
11018 having a regular register file and accurate register pressure classes.
11019 See @file{haifa-sched.c} in the GCC sources for more details.
11021 The default choice depends on the target.
11023 @item max-slsr-cand-scan
11024 Set the maximum number of existing candidates that are considered when
11025 seeking a basis for a new straight-line strength reduction candidate.
11028 Enable buffer overflow detection for global objects. This kind
11029 of protection is enabled by default if you are using
11030 @option{-fsanitize=address} option.
11031 To disable global objects protection use @option{--param asan-globals=0}.
11034 Enable buffer overflow detection for stack objects. This kind of
11035 protection is enabled by default when using @option{-fsanitize=address}.
11036 To disable stack protection use @option{--param asan-stack=0} option.
11038 @item asan-instrument-reads
11039 Enable buffer overflow detection for memory reads. This kind of
11040 protection is enabled by default when using @option{-fsanitize=address}.
11041 To disable memory reads protection use
11042 @option{--param asan-instrument-reads=0}.
11044 @item asan-instrument-writes
11045 Enable buffer overflow detection for memory writes. This kind of
11046 protection is enabled by default when using @option{-fsanitize=address}.
11047 To disable memory writes protection use
11048 @option{--param asan-instrument-writes=0} option.
11050 @item asan-memintrin
11051 Enable detection for built-in functions. This kind of protection
11052 is enabled by default when using @option{-fsanitize=address}.
11053 To disable built-in functions protection use
11054 @option{--param asan-memintrin=0}.
11056 @item asan-use-after-return
11057 Enable detection of use-after-return. This kind of protection
11058 is enabled by default when using the @option{-fsanitize=address} option.
11059 To disable it use @option{--param asan-use-after-return=0}.
11061 Note: By default the check is disabled at run time. To enable it,
11062 add @code{detect_stack_use_after_return=1} to the environment variable
11063 @env{ASAN_OPTIONS}.
11065 @item asan-instrumentation-with-call-threshold
11066 If number of memory accesses in function being instrumented
11067 is greater or equal to this number, use callbacks instead of inline checks.
11068 E.g. to disable inline code use
11069 @option{--param asan-instrumentation-with-call-threshold=0}.
11071 @item use-after-scope-direct-emission-threshold
11072 If the size of a local variable in bytes is smaller or equal to this
11073 number, directly poison (or unpoison) shadow memory instead of using
11074 run-time callbacks. The default value is 256.
11076 @item chkp-max-ctor-size
11077 Static constructors generated by Pointer Bounds Checker may become very
11078 large and significantly increase compile time at optimization level
11079 @option{-O1} and higher. This parameter is a maximum number of statements
11080 in a single generated constructor. Default value is 5000.
11082 @item max-fsm-thread-path-insns
11083 Maximum number of instructions to copy when duplicating blocks on a
11084 finite state automaton jump thread path. The default is 100.
11086 @item max-fsm-thread-length
11087 Maximum number of basic blocks on a finite state automaton jump thread
11088 path. The default is 10.
11090 @item max-fsm-thread-paths
11091 Maximum number of new jump thread paths to create for a finite state
11092 automaton. The default is 50.
11094 @item parloops-chunk-size
11095 Chunk size of omp schedule for loops parallelized by parloops. The default
11098 @item parloops-schedule
11099 Schedule type of omp schedule for loops parallelized by parloops (static,
11100 dynamic, guided, auto, runtime). The default is static.
11102 @item parloops-min-per-thread
11103 The minimum number of iterations per thread of an innermost parallelized
11104 loop for which the parallelized variant is prefered over the single threaded
11105 one. The default is 100. Note that for a parallelized loop nest the
11106 minimum number of iterations of the outermost loop per thread is two.
11108 @item max-ssa-name-query-depth
11109 Maximum depth of recursion when querying properties of SSA names in things
11110 like fold routines. One level of recursion corresponds to following a
11113 @item hsa-gen-debug-stores
11114 Enable emission of special debug stores within HSA kernels which are
11115 then read and reported by libgomp plugin. Generation of these stores
11116 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
11119 @item max-speculative-devirt-maydefs
11120 The maximum number of may-defs we analyze when looking for a must-def
11121 specifying the dynamic type of an object that invokes a virtual call
11122 we may be able to devirtualize speculatively.
11124 @item max-vrp-switch-assertions
11125 The maximum number of assertions to add along the default edge of a switch
11126 statement during VRP. The default is 10.
11128 @item unroll-jam-min-percent
11129 The minimum percentage of memory references that must be optimized
11130 away for the unroll-and-jam transformation to be considered profitable.
11132 @item unroll-jam-max-unroll
11133 The maximum number of times the outer loop should be unrolled by
11134 the unroll-and-jam transformation.
11138 @node Instrumentation Options
11139 @section Program Instrumentation Options
11140 @cindex instrumentation options
11141 @cindex program instrumentation options
11142 @cindex run-time error checking options
11143 @cindex profiling options
11144 @cindex options, program instrumentation
11145 @cindex options, run-time error checking
11146 @cindex options, profiling
11148 GCC supports a number of command-line options that control adding
11149 run-time instrumentation to the code it normally generates.
11150 For example, one purpose of instrumentation is collect profiling
11151 statistics for use in finding program hot spots, code coverage
11152 analysis, or profile-guided optimizations.
11153 Another class of program instrumentation is adding run-time checking
11154 to detect programming errors like invalid pointer
11155 dereferences or out-of-bounds array accesses, as well as deliberately
11156 hostile attacks such as stack smashing or C++ vtable hijacking.
11157 There is also a general hook which can be used to implement other
11158 forms of tracing or function-level instrumentation for debug or
11159 program analysis purposes.
11162 @cindex @command{prof}
11165 Generate extra code to write profile information suitable for the
11166 analysis program @command{prof}. You must use this option when compiling
11167 the source files you want data about, and you must also use it when
11170 @cindex @command{gprof}
11173 Generate extra code to write profile information suitable for the
11174 analysis program @command{gprof}. You must use this option when compiling
11175 the source files you want data about, and you must also use it when
11178 @item -fprofile-arcs
11179 @opindex fprofile-arcs
11180 Add code so that program flow @dfn{arcs} are instrumented. During
11181 execution the program records how many times each branch and call is
11182 executed and how many times it is taken or returns. On targets that support
11183 constructors with priority support, profiling properly handles constructors,
11184 destructors and C++ constructors (and destructors) of classes which are used
11185 as a type of a global variable.
11188 program exits it saves this data to a file called
11189 @file{@var{auxname}.gcda} for each source file. The data may be used for
11190 profile-directed optimizations (@option{-fbranch-probabilities}), or for
11191 test coverage analysis (@option{-ftest-coverage}). Each object file's
11192 @var{auxname} is generated from the name of the output file, if
11193 explicitly specified and it is not the final executable, otherwise it is
11194 the basename of the source file. In both cases any suffix is removed
11195 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
11196 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
11197 @xref{Cross-profiling}.
11199 @cindex @command{gcov}
11203 This option is used to compile and link code instrumented for coverage
11204 analysis. The option is a synonym for @option{-fprofile-arcs}
11205 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
11206 linking). See the documentation for those options for more details.
11211 Compile the source files with @option{-fprofile-arcs} plus optimization
11212 and code generation options. For test coverage analysis, use the
11213 additional @option{-ftest-coverage} option. You do not need to profile
11214 every source file in a program.
11217 Compile the source files additionally with @option{-fprofile-abs-path}
11218 to create absolute path names in the @file{.gcno} files. This allows
11219 @command{gcov} to find the correct sources in projects where compilations
11220 occur with different working directories.
11223 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
11224 (the latter implies the former).
11227 Run the program on a representative workload to generate the arc profile
11228 information. This may be repeated any number of times. You can run
11229 concurrent instances of your program, and provided that the file system
11230 supports locking, the data files will be correctly updated. Unless
11231 a strict ISO C dialect option is in effect, @code{fork} calls are
11232 detected and correctly handled without double counting.
11235 For profile-directed optimizations, compile the source files again with
11236 the same optimization and code generation options plus
11237 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
11238 Control Optimization}).
11241 For test coverage analysis, use @command{gcov} to produce human readable
11242 information from the @file{.gcno} and @file{.gcda} files. Refer to the
11243 @command{gcov} documentation for further information.
11247 With @option{-fprofile-arcs}, for each function of your program GCC
11248 creates a program flow graph, then finds a spanning tree for the graph.
11249 Only arcs that are not on the spanning tree have to be instrumented: the
11250 compiler adds code to count the number of times that these arcs are
11251 executed. When an arc is the only exit or only entrance to a block, the
11252 instrumentation code can be added to the block; otherwise, a new basic
11253 block must be created to hold the instrumentation code.
11256 @item -ftest-coverage
11257 @opindex ftest-coverage
11258 Produce a notes file that the @command{gcov} code-coverage utility
11259 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
11260 show program coverage. Each source file's note file is called
11261 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
11262 above for a description of @var{auxname} and instructions on how to
11263 generate test coverage data. Coverage data matches the source files
11264 more closely if you do not optimize.
11266 @item -fprofile-abs-path
11267 @opindex fprofile-abs-path
11268 Automatically convert relative source file names to absolute path names
11269 in the @file{.gcno} files. This allows @command{gcov} to find the correct
11270 sources in projects where compilations occur with different working
11273 @item -fprofile-dir=@var{path}
11274 @opindex fprofile-dir
11276 Set the directory to search for the profile data files in to @var{path}.
11277 This option affects only the profile data generated by
11278 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
11279 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
11280 and its related options. Both absolute and relative paths can be used.
11281 By default, GCC uses the current directory as @var{path}, thus the
11282 profile data file appears in the same directory as the object file.
11284 @item -fprofile-generate
11285 @itemx -fprofile-generate=@var{path}
11286 @opindex fprofile-generate
11288 Enable options usually used for instrumenting application to produce
11289 profile useful for later recompilation with profile feedback based
11290 optimization. You must use @option{-fprofile-generate} both when
11291 compiling and when linking your program.
11293 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
11295 If @var{path} is specified, GCC looks at the @var{path} to find
11296 the profile feedback data files. See @option{-fprofile-dir}.
11298 To optimize the program based on the collected profile information, use
11299 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
11301 @item -fprofile-update=@var{method}
11302 @opindex fprofile-update
11304 Alter the update method for an application instrumented for profile
11305 feedback based optimization. The @var{method} argument should be one of
11306 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
11307 The first one is useful for single-threaded applications,
11308 while the second one prevents profile corruption by emitting thread-safe code.
11310 @strong{Warning:} When an application does not properly join all threads
11311 (or creates an detached thread), a profile file can be still corrupted.
11313 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
11314 when supported by a target, or to @samp{single} otherwise. The GCC driver
11315 automatically selects @samp{prefer-atomic} when @option{-pthread}
11316 is present in the command line.
11318 @item -fsanitize=address
11319 @opindex fsanitize=address
11320 Enable AddressSanitizer, a fast memory error detector.
11321 Memory access instructions are instrumented to detect
11322 out-of-bounds and use-after-free bugs.
11323 The option enables @option{-fsanitize-address-use-after-scope}.
11324 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
11325 more details. The run-time behavior can be influenced using the
11326 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
11327 the available options are shown at startup of the instrumented program. See
11328 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
11329 for a list of supported options.
11330 The option cannot be combined with @option{-fsanitize=thread}
11331 and/or @option{-fcheck-pointer-bounds}.
11333 @item -fsanitize=kernel-address
11334 @opindex fsanitize=kernel-address
11335 Enable AddressSanitizer for Linux kernel.
11336 See @uref{https://github.com/google/kasan/wiki} for more details.
11337 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11339 @item -fsanitize=pointer-compare
11340 @opindex fsanitize=pointer-compare
11341 Instrument comparison operation (<, <=, >, >=) with pointer operands.
11342 The option must be combined with either @option{-fsanitize=kernel-address} or
11343 @option{-fsanitize=address}
11344 The option cannot be combined with @option{-fsanitize=thread}
11345 and/or @option{-fcheck-pointer-bounds}.
11346 Note: By default the check is disabled at run time. To enable it,
11347 add @code{detect_invalid_pointer_pairs=2} to the environment variable
11348 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
11349 invalid operation only when both pointers are non-null.
11351 @item -fsanitize=pointer-subtract
11352 @opindex fsanitize=pointer-subtract
11353 Instrument subtraction with pointer operands.
11354 The option must be combined with either @option{-fsanitize=kernel-address} or
11355 @option{-fsanitize=address}
11356 The option cannot be combined with @option{-fsanitize=thread}
11357 and/or @option{-fcheck-pointer-bounds}.
11358 Note: By default the check is disabled at run time. To enable it,
11359 add @code{detect_invalid_pointer_pairs=2} to the environment variable
11360 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
11361 invalid operation only when both pointers are non-null.
11363 @item -fsanitize=thread
11364 @opindex fsanitize=thread
11365 Enable ThreadSanitizer, a fast data race detector.
11366 Memory access instructions are instrumented to detect
11367 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
11368 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
11369 environment variable; see
11370 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
11372 The option cannot be combined with @option{-fsanitize=address},
11373 @option{-fsanitize=leak} and/or @option{-fcheck-pointer-bounds}.
11375 Note that sanitized atomic builtins cannot throw exceptions when
11376 operating on invalid memory addresses with non-call exceptions
11377 (@option{-fnon-call-exceptions}).
11379 @item -fsanitize=leak
11380 @opindex fsanitize=leak
11381 Enable LeakSanitizer, a memory leak detector.
11382 This option only matters for linking of executables and
11383 the executable is linked against a library that overrides @code{malloc}
11384 and other allocator functions. See
11385 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
11386 details. The run-time behavior can be influenced using the
11387 @env{LSAN_OPTIONS} environment variable.
11388 The option cannot be combined with @option{-fsanitize=thread}.
11390 @item -fsanitize=undefined
11391 @opindex fsanitize=undefined
11392 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
11393 Various computations are instrumented to detect undefined behavior
11394 at runtime. Current suboptions are:
11398 @item -fsanitize=shift
11399 @opindex fsanitize=shift
11400 This option enables checking that the result of a shift operation is
11401 not undefined. Note that what exactly is considered undefined differs
11402 slightly between C and C++, as well as between ISO C90 and C99, etc.
11403 This option has two suboptions, @option{-fsanitize=shift-base} and
11404 @option{-fsanitize=shift-exponent}.
11406 @item -fsanitize=shift-exponent
11407 @opindex fsanitize=shift-exponent
11408 This option enables checking that the second argument of a shift operation
11409 is not negative and is smaller than the precision of the promoted first
11412 @item -fsanitize=shift-base
11413 @opindex fsanitize=shift-base
11414 If the second argument of a shift operation is within range, check that the
11415 result of a shift operation is not undefined. Note that what exactly is
11416 considered undefined differs slightly between C and C++, as well as between
11417 ISO C90 and C99, etc.
11419 @item -fsanitize=integer-divide-by-zero
11420 @opindex fsanitize=integer-divide-by-zero
11421 Detect integer division by zero as well as @code{INT_MIN / -1} division.
11423 @item -fsanitize=unreachable
11424 @opindex fsanitize=unreachable
11425 With this option, the compiler turns the @code{__builtin_unreachable}
11426 call into a diagnostics message call instead. When reaching the
11427 @code{__builtin_unreachable} call, the behavior is undefined.
11429 @item -fsanitize=vla-bound
11430 @opindex fsanitize=vla-bound
11431 This option instructs the compiler to check that the size of a variable
11432 length array is positive.
11434 @item -fsanitize=null
11435 @opindex fsanitize=null
11436 This option enables pointer checking. Particularly, the application
11437 built with this option turned on will issue an error message when it
11438 tries to dereference a NULL pointer, or if a reference (possibly an
11439 rvalue reference) is bound to a NULL pointer, or if a method is invoked
11440 on an object pointed by a NULL pointer.
11442 @item -fsanitize=return
11443 @opindex fsanitize=return
11444 This option enables return statement checking. Programs
11445 built with this option turned on will issue an error message
11446 when the end of a non-void function is reached without actually
11447 returning a value. This option works in C++ only.
11449 @item -fsanitize=signed-integer-overflow
11450 @opindex fsanitize=signed-integer-overflow
11451 This option enables signed integer overflow checking. We check that
11452 the result of @code{+}, @code{*}, and both unary and binary @code{-}
11453 does not overflow in the signed arithmetics. Note, integer promotion
11454 rules must be taken into account. That is, the following is not an
11457 signed char a = SCHAR_MAX;
11461 @item -fsanitize=bounds
11462 @opindex fsanitize=bounds
11463 This option enables instrumentation of array bounds. Various out of bounds
11464 accesses are detected. Flexible array members, flexible array member-like
11465 arrays, and initializers of variables with static storage are not instrumented.
11466 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11468 @item -fsanitize=bounds-strict
11469 @opindex fsanitize=bounds-strict
11470 This option enables strict instrumentation of array bounds. Most out of bounds
11471 accesses are detected, including flexible array members and flexible array
11472 member-like arrays. Initializers of variables with static storage are not
11473 instrumented. The option cannot be combined
11474 with @option{-fcheck-pointer-bounds}.
11476 @item -fsanitize=alignment
11477 @opindex fsanitize=alignment
11479 This option enables checking of alignment of pointers when they are
11480 dereferenced, or when a reference is bound to insufficiently aligned target,
11481 or when a method or constructor is invoked on insufficiently aligned object.
11483 @item -fsanitize=object-size
11484 @opindex fsanitize=object-size
11485 This option enables instrumentation of memory references using the
11486 @code{__builtin_object_size} function. Various out of bounds pointer
11487 accesses are detected.
11489 @item -fsanitize=float-divide-by-zero
11490 @opindex fsanitize=float-divide-by-zero
11491 Detect floating-point division by zero. Unlike other similar options,
11492 @option{-fsanitize=float-divide-by-zero} is not enabled by
11493 @option{-fsanitize=undefined}, since floating-point division by zero can
11494 be a legitimate way of obtaining infinities and NaNs.
11496 @item -fsanitize=float-cast-overflow
11497 @opindex fsanitize=float-cast-overflow
11498 This option enables floating-point type to integer conversion checking.
11499 We check that the result of the conversion does not overflow.
11500 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
11501 not enabled by @option{-fsanitize=undefined}.
11502 This option does not work well with @code{FE_INVALID} exceptions enabled.
11504 @item -fsanitize=nonnull-attribute
11505 @opindex fsanitize=nonnull-attribute
11507 This option enables instrumentation of calls, checking whether null values
11508 are not passed to arguments marked as requiring a non-null value by the
11509 @code{nonnull} function attribute.
11511 @item -fsanitize=returns-nonnull-attribute
11512 @opindex fsanitize=returns-nonnull-attribute
11514 This option enables instrumentation of return statements in functions
11515 marked with @code{returns_nonnull} function attribute, to detect returning
11516 of null values from such functions.
11518 @item -fsanitize=bool
11519 @opindex fsanitize=bool
11521 This option enables instrumentation of loads from bool. If a value other
11522 than 0/1 is loaded, a run-time error is issued.
11524 @item -fsanitize=enum
11525 @opindex fsanitize=enum
11527 This option enables instrumentation of loads from an enum type. If
11528 a value outside the range of values for the enum type is loaded,
11529 a run-time error is issued.
11531 @item -fsanitize=vptr
11532 @opindex fsanitize=vptr
11534 This option enables instrumentation of C++ member function calls, member
11535 accesses and some conversions between pointers to base and derived classes,
11536 to verify the referenced object has the correct dynamic type.
11538 @item -fsanitize=pointer-overflow
11539 @opindex fsanitize=pointer-overflow
11541 This option enables instrumentation of pointer arithmetics. If the pointer
11542 arithmetics overflows, a run-time error is issued.
11544 @item -fsanitize=builtin
11545 @opindex fsanitize=builtin
11547 This option enables instrumentation of arguments to selected builtin
11548 functions. If an invalid value is passed to such arguments, a run-time
11549 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
11550 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
11555 While @option{-ftrapv} causes traps for signed overflows to be emitted,
11556 @option{-fsanitize=undefined} gives a diagnostic message.
11557 This currently works only for the C family of languages.
11559 @item -fno-sanitize=all
11560 @opindex fno-sanitize=all
11562 This option disables all previously enabled sanitizers.
11563 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
11566 @item -fasan-shadow-offset=@var{number}
11567 @opindex fasan-shadow-offset
11568 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
11569 It is useful for experimenting with different shadow memory layouts in
11570 Kernel AddressSanitizer.
11572 @item -fsanitize-sections=@var{s1},@var{s2},...
11573 @opindex fsanitize-sections
11574 Sanitize global variables in selected user-defined sections. @var{si} may
11577 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
11578 @opindex fsanitize-recover
11579 @opindex fno-sanitize-recover
11580 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
11581 mentioned in comma-separated list of @var{opts}. Enabling this option
11582 for a sanitizer component causes it to attempt to continue
11583 running the program as if no error happened. This means multiple
11584 runtime errors can be reported in a single program run, and the exit
11585 code of the program may indicate success even when errors
11586 have been reported. The @option{-fno-sanitize-recover=} option
11587 can be used to alter
11588 this behavior: only the first detected error is reported
11589 and program then exits with a non-zero exit code.
11591 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
11592 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
11593 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
11594 @option{-fsanitize=bounds-strict},
11595 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
11596 For these sanitizers error recovery is turned on by default,
11597 except @option{-fsanitize=address}, for which this feature is experimental.
11598 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
11599 accepted, the former enables recovery for all sanitizers that support it,
11600 the latter disables recovery for all sanitizers that support it.
11602 Even if a recovery mode is turned on the compiler side, it needs to be also
11603 enabled on the runtime library side, otherwise the failures are still fatal.
11604 The runtime library defaults to @code{halt_on_error=0} for
11605 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
11606 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
11607 setting the @code{halt_on_error} flag in the corresponding environment variable.
11609 Syntax without an explicit @var{opts} parameter is deprecated. It is
11610 equivalent to specifying an @var{opts} list of:
11613 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
11616 @item -fsanitize-address-use-after-scope
11617 @opindex fsanitize-address-use-after-scope
11618 Enable sanitization of local variables to detect use-after-scope bugs.
11619 The option sets @option{-fstack-reuse} to @samp{none}.
11621 @item -fsanitize-undefined-trap-on-error
11622 @opindex fsanitize-undefined-trap-on-error
11623 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
11624 report undefined behavior using @code{__builtin_trap} rather than
11625 a @code{libubsan} library routine. The advantage of this is that the
11626 @code{libubsan} library is not needed and is not linked in, so this
11627 is usable even in freestanding environments.
11629 @item -fsanitize-coverage=trace-pc
11630 @opindex fsanitize-coverage=trace-pc
11631 Enable coverage-guided fuzzing code instrumentation.
11632 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
11634 @item -fsanitize-coverage=trace-cmp
11635 @opindex fsanitize-coverage=trace-cmp
11636 Enable dataflow guided fuzzing code instrumentation.
11637 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
11638 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
11639 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
11640 variable or @code{__sanitizer_cov_trace_const_cmp1},
11641 @code{__sanitizer_cov_trace_const_cmp2},
11642 @code{__sanitizer_cov_trace_const_cmp4} or
11643 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
11644 operand constant, @code{__sanitizer_cov_trace_cmpf} or
11645 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
11646 @code{__sanitizer_cov_trace_switch} for switch statements.
11648 @item -fbounds-check
11649 @opindex fbounds-check
11650 For front ends that support it, generate additional code to check that
11651 indices used to access arrays are within the declared range. This is
11652 currently only supported by the Fortran front end, where this option
11655 @item -fcheck-pointer-bounds
11656 @opindex fcheck-pointer-bounds
11657 @opindex fno-check-pointer-bounds
11658 @cindex Pointer Bounds Checker options
11659 Enable Pointer Bounds Checker instrumentation. Each memory reference
11660 is instrumented with checks of the pointer used for memory access against
11661 bounds associated with that pointer.
11664 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
11665 and @option{-mmpx} are required to enable this feature.
11666 MPX-based instrumentation requires
11667 a runtime library to enable MPX in hardware and handle bounds
11668 violation signals. By default when @option{-fcheck-pointer-bounds}
11669 and @option{-mmpx} options are used to link a program, the GCC driver
11670 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
11671 Bounds checking on calls to dynamic libraries requires a linker
11672 with @option{-z bndplt} support; if GCC was configured with a linker
11673 without support for this option (including the Gold linker and older
11674 versions of ld), a warning is given if you link with @option{-mmpx}
11675 without also specifying @option{-static}, since the overall effectiveness
11676 of the bounds checking protection is reduced.
11677 See also @option{-static-libmpxwrappers}.
11679 MPX-based instrumentation
11680 may be used for debugging and also may be included in production code
11681 to increase program security. Depending on usage, you may
11682 have different requirements for the runtime library. The current version
11683 of the MPX runtime library is more oriented for use as a debugging
11684 tool. MPX runtime library usage implies @option{-lpthread}. See
11685 also @option{-static-libmpx}. The runtime library behavior can be
11686 influenced using various @env{CHKP_RT_*} environment variables. See
11687 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
11690 Generated instrumentation may be controlled by various
11691 @option{-fchkp-*} options and by the @code{bnd_variable_size}
11692 structure field attribute (@pxref{Type Attributes}) and
11693 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
11694 (@pxref{Function Attributes}). GCC also provides a number of built-in
11695 functions for controlling the Pointer Bounds Checker. @xref{Pointer
11696 Bounds Checker builtins}, for more information.
11698 @item -fchkp-check-incomplete-type
11699 @opindex fchkp-check-incomplete-type
11700 @opindex fno-chkp-check-incomplete-type
11701 Generate pointer bounds checks for variables with incomplete type.
11702 Enabled by default.
11704 @item -fchkp-narrow-bounds
11705 @opindex fchkp-narrow-bounds
11706 @opindex fno-chkp-narrow-bounds
11707 Controls bounds used by Pointer Bounds Checker for pointers to object
11708 fields. If narrowing is enabled then field bounds are used. Otherwise
11709 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
11710 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
11712 @item -fchkp-first-field-has-own-bounds
11713 @opindex fchkp-first-field-has-own-bounds
11714 @opindex fno-chkp-first-field-has-own-bounds
11715 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
11716 first field in the structure. By default a pointer to the first field has
11717 the same bounds as a pointer to the whole structure.
11719 @item -fchkp-flexible-struct-trailing-arrays
11720 @opindex fchkp-flexible-struct-trailing-arrays
11721 @opindex fno-chkp-flexible-struct-trailing-arrays
11722 Forces Pointer Bounds Checker to treat all trailing arrays in structures as
11723 possibly flexible. By default only array fields with zero length or that are
11724 marked with attribute bnd_variable_size are treated as flexible.
11726 @item -fchkp-narrow-to-innermost-array
11727 @opindex fchkp-narrow-to-innermost-array
11728 @opindex fno-chkp-narrow-to-innermost-array
11729 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
11730 case of nested static array access. By default this option is disabled and
11731 bounds of the outermost array are used.
11733 @item -fchkp-optimize
11734 @opindex fchkp-optimize
11735 @opindex fno-chkp-optimize
11736 Enables Pointer Bounds Checker optimizations. Enabled by default at
11737 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11739 @item -fchkp-use-fast-string-functions
11740 @opindex fchkp-use-fast-string-functions
11741 @opindex fno-chkp-use-fast-string-functions
11742 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11743 by Pointer Bounds Checker. Disabled by default.
11745 @item -fchkp-use-nochk-string-functions
11746 @opindex fchkp-use-nochk-string-functions
11747 @opindex fno-chkp-use-nochk-string-functions
11748 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11749 by Pointer Bounds Checker. Disabled by default.
11751 @item -fchkp-use-static-bounds
11752 @opindex fchkp-use-static-bounds
11753 @opindex fno-chkp-use-static-bounds
11754 Allow Pointer Bounds Checker to generate static bounds holding
11755 bounds of static variables. Enabled by default.
11757 @item -fchkp-use-static-const-bounds
11758 @opindex fchkp-use-static-const-bounds
11759 @opindex fno-chkp-use-static-const-bounds
11760 Use statically-initialized bounds for constant bounds instead of
11761 generating them each time they are required. By default enabled when
11762 @option{-fchkp-use-static-bounds} is enabled.
11764 @item -fchkp-treat-zero-dynamic-size-as-infinite
11765 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11766 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11767 With this option, objects with incomplete type whose
11768 dynamically-obtained size is zero are treated as having infinite size
11769 instead by Pointer Bounds
11770 Checker. This option may be helpful if a program is linked with a library
11771 missing size information for some symbols. Disabled by default.
11773 @item -fchkp-check-read
11774 @opindex fchkp-check-read
11775 @opindex fno-chkp-check-read
11776 Instructs Pointer Bounds Checker to generate checks for all read
11777 accesses to memory. Enabled by default.
11779 @item -fchkp-check-write
11780 @opindex fchkp-check-write
11781 @opindex fno-chkp-check-write
11782 Instructs Pointer Bounds Checker to generate checks for all write
11783 accesses to memory. Enabled by default.
11785 @item -fchkp-store-bounds
11786 @opindex fchkp-store-bounds
11787 @opindex fno-chkp-store-bounds
11788 Instructs Pointer Bounds Checker to generate bounds stores for
11789 pointer writes. Enabled by default.
11791 @item -fchkp-instrument-calls
11792 @opindex fchkp-instrument-calls
11793 @opindex fno-chkp-instrument-calls
11794 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11795 Enabled by default.
11797 @item -fchkp-instrument-marked-only
11798 @opindex fchkp-instrument-marked-only
11799 @opindex fno-chkp-instrument-marked-only
11800 Instructs Pointer Bounds Checker to instrument only functions
11801 marked with the @code{bnd_instrument} attribute
11802 (@pxref{Function Attributes}). Disabled by default.
11804 @item -fchkp-use-wrappers
11805 @opindex fchkp-use-wrappers
11806 @opindex fno-chkp-use-wrappers
11807 Allows Pointer Bounds Checker to replace calls to built-in functions
11808 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11809 is used to link a program, the GCC driver automatically links
11810 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11811 Enabled by default.
11813 @item -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
11814 @opindex fcf-protection
11815 Enable code instrumentation of control-flow transfers to increase
11816 program security by checking that target addresses of control-flow
11817 transfer instructions (such as indirect function call, function return,
11818 indirect jump) are valid. This prevents diverting the flow of control
11819 to an unexpected target. This is intended to protect against such
11820 threats as Return-oriented Programming (ROP), and similarly
11821 call/jmp-oriented programming (COP/JOP).
11823 The value @code{branch} tells the compiler to implement checking of
11824 validity of control-flow transfer at the point of indirect branch
11825 instructions, i.e. call/jmp instructions. The value @code{return}
11826 implements checking of validity at the point of returning from a
11827 function. The value @code{full} is an alias for specifying both
11828 @code{branch} and @code{return}. The value @code{none} turns off
11831 You can also use the @code{nocf_check} attribute to identify
11832 which functions and calls should be skipped from instrumentation
11833 (@pxref{Function Attributes}).
11835 Currently the x86 GNU/Linux target provides an implementation based
11836 on Intel Control-flow Enforcement Technology (CET). Instrumentation
11837 for x86 is controlled by target-specific options @option{-mcet},
11838 @option{-mibt} and @option{-mshstk} (@pxref{x86 Options}).
11840 @item -fstack-protector
11841 @opindex fstack-protector
11842 Emit extra code to check for buffer overflows, such as stack smashing
11843 attacks. This is done by adding a guard variable to functions with
11844 vulnerable objects. This includes functions that call @code{alloca}, and
11845 functions with buffers larger than 8 bytes. The guards are initialized
11846 when a function is entered and then checked when the function exits.
11847 If a guard check fails, an error message is printed and the program exits.
11849 @item -fstack-protector-all
11850 @opindex fstack-protector-all
11851 Like @option{-fstack-protector} except that all functions are protected.
11853 @item -fstack-protector-strong
11854 @opindex fstack-protector-strong
11855 Like @option{-fstack-protector} but includes additional functions to
11856 be protected --- those that have local array definitions, or have
11857 references to local frame addresses.
11859 @item -fstack-protector-explicit
11860 @opindex fstack-protector-explicit
11861 Like @option{-fstack-protector} but only protects those functions which
11862 have the @code{stack_protect} attribute.
11864 @item -fstack-check
11865 @opindex fstack-check
11866 Generate code to verify that you do not go beyond the boundary of the
11867 stack. You should specify this flag if you are running in an
11868 environment with multiple threads, but you only rarely need to specify it in
11869 a single-threaded environment since stack overflow is automatically
11870 detected on nearly all systems if there is only one stack.
11872 Note that this switch does not actually cause checking to be done; the
11873 operating system or the language runtime must do that. The switch causes
11874 generation of code to ensure that they see the stack being extended.
11876 You can additionally specify a string parameter: @samp{no} means no
11877 checking, @samp{generic} means force the use of old-style checking,
11878 @samp{specific} means use the best checking method and is equivalent
11879 to bare @option{-fstack-check}.
11881 Old-style checking is a generic mechanism that requires no specific
11882 target support in the compiler but comes with the following drawbacks:
11886 Modified allocation strategy for large objects: they are always
11887 allocated dynamically if their size exceeds a fixed threshold. Note this
11888 may change the semantics of some code.
11891 Fixed limit on the size of the static frame of functions: when it is
11892 topped by a particular function, stack checking is not reliable and
11893 a warning is issued by the compiler.
11896 Inefficiency: because of both the modified allocation strategy and the
11897 generic implementation, code performance is hampered.
11900 Note that old-style stack checking is also the fallback method for
11901 @samp{specific} if no target support has been added in the compiler.
11903 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
11904 and stack overflows. @samp{specific} is an excellent choice when compiling
11905 Ada code. It is not generally sufficient to protect against stack-clash
11906 attacks. To protect against those you want @samp{-fstack-clash-protection}.
11908 @item -fstack-clash-protection
11909 @opindex fstack-clash-protection
11910 Generate code to prevent stack clash style attacks. When this option is
11911 enabled, the compiler will only allocate one page of stack space at a time
11912 and each page is accessed immediately after allocation. Thus, it prevents
11913 allocations from jumping over any stack guard page provided by the
11916 Most targets do not fully support stack clash protection. However, on
11917 those targets @option{-fstack-clash-protection} will protect dynamic stack
11918 allocations. @option{-fstack-clash-protection} may also provide limited
11919 protection for static stack allocations if the target supports
11920 @option{-fstack-check=specific}.
11922 @item -fstack-limit-register=@var{reg}
11923 @itemx -fstack-limit-symbol=@var{sym}
11924 @itemx -fno-stack-limit
11925 @opindex fstack-limit-register
11926 @opindex fstack-limit-symbol
11927 @opindex fno-stack-limit
11928 Generate code to ensure that the stack does not grow beyond a certain value,
11929 either the value of a register or the address of a symbol. If a larger
11930 stack is required, a signal is raised at run time. For most targets,
11931 the signal is raised before the stack overruns the boundary, so
11932 it is possible to catch the signal without taking special precautions.
11934 For instance, if the stack starts at absolute address @samp{0x80000000}
11935 and grows downwards, you can use the flags
11936 @option{-fstack-limit-symbol=__stack_limit} and
11937 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11938 of 128KB@. Note that this may only work with the GNU linker.
11940 You can locally override stack limit checking by using the
11941 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11943 @item -fsplit-stack
11944 @opindex fsplit-stack
11945 Generate code to automatically split the stack before it overflows.
11946 The resulting program has a discontiguous stack which can only
11947 overflow if the program is unable to allocate any more memory. This
11948 is most useful when running threaded programs, as it is no longer
11949 necessary to calculate a good stack size to use for each thread. This
11950 is currently only implemented for the x86 targets running
11953 When code compiled with @option{-fsplit-stack} calls code compiled
11954 without @option{-fsplit-stack}, there may not be much stack space
11955 available for the latter code to run. If compiling all code,
11956 including library code, with @option{-fsplit-stack} is not an option,
11957 then the linker can fix up these calls so that the code compiled
11958 without @option{-fsplit-stack} always has a large stack. Support for
11959 this is implemented in the gold linker in GNU binutils release 2.21
11962 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11963 @opindex fvtable-verify
11964 This option is only available when compiling C++ code.
11965 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11966 feature that verifies at run time, for every virtual call, that
11967 the vtable pointer through which the call is made is valid for the type of
11968 the object, and has not been corrupted or overwritten. If an invalid vtable
11969 pointer is detected at run time, an error is reported and execution of the
11970 program is immediately halted.
11972 This option causes run-time data structures to be built at program startup,
11973 which are used for verifying the vtable pointers.
11974 The options @samp{std} and @samp{preinit}
11975 control the timing of when these data structures are built. In both cases the
11976 data structures are built before execution reaches @code{main}. Using
11977 @option{-fvtable-verify=std} causes the data structures to be built after
11978 shared libraries have been loaded and initialized.
11979 @option{-fvtable-verify=preinit} causes them to be built before shared
11980 libraries have been loaded and initialized.
11982 If this option appears multiple times in the command line with different
11983 values specified, @samp{none} takes highest priority over both @samp{std} and
11984 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11987 @opindex fvtv-debug
11988 When used in conjunction with @option{-fvtable-verify=std} or
11989 @option{-fvtable-verify=preinit}, causes debug versions of the
11990 runtime functions for the vtable verification feature to be called.
11991 This flag also causes the compiler to log information about which
11992 vtable pointers it finds for each class.
11993 This information is written to a file named @file{vtv_set_ptr_data.log}
11994 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11995 if that is defined or the current working directory otherwise.
11997 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11998 file, be sure to delete any existing one.
12001 @opindex fvtv-counts
12002 This is a debugging flag. When used in conjunction with
12003 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
12004 causes the compiler to keep track of the total number of virtual calls
12005 it encounters and the number of verifications it inserts. It also
12006 counts the number of calls to certain run-time library functions
12007 that it inserts and logs this information for each compilation unit.
12008 The compiler writes this information to a file named
12009 @file{vtv_count_data.log} in the directory named by the environment
12010 variable @env{VTV_LOGS_DIR} if that is defined or the current working
12011 directory otherwise. It also counts the size of the vtable pointer sets
12012 for each class, and writes this information to @file{vtv_class_set_sizes.log}
12013 in the same directory.
12015 Note: This feature @emph{appends} data to the log files. To get fresh log
12016 files, be sure to delete any existing ones.
12018 @item -finstrument-functions
12019 @opindex finstrument-functions
12020 Generate instrumentation calls for entry and exit to functions. Just
12021 after function entry and just before function exit, the following
12022 profiling functions are called with the address of the current
12023 function and its call site. (On some platforms,
12024 @code{__builtin_return_address} does not work beyond the current
12025 function, so the call site information may not be available to the
12026 profiling functions otherwise.)
12029 void __cyg_profile_func_enter (void *this_fn,
12031 void __cyg_profile_func_exit (void *this_fn,
12035 The first argument is the address of the start of the current function,
12036 which may be looked up exactly in the symbol table.
12038 This instrumentation is also done for functions expanded inline in other
12039 functions. The profiling calls indicate where, conceptually, the
12040 inline function is entered and exited. This means that addressable
12041 versions of such functions must be available. If all your uses of a
12042 function are expanded inline, this may mean an additional expansion of
12043 code size. If you use @code{extern inline} in your C code, an
12044 addressable version of such functions must be provided. (This is
12045 normally the case anyway, but if you get lucky and the optimizer always
12046 expands the functions inline, you might have gotten away without
12047 providing static copies.)
12049 A function may be given the attribute @code{no_instrument_function}, in
12050 which case this instrumentation is not done. This can be used, for
12051 example, for the profiling functions listed above, high-priority
12052 interrupt routines, and any functions from which the profiling functions
12053 cannot safely be called (perhaps signal handlers, if the profiling
12054 routines generate output or allocate memory).
12056 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
12057 @opindex finstrument-functions-exclude-file-list
12059 Set the list of functions that are excluded from instrumentation (see
12060 the description of @option{-finstrument-functions}). If the file that
12061 contains a function definition matches with one of @var{file}, then
12062 that function is not instrumented. The match is done on substrings:
12063 if the @var{file} parameter is a substring of the file name, it is
12064 considered to be a match.
12069 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
12073 excludes any inline function defined in files whose pathnames
12074 contain @file{/bits/stl} or @file{include/sys}.
12076 If, for some reason, you want to include letter @samp{,} in one of
12077 @var{sym}, write @samp{\,}. For example,
12078 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
12079 (note the single quote surrounding the option).
12081 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
12082 @opindex finstrument-functions-exclude-function-list
12084 This is similar to @option{-finstrument-functions-exclude-file-list},
12085 but this option sets the list of function names to be excluded from
12086 instrumentation. The function name to be matched is its user-visible
12087 name, such as @code{vector<int> blah(const vector<int> &)}, not the
12088 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
12089 match is done on substrings: if the @var{sym} parameter is a substring
12090 of the function name, it is considered to be a match. For C99 and C++
12091 extended identifiers, the function name must be given in UTF-8, not
12092 using universal character names.
12094 @item -fpatchable-function-entry=@var{N}[,@var{M}]
12095 @opindex fpatchable-function-entry
12096 Generate @var{N} NOPs right at the beginning
12097 of each function, with the function entry point before the @var{M}th NOP.
12098 If @var{M} is omitted, it defaults to @code{0} so the
12099 function entry points to the address just at the first NOP.
12100 The NOP instructions reserve extra space which can be used to patch in
12101 any desired instrumentation at run time, provided that the code segment
12102 is writable. The amount of space is controllable indirectly via
12103 the number of NOPs; the NOP instruction used corresponds to the instruction
12104 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
12105 is target-specific and may also depend on the architecture variant and/or
12106 other compilation options.
12108 For run-time identification, the starting addresses of these areas,
12109 which correspond to their respective function entries minus @var{M},
12110 are additionally collected in the @code{__patchable_function_entries}
12111 section of the resulting binary.
12113 Note that the value of @code{__attribute__ ((patchable_function_entry
12114 (N,M)))} takes precedence over command-line option
12115 @option{-fpatchable-function-entry=N,M}. This can be used to increase
12116 the area size or to remove it completely on a single function.
12117 If @code{N=0}, no pad location is recorded.
12119 The NOP instructions are inserted at---and maybe before, depending on
12120 @var{M}---the function entry address, even before the prologue.
12125 @node Preprocessor Options
12126 @section Options Controlling the Preprocessor
12127 @cindex preprocessor options
12128 @cindex options, preprocessor
12130 These options control the C preprocessor, which is run on each C source
12131 file before actual compilation.
12133 If you use the @option{-E} option, nothing is done except preprocessing.
12134 Some of these options make sense only together with @option{-E} because
12135 they cause the preprocessor output to be unsuitable for actual
12138 In addition to the options listed here, there are a number of options
12139 to control search paths for include files documented in
12140 @ref{Directory Options}.
12141 Options to control preprocessor diagnostics are listed in
12142 @ref{Warning Options}.
12145 @include cppopts.texi
12147 @item -Wp,@var{option}
12149 You can use @option{-Wp,@var{option}} to bypass the compiler driver
12150 and pass @var{option} directly through to the preprocessor. If
12151 @var{option} contains commas, it is split into multiple options at the
12152 commas. However, many options are modified, translated or interpreted
12153 by the compiler driver before being passed to the preprocessor, and
12154 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
12155 interface is undocumented and subject to change, so whenever possible
12156 you should avoid using @option{-Wp} and let the driver handle the
12159 @item -Xpreprocessor @var{option}
12160 @opindex Xpreprocessor
12161 Pass @var{option} as an option to the preprocessor. You can use this to
12162 supply system-specific preprocessor options that GCC does not
12165 If you want to pass an option that takes an argument, you must use
12166 @option{-Xpreprocessor} twice, once for the option and once for the argument.
12168 @item -no-integrated-cpp
12169 @opindex no-integrated-cpp
12170 Perform preprocessing as a separate pass before compilation.
12171 By default, GCC performs preprocessing as an integrated part of
12172 input tokenization and parsing.
12173 If this option is provided, the appropriate language front end
12174 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
12175 and Objective-C, respectively) is instead invoked twice,
12176 once for preprocessing only and once for actual compilation
12177 of the preprocessed input.
12178 This option may be useful in conjunction with the @option{-B} or
12179 @option{-wrapper} options to specify an alternate preprocessor or
12180 perform additional processing of the program source between
12181 normal preprocessing and compilation.
12185 @node Assembler Options
12186 @section Passing Options to the Assembler
12188 @c prevent bad page break with this line
12189 You can pass options to the assembler.
12192 @item -Wa,@var{option}
12194 Pass @var{option} as an option to the assembler. If @var{option}
12195 contains commas, it is split into multiple options at the commas.
12197 @item -Xassembler @var{option}
12198 @opindex Xassembler
12199 Pass @var{option} as an option to the assembler. You can use this to
12200 supply system-specific assembler options that GCC does not
12203 If you want to pass an option that takes an argument, you must use
12204 @option{-Xassembler} twice, once for the option and once for the argument.
12209 @section Options for Linking
12210 @cindex link options
12211 @cindex options, linking
12213 These options come into play when the compiler links object files into
12214 an executable output file. They are meaningless if the compiler is
12215 not doing a link step.
12219 @item @var{object-file-name}
12220 A file name that does not end in a special recognized suffix is
12221 considered to name an object file or library. (Object files are
12222 distinguished from libraries by the linker according to the file
12223 contents.) If linking is done, these object files are used as input
12232 If any of these options is used, then the linker is not run, and
12233 object file names should not be used as arguments. @xref{Overall
12237 @opindex fuse-ld=bfd
12238 Use the @command{bfd} linker instead of the default linker.
12240 @item -fuse-ld=gold
12241 @opindex fuse-ld=gold
12242 Use the @command{gold} linker instead of the default linker.
12245 @item -l@var{library}
12246 @itemx -l @var{library}
12248 Search the library named @var{library} when linking. (The second
12249 alternative with the library as a separate argument is only for
12250 POSIX compliance and is not recommended.)
12252 It makes a difference where in the command you write this option; the
12253 linker searches and processes libraries and object files in the order they
12254 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
12255 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
12256 to functions in @samp{z}, those functions may not be loaded.
12258 The linker searches a standard list of directories for the library,
12259 which is actually a file named @file{lib@var{library}.a}. The linker
12260 then uses this file as if it had been specified precisely by name.
12262 The directories searched include several standard system directories
12263 plus any that you specify with @option{-L}.
12265 Normally the files found this way are library files---archive files
12266 whose members are object files. The linker handles an archive file by
12267 scanning through it for members which define symbols that have so far
12268 been referenced but not defined. But if the file that is found is an
12269 ordinary object file, it is linked in the usual fashion. The only
12270 difference between using an @option{-l} option and specifying a file name
12271 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
12272 and searches several directories.
12276 You need this special case of the @option{-l} option in order to
12277 link an Objective-C or Objective-C++ program.
12279 @item -nostartfiles
12280 @opindex nostartfiles
12281 Do not use the standard system startup files when linking.
12282 The standard system libraries are used normally, unless @option{-nostdlib}
12283 or @option{-nodefaultlibs} is used.
12285 @item -nodefaultlibs
12286 @opindex nodefaultlibs
12287 Do not use the standard system libraries when linking.
12288 Only the libraries you specify are passed to the linker, and options
12289 specifying linkage of the system libraries, such as @option{-static-libgcc}
12290 or @option{-shared-libgcc}, are ignored.
12291 The standard startup files are used normally, unless @option{-nostartfiles}
12294 The compiler may generate calls to @code{memcmp},
12295 @code{memset}, @code{memcpy} and @code{memmove}.
12296 These entries are usually resolved by entries in
12297 libc. These entry points should be supplied through some other
12298 mechanism when this option is specified.
12302 Do not use the standard system startup files or libraries when linking.
12303 No startup files and only the libraries you specify are passed to
12304 the linker, and options specifying linkage of the system libraries, such as
12305 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
12307 The compiler may generate calls to @code{memcmp}, @code{memset},
12308 @code{memcpy} and @code{memmove}.
12309 These entries are usually resolved by entries in
12310 libc. These entry points should be supplied through some other
12311 mechanism when this option is specified.
12313 @cindex @option{-lgcc}, use with @option{-nostdlib}
12314 @cindex @option{-nostdlib} and unresolved references
12315 @cindex unresolved references and @option{-nostdlib}
12316 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
12317 @cindex @option{-nodefaultlibs} and unresolved references
12318 @cindex unresolved references and @option{-nodefaultlibs}
12319 One of the standard libraries bypassed by @option{-nostdlib} and
12320 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
12321 which GCC uses to overcome shortcomings of particular machines, or special
12322 needs for some languages.
12323 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
12324 Collection (GCC) Internals},
12325 for more discussion of @file{libgcc.a}.)
12326 In most cases, you need @file{libgcc.a} even when you want to avoid
12327 other standard libraries. In other words, when you specify @option{-nostdlib}
12328 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
12329 This ensures that you have no unresolved references to internal GCC
12330 library subroutines.
12331 (An example of such an internal subroutine is @code{__main}, used to ensure C++
12332 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
12333 GNU Compiler Collection (GCC) Internals}.)
12337 Produce a dynamically linked position independent executable on targets
12338 that support it. For predictable results, you must also specify the same
12339 set of options used for compilation (@option{-fpie}, @option{-fPIE},
12340 or model suboptions) when you specify this linker option.
12344 Don't produce a dynamically linked position independent executable.
12347 @opindex static-pie
12348 Produce a static position independent executable on targets that support
12349 it. A static position independent executable is similar to a static
12350 executable, but can be loaded at any address without a dynamic linker.
12351 For predictable results, you must also specify the same set of options
12352 used for compilation (@option{-fpie}, @option{-fPIE}, or model
12353 suboptions) when you specify this linker option.
12357 Link with the POSIX threads library. This option is supported on
12358 GNU/Linux targets, most other Unix derivatives, and also on
12359 x86 Cygwin and MinGW targets. On some targets this option also sets
12360 flags for the preprocessor, so it should be used consistently for both
12361 compilation and linking.
12365 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
12366 that support it. This instructs the linker to add all symbols, not
12367 only used ones, to the dynamic symbol table. This option is needed
12368 for some uses of @code{dlopen} or to allow obtaining backtraces
12369 from within a program.
12373 Remove all symbol table and relocation information from the executable.
12377 On systems that support dynamic linking, this overrides @option{-pie}
12378 and prevents linking with the shared libraries. On other systems, this
12379 option has no effect.
12383 Produce a shared object which can then be linked with other objects to
12384 form an executable. Not all systems support this option. For predictable
12385 results, you must also specify the same set of options used for compilation
12386 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
12387 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
12388 needs to build supplementary stub code for constructors to work. On
12389 multi-libbed systems, @samp{gcc -shared} must select the correct support
12390 libraries to link against. Failing to supply the correct flags may lead
12391 to subtle defects. Supplying them in cases where they are not necessary
12394 @item -shared-libgcc
12395 @itemx -static-libgcc
12396 @opindex shared-libgcc
12397 @opindex static-libgcc
12398 On systems that provide @file{libgcc} as a shared library, these options
12399 force the use of either the shared or static version, respectively.
12400 If no shared version of @file{libgcc} was built when the compiler was
12401 configured, these options have no effect.
12403 There are several situations in which an application should use the
12404 shared @file{libgcc} instead of the static version. The most common
12405 of these is when the application wishes to throw and catch exceptions
12406 across different shared libraries. In that case, each of the libraries
12407 as well as the application itself should use the shared @file{libgcc}.
12409 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
12410 whenever you build a shared library or a main executable, because C++
12411 programs typically use exceptions, so this is the right thing to do.
12413 If, instead, you use the GCC driver to create shared libraries, you may
12414 find that they are not always linked with the shared @file{libgcc}.
12415 If GCC finds, at its configuration time, that you have a non-GNU linker
12416 or a GNU linker that does not support option @option{--eh-frame-hdr},
12417 it links the shared version of @file{libgcc} into shared libraries
12418 by default. Otherwise, it takes advantage of the linker and optimizes
12419 away the linking with the shared version of @file{libgcc}, linking with
12420 the static version of libgcc by default. This allows exceptions to
12421 propagate through such shared libraries, without incurring relocation
12422 costs at library load time.
12424 However, if a library or main executable is supposed to throw or catch
12425 exceptions, you must link it using the G++ driver, as appropriate
12426 for the languages used in the program, or using the option
12427 @option{-shared-libgcc}, such that it is linked with the shared
12430 @item -static-libasan
12431 @opindex static-libasan
12432 When the @option{-fsanitize=address} option is used to link a program,
12433 the GCC driver automatically links against @option{libasan}. If
12434 @file{libasan} is available as a shared library, and the @option{-static}
12435 option is not used, then this links against the shared version of
12436 @file{libasan}. The @option{-static-libasan} option directs the GCC
12437 driver to link @file{libasan} statically, without necessarily linking
12438 other libraries statically.
12440 @item -static-libtsan
12441 @opindex static-libtsan
12442 When the @option{-fsanitize=thread} option is used to link a program,
12443 the GCC driver automatically links against @option{libtsan}. If
12444 @file{libtsan} is available as a shared library, and the @option{-static}
12445 option is not used, then this links against the shared version of
12446 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
12447 driver to link @file{libtsan} statically, without necessarily linking
12448 other libraries statically.
12450 @item -static-liblsan
12451 @opindex static-liblsan
12452 When the @option{-fsanitize=leak} option is used to link a program,
12453 the GCC driver automatically links against @option{liblsan}. If
12454 @file{liblsan} is available as a shared library, and the @option{-static}
12455 option is not used, then this links against the shared version of
12456 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
12457 driver to link @file{liblsan} statically, without necessarily linking
12458 other libraries statically.
12460 @item -static-libubsan
12461 @opindex static-libubsan
12462 When the @option{-fsanitize=undefined} option is used to link a program,
12463 the GCC driver automatically links against @option{libubsan}. If
12464 @file{libubsan} is available as a shared library, and the @option{-static}
12465 option is not used, then this links against the shared version of
12466 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
12467 driver to link @file{libubsan} statically, without necessarily linking
12468 other libraries statically.
12470 @item -static-libmpx
12471 @opindex static-libmpx
12472 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
12473 used to link a program, the GCC driver automatically links against
12474 @file{libmpx}. If @file{libmpx} is available as a shared library,
12475 and the @option{-static} option is not used, then this links against
12476 the shared version of @file{libmpx}. The @option{-static-libmpx}
12477 option directs the GCC driver to link @file{libmpx} statically,
12478 without necessarily linking other libraries statically.
12480 @item -static-libmpxwrappers
12481 @opindex static-libmpxwrappers
12482 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
12483 to link a program without also using @option{-fno-chkp-use-wrappers}, the
12484 GCC driver automatically links against @file{libmpxwrappers}. If
12485 @file{libmpxwrappers} is available as a shared library, and the
12486 @option{-static} option is not used, then this links against the shared
12487 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
12488 option directs the GCC driver to link @file{libmpxwrappers} statically,
12489 without necessarily linking other libraries statically.
12491 @item -static-libstdc++
12492 @opindex static-libstdc++
12493 When the @command{g++} program is used to link a C++ program, it
12494 normally automatically links against @option{libstdc++}. If
12495 @file{libstdc++} is available as a shared library, and the
12496 @option{-static} option is not used, then this links against the
12497 shared version of @file{libstdc++}. That is normally fine. However, it
12498 is sometimes useful to freeze the version of @file{libstdc++} used by
12499 the program without going all the way to a fully static link. The
12500 @option{-static-libstdc++} option directs the @command{g++} driver to
12501 link @file{libstdc++} statically, without necessarily linking other
12502 libraries statically.
12506 Bind references to global symbols when building a shared object. Warn
12507 about any unresolved references (unless overridden by the link editor
12508 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
12511 @item -T @var{script}
12513 @cindex linker script
12514 Use @var{script} as the linker script. This option is supported by most
12515 systems using the GNU linker. On some targets, such as bare-board
12516 targets without an operating system, the @option{-T} option may be required
12517 when linking to avoid references to undefined symbols.
12519 @item -Xlinker @var{option}
12521 Pass @var{option} as an option to the linker. You can use this to
12522 supply system-specific linker options that GCC does not recognize.
12524 If you want to pass an option that takes a separate argument, you must use
12525 @option{-Xlinker} twice, once for the option and once for the argument.
12526 For example, to pass @option{-assert definitions}, you must write
12527 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
12528 @option{-Xlinker "-assert definitions"}, because this passes the entire
12529 string as a single argument, which is not what the linker expects.
12531 When using the GNU linker, it is usually more convenient to pass
12532 arguments to linker options using the @option{@var{option}=@var{value}}
12533 syntax than as separate arguments. For example, you can specify
12534 @option{-Xlinker -Map=output.map} rather than
12535 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
12536 this syntax for command-line options.
12538 @item -Wl,@var{option}
12540 Pass @var{option} as an option to the linker. If @var{option} contains
12541 commas, it is split into multiple options at the commas. You can use this
12542 syntax to pass an argument to the option.
12543 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
12544 linker. When using the GNU linker, you can also get the same effect with
12545 @option{-Wl,-Map=output.map}.
12547 @item -u @var{symbol}
12549 Pretend the symbol @var{symbol} is undefined, to force linking of
12550 library modules to define it. You can use @option{-u} multiple times with
12551 different symbols to force loading of additional library modules.
12553 @item -z @var{keyword}
12555 @option{-z} is passed directly on to the linker along with the keyword
12556 @var{keyword}. See the section in the documentation of your linker for
12557 permitted values and their meanings.
12560 @node Directory Options
12561 @section Options for Directory Search
12562 @cindex directory options
12563 @cindex options, directory search
12564 @cindex search path
12566 These options specify directories to search for header files, for
12567 libraries and for parts of the compiler:
12570 @include cppdiropts.texi
12572 @item -iplugindir=@var{dir}
12573 @opindex iplugindir=
12574 Set the directory to search for plugins that are passed
12575 by @option{-fplugin=@var{name}} instead of
12576 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
12577 to be used by the user, but only passed by the driver.
12581 Add directory @var{dir} to the list of directories to be searched
12584 @item -B@var{prefix}
12586 This option specifies where to find the executables, libraries,
12587 include files, and data files of the compiler itself.
12589 The compiler driver program runs one or more of the subprograms
12590 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
12591 @var{prefix} as a prefix for each program it tries to run, both with and
12592 without @samp{@var{machine}/@var{version}/} for the corresponding target
12593 machine and compiler version.
12595 For each subprogram to be run, the compiler driver first tries the
12596 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
12597 is not specified, the driver tries two standard prefixes,
12598 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
12599 those results in a file name that is found, the unmodified program
12600 name is searched for using the directories specified in your
12601 @env{PATH} environment variable.
12603 The compiler checks to see if the path provided by @option{-B}
12604 refers to a directory, and if necessary it adds a directory
12605 separator character at the end of the path.
12607 @option{-B} prefixes that effectively specify directory names also apply
12608 to libraries in the linker, because the compiler translates these
12609 options into @option{-L} options for the linker. They also apply to
12610 include files in the preprocessor, because the compiler translates these
12611 options into @option{-isystem} options for the preprocessor. In this case,
12612 the compiler appends @samp{include} to the prefix.
12614 The runtime support file @file{libgcc.a} can also be searched for using
12615 the @option{-B} prefix, if needed. If it is not found there, the two
12616 standard prefixes above are tried, and that is all. The file is left
12617 out of the link if it is not found by those means.
12619 Another way to specify a prefix much like the @option{-B} prefix is to use
12620 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
12623 As a special kludge, if the path provided by @option{-B} is
12624 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
12625 9, then it is replaced by @file{[dir/]include}. This is to help
12626 with boot-strapping the compiler.
12628 @item -no-canonical-prefixes
12629 @opindex no-canonical-prefixes
12630 Do not expand any symbolic links, resolve references to @samp{/../}
12631 or @samp{/./}, or make the path absolute when generating a relative
12634 @item --sysroot=@var{dir}
12636 Use @var{dir} as the logical root directory for headers and libraries.
12637 For example, if the compiler normally searches for headers in
12638 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
12639 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
12641 If you use both this option and the @option{-isysroot} option, then
12642 the @option{--sysroot} option applies to libraries, but the
12643 @option{-isysroot} option applies to header files.
12645 The GNU linker (beginning with version 2.16) has the necessary support
12646 for this option. If your linker does not support this option, the
12647 header file aspect of @option{--sysroot} still works, but the
12648 library aspect does not.
12650 @item --no-sysroot-suffix
12651 @opindex no-sysroot-suffix
12652 For some targets, a suffix is added to the root directory specified
12653 with @option{--sysroot}, depending on the other options used, so that
12654 headers may for example be found in
12655 @file{@var{dir}/@var{suffix}/usr/include} instead of
12656 @file{@var{dir}/usr/include}. This option disables the addition of
12661 @node Code Gen Options
12662 @section Options for Code Generation Conventions
12663 @cindex code generation conventions
12664 @cindex options, code generation
12665 @cindex run-time options
12667 These machine-independent options control the interface conventions
12668 used in code generation.
12670 Most of them have both positive and negative forms; the negative form
12671 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
12672 one of the forms is listed---the one that is not the default. You
12673 can figure out the other form by either removing @samp{no-} or adding
12677 @item -fstack-reuse=@var{reuse-level}
12678 @opindex fstack_reuse
12679 This option controls stack space reuse for user declared local/auto variables
12680 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
12681 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
12682 local variables and temporaries, @samp{named_vars} enables the reuse only for
12683 user defined local variables with names, and @samp{none} disables stack reuse
12684 completely. The default value is @samp{all}. The option is needed when the
12685 program extends the lifetime of a scoped local variable or a compiler generated
12686 temporary beyond the end point defined by the language. When a lifetime of
12687 a variable ends, and if the variable lives in memory, the optimizing compiler
12688 has the freedom to reuse its stack space with other temporaries or scoped
12689 local variables whose live range does not overlap with it. Legacy code extending
12690 local lifetime is likely to break with the stack reuse optimization.
12709 if (*p == 10) // out of scope use of local1
12720 A(int k) : i(k), j(k) @{ @}
12727 void foo(const A& ar)
12734 foo(A(10)); // temp object's lifetime ends when foo returns
12740 ap->i+= 10; // ap references out of scope temp whose space
12741 // is reused with a. What is the value of ap->i?
12746 The lifetime of a compiler generated temporary is well defined by the C++
12747 standard. When a lifetime of a temporary ends, and if the temporary lives
12748 in memory, the optimizing compiler has the freedom to reuse its stack
12749 space with other temporaries or scoped local variables whose live range
12750 does not overlap with it. However some of the legacy code relies on
12751 the behavior of older compilers in which temporaries' stack space is
12752 not reused, the aggressive stack reuse can lead to runtime errors. This
12753 option is used to control the temporary stack reuse optimization.
12757 This option generates traps for signed overflow on addition, subtraction,
12758 multiplication operations.
12759 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12760 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12761 @option{-fwrapv} being effective. Note that only active options override, so
12762 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12763 results in @option{-ftrapv} being effective.
12767 This option instructs the compiler to assume that signed arithmetic
12768 overflow of addition, subtraction and multiplication wraps around
12769 using twos-complement representation. This flag enables some optimizations
12770 and disables others.
12771 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12772 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12773 @option{-fwrapv} being effective. Note that only active options override, so
12774 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12775 results in @option{-ftrapv} being effective.
12777 @item -fwrapv-pointer
12778 @opindex fwrapv-pointer
12779 This option instructs the compiler to assume that pointer arithmetic
12780 overflow on addition and subtraction wraps around using twos-complement
12781 representation. This flag disables some optimizations which assume
12782 pointer overflow is invalid.
12784 @item -fstrict-overflow
12785 @opindex fstrict-overflow
12786 This option implies @option{-fno-wrapv} @option{-fno-wrapv-pointer} and when
12787 negated implies @option{-fwrapv} @option{-fwrapv-pointer}.
12790 @opindex fexceptions
12791 Enable exception handling. Generates extra code needed to propagate
12792 exceptions. For some targets, this implies GCC generates frame
12793 unwind information for all functions, which can produce significant data
12794 size overhead, although it does not affect execution. If you do not
12795 specify this option, GCC enables it by default for languages like
12796 C++ that normally require exception handling, and disables it for
12797 languages like C that do not normally require it. However, you may need
12798 to enable this option when compiling C code that needs to interoperate
12799 properly with exception handlers written in C++. You may also wish to
12800 disable this option if you are compiling older C++ programs that don't
12801 use exception handling.
12803 @item -fnon-call-exceptions
12804 @opindex fnon-call-exceptions
12805 Generate code that allows trapping instructions to throw exceptions.
12806 Note that this requires platform-specific runtime support that does
12807 not exist everywhere. Moreover, it only allows @emph{trapping}
12808 instructions to throw exceptions, i.e.@: memory references or floating-point
12809 instructions. It does not allow exceptions to be thrown from
12810 arbitrary signal handlers such as @code{SIGALRM}.
12812 @item -fdelete-dead-exceptions
12813 @opindex fdelete-dead-exceptions
12814 Consider that instructions that may throw exceptions but don't otherwise
12815 contribute to the execution of the program can be optimized away.
12816 This option is enabled by default for the Ada front end, as permitted by
12817 the Ada language specification.
12818 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12820 @item -funwind-tables
12821 @opindex funwind-tables
12822 Similar to @option{-fexceptions}, except that it just generates any needed
12823 static data, but does not affect the generated code in any other way.
12824 You normally do not need to enable this option; instead, a language processor
12825 that needs this handling enables it on your behalf.
12827 @item -fasynchronous-unwind-tables
12828 @opindex fasynchronous-unwind-tables
12829 Generate unwind table in DWARF format, if supported by target machine. The
12830 table is exact at each instruction boundary, so it can be used for stack
12831 unwinding from asynchronous events (such as debugger or garbage collector).
12833 @item -fno-gnu-unique
12834 @opindex fno-gnu-unique
12835 On systems with recent GNU assembler and C library, the C++ compiler
12836 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12837 of template static data members and static local variables in inline
12838 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12839 is necessary to avoid problems with a library used by two different
12840 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12841 therefore disagreeing with the other one about the binding of the
12842 symbol. But this causes @code{dlclose} to be ignored for affected
12843 DSOs; if your program relies on reinitialization of a DSO via
12844 @code{dlclose} and @code{dlopen}, you can use
12845 @option{-fno-gnu-unique}.
12847 @item -fpcc-struct-return
12848 @opindex fpcc-struct-return
12849 Return ``short'' @code{struct} and @code{union} values in memory like
12850 longer ones, rather than in registers. This convention is less
12851 efficient, but it has the advantage of allowing intercallability between
12852 GCC-compiled files and files compiled with other compilers, particularly
12853 the Portable C Compiler (pcc).
12855 The precise convention for returning structures in memory depends
12856 on the target configuration macros.
12858 Short structures and unions are those whose size and alignment match
12859 that of some integer type.
12861 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12862 switch is not binary compatible with code compiled with the
12863 @option{-freg-struct-return} switch.
12864 Use it to conform to a non-default application binary interface.
12866 @item -freg-struct-return
12867 @opindex freg-struct-return
12868 Return @code{struct} and @code{union} values in registers when possible.
12869 This is more efficient for small structures than
12870 @option{-fpcc-struct-return}.
12872 If you specify neither @option{-fpcc-struct-return} nor
12873 @option{-freg-struct-return}, GCC defaults to whichever convention is
12874 standard for the target. If there is no standard convention, GCC
12875 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12876 the principal compiler. In those cases, we can choose the standard, and
12877 we chose the more efficient register return alternative.
12879 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12880 switch is not binary compatible with code compiled with the
12881 @option{-fpcc-struct-return} switch.
12882 Use it to conform to a non-default application binary interface.
12884 @item -fshort-enums
12885 @opindex fshort-enums
12886 Allocate to an @code{enum} type only as many bytes as it needs for the
12887 declared range of possible values. Specifically, the @code{enum} type
12888 is equivalent to the smallest integer type that has enough room.
12890 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12891 code that is not binary compatible with code generated without that switch.
12892 Use it to conform to a non-default application binary interface.
12894 @item -fshort-wchar
12895 @opindex fshort-wchar
12896 Override the underlying type for @code{wchar_t} to be @code{short
12897 unsigned int} instead of the default for the target. This option is
12898 useful for building programs to run under WINE@.
12900 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12901 code that is not binary compatible with code generated without that switch.
12902 Use it to conform to a non-default application binary interface.
12905 @opindex fno-common
12906 @cindex tentative definitions
12907 In C code, this option controls the placement of global variables
12908 defined without an initializer, known as @dfn{tentative definitions}
12909 in the C standard. Tentative definitions are distinct from declarations
12910 of a variable with the @code{extern} keyword, which do not allocate storage.
12912 Unix C compilers have traditionally allocated storage for
12913 uninitialized global variables in a common block. This allows the
12914 linker to resolve all tentative definitions of the same variable
12915 in different compilation units to the same object, or to a non-tentative
12917 This is the behavior specified by @option{-fcommon}, and is the default for
12918 GCC on most targets.
12919 On the other hand, this behavior is not required by ISO
12920 C, and on some targets may carry a speed or code size penalty on
12921 variable references.
12923 The @option{-fno-common} option specifies that the compiler should instead
12924 place uninitialized global variables in the data section of the object file.
12925 This inhibits the merging of tentative definitions by the linker so
12926 you get a multiple-definition error if the same
12927 variable is defined in more than one compilation unit.
12928 Compiling with @option{-fno-common} is useful on targets for which
12929 it provides better performance, or if you wish to verify that the
12930 program will work on other systems that always treat uninitialized
12931 variable definitions this way.
12935 Ignore the @code{#ident} directive.
12937 @item -finhibit-size-directive
12938 @opindex finhibit-size-directive
12939 Don't output a @code{.size} assembler directive, or anything else that
12940 would cause trouble if the function is split in the middle, and the
12941 two halves are placed at locations far apart in memory. This option is
12942 used when compiling @file{crtstuff.c}; you should not need to use it
12945 @item -fverbose-asm
12946 @opindex fverbose-asm
12947 Put extra commentary information in the generated assembly code to
12948 make it more readable. This option is generally only of use to those
12949 who actually need to read the generated assembly code (perhaps while
12950 debugging the compiler itself).
12952 @option{-fno-verbose-asm}, the default, causes the
12953 extra information to be omitted and is useful when comparing two assembler
12956 The added comments include:
12961 information on the compiler version and command-line options,
12964 the source code lines associated with the assembly instructions,
12965 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12968 hints on which high-level expressions correspond to
12969 the various assembly instruction operands.
12973 For example, given this C source file:
12981 for (i = 0; i < n; i++)
12988 compiling to (x86_64) assembly via @option{-S} and emitting the result
12989 direct to stdout via @option{-o} @option{-}
12992 gcc -S test.c -fverbose-asm -Os -o -
12995 gives output similar to this:
12999 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
13006 .type test, @@function
13010 # test.c:4: int total = 0;
13011 xorl %eax, %eax # <retval>
13012 # test.c:6: for (i = 0; i < n; i++)
13013 xorl %edx, %edx # i
13015 # test.c:6: for (i = 0; i < n; i++)
13016 cmpl %edi, %edx # n, i
13018 # test.c:7: total += i * i;
13019 movl %edx, %ecx # i, tmp92
13020 imull %edx, %ecx # i, tmp92
13021 # test.c:6: for (i = 0; i < n; i++)
13023 # test.c:7: total += i * i;
13024 addl %ecx, %eax # tmp92, <retval>
13032 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
13033 .section .note.GNU-stack,"",@@progbits
13036 The comments are intended for humans rather than machines and hence the
13037 precise format of the comments is subject to change.
13039 @item -frecord-gcc-switches
13040 @opindex frecord-gcc-switches
13041 This switch causes the command line used to invoke the
13042 compiler to be recorded into the object file that is being created.
13043 This switch is only implemented on some targets and the exact format
13044 of the recording is target and binary file format dependent, but it
13045 usually takes the form of a section containing ASCII text. This
13046 switch is related to the @option{-fverbose-asm} switch, but that
13047 switch only records information in the assembler output file as
13048 comments, so it never reaches the object file.
13049 See also @option{-grecord-gcc-switches} for another
13050 way of storing compiler options into the object file.
13054 @cindex global offset table
13056 Generate position-independent code (PIC) suitable for use in a shared
13057 library, if supported for the target machine. Such code accesses all
13058 constant addresses through a global offset table (GOT)@. The dynamic
13059 loader resolves the GOT entries when the program starts (the dynamic
13060 loader is not part of GCC; it is part of the operating system). If
13061 the GOT size for the linked executable exceeds a machine-specific
13062 maximum size, you get an error message from the linker indicating that
13063 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13064 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
13065 on the m68k and RS/6000. The x86 has no such limit.)
13067 Position-independent code requires special support, and therefore works
13068 only on certain machines. For the x86, GCC supports PIC for System V
13069 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13070 position-independent.
13072 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13077 If supported for the target machine, emit position-independent code,
13078 suitable for dynamic linking and avoiding any limit on the size of the
13079 global offset table. This option makes a difference on AArch64, m68k,
13080 PowerPC and SPARC@.
13082 Position-independent code requires special support, and therefore works
13083 only on certain machines.
13085 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13092 These options are similar to @option{-fpic} and @option{-fPIC}, but
13093 generated position independent code can be only linked into executables.
13094 Usually these options are used when @option{-pie} GCC option is
13095 used during linking.
13097 @option{-fpie} and @option{-fPIE} both define the macros
13098 @code{__pie__} and @code{__PIE__}. The macros have the value 1
13099 for @option{-fpie} and 2 for @option{-fPIE}.
13103 Do not use the PLT for external function calls in position-independent code.
13104 Instead, load the callee address at call sites from the GOT and branch to it.
13105 This leads to more efficient code by eliminating PLT stubs and exposing
13106 GOT loads to optimizations. On architectures such as 32-bit x86 where
13107 PLT stubs expect the GOT pointer in a specific register, this gives more
13108 register allocation freedom to the compiler.
13109 Lazy binding requires use of the PLT;
13110 with @option{-fno-plt} all external symbols are resolved at load time.
13112 Alternatively, the function attribute @code{noplt} can be used to avoid calls
13113 through the PLT for specific external functions.
13115 In position-dependent code, a few targets also convert calls to
13116 functions that are marked to not use the PLT to use the GOT instead.
13118 @item -fno-jump-tables
13119 @opindex fno-jump-tables
13120 Do not use jump tables for switch statements even where it would be
13121 more efficient than other code generation strategies. This option is
13122 of use in conjunction with @option{-fpic} or @option{-fPIC} for
13123 building code that forms part of a dynamic linker and cannot
13124 reference the address of a jump table. On some targets, jump tables
13125 do not require a GOT and this option is not needed.
13127 @item -ffixed-@var{reg}
13129 Treat the register named @var{reg} as a fixed register; generated code
13130 should never refer to it (except perhaps as a stack pointer, frame
13131 pointer or in some other fixed role).
13133 @var{reg} must be the name of a register. The register names accepted
13134 are machine-specific and are defined in the @code{REGISTER_NAMES}
13135 macro in the machine description macro file.
13137 This flag does not have a negative form, because it specifies a
13140 @item -fcall-used-@var{reg}
13141 @opindex fcall-used
13142 Treat the register named @var{reg} as an allocable register that is
13143 clobbered by function calls. It may be allocated for temporaries or
13144 variables that do not live across a call. Functions compiled this way
13145 do not save and restore the register @var{reg}.
13147 It is an error to use this flag with the frame pointer or stack pointer.
13148 Use of this flag for other registers that have fixed pervasive roles in
13149 the machine's execution model produces disastrous results.
13151 This flag does not have a negative form, because it specifies a
13154 @item -fcall-saved-@var{reg}
13155 @opindex fcall-saved
13156 Treat the register named @var{reg} as an allocable register saved by
13157 functions. It may be allocated even for temporaries or variables that
13158 live across a call. Functions compiled this way save and restore
13159 the register @var{reg} if they use it.
13161 It is an error to use this flag with the frame pointer or stack pointer.
13162 Use of this flag for other registers that have fixed pervasive roles in
13163 the machine's execution model produces disastrous results.
13165 A different sort of disaster results from the use of this flag for
13166 a register in which function values may be returned.
13168 This flag does not have a negative form, because it specifies a
13171 @item -fpack-struct[=@var{n}]
13172 @opindex fpack-struct
13173 Without a value specified, pack all structure members together without
13174 holes. When a value is specified (which must be a small power of two), pack
13175 structure members according to this value, representing the maximum
13176 alignment (that is, objects with default alignment requirements larger than
13177 this are output potentially unaligned at the next fitting location.
13179 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13180 code that is not binary compatible with code generated without that switch.
13181 Additionally, it makes the code suboptimal.
13182 Use it to conform to a non-default application binary interface.
13184 @item -fleading-underscore
13185 @opindex fleading-underscore
13186 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13187 change the way C symbols are represented in the object file. One use
13188 is to help link with legacy assembly code.
13190 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13191 generate code that is not binary compatible with code generated without that
13192 switch. Use it to conform to a non-default application binary interface.
13193 Not all targets provide complete support for this switch.
13195 @item -ftls-model=@var{model}
13196 @opindex ftls-model
13197 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13198 The @var{model} argument should be one of @samp{global-dynamic},
13199 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
13200 Note that the choice is subject to optimization: the compiler may use
13201 a more efficient model for symbols not visible outside of the translation
13202 unit, or if @option{-fpic} is not given on the command line.
13204 The default without @option{-fpic} is @samp{initial-exec}; with
13205 @option{-fpic} the default is @samp{global-dynamic}.
13207 @item -ftrampolines
13208 @opindex ftrampolines
13209 For targets that normally need trampolines for nested functions, always
13210 generate them instead of using descriptors. Otherwise, for targets that
13211 do not need them, like for example HP-PA or IA-64, do nothing.
13213 A trampoline is a small piece of code that is created at run time on the
13214 stack when the address of a nested function is taken, and is used to call
13215 the nested function indirectly. Therefore, it requires the stack to be
13216 made executable in order for the program to work properly.
13218 @option{-fno-trampolines} is enabled by default on a language by language
13219 basis to let the compiler avoid generating them, if it computes that this
13220 is safe, and replace them with descriptors. Descriptors are made up of data
13221 only, but the generated code must be prepared to deal with them. As of this
13222 writing, @option{-fno-trampolines} is enabled by default only for Ada.
13224 Moreover, code compiled with @option{-ftrampolines} and code compiled with
13225 @option{-fno-trampolines} are not binary compatible if nested functions are
13226 present. This option must therefore be used on a program-wide basis and be
13227 manipulated with extreme care.
13229 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
13230 @opindex fvisibility
13231 Set the default ELF image symbol visibility to the specified option---all
13232 symbols are marked with this unless overridden within the code.
13233 Using this feature can very substantially improve linking and
13234 load times of shared object libraries, produce more optimized
13235 code, provide near-perfect API export and prevent symbol clashes.
13236 It is @strong{strongly} recommended that you use this in any shared objects
13239 Despite the nomenclature, @samp{default} always means public; i.e.,
13240 available to be linked against from outside the shared object.
13241 @samp{protected} and @samp{internal} are pretty useless in real-world
13242 usage so the only other commonly used option is @samp{hidden}.
13243 The default if @option{-fvisibility} isn't specified is
13244 @samp{default}, i.e., make every symbol public.
13246 A good explanation of the benefits offered by ensuring ELF
13247 symbols have the correct visibility is given by ``How To Write
13248 Shared Libraries'' by Ulrich Drepper (which can be found at
13249 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
13250 solution made possible by this option to marking things hidden when
13251 the default is public is to make the default hidden and mark things
13252 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
13253 and @code{__attribute__ ((visibility("default")))} instead of
13254 @code{__declspec(dllexport)} you get almost identical semantics with
13255 identical syntax. This is a great boon to those working with
13256 cross-platform projects.
13258 For those adding visibility support to existing code, you may find
13259 @code{#pragma GCC visibility} of use. This works by you enclosing
13260 the declarations you wish to set visibility for with (for example)
13261 @code{#pragma GCC visibility push(hidden)} and
13262 @code{#pragma GCC visibility pop}.
13263 Bear in mind that symbol visibility should be viewed @strong{as
13264 part of the API interface contract} and thus all new code should
13265 always specify visibility when it is not the default; i.e., declarations
13266 only for use within the local DSO should @strong{always} be marked explicitly
13267 as hidden as so to avoid PLT indirection overheads---making this
13268 abundantly clear also aids readability and self-documentation of the code.
13269 Note that due to ISO C++ specification requirements, @code{operator new} and
13270 @code{operator delete} must always be of default visibility.
13272 Be aware that headers from outside your project, in particular system
13273 headers and headers from any other library you use, may not be
13274 expecting to be compiled with visibility other than the default. You
13275 may need to explicitly say @code{#pragma GCC visibility push(default)}
13276 before including any such headers.
13278 @code{extern} declarations are not affected by @option{-fvisibility}, so
13279 a lot of code can be recompiled with @option{-fvisibility=hidden} with
13280 no modifications. However, this means that calls to @code{extern}
13281 functions with no explicit visibility use the PLT, so it is more
13282 effective to use @code{__attribute ((visibility))} and/or
13283 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
13284 declarations should be treated as hidden.
13286 Note that @option{-fvisibility} does affect C++ vague linkage
13287 entities. This means that, for instance, an exception class that is
13288 be thrown between DSOs must be explicitly marked with default
13289 visibility so that the @samp{type_info} nodes are unified between
13292 An overview of these techniques, their benefits and how to use them
13293 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
13295 @item -fstrict-volatile-bitfields
13296 @opindex fstrict-volatile-bitfields
13297 This option should be used if accesses to volatile bit-fields (or other
13298 structure fields, although the compiler usually honors those types
13299 anyway) should use a single access of the width of the
13300 field's type, aligned to a natural alignment if possible. For
13301 example, targets with memory-mapped peripheral registers might require
13302 all such accesses to be 16 bits wide; with this flag you can
13303 declare all peripheral bit-fields as @code{unsigned short} (assuming short
13304 is 16 bits on these targets) to force GCC to use 16-bit accesses
13305 instead of, perhaps, a more efficient 32-bit access.
13307 If this option is disabled, the compiler uses the most efficient
13308 instruction. In the previous example, that might be a 32-bit load
13309 instruction, even though that accesses bytes that do not contain
13310 any portion of the bit-field, or memory-mapped registers unrelated to
13311 the one being updated.
13313 In some cases, such as when the @code{packed} attribute is applied to a
13314 structure field, it may not be possible to access the field with a single
13315 read or write that is correctly aligned for the target machine. In this
13316 case GCC falls back to generating multiple accesses rather than code that
13317 will fault or truncate the result at run time.
13319 Note: Due to restrictions of the C/C++11 memory model, write accesses are
13320 not allowed to touch non bit-field members. It is therefore recommended
13321 to define all bits of the field's type as bit-field members.
13323 The default value of this option is determined by the application binary
13324 interface for the target processor.
13326 @item -fsync-libcalls
13327 @opindex fsync-libcalls
13328 This option controls whether any out-of-line instance of the @code{__sync}
13329 family of functions may be used to implement the C++11 @code{__atomic}
13330 family of functions.
13332 The default value of this option is enabled, thus the only useful form
13333 of the option is @option{-fno-sync-libcalls}. This option is used in
13334 the implementation of the @file{libatomic} runtime library.
13338 @node Developer Options
13339 @section GCC Developer Options
13340 @cindex developer options
13341 @cindex debugging GCC
13342 @cindex debug dump options
13343 @cindex dump options
13344 @cindex compilation statistics
13346 This section describes command-line options that are primarily of
13347 interest to GCC developers, including options to support compiler
13348 testing and investigation of compiler bugs and compile-time
13349 performance problems. This includes options that produce debug dumps
13350 at various points in the compilation; that print statistics such as
13351 memory use and execution time; and that print information about GCC's
13352 configuration, such as where it searches for libraries. You should
13353 rarely need to use any of these options for ordinary compilation and
13358 @item -d@var{letters}
13359 @itemx -fdump-rtl-@var{pass}
13360 @itemx -fdump-rtl-@var{pass}=@var{filename}
13362 @opindex fdump-rtl-@var{pass}
13363 Says to make debugging dumps during compilation at times specified by
13364 @var{letters}. This is used for debugging the RTL-based passes of the
13365 compiler. The file names for most of the dumps are made by appending
13366 a pass number and a word to the @var{dumpname}, and the files are
13367 created in the directory of the output file. In case of
13368 @option{=@var{filename}} option, the dump is output on the given file
13369 instead of the pass numbered dump files. Note that the pass number is
13370 assigned as passes are registered into the pass manager. Most passes
13371 are registered in the order that they will execute and for these passes
13372 the number corresponds to the pass execution order. However, passes
13373 registered by plugins, passes specific to compilation targets, or
13374 passes that are otherwise registered after all the other passes are
13375 numbered higher than a pass named "final", even if they are executed
13376 earlier. @var{dumpname} is generated from the name of the output
13377 file if explicitly specified and not an executable, otherwise it is
13378 the basename of the source file.
13380 Some @option{-d@var{letters}} switches have different meaning when
13381 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
13382 for information about preprocessor-specific dump options.
13384 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
13385 @option{-d} option @var{letters}. Here are the possible
13386 letters for use in @var{pass} and @var{letters}, and their meanings:
13390 @item -fdump-rtl-alignments
13391 @opindex fdump-rtl-alignments
13392 Dump after branch alignments have been computed.
13394 @item -fdump-rtl-asmcons
13395 @opindex fdump-rtl-asmcons
13396 Dump after fixing rtl statements that have unsatisfied in/out constraints.
13398 @item -fdump-rtl-auto_inc_dec
13399 @opindex fdump-rtl-auto_inc_dec
13400 Dump after auto-inc-dec discovery. This pass is only run on
13401 architectures that have auto inc or auto dec instructions.
13403 @item -fdump-rtl-barriers
13404 @opindex fdump-rtl-barriers
13405 Dump after cleaning up the barrier instructions.
13407 @item -fdump-rtl-bbpart
13408 @opindex fdump-rtl-bbpart
13409 Dump after partitioning hot and cold basic blocks.
13411 @item -fdump-rtl-bbro
13412 @opindex fdump-rtl-bbro
13413 Dump after block reordering.
13415 @item -fdump-rtl-btl1
13416 @itemx -fdump-rtl-btl2
13417 @opindex fdump-rtl-btl2
13418 @opindex fdump-rtl-btl2
13419 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
13420 after the two branch
13421 target load optimization passes.
13423 @item -fdump-rtl-bypass
13424 @opindex fdump-rtl-bypass
13425 Dump after jump bypassing and control flow optimizations.
13427 @item -fdump-rtl-combine
13428 @opindex fdump-rtl-combine
13429 Dump after the RTL instruction combination pass.
13431 @item -fdump-rtl-compgotos
13432 @opindex fdump-rtl-compgotos
13433 Dump after duplicating the computed gotos.
13435 @item -fdump-rtl-ce1
13436 @itemx -fdump-rtl-ce2
13437 @itemx -fdump-rtl-ce3
13438 @opindex fdump-rtl-ce1
13439 @opindex fdump-rtl-ce2
13440 @opindex fdump-rtl-ce3
13441 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
13442 @option{-fdump-rtl-ce3} enable dumping after the three
13443 if conversion passes.
13445 @item -fdump-rtl-cprop_hardreg
13446 @opindex fdump-rtl-cprop_hardreg
13447 Dump after hard register copy propagation.
13449 @item -fdump-rtl-csa
13450 @opindex fdump-rtl-csa
13451 Dump after combining stack adjustments.
13453 @item -fdump-rtl-cse1
13454 @itemx -fdump-rtl-cse2
13455 @opindex fdump-rtl-cse1
13456 @opindex fdump-rtl-cse2
13457 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
13458 the two common subexpression elimination passes.
13460 @item -fdump-rtl-dce
13461 @opindex fdump-rtl-dce
13462 Dump after the standalone dead code elimination passes.
13464 @item -fdump-rtl-dbr
13465 @opindex fdump-rtl-dbr
13466 Dump after delayed branch scheduling.
13468 @item -fdump-rtl-dce1
13469 @itemx -fdump-rtl-dce2
13470 @opindex fdump-rtl-dce1
13471 @opindex fdump-rtl-dce2
13472 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
13473 the two dead store elimination passes.
13475 @item -fdump-rtl-eh
13476 @opindex fdump-rtl-eh
13477 Dump after finalization of EH handling code.
13479 @item -fdump-rtl-eh_ranges
13480 @opindex fdump-rtl-eh_ranges
13481 Dump after conversion of EH handling range regions.
13483 @item -fdump-rtl-expand
13484 @opindex fdump-rtl-expand
13485 Dump after RTL generation.
13487 @item -fdump-rtl-fwprop1
13488 @itemx -fdump-rtl-fwprop2
13489 @opindex fdump-rtl-fwprop1
13490 @opindex fdump-rtl-fwprop2
13491 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
13492 dumping after the two forward propagation passes.
13494 @item -fdump-rtl-gcse1
13495 @itemx -fdump-rtl-gcse2
13496 @opindex fdump-rtl-gcse1
13497 @opindex fdump-rtl-gcse2
13498 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
13499 after global common subexpression elimination.
13501 @item -fdump-rtl-init-regs
13502 @opindex fdump-rtl-init-regs
13503 Dump after the initialization of the registers.
13505 @item -fdump-rtl-initvals
13506 @opindex fdump-rtl-initvals
13507 Dump after the computation of the initial value sets.
13509 @item -fdump-rtl-into_cfglayout
13510 @opindex fdump-rtl-into_cfglayout
13511 Dump after converting to cfglayout mode.
13513 @item -fdump-rtl-ira
13514 @opindex fdump-rtl-ira
13515 Dump after iterated register allocation.
13517 @item -fdump-rtl-jump
13518 @opindex fdump-rtl-jump
13519 Dump after the second jump optimization.
13521 @item -fdump-rtl-loop2
13522 @opindex fdump-rtl-loop2
13523 @option{-fdump-rtl-loop2} enables dumping after the rtl
13524 loop optimization passes.
13526 @item -fdump-rtl-mach
13527 @opindex fdump-rtl-mach
13528 Dump after performing the machine dependent reorganization pass, if that
13531 @item -fdump-rtl-mode_sw
13532 @opindex fdump-rtl-mode_sw
13533 Dump after removing redundant mode switches.
13535 @item -fdump-rtl-rnreg
13536 @opindex fdump-rtl-rnreg
13537 Dump after register renumbering.
13539 @item -fdump-rtl-outof_cfglayout
13540 @opindex fdump-rtl-outof_cfglayout
13541 Dump after converting from cfglayout mode.
13543 @item -fdump-rtl-peephole2
13544 @opindex fdump-rtl-peephole2
13545 Dump after the peephole pass.
13547 @item -fdump-rtl-postreload
13548 @opindex fdump-rtl-postreload
13549 Dump after post-reload optimizations.
13551 @item -fdump-rtl-pro_and_epilogue
13552 @opindex fdump-rtl-pro_and_epilogue
13553 Dump after generating the function prologues and epilogues.
13555 @item -fdump-rtl-sched1
13556 @itemx -fdump-rtl-sched2
13557 @opindex fdump-rtl-sched1
13558 @opindex fdump-rtl-sched2
13559 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
13560 after the basic block scheduling passes.
13562 @item -fdump-rtl-ree
13563 @opindex fdump-rtl-ree
13564 Dump after sign/zero extension elimination.
13566 @item -fdump-rtl-seqabstr
13567 @opindex fdump-rtl-seqabstr
13568 Dump after common sequence discovery.
13570 @item -fdump-rtl-shorten
13571 @opindex fdump-rtl-shorten
13572 Dump after shortening branches.
13574 @item -fdump-rtl-sibling
13575 @opindex fdump-rtl-sibling
13576 Dump after sibling call optimizations.
13578 @item -fdump-rtl-split1
13579 @itemx -fdump-rtl-split2
13580 @itemx -fdump-rtl-split3
13581 @itemx -fdump-rtl-split4
13582 @itemx -fdump-rtl-split5
13583 @opindex fdump-rtl-split1
13584 @opindex fdump-rtl-split2
13585 @opindex fdump-rtl-split3
13586 @opindex fdump-rtl-split4
13587 @opindex fdump-rtl-split5
13588 These options enable dumping after five rounds of
13589 instruction splitting.
13591 @item -fdump-rtl-sms
13592 @opindex fdump-rtl-sms
13593 Dump after modulo scheduling. This pass is only run on some
13596 @item -fdump-rtl-stack
13597 @opindex fdump-rtl-stack
13598 Dump after conversion from GCC's ``flat register file'' registers to the
13599 x87's stack-like registers. This pass is only run on x86 variants.
13601 @item -fdump-rtl-subreg1
13602 @itemx -fdump-rtl-subreg2
13603 @opindex fdump-rtl-subreg1
13604 @opindex fdump-rtl-subreg2
13605 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
13606 the two subreg expansion passes.
13608 @item -fdump-rtl-unshare
13609 @opindex fdump-rtl-unshare
13610 Dump after all rtl has been unshared.
13612 @item -fdump-rtl-vartrack
13613 @opindex fdump-rtl-vartrack
13614 Dump after variable tracking.
13616 @item -fdump-rtl-vregs
13617 @opindex fdump-rtl-vregs
13618 Dump after converting virtual registers to hard registers.
13620 @item -fdump-rtl-web
13621 @opindex fdump-rtl-web
13622 Dump after live range splitting.
13624 @item -fdump-rtl-regclass
13625 @itemx -fdump-rtl-subregs_of_mode_init
13626 @itemx -fdump-rtl-subregs_of_mode_finish
13627 @itemx -fdump-rtl-dfinit
13628 @itemx -fdump-rtl-dfinish
13629 @opindex fdump-rtl-regclass
13630 @opindex fdump-rtl-subregs_of_mode_init
13631 @opindex fdump-rtl-subregs_of_mode_finish
13632 @opindex fdump-rtl-dfinit
13633 @opindex fdump-rtl-dfinish
13634 These dumps are defined but always produce empty files.
13637 @itemx -fdump-rtl-all
13639 @opindex fdump-rtl-all
13640 Produce all the dumps listed above.
13644 Annotate the assembler output with miscellaneous debugging information.
13648 Dump all macro definitions, at the end of preprocessing, in addition to
13653 Produce a core dump whenever an error occurs.
13657 Annotate the assembler output with a comment indicating which
13658 pattern and alternative is used. The length and cost of each instruction are
13663 Dump the RTL in the assembler output as a comment before each instruction.
13664 Also turns on @option{-dp} annotation.
13668 Just generate RTL for a function instead of compiling it. Usually used
13669 with @option{-fdump-rtl-expand}.
13672 @item -fdump-noaddr
13673 @opindex fdump-noaddr
13674 When doing debugging dumps, suppress address output. This makes it more
13675 feasible to use diff on debugging dumps for compiler invocations with
13676 different compiler binaries and/or different
13677 text / bss / data / heap / stack / dso start locations.
13680 @opindex freport-bug
13681 Collect and dump debug information into a temporary file if an
13682 internal compiler error (ICE) occurs.
13684 @item -fdump-unnumbered
13685 @opindex fdump-unnumbered
13686 When doing debugging dumps, suppress instruction numbers and address output.
13687 This makes it more feasible to use diff on debugging dumps for compiler
13688 invocations with different options, in particular with and without
13691 @item -fdump-unnumbered-links
13692 @opindex fdump-unnumbered-links
13693 When doing debugging dumps (see @option{-d} option above), suppress
13694 instruction numbers for the links to the previous and next instructions
13697 @item -fdump-ipa-@var{switch}
13699 Control the dumping at various stages of inter-procedural analysis
13700 language tree to a file. The file name is generated by appending a
13701 switch specific suffix to the source file name, and the file is created
13702 in the same directory as the output file. The following dumps are
13707 Enables all inter-procedural analysis dumps.
13710 Dumps information about call-graph optimization, unused function removal,
13711 and inlining decisions.
13714 Dump after function inlining.
13718 @item -fdump-lang-all
13719 @itemx -fdump-lang-@var{switch}
13720 @itemx -fdump-lang-@var{switch}-@var{options}
13721 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
13722 @opindex fdump-lang-all
13723 @opindex fdump-lang
13724 Control the dumping of language-specific information. The @var{options}
13725 and @var{filename} portions behave as described in the
13726 @option{-fdump-tree} option. The following @var{switch} values are
13732 Enable all language-specific dumps.
13735 Dump class hierarchy information. Virtual table information is emitted
13736 unless '@option{slim}' is specified. This option is applicable to C++ only.
13739 Dump the raw internal tree data. This option is applicable to C++ only.
13743 @item -fdump-passes
13744 @opindex fdump-passes
13745 Print on @file{stderr} the list of optimization passes that are turned
13746 on and off by the current command-line options.
13748 @item -fdump-statistics-@var{option}
13749 @opindex fdump-statistics
13750 Enable and control dumping of pass statistics in a separate file. The
13751 file name is generated by appending a suffix ending in
13752 @samp{.statistics} to the source file name, and the file is created in
13753 the same directory as the output file. If the @samp{-@var{option}}
13754 form is used, @samp{-stats} causes counters to be summed over the
13755 whole compilation unit while @samp{-details} dumps every event as
13756 the passes generate them. The default with no option is to sum
13757 counters for each function compiled.
13759 @item -fdump-tree-all
13760 @itemx -fdump-tree-@var{switch}
13761 @itemx -fdump-tree-@var{switch}-@var{options}
13762 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
13763 @opindex fdump-tree-all
13764 @opindex fdump-tree
13765 Control the dumping at various stages of processing the intermediate
13766 language tree to a file. The file name is generated by appending a
13767 switch-specific suffix to the source file name, and the file is
13768 created in the same directory as the output file. In case of
13769 @option{=@var{filename}} option, the dump is output on the given file
13770 instead of the auto named dump files. If the @samp{-@var{options}}
13771 form is used, @var{options} is a list of @samp{-} separated options
13772 which control the details of the dump. Not all options are applicable
13773 to all dumps; those that are not meaningful are ignored. The
13774 following options are available
13778 Print the address of each node. Usually this is not meaningful as it
13779 changes according to the environment and source file. Its primary use
13780 is for tying up a dump file with a debug environment.
13782 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
13783 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
13784 use working backward from mangled names in the assembly file.
13786 When dumping front-end intermediate representations, inhibit dumping
13787 of members of a scope or body of a function merely because that scope
13788 has been reached. Only dump such items when they are directly reachable
13789 by some other path.
13791 When dumping pretty-printed trees, this option inhibits dumping the
13792 bodies of control structures.
13794 When dumping RTL, print the RTL in slim (condensed) form instead of
13795 the default LISP-like representation.
13797 Print a raw representation of the tree. By default, trees are
13798 pretty-printed into a C-like representation.
13800 Enable more detailed dumps (not honored by every dump option). Also
13801 include information from the optimization passes.
13803 Enable dumping various statistics about the pass (not honored by every dump
13806 Enable showing basic block boundaries (disabled in raw dumps).
13808 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13809 dump a representation of the control flow graph suitable for viewing with
13810 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13811 the file is pretty-printed as a subgraph, so that GraphViz can render them
13812 all in a single plot.
13814 This option currently only works for RTL dumps, and the RTL is always
13815 dumped in slim form.
13817 Enable showing virtual operands for every statement.
13819 Enable showing line numbers for statements.
13821 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13823 Enable showing the tree dump for each statement.
13825 Enable showing the EH region number holding each statement.
13827 Enable showing scalar evolution analysis details.
13829 Enable showing optimization information (only available in certain
13832 Enable showing missed optimization information (only available in certain
13835 Enable other detailed optimization information (only available in
13837 @item =@var{filename}
13838 Instead of an auto named dump file, output into the given file
13839 name. The file names @file{stdout} and @file{stderr} are treated
13840 specially and are considered already open standard streams. For
13844 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13845 -fdump-tree-pre=/dev/stderr file.c
13848 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13849 output on to @file{stderr}. If two conflicting dump filenames are
13850 given for the same pass, then the latter option overrides the earlier
13854 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13855 and @option{lineno}.
13858 Turn on all optimization options, i.e., @option{optimized},
13859 @option{missed}, and @option{note}.
13862 To determine what tree dumps are available or find the dump for a pass
13863 of interest follow the steps below.
13867 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
13868 look for a code that corresponds to the pass you are interested in.
13869 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
13870 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
13871 The number at the end distinguishes distinct invocations of the same pass.
13873 To enable the creation of the dump file, append the pass code to
13874 the @option{-fdump-} option prefix and invoke GCC with it. For example,
13875 to enable the dump from the Early Value Range Propagation pass, invoke
13876 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
13877 specify the name of the dump file. If you don't specify one, GCC
13878 creates as described below.
13880 Find the pass dump in a file whose name is composed of three components
13881 separated by a period: the name of the source file GCC was invoked to
13882 compile, a numeric suffix indicating the pass number followed by the
13883 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
13884 and finally the pass code. For example, the Early VRP pass dump might
13885 be in a file named @file{myfile.c.038t.evrp} in the current working
13886 directory. Note that the numeric codes are not stable and may change
13887 from one version of GCC to another.
13891 @itemx -fopt-info-@var{options}
13892 @itemx -fopt-info-@var{options}=@var{filename}
13894 Controls optimization dumps from various optimization passes. If the
13895 @samp{-@var{options}} form is used, @var{options} is a list of
13896 @samp{-} separated option keywords to select the dump details and
13899 The @var{options} can be divided into two groups: options describing the
13900 verbosity of the dump, and options describing which optimizations
13901 should be included. The options from both the groups can be freely
13902 mixed as they are non-overlapping. However, in case of any conflicts,
13903 the later options override the earlier options on the command
13906 The following options control the dump verbosity:
13910 Print information when an optimization is successfully applied. It is
13911 up to a pass to decide which information is relevant. For example, the
13912 vectorizer passes print the source location of loops which are
13913 successfully vectorized.
13915 Print information about missed optimizations. Individual passes
13916 control which information to include in the output.
13918 Print verbose information about optimizations, such as certain
13919 transformations, more detailed messages about decisions etc.
13921 Print detailed optimization information. This includes
13922 @samp{optimized}, @samp{missed}, and @samp{note}.
13925 One or more of the following option keywords can be used to describe a
13926 group of optimizations:
13930 Enable dumps from all interprocedural optimizations.
13932 Enable dumps from all loop optimizations.
13934 Enable dumps from all inlining optimizations.
13936 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
13938 Enable dumps from all vectorization optimizations.
13940 Enable dumps from all optimizations. This is a superset of
13941 the optimization groups listed above.
13944 If @var{options} is
13945 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13946 info about successful optimizations from all the passes.
13948 If the @var{filename} is provided, then the dumps from all the
13949 applicable optimizations are concatenated into the @var{filename}.
13950 Otherwise the dump is output onto @file{stderr}. Though multiple
13951 @option{-fopt-info} options are accepted, only one of them can include
13952 a @var{filename}. If other filenames are provided then all but the
13953 first such option are ignored.
13955 Note that the output @var{filename} is overwritten
13956 in case of multiple translation units. If a combined output from
13957 multiple translation units is desired, @file{stderr} should be used
13960 In the following example, the optimization info is output to
13969 gcc -O3 -fopt-info-missed=missed.all
13973 outputs missed optimization report from all the passes into
13974 @file{missed.all}, and this one:
13977 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13981 prints information about missed optimization opportunities from
13982 vectorization passes on @file{stderr}.
13983 Note that @option{-fopt-info-vec-missed} is equivalent to
13984 @option{-fopt-info-missed-vec}. The order of the optimization group
13985 names and message types listed after @option{-fopt-info} does not matter.
13987 As another example,
13989 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13993 outputs information about missed optimizations as well as
13994 optimized locations from all the inlining passes into
14000 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
14004 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
14005 in conflict since only one output file is allowed. In this case, only
14006 the first option takes effect and the subsequent options are
14007 ignored. Thus only @file{vec.miss} is produced which contains
14008 dumps from the vectorizer about missed opportunities.
14010 @item -fsched-verbose=@var{n}
14011 @opindex fsched-verbose
14012 On targets that use instruction scheduling, this option controls the
14013 amount of debugging output the scheduler prints to the dump files.
14015 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
14016 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
14017 For @var{n} greater than one, it also output basic block probabilities,
14018 detailed ready list information and unit/insn info. For @var{n} greater
14019 than two, it includes RTL at abort point, control-flow and regions info.
14020 And for @var{n} over four, @option{-fsched-verbose} also includes
14025 @item -fenable-@var{kind}-@var{pass}
14026 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
14030 This is a set of options that are used to explicitly disable/enable
14031 optimization passes. These options are intended for use for debugging GCC.
14032 Compiler users should use regular options for enabling/disabling
14037 @item -fdisable-ipa-@var{pass}
14038 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
14039 statically invoked in the compiler multiple times, the pass name should be
14040 appended with a sequential number starting from 1.
14042 @item -fdisable-rtl-@var{pass}
14043 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
14044 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
14045 statically invoked in the compiler multiple times, the pass name should be
14046 appended with a sequential number starting from 1. @var{range-list} is a
14047 comma-separated list of function ranges or assembler names. Each range is a number
14048 pair separated by a colon. The range is inclusive in both ends. If the range
14049 is trivial, the number pair can be simplified as a single number. If the
14050 function's call graph node's @var{uid} falls within one of the specified ranges,
14051 the @var{pass} is disabled for that function. The @var{uid} is shown in the
14052 function header of a dump file, and the pass names can be dumped by using
14053 option @option{-fdump-passes}.
14055 @item -fdisable-tree-@var{pass}
14056 @itemx -fdisable-tree-@var{pass}=@var{range-list}
14057 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
14060 @item -fenable-ipa-@var{pass}
14061 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
14062 statically invoked in the compiler multiple times, the pass name should be
14063 appended with a sequential number starting from 1.
14065 @item -fenable-rtl-@var{pass}
14066 @itemx -fenable-rtl-@var{pass}=@var{range-list}
14067 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
14068 description and examples.
14070 @item -fenable-tree-@var{pass}
14071 @itemx -fenable-tree-@var{pass}=@var{range-list}
14072 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
14073 of option arguments.
14077 Here are some examples showing uses of these options.
14081 # disable ccp1 for all functions
14082 -fdisable-tree-ccp1
14083 # disable complete unroll for function whose cgraph node uid is 1
14084 -fenable-tree-cunroll=1
14085 # disable gcse2 for functions at the following ranges [1,1],
14086 # [300,400], and [400,1000]
14087 # disable gcse2 for functions foo and foo2
14088 -fdisable-rtl-gcse2=foo,foo2
14089 # disable early inlining
14090 -fdisable-tree-einline
14091 # disable ipa inlining
14092 -fdisable-ipa-inline
14093 # enable tree full unroll
14094 -fenable-tree-unroll
14099 @itemx -fchecking=@var{n}
14101 @opindex fno-checking
14102 Enable internal consistency checking. The default depends on
14103 the compiler configuration. @option{-fchecking=2} enables further
14104 internal consistency checking that might affect code generation.
14106 @item -frandom-seed=@var{string}
14107 @opindex frandom-seed
14108 This option provides a seed that GCC uses in place of
14109 random numbers in generating certain symbol names
14110 that have to be different in every compiled file. It is also used to
14111 place unique stamps in coverage data files and the object files that
14112 produce them. You can use the @option{-frandom-seed} option to produce
14113 reproducibly identical object files.
14115 The @var{string} can either be a number (decimal, octal or hex) or an
14116 arbitrary string (in which case it's converted to a number by
14119 The @var{string} should be different for every file you compile.
14122 @itemx -save-temps=cwd
14123 @opindex save-temps
14124 Store the usual ``temporary'' intermediate files permanently; place them
14125 in the current directory and name them based on the source file. Thus,
14126 compiling @file{foo.c} with @option{-c -save-temps} produces files
14127 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
14128 preprocessed @file{foo.i} output file even though the compiler now
14129 normally uses an integrated preprocessor.
14131 When used in combination with the @option{-x} command-line option,
14132 @option{-save-temps} is sensible enough to avoid over writing an
14133 input source file with the same extension as an intermediate file.
14134 The corresponding intermediate file may be obtained by renaming the
14135 source file before using @option{-save-temps}.
14137 If you invoke GCC in parallel, compiling several different source
14138 files that share a common base name in different subdirectories or the
14139 same source file compiled for multiple output destinations, it is
14140 likely that the different parallel compilers will interfere with each
14141 other, and overwrite the temporary files. For instance:
14144 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
14145 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
14148 may result in @file{foo.i} and @file{foo.o} being written to
14149 simultaneously by both compilers.
14151 @item -save-temps=obj
14152 @opindex save-temps=obj
14153 Store the usual ``temporary'' intermediate files permanently. If the
14154 @option{-o} option is used, the temporary files are based on the
14155 object file. If the @option{-o} option is not used, the
14156 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
14161 gcc -save-temps=obj -c foo.c
14162 gcc -save-temps=obj -c bar.c -o dir/xbar.o
14163 gcc -save-temps=obj foobar.c -o dir2/yfoobar
14167 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
14168 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
14169 @file{dir2/yfoobar.o}.
14171 @item -time@r{[}=@var{file}@r{]}
14173 Report the CPU time taken by each subprocess in the compilation
14174 sequence. For C source files, this is the compiler proper and assembler
14175 (plus the linker if linking is done).
14177 Without the specification of an output file, the output looks like this:
14184 The first number on each line is the ``user time'', that is time spent
14185 executing the program itself. The second number is ``system time'',
14186 time spent executing operating system routines on behalf of the program.
14187 Both numbers are in seconds.
14189 With the specification of an output file, the output is appended to the
14190 named file, and it looks like this:
14193 0.12 0.01 cc1 @var{options}
14194 0.00 0.01 as @var{options}
14197 The ``user time'' and the ``system time'' are moved before the program
14198 name, and the options passed to the program are displayed, so that one
14199 can later tell what file was being compiled, and with which options.
14201 @item -fdump-final-insns@r{[}=@var{file}@r{]}
14202 @opindex fdump-final-insns
14203 Dump the final internal representation (RTL) to @var{file}. If the
14204 optional argument is omitted (or if @var{file} is @code{.}), the name
14205 of the dump file is determined by appending @code{.gkd} to the
14206 compilation output file name.
14208 @item -fcompare-debug@r{[}=@var{opts}@r{]}
14209 @opindex fcompare-debug
14210 @opindex fno-compare-debug
14211 If no error occurs during compilation, run the compiler a second time,
14212 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
14213 passed to the second compilation. Dump the final internal
14214 representation in both compilations, and print an error if they differ.
14216 If the equal sign is omitted, the default @option{-gtoggle} is used.
14218 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
14219 and nonzero, implicitly enables @option{-fcompare-debug}. If
14220 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
14221 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
14224 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
14225 is equivalent to @option{-fno-compare-debug}, which disables the dumping
14226 of the final representation and the second compilation, preventing even
14227 @env{GCC_COMPARE_DEBUG} from taking effect.
14229 To verify full coverage during @option{-fcompare-debug} testing, set
14230 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
14231 which GCC rejects as an invalid option in any actual compilation
14232 (rather than preprocessing, assembly or linking). To get just a
14233 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
14234 not overridden} will do.
14236 @item -fcompare-debug-second
14237 @opindex fcompare-debug-second
14238 This option is implicitly passed to the compiler for the second
14239 compilation requested by @option{-fcompare-debug}, along with options to
14240 silence warnings, and omitting other options that would cause the compiler
14241 to produce output to files or to standard output as a side effect. Dump
14242 files and preserved temporary files are renamed so as to contain the
14243 @code{.gk} additional extension during the second compilation, to avoid
14244 overwriting those generated by the first.
14246 When this option is passed to the compiler driver, it causes the
14247 @emph{first} compilation to be skipped, which makes it useful for little
14248 other than debugging the compiler proper.
14252 Turn off generation of debug info, if leaving out this option
14253 generates it, or turn it on at level 2 otherwise. The position of this
14254 argument in the command line does not matter; it takes effect after all
14255 other options are processed, and it does so only once, no matter how
14256 many times it is given. This is mainly intended to be used with
14257 @option{-fcompare-debug}.
14259 @item -fvar-tracking-assignments-toggle
14260 @opindex fvar-tracking-assignments-toggle
14261 @opindex fno-var-tracking-assignments-toggle
14262 Toggle @option{-fvar-tracking-assignments}, in the same way that
14263 @option{-gtoggle} toggles @option{-g}.
14267 Makes the compiler print out each function name as it is compiled, and
14268 print some statistics about each pass when it finishes.
14270 @item -ftime-report
14271 @opindex ftime-report
14272 Makes the compiler print some statistics about the time consumed by each
14273 pass when it finishes.
14275 @item -ftime-report-details
14276 @opindex ftime-report-details
14277 Record the time consumed by infrastructure parts separately for each pass.
14279 @item -fira-verbose=@var{n}
14280 @opindex fira-verbose
14281 Control the verbosity of the dump file for the integrated register allocator.
14282 The default value is 5. If the value @var{n} is greater or equal to 10,
14283 the dump output is sent to stderr using the same format as @var{n} minus 10.
14286 @opindex flto-report
14287 Prints a report with internal details on the workings of the link-time
14288 optimizer. The contents of this report vary from version to version.
14289 It is meant to be useful to GCC developers when processing object
14290 files in LTO mode (via @option{-flto}).
14292 Disabled by default.
14294 @item -flto-report-wpa
14295 @opindex flto-report-wpa
14296 Like @option{-flto-report}, but only print for the WPA phase of Link
14300 @opindex fmem-report
14301 Makes the compiler print some statistics about permanent memory
14302 allocation when it finishes.
14304 @item -fmem-report-wpa
14305 @opindex fmem-report-wpa
14306 Makes the compiler print some statistics about permanent memory
14307 allocation for the WPA phase only.
14309 @item -fpre-ipa-mem-report
14310 @opindex fpre-ipa-mem-report
14311 @item -fpost-ipa-mem-report
14312 @opindex fpost-ipa-mem-report
14313 Makes the compiler print some statistics about permanent memory
14314 allocation before or after interprocedural optimization.
14316 @item -fprofile-report
14317 @opindex fprofile-report
14318 Makes the compiler print some statistics about consistency of the
14319 (estimated) profile and effect of individual passes.
14321 @item -fstack-usage
14322 @opindex fstack-usage
14323 Makes the compiler output stack usage information for the program, on a
14324 per-function basis. The filename for the dump is made by appending
14325 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
14326 the output file, if explicitly specified and it is not an executable,
14327 otherwise it is the basename of the source file. An entry is made up
14332 The name of the function.
14336 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
14339 The qualifier @code{static} means that the function manipulates the stack
14340 statically: a fixed number of bytes are allocated for the frame on function
14341 entry and released on function exit; no stack adjustments are otherwise made
14342 in the function. The second field is this fixed number of bytes.
14344 The qualifier @code{dynamic} means that the function manipulates the stack
14345 dynamically: in addition to the static allocation described above, stack
14346 adjustments are made in the body of the function, for example to push/pop
14347 arguments around function calls. If the qualifier @code{bounded} is also
14348 present, the amount of these adjustments is bounded at compile time and
14349 the second field is an upper bound of the total amount of stack used by
14350 the function. If it is not present, the amount of these adjustments is
14351 not bounded at compile time and the second field only represents the
14356 Emit statistics about front-end processing at the end of the compilation.
14357 This option is supported only by the C++ front end, and
14358 the information is generally only useful to the G++ development team.
14360 @item -fdbg-cnt-list
14361 @opindex fdbg-cnt-list
14362 Print the name and the counter upper bound for all debug counters.
14365 @item -fdbg-cnt=@var{counter-value-list}
14367 Set the internal debug counter upper bound. @var{counter-value-list}
14368 is a comma-separated list of @var{name}:@var{value} pairs
14369 which sets the upper bound of each debug counter @var{name} to @var{value}.
14370 All debug counters have the initial upper bound of @code{UINT_MAX};
14371 thus @code{dbg_cnt} returns true always unless the upper bound
14372 is set by this option.
14373 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
14374 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
14376 @item -print-file-name=@var{library}
14377 @opindex print-file-name
14378 Print the full absolute name of the library file @var{library} that
14379 would be used when linking---and don't do anything else. With this
14380 option, GCC does not compile or link anything; it just prints the
14383 @item -print-multi-directory
14384 @opindex print-multi-directory
14385 Print the directory name corresponding to the multilib selected by any
14386 other switches present in the command line. This directory is supposed
14387 to exist in @env{GCC_EXEC_PREFIX}.
14389 @item -print-multi-lib
14390 @opindex print-multi-lib
14391 Print the mapping from multilib directory names to compiler switches
14392 that enable them. The directory name is separated from the switches by
14393 @samp{;}, and each switch starts with an @samp{@@} instead of the
14394 @samp{-}, without spaces between multiple switches. This is supposed to
14395 ease shell processing.
14397 @item -print-multi-os-directory
14398 @opindex print-multi-os-directory
14399 Print the path to OS libraries for the selected
14400 multilib, relative to some @file{lib} subdirectory. If OS libraries are
14401 present in the @file{lib} subdirectory and no multilibs are used, this is
14402 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
14403 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
14404 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
14405 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
14407 @item -print-multiarch
14408 @opindex print-multiarch
14409 Print the path to OS libraries for the selected multiarch,
14410 relative to some @file{lib} subdirectory.
14412 @item -print-prog-name=@var{program}
14413 @opindex print-prog-name
14414 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
14416 @item -print-libgcc-file-name
14417 @opindex print-libgcc-file-name
14418 Same as @option{-print-file-name=libgcc.a}.
14420 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
14421 but you do want to link with @file{libgcc.a}. You can do:
14424 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
14427 @item -print-search-dirs
14428 @opindex print-search-dirs
14429 Print the name of the configured installation directory and a list of
14430 program and library directories @command{gcc} searches---and don't do anything else.
14432 This is useful when @command{gcc} prints the error message
14433 @samp{installation problem, cannot exec cpp0: No such file or directory}.
14434 To resolve this you either need to put @file{cpp0} and the other compiler
14435 components where @command{gcc} expects to find them, or you can set the environment
14436 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
14437 Don't forget the trailing @samp{/}.
14438 @xref{Environment Variables}.
14440 @item -print-sysroot
14441 @opindex print-sysroot
14442 Print the target sysroot directory that is used during
14443 compilation. This is the target sysroot specified either at configure
14444 time or using the @option{--sysroot} option, possibly with an extra
14445 suffix that depends on compilation options. If no target sysroot is
14446 specified, the option prints nothing.
14448 @item -print-sysroot-headers-suffix
14449 @opindex print-sysroot-headers-suffix
14450 Print the suffix added to the target sysroot when searching for
14451 headers, or give an error if the compiler is not configured with such
14452 a suffix---and don't do anything else.
14455 @opindex dumpmachine
14456 Print the compiler's target machine (for example,
14457 @samp{i686-pc-linux-gnu})---and don't do anything else.
14460 @opindex dumpversion
14461 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
14462 anything else. This is the compiler version used in filesystem paths,
14463 specs, can be depending on how the compiler has been configured just
14464 a single number (major version), two numbers separated by dot (major and
14465 minor version) or three numbers separated by dots (major, minor and patchlevel
14468 @item -dumpfullversion
14469 @opindex dumpfullversion
14470 Print the full compiler version, always 3 numbers separated by dots,
14471 major, minor and patchlevel version.
14475 Print the compiler's built-in specs---and don't do anything else. (This
14476 is used when GCC itself is being built.) @xref{Spec Files}.
14479 @node Submodel Options
14480 @section Machine-Dependent Options
14481 @cindex submodel options
14482 @cindex specifying hardware config
14483 @cindex hardware models and configurations, specifying
14484 @cindex target-dependent options
14485 @cindex machine-dependent options
14487 Each target machine supported by GCC can have its own options---for
14488 example, to allow you to compile for a particular processor variant or
14489 ABI, or to control optimizations specific to that machine. By
14490 convention, the names of machine-specific options start with
14493 Some configurations of the compiler also support additional target-specific
14494 options, usually for compatibility with other compilers on the same
14497 @c This list is ordered alphanumerically by subsection name.
14498 @c It should be the same order and spelling as these options are listed
14499 @c in Machine Dependent Options
14502 * AArch64 Options::
14503 * Adapteva Epiphany Options::
14507 * Blackfin Options::
14512 * DEC Alpha Options::
14516 * GNU/Linux Options::
14526 * MicroBlaze Options::
14529 * MN10300 Options::
14533 * Nios II Options::
14534 * Nvidia PTX Options::
14536 * picoChip Options::
14537 * PowerPC Options::
14538 * PowerPC SPE Options::
14541 * RS/6000 and PowerPC Options::
14543 * S/390 and zSeries Options::
14546 * Solaris 2 Options::
14549 * System V Options::
14550 * TILE-Gx Options::
14551 * TILEPro Options::
14556 * VxWorks Options::
14558 * x86 Windows Options::
14559 * Xstormy16 Options::
14561 * zSeries Options::
14564 @node AArch64 Options
14565 @subsection AArch64 Options
14566 @cindex AArch64 Options
14568 These options are defined for AArch64 implementations:
14572 @item -mabi=@var{name}
14574 Generate code for the specified data model. Permissible values
14575 are @samp{ilp32} for SysV-like data model where int, long int and pointers
14576 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
14577 but long int and pointers are 64 bits.
14579 The default depends on the specific target configuration. Note that
14580 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
14581 entire program with the same ABI, and link with a compatible set of libraries.
14584 @opindex mbig-endian
14585 Generate big-endian code. This is the default when GCC is configured for an
14586 @samp{aarch64_be-*-*} target.
14588 @item -mgeneral-regs-only
14589 @opindex mgeneral-regs-only
14590 Generate code which uses only the general-purpose registers. This will prevent
14591 the compiler from using floating-point and Advanced SIMD registers but will not
14592 impose any restrictions on the assembler.
14594 @item -mlittle-endian
14595 @opindex mlittle-endian
14596 Generate little-endian code. This is the default when GCC is configured for an
14597 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
14599 @item -mcmodel=tiny
14600 @opindex mcmodel=tiny
14601 Generate code for the tiny code model. The program and its statically defined
14602 symbols must be within 1MB of each other. Programs can be statically or
14603 dynamically linked.
14605 @item -mcmodel=small
14606 @opindex mcmodel=small
14607 Generate code for the small code model. The program and its statically defined
14608 symbols must be within 4GB of each other. Programs can be statically or
14609 dynamically linked. This is the default code model.
14611 @item -mcmodel=large
14612 @opindex mcmodel=large
14613 Generate code for the large code model. This makes no assumptions about
14614 addresses and sizes of sections. Programs can be statically linked only.
14616 @item -mstrict-align
14617 @opindex mstrict-align
14618 Avoid generating memory accesses that may not be aligned on a natural object
14619 boundary as described in the architecture specification.
14621 @item -momit-leaf-frame-pointer
14622 @itemx -mno-omit-leaf-frame-pointer
14623 @opindex momit-leaf-frame-pointer
14624 @opindex mno-omit-leaf-frame-pointer
14625 Omit or keep the frame pointer in leaf functions. The former behavior is the
14628 @item -mtls-dialect=desc
14629 @opindex mtls-dialect=desc
14630 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
14631 of TLS variables. This is the default.
14633 @item -mtls-dialect=traditional
14634 @opindex mtls-dialect=traditional
14635 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
14638 @item -mtls-size=@var{size}
14640 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
14641 This option requires binutils 2.26 or newer.
14643 @item -mfix-cortex-a53-835769
14644 @itemx -mno-fix-cortex-a53-835769
14645 @opindex mfix-cortex-a53-835769
14646 @opindex mno-fix-cortex-a53-835769
14647 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
14648 This involves inserting a NOP instruction between memory instructions and
14649 64-bit integer multiply-accumulate instructions.
14651 @item -mfix-cortex-a53-843419
14652 @itemx -mno-fix-cortex-a53-843419
14653 @opindex mfix-cortex-a53-843419
14654 @opindex mno-fix-cortex-a53-843419
14655 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
14656 This erratum workaround is made at link time and this will only pass the
14657 corresponding flag to the linker.
14659 @item -mlow-precision-recip-sqrt
14660 @itemx -mno-low-precision-recip-sqrt
14661 @opindex mlow-precision-recip-sqrt
14662 @opindex mno-low-precision-recip-sqrt
14663 Enable or disable the reciprocal square root approximation.
14664 This option only has an effect if @option{-ffast-math} or
14665 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14666 precision of reciprocal square root results to about 16 bits for
14667 single precision and to 32 bits for double precision.
14669 @item -mlow-precision-sqrt
14670 @itemx -mno-low-precision-sqrt
14671 @opindex -mlow-precision-sqrt
14672 @opindex -mno-low-precision-sqrt
14673 Enable or disable the square root approximation.
14674 This option only has an effect if @option{-ffast-math} or
14675 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14676 precision of square root results to about 16 bits for
14677 single precision and to 32 bits for double precision.
14678 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
14680 @item -mlow-precision-div
14681 @itemx -mno-low-precision-div
14682 @opindex -mlow-precision-div
14683 @opindex -mno-low-precision-div
14684 Enable or disable the division approximation.
14685 This option only has an effect if @option{-ffast-math} or
14686 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14687 precision of division results to about 16 bits for
14688 single precision and to 32 bits for double precision.
14690 @item -march=@var{name}
14692 Specify the name of the target architecture and, optionally, one or
14693 more feature modifiers. This option has the form
14694 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
14696 The permissible values for @var{arch} are @samp{armv8-a},
14697 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @samp{armv8.4-a}
14700 The value @samp{armv8.4-a} implies @samp{armv8.3-a} and enables compiler
14701 support for the ARMv8.4-A architecture extensions.
14703 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
14704 support for the ARMv8.3-A architecture extensions.
14706 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
14707 support for the ARMv8.2-A architecture extensions.
14709 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
14710 support for the ARMv8.1-A architecture extension. In particular, it
14711 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
14713 The value @samp{native} is available on native AArch64 GNU/Linux and
14714 causes the compiler to pick the architecture of the host system. This
14715 option has no effect if the compiler is unable to recognize the
14716 architecture of the host system,
14718 The permissible values for @var{feature} are listed in the sub-section
14719 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14720 Feature Modifiers}. Where conflicting feature modifiers are
14721 specified, the right-most feature is used.
14723 GCC uses @var{name} to determine what kind of instructions it can emit
14724 when generating assembly code. If @option{-march} is specified
14725 without either of @option{-mtune} or @option{-mcpu} also being
14726 specified, the code is tuned to perform well across a range of target
14727 processors implementing the target architecture.
14729 @item -mtune=@var{name}
14731 Specify the name of the target processor for which GCC should tune the
14732 performance of the code. Permissible values for this option are:
14733 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
14734 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
14735 @samp{exynos-m1}, @samp{falkor}, @samp{qdf24xx}, @samp{saphira},
14736 @samp{xgene1}, @samp{vulcan}, @samp{thunderx},
14737 @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
14738 @samp{thunderxt83}, @samp{thunderx2t99}, @samp{cortex-a57.cortex-a53},
14739 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
14740 @samp{cortex-a73.cortex-a53}, @samp{cortex-a75.cortex-a55},
14743 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14744 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
14745 @samp{cortex-a75.cortex-a55} specify that GCC should tune for a
14748 Additionally on native AArch64 GNU/Linux systems the value
14749 @samp{native} tunes performance to the host system. This option has no effect
14750 if the compiler is unable to recognize the processor of the host system.
14752 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
14753 are specified, the code is tuned to perform well across a range
14754 of target processors.
14756 This option cannot be suffixed by feature modifiers.
14758 @item -mcpu=@var{name}
14760 Specify the name of the target processor, optionally suffixed by one
14761 or more feature modifiers. This option has the form
14762 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
14763 the permissible values for @var{cpu} are the same as those available
14764 for @option{-mtune}. The permissible values for @var{feature} are
14765 documented in the sub-section on
14766 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14767 Feature Modifiers}. Where conflicting feature modifiers are
14768 specified, the right-most feature is used.
14770 GCC uses @var{name} to determine what kind of instructions it can emit when
14771 generating assembly code (as if by @option{-march}) and to determine
14772 the target processor for which to tune for performance (as if
14773 by @option{-mtune}). Where this option is used in conjunction
14774 with @option{-march} or @option{-mtune}, those options take precedence
14775 over the appropriate part of this option.
14777 @item -moverride=@var{string}
14779 Override tuning decisions made by the back-end in response to a
14780 @option{-mtune=} switch. The syntax, semantics, and accepted values
14781 for @var{string} in this option are not guaranteed to be consistent
14784 This option is only intended to be useful when developing GCC.
14786 @item -mverbose-cost-dump
14787 @opindex mverbose-cost-dump
14788 Enable verbose cost model dumping in the debug dump files. This option is
14789 provided for use in debugging the compiler.
14791 @item -mpc-relative-literal-loads
14792 @itemx -mno-pc-relative-literal-loads
14793 @opindex mpc-relative-literal-loads
14794 @opindex mno-pc-relative-literal-loads
14795 Enable or disable PC-relative literal loads. With this option literal pools are
14796 accessed using a single instruction and emitted after each function. This
14797 limits the maximum size of functions to 1MB. This is enabled by default for
14798 @option{-mcmodel=tiny}.
14800 @item -msign-return-address=@var{scope}
14801 @opindex msign-return-address
14802 Select the function scope on which return address signing will be applied.
14803 Permissible values are @samp{none}, which disables return address signing,
14804 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
14805 functions, and @samp{all}, which enables pointer signing for all functions. The
14806 default value is @samp{none}.
14808 @item -msve-vector-bits=@var{bits}
14809 @opindex msve-vector-bits
14810 Specify the number of bits in an SVE vector register. This option only has
14811 an effect when SVE is enabled.
14813 GCC supports two forms of SVE code generation: ``vector-length
14814 agnostic'' output that works with any size of vector register and
14815 ``vector-length specific'' output that only works when the vector
14816 registers are a particular size. Replacing @var{bits} with
14817 @samp{scalable} selects vector-length agnostic output while
14818 replacing it with a number selects vector-length specific output.
14819 The possible lengths in the latter case are: 128, 256, 512, 1024
14820 and 2048. @samp{scalable} is the default.
14822 At present, @samp{-msve-vector-bits=128} produces the same output
14823 as @samp{-msve-vector-bits=scalable}.
14827 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14828 @anchor{aarch64-feature-modifiers}
14829 @cindex @option{-march} feature modifiers
14830 @cindex @option{-mcpu} feature modifiers
14831 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14832 the following and their inverses @option{no@var{feature}}:
14836 Enable CRC extension. This is on by default for
14837 @option{-march=armv8.1-a}.
14839 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14842 Enable floating-point instructions. This is on by default for all possible
14843 values for options @option{-march} and @option{-mcpu}.
14845 Enable Advanced SIMD instructions. This also enables floating-point
14846 instructions. This is on by default for all possible values for options
14847 @option{-march} and @option{-mcpu}.
14849 Enable Scalable Vector Extension instructions. This also enables Advanced
14850 SIMD and floating-point instructions.
14852 Enable Large System Extension instructions. This is on by default for
14853 @option{-march=armv8.1-a}.
14855 Enable Round Double Multiply Accumulate instructions. This is on by default
14856 for @option{-march=armv8.1-a}.
14858 Enable FP16 extension. This also enables floating-point instructions.
14860 Enable FP16 fmla extension. This also enables FP16 extensions and
14861 floating-point instructions. This option is enabled by default for @option{-march=armv8.4-a}. Use of this option with architectures prior to Armv8.2-A is not supported.
14864 Enable the RcPc extension. This does not change code generation from GCC,
14865 but is passed on to the assembler, enabling inline asm statements to use
14866 instructions from the RcPc extension.
14868 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
14870 Enable the Armv8-a aes and pmull crypto extension. This also enables Advanced
14873 Enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions.
14875 Enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD
14876 instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
14878 Enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions.
14879 Use of this option with architectures prior to Armv8.2-A is not supported.
14883 Feature @option{crypto} implies @option{aes}, @option{sha2}, and @option{simd},
14884 which implies @option{fp}.
14885 Conversely, @option{nofp} implies @option{nosimd}, which implies
14886 @option{nocrypto}, @option{noaes} and @option{nosha2}.
14888 @node Adapteva Epiphany Options
14889 @subsection Adapteva Epiphany Options
14891 These @samp{-m} options are defined for Adapteva Epiphany:
14894 @item -mhalf-reg-file
14895 @opindex mhalf-reg-file
14896 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14897 That allows code to run on hardware variants that lack these registers.
14899 @item -mprefer-short-insn-regs
14900 @opindex mprefer-short-insn-regs
14901 Preferentially allocate registers that allow short instruction generation.
14902 This can result in increased instruction count, so this may either reduce or
14903 increase overall code size.
14905 @item -mbranch-cost=@var{num}
14906 @opindex mbranch-cost
14907 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14908 This cost is only a heuristic and is not guaranteed to produce
14909 consistent results across releases.
14913 Enable the generation of conditional moves.
14915 @item -mnops=@var{num}
14917 Emit @var{num} NOPs before every other generated instruction.
14919 @item -mno-soft-cmpsf
14920 @opindex mno-soft-cmpsf
14921 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14922 and test the flags. This is faster than a software comparison, but can
14923 get incorrect results in the presence of NaNs, or when two different small
14924 numbers are compared such that their difference is calculated as zero.
14925 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14926 software comparisons.
14928 @item -mstack-offset=@var{num}
14929 @opindex mstack-offset
14930 Set the offset between the top of the stack and the stack pointer.
14931 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14932 can be used by leaf functions without stack allocation.
14933 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14934 Note also that this option changes the ABI; compiling a program with a
14935 different stack offset than the libraries have been compiled with
14936 generally does not work.
14937 This option can be useful if you want to evaluate if a different stack
14938 offset would give you better code, but to actually use a different stack
14939 offset to build working programs, it is recommended to configure the
14940 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14942 @item -mno-round-nearest
14943 @opindex mno-round-nearest
14944 Make the scheduler assume that the rounding mode has been set to
14945 truncating. The default is @option{-mround-nearest}.
14948 @opindex mlong-calls
14949 If not otherwise specified by an attribute, assume all calls might be beyond
14950 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14951 function address into a register before performing a (otherwise direct) call.
14952 This is the default.
14954 @item -mshort-calls
14955 @opindex short-calls
14956 If not otherwise specified by an attribute, assume all direct calls are
14957 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14958 for direct calls. The default is @option{-mlong-calls}.
14962 Assume addresses can be loaded as 16-bit unsigned values. This does not
14963 apply to function addresses for which @option{-mlong-calls} semantics
14966 @item -mfp-mode=@var{mode}
14968 Set the prevailing mode of the floating-point unit.
14969 This determines the floating-point mode that is provided and expected
14970 at function call and return time. Making this mode match the mode you
14971 predominantly need at function start can make your programs smaller and
14972 faster by avoiding unnecessary mode switches.
14974 @var{mode} can be set to one the following values:
14978 Any mode at function entry is valid, and retained or restored when
14979 the function returns, and when it calls other functions.
14980 This mode is useful for compiling libraries or other compilation units
14981 you might want to incorporate into different programs with different
14982 prevailing FPU modes, and the convenience of being able to use a single
14983 object file outweighs the size and speed overhead for any extra
14984 mode switching that might be needed, compared with what would be needed
14985 with a more specific choice of prevailing FPU mode.
14988 This is the mode used for floating-point calculations with
14989 truncating (i.e.@: round towards zero) rounding mode. That includes
14990 conversion from floating point to integer.
14992 @item round-nearest
14993 This is the mode used for floating-point calculations with
14994 round-to-nearest-or-even rounding mode.
14997 This is the mode used to perform integer calculations in the FPU, e.g.@:
14998 integer multiply, or integer multiply-and-accumulate.
15001 The default is @option{-mfp-mode=caller}
15003 @item -mnosplit-lohi
15004 @itemx -mno-postinc
15005 @itemx -mno-postmodify
15006 @opindex mnosplit-lohi
15007 @opindex mno-postinc
15008 @opindex mno-postmodify
15009 Code generation tweaks that disable, respectively, splitting of 32-bit
15010 loads, generation of post-increment addresses, and generation of
15011 post-modify addresses. The defaults are @option{msplit-lohi},
15012 @option{-mpost-inc}, and @option{-mpost-modify}.
15014 @item -mnovect-double
15015 @opindex mno-vect-double
15016 Change the preferred SIMD mode to SImode. The default is
15017 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
15019 @item -max-vect-align=@var{num}
15020 @opindex max-vect-align
15021 The maximum alignment for SIMD vector mode types.
15022 @var{num} may be 4 or 8. The default is 8.
15023 Note that this is an ABI change, even though many library function
15024 interfaces are unaffected if they don't use SIMD vector modes
15025 in places that affect size and/or alignment of relevant types.
15027 @item -msplit-vecmove-early
15028 @opindex msplit-vecmove-early
15029 Split vector moves into single word moves before reload. In theory this
15030 can give better register allocation, but so far the reverse seems to be
15031 generally the case.
15033 @item -m1reg-@var{reg}
15035 Specify a register to hold the constant @minus{}1, which makes loading small negative
15036 constants and certain bitmasks faster.
15037 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
15038 which specify use of that register as a fixed register,
15039 and @samp{none}, which means that no register is used for this
15040 purpose. The default is @option{-m1reg-none}.
15045 @subsection ARC Options
15046 @cindex ARC options
15048 The following options control the architecture variant for which code
15051 @c architecture variants
15054 @item -mbarrel-shifter
15055 @opindex mbarrel-shifter
15056 Generate instructions supported by barrel shifter. This is the default
15057 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
15060 @opindex mjli-alawys
15061 Force to call a function using jli_s instruction. This option is
15062 valid only for ARCv2 architecture.
15064 @item -mcpu=@var{cpu}
15066 Set architecture type, register usage, and instruction scheduling
15067 parameters for @var{cpu}. There are also shortcut alias options
15068 available for backward compatibility and convenience. Supported
15069 values for @var{cpu} are
15075 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
15079 Compile for ARC601. Alias: @option{-mARC601}.
15084 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
15085 This is the default when configured with @option{--with-cpu=arc700}@.
15088 Compile for ARC EM.
15091 Compile for ARC HS.
15094 Compile for ARC EM CPU with no hardware extensions.
15097 Compile for ARC EM4 CPU.
15100 Compile for ARC EM4 DMIPS CPU.
15103 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
15107 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
15108 double assist instructions.
15111 Compile for ARC HS CPU with no hardware extensions except the atomic
15115 Compile for ARC HS34 CPU.
15118 Compile for ARC HS38 CPU.
15121 Compile for ARC HS38 CPU with all hardware extensions on.
15124 Compile for ARC 600 CPU with @code{norm} instructions enabled.
15126 @item arc600_mul32x16
15127 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
15128 instructions enabled.
15131 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
15132 instructions enabled.
15135 Compile for ARC 601 CPU with @code{norm} instructions enabled.
15137 @item arc601_mul32x16
15138 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
15139 instructions enabled.
15142 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
15143 instructions enabled.
15146 Compile for ARC 700 on NPS400 chip.
15149 Compile for ARC EM minimalist configuration featuring reduced register
15156 @itemx -mdpfp-compact
15157 @opindex mdpfp-compact
15158 Generate double-precision FPX instructions, tuned for the compact
15162 @opindex mdpfp-fast
15163 Generate double-precision FPX instructions, tuned for the fast
15166 @item -mno-dpfp-lrsr
15167 @opindex mno-dpfp-lrsr
15168 Disable @code{lr} and @code{sr} instructions from using FPX extension
15173 Generate extended arithmetic instructions. Currently only
15174 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
15175 supported. This is always enabled for @option{-mcpu=ARC700}.
15179 Do not generate @code{mpy}-family instructions for ARC700. This option is
15184 Generate 32x16-bit multiply and multiply-accumulate instructions.
15188 Generate @code{mul64} and @code{mulu64} instructions.
15189 Only valid for @option{-mcpu=ARC600}.
15193 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
15198 @itemx -mspfp-compact
15199 @opindex mspfp-compact
15200 Generate single-precision FPX instructions, tuned for the compact
15204 @opindex mspfp-fast
15205 Generate single-precision FPX instructions, tuned for the fast
15210 Enable generation of ARC SIMD instructions via target-specific
15211 builtins. Only valid for @option{-mcpu=ARC700}.
15214 @opindex msoft-float
15215 This option ignored; it is provided for compatibility purposes only.
15216 Software floating-point code is emitted by default, and this default
15217 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
15218 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
15219 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
15223 Generate @code{swap} instructions.
15227 This enables use of the locked load/store conditional extension to implement
15228 atomic memory built-in functions. Not available for ARC 6xx or ARC
15233 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
15235 @item -mcode-density
15236 @opindex mcode-density
15237 Enable code density instructions for ARC EM.
15238 This option is on by default for ARC HS.
15242 Enable double load/store operations for ARC HS cores.
15244 @item -mtp-regno=@var{regno}
15246 Specify thread pointer register number.
15248 @item -mmpy-option=@var{multo}
15249 @opindex mmpy-option
15250 Compile ARCv2 code with a multiplier design option. You can specify
15251 the option using either a string or numeric value for @var{multo}.
15252 @samp{wlh1} is the default value. The recognized values are:
15257 No multiplier available.
15261 16x16 multiplier, fully pipelined.
15262 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
15266 32x32 multiplier, fully
15267 pipelined (1 stage). The following instructions are additionally
15268 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15272 32x32 multiplier, fully pipelined
15273 (2 stages). The following instructions are additionally enabled: @code{mpy},
15274 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15278 Two 16x16 multipliers, blocking,
15279 sequential. The following instructions are additionally enabled: @code{mpy},
15280 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15284 One 16x16 multiplier, blocking,
15285 sequential. The following instructions are additionally enabled: @code{mpy},
15286 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15290 One 32x4 multiplier, blocking,
15291 sequential. The following instructions are additionally enabled: @code{mpy},
15292 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15296 ARC HS SIMD support.
15300 ARC HS SIMD support.
15304 ARC HS SIMD support.
15308 This option is only available for ARCv2 cores@.
15310 @item -mfpu=@var{fpu}
15312 Enables support for specific floating-point hardware extensions for ARCv2
15313 cores. Supported values for @var{fpu} are:
15318 Enables support for single-precision floating-point hardware
15322 Enables support for double-precision floating-point hardware
15323 extensions. The single-precision floating-point extension is also
15324 enabled. Not available for ARC EM@.
15327 Enables support for double-precision floating-point hardware
15328 extensions using double-precision assist instructions. The single-precision
15329 floating-point extension is also enabled. This option is
15330 only available for ARC EM@.
15333 Enables support for double-precision floating-point hardware
15334 extensions using double-precision assist instructions.
15335 The single-precision floating-point, square-root, and divide
15336 extensions are also enabled. This option is
15337 only available for ARC EM@.
15340 Enables support for double-precision floating-point hardware
15341 extensions using double-precision assist instructions.
15342 The single-precision floating-point and fused multiply and add
15343 hardware extensions are also enabled. This option is
15344 only available for ARC EM@.
15347 Enables support for double-precision floating-point hardware
15348 extensions using double-precision assist instructions.
15349 All single-precision floating-point hardware extensions are also
15350 enabled. This option is only available for ARC EM@.
15353 Enables support for single-precision floating-point, square-root and divide
15354 hardware extensions@.
15357 Enables support for double-precision floating-point, square-root and divide
15358 hardware extensions. This option
15359 includes option @samp{fpus_div}. Not available for ARC EM@.
15362 Enables support for single-precision floating-point and
15363 fused multiply and add hardware extensions@.
15366 Enables support for double-precision floating-point and
15367 fused multiply and add hardware extensions. This option
15368 includes option @samp{fpus_fma}. Not available for ARC EM@.
15371 Enables support for all single-precision floating-point hardware
15375 Enables support for all single- and double-precision floating-point
15376 hardware extensions. Not available for ARC EM@.
15380 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
15381 @opindex mirq-ctrl-saved
15382 Specifies general-purposes registers that the processor automatically
15383 saves/restores on interrupt entry and exit. @var{register-range} is
15384 specified as two registers separated by a dash. The register range
15385 always starts with @code{r0}, the upper limit is @code{fp} register.
15386 @var{blink} and @var{lp_count} are optional. This option is only
15387 valid for ARC EM and ARC HS cores.
15389 @item -mrgf-banked-regs=@var{number}
15390 @opindex mrgf-banked-regs
15391 Specifies the number of registers replicated in second register bank
15392 on entry to fast interrupt. Fast interrupts are interrupts with the
15393 highest priority level P0. These interrupts save only PC and STATUS32
15394 registers to avoid memory transactions during interrupt entry and exit
15395 sequences. Use this option when you are using fast interrupts in an
15396 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
15398 @item -mlpc-width=@var{width}
15399 @opindex mlpc-width
15400 Specify the width of the @code{lp_count} register. Valid values for
15401 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
15402 fixed to 32 bits. If the width is less than 32, the compiler does not
15403 attempt to transform loops in your program to use the zero-delay loop
15404 mechanism unless it is known that the @code{lp_count} register can
15405 hold the required loop-counter value. Depending on the width
15406 specified, the compiler and run-time library might continue to use the
15407 loop mechanism for various needs. This option defines macro
15408 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
15412 This option instructs the compiler to generate code for a 16-entry
15413 register file. This option defines the @code{__ARC_RF16__}
15414 preprocessor macro.
15418 The following options are passed through to the assembler, and also
15419 define preprocessor macro symbols.
15421 @c Flags used by the assembler, but for which we define preprocessor
15422 @c macro symbols as well.
15425 @opindex mdsp-packa
15426 Passed down to the assembler to enable the DSP Pack A extensions.
15427 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
15432 Passed down to the assembler to enable the dual Viterbi butterfly
15433 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
15434 option is deprecated.
15436 @c ARC700 4.10 extension instruction
15439 Passed down to the assembler to enable the locked load/store
15440 conditional extension. Also sets the preprocessor symbol
15445 Passed down to the assembler. Also sets the preprocessor symbol
15446 @code{__Xxmac_d16}. This option is deprecated.
15450 Passed down to the assembler. Also sets the preprocessor symbol
15451 @code{__Xxmac_24}. This option is deprecated.
15453 @c ARC700 4.10 extension instruction
15456 Passed down to the assembler to enable the 64-bit time-stamp counter
15457 extension instruction. Also sets the preprocessor symbol
15458 @code{__Xrtsc}. This option is deprecated.
15460 @c ARC700 4.10 extension instruction
15463 Passed down to the assembler to enable the swap byte ordering
15464 extension instruction. Also sets the preprocessor symbol
15468 @opindex mtelephony
15469 Passed down to the assembler to enable dual- and single-operand
15470 instructions for telephony. Also sets the preprocessor symbol
15471 @code{__Xtelephony}. This option is deprecated.
15475 Passed down to the assembler to enable the XY memory extension. Also
15476 sets the preprocessor symbol @code{__Xxy}.
15480 The following options control how the assembly code is annotated:
15482 @c Assembly annotation options
15486 Annotate assembler instructions with estimated addresses.
15488 @item -mannotate-align
15489 @opindex mannotate-align
15490 Explain what alignment considerations lead to the decision to make an
15491 instruction short or long.
15495 The following options are passed through to the linker:
15497 @c options passed through to the linker
15501 Passed through to the linker, to specify use of the @code{arclinux} emulation.
15502 This option is enabled by default in tool chains built for
15503 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
15504 when profiling is not requested.
15506 @item -marclinux_prof
15507 @opindex marclinux_prof
15508 Passed through to the linker, to specify use of the
15509 @code{arclinux_prof} emulation. This option is enabled by default in
15510 tool chains built for @w{@code{arc-linux-uclibc}} and
15511 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
15515 The following options control the semantics of generated code:
15517 @c semantically relevant code generation options
15520 @opindex mlong-calls
15521 Generate calls as register indirect calls, thus providing access
15522 to the full 32-bit address range.
15524 @item -mmedium-calls
15525 @opindex mmedium-calls
15526 Don't use less than 25-bit addressing range for calls, which is the
15527 offset available for an unconditional branch-and-link
15528 instruction. Conditional execution of function calls is suppressed, to
15529 allow use of the 25-bit range, rather than the 21-bit range with
15530 conditional branch-and-link. This is the default for tool chains built
15531 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
15535 Put definitions of externally-visible data in a small data section if
15536 that data is no bigger than @var{num} bytes. The default value of
15537 @var{num} is 4 for any ARC configuration, or 8 when we have double
15538 load/store operations.
15542 Do not generate sdata references. This is the default for tool chains
15543 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
15546 @item -mvolatile-cache
15547 @opindex mvolatile-cache
15548 Use ordinarily cached memory accesses for volatile references. This is the
15551 @item -mno-volatile-cache
15552 @opindex mno-volatile-cache
15553 Enable cache bypass for volatile references.
15557 The following options fine tune code generation:
15558 @c code generation tuning options
15561 @opindex malign-call
15562 Do alignment optimizations for call instructions.
15564 @item -mauto-modify-reg
15565 @opindex mauto-modify-reg
15566 Enable the use of pre/post modify with register displacement.
15568 @item -mbbit-peephole
15569 @opindex mbbit-peephole
15570 Enable bbit peephole2.
15574 This option disables a target-specific pass in @file{arc_reorg} to
15575 generate compare-and-branch (@code{br@var{cc}}) instructions.
15576 It has no effect on
15577 generation of these instructions driven by the combiner pass.
15579 @item -mcase-vector-pcrel
15580 @opindex mcase-vector-pcrel
15581 Use PC-relative switch case tables to enable case table shortening.
15582 This is the default for @option{-Os}.
15584 @item -mcompact-casesi
15585 @opindex mcompact-casesi
15586 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
15587 and only available for ARCv1 cores.
15589 @item -mno-cond-exec
15590 @opindex mno-cond-exec
15591 Disable the ARCompact-specific pass to generate conditional
15592 execution instructions.
15594 Due to delay slot scheduling and interactions between operand numbers,
15595 literal sizes, instruction lengths, and the support for conditional execution,
15596 the target-independent pass to generate conditional execution is often lacking,
15597 so the ARC port has kept a special pass around that tries to find more
15598 conditional execution generation opportunities after register allocation,
15599 branch shortening, and delay slot scheduling have been done. This pass
15600 generally, but not always, improves performance and code size, at the cost of
15601 extra compilation time, which is why there is an option to switch it off.
15602 If you have a problem with call instructions exceeding their allowable
15603 offset range because they are conditionalized, you should consider using
15604 @option{-mmedium-calls} instead.
15606 @item -mearly-cbranchsi
15607 @opindex mearly-cbranchsi
15608 Enable pre-reload use of the @code{cbranchsi} pattern.
15610 @item -mexpand-adddi
15611 @opindex mexpand-adddi
15612 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
15613 @code{add.f}, @code{adc} etc. This option is deprecated.
15615 @item -mindexed-loads
15616 @opindex mindexed-loads
15617 Enable the use of indexed loads. This can be problematic because some
15618 optimizers then assume that indexed stores exist, which is not
15623 Enable Local Register Allocation. This is still experimental for ARC,
15624 so by default the compiler uses standard reload
15625 (i.e. @option{-mno-lra}).
15627 @item -mlra-priority-none
15628 @opindex mlra-priority-none
15629 Don't indicate any priority for target registers.
15631 @item -mlra-priority-compact
15632 @opindex mlra-priority-compact
15633 Indicate target register priority for r0..r3 / r12..r15.
15635 @item -mlra-priority-noncompact
15636 @opindex mlra-priority-noncompact
15637 Reduce target register priority for r0..r3 / r12..r15.
15639 @item -mno-millicode
15640 @opindex mno-millicode
15641 When optimizing for size (using @option{-Os}), prologues and epilogues
15642 that have to save or restore a large number of registers are often
15643 shortened by using call to a special function in libgcc; this is
15644 referred to as a @emph{millicode} call. As these calls can pose
15645 performance issues, and/or cause linking issues when linking in a
15646 nonstandard way, this option is provided to turn off millicode call
15650 @opindex mmixed-code
15651 Tweak register allocation to help 16-bit instruction generation.
15652 This generally has the effect of decreasing the average instruction size
15653 while increasing the instruction count.
15657 Enable @samp{q} instruction alternatives.
15658 This is the default for @option{-Os}.
15662 Enable @samp{Rcq} constraint handling.
15663 Most short code generation depends on this.
15664 This is the default.
15668 Enable @samp{Rcw} constraint handling.
15669 Most ccfsm condexec mostly depends on this.
15670 This is the default.
15672 @item -msize-level=@var{level}
15673 @opindex msize-level
15674 Fine-tune size optimization with regards to instruction lengths and alignment.
15675 The recognized values for @var{level} are:
15678 No size optimization. This level is deprecated and treated like @samp{1}.
15681 Short instructions are used opportunistically.
15684 In addition, alignment of loops and of code after barriers are dropped.
15687 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
15691 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
15692 the behavior when this is not set is equivalent to level @samp{1}.
15694 @item -mtune=@var{cpu}
15696 Set instruction scheduling parameters for @var{cpu}, overriding any implied
15697 by @option{-mcpu=}.
15699 Supported values for @var{cpu} are
15703 Tune for ARC600 CPU.
15706 Tune for ARC601 CPU.
15709 Tune for ARC700 CPU with standard multiplier block.
15712 Tune for ARC700 CPU with XMAC block.
15715 Tune for ARC725D CPU.
15718 Tune for ARC750D CPU.
15722 @item -mmultcost=@var{num}
15724 Cost to assume for a multiply instruction, with @samp{4} being equal to a
15725 normal instruction.
15727 @item -munalign-prob-threshold=@var{probability}
15728 @opindex munalign-prob-threshold
15729 Set probability threshold for unaligning branches.
15730 When tuning for @samp{ARC700} and optimizing for speed, branches without
15731 filled delay slot are preferably emitted unaligned and long, unless
15732 profiling indicates that the probability for the branch to be taken
15733 is below @var{probability}. @xref{Cross-profiling}.
15734 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
15738 The following options are maintained for backward compatibility, but
15739 are now deprecated and will be removed in a future release:
15741 @c Deprecated options
15749 @opindex mbig-endian
15752 Compile code for big-endian targets. Use of these options is now
15753 deprecated. Big-endian code is supported by configuring GCC to build
15754 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
15755 for which big endian is the default.
15757 @item -mlittle-endian
15758 @opindex mlittle-endian
15761 Compile code for little-endian targets. Use of these options is now
15762 deprecated. Little-endian code is supported by configuring GCC to build
15763 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
15764 for which little endian is the default.
15766 @item -mbarrel_shifter
15767 @opindex mbarrel_shifter
15768 Replaced by @option{-mbarrel-shifter}.
15770 @item -mdpfp_compact
15771 @opindex mdpfp_compact
15772 Replaced by @option{-mdpfp-compact}.
15775 @opindex mdpfp_fast
15776 Replaced by @option{-mdpfp-fast}.
15779 @opindex mdsp_packa
15780 Replaced by @option{-mdsp-packa}.
15784 Replaced by @option{-mea}.
15788 Replaced by @option{-mmac-24}.
15792 Replaced by @option{-mmac-d16}.
15794 @item -mspfp_compact
15795 @opindex mspfp_compact
15796 Replaced by @option{-mspfp-compact}.
15799 @opindex mspfp_fast
15800 Replaced by @option{-mspfp-fast}.
15802 @item -mtune=@var{cpu}
15804 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
15805 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
15806 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
15808 @item -multcost=@var{num}
15810 Replaced by @option{-mmultcost}.
15815 @subsection ARM Options
15816 @cindex ARM options
15818 These @samp{-m} options are defined for the ARM port:
15821 @item -mabi=@var{name}
15823 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
15824 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
15827 @opindex mapcs-frame
15828 Generate a stack frame that is compliant with the ARM Procedure Call
15829 Standard for all functions, even if this is not strictly necessary for
15830 correct execution of the code. Specifying @option{-fomit-frame-pointer}
15831 with this option causes the stack frames not to be generated for
15832 leaf functions. The default is @option{-mno-apcs-frame}.
15833 This option is deprecated.
15837 This is a synonym for @option{-mapcs-frame} and is deprecated.
15840 @c not currently implemented
15841 @item -mapcs-stack-check
15842 @opindex mapcs-stack-check
15843 Generate code to check the amount of stack space available upon entry to
15844 every function (that actually uses some stack space). If there is
15845 insufficient space available then either the function
15846 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
15847 called, depending upon the amount of stack space required. The runtime
15848 system is required to provide these functions. The default is
15849 @option{-mno-apcs-stack-check}, since this produces smaller code.
15851 @c not currently implemented
15852 @item -mapcs-reentrant
15853 @opindex mapcs-reentrant
15854 Generate reentrant, position-independent code. The default is
15855 @option{-mno-apcs-reentrant}.
15858 @item -mthumb-interwork
15859 @opindex mthumb-interwork
15860 Generate code that supports calling between the ARM and Thumb
15861 instruction sets. Without this option, on pre-v5 architectures, the
15862 two instruction sets cannot be reliably used inside one program. The
15863 default is @option{-mno-thumb-interwork}, since slightly larger code
15864 is generated when @option{-mthumb-interwork} is specified. In AAPCS
15865 configurations this option is meaningless.
15867 @item -mno-sched-prolog
15868 @opindex mno-sched-prolog
15869 Prevent the reordering of instructions in the function prologue, or the
15870 merging of those instruction with the instructions in the function's
15871 body. This means that all functions start with a recognizable set
15872 of instructions (or in fact one of a choice from a small set of
15873 different function prologues), and this information can be used to
15874 locate the start of functions inside an executable piece of code. The
15875 default is @option{-msched-prolog}.
15877 @item -mfloat-abi=@var{name}
15878 @opindex mfloat-abi
15879 Specifies which floating-point ABI to use. Permissible values
15880 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15882 Specifying @samp{soft} causes GCC to generate output containing
15883 library calls for floating-point operations.
15884 @samp{softfp} allows the generation of code using hardware floating-point
15885 instructions, but still uses the soft-float calling conventions.
15886 @samp{hard} allows generation of floating-point instructions
15887 and uses FPU-specific calling conventions.
15889 The default depends on the specific target configuration. Note that
15890 the hard-float and soft-float ABIs are not link-compatible; you must
15891 compile your entire program with the same ABI, and link with a
15892 compatible set of libraries.
15894 @item -mlittle-endian
15895 @opindex mlittle-endian
15896 Generate code for a processor running in little-endian mode. This is
15897 the default for all standard configurations.
15900 @opindex mbig-endian
15901 Generate code for a processor running in big-endian mode; the default is
15902 to compile code for a little-endian processor.
15907 When linking a big-endian image select between BE8 and BE32 formats.
15908 The option has no effect for little-endian images and is ignored. The
15909 default is dependent on the selected target architecture. For ARMv6
15910 and later architectures the default is BE8, for older architectures
15911 the default is BE32. BE32 format has been deprecated by ARM.
15913 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
15915 This specifies the name of the target ARM architecture. GCC uses this
15916 name to determine what kind of instructions it can emit when generating
15917 assembly code. This option can be used in conjunction with or instead
15918 of the @option{-mcpu=} option.
15920 Permissible names are:
15922 @samp{armv5t}, @samp{armv5te},
15923 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
15924 @samp{armv6z}, @samp{armv6zk},
15925 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
15926 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
15930 @samp{armv6-m}, @samp{armv6s-m},
15931 @samp{armv7-m}, @samp{armv7e-m},
15932 @samp{armv8-m.base}, @samp{armv8-m.main},
15933 @samp{iwmmxt} and @samp{iwmmxt2}.
15935 Additionally, the following architectures, which lack support for the
15936 Thumb execution state, are recognized but support is deprecated:
15937 @samp{armv2}, @samp{armv2a}, @samp{armv3}, @samp{armv3m},
15938 @samp{armv4}, @samp{armv5} and @samp{armv5e}.
15940 Many of the architectures support extensions. These can be added by
15941 appending @samp{+@var{extension}} to the architecture name. Extension
15942 options are processed in order and capabilities accumulate. An extension
15943 will also enable any necessary base extensions
15944 upon which it depends. For example, the @samp{+crypto} extension
15945 will always enable the @samp{+simd} extension. The exception to the
15946 additive construction is for extensions that are prefixed with
15947 @samp{+no@dots{}}: these extensions disable the specified option and
15948 any other extensions that may depend on the presence of that
15951 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
15952 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
15953 entirely disabled by the @samp{+nofp} option that follows it.
15955 Most extension names are generically named, but have an effect that is
15956 dependent upon the architecture to which it is applied. For example,
15957 the @samp{+simd} option can be applied to both @samp{armv7-a} and
15958 @samp{armv8-a} architectures, but will enable the original ARMv7-A
15959 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
15960 variant for @samp{armv8-a}.
15962 The table below lists the supported extensions for each architecture.
15963 Architectures not mentioned do not support any extensions.
15977 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
15978 used as an alias for this extension.
15981 Disable the floating-point instructions.
15985 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
15988 The VFPv3 floating-point instructions, with 16 double-precision
15989 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15990 for this extension. Note that floating-point is not supported by the
15991 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
15992 ARMv7-R architectures.
15995 Disable the floating-point instructions.
16001 The VFPv3 floating-point instructions, with 16 double-precision
16002 registers. The extension @samp{+vfpv3-d16} can be used as an alias
16003 for this extension.
16006 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
16007 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
16008 for this extension.
16011 The VFPv3 floating-point instructions, with 32 double-precision
16014 @item +vfpv3-d16-fp16
16015 The VFPv3 floating-point instructions, with 16 double-precision
16016 registers and the half-precision floating-point conversion operations.
16019 The VFPv3 floating-point instructions, with 32 double-precision
16020 registers and the half-precision floating-point conversion operations.
16023 The VFPv4 floating-point instructions, with 16 double-precision
16027 The VFPv4 floating-point instructions, with 32 double-precision
16031 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
16032 the half-precision floating-point conversion operations.
16035 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
16038 Disable the Advanced SIMD instructions (does not disable floating point).
16041 Disable the floating-point and Advanced SIMD instructions.
16045 The extended version of the ARMv7-A architecture with support for
16049 The VFPv4 floating-point instructions, with 16 double-precision registers.
16050 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
16053 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
16054 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
16057 The VFPv3 floating-point instructions, with 16 double-precision
16061 The VFPv3 floating-point instructions, with 32 double-precision
16064 @item +vfpv3-d16-fp16
16065 The VFPv3 floating-point instructions, with 16 double-precision
16066 registers and the half-precision floating-point conversion operations.
16069 The VFPv3 floating-point instructions, with 32 double-precision
16070 registers and the half-precision floating-point conversion operations.
16073 The VFPv4 floating-point instructions, with 16 double-precision
16077 The VFPv4 floating-point instructions, with 32 double-precision
16081 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
16082 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
16085 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
16086 the half-precision floating-point conversion operations.
16089 Disable the Advanced SIMD instructions (does not disable floating point).
16092 Disable the floating-point and Advanced SIMD instructions.
16098 The Cyclic Redundancy Check (CRC) instructions.
16100 The ARMv8-A Advanced SIMD and floating-point instructions.
16102 The cryptographic instructions.
16104 Disable the cryptographic instructions.
16106 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16112 The ARMv8.1-A Advanced SIMD and floating-point instructions.
16115 The cryptographic instructions. This also enables the Advanced SIMD and
16116 floating-point instructions.
16119 Disable the cryptographic instructions.
16122 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16129 The half-precision floating-point data processing instructions.
16130 This also enables the Advanced SIMD and floating-point instructions.
16133 The half-precision floating-point fmla extension. This also enables
16134 the half-precision floating-point extension and Advanced SIMD and
16135 floating-point instructions.
16138 The ARMv8.1-A Advanced SIMD and floating-point instructions.
16141 The cryptographic instructions. This also enables the Advanced SIMD and
16142 floating-point instructions.
16145 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
16148 Disable the cryptographic extension.
16151 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16157 The half-precision floating-point data processing instructions.
16158 This also enables the Advanced SIMD and floating-point instructions as well
16159 as the Dot Product extension and the half-precision floating-point fmla
16163 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
16164 Dot Product extension.
16167 The cryptographic instructions. This also enables the Advanced SIMD and
16168 floating-point instructions as well as the Dot Product extension.
16171 Disable the cryptographic extension.
16174 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16180 The single-precision VFPv3 floating-point instructions. The extension
16181 @samp{+vfpv3xd} can be used as an alias for this extension.
16184 The VFPv3 floating-point instructions with 16 double-precision registers.
16185 The extension +vfpv3-d16 can be used as an alias for this extension.
16188 Disable the floating-point extension.
16191 The ARM-state integer division instructions.
16194 Disable the ARM-state integer division extension.
16200 The single-precision VFPv4 floating-point instructions.
16203 The single-precision FPv5 floating-point instructions.
16206 The single- and double-precision FPv5 floating-point instructions.
16209 Disable the floating-point extensions.
16215 The DSP instructions.
16218 Disable the DSP extension.
16221 The single-precision floating-point instructions.
16224 The single- and double-precision floating-point instructions.
16227 Disable the floating-point extension.
16233 The Cyclic Redundancy Check (CRC) instructions.
16235 The single-precision FPv5 floating-point instructions.
16237 The ARMv8-A Advanced SIMD and floating-point instructions.
16239 The cryptographic instructions.
16241 Disable the cryptographic instructions.
16243 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16248 @option{-march=native} causes the compiler to auto-detect the architecture
16249 of the build computer. At present, this feature is only supported on
16250 GNU/Linux, and not all architectures are recognized. If the auto-detect
16251 is unsuccessful the option has no effect.
16253 @item -mtune=@var{name}
16255 This option specifies the name of the target ARM processor for
16256 which GCC should tune the performance of the code.
16257 For some ARM implementations better performance can be obtained by using
16259 Permissible names are: @samp{arm2}, @samp{arm250},
16260 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
16261 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
16262 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
16263 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
16265 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
16266 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
16267 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
16268 @samp{strongarm1110},
16269 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
16270 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
16271 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
16272 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
16273 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
16274 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
16275 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
16276 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
16277 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
16278 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
16279 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
16280 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7},
16281 @samp{cortex-r8}, @samp{cortex-r52},
16289 @samp{cortex-m0plus},
16290 @samp{cortex-m1.small-multiply},
16291 @samp{cortex-m0.small-multiply},
16292 @samp{cortex-m0plus.small-multiply},
16294 @samp{marvell-pj4},
16295 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
16296 @samp{fa526}, @samp{fa626},
16297 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
16300 Additionally, this option can specify that GCC should tune the performance
16301 of the code for a big.LITTLE system. Permissible names are:
16302 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
16303 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
16304 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
16305 @samp{cortex-a75.cortex-a55}.
16307 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
16308 performance for a blend of processors within architecture @var{arch}.
16309 The aim is to generate code that run well on the current most popular
16310 processors, balancing between optimizations that benefit some CPUs in the
16311 range, and avoiding performance pitfalls of other CPUs. The effects of
16312 this option may change in future GCC versions as CPU models come and go.
16314 @option{-mtune} permits the same extension options as @option{-mcpu}, but
16315 the extension options do not affect the tuning of the generated code.
16317 @option{-mtune=native} causes the compiler to auto-detect the CPU
16318 of the build computer. At present, this feature is only supported on
16319 GNU/Linux, and not all architectures are recognized. If the auto-detect is
16320 unsuccessful the option has no effect.
16322 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
16324 This specifies the name of the target ARM processor. GCC uses this name
16325 to derive the name of the target ARM architecture (as if specified
16326 by @option{-march}) and the ARM processor type for which to tune for
16327 performance (as if specified by @option{-mtune}). Where this option
16328 is used in conjunction with @option{-march} or @option{-mtune},
16329 those options take precedence over the appropriate part of this option.
16331 Many of the supported CPUs implement optional architectural
16332 extensions. Where this is so the architectural extensions are
16333 normally enabled by default. If implementations that lack the
16334 extension exist, then the extension syntax can be used to disable
16335 those extensions that have been omitted. For floating-point and
16336 Advanced SIMD (Neon) instructions, the settings of the options
16337 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
16338 floating-point and Advanced SIMD instructions will only be used if
16339 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
16340 @option{-mfpu} other than @samp{auto} will override the available
16341 floating-point and SIMD extension instructions.
16343 For example, @samp{cortex-a9} can be found in three major
16344 configurations: integer only, with just a floating-point unit or with
16345 floating-point and Advanced SIMD. The default is to enable all the
16346 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
16347 be used to disable just the SIMD or both the SIMD and floating-point
16348 instructions respectively.
16350 Permissible names for this option are the same as those for
16353 The following extension options are common to the listed CPUs:
16357 Disable the DSP instructions on @samp{cortex-m33}.
16360 Disables the floating-point instructions on @samp{arm9e},
16361 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
16362 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
16363 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
16364 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
16365 Disables the floating-point and SIMD instructions on
16366 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
16367 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
16368 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
16369 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
16370 @samp{cortex-a53} and @samp{cortex-a55}.
16373 Disables the double-precision component of the floating-point instructions
16374 on @samp{cortex-r5}, @samp{cortex-r52} and @samp{cortex-m7}.
16377 Disables the SIMD (but not floating-point) instructions on
16378 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
16379 and @samp{cortex-a9}.
16382 Enables the cryptographic instructions on @samp{cortex-a32},
16383 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
16384 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
16385 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
16386 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
16387 @samp{cortex-a75.cortex-a55}.
16390 Additionally the @samp{generic-armv7-a} pseudo target defaults to
16391 VFPv3 with 16 double-precision registers. It supports the following
16392 extension options: @samp{vfpv3-d16}, @samp{vfpv3},
16393 @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16}, @samp{vfpv4-d16},
16394 @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3}, @samp{neon-fp16},
16395 @samp{neon-vfpv4}. The meanings are the same as for the extensions to
16396 @option{-march=armv7-a}.
16398 @option{-mcpu=generic-@var{arch}} is also permissible, and is
16399 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
16400 See @option{-mtune} for more information.
16402 @option{-mcpu=native} causes the compiler to auto-detect the CPU
16403 of the build computer. At present, this feature is only supported on
16404 GNU/Linux, and not all architectures are recognized. If the auto-detect
16405 is unsuccessful the option has no effect.
16407 @item -mfpu=@var{name}
16409 This specifies what floating-point hardware (or hardware emulation) is
16410 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
16412 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
16413 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
16414 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
16415 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
16416 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
16417 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
16418 is an alias for @samp{vfpv2}.
16420 The setting @samp{auto} is the default and is special. It causes the
16421 compiler to select the floating-point and Advanced SIMD instructions
16422 based on the settings of @option{-mcpu} and @option{-march}.
16424 If the selected floating-point hardware includes the NEON extension
16425 (e.g. @option{-mfpu=neon}), note that floating-point
16426 operations are not generated by GCC's auto-vectorization pass unless
16427 @option{-funsafe-math-optimizations} is also specified. This is
16428 because NEON hardware does not fully implement the IEEE 754 standard for
16429 floating-point arithmetic (in particular denormal values are treated as
16430 zero), so the use of NEON instructions may lead to a loss of precision.
16432 You can also set the fpu name at function level by using the @code{target("fpu=")} function attributes (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
16434 @item -mfp16-format=@var{name}
16435 @opindex mfp16-format
16436 Specify the format of the @code{__fp16} half-precision floating-point type.
16437 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
16438 the default is @samp{none}, in which case the @code{__fp16} type is not
16439 defined. @xref{Half-Precision}, for more information.
16441 @item -mstructure-size-boundary=@var{n}
16442 @opindex mstructure-size-boundary
16443 The sizes of all structures and unions are rounded up to a multiple
16444 of the number of bits set by this option. Permissible values are 8, 32
16445 and 64. The default value varies for different toolchains. For the COFF
16446 targeted toolchain the default value is 8. A value of 64 is only allowed
16447 if the underlying ABI supports it.
16449 Specifying a larger number can produce faster, more efficient code, but
16450 can also increase the size of the program. Different values are potentially
16451 incompatible. Code compiled with one value cannot necessarily expect to
16452 work with code or libraries compiled with another value, if they exchange
16453 information using structures or unions.
16455 This option is deprecated.
16457 @item -mabort-on-noreturn
16458 @opindex mabort-on-noreturn
16459 Generate a call to the function @code{abort} at the end of a
16460 @code{noreturn} function. It is executed if the function tries to
16464 @itemx -mno-long-calls
16465 @opindex mlong-calls
16466 @opindex mno-long-calls
16467 Tells the compiler to perform function calls by first loading the
16468 address of the function into a register and then performing a subroutine
16469 call on this register. This switch is needed if the target function
16470 lies outside of the 64-megabyte addressing range of the offset-based
16471 version of subroutine call instruction.
16473 Even if this switch is enabled, not all function calls are turned
16474 into long calls. The heuristic is that static functions, functions
16475 that have the @code{short_call} attribute, functions that are inside
16476 the scope of a @code{#pragma no_long_calls} directive, and functions whose
16477 definitions have already been compiled within the current compilation
16478 unit are not turned into long calls. The exceptions to this rule are
16479 that weak function definitions, functions with the @code{long_call}
16480 attribute or the @code{section} attribute, and functions that are within
16481 the scope of a @code{#pragma long_calls} directive are always
16482 turned into long calls.
16484 This feature is not enabled by default. Specifying
16485 @option{-mno-long-calls} restores the default behavior, as does
16486 placing the function calls within the scope of a @code{#pragma
16487 long_calls_off} directive. Note these switches have no effect on how
16488 the compiler generates code to handle function calls via function
16491 @item -msingle-pic-base
16492 @opindex msingle-pic-base
16493 Treat the register used for PIC addressing as read-only, rather than
16494 loading it in the prologue for each function. The runtime system is
16495 responsible for initializing this register with an appropriate value
16496 before execution begins.
16498 @item -mpic-register=@var{reg}
16499 @opindex mpic-register
16500 Specify the register to be used for PIC addressing.
16501 For standard PIC base case, the default is any suitable register
16502 determined by compiler. For single PIC base case, the default is
16503 @samp{R9} if target is EABI based or stack-checking is enabled,
16504 otherwise the default is @samp{R10}.
16506 @item -mpic-data-is-text-relative
16507 @opindex mpic-data-is-text-relative
16508 Assume that the displacement between the text and data segments is fixed
16509 at static link time. This permits using PC-relative addressing
16510 operations to access data known to be in the data segment. For
16511 non-VxWorks RTP targets, this option is enabled by default. When
16512 disabled on such targets, it will enable @option{-msingle-pic-base} by
16515 @item -mpoke-function-name
16516 @opindex mpoke-function-name
16517 Write the name of each function into the text section, directly
16518 preceding the function prologue. The generated code is similar to this:
16522 .ascii "arm_poke_function_name", 0
16525 .word 0xff000000 + (t1 - t0)
16526 arm_poke_function_name
16528 stmfd sp!, @{fp, ip, lr, pc@}
16532 When performing a stack backtrace, code can inspect the value of
16533 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
16534 location @code{pc - 12} and the top 8 bits are set, then we know that
16535 there is a function name embedded immediately preceding this location
16536 and has length @code{((pc[-3]) & 0xff000000)}.
16543 Select between generating code that executes in ARM and Thumb
16544 states. The default for most configurations is to generate code
16545 that executes in ARM state, but the default can be changed by
16546 configuring GCC with the @option{--with-mode=}@var{state}
16549 You can also override the ARM and Thumb mode for each function
16550 by using the @code{target("thumb")} and @code{target("arm")} function attributes
16551 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
16554 @opindex mflip-thumb
16555 Switch ARM/Thumb modes on alternating functions.
16556 This option is provided for regression testing of mixed Thumb/ARM code
16557 generation, and is not intended for ordinary use in compiling code.
16560 @opindex mtpcs-frame
16561 Generate a stack frame that is compliant with the Thumb Procedure Call
16562 Standard for all non-leaf functions. (A leaf function is one that does
16563 not call any other functions.) The default is @option{-mno-tpcs-frame}.
16565 @item -mtpcs-leaf-frame
16566 @opindex mtpcs-leaf-frame
16567 Generate a stack frame that is compliant with the Thumb Procedure Call
16568 Standard for all leaf functions. (A leaf function is one that does
16569 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
16571 @item -mcallee-super-interworking
16572 @opindex mcallee-super-interworking
16573 Gives all externally visible functions in the file being compiled an ARM
16574 instruction set header which switches to Thumb mode before executing the
16575 rest of the function. This allows these functions to be called from
16576 non-interworking code. This option is not valid in AAPCS configurations
16577 because interworking is enabled by default.
16579 @item -mcaller-super-interworking
16580 @opindex mcaller-super-interworking
16581 Allows calls via function pointers (including virtual functions) to
16582 execute correctly regardless of whether the target code has been
16583 compiled for interworking or not. There is a small overhead in the cost
16584 of executing a function pointer if this option is enabled. This option
16585 is not valid in AAPCS configurations because interworking is enabled
16588 @item -mtp=@var{name}
16590 Specify the access model for the thread local storage pointer. The valid
16591 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
16592 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
16593 (supported in the arm6k architecture), and @samp{auto}, which uses the
16594 best available method for the selected processor. The default setting is
16597 @item -mtls-dialect=@var{dialect}
16598 @opindex mtls-dialect
16599 Specify the dialect to use for accessing thread local storage. Two
16600 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
16601 @samp{gnu} dialect selects the original GNU scheme for supporting
16602 local and global dynamic TLS models. The @samp{gnu2} dialect
16603 selects the GNU descriptor scheme, which provides better performance
16604 for shared libraries. The GNU descriptor scheme is compatible with
16605 the original scheme, but does require new assembler, linker and
16606 library support. Initial and local exec TLS models are unaffected by
16607 this option and always use the original scheme.
16609 @item -mword-relocations
16610 @opindex mword-relocations
16611 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
16612 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
16613 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
16616 @item -mfix-cortex-m3-ldrd
16617 @opindex mfix-cortex-m3-ldrd
16618 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
16619 with overlapping destination and base registers are used. This option avoids
16620 generating these instructions. This option is enabled by default when
16621 @option{-mcpu=cortex-m3} is specified.
16623 @item -munaligned-access
16624 @itemx -mno-unaligned-access
16625 @opindex munaligned-access
16626 @opindex mno-unaligned-access
16627 Enables (or disables) reading and writing of 16- and 32- bit values
16628 from addresses that are not 16- or 32- bit aligned. By default
16629 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
16630 ARMv8-M Baseline architectures, and enabled for all other
16631 architectures. If unaligned access is not enabled then words in packed
16632 data structures are accessed a byte at a time.
16634 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
16635 generated object file to either true or false, depending upon the
16636 setting of this option. If unaligned access is enabled then the
16637 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
16640 @item -mneon-for-64bits
16641 @opindex mneon-for-64bits
16642 Enables using Neon to handle scalar 64-bits operations. This is
16643 disabled by default since the cost of moving data from core registers
16646 @item -mslow-flash-data
16647 @opindex mslow-flash-data
16648 Assume loading data from flash is slower than fetching instruction.
16649 Therefore literal load is minimized for better performance.
16650 This option is only supported when compiling for ARMv7 M-profile and
16653 @item -masm-syntax-unified
16654 @opindex masm-syntax-unified
16655 Assume inline assembler is using unified asm syntax. The default is
16656 currently off which implies divided syntax. This option has no impact
16657 on Thumb2. However, this may change in future releases of GCC.
16658 Divided syntax should be considered deprecated.
16660 @item -mrestrict-it
16661 @opindex mrestrict-it
16662 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
16663 IT blocks can only contain a single 16-bit instruction from a select
16664 set of instructions. This option is on by default for ARMv8-A Thumb mode.
16666 @item -mprint-tune-info
16667 @opindex mprint-tune-info
16668 Print CPU tuning information as comment in assembler file. This is
16669 an option used only for regression testing of the compiler and not
16670 intended for ordinary use in compiling code. This option is disabled
16673 @item -mverbose-cost-dump
16674 @opindex mverbose-cost-dump
16675 Enable verbose cost model dumping in the debug dump files. This option is
16676 provided for use in debugging the compiler.
16679 @opindex mpure-code
16680 Do not allow constant data to be placed in code sections.
16681 Additionally, when compiling for ELF object format give all text sections the
16682 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
16683 is only available when generating non-pic code for M-profile targets with the
16688 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
16689 Development Tools Engineering Specification", which can be found on
16690 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
16694 @subsection AVR Options
16695 @cindex AVR Options
16697 These options are defined for AVR implementations:
16700 @item -mmcu=@var{mcu}
16702 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
16704 The default for this option is@tie{}@samp{avr2}.
16706 GCC supports the following AVR devices and ISAs:
16708 @include avr-mmcu.texi
16713 Assume that all data in static storage can be accessed by LDS / STS
16714 instructions. This option has only an effect on reduced Tiny devices like
16715 ATtiny40. See also the @code{absdata}
16716 @ref{AVR Variable Attributes,variable attribute}.
16718 @item -maccumulate-args
16719 @opindex maccumulate-args
16720 Accumulate outgoing function arguments and acquire/release the needed
16721 stack space for outgoing function arguments once in function
16722 prologue/epilogue. Without this option, outgoing arguments are pushed
16723 before calling a function and popped afterwards.
16725 Popping the arguments after the function call can be expensive on
16726 AVR so that accumulating the stack space might lead to smaller
16727 executables because arguments need not be removed from the
16728 stack after such a function call.
16730 This option can lead to reduced code size for functions that perform
16731 several calls to functions that get their arguments on the stack like
16732 calls to printf-like functions.
16734 @item -mbranch-cost=@var{cost}
16735 @opindex mbranch-cost
16736 Set the branch costs for conditional branch instructions to
16737 @var{cost}. Reasonable values for @var{cost} are small, non-negative
16738 integers. The default branch cost is 0.
16740 @item -mcall-prologues
16741 @opindex mcall-prologues
16742 Functions prologues/epilogues are expanded as calls to appropriate
16743 subroutines. Code size is smaller.
16745 @item -mgas-isr-prologues
16746 @opindex mgas-isr-prologues
16747 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
16748 instruction supported by GNU Binutils.
16749 If this option is on, the feature can still be disabled for individual
16750 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
16751 function attribute. This feature is activated per default
16752 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
16753 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
16757 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
16758 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
16759 and @code{long long} is 4 bytes. Please note that this option does not
16760 conform to the C standards, but it results in smaller code
16763 @item -mmain-is-OS_task
16764 @opindex mmain-is-OS_task
16765 Do not save registers in @code{main}. The effect is the same like
16766 attaching attribute @ref{AVR Function Attributes,,@code{OS_task}}
16767 to @code{main}. It is activated per default if optimization is on.
16769 @item -mn-flash=@var{num}
16771 Assume that the flash memory has a size of
16772 @var{num} times 64@tie{}KiB.
16774 @item -mno-interrupts
16775 @opindex mno-interrupts
16776 Generated code is not compatible with hardware interrupts.
16777 Code size is smaller.
16781 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
16782 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
16783 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
16784 the assembler's command line and the @option{--relax} option to the
16785 linker's command line.
16787 Jump relaxing is performed by the linker because jump offsets are not
16788 known before code is located. Therefore, the assembler code generated by the
16789 compiler is the same, but the instructions in the executable may
16790 differ from instructions in the assembler code.
16792 Relaxing must be turned on if linker stubs are needed, see the
16793 section on @code{EIND} and linker stubs below.
16797 Assume that the device supports the Read-Modify-Write
16798 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
16800 @item -mshort-calls
16801 @opindex mshort-calls
16803 Assume that @code{RJMP} and @code{RCALL} can target the whole
16806 This option is used internally for multilib selection. It is
16807 not an optimization option, and you don't need to set it by hand.
16811 Treat the stack pointer register as an 8-bit register,
16812 i.e.@: assume the high byte of the stack pointer is zero.
16813 In general, you don't need to set this option by hand.
16815 This option is used internally by the compiler to select and
16816 build multilibs for architectures @code{avr2} and @code{avr25}.
16817 These architectures mix devices with and without @code{SPH}.
16818 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
16819 the compiler driver adds or removes this option from the compiler
16820 proper's command line, because the compiler then knows if the device
16821 or architecture has an 8-bit stack pointer and thus no @code{SPH}
16826 Use address register @code{X} in a way proposed by the hardware. This means
16827 that @code{X} is only used in indirect, post-increment or
16828 pre-decrement addressing.
16830 Without this option, the @code{X} register may be used in the same way
16831 as @code{Y} or @code{Z} which then is emulated by additional
16833 For example, loading a value with @code{X+const} addressing with a
16834 small non-negative @code{const < 64} to a register @var{Rn} is
16838 adiw r26, const ; X += const
16839 ld @var{Rn}, X ; @var{Rn} = *X
16840 sbiw r26, const ; X -= const
16844 @opindex mtiny-stack
16845 Only change the lower 8@tie{}bits of the stack pointer.
16847 @item -mfract-convert-truncate
16848 @opindex mfract-convert-truncate
16849 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
16852 @opindex nodevicelib
16853 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
16855 @item -Waddr-space-convert
16856 @opindex Waddr-space-convert
16857 Warn about conversions between address spaces in the case where the
16858 resulting address space is not contained in the incoming address space.
16860 @item -Wmisspelled-isr
16861 @opindex Wmisspelled-isr
16862 Warn if the ISR is misspelled, i.e. without __vector prefix.
16863 Enabled by default.
16866 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
16867 @cindex @code{EIND}
16868 Pointers in the implementation are 16@tie{}bits wide.
16869 The address of a function or label is represented as word address so
16870 that indirect jumps and calls can target any code address in the
16871 range of 64@tie{}Ki words.
16873 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
16874 bytes of program memory space, there is a special function register called
16875 @code{EIND} that serves as most significant part of the target address
16876 when @code{EICALL} or @code{EIJMP} instructions are used.
16878 Indirect jumps and calls on these devices are handled as follows by
16879 the compiler and are subject to some limitations:
16884 The compiler never sets @code{EIND}.
16887 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
16888 instructions or might read @code{EIND} directly in order to emulate an
16889 indirect call/jump by means of a @code{RET} instruction.
16892 The compiler assumes that @code{EIND} never changes during the startup
16893 code or during the application. In particular, @code{EIND} is not
16894 saved/restored in function or interrupt service routine
16898 For indirect calls to functions and computed goto, the linker
16899 generates @emph{stubs}. Stubs are jump pads sometimes also called
16900 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
16901 The stub contains a direct jump to the desired address.
16904 Linker relaxation must be turned on so that the linker generates
16905 the stubs correctly in all situations. See the compiler option
16906 @option{-mrelax} and the linker option @option{--relax}.
16907 There are corner cases where the linker is supposed to generate stubs
16908 but aborts without relaxation and without a helpful error message.
16911 The default linker script is arranged for code with @code{EIND = 0}.
16912 If code is supposed to work for a setup with @code{EIND != 0}, a custom
16913 linker script has to be used in order to place the sections whose
16914 name start with @code{.trampolines} into the segment where @code{EIND}
16918 The startup code from libgcc never sets @code{EIND}.
16919 Notice that startup code is a blend of code from libgcc and AVR-LibC.
16920 For the impact of AVR-LibC on @code{EIND}, see the
16921 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
16924 It is legitimate for user-specific startup code to set up @code{EIND}
16925 early, for example by means of initialization code located in
16926 section @code{.init3}. Such code runs prior to general startup code
16927 that initializes RAM and calls constructors, but after the bit
16928 of startup code from AVR-LibC that sets @code{EIND} to the segment
16929 where the vector table is located.
16931 #include <avr/io.h>
16934 __attribute__((section(".init3"),naked,used,no_instrument_function))
16935 init3_set_eind (void)
16937 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
16938 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
16943 The @code{__trampolines_start} symbol is defined in the linker script.
16946 Stubs are generated automatically by the linker if
16947 the following two conditions are met:
16950 @item The address of a label is taken by means of the @code{gs} modifier
16951 (short for @emph{generate stubs}) like so:
16953 LDI r24, lo8(gs(@var{func}))
16954 LDI r25, hi8(gs(@var{func}))
16956 @item The final location of that label is in a code segment
16957 @emph{outside} the segment where the stubs are located.
16961 The compiler emits such @code{gs} modifiers for code labels in the
16962 following situations:
16964 @item Taking address of a function or code label.
16965 @item Computed goto.
16966 @item If prologue-save function is used, see @option{-mcall-prologues}
16967 command-line option.
16968 @item Switch/case dispatch tables. If you do not want such dispatch
16969 tables you can specify the @option{-fno-jump-tables} command-line option.
16970 @item C and C++ constructors/destructors called during startup/shutdown.
16971 @item If the tools hit a @code{gs()} modifier explained above.
16975 Jumping to non-symbolic addresses like so is @emph{not} supported:
16980 /* Call function at word address 0x2 */
16981 return ((int(*)(void)) 0x2)();
16985 Instead, a stub has to be set up, i.e.@: the function has to be called
16986 through a symbol (@code{func_4} in the example):
16991 extern int func_4 (void);
16993 /* Call function at byte address 0x4 */
16998 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
16999 Alternatively, @code{func_4} can be defined in the linker script.
17002 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
17003 @cindex @code{RAMPD}
17004 @cindex @code{RAMPX}
17005 @cindex @code{RAMPY}
17006 @cindex @code{RAMPZ}
17007 Some AVR devices support memories larger than the 64@tie{}KiB range
17008 that can be accessed with 16-bit pointers. To access memory locations
17009 outside this 64@tie{}KiB range, the content of a @code{RAMP}
17010 register is used as high part of the address:
17011 The @code{X}, @code{Y}, @code{Z} address register is concatenated
17012 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
17013 register, respectively, to get a wide address. Similarly,
17014 @code{RAMPD} is used together with direct addressing.
17018 The startup code initializes the @code{RAMP} special function
17019 registers with zero.
17022 If a @ref{AVR Named Address Spaces,named address space} other than
17023 generic or @code{__flash} is used, then @code{RAMPZ} is set
17024 as needed before the operation.
17027 If the device supports RAM larger than 64@tie{}KiB and the compiler
17028 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
17029 is reset to zero after the operation.
17032 If the device comes with a specific @code{RAMP} register, the ISR
17033 prologue/epilogue saves/restores that SFR and initializes it with
17034 zero in case the ISR code might (implicitly) use it.
17037 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
17038 If you use inline assembler to read from locations outside the
17039 16-bit address range and change one of the @code{RAMP} registers,
17040 you must reset it to zero after the access.
17044 @subsubsection AVR Built-in Macros
17046 GCC defines several built-in macros so that the user code can test
17047 for the presence or absence of features. Almost any of the following
17048 built-in macros are deduced from device capabilities and thus
17049 triggered by the @option{-mmcu=} command-line option.
17051 For even more AVR-specific built-in macros see
17052 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
17057 Build-in macro that resolves to a decimal number that identifies the
17058 architecture and depends on the @option{-mmcu=@var{mcu}} option.
17059 Possible values are:
17061 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
17062 @code{4}, @code{5}, @code{51}, @code{6}
17064 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
17065 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
17070 @code{102}, @code{103}, @code{104},
17071 @code{105}, @code{106}, @code{107}
17073 for @var{mcu}=@code{avrtiny},
17074 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
17075 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
17076 If @var{mcu} specifies a device, this built-in macro is set
17077 accordingly. For example, with @option{-mmcu=atmega8} the macro is
17078 defined to @code{4}.
17080 @item __AVR_@var{Device}__
17081 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
17082 the device's name. For example, @option{-mmcu=atmega8} defines the
17083 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
17084 @code{__AVR_ATtiny261A__}, etc.
17086 The built-in macros' names follow
17087 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
17088 the device name as from the AVR user manual. The difference between
17089 @var{Device} in the built-in macro and @var{device} in
17090 @option{-mmcu=@var{device}} is that the latter is always lowercase.
17092 If @var{device} is not a device but only a core architecture like
17093 @samp{avr51}, this macro is not defined.
17095 @item __AVR_DEVICE_NAME__
17096 Setting @option{-mmcu=@var{device}} defines this built-in macro to
17097 the device's name. For example, with @option{-mmcu=atmega8} the macro
17098 is defined to @code{atmega8}.
17100 If @var{device} is not a device but only a core architecture like
17101 @samp{avr51}, this macro is not defined.
17103 @item __AVR_XMEGA__
17104 The device / architecture belongs to the XMEGA family of devices.
17106 @item __AVR_HAVE_ELPM__
17107 The device has the @code{ELPM} instruction.
17109 @item __AVR_HAVE_ELPMX__
17110 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
17111 R@var{n},Z+} instructions.
17113 @item __AVR_HAVE_MOVW__
17114 The device has the @code{MOVW} instruction to perform 16-bit
17115 register-register moves.
17117 @item __AVR_HAVE_LPMX__
17118 The device has the @code{LPM R@var{n},Z} and
17119 @code{LPM R@var{n},Z+} instructions.
17121 @item __AVR_HAVE_MUL__
17122 The device has a hardware multiplier.
17124 @item __AVR_HAVE_JMP_CALL__
17125 The device has the @code{JMP} and @code{CALL} instructions.
17126 This is the case for devices with more than 8@tie{}KiB of program
17129 @item __AVR_HAVE_EIJMP_EICALL__
17130 @itemx __AVR_3_BYTE_PC__
17131 The device has the @code{EIJMP} and @code{EICALL} instructions.
17132 This is the case for devices with more than 128@tie{}KiB of program memory.
17133 This also means that the program counter
17134 (PC) is 3@tie{}bytes wide.
17136 @item __AVR_2_BYTE_PC__
17137 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
17138 with up to 128@tie{}KiB of program memory.
17140 @item __AVR_HAVE_8BIT_SP__
17141 @itemx __AVR_HAVE_16BIT_SP__
17142 The stack pointer (SP) register is treated as 8-bit respectively
17143 16-bit register by the compiler.
17144 The definition of these macros is affected by @option{-mtiny-stack}.
17146 @item __AVR_HAVE_SPH__
17148 The device has the SPH (high part of stack pointer) special function
17149 register or has an 8-bit stack pointer, respectively.
17150 The definition of these macros is affected by @option{-mmcu=} and
17151 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
17154 @item __AVR_HAVE_RAMPD__
17155 @itemx __AVR_HAVE_RAMPX__
17156 @itemx __AVR_HAVE_RAMPY__
17157 @itemx __AVR_HAVE_RAMPZ__
17158 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
17159 @code{RAMPZ} special function register, respectively.
17161 @item __NO_INTERRUPTS__
17162 This macro reflects the @option{-mno-interrupts} command-line option.
17164 @item __AVR_ERRATA_SKIP__
17165 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
17166 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
17167 instructions because of a hardware erratum. Skip instructions are
17168 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
17169 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
17172 @item __AVR_ISA_RMW__
17173 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
17175 @item __AVR_SFR_OFFSET__=@var{offset}
17176 Instructions that can address I/O special function registers directly
17177 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
17178 address as if addressed by an instruction to access RAM like @code{LD}
17179 or @code{STS}. This offset depends on the device architecture and has
17180 to be subtracted from the RAM address in order to get the
17181 respective I/O@tie{}address.
17183 @item __AVR_SHORT_CALLS__
17184 The @option{-mshort-calls} command line option is set.
17186 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
17187 Some devices support reading from flash memory by means of @code{LD*}
17188 instructions. The flash memory is seen in the data address space
17189 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
17190 is not defined, this feature is not available. If defined,
17191 the address space is linear and there is no need to put
17192 @code{.rodata} into RAM. This is handled by the default linker
17193 description file, and is currently available for
17194 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
17195 there is no need to use address spaces like @code{__flash} or
17196 features like attribute @code{progmem} and @code{pgm_read_*}.
17198 @item __WITH_AVRLIBC__
17199 The compiler is configured to be used together with AVR-Libc.
17200 See the @option{--with-avrlibc} configure option.
17204 @node Blackfin Options
17205 @subsection Blackfin Options
17206 @cindex Blackfin Options
17209 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
17211 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
17212 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
17213 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
17214 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
17215 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
17216 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
17217 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
17218 @samp{bf561}, @samp{bf592}.
17220 The optional @var{sirevision} specifies the silicon revision of the target
17221 Blackfin processor. Any workarounds available for the targeted silicon revision
17222 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
17223 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
17224 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
17225 hexadecimal digits representing the major and minor numbers in the silicon
17226 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
17227 is not defined. If @var{sirevision} is @samp{any}, the
17228 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
17229 If this optional @var{sirevision} is not used, GCC assumes the latest known
17230 silicon revision of the targeted Blackfin processor.
17232 GCC defines a preprocessor macro for the specified @var{cpu}.
17233 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
17234 provided by libgloss to be linked in if @option{-msim} is not given.
17236 Without this option, @samp{bf532} is used as the processor by default.
17238 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
17239 only the preprocessor macro is defined.
17243 Specifies that the program will be run on the simulator. This causes
17244 the simulator BSP provided by libgloss to be linked in. This option
17245 has effect only for @samp{bfin-elf} toolchain.
17246 Certain other options, such as @option{-mid-shared-library} and
17247 @option{-mfdpic}, imply @option{-msim}.
17249 @item -momit-leaf-frame-pointer
17250 @opindex momit-leaf-frame-pointer
17251 Don't keep the frame pointer in a register for leaf functions. This
17252 avoids the instructions to save, set up and restore frame pointers and
17253 makes an extra register available in leaf functions.
17255 @item -mspecld-anomaly
17256 @opindex mspecld-anomaly
17257 When enabled, the compiler ensures that the generated code does not
17258 contain speculative loads after jump instructions. If this option is used,
17259 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
17261 @item -mno-specld-anomaly
17262 @opindex mno-specld-anomaly
17263 Don't generate extra code to prevent speculative loads from occurring.
17265 @item -mcsync-anomaly
17266 @opindex mcsync-anomaly
17267 When enabled, the compiler ensures that the generated code does not
17268 contain CSYNC or SSYNC instructions too soon after conditional branches.
17269 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
17271 @item -mno-csync-anomaly
17272 @opindex mno-csync-anomaly
17273 Don't generate extra code to prevent CSYNC or SSYNC instructions from
17274 occurring too soon after a conditional branch.
17278 When enabled, the compiler is free to take advantage of the knowledge that
17279 the entire program fits into the low 64k of memory.
17282 @opindex mno-low-64k
17283 Assume that the program is arbitrarily large. This is the default.
17285 @item -mstack-check-l1
17286 @opindex mstack-check-l1
17287 Do stack checking using information placed into L1 scratchpad memory by the
17290 @item -mid-shared-library
17291 @opindex mid-shared-library
17292 Generate code that supports shared libraries via the library ID method.
17293 This allows for execute in place and shared libraries in an environment
17294 without virtual memory management. This option implies @option{-fPIC}.
17295 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17297 @item -mno-id-shared-library
17298 @opindex mno-id-shared-library
17299 Generate code that doesn't assume ID-based shared libraries are being used.
17300 This is the default.
17302 @item -mleaf-id-shared-library
17303 @opindex mleaf-id-shared-library
17304 Generate code that supports shared libraries via the library ID method,
17305 but assumes that this library or executable won't link against any other
17306 ID shared libraries. That allows the compiler to use faster code for jumps
17309 @item -mno-leaf-id-shared-library
17310 @opindex mno-leaf-id-shared-library
17311 Do not assume that the code being compiled won't link against any ID shared
17312 libraries. Slower code is generated for jump and call insns.
17314 @item -mshared-library-id=n
17315 @opindex mshared-library-id
17316 Specifies the identification number of the ID-based shared library being
17317 compiled. Specifying a value of 0 generates more compact code; specifying
17318 other values forces the allocation of that number to the current
17319 library but is no more space- or time-efficient than omitting this option.
17323 Generate code that allows the data segment to be located in a different
17324 area of memory from the text segment. This allows for execute in place in
17325 an environment without virtual memory management by eliminating relocations
17326 against the text section.
17328 @item -mno-sep-data
17329 @opindex mno-sep-data
17330 Generate code that assumes that the data segment follows the text segment.
17331 This is the default.
17334 @itemx -mno-long-calls
17335 @opindex mlong-calls
17336 @opindex mno-long-calls
17337 Tells the compiler to perform function calls by first loading the
17338 address of the function into a register and then performing a subroutine
17339 call on this register. This switch is needed if the target function
17340 lies outside of the 24-bit addressing range of the offset-based
17341 version of subroutine call instruction.
17343 This feature is not enabled by default. Specifying
17344 @option{-mno-long-calls} restores the default behavior. Note these
17345 switches have no effect on how the compiler generates code to handle
17346 function calls via function pointers.
17350 Link with the fast floating-point library. This library relaxes some of
17351 the IEEE floating-point standard's rules for checking inputs against
17352 Not-a-Number (NAN), in the interest of performance.
17355 @opindex minline-plt
17356 Enable inlining of PLT entries in function calls to functions that are
17357 not known to bind locally. It has no effect without @option{-mfdpic}.
17360 @opindex mmulticore
17361 Build a standalone application for multicore Blackfin processors.
17362 This option causes proper start files and link scripts supporting
17363 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
17364 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
17366 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
17367 selects the one-application-per-core programming model. Without
17368 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
17369 programming model is used. In this model, the main function of Core B
17370 should be named as @code{coreb_main}.
17372 If this option is not used, the single-core application programming
17377 Build a standalone application for Core A of BF561 when using
17378 the one-application-per-core programming model. Proper start files
17379 and link scripts are used to support Core A, and the macro
17380 @code{__BFIN_COREA} is defined.
17381 This option can only be used in conjunction with @option{-mmulticore}.
17385 Build a standalone application for Core B of BF561 when using
17386 the one-application-per-core programming model. Proper start files
17387 and link scripts are used to support Core B, and the macro
17388 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
17389 should be used instead of @code{main}.
17390 This option can only be used in conjunction with @option{-mmulticore}.
17394 Build a standalone application for SDRAM. Proper start files and
17395 link scripts are used to put the application into SDRAM, and the macro
17396 @code{__BFIN_SDRAM} is defined.
17397 The loader should initialize SDRAM before loading the application.
17401 Assume that ICPLBs are enabled at run time. This has an effect on certain
17402 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
17403 are enabled; for standalone applications the default is off.
17407 @subsection C6X Options
17408 @cindex C6X Options
17411 @item -march=@var{name}
17413 This specifies the name of the target architecture. GCC uses this
17414 name to determine what kind of instructions it can emit when generating
17415 assembly code. Permissible names are: @samp{c62x},
17416 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
17419 @opindex mbig-endian
17420 Generate code for a big-endian target.
17422 @item -mlittle-endian
17423 @opindex mlittle-endian
17424 Generate code for a little-endian target. This is the default.
17428 Choose startup files and linker script suitable for the simulator.
17430 @item -msdata=default
17431 @opindex msdata=default
17432 Put small global and static data in the @code{.neardata} section,
17433 which is pointed to by register @code{B14}. Put small uninitialized
17434 global and static data in the @code{.bss} section, which is adjacent
17435 to the @code{.neardata} section. Put small read-only data into the
17436 @code{.rodata} section. The corresponding sections used for large
17437 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
17440 @opindex msdata=all
17441 Put all data, not just small objects, into the sections reserved for
17442 small data, and use addressing relative to the @code{B14} register to
17446 @opindex msdata=none
17447 Make no use of the sections reserved for small data, and use absolute
17448 addresses to access all data. Put all initialized global and static
17449 data in the @code{.fardata} section, and all uninitialized data in the
17450 @code{.far} section. Put all constant data into the @code{.const}
17455 @subsection CRIS Options
17456 @cindex CRIS Options
17458 These options are defined specifically for the CRIS ports.
17461 @item -march=@var{architecture-type}
17462 @itemx -mcpu=@var{architecture-type}
17465 Generate code for the specified architecture. The choices for
17466 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
17467 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
17468 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
17471 @item -mtune=@var{architecture-type}
17473 Tune to @var{architecture-type} everything applicable about the generated
17474 code, except for the ABI and the set of available instructions. The
17475 choices for @var{architecture-type} are the same as for
17476 @option{-march=@var{architecture-type}}.
17478 @item -mmax-stack-frame=@var{n}
17479 @opindex mmax-stack-frame
17480 Warn when the stack frame of a function exceeds @var{n} bytes.
17486 The options @option{-metrax4} and @option{-metrax100} are synonyms for
17487 @option{-march=v3} and @option{-march=v8} respectively.
17489 @item -mmul-bug-workaround
17490 @itemx -mno-mul-bug-workaround
17491 @opindex mmul-bug-workaround
17492 @opindex mno-mul-bug-workaround
17493 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
17494 models where it applies. This option is active by default.
17498 Enable CRIS-specific verbose debug-related information in the assembly
17499 code. This option also has the effect of turning off the @samp{#NO_APP}
17500 formatted-code indicator to the assembler at the beginning of the
17505 Do not use condition-code results from previous instruction; always emit
17506 compare and test instructions before use of condition codes.
17508 @item -mno-side-effects
17509 @opindex mno-side-effects
17510 Do not emit instructions with side effects in addressing modes other than
17513 @item -mstack-align
17514 @itemx -mno-stack-align
17515 @itemx -mdata-align
17516 @itemx -mno-data-align
17517 @itemx -mconst-align
17518 @itemx -mno-const-align
17519 @opindex mstack-align
17520 @opindex mno-stack-align
17521 @opindex mdata-align
17522 @opindex mno-data-align
17523 @opindex mconst-align
17524 @opindex mno-const-align
17525 These options (@samp{no-} options) arrange (eliminate arrangements) for the
17526 stack frame, individual data and constants to be aligned for the maximum
17527 single data access size for the chosen CPU model. The default is to
17528 arrange for 32-bit alignment. ABI details such as structure layout are
17529 not affected by these options.
17537 Similar to the stack- data- and const-align options above, these options
17538 arrange for stack frame, writable data and constants to all be 32-bit,
17539 16-bit or 8-bit aligned. The default is 32-bit alignment.
17541 @item -mno-prologue-epilogue
17542 @itemx -mprologue-epilogue
17543 @opindex mno-prologue-epilogue
17544 @opindex mprologue-epilogue
17545 With @option{-mno-prologue-epilogue}, the normal function prologue and
17546 epilogue which set up the stack frame are omitted and no return
17547 instructions or return sequences are generated in the code. Use this
17548 option only together with visual inspection of the compiled code: no
17549 warnings or errors are generated when call-saved registers must be saved,
17550 or storage for local variables needs to be allocated.
17554 @opindex mno-gotplt
17556 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
17557 instruction sequences that load addresses for functions from the PLT part
17558 of the GOT rather than (traditional on other architectures) calls to the
17559 PLT@. The default is @option{-mgotplt}.
17563 Legacy no-op option only recognized with the cris-axis-elf and
17564 cris-axis-linux-gnu targets.
17568 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
17572 This option, recognized for the cris-axis-elf, arranges
17573 to link with input-output functions from a simulator library. Code,
17574 initialized data and zero-initialized data are allocated consecutively.
17578 Like @option{-sim}, but pass linker options to locate initialized data at
17579 0x40000000 and zero-initialized data at 0x80000000.
17583 @subsection CR16 Options
17584 @cindex CR16 Options
17586 These options are defined specifically for the CR16 ports.
17592 Enable the use of multiply-accumulate instructions. Disabled by default.
17596 @opindex mcr16cplus
17598 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
17603 Links the library libsim.a which is in compatible with simulator. Applicable
17604 to ELF compiler only.
17608 Choose integer type as 32-bit wide.
17612 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
17614 @item -mdata-model=@var{model}
17615 @opindex mdata-model
17616 Choose a data model. The choices for @var{model} are @samp{near},
17617 @samp{far} or @samp{medium}. @samp{medium} is default.
17618 However, @samp{far} is not valid with @option{-mcr16c}, as the
17619 CR16C architecture does not support the far data model.
17622 @node Darwin Options
17623 @subsection Darwin Options
17624 @cindex Darwin options
17626 These options are defined for all architectures running the Darwin operating
17629 FSF GCC on Darwin does not create ``fat'' object files; it creates
17630 an object file for the single architecture that GCC was built to
17631 target. Apple's GCC on Darwin does create ``fat'' files if multiple
17632 @option{-arch} options are used; it does so by running the compiler or
17633 linker multiple times and joining the results together with
17636 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
17637 @samp{i686}) is determined by the flags that specify the ISA
17638 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
17639 @option{-force_cpusubtype_ALL} option can be used to override this.
17641 The Darwin tools vary in their behavior when presented with an ISA
17642 mismatch. The assembler, @file{as}, only permits instructions to
17643 be used that are valid for the subtype of the file it is generating,
17644 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
17645 The linker for shared libraries, @file{/usr/bin/libtool}, fails
17646 and prints an error if asked to create a shared library with a less
17647 restrictive subtype than its input files (for instance, trying to put
17648 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
17649 for executables, @command{ld}, quietly gives the executable the most
17650 restrictive subtype of any of its input files.
17655 Add the framework directory @var{dir} to the head of the list of
17656 directories to be searched for header files. These directories are
17657 interleaved with those specified by @option{-I} options and are
17658 scanned in a left-to-right order.
17660 A framework directory is a directory with frameworks in it. A
17661 framework is a directory with a @file{Headers} and/or
17662 @file{PrivateHeaders} directory contained directly in it that ends
17663 in @file{.framework}. The name of a framework is the name of this
17664 directory excluding the @file{.framework}. Headers associated with
17665 the framework are found in one of those two directories, with
17666 @file{Headers} being searched first. A subframework is a framework
17667 directory that is in a framework's @file{Frameworks} directory.
17668 Includes of subframework headers can only appear in a header of a
17669 framework that contains the subframework, or in a sibling subframework
17670 header. Two subframeworks are siblings if they occur in the same
17671 framework. A subframework should not have the same name as a
17672 framework; a warning is issued if this is violated. Currently a
17673 subframework cannot have subframeworks; in the future, the mechanism
17674 may be extended to support this. The standard frameworks can be found
17675 in @file{/System/Library/Frameworks} and
17676 @file{/Library/Frameworks}. An example include looks like
17677 @code{#include <Framework/header.h>}, where @file{Framework} denotes
17678 the name of the framework and @file{header.h} is found in the
17679 @file{PrivateHeaders} or @file{Headers} directory.
17681 @item -iframework@var{dir}
17682 @opindex iframework
17683 Like @option{-F} except the directory is a treated as a system
17684 directory. The main difference between this @option{-iframework} and
17685 @option{-F} is that with @option{-iframework} the compiler does not
17686 warn about constructs contained within header files found via
17687 @var{dir}. This option is valid only for the C family of languages.
17691 Emit debugging information for symbols that are used. For stabs
17692 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
17693 This is by default ON@.
17697 Emit debugging information for all symbols and types.
17699 @item -mmacosx-version-min=@var{version}
17700 The earliest version of MacOS X that this executable will run on
17701 is @var{version}. Typical values of @var{version} include @code{10.1},
17702 @code{10.2}, and @code{10.3.9}.
17704 If the compiler was built to use the system's headers by default,
17705 then the default for this option is the system version on which the
17706 compiler is running, otherwise the default is to make choices that
17707 are compatible with as many systems and code bases as possible.
17711 Enable kernel development mode. The @option{-mkernel} option sets
17712 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
17713 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
17714 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
17715 applicable. This mode also sets @option{-mno-altivec},
17716 @option{-msoft-float}, @option{-fno-builtin} and
17717 @option{-mlong-branch} for PowerPC targets.
17719 @item -mone-byte-bool
17720 @opindex mone-byte-bool
17721 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
17722 By default @code{sizeof(bool)} is @code{4} when compiling for
17723 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
17724 option has no effect on x86.
17726 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
17727 to generate code that is not binary compatible with code generated
17728 without that switch. Using this switch may require recompiling all
17729 other modules in a program, including system libraries. Use this
17730 switch to conform to a non-default data model.
17732 @item -mfix-and-continue
17733 @itemx -ffix-and-continue
17734 @itemx -findirect-data
17735 @opindex mfix-and-continue
17736 @opindex ffix-and-continue
17737 @opindex findirect-data
17738 Generate code suitable for fast turnaround development, such as to
17739 allow GDB to dynamically load @file{.o} files into already-running
17740 programs. @option{-findirect-data} and @option{-ffix-and-continue}
17741 are provided for backwards compatibility.
17745 Loads all members of static archive libraries.
17746 See man ld(1) for more information.
17748 @item -arch_errors_fatal
17749 @opindex arch_errors_fatal
17750 Cause the errors having to do with files that have the wrong architecture
17753 @item -bind_at_load
17754 @opindex bind_at_load
17755 Causes the output file to be marked such that the dynamic linker will
17756 bind all undefined references when the file is loaded or launched.
17760 Produce a Mach-o bundle format file.
17761 See man ld(1) for more information.
17763 @item -bundle_loader @var{executable}
17764 @opindex bundle_loader
17765 This option specifies the @var{executable} that will load the build
17766 output file being linked. See man ld(1) for more information.
17769 @opindex dynamiclib
17770 When passed this option, GCC produces a dynamic library instead of
17771 an executable when linking, using the Darwin @file{libtool} command.
17773 @item -force_cpusubtype_ALL
17774 @opindex force_cpusubtype_ALL
17775 This causes GCC's output file to have the @samp{ALL} subtype, instead of
17776 one controlled by the @option{-mcpu} or @option{-march} option.
17778 @item -allowable_client @var{client_name}
17779 @itemx -client_name
17780 @itemx -compatibility_version
17781 @itemx -current_version
17783 @itemx -dependency-file
17785 @itemx -dylinker_install_name
17787 @itemx -exported_symbols_list
17790 @itemx -flat_namespace
17791 @itemx -force_flat_namespace
17792 @itemx -headerpad_max_install_names
17795 @itemx -install_name
17796 @itemx -keep_private_externs
17797 @itemx -multi_module
17798 @itemx -multiply_defined
17799 @itemx -multiply_defined_unused
17802 @itemx -no_dead_strip_inits_and_terms
17803 @itemx -nofixprebinding
17804 @itemx -nomultidefs
17806 @itemx -noseglinkedit
17807 @itemx -pagezero_size
17809 @itemx -prebind_all_twolevel_modules
17810 @itemx -private_bundle
17812 @itemx -read_only_relocs
17814 @itemx -sectobjectsymbols
17818 @itemx -sectobjectsymbols
17821 @itemx -segs_read_only_addr
17823 @itemx -segs_read_write_addr
17824 @itemx -seg_addr_table
17825 @itemx -seg_addr_table_filename
17826 @itemx -seglinkedit
17828 @itemx -segs_read_only_addr
17829 @itemx -segs_read_write_addr
17830 @itemx -single_module
17832 @itemx -sub_library
17834 @itemx -sub_umbrella
17835 @itemx -twolevel_namespace
17838 @itemx -unexported_symbols_list
17839 @itemx -weak_reference_mismatches
17840 @itemx -whatsloaded
17841 @opindex allowable_client
17842 @opindex client_name
17843 @opindex compatibility_version
17844 @opindex current_version
17845 @opindex dead_strip
17846 @opindex dependency-file
17847 @opindex dylib_file
17848 @opindex dylinker_install_name
17850 @opindex exported_symbols_list
17852 @opindex flat_namespace
17853 @opindex force_flat_namespace
17854 @opindex headerpad_max_install_names
17855 @opindex image_base
17857 @opindex install_name
17858 @opindex keep_private_externs
17859 @opindex multi_module
17860 @opindex multiply_defined
17861 @opindex multiply_defined_unused
17862 @opindex noall_load
17863 @opindex no_dead_strip_inits_and_terms
17864 @opindex nofixprebinding
17865 @opindex nomultidefs
17867 @opindex noseglinkedit
17868 @opindex pagezero_size
17870 @opindex prebind_all_twolevel_modules
17871 @opindex private_bundle
17872 @opindex read_only_relocs
17874 @opindex sectobjectsymbols
17877 @opindex sectcreate
17878 @opindex sectobjectsymbols
17881 @opindex segs_read_only_addr
17882 @opindex segs_read_write_addr
17883 @opindex seg_addr_table
17884 @opindex seg_addr_table_filename
17885 @opindex seglinkedit
17887 @opindex segs_read_only_addr
17888 @opindex segs_read_write_addr
17889 @opindex single_module
17891 @opindex sub_library
17892 @opindex sub_umbrella
17893 @opindex twolevel_namespace
17896 @opindex unexported_symbols_list
17897 @opindex weak_reference_mismatches
17898 @opindex whatsloaded
17899 These options are passed to the Darwin linker. The Darwin linker man page
17900 describes them in detail.
17903 @node DEC Alpha Options
17904 @subsection DEC Alpha Options
17906 These @samp{-m} options are defined for the DEC Alpha implementations:
17909 @item -mno-soft-float
17910 @itemx -msoft-float
17911 @opindex mno-soft-float
17912 @opindex msoft-float
17913 Use (do not use) the hardware floating-point instructions for
17914 floating-point operations. When @option{-msoft-float} is specified,
17915 functions in @file{libgcc.a} are used to perform floating-point
17916 operations. Unless they are replaced by routines that emulate the
17917 floating-point operations, or compiled in such a way as to call such
17918 emulations routines, these routines issue floating-point
17919 operations. If you are compiling for an Alpha without floating-point
17920 operations, you must ensure that the library is built so as not to call
17923 Note that Alpha implementations without floating-point operations are
17924 required to have floating-point registers.
17927 @itemx -mno-fp-regs
17929 @opindex mno-fp-regs
17930 Generate code that uses (does not use) the floating-point register set.
17931 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
17932 register set is not used, floating-point operands are passed in integer
17933 registers as if they were integers and floating-point results are passed
17934 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
17935 so any function with a floating-point argument or return value called by code
17936 compiled with @option{-mno-fp-regs} must also be compiled with that
17939 A typical use of this option is building a kernel that does not use,
17940 and hence need not save and restore, any floating-point registers.
17944 The Alpha architecture implements floating-point hardware optimized for
17945 maximum performance. It is mostly compliant with the IEEE floating-point
17946 standard. However, for full compliance, software assistance is
17947 required. This option generates code fully IEEE-compliant code
17948 @emph{except} that the @var{inexact-flag} is not maintained (see below).
17949 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
17950 defined during compilation. The resulting code is less efficient but is
17951 able to correctly support denormalized numbers and exceptional IEEE
17952 values such as not-a-number and plus/minus infinity. Other Alpha
17953 compilers call this option @option{-ieee_with_no_inexact}.
17955 @item -mieee-with-inexact
17956 @opindex mieee-with-inexact
17957 This is like @option{-mieee} except the generated code also maintains
17958 the IEEE @var{inexact-flag}. Turning on this option causes the
17959 generated code to implement fully-compliant IEEE math. In addition to
17960 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
17961 macro. On some Alpha implementations the resulting code may execute
17962 significantly slower than the code generated by default. Since there is
17963 very little code that depends on the @var{inexact-flag}, you should
17964 normally not specify this option. Other Alpha compilers call this
17965 option @option{-ieee_with_inexact}.
17967 @item -mfp-trap-mode=@var{trap-mode}
17968 @opindex mfp-trap-mode
17969 This option controls what floating-point related traps are enabled.
17970 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
17971 The trap mode can be set to one of four values:
17975 This is the default (normal) setting. The only traps that are enabled
17976 are the ones that cannot be disabled in software (e.g., division by zero
17980 In addition to the traps enabled by @samp{n}, underflow traps are enabled
17984 Like @samp{u}, but the instructions are marked to be safe for software
17985 completion (see Alpha architecture manual for details).
17988 Like @samp{su}, but inexact traps are enabled as well.
17991 @item -mfp-rounding-mode=@var{rounding-mode}
17992 @opindex mfp-rounding-mode
17993 Selects the IEEE rounding mode. Other Alpha compilers call this option
17994 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
17999 Normal IEEE rounding mode. Floating-point numbers are rounded towards
18000 the nearest machine number or towards the even machine number in case
18004 Round towards minus infinity.
18007 Chopped rounding mode. Floating-point numbers are rounded towards zero.
18010 Dynamic rounding mode. A field in the floating-point control register
18011 (@var{fpcr}, see Alpha architecture reference manual) controls the
18012 rounding mode in effect. The C library initializes this register for
18013 rounding towards plus infinity. Thus, unless your program modifies the
18014 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
18017 @item -mtrap-precision=@var{trap-precision}
18018 @opindex mtrap-precision
18019 In the Alpha architecture, floating-point traps are imprecise. This
18020 means without software assistance it is impossible to recover from a
18021 floating trap and program execution normally needs to be terminated.
18022 GCC can generate code that can assist operating system trap handlers
18023 in determining the exact location that caused a floating-point trap.
18024 Depending on the requirements of an application, different levels of
18025 precisions can be selected:
18029 Program precision. This option is the default and means a trap handler
18030 can only identify which program caused a floating-point exception.
18033 Function precision. The trap handler can determine the function that
18034 caused a floating-point exception.
18037 Instruction precision. The trap handler can determine the exact
18038 instruction that caused a floating-point exception.
18041 Other Alpha compilers provide the equivalent options called
18042 @option{-scope_safe} and @option{-resumption_safe}.
18044 @item -mieee-conformant
18045 @opindex mieee-conformant
18046 This option marks the generated code as IEEE conformant. You must not
18047 use this option unless you also specify @option{-mtrap-precision=i} and either
18048 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
18049 is to emit the line @samp{.eflag 48} in the function prologue of the
18050 generated assembly file.
18052 @item -mbuild-constants
18053 @opindex mbuild-constants
18054 Normally GCC examines a 32- or 64-bit integer constant to
18055 see if it can construct it from smaller constants in two or three
18056 instructions. If it cannot, it outputs the constant as a literal and
18057 generates code to load it from the data segment at run time.
18059 Use this option to require GCC to construct @emph{all} integer constants
18060 using code, even if it takes more instructions (the maximum is six).
18062 You typically use this option to build a shared library dynamic
18063 loader. Itself a shared library, it must relocate itself in memory
18064 before it can find the variables and constants in its own data segment.
18082 Indicate whether GCC should generate code to use the optional BWX,
18083 CIX, FIX and MAX instruction sets. The default is to use the instruction
18084 sets supported by the CPU type specified via @option{-mcpu=} option or that
18085 of the CPU on which GCC was built if none is specified.
18088 @itemx -mfloat-ieee
18089 @opindex mfloat-vax
18090 @opindex mfloat-ieee
18091 Generate code that uses (does not use) VAX F and G floating-point
18092 arithmetic instead of IEEE single and double precision.
18094 @item -mexplicit-relocs
18095 @itemx -mno-explicit-relocs
18096 @opindex mexplicit-relocs
18097 @opindex mno-explicit-relocs
18098 Older Alpha assemblers provided no way to generate symbol relocations
18099 except via assembler macros. Use of these macros does not allow
18100 optimal instruction scheduling. GNU binutils as of version 2.12
18101 supports a new syntax that allows the compiler to explicitly mark
18102 which relocations should apply to which instructions. This option
18103 is mostly useful for debugging, as GCC detects the capabilities of
18104 the assembler when it is built and sets the default accordingly.
18107 @itemx -mlarge-data
18108 @opindex msmall-data
18109 @opindex mlarge-data
18110 When @option{-mexplicit-relocs} is in effect, static data is
18111 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
18112 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
18113 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
18114 16-bit relocations off of the @code{$gp} register. This limits the
18115 size of the small data area to 64KB, but allows the variables to be
18116 directly accessed via a single instruction.
18118 The default is @option{-mlarge-data}. With this option the data area
18119 is limited to just below 2GB@. Programs that require more than 2GB of
18120 data must use @code{malloc} or @code{mmap} to allocate the data in the
18121 heap instead of in the program's data segment.
18123 When generating code for shared libraries, @option{-fpic} implies
18124 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
18127 @itemx -mlarge-text
18128 @opindex msmall-text
18129 @opindex mlarge-text
18130 When @option{-msmall-text} is used, the compiler assumes that the
18131 code of the entire program (or shared library) fits in 4MB, and is
18132 thus reachable with a branch instruction. When @option{-msmall-data}
18133 is used, the compiler can assume that all local symbols share the
18134 same @code{$gp} value, and thus reduce the number of instructions
18135 required for a function call from 4 to 1.
18137 The default is @option{-mlarge-text}.
18139 @item -mcpu=@var{cpu_type}
18141 Set the instruction set and instruction scheduling parameters for
18142 machine type @var{cpu_type}. You can specify either the @samp{EV}
18143 style name or the corresponding chip number. GCC supports scheduling
18144 parameters for the EV4, EV5 and EV6 family of processors and
18145 chooses the default values for the instruction set from the processor
18146 you specify. If you do not specify a processor type, GCC defaults
18147 to the processor on which the compiler was built.
18149 Supported values for @var{cpu_type} are
18155 Schedules as an EV4 and has no instruction set extensions.
18159 Schedules as an EV5 and has no instruction set extensions.
18163 Schedules as an EV5 and supports the BWX extension.
18168 Schedules as an EV5 and supports the BWX and MAX extensions.
18172 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
18176 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
18179 Native toolchains also support the value @samp{native},
18180 which selects the best architecture option for the host processor.
18181 @option{-mcpu=native} has no effect if GCC does not recognize
18184 @item -mtune=@var{cpu_type}
18186 Set only the instruction scheduling parameters for machine type
18187 @var{cpu_type}. The instruction set is not changed.
18189 Native toolchains also support the value @samp{native},
18190 which selects the best architecture option for the host processor.
18191 @option{-mtune=native} has no effect if GCC does not recognize
18194 @item -mmemory-latency=@var{time}
18195 @opindex mmemory-latency
18196 Sets the latency the scheduler should assume for typical memory
18197 references as seen by the application. This number is highly
18198 dependent on the memory access patterns used by the application
18199 and the size of the external cache on the machine.
18201 Valid options for @var{time} are
18205 A decimal number representing clock cycles.
18211 The compiler contains estimates of the number of clock cycles for
18212 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
18213 (also called Dcache, Scache, and Bcache), as well as to main memory.
18214 Note that L3 is only valid for EV5.
18220 @subsection FR30 Options
18221 @cindex FR30 Options
18223 These options are defined specifically for the FR30 port.
18227 @item -msmall-model
18228 @opindex msmall-model
18229 Use the small address space model. This can produce smaller code, but
18230 it does assume that all symbolic values and addresses fit into a
18235 Assume that runtime support has been provided and so there is no need
18236 to include the simulator library (@file{libsim.a}) on the linker
18242 @subsection FT32 Options
18243 @cindex FT32 Options
18245 These options are defined specifically for the FT32 port.
18251 Specifies that the program will be run on the simulator. This causes
18252 an alternate runtime startup and library to be linked.
18253 You must not use this option when generating programs that will run on
18254 real hardware; you must provide your own runtime library for whatever
18255 I/O functions are needed.
18259 Enable Local Register Allocation. This is still experimental for FT32,
18260 so by default the compiler uses standard reload.
18264 Do not use div and mod instructions.
18268 Enable use of the extended instructions of the FT32B processor.
18272 Compress all code using the Ft32B code compression scheme.
18276 Do not generate code that reads program memory.
18281 @subsection FRV Options
18282 @cindex FRV Options
18288 Only use the first 32 general-purpose registers.
18293 Use all 64 general-purpose registers.
18298 Use only the first 32 floating-point registers.
18303 Use all 64 floating-point registers.
18306 @opindex mhard-float
18308 Use hardware instructions for floating-point operations.
18311 @opindex msoft-float
18313 Use library routines for floating-point operations.
18318 Dynamically allocate condition code registers.
18323 Do not try to dynamically allocate condition code registers, only
18324 use @code{icc0} and @code{fcc0}.
18329 Change ABI to use double word insns.
18334 Do not use double word instructions.
18339 Use floating-point double instructions.
18342 @opindex mno-double
18344 Do not use floating-point double instructions.
18349 Use media instructions.
18354 Do not use media instructions.
18359 Use multiply and add/subtract instructions.
18362 @opindex mno-muladd
18364 Do not use multiply and add/subtract instructions.
18369 Select the FDPIC ABI, which uses function descriptors to represent
18370 pointers to functions. Without any PIC/PIE-related options, it
18371 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
18372 assumes GOT entries and small data are within a 12-bit range from the
18373 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
18374 are computed with 32 bits.
18375 With a @samp{bfin-elf} target, this option implies @option{-msim}.
18378 @opindex minline-plt
18380 Enable inlining of PLT entries in function calls to functions that are
18381 not known to bind locally. It has no effect without @option{-mfdpic}.
18382 It's enabled by default if optimizing for speed and compiling for
18383 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
18384 optimization option such as @option{-O3} or above is present in the
18390 Assume a large TLS segment when generating thread-local code.
18395 Do not assume a large TLS segment when generating thread-local code.
18400 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
18401 that is known to be in read-only sections. It's enabled by default,
18402 except for @option{-fpic} or @option{-fpie}: even though it may help
18403 make the global offset table smaller, it trades 1 instruction for 4.
18404 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
18405 one of which may be shared by multiple symbols, and it avoids the need
18406 for a GOT entry for the referenced symbol, so it's more likely to be a
18407 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
18409 @item -multilib-library-pic
18410 @opindex multilib-library-pic
18412 Link with the (library, not FD) pic libraries. It's implied by
18413 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
18414 @option{-fpic} without @option{-mfdpic}. You should never have to use
18418 @opindex mlinked-fp
18420 Follow the EABI requirement of always creating a frame pointer whenever
18421 a stack frame is allocated. This option is enabled by default and can
18422 be disabled with @option{-mno-linked-fp}.
18425 @opindex mlong-calls
18427 Use indirect addressing to call functions outside the current
18428 compilation unit. This allows the functions to be placed anywhere
18429 within the 32-bit address space.
18431 @item -malign-labels
18432 @opindex malign-labels
18434 Try to align labels to an 8-byte boundary by inserting NOPs into the
18435 previous packet. This option only has an effect when VLIW packing
18436 is enabled. It doesn't create new packets; it merely adds NOPs to
18439 @item -mlibrary-pic
18440 @opindex mlibrary-pic
18442 Generate position-independent EABI code.
18447 Use only the first four media accumulator registers.
18452 Use all eight media accumulator registers.
18457 Pack VLIW instructions.
18462 Do not pack VLIW instructions.
18465 @opindex mno-eflags
18467 Do not mark ABI switches in e_flags.
18470 @opindex mcond-move
18472 Enable the use of conditional-move instructions (default).
18474 This switch is mainly for debugging the compiler and will likely be removed
18475 in a future version.
18477 @item -mno-cond-move
18478 @opindex mno-cond-move
18480 Disable the use of conditional-move instructions.
18482 This switch is mainly for debugging the compiler and will likely be removed
18483 in a future version.
18488 Enable the use of conditional set instructions (default).
18490 This switch is mainly for debugging the compiler and will likely be removed
18491 in a future version.
18496 Disable the use of conditional set instructions.
18498 This switch is mainly for debugging the compiler and will likely be removed
18499 in a future version.
18502 @opindex mcond-exec
18504 Enable the use of conditional execution (default).
18506 This switch is mainly for debugging the compiler and will likely be removed
18507 in a future version.
18509 @item -mno-cond-exec
18510 @opindex mno-cond-exec
18512 Disable the use of conditional execution.
18514 This switch is mainly for debugging the compiler and will likely be removed
18515 in a future version.
18517 @item -mvliw-branch
18518 @opindex mvliw-branch
18520 Run a pass to pack branches into VLIW instructions (default).
18522 This switch is mainly for debugging the compiler and will likely be removed
18523 in a future version.
18525 @item -mno-vliw-branch
18526 @opindex mno-vliw-branch
18528 Do not run a pass to pack branches into VLIW instructions.
18530 This switch is mainly for debugging the compiler and will likely be removed
18531 in a future version.
18533 @item -mmulti-cond-exec
18534 @opindex mmulti-cond-exec
18536 Enable optimization of @code{&&} and @code{||} in conditional execution
18539 This switch is mainly for debugging the compiler and will likely be removed
18540 in a future version.
18542 @item -mno-multi-cond-exec
18543 @opindex mno-multi-cond-exec
18545 Disable optimization of @code{&&} and @code{||} in conditional execution.
18547 This switch is mainly for debugging the compiler and will likely be removed
18548 in a future version.
18550 @item -mnested-cond-exec
18551 @opindex mnested-cond-exec
18553 Enable nested conditional execution optimizations (default).
18555 This switch is mainly for debugging the compiler and will likely be removed
18556 in a future version.
18558 @item -mno-nested-cond-exec
18559 @opindex mno-nested-cond-exec
18561 Disable nested conditional execution optimizations.
18563 This switch is mainly for debugging the compiler and will likely be removed
18564 in a future version.
18566 @item -moptimize-membar
18567 @opindex moptimize-membar
18569 This switch removes redundant @code{membar} instructions from the
18570 compiler-generated code. It is enabled by default.
18572 @item -mno-optimize-membar
18573 @opindex mno-optimize-membar
18575 This switch disables the automatic removal of redundant @code{membar}
18576 instructions from the generated code.
18578 @item -mtomcat-stats
18579 @opindex mtomcat-stats
18581 Cause gas to print out tomcat statistics.
18583 @item -mcpu=@var{cpu}
18586 Select the processor type for which to generate code. Possible values are
18587 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
18588 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
18592 @node GNU/Linux Options
18593 @subsection GNU/Linux Options
18595 These @samp{-m} options are defined for GNU/Linux targets:
18600 Use the GNU C library. This is the default except
18601 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
18602 @samp{*-*-linux-*android*} targets.
18606 Use uClibc C library. This is the default on
18607 @samp{*-*-linux-*uclibc*} targets.
18611 Use the musl C library. This is the default on
18612 @samp{*-*-linux-*musl*} targets.
18616 Use Bionic C library. This is the default on
18617 @samp{*-*-linux-*android*} targets.
18621 Compile code compatible with Android platform. This is the default on
18622 @samp{*-*-linux-*android*} targets.
18624 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
18625 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
18626 this option makes the GCC driver pass Android-specific options to the linker.
18627 Finally, this option causes the preprocessor macro @code{__ANDROID__}
18630 @item -tno-android-cc
18631 @opindex tno-android-cc
18632 Disable compilation effects of @option{-mandroid}, i.e., do not enable
18633 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
18634 @option{-fno-rtti} by default.
18636 @item -tno-android-ld
18637 @opindex tno-android-ld
18638 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
18639 linking options to the linker.
18643 @node H8/300 Options
18644 @subsection H8/300 Options
18646 These @samp{-m} options are defined for the H8/300 implementations:
18651 Shorten some address references at link time, when possible; uses the
18652 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
18653 ld, Using ld}, for a fuller description.
18657 Generate code for the H8/300H@.
18661 Generate code for the H8S@.
18665 Generate code for the H8S and H8/300H in the normal mode. This switch
18666 must be used either with @option{-mh} or @option{-ms}.
18670 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
18674 Extended registers are stored on stack before execution of function
18675 with monitor attribute. Default option is @option{-mexr}.
18676 This option is valid only for H8S targets.
18680 Extended registers are not stored on stack before execution of function
18681 with monitor attribute. Default option is @option{-mno-exr}.
18682 This option is valid only for H8S targets.
18686 Make @code{int} data 32 bits by default.
18689 @opindex malign-300
18690 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
18691 The default for the H8/300H and H8S is to align longs and floats on
18693 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
18694 This option has no effect on the H8/300.
18698 @subsection HPPA Options
18699 @cindex HPPA Options
18701 These @samp{-m} options are defined for the HPPA family of computers:
18704 @item -march=@var{architecture-type}
18706 Generate code for the specified architecture. The choices for
18707 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
18708 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
18709 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
18710 architecture option for your machine. Code compiled for lower numbered
18711 architectures runs on higher numbered architectures, but not the
18714 @item -mpa-risc-1-0
18715 @itemx -mpa-risc-1-1
18716 @itemx -mpa-risc-2-0
18717 @opindex mpa-risc-1-0
18718 @opindex mpa-risc-1-1
18719 @opindex mpa-risc-2-0
18720 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
18722 @item -mcaller-copies
18723 @opindex mcaller-copies
18724 The caller copies function arguments passed by hidden reference. This
18725 option should be used with care as it is not compatible with the default
18726 32-bit runtime. However, only aggregates larger than eight bytes are
18727 passed by hidden reference and the option provides better compatibility
18730 @item -mjump-in-delay
18731 @opindex mjump-in-delay
18732 This option is ignored and provided for compatibility purposes only.
18734 @item -mdisable-fpregs
18735 @opindex mdisable-fpregs
18736 Prevent floating-point registers from being used in any manner. This is
18737 necessary for compiling kernels that perform lazy context switching of
18738 floating-point registers. If you use this option and attempt to perform
18739 floating-point operations, the compiler aborts.
18741 @item -mdisable-indexing
18742 @opindex mdisable-indexing
18743 Prevent the compiler from using indexing address modes. This avoids some
18744 rather obscure problems when compiling MIG generated code under MACH@.
18746 @item -mno-space-regs
18747 @opindex mno-space-regs
18748 Generate code that assumes the target has no space registers. This allows
18749 GCC to generate faster indirect calls and use unscaled index address modes.
18751 Such code is suitable for level 0 PA systems and kernels.
18753 @item -mfast-indirect-calls
18754 @opindex mfast-indirect-calls
18755 Generate code that assumes calls never cross space boundaries. This
18756 allows GCC to emit code that performs faster indirect calls.
18758 This option does not work in the presence of shared libraries or nested
18761 @item -mfixed-range=@var{register-range}
18762 @opindex mfixed-range
18763 Generate code treating the given register range as fixed registers.
18764 A fixed register is one that the register allocator cannot use. This is
18765 useful when compiling kernel code. A register range is specified as
18766 two registers separated by a dash. Multiple register ranges can be
18767 specified separated by a comma.
18769 @item -mlong-load-store
18770 @opindex mlong-load-store
18771 Generate 3-instruction load and store sequences as sometimes required by
18772 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
18775 @item -mportable-runtime
18776 @opindex mportable-runtime
18777 Use the portable calling conventions proposed by HP for ELF systems.
18781 Enable the use of assembler directives only GAS understands.
18783 @item -mschedule=@var{cpu-type}
18785 Schedule code according to the constraints for the machine type
18786 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
18787 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
18788 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
18789 proper scheduling option for your machine. The default scheduling is
18793 @opindex mlinker-opt
18794 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
18795 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
18796 linkers in which they give bogus error messages when linking some programs.
18799 @opindex msoft-float
18800 Generate output containing library calls for floating point.
18801 @strong{Warning:} the requisite libraries are not available for all HPPA
18802 targets. Normally the facilities of the machine's usual C compiler are
18803 used, but this cannot be done directly in cross-compilation. You must make
18804 your own arrangements to provide suitable library functions for
18807 @option{-msoft-float} changes the calling convention in the output file;
18808 therefore, it is only useful if you compile @emph{all} of a program with
18809 this option. In particular, you need to compile @file{libgcc.a}, the
18810 library that comes with GCC, with @option{-msoft-float} in order for
18815 Generate the predefine, @code{_SIO}, for server IO@. The default is
18816 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
18817 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
18818 options are available under HP-UX and HI-UX@.
18822 Use options specific to GNU @command{ld}.
18823 This passes @option{-shared} to @command{ld} when
18824 building a shared library. It is the default when GCC is configured,
18825 explicitly or implicitly, with the GNU linker. This option does not
18826 affect which @command{ld} is called; it only changes what parameters
18827 are passed to that @command{ld}.
18828 The @command{ld} that is called is determined by the
18829 @option{--with-ld} configure option, GCC's program search path, and
18830 finally by the user's @env{PATH}. The linker used by GCC can be printed
18831 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
18832 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18836 Use options specific to HP @command{ld}.
18837 This passes @option{-b} to @command{ld} when building
18838 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
18839 links. It is the default when GCC is configured, explicitly or
18840 implicitly, with the HP linker. This option does not affect
18841 which @command{ld} is called; it only changes what parameters are passed to that
18843 The @command{ld} that is called is determined by the @option{--with-ld}
18844 configure option, GCC's program search path, and finally by the user's
18845 @env{PATH}. The linker used by GCC can be printed using @samp{which
18846 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
18847 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18850 @opindex mno-long-calls
18851 Generate code that uses long call sequences. This ensures that a call
18852 is always able to reach linker generated stubs. The default is to generate
18853 long calls only when the distance from the call site to the beginning
18854 of the function or translation unit, as the case may be, exceeds a
18855 predefined limit set by the branch type being used. The limits for
18856 normal calls are 7,600,000 and 240,000 bytes, respectively for the
18857 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
18860 Distances are measured from the beginning of functions when using the
18861 @option{-ffunction-sections} option, or when using the @option{-mgas}
18862 and @option{-mno-portable-runtime} options together under HP-UX with
18865 It is normally not desirable to use this option as it degrades
18866 performance. However, it may be useful in large applications,
18867 particularly when partial linking is used to build the application.
18869 The types of long calls used depends on the capabilities of the
18870 assembler and linker, and the type of code being generated. The
18871 impact on systems that support long absolute calls, and long pic
18872 symbol-difference or pc-relative calls should be relatively small.
18873 However, an indirect call is used on 32-bit ELF systems in pic code
18874 and it is quite long.
18876 @item -munix=@var{unix-std}
18878 Generate compiler predefines and select a startfile for the specified
18879 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
18880 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
18881 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
18882 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
18883 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
18886 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
18887 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
18888 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
18889 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
18890 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
18891 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
18893 It is @emph{important} to note that this option changes the interfaces
18894 for various library routines. It also affects the operational behavior
18895 of the C library. Thus, @emph{extreme} care is needed in using this
18898 Library code that is intended to operate with more than one UNIX
18899 standard must test, set and restore the variable @code{__xpg4_extended_mask}
18900 as appropriate. Most GNU software doesn't provide this capability.
18904 Suppress the generation of link options to search libdld.sl when the
18905 @option{-static} option is specified on HP-UX 10 and later.
18909 The HP-UX implementation of setlocale in libc has a dependency on
18910 libdld.sl. There isn't an archive version of libdld.sl. Thus,
18911 when the @option{-static} option is specified, special link options
18912 are needed to resolve this dependency.
18914 On HP-UX 10 and later, the GCC driver adds the necessary options to
18915 link with libdld.sl when the @option{-static} option is specified.
18916 This causes the resulting binary to be dynamic. On the 64-bit port,
18917 the linkers generate dynamic binaries by default in any case. The
18918 @option{-nolibdld} option can be used to prevent the GCC driver from
18919 adding these link options.
18923 Add support for multithreading with the @dfn{dce thread} library
18924 under HP-UX@. This option sets flags for both the preprocessor and
18928 @node IA-64 Options
18929 @subsection IA-64 Options
18930 @cindex IA-64 Options
18932 These are the @samp{-m} options defined for the Intel IA-64 architecture.
18936 @opindex mbig-endian
18937 Generate code for a big-endian target. This is the default for HP-UX@.
18939 @item -mlittle-endian
18940 @opindex mlittle-endian
18941 Generate code for a little-endian target. This is the default for AIX5
18947 @opindex mno-gnu-as
18948 Generate (or don't) code for the GNU assembler. This is the default.
18949 @c Also, this is the default if the configure option @option{--with-gnu-as}
18955 @opindex mno-gnu-ld
18956 Generate (or don't) code for the GNU linker. This is the default.
18957 @c Also, this is the default if the configure option @option{--with-gnu-ld}
18962 Generate code that does not use a global pointer register. The result
18963 is not position independent code, and violates the IA-64 ABI@.
18965 @item -mvolatile-asm-stop
18966 @itemx -mno-volatile-asm-stop
18967 @opindex mvolatile-asm-stop
18968 @opindex mno-volatile-asm-stop
18969 Generate (or don't) a stop bit immediately before and after volatile asm
18972 @item -mregister-names
18973 @itemx -mno-register-names
18974 @opindex mregister-names
18975 @opindex mno-register-names
18976 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
18977 the stacked registers. This may make assembler output more readable.
18983 Disable (or enable) optimizations that use the small data section. This may
18984 be useful for working around optimizer bugs.
18986 @item -mconstant-gp
18987 @opindex mconstant-gp
18988 Generate code that uses a single constant global pointer value. This is
18989 useful when compiling kernel code.
18993 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
18994 This is useful when compiling firmware code.
18996 @item -minline-float-divide-min-latency
18997 @opindex minline-float-divide-min-latency
18998 Generate code for inline divides of floating-point values
18999 using the minimum latency algorithm.
19001 @item -minline-float-divide-max-throughput
19002 @opindex minline-float-divide-max-throughput
19003 Generate code for inline divides of floating-point values
19004 using the maximum throughput algorithm.
19006 @item -mno-inline-float-divide
19007 @opindex mno-inline-float-divide
19008 Do not generate inline code for divides of floating-point values.
19010 @item -minline-int-divide-min-latency
19011 @opindex minline-int-divide-min-latency
19012 Generate code for inline divides of integer values
19013 using the minimum latency algorithm.
19015 @item -minline-int-divide-max-throughput
19016 @opindex minline-int-divide-max-throughput
19017 Generate code for inline divides of integer values
19018 using the maximum throughput algorithm.
19020 @item -mno-inline-int-divide
19021 @opindex mno-inline-int-divide
19022 Do not generate inline code for divides of integer values.
19024 @item -minline-sqrt-min-latency
19025 @opindex minline-sqrt-min-latency
19026 Generate code for inline square roots
19027 using the minimum latency algorithm.
19029 @item -minline-sqrt-max-throughput
19030 @opindex minline-sqrt-max-throughput
19031 Generate code for inline square roots
19032 using the maximum throughput algorithm.
19034 @item -mno-inline-sqrt
19035 @opindex mno-inline-sqrt
19036 Do not generate inline code for @code{sqrt}.
19039 @itemx -mno-fused-madd
19040 @opindex mfused-madd
19041 @opindex mno-fused-madd
19042 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
19043 instructions. The default is to use these instructions.
19045 @item -mno-dwarf2-asm
19046 @itemx -mdwarf2-asm
19047 @opindex mno-dwarf2-asm
19048 @opindex mdwarf2-asm
19049 Don't (or do) generate assembler code for the DWARF line number debugging
19050 info. This may be useful when not using the GNU assembler.
19052 @item -mearly-stop-bits
19053 @itemx -mno-early-stop-bits
19054 @opindex mearly-stop-bits
19055 @opindex mno-early-stop-bits
19056 Allow stop bits to be placed earlier than immediately preceding the
19057 instruction that triggered the stop bit. This can improve instruction
19058 scheduling, but does not always do so.
19060 @item -mfixed-range=@var{register-range}
19061 @opindex mfixed-range
19062 Generate code treating the given register range as fixed registers.
19063 A fixed register is one that the register allocator cannot use. This is
19064 useful when compiling kernel code. A register range is specified as
19065 two registers separated by a dash. Multiple register ranges can be
19066 specified separated by a comma.
19068 @item -mtls-size=@var{tls-size}
19070 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
19073 @item -mtune=@var{cpu-type}
19075 Tune the instruction scheduling for a particular CPU, Valid values are
19076 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
19077 and @samp{mckinley}.
19083 Generate code for a 32-bit or 64-bit environment.
19084 The 32-bit environment sets int, long and pointer to 32 bits.
19085 The 64-bit environment sets int to 32 bits and long and pointer
19086 to 64 bits. These are HP-UX specific flags.
19088 @item -mno-sched-br-data-spec
19089 @itemx -msched-br-data-spec
19090 @opindex mno-sched-br-data-spec
19091 @opindex msched-br-data-spec
19092 (Dis/En)able data speculative scheduling before reload.
19093 This results in generation of @code{ld.a} instructions and
19094 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
19095 The default setting is disabled.
19097 @item -msched-ar-data-spec
19098 @itemx -mno-sched-ar-data-spec
19099 @opindex msched-ar-data-spec
19100 @opindex mno-sched-ar-data-spec
19101 (En/Dis)able data speculative scheduling after reload.
19102 This results in generation of @code{ld.a} instructions and
19103 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
19104 The default setting is enabled.
19106 @item -mno-sched-control-spec
19107 @itemx -msched-control-spec
19108 @opindex mno-sched-control-spec
19109 @opindex msched-control-spec
19110 (Dis/En)able control speculative scheduling. This feature is
19111 available only during region scheduling (i.e.@: before reload).
19112 This results in generation of the @code{ld.s} instructions and
19113 the corresponding check instructions @code{chk.s}.
19114 The default setting is disabled.
19116 @item -msched-br-in-data-spec
19117 @itemx -mno-sched-br-in-data-spec
19118 @opindex msched-br-in-data-spec
19119 @opindex mno-sched-br-in-data-spec
19120 (En/Dis)able speculative scheduling of the instructions that
19121 are dependent on the data speculative loads before reload.
19122 This is effective only with @option{-msched-br-data-spec} enabled.
19123 The default setting is enabled.
19125 @item -msched-ar-in-data-spec
19126 @itemx -mno-sched-ar-in-data-spec
19127 @opindex msched-ar-in-data-spec
19128 @opindex mno-sched-ar-in-data-spec
19129 (En/Dis)able speculative scheduling of the instructions that
19130 are dependent on the data speculative loads after reload.
19131 This is effective only with @option{-msched-ar-data-spec} enabled.
19132 The default setting is enabled.
19134 @item -msched-in-control-spec
19135 @itemx -mno-sched-in-control-spec
19136 @opindex msched-in-control-spec
19137 @opindex mno-sched-in-control-spec
19138 (En/Dis)able speculative scheduling of the instructions that
19139 are dependent on the control speculative loads.
19140 This is effective only with @option{-msched-control-spec} enabled.
19141 The default setting is enabled.
19143 @item -mno-sched-prefer-non-data-spec-insns
19144 @itemx -msched-prefer-non-data-spec-insns
19145 @opindex mno-sched-prefer-non-data-spec-insns
19146 @opindex msched-prefer-non-data-spec-insns
19147 If enabled, data-speculative instructions are chosen for schedule
19148 only if there are no other choices at the moment. This makes
19149 the use of the data speculation much more conservative.
19150 The default setting is disabled.
19152 @item -mno-sched-prefer-non-control-spec-insns
19153 @itemx -msched-prefer-non-control-spec-insns
19154 @opindex mno-sched-prefer-non-control-spec-insns
19155 @opindex msched-prefer-non-control-spec-insns
19156 If enabled, control-speculative instructions are chosen for schedule
19157 only if there are no other choices at the moment. This makes
19158 the use of the control speculation much more conservative.
19159 The default setting is disabled.
19161 @item -mno-sched-count-spec-in-critical-path
19162 @itemx -msched-count-spec-in-critical-path
19163 @opindex mno-sched-count-spec-in-critical-path
19164 @opindex msched-count-spec-in-critical-path
19165 If enabled, speculative dependencies are considered during
19166 computation of the instructions priorities. This makes the use of the
19167 speculation a bit more conservative.
19168 The default setting is disabled.
19170 @item -msched-spec-ldc
19171 @opindex msched-spec-ldc
19172 Use a simple data speculation check. This option is on by default.
19174 @item -msched-control-spec-ldc
19175 @opindex msched-spec-ldc
19176 Use a simple check for control speculation. This option is on by default.
19178 @item -msched-stop-bits-after-every-cycle
19179 @opindex msched-stop-bits-after-every-cycle
19180 Place a stop bit after every cycle when scheduling. This option is on
19183 @item -msched-fp-mem-deps-zero-cost
19184 @opindex msched-fp-mem-deps-zero-cost
19185 Assume that floating-point stores and loads are not likely to cause a conflict
19186 when placed into the same instruction group. This option is disabled by
19189 @item -msel-sched-dont-check-control-spec
19190 @opindex msel-sched-dont-check-control-spec
19191 Generate checks for control speculation in selective scheduling.
19192 This flag is disabled by default.
19194 @item -msched-max-memory-insns=@var{max-insns}
19195 @opindex msched-max-memory-insns
19196 Limit on the number of memory insns per instruction group, giving lower
19197 priority to subsequent memory insns attempting to schedule in the same
19198 instruction group. Frequently useful to prevent cache bank conflicts.
19199 The default value is 1.
19201 @item -msched-max-memory-insns-hard-limit
19202 @opindex msched-max-memory-insns-hard-limit
19203 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
19204 disallowing more than that number in an instruction group.
19205 Otherwise, the limit is ``soft'', meaning that non-memory operations
19206 are preferred when the limit is reached, but memory operations may still
19212 @subsection LM32 Options
19213 @cindex LM32 options
19215 These @option{-m} options are defined for the LatticeMico32 architecture:
19218 @item -mbarrel-shift-enabled
19219 @opindex mbarrel-shift-enabled
19220 Enable barrel-shift instructions.
19222 @item -mdivide-enabled
19223 @opindex mdivide-enabled
19224 Enable divide and modulus instructions.
19226 @item -mmultiply-enabled
19227 @opindex multiply-enabled
19228 Enable multiply instructions.
19230 @item -msign-extend-enabled
19231 @opindex msign-extend-enabled
19232 Enable sign extend instructions.
19234 @item -muser-enabled
19235 @opindex muser-enabled
19236 Enable user-defined instructions.
19241 @subsection M32C Options
19242 @cindex M32C options
19245 @item -mcpu=@var{name}
19247 Select the CPU for which code is generated. @var{name} may be one of
19248 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
19249 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
19250 the M32C/80 series.
19254 Specifies that the program will be run on the simulator. This causes
19255 an alternate runtime library to be linked in which supports, for
19256 example, file I/O@. You must not use this option when generating
19257 programs that will run on real hardware; you must provide your own
19258 runtime library for whatever I/O functions are needed.
19260 @item -memregs=@var{number}
19262 Specifies the number of memory-based pseudo-registers GCC uses
19263 during code generation. These pseudo-registers are used like real
19264 registers, so there is a tradeoff between GCC's ability to fit the
19265 code into available registers, and the performance penalty of using
19266 memory instead of registers. Note that all modules in a program must
19267 be compiled with the same value for this option. Because of that, you
19268 must not use this option with GCC's default runtime libraries.
19272 @node M32R/D Options
19273 @subsection M32R/D Options
19274 @cindex M32R/D options
19276 These @option{-m} options are defined for Renesas M32R/D architectures:
19281 Generate code for the M32R/2@.
19285 Generate code for the M32R/X@.
19289 Generate code for the M32R@. This is the default.
19291 @item -mmodel=small
19292 @opindex mmodel=small
19293 Assume all objects live in the lower 16MB of memory (so that their addresses
19294 can be loaded with the @code{ld24} instruction), and assume all subroutines
19295 are reachable with the @code{bl} instruction.
19296 This is the default.
19298 The addressability of a particular object can be set with the
19299 @code{model} attribute.
19301 @item -mmodel=medium
19302 @opindex mmodel=medium
19303 Assume objects may be anywhere in the 32-bit address space (the compiler
19304 generates @code{seth/add3} instructions to load their addresses), and
19305 assume all subroutines are reachable with the @code{bl} instruction.
19307 @item -mmodel=large
19308 @opindex mmodel=large
19309 Assume objects may be anywhere in the 32-bit address space (the compiler
19310 generates @code{seth/add3} instructions to load their addresses), and
19311 assume subroutines may not be reachable with the @code{bl} instruction
19312 (the compiler generates the much slower @code{seth/add3/jl}
19313 instruction sequence).
19316 @opindex msdata=none
19317 Disable use of the small data area. Variables are put into
19318 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
19319 @code{section} attribute has been specified).
19320 This is the default.
19322 The small data area consists of sections @code{.sdata} and @code{.sbss}.
19323 Objects may be explicitly put in the small data area with the
19324 @code{section} attribute using one of these sections.
19326 @item -msdata=sdata
19327 @opindex msdata=sdata
19328 Put small global and static data in the small data area, but do not
19329 generate special code to reference them.
19332 @opindex msdata=use
19333 Put small global and static data in the small data area, and generate
19334 special instructions to reference them.
19338 @cindex smaller data references
19339 Put global and static objects less than or equal to @var{num} bytes
19340 into the small data or BSS sections instead of the normal data or BSS
19341 sections. The default value of @var{num} is 8.
19342 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
19343 for this option to have any effect.
19345 All modules should be compiled with the same @option{-G @var{num}} value.
19346 Compiling with different values of @var{num} may or may not work; if it
19347 doesn't the linker gives an error message---incorrect code is not
19352 Makes the M32R-specific code in the compiler display some statistics
19353 that might help in debugging programs.
19355 @item -malign-loops
19356 @opindex malign-loops
19357 Align all loops to a 32-byte boundary.
19359 @item -mno-align-loops
19360 @opindex mno-align-loops
19361 Do not enforce a 32-byte alignment for loops. This is the default.
19363 @item -missue-rate=@var{number}
19364 @opindex missue-rate=@var{number}
19365 Issue @var{number} instructions per cycle. @var{number} can only be 1
19368 @item -mbranch-cost=@var{number}
19369 @opindex mbranch-cost=@var{number}
19370 @var{number} can only be 1 or 2. If it is 1 then branches are
19371 preferred over conditional code, if it is 2, then the opposite applies.
19373 @item -mflush-trap=@var{number}
19374 @opindex mflush-trap=@var{number}
19375 Specifies the trap number to use to flush the cache. The default is
19376 12. Valid numbers are between 0 and 15 inclusive.
19378 @item -mno-flush-trap
19379 @opindex mno-flush-trap
19380 Specifies that the cache cannot be flushed by using a trap.
19382 @item -mflush-func=@var{name}
19383 @opindex mflush-func=@var{name}
19384 Specifies the name of the operating system function to call to flush
19385 the cache. The default is @samp{_flush_cache}, but a function call
19386 is only used if a trap is not available.
19388 @item -mno-flush-func
19389 @opindex mno-flush-func
19390 Indicates that there is no OS function for flushing the cache.
19394 @node M680x0 Options
19395 @subsection M680x0 Options
19396 @cindex M680x0 options
19398 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
19399 The default settings depend on which architecture was selected when
19400 the compiler was configured; the defaults for the most common choices
19404 @item -march=@var{arch}
19406 Generate code for a specific M680x0 or ColdFire instruction set
19407 architecture. Permissible values of @var{arch} for M680x0
19408 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
19409 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
19410 architectures are selected according to Freescale's ISA classification
19411 and the permissible values are: @samp{isaa}, @samp{isaaplus},
19412 @samp{isab} and @samp{isac}.
19414 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
19415 code for a ColdFire target. The @var{arch} in this macro is one of the
19416 @option{-march} arguments given above.
19418 When used together, @option{-march} and @option{-mtune} select code
19419 that runs on a family of similar processors but that is optimized
19420 for a particular microarchitecture.
19422 @item -mcpu=@var{cpu}
19424 Generate code for a specific M680x0 or ColdFire processor.
19425 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
19426 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
19427 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
19428 below, which also classifies the CPUs into families:
19430 @multitable @columnfractions 0.20 0.80
19431 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
19432 @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}
19433 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
19434 @item @samp{5206e} @tab @samp{5206e}
19435 @item @samp{5208} @tab @samp{5207} @samp{5208}
19436 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
19437 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
19438 @item @samp{5216} @tab @samp{5214} @samp{5216}
19439 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
19440 @item @samp{5225} @tab @samp{5224} @samp{5225}
19441 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
19442 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
19443 @item @samp{5249} @tab @samp{5249}
19444 @item @samp{5250} @tab @samp{5250}
19445 @item @samp{5271} @tab @samp{5270} @samp{5271}
19446 @item @samp{5272} @tab @samp{5272}
19447 @item @samp{5275} @tab @samp{5274} @samp{5275}
19448 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
19449 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
19450 @item @samp{5307} @tab @samp{5307}
19451 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
19452 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
19453 @item @samp{5407} @tab @samp{5407}
19454 @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}
19457 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
19458 @var{arch} is compatible with @var{cpu}. Other combinations of
19459 @option{-mcpu} and @option{-march} are rejected.
19461 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
19462 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
19463 where the value of @var{family} is given by the table above.
19465 @item -mtune=@var{tune}
19467 Tune the code for a particular microarchitecture within the
19468 constraints set by @option{-march} and @option{-mcpu}.
19469 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
19470 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
19471 and @samp{cpu32}. The ColdFire microarchitectures
19472 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
19474 You can also use @option{-mtune=68020-40} for code that needs
19475 to run relatively well on 68020, 68030 and 68040 targets.
19476 @option{-mtune=68020-60} is similar but includes 68060 targets
19477 as well. These two options select the same tuning decisions as
19478 @option{-m68020-40} and @option{-m68020-60} respectively.
19480 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
19481 when tuning for 680x0 architecture @var{arch}. It also defines
19482 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
19483 option is used. If GCC is tuning for a range of architectures,
19484 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
19485 it defines the macros for every architecture in the range.
19487 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
19488 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
19489 of the arguments given above.
19495 Generate output for a 68000. This is the default
19496 when the compiler is configured for 68000-based systems.
19497 It is equivalent to @option{-march=68000}.
19499 Use this option for microcontrollers with a 68000 or EC000 core,
19500 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
19504 Generate output for a 68010. This is the default
19505 when the compiler is configured for 68010-based systems.
19506 It is equivalent to @option{-march=68010}.
19512 Generate output for a 68020. This is the default
19513 when the compiler is configured for 68020-based systems.
19514 It is equivalent to @option{-march=68020}.
19518 Generate output for a 68030. This is the default when the compiler is
19519 configured for 68030-based systems. It is equivalent to
19520 @option{-march=68030}.
19524 Generate output for a 68040. This is the default when the compiler is
19525 configured for 68040-based systems. It is equivalent to
19526 @option{-march=68040}.
19528 This option inhibits the use of 68881/68882 instructions that have to be
19529 emulated by software on the 68040. Use this option if your 68040 does not
19530 have code to emulate those instructions.
19534 Generate output for a 68060. This is the default when the compiler is
19535 configured for 68060-based systems. It is equivalent to
19536 @option{-march=68060}.
19538 This option inhibits the use of 68020 and 68881/68882 instructions that
19539 have to be emulated by software on the 68060. Use this option if your 68060
19540 does not have code to emulate those instructions.
19544 Generate output for a CPU32. This is the default
19545 when the compiler is configured for CPU32-based systems.
19546 It is equivalent to @option{-march=cpu32}.
19548 Use this option for microcontrollers with a
19549 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
19550 68336, 68340, 68341, 68349 and 68360.
19554 Generate output for a 520X ColdFire CPU@. This is the default
19555 when the compiler is configured for 520X-based systems.
19556 It is equivalent to @option{-mcpu=5206}, and is now deprecated
19557 in favor of that option.
19559 Use this option for microcontroller with a 5200 core, including
19560 the MCF5202, MCF5203, MCF5204 and MCF5206.
19564 Generate output for a 5206e ColdFire CPU@. The option is now
19565 deprecated in favor of the equivalent @option{-mcpu=5206e}.
19569 Generate output for a member of the ColdFire 528X family.
19570 The option is now deprecated in favor of the equivalent
19571 @option{-mcpu=528x}.
19575 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
19576 in favor of the equivalent @option{-mcpu=5307}.
19580 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
19581 in favor of the equivalent @option{-mcpu=5407}.
19585 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
19586 This includes use of hardware floating-point instructions.
19587 The option is equivalent to @option{-mcpu=547x}, and is now
19588 deprecated in favor of that option.
19592 Generate output for a 68040, without using any of the new instructions.
19593 This results in code that can run relatively efficiently on either a
19594 68020/68881 or a 68030 or a 68040. The generated code does use the
19595 68881 instructions that are emulated on the 68040.
19597 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
19601 Generate output for a 68060, without using any of the new instructions.
19602 This results in code that can run relatively efficiently on either a
19603 68020/68881 or a 68030 or a 68040. The generated code does use the
19604 68881 instructions that are emulated on the 68060.
19606 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
19610 @opindex mhard-float
19612 Generate floating-point instructions. This is the default for 68020
19613 and above, and for ColdFire devices that have an FPU@. It defines the
19614 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
19615 on ColdFire targets.
19618 @opindex msoft-float
19619 Do not generate floating-point instructions; use library calls instead.
19620 This is the default for 68000, 68010, and 68832 targets. It is also
19621 the default for ColdFire devices that have no FPU.
19627 Generate (do not generate) ColdFire hardware divide and remainder
19628 instructions. If @option{-march} is used without @option{-mcpu},
19629 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
19630 architectures. Otherwise, the default is taken from the target CPU
19631 (either the default CPU, or the one specified by @option{-mcpu}). For
19632 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
19633 @option{-mcpu=5206e}.
19635 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
19639 Consider type @code{int} to be 16 bits wide, like @code{short int}.
19640 Additionally, parameters passed on the stack are also aligned to a
19641 16-bit boundary even on targets whose API mandates promotion to 32-bit.
19645 Do not consider type @code{int} to be 16 bits wide. This is the default.
19648 @itemx -mno-bitfield
19649 @opindex mnobitfield
19650 @opindex mno-bitfield
19651 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
19652 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
19656 Do use the bit-field instructions. The @option{-m68020} option implies
19657 @option{-mbitfield}. This is the default if you use a configuration
19658 designed for a 68020.
19662 Use a different function-calling convention, in which functions
19663 that take a fixed number of arguments return with the @code{rtd}
19664 instruction, which pops their arguments while returning. This
19665 saves one instruction in the caller since there is no need to pop
19666 the arguments there.
19668 This calling convention is incompatible with the one normally
19669 used on Unix, so you cannot use it if you need to call libraries
19670 compiled with the Unix compiler.
19672 Also, you must provide function prototypes for all functions that
19673 take variable numbers of arguments (including @code{printf});
19674 otherwise incorrect code is generated for calls to those
19677 In addition, seriously incorrect code results if you call a
19678 function with too many arguments. (Normally, extra arguments are
19679 harmlessly ignored.)
19681 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
19682 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
19686 Do not use the calling conventions selected by @option{-mrtd}.
19687 This is the default.
19690 @itemx -mno-align-int
19691 @opindex malign-int
19692 @opindex mno-align-int
19693 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
19694 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
19695 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
19696 Aligning variables on 32-bit boundaries produces code that runs somewhat
19697 faster on processors with 32-bit busses at the expense of more memory.
19699 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
19700 aligns structures containing the above types differently than
19701 most published application binary interface specifications for the m68k.
19705 Use the pc-relative addressing mode of the 68000 directly, instead of
19706 using a global offset table. At present, this option implies @option{-fpic},
19707 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
19708 not presently supported with @option{-mpcrel}, though this could be supported for
19709 68020 and higher processors.
19711 @item -mno-strict-align
19712 @itemx -mstrict-align
19713 @opindex mno-strict-align
19714 @opindex mstrict-align
19715 Do not (do) assume that unaligned memory references are handled by
19719 Generate code that allows the data segment to be located in a different
19720 area of memory from the text segment. This allows for execute-in-place in
19721 an environment without virtual memory management. This option implies
19724 @item -mno-sep-data
19725 Generate code that assumes that the data segment follows the text segment.
19726 This is the default.
19728 @item -mid-shared-library
19729 Generate code that supports shared libraries via the library ID method.
19730 This allows for execute-in-place and shared libraries in an environment
19731 without virtual memory management. This option implies @option{-fPIC}.
19733 @item -mno-id-shared-library
19734 Generate code that doesn't assume ID-based shared libraries are being used.
19735 This is the default.
19737 @item -mshared-library-id=n
19738 Specifies the identification number of the ID-based shared library being
19739 compiled. Specifying a value of 0 generates more compact code; specifying
19740 other values forces the allocation of that number to the current
19741 library, but is no more space- or time-efficient than omitting this option.
19747 When generating position-independent code for ColdFire, generate code
19748 that works if the GOT has more than 8192 entries. This code is
19749 larger and slower than code generated without this option. On M680x0
19750 processors, this option is not needed; @option{-fPIC} suffices.
19752 GCC normally uses a single instruction to load values from the GOT@.
19753 While this is relatively efficient, it only works if the GOT
19754 is smaller than about 64k. Anything larger causes the linker
19755 to report an error such as:
19757 @cindex relocation truncated to fit (ColdFire)
19759 relocation truncated to fit: R_68K_GOT16O foobar
19762 If this happens, you should recompile your code with @option{-mxgot}.
19763 It should then work with very large GOTs. However, code generated with
19764 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
19765 the value of a global symbol.
19767 Note that some linkers, including newer versions of the GNU linker,
19768 can create multiple GOTs and sort GOT entries. If you have such a linker,
19769 you should only need to use @option{-mxgot} when compiling a single
19770 object file that accesses more than 8192 GOT entries. Very few do.
19772 These options have no effect unless GCC is generating
19773 position-independent code.
19775 @item -mlong-jump-table-offsets
19776 @opindex mlong-jump-table-offsets
19777 Use 32-bit offsets in @code{switch} tables. The default is to use
19782 @node MCore Options
19783 @subsection MCore Options
19784 @cindex MCore options
19786 These are the @samp{-m} options defined for the Motorola M*Core
19792 @itemx -mno-hardlit
19794 @opindex mno-hardlit
19795 Inline constants into the code stream if it can be done in two
19796 instructions or less.
19802 Use the divide instruction. (Enabled by default).
19804 @item -mrelax-immediate
19805 @itemx -mno-relax-immediate
19806 @opindex mrelax-immediate
19807 @opindex mno-relax-immediate
19808 Allow arbitrary-sized immediates in bit operations.
19810 @item -mwide-bitfields
19811 @itemx -mno-wide-bitfields
19812 @opindex mwide-bitfields
19813 @opindex mno-wide-bitfields
19814 Always treat bit-fields as @code{int}-sized.
19816 @item -m4byte-functions
19817 @itemx -mno-4byte-functions
19818 @opindex m4byte-functions
19819 @opindex mno-4byte-functions
19820 Force all functions to be aligned to a 4-byte boundary.
19822 @item -mcallgraph-data
19823 @itemx -mno-callgraph-data
19824 @opindex mcallgraph-data
19825 @opindex mno-callgraph-data
19826 Emit callgraph information.
19829 @itemx -mno-slow-bytes
19830 @opindex mslow-bytes
19831 @opindex mno-slow-bytes
19832 Prefer word access when reading byte quantities.
19834 @item -mlittle-endian
19835 @itemx -mbig-endian
19836 @opindex mlittle-endian
19837 @opindex mbig-endian
19838 Generate code for a little-endian target.
19844 Generate code for the 210 processor.
19848 Assume that runtime support has been provided and so omit the
19849 simulator library (@file{libsim.a)} from the linker command line.
19851 @item -mstack-increment=@var{size}
19852 @opindex mstack-increment
19853 Set the maximum amount for a single stack increment operation. Large
19854 values can increase the speed of programs that contain functions
19855 that need a large amount of stack space, but they can also trigger a
19856 segmentation fault if the stack is extended too much. The default
19862 @subsection MeP Options
19863 @cindex MeP options
19869 Enables the @code{abs} instruction, which is the absolute difference
19870 between two registers.
19874 Enables all the optional instructions---average, multiply, divide, bit
19875 operations, leading zero, absolute difference, min/max, clip, and
19881 Enables the @code{ave} instruction, which computes the average of two
19884 @item -mbased=@var{n}
19886 Variables of size @var{n} bytes or smaller are placed in the
19887 @code{.based} section by default. Based variables use the @code{$tp}
19888 register as a base register, and there is a 128-byte limit to the
19889 @code{.based} section.
19893 Enables the bit operation instructions---bit test (@code{btstm}), set
19894 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
19895 test-and-set (@code{tas}).
19897 @item -mc=@var{name}
19899 Selects which section constant data is placed in. @var{name} may
19900 be @samp{tiny}, @samp{near}, or @samp{far}.
19904 Enables the @code{clip} instruction. Note that @option{-mclip} is not
19905 useful unless you also provide @option{-mminmax}.
19907 @item -mconfig=@var{name}
19909 Selects one of the built-in core configurations. Each MeP chip has
19910 one or more modules in it; each module has a core CPU and a variety of
19911 coprocessors, optional instructions, and peripherals. The
19912 @code{MeP-Integrator} tool, not part of GCC, provides these
19913 configurations through this option; using this option is the same as
19914 using all the corresponding command-line options. The default
19915 configuration is @samp{default}.
19919 Enables the coprocessor instructions. By default, this is a 32-bit
19920 coprocessor. Note that the coprocessor is normally enabled via the
19921 @option{-mconfig=} option.
19925 Enables the 32-bit coprocessor's instructions.
19929 Enables the 64-bit coprocessor's instructions.
19933 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
19937 Causes constant variables to be placed in the @code{.near} section.
19941 Enables the @code{div} and @code{divu} instructions.
19945 Generate big-endian code.
19949 Generate little-endian code.
19951 @item -mio-volatile
19952 @opindex mio-volatile
19953 Tells the compiler that any variable marked with the @code{io}
19954 attribute is to be considered volatile.
19958 Causes variables to be assigned to the @code{.far} section by default.
19962 Enables the @code{leadz} (leading zero) instruction.
19966 Causes variables to be assigned to the @code{.near} section by default.
19970 Enables the @code{min} and @code{max} instructions.
19974 Enables the multiplication and multiply-accumulate instructions.
19978 Disables all the optional instructions enabled by @option{-mall-opts}.
19982 Enables the @code{repeat} and @code{erepeat} instructions, used for
19983 low-overhead looping.
19987 Causes all variables to default to the @code{.tiny} section. Note
19988 that there is a 65536-byte limit to this section. Accesses to these
19989 variables use the @code{%gp} base register.
19993 Enables the saturation instructions. Note that the compiler does not
19994 currently generate these itself, but this option is included for
19995 compatibility with other tools, like @code{as}.
19999 Link the SDRAM-based runtime instead of the default ROM-based runtime.
20003 Link the simulator run-time libraries.
20007 Link the simulator runtime libraries, excluding built-in support
20008 for reset and exception vectors and tables.
20012 Causes all functions to default to the @code{.far} section. Without
20013 this option, functions default to the @code{.near} section.
20015 @item -mtiny=@var{n}
20017 Variables that are @var{n} bytes or smaller are allocated to the
20018 @code{.tiny} section. These variables use the @code{$gp} base
20019 register. The default for this option is 4, but note that there's a
20020 65536-byte limit to the @code{.tiny} section.
20024 @node MicroBlaze Options
20025 @subsection MicroBlaze Options
20026 @cindex MicroBlaze Options
20031 @opindex msoft-float
20032 Use software emulation for floating point (default).
20035 @opindex mhard-float
20036 Use hardware floating-point instructions.
20040 Do not optimize block moves, use @code{memcpy}.
20042 @item -mno-clearbss
20043 @opindex mno-clearbss
20044 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
20046 @item -mcpu=@var{cpu-type}
20048 Use features of, and schedule code for, the given CPU.
20049 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
20050 where @var{X} is a major version, @var{YY} is the minor version, and
20051 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
20052 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
20054 @item -mxl-soft-mul
20055 @opindex mxl-soft-mul
20056 Use software multiply emulation (default).
20058 @item -mxl-soft-div
20059 @opindex mxl-soft-div
20060 Use software emulation for divides (default).
20062 @item -mxl-barrel-shift
20063 @opindex mxl-barrel-shift
20064 Use the hardware barrel shifter.
20066 @item -mxl-pattern-compare
20067 @opindex mxl-pattern-compare
20068 Use pattern compare instructions.
20070 @item -msmall-divides
20071 @opindex msmall-divides
20072 Use table lookup optimization for small signed integer divisions.
20074 @item -mxl-stack-check
20075 @opindex mxl-stack-check
20076 This option is deprecated. Use @option{-fstack-check} instead.
20079 @opindex mxl-gp-opt
20080 Use GP-relative @code{.sdata}/@code{.sbss} sections.
20082 @item -mxl-multiply-high
20083 @opindex mxl-multiply-high
20084 Use multiply high instructions for high part of 32x32 multiply.
20086 @item -mxl-float-convert
20087 @opindex mxl-float-convert
20088 Use hardware floating-point conversion instructions.
20090 @item -mxl-float-sqrt
20091 @opindex mxl-float-sqrt
20092 Use hardware floating-point square root instruction.
20095 @opindex mbig-endian
20096 Generate code for a big-endian target.
20098 @item -mlittle-endian
20099 @opindex mlittle-endian
20100 Generate code for a little-endian target.
20103 @opindex mxl-reorder
20104 Use reorder instructions (swap and byte reversed load/store).
20106 @item -mxl-mode-@var{app-model}
20107 Select application model @var{app-model}. Valid models are
20110 normal executable (default), uses startup code @file{crt0.o}.
20113 for use with Xilinx Microprocessor Debugger (XMD) based
20114 software intrusive debug agent called xmdstub. This uses startup file
20115 @file{crt1.o} and sets the start address of the program to 0x800.
20118 for applications that are loaded using a bootloader.
20119 This model uses startup file @file{crt2.o} which does not contain a processor
20120 reset vector handler. This is suitable for transferring control on a
20121 processor reset to the bootloader rather than the application.
20124 for applications that do not require any of the
20125 MicroBlaze vectors. This option may be useful for applications running
20126 within a monitoring application. This model uses @file{crt3.o} as a startup file.
20129 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
20130 @option{-mxl-mode-@var{app-model}}.
20135 @subsection MIPS Options
20136 @cindex MIPS options
20142 Generate big-endian code.
20146 Generate little-endian code. This is the default for @samp{mips*el-*-*}
20149 @item -march=@var{arch}
20151 Generate code that runs on @var{arch}, which can be the name of a
20152 generic MIPS ISA, or the name of a particular processor.
20154 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
20155 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
20156 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
20157 @samp{mips64r5} and @samp{mips64r6}.
20158 The processor names are:
20159 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
20160 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
20161 @samp{5kc}, @samp{5kf},
20163 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
20164 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
20165 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
20166 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
20167 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
20170 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
20172 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
20173 @samp{m5100}, @samp{m5101},
20174 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
20177 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
20178 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
20179 @samp{rm7000}, @samp{rm9000},
20180 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
20183 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
20184 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
20185 @samp{xlr} and @samp{xlp}.
20186 The special value @samp{from-abi} selects the
20187 most compatible architecture for the selected ABI (that is,
20188 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
20190 The native Linux/GNU toolchain also supports the value @samp{native},
20191 which selects the best architecture option for the host processor.
20192 @option{-march=native} has no effect if GCC does not recognize
20195 In processor names, a final @samp{000} can be abbreviated as @samp{k}
20196 (for example, @option{-march=r2k}). Prefixes are optional, and
20197 @samp{vr} may be written @samp{r}.
20199 Names of the form @samp{@var{n}f2_1} refer to processors with
20200 FPUs clocked at half the rate of the core, names of the form
20201 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
20202 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
20203 processors with FPUs clocked a ratio of 3:2 with respect to the core.
20204 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
20205 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
20206 accepted as synonyms for @samp{@var{n}f1_1}.
20208 GCC defines two macros based on the value of this option. The first
20209 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
20210 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
20211 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
20212 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
20213 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
20215 Note that the @code{_MIPS_ARCH} macro uses the processor names given
20216 above. In other words, it has the full prefix and does not
20217 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
20218 the macro names the resolved architecture (either @code{"mips1"} or
20219 @code{"mips3"}). It names the default architecture when no
20220 @option{-march} option is given.
20222 @item -mtune=@var{arch}
20224 Optimize for @var{arch}. Among other things, this option controls
20225 the way instructions are scheduled, and the perceived cost of arithmetic
20226 operations. The list of @var{arch} values is the same as for
20229 When this option is not used, GCC optimizes for the processor
20230 specified by @option{-march}. By using @option{-march} and
20231 @option{-mtune} together, it is possible to generate code that
20232 runs on a family of processors, but optimize the code for one
20233 particular member of that family.
20235 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
20236 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
20237 @option{-march} ones described above.
20241 Equivalent to @option{-march=mips1}.
20245 Equivalent to @option{-march=mips2}.
20249 Equivalent to @option{-march=mips3}.
20253 Equivalent to @option{-march=mips4}.
20257 Equivalent to @option{-march=mips32}.
20261 Equivalent to @option{-march=mips32r3}.
20265 Equivalent to @option{-march=mips32r5}.
20269 Equivalent to @option{-march=mips32r6}.
20273 Equivalent to @option{-march=mips64}.
20277 Equivalent to @option{-march=mips64r2}.
20281 Equivalent to @option{-march=mips64r3}.
20285 Equivalent to @option{-march=mips64r5}.
20289 Equivalent to @option{-march=mips64r6}.
20294 @opindex mno-mips16
20295 Generate (do not generate) MIPS16 code. If GCC is targeting a
20296 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
20298 MIPS16 code generation can also be controlled on a per-function basis
20299 by means of @code{mips16} and @code{nomips16} attributes.
20300 @xref{Function Attributes}, for more information.
20302 @item -mflip-mips16
20303 @opindex mflip-mips16
20304 Generate MIPS16 code on alternating functions. This option is provided
20305 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
20306 not intended for ordinary use in compiling user code.
20308 @item -minterlink-compressed
20309 @itemx -mno-interlink-compressed
20310 @opindex minterlink-compressed
20311 @opindex mno-interlink-compressed
20312 Require (do not require) that code using the standard (uncompressed) MIPS ISA
20313 be link-compatible with MIPS16 and microMIPS code, and vice versa.
20315 For example, code using the standard ISA encoding cannot jump directly
20316 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
20317 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
20318 knows that the target of the jump is not compressed.
20320 @item -minterlink-mips16
20321 @itemx -mno-interlink-mips16
20322 @opindex minterlink-mips16
20323 @opindex mno-interlink-mips16
20324 Aliases of @option{-minterlink-compressed} and
20325 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
20326 and are retained for backwards compatibility.
20338 Generate code for the given ABI@.
20340 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
20341 generates 64-bit code when you select a 64-bit architecture, but you
20342 can use @option{-mgp32} to get 32-bit code instead.
20344 For information about the O64 ABI, see
20345 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
20347 GCC supports a variant of the o32 ABI in which floating-point registers
20348 are 64 rather than 32 bits wide. You can select this combination with
20349 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
20350 and @code{mfhc1} instructions and is therefore only supported for
20351 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
20353 The register assignments for arguments and return values remain the
20354 same, but each scalar value is passed in a single 64-bit register
20355 rather than a pair of 32-bit registers. For example, scalar
20356 floating-point values are returned in @samp{$f0} only, not a
20357 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
20358 remains the same in that the even-numbered double-precision registers
20361 Two additional variants of the o32 ABI are supported to enable
20362 a transition from 32-bit to 64-bit registers. These are FPXX
20363 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
20364 The FPXX extension mandates that all code must execute correctly
20365 when run using 32-bit or 64-bit registers. The code can be interlinked
20366 with either FP32 or FP64, but not both.
20367 The FP64A extension is similar to the FP64 extension but forbids the
20368 use of odd-numbered single-precision registers. This can be used
20369 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
20370 processors and allows both FP32 and FP64A code to interlink and
20371 run in the same process without changing FPU modes.
20374 @itemx -mno-abicalls
20376 @opindex mno-abicalls
20377 Generate (do not generate) code that is suitable for SVR4-style
20378 dynamic objects. @option{-mabicalls} is the default for SVR4-based
20383 Generate (do not generate) code that is fully position-independent,
20384 and that can therefore be linked into shared libraries. This option
20385 only affects @option{-mabicalls}.
20387 All @option{-mabicalls} code has traditionally been position-independent,
20388 regardless of options like @option{-fPIC} and @option{-fpic}. However,
20389 as an extension, the GNU toolchain allows executables to use absolute
20390 accesses for locally-binding symbols. It can also use shorter GP
20391 initialization sequences and generate direct calls to locally-defined
20392 functions. This mode is selected by @option{-mno-shared}.
20394 @option{-mno-shared} depends on binutils 2.16 or higher and generates
20395 objects that can only be linked by the GNU linker. However, the option
20396 does not affect the ABI of the final executable; it only affects the ABI
20397 of relocatable objects. Using @option{-mno-shared} generally makes
20398 executables both smaller and quicker.
20400 @option{-mshared} is the default.
20406 Assume (do not assume) that the static and dynamic linkers
20407 support PLTs and copy relocations. This option only affects
20408 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
20409 has no effect without @option{-msym32}.
20411 You can make @option{-mplt} the default by configuring
20412 GCC with @option{--with-mips-plt}. The default is
20413 @option{-mno-plt} otherwise.
20419 Lift (do not lift) the usual restrictions on the size of the global
20422 GCC normally uses a single instruction to load values from the GOT@.
20423 While this is relatively efficient, it only works if the GOT
20424 is smaller than about 64k. Anything larger causes the linker
20425 to report an error such as:
20427 @cindex relocation truncated to fit (MIPS)
20429 relocation truncated to fit: R_MIPS_GOT16 foobar
20432 If this happens, you should recompile your code with @option{-mxgot}.
20433 This works with very large GOTs, although the code is also
20434 less efficient, since it takes three instructions to fetch the
20435 value of a global symbol.
20437 Note that some linkers can create multiple GOTs. If you have such a
20438 linker, you should only need to use @option{-mxgot} when a single object
20439 file accesses more than 64k's worth of GOT entries. Very few do.
20441 These options have no effect unless GCC is generating position
20446 Assume that general-purpose registers are 32 bits wide.
20450 Assume that general-purpose registers are 64 bits wide.
20454 Assume that floating-point registers are 32 bits wide.
20458 Assume that floating-point registers are 64 bits wide.
20462 Do not assume the width of floating-point registers.
20465 @opindex mhard-float
20466 Use floating-point coprocessor instructions.
20469 @opindex msoft-float
20470 Do not use floating-point coprocessor instructions. Implement
20471 floating-point calculations using library calls instead.
20475 Equivalent to @option{-msoft-float}, but additionally asserts that the
20476 program being compiled does not perform any floating-point operations.
20477 This option is presently supported only by some bare-metal MIPS
20478 configurations, where it may select a special set of libraries
20479 that lack all floating-point support (including, for example, the
20480 floating-point @code{printf} formats).
20481 If code compiled with @option{-mno-float} accidentally contains
20482 floating-point operations, it is likely to suffer a link-time
20483 or run-time failure.
20485 @item -msingle-float
20486 @opindex msingle-float
20487 Assume that the floating-point coprocessor only supports single-precision
20490 @item -mdouble-float
20491 @opindex mdouble-float
20492 Assume that the floating-point coprocessor supports double-precision
20493 operations. This is the default.
20496 @itemx -mno-odd-spreg
20497 @opindex modd-spreg
20498 @opindex mno-odd-spreg
20499 Enable the use of odd-numbered single-precision floating-point registers
20500 for the o32 ABI. This is the default for processors that are known to
20501 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
20505 @itemx -mabs=legacy
20507 @opindex mabs=legacy
20508 These options control the treatment of the special not-a-number (NaN)
20509 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
20510 @code{neg.@i{fmt}} machine instructions.
20512 By default or when @option{-mabs=legacy} is used the legacy
20513 treatment is selected. In this case these instructions are considered
20514 arithmetic and avoided where correct operation is required and the
20515 input operand might be a NaN. A longer sequence of instructions that
20516 manipulate the sign bit of floating-point datum manually is used
20517 instead unless the @option{-ffinite-math-only} option has also been
20520 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
20521 this case these instructions are considered non-arithmetic and therefore
20522 operating correctly in all cases, including in particular where the
20523 input operand is a NaN. These instructions are therefore always used
20524 for the respective operations.
20527 @itemx -mnan=legacy
20529 @opindex mnan=legacy
20530 These options control the encoding of the special not-a-number (NaN)
20531 IEEE 754 floating-point data.
20533 The @option{-mnan=legacy} option selects the legacy encoding. In this
20534 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
20535 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
20536 by the first bit of their trailing significand field being 1.
20538 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
20539 this case qNaNs are denoted by the first bit of their trailing
20540 significand field being 1, whereas sNaNs are denoted by the first bit of
20541 their trailing significand field being 0.
20543 The default is @option{-mnan=legacy} unless GCC has been configured with
20544 @option{--with-nan=2008}.
20550 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
20551 implement atomic memory built-in functions. When neither option is
20552 specified, GCC uses the instructions if the target architecture
20555 @option{-mllsc} is useful if the runtime environment can emulate the
20556 instructions and @option{-mno-llsc} can be useful when compiling for
20557 nonstandard ISAs. You can make either option the default by
20558 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
20559 respectively. @option{--with-llsc} is the default for some
20560 configurations; see the installation documentation for details.
20566 Use (do not use) revision 1 of the MIPS DSP ASE@.
20567 @xref{MIPS DSP Built-in Functions}. This option defines the
20568 preprocessor macro @code{__mips_dsp}. It also defines
20569 @code{__mips_dsp_rev} to 1.
20575 Use (do not use) revision 2 of the MIPS DSP ASE@.
20576 @xref{MIPS DSP Built-in Functions}. This option defines the
20577 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
20578 It also defines @code{__mips_dsp_rev} to 2.
20581 @itemx -mno-smartmips
20582 @opindex msmartmips
20583 @opindex mno-smartmips
20584 Use (do not use) the MIPS SmartMIPS ASE.
20586 @item -mpaired-single
20587 @itemx -mno-paired-single
20588 @opindex mpaired-single
20589 @opindex mno-paired-single
20590 Use (do not use) paired-single floating-point instructions.
20591 @xref{MIPS Paired-Single Support}. This option requires
20592 hardware floating-point support to be enabled.
20598 Use (do not use) MIPS Digital Media Extension instructions.
20599 This option can only be used when generating 64-bit code and requires
20600 hardware floating-point support to be enabled.
20605 @opindex mno-mips3d
20606 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
20607 The option @option{-mips3d} implies @option{-mpaired-single}.
20610 @itemx -mno-micromips
20611 @opindex mmicromips
20612 @opindex mno-mmicromips
20613 Generate (do not generate) microMIPS code.
20615 MicroMIPS code generation can also be controlled on a per-function basis
20616 by means of @code{micromips} and @code{nomicromips} attributes.
20617 @xref{Function Attributes}, for more information.
20623 Use (do not use) MT Multithreading instructions.
20629 Use (do not use) the MIPS MCU ASE instructions.
20635 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
20641 Use (do not use) the MIPS Virtualization (VZ) instructions.
20647 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
20651 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
20652 an explanation of the default and the way that the pointer size is
20657 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
20659 The default size of @code{int}s, @code{long}s and pointers depends on
20660 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
20661 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
20662 32-bit @code{long}s. Pointers are the same size as @code{long}s,
20663 or the same size as integer registers, whichever is smaller.
20669 Assume (do not assume) that all symbols have 32-bit values, regardless
20670 of the selected ABI@. This option is useful in combination with
20671 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
20672 to generate shorter and faster references to symbolic addresses.
20676 Put definitions of externally-visible data in a small data section
20677 if that data is no bigger than @var{num} bytes. GCC can then generate
20678 more efficient accesses to the data; see @option{-mgpopt} for details.
20680 The default @option{-G} option depends on the configuration.
20682 @item -mlocal-sdata
20683 @itemx -mno-local-sdata
20684 @opindex mlocal-sdata
20685 @opindex mno-local-sdata
20686 Extend (do not extend) the @option{-G} behavior to local data too,
20687 such as to static variables in C@. @option{-mlocal-sdata} is the
20688 default for all configurations.
20690 If the linker complains that an application is using too much small data,
20691 you might want to try rebuilding the less performance-critical parts with
20692 @option{-mno-local-sdata}. You might also want to build large
20693 libraries with @option{-mno-local-sdata}, so that the libraries leave
20694 more room for the main program.
20696 @item -mextern-sdata
20697 @itemx -mno-extern-sdata
20698 @opindex mextern-sdata
20699 @opindex mno-extern-sdata
20700 Assume (do not assume) that externally-defined data is in
20701 a small data section if the size of that data is within the @option{-G} limit.
20702 @option{-mextern-sdata} is the default for all configurations.
20704 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
20705 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
20706 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
20707 is placed in a small data section. If @var{Var} is defined by another
20708 module, you must either compile that module with a high-enough
20709 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
20710 definition. If @var{Var} is common, you must link the application
20711 with a high-enough @option{-G} setting.
20713 The easiest way of satisfying these restrictions is to compile
20714 and link every module with the same @option{-G} option. However,
20715 you may wish to build a library that supports several different
20716 small data limits. You can do this by compiling the library with
20717 the highest supported @option{-G} setting and additionally using
20718 @option{-mno-extern-sdata} to stop the library from making assumptions
20719 about externally-defined data.
20725 Use (do not use) GP-relative accesses for symbols that are known to be
20726 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
20727 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
20730 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
20731 might not hold the value of @code{_gp}. For example, if the code is
20732 part of a library that might be used in a boot monitor, programs that
20733 call boot monitor routines pass an unknown value in @code{$gp}.
20734 (In such situations, the boot monitor itself is usually compiled
20735 with @option{-G0}.)
20737 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
20738 @option{-mno-extern-sdata}.
20740 @item -membedded-data
20741 @itemx -mno-embedded-data
20742 @opindex membedded-data
20743 @opindex mno-embedded-data
20744 Allocate variables to the read-only data section first if possible, then
20745 next in the small data section if possible, otherwise in data. This gives
20746 slightly slower code than the default, but reduces the amount of RAM required
20747 when executing, and thus may be preferred for some embedded systems.
20749 @item -muninit-const-in-rodata
20750 @itemx -mno-uninit-const-in-rodata
20751 @opindex muninit-const-in-rodata
20752 @opindex mno-uninit-const-in-rodata
20753 Put uninitialized @code{const} variables in the read-only data section.
20754 This option is only meaningful in conjunction with @option{-membedded-data}.
20756 @item -mcode-readable=@var{setting}
20757 @opindex mcode-readable
20758 Specify whether GCC may generate code that reads from executable sections.
20759 There are three possible settings:
20762 @item -mcode-readable=yes
20763 Instructions may freely access executable sections. This is the
20766 @item -mcode-readable=pcrel
20767 MIPS16 PC-relative load instructions can access executable sections,
20768 but other instructions must not do so. This option is useful on 4KSc
20769 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
20770 It is also useful on processors that can be configured to have a dual
20771 instruction/data SRAM interface and that, like the M4K, automatically
20772 redirect PC-relative loads to the instruction RAM.
20774 @item -mcode-readable=no
20775 Instructions must not access executable sections. This option can be
20776 useful on targets that are configured to have a dual instruction/data
20777 SRAM interface but that (unlike the M4K) do not automatically redirect
20778 PC-relative loads to the instruction RAM.
20781 @item -msplit-addresses
20782 @itemx -mno-split-addresses
20783 @opindex msplit-addresses
20784 @opindex mno-split-addresses
20785 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
20786 relocation operators. This option has been superseded by
20787 @option{-mexplicit-relocs} but is retained for backwards compatibility.
20789 @item -mexplicit-relocs
20790 @itemx -mno-explicit-relocs
20791 @opindex mexplicit-relocs
20792 @opindex mno-explicit-relocs
20793 Use (do not use) assembler relocation operators when dealing with symbolic
20794 addresses. The alternative, selected by @option{-mno-explicit-relocs},
20795 is to use assembler macros instead.
20797 @option{-mexplicit-relocs} is the default if GCC was configured
20798 to use an assembler that supports relocation operators.
20800 @item -mcheck-zero-division
20801 @itemx -mno-check-zero-division
20802 @opindex mcheck-zero-division
20803 @opindex mno-check-zero-division
20804 Trap (do not trap) on integer division by zero.
20806 The default is @option{-mcheck-zero-division}.
20808 @item -mdivide-traps
20809 @itemx -mdivide-breaks
20810 @opindex mdivide-traps
20811 @opindex mdivide-breaks
20812 MIPS systems check for division by zero by generating either a
20813 conditional trap or a break instruction. Using traps results in
20814 smaller code, but is only supported on MIPS II and later. Also, some
20815 versions of the Linux kernel have a bug that prevents trap from
20816 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
20817 allow conditional traps on architectures that support them and
20818 @option{-mdivide-breaks} to force the use of breaks.
20820 The default is usually @option{-mdivide-traps}, but this can be
20821 overridden at configure time using @option{--with-divide=breaks}.
20822 Divide-by-zero checks can be completely disabled using
20823 @option{-mno-check-zero-division}.
20825 @item -mload-store-pairs
20826 @itemx -mno-load-store-pairs
20827 @opindex mload-store-pairs
20828 @opindex mno-load-store-pairs
20829 Enable (disable) an optimization that pairs consecutive load or store
20830 instructions to enable load/store bonding. This option is enabled by
20831 default but only takes effect when the selected architecture is known
20832 to support bonding.
20837 @opindex mno-memcpy
20838 Force (do not force) the use of @code{memcpy} for non-trivial block
20839 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
20840 most constant-sized copies.
20843 @itemx -mno-long-calls
20844 @opindex mlong-calls
20845 @opindex mno-long-calls
20846 Disable (do not disable) use of the @code{jal} instruction. Calling
20847 functions using @code{jal} is more efficient but requires the caller
20848 and callee to be in the same 256 megabyte segment.
20850 This option has no effect on abicalls code. The default is
20851 @option{-mno-long-calls}.
20857 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
20858 instructions, as provided by the R4650 ISA@.
20864 Enable (disable) use of the @code{madd} and @code{msub} integer
20865 instructions. The default is @option{-mimadd} on architectures
20866 that support @code{madd} and @code{msub} except for the 74k
20867 architecture where it was found to generate slower code.
20870 @itemx -mno-fused-madd
20871 @opindex mfused-madd
20872 @opindex mno-fused-madd
20873 Enable (disable) use of the floating-point multiply-accumulate
20874 instructions, when they are available. The default is
20875 @option{-mfused-madd}.
20877 On the R8000 CPU when multiply-accumulate instructions are used,
20878 the intermediate product is calculated to infinite precision
20879 and is not subject to the FCSR Flush to Zero bit. This may be
20880 undesirable in some circumstances. On other processors the result
20881 is numerically identical to the equivalent computation using
20882 separate multiply, add, subtract and negate instructions.
20886 Tell the MIPS assembler to not run its preprocessor over user
20887 assembler files (with a @samp{.s} suffix) when assembling them.
20890 @itemx -mno-fix-24k
20892 @opindex mno-fix-24k
20893 Work around the 24K E48 (lost data on stores during refill) errata.
20894 The workarounds are implemented by the assembler rather than by GCC@.
20897 @itemx -mno-fix-r4000
20898 @opindex mfix-r4000
20899 @opindex mno-fix-r4000
20900 Work around certain R4000 CPU errata:
20903 A double-word or a variable shift may give an incorrect result if executed
20904 immediately after starting an integer division.
20906 A double-word or a variable shift may give an incorrect result if executed
20907 while an integer multiplication is in progress.
20909 An integer division may give an incorrect result if started in a delay slot
20910 of a taken branch or a jump.
20914 @itemx -mno-fix-r4400
20915 @opindex mfix-r4400
20916 @opindex mno-fix-r4400
20917 Work around certain R4400 CPU errata:
20920 A double-word or a variable shift may give an incorrect result if executed
20921 immediately after starting an integer division.
20925 @itemx -mno-fix-r10000
20926 @opindex mfix-r10000
20927 @opindex mno-fix-r10000
20928 Work around certain R10000 errata:
20931 @code{ll}/@code{sc} sequences may not behave atomically on revisions
20932 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
20935 This option can only be used if the target architecture supports
20936 branch-likely instructions. @option{-mfix-r10000} is the default when
20937 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
20941 @itemx -mno-fix-rm7000
20942 @opindex mfix-rm7000
20943 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
20944 workarounds are implemented by the assembler rather than by GCC@.
20947 @itemx -mno-fix-vr4120
20948 @opindex mfix-vr4120
20949 Work around certain VR4120 errata:
20952 @code{dmultu} does not always produce the correct result.
20954 @code{div} and @code{ddiv} do not always produce the correct result if one
20955 of the operands is negative.
20957 The workarounds for the division errata rely on special functions in
20958 @file{libgcc.a}. At present, these functions are only provided by
20959 the @code{mips64vr*-elf} configurations.
20961 Other VR4120 errata require a NOP to be inserted between certain pairs of
20962 instructions. These errata are handled by the assembler, not by GCC itself.
20965 @opindex mfix-vr4130
20966 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
20967 workarounds are implemented by the assembler rather than by GCC,
20968 although GCC avoids using @code{mflo} and @code{mfhi} if the
20969 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
20970 instructions are available instead.
20973 @itemx -mno-fix-sb1
20975 Work around certain SB-1 CPU core errata.
20976 (This flag currently works around the SB-1 revision 2
20977 ``F1'' and ``F2'' floating-point errata.)
20979 @item -mr10k-cache-barrier=@var{setting}
20980 @opindex mr10k-cache-barrier
20981 Specify whether GCC should insert cache barriers to avoid the
20982 side effects of speculation on R10K processors.
20984 In common with many processors, the R10K tries to predict the outcome
20985 of a conditional branch and speculatively executes instructions from
20986 the ``taken'' branch. It later aborts these instructions if the
20987 predicted outcome is wrong. However, on the R10K, even aborted
20988 instructions can have side effects.
20990 This problem only affects kernel stores and, depending on the system,
20991 kernel loads. As an example, a speculatively-executed store may load
20992 the target memory into cache and mark the cache line as dirty, even if
20993 the store itself is later aborted. If a DMA operation writes to the
20994 same area of memory before the ``dirty'' line is flushed, the cached
20995 data overwrites the DMA-ed data. See the R10K processor manual
20996 for a full description, including other potential problems.
20998 One workaround is to insert cache barrier instructions before every memory
20999 access that might be speculatively executed and that might have side
21000 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
21001 controls GCC's implementation of this workaround. It assumes that
21002 aborted accesses to any byte in the following regions does not have
21007 the memory occupied by the current function's stack frame;
21010 the memory occupied by an incoming stack argument;
21013 the memory occupied by an object with a link-time-constant address.
21016 It is the kernel's responsibility to ensure that speculative
21017 accesses to these regions are indeed safe.
21019 If the input program contains a function declaration such as:
21025 then the implementation of @code{foo} must allow @code{j foo} and
21026 @code{jal foo} to be executed speculatively. GCC honors this
21027 restriction for functions it compiles itself. It expects non-GCC
21028 functions (such as hand-written assembly code) to do the same.
21030 The option has three forms:
21033 @item -mr10k-cache-barrier=load-store
21034 Insert a cache barrier before a load or store that might be
21035 speculatively executed and that might have side effects even
21038 @item -mr10k-cache-barrier=store
21039 Insert a cache barrier before a store that might be speculatively
21040 executed and that might have side effects even if aborted.
21042 @item -mr10k-cache-barrier=none
21043 Disable the insertion of cache barriers. This is the default setting.
21046 @item -mflush-func=@var{func}
21047 @itemx -mno-flush-func
21048 @opindex mflush-func
21049 Specifies the function to call to flush the I and D caches, or to not
21050 call any such function. If called, the function must take the same
21051 arguments as the common @code{_flush_func}, that is, the address of the
21052 memory range for which the cache is being flushed, the size of the
21053 memory range, and the number 3 (to flush both caches). The default
21054 depends on the target GCC was configured for, but commonly is either
21055 @code{_flush_func} or @code{__cpu_flush}.
21057 @item mbranch-cost=@var{num}
21058 @opindex mbranch-cost
21059 Set the cost of branches to roughly @var{num} ``simple'' instructions.
21060 This cost is only a heuristic and is not guaranteed to produce
21061 consistent results across releases. A zero cost redundantly selects
21062 the default, which is based on the @option{-mtune} setting.
21064 @item -mbranch-likely
21065 @itemx -mno-branch-likely
21066 @opindex mbranch-likely
21067 @opindex mno-branch-likely
21068 Enable or disable use of Branch Likely instructions, regardless of the
21069 default for the selected architecture. By default, Branch Likely
21070 instructions may be generated if they are supported by the selected
21071 architecture. An exception is for the MIPS32 and MIPS64 architectures
21072 and processors that implement those architectures; for those, Branch
21073 Likely instructions are not be generated by default because the MIPS32
21074 and MIPS64 architectures specifically deprecate their use.
21076 @item -mcompact-branches=never
21077 @itemx -mcompact-branches=optimal
21078 @itemx -mcompact-branches=always
21079 @opindex mcompact-branches=never
21080 @opindex mcompact-branches=optimal
21081 @opindex mcompact-branches=always
21082 These options control which form of branches will be generated. The
21083 default is @option{-mcompact-branches=optimal}.
21085 The @option{-mcompact-branches=never} option ensures that compact branch
21086 instructions will never be generated.
21088 The @option{-mcompact-branches=always} option ensures that a compact
21089 branch instruction will be generated if available. If a compact branch
21090 instruction is not available, a delay slot form of the branch will be
21093 This option is supported from MIPS Release 6 onwards.
21095 The @option{-mcompact-branches=optimal} option will cause a delay slot
21096 branch to be used if one is available in the current ISA and the delay
21097 slot is successfully filled. If the delay slot is not filled, a compact
21098 branch will be chosen if one is available.
21100 @item -mfp-exceptions
21101 @itemx -mno-fp-exceptions
21102 @opindex mfp-exceptions
21103 Specifies whether FP exceptions are enabled. This affects how
21104 FP instructions are scheduled for some processors.
21105 The default is that FP exceptions are
21108 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
21109 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
21112 @item -mvr4130-align
21113 @itemx -mno-vr4130-align
21114 @opindex mvr4130-align
21115 The VR4130 pipeline is two-way superscalar, but can only issue two
21116 instructions together if the first one is 8-byte aligned. When this
21117 option is enabled, GCC aligns pairs of instructions that it
21118 thinks should execute in parallel.
21120 This option only has an effect when optimizing for the VR4130.
21121 It normally makes code faster, but at the expense of making it bigger.
21122 It is enabled by default at optimization level @option{-O3}.
21127 Enable (disable) generation of @code{synci} instructions on
21128 architectures that support it. The @code{synci} instructions (if
21129 enabled) are generated when @code{__builtin___clear_cache} is
21132 This option defaults to @option{-mno-synci}, but the default can be
21133 overridden by configuring GCC with @option{--with-synci}.
21135 When compiling code for single processor systems, it is generally safe
21136 to use @code{synci}. However, on many multi-core (SMP) systems, it
21137 does not invalidate the instruction caches on all cores and may lead
21138 to undefined behavior.
21140 @item -mrelax-pic-calls
21141 @itemx -mno-relax-pic-calls
21142 @opindex mrelax-pic-calls
21143 Try to turn PIC calls that are normally dispatched via register
21144 @code{$25} into direct calls. This is only possible if the linker can
21145 resolve the destination at link time and if the destination is within
21146 range for a direct call.
21148 @option{-mrelax-pic-calls} is the default if GCC was configured to use
21149 an assembler and a linker that support the @code{.reloc} assembly
21150 directive and @option{-mexplicit-relocs} is in effect. With
21151 @option{-mno-explicit-relocs}, this optimization can be performed by the
21152 assembler and the linker alone without help from the compiler.
21154 @item -mmcount-ra-address
21155 @itemx -mno-mcount-ra-address
21156 @opindex mmcount-ra-address
21157 @opindex mno-mcount-ra-address
21158 Emit (do not emit) code that allows @code{_mcount} to modify the
21159 calling function's return address. When enabled, this option extends
21160 the usual @code{_mcount} interface with a new @var{ra-address}
21161 parameter, which has type @code{intptr_t *} and is passed in register
21162 @code{$12}. @code{_mcount} can then modify the return address by
21163 doing both of the following:
21166 Returning the new address in register @code{$31}.
21168 Storing the new address in @code{*@var{ra-address}},
21169 if @var{ra-address} is nonnull.
21172 The default is @option{-mno-mcount-ra-address}.
21174 @item -mframe-header-opt
21175 @itemx -mno-frame-header-opt
21176 @opindex mframe-header-opt
21177 Enable (disable) frame header optimization in the o32 ABI. When using the
21178 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
21179 function to write out register arguments. When enabled, this optimization
21180 will suppress the allocation of the frame header if it can be determined that
21183 This optimization is off by default at all optimization levels.
21186 @itemx -mno-lxc1-sxc1
21187 @opindex mlxc1-sxc1
21188 When applicable, enable (disable) the generation of @code{lwxc1},
21189 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
21194 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
21195 @code{madd.d} and related instructions. Enabled by default.
21200 @subsection MMIX Options
21201 @cindex MMIX Options
21203 These options are defined for the MMIX:
21207 @itemx -mno-libfuncs
21209 @opindex mno-libfuncs
21210 Specify that intrinsic library functions are being compiled, passing all
21211 values in registers, no matter the size.
21214 @itemx -mno-epsilon
21216 @opindex mno-epsilon
21217 Generate floating-point comparison instructions that compare with respect
21218 to the @code{rE} epsilon register.
21220 @item -mabi=mmixware
21222 @opindex mabi=mmixware
21224 Generate code that passes function parameters and return values that (in
21225 the called function) are seen as registers @code{$0} and up, as opposed to
21226 the GNU ABI which uses global registers @code{$231} and up.
21228 @item -mzero-extend
21229 @itemx -mno-zero-extend
21230 @opindex mzero-extend
21231 @opindex mno-zero-extend
21232 When reading data from memory in sizes shorter than 64 bits, use (do not
21233 use) zero-extending load instructions by default, rather than
21234 sign-extending ones.
21237 @itemx -mno-knuthdiv
21239 @opindex mno-knuthdiv
21240 Make the result of a division yielding a remainder have the same sign as
21241 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
21242 remainder follows the sign of the dividend. Both methods are
21243 arithmetically valid, the latter being almost exclusively used.
21245 @item -mtoplevel-symbols
21246 @itemx -mno-toplevel-symbols
21247 @opindex mtoplevel-symbols
21248 @opindex mno-toplevel-symbols
21249 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
21250 code can be used with the @code{PREFIX} assembly directive.
21254 Generate an executable in the ELF format, rather than the default
21255 @samp{mmo} format used by the @command{mmix} simulator.
21257 @item -mbranch-predict
21258 @itemx -mno-branch-predict
21259 @opindex mbranch-predict
21260 @opindex mno-branch-predict
21261 Use (do not use) the probable-branch instructions, when static branch
21262 prediction indicates a probable branch.
21264 @item -mbase-addresses
21265 @itemx -mno-base-addresses
21266 @opindex mbase-addresses
21267 @opindex mno-base-addresses
21268 Generate (do not generate) code that uses @emph{base addresses}. Using a
21269 base address automatically generates a request (handled by the assembler
21270 and the linker) for a constant to be set up in a global register. The
21271 register is used for one or more base address requests within the range 0
21272 to 255 from the value held in the register. The generally leads to short
21273 and fast code, but the number of different data items that can be
21274 addressed is limited. This means that a program that uses lots of static
21275 data may require @option{-mno-base-addresses}.
21277 @item -msingle-exit
21278 @itemx -mno-single-exit
21279 @opindex msingle-exit
21280 @opindex mno-single-exit
21281 Force (do not force) generated code to have a single exit point in each
21285 @node MN10300 Options
21286 @subsection MN10300 Options
21287 @cindex MN10300 options
21289 These @option{-m} options are defined for Matsushita MN10300 architectures:
21294 Generate code to avoid bugs in the multiply instructions for the MN10300
21295 processors. This is the default.
21297 @item -mno-mult-bug
21298 @opindex mno-mult-bug
21299 Do not generate code to avoid bugs in the multiply instructions for the
21300 MN10300 processors.
21304 Generate code using features specific to the AM33 processor.
21308 Do not generate code using features specific to the AM33 processor. This
21313 Generate code using features specific to the AM33/2.0 processor.
21317 Generate code using features specific to the AM34 processor.
21319 @item -mtune=@var{cpu-type}
21321 Use the timing characteristics of the indicated CPU type when
21322 scheduling instructions. This does not change the targeted processor
21323 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
21324 @samp{am33-2} or @samp{am34}.
21326 @item -mreturn-pointer-on-d0
21327 @opindex mreturn-pointer-on-d0
21328 When generating a function that returns a pointer, return the pointer
21329 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
21330 only in @code{a0}, and attempts to call such functions without a prototype
21331 result in errors. Note that this option is on by default; use
21332 @option{-mno-return-pointer-on-d0} to disable it.
21336 Do not link in the C run-time initialization object file.
21340 Indicate to the linker that it should perform a relaxation optimization pass
21341 to shorten branches, calls and absolute memory addresses. This option only
21342 has an effect when used on the command line for the final link step.
21344 This option makes symbolic debugging impossible.
21348 Allow the compiler to generate @emph{Long Instruction Word}
21349 instructions if the target is the @samp{AM33} or later. This is the
21350 default. This option defines the preprocessor macro @code{__LIW__}.
21354 Do not allow the compiler to generate @emph{Long Instruction Word}
21355 instructions. This option defines the preprocessor macro
21360 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
21361 instructions if the target is the @samp{AM33} or later. This is the
21362 default. This option defines the preprocessor macro @code{__SETLB__}.
21366 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
21367 instructions. This option defines the preprocessor macro
21368 @code{__NO_SETLB__}.
21372 @node Moxie Options
21373 @subsection Moxie Options
21374 @cindex Moxie Options
21380 Generate big-endian code. This is the default for @samp{moxie-*-*}
21385 Generate little-endian code.
21389 Generate mul.x and umul.x instructions. This is the default for
21390 @samp{moxiebox-*-*} configurations.
21394 Do not link in the C run-time initialization object file.
21398 @node MSP430 Options
21399 @subsection MSP430 Options
21400 @cindex MSP430 Options
21402 These options are defined for the MSP430:
21408 Force assembly output to always use hex constants. Normally such
21409 constants are signed decimals, but this option is available for
21410 testsuite and/or aesthetic purposes.
21414 Select the MCU to target. This is used to create a C preprocessor
21415 symbol based upon the MCU name, converted to upper case and pre- and
21416 post-fixed with @samp{__}. This in turn is used by the
21417 @file{msp430.h} header file to select an MCU-specific supplementary
21420 The option also sets the ISA to use. If the MCU name is one that is
21421 known to only support the 430 ISA then that is selected, otherwise the
21422 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
21423 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
21424 name selects the 430X ISA.
21426 In addition an MCU-specific linker script is added to the linker
21427 command line. The script's name is the name of the MCU with
21428 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
21429 command line defines the C preprocessor symbol @code{__XXX__} and
21430 cause the linker to search for a script called @file{xxx.ld}.
21432 This option is also passed on to the assembler.
21435 @itemx -mno-warn-mcu
21437 @opindex mno-warn-mcu
21438 This option enables or disables warnings about conflicts between the
21439 MCU name specified by the @option{-mmcu} option and the ISA set by the
21440 @option{-mcpu} option and/or the hardware multiply support set by the
21441 @option{-mhwmult} option. It also toggles warnings about unrecognized
21442 MCU names. This option is on by default.
21446 Specifies the ISA to use. Accepted values are @samp{msp430},
21447 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
21448 @option{-mmcu=} option should be used to select the ISA.
21452 Link to the simulator runtime libraries and linker script. Overrides
21453 any scripts that would be selected by the @option{-mmcu=} option.
21457 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
21461 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
21465 This option is passed to the assembler and linker, and allows the
21466 linker to perform certain optimizations that cannot be done until
21471 Describes the type of hardware multiply supported by the target.
21472 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
21473 for the original 16-bit-only multiply supported by early MCUs.
21474 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
21475 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
21476 A value of @samp{auto} can also be given. This tells GCC to deduce
21477 the hardware multiply support based upon the MCU name provided by the
21478 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
21479 the MCU name is not recognized then no hardware multiply support is
21480 assumed. @code{auto} is the default setting.
21482 Hardware multiplies are normally performed by calling a library
21483 routine. This saves space in the generated code. When compiling at
21484 @option{-O3} or higher however the hardware multiplier is invoked
21485 inline. This makes for bigger, but faster code.
21487 The hardware multiply routines disable interrupts whilst running and
21488 restore the previous interrupt state when they finish. This makes
21489 them safe to use inside interrupt handlers as well as in normal code.
21493 Enable the use of a minimum runtime environment - no static
21494 initializers or constructors. This is intended for memory-constrained
21495 devices. The compiler includes special symbols in some objects
21496 that tell the linker and runtime which code fragments are required.
21498 @item -mcode-region=
21499 @itemx -mdata-region=
21500 @opindex mcode-region
21501 @opindex mdata-region
21502 These options tell the compiler where to place functions and data that
21503 do not have one of the @code{lower}, @code{upper}, @code{either} or
21504 @code{section} attributes. Possible values are @code{lower},
21505 @code{upper}, @code{either} or @code{any}. The first three behave
21506 like the corresponding attribute. The fourth possible value -
21507 @code{any} - is the default. It leaves placement entirely up to the
21508 linker script and how it assigns the standard sections
21509 (@code{.text}, @code{.data}, etc) to the memory regions.
21511 @item -msilicon-errata=
21512 @opindex msilicon-errata
21513 This option passes on a request to assembler to enable the fixes for
21514 the named silicon errata.
21516 @item -msilicon-errata-warn=
21517 @opindex msilicon-errata-warn
21518 This option passes on a request to the assembler to enable warning
21519 messages when a silicon errata might need to be applied.
21523 @node NDS32 Options
21524 @subsection NDS32 Options
21525 @cindex NDS32 Options
21527 These options are defined for NDS32 implementations:
21532 @opindex mbig-endian
21533 Generate code in big-endian mode.
21535 @item -mlittle-endian
21536 @opindex mlittle-endian
21537 Generate code in little-endian mode.
21539 @item -mreduced-regs
21540 @opindex mreduced-regs
21541 Use reduced-set registers for register allocation.
21544 @opindex mfull-regs
21545 Use full-set registers for register allocation.
21549 Generate conditional move instructions.
21553 Do not generate conditional move instructions.
21557 Generate performance extension instructions.
21559 @item -mno-ext-perf
21560 @opindex mno-perf-ext
21561 Do not generate performance extension instructions.
21565 Generate performance extension 2 instructions.
21567 @item -mno-ext-perf2
21568 @opindex mno-perf-ext
21569 Do not generate performance extension 2 instructions.
21573 Generate string extension instructions.
21575 @item -mno-ext-string
21576 @opindex mno-perf-ext
21577 Do not generate string extension instructions.
21581 Generate v3 push25/pop25 instructions.
21584 @opindex mno-v3push
21585 Do not generate v3 push25/pop25 instructions.
21589 Generate 16-bit instructions.
21592 @opindex mno-16-bit
21593 Do not generate 16-bit instructions.
21595 @item -misr-vector-size=@var{num}
21596 @opindex misr-vector-size
21597 Specify the size of each interrupt vector, which must be 4 or 16.
21599 @item -mcache-block-size=@var{num}
21600 @opindex mcache-block-size
21601 Specify the size of each cache block,
21602 which must be a power of 2 between 4 and 512.
21604 @item -march=@var{arch}
21606 Specify the name of the target architecture.
21608 @item -mcmodel=@var{code-model}
21610 Set the code model to one of
21613 All the data and read-only data segments must be within 512KB addressing space.
21614 The text segment must be within 16MB addressing space.
21615 @item @samp{medium}
21616 The data segment must be within 512KB while the read-only data segment can be
21617 within 4GB addressing space. The text segment should be still within 16MB
21620 All the text and data segments can be within 4GB addressing space.
21624 @opindex mctor-dtor
21625 Enable constructor/destructor feature.
21629 Guide linker to relax instructions.
21633 @node Nios II Options
21634 @subsection Nios II Options
21635 @cindex Nios II options
21636 @cindex Altera Nios II options
21638 These are the options defined for the Altera Nios II processor.
21644 @cindex smaller data references
21645 Put global and static objects less than or equal to @var{num} bytes
21646 into the small data or BSS sections instead of the normal data or BSS
21647 sections. The default value of @var{num} is 8.
21649 @item -mgpopt=@var{option}
21654 Generate (do not generate) GP-relative accesses. The following
21655 @var{option} names are recognized:
21660 Do not generate GP-relative accesses.
21663 Generate GP-relative accesses for small data objects that are not
21664 external, weak, or uninitialized common symbols.
21665 Also use GP-relative addressing for objects that
21666 have been explicitly placed in a small data section via a @code{section}
21670 As for @samp{local}, but also generate GP-relative accesses for
21671 small data objects that are external, weak, or common. If you use this option,
21672 you must ensure that all parts of your program (including libraries) are
21673 compiled with the same @option{-G} setting.
21676 Generate GP-relative accesses for all data objects in the program. If you
21677 use this option, the entire data and BSS segments
21678 of your program must fit in 64K of memory and you must use an appropriate
21679 linker script to allocate them within the addressable range of the
21683 Generate GP-relative addresses for function pointers as well as data
21684 pointers. If you use this option, the entire text, data, and BSS segments
21685 of your program must fit in 64K of memory and you must use an appropriate
21686 linker script to allocate them within the addressable range of the
21691 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
21692 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
21694 The default is @option{-mgpopt} except when @option{-fpic} or
21695 @option{-fPIC} is specified to generate position-independent code.
21696 Note that the Nios II ABI does not permit GP-relative accesses from
21699 You may need to specify @option{-mno-gpopt} explicitly when building
21700 programs that include large amounts of small data, including large
21701 GOT data sections. In this case, the 16-bit offset for GP-relative
21702 addressing may not be large enough to allow access to the entire
21703 small data section.
21705 @item -mgprel-sec=@var{regexp}
21706 @opindex mgprel-sec
21707 This option specifies additional section names that can be accessed via
21708 GP-relative addressing. It is most useful in conjunction with
21709 @code{section} attributes on variable declarations
21710 (@pxref{Common Variable Attributes}) and a custom linker script.
21711 The @var{regexp} is a POSIX Extended Regular Expression.
21713 This option does not affect the behavior of the @option{-G} option, and
21714 and the specified sections are in addition to the standard @code{.sdata}
21715 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
21717 @item -mr0rel-sec=@var{regexp}
21718 @opindex mr0rel-sec
21719 This option specifies names of sections that can be accessed via a
21720 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
21721 of the 32-bit address space. It is most useful in conjunction with
21722 @code{section} attributes on variable declarations
21723 (@pxref{Common Variable Attributes}) and a custom linker script.
21724 The @var{regexp} is a POSIX Extended Regular Expression.
21726 In contrast to the use of GP-relative addressing for small data,
21727 zero-based addressing is never generated by default and there are no
21728 conventional section names used in standard linker scripts for sections
21729 in the low or high areas of memory.
21735 Generate little-endian (default) or big-endian (experimental) code,
21738 @item -march=@var{arch}
21740 This specifies the name of the target Nios II architecture. GCC uses this
21741 name to determine what kind of instructions it can emit when generating
21742 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
21744 The preprocessor macro @code{__nios2_arch__} is available to programs,
21745 with value 1 or 2, indicating the targeted ISA level.
21747 @item -mbypass-cache
21748 @itemx -mno-bypass-cache
21749 @opindex mno-bypass-cache
21750 @opindex mbypass-cache
21751 Force all load and store instructions to always bypass cache by
21752 using I/O variants of the instructions. The default is not to
21755 @item -mno-cache-volatile
21756 @itemx -mcache-volatile
21757 @opindex mcache-volatile
21758 @opindex mno-cache-volatile
21759 Volatile memory access bypass the cache using the I/O variants of
21760 the load and store instructions. The default is not to bypass the cache.
21762 @item -mno-fast-sw-div
21763 @itemx -mfast-sw-div
21764 @opindex mno-fast-sw-div
21765 @opindex mfast-sw-div
21766 Do not use table-based fast divide for small numbers. The default
21767 is to use the fast divide at @option{-O3} and above.
21771 @itemx -mno-hw-mulx
21775 @opindex mno-hw-mul
21777 @opindex mno-hw-mulx
21779 @opindex mno-hw-div
21781 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
21782 instructions by the compiler. The default is to emit @code{mul}
21783 and not emit @code{div} and @code{mulx}.
21789 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
21790 CDX (code density) instructions. Enabling these instructions also
21791 requires @option{-march=r2}. Since these instructions are optional
21792 extensions to the R2 architecture, the default is not to emit them.
21794 @item -mcustom-@var{insn}=@var{N}
21795 @itemx -mno-custom-@var{insn}
21796 @opindex mcustom-@var{insn}
21797 @opindex mno-custom-@var{insn}
21798 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
21799 custom instruction with encoding @var{N} when generating code that uses
21800 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
21801 instruction 253 for single-precision floating-point add operations instead
21802 of the default behavior of using a library call.
21804 The following values of @var{insn} are supported. Except as otherwise
21805 noted, floating-point operations are expected to be implemented with
21806 normal IEEE 754 semantics and correspond directly to the C operators or the
21807 equivalent GCC built-in functions (@pxref{Other Builtins}).
21809 Single-precision floating point:
21812 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
21813 Binary arithmetic operations.
21819 Unary absolute value.
21821 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
21822 Comparison operations.
21824 @item @samp{fmins}, @samp{fmaxs}
21825 Floating-point minimum and maximum. These instructions are only
21826 generated if @option{-ffinite-math-only} is specified.
21828 @item @samp{fsqrts}
21829 Unary square root operation.
21831 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
21832 Floating-point trigonometric and exponential functions. These instructions
21833 are only generated if @option{-funsafe-math-optimizations} is also specified.
21837 Double-precision floating point:
21840 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
21841 Binary arithmetic operations.
21847 Unary absolute value.
21849 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
21850 Comparison operations.
21852 @item @samp{fmind}, @samp{fmaxd}
21853 Double-precision minimum and maximum. These instructions are only
21854 generated if @option{-ffinite-math-only} is specified.
21856 @item @samp{fsqrtd}
21857 Unary square root operation.
21859 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
21860 Double-precision trigonometric and exponential functions. These instructions
21861 are only generated if @option{-funsafe-math-optimizations} is also specified.
21867 @item @samp{fextsd}
21868 Conversion from single precision to double precision.
21870 @item @samp{ftruncds}
21871 Conversion from double precision to single precision.
21873 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
21874 Conversion from floating point to signed or unsigned integer types, with
21875 truncation towards zero.
21878 Conversion from single-precision floating point to signed integer,
21879 rounding to the nearest integer and ties away from zero.
21880 This corresponds to the @code{__builtin_lroundf} function when
21881 @option{-fno-math-errno} is used.
21883 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
21884 Conversion from signed or unsigned integer types to floating-point types.
21888 In addition, all of the following transfer instructions for internal
21889 registers X and Y must be provided to use any of the double-precision
21890 floating-point instructions. Custom instructions taking two
21891 double-precision source operands expect the first operand in the
21892 64-bit register X. The other operand (or only operand of a unary
21893 operation) is given to the custom arithmetic instruction with the
21894 least significant half in source register @var{src1} and the most
21895 significant half in @var{src2}. A custom instruction that returns a
21896 double-precision result returns the most significant 32 bits in the
21897 destination register and the other half in 32-bit register Y.
21898 GCC automatically generates the necessary code sequences to write
21899 register X and/or read register Y when double-precision floating-point
21900 instructions are used.
21905 Write @var{src1} into the least significant half of X and @var{src2} into
21906 the most significant half of X.
21909 Write @var{src1} into Y.
21911 @item @samp{frdxhi}, @samp{frdxlo}
21912 Read the most or least (respectively) significant half of X and store it in
21916 Read the value of Y and store it into @var{dest}.
21919 Note that you can gain more local control over generation of Nios II custom
21920 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
21921 and @code{target("no-custom-@var{insn}")} function attributes
21922 (@pxref{Function Attributes})
21923 or pragmas (@pxref{Function Specific Option Pragmas}).
21925 @item -mcustom-fpu-cfg=@var{name}
21926 @opindex mcustom-fpu-cfg
21928 This option enables a predefined, named set of custom instruction encodings
21929 (see @option{-mcustom-@var{insn}} above).
21930 Currently, the following sets are defined:
21932 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
21933 @gccoptlist{-mcustom-fmuls=252 @gol
21934 -mcustom-fadds=253 @gol
21935 -mcustom-fsubs=254 @gol
21936 -fsingle-precision-constant}
21938 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
21939 @gccoptlist{-mcustom-fmuls=252 @gol
21940 -mcustom-fadds=253 @gol
21941 -mcustom-fsubs=254 @gol
21942 -mcustom-fdivs=255 @gol
21943 -fsingle-precision-constant}
21945 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
21946 @gccoptlist{-mcustom-floatus=243 @gol
21947 -mcustom-fixsi=244 @gol
21948 -mcustom-floatis=245 @gol
21949 -mcustom-fcmpgts=246 @gol
21950 -mcustom-fcmples=249 @gol
21951 -mcustom-fcmpeqs=250 @gol
21952 -mcustom-fcmpnes=251 @gol
21953 -mcustom-fmuls=252 @gol
21954 -mcustom-fadds=253 @gol
21955 -mcustom-fsubs=254 @gol
21956 -mcustom-fdivs=255 @gol
21957 -fsingle-precision-constant}
21959 Custom instruction assignments given by individual
21960 @option{-mcustom-@var{insn}=} options override those given by
21961 @option{-mcustom-fpu-cfg=}, regardless of the
21962 order of the options on the command line.
21964 Note that you can gain more local control over selection of a FPU
21965 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
21966 function attribute (@pxref{Function Attributes})
21967 or pragma (@pxref{Function Specific Option Pragmas}).
21971 These additional @samp{-m} options are available for the Altera Nios II
21972 ELF (bare-metal) target:
21978 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
21979 startup and termination code, and is typically used in conjunction with
21980 @option{-msys-crt0=} to specify the location of the alternate startup code
21981 provided by the HAL BSP.
21985 Link with a limited version of the C library, @option{-lsmallc}, rather than
21988 @item -msys-crt0=@var{startfile}
21990 @var{startfile} is the file name of the startfile (crt0) to use
21991 when linking. This option is only useful in conjunction with @option{-mhal}.
21993 @item -msys-lib=@var{systemlib}
21995 @var{systemlib} is the library name of the library that provides
21996 low-level system calls required by the C library,
21997 e.g. @code{read} and @code{write}.
21998 This option is typically used to link with a library provided by a HAL BSP.
22002 @node Nvidia PTX Options
22003 @subsection Nvidia PTX Options
22004 @cindex Nvidia PTX options
22005 @cindex nvptx options
22007 These options are defined for Nvidia PTX:
22015 Generate code for 32-bit or 64-bit ABI.
22018 @opindex mmainkernel
22019 Link in code for a __main kernel. This is for stand-alone instead of
22020 offloading execution.
22024 Apply partitioned execution optimizations. This is the default when any
22025 level of optimization is selected.
22028 @opindex msoft-stack
22029 Generate code that does not use @code{.local} memory
22030 directly for stack storage. Instead, a per-warp stack pointer is
22031 maintained explicitly. This enables variable-length stack allocation (with
22032 variable-length arrays or @code{alloca}), and when global memory is used for
22033 underlying storage, makes it possible to access automatic variables from other
22034 threads, or with atomic instructions. This code generation variant is used
22035 for OpenMP offloading, but the option is exposed on its own for the purpose
22036 of testing the compiler; to generate code suitable for linking into programs
22037 using OpenMP offloading, use option @option{-mgomp}.
22039 @item -muniform-simt
22040 @opindex muniform-simt
22041 Switch to code generation variant that allows to execute all threads in each
22042 warp, while maintaining memory state and side effects as if only one thread
22043 in each warp was active outside of OpenMP SIMD regions. All atomic operations
22044 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
22045 current lane index equals the master lane index), and the register being
22046 assigned is copied via a shuffle instruction from the master lane. Outside of
22047 SIMD regions lane 0 is the master; inside, each thread sees itself as the
22048 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
22049 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
22050 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
22051 with current lane index to compute the master lane index.
22055 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
22056 @option{-muniform-simt} options, and selects corresponding multilib variant.
22060 @node PDP-11 Options
22061 @subsection PDP-11 Options
22062 @cindex PDP-11 Options
22064 These options are defined for the PDP-11:
22069 Use hardware FPP floating point. This is the default. (FIS floating
22070 point on the PDP-11/40 is not supported.)
22073 @opindex msoft-float
22074 Do not use hardware floating point.
22078 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
22082 Return floating-point results in memory. This is the default.
22086 Generate code for a PDP-11/40.
22090 Generate code for a PDP-11/45. This is the default.
22094 Generate code for a PDP-11/10.
22096 @item -mbcopy-builtin
22097 @opindex mbcopy-builtin
22098 Use inline @code{movmemhi} patterns for copying memory. This is the
22103 Do not use inline @code{movmemhi} patterns for copying memory.
22109 Use 16-bit @code{int}. This is the default.
22115 Use 32-bit @code{int}.
22118 @itemx -mno-float32
22120 @opindex mno-float32
22121 Use 64-bit @code{float}. This is the default.
22124 @itemx -mno-float64
22126 @opindex mno-float64
22127 Use 32-bit @code{float}.
22131 Use @code{abshi2} pattern. This is the default.
22135 Do not use @code{abshi2} pattern.
22137 @item -mbranch-expensive
22138 @opindex mbranch-expensive
22139 Pretend that branches are expensive. This is for experimenting with
22140 code generation only.
22142 @item -mbranch-cheap
22143 @opindex mbranch-cheap
22144 Do not pretend that branches are expensive. This is the default.
22148 Use Unix assembler syntax. This is the default when configured for
22149 @samp{pdp11-*-bsd}.
22153 Use DEC assembler syntax. This is the default when configured for any
22154 PDP-11 target other than @samp{pdp11-*-bsd}.
22157 @node picoChip Options
22158 @subsection picoChip Options
22159 @cindex picoChip options
22161 These @samp{-m} options are defined for picoChip implementations:
22165 @item -mae=@var{ae_type}
22167 Set the instruction set, register set, and instruction scheduling
22168 parameters for array element type @var{ae_type}. Supported values
22169 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
22171 @option{-mae=ANY} selects a completely generic AE type. Code
22172 generated with this option runs on any of the other AE types. The
22173 code is not as efficient as it would be if compiled for a specific
22174 AE type, and some types of operation (e.g., multiplication) do not
22175 work properly on all types of AE.
22177 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
22178 for compiled code, and is the default.
22180 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
22181 option may suffer from poor performance of byte (char) manipulation,
22182 since the DSP AE does not provide hardware support for byte load/stores.
22184 @item -msymbol-as-address
22185 Enable the compiler to directly use a symbol name as an address in a
22186 load/store instruction, without first loading it into a
22187 register. Typically, the use of this option generates larger
22188 programs, which run faster than when the option isn't used. However, the
22189 results vary from program to program, so it is left as a user option,
22190 rather than being permanently enabled.
22192 @item -mno-inefficient-warnings
22193 Disables warnings about the generation of inefficient code. These
22194 warnings can be generated, for example, when compiling code that
22195 performs byte-level memory operations on the MAC AE type. The MAC AE has
22196 no hardware support for byte-level memory operations, so all byte
22197 load/stores must be synthesized from word load/store operations. This is
22198 inefficient and a warning is generated to indicate
22199 that you should rewrite the code to avoid byte operations, or to target
22200 an AE type that has the necessary hardware support. This option disables
22205 @node PowerPC Options
22206 @subsection PowerPC Options
22207 @cindex PowerPC options
22209 These are listed under @xref{RS/6000 and PowerPC Options}.
22211 @node PowerPC SPE Options
22212 @subsection PowerPC SPE Options
22213 @cindex PowerPC SPE options
22215 These @samp{-m} options are defined for PowerPC SPE:
22220 @itemx -mno-popcntb
22224 @opindex mno-popcntb
22225 You use these options to specify which instructions are available on the
22226 processor you are using. The default value of these options is
22227 determined when configuring GCC@. Specifying the
22228 @option{-mcpu=@var{cpu_type}} overrides the specification of these
22229 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
22230 rather than the options listed above.
22232 The @option{-mmfcrf} option allows GCC to generate the move from
22233 condition register field instruction implemented on the POWER4
22234 processor and other processors that support the PowerPC V2.01
22236 The @option{-mpopcntb} option allows GCC to generate the popcount and
22237 double-precision FP reciprocal estimate instruction implemented on the
22238 POWER5 processor and other processors that support the PowerPC V2.02
22241 @item -mcpu=@var{cpu_type}
22243 Set architecture type, register usage, and
22244 instruction scheduling parameters for machine type @var{cpu_type}.
22245 Supported values for @var{cpu_type} are @samp{8540}, @samp{8548},
22248 @option{-mcpu=powerpc} specifies pure 32-bit PowerPC (either
22249 endian), with an appropriate, generic processor model assumed for
22250 scheduling purposes.
22252 Specifying @samp{native} as cpu type detects and selects the
22253 architecture option that corresponds to the host processor of the
22254 system performing the compilation.
22255 @option{-mcpu=native} has no effect if GCC does not recognize the
22258 The other options specify a specific processor. Code generated under
22259 those options runs best on that processor, and may not run at all on
22262 The @option{-mcpu} options automatically enable or disable the
22265 @gccoptlist{-mhard-float -mmfcrf -mmultiple @gol
22266 -mpopcntb -mpopcntd @gol
22267 -msingle-float -mdouble-float @gol
22270 The particular options set for any particular CPU varies between
22271 compiler versions, depending on what setting seems to produce optimal
22272 code for that CPU; it doesn't necessarily reflect the actual hardware's
22273 capabilities. If you wish to set an individual option to a particular
22274 value, you may specify it after the @option{-mcpu} option, like
22275 @option{-mcpu=8548}.
22277 @item -mtune=@var{cpu_type}
22279 Set the instruction scheduling parameters for machine type
22280 @var{cpu_type}, but do not set the architecture type or register usage,
22281 as @option{-mcpu=@var{cpu_type}} does. The same
22282 values for @var{cpu_type} are used for @option{-mtune} as for
22283 @option{-mcpu}. If both are specified, the code generated uses the
22284 architecture and registers set by @option{-mcpu}, but the
22285 scheduling parameters set by @option{-mtune}.
22288 @opindex msecure-plt
22289 Generate code that allows @command{ld} and @command{ld.so}
22290 to build executables and shared
22291 libraries with non-executable @code{.plt} and @code{.got} sections.
22293 32-bit SYSV ABI option.
22297 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
22299 requires @code{.plt} and @code{.got}
22300 sections that are both writable and executable.
22301 This is a PowerPC 32-bit SYSV ABI option.
22307 This switch enables or disables the generation of ISEL instructions.
22309 @item -misel=@var{yes/no}
22310 This switch has been deprecated. Use @option{-misel} and
22311 @option{-mno-isel} instead.
22317 This switch enables or disables the generation of SPE simd
22320 @item -mspe=@var{yes/no}
22321 This option has been deprecated. Use @option{-mspe} and
22322 @option{-mno-spe} instead.
22325 @itemx -mno-float128
22327 @opindex mno-float128
22328 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
22329 and use either software emulation for IEEE 128-bit floating point or
22330 hardware instructions.
22332 @item -mfloat-gprs=@var{yes/single/double/no}
22333 @itemx -mfloat-gprs
22334 @opindex mfloat-gprs
22335 This switch enables or disables the generation of floating-point
22336 operations on the general-purpose registers for architectures that
22339 The argument @samp{yes} or @samp{single} enables the use of
22340 single-precision floating-point operations.
22342 The argument @samp{double} enables the use of single and
22343 double-precision floating-point operations.
22345 The argument @samp{no} disables floating-point operations on the
22346 general-purpose registers.
22348 This option is currently only available on the MPC854x.
22351 @itemx -mno-fp-in-toc
22352 @itemx -mno-sum-in-toc
22353 @itemx -mminimal-toc
22355 @opindex mno-fp-in-toc
22356 @opindex mno-sum-in-toc
22357 @opindex mminimal-toc
22358 Modify generation of the TOC (Table Of Contents), which is created for
22359 every executable file. The @option{-mfull-toc} option is selected by
22360 default. In that case, GCC allocates at least one TOC entry for
22361 each unique non-automatic variable reference in your program. GCC
22362 also places floating-point constants in the TOC@. However, only
22363 16,384 entries are available in the TOC@.
22365 If you receive a linker error message that saying you have overflowed
22366 the available TOC space, you can reduce the amount of TOC space used
22367 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
22368 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
22369 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
22370 generate code to calculate the sum of an address and a constant at
22371 run time instead of putting that sum into the TOC@. You may specify one
22372 or both of these options. Each causes GCC to produce very slightly
22373 slower and larger code at the expense of conserving TOC space.
22375 If you still run out of space in the TOC even when you specify both of
22376 these options, specify @option{-mminimal-toc} instead. This option causes
22377 GCC to make only one TOC entry for every file. When you specify this
22378 option, GCC produces code that is slower and larger but which
22379 uses extremely little TOC space. You may wish to use this option
22380 only on files that contain less frequently-executed code.
22384 Disables the 64-bit ABI. GCC defaults to @option{-maix32}.
22387 @itemx -mno-xl-compat
22388 @opindex mxl-compat
22389 @opindex mno-xl-compat
22390 Produce code that conforms more closely to IBM XL compiler semantics
22391 when using AIX-compatible ABI@. Pass floating-point arguments to
22392 prototyped functions beyond the register save area (RSA) on the stack
22393 in addition to argument FPRs. Do not assume that most significant
22394 double in 128-bit long double value is properly rounded when comparing
22395 values and converting to double. Use XL symbol names for long double
22398 The AIX calling convention was extended but not initially documented to
22399 handle an obscure K&R C case of calling a function that takes the
22400 address of its arguments with fewer arguments than declared. IBM XL
22401 compilers access floating-point arguments that do not fit in the
22402 RSA from the stack when a subroutine is compiled without
22403 optimization. Because always storing floating-point arguments on the
22404 stack is inefficient and rarely needed, this option is not enabled by
22405 default and only is necessary when calling subroutines compiled by IBM
22406 XL compilers without optimization.
22408 @item -malign-natural
22409 @itemx -malign-power
22410 @opindex malign-natural
22411 @opindex malign-power
22412 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
22413 @option{-malign-natural} overrides the ABI-defined alignment of larger
22414 types, such as floating-point doubles, on their natural size-based boundary.
22415 The option @option{-malign-power} instructs GCC to follow the ABI-specified
22416 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
22418 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
22422 @itemx -mhard-float
22423 @opindex msoft-float
22424 @opindex mhard-float
22425 Generate code that does not use (uses) the floating-point register set.
22426 Software floating-point emulation is provided if you use the
22427 @option{-msoft-float} option, and pass the option to GCC when linking.
22429 @item -msingle-float
22430 @itemx -mdouble-float
22431 @opindex msingle-float
22432 @opindex mdouble-float
22433 Generate code for single- or double-precision floating-point operations.
22434 @option{-mdouble-float} implies @option{-msingle-float}.
22437 @itemx -mno-multiple
22439 @opindex mno-multiple
22440 Generate code that uses (does not use) the load multiple word
22441 instructions and the store multiple word instructions. These
22442 instructions are generated by default on POWER systems, and not
22443 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
22444 PowerPC systems, since those instructions do not work when the
22445 processor is in little-endian mode. The exceptions are PPC740 and
22446 PPC750 which permit these instructions in little-endian mode.
22451 @opindex mno-update
22452 Generate code that uses (does not use) the load or store instructions
22453 that update the base register to the address of the calculated memory
22454 location. These instructions are generated by default. If you use
22455 @option{-mno-update}, there is a small window between the time that the
22456 stack pointer is updated and the address of the previous frame is
22457 stored, which means code that walks the stack frame across interrupts or
22458 signals may get corrupted data.
22460 @item -mavoid-indexed-addresses
22461 @itemx -mno-avoid-indexed-addresses
22462 @opindex mavoid-indexed-addresses
22463 @opindex mno-avoid-indexed-addresses
22464 Generate code that tries to avoid (not avoid) the use of indexed load
22465 or store instructions. These instructions can incur a performance
22466 penalty on Power6 processors in certain situations, such as when
22467 stepping through large arrays that cross a 16M boundary. This option
22468 is enabled by default when targeting Power6 and disabled otherwise.
22471 @itemx -mno-fused-madd
22472 @opindex mfused-madd
22473 @opindex mno-fused-madd
22474 Generate code that uses (does not use) the floating-point multiply and
22475 accumulate instructions. These instructions are generated by default
22476 if hardware floating point is used. The machine-dependent
22477 @option{-mfused-madd} option is now mapped to the machine-independent
22478 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22479 mapped to @option{-ffp-contract=off}.
22481 @item -mno-strict-align
22482 @itemx -mstrict-align
22483 @opindex mno-strict-align
22484 @opindex mstrict-align
22485 On System V.4 and embedded PowerPC systems do not (do) assume that
22486 unaligned memory references are handled by the system.
22488 @item -mrelocatable
22489 @itemx -mno-relocatable
22490 @opindex mrelocatable
22491 @opindex mno-relocatable
22492 Generate code that allows (does not allow) a static executable to be
22493 relocated to a different address at run time. A simple embedded
22494 PowerPC system loader should relocate the entire contents of
22495 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
22496 a table of 32-bit addresses generated by this option. For this to
22497 work, all objects linked together must be compiled with
22498 @option{-mrelocatable} or @option{-mrelocatable-lib}.
22499 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
22501 @item -mrelocatable-lib
22502 @itemx -mno-relocatable-lib
22503 @opindex mrelocatable-lib
22504 @opindex mno-relocatable-lib
22505 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
22506 @code{.fixup} section to allow static executables to be relocated at
22507 run time, but @option{-mrelocatable-lib} does not use the smaller stack
22508 alignment of @option{-mrelocatable}. Objects compiled with
22509 @option{-mrelocatable-lib} may be linked with objects compiled with
22510 any combination of the @option{-mrelocatable} options.
22516 On System V.4 and embedded PowerPC systems do not (do) assume that
22517 register 2 contains a pointer to a global area pointing to the addresses
22518 used in the program.
22521 @itemx -mlittle-endian
22523 @opindex mlittle-endian
22524 On System V.4 and embedded PowerPC systems compile code for the
22525 processor in little-endian mode. The @option{-mlittle-endian} option is
22526 the same as @option{-mlittle}.
22529 @itemx -mbig-endian
22531 @opindex mbig-endian
22532 On System V.4 and embedded PowerPC systems compile code for the
22533 processor in big-endian mode. The @option{-mbig-endian} option is
22534 the same as @option{-mbig}.
22536 @item -mdynamic-no-pic
22537 @opindex mdynamic-no-pic
22538 On Darwin and Mac OS X systems, compile code so that it is not
22539 relocatable, but that its external references are relocatable. The
22540 resulting code is suitable for applications, but not shared
22543 @item -msingle-pic-base
22544 @opindex msingle-pic-base
22545 Treat the register used for PIC addressing as read-only, rather than
22546 loading it in the prologue for each function. The runtime system is
22547 responsible for initializing this register with an appropriate value
22548 before execution begins.
22550 @item -mprioritize-restricted-insns=@var{priority}
22551 @opindex mprioritize-restricted-insns
22552 This option controls the priority that is assigned to
22553 dispatch-slot restricted instructions during the second scheduling
22554 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
22555 or @samp{2} to assign no, highest, or second-highest (respectively)
22556 priority to dispatch-slot restricted
22559 @item -msched-costly-dep=@var{dependence_type}
22560 @opindex msched-costly-dep
22561 This option controls which dependences are considered costly
22562 by the target during instruction scheduling. The argument
22563 @var{dependence_type} takes one of the following values:
22567 No dependence is costly.
22570 All dependences are costly.
22572 @item @samp{true_store_to_load}
22573 A true dependence from store to load is costly.
22575 @item @samp{store_to_load}
22576 Any dependence from store to load is costly.
22579 Any dependence for which the latency is greater than or equal to
22580 @var{number} is costly.
22583 @item -minsert-sched-nops=@var{scheme}
22584 @opindex minsert-sched-nops
22585 This option controls which NOP insertion scheme is used during
22586 the second scheduling pass. The argument @var{scheme} takes one of the
22594 Pad with NOPs any dispatch group that has vacant issue slots,
22595 according to the scheduler's grouping.
22597 @item @samp{regroup_exact}
22598 Insert NOPs to force costly dependent insns into
22599 separate groups. Insert exactly as many NOPs as needed to force an insn
22600 to a new group, according to the estimated processor grouping.
22603 Insert NOPs to force costly dependent insns into
22604 separate groups. Insert @var{number} NOPs to force an insn to a new group.
22608 @opindex mcall-sysv
22609 On System V.4 and embedded PowerPC systems compile code using calling
22610 conventions that adhere to the March 1995 draft of the System V
22611 Application Binary Interface, PowerPC processor supplement. This is the
22612 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
22614 @item -mcall-sysv-eabi
22616 @opindex mcall-sysv-eabi
22617 @opindex mcall-eabi
22618 Specify both @option{-mcall-sysv} and @option{-meabi} options.
22620 @item -mcall-sysv-noeabi
22621 @opindex mcall-sysv-noeabi
22622 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
22624 @item -mcall-aixdesc
22626 On System V.4 and embedded PowerPC systems compile code for the AIX
22630 @opindex mcall-linux
22631 On System V.4 and embedded PowerPC systems compile code for the
22632 Linux-based GNU system.
22634 @item -mcall-freebsd
22635 @opindex mcall-freebsd
22636 On System V.4 and embedded PowerPC systems compile code for the
22637 FreeBSD operating system.
22639 @item -mcall-netbsd
22640 @opindex mcall-netbsd
22641 On System V.4 and embedded PowerPC systems compile code for the
22642 NetBSD operating system.
22644 @item -mcall-openbsd
22645 @opindex mcall-netbsd
22646 On System V.4 and embedded PowerPC systems compile code for the
22647 OpenBSD operating system.
22649 @item -maix-struct-return
22650 @opindex maix-struct-return
22651 Return all structures in memory (as specified by the AIX ABI)@.
22653 @item -msvr4-struct-return
22654 @opindex msvr4-struct-return
22655 Return structures smaller than 8 bytes in registers (as specified by the
22658 @item -mabi=@var{abi-type}
22660 Extend the current ABI with a particular extension, or remove such extension.
22661 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
22662 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
22663 @samp{elfv1}, @samp{elfv2}@.
22667 Extend the current ABI with SPE ABI extensions. This does not change
22668 the default ABI, instead it adds the SPE ABI extensions to the current
22672 @opindex mabi=no-spe
22673 Disable Book-E SPE ABI extensions for the current ABI@.
22675 @item -mabi=ibmlongdouble
22676 @opindex mabi=ibmlongdouble
22677 Change the current ABI to use IBM extended-precision long double.
22678 This is not likely to work if your system defaults to using IEEE
22679 extended-precision long double. If you change the long double type
22680 from IEEE extended-precision, the compiler will issue a warning unless
22681 you use the @option{-Wno-psabi} option.
22683 @item -mabi=ieeelongdouble
22684 @opindex mabi=ieeelongdouble
22685 Change the current ABI to use IEEE extended-precision long double.
22686 This is not likely to work if your system defaults to using IBM
22687 extended-precision long double. If you change the long double type
22688 from IBM extended-precision, the compiler will issue a warning unless
22689 you use the @option{-Wno-psabi} option.
22692 @opindex mabi=elfv1
22693 Change the current ABI to use the ELFv1 ABI.
22694 This is the default ABI for big-endian PowerPC 64-bit Linux.
22695 Overriding the default ABI requires special system support and is
22696 likely to fail in spectacular ways.
22699 @opindex mabi=elfv2
22700 Change the current ABI to use the ELFv2 ABI.
22701 This is the default ABI for little-endian PowerPC 64-bit Linux.
22702 Overriding the default ABI requires special system support and is
22703 likely to fail in spectacular ways.
22705 @item -mgnu-attribute
22706 @itemx -mno-gnu-attribute
22707 @opindex mgnu-attribute
22708 @opindex mno-gnu-attribute
22709 Emit .gnu_attribute assembly directives to set tag/value pairs in a
22710 .gnu.attributes section that specify ABI variations in function
22711 parameters or return values.
22714 @itemx -mno-prototype
22715 @opindex mprototype
22716 @opindex mno-prototype
22717 On System V.4 and embedded PowerPC systems assume that all calls to
22718 variable argument functions are properly prototyped. Otherwise, the
22719 compiler must insert an instruction before every non-prototyped call to
22720 set or clear bit 6 of the condition code register (@code{CR}) to
22721 indicate whether floating-point values are passed in the floating-point
22722 registers in case the function takes variable arguments. With
22723 @option{-mprototype}, only calls to prototyped variable argument functions
22724 set or clear the bit.
22728 On embedded PowerPC systems, assume that the startup module is called
22729 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
22730 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
22735 On embedded PowerPC systems, assume that the startup module is called
22736 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
22741 On embedded PowerPC systems, assume that the startup module is called
22742 @file{crt0.o} and the standard C libraries are @file{libads.a} and
22745 @item -myellowknife
22746 @opindex myellowknife
22747 On embedded PowerPC systems, assume that the startup module is called
22748 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
22753 On System V.4 and embedded PowerPC systems, specify that you are
22754 compiling for a VxWorks system.
22758 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
22759 header to indicate that @samp{eabi} extended relocations are used.
22765 On System V.4 and embedded PowerPC systems do (do not) adhere to the
22766 Embedded Applications Binary Interface (EABI), which is a set of
22767 modifications to the System V.4 specifications. Selecting @option{-meabi}
22768 means that the stack is aligned to an 8-byte boundary, a function
22769 @code{__eabi} is called from @code{main} to set up the EABI
22770 environment, and the @option{-msdata} option can use both @code{r2} and
22771 @code{r13} to point to two separate small data areas. Selecting
22772 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
22773 no EABI initialization function is called from @code{main}, and the
22774 @option{-msdata} option only uses @code{r13} to point to a single
22775 small data area. The @option{-meabi} option is on by default if you
22776 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
22779 @opindex msdata=eabi
22780 On System V.4 and embedded PowerPC systems, put small initialized
22781 @code{const} global and static data in the @code{.sdata2} section, which
22782 is pointed to by register @code{r2}. Put small initialized
22783 non-@code{const} global and static data in the @code{.sdata} section,
22784 which is pointed to by register @code{r13}. Put small uninitialized
22785 global and static data in the @code{.sbss} section, which is adjacent to
22786 the @code{.sdata} section. The @option{-msdata=eabi} option is
22787 incompatible with the @option{-mrelocatable} option. The
22788 @option{-msdata=eabi} option also sets the @option{-memb} option.
22791 @opindex msdata=sysv
22792 On System V.4 and embedded PowerPC systems, put small global and static
22793 data in the @code{.sdata} section, which is pointed to by register
22794 @code{r13}. Put small uninitialized global and static data in the
22795 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
22796 The @option{-msdata=sysv} option is incompatible with the
22797 @option{-mrelocatable} option.
22799 @item -msdata=default
22801 @opindex msdata=default
22803 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
22804 compile code the same as @option{-msdata=eabi}, otherwise compile code the
22805 same as @option{-msdata=sysv}.
22808 @opindex msdata=data
22809 On System V.4 and embedded PowerPC systems, put small global
22810 data in the @code{.sdata} section. Put small uninitialized global
22811 data in the @code{.sbss} section. Do not use register @code{r13}
22812 to address small data however. This is the default behavior unless
22813 other @option{-msdata} options are used.
22817 @opindex msdata=none
22819 On embedded PowerPC systems, put all initialized global and static data
22820 in the @code{.data} section, and all uninitialized data in the
22821 @code{.bss} section.
22823 @item -mblock-move-inline-limit=@var{num}
22824 @opindex mblock-move-inline-limit
22825 Inline all block moves (such as calls to @code{memcpy} or structure
22826 copies) less than or equal to @var{num} bytes. The minimum value for
22827 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
22828 targets. The default value is target-specific.
22832 @cindex smaller data references (PowerPC)
22833 @cindex .sdata/.sdata2 references (PowerPC)
22834 On embedded PowerPC systems, put global and static items less than or
22835 equal to @var{num} bytes into the small data or BSS sections instead of
22836 the normal data or BSS section. By default, @var{num} is 8. The
22837 @option{-G @var{num}} switch is also passed to the linker.
22838 All modules should be compiled with the same @option{-G @var{num}} value.
22841 @itemx -mno-regnames
22843 @opindex mno-regnames
22844 On System V.4 and embedded PowerPC systems do (do not) emit register
22845 names in the assembly language output using symbolic forms.
22848 @itemx -mno-longcall
22850 @opindex mno-longcall
22851 By default assume that all calls are far away so that a longer and more
22852 expensive calling sequence is required. This is required for calls
22853 farther than 32 megabytes (33,554,432 bytes) from the current location.
22854 A short call is generated if the compiler knows
22855 the call cannot be that far away. This setting can be overridden by
22856 the @code{shortcall} function attribute, or by @code{#pragma
22859 Some linkers are capable of detecting out-of-range calls and generating
22860 glue code on the fly. On these systems, long calls are unnecessary and
22861 generate slower code. As of this writing, the AIX linker can do this,
22862 as can the GNU linker for PowerPC/64. It is planned to add this feature
22863 to the GNU linker for 32-bit PowerPC systems as well.
22865 In the future, GCC may ignore all longcall specifications
22866 when the linker is known to generate glue.
22868 @item -mtls-markers
22869 @itemx -mno-tls-markers
22870 @opindex mtls-markers
22871 @opindex mno-tls-markers
22872 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22873 specifying the function argument. The relocation allows the linker to
22874 reliably associate function call with argument setup instructions for
22875 TLS optimization, which in turn allows GCC to better schedule the
22881 This option enables use of the reciprocal estimate and
22882 reciprocal square root estimate instructions with additional
22883 Newton-Raphson steps to increase precision instead of doing a divide or
22884 square root and divide for floating-point arguments. You should use
22885 the @option{-ffast-math} option when using @option{-mrecip} (or at
22886 least @option{-funsafe-math-optimizations},
22887 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22888 @option{-fno-trapping-math}). Note that while the throughput of the
22889 sequence is generally higher than the throughput of the non-reciprocal
22890 instruction, the precision of the sequence can be decreased by up to 2
22891 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
22894 @item -mrecip=@var{opt}
22895 @opindex mrecip=opt
22896 This option controls which reciprocal estimate instructions
22897 may be used. @var{opt} is a comma-separated list of options, which may
22898 be preceded by a @code{!} to invert the option:
22903 Enable all estimate instructions.
22906 Enable the default instructions, equivalent to @option{-mrecip}.
22909 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22912 Enable the reciprocal approximation instructions for both
22913 single and double precision.
22916 Enable the single-precision reciprocal approximation instructions.
22919 Enable the double-precision reciprocal approximation instructions.
22922 Enable the reciprocal square root approximation instructions for both
22923 single and double precision.
22926 Enable the single-precision reciprocal square root approximation instructions.
22929 Enable the double-precision reciprocal square root approximation instructions.
22933 So, for example, @option{-mrecip=all,!rsqrtd} enables
22934 all of the reciprocal estimate instructions, except for the
22935 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
22936 which handle the double-precision reciprocal square root calculations.
22938 @item -mrecip-precision
22939 @itemx -mno-recip-precision
22940 @opindex mrecip-precision
22941 Assume (do not assume) that the reciprocal estimate instructions
22942 provide higher-precision estimates than is mandated by the PowerPC
22943 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
22944 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
22945 The double-precision square root estimate instructions are not generated by
22946 default on low-precision machines, since they do not provide an
22947 estimate that converges after three steps.
22949 @item -mpointers-to-nested-functions
22950 @itemx -mno-pointers-to-nested-functions
22951 @opindex mpointers-to-nested-functions
22952 Generate (do not generate) code to load up the static chain register
22953 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
22954 systems where a function pointer points to a 3-word descriptor giving
22955 the function address, TOC value to be loaded in register @code{r2}, and
22956 static chain value to be loaded in register @code{r11}. The
22957 @option{-mpointers-to-nested-functions} is on by default. You cannot
22958 call through pointers to nested functions or pointers
22959 to functions compiled in other languages that use the static chain if
22960 you use @option{-mno-pointers-to-nested-functions}.
22962 @item -msave-toc-indirect
22963 @itemx -mno-save-toc-indirect
22964 @opindex msave-toc-indirect
22965 Generate (do not generate) code to save the TOC value in the reserved
22966 stack location in the function prologue if the function calls through
22967 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
22968 saved in the prologue, it is saved just before the call through the
22969 pointer. The @option{-mno-save-toc-indirect} option is the default.
22971 @item -mcompat-align-parm
22972 @itemx -mno-compat-align-parm
22973 @opindex mcompat-align-parm
22974 Generate (do not generate) code to pass structure parameters with a
22975 maximum alignment of 64 bits, for compatibility with older versions
22978 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
22979 structure parameter on a 128-bit boundary when that structure contained
22980 a member requiring 128-bit alignment. This is corrected in more
22981 recent versions of GCC. This option may be used to generate code
22982 that is compatible with functions compiled with older versions of
22985 The @option{-mno-compat-align-parm} option is the default.
22987 @item -mstack-protector-guard=@var{guard}
22988 @itemx -mstack-protector-guard-reg=@var{reg}
22989 @itemx -mstack-protector-guard-offset=@var{offset}
22990 @itemx -mstack-protector-guard-symbol=@var{symbol}
22991 @opindex mstack-protector-guard
22992 @opindex mstack-protector-guard-reg
22993 @opindex mstack-protector-guard-offset
22994 @opindex mstack-protector-guard-symbol
22995 Generate stack protection code using canary at @var{guard}. Supported
22996 locations are @samp{global} for global canary or @samp{tls} for per-thread
22997 canary in the TLS block (the default with GNU libc version 2.4 or later).
22999 With the latter choice the options
23000 @option{-mstack-protector-guard-reg=@var{reg}} and
23001 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
23002 which register to use as base register for reading the canary, and from what
23003 offset from that base register. The default for those is as specified in the
23004 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
23005 the offset with a symbol reference to a canary in the TLS block.
23009 @node RISC-V Options
23010 @subsection RISC-V Options
23011 @cindex RISC-V Options
23013 These command-line options are defined for RISC-V targets:
23016 @item -mbranch-cost=@var{n}
23017 @opindex mbranch-cost
23018 Set the cost of branches to roughly @var{n} instructions.
23023 When generating PIC code, do or don't allow the use of PLTs. Ignored for
23024 non-PIC. The default is @option{-mplt}.
23026 @item -mabi=@var{ABI-string}
23028 Specify integer and floating-point calling convention. @var{ABI-string}
23029 contains two parts: the size of integer types and the registers used for
23030 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
23031 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
23032 32-bit), and that floating-point values up to 64 bits wide are passed in F
23033 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
23034 allows the compiler to generate code that uses the F and D extensions but only
23035 allows floating-point values up to 32 bits long to be passed in registers; or
23036 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
23037 passed in registers.
23039 The default for this argument is system dependent, users who want a specific
23040 calling convention should specify one explicitly. The valid calling
23041 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
23042 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
23043 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
23044 invalid because the ABI requires 64-bit values be passed in F registers, but F
23045 registers are only 32 bits wide.
23050 Do or don't use hardware floating-point divide and square root instructions.
23051 This requires the F or D extensions for floating-point registers. The default
23052 is to use them if the specified architecture has these instructions.
23057 Do or don't use hardware instructions for integer division. This requires the
23058 M extension. The default is to use them if the specified architecture has
23059 these instructions.
23061 @item -march=@var{ISA-string}
23063 Generate code for given RISC-V ISA (e.g.@ @samp{rv64im}). ISA strings must be
23064 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, and @samp{rv32imaf}.
23066 @item -mtune=@var{processor-string}
23068 Optimize the output for the given processor, specified by microarchitecture
23071 @item -mpreferred-stack-boundary=@var{num}
23072 @opindex mpreferred-stack-boundary
23073 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23074 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23075 the default is 4 (16 bytes or 128-bits).
23077 @strong{Warning:} If you use this switch, then you must build all modules with
23078 the same value, including any libraries. This includes the system libraries
23079 and startup modules.
23081 @item -msmall-data-limit=@var{n}
23082 @opindex msmall-data-limit
23083 Put global and static data smaller than @var{n} bytes into a special section
23086 @item -msave-restore
23087 @itemx -mno-save-restore
23088 @opindex msave-restore
23089 Do or don't use smaller but slower prologue and epilogue code that uses
23090 library function calls. The default is to use fast inline prologues and
23093 @item -mstrict-align
23094 @itemx -mno-strict-align
23095 @opindex mstrict-align
23096 Do not or do generate unaligned memory accesses. The default is set depending
23097 on whether the processor we are optimizing for supports fast unaligned access
23100 @item -mcmodel=medlow
23101 @opindex mcmodel=medlow
23102 Generate code for the medium-low code model. The program and its statically
23103 defined symbols must lie within a single 2 GiB address range and must lie
23104 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
23105 statically or dynamically linked. This is the default code model.
23107 @item -mcmodel=medany
23108 @opindex mcmodel=medany
23109 Generate code for the medium-any code model. The program and its statically
23110 defined symbols must be within any single 2 GiB address range. Programs can be
23111 statically or dynamically linked.
23113 @item -mexplicit-relocs
23114 @itemx -mno-exlicit-relocs
23115 Use or do not use assembler relocation operators when dealing with symbolic
23116 addresses. The alternative is to use assembler macros instead, which may
23117 limit optimization.
23121 Take advantage of linker relaxations to reduce the number of instructions
23122 required to materialize symbol addresses. The default is to take advantage of
23123 linker relaxations.
23128 @subsection RL78 Options
23129 @cindex RL78 Options
23135 Links in additional target libraries to support operation within a
23144 Specifies the type of hardware multiplication and division support to
23145 be used. The simplest is @code{none}, which uses software for both
23146 multiplication and division. This is the default. The @code{g13}
23147 value is for the hardware multiply/divide peripheral found on the
23148 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
23149 the multiplication and division instructions supported by the RL78/G14
23150 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
23151 the value @code{mg10} is an alias for @code{none}.
23153 In addition a C preprocessor macro is defined, based upon the setting
23154 of this option. Possible values are: @code{__RL78_MUL_NONE__},
23155 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
23162 Specifies the RL78 core to target. The default is the G14 core, also
23163 known as an S3 core or just RL78. The G13 or S2 core does not have
23164 multiply or divide instructions, instead it uses a hardware peripheral
23165 for these operations. The G10 or S1 core does not have register
23166 banks, so it uses a different calling convention.
23168 If this option is set it also selects the type of hardware multiply
23169 support to use, unless this is overridden by an explicit
23170 @option{-mmul=none} option on the command line. Thus specifying
23171 @option{-mcpu=g13} enables the use of the G13 hardware multiply
23172 peripheral and specifying @option{-mcpu=g10} disables the use of
23173 hardware multiplications altogether.
23175 Note, although the RL78/G14 core is the default target, specifying
23176 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
23177 change the behavior of the toolchain since it also enables G14
23178 hardware multiply support. If these options are not specified on the
23179 command line then software multiplication routines will be used even
23180 though the code targets the RL78 core. This is for backwards
23181 compatibility with older toolchains which did not have hardware
23182 multiply and divide support.
23184 In addition a C preprocessor macro is defined, based upon the setting
23185 of this option. Possible values are: @code{__RL78_G10__},
23186 @code{__RL78_G13__} or @code{__RL78_G14__}.
23196 These are aliases for the corresponding @option{-mcpu=} option. They
23197 are provided for backwards compatibility.
23201 Allow the compiler to use all of the available registers. By default
23202 registers @code{r24..r31} are reserved for use in interrupt handlers.
23203 With this option enabled these registers can be used in ordinary
23206 @item -m64bit-doubles
23207 @itemx -m32bit-doubles
23208 @opindex m64bit-doubles
23209 @opindex m32bit-doubles
23210 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23211 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23212 @option{-m32bit-doubles}.
23214 @item -msave-mduc-in-interrupts
23215 @itemx -mno-save-mduc-in-interrupts
23216 @opindex msave-mduc-in-interrupts
23217 @opindex mno-save-mduc-in-interrupts
23218 Specifies that interrupt handler functions should preserve the
23219 MDUC registers. This is only necessary if normal code might use
23220 the MDUC registers, for example because it performs multiplication
23221 and division operations. The default is to ignore the MDUC registers
23222 as this makes the interrupt handlers faster. The target option -mg13
23223 needs to be passed for this to work as this feature is only available
23224 on the G13 target (S2 core). The MDUC registers will only be saved
23225 if the interrupt handler performs a multiplication or division
23226 operation or it calls another function.
23230 @node RS/6000 and PowerPC Options
23231 @subsection IBM RS/6000 and PowerPC Options
23232 @cindex RS/6000 and PowerPC Options
23233 @cindex IBM RS/6000 and PowerPC Options
23235 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
23237 @item -mpowerpc-gpopt
23238 @itemx -mno-powerpc-gpopt
23239 @itemx -mpowerpc-gfxopt
23240 @itemx -mno-powerpc-gfxopt
23243 @itemx -mno-powerpc64
23247 @itemx -mno-popcntb
23249 @itemx -mno-popcntd
23258 @itemx -mno-hard-dfp
23259 @opindex mpowerpc-gpopt
23260 @opindex mno-powerpc-gpopt
23261 @opindex mpowerpc-gfxopt
23262 @opindex mno-powerpc-gfxopt
23263 @opindex mpowerpc64
23264 @opindex mno-powerpc64
23268 @opindex mno-popcntb
23270 @opindex mno-popcntd
23276 @opindex mno-mfpgpr
23278 @opindex mno-hard-dfp
23279 You use these options to specify which instructions are available on the
23280 processor you are using. The default value of these options is
23281 determined when configuring GCC@. Specifying the
23282 @option{-mcpu=@var{cpu_type}} overrides the specification of these
23283 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
23284 rather than the options listed above.
23286 Specifying @option{-mpowerpc-gpopt} allows
23287 GCC to use the optional PowerPC architecture instructions in the
23288 General Purpose group, including floating-point square root. Specifying
23289 @option{-mpowerpc-gfxopt} allows GCC to
23290 use the optional PowerPC architecture instructions in the Graphics
23291 group, including floating-point select.
23293 The @option{-mmfcrf} option allows GCC to generate the move from
23294 condition register field instruction implemented on the POWER4
23295 processor and other processors that support the PowerPC V2.01
23297 The @option{-mpopcntb} option allows GCC to generate the popcount and
23298 double-precision FP reciprocal estimate instruction implemented on the
23299 POWER5 processor and other processors that support the PowerPC V2.02
23301 The @option{-mpopcntd} option allows GCC to generate the popcount
23302 instruction implemented on the POWER7 processor and other processors
23303 that support the PowerPC V2.06 architecture.
23304 The @option{-mfprnd} option allows GCC to generate the FP round to
23305 integer instructions implemented on the POWER5+ processor and other
23306 processors that support the PowerPC V2.03 architecture.
23307 The @option{-mcmpb} option allows GCC to generate the compare bytes
23308 instruction implemented on the POWER6 processor and other processors
23309 that support the PowerPC V2.05 architecture.
23310 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
23311 general-purpose register instructions implemented on the POWER6X
23312 processor and other processors that support the extended PowerPC V2.05
23314 The @option{-mhard-dfp} option allows GCC to generate the decimal
23315 floating-point instructions implemented on some POWER processors.
23317 The @option{-mpowerpc64} option allows GCC to generate the additional
23318 64-bit instructions that are found in the full PowerPC64 architecture
23319 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
23320 @option{-mno-powerpc64}.
23322 @item -mcpu=@var{cpu_type}
23324 Set architecture type, register usage, and
23325 instruction scheduling parameters for machine type @var{cpu_type}.
23326 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
23327 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
23328 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
23329 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
23330 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
23331 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
23332 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
23333 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
23334 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
23335 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
23336 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
23337 @samp{rs64}, and @samp{native}.
23339 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
23340 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
23341 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
23342 architecture machine types, with an appropriate, generic processor
23343 model assumed for scheduling purposes.
23345 Specifying @samp{native} as cpu type detects and selects the
23346 architecture option that corresponds to the host processor of the
23347 system performing the compilation.
23348 @option{-mcpu=native} has no effect if GCC does not recognize the
23351 The other options specify a specific processor. Code generated under
23352 those options runs best on that processor, and may not run at all on
23355 The @option{-mcpu} options automatically enable or disable the
23358 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
23359 -mpopcntb -mpopcntd -mpowerpc64 @gol
23360 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
23361 -msimple-fpu -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
23362 -mcrypto -mhtm -mpower8-fusion -mpower8-vector @gol
23363 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
23365 The particular options set for any particular CPU varies between
23366 compiler versions, depending on what setting seems to produce optimal
23367 code for that CPU; it doesn't necessarily reflect the actual hardware's
23368 capabilities. If you wish to set an individual option to a particular
23369 value, you may specify it after the @option{-mcpu} option, like
23370 @option{-mcpu=970 -mno-altivec}.
23372 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
23373 not enabled or disabled by the @option{-mcpu} option at present because
23374 AIX does not have full support for these options. You may still
23375 enable or disable them individually if you're sure it'll work in your
23378 @item -mtune=@var{cpu_type}
23380 Set the instruction scheduling parameters for machine type
23381 @var{cpu_type}, but do not set the architecture type or register usage,
23382 as @option{-mcpu=@var{cpu_type}} does. The same
23383 values for @var{cpu_type} are used for @option{-mtune} as for
23384 @option{-mcpu}. If both are specified, the code generated uses the
23385 architecture and registers set by @option{-mcpu}, but the
23386 scheduling parameters set by @option{-mtune}.
23388 @item -mcmodel=small
23389 @opindex mcmodel=small
23390 Generate PowerPC64 code for the small model: The TOC is limited to
23393 @item -mcmodel=medium
23394 @opindex mcmodel=medium
23395 Generate PowerPC64 code for the medium model: The TOC and other static
23396 data may be up to a total of 4G in size. This is the default for 64-bit
23399 @item -mcmodel=large
23400 @opindex mcmodel=large
23401 Generate PowerPC64 code for the large model: The TOC may be up to 4G
23402 in size. Other data and code is only limited by the 64-bit address
23406 @itemx -mno-altivec
23408 @opindex mno-altivec
23409 Generate code that uses (does not use) AltiVec instructions, and also
23410 enable the use of built-in functions that allow more direct access to
23411 the AltiVec instruction set. You may also need to set
23412 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
23415 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
23416 @option{-maltivec=be}, the element order for AltiVec intrinsics such
23417 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
23418 match array element order corresponding to the endianness of the
23419 target. That is, element zero identifies the leftmost element in a
23420 vector register when targeting a big-endian platform, and identifies
23421 the rightmost element in a vector register when targeting a
23422 little-endian platform.
23425 @opindex maltivec=be
23426 Generate AltiVec instructions using big-endian element order,
23427 regardless of whether the target is big- or little-endian. This is
23428 the default when targeting a big-endian platform. Using this option
23429 is currently deprecated. Support for this feature will be removed in
23432 The element order is used to interpret element numbers in AltiVec
23433 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
23434 @code{vec_insert}. By default, these match array element order
23435 corresponding to the endianness for the target.
23438 @opindex maltivec=le
23439 Generate AltiVec instructions using little-endian element order,
23440 regardless of whether the target is big- or little-endian. This is
23441 the default when targeting a little-endian platform. This option is
23442 currently ignored when targeting a big-endian platform.
23444 The element order is used to interpret element numbers in AltiVec
23445 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
23446 @code{vec_insert}. By default, these match array element order
23447 corresponding to the endianness for the target.
23452 @opindex mno-vrsave
23453 Generate VRSAVE instructions when generating AltiVec code.
23456 @opindex msecure-plt
23457 Generate code that allows @command{ld} and @command{ld.so}
23458 to build executables and shared
23459 libraries with non-executable @code{.plt} and @code{.got} sections.
23461 32-bit SYSV ABI option.
23465 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
23467 requires @code{.plt} and @code{.got}
23468 sections that are both writable and executable.
23469 This is a PowerPC 32-bit SYSV ABI option.
23475 This switch enables or disables the generation of ISEL instructions.
23477 @item -misel=@var{yes/no}
23478 This switch has been deprecated. Use @option{-misel} and
23479 @option{-mno-isel} instead.
23484 @opindex mno-paired
23485 This switch enables or disables the generation of PAIRED simd
23492 Generate code that uses (does not use) vector/scalar (VSX)
23493 instructions, and also enable the use of built-in functions that allow
23494 more direct access to the VSX instruction set.
23499 @opindex mno-crypto
23500 Enable the use (disable) of the built-in functions that allow direct
23501 access to the cryptographic instructions that were added in version
23502 2.07 of the PowerPC ISA.
23508 Enable (disable) the use of the built-in functions that allow direct
23509 access to the Hardware Transactional Memory (HTM) instructions that
23510 were added in version 2.07 of the PowerPC ISA.
23512 @item -mpower8-fusion
23513 @itemx -mno-power8-fusion
23514 @opindex mpower8-fusion
23515 @opindex mno-power8-fusion
23516 Generate code that keeps (does not keeps) some integer operations
23517 adjacent so that the instructions can be fused together on power8 and
23520 @item -mpower8-vector
23521 @itemx -mno-power8-vector
23522 @opindex mpower8-vector
23523 @opindex mno-power8-vector
23524 Generate code that uses (does not use) the vector and scalar
23525 instructions that were added in version 2.07 of the PowerPC ISA. Also
23526 enable the use of built-in functions that allow more direct access to
23527 the vector instructions.
23529 @item -mquad-memory
23530 @itemx -mno-quad-memory
23531 @opindex mquad-memory
23532 @opindex mno-quad-memory
23533 Generate code that uses (does not use) the non-atomic quad word memory
23534 instructions. The @option{-mquad-memory} option requires use of
23537 @item -mquad-memory-atomic
23538 @itemx -mno-quad-memory-atomic
23539 @opindex mquad-memory-atomic
23540 @opindex mno-quad-memory-atomic
23541 Generate code that uses (does not use) the atomic quad word memory
23542 instructions. The @option{-mquad-memory-atomic} option requires use of
23546 @itemx -mno-float128
23548 @opindex mno-float128
23549 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
23550 and use either software emulation for IEEE 128-bit floating point or
23551 hardware instructions.
23553 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
23554 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
23555 use the IEEE 128-bit floating point support. The IEEE 128-bit
23556 floating point support only works on PowerPC Linux systems.
23558 The default for @option{-mfloat128} is enabled on PowerPC Linux
23559 systems using the VSX instruction set, and disabled on other systems.
23561 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
23562 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
23563 point support will also enable the generation of ISA 3.0 IEEE 128-bit
23564 floating point instructions. Otherwise, if you do not specify to
23565 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
23566 system, IEEE 128-bit floating point will be done with software
23569 @item -mfloat128-hardware
23570 @itemx -mno-float128-hardware
23571 @opindex mfloat128-hardware
23572 @opindex mno-float128-hardware
23573 Enable/disable using ISA 3.0 hardware instructions to support the
23574 @var{__float128} data type.
23576 The default for @option{-mfloat128-hardware} is enabled on PowerPC
23577 Linux systems using the ISA 3.0 instruction set, and disabled on other
23584 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
23585 targets (including GNU/Linux). The 32-bit environment sets int, long
23586 and pointer to 32 bits and generates code that runs on any PowerPC
23587 variant. The 64-bit environment sets int to 32 bits and long and
23588 pointer to 64 bits, and generates code for PowerPC64, as for
23589 @option{-mpowerpc64}.
23592 @itemx -mno-fp-in-toc
23593 @itemx -mno-sum-in-toc
23594 @itemx -mminimal-toc
23596 @opindex mno-fp-in-toc
23597 @opindex mno-sum-in-toc
23598 @opindex mminimal-toc
23599 Modify generation of the TOC (Table Of Contents), which is created for
23600 every executable file. The @option{-mfull-toc} option is selected by
23601 default. In that case, GCC allocates at least one TOC entry for
23602 each unique non-automatic variable reference in your program. GCC
23603 also places floating-point constants in the TOC@. However, only
23604 16,384 entries are available in the TOC@.
23606 If you receive a linker error message that saying you have overflowed
23607 the available TOC space, you can reduce the amount of TOC space used
23608 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
23609 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
23610 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
23611 generate code to calculate the sum of an address and a constant at
23612 run time instead of putting that sum into the TOC@. You may specify one
23613 or both of these options. Each causes GCC to produce very slightly
23614 slower and larger code at the expense of conserving TOC space.
23616 If you still run out of space in the TOC even when you specify both of
23617 these options, specify @option{-mminimal-toc} instead. This option causes
23618 GCC to make only one TOC entry for every file. When you specify this
23619 option, GCC produces code that is slower and larger but which
23620 uses extremely little TOC space. You may wish to use this option
23621 only on files that contain less frequently-executed code.
23627 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
23628 @code{long} type, and the infrastructure needed to support them.
23629 Specifying @option{-maix64} implies @option{-mpowerpc64},
23630 while @option{-maix32} disables the 64-bit ABI and
23631 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
23634 @itemx -mno-xl-compat
23635 @opindex mxl-compat
23636 @opindex mno-xl-compat
23637 Produce code that conforms more closely to IBM XL compiler semantics
23638 when using AIX-compatible ABI@. Pass floating-point arguments to
23639 prototyped functions beyond the register save area (RSA) on the stack
23640 in addition to argument FPRs. Do not assume that most significant
23641 double in 128-bit long double value is properly rounded when comparing
23642 values and converting to double. Use XL symbol names for long double
23645 The AIX calling convention was extended but not initially documented to
23646 handle an obscure K&R C case of calling a function that takes the
23647 address of its arguments with fewer arguments than declared. IBM XL
23648 compilers access floating-point arguments that do not fit in the
23649 RSA from the stack when a subroutine is compiled without
23650 optimization. Because always storing floating-point arguments on the
23651 stack is inefficient and rarely needed, this option is not enabled by
23652 default and only is necessary when calling subroutines compiled by IBM
23653 XL compilers without optimization.
23657 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
23658 application written to use message passing with special startup code to
23659 enable the application to run. The system must have PE installed in the
23660 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
23661 must be overridden with the @option{-specs=} option to specify the
23662 appropriate directory location. The Parallel Environment does not
23663 support threads, so the @option{-mpe} option and the @option{-pthread}
23664 option are incompatible.
23666 @item -malign-natural
23667 @itemx -malign-power
23668 @opindex malign-natural
23669 @opindex malign-power
23670 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
23671 @option{-malign-natural} overrides the ABI-defined alignment of larger
23672 types, such as floating-point doubles, on their natural size-based boundary.
23673 The option @option{-malign-power} instructs GCC to follow the ABI-specified
23674 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
23676 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
23680 @itemx -mhard-float
23681 @opindex msoft-float
23682 @opindex mhard-float
23683 Generate code that does not use (uses) the floating-point register set.
23684 Software floating-point emulation is provided if you use the
23685 @option{-msoft-float} option, and pass the option to GCC when linking.
23687 @item -msingle-float
23688 @itemx -mdouble-float
23689 @opindex msingle-float
23690 @opindex mdouble-float
23691 Generate code for single- or double-precision floating-point operations.
23692 @option{-mdouble-float} implies @option{-msingle-float}.
23695 @opindex msimple-fpu
23696 Do not generate @code{sqrt} and @code{div} instructions for hardware
23697 floating-point unit.
23699 @item -mfpu=@var{name}
23701 Specify type of floating-point unit. Valid values for @var{name} are
23702 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
23703 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
23704 @samp{sp_full} (equivalent to @option{-msingle-float}),
23705 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
23708 @opindex mxilinx-fpu
23709 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
23712 @itemx -mno-multiple
23714 @opindex mno-multiple
23715 Generate code that uses (does not use) the load multiple word
23716 instructions and the store multiple word instructions. These
23717 instructions are generated by default on POWER systems, and not
23718 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
23719 PowerPC systems, since those instructions do not work when the
23720 processor is in little-endian mode. The exceptions are PPC740 and
23721 PPC750 which permit these instructions in little-endian mode.
23726 @opindex mno-update
23727 Generate code that uses (does not use) the load or store instructions
23728 that update the base register to the address of the calculated memory
23729 location. These instructions are generated by default. If you use
23730 @option{-mno-update}, there is a small window between the time that the
23731 stack pointer is updated and the address of the previous frame is
23732 stored, which means code that walks the stack frame across interrupts or
23733 signals may get corrupted data.
23735 @item -mavoid-indexed-addresses
23736 @itemx -mno-avoid-indexed-addresses
23737 @opindex mavoid-indexed-addresses
23738 @opindex mno-avoid-indexed-addresses
23739 Generate code that tries to avoid (not avoid) the use of indexed load
23740 or store instructions. These instructions can incur a performance
23741 penalty on Power6 processors in certain situations, such as when
23742 stepping through large arrays that cross a 16M boundary. This option
23743 is enabled by default when targeting Power6 and disabled otherwise.
23746 @itemx -mno-fused-madd
23747 @opindex mfused-madd
23748 @opindex mno-fused-madd
23749 Generate code that uses (does not use) the floating-point multiply and
23750 accumulate instructions. These instructions are generated by default
23751 if hardware floating point is used. The machine-dependent
23752 @option{-mfused-madd} option is now mapped to the machine-independent
23753 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
23754 mapped to @option{-ffp-contract=off}.
23760 Generate code that uses (does not use) the half-word multiply and
23761 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
23762 These instructions are generated by default when targeting those
23769 Generate code that uses (does not use) the string-search @samp{dlmzb}
23770 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
23771 generated by default when targeting those processors.
23773 @item -mno-bit-align
23775 @opindex mno-bit-align
23776 @opindex mbit-align
23777 On System V.4 and embedded PowerPC systems do not (do) force structures
23778 and unions that contain bit-fields to be aligned to the base type of the
23781 For example, by default a structure containing nothing but 8
23782 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
23783 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
23784 the structure is aligned to a 1-byte boundary and is 1 byte in
23787 @item -mno-strict-align
23788 @itemx -mstrict-align
23789 @opindex mno-strict-align
23790 @opindex mstrict-align
23791 On System V.4 and embedded PowerPC systems do not (do) assume that
23792 unaligned memory references are handled by the system.
23794 @item -mrelocatable
23795 @itemx -mno-relocatable
23796 @opindex mrelocatable
23797 @opindex mno-relocatable
23798 Generate code that allows (does not allow) a static executable to be
23799 relocated to a different address at run time. A simple embedded
23800 PowerPC system loader should relocate the entire contents of
23801 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
23802 a table of 32-bit addresses generated by this option. For this to
23803 work, all objects linked together must be compiled with
23804 @option{-mrelocatable} or @option{-mrelocatable-lib}.
23805 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
23807 @item -mrelocatable-lib
23808 @itemx -mno-relocatable-lib
23809 @opindex mrelocatable-lib
23810 @opindex mno-relocatable-lib
23811 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
23812 @code{.fixup} section to allow static executables to be relocated at
23813 run time, but @option{-mrelocatable-lib} does not use the smaller stack
23814 alignment of @option{-mrelocatable}. Objects compiled with
23815 @option{-mrelocatable-lib} may be linked with objects compiled with
23816 any combination of the @option{-mrelocatable} options.
23822 On System V.4 and embedded PowerPC systems do not (do) assume that
23823 register 2 contains a pointer to a global area pointing to the addresses
23824 used in the program.
23827 @itemx -mlittle-endian
23829 @opindex mlittle-endian
23830 On System V.4 and embedded PowerPC systems compile code for the
23831 processor in little-endian mode. The @option{-mlittle-endian} option is
23832 the same as @option{-mlittle}.
23835 @itemx -mbig-endian
23837 @opindex mbig-endian
23838 On System V.4 and embedded PowerPC systems compile code for the
23839 processor in big-endian mode. The @option{-mbig-endian} option is
23840 the same as @option{-mbig}.
23842 @item -mdynamic-no-pic
23843 @opindex mdynamic-no-pic
23844 On Darwin and Mac OS X systems, compile code so that it is not
23845 relocatable, but that its external references are relocatable. The
23846 resulting code is suitable for applications, but not shared
23849 @item -msingle-pic-base
23850 @opindex msingle-pic-base
23851 Treat the register used for PIC addressing as read-only, rather than
23852 loading it in the prologue for each function. The runtime system is
23853 responsible for initializing this register with an appropriate value
23854 before execution begins.
23856 @item -mprioritize-restricted-insns=@var{priority}
23857 @opindex mprioritize-restricted-insns
23858 This option controls the priority that is assigned to
23859 dispatch-slot restricted instructions during the second scheduling
23860 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
23861 or @samp{2} to assign no, highest, or second-highest (respectively)
23862 priority to dispatch-slot restricted
23865 @item -msched-costly-dep=@var{dependence_type}
23866 @opindex msched-costly-dep
23867 This option controls which dependences are considered costly
23868 by the target during instruction scheduling. The argument
23869 @var{dependence_type} takes one of the following values:
23873 No dependence is costly.
23876 All dependences are costly.
23878 @item @samp{true_store_to_load}
23879 A true dependence from store to load is costly.
23881 @item @samp{store_to_load}
23882 Any dependence from store to load is costly.
23885 Any dependence for which the latency is greater than or equal to
23886 @var{number} is costly.
23889 @item -minsert-sched-nops=@var{scheme}
23890 @opindex minsert-sched-nops
23891 This option controls which NOP insertion scheme is used during
23892 the second scheduling pass. The argument @var{scheme} takes one of the
23900 Pad with NOPs any dispatch group that has vacant issue slots,
23901 according to the scheduler's grouping.
23903 @item @samp{regroup_exact}
23904 Insert NOPs to force costly dependent insns into
23905 separate groups. Insert exactly as many NOPs as needed to force an insn
23906 to a new group, according to the estimated processor grouping.
23909 Insert NOPs to force costly dependent insns into
23910 separate groups. Insert @var{number} NOPs to force an insn to a new group.
23914 @opindex mcall-sysv
23915 On System V.4 and embedded PowerPC systems compile code using calling
23916 conventions that adhere to the March 1995 draft of the System V
23917 Application Binary Interface, PowerPC processor supplement. This is the
23918 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
23920 @item -mcall-sysv-eabi
23922 @opindex mcall-sysv-eabi
23923 @opindex mcall-eabi
23924 Specify both @option{-mcall-sysv} and @option{-meabi} options.
23926 @item -mcall-sysv-noeabi
23927 @opindex mcall-sysv-noeabi
23928 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
23930 @item -mcall-aixdesc
23932 On System V.4 and embedded PowerPC systems compile code for the AIX
23936 @opindex mcall-linux
23937 On System V.4 and embedded PowerPC systems compile code for the
23938 Linux-based GNU system.
23940 @item -mcall-freebsd
23941 @opindex mcall-freebsd
23942 On System V.4 and embedded PowerPC systems compile code for the
23943 FreeBSD operating system.
23945 @item -mcall-netbsd
23946 @opindex mcall-netbsd
23947 On System V.4 and embedded PowerPC systems compile code for the
23948 NetBSD operating system.
23950 @item -mcall-openbsd
23951 @opindex mcall-netbsd
23952 On System V.4 and embedded PowerPC systems compile code for the
23953 OpenBSD operating system.
23955 @item -mtraceback=@var{traceback_type}
23956 @opindex mtraceback
23957 Select the type of traceback table. Valid values for @var{traceback_type}
23958 are @samp{full}, @samp{part}, and @samp{no}.
23960 @item -maix-struct-return
23961 @opindex maix-struct-return
23962 Return all structures in memory (as specified by the AIX ABI)@.
23964 @item -msvr4-struct-return
23965 @opindex msvr4-struct-return
23966 Return structures smaller than 8 bytes in registers (as specified by the
23969 @item -mabi=@var{abi-type}
23971 Extend the current ABI with a particular extension, or remove such extension.
23972 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
23973 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
23974 @samp{elfv1}, @samp{elfv2}@.
23976 @item -mabi=ibmlongdouble
23977 @opindex mabi=ibmlongdouble
23978 Change the current ABI to use IBM extended-precision long double.
23979 This is not likely to work if your system defaults to using IEEE
23980 extended-precision long double. If you change the long double type
23981 from IEEE extended-precision, the compiler will issue a warning unless
23982 you use the @option{-Wno-psabi} option.
23984 @item -mabi=ieeelongdouble
23985 @opindex mabi=ieeelongdouble
23986 Change the current ABI to use IEEE extended-precision long double.
23987 This is not likely to work if your system defaults to using IBM
23988 extended-precision long double. If you change the long double type
23989 from IBM extended-precision, the compiler will issue a warning unless
23990 you use the @option{-Wno-psabi} option.
23993 @opindex mabi=elfv1
23994 Change the current ABI to use the ELFv1 ABI.
23995 This is the default ABI for big-endian PowerPC 64-bit Linux.
23996 Overriding the default ABI requires special system support and is
23997 likely to fail in spectacular ways.
24000 @opindex mabi=elfv2
24001 Change the current ABI to use the ELFv2 ABI.
24002 This is the default ABI for little-endian PowerPC 64-bit Linux.
24003 Overriding the default ABI requires special system support and is
24004 likely to fail in spectacular ways.
24006 @item -mgnu-attribute
24007 @itemx -mno-gnu-attribute
24008 @opindex mgnu-attribute
24009 @opindex mno-gnu-attribute
24010 Emit .gnu_attribute assembly directives to set tag/value pairs in a
24011 .gnu.attributes section that specify ABI variations in function
24012 parameters or return values.
24015 @itemx -mno-prototype
24016 @opindex mprototype
24017 @opindex mno-prototype
24018 On System V.4 and embedded PowerPC systems assume that all calls to
24019 variable argument functions are properly prototyped. Otherwise, the
24020 compiler must insert an instruction before every non-prototyped call to
24021 set or clear bit 6 of the condition code register (@code{CR}) to
24022 indicate whether floating-point values are passed in the floating-point
24023 registers in case the function takes variable arguments. With
24024 @option{-mprototype}, only calls to prototyped variable argument functions
24025 set or clear the bit.
24029 On embedded PowerPC systems, assume that the startup module is called
24030 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
24031 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
24036 On embedded PowerPC systems, assume that the startup module is called
24037 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
24042 On embedded PowerPC systems, assume that the startup module is called
24043 @file{crt0.o} and the standard C libraries are @file{libads.a} and
24046 @item -myellowknife
24047 @opindex myellowknife
24048 On embedded PowerPC systems, assume that the startup module is called
24049 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
24054 On System V.4 and embedded PowerPC systems, specify that you are
24055 compiling for a VxWorks system.
24059 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
24060 header to indicate that @samp{eabi} extended relocations are used.
24066 On System V.4 and embedded PowerPC systems do (do not) adhere to the
24067 Embedded Applications Binary Interface (EABI), which is a set of
24068 modifications to the System V.4 specifications. Selecting @option{-meabi}
24069 means that the stack is aligned to an 8-byte boundary, a function
24070 @code{__eabi} is called from @code{main} to set up the EABI
24071 environment, and the @option{-msdata} option can use both @code{r2} and
24072 @code{r13} to point to two separate small data areas. Selecting
24073 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
24074 no EABI initialization function is called from @code{main}, and the
24075 @option{-msdata} option only uses @code{r13} to point to a single
24076 small data area. The @option{-meabi} option is on by default if you
24077 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
24080 @opindex msdata=eabi
24081 On System V.4 and embedded PowerPC systems, put small initialized
24082 @code{const} global and static data in the @code{.sdata2} section, which
24083 is pointed to by register @code{r2}. Put small initialized
24084 non-@code{const} global and static data in the @code{.sdata} section,
24085 which is pointed to by register @code{r13}. Put small uninitialized
24086 global and static data in the @code{.sbss} section, which is adjacent to
24087 the @code{.sdata} section. The @option{-msdata=eabi} option is
24088 incompatible with the @option{-mrelocatable} option. The
24089 @option{-msdata=eabi} option also sets the @option{-memb} option.
24092 @opindex msdata=sysv
24093 On System V.4 and embedded PowerPC systems, put small global and static
24094 data in the @code{.sdata} section, which is pointed to by register
24095 @code{r13}. Put small uninitialized global and static data in the
24096 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
24097 The @option{-msdata=sysv} option is incompatible with the
24098 @option{-mrelocatable} option.
24100 @item -msdata=default
24102 @opindex msdata=default
24104 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
24105 compile code the same as @option{-msdata=eabi}, otherwise compile code the
24106 same as @option{-msdata=sysv}.
24109 @opindex msdata=data
24110 On System V.4 and embedded PowerPC systems, put small global
24111 data in the @code{.sdata} section. Put small uninitialized global
24112 data in the @code{.sbss} section. Do not use register @code{r13}
24113 to address small data however. This is the default behavior unless
24114 other @option{-msdata} options are used.
24118 @opindex msdata=none
24120 On embedded PowerPC systems, put all initialized global and static data
24121 in the @code{.data} section, and all uninitialized data in the
24122 @code{.bss} section.
24124 @item -mreadonly-in-sdata
24125 @itemx -mreadonly-in-sdata
24126 @opindex mreadonly-in-sdata
24127 @opindex mno-readonly-in-sdata
24128 Put read-only objects in the @code{.sdata} section as well. This is the
24131 @item -mblock-move-inline-limit=@var{num}
24132 @opindex mblock-move-inline-limit
24133 Inline all block moves (such as calls to @code{memcpy} or structure
24134 copies) less than or equal to @var{num} bytes. The minimum value for
24135 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
24136 targets. The default value is target-specific.
24138 @item -mblock-compare-inline-limit=@var{num}
24139 @opindex mblock-compare-inline-limit
24140 Generate non-looping inline code for all block compares (such as calls
24141 to @code{memcmp} or structure compares) less than or equal to @var{num}
24142 bytes. If @var{num} is 0, all inline expansion (non-loop and loop) of
24143 block compare is disabled. The default value is target-specific.
24145 @item -mblock-compare-inline-loop-limit=@var{num}
24146 @opindex mblock-compare-inline-loop-limit
24147 Generate an inline expansion using loop code for all block compares that
24148 are less than or equal to @var{num} bytes, but greater than the limit
24149 for non-loop inline block compare expansion. If the block length is not
24150 constant, at most @var{num} bytes will be compared before @code{memcmp}
24151 is called to compare the remainder of the block. The default value is
24154 @item -mstring-compare-inline-limit=@var{num}
24155 @opindex mstring-compare-inline-limit
24156 Generate at most @var{num} pairs of load instructions to compare the
24157 string inline. If the difference or end of string is not found at the
24158 end of the inline compare a call to @code{strcmp} or @code{strncmp} will
24159 take care of the rest of the comparison. The default is 8 pairs of
24160 loads, which will compare 64 bytes on a 64-bit target and 32 bytes on a
24165 @cindex smaller data references (PowerPC)
24166 @cindex .sdata/.sdata2 references (PowerPC)
24167 On embedded PowerPC systems, put global and static items less than or
24168 equal to @var{num} bytes into the small data or BSS sections instead of
24169 the normal data or BSS section. By default, @var{num} is 8. The
24170 @option{-G @var{num}} switch is also passed to the linker.
24171 All modules should be compiled with the same @option{-G @var{num}} value.
24174 @itemx -mno-regnames
24176 @opindex mno-regnames
24177 On System V.4 and embedded PowerPC systems do (do not) emit register
24178 names in the assembly language output using symbolic forms.
24181 @itemx -mno-longcall
24183 @opindex mno-longcall
24184 By default assume that all calls are far away so that a longer and more
24185 expensive calling sequence is required. This is required for calls
24186 farther than 32 megabytes (33,554,432 bytes) from the current location.
24187 A short call is generated if the compiler knows
24188 the call cannot be that far away. This setting can be overridden by
24189 the @code{shortcall} function attribute, or by @code{#pragma
24192 Some linkers are capable of detecting out-of-range calls and generating
24193 glue code on the fly. On these systems, long calls are unnecessary and
24194 generate slower code. As of this writing, the AIX linker can do this,
24195 as can the GNU linker for PowerPC/64. It is planned to add this feature
24196 to the GNU linker for 32-bit PowerPC systems as well.
24198 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
24199 callee, L42}, plus a @dfn{branch island} (glue code). The two target
24200 addresses represent the callee and the branch island. The
24201 Darwin/PPC linker prefers the first address and generates a @code{bl
24202 callee} if the PPC @code{bl} instruction reaches the callee directly;
24203 otherwise, the linker generates @code{bl L42} to call the branch
24204 island. The branch island is appended to the body of the
24205 calling function; it computes the full 32-bit address of the callee
24208 On Mach-O (Darwin) systems, this option directs the compiler emit to
24209 the glue for every direct call, and the Darwin linker decides whether
24210 to use or discard it.
24212 In the future, GCC may ignore all longcall specifications
24213 when the linker is known to generate glue.
24215 @item -mtls-markers
24216 @itemx -mno-tls-markers
24217 @opindex mtls-markers
24218 @opindex mno-tls-markers
24219 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
24220 specifying the function argument. The relocation allows the linker to
24221 reliably associate function call with argument setup instructions for
24222 TLS optimization, which in turn allows GCC to better schedule the
24228 This option enables use of the reciprocal estimate and
24229 reciprocal square root estimate instructions with additional
24230 Newton-Raphson steps to increase precision instead of doing a divide or
24231 square root and divide for floating-point arguments. You should use
24232 the @option{-ffast-math} option when using @option{-mrecip} (or at
24233 least @option{-funsafe-math-optimizations},
24234 @option{-ffinite-math-only}, @option{-freciprocal-math} and
24235 @option{-fno-trapping-math}). Note that while the throughput of the
24236 sequence is generally higher than the throughput of the non-reciprocal
24237 instruction, the precision of the sequence can be decreased by up to 2
24238 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
24241 @item -mrecip=@var{opt}
24242 @opindex mrecip=opt
24243 This option controls which reciprocal estimate instructions
24244 may be used. @var{opt} is a comma-separated list of options, which may
24245 be preceded by a @code{!} to invert the option:
24250 Enable all estimate instructions.
24253 Enable the default instructions, equivalent to @option{-mrecip}.
24256 Disable all estimate instructions, equivalent to @option{-mno-recip}.
24259 Enable the reciprocal approximation instructions for both
24260 single and double precision.
24263 Enable the single-precision reciprocal approximation instructions.
24266 Enable the double-precision reciprocal approximation instructions.
24269 Enable the reciprocal square root approximation instructions for both
24270 single and double precision.
24273 Enable the single-precision reciprocal square root approximation instructions.
24276 Enable the double-precision reciprocal square root approximation instructions.
24280 So, for example, @option{-mrecip=all,!rsqrtd} enables
24281 all of the reciprocal estimate instructions, except for the
24282 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
24283 which handle the double-precision reciprocal square root calculations.
24285 @item -mrecip-precision
24286 @itemx -mno-recip-precision
24287 @opindex mrecip-precision
24288 Assume (do not assume) that the reciprocal estimate instructions
24289 provide higher-precision estimates than is mandated by the PowerPC
24290 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
24291 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
24292 The double-precision square root estimate instructions are not generated by
24293 default on low-precision machines, since they do not provide an
24294 estimate that converges after three steps.
24296 @item -mveclibabi=@var{type}
24297 @opindex mveclibabi
24298 Specifies the ABI type to use for vectorizing intrinsics using an
24299 external library. The only type supported at present is @samp{mass},
24300 which specifies to use IBM's Mathematical Acceleration Subsystem
24301 (MASS) libraries for vectorizing intrinsics using external libraries.
24302 GCC currently emits calls to @code{acosd2}, @code{acosf4},
24303 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
24304 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
24305 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
24306 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
24307 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
24308 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
24309 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
24310 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
24311 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
24312 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
24313 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
24314 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
24315 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
24316 for power7. Both @option{-ftree-vectorize} and
24317 @option{-funsafe-math-optimizations} must also be enabled. The MASS
24318 libraries must be specified at link time.
24323 Generate (do not generate) the @code{friz} instruction when the
24324 @option{-funsafe-math-optimizations} option is used to optimize
24325 rounding of floating-point values to 64-bit integer and back to floating
24326 point. The @code{friz} instruction does not return the same value if
24327 the floating-point number is too large to fit in an integer.
24329 @item -mpointers-to-nested-functions
24330 @itemx -mno-pointers-to-nested-functions
24331 @opindex mpointers-to-nested-functions
24332 Generate (do not generate) code to load up the static chain register
24333 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
24334 systems where a function pointer points to a 3-word descriptor giving
24335 the function address, TOC value to be loaded in register @code{r2}, and
24336 static chain value to be loaded in register @code{r11}. The
24337 @option{-mpointers-to-nested-functions} is on by default. You cannot
24338 call through pointers to nested functions or pointers
24339 to functions compiled in other languages that use the static chain if
24340 you use @option{-mno-pointers-to-nested-functions}.
24342 @item -msave-toc-indirect
24343 @itemx -mno-save-toc-indirect
24344 @opindex msave-toc-indirect
24345 Generate (do not generate) code to save the TOC value in the reserved
24346 stack location in the function prologue if the function calls through
24347 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
24348 saved in the prologue, it is saved just before the call through the
24349 pointer. The @option{-mno-save-toc-indirect} option is the default.
24351 @item -mcompat-align-parm
24352 @itemx -mno-compat-align-parm
24353 @opindex mcompat-align-parm
24354 Generate (do not generate) code to pass structure parameters with a
24355 maximum alignment of 64 bits, for compatibility with older versions
24358 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
24359 structure parameter on a 128-bit boundary when that structure contained
24360 a member requiring 128-bit alignment. This is corrected in more
24361 recent versions of GCC. This option may be used to generate code
24362 that is compatible with functions compiled with older versions of
24365 The @option{-mno-compat-align-parm} option is the default.
24367 @item -mstack-protector-guard=@var{guard}
24368 @itemx -mstack-protector-guard-reg=@var{reg}
24369 @itemx -mstack-protector-guard-offset=@var{offset}
24370 @itemx -mstack-protector-guard-symbol=@var{symbol}
24371 @opindex mstack-protector-guard
24372 @opindex mstack-protector-guard-reg
24373 @opindex mstack-protector-guard-offset
24374 @opindex mstack-protector-guard-symbol
24375 Generate stack protection code using canary at @var{guard}. Supported
24376 locations are @samp{global} for global canary or @samp{tls} for per-thread
24377 canary in the TLS block (the default with GNU libc version 2.4 or later).
24379 With the latter choice the options
24380 @option{-mstack-protector-guard-reg=@var{reg}} and
24381 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
24382 which register to use as base register for reading the canary, and from what
24383 offset from that base register. The default for those is as specified in the
24384 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
24385 the offset with a symbol reference to a canary in the TLS block.
24389 @subsection RX Options
24392 These command-line options are defined for RX targets:
24395 @item -m64bit-doubles
24396 @itemx -m32bit-doubles
24397 @opindex m64bit-doubles
24398 @opindex m32bit-doubles
24399 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
24400 or 32 bits (@option{-m32bit-doubles}) in size. The default is
24401 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
24402 works on 32-bit values, which is why the default is
24403 @option{-m32bit-doubles}.
24409 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
24410 floating-point hardware. The default is enabled for the RX600
24411 series and disabled for the RX200 series.
24413 Floating-point instructions are only generated for 32-bit floating-point
24414 values, however, so the FPU hardware is not used for doubles if the
24415 @option{-m64bit-doubles} option is used.
24417 @emph{Note} If the @option{-fpu} option is enabled then
24418 @option{-funsafe-math-optimizations} is also enabled automatically.
24419 This is because the RX FPU instructions are themselves unsafe.
24421 @item -mcpu=@var{name}
24423 Selects the type of RX CPU to be targeted. Currently three types are
24424 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
24425 the specific @samp{RX610} CPU. The default is @samp{RX600}.
24427 The only difference between @samp{RX600} and @samp{RX610} is that the
24428 @samp{RX610} does not support the @code{MVTIPL} instruction.
24430 The @samp{RX200} series does not have a hardware floating-point unit
24431 and so @option{-nofpu} is enabled by default when this type is
24434 @item -mbig-endian-data
24435 @itemx -mlittle-endian-data
24436 @opindex mbig-endian-data
24437 @opindex mlittle-endian-data
24438 Store data (but not code) in the big-endian format. The default is
24439 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
24442 @item -msmall-data-limit=@var{N}
24443 @opindex msmall-data-limit
24444 Specifies the maximum size in bytes of global and static variables
24445 which can be placed into the small data area. Using the small data
24446 area can lead to smaller and faster code, but the size of area is
24447 limited and it is up to the programmer to ensure that the area does
24448 not overflow. Also when the small data area is used one of the RX's
24449 registers (usually @code{r13}) is reserved for use pointing to this
24450 area, so it is no longer available for use by the compiler. This
24451 could result in slower and/or larger code if variables are pushed onto
24452 the stack instead of being held in this register.
24454 Note, common variables (variables that have not been initialized) and
24455 constants are not placed into the small data area as they are assigned
24456 to other sections in the output executable.
24458 The default value is zero, which disables this feature. Note, this
24459 feature is not enabled by default with higher optimization levels
24460 (@option{-O2} etc) because of the potentially detrimental effects of
24461 reserving a register. It is up to the programmer to experiment and
24462 discover whether this feature is of benefit to their program. See the
24463 description of the @option{-mpid} option for a description of how the
24464 actual register to hold the small data area pointer is chosen.
24470 Use the simulator runtime. The default is to use the libgloss
24471 board-specific runtime.
24473 @item -mas100-syntax
24474 @itemx -mno-as100-syntax
24475 @opindex mas100-syntax
24476 @opindex mno-as100-syntax
24477 When generating assembler output use a syntax that is compatible with
24478 Renesas's AS100 assembler. This syntax can also be handled by the GAS
24479 assembler, but it has some restrictions so it is not generated by default.
24481 @item -mmax-constant-size=@var{N}
24482 @opindex mmax-constant-size
24483 Specifies the maximum size, in bytes, of a constant that can be used as
24484 an operand in a RX instruction. Although the RX instruction set does
24485 allow constants of up to 4 bytes in length to be used in instructions,
24486 a longer value equates to a longer instruction. Thus in some
24487 circumstances it can be beneficial to restrict the size of constants
24488 that are used in instructions. Constants that are too big are instead
24489 placed into a constant pool and referenced via register indirection.
24491 The value @var{N} can be between 0 and 4. A value of 0 (the default)
24492 or 4 means that constants of any size are allowed.
24496 Enable linker relaxation. Linker relaxation is a process whereby the
24497 linker attempts to reduce the size of a program by finding shorter
24498 versions of various instructions. Disabled by default.
24500 @item -mint-register=@var{N}
24501 @opindex mint-register
24502 Specify the number of registers to reserve for fast interrupt handler
24503 functions. The value @var{N} can be between 0 and 4. A value of 1
24504 means that register @code{r13} is reserved for the exclusive use
24505 of fast interrupt handlers. A value of 2 reserves @code{r13} and
24506 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
24507 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
24508 A value of 0, the default, does not reserve any registers.
24510 @item -msave-acc-in-interrupts
24511 @opindex msave-acc-in-interrupts
24512 Specifies that interrupt handler functions should preserve the
24513 accumulator register. This is only necessary if normal code might use
24514 the accumulator register, for example because it performs 64-bit
24515 multiplications. The default is to ignore the accumulator as this
24516 makes the interrupt handlers faster.
24522 Enables the generation of position independent data. When enabled any
24523 access to constant data is done via an offset from a base address
24524 held in a register. This allows the location of constant data to be
24525 determined at run time without requiring the executable to be
24526 relocated, which is a benefit to embedded applications with tight
24527 memory constraints. Data that can be modified is not affected by this
24530 Note, using this feature reserves a register, usually @code{r13}, for
24531 the constant data base address. This can result in slower and/or
24532 larger code, especially in complicated functions.
24534 The actual register chosen to hold the constant data base address
24535 depends upon whether the @option{-msmall-data-limit} and/or the
24536 @option{-mint-register} command-line options are enabled. Starting
24537 with register @code{r13} and proceeding downwards, registers are
24538 allocated first to satisfy the requirements of @option{-mint-register},
24539 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
24540 is possible for the small data area register to be @code{r8} if both
24541 @option{-mint-register=4} and @option{-mpid} are specified on the
24544 By default this feature is not enabled. The default can be restored
24545 via the @option{-mno-pid} command-line option.
24547 @item -mno-warn-multiple-fast-interrupts
24548 @itemx -mwarn-multiple-fast-interrupts
24549 @opindex mno-warn-multiple-fast-interrupts
24550 @opindex mwarn-multiple-fast-interrupts
24551 Prevents GCC from issuing a warning message if it finds more than one
24552 fast interrupt handler when it is compiling a file. The default is to
24553 issue a warning for each extra fast interrupt handler found, as the RX
24554 only supports one such interrupt.
24556 @item -mallow-string-insns
24557 @itemx -mno-allow-string-insns
24558 @opindex mallow-string-insns
24559 @opindex mno-allow-string-insns
24560 Enables or disables the use of the string manipulation instructions
24561 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
24562 @code{SWHILE} and also the @code{RMPA} instruction. These
24563 instructions may prefetch data, which is not safe to do if accessing
24564 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
24565 for more information).
24567 The default is to allow these instructions, but it is not possible for
24568 GCC to reliably detect all circumstances where a string instruction
24569 might be used to access an I/O register, so their use cannot be
24570 disabled automatically. Instead it is reliant upon the programmer to
24571 use the @option{-mno-allow-string-insns} option if their program
24572 accesses I/O space.
24574 When the instructions are enabled GCC defines the C preprocessor
24575 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
24576 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
24582 Use only (or not only) @code{JSR} instructions to access functions.
24583 This option can be used when code size exceeds the range of @code{BSR}
24584 instructions. Note that @option{-mno-jsr} does not mean to not use
24585 @code{JSR} but instead means that any type of branch may be used.
24588 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
24589 has special significance to the RX port when used with the
24590 @code{interrupt} function attribute. This attribute indicates a
24591 function intended to process fast interrupts. GCC ensures
24592 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
24593 and/or @code{r13} and only provided that the normal use of the
24594 corresponding registers have been restricted via the
24595 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
24598 @node S/390 and zSeries Options
24599 @subsection S/390 and zSeries Options
24600 @cindex S/390 and zSeries Options
24602 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
24606 @itemx -msoft-float
24607 @opindex mhard-float
24608 @opindex msoft-float
24609 Use (do not use) the hardware floating-point instructions and registers
24610 for floating-point operations. When @option{-msoft-float} is specified,
24611 functions in @file{libgcc.a} are used to perform floating-point
24612 operations. When @option{-mhard-float} is specified, the compiler
24613 generates IEEE floating-point instructions. This is the default.
24616 @itemx -mno-hard-dfp
24618 @opindex mno-hard-dfp
24619 Use (do not use) the hardware decimal-floating-point instructions for
24620 decimal-floating-point operations. When @option{-mno-hard-dfp} is
24621 specified, functions in @file{libgcc.a} are used to perform
24622 decimal-floating-point operations. When @option{-mhard-dfp} is
24623 specified, the compiler generates decimal-floating-point hardware
24624 instructions. This is the default for @option{-march=z9-ec} or higher.
24626 @item -mlong-double-64
24627 @itemx -mlong-double-128
24628 @opindex mlong-double-64
24629 @opindex mlong-double-128
24630 These switches control the size of @code{long double} type. A size
24631 of 64 bits makes the @code{long double} type equivalent to the @code{double}
24632 type. This is the default.
24635 @itemx -mno-backchain
24636 @opindex mbackchain
24637 @opindex mno-backchain
24638 Store (do not store) the address of the caller's frame as backchain pointer
24639 into the callee's stack frame.
24640 A backchain may be needed to allow debugging using tools that do not understand
24641 DWARF call frame information.
24642 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
24643 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
24644 the backchain is placed into the topmost word of the 96/160 byte register
24647 In general, code compiled with @option{-mbackchain} is call-compatible with
24648 code compiled with @option{-mmo-backchain}; however, use of the backchain
24649 for debugging purposes usually requires that the whole binary is built with
24650 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
24651 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
24652 to build a linux kernel use @option{-msoft-float}.
24654 The default is to not maintain the backchain.
24656 @item -mpacked-stack
24657 @itemx -mno-packed-stack
24658 @opindex mpacked-stack
24659 @opindex mno-packed-stack
24660 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
24661 specified, the compiler uses the all fields of the 96/160 byte register save
24662 area only for their default purpose; unused fields still take up stack space.
24663 When @option{-mpacked-stack} is specified, register save slots are densely
24664 packed at the top of the register save area; unused space is reused for other
24665 purposes, allowing for more efficient use of the available stack space.
24666 However, when @option{-mbackchain} is also in effect, the topmost word of
24667 the save area is always used to store the backchain, and the return address
24668 register is always saved two words below the backchain.
24670 As long as the stack frame backchain is not used, code generated with
24671 @option{-mpacked-stack} is call-compatible with code generated with
24672 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
24673 S/390 or zSeries generated code that uses the stack frame backchain at run
24674 time, not just for debugging purposes. Such code is not call-compatible
24675 with code compiled with @option{-mpacked-stack}. Also, note that the
24676 combination of @option{-mbackchain},
24677 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
24678 to build a linux kernel use @option{-msoft-float}.
24680 The default is to not use the packed stack layout.
24683 @itemx -mno-small-exec
24684 @opindex msmall-exec
24685 @opindex mno-small-exec
24686 Generate (or do not generate) code using the @code{bras} instruction
24687 to do subroutine calls.
24688 This only works reliably if the total executable size does not
24689 exceed 64k. The default is to use the @code{basr} instruction instead,
24690 which does not have this limitation.
24696 When @option{-m31} is specified, generate code compliant to the
24697 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
24698 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
24699 particular to generate 64-bit instructions. For the @samp{s390}
24700 targets, the default is @option{-m31}, while the @samp{s390x}
24701 targets default to @option{-m64}.
24707 When @option{-mzarch} is specified, generate code using the
24708 instructions available on z/Architecture.
24709 When @option{-mesa} is specified, generate code using the
24710 instructions available on ESA/390. Note that @option{-mesa} is
24711 not possible with @option{-m64}.
24712 When generating code compliant to the GNU/Linux for S/390 ABI,
24713 the default is @option{-mesa}. When generating code compliant
24714 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
24720 The @option{-mhtm} option enables a set of builtins making use of
24721 instructions available with the transactional execution facility
24722 introduced with the IBM zEnterprise EC12 machine generation
24723 @ref{S/390 System z Built-in Functions}.
24724 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
24730 When @option{-mvx} is specified, generate code using the instructions
24731 available with the vector extension facility introduced with the IBM
24732 z13 machine generation.
24733 This option changes the ABI for some vector type values with regard to
24734 alignment and calling conventions. In case vector type values are
24735 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
24736 command will be added to mark the resulting binary with the ABI used.
24737 @option{-mvx} is enabled by default when using @option{-march=z13}.
24740 @itemx -mno-zvector
24742 @opindex mno-zvector
24743 The @option{-mzvector} option enables vector language extensions and
24744 builtins using instructions available with the vector extension
24745 facility introduced with the IBM z13 machine generation.
24746 This option adds support for @samp{vector} to be used as a keyword to
24747 define vector type variables and arguments. @samp{vector} is only
24748 available when GNU extensions are enabled. It will not be expanded
24749 when requesting strict standard compliance e.g. with @option{-std=c99}.
24750 In addition to the GCC low-level builtins @option{-mzvector} enables
24751 a set of builtins added for compatibility with AltiVec-style
24752 implementations like Power and Cell. In order to make use of these
24753 builtins the header file @file{vecintrin.h} needs to be included.
24754 @option{-mzvector} is disabled by default.
24760 Generate (or do not generate) code using the @code{mvcle} instruction
24761 to perform block moves. When @option{-mno-mvcle} is specified,
24762 use a @code{mvc} loop instead. This is the default unless optimizing for
24769 Print (or do not print) additional debug information when compiling.
24770 The default is to not print debug information.
24772 @item -march=@var{cpu-type}
24774 Generate code that runs on @var{cpu-type}, which is the name of a
24775 system representing a certain processor type. Possible values for
24776 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
24777 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
24778 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
24781 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
24782 @samp{g6} are deprecated and will be removed with future releases.
24784 Specifying @samp{native} as cpu type can be used to select the best
24785 architecture option for the host processor.
24786 @option{-march=native} has no effect if GCC does not recognize the
24789 @item -mtune=@var{cpu-type}
24791 Tune to @var{cpu-type} everything applicable about the generated code,
24792 except for the ABI and the set of available instructions.
24793 The list of @var{cpu-type} values is the same as for @option{-march}.
24794 The default is the value used for @option{-march}.
24797 @itemx -mno-tpf-trace
24798 @opindex mtpf-trace
24799 @opindex mno-tpf-trace
24800 Generate code that adds (does not add) in TPF OS specific branches to trace
24801 routines in the operating system. This option is off by default, even
24802 when compiling for the TPF OS@.
24805 @itemx -mno-fused-madd
24806 @opindex mfused-madd
24807 @opindex mno-fused-madd
24808 Generate code that uses (does not use) the floating-point multiply and
24809 accumulate instructions. These instructions are generated by default if
24810 hardware floating point is used.
24812 @item -mwarn-framesize=@var{framesize}
24813 @opindex mwarn-framesize
24814 Emit a warning if the current function exceeds the given frame size. Because
24815 this is a compile-time check it doesn't need to be a real problem when the program
24816 runs. It is intended to identify functions that most probably cause
24817 a stack overflow. It is useful to be used in an environment with limited stack
24818 size e.g.@: the linux kernel.
24820 @item -mwarn-dynamicstack
24821 @opindex mwarn-dynamicstack
24822 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
24823 arrays. This is generally a bad idea with a limited stack size.
24825 @item -mstack-guard=@var{stack-guard}
24826 @itemx -mstack-size=@var{stack-size}
24827 @opindex mstack-guard
24828 @opindex mstack-size
24829 If these options are provided the S/390 back end emits additional instructions in
24830 the function prologue that trigger a trap if the stack size is @var{stack-guard}
24831 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
24832 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
24833 the frame size of the compiled function is chosen.
24834 These options are intended to be used to help debugging stack overflow problems.
24835 The additionally emitted code causes only little overhead and hence can also be
24836 used in production-like systems without greater performance degradation. The given
24837 values have to be exact powers of 2 and @var{stack-size} has to be greater than
24838 @var{stack-guard} without exceeding 64k.
24839 In order to be efficient the extra code makes the assumption that the stack starts
24840 at an address aligned to the value given by @var{stack-size}.
24841 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
24843 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
24845 If the hotpatch option is enabled, a ``hot-patching'' function
24846 prologue is generated for all functions in the compilation unit.
24847 The funtion label is prepended with the given number of two-byte
24848 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
24849 the label, 2 * @var{post-halfwords} bytes are appended, using the
24850 largest NOP like instructions the architecture allows (maximum
24853 If both arguments are zero, hotpatching is disabled.
24855 This option can be overridden for individual functions with the
24856 @code{hotpatch} attribute.
24859 @node Score Options
24860 @subsection Score Options
24861 @cindex Score Options
24863 These options are defined for Score implementations:
24868 Compile code for big-endian mode. This is the default.
24872 Compile code for little-endian mode.
24876 Disable generation of @code{bcnz} instructions.
24880 Enable generation of unaligned load and store instructions.
24884 Enable the use of multiply-accumulate instructions. Disabled by default.
24888 Specify the SCORE5 as the target architecture.
24892 Specify the SCORE5U of the target architecture.
24896 Specify the SCORE7 as the target architecture. This is the default.
24900 Specify the SCORE7D as the target architecture.
24904 @subsection SH Options
24906 These @samp{-m} options are defined for the SH implementations:
24911 Generate code for the SH1.
24915 Generate code for the SH2.
24918 Generate code for the SH2e.
24922 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
24923 that the floating-point unit is not used.
24925 @item -m2a-single-only
24926 @opindex m2a-single-only
24927 Generate code for the SH2a-FPU, in such a way that no double-precision
24928 floating-point operations are used.
24931 @opindex m2a-single
24932 Generate code for the SH2a-FPU assuming the floating-point unit is in
24933 single-precision mode by default.
24937 Generate code for the SH2a-FPU assuming the floating-point unit is in
24938 double-precision mode by default.
24942 Generate code for the SH3.
24946 Generate code for the SH3e.
24950 Generate code for the SH4 without a floating-point unit.
24952 @item -m4-single-only
24953 @opindex m4-single-only
24954 Generate code for the SH4 with a floating-point unit that only
24955 supports single-precision arithmetic.
24959 Generate code for the SH4 assuming the floating-point unit is in
24960 single-precision mode by default.
24964 Generate code for the SH4.
24968 Generate code for SH4-100.
24970 @item -m4-100-nofpu
24971 @opindex m4-100-nofpu
24972 Generate code for SH4-100 in such a way that the
24973 floating-point unit is not used.
24975 @item -m4-100-single
24976 @opindex m4-100-single
24977 Generate code for SH4-100 assuming the floating-point unit is in
24978 single-precision mode by default.
24980 @item -m4-100-single-only
24981 @opindex m4-100-single-only
24982 Generate code for SH4-100 in such a way that no double-precision
24983 floating-point operations are used.
24987 Generate code for SH4-200.
24989 @item -m4-200-nofpu
24990 @opindex m4-200-nofpu
24991 Generate code for SH4-200 without in such a way that the
24992 floating-point unit is not used.
24994 @item -m4-200-single
24995 @opindex m4-200-single
24996 Generate code for SH4-200 assuming the floating-point unit is in
24997 single-precision mode by default.
24999 @item -m4-200-single-only
25000 @opindex m4-200-single-only
25001 Generate code for SH4-200 in such a way that no double-precision
25002 floating-point operations are used.
25006 Generate code for SH4-300.
25008 @item -m4-300-nofpu
25009 @opindex m4-300-nofpu
25010 Generate code for SH4-300 without in such a way that the
25011 floating-point unit is not used.
25013 @item -m4-300-single
25014 @opindex m4-300-single
25015 Generate code for SH4-300 in such a way that no double-precision
25016 floating-point operations are used.
25018 @item -m4-300-single-only
25019 @opindex m4-300-single-only
25020 Generate code for SH4-300 in such a way that no double-precision
25021 floating-point operations are used.
25025 Generate code for SH4-340 (no MMU, no FPU).
25029 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
25034 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
25035 floating-point unit is not used.
25037 @item -m4a-single-only
25038 @opindex m4a-single-only
25039 Generate code for the SH4a, in such a way that no double-precision
25040 floating-point operations are used.
25043 @opindex m4a-single
25044 Generate code for the SH4a assuming the floating-point unit is in
25045 single-precision mode by default.
25049 Generate code for the SH4a.
25053 Same as @option{-m4a-nofpu}, except that it implicitly passes
25054 @option{-dsp} to the assembler. GCC doesn't generate any DSP
25055 instructions at the moment.
25059 Compile code for the processor in big-endian mode.
25063 Compile code for the processor in little-endian mode.
25067 Align doubles at 64-bit boundaries. Note that this changes the calling
25068 conventions, and thus some functions from the standard C library do
25069 not work unless you recompile it first with @option{-mdalign}.
25073 Shorten some address references at link time, when possible; uses the
25074 linker option @option{-relax}.
25078 Use 32-bit offsets in @code{switch} tables. The default is to use
25083 Enable the use of bit manipulation instructions on SH2A.
25087 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
25088 alignment constraints.
25092 Comply with the calling conventions defined by Renesas.
25095 @opindex mno-renesas
25096 Comply with the calling conventions defined for GCC before the Renesas
25097 conventions were available. This option is the default for all
25098 targets of the SH toolchain.
25101 @opindex mnomacsave
25102 Mark the @code{MAC} register as call-clobbered, even if
25103 @option{-mrenesas} is given.
25109 Control the IEEE compliance of floating-point comparisons, which affects the
25110 handling of cases where the result of a comparison is unordered. By default
25111 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
25112 enabled @option{-mno-ieee} is implicitly set, which results in faster
25113 floating-point greater-equal and less-equal comparisons. The implicit settings
25114 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
25116 @item -minline-ic_invalidate
25117 @opindex minline-ic_invalidate
25118 Inline code to invalidate instruction cache entries after setting up
25119 nested function trampolines.
25120 This option has no effect if @option{-musermode} is in effect and the selected
25121 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
25123 If the selected code generation option does not allow the use of the @code{icbi}
25124 instruction, and @option{-musermode} is not in effect, the inlined code
25125 manipulates the instruction cache address array directly with an associative
25126 write. This not only requires privileged mode at run time, but it also
25127 fails if the cache line had been mapped via the TLB and has become unmapped.
25131 Dump instruction size and location in the assembly code.
25134 @opindex mpadstruct
25135 This option is deprecated. It pads structures to multiple of 4 bytes,
25136 which is incompatible with the SH ABI@.
25138 @item -matomic-model=@var{model}
25139 @opindex matomic-model=@var{model}
25140 Sets the model of atomic operations and additional parameters as a comma
25141 separated list. For details on the atomic built-in functions see
25142 @ref{__atomic Builtins}. The following models and parameters are supported:
25147 Disable compiler generated atomic sequences and emit library calls for atomic
25148 operations. This is the default if the target is not @code{sh*-*-linux*}.
25151 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
25152 built-in functions. The generated atomic sequences require additional support
25153 from the interrupt/exception handling code of the system and are only suitable
25154 for SH3* and SH4* single-core systems. This option is enabled by default when
25155 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
25156 this option also partially utilizes the hardware atomic instructions
25157 @code{movli.l} and @code{movco.l} to create more efficient code, unless
25158 @samp{strict} is specified.
25161 Generate software atomic sequences that use a variable in the thread control
25162 block. This is a variation of the gUSA sequences which can also be used on
25163 SH1* and SH2* targets. The generated atomic sequences require additional
25164 support from the interrupt/exception handling code of the system and are only
25165 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
25166 parameter has to be specified as well.
25169 Generate software atomic sequences that temporarily disable interrupts by
25170 setting @code{SR.IMASK = 1111}. This model works only when the program runs
25171 in privileged mode and is only suitable for single-core systems. Additional
25172 support from the interrupt/exception handling code of the system is not
25173 required. This model is enabled by default when the target is
25174 @code{sh*-*-linux*} and SH1* or SH2*.
25177 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
25178 instructions only. This is only available on SH4A and is suitable for
25179 multi-core systems. Since the hardware instructions support only 32 bit atomic
25180 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
25181 Code compiled with this option is also compatible with other software
25182 atomic model interrupt/exception handling systems if executed on an SH4A
25183 system. Additional support from the interrupt/exception handling code of the
25184 system is not required for this model.
25187 This parameter specifies the offset in bytes of the variable in the thread
25188 control block structure that should be used by the generated atomic sequences
25189 when the @samp{soft-tcb} model has been selected. For other models this
25190 parameter is ignored. The specified value must be an integer multiple of four
25191 and in the range 0-1020.
25194 This parameter prevents mixed usage of multiple atomic models, even if they
25195 are compatible, and makes the compiler generate atomic sequences of the
25196 specified model only.
25202 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
25203 Notice that depending on the particular hardware and software configuration
25204 this can degrade overall performance due to the operand cache line flushes
25205 that are implied by the @code{tas.b} instruction. On multi-core SH4A
25206 processors the @code{tas.b} instruction must be used with caution since it
25207 can result in data corruption for certain cache configurations.
25210 @opindex mprefergot
25211 When generating position-independent code, emit function calls using
25212 the Global Offset Table instead of the Procedure Linkage Table.
25215 @itemx -mno-usermode
25217 @opindex mno-usermode
25218 Don't allow (allow) the compiler generating privileged mode code. Specifying
25219 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
25220 inlined code would not work in user mode. @option{-musermode} is the default
25221 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
25222 @option{-musermode} has no effect, since there is no user mode.
25224 @item -multcost=@var{number}
25225 @opindex multcost=@var{number}
25226 Set the cost to assume for a multiply insn.
25228 @item -mdiv=@var{strategy}
25229 @opindex mdiv=@var{strategy}
25230 Set the division strategy to be used for integer division operations.
25231 @var{strategy} can be one of:
25236 Calls a library function that uses the single-step division instruction
25237 @code{div1} to perform the operation. Division by zero calculates an
25238 unspecified result and does not trap. This is the default except for SH4,
25239 SH2A and SHcompact.
25242 Calls a library function that performs the operation in double precision
25243 floating point. Division by zero causes a floating-point exception. This is
25244 the default for SHcompact with FPU. Specifying this for targets that do not
25245 have a double precision FPU defaults to @code{call-div1}.
25248 Calls a library function that uses a lookup table for small divisors and
25249 the @code{div1} instruction with case distinction for larger divisors. Division
25250 by zero calculates an unspecified result and does not trap. This is the default
25251 for SH4. Specifying this for targets that do not have dynamic shift
25252 instructions defaults to @code{call-div1}.
25256 When a division strategy has not been specified the default strategy is
25257 selected based on the current target. For SH2A the default strategy is to
25258 use the @code{divs} and @code{divu} instructions instead of library function
25261 @item -maccumulate-outgoing-args
25262 @opindex maccumulate-outgoing-args
25263 Reserve space once for outgoing arguments in the function prologue rather
25264 than around each call. Generally beneficial for performance and size. Also
25265 needed for unwinding to avoid changing the stack frame around conditional code.
25267 @item -mdivsi3_libfunc=@var{name}
25268 @opindex mdivsi3_libfunc=@var{name}
25269 Set the name of the library function used for 32-bit signed division to
25271 This only affects the name used in the @samp{call} division strategies, and
25272 the compiler still expects the same sets of input/output/clobbered registers as
25273 if this option were not present.
25275 @item -mfixed-range=@var{register-range}
25276 @opindex mfixed-range
25277 Generate code treating the given register range as fixed registers.
25278 A fixed register is one that the register allocator can not use. This is
25279 useful when compiling kernel code. A register range is specified as
25280 two registers separated by a dash. Multiple register ranges can be
25281 specified separated by a comma.
25283 @item -mbranch-cost=@var{num}
25284 @opindex mbranch-cost=@var{num}
25285 Assume @var{num} to be the cost for a branch instruction. Higher numbers
25286 make the compiler try to generate more branch-free code if possible.
25287 If not specified the value is selected depending on the processor type that
25288 is being compiled for.
25291 @itemx -mno-zdcbranch
25292 @opindex mzdcbranch
25293 @opindex mno-zdcbranch
25294 Assume (do not assume) that zero displacement conditional branch instructions
25295 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
25296 compiler prefers zero displacement branch code sequences. This is
25297 enabled by default when generating code for SH4 and SH4A. It can be explicitly
25298 disabled by specifying @option{-mno-zdcbranch}.
25300 @item -mcbranch-force-delay-slot
25301 @opindex mcbranch-force-delay-slot
25302 Force the usage of delay slots for conditional branches, which stuffs the delay
25303 slot with a @code{nop} if a suitable instruction cannot be found. By default
25304 this option is disabled. It can be enabled to work around hardware bugs as
25305 found in the original SH7055.
25308 @itemx -mno-fused-madd
25309 @opindex mfused-madd
25310 @opindex mno-fused-madd
25311 Generate code that uses (does not use) the floating-point multiply and
25312 accumulate instructions. These instructions are generated by default
25313 if hardware floating point is used. The machine-dependent
25314 @option{-mfused-madd} option is now mapped to the machine-independent
25315 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
25316 mapped to @option{-ffp-contract=off}.
25322 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
25323 and cosine approximations. The option @option{-mfsca} must be used in
25324 combination with @option{-funsafe-math-optimizations}. It is enabled by default
25325 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
25326 approximations even if @option{-funsafe-math-optimizations} is in effect.
25332 Allow or disallow the compiler to emit the @code{fsrra} instruction for
25333 reciprocal square root approximations. The option @option{-mfsrra} must be used
25334 in combination with @option{-funsafe-math-optimizations} and
25335 @option{-ffinite-math-only}. It is enabled by default when generating code for
25336 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
25337 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
25340 @item -mpretend-cmove
25341 @opindex mpretend-cmove
25342 Prefer zero-displacement conditional branches for conditional move instruction
25343 patterns. This can result in faster code on the SH4 processor.
25347 Generate code using the FDPIC ABI.
25351 @node Solaris 2 Options
25352 @subsection Solaris 2 Options
25353 @cindex Solaris 2 options
25355 These @samp{-m} options are supported on Solaris 2:
25358 @item -mclear-hwcap
25359 @opindex mclear-hwcap
25360 @option{-mclear-hwcap} tells the compiler to remove the hardware
25361 capabilities generated by the Solaris assembler. This is only necessary
25362 when object files use ISA extensions not supported by the current
25363 machine, but check at runtime whether or not to use them.
25365 @item -mimpure-text
25366 @opindex mimpure-text
25367 @option{-mimpure-text}, used in addition to @option{-shared}, tells
25368 the compiler to not pass @option{-z text} to the linker when linking a
25369 shared object. Using this option, you can link position-dependent
25370 code into a shared object.
25372 @option{-mimpure-text} suppresses the ``relocations remain against
25373 allocatable but non-writable sections'' linker error message.
25374 However, the necessary relocations trigger copy-on-write, and the
25375 shared object is not actually shared across processes. Instead of
25376 using @option{-mimpure-text}, you should compile all source code with
25377 @option{-fpic} or @option{-fPIC}.
25381 These switches are supported in addition to the above on Solaris 2:
25386 This is a synonym for @option{-pthread}.
25389 @node SPARC Options
25390 @subsection SPARC Options
25391 @cindex SPARC options
25393 These @samp{-m} options are supported on the SPARC:
25396 @item -mno-app-regs
25398 @opindex mno-app-regs
25400 Specify @option{-mapp-regs} to generate output using the global registers
25401 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
25402 global register 1, each global register 2 through 4 is then treated as an
25403 allocable register that is clobbered by function calls. This is the default.
25405 To be fully SVR4 ABI-compliant at the cost of some performance loss,
25406 specify @option{-mno-app-regs}. You should compile libraries and system
25407 software with this option.
25413 With @option{-mflat}, the compiler does not generate save/restore instructions
25414 and uses a ``flat'' or single register window model. This model is compatible
25415 with the regular register window model. The local registers and the input
25416 registers (0--5) are still treated as ``call-saved'' registers and are
25417 saved on the stack as needed.
25419 With @option{-mno-flat} (the default), the compiler generates save/restore
25420 instructions (except for leaf functions). This is the normal operating mode.
25423 @itemx -mhard-float
25425 @opindex mhard-float
25426 Generate output containing floating-point instructions. This is the
25430 @itemx -msoft-float
25432 @opindex msoft-float
25433 Generate output containing library calls for floating point.
25434 @strong{Warning:} the requisite libraries are not available for all SPARC
25435 targets. Normally the facilities of the machine's usual C compiler are
25436 used, but this cannot be done directly in cross-compilation. You must make
25437 your own arrangements to provide suitable library functions for
25438 cross-compilation. The embedded targets @samp{sparc-*-aout} and
25439 @samp{sparclite-*-*} do provide software floating-point support.
25441 @option{-msoft-float} changes the calling convention in the output file;
25442 therefore, it is only useful if you compile @emph{all} of a program with
25443 this option. In particular, you need to compile @file{libgcc.a}, the
25444 library that comes with GCC, with @option{-msoft-float} in order for
25447 @item -mhard-quad-float
25448 @opindex mhard-quad-float
25449 Generate output containing quad-word (long double) floating-point
25452 @item -msoft-quad-float
25453 @opindex msoft-quad-float
25454 Generate output containing library calls for quad-word (long double)
25455 floating-point instructions. The functions called are those specified
25456 in the SPARC ABI@. This is the default.
25458 As of this writing, there are no SPARC implementations that have hardware
25459 support for the quad-word floating-point instructions. They all invoke
25460 a trap handler for one of these instructions, and then the trap handler
25461 emulates the effect of the instruction. Because of the trap handler overhead,
25462 this is much slower than calling the ABI library routines. Thus the
25463 @option{-msoft-quad-float} option is the default.
25465 @item -mno-unaligned-doubles
25466 @itemx -munaligned-doubles
25467 @opindex mno-unaligned-doubles
25468 @opindex munaligned-doubles
25469 Assume that doubles have 8-byte alignment. This is the default.
25471 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
25472 alignment only if they are contained in another type, or if they have an
25473 absolute address. Otherwise, it assumes they have 4-byte alignment.
25474 Specifying this option avoids some rare compatibility problems with code
25475 generated by other compilers. It is not the default because it results
25476 in a performance loss, especially for floating-point code.
25479 @itemx -mno-user-mode
25480 @opindex muser-mode
25481 @opindex mno-user-mode
25482 Do not generate code that can only run in supervisor mode. This is relevant
25483 only for the @code{casa} instruction emitted for the LEON3 processor. This
25486 @item -mfaster-structs
25487 @itemx -mno-faster-structs
25488 @opindex mfaster-structs
25489 @opindex mno-faster-structs
25490 With @option{-mfaster-structs}, the compiler assumes that structures
25491 should have 8-byte alignment. This enables the use of pairs of
25492 @code{ldd} and @code{std} instructions for copies in structure
25493 assignment, in place of twice as many @code{ld} and @code{st} pairs.
25494 However, the use of this changed alignment directly violates the SPARC
25495 ABI@. Thus, it's intended only for use on targets where the developer
25496 acknowledges that their resulting code is not directly in line with
25497 the rules of the ABI@.
25499 @item -mstd-struct-return
25500 @itemx -mno-std-struct-return
25501 @opindex mstd-struct-return
25502 @opindex mno-std-struct-return
25503 With @option{-mstd-struct-return}, the compiler generates checking code
25504 in functions returning structures or unions to detect size mismatches
25505 between the two sides of function calls, as per the 32-bit ABI@.
25507 The default is @option{-mno-std-struct-return}. This option has no effect
25514 Enable Local Register Allocation. This is the default for SPARC since GCC 7
25515 so @option{-mno-lra} needs to be passed to get old Reload.
25517 @item -mcpu=@var{cpu_type}
25519 Set the instruction set, register set, and instruction scheduling parameters
25520 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
25521 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
25522 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
25523 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
25524 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
25525 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
25527 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
25528 which selects the best architecture option for the host processor.
25529 @option{-mcpu=native} has no effect if GCC does not recognize
25532 Default instruction scheduling parameters are used for values that select
25533 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
25534 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
25536 Here is a list of each supported architecture and their supported
25544 supersparc, hypersparc, leon, leon3
25547 f930, f934, sparclite86x
25553 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
25557 By default (unless configured otherwise), GCC generates code for the V7
25558 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
25559 additionally optimizes it for the Cypress CY7C602 chip, as used in the
25560 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
25561 SPARCStation 1, 2, IPX etc.
25563 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
25564 architecture. The only difference from V7 code is that the compiler emits
25565 the integer multiply and integer divide instructions which exist in SPARC-V8
25566 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
25567 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
25570 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
25571 the SPARC architecture. This adds the integer multiply, integer divide step
25572 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
25573 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
25574 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
25575 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
25576 MB86934 chip, which is the more recent SPARClite with FPU@.
25578 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
25579 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
25580 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
25581 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
25582 optimizes it for the TEMIC SPARClet chip.
25584 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
25585 architecture. This adds 64-bit integer and floating-point move instructions,
25586 3 additional floating-point condition code registers and conditional move
25587 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
25588 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
25589 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
25590 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
25591 @option{-mcpu=niagara}, the compiler additionally optimizes it for
25592 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
25593 additionally optimizes it for Sun UltraSPARC T2 chips. With
25594 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
25595 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
25596 additionally optimizes it for Sun UltraSPARC T4 chips. With
25597 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
25598 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
25599 additionally optimizes it for Oracle M8 chips.
25601 @item -mtune=@var{cpu_type}
25603 Set the instruction scheduling parameters for machine type
25604 @var{cpu_type}, but do not set the instruction set or register set that the
25605 option @option{-mcpu=@var{cpu_type}} does.
25607 The same values for @option{-mcpu=@var{cpu_type}} can be used for
25608 @option{-mtune=@var{cpu_type}}, but the only useful values are those
25609 that select a particular CPU implementation. Those are
25610 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
25611 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
25612 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
25613 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
25614 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
25615 and GNU/Linux toolchains, @samp{native} can also be used.
25620 @opindex mno-v8plus
25621 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
25622 difference from the V8 ABI is that the global and out registers are
25623 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
25624 mode for all SPARC-V9 processors.
25630 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
25631 Visual Instruction Set extensions. The default is @option{-mno-vis}.
25637 With @option{-mvis2}, GCC generates code that takes advantage of
25638 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
25639 default is @option{-mvis2} when targeting a cpu that supports such
25640 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
25641 also sets @option{-mvis}.
25647 With @option{-mvis3}, GCC generates code that takes advantage of
25648 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
25649 default is @option{-mvis3} when targeting a cpu that supports such
25650 instructions, such as niagara-3 and later. Setting @option{-mvis3}
25651 also sets @option{-mvis2} and @option{-mvis}.
25657 With @option{-mvis4}, GCC generates code that takes advantage of
25658 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
25659 default is @option{-mvis4} when targeting a cpu that supports such
25660 instructions, such as niagara-7 and later. Setting @option{-mvis4}
25661 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
25667 With @option{-mvis4b}, GCC generates code that takes advantage of
25668 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
25669 the additional VIS instructions introduced in the Oracle SPARC
25670 Architecture 2017. The default is @option{-mvis4b} when targeting a
25671 cpu that supports such instructions, such as m8 and later. Setting
25672 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
25673 @option{-mvis2} and @option{-mvis}.
25678 @opindex mno-cbcond
25679 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
25680 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
25681 when targeting a CPU that supports such instructions, such as Niagara-4 and
25688 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
25689 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
25690 when targeting a CPU that supports such instructions, such as Niagara-3 and
25696 @opindex mno-fsmuld
25697 With @option{-mfsmuld}, GCC generates code that takes advantage of the
25698 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
25699 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
25700 or V9 with FPU except @option{-mcpu=leon}.
25706 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
25707 Population Count instruction. The default is @option{-mpopc}
25708 when targeting a CPU that supports such an instruction, such as Niagara-2 and
25715 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
25716 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
25717 when targeting a CPU that supports such an instruction, such as Niagara-7 and
25721 @opindex mfix-at697f
25722 Enable the documented workaround for the single erratum of the Atmel AT697F
25723 processor (which corresponds to erratum #13 of the AT697E processor).
25726 @opindex mfix-ut699
25727 Enable the documented workarounds for the floating-point errata and the data
25728 cache nullify errata of the UT699 processor.
25731 @opindex mfix-ut700
25732 Enable the documented workaround for the back-to-back store errata of
25733 the UT699E/UT700 processor.
25735 @item -mfix-gr712rc
25736 @opindex mfix-gr712rc
25737 Enable the documented workaround for the back-to-back store errata of
25738 the GR712RC processor.
25741 These @samp{-m} options are supported in addition to the above
25742 on SPARC-V9 processors in 64-bit environments:
25749 Generate code for a 32-bit or 64-bit environment.
25750 The 32-bit environment sets int, long and pointer to 32 bits.
25751 The 64-bit environment sets int to 32 bits and long and pointer
25754 @item -mcmodel=@var{which}
25756 Set the code model to one of
25760 The Medium/Low code model: 64-bit addresses, programs
25761 must be linked in the low 32 bits of memory. Programs can be statically
25762 or dynamically linked.
25765 The Medium/Middle code model: 64-bit addresses, programs
25766 must be linked in the low 44 bits of memory, the text and data segments must
25767 be less than 2GB in size and the data segment must be located within 2GB of
25771 The Medium/Anywhere code model: 64-bit addresses, programs
25772 may be linked anywhere in memory, the text and data segments must be less
25773 than 2GB in size and the data segment must be located within 2GB of the
25777 The Medium/Anywhere code model for embedded systems:
25778 64-bit addresses, the text and data segments must be less than 2GB in
25779 size, both starting anywhere in memory (determined at link time). The
25780 global register %g4 points to the base of the data segment. Programs
25781 are statically linked and PIC is not supported.
25784 @item -mmemory-model=@var{mem-model}
25785 @opindex mmemory-model
25786 Set the memory model in force on the processor to one of
25790 The default memory model for the processor and operating system.
25793 Relaxed Memory Order
25796 Partial Store Order
25802 Sequential Consistency
25805 These memory models are formally defined in Appendix D of the SPARC-V9
25806 architecture manual, as set in the processor's @code{PSTATE.MM} field.
25809 @itemx -mno-stack-bias
25810 @opindex mstack-bias
25811 @opindex mno-stack-bias
25812 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
25813 frame pointer if present, are offset by @minus{}2047 which must be added back
25814 when making stack frame references. This is the default in 64-bit mode.
25815 Otherwise, assume no such offset is present.
25819 @subsection SPU Options
25820 @cindex SPU options
25822 These @samp{-m} options are supported on the SPU:
25826 @itemx -merror-reloc
25827 @opindex mwarn-reloc
25828 @opindex merror-reloc
25830 The loader for SPU does not handle dynamic relocations. By default, GCC
25831 gives an error when it generates code that requires a dynamic
25832 relocation. @option{-mno-error-reloc} disables the error,
25833 @option{-mwarn-reloc} generates a warning instead.
25836 @itemx -munsafe-dma
25838 @opindex munsafe-dma
25840 Instructions that initiate or test completion of DMA must not be
25841 reordered with respect to loads and stores of the memory that is being
25843 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
25844 memory accesses, but that can lead to inefficient code in places where the
25845 memory is known to not change. Rather than mark the memory as volatile,
25846 you can use @option{-msafe-dma} to tell the compiler to treat
25847 the DMA instructions as potentially affecting all memory.
25849 @item -mbranch-hints
25850 @opindex mbranch-hints
25852 By default, GCC generates a branch hint instruction to avoid
25853 pipeline stalls for always-taken or probably-taken branches. A hint
25854 is not generated closer than 8 instructions away from its branch.
25855 There is little reason to disable them, except for debugging purposes,
25856 or to make an object a little bit smaller.
25860 @opindex msmall-mem
25861 @opindex mlarge-mem
25863 By default, GCC generates code assuming that addresses are never larger
25864 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
25865 a full 32-bit address.
25870 By default, GCC links against startup code that assumes the SPU-style
25871 main function interface (which has an unconventional parameter list).
25872 With @option{-mstdmain}, GCC links your program against startup
25873 code that assumes a C99-style interface to @code{main}, including a
25874 local copy of @code{argv} strings.
25876 @item -mfixed-range=@var{register-range}
25877 @opindex mfixed-range
25878 Generate code treating the given register range as fixed registers.
25879 A fixed register is one that the register allocator cannot use. This is
25880 useful when compiling kernel code. A register range is specified as
25881 two registers separated by a dash. Multiple register ranges can be
25882 specified separated by a comma.
25888 Compile code assuming that pointers to the PPU address space accessed
25889 via the @code{__ea} named address space qualifier are either 32 or 64
25890 bits wide. The default is 32 bits. As this is an ABI-changing option,
25891 all object code in an executable must be compiled with the same setting.
25893 @item -maddress-space-conversion
25894 @itemx -mno-address-space-conversion
25895 @opindex maddress-space-conversion
25896 @opindex mno-address-space-conversion
25897 Allow/disallow treating the @code{__ea} address space as superset
25898 of the generic address space. This enables explicit type casts
25899 between @code{__ea} and generic pointer as well as implicit
25900 conversions of generic pointers to @code{__ea} pointers. The
25901 default is to allow address space pointer conversions.
25903 @item -mcache-size=@var{cache-size}
25904 @opindex mcache-size
25905 This option controls the version of libgcc that the compiler links to an
25906 executable and selects a software-managed cache for accessing variables
25907 in the @code{__ea} address space with a particular cache size. Possible
25908 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
25909 and @samp{128}. The default cache size is 64KB.
25911 @item -matomic-updates
25912 @itemx -mno-atomic-updates
25913 @opindex matomic-updates
25914 @opindex mno-atomic-updates
25915 This option controls the version of libgcc that the compiler links to an
25916 executable and selects whether atomic updates to the software-managed
25917 cache of PPU-side variables are used. If you use atomic updates, changes
25918 to a PPU variable from SPU code using the @code{__ea} named address space
25919 qualifier do not interfere with changes to other PPU variables residing
25920 in the same cache line from PPU code. If you do not use atomic updates,
25921 such interference may occur; however, writing back cache lines is
25922 more efficient. The default behavior is to use atomic updates.
25925 @itemx -mdual-nops=@var{n}
25926 @opindex mdual-nops
25927 By default, GCC inserts NOPs to increase dual issue when it expects
25928 it to increase performance. @var{n} can be a value from 0 to 10. A
25929 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
25930 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
25932 @item -mhint-max-nops=@var{n}
25933 @opindex mhint-max-nops
25934 Maximum number of NOPs to insert for a branch hint. A branch hint must
25935 be at least 8 instructions away from the branch it is affecting. GCC
25936 inserts up to @var{n} NOPs to enforce this, otherwise it does not
25937 generate the branch hint.
25939 @item -mhint-max-distance=@var{n}
25940 @opindex mhint-max-distance
25941 The encoding of the branch hint instruction limits the hint to be within
25942 256 instructions of the branch it is affecting. By default, GCC makes
25943 sure it is within 125.
25946 @opindex msafe-hints
25947 Work around a hardware bug that causes the SPU to stall indefinitely.
25948 By default, GCC inserts the @code{hbrp} instruction to make sure
25949 this stall won't happen.
25953 @node System V Options
25954 @subsection Options for System V
25956 These additional options are available on System V Release 4 for
25957 compatibility with other compilers on those systems:
25962 Create a shared object.
25963 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
25967 Identify the versions of each tool used by the compiler, in a
25968 @code{.ident} assembler directive in the output.
25972 Refrain from adding @code{.ident} directives to the output file (this is
25975 @item -YP,@var{dirs}
25977 Search the directories @var{dirs}, and no others, for libraries
25978 specified with @option{-l}.
25980 @item -Ym,@var{dir}
25982 Look in the directory @var{dir} to find the M4 preprocessor.
25983 The assembler uses this option.
25984 @c This is supposed to go with a -Yd for predefined M4 macro files, but
25985 @c the generic assembler that comes with Solaris takes just -Ym.
25988 @node TILE-Gx Options
25989 @subsection TILE-Gx Options
25990 @cindex TILE-Gx options
25992 These @samp{-m} options are supported on the TILE-Gx:
25995 @item -mcmodel=small
25996 @opindex mcmodel=small
25997 Generate code for the small model. The distance for direct calls is
25998 limited to 500M in either direction. PC-relative addresses are 32
25999 bits. Absolute addresses support the full address range.
26001 @item -mcmodel=large
26002 @opindex mcmodel=large
26003 Generate code for the large model. There is no limitation on call
26004 distance, pc-relative addresses, or absolute addresses.
26006 @item -mcpu=@var{name}
26008 Selects the type of CPU to be targeted. Currently the only supported
26009 type is @samp{tilegx}.
26015 Generate code for a 32-bit or 64-bit environment. The 32-bit
26016 environment sets int, long, and pointer to 32 bits. The 64-bit
26017 environment sets int to 32 bits and long and pointer to 64 bits.
26020 @itemx -mlittle-endian
26021 @opindex mbig-endian
26022 @opindex mlittle-endian
26023 Generate code in big/little endian mode, respectively.
26026 @node TILEPro Options
26027 @subsection TILEPro Options
26028 @cindex TILEPro options
26030 These @samp{-m} options are supported on the TILEPro:
26033 @item -mcpu=@var{name}
26035 Selects the type of CPU to be targeted. Currently the only supported
26036 type is @samp{tilepro}.
26040 Generate code for a 32-bit environment, which sets int, long, and
26041 pointer to 32 bits. This is the only supported behavior so the flag
26042 is essentially ignored.
26046 @subsection V850 Options
26047 @cindex V850 Options
26049 These @samp{-m} options are defined for V850 implementations:
26053 @itemx -mno-long-calls
26054 @opindex mlong-calls
26055 @opindex mno-long-calls
26056 Treat all calls as being far away (near). If calls are assumed to be
26057 far away, the compiler always loads the function's address into a
26058 register, and calls indirect through the pointer.
26064 Do not optimize (do optimize) basic blocks that use the same index
26065 pointer 4 or more times to copy pointer into the @code{ep} register, and
26066 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
26067 option is on by default if you optimize.
26069 @item -mno-prolog-function
26070 @itemx -mprolog-function
26071 @opindex mno-prolog-function
26072 @opindex mprolog-function
26073 Do not use (do use) external functions to save and restore registers
26074 at the prologue and epilogue of a function. The external functions
26075 are slower, but use less code space if more than one function saves
26076 the same number of registers. The @option{-mprolog-function} option
26077 is on by default if you optimize.
26081 Try to make the code as small as possible. At present, this just turns
26082 on the @option{-mep} and @option{-mprolog-function} options.
26084 @item -mtda=@var{n}
26086 Put static or global variables whose size is @var{n} bytes or less into
26087 the tiny data area that register @code{ep} points to. The tiny data
26088 area can hold up to 256 bytes in total (128 bytes for byte references).
26090 @item -msda=@var{n}
26092 Put static or global variables whose size is @var{n} bytes or less into
26093 the small data area that register @code{gp} points to. The small data
26094 area can hold up to 64 kilobytes.
26096 @item -mzda=@var{n}
26098 Put static or global variables whose size is @var{n} bytes or less into
26099 the first 32 kilobytes of memory.
26103 Specify that the target processor is the V850.
26107 Specify that the target processor is the V850E3V5. The preprocessor
26108 constant @code{__v850e3v5__} is defined if this option is used.
26112 Specify that the target processor is the V850E3V5. This is an alias for
26113 the @option{-mv850e3v5} option.
26117 Specify that the target processor is the V850E2V3. The preprocessor
26118 constant @code{__v850e2v3__} is defined if this option is used.
26122 Specify that the target processor is the V850E2. The preprocessor
26123 constant @code{__v850e2__} is defined if this option is used.
26127 Specify that the target processor is the V850E1. The preprocessor
26128 constants @code{__v850e1__} and @code{__v850e__} are defined if
26129 this option is used.
26133 Specify that the target processor is the V850ES. This is an alias for
26134 the @option{-mv850e1} option.
26138 Specify that the target processor is the V850E@. The preprocessor
26139 constant @code{__v850e__} is defined if this option is used.
26141 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
26142 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
26143 are defined then a default target processor is chosen and the
26144 relevant @samp{__v850*__} preprocessor constant is defined.
26146 The preprocessor constants @code{__v850} and @code{__v851__} are always
26147 defined, regardless of which processor variant is the target.
26149 @item -mdisable-callt
26150 @itemx -mno-disable-callt
26151 @opindex mdisable-callt
26152 @opindex mno-disable-callt
26153 This option suppresses generation of the @code{CALLT} instruction for the
26154 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
26157 This option is enabled by default when the RH850 ABI is
26158 in use (see @option{-mrh850-abi}), and disabled by default when the
26159 GCC ABI is in use. If @code{CALLT} instructions are being generated
26160 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
26166 Pass on (or do not pass on) the @option{-mrelax} command-line option
26170 @itemx -mno-long-jumps
26171 @opindex mlong-jumps
26172 @opindex mno-long-jumps
26173 Disable (or re-enable) the generation of PC-relative jump instructions.
26176 @itemx -mhard-float
26177 @opindex msoft-float
26178 @opindex mhard-float
26179 Disable (or re-enable) the generation of hardware floating point
26180 instructions. This option is only significant when the target
26181 architecture is @samp{V850E2V3} or higher. If hardware floating point
26182 instructions are being generated then the C preprocessor symbol
26183 @code{__FPU_OK__} is defined, otherwise the symbol
26184 @code{__NO_FPU__} is defined.
26188 Enables the use of the e3v5 LOOP instruction. The use of this
26189 instruction is not enabled by default when the e3v5 architecture is
26190 selected because its use is still experimental.
26194 @opindex mrh850-abi
26196 Enables support for the RH850 version of the V850 ABI. This is the
26197 default. With this version of the ABI the following rules apply:
26201 Integer sized structures and unions are returned via a memory pointer
26202 rather than a register.
26205 Large structures and unions (more than 8 bytes in size) are passed by
26209 Functions are aligned to 16-bit boundaries.
26212 The @option{-m8byte-align} command-line option is supported.
26215 The @option{-mdisable-callt} command-line option is enabled by
26216 default. The @option{-mno-disable-callt} command-line option is not
26220 When this version of the ABI is enabled the C preprocessor symbol
26221 @code{__V850_RH850_ABI__} is defined.
26225 Enables support for the old GCC version of the V850 ABI. With this
26226 version of the ABI the following rules apply:
26230 Integer sized structures and unions are returned in register @code{r10}.
26233 Large structures and unions (more than 8 bytes in size) are passed by
26237 Functions are aligned to 32-bit boundaries, unless optimizing for
26241 The @option{-m8byte-align} command-line option is not supported.
26244 The @option{-mdisable-callt} command-line option is supported but not
26245 enabled by default.
26248 When this version of the ABI is enabled the C preprocessor symbol
26249 @code{__V850_GCC_ABI__} is defined.
26251 @item -m8byte-align
26252 @itemx -mno-8byte-align
26253 @opindex m8byte-align
26254 @opindex mno-8byte-align
26255 Enables support for @code{double} and @code{long long} types to be
26256 aligned on 8-byte boundaries. The default is to restrict the
26257 alignment of all objects to at most 4-bytes. When
26258 @option{-m8byte-align} is in effect the C preprocessor symbol
26259 @code{__V850_8BYTE_ALIGN__} is defined.
26262 @opindex mbig-switch
26263 Generate code suitable for big switch tables. Use this option only if
26264 the assembler/linker complain about out of range branches within a switch
26269 This option causes r2 and r5 to be used in the code generated by
26270 the compiler. This setting is the default.
26272 @item -mno-app-regs
26273 @opindex mno-app-regs
26274 This option causes r2 and r5 to be treated as fixed registers.
26279 @subsection VAX Options
26280 @cindex VAX options
26282 These @samp{-m} options are defined for the VAX:
26287 Do not output certain jump instructions (@code{aobleq} and so on)
26288 that the Unix assembler for the VAX cannot handle across long
26293 Do output those jump instructions, on the assumption that the
26294 GNU assembler is being used.
26298 Output code for G-format floating-point numbers instead of D-format.
26301 @node Visium Options
26302 @subsection Visium Options
26303 @cindex Visium options
26309 A program which performs file I/O and is destined to run on an MCM target
26310 should be linked with this option. It causes the libraries libc.a and
26311 libdebug.a to be linked. The program should be run on the target under
26312 the control of the GDB remote debugging stub.
26316 A program which performs file I/O and is destined to run on the simulator
26317 should be linked with option. This causes libraries libc.a and libsim.a to
26321 @itemx -mhard-float
26323 @opindex mhard-float
26324 Generate code containing floating-point instructions. This is the
26328 @itemx -msoft-float
26330 @opindex msoft-float
26331 Generate code containing library calls for floating-point.
26333 @option{-msoft-float} changes the calling convention in the output file;
26334 therefore, it is only useful if you compile @emph{all} of a program with
26335 this option. In particular, you need to compile @file{libgcc.a}, the
26336 library that comes with GCC, with @option{-msoft-float} in order for
26339 @item -mcpu=@var{cpu_type}
26341 Set the instruction set, register set, and instruction scheduling parameters
26342 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
26343 @samp{mcm}, @samp{gr5} and @samp{gr6}.
26345 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
26347 By default (unless configured otherwise), GCC generates code for the GR5
26348 variant of the Visium architecture.
26350 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
26351 architecture. The only difference from GR5 code is that the compiler will
26352 generate block move instructions.
26354 @item -mtune=@var{cpu_type}
26356 Set the instruction scheduling parameters for machine type @var{cpu_type},
26357 but do not set the instruction set or register set that the option
26358 @option{-mcpu=@var{cpu_type}} would.
26362 Generate code for the supervisor mode, where there are no restrictions on
26363 the access to general registers. This is the default.
26366 @opindex muser-mode
26367 Generate code for the user mode, where the access to some general registers
26368 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
26369 mode; on the GR6, only registers r29 to r31 are affected.
26373 @subsection VMS Options
26375 These @samp{-m} options are defined for the VMS implementations:
26378 @item -mvms-return-codes
26379 @opindex mvms-return-codes
26380 Return VMS condition codes from @code{main}. The default is to return POSIX-style
26381 condition (e.g.@ error) codes.
26383 @item -mdebug-main=@var{prefix}
26384 @opindex mdebug-main=@var{prefix}
26385 Flag the first routine whose name starts with @var{prefix} as the main
26386 routine for the debugger.
26390 Default to 64-bit memory allocation routines.
26392 @item -mpointer-size=@var{size}
26393 @opindex mpointer-size=@var{size}
26394 Set the default size of pointers. Possible options for @var{size} are
26395 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
26396 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
26397 The later option disables @code{pragma pointer_size}.
26400 @node VxWorks Options
26401 @subsection VxWorks Options
26402 @cindex VxWorks Options
26404 The options in this section are defined for all VxWorks targets.
26405 Options specific to the target hardware are listed with the other
26406 options for that target.
26411 GCC can generate code for both VxWorks kernels and real time processes
26412 (RTPs). This option switches from the former to the latter. It also
26413 defines the preprocessor macro @code{__RTP__}.
26416 @opindex non-static
26417 Link an RTP executable against shared libraries rather than static
26418 libraries. The options @option{-static} and @option{-shared} can
26419 also be used for RTPs (@pxref{Link Options}); @option{-static}
26426 These options are passed down to the linker. They are defined for
26427 compatibility with Diab.
26430 @opindex Xbind-lazy
26431 Enable lazy binding of function calls. This option is equivalent to
26432 @option{-Wl,-z,now} and is defined for compatibility with Diab.
26436 Disable lazy binding of function calls. This option is the default and
26437 is defined for compatibility with Diab.
26441 @subsection x86 Options
26442 @cindex x86 Options
26444 These @samp{-m} options are defined for the x86 family of computers.
26448 @item -march=@var{cpu-type}
26450 Generate instructions for the machine type @var{cpu-type}. In contrast to
26451 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
26452 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
26453 to generate code that may not run at all on processors other than the one
26454 indicated. Specifying @option{-march=@var{cpu-type}} implies
26455 @option{-mtune=@var{cpu-type}}.
26457 The choices for @var{cpu-type} are:
26461 This selects the CPU to generate code for at compilation time by determining
26462 the processor type of the compiling machine. Using @option{-march=native}
26463 enables all instruction subsets supported by the local machine (hence
26464 the result might not run on different machines). Using @option{-mtune=native}
26465 produces code optimized for the local machine under the constraints
26466 of the selected instruction set.
26469 A generic CPU with 64-bit extensions.
26472 Original Intel i386 CPU@.
26475 Intel i486 CPU@. (No scheduling is implemented for this chip.)
26479 Intel Pentium CPU with no MMX support.
26482 Intel Lakemont MCU, based on Intel Pentium CPU.
26485 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
26488 Intel Pentium Pro CPU@.
26491 When used with @option{-march}, the Pentium Pro
26492 instruction set is used, so the code runs on all i686 family chips.
26493 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
26496 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
26501 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
26505 Intel Pentium M; low-power version of Intel Pentium III CPU
26506 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
26510 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
26513 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
26517 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
26518 SSE2 and SSE3 instruction set support.
26521 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
26522 instruction set support.
26525 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
26526 SSE4.1, SSE4.2 and POPCNT instruction set support.
26529 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
26530 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
26533 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
26534 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
26537 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
26538 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
26539 instruction set support.
26542 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
26543 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
26544 BMI, BMI2 and F16C instruction set support.
26547 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
26548 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
26549 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
26552 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
26553 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
26554 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
26555 XSAVES instruction set support.
26558 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
26559 instruction set support.
26562 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
26563 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
26566 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
26567 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
26568 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
26569 AVX512CD instruction set support.
26572 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
26573 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
26574 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
26575 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
26577 @item skylake-avx512
26578 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
26579 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
26580 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
26581 CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
26584 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
26585 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
26586 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
26587 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
26588 AVX512IFMA, SHA and UMIP instruction set support.
26590 @item icelake-client
26591 Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
26592 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
26593 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
26594 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
26595 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
26596 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support.
26598 @item icelake-server
26599 Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
26600 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
26601 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
26602 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
26603 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
26604 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, PCONFIG and WBNOINVD instruction
26608 AMD K6 CPU with MMX instruction set support.
26612 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
26615 @itemx athlon-tbird
26616 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
26622 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
26623 instruction set support.
26629 Processors based on the AMD K8 core with x86-64 instruction set support,
26630 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
26631 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
26632 instruction set extensions.)
26635 @itemx opteron-sse3
26636 @itemx athlon64-sse3
26637 Improved versions of AMD K8 cores with SSE3 instruction set support.
26641 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
26642 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
26643 instruction set extensions.)
26646 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
26647 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
26648 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
26650 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
26651 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
26652 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
26655 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
26656 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
26657 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
26658 64-bit instruction set extensions.
26660 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
26661 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
26662 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
26663 SSE4.2, ABM and 64-bit instruction set extensions.
26666 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
26667 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
26668 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
26669 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
26670 instruction set extensions.
26673 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
26674 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
26675 instruction set extensions.)
26678 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
26679 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
26680 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
26683 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
26687 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
26688 instruction set support.
26691 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
26692 (No scheduling is implemented for this chip.)
26695 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
26696 (No scheduling is implemented for this chip.)
26699 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
26700 (No scheduling is implemented for this chip.)
26703 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
26704 (No scheduling is implemented for this chip.)
26707 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
26708 (No scheduling is implemented for this chip.)
26711 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
26712 (No scheduling is implemented for this chip.)
26715 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
26716 (No scheduling is implemented for this chip.)
26719 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
26720 AVX and AVX2 instruction set support.
26721 (No scheduling is implemented for this chip.)
26724 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
26725 instruction set support.
26726 (No scheduling is implemented for this chip.)
26729 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
26730 instruction set support.
26731 (No scheduling is implemented for this chip.)
26734 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
26735 instruction set support.
26736 (No scheduling is implemented for this chip.)
26739 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
26740 instruction set support.
26741 (No scheduling is implemented for this chip.)
26744 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
26745 instruction set support.
26746 (No scheduling is implemented for this chip.)
26749 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
26750 instruction set support.
26751 (No scheduling is implemented for this chip.)
26754 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
26757 @item -mtune=@var{cpu-type}
26759 Tune to @var{cpu-type} everything applicable about the generated code, except
26760 for the ABI and the set of available instructions.
26761 While picking a specific @var{cpu-type} schedules things appropriately
26762 for that particular chip, the compiler does not generate any code that
26763 cannot run on the default machine type unless you use a
26764 @option{-march=@var{cpu-type}} option.
26765 For example, if GCC is configured for i686-pc-linux-gnu
26766 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
26767 but still runs on i686 machines.
26769 The choices for @var{cpu-type} are the same as for @option{-march}.
26770 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
26774 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
26775 If you know the CPU on which your code will run, then you should use
26776 the corresponding @option{-mtune} or @option{-march} option instead of
26777 @option{-mtune=generic}. But, if you do not know exactly what CPU users
26778 of your application will have, then you should use this option.
26780 As new processors are deployed in the marketplace, the behavior of this
26781 option will change. Therefore, if you upgrade to a newer version of
26782 GCC, code generation controlled by this option will change to reflect
26784 that are most common at the time that version of GCC is released.
26786 There is no @option{-march=generic} option because @option{-march}
26787 indicates the instruction set the compiler can use, and there is no
26788 generic instruction set applicable to all processors. In contrast,
26789 @option{-mtune} indicates the processor (or, in this case, collection of
26790 processors) for which the code is optimized.
26793 Produce code optimized for the most current Intel processors, which are
26794 Haswell and Silvermont for this version of GCC. If you know the CPU
26795 on which your code will run, then you should use the corresponding
26796 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
26797 But, if you want your application performs better on both Haswell and
26798 Silvermont, then you should use this option.
26800 As new Intel processors are deployed in the marketplace, the behavior of
26801 this option will change. Therefore, if you upgrade to a newer version of
26802 GCC, code generation controlled by this option will change to reflect
26803 the most current Intel processors at the time that version of GCC is
26806 There is no @option{-march=intel} option because @option{-march} indicates
26807 the instruction set the compiler can use, and there is no common
26808 instruction set applicable to all processors. In contrast,
26809 @option{-mtune} indicates the processor (or, in this case, collection of
26810 processors) for which the code is optimized.
26813 @item -mcpu=@var{cpu-type}
26815 A deprecated synonym for @option{-mtune}.
26817 @item -mfpmath=@var{unit}
26819 Generate floating-point arithmetic for selected unit @var{unit}. The choices
26820 for @var{unit} are:
26824 Use the standard 387 floating-point coprocessor present on the majority of chips and
26825 emulated otherwise. Code compiled with this option runs almost everywhere.
26826 The temporary results are computed in 80-bit precision instead of the precision
26827 specified by the type, resulting in slightly different results compared to most
26828 of other chips. See @option{-ffloat-store} for more detailed description.
26830 This is the default choice for non-Darwin x86-32 targets.
26833 Use scalar floating-point instructions present in the SSE instruction set.
26834 This instruction set is supported by Pentium III and newer chips,
26835 and in the AMD line
26836 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
26837 instruction set supports only single-precision arithmetic, thus the double and
26838 extended-precision arithmetic are still done using 387. A later version, present
26839 only in Pentium 4 and AMD x86-64 chips, supports double-precision
26842 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
26843 or @option{-msse2} switches to enable SSE extensions and make this option
26844 effective. For the x86-64 compiler, these extensions are enabled by default.
26846 The resulting code should be considerably faster in the majority of cases and avoid
26847 the numerical instability problems of 387 code, but may break some existing
26848 code that expects temporaries to be 80 bits.
26850 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
26851 and the default choice for x86-32 targets with the SSE2 instruction set
26852 when @option{-ffast-math} is enabled.
26857 Attempt to utilize both instruction sets at once. This effectively doubles the
26858 amount of available registers, and on chips with separate execution units for
26859 387 and SSE the execution resources too. Use this option with care, as it is
26860 still experimental, because the GCC register allocator does not model separate
26861 functional units well, resulting in unstable performance.
26864 @item -masm=@var{dialect}
26865 @opindex masm=@var{dialect}
26866 Output assembly instructions using selected @var{dialect}. Also affects
26867 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
26868 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
26869 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
26870 not support @samp{intel}.
26873 @itemx -mno-ieee-fp
26875 @opindex mno-ieee-fp
26876 Control whether or not the compiler uses IEEE floating-point
26877 comparisons. These correctly handle the case where the result of a
26878 comparison is unordered.
26881 @itemx -mhard-float
26883 @opindex mhard-float
26884 Generate output containing 80387 instructions for floating point.
26887 @itemx -msoft-float
26889 @opindex msoft-float
26890 Generate output containing library calls for floating point.
26892 @strong{Warning:} the requisite libraries are not part of GCC@.
26893 Normally the facilities of the machine's usual C compiler are used, but
26894 this cannot be done directly in cross-compilation. You must make your
26895 own arrangements to provide suitable library functions for
26898 On machines where a function returns floating-point results in the 80387
26899 register stack, some floating-point opcodes may be emitted even if
26900 @option{-msoft-float} is used.
26902 @item -mno-fp-ret-in-387
26903 @opindex mno-fp-ret-in-387
26904 Do not use the FPU registers for return values of functions.
26906 The usual calling convention has functions return values of types
26907 @code{float} and @code{double} in an FPU register, even if there
26908 is no FPU@. The idea is that the operating system should emulate
26911 The option @option{-mno-fp-ret-in-387} causes such values to be returned
26912 in ordinary CPU registers instead.
26914 @item -mno-fancy-math-387
26915 @opindex mno-fancy-math-387
26916 Some 387 emulators do not support the @code{sin}, @code{cos} and
26917 @code{sqrt} instructions for the 387. Specify this option to avoid
26918 generating those instructions. This option is the default on
26919 OpenBSD and NetBSD@. This option is overridden when @option{-march}
26920 indicates that the target CPU always has an FPU and so the
26921 instruction does not need emulation. These
26922 instructions are not generated unless you also use the
26923 @option{-funsafe-math-optimizations} switch.
26925 @item -malign-double
26926 @itemx -mno-align-double
26927 @opindex malign-double
26928 @opindex mno-align-double
26929 Control whether GCC aligns @code{double}, @code{long double}, and
26930 @code{long long} variables on a two-word boundary or a one-word
26931 boundary. Aligning @code{double} variables on a two-word boundary
26932 produces code that runs somewhat faster on a Pentium at the
26933 expense of more memory.
26935 On x86-64, @option{-malign-double} is enabled by default.
26937 @strong{Warning:} if you use the @option{-malign-double} switch,
26938 structures containing the above types are aligned differently than
26939 the published application binary interface specifications for the x86-32
26940 and are not binary compatible with structures in code compiled
26941 without that switch.
26943 @item -m96bit-long-double
26944 @itemx -m128bit-long-double
26945 @opindex m96bit-long-double
26946 @opindex m128bit-long-double
26947 These switches control the size of @code{long double} type. The x86-32
26948 application binary interface specifies the size to be 96 bits,
26949 so @option{-m96bit-long-double} is the default in 32-bit mode.
26951 Modern architectures (Pentium and newer) prefer @code{long double}
26952 to be aligned to an 8- or 16-byte boundary. In arrays or structures
26953 conforming to the ABI, this is not possible. So specifying
26954 @option{-m128bit-long-double} aligns @code{long double}
26955 to a 16-byte boundary by padding the @code{long double} with an additional
26958 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
26959 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
26961 Notice that neither of these options enable any extra precision over the x87
26962 standard of 80 bits for a @code{long double}.
26964 @strong{Warning:} if you override the default value for your target ABI, this
26965 changes the size of
26966 structures and arrays containing @code{long double} variables,
26967 as well as modifying the function calling convention for functions taking
26968 @code{long double}. Hence they are not binary-compatible
26969 with code compiled without that switch.
26971 @item -mlong-double-64
26972 @itemx -mlong-double-80
26973 @itemx -mlong-double-128
26974 @opindex mlong-double-64
26975 @opindex mlong-double-80
26976 @opindex mlong-double-128
26977 These switches control the size of @code{long double} type. A size
26978 of 64 bits makes the @code{long double} type equivalent to the @code{double}
26979 type. This is the default for 32-bit Bionic C library. A size
26980 of 128 bits makes the @code{long double} type equivalent to the
26981 @code{__float128} type. This is the default for 64-bit Bionic C library.
26983 @strong{Warning:} if you override the default value for your target ABI, this
26984 changes the size of
26985 structures and arrays containing @code{long double} variables,
26986 as well as modifying the function calling convention for functions taking
26987 @code{long double}. Hence they are not binary-compatible
26988 with code compiled without that switch.
26990 @item -malign-data=@var{type}
26991 @opindex malign-data
26992 Control how GCC aligns variables. Supported values for @var{type} are
26993 @samp{compat} uses increased alignment value compatible uses GCC 4.8
26994 and earlier, @samp{abi} uses alignment value as specified by the
26995 psABI, and @samp{cacheline} uses increased alignment value to match
26996 the cache line size. @samp{compat} is the default.
26998 @item -mlarge-data-threshold=@var{threshold}
26999 @opindex mlarge-data-threshold
27000 When @option{-mcmodel=medium} is specified, data objects larger than
27001 @var{threshold} are placed in the large data section. This value must be the
27002 same across all objects linked into the binary, and defaults to 65535.
27006 Use a different function-calling convention, in which functions that
27007 take a fixed number of arguments return with the @code{ret @var{num}}
27008 instruction, which pops their arguments while returning. This saves one
27009 instruction in the caller since there is no need to pop the arguments
27012 You can specify that an individual function is called with this calling
27013 sequence with the function attribute @code{stdcall}. You can also
27014 override the @option{-mrtd} option by using the function attribute
27015 @code{cdecl}. @xref{Function Attributes}.
27017 @strong{Warning:} this calling convention is incompatible with the one
27018 normally used on Unix, so you cannot use it if you need to call
27019 libraries compiled with the Unix compiler.
27021 Also, you must provide function prototypes for all functions that
27022 take variable numbers of arguments (including @code{printf});
27023 otherwise incorrect code is generated for calls to those
27026 In addition, seriously incorrect code results if you call a
27027 function with too many arguments. (Normally, extra arguments are
27028 harmlessly ignored.)
27030 @item -mregparm=@var{num}
27032 Control how many registers are used to pass integer arguments. By
27033 default, no registers are used to pass arguments, and at most 3
27034 registers can be used. You can control this behavior for a specific
27035 function by using the function attribute @code{regparm}.
27036 @xref{Function Attributes}.
27038 @strong{Warning:} if you use this switch, and
27039 @var{num} is nonzero, then you must build all modules with the same
27040 value, including any libraries. This includes the system libraries and
27044 @opindex msseregparm
27045 Use SSE register passing conventions for float and double arguments
27046 and return values. You can control this behavior for a specific
27047 function by using the function attribute @code{sseregparm}.
27048 @xref{Function Attributes}.
27050 @strong{Warning:} if you use this switch then you must build all
27051 modules with the same value, including any libraries. This includes
27052 the system libraries and startup modules.
27054 @item -mvect8-ret-in-mem
27055 @opindex mvect8-ret-in-mem
27056 Return 8-byte vectors in memory instead of MMX registers. This is the
27057 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
27058 Studio compilers until version 12. Later compiler versions (starting
27059 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
27060 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
27061 you need to remain compatible with existing code produced by those
27062 previous compiler versions or older versions of GCC@.
27071 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
27072 is specified, the significands of results of floating-point operations are
27073 rounded to 24 bits (single precision); @option{-mpc64} rounds the
27074 significands of results of floating-point operations to 53 bits (double
27075 precision) and @option{-mpc80} rounds the significands of results of
27076 floating-point operations to 64 bits (extended double precision), which is
27077 the default. When this option is used, floating-point operations in higher
27078 precisions are not available to the programmer without setting the FPU
27079 control word explicitly.
27081 Setting the rounding of floating-point operations to less than the default
27082 80 bits can speed some programs by 2% or more. Note that some mathematical
27083 libraries assume that extended-precision (80-bit) floating-point operations
27084 are enabled by default; routines in such libraries could suffer significant
27085 loss of accuracy, typically through so-called ``catastrophic cancellation'',
27086 when this option is used to set the precision to less than extended precision.
27088 @item -mstackrealign
27089 @opindex mstackrealign
27090 Realign the stack at entry. On the x86, the @option{-mstackrealign}
27091 option generates an alternate prologue and epilogue that realigns the
27092 run-time stack if necessary. This supports mixing legacy codes that keep
27093 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
27094 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
27095 applicable to individual functions.
27097 @item -mpreferred-stack-boundary=@var{num}
27098 @opindex mpreferred-stack-boundary
27099 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
27100 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
27101 the default is 4 (16 bytes or 128 bits).
27103 @strong{Warning:} When generating code for the x86-64 architecture with
27104 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
27105 used to keep the stack boundary aligned to 8 byte boundary. Since
27106 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
27107 intended to be used in controlled environment where stack space is
27108 important limitation. This option leads to wrong code when functions
27109 compiled with 16 byte stack alignment (such as functions from a standard
27110 library) are called with misaligned stack. In this case, SSE
27111 instructions may lead to misaligned memory access traps. In addition,
27112 variable arguments are handled incorrectly for 16 byte aligned
27113 objects (including x87 long double and __int128), leading to wrong
27114 results. You must build all modules with
27115 @option{-mpreferred-stack-boundary=3}, including any libraries. This
27116 includes the system libraries and startup modules.
27118 @item -mincoming-stack-boundary=@var{num}
27119 @opindex mincoming-stack-boundary
27120 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
27121 boundary. If @option{-mincoming-stack-boundary} is not specified,
27122 the one specified by @option{-mpreferred-stack-boundary} is used.
27124 On Pentium and Pentium Pro, @code{double} and @code{long double} values
27125 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
27126 suffer significant run time performance penalties. On Pentium III, the
27127 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
27128 properly if it is not 16-byte aligned.
27130 To ensure proper alignment of this values on the stack, the stack boundary
27131 must be as aligned as that required by any value stored on the stack.
27132 Further, every function must be generated such that it keeps the stack
27133 aligned. Thus calling a function compiled with a higher preferred
27134 stack boundary from a function compiled with a lower preferred stack
27135 boundary most likely misaligns the stack. It is recommended that
27136 libraries that use callbacks always use the default setting.
27138 This extra alignment does consume extra stack space, and generally
27139 increases code size. Code that is sensitive to stack space usage, such
27140 as embedded systems and operating system kernels, may want to reduce the
27141 preferred alignment to @option{-mpreferred-stack-boundary=2}.
27198 @itemx -mavx512ifma
27199 @opindex mavx512ifma
27201 @itemx -mavx512vbmi
27202 @opindex mavx512vbmi
27213 @itemx -mclflushopt
27214 @opindex mclflushopt
27237 @itemx -mprefetchwt1
27238 @opindex mprefetchwt1
27302 @itemx -mavx512vbmi2
27303 @opindex mavx512vbmi2
27311 @itemx -mvpclmulqdq
27312 @opindex mvpclmulqdq
27314 @itemx -mavx512bitalg
27315 @opindex mavx512bitalg
27317 @itemx -mavx512vpopcntdq
27318 @opindex mavx512vpopcntdq
27319 These switches enable the use of instructions in the MMX, SSE,
27320 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
27321 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
27322 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, BMI, BMI2, VAES,
27323 FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, IBT, SHSTK, AVX512VBMI2,
27324 GFNI, VPCLMULQDQ, AVX512BITALG, AVX512VPOPCNTDQ3DNow!@: or enhanced 3DNow!@:
27325 extended instruction sets.
27326 Each has a corresponding @option{-mno-} option to disable use of these
27329 These extensions are also available as built-in functions: see
27330 @ref{x86 Built-in Functions}, for details of the functions enabled and
27331 disabled by these switches.
27333 To generate SSE/SSE2 instructions automatically from floating-point
27334 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
27336 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
27337 generates new AVX instructions or AVX equivalence for all SSEx instructions
27340 These options enable GCC to use these extended instructions in
27341 generated code, even without @option{-mfpmath=sse}. Applications that
27342 perform run-time CPU detection must compile separate files for each
27343 supported architecture, using the appropriate flags. In particular,
27344 the file containing the CPU detection code should be compiled without
27347 The @option{-mcet} option turns on the @option{-mibt} and @option{-mshstk}
27348 options. The @option{-mibt} option enables indirect branch tracking support
27349 and the @option{-mshstk} option enables shadow stack support from
27350 Intel Control-flow Enforcement Technology (CET). The compiler also provides
27351 a number of built-in functions for fine-grained control in a CET-based
27352 application. See @xref{x86 Built-in Functions}, for more information.
27354 @item -mdump-tune-features
27355 @opindex mdump-tune-features
27356 This option instructs GCC to dump the names of the x86 performance
27357 tuning features and default settings. The names can be used in
27358 @option{-mtune-ctrl=@var{feature-list}}.
27360 @item -mtune-ctrl=@var{feature-list}
27361 @opindex mtune-ctrl=@var{feature-list}
27362 This option is used to do fine grain control of x86 code generation features.
27363 @var{feature-list} is a comma separated list of @var{feature} names. See also
27364 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
27365 on if it is not preceded with @samp{^}, otherwise, it is turned off.
27366 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
27367 developers. Using it may lead to code paths not covered by testing and can
27368 potentially result in compiler ICEs or runtime errors.
27371 @opindex mno-default
27372 This option instructs GCC to turn off all tunable features. See also
27373 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
27377 This option instructs GCC to emit a @code{cld} instruction in the prologue
27378 of functions that use string instructions. String instructions depend on
27379 the DF flag to select between autoincrement or autodecrement mode. While the
27380 ABI specifies the DF flag to be cleared on function entry, some operating
27381 systems violate this specification by not clearing the DF flag in their
27382 exception dispatchers. The exception handler can be invoked with the DF flag
27383 set, which leads to wrong direction mode when string instructions are used.
27384 This option can be enabled by default on 32-bit x86 targets by configuring
27385 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
27386 instructions can be suppressed with the @option{-mno-cld} compiler option
27390 @opindex mvzeroupper
27391 This option instructs GCC to emit a @code{vzeroupper} instruction
27392 before a transfer of control flow out of the function to minimize
27393 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
27396 @item -mprefer-avx128
27397 @opindex mprefer-avx128
27398 This option instructs GCC to use 128-bit AVX instructions instead of
27399 256-bit AVX instructions in the auto-vectorizer.
27401 @item -mprefer-vector-width=@var{opt}
27402 @opindex mprefer-vector-width
27403 This option instructs GCC to use @var{opt}-bit vector width in instructions
27404 instead of default on the selected platform.
27408 No extra limitations applied to GCC other than defined by the selected platform.
27411 Prefer 128-bit vector width for instructions.
27414 Prefer 256-bit vector width for instructions.
27417 Prefer 512-bit vector width for instructions.
27422 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
27423 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
27424 objects. This is useful for atomic updates of data structures exceeding one
27425 machine word in size. The compiler uses this instruction to implement
27426 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
27427 128-bit integers, a library call is always used.
27431 This option enables generation of @code{SAHF} instructions in 64-bit code.
27432 Early Intel Pentium 4 CPUs with Intel 64 support,
27433 prior to the introduction of Pentium 4 G1 step in December 2005,
27434 lacked the @code{LAHF} and @code{SAHF} instructions
27435 which are supported by AMD64.
27436 These are load and store instructions, respectively, for certain status flags.
27437 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
27438 @code{drem}, and @code{remainder} built-in functions;
27439 see @ref{Other Builtins} for details.
27443 This option enables use of the @code{movbe} instruction to implement
27444 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
27448 This option tells the compiler to use indirect branch tracking support
27449 (for indirect calls and jumps) from x86 Control-flow Enforcement
27450 Technology (CET). The option has effect only if the
27451 @option{-fcf-protection=full} or @option{-fcf-protection=branch} option
27452 is specified. The option @option{-mibt} is on by default when the
27453 @code{-mcet} option is specified.
27457 This option tells the compiler to use shadow stack support (return
27458 address tracking) from x86 Control-flow Enforcement Technology (CET).
27459 The option has effect only if the @option{-fcf-protection=full} or
27460 @option{-fcf-protection=return} option is specified. The option
27461 @option{-mshstk} is on by default when the @option{-mcet} option is
27466 This option enables built-in functions @code{__builtin_ia32_crc32qi},
27467 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
27468 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
27472 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
27473 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
27474 with an additional Newton-Raphson step
27475 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
27476 (and their vectorized
27477 variants) for single-precision floating-point arguments. These instructions
27478 are generated only when @option{-funsafe-math-optimizations} is enabled
27479 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
27480 Note that while the throughput of the sequence is higher than the throughput
27481 of the non-reciprocal instruction, the precision of the sequence can be
27482 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
27484 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
27485 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
27486 combination), and doesn't need @option{-mrecip}.
27488 Also note that GCC emits the above sequence with additional Newton-Raphson step
27489 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
27490 already with @option{-ffast-math} (or the above option combination), and
27491 doesn't need @option{-mrecip}.
27493 @item -mrecip=@var{opt}
27494 @opindex mrecip=opt
27495 This option controls which reciprocal estimate instructions
27496 may be used. @var{opt} is a comma-separated list of options, which may
27497 be preceded by a @samp{!} to invert the option:
27501 Enable all estimate instructions.
27504 Enable the default instructions, equivalent to @option{-mrecip}.
27507 Disable all estimate instructions, equivalent to @option{-mno-recip}.
27510 Enable the approximation for scalar division.
27513 Enable the approximation for vectorized division.
27516 Enable the approximation for scalar square root.
27519 Enable the approximation for vectorized square root.
27522 So, for example, @option{-mrecip=all,!sqrt} enables
27523 all of the reciprocal approximations, except for square root.
27525 @item -mveclibabi=@var{type}
27526 @opindex mveclibabi
27527 Specifies the ABI type to use for vectorizing intrinsics using an
27528 external library. Supported values for @var{type} are @samp{svml}
27529 for the Intel short
27530 vector math library and @samp{acml} for the AMD math core library.
27531 To use this option, both @option{-ftree-vectorize} and
27532 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
27533 ABI-compatible library must be specified at link time.
27535 GCC currently emits calls to @code{vmldExp2},
27536 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
27537 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
27538 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
27539 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
27540 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
27541 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
27542 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
27543 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
27544 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
27545 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
27546 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
27547 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
27548 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
27549 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
27550 when @option{-mveclibabi=acml} is used.
27552 @item -mabi=@var{name}
27554 Generate code for the specified calling convention. Permissible values
27555 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
27556 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
27557 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
27558 You can control this behavior for specific functions by
27559 using the function attributes @code{ms_abi} and @code{sysv_abi}.
27560 @xref{Function Attributes}.
27562 @item -mforce-indirect-call
27563 @opindex mforce-indirect-call
27564 Force all calls to functions to be indirect. This is useful
27565 when using Intel Processor Trace where it generates more precise timing
27566 information for function calls.
27568 @item -mcall-ms2sysv-xlogues
27569 @opindex mcall-ms2sysv-xlogues
27570 @opindex mno-call-ms2sysv-xlogues
27571 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
27572 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
27573 default, the code for saving and restoring these registers is emitted inline,
27574 resulting in fairly lengthy prologues and epilogues. Using
27575 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
27576 use stubs in the static portion of libgcc to perform these saves and restores,
27577 thus reducing function size at the cost of a few extra instructions.
27579 @item -mtls-dialect=@var{type}
27580 @opindex mtls-dialect
27581 Generate code to access thread-local storage using the @samp{gnu} or
27582 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
27583 @samp{gnu2} is more efficient, but it may add compile- and run-time
27584 requirements that cannot be satisfied on all systems.
27587 @itemx -mno-push-args
27588 @opindex mpush-args
27589 @opindex mno-push-args
27590 Use PUSH operations to store outgoing parameters. This method is shorter
27591 and usually equally fast as method using SUB/MOV operations and is enabled
27592 by default. In some cases disabling it may improve performance because of
27593 improved scheduling and reduced dependencies.
27595 @item -maccumulate-outgoing-args
27596 @opindex maccumulate-outgoing-args
27597 If enabled, the maximum amount of space required for outgoing arguments is
27598 computed in the function prologue. This is faster on most modern CPUs
27599 because of reduced dependencies, improved scheduling and reduced stack usage
27600 when the preferred stack boundary is not equal to 2. The drawback is a notable
27601 increase in code size. This switch implies @option{-mno-push-args}.
27605 Support thread-safe exception handling on MinGW. Programs that rely
27606 on thread-safe exception handling must compile and link all code with the
27607 @option{-mthreads} option. When compiling, @option{-mthreads} defines
27608 @option{-D_MT}; when linking, it links in a special thread helper library
27609 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
27611 @item -mms-bitfields
27612 @itemx -mno-ms-bitfields
27613 @opindex mms-bitfields
27614 @opindex mno-ms-bitfields
27616 Enable/disable bit-field layout compatible with the native Microsoft
27619 If @code{packed} is used on a structure, or if bit-fields are used,
27620 it may be that the Microsoft ABI lays out the structure differently
27621 than the way GCC normally does. Particularly when moving packed
27622 data between functions compiled with GCC and the native Microsoft compiler
27623 (either via function call or as data in a file), it may be necessary to access
27626 This option is enabled by default for Microsoft Windows
27627 targets. This behavior can also be controlled locally by use of variable
27628 or type attributes. For more information, see @ref{x86 Variable Attributes}
27629 and @ref{x86 Type Attributes}.
27631 The Microsoft structure layout algorithm is fairly simple with the exception
27632 of the bit-field packing.
27633 The padding and alignment of members of structures and whether a bit-field
27634 can straddle a storage-unit boundary are determine by these rules:
27637 @item Structure members are stored sequentially in the order in which they are
27638 declared: the first member has the lowest memory address and the last member
27641 @item Every data object has an alignment requirement. The alignment requirement
27642 for all data except structures, unions, and arrays is either the size of the
27643 object or the current packing size (specified with either the
27644 @code{aligned} attribute or the @code{pack} pragma),
27645 whichever is less. For structures, unions, and arrays,
27646 the alignment requirement is the largest alignment requirement of its members.
27647 Every object is allocated an offset so that:
27650 offset % alignment_requirement == 0
27653 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
27654 unit if the integral types are the same size and if the next bit-field fits
27655 into the current allocation unit without crossing the boundary imposed by the
27656 common alignment requirements of the bit-fields.
27659 MSVC interprets zero-length bit-fields in the following ways:
27662 @item If a zero-length bit-field is inserted between two bit-fields that
27663 are normally coalesced, the bit-fields are not coalesced.
27670 unsigned long bf_1 : 12;
27672 unsigned long bf_2 : 12;
27677 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
27678 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
27680 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
27681 alignment of the zero-length bit-field is greater than the member that follows it,
27682 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
27703 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
27704 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
27705 bit-field does not affect the alignment of @code{bar} or, as a result, the size
27708 Taking this into account, it is important to note the following:
27711 @item If a zero-length bit-field follows a normal bit-field, the type of the
27712 zero-length bit-field may affect the alignment of the structure as whole. For
27713 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
27714 normal bit-field, and is of type short.
27716 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
27717 still affect the alignment of the structure:
27728 Here, @code{t4} takes up 4 bytes.
27731 @item Zero-length bit-fields following non-bit-field members are ignored:
27743 Here, @code{t5} takes up 2 bytes.
27747 @item -mno-align-stringops
27748 @opindex mno-align-stringops
27749 Do not align the destination of inlined string operations. This switch reduces
27750 code size and improves performance in case the destination is already aligned,
27751 but GCC doesn't know about it.
27753 @item -minline-all-stringops
27754 @opindex minline-all-stringops
27755 By default GCC inlines string operations only when the destination is
27756 known to be aligned to least a 4-byte boundary.
27757 This enables more inlining and increases code
27758 size, but may improve performance of code that depends on fast
27759 @code{memcpy}, @code{strlen},
27760 and @code{memset} for short lengths.
27762 @item -minline-stringops-dynamically
27763 @opindex minline-stringops-dynamically
27764 For string operations of unknown size, use run-time checks with
27765 inline code for small blocks and a library call for large blocks.
27767 @item -mstringop-strategy=@var{alg}
27768 @opindex mstringop-strategy=@var{alg}
27769 Override the internal decision heuristic for the particular algorithm to use
27770 for inlining string operations. The allowed values for @var{alg} are:
27776 Expand using i386 @code{rep} prefix of the specified size.
27780 @itemx unrolled_loop
27781 Expand into an inline loop.
27784 Always use a library call.
27787 @item -mmemcpy-strategy=@var{strategy}
27788 @opindex mmemcpy-strategy=@var{strategy}
27789 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
27790 should be inlined and what inline algorithm to use when the expected size
27791 of the copy operation is known. @var{strategy}
27792 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
27793 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
27794 the max byte size with which inline algorithm @var{alg} is allowed. For the last
27795 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
27796 in the list must be specified in increasing order. The minimal byte size for
27797 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
27800 @item -mmemset-strategy=@var{strategy}
27801 @opindex mmemset-strategy=@var{strategy}
27802 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
27803 @code{__builtin_memset} expansion.
27805 @item -momit-leaf-frame-pointer
27806 @opindex momit-leaf-frame-pointer
27807 Don't keep the frame pointer in a register for leaf functions. This
27808 avoids the instructions to save, set up, and restore frame pointers and
27809 makes an extra register available in leaf functions. The option
27810 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
27811 which might make debugging harder.
27813 @item -mtls-direct-seg-refs
27814 @itemx -mno-tls-direct-seg-refs
27815 @opindex mtls-direct-seg-refs
27816 Controls whether TLS variables may be accessed with offsets from the
27817 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
27818 or whether the thread base pointer must be added. Whether or not this
27819 is valid depends on the operating system, and whether it maps the
27820 segment to cover the entire TLS area.
27822 For systems that use the GNU C Library, the default is on.
27825 @itemx -mno-sse2avx
27827 Specify that the assembler should encode SSE instructions with VEX
27828 prefix. The option @option{-mavx} turns this on by default.
27833 If profiling is active (@option{-pg}), put the profiling
27834 counter call before the prologue.
27835 Note: On x86 architectures the attribute @code{ms_hook_prologue}
27836 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
27838 @item -mrecord-mcount
27839 @itemx -mno-record-mcount
27840 @opindex mrecord-mcount
27841 If profiling is active (@option{-pg}), generate a __mcount_loc section
27842 that contains pointers to each profiling call. This is useful for
27843 automatically patching and out calls.
27846 @itemx -mno-nop-mcount
27847 @opindex mnop-mcount
27848 If profiling is active (@option{-pg}), generate the calls to
27849 the profiling functions as NOPs. This is useful when they
27850 should be patched in later dynamically. This is likely only
27851 useful together with @option{-mrecord-mcount}.
27853 @item -mskip-rax-setup
27854 @itemx -mno-skip-rax-setup
27855 @opindex mskip-rax-setup
27856 When generating code for the x86-64 architecture with SSE extensions
27857 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
27858 register when there are no variable arguments passed in vector registers.
27860 @strong{Warning:} Since RAX register is used to avoid unnecessarily
27861 saving vector registers on stack when passing variable arguments, the
27862 impacts of this option are callees may waste some stack space,
27863 misbehave or jump to a random location. GCC 4.4 or newer don't have
27864 those issues, regardless the RAX register value.
27867 @itemx -mno-8bit-idiv
27868 @opindex m8bit-idiv
27869 On some processors, like Intel Atom, 8-bit unsigned integer divide is
27870 much faster than 32-bit/64-bit integer divide. This option generates a
27871 run-time check. If both dividend and divisor are within range of 0
27872 to 255, 8-bit unsigned integer divide is used instead of
27873 32-bit/64-bit integer divide.
27875 @item -mavx256-split-unaligned-load
27876 @itemx -mavx256-split-unaligned-store
27877 @opindex mavx256-split-unaligned-load
27878 @opindex mavx256-split-unaligned-store
27879 Split 32-byte AVX unaligned load and store.
27881 @item -mstack-protector-guard=@var{guard}
27882 @itemx -mstack-protector-guard-reg=@var{reg}
27883 @itemx -mstack-protector-guard-offset=@var{offset}
27884 @opindex mstack-protector-guard
27885 @opindex mstack-protector-guard-reg
27886 @opindex mstack-protector-guard-offset
27887 Generate stack protection code using canary at @var{guard}. Supported
27888 locations are @samp{global} for global canary or @samp{tls} for per-thread
27889 canary in the TLS block (the default). This option has effect only when
27890 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
27892 With the latter choice the options
27893 @option{-mstack-protector-guard-reg=@var{reg}} and
27894 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
27895 which segment register (@code{%fs} or @code{%gs}) to use as base register
27896 for reading the canary, and from what offset from that base register.
27897 The default for those is as specified in the relevant ABI.
27899 @item -mmitigate-rop
27900 @opindex mmitigate-rop
27901 Try to avoid generating code sequences that contain unintended return
27902 opcodes, to mitigate against certain forms of attack. At the moment,
27903 this option is limited in what it can do and should not be relied
27904 on to provide serious protection.
27906 @item -mgeneral-regs-only
27907 @opindex mgeneral-regs-only
27908 Generate code that uses only the general-purpose registers. This
27909 prevents the compiler from using floating-point, vector, mask and bound
27912 @item -mindirect-branch=@var{choice}
27913 @opindex -mindirect-branch
27914 Convert indirect call and jump with @var{choice}. The default is
27915 @samp{keep}, which keeps indirect call and jump unmodified.
27916 @samp{thunk} converts indirect call and jump to call and return thunk.
27917 @samp{thunk-inline} converts indirect call and jump to inlined call
27918 and return thunk. @samp{thunk-extern} converts indirect call and jump
27919 to external call and return thunk provided in a separate object file.
27920 You can control this behavior for a specific function by using the
27921 function attribute @code{indirect_branch}. @xref{Function Attributes}.
27923 Note that @option{-mcmodel=large} is incompatible with
27924 @option{-mindirect-branch=thunk} and
27925 @option{-mindirect-branch=thunk-extern} since the thunk function may
27926 not be reachable in the large code model.
27928 Note that @option{-mindirect-branch=thunk-extern} is incompatible with
27929 @option{-fcf-protection=branch} and @option{-fcheck-pointer-bounds}
27930 since the external thunk can not be modified to disable control-flow
27933 @item -mfunction-return=@var{choice}
27934 @opindex -mfunction-return
27935 Convert function return with @var{choice}. The default is @samp{keep},
27936 which keeps function return unmodified. @samp{thunk} converts function
27937 return to call and return thunk. @samp{thunk-inline} converts function
27938 return to inlined call and return thunk. @samp{thunk-extern} converts
27939 function return to external call and return thunk provided in a separate
27940 object file. You can control this behavior for a specific function by
27941 using the function attribute @code{function_return}.
27942 @xref{Function Attributes}.
27944 Note that @option{-mcmodel=large} is incompatible with
27945 @option{-mfunction-return=thunk} and
27946 @option{-mfunction-return=thunk-extern} since the thunk function may
27947 not be reachable in the large code model.
27950 @item -mindirect-branch-register
27951 @opindex -mindirect-branch-register
27952 Force indirect call and jump via register.
27956 These @samp{-m} switches are supported in addition to the above
27957 on x86-64 processors in 64-bit environments.
27970 Generate code for a 16-bit, 32-bit or 64-bit environment.
27971 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
27973 generates code that runs on any i386 system.
27975 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
27976 types to 64 bits, and generates code for the x86-64 architecture.
27977 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
27978 and @option{-mdynamic-no-pic} options.
27980 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
27982 generates code for the x86-64 architecture.
27984 The @option{-m16} option is the same as @option{-m32}, except for that
27985 it outputs the @code{.code16gcc} assembly directive at the beginning of
27986 the assembly output so that the binary can run in 16-bit mode.
27988 The @option{-miamcu} option generates code which conforms to Intel MCU
27989 psABI. It requires the @option{-m32} option to be turned on.
27991 @item -mno-red-zone
27992 @opindex mno-red-zone
27993 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
27994 by the x86-64 ABI; it is a 128-byte area beyond the location of the
27995 stack pointer that is not modified by signal or interrupt handlers
27996 and therefore can be used for temporary data without adjusting the stack
27997 pointer. The flag @option{-mno-red-zone} disables this red zone.
27999 @item -mcmodel=small
28000 @opindex mcmodel=small
28001 Generate code for the small code model: the program and its symbols must
28002 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
28003 Programs can be statically or dynamically linked. This is the default
28006 @item -mcmodel=kernel
28007 @opindex mcmodel=kernel
28008 Generate code for the kernel code model. The kernel runs in the
28009 negative 2 GB of the address space.
28010 This model has to be used for Linux kernel code.
28012 @item -mcmodel=medium
28013 @opindex mcmodel=medium
28014 Generate code for the medium model: the program is linked in the lower 2
28015 GB of the address space. Small symbols are also placed there. Symbols
28016 with sizes larger than @option{-mlarge-data-threshold} are put into
28017 large data or BSS sections and can be located above 2GB. Programs can
28018 be statically or dynamically linked.
28020 @item -mcmodel=large
28021 @opindex mcmodel=large
28022 Generate code for the large model. This model makes no assumptions
28023 about addresses and sizes of sections.
28025 @item -maddress-mode=long
28026 @opindex maddress-mode=long
28027 Generate code for long address mode. This is only supported for 64-bit
28028 and x32 environments. It is the default address mode for 64-bit
28031 @item -maddress-mode=short
28032 @opindex maddress-mode=short
28033 Generate code for short address mode. This is only supported for 32-bit
28034 and x32 environments. It is the default address mode for 32-bit and
28038 @node x86 Windows Options
28039 @subsection x86 Windows Options
28040 @cindex x86 Windows Options
28041 @cindex Windows Options for x86
28043 These additional options are available for Microsoft Windows targets:
28049 specifies that a console application is to be generated, by
28050 instructing the linker to set the PE header subsystem type
28051 required for console applications.
28052 This option is available for Cygwin and MinGW targets and is
28053 enabled by default on those targets.
28057 This option is available for Cygwin and MinGW targets. It
28058 specifies that a DLL---a dynamic link library---is to be
28059 generated, enabling the selection of the required runtime
28060 startup object and entry point.
28062 @item -mnop-fun-dllimport
28063 @opindex mnop-fun-dllimport
28064 This option is available for Cygwin and MinGW targets. It
28065 specifies that the @code{dllimport} attribute should be ignored.
28069 This option is available for MinGW targets. It specifies
28070 that MinGW-specific thread support is to be used.
28074 This option is available for MinGW-w64 targets. It causes
28075 the @code{UNICODE} preprocessor macro to be predefined, and
28076 chooses Unicode-capable runtime startup code.
28080 This option is available for Cygwin and MinGW targets. It
28081 specifies that the typical Microsoft Windows predefined macros are to
28082 be set in the pre-processor, but does not influence the choice
28083 of runtime library/startup code.
28087 This option is available for Cygwin and MinGW targets. It
28088 specifies that a GUI application is to be generated by
28089 instructing the linker to set the PE header subsystem type
28092 @item -fno-set-stack-executable
28093 @opindex fno-set-stack-executable
28094 This option is available for MinGW targets. It specifies that
28095 the executable flag for the stack used by nested functions isn't
28096 set. This is necessary for binaries running in kernel mode of
28097 Microsoft Windows, as there the User32 API, which is used to set executable
28098 privileges, isn't available.
28100 @item -fwritable-relocated-rdata
28101 @opindex fno-writable-relocated-rdata
28102 This option is available for MinGW and Cygwin targets. It specifies
28103 that relocated-data in read-only section is put into the @code{.data}
28104 section. This is a necessary for older runtimes not supporting
28105 modification of @code{.rdata} sections for pseudo-relocation.
28107 @item -mpe-aligned-commons
28108 @opindex mpe-aligned-commons
28109 This option is available for Cygwin and MinGW targets. It
28110 specifies that the GNU extension to the PE file format that
28111 permits the correct alignment of COMMON variables should be
28112 used when generating code. It is enabled by default if
28113 GCC detects that the target assembler found during configuration
28114 supports the feature.
28117 See also under @ref{x86 Options} for standard options.
28119 @node Xstormy16 Options
28120 @subsection Xstormy16 Options
28121 @cindex Xstormy16 Options
28123 These options are defined for Xstormy16:
28128 Choose startup files and linker script suitable for the simulator.
28131 @node Xtensa Options
28132 @subsection Xtensa Options
28133 @cindex Xtensa Options
28135 These options are supported for Xtensa targets:
28139 @itemx -mno-const16
28141 @opindex mno-const16
28142 Enable or disable use of @code{CONST16} instructions for loading
28143 constant values. The @code{CONST16} instruction is currently not a
28144 standard option from Tensilica. When enabled, @code{CONST16}
28145 instructions are always used in place of the standard @code{L32R}
28146 instructions. The use of @code{CONST16} is enabled by default only if
28147 the @code{L32R} instruction is not available.
28150 @itemx -mno-fused-madd
28151 @opindex mfused-madd
28152 @opindex mno-fused-madd
28153 Enable or disable use of fused multiply/add and multiply/subtract
28154 instructions in the floating-point option. This has no effect if the
28155 floating-point option is not also enabled. Disabling fused multiply/add
28156 and multiply/subtract instructions forces the compiler to use separate
28157 instructions for the multiply and add/subtract operations. This may be
28158 desirable in some cases where strict IEEE 754-compliant results are
28159 required: the fused multiply add/subtract instructions do not round the
28160 intermediate result, thereby producing results with @emph{more} bits of
28161 precision than specified by the IEEE standard. Disabling fused multiply
28162 add/subtract instructions also ensures that the program output is not
28163 sensitive to the compiler's ability to combine multiply and add/subtract
28166 @item -mserialize-volatile
28167 @itemx -mno-serialize-volatile
28168 @opindex mserialize-volatile
28169 @opindex mno-serialize-volatile
28170 When this option is enabled, GCC inserts @code{MEMW} instructions before
28171 @code{volatile} memory references to guarantee sequential consistency.
28172 The default is @option{-mserialize-volatile}. Use
28173 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
28175 @item -mforce-no-pic
28176 @opindex mforce-no-pic
28177 For targets, like GNU/Linux, where all user-mode Xtensa code must be
28178 position-independent code (PIC), this option disables PIC for compiling
28181 @item -mtext-section-literals
28182 @itemx -mno-text-section-literals
28183 @opindex mtext-section-literals
28184 @opindex mno-text-section-literals
28185 These options control the treatment of literal pools. The default is
28186 @option{-mno-text-section-literals}, which places literals in a separate
28187 section in the output file. This allows the literal pool to be placed
28188 in a data RAM/ROM, and it also allows the linker to combine literal
28189 pools from separate object files to remove redundant literals and
28190 improve code size. With @option{-mtext-section-literals}, the literals
28191 are interspersed in the text section in order to keep them as close as
28192 possible to their references. This may be necessary for large assembly
28193 files. Literals for each function are placed right before that function.
28195 @item -mauto-litpools
28196 @itemx -mno-auto-litpools
28197 @opindex mauto-litpools
28198 @opindex mno-auto-litpools
28199 These options control the treatment of literal pools. The default is
28200 @option{-mno-auto-litpools}, which places literals in a separate
28201 section in the output file unless @option{-mtext-section-literals} is
28202 used. With @option{-mauto-litpools} the literals are interspersed in
28203 the text section by the assembler. Compiler does not produce explicit
28204 @code{.literal} directives and loads literals into registers with
28205 @code{MOVI} instructions instead of @code{L32R} to let the assembler
28206 do relaxation and place literals as necessary. This option allows
28207 assembler to create several literal pools per function and assemble
28208 very big functions, which may not be possible with
28209 @option{-mtext-section-literals}.
28211 @item -mtarget-align
28212 @itemx -mno-target-align
28213 @opindex mtarget-align
28214 @opindex mno-target-align
28215 When this option is enabled, GCC instructs the assembler to
28216 automatically align instructions to reduce branch penalties at the
28217 expense of some code density. The assembler attempts to widen density
28218 instructions to align branch targets and the instructions following call
28219 instructions. If there are not enough preceding safe density
28220 instructions to align a target, no widening is performed. The
28221 default is @option{-mtarget-align}. These options do not affect the
28222 treatment of auto-aligned instructions like @code{LOOP}, which the
28223 assembler always aligns, either by widening density instructions or
28224 by inserting NOP instructions.
28227 @itemx -mno-longcalls
28228 @opindex mlongcalls
28229 @opindex mno-longcalls
28230 When this option is enabled, GCC instructs the assembler to translate
28231 direct calls to indirect calls unless it can determine that the target
28232 of a direct call is in the range allowed by the call instruction. This
28233 translation typically occurs for calls to functions in other source
28234 files. Specifically, the assembler translates a direct @code{CALL}
28235 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
28236 The default is @option{-mno-longcalls}. This option should be used in
28237 programs where the call target can potentially be out of range. This
28238 option is implemented in the assembler, not the compiler, so the
28239 assembly code generated by GCC still shows direct call
28240 instructions---look at the disassembled object code to see the actual
28241 instructions. Note that the assembler uses an indirect call for
28242 every cross-file call, not just those that really are out of range.
28245 @node zSeries Options
28246 @subsection zSeries Options
28247 @cindex zSeries options
28249 These are listed under @xref{S/390 and zSeries Options}.
28255 @section Specifying Subprocesses and the Switches to Pass to Them
28258 @command{gcc} is a driver program. It performs its job by invoking a
28259 sequence of other programs to do the work of compiling, assembling and
28260 linking. GCC interprets its command-line parameters and uses these to
28261 deduce which programs it should invoke, and which command-line options
28262 it ought to place on their command lines. This behavior is controlled
28263 by @dfn{spec strings}. In most cases there is one spec string for each
28264 program that GCC can invoke, but a few programs have multiple spec
28265 strings to control their behavior. The spec strings built into GCC can
28266 be overridden by using the @option{-specs=} command-line switch to specify
28269 @dfn{Spec files} are plain-text files that are used to construct spec
28270 strings. They consist of a sequence of directives separated by blank
28271 lines. The type of directive is determined by the first non-whitespace
28272 character on the line, which can be one of the following:
28275 @item %@var{command}
28276 Issues a @var{command} to the spec file processor. The commands that can
28280 @item %include <@var{file}>
28281 @cindex @code{%include}
28282 Search for @var{file} and insert its text at the current point in the
28285 @item %include_noerr <@var{file}>
28286 @cindex @code{%include_noerr}
28287 Just like @samp{%include}, but do not generate an error message if the include
28288 file cannot be found.
28290 @item %rename @var{old_name} @var{new_name}
28291 @cindex @code{%rename}
28292 Rename the spec string @var{old_name} to @var{new_name}.
28296 @item *[@var{spec_name}]:
28297 This tells the compiler to create, override or delete the named spec
28298 string. All lines after this directive up to the next directive or
28299 blank line are considered to be the text for the spec string. If this
28300 results in an empty string then the spec is deleted. (Or, if the
28301 spec did not exist, then nothing happens.) Otherwise, if the spec
28302 does not currently exist a new spec is created. If the spec does
28303 exist then its contents are overridden by the text of this
28304 directive, unless the first character of that text is the @samp{+}
28305 character, in which case the text is appended to the spec.
28307 @item [@var{suffix}]:
28308 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
28309 and up to the next directive or blank line are considered to make up the
28310 spec string for the indicated suffix. When the compiler encounters an
28311 input file with the named suffix, it processes the spec string in
28312 order to work out how to compile that file. For example:
28316 z-compile -input %i
28319 This says that any input file whose name ends in @samp{.ZZ} should be
28320 passed to the program @samp{z-compile}, which should be invoked with the
28321 command-line switch @option{-input} and with the result of performing the
28322 @samp{%i} substitution. (See below.)
28324 As an alternative to providing a spec string, the text following a
28325 suffix directive can be one of the following:
28328 @item @@@var{language}
28329 This says that the suffix is an alias for a known @var{language}. This is
28330 similar to using the @option{-x} command-line switch to GCC to specify a
28331 language explicitly. For example:
28338 Says that .ZZ files are, in fact, C++ source files.
28341 This causes an error messages saying:
28344 @var{name} compiler not installed on this system.
28348 GCC already has an extensive list of suffixes built into it.
28349 This directive adds an entry to the end of the list of suffixes, but
28350 since the list is searched from the end backwards, it is effectively
28351 possible to override earlier entries using this technique.
28355 GCC has the following spec strings built into it. Spec files can
28356 override these strings or create their own. Note that individual
28357 targets can also add their own spec strings to this list.
28360 asm Options to pass to the assembler
28361 asm_final Options to pass to the assembler post-processor
28362 cpp Options to pass to the C preprocessor
28363 cc1 Options to pass to the C compiler
28364 cc1plus Options to pass to the C++ compiler
28365 endfile Object files to include at the end of the link
28366 link Options to pass to the linker
28367 lib Libraries to include on the command line to the linker
28368 libgcc Decides which GCC support library to pass to the linker
28369 linker Sets the name of the linker
28370 predefines Defines to be passed to the C preprocessor
28371 signed_char Defines to pass to CPP to say whether @code{char} is signed
28373 startfile Object files to include at the start of the link
28376 Here is a small example of a spec file:
28379 %rename lib old_lib
28382 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
28385 This example renames the spec called @samp{lib} to @samp{old_lib} and
28386 then overrides the previous definition of @samp{lib} with a new one.
28387 The new definition adds in some extra command-line options before
28388 including the text of the old definition.
28390 @dfn{Spec strings} are a list of command-line options to be passed to their
28391 corresponding program. In addition, the spec strings can contain
28392 @samp{%}-prefixed sequences to substitute variable text or to
28393 conditionally insert text into the command line. Using these constructs
28394 it is possible to generate quite complex command lines.
28396 Here is a table of all defined @samp{%}-sequences for spec
28397 strings. Note that spaces are not generated automatically around the
28398 results of expanding these sequences. Therefore you can concatenate them
28399 together or combine them with constant text in a single argument.
28403 Substitute one @samp{%} into the program name or argument.
28406 Substitute the name of the input file being processed.
28409 Substitute the basename of the input file being processed.
28410 This is the substring up to (and not including) the last period
28411 and not including the directory.
28414 This is the same as @samp{%b}, but include the file suffix (text after
28418 Marks the argument containing or following the @samp{%d} as a
28419 temporary file name, so that that file is deleted if GCC exits
28420 successfully. Unlike @samp{%g}, this contributes no text to the
28423 @item %g@var{suffix}
28424 Substitute a file name that has suffix @var{suffix} and is chosen
28425 once per compilation, and mark the argument in the same way as
28426 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
28427 name is now chosen in a way that is hard to predict even when previously
28428 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
28429 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
28430 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
28431 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
28432 was simply substituted with a file name chosen once per compilation,
28433 without regard to any appended suffix (which was therefore treated
28434 just like ordinary text), making such attacks more likely to succeed.
28436 @item %u@var{suffix}
28437 Like @samp{%g}, but generates a new temporary file name
28438 each time it appears instead of once per compilation.
28440 @item %U@var{suffix}
28441 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
28442 new one if there is no such last file name. In the absence of any
28443 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
28444 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
28445 involves the generation of two distinct file names, one
28446 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
28447 simply substituted with a file name chosen for the previous @samp{%u},
28448 without regard to any appended suffix.
28450 @item %j@var{suffix}
28451 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
28452 writable, and if @option{-save-temps} is not used;
28453 otherwise, substitute the name
28454 of a temporary file, just like @samp{%u}. This temporary file is not
28455 meant for communication between processes, but rather as a junk
28456 disposal mechanism.
28458 @item %|@var{suffix}
28459 @itemx %m@var{suffix}
28460 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
28461 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
28462 all. These are the two most common ways to instruct a program that it
28463 should read from standard input or write to standard output. If you
28464 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
28465 construct: see for example @file{f/lang-specs.h}.
28467 @item %.@var{SUFFIX}
28468 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
28469 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
28470 terminated by the next space or %.
28473 Marks the argument containing or following the @samp{%w} as the
28474 designated output file of this compilation. This puts the argument
28475 into the sequence of arguments that @samp{%o} substitutes.
28478 Substitutes the names of all the output files, with spaces
28479 automatically placed around them. You should write spaces
28480 around the @samp{%o} as well or the results are undefined.
28481 @samp{%o} is for use in the specs for running the linker.
28482 Input files whose names have no recognized suffix are not compiled
28483 at all, but they are included among the output files, so they are
28487 Substitutes the suffix for object files. Note that this is
28488 handled specially when it immediately follows @samp{%g, %u, or %U},
28489 because of the need for those to form complete file names. The
28490 handling is such that @samp{%O} is treated exactly as if it had already
28491 been substituted, except that @samp{%g, %u, and %U} do not currently
28492 support additional @var{suffix} characters following @samp{%O} as they do
28493 following, for example, @samp{.o}.
28496 Substitutes the standard macro predefinitions for the
28497 current target machine. Use this when running @command{cpp}.
28500 Like @samp{%p}, but puts @samp{__} before and after the name of each
28501 predefined macro, except for macros that start with @samp{__} or with
28502 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
28506 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
28507 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
28508 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
28509 and @option{-imultilib} as necessary.
28512 Current argument is the name of a library or startup file of some sort.
28513 Search for that file in a standard list of directories and substitute
28514 the full name found. The current working directory is included in the
28515 list of directories scanned.
28518 Current argument is the name of a linker script. Search for that file
28519 in the current list of directories to scan for libraries. If the file
28520 is located insert a @option{--script} option into the command line
28521 followed by the full path name found. If the file is not found then
28522 generate an error message. Note: the current working directory is not
28526 Print @var{str} as an error message. @var{str} is terminated by a newline.
28527 Use this when inconsistent options are detected.
28529 @item %(@var{name})
28530 Substitute the contents of spec string @var{name} at this point.
28532 @item %x@{@var{option}@}
28533 Accumulate an option for @samp{%X}.
28536 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
28540 Output the accumulated assembler options specified by @option{-Wa}.
28543 Output the accumulated preprocessor options specified by @option{-Wp}.
28546 Process the @code{asm} spec. This is used to compute the
28547 switches to be passed to the assembler.
28550 Process the @code{asm_final} spec. This is a spec string for
28551 passing switches to an assembler post-processor, if such a program is
28555 Process the @code{link} spec. This is the spec for computing the
28556 command line passed to the linker. Typically it makes use of the
28557 @samp{%L %G %S %D and %E} sequences.
28560 Dump out a @option{-L} option for each directory that GCC believes might
28561 contain startup files. If the target supports multilibs then the
28562 current multilib directory is prepended to each of these paths.
28565 Process the @code{lib} spec. This is a spec string for deciding which
28566 libraries are included on the command line to the linker.
28569 Process the @code{libgcc} spec. This is a spec string for deciding
28570 which GCC support library is included on the command line to the linker.
28573 Process the @code{startfile} spec. This is a spec for deciding which
28574 object files are the first ones passed to the linker. Typically
28575 this might be a file named @file{crt0.o}.
28578 Process the @code{endfile} spec. This is a spec string that specifies
28579 the last object files that are passed to the linker.
28582 Process the @code{cpp} spec. This is used to construct the arguments
28583 to be passed to the C preprocessor.
28586 Process the @code{cc1} spec. This is used to construct the options to be
28587 passed to the actual C compiler (@command{cc1}).
28590 Process the @code{cc1plus} spec. This is used to construct the options to be
28591 passed to the actual C++ compiler (@command{cc1plus}).
28594 Substitute the variable part of a matched option. See below.
28595 Note that each comma in the substituted string is replaced by
28599 Remove all occurrences of @code{-S} from the command line. Note---this
28600 command is position dependent. @samp{%} commands in the spec string
28601 before this one see @code{-S}, @samp{%} commands in the spec string
28602 after this one do not.
28604 @item %:@var{function}(@var{args})
28605 Call the named function @var{function}, passing it @var{args}.
28606 @var{args} is first processed as a nested spec string, then split
28607 into an argument vector in the usual fashion. The function returns
28608 a string which is processed as if it had appeared literally as part
28609 of the current spec.
28611 The following built-in spec functions are provided:
28614 @item @code{getenv}
28615 The @code{getenv} spec function takes two arguments: an environment
28616 variable name and a string. If the environment variable is not
28617 defined, a fatal error is issued. Otherwise, the return value is the
28618 value of the environment variable concatenated with the string. For
28619 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
28622 %:getenv(TOPDIR /include)
28625 expands to @file{/path/to/top/include}.
28627 @item @code{if-exists}
28628 The @code{if-exists} spec function takes one argument, an absolute
28629 pathname to a file. If the file exists, @code{if-exists} returns the
28630 pathname. Here is a small example of its usage:
28634 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
28637 @item @code{if-exists-else}
28638 The @code{if-exists-else} spec function is similar to the @code{if-exists}
28639 spec function, except that it takes two arguments. The first argument is
28640 an absolute pathname to a file. If the file exists, @code{if-exists-else}
28641 returns the pathname. If it does not exist, it returns the second argument.
28642 This way, @code{if-exists-else} can be used to select one file or another,
28643 based on the existence of the first. Here is a small example of its usage:
28647 crt0%O%s %:if-exists(crti%O%s) \
28648 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
28651 @item @code{replace-outfile}
28652 The @code{replace-outfile} spec function takes two arguments. It looks for the
28653 first argument in the outfiles array and replaces it with the second argument. Here
28654 is a small example of its usage:
28657 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
28660 @item @code{remove-outfile}
28661 The @code{remove-outfile} spec function takes one argument. It looks for the
28662 first argument in the outfiles array and removes it. Here is a small example
28666 %:remove-outfile(-lm)
28669 @item @code{pass-through-libs}
28670 The @code{pass-through-libs} spec function takes any number of arguments. It
28671 finds any @option{-l} options and any non-options ending in @file{.a} (which it
28672 assumes are the names of linker input library archive files) and returns a
28673 result containing all the found arguments each prepended by
28674 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
28675 intended to be passed to the LTO linker plugin.
28678 %:pass-through-libs(%G %L %G)
28681 @item @code{print-asm-header}
28682 The @code{print-asm-header} function takes no arguments and simply
28683 prints a banner like:
28689 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
28692 It is used to separate compiler options from assembler options
28693 in the @option{--target-help} output.
28697 Substitutes the @code{-S} switch, if that switch is given to GCC@.
28698 If that switch is not specified, this substitutes nothing. Note that
28699 the leading dash is omitted when specifying this option, and it is
28700 automatically inserted if the substitution is performed. Thus the spec
28701 string @samp{%@{foo@}} matches the command-line option @option{-foo}
28702 and outputs the command-line option @option{-foo}.
28705 Like %@{@code{S}@} but mark last argument supplied within as a file to be
28706 deleted on failure.
28709 Substitutes all the switches specified to GCC whose names start
28710 with @code{-S}, but which also take an argument. This is used for
28711 switches like @option{-o}, @option{-D}, @option{-I}, etc.
28712 GCC considers @option{-o foo} as being
28713 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
28714 text, including the space. Thus two arguments are generated.
28717 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
28718 (the order of @code{S} and @code{T} in the spec is not significant).
28719 There can be any number of ampersand-separated variables; for each the
28720 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
28723 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
28726 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
28729 Substitutes @code{X} if one or more switches whose names start with
28730 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
28731 once, no matter how many such switches appeared. However, if @code{%*}
28732 appears somewhere in @code{X}, then @code{X} is substituted once
28733 for each matching switch, with the @code{%*} replaced by the part of
28734 that switch matching the @code{*}.
28736 If @code{%*} appears as the last part of a spec sequence then a space
28737 is added after the end of the last substitution. If there is more
28738 text in the sequence, however, then a space is not generated. This
28739 allows the @code{%*} substitution to be used as part of a larger
28740 string. For example, a spec string like this:
28743 %@{mcu=*:--script=%*/memory.ld@}
28747 when matching an option like @option{-mcu=newchip} produces:
28750 --script=newchip/memory.ld
28754 Substitutes @code{X}, if processing a file with suffix @code{S}.
28757 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
28760 Substitutes @code{X}, if processing a file for language @code{S}.
28763 Substitutes @code{X}, if not processing a file for language @code{S}.
28766 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
28767 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
28768 @code{*} sequences as well, although they have a stronger binding than
28769 the @samp{|}. If @code{%*} appears in @code{X}, all of the
28770 alternatives must be starred, and only the first matching alternative
28773 For example, a spec string like this:
28776 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
28780 outputs the following command-line options from the following input
28781 command-line options:
28786 -d fred.c -foo -baz -boggle
28787 -d jim.d -bar -baz -boggle
28790 @item %@{S:X; T:Y; :D@}
28792 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
28793 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
28794 be as many clauses as you need. This may be combined with @code{.},
28795 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
28800 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
28801 or similar construct can use a backslash to ignore the special meaning
28802 of the character following it, thus allowing literal matching of a
28803 character that is otherwise specially treated. For example,
28804 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
28805 @option{-std=iso9899:1999} option is given.
28807 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
28808 construct may contain other nested @samp{%} constructs or spaces, or
28809 even newlines. They are processed as usual, as described above.
28810 Trailing white space in @code{X} is ignored. White space may also
28811 appear anywhere on the left side of the colon in these constructs,
28812 except between @code{.} or @code{*} and the corresponding word.
28814 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
28815 handled specifically in these constructs. If another value of
28816 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
28817 @option{-W} switch is found later in the command line, the earlier
28818 switch value is ignored, except with @{@code{S}*@} where @code{S} is
28819 just one letter, which passes all matching options.
28821 The character @samp{|} at the beginning of the predicate text is used to
28822 indicate that a command should be piped to the following command, but
28823 only if @option{-pipe} is specified.
28825 It is built into GCC which switches take arguments and which do not.
28826 (You might think it would be useful to generalize this to allow each
28827 compiler's spec to say which switches take arguments. But this cannot
28828 be done in a consistent fashion. GCC cannot even decide which input
28829 files have been specified without knowing which switches take arguments,
28830 and it must know which input files to compile in order to tell which
28833 GCC also knows implicitly that arguments starting in @option{-l} are to be
28834 treated as compiler output files, and passed to the linker in their
28835 proper position among the other output files.
28837 @node Environment Variables
28838 @section Environment Variables Affecting GCC
28839 @cindex environment variables
28841 @c man begin ENVIRONMENT
28842 This section describes several environment variables that affect how GCC
28843 operates. Some of them work by specifying directories or prefixes to use
28844 when searching for various kinds of files. Some are used to specify other
28845 aspects of the compilation environment.
28847 Note that you can also specify places to search using options such as
28848 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
28849 take precedence over places specified using environment variables, which
28850 in turn take precedence over those specified by the configuration of GCC@.
28851 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
28852 GNU Compiler Collection (GCC) Internals}.
28857 @c @itemx LC_COLLATE
28859 @c @itemx LC_MONETARY
28860 @c @itemx LC_NUMERIC
28865 @c @findex LC_COLLATE
28866 @findex LC_MESSAGES
28867 @c @findex LC_MONETARY
28868 @c @findex LC_NUMERIC
28872 These environment variables control the way that GCC uses
28873 localization information which allows GCC to work with different
28874 national conventions. GCC inspects the locale categories
28875 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
28876 so. These locale categories can be set to any value supported by your
28877 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
28878 Kingdom encoded in UTF-8.
28880 The @env{LC_CTYPE} environment variable specifies character
28881 classification. GCC uses it to determine the character boundaries in
28882 a string; this is needed for some multibyte encodings that contain quote
28883 and escape characters that are otherwise interpreted as a string
28886 The @env{LC_MESSAGES} environment variable specifies the language to
28887 use in diagnostic messages.
28889 If the @env{LC_ALL} environment variable is set, it overrides the value
28890 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
28891 and @env{LC_MESSAGES} default to the value of the @env{LANG}
28892 environment variable. If none of these variables are set, GCC
28893 defaults to traditional C English behavior.
28897 If @env{TMPDIR} is set, it specifies the directory to use for temporary
28898 files. GCC uses temporary files to hold the output of one stage of
28899 compilation which is to be used as input to the next stage: for example,
28900 the output of the preprocessor, which is the input to the compiler
28903 @item GCC_COMPARE_DEBUG
28904 @findex GCC_COMPARE_DEBUG
28905 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
28906 @option{-fcompare-debug} to the compiler driver. See the documentation
28907 of this option for more details.
28909 @item GCC_EXEC_PREFIX
28910 @findex GCC_EXEC_PREFIX
28911 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
28912 names of the subprograms executed by the compiler. No slash is added
28913 when this prefix is combined with the name of a subprogram, but you can
28914 specify a prefix that ends with a slash if you wish.
28916 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
28917 an appropriate prefix to use based on the pathname it is invoked with.
28919 If GCC cannot find the subprogram using the specified prefix, it
28920 tries looking in the usual places for the subprogram.
28922 The default value of @env{GCC_EXEC_PREFIX} is
28923 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
28924 the installed compiler. In many cases @var{prefix} is the value
28925 of @code{prefix} when you ran the @file{configure} script.
28927 Other prefixes specified with @option{-B} take precedence over this prefix.
28929 This prefix is also used for finding files such as @file{crt0.o} that are
28932 In addition, the prefix is used in an unusual way in finding the
28933 directories to search for header files. For each of the standard
28934 directories whose name normally begins with @samp{/usr/local/lib/gcc}
28935 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
28936 replacing that beginning with the specified prefix to produce an
28937 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
28938 @file{foo/bar} just before it searches the standard directory
28939 @file{/usr/local/lib/bar}.
28940 If a standard directory begins with the configured
28941 @var{prefix} then the value of @var{prefix} is replaced by
28942 @env{GCC_EXEC_PREFIX} when looking for header files.
28944 @item COMPILER_PATH
28945 @findex COMPILER_PATH
28946 The value of @env{COMPILER_PATH} is a colon-separated list of
28947 directories, much like @env{PATH}. GCC tries the directories thus
28948 specified when searching for subprograms, if it cannot find the
28949 subprograms using @env{GCC_EXEC_PREFIX}.
28952 @findex LIBRARY_PATH
28953 The value of @env{LIBRARY_PATH} is a colon-separated list of
28954 directories, much like @env{PATH}. When configured as a native compiler,
28955 GCC tries the directories thus specified when searching for special
28956 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
28957 using GCC also uses these directories when searching for ordinary
28958 libraries for the @option{-l} option (but directories specified with
28959 @option{-L} come first).
28963 @cindex locale definition
28964 This variable is used to pass locale information to the compiler. One way in
28965 which this information is used is to determine the character set to be used
28966 when character literals, string literals and comments are parsed in C and C++.
28967 When the compiler is configured to allow multibyte characters,
28968 the following values for @env{LANG} are recognized:
28972 Recognize JIS characters.
28974 Recognize SJIS characters.
28976 Recognize EUCJP characters.
28979 If @env{LANG} is not defined, or if it has some other value, then the
28980 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
28981 recognize and translate multibyte characters.
28985 Some additional environment variables affect the behavior of the
28988 @include cppenv.texi
28992 @node Precompiled Headers
28993 @section Using Precompiled Headers
28994 @cindex precompiled headers
28995 @cindex speed of compilation
28997 Often large projects have many header files that are included in every
28998 source file. The time the compiler takes to process these header files
28999 over and over again can account for nearly all of the time required to
29000 build the project. To make builds faster, GCC allows you to
29001 @dfn{precompile} a header file.
29003 To create a precompiled header file, simply compile it as you would any
29004 other file, if necessary using the @option{-x} option to make the driver
29005 treat it as a C or C++ header file. You may want to use a
29006 tool like @command{make} to keep the precompiled header up-to-date when
29007 the headers it contains change.
29009 A precompiled header file is searched for when @code{#include} is
29010 seen in the compilation. As it searches for the included file
29011 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
29012 compiler looks for a precompiled header in each directory just before it
29013 looks for the include file in that directory. The name searched for is
29014 the name specified in the @code{#include} with @samp{.gch} appended. If
29015 the precompiled header file cannot be used, it is ignored.
29017 For instance, if you have @code{#include "all.h"}, and you have
29018 @file{all.h.gch} in the same directory as @file{all.h}, then the
29019 precompiled header file is used if possible, and the original
29020 header is used otherwise.
29022 Alternatively, you might decide to put the precompiled header file in a
29023 directory and use @option{-I} to ensure that directory is searched
29024 before (or instead of) the directory containing the original header.
29025 Then, if you want to check that the precompiled header file is always
29026 used, you can put a file of the same name as the original header in this
29027 directory containing an @code{#error} command.
29029 This also works with @option{-include}. So yet another way to use
29030 precompiled headers, good for projects not designed with precompiled
29031 header files in mind, is to simply take most of the header files used by
29032 a project, include them from another header file, precompile that header
29033 file, and @option{-include} the precompiled header. If the header files
29034 have guards against multiple inclusion, they are skipped because
29035 they've already been included (in the precompiled header).
29037 If you need to precompile the same header file for different
29038 languages, targets, or compiler options, you can instead make a
29039 @emph{directory} named like @file{all.h.gch}, and put each precompiled
29040 header in the directory, perhaps using @option{-o}. It doesn't matter
29041 what you call the files in the directory; every precompiled header in
29042 the directory is considered. The first precompiled header
29043 encountered in the directory that is valid for this compilation is
29044 used; they're searched in no particular order.
29046 There are many other possibilities, limited only by your imagination,
29047 good sense, and the constraints of your build system.
29049 A precompiled header file can be used only when these conditions apply:
29053 Only one precompiled header can be used in a particular compilation.
29056 A precompiled header cannot be used once the first C token is seen. You
29057 can have preprocessor directives before a precompiled header; you cannot
29058 include a precompiled header from inside another header.
29061 The precompiled header file must be produced for the same language as
29062 the current compilation. You cannot use a C precompiled header for a C++
29066 The precompiled header file must have been produced by the same compiler
29067 binary as the current compilation is using.
29070 Any macros defined before the precompiled header is included must
29071 either be defined in the same way as when the precompiled header was
29072 generated, or must not affect the precompiled header, which usually
29073 means that they don't appear in the precompiled header at all.
29075 The @option{-D} option is one way to define a macro before a
29076 precompiled header is included; using a @code{#define} can also do it.
29077 There are also some options that define macros implicitly, like
29078 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
29081 @item If debugging information is output when using the precompiled
29082 header, using @option{-g} or similar, the same kind of debugging information
29083 must have been output when building the precompiled header. However,
29084 a precompiled header built using @option{-g} can be used in a compilation
29085 when no debugging information is being output.
29087 @item The same @option{-m} options must generally be used when building
29088 and using the precompiled header. @xref{Submodel Options},
29089 for any cases where this rule is relaxed.
29091 @item Each of the following options must be the same when building and using
29092 the precompiled header:
29094 @gccoptlist{-fexceptions}
29097 Some other command-line options starting with @option{-f},
29098 @option{-p}, or @option{-O} must be defined in the same way as when
29099 the precompiled header was generated. At present, it's not clear
29100 which options are safe to change and which are not; the safest choice
29101 is to use exactly the same options when generating and using the
29102 precompiled header. The following are known to be safe:
29104 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
29105 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
29106 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
29111 For all of these except the last, the compiler automatically
29112 ignores the precompiled header if the conditions aren't met. If you
29113 find an option combination that doesn't work and doesn't cause the
29114 precompiled header to be ignored, please consider filing a bug report,
29117 If you do use differing options when generating and using the
29118 precompiled header, the actual behavior is a mixture of the
29119 behavior for the options. For instance, if you use @option{-g} to
29120 generate the precompiled header but not when using it, you may or may
29121 not get debugging information for routines in the precompiled header.