1 @c Copyright (C) 1988-2017 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{} 1988-2017 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
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} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
176 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
178 @item C Language Options
179 @xref{C Dialect Options,,Options Controlling C Dialect}.
180 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
181 -fpermitted-flt-eval-methods=@var{standard} @gol
182 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
183 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
184 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
185 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
186 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
187 -fsigned-bitfields -fsigned-char @gol
188 -funsigned-bitfields -funsigned-char}
190 @item C++ Language Options
191 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
192 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
193 -faligned-new=@var{n} -fargs-in-order=@var{n} -fcheck-new @gol
194 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
195 -ffriend-injection @gol
196 -fno-elide-constructors @gol
197 -fno-enforce-eh-specs @gol
198 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
199 -fno-implicit-templates @gol
200 -fno-implicit-inline-templates @gol
201 -fno-implement-inlines -fms-extensions @gol
202 -fnew-inheriting-ctors @gol
203 -fnew-ttp-matching @gol
204 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
205 -fno-optional-diags -fpermissive @gol
206 -fno-pretty-templates @gol
207 -frepo -fno-rtti -fsized-deallocation @gol
208 -ftemplate-backtrace-limit=@var{n} @gol
209 -ftemplate-depth=@var{n} @gol
210 -fno-threadsafe-statics -fuse-cxa-atexit @gol
211 -fno-weak -nostdinc++ @gol
212 -fvisibility-inlines-hidden @gol
213 -fvisibility-ms-compat @gol
214 -fext-numeric-literals @gol
215 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
216 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
217 -Wnamespaces -Wnarrowing @gol
218 -Wnoexcept -Wnoexcept-type -Wclass-memaccess @gol
219 -Wnon-virtual-dtor -Wreorder -Wregister @gol
220 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
221 -Wno-non-template-friend -Wold-style-cast @gol
222 -Woverloaded-virtual -Wno-pmf-conversions @gol
223 -Wsign-promo -Wvirtual-inheritance}
225 @item Objective-C and Objective-C++ Language Options
226 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
227 Objective-C and Objective-C++ Dialects}.
228 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
229 -fgnu-runtime -fnext-runtime @gol
230 -fno-nil-receivers @gol
231 -fobjc-abi-version=@var{n} @gol
232 -fobjc-call-cxx-cdtors @gol
233 -fobjc-direct-dispatch @gol
234 -fobjc-exceptions @gol
237 -fobjc-std=objc1 @gol
238 -fno-local-ivars @gol
239 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
240 -freplace-objc-classes @gol
243 -Wassign-intercept @gol
244 -Wno-protocol -Wselector @gol
245 -Wstrict-selector-match @gol
246 -Wundeclared-selector}
248 @item Diagnostic Message Formatting Options
249 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
250 @gccoptlist{-fmessage-length=@var{n} @gol
251 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
252 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
253 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
254 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
255 -fdiagnostics-show-template-tree -fno-elide-type @gol
258 @item Warning Options
259 @xref{Warning Options,,Options to Request or Suppress Warnings}.
260 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
261 -pedantic-errors @gol
262 -w -Wextra -Wall -Waddress -Waggregate-return @gol
263 -Walloc-zero -Walloc-size-larger-than=@var{n}
264 -Walloca -Walloca-larger-than=@var{n} @gol
265 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
266 -Wno-attributes -Wbool-compare -Wbool-operation @gol
267 -Wno-builtin-declaration-mismatch @gol
268 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
269 -Wc++-compat -Wc++11-compat -Wc++14-compat @gol
270 -Wcast-align -Wcast-align=strict -Wcast-qual @gol
271 -Wchar-subscripts -Wchkp -Wcatch-value -Wcatch-value=@var{n} @gol
272 -Wclobbered -Wcomment -Wconditionally-supported @gol
273 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
274 -Wdelete-incomplete @gol
275 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
276 -Wdisabled-optimization @gol
277 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
278 -Wno-div-by-zero -Wdouble-promotion @gol
279 -Wduplicated-branches -Wduplicated-cond @gol
280 -Wempty-body -Wenum-compare -Wno-endif-labels -Wexpansion-to-defined @gol
281 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
282 -Wfloat-equal -Wformat -Wformat=2 @gol
283 -Wno-format-contains-nul -Wno-format-extra-args @gol
284 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
285 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
286 -Wformat-y2k -Wframe-address @gol
287 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
288 -Wif-not-aligned @gol
289 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
290 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
291 -Wimplicit-function-declaration -Wimplicit-int @gol
292 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
293 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
294 -Winvalid-pch -Wlarger-than=@var{len} @gol
295 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
296 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
297 -Wmisleading-indentation -Wmissing-braces @gol
298 -Wmissing-field-initializers -Wmissing-include-dirs @gol
299 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol
300 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
301 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
302 -Woverride-init-side-effects -Woverlength-strings @gol
303 -Wpacked -Wpacked-bitfield-compat -Wpacked-not-aligned -Wpadded @gol
304 -Wparentheses -Wno-pedantic-ms-format @gol
305 -Wplacement-new -Wplacement-new=@var{n} @gol
306 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
307 -Wno-pragmas -Wredundant-decls -Wrestrict -Wno-return-local-addr @gol
308 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
309 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
310 -Wshift-overflow -Wshift-overflow=@var{n} @gol
311 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
312 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
313 -Wno-scalar-storage-order -Wsizeof-pointer-div @gol
314 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
315 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
316 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
317 -Wstringop-overflow=@var{n} @gol
318 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
319 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
320 -Wmissing-format-attribute -Wsubobject-linkage @gol
321 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
322 -Wswitch-unreachable -Wsync-nand @gol
323 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
324 -Wtype-limits -Wundef @gol
325 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
326 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
327 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
328 -Wunused-parameter -Wno-unused-result @gol
329 -Wunused-value -Wunused-variable @gol
330 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
331 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
332 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
333 -Wvla -Wvla-larger-than=@var{n} -Wvolatile-register-var -Wwrite-strings @gol
334 -Wzero-as-null-pointer-constant -Whsa}
336 @item C and Objective-C-only Warning Options
337 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
338 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
339 -Wold-style-declaration -Wold-style-definition @gol
340 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
341 -Wdeclaration-after-statement -Wpointer-sign}
343 @item Debugging Options
344 @xref{Debugging Options,,Options for Debugging Your Program}.
345 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
346 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
347 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
348 -gcolumn-info -gno-column-info @gol
349 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
350 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
351 -fno-eliminate-unused-debug-types @gol
352 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
353 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
354 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
355 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
356 -fvar-tracking -fvar-tracking-assignments}
358 @item Optimization Options
359 @xref{Optimize Options,,Options that Control Optimization}.
360 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
361 -falign-jumps[=@var{n}] @gol
362 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
363 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
364 -fauto-inc-dec -fbranch-probabilities @gol
365 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
366 -fbtr-bb-exclusive -fcaller-saves @gol
367 -fcombine-stack-adjustments -fconserve-stack @gol
368 -fcompare-elim -fcprop-registers -fcrossjumping @gol
369 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
370 -fcx-limited-range @gol
371 -fdata-sections -fdce -fdelayed-branch @gol
372 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
373 -fdevirtualize-at-ltrans -fdse @gol
374 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
375 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
376 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
377 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
378 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
379 -fif-conversion2 -findirect-inlining @gol
380 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
381 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
382 -fipa-bit-cp -fipa-vrp @gol
383 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
384 -fira-algorithm=@var{algorithm} @gol
385 -fira-region=@var{region} -fira-hoist-pressure @gol
386 -fira-loop-pressure -fno-ira-share-save-slots @gol
387 -fno-ira-share-spill-slots @gol
388 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
389 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
390 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
391 -floop-block -floop-interchange -floop-strip-mine @gol
392 -floop-unroll-and-jam -floop-nest-optimize @gol
393 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
394 -flto-partition=@var{alg} -fmerge-all-constants @gol
395 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
396 -fmove-loop-invariants -fno-branch-count-reg @gol
397 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
398 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
399 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
400 -fno-sched-spec -fno-signed-zeros @gol
401 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
402 -fomit-frame-pointer -foptimize-sibling-calls @gol
403 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
404 -fprefetch-loop-arrays @gol
405 -fprofile-correction @gol
406 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
407 -fprofile-reorder-functions @gol
408 -freciprocal-math -free -frename-registers -freorder-blocks @gol
409 -freorder-blocks-algorithm=@var{algorithm} @gol
410 -freorder-blocks-and-partition -freorder-functions @gol
411 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
412 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
413 -fsched-spec-load -fsched-spec-load-dangerous @gol
414 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
415 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
416 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
417 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
418 -fschedule-fusion @gol
419 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
420 -fselective-scheduling -fselective-scheduling2 @gol
421 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
422 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
423 -fsignaling-nans @gol
424 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
426 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
427 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
428 -fthread-jumps -ftracer -ftree-bit-ccp @gol
429 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
430 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
431 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
432 -ftree-loop-if-convert -ftree-loop-im @gol
433 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
434 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
435 -ftree-loop-vectorize @gol
436 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
437 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
438 -ftree-switch-conversion -ftree-tail-merge @gol
439 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
440 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
441 -funsafe-math-optimizations -funswitch-loops @gol
442 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
443 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
444 --param @var{name}=@var{value}
445 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
447 @item Program Instrumentation Options
448 @xref{Instrumentation Options,,Program Instrumentation Options}.
449 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
450 -fprofile-abs-path @gol
451 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
452 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
453 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
454 -fsanitize-undefined-trap-on-error -fbounds-check @gol
455 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
456 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
457 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
458 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
459 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
460 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
461 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
462 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
463 -fchkp-use-wrappers -fchkp-flexible-struct-trailing-arrays@gol
464 -fcf-protection==@r{[}full@r{|}branch@r{|}return@r{|}none@r{]} @gol
465 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
466 -fstack-protector-explicit -fstack-check @gol
467 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
468 -fno-stack-limit -fsplit-stack @gol
469 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
470 -fvtv-counts -fvtv-debug @gol
471 -finstrument-functions @gol
472 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
473 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
475 @item Preprocessor Options
476 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
477 @gccoptlist{-A@var{question}=@var{answer} @gol
478 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
479 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
480 -dD -dI -dM -dN -dU @gol
481 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
482 -fexec-charset=@var{charset} -fextended-identifiers @gol
483 -finput-charset=@var{charset} -fno-canonical-system-headers @gol
484 -fpch-deps -fpch-preprocess -fpreprocessed @gol
485 -ftabstop=@var{width} -ftrack-macro-expansion @gol
486 -fwide-exec-charset=@var{charset} -fworking-directory @gol
487 -H -imacros @var{file} -include @var{file} @gol
488 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
489 -no-integrated-cpp -P -pthread -remap @gol
490 -traditional -traditional-cpp -trigraphs @gol
491 -U@var{macro} -undef @gol
492 -Wp,@var{option} -Xpreprocessor @var{option}}
494 @item Assembler Options
495 @xref{Assembler Options,,Passing Options to the Assembler}.
496 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
499 @xref{Link Options,,Options for Linking}.
500 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
501 -nostartfiles -nodefaultlibs -nostdlib -pie -pthread -rdynamic @gol
502 -s -static -static-pie -static-libgcc -static-libstdc++ @gol
503 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
504 -static-libmpx -static-libmpxwrappers @gol
505 -shared -shared-libgcc -symbolic @gol
506 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
507 -u @var{symbol} -z @var{keyword}}
509 @item Directory Options
510 @xref{Directory Options,,Options for Directory Search}.
511 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
512 -idirafter @var{dir} @gol
513 -imacros @var{file} -imultilib @var{dir} @gol
514 -iplugindir=@var{dir} -iprefix @var{file} @gol
515 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
516 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
517 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
518 -nostdinc -nostdinc++ --sysroot=@var{dir}}
520 @item Code Generation Options
521 @xref{Code Gen Options,,Options for Code Generation Conventions}.
522 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
523 -ffixed-@var{reg} -fexceptions @gol
524 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
525 -fasynchronous-unwind-tables @gol
527 -finhibit-size-directive -fno-common -fno-ident @gol
528 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
529 -fno-jump-tables @gol
530 -frecord-gcc-switches @gol
531 -freg-struct-return -fshort-enums -fshort-wchar @gol
532 -fverbose-asm -fpack-struct[=@var{n}] @gol
533 -fleading-underscore -ftls-model=@var{model} @gol
534 -fstack-reuse=@var{reuse_level} @gol
535 -ftrampolines -ftrapv -fwrapv @gol
536 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
537 -fstrict-volatile-bitfields -fsync-libcalls}
539 @item Developer Options
540 @xref{Developer Options,,GCC Developer Options}.
541 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
542 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
543 -fdbg-cnt=@var{counter-value-list} @gol
544 -fdisable-ipa-@var{pass_name} @gol
545 -fdisable-rtl-@var{pass_name} @gol
546 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
547 -fdisable-tree-@var{pass_name} @gol
548 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
549 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
550 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
551 -fdump-final-insns@r{[}=@var{file}@r{]}
552 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
554 -fdump-lang-@var{switch} @gol
555 -fdump-lang-@var{switch}-@var{options} @gol
556 -fdump-lang-@var{switch}-@var{options}=@var{filename} @gol
558 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
559 -fdump-statistics @gol
561 -fdump-tree-@var{switch} @gol
562 -fdump-tree-@var{switch}-@var{options} @gol
563 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
564 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
565 -fenable-@var{kind}-@var{pass} @gol
566 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
567 -fira-verbose=@var{n} @gol
568 -flto-report -flto-report-wpa -fmem-report-wpa @gol
569 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
570 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
571 -fprofile-report @gol
572 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
573 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
574 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
575 -fvar-tracking-assignments-toggle -gtoggle @gol
576 -print-file-name=@var{library} -print-libgcc-file-name @gol
577 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
578 -print-prog-name=@var{program} -print-search-dirs -Q @gol
579 -print-sysroot -print-sysroot-headers-suffix @gol
580 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
582 @item Machine-Dependent Options
583 @xref{Submodel Options,,Machine-Dependent Options}.
584 @c This list is ordered alphanumerically by subsection name.
585 @c Try and put the significant identifier (CPU or system) first,
586 @c so users have a clue at guessing where the ones they want will be.
588 @emph{AArch64 Options}
589 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
590 -mgeneral-regs-only @gol
591 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
593 -momit-leaf-frame-pointer @gol
594 -mtls-dialect=desc -mtls-dialect=traditional @gol
595 -mtls-size=@var{size} @gol
596 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol
597 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol
598 -mpc-relative-literal-loads @gol
599 -msign-return-address=@var{scope} @gol
600 -march=@var{name} -mcpu=@var{name} -mtune=@var{name} -moverride=@var{string}}
602 @emph{Adapteva Epiphany Options}
603 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
604 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
605 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
606 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
607 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
608 -msplit-vecmove-early -m1reg-@var{reg}}
611 @gccoptlist{-mbarrel-shifter @gol
612 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
613 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
614 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
615 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
616 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
617 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
618 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol
619 -mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol
620 -mvolatile-cache -mtp-regno=@var{regno} @gol
621 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
622 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
623 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
624 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
625 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
626 -mtune=@var{cpu} -mmultcost=@var{num} @gol
627 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
628 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
631 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
632 -mabi=@var{name} @gol
633 -mapcs-stack-check -mno-apcs-stack-check @gol
634 -mapcs-reentrant -mno-apcs-reentrant @gol
635 -msched-prolog -mno-sched-prolog @gol
636 -mlittle-endian -mbig-endian @gol
638 -mfloat-abi=@var{name} @gol
639 -mfp16-format=@var{name}
640 -mthumb-interwork -mno-thumb-interwork @gol
641 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
642 -mtune=@var{name} -mprint-tune-info @gol
643 -mstructure-size-boundary=@var{n} @gol
644 -mabort-on-noreturn @gol
645 -mlong-calls -mno-long-calls @gol
646 -msingle-pic-base -mno-single-pic-base @gol
647 -mpic-register=@var{reg} @gol
648 -mnop-fun-dllimport @gol
649 -mpoke-function-name @gol
651 -mtpcs-frame -mtpcs-leaf-frame @gol
652 -mcaller-super-interworking -mcallee-super-interworking @gol
653 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
654 -mword-relocations @gol
655 -mfix-cortex-m3-ldrd @gol
656 -munaligned-access @gol
657 -mneon-for-64bits @gol
658 -mslow-flash-data @gol
659 -masm-syntax-unified @gol
665 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
666 -mbranch-cost=@var{cost} @gol
667 -mcall-prologues -mgas-isr-prologues -mint8 @gol
668 -mn_flash=@var{size} -mno-interrupts @gol
669 -mrelax -mrmw -mstrict-X -mtiny-stack -mfract-convert-truncate @gol
670 -mshort-calls -nodevicelib @gol
671 -Waddr-space-convert -Wmisspelled-isr}
673 @emph{Blackfin Options}
674 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
675 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
676 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
677 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
678 -mno-id-shared-library -mshared-library-id=@var{n} @gol
679 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
680 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
681 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
685 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
686 -msim -msdata=@var{sdata-type}}
689 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
690 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
691 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
692 -mstack-align -mdata-align -mconst-align @gol
693 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
694 -melf -maout -melinux -mlinux -sim -sim2 @gol
695 -mmul-bug-workaround -mno-mul-bug-workaround}
698 @gccoptlist{-mmac @gol
699 -mcr16cplus -mcr16c @gol
700 -msim -mint32 -mbit-ops
701 -mdata-model=@var{model}}
703 @emph{Darwin Options}
704 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
705 -arch_only -bind_at_load -bundle -bundle_loader @gol
706 -client_name -compatibility_version -current_version @gol
708 -dependency-file -dylib_file -dylinker_install_name @gol
709 -dynamic -dynamiclib -exported_symbols_list @gol
710 -filelist -flat_namespace -force_cpusubtype_ALL @gol
711 -force_flat_namespace -headerpad_max_install_names @gol
713 -image_base -init -install_name -keep_private_externs @gol
714 -multi_module -multiply_defined -multiply_defined_unused @gol
715 -noall_load -no_dead_strip_inits_and_terms @gol
716 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
717 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
718 -private_bundle -read_only_relocs -sectalign @gol
719 -sectobjectsymbols -whyload -seg1addr @gol
720 -sectcreate -sectobjectsymbols -sectorder @gol
721 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
722 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
723 -segprot -segs_read_only_addr -segs_read_write_addr @gol
724 -single_module -static -sub_library -sub_umbrella @gol
725 -twolevel_namespace -umbrella -undefined @gol
726 -unexported_symbols_list -weak_reference_mismatches @gol
727 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
728 -mkernel -mone-byte-bool}
730 @emph{DEC Alpha Options}
731 @gccoptlist{-mno-fp-regs -msoft-float @gol
732 -mieee -mieee-with-inexact -mieee-conformant @gol
733 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
734 -mtrap-precision=@var{mode} -mbuild-constants @gol
735 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
736 -mbwx -mmax -mfix -mcix @gol
737 -mfloat-vax -mfloat-ieee @gol
738 -mexplicit-relocs -msmall-data -mlarge-data @gol
739 -msmall-text -mlarge-text @gol
740 -mmemory-latency=@var{time}}
743 @gccoptlist{-msmall-model -mno-lsim}
746 @gccoptlist{-msim -mlra -mnodiv}
749 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
750 -mhard-float -msoft-float @gol
751 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
752 -mdouble -mno-double @gol
753 -mmedia -mno-media -mmuladd -mno-muladd @gol
754 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
755 -mlinked-fp -mlong-calls -malign-labels @gol
756 -mlibrary-pic -macc-4 -macc-8 @gol
757 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
758 -moptimize-membar -mno-optimize-membar @gol
759 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
760 -mvliw-branch -mno-vliw-branch @gol
761 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
762 -mno-nested-cond-exec -mtomcat-stats @gol
766 @emph{GNU/Linux Options}
767 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
768 -tno-android-cc -tno-android-ld}
770 @emph{H8/300 Options}
771 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
774 @gccoptlist{-march=@var{architecture-type} @gol
775 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
776 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
777 -mfixed-range=@var{register-range} @gol
778 -mjump-in-delay -mlinker-opt -mlong-calls @gol
779 -mlong-load-store -mno-disable-fpregs @gol
780 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
781 -mno-jump-in-delay -mno-long-load-store @gol
782 -mno-portable-runtime -mno-soft-float @gol
783 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
784 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
785 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
786 -munix=@var{unix-std} -nolibdld -static -threads}
789 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
790 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
791 -mconstant-gp -mauto-pic -mfused-madd @gol
792 -minline-float-divide-min-latency @gol
793 -minline-float-divide-max-throughput @gol
794 -mno-inline-float-divide @gol
795 -minline-int-divide-min-latency @gol
796 -minline-int-divide-max-throughput @gol
797 -mno-inline-int-divide @gol
798 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
799 -mno-inline-sqrt @gol
800 -mdwarf2-asm -mearly-stop-bits @gol
801 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
802 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
803 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
804 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
805 -msched-spec-ldc -msched-spec-control-ldc @gol
806 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
807 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
808 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
809 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
812 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
813 -msign-extend-enabled -muser-enabled}
815 @emph{M32R/D Options}
816 @gccoptlist{-m32r2 -m32rx -m32r @gol
818 -malign-loops -mno-align-loops @gol
819 -missue-rate=@var{number} @gol
820 -mbranch-cost=@var{number} @gol
821 -mmodel=@var{code-size-model-type} @gol
822 -msdata=@var{sdata-type} @gol
823 -mno-flush-func -mflush-func=@var{name} @gol
824 -mno-flush-trap -mflush-trap=@var{number} @gol
828 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
830 @emph{M680x0 Options}
831 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
832 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
833 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
834 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
835 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
836 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
837 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
838 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
839 -mxgot -mno-xgot -mlong-jump-table-offsets}
842 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
843 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
844 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
845 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
846 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
849 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
850 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
851 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
852 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
855 @emph{MicroBlaze Options}
856 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
857 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
858 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
859 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
860 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
863 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
864 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
865 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
866 -mips16 -mno-mips16 -mflip-mips16 @gol
867 -minterlink-compressed -mno-interlink-compressed @gol
868 -minterlink-mips16 -mno-interlink-mips16 @gol
869 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
870 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
871 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
872 -mno-float -msingle-float -mdouble-float @gol
873 -modd-spreg -mno-odd-spreg @gol
874 -mabs=@var{mode} -mnan=@var{encoding} @gol
875 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
878 -mvirt -mno-virt @gol
880 -mmicromips -mno-micromips @gol
882 -mfpu=@var{fpu-type} @gol
883 -msmartmips -mno-smartmips @gol
884 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
885 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
886 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
887 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
888 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
889 -membedded-data -mno-embedded-data @gol
890 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
891 -mcode-readable=@var{setting} @gol
892 -msplit-addresses -mno-split-addresses @gol
893 -mexplicit-relocs -mno-explicit-relocs @gol
894 -mcheck-zero-division -mno-check-zero-division @gol
895 -mdivide-traps -mdivide-breaks @gol
896 -mload-store-pairs -mno-load-store-pairs @gol
897 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
898 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
899 -mfix-24k -mno-fix-24k @gol
900 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
901 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
902 -mfix-vr4120 -mno-fix-vr4120 @gol
903 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
904 -mflush-func=@var{func} -mno-flush-func @gol
905 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
906 -mcompact-branches=@var{policy} @gol
907 -mfp-exceptions -mno-fp-exceptions @gol
908 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
909 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
910 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
911 -mframe-header-opt -mno-frame-header-opt}
914 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
915 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
916 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
917 -mno-base-addresses -msingle-exit -mno-single-exit}
919 @emph{MN10300 Options}
920 @gccoptlist{-mmult-bug -mno-mult-bug @gol
921 -mno-am33 -mam33 -mam33-2 -mam34 @gol
922 -mtune=@var{cpu-type} @gol
923 -mreturn-pointer-on-d0 @gol
924 -mno-crt0 -mrelax -mliw -msetlb}
927 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
929 @emph{MSP430 Options}
930 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
932 -mcode-region= -mdata-region= @gol
933 -msilicon-errata= -msilicon-errata-warn= @gol
937 @gccoptlist{-mbig-endian -mlittle-endian @gol
938 -mreduced-regs -mfull-regs @gol
939 -mcmov -mno-cmov @gol
940 -mperf-ext -mno-perf-ext @gol
941 -mv3push -mno-v3push @gol
942 -m16bit -mno-16bit @gol
943 -misr-vector-size=@var{num} @gol
944 -mcache-block-size=@var{num} @gol
945 -march=@var{arch} @gol
946 -mcmodel=@var{code-model} @gol
949 @emph{Nios II Options}
950 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
952 -mno-bypass-cache -mbypass-cache @gol
953 -mno-cache-volatile -mcache-volatile @gol
954 -mno-fast-sw-div -mfast-sw-div @gol
955 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
956 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
957 -mcustom-fpu-cfg=@var{name} @gol
958 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
959 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
961 @emph{Nvidia PTX Options}
962 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
964 @emph{PDP-11 Options}
965 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
966 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
967 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
968 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
969 -mbranch-expensive -mbranch-cheap @gol
970 -munix-asm -mdec-asm}
972 @emph{picoChip Options}
973 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
974 -msymbol-as-address -mno-inefficient-warnings}
976 @emph{PowerPC Options}
977 See RS/6000 and PowerPC Options.
979 @emph{RISC-V Options}
980 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
981 -mmemcpy -mno-memcpy @gol
983 -mabi=@var{ABI-string} @gol
984 -mfdiv -mno-fdiv @gol
986 -march=@var{ISA-string} @gol
987 -mtune=@var{processor-string} @gol
988 -msmall-data-limit=@var{N-bytes} @gol
989 -msave-restore -mno-save-restore @gol
990 -mstrict-align -mno-strict-align @gol
991 -mcmodel=@var{code-model} @gol
992 -mexplicit-relocs -mno-explicit-relocs @gol}
995 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
996 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
997 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
999 @emph{RS/6000 and PowerPC Options}
1000 @gccoptlist{-mcpu=@var{cpu-type} @gol
1001 -mtune=@var{cpu-type} @gol
1002 -mcmodel=@var{code-model} @gol
1004 -maltivec -mno-altivec @gol
1005 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1006 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1007 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1008 -mfprnd -mno-fprnd @gol
1009 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
1010 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1011 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1012 -malign-power -malign-natural @gol
1013 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1014 -msingle-float -mdouble-float -msimple-fpu @gol
1015 -mstring -mno-string -mupdate -mno-update @gol
1016 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1017 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1018 -mstrict-align -mno-strict-align -mrelocatable @gol
1019 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1020 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1021 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
1022 -mprioritize-restricted-insns=@var{priority} @gol
1023 -msched-costly-dep=@var{dependence_type} @gol
1024 -minsert-sched-nops=@var{scheme} @gol
1025 -mcall-sysv -mcall-netbsd @gol
1026 -maix-struct-return -msvr4-struct-return @gol
1027 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1028 -mblock-move-inline-limit=@var{num} @gol
1029 -misel -mno-isel @gol
1030 -misel=yes -misel=no @gol
1032 -mspe=yes -mspe=no @gol
1034 -mvrsave -mno-vrsave @gol
1035 -mmulhw -mno-mulhw @gol
1036 -mdlmzb -mno-dlmzb @gol
1037 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1038 -mprototype -mno-prototype @gol
1039 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1040 -msdata=@var{opt} -mvxworks -G @var{num} @gol
1041 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1042 -mno-recip-precision @gol
1043 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1044 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1045 -msave-toc-indirect -mno-save-toc-indirect @gol
1046 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1047 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1048 -mquad-memory -mno-quad-memory @gol
1049 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1050 -mcompat-align-parm -mno-compat-align-parm @gol
1051 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1052 -mgnu-attribute -mno-gnu-attribute @gol
1053 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1054 -mstack-protector-guard-offset=@var{offset}}
1057 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1059 -mbig-endian-data -mlittle-endian-data @gol
1062 -mas100-syntax -mno-as100-syntax@gol
1064 -mmax-constant-size=@gol
1067 -mallow-string-insns -mno-allow-string-insns@gol
1069 -mno-warn-multiple-fast-interrupts@gol
1070 -msave-acc-in-interrupts}
1072 @emph{S/390 and zSeries Options}
1073 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1074 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1075 -mlong-double-64 -mlong-double-128 @gol
1076 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1077 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1078 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1079 -mhtm -mvx -mzvector @gol
1080 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1081 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1082 -mhotpatch=@var{halfwords},@var{halfwords}}
1084 @emph{Score Options}
1085 @gccoptlist{-meb -mel @gol
1089 -mscore5 -mscore5u -mscore7 -mscore7d}
1092 @gccoptlist{-m1 -m2 -m2e @gol
1093 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1095 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1096 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1097 -mb -ml -mdalign -mrelax @gol
1098 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1099 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1100 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1101 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1102 -maccumulate-outgoing-args @gol
1103 -matomic-model=@var{atomic-model} @gol
1104 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1105 -mcbranch-force-delay-slot @gol
1106 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1107 -mpretend-cmove -mtas}
1109 @emph{Solaris 2 Options}
1110 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1113 @emph{SPARC Options}
1114 @gccoptlist{-mcpu=@var{cpu-type} @gol
1115 -mtune=@var{cpu-type} @gol
1116 -mcmodel=@var{code-model} @gol
1117 -mmemory-model=@var{mem-model} @gol
1118 -m32 -m64 -mapp-regs -mno-app-regs @gol
1119 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1120 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1121 -mhard-quad-float -msoft-quad-float @gol
1122 -mstack-bias -mno-stack-bias @gol
1123 -mstd-struct-return -mno-std-struct-return @gol
1124 -munaligned-doubles -mno-unaligned-doubles @gol
1125 -muser-mode -mno-user-mode @gol
1126 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1127 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1128 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1129 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1130 -mpopc -mno-popc -msubxc -mno-subxc @gol
1131 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1135 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1136 -msafe-dma -munsafe-dma @gol
1138 -msmall-mem -mlarge-mem -mstdmain @gol
1139 -mfixed-range=@var{register-range} @gol
1141 -maddress-space-conversion -mno-address-space-conversion @gol
1142 -mcache-size=@var{cache-size} @gol
1143 -matomic-updates -mno-atomic-updates}
1145 @emph{System V Options}
1146 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1148 @emph{TILE-Gx Options}
1149 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1150 -mcmodel=@var{code-model}}
1152 @emph{TILEPro Options}
1153 @gccoptlist{-mcpu=@var{cpu} -m32}
1156 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1157 -mprolog-function -mno-prolog-function -mspace @gol
1158 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1159 -mapp-regs -mno-app-regs @gol
1160 -mdisable-callt -mno-disable-callt @gol
1161 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1162 -mv850e -mv850 -mv850e3v5 @gol
1173 @gccoptlist{-mg -mgnu -munix}
1175 @emph{Visium Options}
1176 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1177 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1180 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1181 -mpointer-size=@var{size}}
1183 @emph{VxWorks Options}
1184 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1185 -Xbind-lazy -Xbind-now}
1188 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1189 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1190 -mfpmath=@var{unit} @gol
1191 -masm=@var{dialect} -mno-fancy-math-387 @gol
1192 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1193 -mno-wide-multiply -mrtd -malign-double @gol
1194 -mpreferred-stack-boundary=@var{num} @gol
1195 -mincoming-stack-boundary=@var{num} @gol
1196 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1197 -mrecip -mrecip=@var{opt} @gol
1198 -mvzeroupper -mprefer-avx128 -mprefer-avx256 @gol
1199 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1200 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1201 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1202 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1203 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1204 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1205 -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx @gol
1206 -mmwaitx -mclzero -mpku -mthreads @gol
1207 -mcet -mibt -mshstk @gol
1208 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1209 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1210 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1211 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1212 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1213 -mregparm=@var{num} -msseregparm @gol
1214 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1215 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1216 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1217 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1218 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1219 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1220 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1221 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1222 -mstack-protector-guard-reg=@var{reg} @gol
1223 -mstack-protector-guard-offset=@var{offset} @gol
1224 -mstack-protector-guard-symbol=@var{symbol} -mmitigate-rop @gol
1225 -mgeneral-regs-only -mcall-ms2sysv-xlogues}
1227 @emph{x86 Windows Options}
1228 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1229 -mnop-fun-dllimport -mthread @gol
1230 -municode -mwin32 -mwindows -fno-set-stack-executable}
1232 @emph{Xstormy16 Options}
1235 @emph{Xtensa Options}
1236 @gccoptlist{-mconst16 -mno-const16 @gol
1237 -mfused-madd -mno-fused-madd @gol
1239 -mserialize-volatile -mno-serialize-volatile @gol
1240 -mtext-section-literals -mno-text-section-literals @gol
1241 -mauto-litpools -mno-auto-litpools @gol
1242 -mtarget-align -mno-target-align @gol
1243 -mlongcalls -mno-longcalls}
1245 @emph{zSeries Options}
1246 See S/390 and zSeries Options.
1250 @node Overall Options
1251 @section Options Controlling the Kind of Output
1253 Compilation can involve up to four stages: preprocessing, compilation
1254 proper, assembly and linking, always in that order. GCC is capable of
1255 preprocessing and compiling several files either into several
1256 assembler input files, or into one assembler input file; then each
1257 assembler input file produces an object file, and linking combines all
1258 the object files (those newly compiled, and those specified as input)
1259 into an executable file.
1261 @cindex file name suffix
1262 For any given input file, the file name suffix determines what kind of
1263 compilation is done:
1267 C source code that must be preprocessed.
1270 C source code that should not be preprocessed.
1273 C++ source code that should not be preprocessed.
1276 Objective-C source code. Note that you must link with the @file{libobjc}
1277 library to make an Objective-C program work.
1280 Objective-C source code that should not be preprocessed.
1284 Objective-C++ source code. Note that you must link with the @file{libobjc}
1285 library to make an Objective-C++ program work. Note that @samp{.M} refers
1286 to a literal capital M@.
1288 @item @var{file}.mii
1289 Objective-C++ source code that should not be preprocessed.
1292 C, C++, Objective-C or Objective-C++ header file to be turned into a
1293 precompiled header (default), or C, C++ header file to be turned into an
1294 Ada spec (via the @option{-fdump-ada-spec} switch).
1297 @itemx @var{file}.cp
1298 @itemx @var{file}.cxx
1299 @itemx @var{file}.cpp
1300 @itemx @var{file}.CPP
1301 @itemx @var{file}.c++
1303 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1304 the last two letters must both be literally @samp{x}. Likewise,
1305 @samp{.C} refers to a literal capital C@.
1309 Objective-C++ source code that must be preprocessed.
1311 @item @var{file}.mii
1312 Objective-C++ source code that should not be preprocessed.
1316 @itemx @var{file}.hp
1317 @itemx @var{file}.hxx
1318 @itemx @var{file}.hpp
1319 @itemx @var{file}.HPP
1320 @itemx @var{file}.h++
1321 @itemx @var{file}.tcc
1322 C++ header file to be turned into a precompiled header or Ada spec.
1325 @itemx @var{file}.for
1326 @itemx @var{file}.ftn
1327 Fixed form Fortran source code that should not be preprocessed.
1330 @itemx @var{file}.FOR
1331 @itemx @var{file}.fpp
1332 @itemx @var{file}.FPP
1333 @itemx @var{file}.FTN
1334 Fixed form Fortran source code that must be preprocessed (with the traditional
1337 @item @var{file}.f90
1338 @itemx @var{file}.f95
1339 @itemx @var{file}.f03
1340 @itemx @var{file}.f08
1341 Free form Fortran source code that should not be preprocessed.
1343 @item @var{file}.F90
1344 @itemx @var{file}.F95
1345 @itemx @var{file}.F03
1346 @itemx @var{file}.F08
1347 Free form Fortran source code that must be preprocessed (with the
1348 traditional preprocessor).
1353 @item @var{file}.brig
1354 BRIG files (binary representation of HSAIL).
1356 @item @var{file}.ads
1357 Ada source code file that contains a library unit declaration (a
1358 declaration of a package, subprogram, or generic, or a generic
1359 instantiation), or a library unit renaming declaration (a package,
1360 generic, or subprogram renaming declaration). Such files are also
1363 @item @var{file}.adb
1364 Ada source code file containing a library unit body (a subprogram or
1365 package body). Such files are also called @dfn{bodies}.
1367 @c GCC also knows about some suffixes for languages not yet included:
1378 @itemx @var{file}.sx
1379 Assembler code that must be preprocessed.
1382 An object file to be fed straight into linking.
1383 Any file name with no recognized suffix is treated this way.
1387 You can specify the input language explicitly with the @option{-x} option:
1390 @item -x @var{language}
1391 Specify explicitly the @var{language} for the following input files
1392 (rather than letting the compiler choose a default based on the file
1393 name suffix). This option applies to all following input files until
1394 the next @option{-x} option. Possible values for @var{language} are:
1396 c c-header cpp-output
1397 c++ c++-header c++-cpp-output
1398 objective-c objective-c-header objective-c-cpp-output
1399 objective-c++ objective-c++-header objective-c++-cpp-output
1400 assembler assembler-with-cpp
1402 f77 f77-cpp-input f95 f95-cpp-input
1408 Turn off any specification of a language, so that subsequent files are
1409 handled according to their file name suffixes (as they are if @option{-x}
1410 has not been used at all).
1413 If you only want some of the stages of compilation, you can use
1414 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1415 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1416 @command{gcc} is to stop. Note that some combinations (for example,
1417 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1422 Compile or assemble the source files, but do not link. The linking
1423 stage simply is not done. The ultimate output is in the form of an
1424 object file for each source file.
1426 By default, the object file name for a source file is made by replacing
1427 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1429 Unrecognized input files, not requiring compilation or assembly, are
1434 Stop after the stage of compilation proper; do not assemble. The output
1435 is in the form of an assembler code file for each non-assembler input
1438 By default, the assembler file name for a source file is made by
1439 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1441 Input files that don't require compilation are ignored.
1445 Stop after the preprocessing stage; do not run the compiler proper. The
1446 output is in the form of preprocessed source code, which is sent to the
1449 Input files that don't require preprocessing are ignored.
1451 @cindex output file option
1454 Place output in file @var{file}. This applies to whatever
1455 sort of output is being produced, whether it be an executable file,
1456 an object file, an assembler file or preprocessed C code.
1458 If @option{-o} is not specified, the default is to put an executable
1459 file in @file{a.out}, the object file for
1460 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1461 assembler file in @file{@var{source}.s}, a precompiled header file in
1462 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1467 Print (on standard error output) the commands executed to run the stages
1468 of compilation. Also print the version number of the compiler driver
1469 program and of the preprocessor and the compiler proper.
1473 Like @option{-v} except the commands are not executed and arguments
1474 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1475 This is useful for shell scripts to capture the driver-generated command lines.
1479 Print (on the standard output) a description of the command-line options
1480 understood by @command{gcc}. If the @option{-v} option is also specified
1481 then @option{--help} is also passed on to the various processes
1482 invoked by @command{gcc}, so that they can display the command-line options
1483 they accept. If the @option{-Wextra} option has also been specified
1484 (prior to the @option{--help} option), then command-line options that
1485 have no documentation associated with them are also displayed.
1488 @opindex target-help
1489 Print (on the standard output) a description of target-specific command-line
1490 options for each tool. For some targets extra target-specific
1491 information may also be printed.
1493 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1494 Print (on the standard output) a description of the command-line
1495 options understood by the compiler that fit into all specified classes
1496 and qualifiers. These are the supported classes:
1499 @item @samp{optimizers}
1500 Display all of the optimization options supported by the
1503 @item @samp{warnings}
1504 Display all of the options controlling warning messages
1505 produced by the compiler.
1508 Display target-specific options. Unlike the
1509 @option{--target-help} option however, target-specific options of the
1510 linker and assembler are not displayed. This is because those
1511 tools do not currently support the extended @option{--help=} syntax.
1514 Display the values recognized by the @option{--param}
1517 @item @var{language}
1518 Display the options supported for @var{language}, where
1519 @var{language} is the name of one of the languages supported in this
1523 Display the options that are common to all languages.
1526 These are the supported qualifiers:
1529 @item @samp{undocumented}
1530 Display only those options that are undocumented.
1533 Display options taking an argument that appears after an equal
1534 sign in the same continuous piece of text, such as:
1535 @samp{--help=target}.
1537 @item @samp{separate}
1538 Display options taking an argument that appears as a separate word
1539 following the original option, such as: @samp{-o output-file}.
1542 Thus for example to display all the undocumented target-specific
1543 switches supported by the compiler, use:
1546 --help=target,undocumented
1549 The sense of a qualifier can be inverted by prefixing it with the
1550 @samp{^} character, so for example to display all binary warning
1551 options (i.e., ones that are either on or off and that do not take an
1552 argument) that have a description, use:
1555 --help=warnings,^joined,^undocumented
1558 The argument to @option{--help=} should not consist solely of inverted
1561 Combining several classes is possible, although this usually
1562 restricts the output so much that there is nothing to display. One
1563 case where it does work, however, is when one of the classes is
1564 @var{target}. For example, to display all the target-specific
1565 optimization options, use:
1568 --help=target,optimizers
1571 The @option{--help=} option can be repeated on the command line. Each
1572 successive use displays its requested class of options, skipping
1573 those that have already been displayed.
1575 If the @option{-Q} option appears on the command line before the
1576 @option{--help=} option, then the descriptive text displayed by
1577 @option{--help=} is changed. Instead of describing the displayed
1578 options, an indication is given as to whether the option is enabled,
1579 disabled or set to a specific value (assuming that the compiler
1580 knows this at the point where the @option{--help=} option is used).
1582 Here is a truncated example from the ARM port of @command{gcc}:
1585 % gcc -Q -mabi=2 --help=target -c
1586 The following options are target specific:
1588 -mabort-on-noreturn [disabled]
1592 The output is sensitive to the effects of previous command-line
1593 options, so for example it is possible to find out which optimizations
1594 are enabled at @option{-O2} by using:
1597 -Q -O2 --help=optimizers
1600 Alternatively you can discover which binary optimizations are enabled
1601 by @option{-O3} by using:
1604 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1605 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1606 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1611 Display the version number and copyrights of the invoked GCC@.
1613 @item -pass-exit-codes
1614 @opindex pass-exit-codes
1615 Normally the @command{gcc} program exits with the code of 1 if any
1616 phase of the compiler returns a non-success return code. If you specify
1617 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1618 the numerically highest error produced by any phase returning an error
1619 indication. The C, C++, and Fortran front ends return 4 if an internal
1620 compiler error is encountered.
1624 Use pipes rather than temporary files for communication between the
1625 various stages of compilation. This fails to work on some systems where
1626 the assembler is unable to read from a pipe; but the GNU assembler has
1629 @item -specs=@var{file}
1631 Process @var{file} after the compiler reads in the standard @file{specs}
1632 file, in order to override the defaults which the @command{gcc} driver
1633 program uses when determining what switches to pass to @command{cc1},
1634 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1635 @option{-specs=@var{file}} can be specified on the command line, and they
1636 are processed in order, from left to right. @xref{Spec Files}, for
1637 information about the format of the @var{file}.
1641 Invoke all subcommands under a wrapper program. The name of the
1642 wrapper program and its parameters are passed as a comma separated
1646 gcc -c t.c -wrapper gdb,--args
1650 This invokes all subprograms of @command{gcc} under
1651 @samp{gdb --args}, thus the invocation of @command{cc1} is
1652 @samp{gdb --args cc1 @dots{}}.
1654 @item -fplugin=@var{name}.so
1656 Load the plugin code in file @var{name}.so, assumed to be a
1657 shared object to be dlopen'd by the compiler. The base name of
1658 the shared object file is used to identify the plugin for the
1659 purposes of argument parsing (See
1660 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1661 Each plugin should define the callback functions specified in the
1664 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1665 @opindex fplugin-arg
1666 Define an argument called @var{key} with a value of @var{value}
1667 for the plugin called @var{name}.
1669 @item -fdump-ada-spec@r{[}-slim@r{]}
1670 @opindex fdump-ada-spec
1671 For C and C++ source and include files, generate corresponding Ada specs.
1672 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1673 GNAT User's Guide}, which provides detailed documentation on this feature.
1675 @item -fada-spec-parent=@var{unit}
1676 @opindex fada-spec-parent
1677 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1678 Ada specs as child units of parent @var{unit}.
1680 @item -fdump-go-spec=@var{file}
1681 @opindex fdump-go-spec
1682 For input files in any language, generate corresponding Go
1683 declarations in @var{file}. This generates Go @code{const},
1684 @code{type}, @code{var}, and @code{func} declarations which may be a
1685 useful way to start writing a Go interface to code written in some
1688 @include @value{srcdir}/../libiberty/at-file.texi
1692 @section Compiling C++ Programs
1694 @cindex suffixes for C++ source
1695 @cindex C++ source file suffixes
1696 C++ source files conventionally use one of the suffixes @samp{.C},
1697 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1698 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1699 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1700 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1701 files with these names and compiles them as C++ programs even if you
1702 call the compiler the same way as for compiling C programs (usually
1703 with the name @command{gcc}).
1707 However, the use of @command{gcc} does not add the C++ library.
1708 @command{g++} is a program that calls GCC and automatically specifies linking
1709 against the C++ library. It treats @samp{.c},
1710 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1711 files unless @option{-x} is used. This program is also useful when
1712 precompiling a C header file with a @samp{.h} extension for use in C++
1713 compilations. On many systems, @command{g++} is also installed with
1714 the name @command{c++}.
1716 @cindex invoking @command{g++}
1717 When you compile C++ programs, you may specify many of the same
1718 command-line options that you use for compiling programs in any
1719 language; or command-line options meaningful for C and related
1720 languages; or options that are meaningful only for C++ programs.
1721 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1722 explanations of options for languages related to C@.
1723 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1724 explanations of options that are meaningful only for C++ programs.
1726 @node C Dialect Options
1727 @section Options Controlling C Dialect
1728 @cindex dialect options
1729 @cindex language dialect options
1730 @cindex options, dialect
1732 The following options control the dialect of C (or languages derived
1733 from C, such as C++, Objective-C and Objective-C++) that the compiler
1737 @cindex ANSI support
1741 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1742 equivalent to @option{-std=c++98}.
1744 This turns off certain features of GCC that are incompatible with ISO
1745 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1746 such as the @code{asm} and @code{typeof} keywords, and
1747 predefined macros such as @code{unix} and @code{vax} that identify the
1748 type of system you are using. It also enables the undesirable and
1749 rarely used ISO trigraph feature. For the C compiler,
1750 it disables recognition of C++ style @samp{//} comments as well as
1751 the @code{inline} keyword.
1753 The alternate keywords @code{__asm__}, @code{__extension__},
1754 @code{__inline__} and @code{__typeof__} continue to work despite
1755 @option{-ansi}. You would not want to use them in an ISO C program, of
1756 course, but it is useful to put them in header files that might be included
1757 in compilations done with @option{-ansi}. Alternate predefined macros
1758 such as @code{__unix__} and @code{__vax__} are also available, with or
1759 without @option{-ansi}.
1761 The @option{-ansi} option does not cause non-ISO programs to be
1762 rejected gratuitously. For that, @option{-Wpedantic} is required in
1763 addition to @option{-ansi}. @xref{Warning Options}.
1765 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1766 option is used. Some header files may notice this macro and refrain
1767 from declaring certain functions or defining certain macros that the
1768 ISO standard doesn't call for; this is to avoid interfering with any
1769 programs that might use these names for other things.
1771 Functions that are normally built in but do not have semantics
1772 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1773 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1774 built-in functions provided by GCC}, for details of the functions
1779 Determine the language standard. @xref{Standards,,Language Standards
1780 Supported by GCC}, for details of these standard versions. This option
1781 is currently only supported when compiling C or C++.
1783 The compiler can accept several base standards, such as @samp{c90} or
1784 @samp{c++98}, and GNU dialects of those standards, such as
1785 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1786 compiler accepts all programs following that standard plus those
1787 using GNU extensions that do not contradict it. For example,
1788 @option{-std=c90} turns off certain features of GCC that are
1789 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1790 keywords, but not other GNU extensions that do not have a meaning in
1791 ISO C90, such as omitting the middle term of a @code{?:}
1792 expression. On the other hand, when a GNU dialect of a standard is
1793 specified, all features supported by the compiler are enabled, even when
1794 those features change the meaning of the base standard. As a result, some
1795 strict-conforming programs may be rejected. The particular standard
1796 is used by @option{-Wpedantic} to identify which features are GNU
1797 extensions given that version of the standard. For example
1798 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1799 comments, while @option{-std=gnu99 -Wpedantic} does not.
1801 A value for this option must be provided; possible values are
1807 Support all ISO C90 programs (certain GNU extensions that conflict
1808 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1810 @item iso9899:199409
1811 ISO C90 as modified in amendment 1.
1817 ISO C99. This standard is substantially completely supported, modulo
1818 bugs and floating-point issues
1819 (mainly but not entirely relating to optional C99 features from
1820 Annexes F and G). See
1821 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1822 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1827 ISO C11, the 2011 revision of the ISO C standard. This standard is
1828 substantially completely supported, modulo bugs, floating-point issues
1829 (mainly but not entirely relating to optional C11 features from
1830 Annexes F and G) and the optional Annexes K (Bounds-checking
1831 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1835 GNU dialect of ISO C90 (including some C99 features).
1839 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1843 GNU dialect of ISO C11. This is the default for C code.
1844 The name @samp{gnu1x} is deprecated.
1848 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1849 additional defect reports. Same as @option{-ansi} for C++ code.
1853 GNU dialect of @option{-std=c++98}.
1857 The 2011 ISO C++ standard plus amendments.
1858 The name @samp{c++0x} is deprecated.
1862 GNU dialect of @option{-std=c++11}.
1863 The name @samp{gnu++0x} is deprecated.
1867 The 2014 ISO C++ standard plus amendments.
1868 The name @samp{c++1y} is deprecated.
1872 GNU dialect of @option{-std=c++14}.
1873 This is the default for C++ code.
1874 The name @samp{gnu++1y} is deprecated.
1878 The 2017 ISO C++ standard plus amendments.
1879 The name @samp{c++1z} is deprecated.
1883 GNU dialect of @option{-std=c++17}.
1884 The name @samp{gnu++1z} is deprecated.
1887 The next revision of the ISO C++ standard, tentatively planned for
1888 2020. Support is highly experimental, and will almost certainly
1889 change in incompatible ways in future releases.
1892 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
1893 and will almost certainly change in incompatible ways in future
1897 @item -fgnu89-inline
1898 @opindex fgnu89-inline
1899 The option @option{-fgnu89-inline} tells GCC to use the traditional
1900 GNU semantics for @code{inline} functions when in C99 mode.
1901 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1902 Using this option is roughly equivalent to adding the
1903 @code{gnu_inline} function attribute to all inline functions
1904 (@pxref{Function Attributes}).
1906 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1907 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1908 specifies the default behavior).
1909 This option is not supported in @option{-std=c90} or
1910 @option{-std=gnu90} mode.
1912 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1913 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1914 in effect for @code{inline} functions. @xref{Common Predefined
1915 Macros,,,cpp,The C Preprocessor}.
1917 @item -fpermitted-flt-eval-methods=@var{style}
1918 @opindex fpermitted-flt-eval-methods
1919 @opindex fpermitted-flt-eval-methods=c11
1920 @opindex fpermitted-flt-eval-methods=ts-18661-3
1921 ISO/IEC TS 18661-3 defines new permissible values for
1922 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
1923 a semantic type that is an interchange or extended format should be
1924 evaluated to the precision and range of that type. These new values are
1925 a superset of those permitted under C99/C11, which does not specify the
1926 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
1927 conforming to C11 may not have been written expecting the possibility of
1930 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
1931 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
1932 or the extended set of values specified in ISO/IEC TS 18661-3.
1934 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
1936 The default when in a standards compliant mode (@option{-std=c11} or similar)
1937 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
1938 dialect (@option{-std=gnu11} or similar) is
1939 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
1941 @item -aux-info @var{filename}
1943 Output to the given filename prototyped declarations for all functions
1944 declared and/or defined in a translation unit, including those in header
1945 files. This option is silently ignored in any language other than C@.
1947 Besides declarations, the file indicates, in comments, the origin of
1948 each declaration (source file and line), whether the declaration was
1949 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1950 @samp{O} for old, respectively, in the first character after the line
1951 number and the colon), and whether it came from a declaration or a
1952 definition (@samp{C} or @samp{F}, respectively, in the following
1953 character). In the case of function definitions, a K&R-style list of
1954 arguments followed by their declarations is also provided, inside
1955 comments, after the declaration.
1957 @item -fallow-parameterless-variadic-functions
1958 @opindex fallow-parameterless-variadic-functions
1959 Accept variadic functions without named parameters.
1961 Although it is possible to define such a function, this is not very
1962 useful as it is not possible to read the arguments. This is only
1963 supported for C as this construct is allowed by C++.
1967 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1968 keyword, so that code can use these words as identifiers. You can use
1969 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1970 instead. @option{-ansi} implies @option{-fno-asm}.
1972 In C++, this switch only affects the @code{typeof} keyword, since
1973 @code{asm} and @code{inline} are standard keywords. You may want to
1974 use the @option{-fno-gnu-keywords} flag instead, which has the same
1975 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1976 switch only affects the @code{asm} and @code{typeof} keywords, since
1977 @code{inline} is a standard keyword in ISO C99.
1980 @itemx -fno-builtin-@var{function}
1981 @opindex fno-builtin
1982 @cindex built-in functions
1983 Don't recognize built-in functions that do not begin with
1984 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1985 functions provided by GCC}, for details of the functions affected,
1986 including those which are not built-in functions when @option{-ansi} or
1987 @option{-std} options for strict ISO C conformance are used because they
1988 do not have an ISO standard meaning.
1990 GCC normally generates special code to handle certain built-in functions
1991 more efficiently; for instance, calls to @code{alloca} may become single
1992 instructions which adjust the stack directly, and calls to @code{memcpy}
1993 may become inline copy loops. The resulting code is often both smaller
1994 and faster, but since the function calls no longer appear as such, you
1995 cannot set a breakpoint on those calls, nor can you change the behavior
1996 of the functions by linking with a different library. In addition,
1997 when a function is recognized as a built-in function, GCC may use
1998 information about that function to warn about problems with calls to
1999 that function, or to generate more efficient code, even if the
2000 resulting code still contains calls to that function. For example,
2001 warnings are given with @option{-Wformat} for bad calls to
2002 @code{printf} when @code{printf} is built in and @code{strlen} is
2003 known not to modify global memory.
2005 With the @option{-fno-builtin-@var{function}} option
2006 only the built-in function @var{function} is
2007 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2008 function is named that is not built-in in this version of GCC, this
2009 option is ignored. There is no corresponding
2010 @option{-fbuiltin-@var{function}} option; if you wish to enable
2011 built-in functions selectively when using @option{-fno-builtin} or
2012 @option{-ffreestanding}, you may define macros such as:
2015 #define abs(n) __builtin_abs ((n))
2016 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2022 Enable parsing of function definitions marked with @code{__GIMPLE}.
2023 This is an experimental feature that allows unit testing of GIMPLE
2028 @cindex hosted environment
2030 Assert that compilation targets a hosted environment. This implies
2031 @option{-fbuiltin}. A hosted environment is one in which the
2032 entire standard library is available, and in which @code{main} has a return
2033 type of @code{int}. Examples are nearly everything except a kernel.
2034 This is equivalent to @option{-fno-freestanding}.
2036 @item -ffreestanding
2037 @opindex ffreestanding
2038 @cindex hosted environment
2040 Assert that compilation targets a freestanding environment. This
2041 implies @option{-fno-builtin}. A freestanding environment
2042 is one in which the standard library may not exist, and program startup may
2043 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2044 This is equivalent to @option{-fno-hosted}.
2046 @xref{Standards,,Language Standards Supported by GCC}, for details of
2047 freestanding and hosted environments.
2051 @cindex OpenACC accelerator programming
2052 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2053 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2054 compiler generates accelerated code according to the OpenACC Application
2055 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2056 implies @option{-pthread}, and thus is only supported on targets that
2057 have support for @option{-pthread}.
2059 @item -fopenacc-dim=@var{geom}
2060 @opindex fopenacc-dim
2061 @cindex OpenACC accelerator programming
2062 Specify default compute dimensions for parallel offload regions that do
2063 not explicitly specify. The @var{geom} value is a triple of
2064 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2065 can be omitted, to use a target-specific default value.
2069 @cindex OpenMP parallel
2070 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2071 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2072 compiler generates parallel code according to the OpenMP Application
2073 Program Interface v4.5 @w{@uref{http://www.openmp.org/}}. This option
2074 implies @option{-pthread}, and thus is only supported on targets that
2075 have support for @option{-pthread}. @option{-fopenmp} implies
2076 @option{-fopenmp-simd}.
2079 @opindex fopenmp-simd
2082 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2083 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2088 @cindex Enable Cilk Plus
2089 Enable the usage of Cilk Plus language extension features for C/C++.
2090 When the option @option{-fcilkplus} is specified, enable the usage of
2091 the Cilk Plus Language extension features for C/C++. The present
2092 implementation follows ABI version 1.2. This is an experimental
2093 feature that is only partially complete, and whose interface may
2094 change in future versions of GCC as the official specification
2095 changes. Currently, all features but @code{_Cilk_for} have been
2100 When the option @option{-fgnu-tm} is specified, the compiler
2101 generates code for the Linux variant of Intel's current Transactional
2102 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2103 an experimental feature whose interface may change in future versions
2104 of GCC, as the official specification changes. Please note that not
2105 all architectures are supported for this feature.
2107 For more information on GCC's support for transactional memory,
2108 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2109 Transactional Memory Library}.
2111 Note that the transactional memory feature is not supported with
2112 non-call exceptions (@option{-fnon-call-exceptions}).
2114 @item -fms-extensions
2115 @opindex fms-extensions
2116 Accept some non-standard constructs used in Microsoft header files.
2118 In C++ code, this allows member names in structures to be similar
2119 to previous types declarations.
2128 Some cases of unnamed fields in structures and unions are only
2129 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2130 fields within structs/unions}, for details.
2132 Note that this option is off for all targets but x86
2133 targets using ms-abi.
2135 @item -fplan9-extensions
2136 @opindex fplan9-extensions
2137 Accept some non-standard constructs used in Plan 9 code.
2139 This enables @option{-fms-extensions}, permits passing pointers to
2140 structures with anonymous fields to functions that expect pointers to
2141 elements of the type of the field, and permits referring to anonymous
2142 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2143 struct/union fields within structs/unions}, for details. This is only
2144 supported for C, not C++.
2146 @item -fcond-mismatch
2147 @opindex fcond-mismatch
2148 Allow conditional expressions with mismatched types in the second and
2149 third arguments. The value of such an expression is void. This option
2150 is not supported for C++.
2152 @item -flax-vector-conversions
2153 @opindex flax-vector-conversions
2154 Allow implicit conversions between vectors with differing numbers of
2155 elements and/or incompatible element types. This option should not be
2158 @item -funsigned-char
2159 @opindex funsigned-char
2160 Let the type @code{char} be unsigned, like @code{unsigned char}.
2162 Each kind of machine has a default for what @code{char} should
2163 be. It is either like @code{unsigned char} by default or like
2164 @code{signed char} by default.
2166 Ideally, a portable program should always use @code{signed char} or
2167 @code{unsigned char} when it depends on the signedness of an object.
2168 But many programs have been written to use plain @code{char} and
2169 expect it to be signed, or expect it to be unsigned, depending on the
2170 machines they were written for. This option, and its inverse, let you
2171 make such a program work with the opposite default.
2173 The type @code{char} is always a distinct type from each of
2174 @code{signed char} or @code{unsigned char}, even though its behavior
2175 is always just like one of those two.
2178 @opindex fsigned-char
2179 Let the type @code{char} be signed, like @code{signed char}.
2181 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2182 the negative form of @option{-funsigned-char}. Likewise, the option
2183 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2185 @item -fsigned-bitfields
2186 @itemx -funsigned-bitfields
2187 @itemx -fno-signed-bitfields
2188 @itemx -fno-unsigned-bitfields
2189 @opindex fsigned-bitfields
2190 @opindex funsigned-bitfields
2191 @opindex fno-signed-bitfields
2192 @opindex fno-unsigned-bitfields
2193 These options control whether a bit-field is signed or unsigned, when the
2194 declaration does not use either @code{signed} or @code{unsigned}. By
2195 default, such a bit-field is signed, because this is consistent: the
2196 basic integer types such as @code{int} are signed types.
2198 @item -fsso-struct=@var{endianness}
2199 @opindex fsso-struct
2200 Set the default scalar storage order of structures and unions to the
2201 specified endianness. The accepted values are @samp{big-endian},
2202 @samp{little-endian} and @samp{native} for the native endianness of
2203 the target (the default). This option is not supported for C++.
2205 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2206 code that is not binary compatible with code generated without it if the
2207 specified endianness is not the native endianness of the target.
2210 @node C++ Dialect Options
2211 @section Options Controlling C++ Dialect
2213 @cindex compiler options, C++
2214 @cindex C++ options, command-line
2215 @cindex options, C++
2216 This section describes the command-line options that are only meaningful
2217 for C++ programs. You can also use most of the GNU compiler options
2218 regardless of what language your program is in. For example, you
2219 might compile a file @file{firstClass.C} like this:
2222 g++ -g -fstrict-enums -O -c firstClass.C
2226 In this example, only @option{-fstrict-enums} is an option meant
2227 only for C++ programs; you can use the other options with any
2228 language supported by GCC@.
2230 Some options for compiling C programs, such as @option{-std}, are also
2231 relevant for C++ programs.
2232 @xref{C Dialect Options,,Options Controlling C Dialect}.
2234 Here is a list of options that are @emph{only} for compiling C++ programs:
2238 @item -fabi-version=@var{n}
2239 @opindex fabi-version
2240 Use version @var{n} of the C++ ABI@. The default is version 0.
2242 Version 0 refers to the version conforming most closely to
2243 the C++ ABI specification. Therefore, the ABI obtained using version 0
2244 will change in different versions of G++ as ABI bugs are fixed.
2246 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2248 Version 2 is the version of the C++ ABI that first appeared in G++
2249 3.4, and was the default through G++ 4.9.
2251 Version 3 corrects an error in mangling a constant address as a
2254 Version 4, which first appeared in G++ 4.5, implements a standard
2255 mangling for vector types.
2257 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2258 attribute const/volatile on function pointer types, decltype of a
2259 plain decl, and use of a function parameter in the declaration of
2262 Version 6, which first appeared in G++ 4.7, corrects the promotion
2263 behavior of C++11 scoped enums and the mangling of template argument
2264 packs, const/static_cast, prefix ++ and --, and a class scope function
2265 used as a template argument.
2267 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2268 builtin type and corrects the mangling of lambdas in default argument
2271 Version 8, which first appeared in G++ 4.9, corrects the substitution
2272 behavior of function types with function-cv-qualifiers.
2274 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2277 Version 10, which first appeared in G++ 6.1, adds mangling of
2278 attributes that affect type identity, such as ia32 calling convention
2279 attributes (e.g. @samp{stdcall}).
2281 Version 11, which first appeared in G++ 7, corrects the mangling of
2282 sizeof... expressions and operator names. For multiple entities with
2283 the same name within a function, that are declared in different scopes,
2284 the mangling now changes starting with the twelfth occurrence. It also
2285 implies @option{-fnew-inheriting-ctors}.
2287 See also @option{-Wabi}.
2289 @item -fabi-compat-version=@var{n}
2290 @opindex fabi-compat-version
2291 On targets that support strong aliases, G++
2292 works around mangling changes by creating an alias with the correct
2293 mangled name when defining a symbol with an incorrect mangled name.
2294 This switch specifies which ABI version to use for the alias.
2296 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2297 compatibility). If another ABI version is explicitly selected, this
2298 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2299 use @option{-fabi-compat-version=2}.
2301 If this option is not provided but @option{-Wabi=@var{n}} is, that
2302 version is used for compatibility aliases. If this option is provided
2303 along with @option{-Wabi} (without the version), the version from this
2304 option is used for the warning.
2306 @item -fno-access-control
2307 @opindex fno-access-control
2308 Turn off all access checking. This switch is mainly useful for working
2309 around bugs in the access control code.
2312 @opindex faligned-new
2313 Enable support for C++17 @code{new} of types that require more
2314 alignment than @code{void* ::operator new(std::size_t)} provides. A
2315 numeric argument such as @code{-faligned-new=32} can be used to
2316 specify how much alignment (in bytes) is provided by that function,
2317 but few users will need to override the default of
2318 @code{alignof(std::max_align_t)}.
2320 This flag is enabled by default for @option{-std=c++17}.
2324 Check that the pointer returned by @code{operator new} is non-null
2325 before attempting to modify the storage allocated. This check is
2326 normally unnecessary because the C++ standard specifies that
2327 @code{operator new} only returns @code{0} if it is declared
2328 @code{throw()}, in which case the compiler always checks the
2329 return value even without this option. In all other cases, when
2330 @code{operator new} has a non-empty exception specification, memory
2331 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2332 @samp{new (nothrow)}.
2336 Enable support for the C++ Extensions for Concepts Technical
2337 Specification, ISO 19217 (2015), which allows code like
2340 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2341 template <Addable T> T add (T a, T b) @{ return a + b; @}
2344 @item -fconstexpr-depth=@var{n}
2345 @opindex fconstexpr-depth
2346 Set the maximum nested evaluation depth for C++11 constexpr functions
2347 to @var{n}. A limit is needed to detect endless recursion during
2348 constant expression evaluation. The minimum specified by the standard
2351 @item -fconstexpr-loop-limit=@var{n}
2352 @opindex fconstexpr-loop-limit
2353 Set the maximum number of iterations for a loop in C++14 constexpr functions
2354 to @var{n}. A limit is needed to detect infinite loops during
2355 constant expression evaluation. The default is 262144 (1<<18).
2357 @item -fdeduce-init-list
2358 @opindex fdeduce-init-list
2359 Enable deduction of a template type parameter as
2360 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2363 template <class T> auto forward(T t) -> decltype (realfn (t))
2370 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2374 This deduction was implemented as a possible extension to the
2375 originally proposed semantics for the C++11 standard, but was not part
2376 of the final standard, so it is disabled by default. This option is
2377 deprecated, and may be removed in a future version of G++.
2379 @item -ffriend-injection
2380 @opindex ffriend-injection
2381 Inject friend functions into the enclosing namespace, so that they are
2382 visible outside the scope of the class in which they are declared.
2383 Friend functions were documented to work this way in the old Annotated
2384 C++ Reference Manual.
2385 However, in ISO C++ a friend function that is not declared
2386 in an enclosing scope can only be found using argument dependent
2387 lookup. GCC defaults to the standard behavior.
2389 This option is for compatibility, and may be removed in a future
2392 @item -fno-elide-constructors
2393 @opindex fno-elide-constructors
2394 The C++ standard allows an implementation to omit creating a temporary
2395 that is only used to initialize another object of the same type.
2396 Specifying this option disables that optimization, and forces G++ to
2397 call the copy constructor in all cases. This option also causes G++
2398 to call trivial member functions which otherwise would be expanded inline.
2400 In C++17, the compiler is required to omit these temporaries, but this
2401 option still affects trivial member functions.
2403 @item -fno-enforce-eh-specs
2404 @opindex fno-enforce-eh-specs
2405 Don't generate code to check for violation of exception specifications
2406 at run time. This option violates the C++ standard, but may be useful
2407 for reducing code size in production builds, much like defining
2408 @code{NDEBUG}. This does not give user code permission to throw
2409 exceptions in violation of the exception specifications; the compiler
2410 still optimizes based on the specifications, so throwing an
2411 unexpected exception results in undefined behavior at run time.
2413 @item -fextern-tls-init
2414 @itemx -fno-extern-tls-init
2415 @opindex fextern-tls-init
2416 @opindex fno-extern-tls-init
2417 The C++11 and OpenMP standards allow @code{thread_local} and
2418 @code{threadprivate} variables to have dynamic (runtime)
2419 initialization. To support this, any use of such a variable goes
2420 through a wrapper function that performs any necessary initialization.
2421 When the use and definition of the variable are in the same
2422 translation unit, this overhead can be optimized away, but when the
2423 use is in a different translation unit there is significant overhead
2424 even if the variable doesn't actually need dynamic initialization. If
2425 the programmer can be sure that no use of the variable in a
2426 non-defining TU needs to trigger dynamic initialization (either
2427 because the variable is statically initialized, or a use of the
2428 variable in the defining TU will be executed before any uses in
2429 another TU), they can avoid this overhead with the
2430 @option{-fno-extern-tls-init} option.
2432 On targets that support symbol aliases, the default is
2433 @option{-fextern-tls-init}. On targets that do not support symbol
2434 aliases, the default is @option{-fno-extern-tls-init}.
2437 @itemx -fno-for-scope
2439 @opindex fno-for-scope
2440 If @option{-ffor-scope} is specified, the scope of variables declared in
2441 a @i{for-init-statement} is limited to the @code{for} loop itself,
2442 as specified by the C++ standard.
2443 If @option{-fno-for-scope} is specified, the scope of variables declared in
2444 a @i{for-init-statement} extends to the end of the enclosing scope,
2445 as was the case in old versions of G++, and other (traditional)
2446 implementations of C++.
2448 If neither flag is given, the default is to follow the standard,
2449 but to allow and give a warning for old-style code that would
2450 otherwise be invalid, or have different behavior.
2452 @item -fno-gnu-keywords
2453 @opindex fno-gnu-keywords
2454 Do not recognize @code{typeof} as a keyword, so that code can use this
2455 word as an identifier. You can use the keyword @code{__typeof__} instead.
2456 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2457 @option{-std=c++98}, @option{-std=c++11}, etc.
2459 @item -fno-implicit-templates
2460 @opindex fno-implicit-templates
2461 Never emit code for non-inline templates that are instantiated
2462 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2463 @xref{Template Instantiation}, for more information.
2465 @item -fno-implicit-inline-templates
2466 @opindex fno-implicit-inline-templates
2467 Don't emit code for implicit instantiations of inline templates, either.
2468 The default is to handle inlines differently so that compiles with and
2469 without optimization need the same set of explicit instantiations.
2471 @item -fno-implement-inlines
2472 @opindex fno-implement-inlines
2473 To save space, do not emit out-of-line copies of inline functions
2474 controlled by @code{#pragma implementation}. This causes linker
2475 errors if these functions are not inlined everywhere they are called.
2477 @item -fms-extensions
2478 @opindex fms-extensions
2479 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2480 int and getting a pointer to member function via non-standard syntax.
2482 @item -fnew-inheriting-ctors
2483 @opindex fnew-inheriting-ctors
2484 Enable the P0136 adjustment to the semantics of C++11 constructor
2485 inheritance. This is part of C++17 but also considered to be a Defect
2486 Report against C++11 and C++14. This flag is enabled by default
2487 unless @option{-fabi-version=10} or lower is specified.
2489 @item -fnew-ttp-matching
2490 @opindex fnew-ttp-matching
2491 Enable the P0522 resolution to Core issue 150, template template
2492 parameters and default arguments: this allows a template with default
2493 template arguments as an argument for a template template parameter
2494 with fewer template parameters. This flag is enabled by default for
2495 @option{-std=c++17}.
2497 @item -fno-nonansi-builtins
2498 @opindex fno-nonansi-builtins
2499 Disable built-in declarations of functions that are not mandated by
2500 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2501 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2504 @opindex fnothrow-opt
2505 Treat a @code{throw()} exception specification as if it were a
2506 @code{noexcept} specification to reduce or eliminate the text size
2507 overhead relative to a function with no exception specification. If
2508 the function has local variables of types with non-trivial
2509 destructors, the exception specification actually makes the
2510 function smaller because the EH cleanups for those variables can be
2511 optimized away. The semantic effect is that an exception thrown out of
2512 a function with such an exception specification results in a call
2513 to @code{terminate} rather than @code{unexpected}.
2515 @item -fno-operator-names
2516 @opindex fno-operator-names
2517 Do not treat the operator name keywords @code{and}, @code{bitand},
2518 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2519 synonyms as keywords.
2521 @item -fno-optional-diags
2522 @opindex fno-optional-diags
2523 Disable diagnostics that the standard says a compiler does not need to
2524 issue. Currently, the only such diagnostic issued by G++ is the one for
2525 a name having multiple meanings within a class.
2528 @opindex fpermissive
2529 Downgrade some diagnostics about nonconformant code from errors to
2530 warnings. Thus, using @option{-fpermissive} allows some
2531 nonconforming code to compile.
2533 @item -fno-pretty-templates
2534 @opindex fno-pretty-templates
2535 When an error message refers to a specialization of a function
2536 template, the compiler normally prints the signature of the
2537 template followed by the template arguments and any typedefs or
2538 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2539 rather than @code{void f(int)}) so that it's clear which template is
2540 involved. When an error message refers to a specialization of a class
2541 template, the compiler omits any template arguments that match
2542 the default template arguments for that template. If either of these
2543 behaviors make it harder to understand the error message rather than
2544 easier, you can use @option{-fno-pretty-templates} to disable them.
2548 Enable automatic template instantiation at link time. This option also
2549 implies @option{-fno-implicit-templates}. @xref{Template
2550 Instantiation}, for more information.
2554 Disable generation of information about every class with virtual
2555 functions for use by the C++ run-time type identification features
2556 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2557 of the language, you can save some space by using this flag. Note that
2558 exception handling uses the same information, but G++ generates it as
2559 needed. The @code{dynamic_cast} operator can still be used for casts that
2560 do not require run-time type information, i.e.@: casts to @code{void *} or to
2561 unambiguous base classes.
2563 @item -fsized-deallocation
2564 @opindex fsized-deallocation
2565 Enable the built-in global declarations
2567 void operator delete (void *, std::size_t) noexcept;
2568 void operator delete[] (void *, std::size_t) noexcept;
2570 as introduced in C++14. This is useful for user-defined replacement
2571 deallocation functions that, for example, use the size of the object
2572 to make deallocation faster. Enabled by default under
2573 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2574 warns about places that might want to add a definition.
2576 @item -fstrict-enums
2577 @opindex fstrict-enums
2578 Allow the compiler to optimize using the assumption that a value of
2579 enumerated type can only be one of the values of the enumeration (as
2580 defined in the C++ standard; basically, a value that can be
2581 represented in the minimum number of bits needed to represent all the
2582 enumerators). This assumption may not be valid if the program uses a
2583 cast to convert an arbitrary integer value to the enumerated type.
2585 @item -fstrong-eval-order
2586 @opindex fstrong-eval-order
2587 Evaluate member access, array subscripting, and shift expressions in
2588 left-to-right order, and evaluate assignment in right-to-left order,
2589 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2590 @option{-fstrong-eval-order=some} enables just the ordering of member
2591 access and shift expressions, and is the default without
2592 @option{-std=c++17}.
2594 @item -ftemplate-backtrace-limit=@var{n}
2595 @opindex ftemplate-backtrace-limit
2596 Set the maximum number of template instantiation notes for a single
2597 warning or error to @var{n}. The default value is 10.
2599 @item -ftemplate-depth=@var{n}
2600 @opindex ftemplate-depth
2601 Set the maximum instantiation depth for template classes to @var{n}.
2602 A limit on the template instantiation depth is needed to detect
2603 endless recursions during template class instantiation. ANSI/ISO C++
2604 conforming programs must not rely on a maximum depth greater than 17
2605 (changed to 1024 in C++11). The default value is 900, as the compiler
2606 can run out of stack space before hitting 1024 in some situations.
2608 @item -fno-threadsafe-statics
2609 @opindex fno-threadsafe-statics
2610 Do not emit the extra code to use the routines specified in the C++
2611 ABI for thread-safe initialization of local statics. You can use this
2612 option to reduce code size slightly in code that doesn't need to be
2615 @item -fuse-cxa-atexit
2616 @opindex fuse-cxa-atexit
2617 Register destructors for objects with static storage duration with the
2618 @code{__cxa_atexit} function rather than the @code{atexit} function.
2619 This option is required for fully standards-compliant handling of static
2620 destructors, but only works if your C library supports
2621 @code{__cxa_atexit}.
2623 @item -fno-use-cxa-get-exception-ptr
2624 @opindex fno-use-cxa-get-exception-ptr
2625 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2626 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2627 if the runtime routine is not available.
2629 @item -fvisibility-inlines-hidden
2630 @opindex fvisibility-inlines-hidden
2631 This switch declares that the user does not attempt to compare
2632 pointers to inline functions or methods where the addresses of the two functions
2633 are taken in different shared objects.
2635 The effect of this is that GCC may, effectively, mark inline methods with
2636 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2637 appear in the export table of a DSO and do not require a PLT indirection
2638 when used within the DSO@. Enabling this option can have a dramatic effect
2639 on load and link times of a DSO as it massively reduces the size of the
2640 dynamic export table when the library makes heavy use of templates.
2642 The behavior of this switch is not quite the same as marking the
2643 methods as hidden directly, because it does not affect static variables
2644 local to the function or cause the compiler to deduce that
2645 the function is defined in only one shared object.
2647 You may mark a method as having a visibility explicitly to negate the
2648 effect of the switch for that method. For example, if you do want to
2649 compare pointers to a particular inline method, you might mark it as
2650 having default visibility. Marking the enclosing class with explicit
2651 visibility has no effect.
2653 Explicitly instantiated inline methods are unaffected by this option
2654 as their linkage might otherwise cross a shared library boundary.
2655 @xref{Template Instantiation}.
2657 @item -fvisibility-ms-compat
2658 @opindex fvisibility-ms-compat
2659 This flag attempts to use visibility settings to make GCC's C++
2660 linkage model compatible with that of Microsoft Visual Studio.
2662 The flag makes these changes to GCC's linkage model:
2666 It sets the default visibility to @code{hidden}, like
2667 @option{-fvisibility=hidden}.
2670 Types, but not their members, are not hidden by default.
2673 The One Definition Rule is relaxed for types without explicit
2674 visibility specifications that are defined in more than one
2675 shared object: those declarations are permitted if they are
2676 permitted when this option is not used.
2679 In new code it is better to use @option{-fvisibility=hidden} and
2680 export those classes that are intended to be externally visible.
2681 Unfortunately it is possible for code to rely, perhaps accidentally,
2682 on the Visual Studio behavior.
2684 Among the consequences of these changes are that static data members
2685 of the same type with the same name but defined in different shared
2686 objects are different, so changing one does not change the other;
2687 and that pointers to function members defined in different shared
2688 objects may not compare equal. When this flag is given, it is a
2689 violation of the ODR to define types with the same name differently.
2693 Do not use weak symbol support, even if it is provided by the linker.
2694 By default, G++ uses weak symbols if they are available. This
2695 option exists only for testing, and should not be used by end-users;
2696 it results in inferior code and has no benefits. This option may
2697 be removed in a future release of G++.
2701 Do not search for header files in the standard directories specific to
2702 C++, but do still search the other standard directories. (This option
2703 is used when building the C++ library.)
2706 In addition, these optimization, warning, and code generation options
2707 have meanings only for C++ programs:
2710 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2713 Warn when G++ it generates code that is probably not compatible with
2714 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2715 ABI with each major release, normally @option{-Wabi} will warn only if
2716 there is a check added later in a release series for an ABI issue
2717 discovered since the initial release. @option{-Wabi} will warn about
2718 more things if an older ABI version is selected (with
2719 @option{-fabi-version=@var{n}}).
2721 @option{-Wabi} can also be used with an explicit version number to
2722 warn about compatibility with a particular @option{-fabi-version}
2723 level, e.g. @option{-Wabi=2} to warn about changes relative to
2724 @option{-fabi-version=2}.
2726 If an explicit version number is provided and
2727 @option{-fabi-compat-version} is not specified, the version number
2728 from this option is used for compatibility aliases. If no explicit
2729 version number is provided with this option, but
2730 @option{-fabi-compat-version} is specified, that version number is
2731 used for ABI warnings.
2733 Although an effort has been made to warn about
2734 all such cases, there are probably some cases that are not warned about,
2735 even though G++ is generating incompatible code. There may also be
2736 cases where warnings are emitted even though the code that is generated
2739 You should rewrite your code to avoid these warnings if you are
2740 concerned about the fact that code generated by G++ may not be binary
2741 compatible with code generated by other compilers.
2743 Known incompatibilities in @option{-fabi-version=2} (which was the
2744 default from GCC 3.4 to 4.9) include:
2749 A template with a non-type template parameter of reference type was
2750 mangled incorrectly:
2753 template <int &> struct S @{@};
2757 This was fixed in @option{-fabi-version=3}.
2760 SIMD vector types declared using @code{__attribute ((vector_size))} were
2761 mangled in a non-standard way that does not allow for overloading of
2762 functions taking vectors of different sizes.
2764 The mangling was changed in @option{-fabi-version=4}.
2767 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2768 qualifiers, and @code{decltype} of a plain declaration was folded away.
2770 These mangling issues were fixed in @option{-fabi-version=5}.
2773 Scoped enumerators passed as arguments to a variadic function are
2774 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2775 On most targets this does not actually affect the parameter passing
2776 ABI, as there is no way to pass an argument smaller than @code{int}.
2778 Also, the ABI changed the mangling of template argument packs,
2779 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2780 a class scope function used as a template argument.
2782 These issues were corrected in @option{-fabi-version=6}.
2785 Lambdas in default argument scope were mangled incorrectly, and the
2786 ABI changed the mangling of @code{nullptr_t}.
2788 These issues were corrected in @option{-fabi-version=7}.
2791 When mangling a function type with function-cv-qualifiers, the
2792 un-qualified function type was incorrectly treated as a substitution
2795 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2798 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2799 unaligned accesses. Note that this did not affect the ABI of a
2800 function with a @code{nullptr_t} parameter, as parameters have a
2803 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2806 Target-specific attributes that affect the identity of a type, such as
2807 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2808 did not affect the mangled name, leading to name collisions when
2809 function pointers were used as template arguments.
2811 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2815 It also warns about psABI-related changes. The known psABI changes at this
2821 For SysV/x86-64, unions with @code{long double} members are
2822 passed in memory as specified in psABI. For example:
2832 @code{union U} is always passed in memory.
2836 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2839 Warn when a type with an ABI tag is used in a context that does not
2840 have that ABI tag. See @ref{C++ Attributes} for more information
2843 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2844 @opindex Wctor-dtor-privacy
2845 @opindex Wno-ctor-dtor-privacy
2846 Warn when a class seems unusable because all the constructors or
2847 destructors in that class are private, and it has neither friends nor
2848 public static member functions. Also warn if there are no non-private
2849 methods, and there's at least one private member function that isn't
2850 a constructor or destructor.
2852 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2853 @opindex Wdelete-non-virtual-dtor
2854 @opindex Wno-delete-non-virtual-dtor
2855 Warn when @code{delete} is used to destroy an instance of a class that
2856 has virtual functions and non-virtual destructor. It is unsafe to delete
2857 an instance of a derived class through a pointer to a base class if the
2858 base class does not have a virtual destructor. This warning is enabled
2861 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2862 @opindex Wliteral-suffix
2863 @opindex Wno-literal-suffix
2864 Warn when a string or character literal is followed by a ud-suffix which does
2865 not begin with an underscore. As a conforming extension, GCC treats such
2866 suffixes as separate preprocessing tokens in order to maintain backwards
2867 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2871 #define __STDC_FORMAT_MACROS
2872 #include <inttypes.h>
2877 printf("My int64: %" PRId64"\n", i64);
2881 In this case, @code{PRId64} is treated as a separate preprocessing token.
2883 Additionally, warn when a user-defined literal operator is declared with
2884 a literal suffix identifier that doesn't begin with an underscore. Literal
2885 suffix identifiers that don't begin with an underscore are reserved for
2886 future standardization.
2888 This warning is enabled by default.
2890 @item -Wlto-type-mismatch
2891 @opindex Wlto-type-mismatch
2892 @opindex Wno-lto-type-mismatch
2894 During the link-time optimization warn about type mismatches in
2895 global declarations from different compilation units.
2896 Requires @option{-flto} to be enabled. Enabled by default.
2898 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2900 @opindex Wno-narrowing
2901 For C++11 and later standards, narrowing conversions are diagnosed by default,
2902 as required by the standard. A narrowing conversion from a constant produces
2903 an error, and a narrowing conversion from a non-constant produces a warning,
2904 but @option{-Wno-narrowing} suppresses the diagnostic.
2905 Note that this does not affect the meaning of well-formed code;
2906 narrowing conversions are still considered ill-formed in SFINAE contexts.
2908 With @option{-Wnarrowing} in C++98, warn when a narrowing
2909 conversion prohibited by C++11 occurs within
2913 int i = @{ 2.2 @}; // error: narrowing from double to int
2916 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2918 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2920 @opindex Wno-noexcept
2921 Warn when a noexcept-expression evaluates to false because of a call
2922 to a function that does not have a non-throwing exception
2923 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2924 the compiler to never throw an exception.
2926 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
2927 @opindex Wnoexcept-type
2928 @opindex Wno-noexcept-type
2929 Warn if the C++17 feature making @code{noexcept} part of a function
2930 type changes the mangled name of a symbol relative to C++14. Enabled
2931 by @option{-Wabi} and @option{-Wc++17-compat}.
2934 template <class T> void f(T t) @{ t(); @};
2936 void h() @{ f(g); @} // in C++14 calls f<void(*)()>, in C++17 calls f<void(*)()noexcept>
2939 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
2940 @opindex Wclass-memaccess
2941 Warn when the destination of a call to a raw memory function such as
2942 @code{memset} or @code{memcpy} is an object of class type writing into which
2943 might bypass the class non-trivial or deleted constructor or copy assignment,
2944 violate const-correctness or encapsulation, or corrupt the virtual table.
2945 Modifying the representation of such objects may violate invariants maintained
2946 by member functions of the class. For example, the call to @code{memset}
2947 below is undefined becase it modifies a non-trivial class object and is,
2948 therefore, diagnosed. The safe way to either initialize or clear the storage
2949 of objects of such types is by using the appropriate constructor or assignment
2950 operator, if one is available.
2952 std::string str = "abc";
2953 memset (&str, 0, 3);
2955 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
2957 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2958 @opindex Wnon-virtual-dtor
2959 @opindex Wno-non-virtual-dtor
2960 Warn when a class has virtual functions and an accessible non-virtual
2961 destructor itself or in an accessible polymorphic base class, in which
2962 case it is possible but unsafe to delete an instance of a derived
2963 class through a pointer to the class itself or base class. This
2964 warning is automatically enabled if @option{-Weffc++} is specified.
2966 @item -Wregister @r{(C++ and Objective-C++ only)}
2968 @opindex Wno-register
2969 Warn on uses of the @code{register} storage class specifier, except
2970 when it is part of the GNU @ref{Explicit Register Variables} extension.
2971 The use of the @code{register} keyword as storage class specifier has
2972 been deprecated in C++11 and removed in C++17.
2973 Enabled by default with @option{-std=c++17}.
2975 @item -Wreorder @r{(C++ and Objective-C++ only)}
2977 @opindex Wno-reorder
2978 @cindex reordering, warning
2979 @cindex warning for reordering of member initializers
2980 Warn when the order of member initializers given in the code does not
2981 match the order in which they must be executed. For instance:
2987 A(): j (0), i (1) @{ @}
2992 The compiler rearranges the member initializers for @code{i}
2993 and @code{j} to match the declaration order of the members, emitting
2994 a warning to that effect. This warning is enabled by @option{-Wall}.
2996 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2997 @opindex fext-numeric-literals
2998 @opindex fno-ext-numeric-literals
2999 Accept imaginary, fixed-point, or machine-defined
3000 literal number suffixes as GNU extensions.
3001 When this option is turned off these suffixes are treated
3002 as C++11 user-defined literal numeric suffixes.
3003 This is on by default for all pre-C++11 dialects and all GNU dialects:
3004 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3005 @option{-std=gnu++14}.
3006 This option is off by default
3007 for ISO C++11 onwards (@option{-std=c++11}, ...).
3010 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3013 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3016 Warn about violations of the following style guidelines from Scott Meyers'
3017 @cite{Effective C++} series of books:
3021 Define a copy constructor and an assignment operator for classes
3022 with dynamically-allocated memory.
3025 Prefer initialization to assignment in constructors.
3028 Have @code{operator=} return a reference to @code{*this}.
3031 Don't try to return a reference when you must return an object.
3034 Distinguish between prefix and postfix forms of increment and
3035 decrement operators.
3038 Never overload @code{&&}, @code{||}, or @code{,}.
3042 This option also enables @option{-Wnon-virtual-dtor}, which is also
3043 one of the effective C++ recommendations. However, the check is
3044 extended to warn about the lack of virtual destructor in accessible
3045 non-polymorphic bases classes too.
3047 When selecting this option, be aware that the standard library
3048 headers do not obey all of these guidelines; use @samp{grep -v}
3049 to filter out those warnings.
3051 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3052 @opindex Wstrict-null-sentinel
3053 @opindex Wno-strict-null-sentinel
3054 Warn about the use of an uncasted @code{NULL} as sentinel. When
3055 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3056 to @code{__null}. Although it is a null pointer constant rather than a
3057 null pointer, it is guaranteed to be of the same size as a pointer.
3058 But this use is not portable across different compilers.
3060 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3061 @opindex Wno-non-template-friend
3062 @opindex Wnon-template-friend
3063 Disable warnings when non-template friend functions are declared
3064 within a template. In very old versions of GCC that predate implementation
3065 of the ISO standard, declarations such as
3066 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3067 could be interpreted as a particular specialization of a template
3068 function; the warning exists to diagnose compatibility problems,
3069 and is enabled by default.
3071 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3072 @opindex Wold-style-cast
3073 @opindex Wno-old-style-cast
3074 Warn if an old-style (C-style) cast to a non-void type is used within
3075 a C++ program. The new-style casts (@code{dynamic_cast},
3076 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3077 less vulnerable to unintended effects and much easier to search for.
3079 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3080 @opindex Woverloaded-virtual
3081 @opindex Wno-overloaded-virtual
3082 @cindex overloaded virtual function, warning
3083 @cindex warning for overloaded virtual function
3084 Warn when a function declaration hides virtual functions from a
3085 base class. For example, in:
3092 struct B: public A @{
3097 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3108 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3109 @opindex Wno-pmf-conversions
3110 @opindex Wpmf-conversions
3111 Disable the diagnostic for converting a bound pointer to member function
3114 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3115 @opindex Wsign-promo
3116 @opindex Wno-sign-promo
3117 Warn when overload resolution chooses a promotion from unsigned or
3118 enumerated type to a signed type, over a conversion to an unsigned type of
3119 the same size. Previous versions of G++ tried to preserve
3120 unsignedness, but the standard mandates the current behavior.
3122 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3124 Warn when a primary template declaration is encountered. Some coding
3125 rules disallow templates, and this may be used to enforce that rule.
3126 The warning is inactive inside a system header file, such as the STL, so
3127 one can still use the STL. One may also instantiate or specialize
3130 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3131 @opindex Wmultiple-inheritance
3132 Warn when a class is defined with multiple direct base classes. Some
3133 coding rules disallow multiple inheritance, and this may be used to
3134 enforce that rule. The warning is inactive inside a system header file,
3135 such as the STL, so one can still use the STL. One may also define
3136 classes that indirectly use multiple inheritance.
3138 @item -Wvirtual-inheritance
3139 @opindex Wvirtual-inheritance
3140 Warn when a class is defined with a virtual direct base class. Some
3141 coding rules disallow multiple inheritance, and this may be used to
3142 enforce that rule. The warning is inactive inside a system header file,
3143 such as the STL, so one can still use the STL. One may also define
3144 classes that indirectly use virtual inheritance.
3147 @opindex Wnamespaces
3148 Warn when a namespace definition is opened. Some coding rules disallow
3149 namespaces, and this may be used to enforce that rule. The warning is
3150 inactive inside a system header file, such as the STL, so one can still
3151 use the STL. One may also use using directives and qualified names.
3153 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3155 @opindex Wno-terminate
3156 Disable the warning about a throw-expression that will immediately
3157 result in a call to @code{terminate}.
3160 @node Objective-C and Objective-C++ Dialect Options
3161 @section Options Controlling Objective-C and Objective-C++ Dialects
3163 @cindex compiler options, Objective-C and Objective-C++
3164 @cindex Objective-C and Objective-C++ options, command-line
3165 @cindex options, Objective-C and Objective-C++
3166 (NOTE: This manual does not describe the Objective-C and Objective-C++
3167 languages themselves. @xref{Standards,,Language Standards
3168 Supported by GCC}, for references.)
3170 This section describes the command-line options that are only meaningful
3171 for Objective-C and Objective-C++ programs. You can also use most of
3172 the language-independent GNU compiler options.
3173 For example, you might compile a file @file{some_class.m} like this:
3176 gcc -g -fgnu-runtime -O -c some_class.m
3180 In this example, @option{-fgnu-runtime} is an option meant only for
3181 Objective-C and Objective-C++ programs; you can use the other options with
3182 any language supported by GCC@.
3184 Note that since Objective-C is an extension of the C language, Objective-C
3185 compilations may also use options specific to the C front-end (e.g.,
3186 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3187 C++-specific options (e.g., @option{-Wabi}).
3189 Here is a list of options that are @emph{only} for compiling Objective-C
3190 and Objective-C++ programs:
3193 @item -fconstant-string-class=@var{class-name}
3194 @opindex fconstant-string-class
3195 Use @var{class-name} as the name of the class to instantiate for each
3196 literal string specified with the syntax @code{@@"@dots{}"}. The default
3197 class name is @code{NXConstantString} if the GNU runtime is being used, and
3198 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3199 @option{-fconstant-cfstrings} option, if also present, overrides the
3200 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3201 to be laid out as constant CoreFoundation strings.
3204 @opindex fgnu-runtime
3205 Generate object code compatible with the standard GNU Objective-C
3206 runtime. This is the default for most types of systems.
3208 @item -fnext-runtime
3209 @opindex fnext-runtime
3210 Generate output compatible with the NeXT runtime. This is the default
3211 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3212 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3215 @item -fno-nil-receivers
3216 @opindex fno-nil-receivers
3217 Assume that all Objective-C message dispatches (@code{[receiver
3218 message:arg]}) in this translation unit ensure that the receiver is
3219 not @code{nil}. This allows for more efficient entry points in the
3220 runtime to be used. This option is only available in conjunction with
3221 the NeXT runtime and ABI version 0 or 1.
3223 @item -fobjc-abi-version=@var{n}
3224 @opindex fobjc-abi-version
3225 Use version @var{n} of the Objective-C ABI for the selected runtime.
3226 This option is currently supported only for the NeXT runtime. In that
3227 case, Version 0 is the traditional (32-bit) ABI without support for
3228 properties and other Objective-C 2.0 additions. Version 1 is the
3229 traditional (32-bit) ABI with support for properties and other
3230 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3231 nothing is specified, the default is Version 0 on 32-bit target
3232 machines, and Version 2 on 64-bit target machines.
3234 @item -fobjc-call-cxx-cdtors
3235 @opindex fobjc-call-cxx-cdtors
3236 For each Objective-C class, check if any of its instance variables is a
3237 C++ object with a non-trivial default constructor. If so, synthesize a
3238 special @code{- (id) .cxx_construct} instance method which runs
3239 non-trivial default constructors on any such instance variables, in order,
3240 and then return @code{self}. Similarly, check if any instance variable
3241 is a C++ object with a non-trivial destructor, and if so, synthesize a
3242 special @code{- (void) .cxx_destruct} method which runs
3243 all such default destructors, in reverse order.
3245 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3246 methods thusly generated only operate on instance variables
3247 declared in the current Objective-C class, and not those inherited
3248 from superclasses. It is the responsibility of the Objective-C
3249 runtime to invoke all such methods in an object's inheritance
3250 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3251 by the runtime immediately after a new object instance is allocated;
3252 the @code{- (void) .cxx_destruct} methods are invoked immediately
3253 before the runtime deallocates an object instance.
3255 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3256 support for invoking the @code{- (id) .cxx_construct} and
3257 @code{- (void) .cxx_destruct} methods.
3259 @item -fobjc-direct-dispatch
3260 @opindex fobjc-direct-dispatch
3261 Allow fast jumps to the message dispatcher. On Darwin this is
3262 accomplished via the comm page.
3264 @item -fobjc-exceptions
3265 @opindex fobjc-exceptions
3266 Enable syntactic support for structured exception handling in
3267 Objective-C, similar to what is offered by C++. This option
3268 is required to use the Objective-C keywords @code{@@try},
3269 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3270 @code{@@synchronized}. This option is available with both the GNU
3271 runtime and the NeXT runtime (but not available in conjunction with
3272 the NeXT runtime on Mac OS X 10.2 and earlier).
3276 Enable garbage collection (GC) in Objective-C and Objective-C++
3277 programs. This option is only available with the NeXT runtime; the
3278 GNU runtime has a different garbage collection implementation that
3279 does not require special compiler flags.
3281 @item -fobjc-nilcheck
3282 @opindex fobjc-nilcheck
3283 For the NeXT runtime with version 2 of the ABI, check for a nil
3284 receiver in method invocations before doing the actual method call.
3285 This is the default and can be disabled using
3286 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3287 checked for nil in this way no matter what this flag is set to.
3288 Currently this flag does nothing when the GNU runtime, or an older
3289 version of the NeXT runtime ABI, is used.
3291 @item -fobjc-std=objc1
3293 Conform to the language syntax of Objective-C 1.0, the language
3294 recognized by GCC 4.0. This only affects the Objective-C additions to
3295 the C/C++ language; it does not affect conformance to C/C++ standards,
3296 which is controlled by the separate C/C++ dialect option flags. When
3297 this option is used with the Objective-C or Objective-C++ compiler,
3298 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3299 This is useful if you need to make sure that your Objective-C code can
3300 be compiled with older versions of GCC@.
3302 @item -freplace-objc-classes
3303 @opindex freplace-objc-classes
3304 Emit a special marker instructing @command{ld(1)} not to statically link in
3305 the resulting object file, and allow @command{dyld(1)} to load it in at
3306 run time instead. This is used in conjunction with the Fix-and-Continue
3307 debugging mode, where the object file in question may be recompiled and
3308 dynamically reloaded in the course of program execution, without the need
3309 to restart the program itself. Currently, Fix-and-Continue functionality
3310 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3315 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3316 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3317 compile time) with static class references that get initialized at load time,
3318 which improves run-time performance. Specifying the @option{-fzero-link} flag
3319 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3320 to be retained. This is useful in Zero-Link debugging mode, since it allows
3321 for individual class implementations to be modified during program execution.
3322 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3323 regardless of command-line options.
3325 @item -fno-local-ivars
3326 @opindex fno-local-ivars
3327 @opindex flocal-ivars
3328 By default instance variables in Objective-C can be accessed as if
3329 they were local variables from within the methods of the class they're
3330 declared in. This can lead to shadowing between instance variables
3331 and other variables declared either locally inside a class method or
3332 globally with the same name. Specifying the @option{-fno-local-ivars}
3333 flag disables this behavior thus avoiding variable shadowing issues.
3335 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3336 @opindex fivar-visibility
3337 Set the default instance variable visibility to the specified option
3338 so that instance variables declared outside the scope of any access
3339 modifier directives default to the specified visibility.
3343 Dump interface declarations for all classes seen in the source file to a
3344 file named @file{@var{sourcename}.decl}.
3346 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3347 @opindex Wassign-intercept
3348 @opindex Wno-assign-intercept
3349 Warn whenever an Objective-C assignment is being intercepted by the
3352 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3353 @opindex Wno-protocol
3355 If a class is declared to implement a protocol, a warning is issued for
3356 every method in the protocol that is not implemented by the class. The
3357 default behavior is to issue a warning for every method not explicitly
3358 implemented in the class, even if a method implementation is inherited
3359 from the superclass. If you use the @option{-Wno-protocol} option, then
3360 methods inherited from the superclass are considered to be implemented,
3361 and no warning is issued for them.
3363 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3365 @opindex Wno-selector
3366 Warn if multiple methods of different types for the same selector are
3367 found during compilation. The check is performed on the list of methods
3368 in the final stage of compilation. Additionally, a check is performed
3369 for each selector appearing in a @code{@@selector(@dots{})}
3370 expression, and a corresponding method for that selector has been found
3371 during compilation. Because these checks scan the method table only at
3372 the end of compilation, these warnings are not produced if the final
3373 stage of compilation is not reached, for example because an error is
3374 found during compilation, or because the @option{-fsyntax-only} option is
3377 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3378 @opindex Wstrict-selector-match
3379 @opindex Wno-strict-selector-match
3380 Warn if multiple methods with differing argument and/or return types are
3381 found for a given selector when attempting to send a message using this
3382 selector to a receiver of type @code{id} or @code{Class}. When this flag
3383 is off (which is the default behavior), the compiler omits such warnings
3384 if any differences found are confined to types that share the same size
3387 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3388 @opindex Wundeclared-selector
3389 @opindex Wno-undeclared-selector
3390 Warn if a @code{@@selector(@dots{})} expression referring to an
3391 undeclared selector is found. A selector is considered undeclared if no
3392 method with that name has been declared before the
3393 @code{@@selector(@dots{})} expression, either explicitly in an
3394 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3395 an @code{@@implementation} section. This option always performs its
3396 checks as soon as a @code{@@selector(@dots{})} expression is found,
3397 while @option{-Wselector} only performs its checks in the final stage of
3398 compilation. This also enforces the coding style convention
3399 that methods and selectors must be declared before being used.
3401 @item -print-objc-runtime-info
3402 @opindex print-objc-runtime-info
3403 Generate C header describing the largest structure that is passed by
3408 @node Diagnostic Message Formatting Options
3409 @section Options to Control Diagnostic Messages Formatting
3410 @cindex options to control diagnostics formatting
3411 @cindex diagnostic messages
3412 @cindex message formatting
3414 Traditionally, diagnostic messages have been formatted irrespective of
3415 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3416 options described below
3417 to control the formatting algorithm for diagnostic messages,
3418 e.g.@: how many characters per line, how often source location
3419 information should be reported. Note that some language front ends may not
3420 honor these options.
3423 @item -fmessage-length=@var{n}
3424 @opindex fmessage-length
3425 Try to format error messages so that they fit on lines of about
3426 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3427 done; each error message appears on a single line. This is the
3428 default for all front ends.
3430 @item -fdiagnostics-show-location=once
3431 @opindex fdiagnostics-show-location
3432 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3433 reporter to emit source location information @emph{once}; that is, in
3434 case the message is too long to fit on a single physical line and has to
3435 be wrapped, the source location won't be emitted (as prefix) again,
3436 over and over, in subsequent continuation lines. This is the default
3439 @item -fdiagnostics-show-location=every-line
3440 Only meaningful in line-wrapping mode. Instructs the diagnostic
3441 messages reporter to emit the same source location information (as
3442 prefix) for physical lines that result from the process of breaking
3443 a message which is too long to fit on a single line.
3445 @item -fdiagnostics-color[=@var{WHEN}]
3446 @itemx -fno-diagnostics-color
3447 @opindex fdiagnostics-color
3448 @cindex highlight, color
3449 @vindex GCC_COLORS @r{environment variable}
3450 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3451 or @samp{auto}. The default depends on how the compiler has been configured,
3452 it can be any of the above @var{WHEN} options or also @samp{never}
3453 if @env{GCC_COLORS} environment variable isn't present in the environment,
3454 and @samp{auto} otherwise.
3455 @samp{auto} means to use color only when the standard error is a terminal.
3456 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3457 aliases for @option{-fdiagnostics-color=always} and
3458 @option{-fdiagnostics-color=never}, respectively.
3460 The colors are defined by the environment variable @env{GCC_COLORS}.
3461 Its value is a colon-separated list of capabilities and Select Graphic
3462 Rendition (SGR) substrings. SGR commands are interpreted by the
3463 terminal or terminal emulator. (See the section in the documentation
3464 of your text terminal for permitted values and their meanings as
3465 character attributes.) These substring values are integers in decimal
3466 representation and can be concatenated with semicolons.
3467 Common values to concatenate include
3469 @samp{4} for underline,
3471 @samp{7} for inverse,
3472 @samp{39} for default foreground color,
3473 @samp{30} to @samp{37} for foreground colors,
3474 @samp{90} to @samp{97} for 16-color mode foreground colors,
3475 @samp{38;5;0} to @samp{38;5;255}
3476 for 88-color and 256-color modes foreground colors,
3477 @samp{49} for default background color,
3478 @samp{40} to @samp{47} for background colors,
3479 @samp{100} to @samp{107} for 16-color mode background colors,
3480 and @samp{48;5;0} to @samp{48;5;255}
3481 for 88-color and 256-color modes background colors.
3483 The default @env{GCC_COLORS} is
3485 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3486 quote=01:fixit-insert=32:fixit-delete=31:\
3487 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3491 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3492 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3493 @samp{01} is bold, and @samp{31} is red.
3494 Setting @env{GCC_COLORS} to the empty string disables colors.
3495 Supported capabilities are as follows.
3499 @vindex error GCC_COLORS @r{capability}
3500 SGR substring for error: markers.
3503 @vindex warning GCC_COLORS @r{capability}
3504 SGR substring for warning: markers.
3507 @vindex note GCC_COLORS @r{capability}
3508 SGR substring for note: markers.
3511 @vindex range1 GCC_COLORS @r{capability}
3512 SGR substring for first additional range.
3515 @vindex range2 GCC_COLORS @r{capability}
3516 SGR substring for second additional range.
3519 @vindex locus GCC_COLORS @r{capability}
3520 SGR substring for location information, @samp{file:line} or
3521 @samp{file:line:column} etc.
3524 @vindex quote GCC_COLORS @r{capability}
3525 SGR substring for information printed within quotes.
3528 @vindex fixit-insert GCC_COLORS @r{capability}
3529 SGR substring for fix-it hints suggesting text to
3530 be inserted or replaced.
3533 @vindex fixit-delete GCC_COLORS @r{capability}
3534 SGR substring for fix-it hints suggesting text to
3537 @item diff-filename=
3538 @vindex diff-filename GCC_COLORS @r{capability}
3539 SGR substring for filename headers within generated patches.
3542 @vindex diff-hunk GCC_COLORS @r{capability}
3543 SGR substring for the starts of hunks within generated patches.
3546 @vindex diff-delete GCC_COLORS @r{capability}
3547 SGR substring for deleted lines within generated patches.
3550 @vindex diff-insert GCC_COLORS @r{capability}
3551 SGR substring for inserted lines within generated patches.
3554 @vindex type-diff GCC_COLORS @r{capability}
3555 SGR substring for highlighting mismatching types within template
3556 arguments in the C++ frontend.
3559 @item -fno-diagnostics-show-option
3560 @opindex fno-diagnostics-show-option
3561 @opindex fdiagnostics-show-option
3562 By default, each diagnostic emitted includes text indicating the
3563 command-line option that directly controls the diagnostic (if such an
3564 option is known to the diagnostic machinery). Specifying the
3565 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3567 @item -fno-diagnostics-show-caret
3568 @opindex fno-diagnostics-show-caret
3569 @opindex fdiagnostics-show-caret
3570 By default, each diagnostic emitted includes the original source line
3571 and a caret @samp{^} indicating the column. This option suppresses this
3572 information. The source line is truncated to @var{n} characters, if
3573 the @option{-fmessage-length=n} option is given. When the output is done
3574 to the terminal, the width is limited to the width given by the
3575 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3577 @item -fdiagnostics-parseable-fixits
3578 @opindex fdiagnostics-parseable-fixits
3579 Emit fix-it hints in a machine-parseable format, suitable for consumption
3580 by IDEs. For each fix-it, a line will be printed after the relevant
3581 diagnostic, starting with the string ``fix-it:''. For example:
3584 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3587 The location is expressed as a half-open range, expressed as a count of
3588 bytes, starting at byte 1 for the initial column. In the above example,
3589 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3593 00000000011111111112222222222
3594 12345678901234567890123456789
3595 gtk_widget_showall (dlg);
3600 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3601 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3602 (e.g. vertical tab as ``\013'').
3604 An empty replacement string indicates that the given range is to be removed.
3605 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3606 be inserted at the given position.
3608 @item -fdiagnostics-generate-patch
3609 @opindex fdiagnostics-generate-patch
3610 Print fix-it hints to stderr in unified diff format, after any diagnostics
3611 are printed. For example:
3618 void show_cb(GtkDialog *dlg)
3620 - gtk_widget_showall(dlg);
3621 + gtk_widget_show_all(dlg);
3626 The diff may or may not be colorized, following the same rules
3627 as for diagnostics (see @option{-fdiagnostics-color}).
3629 @item -fdiagnostics-show-template-tree
3630 @opindex fdiagnostics-show-template-tree
3632 In the C++ frontend, when printing diagnostics showing mismatching
3633 template types, such as:
3636 could not convert 'std::map<int, std::vector<double> >()'
3637 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3640 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3641 tree-like structure showing the common and differing parts of the types,
3651 The parts that differ are highlighted with color (``double'' and
3652 ``float'' in this case).
3654 @item -fno-elide-type
3655 @opindex fno-elide-type
3656 @opindex felide-type
3657 By default when the C++ frontend prints diagnostics showing mismatching
3658 template types, common parts of the types are printed as ``[...]'' to
3659 simplify the error message. For example:
3662 could not convert 'std::map<int, std::vector<double> >()'
3663 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3666 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
3667 This flag also affects the output of the
3668 @option{-fdiagnostics-show-template-tree} flag.
3670 @item -fno-show-column
3671 @opindex fno-show-column
3672 Do not print column numbers in diagnostics. This may be necessary if
3673 diagnostics are being scanned by a program that does not understand the
3674 column numbers, such as @command{dejagnu}.
3678 @node Warning Options
3679 @section Options to Request or Suppress Warnings
3680 @cindex options to control warnings
3681 @cindex warning messages
3682 @cindex messages, warning
3683 @cindex suppressing warnings
3685 Warnings are diagnostic messages that report constructions that
3686 are not inherently erroneous but that are risky or suggest there
3687 may have been an error.
3689 The following language-independent options do not enable specific
3690 warnings but control the kinds of diagnostics produced by GCC@.
3693 @cindex syntax checking
3695 @opindex fsyntax-only
3696 Check the code for syntax errors, but don't do anything beyond that.
3698 @item -fmax-errors=@var{n}
3699 @opindex fmax-errors
3700 Limits the maximum number of error messages to @var{n}, at which point
3701 GCC bails out rather than attempting to continue processing the source
3702 code. If @var{n} is 0 (the default), there is no limit on the number
3703 of error messages produced. If @option{-Wfatal-errors} is also
3704 specified, then @option{-Wfatal-errors} takes precedence over this
3709 Inhibit all warning messages.
3714 Make all warnings into errors.
3719 Make the specified warning into an error. The specifier for a warning
3720 is appended; for example @option{-Werror=switch} turns the warnings
3721 controlled by @option{-Wswitch} into errors. This switch takes a
3722 negative form, to be used to negate @option{-Werror} for specific
3723 warnings; for example @option{-Wno-error=switch} makes
3724 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3727 The warning message for each controllable warning includes the
3728 option that controls the warning. That option can then be used with
3729 @option{-Werror=} and @option{-Wno-error=} as described above.
3730 (Printing of the option in the warning message can be disabled using the
3731 @option{-fno-diagnostics-show-option} flag.)
3733 Note that specifying @option{-Werror=}@var{foo} automatically implies
3734 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3737 @item -Wfatal-errors
3738 @opindex Wfatal-errors
3739 @opindex Wno-fatal-errors
3740 This option causes the compiler to abort compilation on the first error
3741 occurred rather than trying to keep going and printing further error
3746 You can request many specific warnings with options beginning with
3747 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3748 implicit declarations. Each of these specific warning options also
3749 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3750 example, @option{-Wno-implicit}. This manual lists only one of the
3751 two forms, whichever is not the default. For further
3752 language-specific options also refer to @ref{C++ Dialect Options} and
3753 @ref{Objective-C and Objective-C++ Dialect Options}.
3755 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3756 options, such as @option{-Wunused}, which may turn on further options,
3757 such as @option{-Wunused-value}. The combined effect of positive and
3758 negative forms is that more specific options have priority over less
3759 specific ones, independently of their position in the command-line. For
3760 options of the same specificity, the last one takes effect. Options
3761 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3762 as if they appeared at the end of the command-line.
3764 When an unrecognized warning option is requested (e.g.,
3765 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3766 that the option is not recognized. However, if the @option{-Wno-} form
3767 is used, the behavior is slightly different: no diagnostic is
3768 produced for @option{-Wno-unknown-warning} unless other diagnostics
3769 are being produced. This allows the use of new @option{-Wno-} options
3770 with old compilers, but if something goes wrong, the compiler
3771 warns that an unrecognized option is present.
3778 Issue all the warnings demanded by strict ISO C and ISO C++;
3779 reject all programs that use forbidden extensions, and some other
3780 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3781 version of the ISO C standard specified by any @option{-std} option used.
3783 Valid ISO C and ISO C++ programs should compile properly with or without
3784 this option (though a rare few require @option{-ansi} or a
3785 @option{-std} option specifying the required version of ISO C)@. However,
3786 without this option, certain GNU extensions and traditional C and C++
3787 features are supported as well. With this option, they are rejected.
3789 @option{-Wpedantic} does not cause warning messages for use of the
3790 alternate keywords whose names begin and end with @samp{__}. Pedantic
3791 warnings are also disabled in the expression that follows
3792 @code{__extension__}. However, only system header files should use
3793 these escape routes; application programs should avoid them.
3794 @xref{Alternate Keywords}.
3796 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3797 C conformance. They soon find that it does not do quite what they want:
3798 it finds some non-ISO practices, but not all---only those for which
3799 ISO C @emph{requires} a diagnostic, and some others for which
3800 diagnostics have been added.
3802 A feature to report any failure to conform to ISO C might be useful in
3803 some instances, but would require considerable additional work and would
3804 be quite different from @option{-Wpedantic}. We don't have plans to
3805 support such a feature in the near future.
3807 Where the standard specified with @option{-std} represents a GNU
3808 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3809 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3810 extended dialect is based. Warnings from @option{-Wpedantic} are given
3811 where they are required by the base standard. (It does not make sense
3812 for such warnings to be given only for features not in the specified GNU
3813 C dialect, since by definition the GNU dialects of C include all
3814 features the compiler supports with the given option, and there would be
3815 nothing to warn about.)
3817 @item -pedantic-errors
3818 @opindex pedantic-errors
3819 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3820 requires a diagnostic, in some cases where there is undefined behavior
3821 at compile-time and in some other cases that do not prevent compilation
3822 of programs that are valid according to the standard. This is not
3823 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3824 by this option and not enabled by the latter and vice versa.
3829 This enables all the warnings about constructions that some users
3830 consider questionable, and that are easy to avoid (or modify to
3831 prevent the warning), even in conjunction with macros. This also
3832 enables some language-specific warnings described in @ref{C++ Dialect
3833 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3835 @option{-Wall} turns on the following warning flags:
3837 @gccoptlist{-Waddress @gol
3838 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3840 -Wbool-operation @gol
3841 -Wc++11-compat -Wc++14-compat @gol
3842 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
3843 -Wchar-subscripts @gol
3845 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3846 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3848 -Wint-in-bool-context @gol
3849 -Wimplicit @r{(C and Objective-C only)} @gol
3850 -Wimplicit-int @r{(C and Objective-C only)} @gol
3851 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3852 -Winit-self @r{(only for C++)} @gol
3853 -Wlogical-not-parentheses @gol
3854 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3855 -Wmaybe-uninitialized @gol
3856 -Wmemset-elt-size @gol
3857 -Wmemset-transposed-args @gol
3858 -Wmisleading-indentation @r{(only for C/C++)} @gol
3859 -Wmissing-braces @r{(only for C/ObjC)} @gol
3860 -Wmultistatement-macros @gol
3861 -Wnarrowing @r{(only for C++)} @gol
3863 -Wnonnull-compare @gol
3869 -Wsequence-point @gol
3870 -Wsign-compare @r{(only in C++)} @gol
3871 -Wsizeof-pointer-div @gol
3872 -Wsizeof-pointer-memaccess @gol
3873 -Wstrict-aliasing @gol
3874 -Wstrict-overflow=1 @gol
3876 -Wtautological-compare @gol
3878 -Wuninitialized @gol
3879 -Wunknown-pragmas @gol
3880 -Wunused-function @gol
3883 -Wunused-variable @gol
3884 -Wvolatile-register-var @gol
3887 Note that some warning flags are not implied by @option{-Wall}. Some of
3888 them warn about constructions that users generally do not consider
3889 questionable, but which occasionally you might wish to check for;
3890 others warn about constructions that are necessary or hard to avoid in
3891 some cases, and there is no simple way to modify the code to suppress
3892 the warning. Some of them are enabled by @option{-Wextra} but many of
3893 them must be enabled individually.
3899 This enables some extra warning flags that are not enabled by
3900 @option{-Wall}. (This option used to be called @option{-W}. The older
3901 name is still supported, but the newer name is more descriptive.)
3903 @gccoptlist{-Wclobbered @gol
3905 -Wignored-qualifiers @gol
3906 -Wimplicit-fallthrough=3 @gol
3907 -Wmissing-field-initializers @gol
3908 -Wmissing-parameter-type @r{(C only)} @gol
3909 -Wold-style-declaration @r{(C only)} @gol
3910 -Woverride-init @gol
3911 -Wsign-compare @r{(C only)} @gol
3913 -Wuninitialized @gol
3914 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3915 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3916 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3919 The option @option{-Wextra} also prints warning messages for the
3925 A pointer is compared against integer zero with @code{<}, @code{<=},
3926 @code{>}, or @code{>=}.
3929 (C++ only) An enumerator and a non-enumerator both appear in a
3930 conditional expression.
3933 (C++ only) Ambiguous virtual bases.
3936 (C++ only) Subscripting an array that has been declared @code{register}.
3939 (C++ only) Taking the address of a variable that has been declared
3943 (C++ only) A base class is not initialized in the copy constructor
3948 @item -Wchar-subscripts
3949 @opindex Wchar-subscripts
3950 @opindex Wno-char-subscripts
3951 Warn if an array subscript has type @code{char}. This is a common cause
3952 of error, as programmers often forget that this type is signed on some
3954 This warning is enabled by @option{-Wall}.
3958 Warn about an invalid memory access that is found by Pointer Bounds Checker
3959 (@option{-fcheck-pointer-bounds}).
3961 @item -Wno-coverage-mismatch
3962 @opindex Wno-coverage-mismatch
3963 Warn if feedback profiles do not match when using the
3964 @option{-fprofile-use} option.
3965 If a source file is changed between compiling with @option{-fprofile-gen} and
3966 with @option{-fprofile-use}, the files with the profile feedback can fail
3967 to match the source file and GCC cannot use the profile feedback
3968 information. By default, this warning is enabled and is treated as an
3969 error. @option{-Wno-coverage-mismatch} can be used to disable the
3970 warning or @option{-Wno-error=coverage-mismatch} can be used to
3971 disable the error. Disabling the error for this warning can result in
3972 poorly optimized code and is useful only in the
3973 case of very minor changes such as bug fixes to an existing code-base.
3974 Completely disabling the warning is not recommended.
3977 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3979 Suppress warning messages emitted by @code{#warning} directives.
3981 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3982 @opindex Wdouble-promotion
3983 @opindex Wno-double-promotion
3984 Give a warning when a value of type @code{float} is implicitly
3985 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3986 floating-point unit implement @code{float} in hardware, but emulate
3987 @code{double} in software. On such a machine, doing computations
3988 using @code{double} values is much more expensive because of the
3989 overhead required for software emulation.
3991 It is easy to accidentally do computations with @code{double} because
3992 floating-point literals are implicitly of type @code{double}. For
3996 float area(float radius)
3998 return 3.14159 * radius * radius;
4002 the compiler performs the entire computation with @code{double}
4003 because the floating-point literal is a @code{double}.
4005 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4006 @opindex Wduplicate-decl-specifier
4007 @opindex Wno-duplicate-decl-specifier
4008 Warn if a declaration has duplicate @code{const}, @code{volatile},
4009 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4013 @itemx -Wformat=@var{n}
4016 @opindex ffreestanding
4017 @opindex fno-builtin
4019 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4020 the arguments supplied have types appropriate to the format string
4021 specified, and that the conversions specified in the format string make
4022 sense. This includes standard functions, and others specified by format
4023 attributes (@pxref{Function Attributes}), in the @code{printf},
4024 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4025 not in the C standard) families (or other target-specific families).
4026 Which functions are checked without format attributes having been
4027 specified depends on the standard version selected, and such checks of
4028 functions without the attribute specified are disabled by
4029 @option{-ffreestanding} or @option{-fno-builtin}.
4031 The formats are checked against the format features supported by GNU
4032 libc version 2.2. These include all ISO C90 and C99 features, as well
4033 as features from the Single Unix Specification and some BSD and GNU
4034 extensions. Other library implementations may not support all these
4035 features; GCC does not support warning about features that go beyond a
4036 particular library's limitations. However, if @option{-Wpedantic} is used
4037 with @option{-Wformat}, warnings are given about format features not
4038 in the selected standard version (but not for @code{strfmon} formats,
4039 since those are not in any version of the C standard). @xref{C Dialect
4040 Options,,Options Controlling C Dialect}.
4047 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4048 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4049 @option{-Wformat} also checks for null format arguments for several
4050 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4051 aspects of this level of format checking can be disabled by the
4052 options: @option{-Wno-format-contains-nul},
4053 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4054 @option{-Wformat} is enabled by @option{-Wall}.
4056 @item -Wno-format-contains-nul
4057 @opindex Wno-format-contains-nul
4058 @opindex Wformat-contains-nul
4059 If @option{-Wformat} is specified, do not warn about format strings that
4062 @item -Wno-format-extra-args
4063 @opindex Wno-format-extra-args
4064 @opindex Wformat-extra-args
4065 If @option{-Wformat} is specified, do not warn about excess arguments to a
4066 @code{printf} or @code{scanf} format function. The C standard specifies
4067 that such arguments are ignored.
4069 Where the unused arguments lie between used arguments that are
4070 specified with @samp{$} operand number specifications, normally
4071 warnings are still given, since the implementation could not know what
4072 type to pass to @code{va_arg} to skip the unused arguments. However,
4073 in the case of @code{scanf} formats, this option suppresses the
4074 warning if the unused arguments are all pointers, since the Single
4075 Unix Specification says that such unused arguments are allowed.
4077 @item -Wformat-overflow
4078 @itemx -Wformat-overflow=@var{level}
4079 @opindex Wformat-overflow
4080 @opindex Wno-format-overflow
4081 Warn about calls to formatted input/output functions such as @code{sprintf}
4082 and @code{vsprintf} that might overflow the destination buffer. When the
4083 exact number of bytes written by a format directive cannot be determined
4084 at compile-time it is estimated based on heuristics that depend on the
4085 @var{level} argument and on optimization. While enabling optimization
4086 will in most cases improve the accuracy of the warning, it may also
4087 result in false positives.
4090 @item -Wformat-overflow
4091 @item -Wformat-overflow=1
4092 @opindex Wformat-overflow
4093 @opindex Wno-format-overflow
4094 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4095 employs a conservative approach that warns only about calls that most
4096 likely overflow the buffer. At this level, numeric arguments to format
4097 directives with unknown values are assumed to have the value of one, and
4098 strings of unknown length to be empty. Numeric arguments that are known
4099 to be bounded to a subrange of their type, or string arguments whose output
4100 is bounded either by their directive's precision or by a finite set of
4101 string literals, are assumed to take on the value within the range that
4102 results in the most bytes on output. For example, the call to @code{sprintf}
4103 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4104 the terminating NUL character (@code{'\0'}) appended by the function
4105 to the destination buffer will be written past its end. Increasing
4106 the size of the buffer by a single byte is sufficient to avoid the
4107 warning, though it may not be sufficient to avoid the overflow.
4110 void f (int a, int b)
4113 sprintf (buf, "a = %i, b = %i\n", a, b);
4117 @item -Wformat-overflow=2
4118 Level @var{2} warns also about calls that might overflow the destination
4119 buffer given an argument of sufficient length or magnitude. At level
4120 @var{2}, unknown numeric arguments are assumed to have the minimum
4121 representable value for signed types with a precision greater than 1, and
4122 the maximum representable value otherwise. Unknown string arguments whose
4123 length cannot be assumed to be bounded either by the directive's precision,
4124 or by a finite set of string literals they may evaluate to, or the character
4125 array they may point to, are assumed to be 1 character long.
4127 At level @var{2}, the call in the example above is again diagnosed, but
4128 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4129 @code{%i} directive will write some of its digits beyond the end of
4130 the destination buffer. To make the call safe regardless of the values
4131 of the two variables, the size of the destination buffer must be increased
4132 to at least 34 bytes. GCC includes the minimum size of the buffer in
4133 an informational note following the warning.
4135 An alternative to increasing the size of the destination buffer is to
4136 constrain the range of formatted values. The maximum length of string
4137 arguments can be bounded by specifying the precision in the format
4138 directive. When numeric arguments of format directives can be assumed
4139 to be bounded by less than the precision of their type, choosing
4140 an appropriate length modifier to the format specifier will reduce
4141 the required buffer size. For example, if @var{a} and @var{b} in the
4142 example above can be assumed to be within the precision of
4143 the @code{short int} type then using either the @code{%hi} format
4144 directive or casting the argument to @code{short} reduces the maximum
4145 required size of the buffer to 24 bytes.
4148 void f (int a, int b)
4151 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4156 @item -Wno-format-zero-length
4157 @opindex Wno-format-zero-length
4158 @opindex Wformat-zero-length
4159 If @option{-Wformat} is specified, do not warn about zero-length formats.
4160 The C standard specifies that zero-length formats are allowed.
4165 Enable @option{-Wformat} plus additional format checks. Currently
4166 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4169 @item -Wformat-nonliteral
4170 @opindex Wformat-nonliteral
4171 @opindex Wno-format-nonliteral
4172 If @option{-Wformat} is specified, also warn if the format string is not a
4173 string literal and so cannot be checked, unless the format function
4174 takes its format arguments as a @code{va_list}.
4176 @item -Wformat-security
4177 @opindex Wformat-security
4178 @opindex Wno-format-security
4179 If @option{-Wformat} is specified, also warn about uses of format
4180 functions that represent possible security problems. At present, this
4181 warns about calls to @code{printf} and @code{scanf} functions where the
4182 format string is not a string literal and there are no format arguments,
4183 as in @code{printf (foo);}. This may be a security hole if the format
4184 string came from untrusted input and contains @samp{%n}. (This is
4185 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4186 in future warnings may be added to @option{-Wformat-security} that are not
4187 included in @option{-Wformat-nonliteral}.)
4189 @item -Wformat-signedness
4190 @opindex Wformat-signedness
4191 @opindex Wno-format-signedness
4192 If @option{-Wformat} is specified, also warn if the format string
4193 requires an unsigned argument and the argument is signed and vice versa.
4195 @item -Wformat-truncation
4196 @itemx -Wformat-truncation=@var{level}
4197 @opindex Wformat-truncation
4198 @opindex Wno-format-truncation
4199 Warn about calls to formatted input/output functions such as @code{snprintf}
4200 and @code{vsnprintf} that might result in output truncation. When the exact
4201 number of bytes written by a format directive cannot be determined at
4202 compile-time it is estimated based on heuristics that depend on
4203 the @var{level} argument and on optimization. While enabling optimization
4204 will in most cases improve the accuracy of the warning, it may also result
4205 in false positives. Except as noted otherwise, the option uses the same
4206 logic @option{-Wformat-overflow}.
4209 @item -Wformat-truncation
4210 @item -Wformat-truncation=1
4211 @opindex Wformat-truncation
4212 @opindex Wno-format-overflow
4213 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4214 employs a conservative approach that warns only about calls to bounded
4215 functions whose return value is unused and that will most likely result
4216 in output truncation.
4218 @item -Wformat-truncation=2
4219 Level @var{2} warns also about calls to bounded functions whose return
4220 value is used and that might result in truncation given an argument of
4221 sufficient length or magnitude.
4225 @opindex Wformat-y2k
4226 @opindex Wno-format-y2k
4227 If @option{-Wformat} is specified, also warn about @code{strftime}
4228 formats that may yield only a two-digit year.
4233 @opindex Wno-nonnull
4234 Warn about passing a null pointer for arguments marked as
4235 requiring a non-null value by the @code{nonnull} function attribute.
4237 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4238 can be disabled with the @option{-Wno-nonnull} option.
4240 @item -Wnonnull-compare
4241 @opindex Wnonnull-compare
4242 @opindex Wno-nonnull-compare
4243 Warn when comparing an argument marked with the @code{nonnull}
4244 function attribute against null inside the function.
4246 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4247 can be disabled with the @option{-Wno-nonnull-compare} option.
4249 @item -Wnull-dereference
4250 @opindex Wnull-dereference
4251 @opindex Wno-null-dereference
4252 Warn if the compiler detects paths that trigger erroneous or
4253 undefined behavior due to dereferencing a null pointer. This option
4254 is only active when @option{-fdelete-null-pointer-checks} is active,
4255 which is enabled by optimizations in most targets. The precision of
4256 the warnings depends on the optimization options used.
4258 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4260 @opindex Wno-init-self
4261 Warn about uninitialized variables that are initialized with themselves.
4262 Note this option can only be used with the @option{-Wuninitialized} option.
4264 For example, GCC warns about @code{i} being uninitialized in the
4265 following snippet only when @option{-Winit-self} has been specified:
4276 This warning is enabled by @option{-Wall} in C++.
4278 @item -Wimplicit-int @r{(C and Objective-C only)}
4279 @opindex Wimplicit-int
4280 @opindex Wno-implicit-int
4281 Warn when a declaration does not specify a type.
4282 This warning is enabled by @option{-Wall}.
4284 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4285 @opindex Wimplicit-function-declaration
4286 @opindex Wno-implicit-function-declaration
4287 Give a warning whenever a function is used before being declared. In
4288 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4289 enabled by default and it is made into an error by
4290 @option{-pedantic-errors}. This warning is also enabled by
4293 @item -Wimplicit @r{(C and Objective-C only)}
4295 @opindex Wno-implicit
4296 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4297 This warning is enabled by @option{-Wall}.
4299 @item -Wimplicit-fallthrough
4300 @opindex Wimplicit-fallthrough
4301 @opindex Wno-implicit-fallthrough
4302 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4303 and @option{-Wno-implicit-fallthrough} is the same as
4304 @option{-Wimplicit-fallthrough=0}.
4306 @item -Wimplicit-fallthrough=@var{n}
4307 @opindex Wimplicit-fallthrough=
4308 Warn when a switch case falls through. For example:
4326 This warning does not warn when the last statement of a case cannot
4327 fall through, e.g. when there is a return statement or a call to function
4328 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4329 also takes into account control flow statements, such as ifs, and only
4330 warns when appropriate. E.g.@:
4340 @} else if (i < 1) @{
4350 Since there are occasions where a switch case fall through is desirable,
4351 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4352 to be used along with a null statement to suppress this warning that
4353 would normally occur:
4361 __attribute__ ((fallthrough));
4368 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4369 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4370 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4371 Instead of these attributes, it is also possible to add a fallthrough comment
4372 to silence the warning. The whole body of the C or C++ style comment should
4373 match the given regular expressions listed below. The option argument @var{n}
4374 specifies what kind of comments are accepted:
4378 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4380 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4381 expression, any comment is used as fallthrough comment.
4383 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4384 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4386 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4387 following regular expressions:
4391 @item @code{-fallthrough}
4393 @item @code{@@fallthrough@@}
4395 @item @code{lint -fallthrough[ \t]*}
4397 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4399 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4401 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4405 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4406 following regular expressions:
4410 @item @code{-fallthrough}
4412 @item @code{@@fallthrough@@}
4414 @item @code{lint -fallthrough[ \t]*}
4416 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4420 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4421 fallthrough comments, only attributes disable the warning.
4425 The comment needs to be followed after optional whitespace and other comments
4426 by @code{case} or @code{default} keywords or by a user label that precedes some
4427 @code{case} or @code{default} label.
4442 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4444 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4445 @opindex Wif-not-aligned
4446 @opindex Wno-if-not-aligned
4447 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4448 should be issued. This is is enabled by default.
4449 Use @option{-Wno-if-not-aligned} to disable it.
4451 @item -Wignored-qualifiers @r{(C and C++ only)}
4452 @opindex Wignored-qualifiers
4453 @opindex Wno-ignored-qualifiers
4454 Warn if the return type of a function has a type qualifier
4455 such as @code{const}. For ISO C such a type qualifier has no effect,
4456 since the value returned by a function is not an lvalue.
4457 For C++, the warning is only emitted for scalar types or @code{void}.
4458 ISO C prohibits qualified @code{void} return types on function
4459 definitions, so such return types always receive a warning
4460 even without this option.
4462 This warning is also enabled by @option{-Wextra}.
4464 @item -Wignored-attributes @r{(C and C++ only)}
4465 @opindex Wignored-attributes
4466 @opindex Wno-ignored-attributes
4467 Warn when an attribute is ignored. This is different from the
4468 @option{-Wattributes} option in that it warns whenever the compiler decides
4469 to drop an attribute, not that the attribute is either unknown, used in a
4470 wrong place, etc. This warning is enabled by default.
4475 Warn if the type of @code{main} is suspicious. @code{main} should be
4476 a function with external linkage, returning int, taking either zero
4477 arguments, two, or three arguments of appropriate types. This warning
4478 is enabled by default in C++ and is enabled by either @option{-Wall}
4479 or @option{-Wpedantic}.
4481 @item -Wmisleading-indentation @r{(C and C++ only)}
4482 @opindex Wmisleading-indentation
4483 @opindex Wno-misleading-indentation
4484 Warn when the indentation of the code does not reflect the block structure.
4485 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4486 @code{for} clauses with a guarded statement that does not use braces,
4487 followed by an unguarded statement with the same indentation.
4489 In the following example, the call to ``bar'' is misleadingly indented as
4490 if it were guarded by the ``if'' conditional.
4493 if (some_condition ())
4495 bar (); /* Gotcha: this is not guarded by the "if". */
4498 In the case of mixed tabs and spaces, the warning uses the
4499 @option{-ftabstop=} option to determine if the statements line up
4502 The warning is not issued for code involving multiline preprocessor logic
4503 such as the following example.
4508 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4514 The warning is not issued after a @code{#line} directive, since this
4515 typically indicates autogenerated code, and no assumptions can be made
4516 about the layout of the file that the directive references.
4518 This warning is enabled by @option{-Wall} in C and C++.
4520 @item -Wmissing-braces
4521 @opindex Wmissing-braces
4522 @opindex Wno-missing-braces
4523 Warn if an aggregate or union initializer is not fully bracketed. In
4524 the following example, the initializer for @code{a} is not fully
4525 bracketed, but that for @code{b} is fully bracketed. This warning is
4526 enabled by @option{-Wall} in C.
4529 int a[2][2] = @{ 0, 1, 2, 3 @};
4530 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4533 This warning is enabled by @option{-Wall}.
4535 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4536 @opindex Wmissing-include-dirs
4537 @opindex Wno-missing-include-dirs
4538 Warn if a user-supplied include directory does not exist.
4540 @item -Wmultistatement-macros
4541 @opindex Wmultistatement-macros
4542 @opindex Wno-multistatement-macros
4543 Warn about unsafe multiple statement macros that appear to be guarded
4544 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
4545 @code{while}, in which only the first statement is actually guarded after
4546 the macro is expanded.
4551 #define DOIT x++; y++
4556 will increment @code{y} unconditionally, not just when @code{c} holds.
4557 The can usually be fixed by wrapping the macro in a do-while loop:
4559 #define DOIT do @{ x++; y++; @} while (0)
4564 This warning is enabled by @option{-Wall} in C and C++.
4567 @opindex Wparentheses
4568 @opindex Wno-parentheses
4569 Warn if parentheses are omitted in certain contexts, such
4570 as when there is an assignment in a context where a truth value
4571 is expected, or when operators are nested whose precedence people
4572 often get confused about.
4574 Also warn if a comparison like @code{x<=y<=z} appears; this is
4575 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4576 interpretation from that of ordinary mathematical notation.
4578 Also warn for dangerous uses of the GNU extension to
4579 @code{?:} with omitted middle operand. When the condition
4580 in the @code{?}: operator is a boolean expression, the omitted value is
4581 always 1. Often programmers expect it to be a value computed
4582 inside the conditional expression instead.
4584 For C++ this also warns for some cases of unnecessary parentheses in
4585 declarations, which can indicate an attempt at a function call instead
4589 // Declares a local variable called mymutex.
4590 std::unique_lock<std::mutex> (mymutex);
4591 // User meant std::unique_lock<std::mutex> lock (mymutex);
4595 This warning is enabled by @option{-Wall}.
4597 @item -Wsequence-point
4598 @opindex Wsequence-point
4599 @opindex Wno-sequence-point
4600 Warn about code that may have undefined semantics because of violations
4601 of sequence point rules in the C and C++ standards.
4603 The C and C++ standards define the order in which expressions in a C/C++
4604 program are evaluated in terms of @dfn{sequence points}, which represent
4605 a partial ordering between the execution of parts of the program: those
4606 executed before the sequence point, and those executed after it. These
4607 occur after the evaluation of a full expression (one which is not part
4608 of a larger expression), after the evaluation of the first operand of a
4609 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4610 function is called (but after the evaluation of its arguments and the
4611 expression denoting the called function), and in certain other places.
4612 Other than as expressed by the sequence point rules, the order of
4613 evaluation of subexpressions of an expression is not specified. All
4614 these rules describe only a partial order rather than a total order,
4615 since, for example, if two functions are called within one expression
4616 with no sequence point between them, the order in which the functions
4617 are called is not specified. However, the standards committee have
4618 ruled that function calls do not overlap.
4620 It is not specified when between sequence points modifications to the
4621 values of objects take effect. Programs whose behavior depends on this
4622 have undefined behavior; the C and C++ standards specify that ``Between
4623 the previous and next sequence point an object shall have its stored
4624 value modified at most once by the evaluation of an expression.
4625 Furthermore, the prior value shall be read only to determine the value
4626 to be stored.''. If a program breaks these rules, the results on any
4627 particular implementation are entirely unpredictable.
4629 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4630 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4631 diagnosed by this option, and it may give an occasional false positive
4632 result, but in general it has been found fairly effective at detecting
4633 this sort of problem in programs.
4635 The C++17 standard will define the order of evaluation of operands in
4636 more cases: in particular it requires that the right-hand side of an
4637 assignment be evaluated before the left-hand side, so the above
4638 examples are no longer undefined. But this warning will still warn
4639 about them, to help people avoid writing code that is undefined in C
4640 and earlier revisions of C++.
4642 The standard is worded confusingly, therefore there is some debate
4643 over the precise meaning of the sequence point rules in subtle cases.
4644 Links to discussions of the problem, including proposed formal
4645 definitions, may be found on the GCC readings page, at
4646 @uref{http://gcc.gnu.org/@/readings.html}.
4648 This warning is enabled by @option{-Wall} for C and C++.
4650 @item -Wno-return-local-addr
4651 @opindex Wno-return-local-addr
4652 @opindex Wreturn-local-addr
4653 Do not warn about returning a pointer (or in C++, a reference) to a
4654 variable that goes out of scope after the function returns.
4657 @opindex Wreturn-type
4658 @opindex Wno-return-type
4659 Warn whenever a function is defined with a return type that defaults
4660 to @code{int}. Also warn about any @code{return} statement with no
4661 return value in a function whose return type is not @code{void}
4662 (falling off the end of the function body is considered returning
4665 For C only, warn about a @code{return} statement with an expression in a
4666 function whose return type is @code{void}, unless the expression type is
4667 also @code{void}. As a GNU extension, the latter case is accepted
4668 without a warning unless @option{-Wpedantic} is used.
4670 For C++, a function without return type always produces a diagnostic
4671 message, even when @option{-Wno-return-type} is specified. The only
4672 exceptions are @code{main} and functions defined in system headers.
4674 This warning is enabled by @option{-Wall}.
4676 @item -Wshift-count-negative
4677 @opindex Wshift-count-negative
4678 @opindex Wno-shift-count-negative
4679 Warn if shift count is negative. This warning is enabled by default.
4681 @item -Wshift-count-overflow
4682 @opindex Wshift-count-overflow
4683 @opindex Wno-shift-count-overflow
4684 Warn if shift count >= width of type. This warning is enabled by default.
4686 @item -Wshift-negative-value
4687 @opindex Wshift-negative-value
4688 @opindex Wno-shift-negative-value
4689 Warn if left shifting a negative value. This warning is enabled by
4690 @option{-Wextra} in C99 and C++11 modes (and newer).
4692 @item -Wshift-overflow
4693 @itemx -Wshift-overflow=@var{n}
4694 @opindex Wshift-overflow
4695 @opindex Wno-shift-overflow
4696 Warn about left shift overflows. This warning is enabled by
4697 default in C99 and C++11 modes (and newer).
4700 @item -Wshift-overflow=1
4701 This is the warning level of @option{-Wshift-overflow} and is enabled
4702 by default in C99 and C++11 modes (and newer). This warning level does
4703 not warn about left-shifting 1 into the sign bit. (However, in C, such
4704 an overflow is still rejected in contexts where an integer constant expression
4707 @item -Wshift-overflow=2
4708 This warning level also warns about left-shifting 1 into the sign bit,
4709 unless C++14 mode is active.
4715 Warn whenever a @code{switch} statement has an index of enumerated type
4716 and lacks a @code{case} for one or more of the named codes of that
4717 enumeration. (The presence of a @code{default} label prevents this
4718 warning.) @code{case} labels outside the enumeration range also
4719 provoke warnings when this option is used (even if there is a
4720 @code{default} label).
4721 This warning is enabled by @option{-Wall}.
4723 @item -Wswitch-default
4724 @opindex Wswitch-default
4725 @opindex Wno-switch-default
4726 Warn whenever a @code{switch} statement does not have a @code{default}
4730 @opindex Wswitch-enum
4731 @opindex Wno-switch-enum
4732 Warn whenever a @code{switch} statement has an index of enumerated type
4733 and lacks a @code{case} for one or more of the named codes of that
4734 enumeration. @code{case} labels outside the enumeration range also
4735 provoke warnings when this option is used. The only difference
4736 between @option{-Wswitch} and this option is that this option gives a
4737 warning about an omitted enumeration code even if there is a
4738 @code{default} label.
4741 @opindex Wswitch-bool
4742 @opindex Wno-switch-bool
4743 Warn whenever a @code{switch} statement has an index of boolean type
4744 and the case values are outside the range of a boolean type.
4745 It is possible to suppress this warning by casting the controlling
4746 expression to a type other than @code{bool}. For example:
4749 switch ((int) (a == 4))
4755 This warning is enabled by default for C and C++ programs.
4757 @item -Wswitch-unreachable
4758 @opindex Wswitch-unreachable
4759 @opindex Wno-switch-unreachable
4760 Warn whenever a @code{switch} statement contains statements between the
4761 controlling expression and the first case label, which will never be
4762 executed. For example:
4774 @option{-Wswitch-unreachable} does not warn if the statement between the
4775 controlling expression and the first case label is just a declaration:
4788 This warning is enabled by default for C and C++ programs.
4790 @item -Wsync-nand @r{(C and C++ only)}
4792 @opindex Wno-sync-nand
4793 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4794 built-in functions are used. These functions changed semantics in GCC 4.4.
4796 @item -Wunused-but-set-parameter
4797 @opindex Wunused-but-set-parameter
4798 @opindex Wno-unused-but-set-parameter
4799 Warn whenever a function parameter is assigned to, but otherwise unused
4800 (aside from its declaration).
4802 To suppress this warning use the @code{unused} attribute
4803 (@pxref{Variable Attributes}).
4805 This warning is also enabled by @option{-Wunused} together with
4808 @item -Wunused-but-set-variable
4809 @opindex Wunused-but-set-variable
4810 @opindex Wno-unused-but-set-variable
4811 Warn whenever a local variable is assigned to, but otherwise unused
4812 (aside from its declaration).
4813 This warning is enabled by @option{-Wall}.
4815 To suppress this warning use the @code{unused} attribute
4816 (@pxref{Variable Attributes}).
4818 This warning is also enabled by @option{-Wunused}, which is enabled
4821 @item -Wunused-function
4822 @opindex Wunused-function
4823 @opindex Wno-unused-function
4824 Warn whenever a static function is declared but not defined or a
4825 non-inline static function is unused.
4826 This warning is enabled by @option{-Wall}.
4828 @item -Wunused-label
4829 @opindex Wunused-label
4830 @opindex Wno-unused-label
4831 Warn whenever a label is declared but not used.
4832 This warning is enabled by @option{-Wall}.
4834 To suppress this warning use the @code{unused} attribute
4835 (@pxref{Variable Attributes}).
4837 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4838 @opindex Wunused-local-typedefs
4839 Warn when a typedef locally defined in a function is not used.
4840 This warning is enabled by @option{-Wall}.
4842 @item -Wunused-parameter
4843 @opindex Wunused-parameter
4844 @opindex Wno-unused-parameter
4845 Warn whenever a function parameter is unused aside from its declaration.
4847 To suppress this warning use the @code{unused} attribute
4848 (@pxref{Variable Attributes}).
4850 @item -Wno-unused-result
4851 @opindex Wunused-result
4852 @opindex Wno-unused-result
4853 Do not warn if a caller of a function marked with attribute
4854 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4855 its return value. The default is @option{-Wunused-result}.
4857 @item -Wunused-variable
4858 @opindex Wunused-variable
4859 @opindex Wno-unused-variable
4860 Warn whenever a local or static variable is unused aside from its
4861 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4862 but not for C++. This warning is enabled by @option{-Wall}.
4864 To suppress this warning use the @code{unused} attribute
4865 (@pxref{Variable Attributes}).
4867 @item -Wunused-const-variable
4868 @itemx -Wunused-const-variable=@var{n}
4869 @opindex Wunused-const-variable
4870 @opindex Wno-unused-const-variable
4871 Warn whenever a constant static variable is unused aside from its declaration.
4872 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4873 for C, but not for C++. In C this declares variable storage, but in C++ this
4874 is not an error since const variables take the place of @code{#define}s.
4876 To suppress this warning use the @code{unused} attribute
4877 (@pxref{Variable Attributes}).
4880 @item -Wunused-const-variable=1
4881 This is the warning level that is enabled by @option{-Wunused-variable} for
4882 C. It warns only about unused static const variables defined in the main
4883 compilation unit, but not about static const variables declared in any
4886 @item -Wunused-const-variable=2
4887 This warning level also warns for unused constant static variables in
4888 headers (excluding system headers). This is the warning level of
4889 @option{-Wunused-const-variable} and must be explicitly requested since
4890 in C++ this isn't an error and in C it might be harder to clean up all
4894 @item -Wunused-value
4895 @opindex Wunused-value
4896 @opindex Wno-unused-value
4897 Warn whenever a statement computes a result that is explicitly not
4898 used. To suppress this warning cast the unused expression to
4899 @code{void}. This includes an expression-statement or the left-hand
4900 side of a comma expression that contains no side effects. For example,
4901 an expression such as @code{x[i,j]} causes a warning, while
4902 @code{x[(void)i,j]} does not.
4904 This warning is enabled by @option{-Wall}.
4909 All the above @option{-Wunused} options combined.
4911 In order to get a warning about an unused function parameter, you must
4912 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4913 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4915 @item -Wuninitialized
4916 @opindex Wuninitialized
4917 @opindex Wno-uninitialized
4918 Warn if an automatic variable is used without first being initialized
4919 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4920 warn if a non-static reference or non-static @code{const} member
4921 appears in a class without constructors.
4923 If you want to warn about code that uses the uninitialized value of the
4924 variable in its own initializer, use the @option{-Winit-self} option.
4926 These warnings occur for individual uninitialized or clobbered
4927 elements of structure, union or array variables as well as for
4928 variables that are uninitialized or clobbered as a whole. They do
4929 not occur for variables or elements declared @code{volatile}. Because
4930 these warnings depend on optimization, the exact variables or elements
4931 for which there are warnings depends on the precise optimization
4932 options and version of GCC used.
4934 Note that there may be no warning about a variable that is used only
4935 to compute a value that itself is never used, because such
4936 computations may be deleted by data flow analysis before the warnings
4939 @item -Winvalid-memory-model
4940 @opindex Winvalid-memory-model
4941 @opindex Wno-invalid-memory-model
4942 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4943 and the C11 atomic generic functions with a memory consistency argument
4944 that is either invalid for the operation or outside the range of values
4945 of the @code{memory_order} enumeration. For example, since the
4946 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4947 defined for the relaxed, release, and sequentially consistent memory
4948 orders the following code is diagnosed:
4953 __atomic_store_n (i, 0, memory_order_consume);
4957 @option{-Winvalid-memory-model} is enabled by default.
4959 @item -Wmaybe-uninitialized
4960 @opindex Wmaybe-uninitialized
4961 @opindex Wno-maybe-uninitialized
4962 For an automatic variable, if there exists a path from the function
4963 entry to a use of the variable that is initialized, but there exist
4964 some other paths for which the variable is not initialized, the compiler
4965 emits a warning if it cannot prove the uninitialized paths are not
4966 executed at run time. These warnings are made optional because GCC is
4967 not smart enough to see all the reasons why the code might be correct
4968 in spite of appearing to have an error. Here is one example of how
4989 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4990 always initialized, but GCC doesn't know this. To suppress the
4991 warning, you need to provide a default case with assert(0) or
4994 @cindex @code{longjmp} warnings
4995 This option also warns when a non-volatile automatic variable might be
4996 changed by a call to @code{longjmp}. These warnings as well are possible
4997 only in optimizing compilation.
4999 The compiler sees only the calls to @code{setjmp}. It cannot know
5000 where @code{longjmp} will be called; in fact, a signal handler could
5001 call it at any point in the code. As a result, you may get a warning
5002 even when there is in fact no problem because @code{longjmp} cannot
5003 in fact be called at the place that would cause a problem.
5005 Some spurious warnings can be avoided if you declare all the functions
5006 you use that never return as @code{noreturn}. @xref{Function
5009 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5011 @item -Wunknown-pragmas
5012 @opindex Wunknown-pragmas
5013 @opindex Wno-unknown-pragmas
5014 @cindex warning for unknown pragmas
5015 @cindex unknown pragmas, warning
5016 @cindex pragmas, warning of unknown
5017 Warn when a @code{#pragma} directive is encountered that is not understood by
5018 GCC@. If this command-line option is used, warnings are even issued
5019 for unknown pragmas in system header files. This is not the case if
5020 the warnings are only enabled by the @option{-Wall} command-line option.
5023 @opindex Wno-pragmas
5025 Do not warn about misuses of pragmas, such as incorrect parameters,
5026 invalid syntax, or conflicts between pragmas. See also
5027 @option{-Wunknown-pragmas}.
5029 @item -Wstrict-aliasing
5030 @opindex Wstrict-aliasing
5031 @opindex Wno-strict-aliasing
5032 This option is only active when @option{-fstrict-aliasing} is active.
5033 It warns about code that might break the strict aliasing rules that the
5034 compiler is using for optimization. The warning does not catch all
5035 cases, but does attempt to catch the more common pitfalls. It is
5036 included in @option{-Wall}.
5037 It is equivalent to @option{-Wstrict-aliasing=3}
5039 @item -Wstrict-aliasing=n
5040 @opindex Wstrict-aliasing=n
5041 This option is only active when @option{-fstrict-aliasing} is active.
5042 It warns about code that might break the strict aliasing rules that the
5043 compiler is using for optimization.
5044 Higher levels correspond to higher accuracy (fewer false positives).
5045 Higher levels also correspond to more effort, similar to the way @option{-O}
5047 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5049 Level 1: Most aggressive, quick, least accurate.
5050 Possibly useful when higher levels
5051 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5052 false negatives. However, it has many false positives.
5053 Warns for all pointer conversions between possibly incompatible types,
5054 even if never dereferenced. Runs in the front end only.
5056 Level 2: Aggressive, quick, not too precise.
5057 May still have many false positives (not as many as level 1 though),
5058 and few false negatives (but possibly more than level 1).
5059 Unlike level 1, it only warns when an address is taken. Warns about
5060 incomplete types. Runs in the front end only.
5062 Level 3 (default for @option{-Wstrict-aliasing}):
5063 Should have very few false positives and few false
5064 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5065 Takes care of the common pun+dereference pattern in the front end:
5066 @code{*(int*)&some_float}.
5067 If optimization is enabled, it also runs in the back end, where it deals
5068 with multiple statement cases using flow-sensitive points-to information.
5069 Only warns when the converted pointer is dereferenced.
5070 Does not warn about incomplete types.
5072 @item -Wstrict-overflow
5073 @itemx -Wstrict-overflow=@var{n}
5074 @opindex Wstrict-overflow
5075 @opindex Wno-strict-overflow
5076 This option is only active when signed overflow is undefined.
5077 It warns about cases where the compiler optimizes based on the
5078 assumption that signed overflow does not occur. Note that it does not
5079 warn about all cases where the code might overflow: it only warns
5080 about cases where the compiler implements some optimization. Thus
5081 this warning depends on the optimization level.
5083 An optimization that assumes that signed overflow does not occur is
5084 perfectly safe if the values of the variables involved are such that
5085 overflow never does, in fact, occur. Therefore this warning can
5086 easily give a false positive: a warning about code that is not
5087 actually a problem. To help focus on important issues, several
5088 warning levels are defined. No warnings are issued for the use of
5089 undefined signed overflow when estimating how many iterations a loop
5090 requires, in particular when determining whether a loop will be
5094 @item -Wstrict-overflow=1
5095 Warn about cases that are both questionable and easy to avoid. For
5096 example the compiler simplifies
5097 @code{x + 1 > x} to @code{1}. This level of
5098 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5099 are not, and must be explicitly requested.
5101 @item -Wstrict-overflow=2
5102 Also warn about other cases where a comparison is simplified to a
5103 constant. For example: @code{abs (x) >= 0}. This can only be
5104 simplified when signed integer overflow is undefined, because
5105 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5106 zero. @option{-Wstrict-overflow} (with no level) is the same as
5107 @option{-Wstrict-overflow=2}.
5109 @item -Wstrict-overflow=3
5110 Also warn about other cases where a comparison is simplified. For
5111 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5113 @item -Wstrict-overflow=4
5114 Also warn about other simplifications not covered by the above cases.
5115 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5117 @item -Wstrict-overflow=5
5118 Also warn about cases where the compiler reduces the magnitude of a
5119 constant involved in a comparison. For example: @code{x + 2 > y} is
5120 simplified to @code{x + 1 >= y}. This is reported only at the
5121 highest warning level because this simplification applies to many
5122 comparisons, so this warning level gives a very large number of
5126 @item -Wstringop-overflow
5127 @itemx -Wstringop-overflow=@var{type}
5128 @opindex Wstringop-overflow
5129 @opindex Wno-stringop-overflow
5130 Warn for calls to string manipulation functions such as @code{memcpy} and
5131 @code{strcpy} that are determined to overflow the destination buffer. The
5132 optional argument is one greater than the type of Object Size Checking to
5133 perform to determine the size of the destination. @xref{Object Size Checking}.
5134 The argument is meaningful only for functions that operate on character arrays
5135 but not for raw memory functions like @code{memcpy} which always make use
5136 of Object Size type-0. The option also warns for calls that specify a size
5137 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5138 The option produces the best results with optimization enabled but can detect
5139 a small subset of simple buffer overflows even without optimization in
5140 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5141 correspond to the standard functions. In any case, the option warns about
5142 just a subset of buffer overflows detected by the corresponding overflow
5143 checking built-ins. For example, the option will issue a warning for
5144 the @code{strcpy} call below because it copies at least 5 characters
5145 (the string @code{"blue"} including the terminating NUL) into the buffer
5149 enum Color @{ blue, purple, yellow @};
5150 const char* f (enum Color clr)
5152 static char buf [4];
5156 case blue: str = "blue"; break;
5157 case purple: str = "purple"; break;
5158 case yellow: str = "yellow"; break;
5161 return strcpy (buf, str); // warning here
5165 Option @option{-Wstringop-overflow=2} is enabled by default.
5168 @item -Wstringop-overflow
5169 @item -Wstringop-overflow=1
5170 @opindex Wstringop-overflow
5171 @opindex Wno-stringop-overflow
5172 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5173 to determine the sizes of destination objects. This is the default setting
5174 of the option. At this setting the option will not warn for writes past
5175 the end of subobjects of larger objects accessed by pointers unless the
5176 size of the largest surrounding object is known. When the destination may
5177 be one of several objects it is assumed to be the largest one of them. On
5178 Linux systems, when optimization is enabled at this setting the option warns
5179 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5182 @item -Wstringop-overflow=2
5183 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5184 to determine the sizes of destination objects. At this setting the option
5185 will warn about overflows when writing to members of the largest complete
5186 objects whose exact size is known. It will, however, not warn for excessive
5187 writes to the same members of unknown objects referenced by pointers since
5188 they may point to arrays containing unknown numbers of elements.
5190 @item -Wstringop-overflow=3
5191 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5192 to determine the sizes of destination objects. At this setting the option
5193 warns about overflowing the smallest object or data member. This is the
5194 most restrictive setting of the option that may result in warnings for safe
5197 @item -Wstringop-overflow=4
5198 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5199 to determine the sizes of destination objects. At this setting the option
5200 will warn about overflowing any data members, and when the destination is
5201 one of several objects it uses the size of the largest of them to decide
5202 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5203 setting of the option may result in warnings for benign code.
5206 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{]}
5207 @opindex Wsuggest-attribute=
5208 @opindex Wno-suggest-attribute=
5209 Warn for cases where adding an attribute may be beneficial. The
5210 attributes currently supported are listed below.
5213 @item -Wsuggest-attribute=pure
5214 @itemx -Wsuggest-attribute=const
5215 @itemx -Wsuggest-attribute=noreturn
5216 @opindex Wsuggest-attribute=pure
5217 @opindex Wno-suggest-attribute=pure
5218 @opindex Wsuggest-attribute=const
5219 @opindex Wno-suggest-attribute=const
5220 @opindex Wsuggest-attribute=noreturn
5221 @opindex Wno-suggest-attribute=noreturn
5223 Warn about functions that might be candidates for attributes
5224 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
5225 functions visible in other compilation units or (in the case of @code{pure} and
5226 @code{const}) if it cannot prove that the function returns normally. A function
5227 returns normally if it doesn't contain an infinite loop or return abnormally
5228 by throwing, calling @code{abort} or trapping. This analysis requires option
5229 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
5230 higher. Higher optimization levels improve the accuracy of the analysis.
5232 @item -Wsuggest-attribute=format
5233 @itemx -Wmissing-format-attribute
5234 @opindex Wsuggest-attribute=format
5235 @opindex Wmissing-format-attribute
5236 @opindex Wno-suggest-attribute=format
5237 @opindex Wno-missing-format-attribute
5241 Warn about function pointers that might be candidates for @code{format}
5242 attributes. Note these are only possible candidates, not absolute ones.
5243 GCC guesses that function pointers with @code{format} attributes that
5244 are used in assignment, initialization, parameter passing or return
5245 statements should have a corresponding @code{format} attribute in the
5246 resulting type. I.e.@: the left-hand side of the assignment or
5247 initialization, the type of the parameter variable, or the return type
5248 of the containing function respectively should also have a @code{format}
5249 attribute to avoid the warning.
5251 GCC also warns about function definitions that might be
5252 candidates for @code{format} attributes. Again, these are only
5253 possible candidates. GCC guesses that @code{format} attributes
5254 might be appropriate for any function that calls a function like
5255 @code{vprintf} or @code{vscanf}, but this might not always be the
5256 case, and some functions for which @code{format} attributes are
5257 appropriate may not be detected.
5259 @item -Wsuggest-attribute=cold
5260 @opindex Wsuggest-attribute=cold
5261 @opindex Wno-suggest-attribute=cold
5263 Warn about functions that might be candidates for @code{cold} attribute. This
5264 is based on static detection and generally will only warn about functions which
5265 always leads to a call to another @code{cold} function such as wrappers of
5266 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5269 @item -Wsuggest-final-types
5270 @opindex Wno-suggest-final-types
5271 @opindex Wsuggest-final-types
5272 Warn about types with virtual methods where code quality would be improved
5273 if the type were declared with the C++11 @code{final} specifier,
5275 declared in an anonymous namespace. This allows GCC to more aggressively
5276 devirtualize the polymorphic calls. This warning is more effective with link
5277 time optimization, where the information about the class hierarchy graph is
5280 @item -Wsuggest-final-methods
5281 @opindex Wno-suggest-final-methods
5282 @opindex Wsuggest-final-methods
5283 Warn about virtual methods where code quality would be improved if the method
5284 were declared with the C++11 @code{final} specifier,
5285 or, if possible, its type were
5286 declared in an anonymous namespace or with the @code{final} specifier.
5288 more effective with link-time optimization, where the information about the
5289 class hierarchy graph is more complete. It is recommended to first consider
5290 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5293 @item -Wsuggest-override
5294 Warn about overriding virtual functions that are not marked with the override
5298 @opindex Wno-alloc-zero
5299 @opindex Walloc-zero
5300 Warn about calls to allocation functions decorated with attribute
5301 @code{alloc_size} that specify zero bytes, including those to the built-in
5302 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5303 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5304 when called with a zero size differs among implementations (and in the case
5305 of @code{realloc} has been deprecated) relying on it may result in subtle
5306 portability bugs and should be avoided.
5308 @item -Walloc-size-larger-than=@var{n}
5309 Warn about calls to functions decorated with attribute @code{alloc_size}
5310 that attempt to allocate objects larger than the specified number of bytes,
5311 or where the result of the size computation in an integer type with infinite
5312 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5313 may end in one of the standard suffixes designating a multiple of bytes
5314 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5315 @code{MB} and @code{MiB} for megabyte and mebibyte, and so on.
5316 @xref{Function Attributes}.
5321 This option warns on all uses of @code{alloca} in the source.
5323 @item -Walloca-larger-than=@var{n}
5324 This option warns on calls to @code{alloca} that are not bounded by a
5325 controlling predicate limiting its argument of integer type to at most
5326 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5327 Arguments of non-integer types are considered unbounded even if they
5328 appear to be constrained to the expected range.
5330 For example, a bounded case of @code{alloca} could be:
5333 void func (size_t n)
5344 In the above example, passing @code{-Walloca-larger-than=1000} would not
5345 issue a warning because the call to @code{alloca} is known to be at most
5346 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5347 the compiler would emit a warning.
5349 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5350 controlling predicate constraining its integer argument. For example:
5355 void *p = alloca (n);
5360 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5361 a warning, but this time because of the lack of bounds checking.
5363 Note, that even seemingly correct code involving signed integers could
5367 void func (signed int n)
5377 In the above example, @var{n} could be negative, causing a larger than
5378 expected argument to be implicitly cast into the @code{alloca} call.
5380 This option also warns when @code{alloca} is used in a loop.
5382 This warning is not enabled by @option{-Wall}, and is only active when
5383 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5385 See also @option{-Wvla-larger-than=@var{n}}.
5387 @item -Warray-bounds
5388 @itemx -Warray-bounds=@var{n}
5389 @opindex Wno-array-bounds
5390 @opindex Warray-bounds
5391 This option is only active when @option{-ftree-vrp} is active
5392 (default for @option{-O2} and above). It warns about subscripts to arrays
5393 that are always out of bounds. This warning is enabled by @option{-Wall}.
5396 @item -Warray-bounds=1
5397 This is the warning level of @option{-Warray-bounds} and is enabled
5398 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5400 @item -Warray-bounds=2
5401 This warning level also warns about out of bounds access for
5402 arrays at the end of a struct and for arrays accessed through
5403 pointers. This warning level may give a larger number of
5404 false positives and is deactivated by default.
5407 @item -Wattribute-alias
5408 Warn about declarations using the @code{alias} and similar attributes whose
5409 target is incompatible with the type of the alias. @xref{Function Attributes,
5410 ,Declaring Attributes of Functions}.
5412 @item -Wbool-compare
5413 @opindex Wno-bool-compare
5414 @opindex Wbool-compare
5415 Warn about boolean expression compared with an integer value different from
5416 @code{true}/@code{false}. For instance, the following comparison is
5421 if ((n > 1) == 2) @{ @dots{} @}
5423 This warning is enabled by @option{-Wall}.
5425 @item -Wbool-operation
5426 @opindex Wno-bool-operation
5427 @opindex Wbool-operation
5428 Warn about suspicious operations on expressions of a boolean type. For
5429 instance, bitwise negation of a boolean is very likely a bug in the program.
5430 For C, this warning also warns about incrementing or decrementing a boolean,
5431 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5432 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
5434 This warning is enabled by @option{-Wall}.
5436 @item -Wduplicated-branches
5437 @opindex Wno-duplicated-branches
5438 @opindex Wduplicated-branches
5439 Warn when an if-else has identical branches. This warning detects cases like
5446 It doesn't warn when both branches contain just a null statement. This warning
5447 also warn for conditional operators:
5449 int i = x ? *p : *p;
5452 @item -Wduplicated-cond
5453 @opindex Wno-duplicated-cond
5454 @opindex Wduplicated-cond
5455 Warn about duplicated conditions in an if-else-if chain. For instance,
5456 warn for the following code:
5458 if (p->q != NULL) @{ @dots{} @}
5459 else if (p->q != NULL) @{ @dots{} @}
5462 @item -Wframe-address
5463 @opindex Wno-frame-address
5464 @opindex Wframe-address
5465 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5466 is called with an argument greater than 0. Such calls may return indeterminate
5467 values or crash the program. The warning is included in @option{-Wall}.
5469 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5470 @opindex Wno-discarded-qualifiers
5471 @opindex Wdiscarded-qualifiers
5472 Do not warn if type qualifiers on pointers are being discarded.
5473 Typically, the compiler warns if a @code{const char *} variable is
5474 passed to a function that takes a @code{char *} parameter. This option
5475 can be used to suppress such a warning.
5477 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5478 @opindex Wno-discarded-array-qualifiers
5479 @opindex Wdiscarded-array-qualifiers
5480 Do not warn if type qualifiers on arrays which are pointer targets
5481 are being discarded. Typically, the compiler warns if a
5482 @code{const int (*)[]} variable is passed to a function that
5483 takes a @code{int (*)[]} parameter. This option can be used to
5484 suppress such a warning.
5486 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5487 @opindex Wno-incompatible-pointer-types
5488 @opindex Wincompatible-pointer-types
5489 Do not warn when there is a conversion between pointers that have incompatible
5490 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5491 which warns for pointer argument passing or assignment with different
5494 @item -Wno-int-conversion @r{(C and Objective-C only)}
5495 @opindex Wno-int-conversion
5496 @opindex Wint-conversion
5497 Do not warn about incompatible integer to pointer and pointer to integer
5498 conversions. This warning is about implicit conversions; for explicit
5499 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5500 @option{-Wno-pointer-to-int-cast} may be used.
5502 @item -Wno-div-by-zero
5503 @opindex Wno-div-by-zero
5504 @opindex Wdiv-by-zero
5505 Do not warn about compile-time integer division by zero. Floating-point
5506 division by zero is not warned about, as it can be a legitimate way of
5507 obtaining infinities and NaNs.
5509 @item -Wsystem-headers
5510 @opindex Wsystem-headers
5511 @opindex Wno-system-headers
5512 @cindex warnings from system headers
5513 @cindex system headers, warnings from
5514 Print warning messages for constructs found in system header files.
5515 Warnings from system headers are normally suppressed, on the assumption
5516 that they usually do not indicate real problems and would only make the
5517 compiler output harder to read. Using this command-line option tells
5518 GCC to emit warnings from system headers as if they occurred in user
5519 code. However, note that using @option{-Wall} in conjunction with this
5520 option does @emph{not} warn about unknown pragmas in system
5521 headers---for that, @option{-Wunknown-pragmas} must also be used.
5523 @item -Wtautological-compare
5524 @opindex Wtautological-compare
5525 @opindex Wno-tautological-compare
5526 Warn if a self-comparison always evaluates to true or false. This
5527 warning detects various mistakes such as:
5531 if (i > i) @{ @dots{} @}
5534 This warning also warns about bitwise comparisons that always evaluate
5535 to true or false, for instance:
5537 if ((a & 16) == 10) @{ @dots{} @}
5539 will always be false.
5541 This warning is enabled by @option{-Wall}.
5544 @opindex Wtrampolines
5545 @opindex Wno-trampolines
5546 Warn about trampolines generated for pointers to nested functions.
5547 A trampoline is a small piece of data or code that is created at run
5548 time on the stack when the address of a nested function is taken, and is
5549 used to call the nested function indirectly. For some targets, it is
5550 made up of data only and thus requires no special treatment. But, for
5551 most targets, it is made up of code and thus requires the stack to be
5552 made executable in order for the program to work properly.
5555 @opindex Wfloat-equal
5556 @opindex Wno-float-equal
5557 Warn if floating-point values are used in equality comparisons.
5559 The idea behind this is that sometimes it is convenient (for the
5560 programmer) to consider floating-point values as approximations to
5561 infinitely precise real numbers. If you are doing this, then you need
5562 to compute (by analyzing the code, or in some other way) the maximum or
5563 likely maximum error that the computation introduces, and allow for it
5564 when performing comparisons (and when producing output, but that's a
5565 different problem). In particular, instead of testing for equality, you
5566 should check to see whether the two values have ranges that overlap; and
5567 this is done with the relational operators, so equality comparisons are
5570 @item -Wtraditional @r{(C and Objective-C only)}
5571 @opindex Wtraditional
5572 @opindex Wno-traditional
5573 Warn about certain constructs that behave differently in traditional and
5574 ISO C@. Also warn about ISO C constructs that have no traditional C
5575 equivalent, and/or problematic constructs that should be avoided.
5579 Macro parameters that appear within string literals in the macro body.
5580 In traditional C macro replacement takes place within string literals,
5581 but in ISO C it does not.
5584 In traditional C, some preprocessor directives did not exist.
5585 Traditional preprocessors only considered a line to be a directive
5586 if the @samp{#} appeared in column 1 on the line. Therefore
5587 @option{-Wtraditional} warns about directives that traditional C
5588 understands but ignores because the @samp{#} does not appear as the
5589 first character on the line. It also suggests you hide directives like
5590 @code{#pragma} not understood by traditional C by indenting them. Some
5591 traditional implementations do not recognize @code{#elif}, so this option
5592 suggests avoiding it altogether.
5595 A function-like macro that appears without arguments.
5598 The unary plus operator.
5601 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5602 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5603 constants.) Note, these suffixes appear in macros defined in the system
5604 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5605 Use of these macros in user code might normally lead to spurious
5606 warnings, however GCC's integrated preprocessor has enough context to
5607 avoid warning in these cases.
5610 A function declared external in one block and then used after the end of
5614 A @code{switch} statement has an operand of type @code{long}.
5617 A non-@code{static} function declaration follows a @code{static} one.
5618 This construct is not accepted by some traditional C compilers.
5621 The ISO type of an integer constant has a different width or
5622 signedness from its traditional type. This warning is only issued if
5623 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5624 typically represent bit patterns, are not warned about.
5627 Usage of ISO string concatenation is detected.
5630 Initialization of automatic aggregates.
5633 Identifier conflicts with labels. Traditional C lacks a separate
5634 namespace for labels.
5637 Initialization of unions. If the initializer is zero, the warning is
5638 omitted. This is done under the assumption that the zero initializer in
5639 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5640 initializer warnings and relies on default initialization to zero in the
5644 Conversions by prototypes between fixed/floating-point values and vice
5645 versa. The absence of these prototypes when compiling with traditional
5646 C causes serious problems. This is a subset of the possible
5647 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5650 Use of ISO C style function definitions. This warning intentionally is
5651 @emph{not} issued for prototype declarations or variadic functions
5652 because these ISO C features appear in your code when using
5653 libiberty's traditional C compatibility macros, @code{PARAMS} and
5654 @code{VPARAMS}. This warning is also bypassed for nested functions
5655 because that feature is already a GCC extension and thus not relevant to
5656 traditional C compatibility.
5659 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5660 @opindex Wtraditional-conversion
5661 @opindex Wno-traditional-conversion
5662 Warn if a prototype causes a type conversion that is different from what
5663 would happen to the same argument in the absence of a prototype. This
5664 includes conversions of fixed point to floating and vice versa, and
5665 conversions changing the width or signedness of a fixed-point argument
5666 except when the same as the default promotion.
5668 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5669 @opindex Wdeclaration-after-statement
5670 @opindex Wno-declaration-after-statement
5671 Warn when a declaration is found after a statement in a block. This
5672 construct, known from C++, was introduced with ISO C99 and is by default
5673 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5678 Warn whenever a local variable or type declaration shadows another
5679 variable, parameter, type, class member (in C++), or instance variable
5680 (in Objective-C) or whenever a built-in function is shadowed. Note
5681 that in C++, the compiler warns if a local variable shadows an
5682 explicit typedef, but not if it shadows a struct/class/enum.
5683 Same as @option{-Wshadow=global}.
5685 @item -Wno-shadow-ivar @r{(Objective-C only)}
5686 @opindex Wno-shadow-ivar
5687 @opindex Wshadow-ivar
5688 Do not warn whenever a local variable shadows an instance variable in an
5691 @item -Wshadow=global
5692 @opindex Wshadow=local
5693 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5695 @item -Wshadow=local
5696 @opindex Wshadow=local
5697 Warn when a local variable shadows another local variable or parameter.
5698 This warning is enabled by @option{-Wshadow=global}.
5700 @item -Wshadow=compatible-local
5701 @opindex Wshadow=compatible-local
5702 Warn when a local variable shadows another local variable or parameter
5703 whose type is compatible with that of the shadowing variable. In C++,
5704 type compatibility here means the type of the shadowing variable can be
5705 converted to that of the shadowed variable. The creation of this flag
5706 (in addition to @option{-Wshadow=local}) is based on the idea that when
5707 a local variable shadows another one of incompatible type, it is most
5708 likely intentional, not a bug or typo, as shown in the following example:
5712 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5714 for (int i = 0; i < N; ++i)
5723 Since the two variable @code{i} in the example above have incompatible types,
5724 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5725 Because their types are incompatible, if a programmer accidentally uses one
5726 in place of the other, type checking will catch that and emit an error or
5727 warning. So not warning (about shadowing) in this case will not lead to
5728 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5729 possibly reduce the number of warnings triggered by intentional shadowing.
5731 This warning is enabled by @option{-Wshadow=local}.
5733 @item -Wlarger-than=@var{len}
5734 @opindex Wlarger-than=@var{len}
5735 @opindex Wlarger-than-@var{len}
5736 Warn whenever an object of larger than @var{len} bytes is defined.
5738 @item -Wframe-larger-than=@var{len}
5739 @opindex Wframe-larger-than
5740 Warn if the size of a function frame is larger than @var{len} bytes.
5741 The computation done to determine the stack frame size is approximate
5742 and not conservative.
5743 The actual requirements may be somewhat greater than @var{len}
5744 even if you do not get a warning. In addition, any space allocated
5745 via @code{alloca}, variable-length arrays, or related constructs
5746 is not included by the compiler when determining
5747 whether or not to issue a warning.
5749 @item -Wno-free-nonheap-object
5750 @opindex Wno-free-nonheap-object
5751 @opindex Wfree-nonheap-object
5752 Do not warn when attempting to free an object that was not allocated
5755 @item -Wstack-usage=@var{len}
5756 @opindex Wstack-usage
5757 Warn if the stack usage of a function might be larger than @var{len} bytes.
5758 The computation done to determine the stack usage is conservative.
5759 Any space allocated via @code{alloca}, variable-length arrays, or related
5760 constructs is included by the compiler when determining whether or not to
5763 The message is in keeping with the output of @option{-fstack-usage}.
5767 If the stack usage is fully static but exceeds the specified amount, it's:
5770 warning: stack usage is 1120 bytes
5773 If the stack usage is (partly) dynamic but bounded, it's:
5776 warning: stack usage might be 1648 bytes
5779 If the stack usage is (partly) dynamic and not bounded, it's:
5782 warning: stack usage might be unbounded
5786 @item -Wunsafe-loop-optimizations
5787 @opindex Wunsafe-loop-optimizations
5788 @opindex Wno-unsafe-loop-optimizations
5789 Warn if the loop cannot be optimized because the compiler cannot
5790 assume anything on the bounds of the loop indices. With
5791 @option{-funsafe-loop-optimizations} warn if the compiler makes
5794 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5795 @opindex Wno-pedantic-ms-format
5796 @opindex Wpedantic-ms-format
5797 When used in combination with @option{-Wformat}
5798 and @option{-pedantic} without GNU extensions, this option
5799 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5800 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5801 which depend on the MS runtime.
5804 @opindex Waligned-new
5805 @opindex Wno-aligned-new
5806 Warn about a new-expression of a type that requires greater alignment
5807 than the @code{alignof(std::max_align_t)} but uses an allocation
5808 function without an explicit alignment parameter. This option is
5809 enabled by @option{-Wall}.
5811 Normally this only warns about global allocation functions, but
5812 @option{-Waligned-new=all} also warns about class member allocation
5815 @item -Wplacement-new
5816 @itemx -Wplacement-new=@var{n}
5817 @opindex Wplacement-new
5818 @opindex Wno-placement-new
5819 Warn about placement new expressions with undefined behavior, such as
5820 constructing an object in a buffer that is smaller than the type of
5821 the object. For example, the placement new expression below is diagnosed
5822 because it attempts to construct an array of 64 integers in a buffer only
5828 This warning is enabled by default.
5831 @item -Wplacement-new=1
5832 This is the default warning level of @option{-Wplacement-new}. At this
5833 level the warning is not issued for some strictly undefined constructs that
5834 GCC allows as extensions for compatibility with legacy code. For example,
5835 the following @code{new} expression is not diagnosed at this level even
5836 though it has undefined behavior according to the C++ standard because
5837 it writes past the end of the one-element array.
5839 struct S @{ int n, a[1]; @};
5840 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5841 new (s->a)int [32]();
5844 @item -Wplacement-new=2
5845 At this level, in addition to diagnosing all the same constructs as at level
5846 1, a diagnostic is also issued for placement new expressions that construct
5847 an object in the last member of structure whose type is an array of a single
5848 element and whose size is less than the size of the object being constructed.
5849 While the previous example would be diagnosed, the following construct makes
5850 use of the flexible member array extension to avoid the warning at level 2.
5852 struct S @{ int n, a[]; @};
5853 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5854 new (s->a)int [32]();
5859 @item -Wpointer-arith
5860 @opindex Wpointer-arith
5861 @opindex Wno-pointer-arith
5862 Warn about anything that depends on the ``size of'' a function type or
5863 of @code{void}. GNU C assigns these types a size of 1, for
5864 convenience in calculations with @code{void *} pointers and pointers
5865 to functions. In C++, warn also when an arithmetic operation involves
5866 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5868 @item -Wpointer-compare
5869 @opindex Wpointer-compare
5870 @opindex Wno-pointer-compare
5871 Warn if a pointer is compared with a zero character constant. This usually
5872 means that the pointer was meant to be dereferenced. For example:
5875 const char *p = foo ();
5880 Note that the code above is invalid in C++11.
5882 This warning is enabled by default.
5885 @opindex Wtype-limits
5886 @opindex Wno-type-limits
5887 Warn if a comparison is always true or always false due to the limited
5888 range of the data type, but do not warn for constant expressions. For
5889 example, warn if an unsigned variable is compared against zero with
5890 @code{<} or @code{>=}. This warning is also enabled by
5893 @include cppwarnopts.texi
5895 @item -Wbad-function-cast @r{(C and Objective-C only)}
5896 @opindex Wbad-function-cast
5897 @opindex Wno-bad-function-cast
5898 Warn when a function call is cast to a non-matching type.
5899 For example, warn if a call to a function returning an integer type
5900 is cast to a pointer type.
5902 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5903 @opindex Wc90-c99-compat
5904 @opindex Wno-c90-c99-compat
5905 Warn about features not present in ISO C90, but present in ISO C99.
5906 For instance, warn about use of variable length arrays, @code{long long}
5907 type, @code{bool} type, compound literals, designated initializers, and so
5908 on. This option is independent of the standards mode. Warnings are disabled
5909 in the expression that follows @code{__extension__}.
5911 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5912 @opindex Wc99-c11-compat
5913 @opindex Wno-c99-c11-compat
5914 Warn about features not present in ISO C99, but present in ISO C11.
5915 For instance, warn about use of anonymous structures and unions,
5916 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5917 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5918 and so on. This option is independent of the standards mode. Warnings are
5919 disabled in the expression that follows @code{__extension__}.
5921 @item -Wc++-compat @r{(C and Objective-C only)}
5922 @opindex Wc++-compat
5923 Warn about ISO C constructs that are outside of the common subset of
5924 ISO C and ISO C++, e.g.@: request for implicit conversion from
5925 @code{void *} to a pointer to non-@code{void} type.
5927 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5928 @opindex Wc++11-compat
5929 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5930 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5931 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5932 enabled by @option{-Wall}.
5934 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5935 @opindex Wc++14-compat
5936 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5937 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5939 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
5940 @opindex Wc++17-compat
5941 Warn about C++ constructs whose meaning differs between ISO C++ 2014
5942 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
5946 @opindex Wno-cast-qual
5947 Warn whenever a pointer is cast so as to remove a type qualifier from
5948 the target type. For example, warn if a @code{const char *} is cast
5949 to an ordinary @code{char *}.
5951 Also warn when making a cast that introduces a type qualifier in an
5952 unsafe way. For example, casting @code{char **} to @code{const char **}
5953 is unsafe, as in this example:
5956 /* p is char ** value. */
5957 const char **q = (const char **) p;
5958 /* Assignment of readonly string to const char * is OK. */
5960 /* Now char** pointer points to read-only memory. */
5965 @opindex Wcast-align
5966 @opindex Wno-cast-align
5967 Warn whenever a pointer is cast such that the required alignment of the
5968 target is increased. For example, warn if a @code{char *} is cast to
5969 an @code{int *} on machines where integers can only be accessed at
5970 two- or four-byte boundaries.
5972 @item -Wcast-align=strict
5973 @opindex Wcast-align=strict
5974 Warn whenever a pointer is cast such that the required alignment of the
5975 target is increased. For example, warn if a @code{char *} is cast to
5976 an @code{int *} regardless of the target machine.
5978 @item -Wwrite-strings
5979 @opindex Wwrite-strings
5980 @opindex Wno-write-strings
5981 When compiling C, give string constants the type @code{const
5982 char[@var{length}]} so that copying the address of one into a
5983 non-@code{const} @code{char *} pointer produces a warning. These
5984 warnings help you find at compile time code that can try to write
5985 into a string constant, but only if you have been very careful about
5986 using @code{const} in declarations and prototypes. Otherwise, it is
5987 just a nuisance. This is why we did not make @option{-Wall} request
5990 When compiling C++, warn about the deprecated conversion from string
5991 literals to @code{char *}. This warning is enabled by default for C++
5995 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
5996 @opindex Wcatch-value
5997 @opindex Wno-catch-value
5998 Warn about catch handlers that do not catch via reference.
5999 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6000 warn about polymorphic class types that are caught by value.
6001 With @option{-Wcatch-value=2} warn about all class types that are caught
6002 by value. With @option{-Wcatch-value=3} warn about all types that are
6003 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6007 @opindex Wno-clobbered
6008 Warn for variables that might be changed by @code{longjmp} or
6009 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6011 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6012 @opindex Wconditionally-supported
6013 @opindex Wno-conditionally-supported
6014 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6017 @opindex Wconversion
6018 @opindex Wno-conversion
6019 Warn for implicit conversions that may alter a value. This includes
6020 conversions between real and integer, like @code{abs (x)} when
6021 @code{x} is @code{double}; conversions between signed and unsigned,
6022 like @code{unsigned ui = -1}; and conversions to smaller types, like
6023 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6024 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6025 changed by the conversion like in @code{abs (2.0)}. Warnings about
6026 conversions between signed and unsigned integers can be disabled by
6027 using @option{-Wno-sign-conversion}.
6029 For C++, also warn for confusing overload resolution for user-defined
6030 conversions; and conversions that never use a type conversion
6031 operator: conversions to @code{void}, the same type, a base class or a
6032 reference to them. Warnings about conversions between signed and
6033 unsigned integers are disabled by default in C++ unless
6034 @option{-Wsign-conversion} is explicitly enabled.
6036 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6037 @opindex Wconversion-null
6038 @opindex Wno-conversion-null
6039 Do not warn for conversions between @code{NULL} and non-pointer
6040 types. @option{-Wconversion-null} is enabled by default.
6042 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6043 @opindex Wzero-as-null-pointer-constant
6044 @opindex Wno-zero-as-null-pointer-constant
6045 Warn when a literal @samp{0} is used as null pointer constant. This can
6046 be useful to facilitate the conversion to @code{nullptr} in C++11.
6048 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6049 @opindex Wsubobject-linkage
6050 @opindex Wno-subobject-linkage
6051 Warn if a class type has a base or a field whose type uses the anonymous
6052 namespace or depends on a type with no linkage. If a type A depends on
6053 a type B with no or internal linkage, defining it in multiple
6054 translation units would be an ODR violation because the meaning of B
6055 is different in each translation unit. If A only appears in a single
6056 translation unit, the best way to silence the warning is to give it
6057 internal linkage by putting it in an anonymous namespace as well. The
6058 compiler doesn't give this warning for types defined in the main .C
6059 file, as those are unlikely to have multiple definitions.
6060 @option{-Wsubobject-linkage} is enabled by default.
6062 @item -Wdangling-else
6063 @opindex Wdangling-else
6064 @opindex Wno-dangling-else
6065 Warn about constructions where there may be confusion to which
6066 @code{if} statement an @code{else} branch belongs. Here is an example of
6081 In C/C++, every @code{else} branch belongs to the innermost possible
6082 @code{if} statement, which in this example is @code{if (b)}. This is
6083 often not what the programmer expected, as illustrated in the above
6084 example by indentation the programmer chose. When there is the
6085 potential for this confusion, GCC issues a warning when this flag
6086 is specified. To eliminate the warning, add explicit braces around
6087 the innermost @code{if} statement so there is no way the @code{else}
6088 can belong to the enclosing @code{if}. The resulting code
6105 This warning is enabled by @option{-Wparentheses}.
6109 @opindex Wno-date-time
6110 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6111 are encountered as they might prevent bit-wise-identical reproducible
6114 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6115 @opindex Wdelete-incomplete
6116 @opindex Wno-delete-incomplete
6117 Warn when deleting a pointer to incomplete type, which may cause
6118 undefined behavior at runtime. This warning is enabled by default.
6120 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6121 @opindex Wuseless-cast
6122 @opindex Wno-useless-cast
6123 Warn when an expression is casted to its own type.
6126 @opindex Wempty-body
6127 @opindex Wno-empty-body
6128 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6129 while} statement. This warning is also enabled by @option{-Wextra}.
6131 @item -Wenum-compare
6132 @opindex Wenum-compare
6133 @opindex Wno-enum-compare
6134 Warn about a comparison between values of different enumerated types.
6135 In C++ enumerated type mismatches in conditional expressions are also
6136 diagnosed and the warning is enabled by default. In C this warning is
6137 enabled by @option{-Wall}.
6139 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6140 @opindex Wextra-semi
6141 @opindex Wno-extra-semi
6142 Warn about redundant semicolon after in-class function definition.
6144 @item -Wjump-misses-init @r{(C, Objective-C only)}
6145 @opindex Wjump-misses-init
6146 @opindex Wno-jump-misses-init
6147 Warn if a @code{goto} statement or a @code{switch} statement jumps
6148 forward across the initialization of a variable, or jumps backward to a
6149 label after the variable has been initialized. This only warns about
6150 variables that are initialized when they are declared. This warning is
6151 only supported for C and Objective-C; in C++ this sort of branch is an
6154 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6155 can be disabled with the @option{-Wno-jump-misses-init} option.
6157 @item -Wsign-compare
6158 @opindex Wsign-compare
6159 @opindex Wno-sign-compare
6160 @cindex warning for comparison of signed and unsigned values
6161 @cindex comparison of signed and unsigned values, warning
6162 @cindex signed and unsigned values, comparison warning
6163 Warn when a comparison between signed and unsigned values could produce
6164 an incorrect result when the signed value is converted to unsigned.
6165 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6166 also enabled by @option{-Wextra}.
6168 @item -Wsign-conversion
6169 @opindex Wsign-conversion
6170 @opindex Wno-sign-conversion
6171 Warn for implicit conversions that may change the sign of an integer
6172 value, like assigning a signed integer expression to an unsigned
6173 integer variable. An explicit cast silences the warning. In C, this
6174 option is enabled also by @option{-Wconversion}.
6176 @item -Wfloat-conversion
6177 @opindex Wfloat-conversion
6178 @opindex Wno-float-conversion
6179 Warn for implicit conversions that reduce the precision of a real value.
6180 This includes conversions from real to integer, and from higher precision
6181 real to lower precision real values. This option is also enabled by
6182 @option{-Wconversion}.
6184 @item -Wno-scalar-storage-order
6185 @opindex -Wno-scalar-storage-order
6186 @opindex -Wscalar-storage-order
6187 Do not warn on suspicious constructs involving reverse scalar storage order.
6189 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6190 @opindex Wsized-deallocation
6191 @opindex Wno-sized-deallocation
6192 Warn about a definition of an unsized deallocation function
6194 void operator delete (void *) noexcept;
6195 void operator delete[] (void *) noexcept;
6197 without a definition of the corresponding sized deallocation function
6199 void operator delete (void *, std::size_t) noexcept;
6200 void operator delete[] (void *, std::size_t) noexcept;
6202 or vice versa. Enabled by @option{-Wextra} along with
6203 @option{-fsized-deallocation}.
6205 @item -Wsizeof-pointer-div
6206 @opindex Wsizeof-pointer-div
6207 @opindex Wno-sizeof-pointer-div
6208 Warn for suspicious divisions of two sizeof expressions that divide
6209 the pointer size by the element size, which is the usual way to compute
6210 the array size but won't work out correctly with pointers. This warning
6211 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
6212 not an array, but a pointer. This warning is enabled by @option{-Wall}.
6214 @item -Wsizeof-pointer-memaccess
6215 @opindex Wsizeof-pointer-memaccess
6216 @opindex Wno-sizeof-pointer-memaccess
6217 Warn for suspicious length parameters to certain string and memory built-in
6218 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
6219 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
6220 but a pointer, and suggests a possible fix, or about
6221 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
6224 @item -Wsizeof-array-argument
6225 @opindex Wsizeof-array-argument
6226 @opindex Wno-sizeof-array-argument
6227 Warn when the @code{sizeof} operator is applied to a parameter that is
6228 declared as an array in a function definition. This warning is enabled by
6229 default for C and C++ programs.
6231 @item -Wmemset-elt-size
6232 @opindex Wmemset-elt-size
6233 @opindex Wno-memset-elt-size
6234 Warn for suspicious calls to the @code{memset} built-in function, if the
6235 first argument references an array, and the third argument is a number
6236 equal to the number of elements, but not equal to the size of the array
6237 in memory. This indicates that the user has omitted a multiplication by
6238 the element size. This warning is enabled by @option{-Wall}.
6240 @item -Wmemset-transposed-args
6241 @opindex Wmemset-transposed-args
6242 @opindex Wno-memset-transposed-args
6243 Warn for suspicious calls to the @code{memset} built-in function, if the
6244 second argument is not zero and the third argument is zero. This warns e.g.@
6245 about @code{memset (buf, sizeof buf, 0)} where most probably
6246 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6247 is only emitted if the third argument is literal zero. If it is some
6248 expression that is folded to zero, a cast of zero to some type, etc.,
6249 it is far less likely that the user has mistakenly exchanged the arguments
6250 and no warning is emitted. This warning is enabled by @option{-Wall}.
6254 @opindex Wno-address
6255 Warn about suspicious uses of memory addresses. These include using
6256 the address of a function in a conditional expression, such as
6257 @code{void func(void); if (func)}, and comparisons against the memory
6258 address of a string literal, such as @code{if (x == "abc")}. Such
6259 uses typically indicate a programmer error: the address of a function
6260 always evaluates to true, so their use in a conditional usually
6261 indicate that the programmer forgot the parentheses in a function
6262 call; and comparisons against string literals result in unspecified
6263 behavior and are not portable in C, so they usually indicate that the
6264 programmer intended to use @code{strcmp}. This warning is enabled by
6268 @opindex Wlogical-op
6269 @opindex Wno-logical-op
6270 Warn about suspicious uses of logical operators in expressions.
6271 This includes using logical operators in contexts where a
6272 bit-wise operator is likely to be expected. Also warns when
6273 the operands of a logical operator are the same:
6276 if (a < 0 && a < 0) @{ @dots{} @}
6279 @item -Wlogical-not-parentheses
6280 @opindex Wlogical-not-parentheses
6281 @opindex Wno-logical-not-parentheses
6282 Warn about logical not used on the left hand side operand of a comparison.
6283 This option does not warn if the right operand is considered to be a boolean
6284 expression. Its purpose is to detect suspicious code like the following:
6288 if (!a > 1) @{ @dots{} @}
6291 It is possible to suppress the warning by wrapping the LHS into
6294 if ((!a) > 1) @{ @dots{} @}
6297 This warning is enabled by @option{-Wall}.
6299 @item -Waggregate-return
6300 @opindex Waggregate-return
6301 @opindex Wno-aggregate-return
6302 Warn if any functions that return structures or unions are defined or
6303 called. (In languages where you can return an array, this also elicits
6306 @item -Wno-aggressive-loop-optimizations
6307 @opindex Wno-aggressive-loop-optimizations
6308 @opindex Waggressive-loop-optimizations
6309 Warn if in a loop with constant number of iterations the compiler detects
6310 undefined behavior in some statement during one or more of the iterations.
6312 @item -Wno-attributes
6313 @opindex Wno-attributes
6314 @opindex Wattributes
6315 Do not warn if an unexpected @code{__attribute__} is used, such as
6316 unrecognized attributes, function attributes applied to variables,
6317 etc. This does not stop errors for incorrect use of supported
6320 @item -Wno-builtin-declaration-mismatch
6321 @opindex Wno-builtin-declaration-mismatch
6322 @opindex Wbuiltin-declaration-mismatch
6323 Warn if a built-in function is declared with the wrong signature or
6325 This warning is enabled by default.
6327 @item -Wno-builtin-macro-redefined
6328 @opindex Wno-builtin-macro-redefined
6329 @opindex Wbuiltin-macro-redefined
6330 Do not warn if certain built-in macros are redefined. This suppresses
6331 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6332 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6334 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6335 @opindex Wstrict-prototypes
6336 @opindex Wno-strict-prototypes
6337 Warn if a function is declared or defined without specifying the
6338 argument types. (An old-style function definition is permitted without
6339 a warning if preceded by a declaration that specifies the argument
6342 @item -Wold-style-declaration @r{(C and Objective-C only)}
6343 @opindex Wold-style-declaration
6344 @opindex Wno-old-style-declaration
6345 Warn for obsolescent usages, according to the C Standard, in a
6346 declaration. For example, warn if storage-class specifiers like
6347 @code{static} are not the first things in a declaration. This warning
6348 is also enabled by @option{-Wextra}.
6350 @item -Wold-style-definition @r{(C and Objective-C only)}
6351 @opindex Wold-style-definition
6352 @opindex Wno-old-style-definition
6353 Warn if an old-style function definition is used. A warning is given
6354 even if there is a previous prototype.
6356 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6357 @opindex Wmissing-parameter-type
6358 @opindex Wno-missing-parameter-type
6359 A function parameter is declared without a type specifier in K&R-style
6366 This warning is also enabled by @option{-Wextra}.
6368 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6369 @opindex Wmissing-prototypes
6370 @opindex Wno-missing-prototypes
6371 Warn if a global function is defined without a previous prototype
6372 declaration. This warning is issued even if the definition itself
6373 provides a prototype. Use this option to detect global functions
6374 that do not have a matching prototype declaration in a header file.
6375 This option is not valid for C++ because all function declarations
6376 provide prototypes and a non-matching declaration declares an
6377 overload rather than conflict with an earlier declaration.
6378 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6380 @item -Wmissing-declarations
6381 @opindex Wmissing-declarations
6382 @opindex Wno-missing-declarations
6383 Warn if a global function is defined without a previous declaration.
6384 Do so even if the definition itself provides a prototype.
6385 Use this option to detect global functions that are not declared in
6386 header files. In C, no warnings are issued for functions with previous
6387 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6388 missing prototypes. In C++, no warnings are issued for function templates,
6389 or for inline functions, or for functions in anonymous namespaces.
6391 @item -Wmissing-field-initializers
6392 @opindex Wmissing-field-initializers
6393 @opindex Wno-missing-field-initializers
6397 Warn if a structure's initializer has some fields missing. For
6398 example, the following code causes such a warning, because
6399 @code{x.h} is implicitly zero:
6402 struct s @{ int f, g, h; @};
6403 struct s x = @{ 3, 4 @};
6406 This option does not warn about designated initializers, so the following
6407 modification does not trigger a warning:
6410 struct s @{ int f, g, h; @};
6411 struct s x = @{ .f = 3, .g = 4 @};
6414 In C this option does not warn about the universal zero initializer
6418 struct s @{ int f, g, h; @};
6419 struct s x = @{ 0 @};
6422 Likewise, in C++ this option does not warn about the empty @{ @}
6423 initializer, for example:
6426 struct s @{ int f, g, h; @};
6430 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6431 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6433 @item -Wno-multichar
6434 @opindex Wno-multichar
6436 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6437 Usually they indicate a typo in the user's code, as they have
6438 implementation-defined values, and should not be used in portable code.
6440 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
6441 @opindex Wnormalized=
6442 @opindex Wnormalized
6443 @opindex Wno-normalized
6446 @cindex character set, input normalization
6447 In ISO C and ISO C++, two identifiers are different if they are
6448 different sequences of characters. However, sometimes when characters
6449 outside the basic ASCII character set are used, you can have two
6450 different character sequences that look the same. To avoid confusion,
6451 the ISO 10646 standard sets out some @dfn{normalization rules} which
6452 when applied ensure that two sequences that look the same are turned into
6453 the same sequence. GCC can warn you if you are using identifiers that
6454 have not been normalized; this option controls that warning.
6456 There are four levels of warning supported by GCC@. The default is
6457 @option{-Wnormalized=nfc}, which warns about any identifier that is
6458 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6459 recommended form for most uses. It is equivalent to
6460 @option{-Wnormalized}.
6462 Unfortunately, there are some characters allowed in identifiers by
6463 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6464 identifiers. That is, there's no way to use these symbols in portable
6465 ISO C or C++ and have all your identifiers in NFC@.
6466 @option{-Wnormalized=id} suppresses the warning for these characters.
6467 It is hoped that future versions of the standards involved will correct
6468 this, which is why this option is not the default.
6470 You can switch the warning off for all characters by writing
6471 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6472 only do this if you are using some other normalization scheme (like
6473 ``D''), because otherwise you can easily create bugs that are
6474 literally impossible to see.
6476 Some characters in ISO 10646 have distinct meanings but look identical
6477 in some fonts or display methodologies, especially once formatting has
6478 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6479 LETTER N'', displays just like a regular @code{n} that has been
6480 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6481 normalization scheme to convert all these into a standard form as
6482 well, and GCC warns if your code is not in NFKC if you use
6483 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6484 about every identifier that contains the letter O because it might be
6485 confused with the digit 0, and so is not the default, but may be
6486 useful as a local coding convention if the programming environment
6487 cannot be fixed to display these characters distinctly.
6489 @item -Wno-deprecated
6490 @opindex Wno-deprecated
6491 @opindex Wdeprecated
6492 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6494 @item -Wno-deprecated-declarations
6495 @opindex Wno-deprecated-declarations
6496 @opindex Wdeprecated-declarations
6497 Do not warn about uses of functions (@pxref{Function Attributes}),
6498 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6499 Attributes}) marked as deprecated by using the @code{deprecated}
6503 @opindex Wno-overflow
6505 Do not warn about compile-time overflow in constant expressions.
6510 Warn about One Definition Rule violations during link-time optimization.
6511 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6514 @opindex Wopenm-simd
6515 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6516 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6517 option can be used to relax the cost model.
6519 @item -Woverride-init @r{(C and Objective-C only)}
6520 @opindex Woverride-init
6521 @opindex Wno-override-init
6525 Warn if an initialized field without side effects is overridden when
6526 using designated initializers (@pxref{Designated Inits, , Designated
6529 This warning is included in @option{-Wextra}. To get other
6530 @option{-Wextra} warnings without this one, use @option{-Wextra
6531 -Wno-override-init}.
6533 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6534 @opindex Woverride-init-side-effects
6535 @opindex Wno-override-init-side-effects
6536 Warn if an initialized field with side effects is overridden when
6537 using designated initializers (@pxref{Designated Inits, , Designated
6538 Initializers}). This warning is enabled by default.
6543 Warn if a structure is given the packed attribute, but the packed
6544 attribute has no effect on the layout or size of the structure.
6545 Such structures may be mis-aligned for little benefit. For
6546 instance, in this code, the variable @code{f.x} in @code{struct bar}
6547 is misaligned even though @code{struct bar} does not itself
6548 have the packed attribute:
6555 @} __attribute__((packed));
6563 @item -Wpacked-bitfield-compat
6564 @opindex Wpacked-bitfield-compat
6565 @opindex Wno-packed-bitfield-compat
6566 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6567 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6568 the change can lead to differences in the structure layout. GCC
6569 informs you when the offset of such a field has changed in GCC 4.4.
6570 For example there is no longer a 4-bit padding between field @code{a}
6571 and @code{b} in this structure:
6578 @} __attribute__ ((packed));
6581 This warning is enabled by default. Use
6582 @option{-Wno-packed-bitfield-compat} to disable this warning.
6584 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
6585 @opindex Wpacked-not-aligned
6586 @opindex Wno-packed-not-aligned
6587 Warn if a structure field with explicitly specified alignment in a
6588 packed struct or union is misaligned. For example, a warning will
6589 be issued on @code{struct S}, like, @code{warning: alignment 1 of
6590 'struct S' is less than 8}, in this code:
6594 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
6595 struct __attribute__ ((packed)) S @{
6601 This warning is enabled by @option{-Wall}.
6606 Warn if padding is included in a structure, either to align an element
6607 of the structure or to align the whole structure. Sometimes when this
6608 happens it is possible to rearrange the fields of the structure to
6609 reduce the padding and so make the structure smaller.
6611 @item -Wredundant-decls
6612 @opindex Wredundant-decls
6613 @opindex Wno-redundant-decls
6614 Warn if anything is declared more than once in the same scope, even in
6615 cases where multiple declaration is valid and changes nothing.
6619 @opindex Wno-restrict
6620 Warn when an argument passed to a restrict-qualified parameter
6621 aliases with another argument.
6623 @item -Wnested-externs @r{(C and Objective-C only)}
6624 @opindex Wnested-externs
6625 @opindex Wno-nested-externs
6626 Warn if an @code{extern} declaration is encountered within a function.
6628 @item -Wno-inherited-variadic-ctor
6629 @opindex Winherited-variadic-ctor
6630 @opindex Wno-inherited-variadic-ctor
6631 Suppress warnings about use of C++11 inheriting constructors when the
6632 base class inherited from has a C variadic constructor; the warning is
6633 on by default because the ellipsis is not inherited.
6638 Warn if a function that is declared as inline cannot be inlined.
6639 Even with this option, the compiler does not warn about failures to
6640 inline functions declared in system headers.
6642 The compiler uses a variety of heuristics to determine whether or not
6643 to inline a function. For example, the compiler takes into account
6644 the size of the function being inlined and the amount of inlining
6645 that has already been done in the current function. Therefore,
6646 seemingly insignificant changes in the source program can cause the
6647 warnings produced by @option{-Winline} to appear or disappear.
6649 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6650 @opindex Wno-invalid-offsetof
6651 @opindex Winvalid-offsetof
6652 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6653 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6654 to a non-standard-layout type is undefined. In existing C++ implementations,
6655 however, @code{offsetof} typically gives meaningful results.
6656 This flag is for users who are aware that they are
6657 writing nonportable code and who have deliberately chosen to ignore the
6660 The restrictions on @code{offsetof} may be relaxed in a future version
6661 of the C++ standard.
6663 @item -Wint-in-bool-context
6664 @opindex Wint-in-bool-context
6665 @opindex Wno-int-in-bool-context
6666 Warn for suspicious use of integer values where boolean values are expected,
6667 such as conditional expressions (?:) using non-boolean integer constants in
6668 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6669 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6670 for all kinds of multiplications regardless of the data type.
6671 This warning is enabled by @option{-Wall}.
6673 @item -Wno-int-to-pointer-cast
6674 @opindex Wno-int-to-pointer-cast
6675 @opindex Wint-to-pointer-cast
6676 Suppress warnings from casts to pointer type of an integer of a
6677 different size. In C++, casting to a pointer type of smaller size is
6678 an error. @option{Wint-to-pointer-cast} is enabled by default.
6681 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6682 @opindex Wno-pointer-to-int-cast
6683 @opindex Wpointer-to-int-cast
6684 Suppress warnings from casts from a pointer to an integer type of a
6688 @opindex Winvalid-pch
6689 @opindex Wno-invalid-pch
6690 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6691 the search path but cannot be used.
6695 @opindex Wno-long-long
6696 Warn if @code{long long} type is used. This is enabled by either
6697 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6698 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6700 @item -Wvariadic-macros
6701 @opindex Wvariadic-macros
6702 @opindex Wno-variadic-macros
6703 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6704 alternate syntax is used in ISO C99 mode. This is enabled by either
6705 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6706 messages, use @option{-Wno-variadic-macros}.
6710 @opindex Wno-varargs
6711 Warn upon questionable usage of the macros used to handle variable
6712 arguments like @code{va_start}. This is default. To inhibit the
6713 warning messages, use @option{-Wno-varargs}.
6715 @item -Wvector-operation-performance
6716 @opindex Wvector-operation-performance
6717 @opindex Wno-vector-operation-performance
6718 Warn if vector operation is not implemented via SIMD capabilities of the
6719 architecture. Mainly useful for the performance tuning.
6720 Vector operation can be implemented @code{piecewise}, which means that the
6721 scalar operation is performed on every vector element;
6722 @code{in parallel}, which means that the vector operation is implemented
6723 using scalars of wider type, which normally is more performance efficient;
6724 and @code{as a single scalar}, which means that vector fits into a
6727 @item -Wno-virtual-move-assign
6728 @opindex Wvirtual-move-assign
6729 @opindex Wno-virtual-move-assign
6730 Suppress warnings about inheriting from a virtual base with a
6731 non-trivial C++11 move assignment operator. This is dangerous because
6732 if the virtual base is reachable along more than one path, it is
6733 moved multiple times, which can mean both objects end up in the
6734 moved-from state. If the move assignment operator is written to avoid
6735 moving from a moved-from object, this warning can be disabled.
6740 Warn if a variable-length array is used in the code.
6741 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6742 the variable-length array.
6744 @item -Wvla-larger-than=@var{n}
6745 If this option is used, the compiler will warn on uses of
6746 variable-length arrays where the size is either unbounded, or bounded
6747 by an argument that can be larger than @var{n} bytes. This is similar
6748 to how @option{-Walloca-larger-than=@var{n}} works, but with
6749 variable-length arrays.
6751 Note that GCC may optimize small variable-length arrays of a known
6752 value into plain arrays, so this warning may not get triggered for
6755 This warning is not enabled by @option{-Wall}, and is only active when
6756 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6758 See also @option{-Walloca-larger-than=@var{n}}.
6760 @item -Wvolatile-register-var
6761 @opindex Wvolatile-register-var
6762 @opindex Wno-volatile-register-var
6763 Warn if a register variable is declared volatile. The volatile
6764 modifier does not inhibit all optimizations that may eliminate reads
6765 and/or writes to register variables. This warning is enabled by
6768 @item -Wdisabled-optimization
6769 @opindex Wdisabled-optimization
6770 @opindex Wno-disabled-optimization
6771 Warn if a requested optimization pass is disabled. This warning does
6772 not generally indicate that there is anything wrong with your code; it
6773 merely indicates that GCC's optimizers are unable to handle the code
6774 effectively. Often, the problem is that your code is too big or too
6775 complex; GCC refuses to optimize programs when the optimization
6776 itself is likely to take inordinate amounts of time.
6778 @item -Wpointer-sign @r{(C and Objective-C only)}
6779 @opindex Wpointer-sign
6780 @opindex Wno-pointer-sign
6781 Warn for pointer argument passing or assignment with different signedness.
6782 This option is only supported for C and Objective-C@. It is implied by
6783 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6784 @option{-Wno-pointer-sign}.
6786 @item -Wstack-protector
6787 @opindex Wstack-protector
6788 @opindex Wno-stack-protector
6789 This option is only active when @option{-fstack-protector} is active. It
6790 warns about functions that are not protected against stack smashing.
6792 @item -Woverlength-strings
6793 @opindex Woverlength-strings
6794 @opindex Wno-overlength-strings
6795 Warn about string constants that are longer than the ``minimum
6796 maximum'' length specified in the C standard. Modern compilers
6797 generally allow string constants that are much longer than the
6798 standard's minimum limit, but very portable programs should avoid
6799 using longer strings.
6801 The limit applies @emph{after} string constant concatenation, and does
6802 not count the trailing NUL@. In C90, the limit was 509 characters; in
6803 C99, it was raised to 4095. C++98 does not specify a normative
6804 minimum maximum, so we do not diagnose overlength strings in C++@.
6806 This option is implied by @option{-Wpedantic}, and can be disabled with
6807 @option{-Wno-overlength-strings}.
6809 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6810 @opindex Wunsuffixed-float-constants
6812 Issue a warning for any floating constant that does not have
6813 a suffix. When used together with @option{-Wsystem-headers} it
6814 warns about such constants in system header files. This can be useful
6815 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6816 from the decimal floating-point extension to C99.
6818 @item -Wno-designated-init @r{(C and Objective-C only)}
6819 Suppress warnings when a positional initializer is used to initialize
6820 a structure that has been marked with the @code{designated_init}
6824 Issue a warning when HSAIL cannot be emitted for the compiled function or
6829 @node Debugging Options
6830 @section Options for Debugging Your Program
6831 @cindex options, debugging
6832 @cindex debugging information options
6834 To tell GCC to emit extra information for use by a debugger, in almost
6835 all cases you need only to add @option{-g} to your other options.
6837 GCC allows you to use @option{-g} with
6838 @option{-O}. The shortcuts taken by optimized code may occasionally
6839 be surprising: some variables you declared may not exist
6840 at all; flow of control may briefly move where you did not expect it;
6841 some statements may not be executed because they compute constant
6842 results or their values are already at hand; some statements may
6843 execute in different places because they have been moved out of loops.
6844 Nevertheless it is possible to debug optimized output. This makes
6845 it reasonable to use the optimizer for programs that might have bugs.
6847 If you are not using some other optimization option, consider
6848 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6849 With no @option{-O} option at all, some compiler passes that collect
6850 information useful for debugging do not run at all, so that
6851 @option{-Og} may result in a better debugging experience.
6856 Produce debugging information in the operating system's native format
6857 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6860 On most systems that use stabs format, @option{-g} enables use of extra
6861 debugging information that only GDB can use; this extra information
6862 makes debugging work better in GDB but probably makes other debuggers
6864 refuse to read the program. If you want to control for certain whether
6865 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6866 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6870 Produce debugging information for use by GDB@. This means to use the
6871 most expressive format available (DWARF, stabs, or the native format
6872 if neither of those are supported), including GDB extensions if at all
6876 @itemx -gdwarf-@var{version}
6878 Produce debugging information in DWARF format (if that is supported).
6879 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6880 for most targets is 4. DWARF Version 5 is only experimental.
6882 Note that with DWARF Version 2, some ports require and always
6883 use some non-conflicting DWARF 3 extensions in the unwind tables.
6885 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6886 for maximum benefit.
6888 GCC no longer supports DWARF Version 1, which is substantially
6889 different than Version 2 and later. For historical reasons, some
6890 other DWARF-related options such as
6891 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6892 in their names, but apply to all currently-supported versions of DWARF.
6896 Produce debugging information in stabs format (if that is supported),
6897 without GDB extensions. This is the format used by DBX on most BSD
6898 systems. On MIPS, Alpha and System V Release 4 systems this option
6899 produces stabs debugging output that is not understood by DBX or SDB@.
6900 On System V Release 4 systems this option requires the GNU assembler.
6904 Produce debugging information in stabs format (if that is supported),
6905 using GNU extensions understood only by the GNU debugger (GDB)@. The
6906 use of these extensions is likely to make other debuggers crash or
6907 refuse to read the program.
6911 Produce debugging information in COFF format (if that is supported).
6912 This is the format used by SDB on most System V systems prior to
6917 Produce debugging information in XCOFF format (if that is supported).
6918 This is the format used by the DBX debugger on IBM RS/6000 systems.
6922 Produce debugging information in XCOFF format (if that is supported),
6923 using GNU extensions understood only by the GNU debugger (GDB)@. The
6924 use of these extensions is likely to make other debuggers crash or
6925 refuse to read the program, and may cause assemblers other than the GNU
6926 assembler (GAS) to fail with an error.
6930 Produce debugging information in Alpha/VMS debug format (if that is
6931 supported). This is the format used by DEBUG on Alpha/VMS systems.
6934 @itemx -ggdb@var{level}
6935 @itemx -gstabs@var{level}
6936 @itemx -gcoff@var{level}
6937 @itemx -gxcoff@var{level}
6938 @itemx -gvms@var{level}
6939 Request debugging information and also use @var{level} to specify how
6940 much information. The default level is 2.
6942 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6945 Level 1 produces minimal information, enough for making backtraces in
6946 parts of the program that you don't plan to debug. This includes
6947 descriptions of functions and external variables, and line number
6948 tables, but no information about local variables.
6950 Level 3 includes extra information, such as all the macro definitions
6951 present in the program. Some debuggers support macro expansion when
6952 you use @option{-g3}.
6954 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6955 confusion with @option{-gdwarf-@var{level}}.
6956 Instead use an additional @option{-g@var{level}} option to change the
6957 debug level for DWARF.
6959 @item -feliminate-unused-debug-symbols
6960 @opindex feliminate-unused-debug-symbols
6961 Produce debugging information in stabs format (if that is supported),
6962 for only symbols that are actually used.
6964 @item -femit-class-debug-always
6965 @opindex femit-class-debug-always
6966 Instead of emitting debugging information for a C++ class in only one
6967 object file, emit it in all object files using the class. This option
6968 should be used only with debuggers that are unable to handle the way GCC
6969 normally emits debugging information for classes because using this
6970 option increases the size of debugging information by as much as a
6973 @item -fno-merge-debug-strings
6974 @opindex fmerge-debug-strings
6975 @opindex fno-merge-debug-strings
6976 Direct the linker to not merge together strings in the debugging
6977 information that are identical in different object files. Merging is
6978 not supported by all assemblers or linkers. Merging decreases the size
6979 of the debug information in the output file at the cost of increasing
6980 link processing time. Merging is enabled by default.
6982 @item -fdebug-prefix-map=@var{old}=@var{new}
6983 @opindex fdebug-prefix-map
6984 When compiling files in directory @file{@var{old}}, record debugging
6985 information describing them as in @file{@var{new}} instead. This can be
6986 used to replace a build time path with an install time path in the debug info.
6987 It can also be used to change an absolute path to a relative path by using
6988 @file{.} for @var{new}. This can give more reproducible builds, which are
6989 location independent, but may require an extra command to tell GDB where to
6990 find the source files.
6992 @item -fvar-tracking
6993 @opindex fvar-tracking
6994 Run variable tracking pass. It computes where variables are stored at each
6995 position in code. Better debugging information is then generated
6996 (if the debugging information format supports this information).
6998 It is enabled by default when compiling with optimization (@option{-Os},
6999 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7000 the debug info format supports it.
7002 @item -fvar-tracking-assignments
7003 @opindex fvar-tracking-assignments
7004 @opindex fno-var-tracking-assignments
7005 Annotate assignments to user variables early in the compilation and
7006 attempt to carry the annotations over throughout the compilation all the
7007 way to the end, in an attempt to improve debug information while
7008 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7010 It can be enabled even if var-tracking is disabled, in which case
7011 annotations are created and maintained, but discarded at the end.
7012 By default, this flag is enabled together with @option{-fvar-tracking},
7013 except when selective scheduling is enabled.
7016 @opindex gsplit-dwarf
7017 Separate as much DWARF debugging information as possible into a
7018 separate output file with the extension @file{.dwo}. This option allows
7019 the build system to avoid linking files with debug information. To
7020 be useful, this option requires a debugger capable of reading @file{.dwo}
7025 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7027 @item -ggnu-pubnames
7028 @opindex ggnu-pubnames
7029 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7030 suitable for conversion into a GDB@ index. This option is only useful
7031 with a linker that can produce GDB@ index version 7.
7033 @item -fdebug-types-section
7034 @opindex fdebug-types-section
7035 @opindex fno-debug-types-section
7036 When using DWARF Version 4 or higher, type DIEs can be put into
7037 their own @code{.debug_types} section instead of making them part of the
7038 @code{.debug_info} section. It is more efficient to put them in a separate
7039 comdat sections since the linker can then remove duplicates.
7040 But not all DWARF consumers support @code{.debug_types} sections yet
7041 and on some objects @code{.debug_types} produces larger instead of smaller
7042 debugging information.
7044 @item -grecord-gcc-switches
7045 @item -gno-record-gcc-switches
7046 @opindex grecord-gcc-switches
7047 @opindex gno-record-gcc-switches
7048 This switch causes the command-line options used to invoke the
7049 compiler that may affect code generation to be appended to the
7050 DW_AT_producer attribute in DWARF debugging information. The options
7051 are concatenated with spaces separating them from each other and from
7052 the compiler version.
7053 It is enabled by default.
7054 See also @option{-frecord-gcc-switches} for another
7055 way of storing compiler options into the object file.
7057 @item -gstrict-dwarf
7058 @opindex gstrict-dwarf
7059 Disallow using extensions of later DWARF standard version than selected
7060 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7061 DWARF extensions from later standard versions is allowed.
7063 @item -gno-strict-dwarf
7064 @opindex gno-strict-dwarf
7065 Allow using extensions of later DWARF standard version than selected with
7066 @option{-gdwarf-@var{version}}.
7069 @item -gno-column-info
7070 @opindex gcolumn-info
7071 @opindex gno-column-info
7072 Emit location column information into DWARF debugging information, rather
7073 than just file and line.
7074 This option is enabled by default.
7076 @item -gz@r{[}=@var{type}@r{]}
7078 Produce compressed debug sections in DWARF format, if that is supported.
7079 If @var{type} is not given, the default type depends on the capabilities
7080 of the assembler and linker used. @var{type} may be one of
7081 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
7082 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
7083 compression in traditional GNU format). If the linker doesn't support
7084 writing compressed debug sections, the option is rejected. Otherwise,
7085 if the assembler does not support them, @option{-gz} is silently ignored
7086 when producing object files.
7088 @item -femit-struct-debug-baseonly
7089 @opindex femit-struct-debug-baseonly
7090 Emit debug information for struct-like types
7091 only when the base name of the compilation source file
7092 matches the base name of file in which the struct is defined.
7094 This option substantially reduces the size of debugging information,
7095 but at significant potential loss in type information to the debugger.
7096 See @option{-femit-struct-debug-reduced} for a less aggressive option.
7097 See @option{-femit-struct-debug-detailed} for more detailed control.
7099 This option works only with DWARF debug output.
7101 @item -femit-struct-debug-reduced
7102 @opindex femit-struct-debug-reduced
7103 Emit debug information for struct-like types
7104 only when the base name of the compilation source file
7105 matches the base name of file in which the type is defined,
7106 unless the struct is a template or defined in a system header.
7108 This option significantly reduces the size of debugging information,
7109 with some potential loss in type information to the debugger.
7110 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
7111 See @option{-femit-struct-debug-detailed} for more detailed control.
7113 This option works only with DWARF debug output.
7115 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
7116 @opindex femit-struct-debug-detailed
7117 Specify the struct-like types
7118 for which the compiler generates debug information.
7119 The intent is to reduce duplicate struct debug information
7120 between different object files within the same program.
7122 This option is a detailed version of
7123 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
7124 which serves for most needs.
7126 A specification has the syntax@*
7127 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
7129 The optional first word limits the specification to
7130 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
7131 A struct type is used directly when it is the type of a variable, member.
7132 Indirect uses arise through pointers to structs.
7133 That is, when use of an incomplete struct is valid, the use is indirect.
7135 @samp{struct one direct; struct two * indirect;}.
7137 The optional second word limits the specification to
7138 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
7139 Generic structs are a bit complicated to explain.
7140 For C++, these are non-explicit specializations of template classes,
7141 or non-template classes within the above.
7142 Other programming languages have generics,
7143 but @option{-femit-struct-debug-detailed} does not yet implement them.
7145 The third word specifies the source files for those
7146 structs for which the compiler should emit debug information.
7147 The values @samp{none} and @samp{any} have the normal meaning.
7148 The value @samp{base} means that
7149 the base of name of the file in which the type declaration appears
7150 must match the base of the name of the main compilation file.
7151 In practice, this means that when compiling @file{foo.c}, debug information
7152 is generated for types declared in that file and @file{foo.h},
7153 but not other header files.
7154 The value @samp{sys} means those types satisfying @samp{base}
7155 or declared in system or compiler headers.
7157 You may need to experiment to determine the best settings for your application.
7159 The default is @option{-femit-struct-debug-detailed=all}.
7161 This option works only with DWARF debug output.
7163 @item -fno-dwarf2-cfi-asm
7164 @opindex fdwarf2-cfi-asm
7165 @opindex fno-dwarf2-cfi-asm
7166 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
7167 instead of using GAS @code{.cfi_*} directives.
7169 @item -fno-eliminate-unused-debug-types
7170 @opindex feliminate-unused-debug-types
7171 @opindex fno-eliminate-unused-debug-types
7172 Normally, when producing DWARF output, GCC avoids producing debug symbol
7173 output for types that are nowhere used in the source file being compiled.
7174 Sometimes it is useful to have GCC emit debugging
7175 information for all types declared in a compilation
7176 unit, regardless of whether or not they are actually used
7177 in that compilation unit, for example
7178 if, in the debugger, you want to cast a value to a type that is
7179 not actually used in your program (but is declared). More often,
7180 however, this results in a significant amount of wasted space.
7183 @node Optimize Options
7184 @section Options That Control Optimization
7185 @cindex optimize options
7186 @cindex options, optimization
7188 These options control various sorts of optimizations.
7190 Without any optimization option, the compiler's goal is to reduce the
7191 cost of compilation and to make debugging produce the expected
7192 results. Statements are independent: if you stop the program with a
7193 breakpoint between statements, you can then assign a new value to any
7194 variable or change the program counter to any other statement in the
7195 function and get exactly the results you expect from the source
7198 Turning on optimization flags makes the compiler attempt to improve
7199 the performance and/or code size at the expense of compilation time
7200 and possibly the ability to debug the program.
7202 The compiler performs optimization based on the knowledge it has of the
7203 program. Compiling multiple files at once to a single output file mode allows
7204 the compiler to use information gained from all of the files when compiling
7207 Not all optimizations are controlled directly by a flag. Only
7208 optimizations that have a flag are listed in this section.
7210 Most optimizations are only enabled if an @option{-O} level is set on
7211 the command line. Otherwise they are disabled, even if individual
7212 optimization flags are specified.
7214 Depending on the target and how GCC was configured, a slightly different
7215 set of optimizations may be enabled at each @option{-O} level than
7216 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7217 to find out the exact set of optimizations that are enabled at each level.
7218 @xref{Overall Options}, for examples.
7225 Optimize. Optimizing compilation takes somewhat more time, and a lot
7226 more memory for a large function.
7228 With @option{-O}, the compiler tries to reduce code size and execution
7229 time, without performing any optimizations that take a great deal of
7232 @option{-O} turns on the following optimization flags:
7235 -fbranch-count-reg @gol
7236 -fcombine-stack-adjustments @gol
7238 -fcprop-registers @gol
7241 -fdelayed-branch @gol
7243 -fforward-propagate @gol
7244 -fguess-branch-probability @gol
7245 -fif-conversion2 @gol
7246 -fif-conversion @gol
7247 -finline-functions-called-once @gol
7248 -fipa-pure-const @gol
7250 -fipa-reference @gol
7251 -fmerge-constants @gol
7252 -fmove-loop-invariants @gol
7253 -freorder-blocks @gol
7255 -fshrink-wrap-separate @gol
7256 -fsplit-wide-types @gol
7262 -ftree-coalesce-vars @gol
7263 -ftree-copy-prop @gol
7265 -ftree-dominator-opts @gol
7267 -ftree-forwprop @gol
7277 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7278 where doing so does not interfere with debugging.
7282 Optimize even more. GCC performs nearly all supported optimizations
7283 that do not involve a space-speed tradeoff.
7284 As compared to @option{-O}, this option increases both compilation time
7285 and the performance of the generated code.
7287 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7288 also turns on the following optimization flags:
7289 @gccoptlist{-fthread-jumps @gol
7290 -falign-functions -falign-jumps @gol
7291 -falign-loops -falign-labels @gol
7294 -fcse-follow-jumps -fcse-skip-blocks @gol
7295 -fdelete-null-pointer-checks @gol
7296 -fdevirtualize -fdevirtualize-speculatively @gol
7297 -fexpensive-optimizations @gol
7298 -fgcse -fgcse-lm @gol
7299 -fhoist-adjacent-loads @gol
7300 -finline-small-functions @gol
7301 -findirect-inlining @gol
7307 -fisolate-erroneous-paths-dereference @gol
7309 -foptimize-sibling-calls @gol
7310 -foptimize-strlen @gol
7311 -fpartial-inlining @gol
7313 -freorder-blocks-algorithm=stc @gol
7314 -freorder-blocks-and-partition -freorder-functions @gol
7315 -frerun-cse-after-loop @gol
7316 -fsched-interblock -fsched-spec @gol
7317 -fschedule-insns -fschedule-insns2 @gol
7318 -fstore-merging @gol
7319 -fstrict-aliasing @gol
7320 -ftree-builtin-call-dce @gol
7321 -ftree-switch-conversion -ftree-tail-merge @gol
7322 -fcode-hoisting @gol
7327 Please note the warning under @option{-fgcse} about
7328 invoking @option{-O2} on programs that use computed gotos.
7332 Optimize yet more. @option{-O3} turns on all optimizations specified
7333 by @option{-O2} and also turns on the following optimization flags:
7334 @gccoptlist{-finline-functions @gol
7335 -funswitch-loops @gol
7336 -fpredictive-commoning @gol
7337 -fgcse-after-reload @gol
7338 -ftree-loop-vectorize @gol
7339 -ftree-loop-distribution @gol
7340 -ftree-loop-distribute-patterns @gol
7342 -ftree-slp-vectorize @gol
7343 -fvect-cost-model @gol
7344 -ftree-partial-pre @gol
7350 Reduce compilation time and make debugging produce the expected
7351 results. This is the default.
7355 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7356 do not typically increase code size. It also performs further
7357 optimizations designed to reduce code size.
7359 @option{-Os} disables the following optimization flags:
7360 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7361 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7362 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7366 Disregard strict standards compliance. @option{-Ofast} enables all
7367 @option{-O3} optimizations. It also enables optimizations that are not
7368 valid for all standard-compliant programs.
7369 It turns on @option{-ffast-math} and the Fortran-specific
7370 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
7371 specified, and @option{-fno-protect-parens}.
7375 Optimize debugging experience. @option{-Og} enables optimizations
7376 that do not interfere with debugging. It should be the optimization
7377 level of choice for the standard edit-compile-debug cycle, offering
7378 a reasonable level of optimization while maintaining fast compilation
7379 and a good debugging experience.
7382 If you use multiple @option{-O} options, with or without level numbers,
7383 the last such option is the one that is effective.
7385 Options of the form @option{-f@var{flag}} specify machine-independent
7386 flags. Most flags have both positive and negative forms; the negative
7387 form of @option{-ffoo} is @option{-fno-foo}. In the table
7388 below, only one of the forms is listed---the one you typically
7389 use. You can figure out the other form by either removing @samp{no-}
7392 The following options control specific optimizations. They are either
7393 activated by @option{-O} options or are related to ones that are. You
7394 can use the following flags in the rare cases when ``fine-tuning'' of
7395 optimizations to be performed is desired.
7398 @item -fno-defer-pop
7399 @opindex fno-defer-pop
7400 Always pop the arguments to each function call as soon as that function
7401 returns. For machines that must pop arguments after a function call,
7402 the compiler normally lets arguments accumulate on the stack for several
7403 function calls and pops them all at once.
7405 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7407 @item -fforward-propagate
7408 @opindex fforward-propagate
7409 Perform a forward propagation pass on RTL@. The pass tries to combine two
7410 instructions and checks if the result can be simplified. If loop unrolling
7411 is active, two passes are performed and the second is scheduled after
7414 This option is enabled by default at optimization levels @option{-O},
7415 @option{-O2}, @option{-O3}, @option{-Os}.
7417 @item -ffp-contract=@var{style}
7418 @opindex ffp-contract
7419 @option{-ffp-contract=off} disables floating-point expression contraction.
7420 @option{-ffp-contract=fast} enables floating-point expression contraction
7421 such as forming of fused multiply-add operations if the target has
7422 native support for them.
7423 @option{-ffp-contract=on} enables floating-point expression contraction
7424 if allowed by the language standard. This is currently not implemented
7425 and treated equal to @option{-ffp-contract=off}.
7427 The default is @option{-ffp-contract=fast}.
7429 @item -fomit-frame-pointer
7430 @opindex fomit-frame-pointer
7431 Don't keep the frame pointer in a register for functions that
7432 don't need one. This avoids the instructions to save, set up and
7433 restore frame pointers; it also makes an extra register available
7434 in many functions. @strong{It also makes debugging impossible on
7437 On some machines, such as the VAX, this flag has no effect, because
7438 the standard calling sequence automatically handles the frame pointer
7439 and nothing is saved by pretending it doesn't exist. The
7440 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7441 whether a target machine supports this flag. @xref{Registers,,Register
7442 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7444 The default setting (when not optimizing for
7445 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7446 @option{-fomit-frame-pointer}. You can configure GCC with the
7447 @option{--enable-frame-pointer} configure option to change the default.
7449 Note that @option{-fno-omit-frame-pointer} doesn't force a new stack
7450 frame for all functions if it isn't otherwise needed, and hence doesn't
7451 guarantee a new frame pointer for all functions.
7453 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7455 @item -foptimize-sibling-calls
7456 @opindex foptimize-sibling-calls
7457 Optimize sibling and tail recursive calls.
7459 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7461 @item -foptimize-strlen
7462 @opindex foptimize-strlen
7463 Optimize various standard C string functions (e.g. @code{strlen},
7464 @code{strchr} or @code{strcpy}) and
7465 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7467 Enabled at levels @option{-O2}, @option{-O3}.
7471 Do not expand any functions inline apart from those marked with
7472 the @code{always_inline} attribute. This is the default when not
7475 Single functions can be exempted from inlining by marking them
7476 with the @code{noinline} attribute.
7478 @item -finline-small-functions
7479 @opindex finline-small-functions
7480 Integrate functions into their callers when their body is smaller than expected
7481 function call code (so overall size of program gets smaller). The compiler
7482 heuristically decides which functions are simple enough to be worth integrating
7483 in this way. This inlining applies to all functions, even those not declared
7486 Enabled at level @option{-O2}.
7488 @item -findirect-inlining
7489 @opindex findirect-inlining
7490 Inline also indirect calls that are discovered to be known at compile
7491 time thanks to previous inlining. This option has any effect only
7492 when inlining itself is turned on by the @option{-finline-functions}
7493 or @option{-finline-small-functions} options.
7495 Enabled at level @option{-O2}.
7497 @item -finline-functions
7498 @opindex finline-functions
7499 Consider all functions for inlining, even if they are not declared inline.
7500 The compiler heuristically decides which functions are worth integrating
7503 If all calls to a given function are integrated, and the function is
7504 declared @code{static}, then the function is normally not output as
7505 assembler code in its own right.
7507 Enabled at level @option{-O3}.
7509 @item -finline-functions-called-once
7510 @opindex finline-functions-called-once
7511 Consider all @code{static} functions called once for inlining into their
7512 caller even if they are not marked @code{inline}. If a call to a given
7513 function is integrated, then the function is not output as assembler code
7516 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7518 @item -fearly-inlining
7519 @opindex fearly-inlining
7520 Inline functions marked by @code{always_inline} and functions whose body seems
7521 smaller than the function call overhead early before doing
7522 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7523 makes profiling significantly cheaper and usually inlining faster on programs
7524 having large chains of nested wrapper functions.
7530 Perform interprocedural scalar replacement of aggregates, removal of
7531 unused parameters and replacement of parameters passed by reference
7532 by parameters passed by value.
7534 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7536 @item -finline-limit=@var{n}
7537 @opindex finline-limit
7538 By default, GCC limits the size of functions that can be inlined. This flag
7539 allows coarse control of this limit. @var{n} is the size of functions that
7540 can be inlined in number of pseudo instructions.
7542 Inlining is actually controlled by a number of parameters, which may be
7543 specified individually by using @option{--param @var{name}=@var{value}}.
7544 The @option{-finline-limit=@var{n}} option sets some of these parameters
7548 @item max-inline-insns-single
7549 is set to @var{n}/2.
7550 @item max-inline-insns-auto
7551 is set to @var{n}/2.
7554 See below for a documentation of the individual
7555 parameters controlling inlining and for the defaults of these parameters.
7557 @emph{Note:} there may be no value to @option{-finline-limit} that results
7558 in default behavior.
7560 @emph{Note:} pseudo instruction represents, in this particular context, an
7561 abstract measurement of function's size. In no way does it represent a count
7562 of assembly instructions and as such its exact meaning might change from one
7563 release to an another.
7565 @item -fno-keep-inline-dllexport
7566 @opindex fno-keep-inline-dllexport
7567 This is a more fine-grained version of @option{-fkeep-inline-functions},
7568 which applies only to functions that are declared using the @code{dllexport}
7569 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
7572 @item -fkeep-inline-functions
7573 @opindex fkeep-inline-functions
7574 In C, emit @code{static} functions that are declared @code{inline}
7575 into the object file, even if the function has been inlined into all
7576 of its callers. This switch does not affect functions using the
7577 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7578 inline functions into the object file.
7580 @item -fkeep-static-functions
7581 @opindex fkeep-static-functions
7582 Emit @code{static} functions into the object file, even if the function
7585 @item -fkeep-static-consts
7586 @opindex fkeep-static-consts
7587 Emit variables declared @code{static const} when optimization isn't turned
7588 on, even if the variables aren't referenced.
7590 GCC enables this option by default. If you want to force the compiler to
7591 check if a variable is referenced, regardless of whether or not
7592 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7594 @item -fmerge-constants
7595 @opindex fmerge-constants
7596 Attempt to merge identical constants (string constants and floating-point
7597 constants) across compilation units.
7599 This option is the default for optimized compilation if the assembler and
7600 linker support it. Use @option{-fno-merge-constants} to inhibit this
7603 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7605 @item -fmerge-all-constants
7606 @opindex fmerge-all-constants
7607 Attempt to merge identical constants and identical variables.
7609 This option implies @option{-fmerge-constants}. In addition to
7610 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7611 arrays or initialized constant variables with integral or floating-point
7612 types. Languages like C or C++ require each variable, including multiple
7613 instances of the same variable in recursive calls, to have distinct locations,
7614 so using this option results in non-conforming
7617 @item -fmodulo-sched
7618 @opindex fmodulo-sched
7619 Perform swing modulo scheduling immediately before the first scheduling
7620 pass. This pass looks at innermost loops and reorders their
7621 instructions by overlapping different iterations.
7623 @item -fmodulo-sched-allow-regmoves
7624 @opindex fmodulo-sched-allow-regmoves
7625 Perform more aggressive SMS-based modulo scheduling with register moves
7626 allowed. By setting this flag certain anti-dependences edges are
7627 deleted, which triggers the generation of reg-moves based on the
7628 life-range analysis. This option is effective only with
7629 @option{-fmodulo-sched} enabled.
7631 @item -fno-branch-count-reg
7632 @opindex fno-branch-count-reg
7633 Avoid running a pass scanning for opportunities to use ``decrement and
7634 branch'' instructions on a count register instead of generating sequences
7635 of instructions that decrement a register, compare it against zero, and
7636 then branch based upon the result. This option is only meaningful on
7637 architectures that support such instructions, which include x86, PowerPC,
7638 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7639 doesn't remove the decrement and branch instructions from the generated
7640 instruction stream introduced by other optimization passes.
7642 Enabled by default at @option{-O1} and higher.
7644 The default is @option{-fbranch-count-reg}.
7646 @item -fno-function-cse
7647 @opindex fno-function-cse
7648 Do not put function addresses in registers; make each instruction that
7649 calls a constant function contain the function's address explicitly.
7651 This option results in less efficient code, but some strange hacks
7652 that alter the assembler output may be confused by the optimizations
7653 performed when this option is not used.
7655 The default is @option{-ffunction-cse}
7657 @item -fno-zero-initialized-in-bss
7658 @opindex fno-zero-initialized-in-bss
7659 If the target supports a BSS section, GCC by default puts variables that
7660 are initialized to zero into BSS@. This can save space in the resulting
7663 This option turns off this behavior because some programs explicitly
7664 rely on variables going to the data section---e.g., so that the
7665 resulting executable can find the beginning of that section and/or make
7666 assumptions based on that.
7668 The default is @option{-fzero-initialized-in-bss}.
7670 @item -fthread-jumps
7671 @opindex fthread-jumps
7672 Perform optimizations that check to see if a jump branches to a
7673 location where another comparison subsumed by the first is found. If
7674 so, the first branch is redirected to either the destination of the
7675 second branch or a point immediately following it, depending on whether
7676 the condition is known to be true or false.
7678 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7680 @item -fsplit-wide-types
7681 @opindex fsplit-wide-types
7682 When using a type that occupies multiple registers, such as @code{long
7683 long} on a 32-bit system, split the registers apart and allocate them
7684 independently. This normally generates better code for those types,
7685 but may make debugging more difficult.
7687 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7690 @item -fcse-follow-jumps
7691 @opindex fcse-follow-jumps
7692 In common subexpression elimination (CSE), scan through jump instructions
7693 when the target of the jump is not reached by any other path. For
7694 example, when CSE encounters an @code{if} statement with an
7695 @code{else} clause, CSE follows the jump when the condition
7698 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7700 @item -fcse-skip-blocks
7701 @opindex fcse-skip-blocks
7702 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7703 follow jumps that conditionally skip over blocks. When CSE
7704 encounters a simple @code{if} statement with no else clause,
7705 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7706 body of the @code{if}.
7708 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7710 @item -frerun-cse-after-loop
7711 @opindex frerun-cse-after-loop
7712 Re-run common subexpression elimination after loop optimizations are
7715 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7719 Perform a global common subexpression elimination pass.
7720 This pass also performs global constant and copy propagation.
7722 @emph{Note:} When compiling a program using computed gotos, a GCC
7723 extension, you may get better run-time performance if you disable
7724 the global common subexpression elimination pass by adding
7725 @option{-fno-gcse} to the command line.
7727 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7731 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7732 attempts to move loads that are only killed by stores into themselves. This
7733 allows a loop containing a load/store sequence to be changed to a load outside
7734 the loop, and a copy/store within the loop.
7736 Enabled by default when @option{-fgcse} is enabled.
7740 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7741 global common subexpression elimination. This pass attempts to move
7742 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7743 loops containing a load/store sequence can be changed to a load before
7744 the loop and a store after the loop.
7746 Not enabled at any optimization level.
7750 When @option{-fgcse-las} is enabled, the global common subexpression
7751 elimination pass eliminates redundant loads that come after stores to the
7752 same memory location (both partial and full redundancies).
7754 Not enabled at any optimization level.
7756 @item -fgcse-after-reload
7757 @opindex fgcse-after-reload
7758 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7759 pass is performed after reload. The purpose of this pass is to clean up
7762 @item -faggressive-loop-optimizations
7763 @opindex faggressive-loop-optimizations
7764 This option tells the loop optimizer to use language constraints to
7765 derive bounds for the number of iterations of a loop. This assumes that
7766 loop code does not invoke undefined behavior by for example causing signed
7767 integer overflows or out-of-bound array accesses. The bounds for the
7768 number of iterations of a loop are used to guide loop unrolling and peeling
7769 and loop exit test optimizations.
7770 This option is enabled by default.
7772 @item -funconstrained-commons
7773 @opindex funconstrained-commons
7774 This option tells the compiler that variables declared in common blocks
7775 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7776 prevents certain optimizations that depend on knowing the array bounds.
7778 @item -fcrossjumping
7779 @opindex fcrossjumping
7780 Perform cross-jumping transformation.
7781 This transformation unifies equivalent code and saves code size. The
7782 resulting code may or may not perform better than without cross-jumping.
7784 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7786 @item -fauto-inc-dec
7787 @opindex fauto-inc-dec
7788 Combine increments or decrements of addresses with memory accesses.
7789 This pass is always skipped on architectures that do not have
7790 instructions to support this. Enabled by default at @option{-O} and
7791 higher on architectures that support this.
7795 Perform dead code elimination (DCE) on RTL@.
7796 Enabled by default at @option{-O} and higher.
7800 Perform dead store elimination (DSE) on RTL@.
7801 Enabled by default at @option{-O} and higher.
7803 @item -fif-conversion
7804 @opindex fif-conversion
7805 Attempt to transform conditional jumps into branch-less equivalents. This
7806 includes use of conditional moves, min, max, set flags and abs instructions, and
7807 some tricks doable by standard arithmetics. The use of conditional execution
7808 on chips where it is available is controlled by @option{-fif-conversion2}.
7810 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7812 @item -fif-conversion2
7813 @opindex fif-conversion2
7814 Use conditional execution (where available) to transform conditional jumps into
7815 branch-less equivalents.
7817 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7819 @item -fdeclone-ctor-dtor
7820 @opindex fdeclone-ctor-dtor
7821 The C++ ABI requires multiple entry points for constructors and
7822 destructors: one for a base subobject, one for a complete object, and
7823 one for a virtual destructor that calls operator delete afterwards.
7824 For a hierarchy with virtual bases, the base and complete variants are
7825 clones, which means two copies of the function. With this option, the
7826 base and complete variants are changed to be thunks that call a common
7829 Enabled by @option{-Os}.
7831 @item -fdelete-null-pointer-checks
7832 @opindex fdelete-null-pointer-checks
7833 Assume that programs cannot safely dereference null pointers, and that
7834 no code or data element resides at address zero.
7835 This option enables simple constant
7836 folding optimizations at all optimization levels. In addition, other
7837 optimization passes in GCC use this flag to control global dataflow
7838 analyses that eliminate useless checks for null pointers; these assume
7839 that a memory access to address zero always results in a trap, so
7840 that if a pointer is checked after it has already been dereferenced,
7843 Note however that in some environments this assumption is not true.
7844 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7845 for programs that depend on that behavior.
7847 This option is enabled by default on most targets. On Nios II ELF, it
7848 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
7850 Passes that use the dataflow information
7851 are enabled independently at different optimization levels.
7853 @item -fdevirtualize
7854 @opindex fdevirtualize
7855 Attempt to convert calls to virtual functions to direct calls. This
7856 is done both within a procedure and interprocedurally as part of
7857 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7858 propagation (@option{-fipa-cp}).
7859 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7861 @item -fdevirtualize-speculatively
7862 @opindex fdevirtualize-speculatively
7863 Attempt to convert calls to virtual functions to speculative direct calls.
7864 Based on the analysis of the type inheritance graph, determine for a given call
7865 the set of likely targets. If the set is small, preferably of size 1, change
7866 the call into a conditional deciding between direct and indirect calls. The
7867 speculative calls enable more optimizations, such as inlining. When they seem
7868 useless after further optimization, they are converted back into original form.
7870 @item -fdevirtualize-at-ltrans
7871 @opindex fdevirtualize-at-ltrans
7872 Stream extra information needed for aggressive devirtualization when running
7873 the link-time optimizer in local transformation mode.
7874 This option enables more devirtualization but
7875 significantly increases the size of streamed data. For this reason it is
7876 disabled by default.
7878 @item -fexpensive-optimizations
7879 @opindex fexpensive-optimizations
7880 Perform a number of minor optimizations that are relatively expensive.
7882 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7886 Attempt to remove redundant extension instructions. This is especially
7887 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7888 registers after writing to their lower 32-bit half.
7890 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7891 @option{-O3}, @option{-Os}.
7893 @item -fno-lifetime-dse
7894 @opindex fno-lifetime-dse
7895 In C++ the value of an object is only affected by changes within its
7896 lifetime: when the constructor begins, the object has an indeterminate
7897 value, and any changes during the lifetime of the object are dead when
7898 the object is destroyed. Normally dead store elimination will take
7899 advantage of this; if your code relies on the value of the object
7900 storage persisting beyond the lifetime of the object, you can use this
7901 flag to disable this optimization. To preserve stores before the
7902 constructor starts (e.g. because your operator new clears the object
7903 storage) but still treat the object as dead after the destructor you,
7904 can use @option{-flifetime-dse=1}. The default behavior can be
7905 explicitly selected with @option{-flifetime-dse=2}.
7906 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7908 @item -flive-range-shrinkage
7909 @opindex flive-range-shrinkage
7910 Attempt to decrease register pressure through register live range
7911 shrinkage. This is helpful for fast processors with small or moderate
7914 @item -fira-algorithm=@var{algorithm}
7915 @opindex fira-algorithm
7916 Use the specified coloring algorithm for the integrated register
7917 allocator. The @var{algorithm} argument can be @samp{priority}, which
7918 specifies Chow's priority coloring, or @samp{CB}, which specifies
7919 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7920 for all architectures, but for those targets that do support it, it is
7921 the default because it generates better code.
7923 @item -fira-region=@var{region}
7924 @opindex fira-region
7925 Use specified regions for the integrated register allocator. The
7926 @var{region} argument should be one of the following:
7931 Use all loops as register allocation regions.
7932 This can give the best results for machines with a small and/or
7933 irregular register set.
7936 Use all loops except for loops with small register pressure
7937 as the regions. This value usually gives
7938 the best results in most cases and for most architectures,
7939 and is enabled by default when compiling with optimization for speed
7940 (@option{-O}, @option{-O2}, @dots{}).
7943 Use all functions as a single region.
7944 This typically results in the smallest code size, and is enabled by default for
7945 @option{-Os} or @option{-O0}.
7949 @item -fira-hoist-pressure
7950 @opindex fira-hoist-pressure
7951 Use IRA to evaluate register pressure in the code hoisting pass for
7952 decisions to hoist expressions. This option usually results in smaller
7953 code, but it can slow the compiler down.
7955 This option is enabled at level @option{-Os} for all targets.
7957 @item -fira-loop-pressure
7958 @opindex fira-loop-pressure
7959 Use IRA to evaluate register pressure in loops for decisions to move
7960 loop invariants. This option usually results in generation
7961 of faster and smaller code on machines with large register files (>= 32
7962 registers), but it can slow the compiler down.
7964 This option is enabled at level @option{-O3} for some targets.
7966 @item -fno-ira-share-save-slots
7967 @opindex fno-ira-share-save-slots
7968 Disable sharing of stack slots used for saving call-used hard
7969 registers living through a call. Each hard register gets a
7970 separate stack slot, and as a result function stack frames are
7973 @item -fno-ira-share-spill-slots
7974 @opindex fno-ira-share-spill-slots
7975 Disable sharing of stack slots allocated for pseudo-registers. Each
7976 pseudo-register that does not get a hard register gets a separate
7977 stack slot, and as a result function stack frames are larger.
7981 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7982 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7983 values if it is profitable.
7985 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7987 @item -fdelayed-branch
7988 @opindex fdelayed-branch
7989 If supported for the target machine, attempt to reorder instructions
7990 to exploit instruction slots available after delayed branch
7993 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7995 @item -fschedule-insns
7996 @opindex fschedule-insns
7997 If supported for the target machine, attempt to reorder instructions to
7998 eliminate execution stalls due to required data being unavailable. This
7999 helps machines that have slow floating point or memory load instructions
8000 by allowing other instructions to be issued until the result of the load
8001 or floating-point instruction is required.
8003 Enabled at levels @option{-O2}, @option{-O3}.
8005 @item -fschedule-insns2
8006 @opindex fschedule-insns2
8007 Similar to @option{-fschedule-insns}, but requests an additional pass of
8008 instruction scheduling after register allocation has been done. This is
8009 especially useful on machines with a relatively small number of
8010 registers and where memory load instructions take more than one cycle.
8012 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8014 @item -fno-sched-interblock
8015 @opindex fno-sched-interblock
8016 Don't schedule instructions across basic blocks. This is normally
8017 enabled by default when scheduling before register allocation, i.e.@:
8018 with @option{-fschedule-insns} or at @option{-O2} or higher.
8020 @item -fno-sched-spec
8021 @opindex fno-sched-spec
8022 Don't allow speculative motion of non-load instructions. This is normally
8023 enabled by default when scheduling before register allocation, i.e.@:
8024 with @option{-fschedule-insns} or at @option{-O2} or higher.
8026 @item -fsched-pressure
8027 @opindex fsched-pressure
8028 Enable register pressure sensitive insn scheduling before register
8029 allocation. This only makes sense when scheduling before register
8030 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8031 @option{-O2} or higher. Usage of this option can improve the
8032 generated code and decrease its size by preventing register pressure
8033 increase above the number of available hard registers and subsequent
8034 spills in register allocation.
8036 @item -fsched-spec-load
8037 @opindex fsched-spec-load
8038 Allow speculative motion of some load instructions. This only makes
8039 sense when scheduling before register allocation, i.e.@: with
8040 @option{-fschedule-insns} or at @option{-O2} or higher.
8042 @item -fsched-spec-load-dangerous
8043 @opindex fsched-spec-load-dangerous
8044 Allow speculative motion of more load instructions. This only makes
8045 sense when scheduling before register allocation, i.e.@: with
8046 @option{-fschedule-insns} or at @option{-O2} or higher.
8048 @item -fsched-stalled-insns
8049 @itemx -fsched-stalled-insns=@var{n}
8050 @opindex fsched-stalled-insns
8051 Define how many insns (if any) can be moved prematurely from the queue
8052 of stalled insns into the ready list during the second scheduling pass.
8053 @option{-fno-sched-stalled-insns} means that no insns are moved
8054 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8055 on how many queued insns can be moved prematurely.
8056 @option{-fsched-stalled-insns} without a value is equivalent to
8057 @option{-fsched-stalled-insns=1}.
8059 @item -fsched-stalled-insns-dep
8060 @itemx -fsched-stalled-insns-dep=@var{n}
8061 @opindex fsched-stalled-insns-dep
8062 Define how many insn groups (cycles) are examined for a dependency
8063 on a stalled insn that is a candidate for premature removal from the queue
8064 of stalled insns. This has an effect only during the second scheduling pass,
8065 and only if @option{-fsched-stalled-insns} is used.
8066 @option{-fno-sched-stalled-insns-dep} is equivalent to
8067 @option{-fsched-stalled-insns-dep=0}.
8068 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8069 @option{-fsched-stalled-insns-dep=1}.
8071 @item -fsched2-use-superblocks
8072 @opindex fsched2-use-superblocks
8073 When scheduling after register allocation, use superblock scheduling.
8074 This allows motion across basic block boundaries,
8075 resulting in faster schedules. This option is experimental, as not all machine
8076 descriptions used by GCC model the CPU closely enough to avoid unreliable
8077 results from the algorithm.
8079 This only makes sense when scheduling after register allocation, i.e.@: with
8080 @option{-fschedule-insns2} or at @option{-O2} or higher.
8082 @item -fsched-group-heuristic
8083 @opindex fsched-group-heuristic
8084 Enable the group heuristic in the scheduler. This heuristic favors
8085 the instruction that belongs to a schedule group. This is enabled
8086 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8087 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8089 @item -fsched-critical-path-heuristic
8090 @opindex fsched-critical-path-heuristic
8091 Enable the critical-path heuristic in the scheduler. This heuristic favors
8092 instructions on the critical path. This is enabled by default when
8093 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8094 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8096 @item -fsched-spec-insn-heuristic
8097 @opindex fsched-spec-insn-heuristic
8098 Enable the speculative instruction heuristic in the scheduler. This
8099 heuristic favors speculative instructions with greater dependency weakness.
8100 This is enabled by default when scheduling is enabled, i.e.@:
8101 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8102 or at @option{-O2} or higher.
8104 @item -fsched-rank-heuristic
8105 @opindex fsched-rank-heuristic
8106 Enable the rank heuristic in the scheduler. This heuristic favors
8107 the instruction belonging to a basic block with greater size or frequency.
8108 This is enabled by default when scheduling is enabled, i.e.@:
8109 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8110 at @option{-O2} or higher.
8112 @item -fsched-last-insn-heuristic
8113 @opindex fsched-last-insn-heuristic
8114 Enable the last-instruction heuristic in the scheduler. This heuristic
8115 favors the instruction that is less dependent on the last instruction
8116 scheduled. This is enabled by default when scheduling is enabled,
8117 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8118 at @option{-O2} or higher.
8120 @item -fsched-dep-count-heuristic
8121 @opindex fsched-dep-count-heuristic
8122 Enable the dependent-count heuristic in the scheduler. This heuristic
8123 favors the instruction that has more instructions depending on it.
8124 This is enabled by default when scheduling is enabled, i.e.@:
8125 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8126 at @option{-O2} or higher.
8128 @item -freschedule-modulo-scheduled-loops
8129 @opindex freschedule-modulo-scheduled-loops
8130 Modulo scheduling is performed before traditional scheduling. If a loop
8131 is modulo scheduled, later scheduling passes may change its schedule.
8132 Use this option to control that behavior.
8134 @item -fselective-scheduling
8135 @opindex fselective-scheduling
8136 Schedule instructions using selective scheduling algorithm. Selective
8137 scheduling runs instead of the first scheduler pass.
8139 @item -fselective-scheduling2
8140 @opindex fselective-scheduling2
8141 Schedule instructions using selective scheduling algorithm. Selective
8142 scheduling runs instead of the second scheduler pass.
8144 @item -fsel-sched-pipelining
8145 @opindex fsel-sched-pipelining
8146 Enable software pipelining of innermost loops during selective scheduling.
8147 This option has no effect unless one of @option{-fselective-scheduling} or
8148 @option{-fselective-scheduling2} is turned on.
8150 @item -fsel-sched-pipelining-outer-loops
8151 @opindex fsel-sched-pipelining-outer-loops
8152 When pipelining loops during selective scheduling, also pipeline outer loops.
8153 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8155 @item -fsemantic-interposition
8156 @opindex fsemantic-interposition
8157 Some object formats, like ELF, allow interposing of symbols by the
8159 This means that for symbols exported from the DSO, the compiler cannot perform
8160 interprocedural propagation, inlining and other optimizations in anticipation
8161 that the function or variable in question may change. While this feature is
8162 useful, for example, to rewrite memory allocation functions by a debugging
8163 implementation, it is expensive in the terms of code quality.
8164 With @option{-fno-semantic-interposition} the compiler assumes that
8165 if interposition happens for functions the overwriting function will have
8166 precisely the same semantics (and side effects).
8167 Similarly if interposition happens
8168 for variables, the constructor of the variable will be the same. The flag
8169 has no effect for functions explicitly declared inline
8170 (where it is never allowed for interposition to change semantics)
8171 and for symbols explicitly declared weak.
8174 @opindex fshrink-wrap
8175 Emit function prologues only before parts of the function that need it,
8176 rather than at the top of the function. This flag is enabled by default at
8177 @option{-O} and higher.
8179 @item -fshrink-wrap-separate
8180 @opindex fshrink-wrap-separate
8181 Shrink-wrap separate parts of the prologue and epilogue separately, so that
8182 those parts are only executed when needed.
8183 This option is on by default, but has no effect unless @option{-fshrink-wrap}
8184 is also turned on and the target supports this.
8186 @item -fcaller-saves
8187 @opindex fcaller-saves
8188 Enable allocation of values to registers that are clobbered by
8189 function calls, by emitting extra instructions to save and restore the
8190 registers around such calls. Such allocation is done only when it
8191 seems to result in better code.
8193 This option is always enabled by default on certain machines, usually
8194 those which have no call-preserved registers to use instead.
8196 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8198 @item -fcombine-stack-adjustments
8199 @opindex fcombine-stack-adjustments
8200 Tracks stack adjustments (pushes and pops) and stack memory references
8201 and then tries to find ways to combine them.
8203 Enabled by default at @option{-O1} and higher.
8207 Use caller save registers for allocation if those registers are not used by
8208 any called function. In that case it is not necessary to save and restore
8209 them around calls. This is only possible if called functions are part of
8210 same compilation unit as current function and they are compiled before it.
8212 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
8213 is disabled if generated code will be instrumented for profiling
8214 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
8215 exactly (this happens on targets that do not expose prologues
8216 and epilogues in RTL).
8218 @item -fconserve-stack
8219 @opindex fconserve-stack
8220 Attempt to minimize stack usage. The compiler attempts to use less
8221 stack space, even if that makes the program slower. This option
8222 implies setting the @option{large-stack-frame} parameter to 100
8223 and the @option{large-stack-frame-growth} parameter to 400.
8225 @item -ftree-reassoc
8226 @opindex ftree-reassoc
8227 Perform reassociation on trees. This flag is enabled by default
8228 at @option{-O} and higher.
8230 @item -fcode-hoisting
8231 @opindex fcode-hoisting
8232 Perform code hoisting. Code hoisting tries to move the
8233 evaluation of expressions executed on all paths to the function exit
8234 as early as possible. This is especially useful as a code size
8235 optimization, but it often helps for code speed as well.
8236 This flag is enabled by default at @option{-O2} and higher.
8240 Perform partial redundancy elimination (PRE) on trees. This flag is
8241 enabled by default at @option{-O2} and @option{-O3}.
8243 @item -ftree-partial-pre
8244 @opindex ftree-partial-pre
8245 Make partial redundancy elimination (PRE) more aggressive. This flag is
8246 enabled by default at @option{-O3}.
8248 @item -ftree-forwprop
8249 @opindex ftree-forwprop
8250 Perform forward propagation on trees. This flag is enabled by default
8251 at @option{-O} and higher.
8255 Perform full redundancy elimination (FRE) on trees. The difference
8256 between FRE and PRE is that FRE only considers expressions
8257 that are computed on all paths leading to the redundant computation.
8258 This analysis is faster than PRE, though it exposes fewer redundancies.
8259 This flag is enabled by default at @option{-O} and higher.
8261 @item -ftree-phiprop
8262 @opindex ftree-phiprop
8263 Perform hoisting of loads from conditional pointers on trees. This
8264 pass is enabled by default at @option{-O} and higher.
8266 @item -fhoist-adjacent-loads
8267 @opindex fhoist-adjacent-loads
8268 Speculatively hoist loads from both branches of an if-then-else if the
8269 loads are from adjacent locations in the same structure and the target
8270 architecture has a conditional move instruction. This flag is enabled
8271 by default at @option{-O2} and higher.
8273 @item -ftree-copy-prop
8274 @opindex ftree-copy-prop
8275 Perform copy propagation on trees. This pass eliminates unnecessary
8276 copy operations. This flag is enabled by default at @option{-O} and
8279 @item -fipa-pure-const
8280 @opindex fipa-pure-const
8281 Discover which functions are pure or constant.
8282 Enabled by default at @option{-O} and higher.
8284 @item -fipa-reference
8285 @opindex fipa-reference
8286 Discover which static variables do not escape the
8288 Enabled by default at @option{-O} and higher.
8292 Perform interprocedural pointer analysis and interprocedural modification
8293 and reference analysis. This option can cause excessive memory and
8294 compile-time usage on large compilation units. It is not enabled by
8295 default at any optimization level.
8298 @opindex fipa-profile
8299 Perform interprocedural profile propagation. The functions called only from
8300 cold functions are marked as cold. Also functions executed once (such as
8301 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8302 functions and loop less parts of functions executed once are then optimized for
8304 Enabled by default at @option{-O} and higher.
8308 Perform interprocedural constant propagation.
8309 This optimization analyzes the program to determine when values passed
8310 to functions are constants and then optimizes accordingly.
8311 This optimization can substantially increase performance
8312 if the application has constants passed to functions.
8313 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8315 @item -fipa-cp-clone
8316 @opindex fipa-cp-clone
8317 Perform function cloning to make interprocedural constant propagation stronger.
8318 When enabled, interprocedural constant propagation performs function cloning
8319 when externally visible function can be called with constant arguments.
8320 Because this optimization can create multiple copies of functions,
8321 it may significantly increase code size
8322 (see @option{--param ipcp-unit-growth=@var{value}}).
8323 This flag is enabled by default at @option{-O3}.
8326 @opindex -fipa-bit-cp
8327 When enabled, perform interprocedural bitwise constant
8328 propagation. This flag is enabled by default at @option{-O2}. It
8329 requires that @option{-fipa-cp} is enabled.
8333 When enabled, perform interprocedural propagation of value
8334 ranges. This flag is enabled by default at @option{-O2}. It requires
8335 that @option{-fipa-cp} is enabled.
8339 Perform Identical Code Folding for functions and read-only variables.
8340 The optimization reduces code size and may disturb unwind stacks by replacing
8341 a function by equivalent one with a different name. The optimization works
8342 more effectively with link-time optimization enabled.
8344 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8345 works on different levels and thus the optimizations are not same - there are
8346 equivalences that are found only by GCC and equivalences found only by Gold.
8348 This flag is enabled by default at @option{-O2} and @option{-Os}.
8350 @item -fisolate-erroneous-paths-dereference
8351 @opindex fisolate-erroneous-paths-dereference
8352 Detect paths that trigger erroneous or undefined behavior due to
8353 dereferencing a null pointer. Isolate those paths from the main control
8354 flow and turn the statement with erroneous or undefined behavior into a trap.
8355 This flag is enabled by default at @option{-O2} and higher and depends on
8356 @option{-fdelete-null-pointer-checks} also being enabled.
8358 @item -fisolate-erroneous-paths-attribute
8359 @opindex fisolate-erroneous-paths-attribute
8360 Detect paths that trigger erroneous or undefined behavior due a null value
8361 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8362 attribute. Isolate those paths from the main control flow and turn the
8363 statement with erroneous or undefined behavior into a trap. This is not
8364 currently enabled, but may be enabled by @option{-O2} in the future.
8368 Perform forward store motion on trees. This flag is
8369 enabled by default at @option{-O} and higher.
8371 @item -ftree-bit-ccp
8372 @opindex ftree-bit-ccp
8373 Perform sparse conditional bit constant propagation on trees and propagate
8374 pointer alignment information.
8375 This pass only operates on local scalar variables and is enabled by default
8376 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8380 Perform sparse conditional constant propagation (CCP) on trees. This
8381 pass only operates on local scalar variables and is enabled by default
8382 at @option{-O} and higher.
8384 @item -fssa-backprop
8385 @opindex fssa-backprop
8386 Propagate information about uses of a value up the definition chain
8387 in order to simplify the definitions. For example, this pass strips
8388 sign operations if the sign of a value never matters. The flag is
8389 enabled by default at @option{-O} and higher.
8392 @opindex fssa-phiopt
8393 Perform pattern matching on SSA PHI nodes to optimize conditional
8394 code. This pass is enabled by default at @option{-O} and higher.
8396 @item -ftree-switch-conversion
8397 @opindex ftree-switch-conversion
8398 Perform conversion of simple initializations in a switch to
8399 initializations from a scalar array. This flag is enabled by default
8400 at @option{-O2} and higher.
8402 @item -ftree-tail-merge
8403 @opindex ftree-tail-merge
8404 Look for identical code sequences. When found, replace one with a jump to the
8405 other. This optimization is known as tail merging or cross jumping. This flag
8406 is enabled by default at @option{-O2} and higher. The compilation time
8408 be limited using @option{max-tail-merge-comparisons} parameter and
8409 @option{max-tail-merge-iterations} parameter.
8413 Perform dead code elimination (DCE) on trees. This flag is enabled by
8414 default at @option{-O} and higher.
8416 @item -ftree-builtin-call-dce
8417 @opindex ftree-builtin-call-dce
8418 Perform conditional dead code elimination (DCE) for calls to built-in functions
8419 that may set @code{errno} but are otherwise side-effect free. This flag is
8420 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8423 @item -ftree-dominator-opts
8424 @opindex ftree-dominator-opts
8425 Perform a variety of simple scalar cleanups (constant/copy
8426 propagation, redundancy elimination, range propagation and expression
8427 simplification) based on a dominator tree traversal. This also
8428 performs jump threading (to reduce jumps to jumps). This flag is
8429 enabled by default at @option{-O} and higher.
8433 Perform dead store elimination (DSE) on trees. A dead store is a store into
8434 a memory location that is later overwritten by another store without
8435 any intervening loads. In this case the earlier store can be deleted. This
8436 flag is enabled by default at @option{-O} and higher.
8440 Perform loop header copying on trees. This is beneficial since it increases
8441 effectiveness of code motion optimizations. It also saves one jump. This flag
8442 is enabled by default at @option{-O} and higher. It is not enabled
8443 for @option{-Os}, since it usually increases code size.
8445 @item -ftree-loop-optimize
8446 @opindex ftree-loop-optimize
8447 Perform loop optimizations on trees. This flag is enabled by default
8448 at @option{-O} and higher.
8450 @item -ftree-loop-linear
8451 @itemx -floop-interchange
8452 @itemx -floop-strip-mine
8454 @itemx -floop-unroll-and-jam
8455 @opindex ftree-loop-linear
8456 @opindex floop-interchange
8457 @opindex floop-strip-mine
8458 @opindex floop-block
8459 @opindex floop-unroll-and-jam
8460 Perform loop nest optimizations. Same as
8461 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8462 to be configured with @option{--with-isl} to enable the Graphite loop
8463 transformation infrastructure.
8465 @item -fgraphite-identity
8466 @opindex fgraphite-identity
8467 Enable the identity transformation for graphite. For every SCoP we generate
8468 the polyhedral representation and transform it back to gimple. Using
8469 @option{-fgraphite-identity} we can check the costs or benefits of the
8470 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8471 are also performed by the code generator isl, like index splitting and
8472 dead code elimination in loops.
8474 @item -floop-nest-optimize
8475 @opindex floop-nest-optimize
8476 Enable the isl based loop nest optimizer. This is a generic loop nest
8477 optimizer based on the Pluto optimization algorithms. It calculates a loop
8478 structure optimized for data-locality and parallelism. This option
8481 @item -floop-parallelize-all
8482 @opindex floop-parallelize-all
8483 Use the Graphite data dependence analysis to identify loops that can
8484 be parallelized. Parallelize all the loops that can be analyzed to
8485 not contain loop carried dependences without checking that it is
8486 profitable to parallelize the loops.
8488 @item -ftree-coalesce-vars
8489 @opindex ftree-coalesce-vars
8490 While transforming the program out of the SSA representation, attempt to
8491 reduce copying by coalescing versions of different user-defined
8492 variables, instead of just compiler temporaries. This may severely
8493 limit the ability to debug an optimized program compiled with
8494 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8495 prevents SSA coalescing of user variables. This option is enabled by
8496 default if optimization is enabled, and it does very little otherwise.
8498 @item -ftree-loop-if-convert
8499 @opindex ftree-loop-if-convert
8500 Attempt to transform conditional jumps in the innermost loops to
8501 branch-less equivalents. The intent is to remove control-flow from
8502 the innermost loops in order to improve the ability of the
8503 vectorization pass to handle these loops. This is enabled by default
8504 if vectorization is enabled.
8506 @item -ftree-loop-distribution
8507 @opindex ftree-loop-distribution
8508 Perform loop distribution. This flag can improve cache performance on
8509 big loop bodies and allow further loop optimizations, like
8510 parallelization or vectorization, to take place. For example, the loop
8527 @item -ftree-loop-distribute-patterns
8528 @opindex ftree-loop-distribute-patterns
8529 Perform loop distribution of patterns that can be code generated with
8530 calls to a library. This flag is enabled by default at @option{-O3}.
8532 This pass distributes the initialization loops and generates a call to
8533 memset zero. For example, the loop
8549 and the initialization loop is transformed into a call to memset zero.
8551 @item -ftree-loop-im
8552 @opindex ftree-loop-im
8553 Perform loop invariant motion on trees. This pass moves only invariants that
8554 are hard to handle at RTL level (function calls, operations that expand to
8555 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8556 operands of conditions that are invariant out of the loop, so that we can use
8557 just trivial invariantness analysis in loop unswitching. The pass also includes
8560 @item -ftree-loop-ivcanon
8561 @opindex ftree-loop-ivcanon
8562 Create a canonical counter for number of iterations in loops for which
8563 determining number of iterations requires complicated analysis. Later
8564 optimizations then may determine the number easily. Useful especially
8565 in connection with unrolling.
8569 Perform induction variable optimizations (strength reduction, induction
8570 variable merging and induction variable elimination) on trees.
8572 @item -ftree-parallelize-loops=n
8573 @opindex ftree-parallelize-loops
8574 Parallelize loops, i.e., split their iteration space to run in n threads.
8575 This is only possible for loops whose iterations are independent
8576 and can be arbitrarily reordered. The optimization is only
8577 profitable on multiprocessor machines, for loops that are CPU-intensive,
8578 rather than constrained e.g.@: by memory bandwidth. This option
8579 implies @option{-pthread}, and thus is only supported on targets
8580 that have support for @option{-pthread}.
8584 Perform function-local points-to analysis on trees. This flag is
8585 enabled by default at @option{-O} and higher.
8589 Perform scalar replacement of aggregates. This pass replaces structure
8590 references with scalars to prevent committing structures to memory too
8591 early. This flag is enabled by default at @option{-O} and higher.
8593 @item -fstore-merging
8594 @opindex fstore-merging
8595 Perform merging of narrow stores to consecutive memory addresses. This pass
8596 merges contiguous stores of immediate values narrower than a word into fewer
8597 wider stores to reduce the number of instructions. This is enabled by default
8598 at @option{-O2} and higher as well as @option{-Os}.
8602 Perform temporary expression replacement during the SSA->normal phase. Single
8603 use/single def temporaries are replaced at their use location with their
8604 defining expression. This results in non-GIMPLE code, but gives the expanders
8605 much more complex trees to work on resulting in better RTL generation. This is
8606 enabled by default at @option{-O} and higher.
8610 Perform straight-line strength reduction on trees. This recognizes related
8611 expressions involving multiplications and replaces them by less expensive
8612 calculations when possible. This is enabled by default at @option{-O} and
8615 @item -ftree-vectorize
8616 @opindex ftree-vectorize
8617 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8618 and @option{-ftree-slp-vectorize} if not explicitly specified.
8620 @item -ftree-loop-vectorize
8621 @opindex ftree-loop-vectorize
8622 Perform loop vectorization on trees. This flag is enabled by default at
8623 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8625 @item -ftree-slp-vectorize
8626 @opindex ftree-slp-vectorize
8627 Perform basic block vectorization on trees. This flag is enabled by default at
8628 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8630 @item -fvect-cost-model=@var{model}
8631 @opindex fvect-cost-model
8632 Alter the cost model used for vectorization. The @var{model} argument
8633 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8634 With the @samp{unlimited} model the vectorized code-path is assumed
8635 to be profitable while with the @samp{dynamic} model a runtime check
8636 guards the vectorized code-path to enable it only for iteration
8637 counts that will likely execute faster than when executing the original
8638 scalar loop. The @samp{cheap} model disables vectorization of
8639 loops where doing so would be cost prohibitive for example due to
8640 required runtime checks for data dependence or alignment but otherwise
8641 is equal to the @samp{dynamic} model.
8642 The default cost model depends on other optimization flags and is
8643 either @samp{dynamic} or @samp{cheap}.
8645 @item -fsimd-cost-model=@var{model}
8646 @opindex fsimd-cost-model
8647 Alter the cost model used for vectorization of loops marked with the OpenMP
8648 or Cilk Plus simd directive. The @var{model} argument should be one of
8649 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8650 have the same meaning as described in @option{-fvect-cost-model} and by
8651 default a cost model defined with @option{-fvect-cost-model} is used.
8655 Perform Value Range Propagation on trees. This is similar to the
8656 constant propagation pass, but instead of values, ranges of values are
8657 propagated. This allows the optimizers to remove unnecessary range
8658 checks like array bound checks and null pointer checks. This is
8659 enabled by default at @option{-O2} and higher. Null pointer check
8660 elimination is only done if @option{-fdelete-null-pointer-checks} is
8664 @opindex fsplit-paths
8665 Split paths leading to loop backedges. This can improve dead code
8666 elimination and common subexpression elimination. This is enabled by
8667 default at @option{-O2} and above.
8669 @item -fsplit-ivs-in-unroller
8670 @opindex fsplit-ivs-in-unroller
8671 Enables expression of values of induction variables in later iterations
8672 of the unrolled loop using the value in the first iteration. This breaks
8673 long dependency chains, thus improving efficiency of the scheduling passes.
8675 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8676 same effect. However, that is not reliable in cases where the loop body
8677 is more complicated than a single basic block. It also does not work at all
8678 on some architectures due to restrictions in the CSE pass.
8680 This optimization is enabled by default.
8682 @item -fvariable-expansion-in-unroller
8683 @opindex fvariable-expansion-in-unroller
8684 With this option, the compiler creates multiple copies of some
8685 local variables when unrolling a loop, which can result in superior code.
8687 @item -fpartial-inlining
8688 @opindex fpartial-inlining
8689 Inline parts of functions. This option has any effect only
8690 when inlining itself is turned on by the @option{-finline-functions}
8691 or @option{-finline-small-functions} options.
8693 Enabled at level @option{-O2}.
8695 @item -fpredictive-commoning
8696 @opindex fpredictive-commoning
8697 Perform predictive commoning optimization, i.e., reusing computations
8698 (especially memory loads and stores) performed in previous
8699 iterations of loops.
8701 This option is enabled at level @option{-O3}.
8703 @item -fprefetch-loop-arrays
8704 @opindex fprefetch-loop-arrays
8705 If supported by the target machine, generate instructions to prefetch
8706 memory to improve the performance of loops that access large arrays.
8708 This option may generate better or worse code; results are highly
8709 dependent on the structure of loops within the source code.
8711 Disabled at level @option{-Os}.
8713 @item -fno-printf-return-value
8714 @opindex fno-printf-return-value
8715 Do not substitute constants for known return value of formatted output
8716 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
8717 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
8718 transformation allows GCC to optimize or even eliminate branches based
8719 on the known return value of these functions called with arguments that
8720 are either constant, or whose values are known to be in a range that
8721 makes determining the exact return value possible. For example, when
8722 @option{-fprintf-return-value} is in effect, both the branch and the
8723 body of the @code{if} statement (but not the call to @code{snprint})
8724 can be optimized away when @code{i} is a 32-bit or smaller integer
8725 because the return value is guaranteed to be at most 8.
8729 if (snprintf (buf, "%08x", i) >= sizeof buf)
8733 The @option{-fprintf-return-value} option relies on other optimizations
8734 and yields best results with @option{-O2}. It works in tandem with the
8735 @option{-Wformat-overflow} and @option{-Wformat-truncation} options.
8736 The @option{-fprintf-return-value} option is enabled by default.
8739 @itemx -fno-peephole2
8740 @opindex fno-peephole
8741 @opindex fno-peephole2
8742 Disable any machine-specific peephole optimizations. The difference
8743 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8744 are implemented in the compiler; some targets use one, some use the
8745 other, a few use both.
8747 @option{-fpeephole} is enabled by default.
8748 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8750 @item -fno-guess-branch-probability
8751 @opindex fno-guess-branch-probability
8752 Do not guess branch probabilities using heuristics.
8754 GCC uses heuristics to guess branch probabilities if they are
8755 not provided by profiling feedback (@option{-fprofile-arcs}). These
8756 heuristics are based on the control flow graph. If some branch probabilities
8757 are specified by @code{__builtin_expect}, then the heuristics are
8758 used to guess branch probabilities for the rest of the control flow graph,
8759 taking the @code{__builtin_expect} info into account. The interactions
8760 between the heuristics and @code{__builtin_expect} can be complex, and in
8761 some cases, it may be useful to disable the heuristics so that the effects
8762 of @code{__builtin_expect} are easier to understand.
8764 The default is @option{-fguess-branch-probability} at levels
8765 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8767 @item -freorder-blocks
8768 @opindex freorder-blocks
8769 Reorder basic blocks in the compiled function in order to reduce number of
8770 taken branches and improve code locality.
8772 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8774 @item -freorder-blocks-algorithm=@var{algorithm}
8775 @opindex freorder-blocks-algorithm
8776 Use the specified algorithm for basic block reordering. The
8777 @var{algorithm} argument can be @samp{simple}, which does not increase
8778 code size (except sometimes due to secondary effects like alignment),
8779 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8780 put all often executed code together, minimizing the number of branches
8781 executed by making extra copies of code.
8783 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8784 @samp{stc} at levels @option{-O2}, @option{-O3}.
8786 @item -freorder-blocks-and-partition
8787 @opindex freorder-blocks-and-partition
8788 In addition to reordering basic blocks in the compiled function, in order
8789 to reduce number of taken branches, partitions hot and cold basic blocks
8790 into separate sections of the assembly and @file{.o} files, to improve
8791 paging and cache locality performance.
8793 This optimization is automatically turned off in the presence of
8794 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
8795 section attribute and on any architecture that does not support named
8796 sections. When @option{-fsplit-stack} is used this option is not
8797 enabled by default (to avoid linker errors), but may be enabled
8798 explicitly (if using a working linker).
8800 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8802 @item -freorder-functions
8803 @opindex freorder-functions
8804 Reorder functions in the object file in order to
8805 improve code locality. This is implemented by using special
8806 subsections @code{.text.hot} for most frequently executed functions and
8807 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8808 the linker so object file format must support named sections and linker must
8809 place them in a reasonable way.
8811 Also profile feedback must be available to make this option effective. See
8812 @option{-fprofile-arcs} for details.
8814 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8816 @item -fstrict-aliasing
8817 @opindex fstrict-aliasing
8818 Allow the compiler to assume the strictest aliasing rules applicable to
8819 the language being compiled. For C (and C++), this activates
8820 optimizations based on the type of expressions. In particular, an
8821 object of one type is assumed never to reside at the same address as an
8822 object of a different type, unless the types are almost the same. For
8823 example, an @code{unsigned int} can alias an @code{int}, but not a
8824 @code{void*} or a @code{double}. A character type may alias any other
8827 @anchor{Type-punning}Pay special attention to code like this:
8840 The practice of reading from a different union member than the one most
8841 recently written to (called ``type-punning'') is common. Even with
8842 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8843 is accessed through the union type. So, the code above works as
8844 expected. @xref{Structures unions enumerations and bit-fields
8845 implementation}. However, this code might not:
8856 Similarly, access by taking the address, casting the resulting pointer
8857 and dereferencing the result has undefined behavior, even if the cast
8858 uses a union type, e.g.:
8862 return ((union a_union *) &d)->i;
8866 The @option{-fstrict-aliasing} option is enabled at levels
8867 @option{-O2}, @option{-O3}, @option{-Os}.
8869 @item -falign-functions
8870 @itemx -falign-functions=@var{n}
8871 @opindex falign-functions
8872 Align the start of functions to the next power-of-two greater than
8873 @var{n}, skipping up to @var{n} bytes. For instance,
8874 @option{-falign-functions=32} aligns functions to the next 32-byte
8875 boundary, but @option{-falign-functions=24} aligns to the next
8876 32-byte boundary only if this can be done by skipping 23 bytes or less.
8878 @option{-fno-align-functions} and @option{-falign-functions=1} are
8879 equivalent and mean that functions are not aligned.
8881 Some assemblers only support this flag when @var{n} is a power of two;
8882 in that case, it is rounded up.
8884 If @var{n} is not specified or is zero, use a machine-dependent default.
8886 Enabled at levels @option{-O2}, @option{-O3}.
8888 @item -flimit-function-alignment
8889 If this option is enabled, the compiler tries to avoid unnecessarily
8890 overaligning functions. It attempts to instruct the assembler to align
8891 by the amount specified by @option{-falign-functions}, but not to
8892 skip more bytes than the size of the function.
8894 @item -falign-labels
8895 @itemx -falign-labels=@var{n}
8896 @opindex falign-labels
8897 Align all branch targets to a power-of-two boundary, skipping up to
8898 @var{n} bytes like @option{-falign-functions}. This option can easily
8899 make code slower, because it must insert dummy operations for when the
8900 branch target is reached in the usual flow of the code.
8902 @option{-fno-align-labels} and @option{-falign-labels=1} are
8903 equivalent and mean that labels are not aligned.
8905 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8906 are greater than this value, then their values are used instead.
8908 If @var{n} is not specified or is zero, use a machine-dependent default
8909 which is very likely to be @samp{1}, meaning no alignment.
8911 Enabled at levels @option{-O2}, @option{-O3}.
8914 @itemx -falign-loops=@var{n}
8915 @opindex falign-loops
8916 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8917 like @option{-falign-functions}. If the loops are
8918 executed many times, this makes up for any execution of the dummy
8921 @option{-fno-align-loops} and @option{-falign-loops=1} are
8922 equivalent and mean that loops are not aligned.
8924 If @var{n} is not specified or is zero, use a machine-dependent default.
8926 Enabled at levels @option{-O2}, @option{-O3}.
8929 @itemx -falign-jumps=@var{n}
8930 @opindex falign-jumps
8931 Align branch targets to a power-of-two boundary, for branch targets
8932 where the targets can only be reached by jumping, skipping up to @var{n}
8933 bytes like @option{-falign-functions}. In this case, no dummy operations
8936 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8937 equivalent and mean that loops are not aligned.
8939 If @var{n} is not specified or is zero, use a machine-dependent default.
8941 Enabled at levels @option{-O2}, @option{-O3}.
8943 @item -funit-at-a-time
8944 @opindex funit-at-a-time
8945 This option is left for compatibility reasons. @option{-funit-at-a-time}
8946 has no effect, while @option{-fno-unit-at-a-time} implies
8947 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8951 @item -fno-toplevel-reorder
8952 @opindex fno-toplevel-reorder
8953 Do not reorder top-level functions, variables, and @code{asm}
8954 statements. Output them in the same order that they appear in the
8955 input file. When this option is used, unreferenced static variables
8956 are not removed. This option is intended to support existing code
8957 that relies on a particular ordering. For new code, it is better to
8958 use attributes when possible.
8960 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8961 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8966 Constructs webs as commonly used for register allocation purposes and assign
8967 each web individual pseudo register. This allows the register allocation pass
8968 to operate on pseudos directly, but also strengthens several other optimization
8969 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8970 however, make debugging impossible, since variables no longer stay in a
8973 Enabled by default with @option{-funroll-loops}.
8975 @item -fwhole-program
8976 @opindex fwhole-program
8977 Assume that the current compilation unit represents the whole program being
8978 compiled. All public functions and variables with the exception of @code{main}
8979 and those merged by attribute @code{externally_visible} become static functions
8980 and in effect are optimized more aggressively by interprocedural optimizers.
8982 This option should not be used in combination with @option{-flto}.
8983 Instead relying on a linker plugin should provide safer and more precise
8986 @item -flto[=@var{n}]
8988 This option runs the standard link-time optimizer. When invoked
8989 with source code, it generates GIMPLE (one of GCC's internal
8990 representations) and writes it to special ELF sections in the object
8991 file. When the object files are linked together, all the function
8992 bodies are read from these ELF sections and instantiated as if they
8993 had been part of the same translation unit.
8995 To use the link-time optimizer, @option{-flto} and optimization
8996 options should be specified at compile time and during the final link.
8997 It is recommended that you compile all the files participating in the
8998 same link with the same options and also specify those options at
9003 gcc -c -O2 -flto foo.c
9004 gcc -c -O2 -flto bar.c
9005 gcc -o myprog -flto -O2 foo.o bar.o
9008 The first two invocations to GCC save a bytecode representation
9009 of GIMPLE into special ELF sections inside @file{foo.o} and
9010 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9011 @file{foo.o} and @file{bar.o}, merges the two files into a single
9012 internal image, and compiles the result as usual. Since both
9013 @file{foo.o} and @file{bar.o} are merged into a single image, this
9014 causes all the interprocedural analyses and optimizations in GCC to
9015 work across the two files as if they were a single one. This means,
9016 for example, that the inliner is able to inline functions in
9017 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9019 Another (simpler) way to enable link-time optimization is:
9022 gcc -o myprog -flto -O2 foo.c bar.c
9025 The above generates bytecode for @file{foo.c} and @file{bar.c},
9026 merges them together into a single GIMPLE representation and optimizes
9027 them as usual to produce @file{myprog}.
9029 The only important thing to keep in mind is that to enable link-time
9030 optimizations you need to use the GCC driver to perform the link step.
9031 GCC then automatically performs link-time optimization if any of the
9032 objects involved were compiled with the @option{-flto} command-line option.
9034 should specify the optimization options to be used for link-time
9035 optimization though GCC tries to be clever at guessing an
9036 optimization level to use from the options used at compile time
9037 if you fail to specify one at link time. You can always override
9038 the automatic decision to do link-time optimization
9039 by passing @option{-fno-lto} to the link command.
9041 To make whole program optimization effective, it is necessary to make
9042 certain whole program assumptions. The compiler needs to know
9043 what functions and variables can be accessed by libraries and runtime
9044 outside of the link-time optimized unit. When supported by the linker,
9045 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9046 to the compiler about used and externally visible symbols. When
9047 the linker plugin is not available, @option{-fwhole-program} should be
9048 used to allow the compiler to make these assumptions, which leads
9049 to more aggressive optimization decisions.
9051 When @option{-fuse-linker-plugin} is not enabled, when a file is
9052 compiled with @option{-flto}, the generated object file is larger than
9053 a regular object file because it contains GIMPLE bytecodes and the usual
9054 final code (see @option{-ffat-lto-objects}. This means that
9055 object files with LTO information can be linked as normal object
9056 files; if @option{-fno-lto} is passed to the linker, no
9057 interprocedural optimizations are applied. Note that when
9058 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
9059 but you cannot perform a regular, non-LTO link on them.
9061 Additionally, the optimization flags used to compile individual files
9062 are not necessarily related to those used at link time. For instance,
9065 gcc -c -O0 -ffat-lto-objects -flto foo.c
9066 gcc -c -O0 -ffat-lto-objects -flto bar.c
9067 gcc -o myprog -O3 foo.o bar.o
9070 This produces individual object files with unoptimized assembler
9071 code, but the resulting binary @file{myprog} is optimized at
9072 @option{-O3}. If, instead, the final binary is generated with
9073 @option{-fno-lto}, then @file{myprog} is not optimized.
9075 When producing the final binary, GCC only
9076 applies link-time optimizations to those files that contain bytecode.
9077 Therefore, you can mix and match object files and libraries with
9078 GIMPLE bytecodes and final object code. GCC automatically selects
9079 which files to optimize in LTO mode and which files to link without
9082 There are some code generation flags preserved by GCC when
9083 generating bytecodes, as they need to be used during the final link
9084 stage. Generally options specified at link time override those
9085 specified at compile time.
9087 If you do not specify an optimization level option @option{-O} at
9088 link time, then GCC uses the highest optimization level
9089 used when compiling the object files.
9091 Currently, the following options and their settings are taken from
9092 the first object file that explicitly specifies them:
9093 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9094 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9095 and all the @option{-m} target flags.
9097 Certain ABI-changing flags are required to match in all compilation units,
9098 and trying to override this at link time with a conflicting value
9099 is ignored. This includes options such as @option{-freg-struct-return}
9100 and @option{-fpcc-struct-return}.
9102 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9103 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9104 are passed through to the link stage and merged conservatively for
9105 conflicting translation units. Specifically
9106 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9107 precedence; and for example @option{-ffp-contract=off} takes precedence
9108 over @option{-ffp-contract=fast}. You can override them at link time.
9110 If LTO encounters objects with C linkage declared with incompatible
9111 types in separate translation units to be linked together (undefined
9112 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9113 issued. The behavior is still undefined at run time. Similar
9114 diagnostics may be raised for other languages.
9116 Another feature of LTO is that it is possible to apply interprocedural
9117 optimizations on files written in different languages:
9122 gfortran -c -flto baz.f90
9123 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9126 Notice that the final link is done with @command{g++} to get the C++
9127 runtime libraries and @option{-lgfortran} is added to get the Fortran
9128 runtime libraries. In general, when mixing languages in LTO mode, you
9129 should use the same link command options as when mixing languages in a
9130 regular (non-LTO) compilation.
9132 If object files containing GIMPLE bytecode are stored in a library archive, say
9133 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9134 are using a linker with plugin support. To create static libraries suitable
9135 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9136 and @command{ranlib};
9137 to show the symbols of object files with GIMPLE bytecode, use
9138 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9139 and @command{nm} have been compiled with plugin support. At link time, use the the
9140 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9141 the LTO optimization process:
9144 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9147 With the linker plugin enabled, the linker extracts the needed
9148 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9149 to make them part of the aggregated GIMPLE image to be optimized.
9151 If you are not using a linker with plugin support and/or do not
9152 enable the linker plugin, then the objects inside @file{libfoo.a}
9153 are extracted and linked as usual, but they do not participate
9154 in the LTO optimization process. In order to make a static library suitable
9155 for both LTO optimization and usual linkage, compile its object files with
9156 @option{-flto} @option{-ffat-lto-objects}.
9158 Link-time optimizations do not require the presence of the whole program to
9159 operate. If the program does not require any symbols to be exported, it is
9160 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9161 the interprocedural optimizers to use more aggressive assumptions which may
9162 lead to improved optimization opportunities.
9163 Use of @option{-fwhole-program} is not needed when linker plugin is
9164 active (see @option{-fuse-linker-plugin}).
9166 The current implementation of LTO makes no
9167 attempt to generate bytecode that is portable between different
9168 types of hosts. The bytecode files are versioned and there is a
9169 strict version check, so bytecode files generated in one version of
9170 GCC do not work with an older or newer version of GCC.
9172 Link-time optimization does not work well with generation of debugging
9173 information. Combining @option{-flto} with
9174 @option{-g} is currently experimental and expected to produce unexpected
9177 If you specify the optional @var{n}, the optimization and code
9178 generation done at link time is executed in parallel using @var{n}
9179 parallel jobs by utilizing an installed @command{make} program. The
9180 environment variable @env{MAKE} may be used to override the program
9181 used. The default value for @var{n} is 1.
9183 You can also specify @option{-flto=jobserver} to use GNU make's
9184 job server mode to determine the number of parallel jobs. This
9185 is useful when the Makefile calling GCC is already executing in parallel.
9186 You must prepend a @samp{+} to the command recipe in the parent Makefile
9187 for this to work. This option likely only works if @env{MAKE} is
9190 @item -flto-partition=@var{alg}
9191 @opindex flto-partition
9192 Specify the partitioning algorithm used by the link-time optimizer.
9193 The value is either @samp{1to1} to specify a partitioning mirroring
9194 the original source files or @samp{balanced} to specify partitioning
9195 into equally sized chunks (whenever possible) or @samp{max} to create
9196 new partition for every symbol where possible. Specifying @samp{none}
9197 as an algorithm disables partitioning and streaming completely.
9198 The default value is @samp{balanced}. While @samp{1to1} can be used
9199 as an workaround for various code ordering issues, the @samp{max}
9200 partitioning is intended for internal testing only.
9201 The value @samp{one} specifies that exactly one partition should be
9202 used while the value @samp{none} bypasses partitioning and executes
9203 the link-time optimization step directly from the WPA phase.
9205 @item -flto-odr-type-merging
9206 @opindex flto-odr-type-merging
9207 Enable streaming of mangled types names of C++ types and their unification
9208 at link time. This increases size of LTO object files, but enables
9209 diagnostics about One Definition Rule violations.
9211 @item -flto-compression-level=@var{n}
9212 @opindex flto-compression-level
9213 This option specifies the level of compression used for intermediate
9214 language written to LTO object files, and is only meaningful in
9215 conjunction with LTO mode (@option{-flto}). Valid
9216 values are 0 (no compression) to 9 (maximum compression). Values
9217 outside this range are clamped to either 0 or 9. If the option is not
9218 given, a default balanced compression setting is used.
9220 @item -fuse-linker-plugin
9221 @opindex fuse-linker-plugin
9222 Enables the use of a linker plugin during link-time optimization. This
9223 option relies on plugin support in the linker, which is available in gold
9224 or in GNU ld 2.21 or newer.
9226 This option enables the extraction of object files with GIMPLE bytecode out
9227 of library archives. This improves the quality of optimization by exposing
9228 more code to the link-time optimizer. This information specifies what
9229 symbols can be accessed externally (by non-LTO object or during dynamic
9230 linking). Resulting code quality improvements on binaries (and shared
9231 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9232 See @option{-flto} for a description of the effect of this flag and how to
9235 This option is enabled by default when LTO support in GCC is enabled
9236 and GCC was configured for use with
9237 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9239 @item -ffat-lto-objects
9240 @opindex ffat-lto-objects
9241 Fat LTO objects are object files that contain both the intermediate language
9242 and the object code. This makes them usable for both LTO linking and normal
9243 linking. This option is effective only when compiling with @option{-flto}
9244 and is ignored at link time.
9246 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9247 requires the complete toolchain to be aware of LTO. It requires a linker with
9248 linker plugin support for basic functionality. Additionally,
9249 @command{nm}, @command{ar} and @command{ranlib}
9250 need to support linker plugins to allow a full-featured build environment
9251 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9252 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9253 to these tools. With non fat LTO makefiles need to be modified to use them.
9255 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9258 @item -fcompare-elim
9259 @opindex fcompare-elim
9260 After register allocation and post-register allocation instruction splitting,
9261 identify arithmetic instructions that compute processor flags similar to a
9262 comparison operation based on that arithmetic. If possible, eliminate the
9263 explicit comparison operation.
9265 This pass only applies to certain targets that cannot explicitly represent
9266 the comparison operation before register allocation is complete.
9268 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9270 @item -fcprop-registers
9271 @opindex fcprop-registers
9272 After register allocation and post-register allocation instruction splitting,
9273 perform a copy-propagation pass to try to reduce scheduling dependencies
9274 and occasionally eliminate the copy.
9276 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9278 @item -fprofile-correction
9279 @opindex fprofile-correction
9280 Profiles collected using an instrumented binary for multi-threaded programs may
9281 be inconsistent due to missed counter updates. When this option is specified,
9282 GCC uses heuristics to correct or smooth out such inconsistencies. By
9283 default, GCC emits an error message when an inconsistent profile is detected.
9286 @itemx -fprofile-use=@var{path}
9287 @opindex fprofile-use
9288 Enable profile feedback-directed optimizations,
9289 and the following optimizations
9290 which are generally profitable only with profile feedback available:
9291 @option{-fbranch-probabilities}, @option{-fvpt},
9292 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9293 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9295 Before you can use this option, you must first generate profiling information.
9296 @xref{Instrumentation Options}, for information about the
9297 @option{-fprofile-generate} option.
9299 By default, GCC emits an error message if the feedback profiles do not
9300 match the source code. This error can be turned into a warning by using
9301 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9304 If @var{path} is specified, GCC looks at the @var{path} to find
9305 the profile feedback data files. See @option{-fprofile-dir}.
9307 @item -fauto-profile
9308 @itemx -fauto-profile=@var{path}
9309 @opindex fauto-profile
9310 Enable sampling-based feedback-directed optimizations,
9311 and the following optimizations
9312 which are generally profitable only with profile feedback available:
9313 @option{-fbranch-probabilities}, @option{-fvpt},
9314 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9315 @option{-ftree-vectorize},
9316 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9317 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9318 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9320 @var{path} is the name of a file containing AutoFDO profile information.
9321 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9323 Producing an AutoFDO profile data file requires running your program
9324 with the @command{perf} utility on a supported GNU/Linux target system.
9325 For more information, see @uref{https://perf.wiki.kernel.org/}.
9329 perf record -e br_inst_retired:near_taken -b -o perf.data \
9333 Then use the @command{create_gcov} tool to convert the raw profile data
9334 to a format that can be used by GCC.@ You must also supply the
9335 unstripped binary for your program to this tool.
9336 See @uref{https://github.com/google/autofdo}.
9340 create_gcov --binary=your_program.unstripped --profile=perf.data \
9345 The following options control compiler behavior regarding floating-point
9346 arithmetic. These options trade off between speed and
9347 correctness. All must be specifically enabled.
9351 @opindex ffloat-store
9352 Do not store floating-point variables in registers, and inhibit other
9353 options that might change whether a floating-point value is taken from a
9356 @cindex floating-point precision
9357 This option prevents undesirable excess precision on machines such as
9358 the 68000 where the floating registers (of the 68881) keep more
9359 precision than a @code{double} is supposed to have. Similarly for the
9360 x86 architecture. For most programs, the excess precision does only
9361 good, but a few programs rely on the precise definition of IEEE floating
9362 point. Use @option{-ffloat-store} for such programs, after modifying
9363 them to store all pertinent intermediate computations into variables.
9365 @item -fexcess-precision=@var{style}
9366 @opindex fexcess-precision
9367 This option allows further control over excess precision on machines
9368 where floating-point operations occur in a format with more precision or
9369 range than the IEEE standard and interchange floating-point types. By
9370 default, @option{-fexcess-precision=fast} is in effect; this means that
9371 operations may be carried out in a wider precision than the types specified
9372 in the source if that would result in faster code, and it is unpredictable
9373 when rounding to the types specified in the source code takes place.
9374 When compiling C, if @option{-fexcess-precision=standard} is specified then
9375 excess precision follows the rules specified in ISO C99; in particular,
9376 both casts and assignments cause values to be rounded to their
9377 semantic types (whereas @option{-ffloat-store} only affects
9378 assignments). This option is enabled by default for C if a strict
9379 conformance option such as @option{-std=c99} is used.
9380 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
9381 regardless of whether a strict conformance option is used.
9384 @option{-fexcess-precision=standard} is not implemented for languages
9385 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
9386 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9387 semantics apply without excess precision, and in the latter, rounding
9392 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9393 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9394 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
9395 @option{-fexcess-precision=fast}.
9397 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9399 This option is not turned on by any @option{-O} option besides
9400 @option{-Ofast} since it can result in incorrect output for programs
9401 that depend on an exact implementation of IEEE or ISO rules/specifications
9402 for math functions. It may, however, yield faster code for programs
9403 that do not require the guarantees of these specifications.
9405 @item -fno-math-errno
9406 @opindex fno-math-errno
9407 Do not set @code{errno} after calling math functions that are executed
9408 with a single instruction, e.g., @code{sqrt}. A program that relies on
9409 IEEE exceptions for math error handling may want to use this flag
9410 for speed while maintaining IEEE arithmetic compatibility.
9412 This option is not turned on by any @option{-O} option since
9413 it can result in incorrect output for programs that depend on
9414 an exact implementation of IEEE or ISO rules/specifications for
9415 math functions. It may, however, yield faster code for programs
9416 that do not require the guarantees of these specifications.
9418 The default is @option{-fmath-errno}.
9420 On Darwin systems, the math library never sets @code{errno}. There is
9421 therefore no reason for the compiler to consider the possibility that
9422 it might, and @option{-fno-math-errno} is the default.
9424 @item -funsafe-math-optimizations
9425 @opindex funsafe-math-optimizations
9427 Allow optimizations for floating-point arithmetic that (a) assume
9428 that arguments and results are valid and (b) may violate IEEE or
9429 ANSI standards. When used at link time, it may include libraries
9430 or startup files that change the default FPU control word or other
9431 similar optimizations.
9433 This option is not turned on by any @option{-O} option since
9434 it can result in incorrect output for programs that depend on
9435 an exact implementation of IEEE or ISO rules/specifications for
9436 math functions. It may, however, yield faster code for programs
9437 that do not require the guarantees of these specifications.
9438 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9439 @option{-fassociative-math} and @option{-freciprocal-math}.
9441 The default is @option{-fno-unsafe-math-optimizations}.
9443 @item -fassociative-math
9444 @opindex fassociative-math
9446 Allow re-association of operands in series of floating-point operations.
9447 This violates the ISO C and C++ language standard by possibly changing
9448 computation result. NOTE: re-ordering may change the sign of zero as
9449 well as ignore NaNs and inhibit or create underflow or overflow (and
9450 thus cannot be used on code that relies on rounding behavior like
9451 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9452 and thus may not be used when ordered comparisons are required.
9453 This option requires that both @option{-fno-signed-zeros} and
9454 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9455 much sense with @option{-frounding-math}. For Fortran the option
9456 is automatically enabled when both @option{-fno-signed-zeros} and
9457 @option{-fno-trapping-math} are in effect.
9459 The default is @option{-fno-associative-math}.
9461 @item -freciprocal-math
9462 @opindex freciprocal-math
9464 Allow the reciprocal of a value to be used instead of dividing by
9465 the value if this enables optimizations. For example @code{x / y}
9466 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9467 is subject to common subexpression elimination. Note that this loses
9468 precision and increases the number of flops operating on the value.
9470 The default is @option{-fno-reciprocal-math}.
9472 @item -ffinite-math-only
9473 @opindex ffinite-math-only
9474 Allow optimizations for floating-point arithmetic that assume
9475 that arguments and results are not NaNs or +-Infs.
9477 This option is not turned on by any @option{-O} option since
9478 it can result in incorrect output for programs that depend on
9479 an exact implementation of IEEE or ISO rules/specifications for
9480 math functions. It may, however, yield faster code for programs
9481 that do not require the guarantees of these specifications.
9483 The default is @option{-fno-finite-math-only}.
9485 @item -fno-signed-zeros
9486 @opindex fno-signed-zeros
9487 Allow optimizations for floating-point arithmetic that ignore the
9488 signedness of zero. IEEE arithmetic specifies the behavior of
9489 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9490 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9491 This option implies that the sign of a zero result isn't significant.
9493 The default is @option{-fsigned-zeros}.
9495 @item -fno-trapping-math
9496 @opindex fno-trapping-math
9497 Compile code assuming that floating-point operations cannot generate
9498 user-visible traps. These traps include division by zero, overflow,
9499 underflow, inexact result and invalid operation. This option requires
9500 that @option{-fno-signaling-nans} be in effect. Setting this option may
9501 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9503 This option should never be turned on by any @option{-O} option since
9504 it can result in incorrect output for programs that depend on
9505 an exact implementation of IEEE or ISO rules/specifications for
9508 The default is @option{-ftrapping-math}.
9510 @item -frounding-math
9511 @opindex frounding-math
9512 Disable transformations and optimizations that assume default floating-point
9513 rounding behavior. This is round-to-zero for all floating point
9514 to integer conversions, and round-to-nearest for all other arithmetic
9515 truncations. This option should be specified for programs that change
9516 the FP rounding mode dynamically, or that may be executed with a
9517 non-default rounding mode. This option disables constant folding of
9518 floating-point expressions at compile time (which may be affected by
9519 rounding mode) and arithmetic transformations that are unsafe in the
9520 presence of sign-dependent rounding modes.
9522 The default is @option{-fno-rounding-math}.
9524 This option is experimental and does not currently guarantee to
9525 disable all GCC optimizations that are affected by rounding mode.
9526 Future versions of GCC may provide finer control of this setting
9527 using C99's @code{FENV_ACCESS} pragma. This command-line option
9528 will be used to specify the default state for @code{FENV_ACCESS}.
9530 @item -fsignaling-nans
9531 @opindex fsignaling-nans
9532 Compile code assuming that IEEE signaling NaNs may generate user-visible
9533 traps during floating-point operations. Setting this option disables
9534 optimizations that may change the number of exceptions visible with
9535 signaling NaNs. This option implies @option{-ftrapping-math}.
9537 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9540 The default is @option{-fno-signaling-nans}.
9542 This option is experimental and does not currently guarantee to
9543 disable all GCC optimizations that affect signaling NaN behavior.
9545 @item -fno-fp-int-builtin-inexact
9546 @opindex fno-fp-int-builtin-inexact
9547 Do not allow the built-in functions @code{ceil}, @code{floor},
9548 @code{round} and @code{trunc}, and their @code{float} and @code{long
9549 double} variants, to generate code that raises the ``inexact''
9550 floating-point exception for noninteger arguments. ISO C99 and C11
9551 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9552 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9555 The default is @option{-ffp-int-builtin-inexact}, allowing the
9556 exception to be raised. This option does nothing unless
9557 @option{-ftrapping-math} is in effect.
9559 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9560 generate a call to a library function then the ``inexact'' exception
9561 may be raised if the library implementation does not follow TS 18661.
9563 @item -fsingle-precision-constant
9564 @opindex fsingle-precision-constant
9565 Treat floating-point constants as single precision instead of
9566 implicitly converting them to double-precision constants.
9568 @item -fcx-limited-range
9569 @opindex fcx-limited-range
9570 When enabled, this option states that a range reduction step is not
9571 needed when performing complex division. Also, there is no checking
9572 whether the result of a complex multiplication or division is @code{NaN
9573 + I*NaN}, with an attempt to rescue the situation in that case. The
9574 default is @option{-fno-cx-limited-range}, but is enabled by
9575 @option{-ffast-math}.
9577 This option controls the default setting of the ISO C99
9578 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9581 @item -fcx-fortran-rules
9582 @opindex fcx-fortran-rules
9583 Complex multiplication and division follow Fortran rules. Range
9584 reduction is done as part of complex division, but there is no checking
9585 whether the result of a complex multiplication or division is @code{NaN
9586 + I*NaN}, with an attempt to rescue the situation in that case.
9588 The default is @option{-fno-cx-fortran-rules}.
9592 The following options control optimizations that may improve
9593 performance, but are not enabled by any @option{-O} options. This
9594 section includes experimental options that may produce broken code.
9597 @item -fbranch-probabilities
9598 @opindex fbranch-probabilities
9599 After running a program compiled with @option{-fprofile-arcs}
9600 (@pxref{Instrumentation Options}),
9601 you can compile it a second time using
9602 @option{-fbranch-probabilities}, to improve optimizations based on
9603 the number of times each branch was taken. When a program
9604 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9605 counts to a file called @file{@var{sourcename}.gcda} for each source
9606 file. The information in this data file is very dependent on the
9607 structure of the generated code, so you must use the same source code
9608 and the same optimization options for both compilations.
9610 With @option{-fbranch-probabilities}, GCC puts a
9611 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9612 These can be used to improve optimization. Currently, they are only
9613 used in one place: in @file{reorg.c}, instead of guessing which path a
9614 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9615 exactly determine which path is taken more often.
9617 @item -fprofile-values
9618 @opindex fprofile-values
9619 If combined with @option{-fprofile-arcs}, it adds code so that some
9620 data about values of expressions in the program is gathered.
9622 With @option{-fbranch-probabilities}, it reads back the data gathered
9623 from profiling values of expressions for usage in optimizations.
9625 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9627 @item -fprofile-reorder-functions
9628 @opindex fprofile-reorder-functions
9629 Function reordering based on profile instrumentation collects
9630 first time of execution of a function and orders these functions
9633 Enabled with @option{-fprofile-use}.
9637 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9638 to add code to gather information about values of expressions.
9640 With @option{-fbranch-probabilities}, it reads back the data gathered
9641 and actually performs the optimizations based on them.
9642 Currently the optimizations include specialization of division operations
9643 using the knowledge about the value of the denominator.
9645 @item -frename-registers
9646 @opindex frename-registers
9647 Attempt to avoid false dependencies in scheduled code by making use
9648 of registers left over after register allocation. This optimization
9649 most benefits processors with lots of registers. Depending on the
9650 debug information format adopted by the target, however, it can
9651 make debugging impossible, since variables no longer stay in
9652 a ``home register''.
9654 Enabled by default with @option{-funroll-loops}.
9656 @item -fschedule-fusion
9657 @opindex fschedule-fusion
9658 Performs a target dependent pass over the instruction stream to schedule
9659 instructions of same type together because target machine can execute them
9660 more efficiently if they are adjacent to each other in the instruction flow.
9662 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9666 Perform tail duplication to enlarge superblock size. This transformation
9667 simplifies the control flow of the function allowing other optimizations to do
9670 Enabled with @option{-fprofile-use}.
9672 @item -funroll-loops
9673 @opindex funroll-loops
9674 Unroll loops whose number of iterations can be determined at compile time or
9675 upon entry to the loop. @option{-funroll-loops} implies
9676 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9677 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9678 a small constant number of iterations). This option makes code larger, and may
9679 or may not make it run faster.
9681 Enabled with @option{-fprofile-use}.
9683 @item -funroll-all-loops
9684 @opindex funroll-all-loops
9685 Unroll all loops, even if their number of iterations is uncertain when
9686 the loop is entered. This usually makes programs run more slowly.
9687 @option{-funroll-all-loops} implies the same options as
9688 @option{-funroll-loops}.
9691 @opindex fpeel-loops
9692 Peels loops for which there is enough information that they do not
9693 roll much (from profile feedback or static analysis). It also turns on
9694 complete loop peeling (i.e.@: complete removal of loops with small constant
9695 number of iterations).
9697 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9699 @item -fmove-loop-invariants
9700 @opindex fmove-loop-invariants
9701 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9702 at level @option{-O1}
9705 @opindex fsplit-loops
9706 Split a loop into two if it contains a condition that's always true
9707 for one side of the iteration space and false for the other.
9709 @item -funswitch-loops
9710 @opindex funswitch-loops
9711 Move branches with loop invariant conditions out of the loop, with duplicates
9712 of the loop on both branches (modified according to result of the condition).
9714 @item -ffunction-sections
9715 @itemx -fdata-sections
9716 @opindex ffunction-sections
9717 @opindex fdata-sections
9718 Place each function or data item into its own section in the output
9719 file if the target supports arbitrary sections. The name of the
9720 function or the name of the data item determines the section's name
9723 Use these options on systems where the linker can perform optimizations to
9724 improve locality of reference in the instruction space. Most systems using the
9725 ELF object format have linkers with such optimizations. On AIX, the linker
9726 rearranges sections (CSECTs) based on the call graph. The performance impact
9729 Together with a linker garbage collection (linker @option{--gc-sections}
9730 option) these options may lead to smaller statically-linked executables (after
9733 On ELF/DWARF systems these options do not degenerate the quality of the debug
9734 information. There could be issues with other object files/debug info formats.
9736 Only use these options when there are significant benefits from doing so. When
9737 you specify these options, the assembler and linker create larger object and
9738 executable files and are also slower. These options affect code generation.
9739 They prevent optimizations by the compiler and assembler using relative
9740 locations inside a translation unit since the locations are unknown until
9741 link time. An example of such an optimization is relaxing calls to short call
9744 @item -fbranch-target-load-optimize
9745 @opindex fbranch-target-load-optimize
9746 Perform branch target register load optimization before prologue / epilogue
9748 The use of target registers can typically be exposed only during reload,
9749 thus hoisting loads out of loops and doing inter-block scheduling needs
9750 a separate optimization pass.
9752 @item -fbranch-target-load-optimize2
9753 @opindex fbranch-target-load-optimize2
9754 Perform branch target register load optimization after prologue / epilogue
9757 @item -fbtr-bb-exclusive
9758 @opindex fbtr-bb-exclusive
9759 When performing branch target register load optimization, don't reuse
9760 branch target registers within any basic block.
9763 @opindex fstdarg-opt
9764 Optimize the prologue of variadic argument functions with respect to usage of
9767 @item -fsection-anchors
9768 @opindex fsection-anchors
9769 Try to reduce the number of symbolic address calculations by using
9770 shared ``anchor'' symbols to address nearby objects. This transformation
9771 can help to reduce the number of GOT entries and GOT accesses on some
9774 For example, the implementation of the following function @code{foo}:
9778 int foo (void) @{ return a + b + c; @}
9782 usually calculates the addresses of all three variables, but if you
9783 compile it with @option{-fsection-anchors}, it accesses the variables
9784 from a common anchor point instead. The effect is similar to the
9785 following pseudocode (which isn't valid C):
9790 register int *xr = &x;
9791 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9795 Not all targets support this option.
9797 @item --param @var{name}=@var{value}
9799 In some places, GCC uses various constants to control the amount of
9800 optimization that is done. For example, GCC does not inline functions
9801 that contain more than a certain number of instructions. You can
9802 control some of these constants on the command line using the
9803 @option{--param} option.
9805 The names of specific parameters, and the meaning of the values, are
9806 tied to the internals of the compiler, and are subject to change
9807 without notice in future releases.
9809 In each case, the @var{value} is an integer. The allowable choices for
9813 @item predictable-branch-outcome
9814 When branch is predicted to be taken with probability lower than this threshold
9815 (in percent), then it is considered well predictable. The default is 10.
9817 @item max-rtl-if-conversion-insns
9818 RTL if-conversion tries to remove conditional branches around a block and
9819 replace them with conditionally executed instructions. This parameter
9820 gives the maximum number of instructions in a block which should be
9821 considered for if-conversion. The default is 10, though the compiler will
9822 also use other heuristics to decide whether if-conversion is likely to be
9825 @item max-rtl-if-conversion-predictable-cost
9826 @item max-rtl-if-conversion-unpredictable-cost
9827 RTL if-conversion will try to remove conditional branches around a block
9828 and replace them with conditionally executed instructions. These parameters
9829 give the maximum permissible cost for the sequence that would be generated
9830 by if-conversion depending on whether the branch is statically determined
9831 to be predictable or not. The units for this parameter are the same as
9832 those for the GCC internal seq_cost metric. The compiler will try to
9833 provide a reasonable default for this parameter using the BRANCH_COST
9836 @item max-crossjump-edges
9837 The maximum number of incoming edges to consider for cross-jumping.
9838 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9839 the number of edges incoming to each block. Increasing values mean
9840 more aggressive optimization, making the compilation time increase with
9841 probably small improvement in executable size.
9843 @item min-crossjump-insns
9844 The minimum number of instructions that must be matched at the end
9845 of two blocks before cross-jumping is performed on them. This
9846 value is ignored in the case where all instructions in the block being
9847 cross-jumped from are matched. The default value is 5.
9849 @item max-grow-copy-bb-insns
9850 The maximum code size expansion factor when copying basic blocks
9851 instead of jumping. The expansion is relative to a jump instruction.
9852 The default value is 8.
9854 @item max-goto-duplication-insns
9855 The maximum number of instructions to duplicate to a block that jumps
9856 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9857 passes, GCC factors computed gotos early in the compilation process,
9858 and unfactors them as late as possible. Only computed jumps at the
9859 end of a basic blocks with no more than max-goto-duplication-insns are
9860 unfactored. The default value is 8.
9862 @item max-delay-slot-insn-search
9863 The maximum number of instructions to consider when looking for an
9864 instruction to fill a delay slot. If more than this arbitrary number of
9865 instructions are searched, the time savings from filling the delay slot
9866 are minimal, so stop searching. Increasing values mean more
9867 aggressive optimization, making the compilation time increase with probably
9868 small improvement in execution time.
9870 @item max-delay-slot-live-search
9871 When trying to fill delay slots, the maximum number of instructions to
9872 consider when searching for a block with valid live register
9873 information. Increasing this arbitrarily chosen value means more
9874 aggressive optimization, increasing the compilation time. This parameter
9875 should be removed when the delay slot code is rewritten to maintain the
9878 @item max-gcse-memory
9879 The approximate maximum amount of memory that can be allocated in
9880 order to perform the global common subexpression elimination
9881 optimization. If more memory than specified is required, the
9882 optimization is not done.
9884 @item max-gcse-insertion-ratio
9885 If the ratio of expression insertions to deletions is larger than this value
9886 for any expression, then RTL PRE inserts or removes the expression and thus
9887 leaves partially redundant computations in the instruction stream. The default value is 20.
9889 @item max-pending-list-length
9890 The maximum number of pending dependencies scheduling allows
9891 before flushing the current state and starting over. Large functions
9892 with few branches or calls can create excessively large lists which
9893 needlessly consume memory and resources.
9895 @item max-modulo-backtrack-attempts
9896 The maximum number of backtrack attempts the scheduler should make
9897 when modulo scheduling a loop. Larger values can exponentially increase
9900 @item max-inline-insns-single
9901 Several parameters control the tree inliner used in GCC@.
9902 This number sets the maximum number of instructions (counted in GCC's
9903 internal representation) in a single function that the tree inliner
9904 considers for inlining. This only affects functions declared
9905 inline and methods implemented in a class declaration (C++).
9906 The default value is 400.
9908 @item max-inline-insns-auto
9909 When you use @option{-finline-functions} (included in @option{-O3}),
9910 a lot of functions that would otherwise not be considered for inlining
9911 by the compiler are investigated. To those functions, a different
9912 (more restrictive) limit compared to functions declared inline can
9914 The default value is 40.
9916 @item inline-min-speedup
9917 When estimated performance improvement of caller + callee runtime exceeds this
9918 threshold (in percent), the function can be inlined regardless of the limit on
9919 @option{--param max-inline-insns-single} and @option{--param
9920 max-inline-insns-auto}.
9922 @item large-function-insns
9923 The limit specifying really large functions. For functions larger than this
9924 limit after inlining, inlining is constrained by
9925 @option{--param large-function-growth}. This parameter is useful primarily
9926 to avoid extreme compilation time caused by non-linear algorithms used by the
9928 The default value is 2700.
9930 @item large-function-growth
9931 Specifies maximal growth of large function caused by inlining in percents.
9932 The default value is 100 which limits large function growth to 2.0 times
9935 @item large-unit-insns
9936 The limit specifying large translation unit. Growth caused by inlining of
9937 units larger than this limit is limited by @option{--param inline-unit-growth}.
9938 For small units this might be too tight.
9939 For example, consider a unit consisting of function A
9940 that is inline and B that just calls A three times. If B is small relative to
9941 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9942 large units consisting of small inlineable functions, however, the overall unit
9943 growth limit is needed to avoid exponential explosion of code size. Thus for
9944 smaller units, the size is increased to @option{--param large-unit-insns}
9945 before applying @option{--param inline-unit-growth}. The default is 10000.
9947 @item inline-unit-growth
9948 Specifies maximal overall growth of the compilation unit caused by inlining.
9949 The default value is 20 which limits unit growth to 1.2 times the original
9950 size. Cold functions (either marked cold via an attribute or by profile
9951 feedback) are not accounted into the unit size.
9953 @item ipcp-unit-growth
9954 Specifies maximal overall growth of the compilation unit caused by
9955 interprocedural constant propagation. The default value is 10 which limits
9956 unit growth to 1.1 times the original size.
9958 @item large-stack-frame
9959 The limit specifying large stack frames. While inlining the algorithm is trying
9960 to not grow past this limit too much. The default value is 256 bytes.
9962 @item large-stack-frame-growth
9963 Specifies maximal growth of large stack frames caused by inlining in percents.
9964 The default value is 1000 which limits large stack frame growth to 11 times
9967 @item max-inline-insns-recursive
9968 @itemx max-inline-insns-recursive-auto
9969 Specifies the maximum number of instructions an out-of-line copy of a
9970 self-recursive inline
9971 function can grow into by performing recursive inlining.
9973 @option{--param max-inline-insns-recursive} applies to functions
9975 For functions not declared inline, recursive inlining
9976 happens only when @option{-finline-functions} (included in @option{-O3}) is
9977 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
9978 default value is 450.
9980 @item max-inline-recursive-depth
9981 @itemx max-inline-recursive-depth-auto
9982 Specifies the maximum recursion depth used for recursive inlining.
9984 @option{--param max-inline-recursive-depth} applies to functions
9985 declared inline. For functions not declared inline, recursive inlining
9986 happens only when @option{-finline-functions} (included in @option{-O3}) is
9987 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9990 @item min-inline-recursive-probability
9991 Recursive inlining is profitable only for function having deep recursion
9992 in average and can hurt for function having little recursion depth by
9993 increasing the prologue size or complexity of function body to other
9996 When profile feedback is available (see @option{-fprofile-generate}) the actual
9997 recursion depth can be guessed from the probability that function recurses
9998 via a given call expression. This parameter limits inlining only to call
9999 expressions whose probability exceeds the given threshold (in percents).
10000 The default value is 10.
10002 @item early-inlining-insns
10003 Specify growth that the early inliner can make. In effect it increases
10004 the amount of inlining for code having a large abstraction penalty.
10005 The default value is 14.
10007 @item max-early-inliner-iterations
10008 Limit of iterations of the early inliner. This basically bounds
10009 the number of nested indirect calls the early inliner can resolve.
10010 Deeper chains are still handled by late inlining.
10012 @item comdat-sharing-probability
10013 Probability (in percent) that C++ inline function with comdat visibility
10014 are shared across multiple compilation units. The default value is 20.
10016 @item profile-func-internal-id
10017 A parameter to control whether to use function internal id in profile
10018 database lookup. If the value is 0, the compiler uses an id that
10019 is based on function assembler name and filename, which makes old profile
10020 data more tolerant to source changes such as function reordering etc.
10021 The default value is 0.
10023 @item min-vect-loop-bound
10024 The minimum number of iterations under which loops are not vectorized
10025 when @option{-ftree-vectorize} is used. The number of iterations after
10026 vectorization needs to be greater than the value specified by this option
10027 to allow vectorization. The default value is 0.
10029 @item gcse-cost-distance-ratio
10030 Scaling factor in calculation of maximum distance an expression
10031 can be moved by GCSE optimizations. This is currently supported only in the
10032 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10033 is with simple expressions, i.e., the expressions that have cost
10034 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10035 hoisting of simple expressions. The default value is 10.
10037 @item gcse-unrestricted-cost
10038 Cost, roughly measured as the cost of a single typical machine
10039 instruction, at which GCSE optimizations do not constrain
10040 the distance an expression can travel. This is currently
10041 supported only in the code hoisting pass. The lesser the cost,
10042 the more aggressive code hoisting is. Specifying 0
10043 allows all expressions to travel unrestricted distances.
10044 The default value is 3.
10046 @item max-hoist-depth
10047 The depth of search in the dominator tree for expressions to hoist.
10048 This is used to avoid quadratic behavior in hoisting algorithm.
10049 The value of 0 does not limit on the search, but may slow down compilation
10050 of huge functions. The default value is 30.
10052 @item max-tail-merge-comparisons
10053 The maximum amount of similar bbs to compare a bb with. This is used to
10054 avoid quadratic behavior in tree tail merging. The default value is 10.
10056 @item max-tail-merge-iterations
10057 The maximum amount of iterations of the pass over the function. This is used to
10058 limit compilation time in tree tail merging. The default value is 2.
10060 @item store-merging-allow-unaligned
10061 Allow the store merging pass to introduce unaligned stores if it is legal to
10062 do so. The default value is 1.
10064 @item max-stores-to-merge
10065 The maximum number of stores to attempt to merge into wider stores in the store
10066 merging pass. The minimum value is 2 and the default is 64.
10068 @item max-unrolled-insns
10069 The maximum number of instructions that a loop may have to be unrolled.
10070 If a loop is unrolled, this parameter also determines how many times
10071 the loop code is unrolled.
10073 @item max-average-unrolled-insns
10074 The maximum number of instructions biased by probabilities of their execution
10075 that a loop may have to be unrolled. If a loop is unrolled,
10076 this parameter also determines how many times the loop code is unrolled.
10078 @item max-unroll-times
10079 The maximum number of unrollings of a single loop.
10081 @item max-peeled-insns
10082 The maximum number of instructions that a loop may have to be peeled.
10083 If a loop is peeled, this parameter also determines how many times
10084 the loop code is peeled.
10086 @item max-peel-times
10087 The maximum number of peelings of a single loop.
10089 @item max-peel-branches
10090 The maximum number of branches on the hot path through the peeled sequence.
10092 @item max-completely-peeled-insns
10093 The maximum number of insns of a completely peeled loop.
10095 @item max-completely-peel-times
10096 The maximum number of iterations of a loop to be suitable for complete peeling.
10098 @item max-completely-peel-loop-nest-depth
10099 The maximum depth of a loop nest suitable for complete peeling.
10101 @item max-unswitch-insns
10102 The maximum number of insns of an unswitched loop.
10104 @item max-unswitch-level
10105 The maximum number of branches unswitched in a single loop.
10107 @item max-loop-headers-insns
10108 The maximum number of insns in loop header duplicated by the copy loop headers
10111 @item lim-expensive
10112 The minimum cost of an expensive expression in the loop invariant motion.
10114 @item iv-consider-all-candidates-bound
10115 Bound on number of candidates for induction variables, below which
10116 all candidates are considered for each use in induction variable
10117 optimizations. If there are more candidates than this,
10118 only the most relevant ones are considered to avoid quadratic time complexity.
10120 @item iv-max-considered-uses
10121 The induction variable optimizations give up on loops that contain more
10122 induction variable uses.
10124 @item iv-always-prune-cand-set-bound
10125 If the number of candidates in the set is smaller than this value,
10126 always try to remove unnecessary ivs from the set
10127 when adding a new one.
10129 @item avg-loop-niter
10130 Average number of iterations of a loop.
10132 @item dse-max-object-size
10133 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
10134 Larger values may result in larger compilation times.
10136 @item scev-max-expr-size
10137 Bound on size of expressions used in the scalar evolutions analyzer.
10138 Large expressions slow the analyzer.
10140 @item scev-max-expr-complexity
10141 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10142 Complex expressions slow the analyzer.
10144 @item max-tree-if-conversion-phi-args
10145 Maximum number of arguments in a PHI supported by TREE if conversion
10146 unless the loop is marked with simd pragma.
10148 @item vect-max-version-for-alignment-checks
10149 The maximum number of run-time checks that can be performed when
10150 doing loop versioning for alignment in the vectorizer.
10152 @item vect-max-version-for-alias-checks
10153 The maximum number of run-time checks that can be performed when
10154 doing loop versioning for alias in the vectorizer.
10156 @item vect-max-peeling-for-alignment
10157 The maximum number of loop peels to enhance access alignment
10158 for vectorizer. Value -1 means no limit.
10160 @item max-iterations-to-track
10161 The maximum number of iterations of a loop the brute-force algorithm
10162 for analysis of the number of iterations of the loop tries to evaluate.
10164 @item hot-bb-count-ws-permille
10165 A basic block profile count is considered hot if it contributes to
10166 the given permillage (i.e. 0...1000) of the entire profiled execution.
10168 @item hot-bb-frequency-fraction
10169 Select fraction of the entry block frequency of executions of basic block in
10170 function given basic block needs to have to be considered hot.
10172 @item max-predicted-iterations
10173 The maximum number of loop iterations we predict statically. This is useful
10174 in cases where a function contains a single loop with known bound and
10175 another loop with unknown bound.
10176 The known number of iterations is predicted correctly, while
10177 the unknown number of iterations average to roughly 10. This means that the
10178 loop without bounds appears artificially cold relative to the other one.
10180 @item builtin-expect-probability
10181 Control the probability of the expression having the specified value. This
10182 parameter takes a percentage (i.e. 0 ... 100) as input.
10183 The default probability of 90 is obtained empirically.
10185 @item align-threshold
10187 Select fraction of the maximal frequency of executions of a basic block in
10188 a function to align the basic block.
10190 @item align-loop-iterations
10192 A loop expected to iterate at least the selected number of iterations is
10195 @item tracer-dynamic-coverage
10196 @itemx tracer-dynamic-coverage-feedback
10198 This value is used to limit superblock formation once the given percentage of
10199 executed instructions is covered. This limits unnecessary code size
10202 The @option{tracer-dynamic-coverage-feedback} parameter
10203 is used only when profile
10204 feedback is available. The real profiles (as opposed to statically estimated
10205 ones) are much less balanced allowing the threshold to be larger value.
10207 @item tracer-max-code-growth
10208 Stop tail duplication once code growth has reached given percentage. This is
10209 a rather artificial limit, as most of the duplicates are eliminated later in
10210 cross jumping, so it may be set to much higher values than is the desired code
10213 @item tracer-min-branch-ratio
10215 Stop reverse growth when the reverse probability of best edge is less than this
10216 threshold (in percent).
10218 @item tracer-min-branch-probability
10219 @itemx tracer-min-branch-probability-feedback
10221 Stop forward growth if the best edge has probability lower than this
10224 Similarly to @option{tracer-dynamic-coverage} two parameters are
10225 provided. @option{tracer-min-branch-probability-feedback} is used for
10226 compilation with profile feedback and @option{tracer-min-branch-probability}
10227 compilation without. The value for compilation with profile feedback
10228 needs to be more conservative (higher) in order to make tracer
10231 @item stack-clash-protection-guard-size
10232 Specify the size of the operating system provided stack guard as
10233 2 raised to @var{num} bytes. The default value is 12 (4096 bytes).
10234 Acceptable values are between 12 and 30. Higher values may reduce the
10235 number of explicit probes, but a value larger than the operating system
10236 provided guard will leave code vulnerable to stack clash style attacks.
10238 @item stack-clash-protection-probe-interval
10239 Stack clash protection involves probing stack space as it is allocated. This
10240 param controls the maximum distance between probes into the stack as 2 raised
10241 to @var{num} bytes. Acceptable values are between 10 and 16 and defaults to
10242 12. Higher values may reduce the number of explicit probes, but a value
10243 larger than the operating system provided guard will leave code vulnerable to
10244 stack clash style attacks.
10246 @item max-cse-path-length
10248 The maximum number of basic blocks on path that CSE considers.
10251 @item max-cse-insns
10252 The maximum number of instructions CSE processes before flushing.
10253 The default is 1000.
10255 @item ggc-min-expand
10257 GCC uses a garbage collector to manage its own memory allocation. This
10258 parameter specifies the minimum percentage by which the garbage
10259 collector's heap should be allowed to expand between collections.
10260 Tuning this may improve compilation speed; it has no effect on code
10263 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10264 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10265 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10266 GCC is not able to calculate RAM on a particular platform, the lower
10267 bound of 30% is used. Setting this parameter and
10268 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10269 every opportunity. This is extremely slow, but can be useful for
10272 @item ggc-min-heapsize
10274 Minimum size of the garbage collector's heap before it begins bothering
10275 to collect garbage. The first collection occurs after the heap expands
10276 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10277 tuning this may improve compilation speed, and has no effect on code
10280 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10281 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10282 with a lower bound of 4096 (four megabytes) and an upper bound of
10283 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10284 particular platform, the lower bound is used. Setting this parameter
10285 very large effectively disables garbage collection. Setting this
10286 parameter and @option{ggc-min-expand} to zero causes a full collection
10287 to occur at every opportunity.
10289 @item max-reload-search-insns
10290 The maximum number of instruction reload should look backward for equivalent
10291 register. Increasing values mean more aggressive optimization, making the
10292 compilation time increase with probably slightly better performance.
10293 The default value is 100.
10295 @item max-cselib-memory-locations
10296 The maximum number of memory locations cselib should take into account.
10297 Increasing values mean more aggressive optimization, making the compilation time
10298 increase with probably slightly better performance. The default value is 500.
10300 @item max-sched-ready-insns
10301 The maximum number of instructions ready to be issued the scheduler should
10302 consider at any given time during the first scheduling pass. Increasing
10303 values mean more thorough searches, making the compilation time increase
10304 with probably little benefit. The default value is 100.
10306 @item max-sched-region-blocks
10307 The maximum number of blocks in a region to be considered for
10308 interblock scheduling. The default value is 10.
10310 @item max-pipeline-region-blocks
10311 The maximum number of blocks in a region to be considered for
10312 pipelining in the selective scheduler. The default value is 15.
10314 @item max-sched-region-insns
10315 The maximum number of insns in a region to be considered for
10316 interblock scheduling. The default value is 100.
10318 @item max-pipeline-region-insns
10319 The maximum number of insns in a region to be considered for
10320 pipelining in the selective scheduler. The default value is 200.
10322 @item min-spec-prob
10323 The minimum probability (in percents) of reaching a source block
10324 for interblock speculative scheduling. The default value is 40.
10326 @item max-sched-extend-regions-iters
10327 The maximum number of iterations through CFG to extend regions.
10328 A value of 0 (the default) disables region extensions.
10330 @item max-sched-insn-conflict-delay
10331 The maximum conflict delay for an insn to be considered for speculative motion.
10332 The default value is 3.
10334 @item sched-spec-prob-cutoff
10335 The minimal probability of speculation success (in percents), so that
10336 speculative insns are scheduled.
10337 The default value is 40.
10339 @item sched-state-edge-prob-cutoff
10340 The minimum probability an edge must have for the scheduler to save its
10342 The default value is 10.
10344 @item sched-mem-true-dep-cost
10345 Minimal distance (in CPU cycles) between store and load targeting same
10346 memory locations. The default value is 1.
10348 @item selsched-max-lookahead
10349 The maximum size of the lookahead window of selective scheduling. It is a
10350 depth of search for available instructions.
10351 The default value is 50.
10353 @item selsched-max-sched-times
10354 The maximum number of times that an instruction is scheduled during
10355 selective scheduling. This is the limit on the number of iterations
10356 through which the instruction may be pipelined. The default value is 2.
10358 @item selsched-insns-to-rename
10359 The maximum number of best instructions in the ready list that are considered
10360 for renaming in the selective scheduler. The default value is 2.
10363 The minimum value of stage count that swing modulo scheduler
10364 generates. The default value is 2.
10366 @item max-last-value-rtl
10367 The maximum size measured as number of RTLs that can be recorded in an expression
10368 in combiner for a pseudo register as last known value of that register. The default
10371 @item max-combine-insns
10372 The maximum number of instructions the RTL combiner tries to combine.
10373 The default value is 2 at @option{-Og} and 4 otherwise.
10375 @item integer-share-limit
10376 Small integer constants can use a shared data structure, reducing the
10377 compiler's memory usage and increasing its speed. This sets the maximum
10378 value of a shared integer constant. The default value is 256.
10380 @item ssp-buffer-size
10381 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10382 protection when @option{-fstack-protection} is used.
10384 @item min-size-for-stack-sharing
10385 The minimum size of variables taking part in stack slot sharing when not
10386 optimizing. The default value is 32.
10388 @item max-jump-thread-duplication-stmts
10389 Maximum number of statements allowed in a block that needs to be
10390 duplicated when threading jumps.
10392 @item max-fields-for-field-sensitive
10393 Maximum number of fields in a structure treated in
10394 a field sensitive manner during pointer analysis. The default is zero
10395 for @option{-O0} and @option{-O1},
10396 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10398 @item prefetch-latency
10399 Estimate on average number of instructions that are executed before
10400 prefetch finishes. The distance prefetched ahead is proportional
10401 to this constant. Increasing this number may also lead to less
10402 streams being prefetched (see @option{simultaneous-prefetches}).
10404 @item simultaneous-prefetches
10405 Maximum number of prefetches that can run at the same time.
10407 @item l1-cache-line-size
10408 The size of cache line in L1 cache, in bytes.
10410 @item l1-cache-size
10411 The size of L1 cache, in kilobytes.
10413 @item l2-cache-size
10414 The size of L2 cache, in kilobytes.
10416 @item min-insn-to-prefetch-ratio
10417 The minimum ratio between the number of instructions and the
10418 number of prefetches to enable prefetching in a loop.
10420 @item prefetch-min-insn-to-mem-ratio
10421 The minimum ratio between the number of instructions and the
10422 number of memory references to enable prefetching in a loop.
10424 @item use-canonical-types
10425 Whether the compiler should use the ``canonical'' type system. By
10426 default, this should always be 1, which uses a more efficient internal
10427 mechanism for comparing types in C++ and Objective-C++. However, if
10428 bugs in the canonical type system are causing compilation failures,
10429 set this value to 0 to disable canonical types.
10431 @item switch-conversion-max-branch-ratio
10432 Switch initialization conversion refuses to create arrays that are
10433 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10434 branches in the switch.
10436 @item max-partial-antic-length
10437 Maximum length of the partial antic set computed during the tree
10438 partial redundancy elimination optimization (@option{-ftree-pre}) when
10439 optimizing at @option{-O3} and above. For some sorts of source code
10440 the enhanced partial redundancy elimination optimization can run away,
10441 consuming all of the memory available on the host machine. This
10442 parameter sets a limit on the length of the sets that are computed,
10443 which prevents the runaway behavior. Setting a value of 0 for
10444 this parameter allows an unlimited set length.
10446 @item sccvn-max-scc-size
10447 Maximum size of a strongly connected component (SCC) during SCCVN
10448 processing. If this limit is hit, SCCVN processing for the whole
10449 function is not done and optimizations depending on it are
10450 disabled. The default maximum SCC size is 10000.
10452 @item sccvn-max-alias-queries-per-access
10453 Maximum number of alias-oracle queries we perform when looking for
10454 redundancies for loads and stores. If this limit is hit the search
10455 is aborted and the load or store is not considered redundant. The
10456 number of queries is algorithmically limited to the number of
10457 stores on all paths from the load to the function entry.
10458 The default maximum number of queries is 1000.
10460 @item ira-max-loops-num
10461 IRA uses regional register allocation by default. If a function
10462 contains more loops than the number given by this parameter, only at most
10463 the given number of the most frequently-executed loops form regions
10464 for regional register allocation. The default value of the
10467 @item ira-max-conflict-table-size
10468 Although IRA uses a sophisticated algorithm to compress the conflict
10469 table, the table can still require excessive amounts of memory for
10470 huge functions. If the conflict table for a function could be more
10471 than the size in MB given by this parameter, the register allocator
10472 instead uses a faster, simpler, and lower-quality
10473 algorithm that does not require building a pseudo-register conflict table.
10474 The default value of the parameter is 2000.
10476 @item ira-loop-reserved-regs
10477 IRA can be used to evaluate more accurate register pressure in loops
10478 for decisions to move loop invariants (see @option{-O3}). The number
10479 of available registers reserved for some other purposes is given
10480 by this parameter. The default value of the parameter is 2, which is
10481 the minimal number of registers needed by typical instructions.
10482 This value is the best found from numerous experiments.
10484 @item lra-inheritance-ebb-probability-cutoff
10485 LRA tries to reuse values reloaded in registers in subsequent insns.
10486 This optimization is called inheritance. EBB is used as a region to
10487 do this optimization. The parameter defines a minimal fall-through
10488 edge probability in percentage used to add BB to inheritance EBB in
10489 LRA. The default value of the parameter is 40. The value was chosen
10490 from numerous runs of SPEC2000 on x86-64.
10492 @item loop-invariant-max-bbs-in-loop
10493 Loop invariant motion can be very expensive, both in compilation time and
10494 in amount of needed compile-time memory, with very large loops. Loops
10495 with more basic blocks than this parameter won't have loop invariant
10496 motion optimization performed on them. The default value of the
10497 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10499 @item loop-max-datarefs-for-datadeps
10500 Building data dependencies is expensive for very large loops. This
10501 parameter limits the number of data references in loops that are
10502 considered for data dependence analysis. These large loops are no
10503 handled by the optimizations using loop data dependencies.
10504 The default value is 1000.
10506 @item max-vartrack-size
10507 Sets a maximum number of hash table slots to use during variable
10508 tracking dataflow analysis of any function. If this limit is exceeded
10509 with variable tracking at assignments enabled, analysis for that
10510 function is retried without it, after removing all debug insns from
10511 the function. If the limit is exceeded even without debug insns, var
10512 tracking analysis is completely disabled for the function. Setting
10513 the parameter to zero makes it unlimited.
10515 @item max-vartrack-expr-depth
10516 Sets a maximum number of recursion levels when attempting to map
10517 variable names or debug temporaries to value expressions. This trades
10518 compilation time for more complete debug information. If this is set too
10519 low, value expressions that are available and could be represented in
10520 debug information may end up not being used; setting this higher may
10521 enable the compiler to find more complex debug expressions, but compile
10522 time and memory use may grow. The default is 12.
10524 @item min-nondebug-insn-uid
10525 Use uids starting at this parameter for nondebug insns. The range below
10526 the parameter is reserved exclusively for debug insns created by
10527 @option{-fvar-tracking-assignments}, but debug insns may get
10528 (non-overlapping) uids above it if the reserved range is exhausted.
10530 @item ipa-sra-ptr-growth-factor
10531 IPA-SRA replaces a pointer to an aggregate with one or more new
10532 parameters only when their cumulative size is less or equal to
10533 @option{ipa-sra-ptr-growth-factor} times the size of the original
10536 @item sra-max-scalarization-size-Ospeed
10537 @item sra-max-scalarization-size-Osize
10538 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10539 replace scalar parts of aggregates with uses of independent scalar
10540 variables. These parameters control the maximum size, in storage units,
10541 of aggregate which is considered for replacement when compiling for
10543 (@option{sra-max-scalarization-size-Ospeed}) or size
10544 (@option{sra-max-scalarization-size-Osize}) respectively.
10546 @item tm-max-aggregate-size
10547 When making copies of thread-local variables in a transaction, this
10548 parameter specifies the size in bytes after which variables are
10549 saved with the logging functions as opposed to save/restore code
10550 sequence pairs. This option only applies when using
10553 @item graphite-max-nb-scop-params
10554 To avoid exponential effects in the Graphite loop transforms, the
10555 number of parameters in a Static Control Part (SCoP) is bounded. The
10556 default value is 10 parameters, a value of zero can be used to lift
10557 the bound. A variable whose value is unknown at compilation time and
10558 defined outside a SCoP is a parameter of the SCoP.
10560 @item loop-block-tile-size
10561 Loop blocking or strip mining transforms, enabled with
10562 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10563 loop in the loop nest by a given number of iterations. The strip
10564 length can be changed using the @option{loop-block-tile-size}
10565 parameter. The default value is 51 iterations.
10567 @item loop-unroll-jam-size
10568 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10569 default value is 4.
10571 @item loop-unroll-jam-depth
10572 Specify the dimension to be unrolled (counting from the most inner loop)
10573 for the @option{-floop-unroll-and-jam}. The default value is 2.
10575 @item ipa-cp-value-list-size
10576 IPA-CP attempts to track all possible values and types passed to a function's
10577 parameter in order to propagate them and perform devirtualization.
10578 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10579 stores per one formal parameter of a function.
10581 @item ipa-cp-eval-threshold
10582 IPA-CP calculates its own score of cloning profitability heuristics
10583 and performs those cloning opportunities with scores that exceed
10584 @option{ipa-cp-eval-threshold}.
10586 @item ipa-cp-recursion-penalty
10587 Percentage penalty the recursive functions will receive when they
10588 are evaluated for cloning.
10590 @item ipa-cp-single-call-penalty
10591 Percentage penalty functions containing a single call to another
10592 function will receive when they are evaluated for cloning.
10595 @item ipa-max-agg-items
10596 IPA-CP is also capable to propagate a number of scalar values passed
10597 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10598 number of such values per one parameter.
10600 @item ipa-cp-loop-hint-bonus
10601 When IPA-CP determines that a cloning candidate would make the number
10602 of iterations of a loop known, it adds a bonus of
10603 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10606 @item ipa-cp-array-index-hint-bonus
10607 When IPA-CP determines that a cloning candidate would make the index of
10608 an array access known, it adds a bonus of
10609 @option{ipa-cp-array-index-hint-bonus} to the profitability
10610 score of the candidate.
10612 @item ipa-max-aa-steps
10613 During its analysis of function bodies, IPA-CP employs alias analysis
10614 in order to track values pointed to by function parameters. In order
10615 not spend too much time analyzing huge functions, it gives up and
10616 consider all memory clobbered after examining
10617 @option{ipa-max-aa-steps} statements modifying memory.
10619 @item lto-partitions
10620 Specify desired number of partitions produced during WHOPR compilation.
10621 The number of partitions should exceed the number of CPUs used for compilation.
10622 The default value is 32.
10624 @item lto-min-partition
10625 Size of minimal partition for WHOPR (in estimated instructions).
10626 This prevents expenses of splitting very small programs into too many
10629 @item lto-max-partition
10630 Size of max partition for WHOPR (in estimated instructions).
10631 to provide an upper bound for individual size of partition.
10632 Meant to be used only with balanced partitioning.
10634 @item cxx-max-namespaces-for-diagnostic-help
10635 The maximum number of namespaces to consult for suggestions when C++
10636 name lookup fails for an identifier. The default is 1000.
10638 @item sink-frequency-threshold
10639 The maximum relative execution frequency (in percents) of the target block
10640 relative to a statement's original block to allow statement sinking of a
10641 statement. Larger numbers result in more aggressive statement sinking.
10642 The default value is 75. A small positive adjustment is applied for
10643 statements with memory operands as those are even more profitable so sink.
10645 @item max-stores-to-sink
10646 The maximum number of conditional store pairs that can be sunk. Set to 0
10647 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10648 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10650 @item allow-store-data-races
10651 Allow optimizers to introduce new data races on stores.
10652 Set to 1 to allow, otherwise to 0. This option is enabled by default
10653 at optimization level @option{-Ofast}.
10655 @item case-values-threshold
10656 The smallest number of different values for which it is best to use a
10657 jump-table instead of a tree of conditional branches. If the value is
10658 0, use the default for the machine. The default is 0.
10660 @item tree-reassoc-width
10661 Set the maximum number of instructions executed in parallel in
10662 reassociated tree. This parameter overrides target dependent
10663 heuristics used by default if has non zero value.
10665 @item sched-pressure-algorithm
10666 Choose between the two available implementations of
10667 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10668 and is the more likely to prevent instructions from being reordered.
10669 Algorithm 2 was designed to be a compromise between the relatively
10670 conservative approach taken by algorithm 1 and the rather aggressive
10671 approach taken by the default scheduler. It relies more heavily on
10672 having a regular register file and accurate register pressure classes.
10673 See @file{haifa-sched.c} in the GCC sources for more details.
10675 The default choice depends on the target.
10677 @item max-slsr-cand-scan
10678 Set the maximum number of existing candidates that are considered when
10679 seeking a basis for a new straight-line strength reduction candidate.
10682 Enable buffer overflow detection for global objects. This kind
10683 of protection is enabled by default if you are using
10684 @option{-fsanitize=address} option.
10685 To disable global objects protection use @option{--param asan-globals=0}.
10688 Enable buffer overflow detection for stack objects. This kind of
10689 protection is enabled by default when using @option{-fsanitize=address}.
10690 To disable stack protection use @option{--param asan-stack=0} option.
10692 @item asan-instrument-reads
10693 Enable buffer overflow detection for memory reads. This kind of
10694 protection is enabled by default when using @option{-fsanitize=address}.
10695 To disable memory reads protection use
10696 @option{--param asan-instrument-reads=0}.
10698 @item asan-instrument-writes
10699 Enable buffer overflow detection for memory writes. This kind of
10700 protection is enabled by default when using @option{-fsanitize=address}.
10701 To disable memory writes protection use
10702 @option{--param asan-instrument-writes=0} option.
10704 @item asan-memintrin
10705 Enable detection for built-in functions. This kind of protection
10706 is enabled by default when using @option{-fsanitize=address}.
10707 To disable built-in functions protection use
10708 @option{--param asan-memintrin=0}.
10710 @item asan-use-after-return
10711 Enable detection of use-after-return. This kind of protection
10712 is enabled by default when using the @option{-fsanitize=address} option.
10713 To disable it use @option{--param asan-use-after-return=0}.
10715 Note: By default the check is disabled at run time. To enable it,
10716 add @code{detect_stack_use_after_return=1} to the environment variable
10717 @env{ASAN_OPTIONS}.
10719 @item asan-instrumentation-with-call-threshold
10720 If number of memory accesses in function being instrumented
10721 is greater or equal to this number, use callbacks instead of inline checks.
10722 E.g. to disable inline code use
10723 @option{--param asan-instrumentation-with-call-threshold=0}.
10725 @item use-after-scope-direct-emission-threshold
10726 If the size of a local variable in bytes is smaller or equal to this
10727 number, directly poison (or unpoison) shadow memory instead of using
10728 run-time callbacks. The default value is 256.
10730 @item chkp-max-ctor-size
10731 Static constructors generated by Pointer Bounds Checker may become very
10732 large and significantly increase compile time at optimization level
10733 @option{-O1} and higher. This parameter is a maximum number of statements
10734 in a single generated constructor. Default value is 5000.
10736 @item max-fsm-thread-path-insns
10737 Maximum number of instructions to copy when duplicating blocks on a
10738 finite state automaton jump thread path. The default is 100.
10740 @item max-fsm-thread-length
10741 Maximum number of basic blocks on a finite state automaton jump thread
10742 path. The default is 10.
10744 @item max-fsm-thread-paths
10745 Maximum number of new jump thread paths to create for a finite state
10746 automaton. The default is 50.
10748 @item parloops-chunk-size
10749 Chunk size of omp schedule for loops parallelized by parloops. The default
10752 @item parloops-schedule
10753 Schedule type of omp schedule for loops parallelized by parloops (static,
10754 dynamic, guided, auto, runtime). The default is static.
10756 @item max-ssa-name-query-depth
10757 Maximum depth of recursion when querying properties of SSA names in things
10758 like fold routines. One level of recursion corresponds to following a
10761 @item hsa-gen-debug-stores
10762 Enable emission of special debug stores within HSA kernels which are
10763 then read and reported by libgomp plugin. Generation of these stores
10764 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10767 @item max-speculative-devirt-maydefs
10768 The maximum number of may-defs we analyze when looking for a must-def
10769 specifying the dynamic type of an object that invokes a virtual call
10770 we may be able to devirtualize speculatively.
10772 @item max-vrp-switch-assertions
10773 The maximum number of assertions to add along the default edge of a switch
10774 statement during VRP. The default is 10.
10778 @node Instrumentation Options
10779 @section Program Instrumentation Options
10780 @cindex instrumentation options
10781 @cindex program instrumentation options
10782 @cindex run-time error checking options
10783 @cindex profiling options
10784 @cindex options, program instrumentation
10785 @cindex options, run-time error checking
10786 @cindex options, profiling
10788 GCC supports a number of command-line options that control adding
10789 run-time instrumentation to the code it normally generates.
10790 For example, one purpose of instrumentation is collect profiling
10791 statistics for use in finding program hot spots, code coverage
10792 analysis, or profile-guided optimizations.
10793 Another class of program instrumentation is adding run-time checking
10794 to detect programming errors like invalid pointer
10795 dereferences or out-of-bounds array accesses, as well as deliberately
10796 hostile attacks such as stack smashing or C++ vtable hijacking.
10797 There is also a general hook which can be used to implement other
10798 forms of tracing or function-level instrumentation for debug or
10799 program analysis purposes.
10802 @cindex @command{prof}
10805 Generate extra code to write profile information suitable for the
10806 analysis program @command{prof}. You must use this option when compiling
10807 the source files you want data about, and you must also use it when
10810 @cindex @command{gprof}
10813 Generate extra code to write profile information suitable for the
10814 analysis program @command{gprof}. You must use this option when compiling
10815 the source files you want data about, and you must also use it when
10818 @item -fprofile-arcs
10819 @opindex fprofile-arcs
10820 Add code so that program flow @dfn{arcs} are instrumented. During
10821 execution the program records how many times each branch and call is
10822 executed and how many times it is taken or returns. On targets that support
10823 constructors with priority support, profiling properly handles constructors,
10824 destructors and C++ constructors (and destructors) of classes which are used
10825 as a type of a global variable.
10828 program exits it saves this data to a file called
10829 @file{@var{auxname}.gcda} for each source file. The data may be used for
10830 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10831 test coverage analysis (@option{-ftest-coverage}). Each object file's
10832 @var{auxname} is generated from the name of the output file, if
10833 explicitly specified and it is not the final executable, otherwise it is
10834 the basename of the source file. In both cases any suffix is removed
10835 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10836 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10837 @xref{Cross-profiling}.
10839 @cindex @command{gcov}
10843 This option is used to compile and link code instrumented for coverage
10844 analysis. The option is a synonym for @option{-fprofile-arcs}
10845 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10846 linking). See the documentation for those options for more details.
10851 Compile the source files with @option{-fprofile-arcs} plus optimization
10852 and code generation options. For test coverage analysis, use the
10853 additional @option{-ftest-coverage} option. You do not need to profile
10854 every source file in a program.
10857 Compile the source files additionally with @option{-fprofile-abs-path}
10858 to create absolute path names in the @file{.gcno} files. This allows
10859 @command{gcov} to find the correct sources in projects where compilations
10860 occur with different working directories.
10863 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10864 (the latter implies the former).
10867 Run the program on a representative workload to generate the arc profile
10868 information. This may be repeated any number of times. You can run
10869 concurrent instances of your program, and provided that the file system
10870 supports locking, the data files will be correctly updated. Also
10871 @code{fork} calls are detected and correctly handled (double counting
10875 For profile-directed optimizations, compile the source files again with
10876 the same optimization and code generation options plus
10877 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10878 Control Optimization}).
10881 For test coverage analysis, use @command{gcov} to produce human readable
10882 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10883 @command{gcov} documentation for further information.
10887 With @option{-fprofile-arcs}, for each function of your program GCC
10888 creates a program flow graph, then finds a spanning tree for the graph.
10889 Only arcs that are not on the spanning tree have to be instrumented: the
10890 compiler adds code to count the number of times that these arcs are
10891 executed. When an arc is the only exit or only entrance to a block, the
10892 instrumentation code can be added to the block; otherwise, a new basic
10893 block must be created to hold the instrumentation code.
10896 @item -ftest-coverage
10897 @opindex ftest-coverage
10898 Produce a notes file that the @command{gcov} code-coverage utility
10899 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10900 show program coverage. Each source file's note file is called
10901 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10902 above for a description of @var{auxname} and instructions on how to
10903 generate test coverage data. Coverage data matches the source files
10904 more closely if you do not optimize.
10906 @item -fprofile-abs-path
10907 @opindex fprofile-abs-path
10908 Automatically convert relative source file names to absolute path names
10909 in the @file{.gcno} files. This allows @command{gcov} to find the correct
10910 sources in projects where compilations occur with different working
10913 @item -fprofile-dir=@var{path}
10914 @opindex fprofile-dir
10916 Set the directory to search for the profile data files in to @var{path}.
10917 This option affects only the profile data generated by
10918 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10919 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10920 and its related options. Both absolute and relative paths can be used.
10921 By default, GCC uses the current directory as @var{path}, thus the
10922 profile data file appears in the same directory as the object file.
10924 @item -fprofile-generate
10925 @itemx -fprofile-generate=@var{path}
10926 @opindex fprofile-generate
10928 Enable options usually used for instrumenting application to produce
10929 profile useful for later recompilation with profile feedback based
10930 optimization. You must use @option{-fprofile-generate} both when
10931 compiling and when linking your program.
10933 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10935 If @var{path} is specified, GCC looks at the @var{path} to find
10936 the profile feedback data files. See @option{-fprofile-dir}.
10938 To optimize the program based on the collected profile information, use
10939 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10941 @item -fprofile-update=@var{method}
10942 @opindex fprofile-update
10944 Alter the update method for an application instrumented for profile
10945 feedback based optimization. The @var{method} argument should be one of
10946 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
10947 The first one is useful for single-threaded applications,
10948 while the second one prevents profile corruption by emitting thread-safe code.
10950 @strong{Warning:} When an application does not properly join all threads
10951 (or creates an detached thread), a profile file can be still corrupted.
10953 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
10954 when supported by a target, or to @samp{single} otherwise. The GCC driver
10955 automatically selects @samp{prefer-atomic} when @option{-pthread}
10956 is present in the command line.
10958 @item -fsanitize=address
10959 @opindex fsanitize=address
10960 Enable AddressSanitizer, a fast memory error detector.
10961 Memory access instructions are instrumented to detect
10962 out-of-bounds and use-after-free bugs.
10963 The option enables @option{-fsanitize-address-use-after-scope}.
10964 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
10965 more details. The run-time behavior can be influenced using the
10966 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
10967 the available options are shown at startup of the instrumented program. See
10968 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
10969 for a list of supported options.
10970 The option cannot be combined with @option{-fsanitize=thread}
10971 and/or @option{-fcheck-pointer-bounds}.
10973 @item -fsanitize=kernel-address
10974 @opindex fsanitize=kernel-address
10975 Enable AddressSanitizer for Linux kernel.
10976 See @uref{https://github.com/google/kasan/wiki} for more details.
10977 The option cannot be combined with @option{-fcheck-pointer-bounds}.
10979 @item -fsanitize=thread
10980 @opindex fsanitize=thread
10981 Enable ThreadSanitizer, a fast data race detector.
10982 Memory access instructions are instrumented to detect
10983 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
10984 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
10985 environment variable; see
10986 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
10988 The option cannot be combined with @option{-fsanitize=address},
10989 @option{-fsanitize=leak} and/or @option{-fcheck-pointer-bounds}.
10991 Note that sanitized atomic builtins cannot throw exceptions when
10992 operating on invalid memory addresses with non-call exceptions
10993 (@option{-fnon-call-exceptions}).
10995 @item -fsanitize=leak
10996 @opindex fsanitize=leak
10997 Enable LeakSanitizer, a memory leak detector.
10998 This option only matters for linking of executables and
10999 the executable is linked against a library that overrides @code{malloc}
11000 and other allocator functions. See
11001 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
11002 details. The run-time behavior can be influenced using the
11003 @env{LSAN_OPTIONS} environment variable.
11004 The option cannot be combined with @option{-fsanitize=thread}.
11006 @item -fsanitize=undefined
11007 @opindex fsanitize=undefined
11008 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
11009 Various computations are instrumented to detect undefined behavior
11010 at runtime. Current suboptions are:
11014 @item -fsanitize=shift
11015 @opindex fsanitize=shift
11016 This option enables checking that the result of a shift operation is
11017 not undefined. Note that what exactly is considered undefined differs
11018 slightly between C and C++, as well as between ISO C90 and C99, etc.
11019 This option has two suboptions, @option{-fsanitize=shift-base} and
11020 @option{-fsanitize=shift-exponent}.
11022 @item -fsanitize=shift-exponent
11023 @opindex fsanitize=shift-exponent
11024 This option enables checking that the second argument of a shift operation
11025 is not negative and is smaller than the precision of the promoted first
11028 @item -fsanitize=shift-base
11029 @opindex fsanitize=shift-base
11030 If the second argument of a shift operation is within range, check that the
11031 result of a shift operation is not undefined. Note that what exactly is
11032 considered undefined differs slightly between C and C++, as well as between
11033 ISO C90 and C99, etc.
11035 @item -fsanitize=integer-divide-by-zero
11036 @opindex fsanitize=integer-divide-by-zero
11037 Detect integer division by zero as well as @code{INT_MIN / -1} division.
11039 @item -fsanitize=unreachable
11040 @opindex fsanitize=unreachable
11041 With this option, the compiler turns the @code{__builtin_unreachable}
11042 call into a diagnostics message call instead. When reaching the
11043 @code{__builtin_unreachable} call, the behavior is undefined.
11045 @item -fsanitize=vla-bound
11046 @opindex fsanitize=vla-bound
11047 This option instructs the compiler to check that the size of a variable
11048 length array is positive.
11050 @item -fsanitize=null
11051 @opindex fsanitize=null
11052 This option enables pointer checking. Particularly, the application
11053 built with this option turned on will issue an error message when it
11054 tries to dereference a NULL pointer, or if a reference (possibly an
11055 rvalue reference) is bound to a NULL pointer, or if a method is invoked
11056 on an object pointed by a NULL pointer.
11058 @item -fsanitize=return
11059 @opindex fsanitize=return
11060 This option enables return statement checking. Programs
11061 built with this option turned on will issue an error message
11062 when the end of a non-void function is reached without actually
11063 returning a value. This option works in C++ only.
11065 @item -fsanitize=signed-integer-overflow
11066 @opindex fsanitize=signed-integer-overflow
11067 This option enables signed integer overflow checking. We check that
11068 the result of @code{+}, @code{*}, and both unary and binary @code{-}
11069 does not overflow in the signed arithmetics. Note, integer promotion
11070 rules must be taken into account. That is, the following is not an
11073 signed char a = SCHAR_MAX;
11077 @item -fsanitize=bounds
11078 @opindex fsanitize=bounds
11079 This option enables instrumentation of array bounds. Various out of bounds
11080 accesses are detected. Flexible array members, flexible array member-like
11081 arrays, and initializers of variables with static storage are not instrumented.
11082 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11084 @item -fsanitize=bounds-strict
11085 @opindex fsanitize=bounds-strict
11086 This option enables strict instrumentation of array bounds. Most out of bounds
11087 accesses are detected, including flexible array members and flexible array
11088 member-like arrays. Initializers of variables with static storage are not
11089 instrumented. The option cannot be combined
11090 with @option{-fcheck-pointer-bounds}.
11092 @item -fsanitize=alignment
11093 @opindex fsanitize=alignment
11095 This option enables checking of alignment of pointers when they are
11096 dereferenced, or when a reference is bound to insufficiently aligned target,
11097 or when a method or constructor is invoked on insufficiently aligned object.
11099 @item -fsanitize=object-size
11100 @opindex fsanitize=object-size
11101 This option enables instrumentation of memory references using the
11102 @code{__builtin_object_size} function. Various out of bounds pointer
11103 accesses are detected.
11105 @item -fsanitize=float-divide-by-zero
11106 @opindex fsanitize=float-divide-by-zero
11107 Detect floating-point division by zero. Unlike other similar options,
11108 @option{-fsanitize=float-divide-by-zero} is not enabled by
11109 @option{-fsanitize=undefined}, since floating-point division by zero can
11110 be a legitimate way of obtaining infinities and NaNs.
11112 @item -fsanitize=float-cast-overflow
11113 @opindex fsanitize=float-cast-overflow
11114 This option enables floating-point type to integer conversion checking.
11115 We check that the result of the conversion does not overflow.
11116 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
11117 not enabled by @option{-fsanitize=undefined}.
11118 This option does not work well with @code{FE_INVALID} exceptions enabled.
11120 @item -fsanitize=nonnull-attribute
11121 @opindex fsanitize=nonnull-attribute
11123 This option enables instrumentation of calls, checking whether null values
11124 are not passed to arguments marked as requiring a non-null value by the
11125 @code{nonnull} function attribute.
11127 @item -fsanitize=returns-nonnull-attribute
11128 @opindex fsanitize=returns-nonnull-attribute
11130 This option enables instrumentation of return statements in functions
11131 marked with @code{returns_nonnull} function attribute, to detect returning
11132 of null values from such functions.
11134 @item -fsanitize=bool
11135 @opindex fsanitize=bool
11137 This option enables instrumentation of loads from bool. If a value other
11138 than 0/1 is loaded, a run-time error is issued.
11140 @item -fsanitize=enum
11141 @opindex fsanitize=enum
11143 This option enables instrumentation of loads from an enum type. If
11144 a value outside the range of values for the enum type is loaded,
11145 a run-time error is issued.
11147 @item -fsanitize=vptr
11148 @opindex fsanitize=vptr
11150 This option enables instrumentation of C++ member function calls, member
11151 accesses and some conversions between pointers to base and derived classes,
11152 to verify the referenced object has the correct dynamic type.
11154 @item -fsanitize=pointer-overflow
11155 @opindex fsanitize=pointer-overflow
11157 This option enables instrumentation of pointer arithmetics. If the pointer
11158 arithmetics overflows, a run-time error is issued.
11160 @item -fsanitize=builtin
11161 @opindex fsanitize=builtin
11163 This option enables instrumentation of arguments to selected builtin
11164 functions. If an invalid value is passed to such arguments, a run-time
11165 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
11166 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
11171 While @option{-ftrapv} causes traps for signed overflows to be emitted,
11172 @option{-fsanitize=undefined} gives a diagnostic message.
11173 This currently works only for the C family of languages.
11175 @item -fno-sanitize=all
11176 @opindex fno-sanitize=all
11178 This option disables all previously enabled sanitizers.
11179 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
11182 @item -fasan-shadow-offset=@var{number}
11183 @opindex fasan-shadow-offset
11184 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
11185 It is useful for experimenting with different shadow memory layouts in
11186 Kernel AddressSanitizer.
11188 @item -fsanitize-sections=@var{s1},@var{s2},...
11189 @opindex fsanitize-sections
11190 Sanitize global variables in selected user-defined sections. @var{si} may
11193 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
11194 @opindex fsanitize-recover
11195 @opindex fno-sanitize-recover
11196 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
11197 mentioned in comma-separated list of @var{opts}. Enabling this option
11198 for a sanitizer component causes it to attempt to continue
11199 running the program as if no error happened. This means multiple
11200 runtime errors can be reported in a single program run, and the exit
11201 code of the program may indicate success even when errors
11202 have been reported. The @option{-fno-sanitize-recover=} option
11203 can be used to alter
11204 this behavior: only the first detected error is reported
11205 and program then exits with a non-zero exit code.
11207 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
11208 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
11209 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
11210 @option{-fsanitize=bounds-strict},
11211 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
11212 For these sanitizers error recovery is turned on by default,
11213 except @option{-fsanitize=address}, for which this feature is experimental.
11214 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
11215 accepted, the former enables recovery for all sanitizers that support it,
11216 the latter disables recovery for all sanitizers that support it.
11218 Even if a recovery mode is turned on the compiler side, it needs to be also
11219 enabled on the runtime library side, otherwise the failures are still fatal.
11220 The runtime library defaults to @code{halt_on_error=0} for
11221 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
11222 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
11223 setting the @code{halt_on_error} flag in the corresponding environment variable.
11225 Syntax without an explicit @var{opts} parameter is deprecated. It is
11226 equivalent to specifying an @var{opts} list of:
11229 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
11232 @item -fsanitize-address-use-after-scope
11233 @opindex fsanitize-address-use-after-scope
11234 Enable sanitization of local variables to detect use-after-scope bugs.
11235 The option sets @option{-fstack-reuse} to @samp{none}.
11237 @item -fsanitize-undefined-trap-on-error
11238 @opindex fsanitize-undefined-trap-on-error
11239 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
11240 report undefined behavior using @code{__builtin_trap} rather than
11241 a @code{libubsan} library routine. The advantage of this is that the
11242 @code{libubsan} library is not needed and is not linked in, so this
11243 is usable even in freestanding environments.
11245 @item -fsanitize-coverage=trace-pc
11246 @opindex fsanitize-coverage=trace-pc
11247 Enable coverage-guided fuzzing code instrumentation.
11248 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
11250 @item -fsanitize-coverage=trace-cmp
11251 @opindex fsanitize-coverage=trace-cmp
11252 Enable dataflow guided fuzzing code instrumentation.
11253 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
11254 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
11255 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
11256 variable or @code{__sanitizer_cov_trace_const_cmp1},
11257 @code{__sanitizer_cov_trace_const_cmp2},
11258 @code{__sanitizer_cov_trace_const_cmp4} or
11259 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
11260 operand constant, @code{__sanitizer_cov_trace_cmpf} or
11261 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
11262 @code{__sanitizer_cov_trace_switch} for switch statements.
11264 @item -fbounds-check
11265 @opindex fbounds-check
11266 For front ends that support it, generate additional code to check that
11267 indices used to access arrays are within the declared range. This is
11268 currently only supported by the Fortran front end, where this option
11271 @item -fcheck-pointer-bounds
11272 @opindex fcheck-pointer-bounds
11273 @opindex fno-check-pointer-bounds
11274 @cindex Pointer Bounds Checker options
11275 Enable Pointer Bounds Checker instrumentation. Each memory reference
11276 is instrumented with checks of the pointer used for memory access against
11277 bounds associated with that pointer.
11280 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
11281 and @option{-mmpx} are required to enable this feature.
11282 MPX-based instrumentation requires
11283 a runtime library to enable MPX in hardware and handle bounds
11284 violation signals. By default when @option{-fcheck-pointer-bounds}
11285 and @option{-mmpx} options are used to link a program, the GCC driver
11286 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
11287 Bounds checking on calls to dynamic libraries requires a linker
11288 with @option{-z bndplt} support; if GCC was configured with a linker
11289 without support for this option (including the Gold linker and older
11290 versions of ld), a warning is given if you link with @option{-mmpx}
11291 without also specifying @option{-static}, since the overall effectiveness
11292 of the bounds checking protection is reduced.
11293 See also @option{-static-libmpxwrappers}.
11295 MPX-based instrumentation
11296 may be used for debugging and also may be included in production code
11297 to increase program security. Depending on usage, you may
11298 have different requirements for the runtime library. The current version
11299 of the MPX runtime library is more oriented for use as a debugging
11300 tool. MPX runtime library usage implies @option{-lpthread}. See
11301 also @option{-static-libmpx}. The runtime library behavior can be
11302 influenced using various @env{CHKP_RT_*} environment variables. See
11303 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
11306 Generated instrumentation may be controlled by various
11307 @option{-fchkp-*} options and by the @code{bnd_variable_size}
11308 structure field attribute (@pxref{Type Attributes}) and
11309 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
11310 (@pxref{Function Attributes}). GCC also provides a number of built-in
11311 functions for controlling the Pointer Bounds Checker. @xref{Pointer
11312 Bounds Checker builtins}, for more information.
11314 @item -fchkp-check-incomplete-type
11315 @opindex fchkp-check-incomplete-type
11316 @opindex fno-chkp-check-incomplete-type
11317 Generate pointer bounds checks for variables with incomplete type.
11318 Enabled by default.
11320 @item -fchkp-narrow-bounds
11321 @opindex fchkp-narrow-bounds
11322 @opindex fno-chkp-narrow-bounds
11323 Controls bounds used by Pointer Bounds Checker for pointers to object
11324 fields. If narrowing is enabled then field bounds are used. Otherwise
11325 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
11326 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
11328 @item -fchkp-first-field-has-own-bounds
11329 @opindex fchkp-first-field-has-own-bounds
11330 @opindex fno-chkp-first-field-has-own-bounds
11331 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
11332 first field in the structure. By default a pointer to the first field has
11333 the same bounds as a pointer to the whole structure.
11335 @item -fchkp-flexible-struct-trailing-arrays
11336 @opindex fchkp-flexible-struct-trailing-arrays
11337 @opindex fno-chkp-flexible-struct-trailing-arrays
11338 Forces Pointer Bounds Checker to treat all trailing arrays in structures as
11339 possibly flexible. By default only array fields with zero length or that are
11340 marked with attribute bnd_variable_size are treated as flexible.
11342 @item -fchkp-narrow-to-innermost-array
11343 @opindex fchkp-narrow-to-innermost-array
11344 @opindex fno-chkp-narrow-to-innermost-array
11345 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
11346 case of nested static array access. By default this option is disabled and
11347 bounds of the outermost array are used.
11349 @item -fchkp-optimize
11350 @opindex fchkp-optimize
11351 @opindex fno-chkp-optimize
11352 Enables Pointer Bounds Checker optimizations. Enabled by default at
11353 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11355 @item -fchkp-use-fast-string-functions
11356 @opindex fchkp-use-fast-string-functions
11357 @opindex fno-chkp-use-fast-string-functions
11358 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11359 by Pointer Bounds Checker. Disabled by default.
11361 @item -fchkp-use-nochk-string-functions
11362 @opindex fchkp-use-nochk-string-functions
11363 @opindex fno-chkp-use-nochk-string-functions
11364 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11365 by Pointer Bounds Checker. Disabled by default.
11367 @item -fchkp-use-static-bounds
11368 @opindex fchkp-use-static-bounds
11369 @opindex fno-chkp-use-static-bounds
11370 Allow Pointer Bounds Checker to generate static bounds holding
11371 bounds of static variables. Enabled by default.
11373 @item -fchkp-use-static-const-bounds
11374 @opindex fchkp-use-static-const-bounds
11375 @opindex fno-chkp-use-static-const-bounds
11376 Use statically-initialized bounds for constant bounds instead of
11377 generating them each time they are required. By default enabled when
11378 @option{-fchkp-use-static-bounds} is enabled.
11380 @item -fchkp-treat-zero-dynamic-size-as-infinite
11381 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11382 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11383 With this option, objects with incomplete type whose
11384 dynamically-obtained size is zero are treated as having infinite size
11385 instead by Pointer Bounds
11386 Checker. This option may be helpful if a program is linked with a library
11387 missing size information for some symbols. Disabled by default.
11389 @item -fchkp-check-read
11390 @opindex fchkp-check-read
11391 @opindex fno-chkp-check-read
11392 Instructs Pointer Bounds Checker to generate checks for all read
11393 accesses to memory. Enabled by default.
11395 @item -fchkp-check-write
11396 @opindex fchkp-check-write
11397 @opindex fno-chkp-check-write
11398 Instructs Pointer Bounds Checker to generate checks for all write
11399 accesses to memory. Enabled by default.
11401 @item -fchkp-store-bounds
11402 @opindex fchkp-store-bounds
11403 @opindex fno-chkp-store-bounds
11404 Instructs Pointer Bounds Checker to generate bounds stores for
11405 pointer writes. Enabled by default.
11407 @item -fchkp-instrument-calls
11408 @opindex fchkp-instrument-calls
11409 @opindex fno-chkp-instrument-calls
11410 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11411 Enabled by default.
11413 @item -fchkp-instrument-marked-only
11414 @opindex fchkp-instrument-marked-only
11415 @opindex fno-chkp-instrument-marked-only
11416 Instructs Pointer Bounds Checker to instrument only functions
11417 marked with the @code{bnd_instrument} attribute
11418 (@pxref{Function Attributes}). Disabled by default.
11420 @item -fchkp-use-wrappers
11421 @opindex fchkp-use-wrappers
11422 @opindex fno-chkp-use-wrappers
11423 Allows Pointer Bounds Checker to replace calls to built-in functions
11424 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11425 is used to link a program, the GCC driver automatically links
11426 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11427 Enabled by default.
11429 @item -fcf-protection==@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
11430 @opindex fcf-protection
11431 Enable code instrumentation of control-flow transfers to increase
11432 program security by checking that target addresses of control-flow
11433 transfer instructions (such as indirect function call, function return,
11434 indirect jump) are valid. This prevents diverting the flow of control
11435 to an unexpected target. This is intended to protect against such
11436 threats as Return-oriented Programming (ROP), and similarly
11437 call/jmp-oriented programming (COP/JOP).
11439 The value @code{branch} tells the compiler to implement checking of
11440 validity of control-flow transfer at the point of indirect branch
11441 instructions, i.e. call/jmp instructions. The value @code{return}
11442 implements checking of validity at the point of returning from a
11443 function. The value @code{full} is an alias for specifying both
11444 @code{branch} and @code{return}. The value @code{none} turns off
11447 You can also use the @code{nocf_check} attribute to identify
11448 which functions and calls should be skipped from instrumentation
11449 (@pxref{Function Attributes}).
11451 Currently the x86 GNU/Linux target provides an implementation based
11452 on Intel Control-flow Enforcement Technology (CET). Instrumentation
11453 for x86 is controlled by target-specific options @option{-mcet},
11454 @option{-mibt} and @option{-mshstk} (@pxref{x86 Options}).
11456 @item -fstack-protector
11457 @opindex fstack-protector
11458 Emit extra code to check for buffer overflows, such as stack smashing
11459 attacks. This is done by adding a guard variable to functions with
11460 vulnerable objects. This includes functions that call @code{alloca}, and
11461 functions with buffers larger than 8 bytes. The guards are initialized
11462 when a function is entered and then checked when the function exits.
11463 If a guard check fails, an error message is printed and the program exits.
11465 @item -fstack-protector-all
11466 @opindex fstack-protector-all
11467 Like @option{-fstack-protector} except that all functions are protected.
11469 @item -fstack-protector-strong
11470 @opindex fstack-protector-strong
11471 Like @option{-fstack-protector} but includes additional functions to
11472 be protected --- those that have local array definitions, or have
11473 references to local frame addresses.
11475 @item -fstack-protector-explicit
11476 @opindex fstack-protector-explicit
11477 Like @option{-fstack-protector} but only protects those functions which
11478 have the @code{stack_protect} attribute.
11480 @item -fstack-check
11481 @opindex fstack-check
11482 Generate code to verify that you do not go beyond the boundary of the
11483 stack. You should specify this flag if you are running in an
11484 environment with multiple threads, but you only rarely need to specify it in
11485 a single-threaded environment since stack overflow is automatically
11486 detected on nearly all systems if there is only one stack.
11488 Note that this switch does not actually cause checking to be done; the
11489 operating system or the language runtime must do that. The switch causes
11490 generation of code to ensure that they see the stack being extended.
11492 You can additionally specify a string parameter: @samp{no} means no
11493 checking, @samp{generic} means force the use of old-style checking,
11494 @samp{specific} means use the best checking method and is equivalent
11495 to bare @option{-fstack-check}.
11497 Old-style checking is a generic mechanism that requires no specific
11498 target support in the compiler but comes with the following drawbacks:
11502 Modified allocation strategy for large objects: they are always
11503 allocated dynamically if their size exceeds a fixed threshold. Note this
11504 may change the semantics of some code.
11507 Fixed limit on the size of the static frame of functions: when it is
11508 topped by a particular function, stack checking is not reliable and
11509 a warning is issued by the compiler.
11512 Inefficiency: because of both the modified allocation strategy and the
11513 generic implementation, code performance is hampered.
11516 Note that old-style stack checking is also the fallback method for
11517 @samp{specific} if no target support has been added in the compiler.
11519 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
11520 and stack overflows. @samp{specific} is an excellent choice when compiling
11521 Ada code. It is not generally sufficient to protect against stack-clash
11522 attacks. To protect against those you want @samp{-fstack-clash-protection}.
11524 @item -fstack-clash-protection
11525 @opindex fstack-clash-protection
11526 Generate code to prevent stack clash style attacks. When this option is
11527 enabled, the compiler will only allocate one page of stack space at a time
11528 and each page is accessed immediately after allocation. Thus, it prevents
11529 allocations from jumping over any stack guard page provided by the
11532 Most targets do not fully support stack clash protection. However, on
11533 those targets @option{-fstack-clash-protection} will protect dynamic stack
11534 allocations. @option{-fstack-clash-protection} may also provide limited
11535 protection for static stack allocations if the target supports
11536 @option{-fstack-check=specific}.
11538 @item -fstack-limit-register=@var{reg}
11539 @itemx -fstack-limit-symbol=@var{sym}
11540 @itemx -fno-stack-limit
11541 @opindex fstack-limit-register
11542 @opindex fstack-limit-symbol
11543 @opindex fno-stack-limit
11544 Generate code to ensure that the stack does not grow beyond a certain value,
11545 either the value of a register or the address of a symbol. If a larger
11546 stack is required, a signal is raised at run time. For most targets,
11547 the signal is raised before the stack overruns the boundary, so
11548 it is possible to catch the signal without taking special precautions.
11550 For instance, if the stack starts at absolute address @samp{0x80000000}
11551 and grows downwards, you can use the flags
11552 @option{-fstack-limit-symbol=__stack_limit} and
11553 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11554 of 128KB@. Note that this may only work with the GNU linker.
11556 You can locally override stack limit checking by using the
11557 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11559 @item -fsplit-stack
11560 @opindex fsplit-stack
11561 Generate code to automatically split the stack before it overflows.
11562 The resulting program has a discontiguous stack which can only
11563 overflow if the program is unable to allocate any more memory. This
11564 is most useful when running threaded programs, as it is no longer
11565 necessary to calculate a good stack size to use for each thread. This
11566 is currently only implemented for the x86 targets running
11569 When code compiled with @option{-fsplit-stack} calls code compiled
11570 without @option{-fsplit-stack}, there may not be much stack space
11571 available for the latter code to run. If compiling all code,
11572 including library code, with @option{-fsplit-stack} is not an option,
11573 then the linker can fix up these calls so that the code compiled
11574 without @option{-fsplit-stack} always has a large stack. Support for
11575 this is implemented in the gold linker in GNU binutils release 2.21
11578 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11579 @opindex fvtable-verify
11580 This option is only available when compiling C++ code.
11581 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11582 feature that verifies at run time, for every virtual call, that
11583 the vtable pointer through which the call is made is valid for the type of
11584 the object, and has not been corrupted or overwritten. If an invalid vtable
11585 pointer is detected at run time, an error is reported and execution of the
11586 program is immediately halted.
11588 This option causes run-time data structures to be built at program startup,
11589 which are used for verifying the vtable pointers.
11590 The options @samp{std} and @samp{preinit}
11591 control the timing of when these data structures are built. In both cases the
11592 data structures are built before execution reaches @code{main}. Using
11593 @option{-fvtable-verify=std} causes the data structures to be built after
11594 shared libraries have been loaded and initialized.
11595 @option{-fvtable-verify=preinit} causes them to be built before shared
11596 libraries have been loaded and initialized.
11598 If this option appears multiple times in the command line with different
11599 values specified, @samp{none} takes highest priority over both @samp{std} and
11600 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11603 @opindex fvtv-debug
11604 When used in conjunction with @option{-fvtable-verify=std} or
11605 @option{-fvtable-verify=preinit}, causes debug versions of the
11606 runtime functions for the vtable verification feature to be called.
11607 This flag also causes the compiler to log information about which
11608 vtable pointers it finds for each class.
11609 This information is written to a file named @file{vtv_set_ptr_data.log}
11610 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11611 if that is defined or the current working directory otherwise.
11613 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11614 file, be sure to delete any existing one.
11617 @opindex fvtv-counts
11618 This is a debugging flag. When used in conjunction with
11619 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11620 causes the compiler to keep track of the total number of virtual calls
11621 it encounters and the number of verifications it inserts. It also
11622 counts the number of calls to certain run-time library functions
11623 that it inserts and logs this information for each compilation unit.
11624 The compiler writes this information to a file named
11625 @file{vtv_count_data.log} in the directory named by the environment
11626 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11627 directory otherwise. It also counts the size of the vtable pointer sets
11628 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11629 in the same directory.
11631 Note: This feature @emph{appends} data to the log files. To get fresh log
11632 files, be sure to delete any existing ones.
11634 @item -finstrument-functions
11635 @opindex finstrument-functions
11636 Generate instrumentation calls for entry and exit to functions. Just
11637 after function entry and just before function exit, the following
11638 profiling functions are called with the address of the current
11639 function and its call site. (On some platforms,
11640 @code{__builtin_return_address} does not work beyond the current
11641 function, so the call site information may not be available to the
11642 profiling functions otherwise.)
11645 void __cyg_profile_func_enter (void *this_fn,
11647 void __cyg_profile_func_exit (void *this_fn,
11651 The first argument is the address of the start of the current function,
11652 which may be looked up exactly in the symbol table.
11654 This instrumentation is also done for functions expanded inline in other
11655 functions. The profiling calls indicate where, conceptually, the
11656 inline function is entered and exited. This means that addressable
11657 versions of such functions must be available. If all your uses of a
11658 function are expanded inline, this may mean an additional expansion of
11659 code size. If you use @code{extern inline} in your C code, an
11660 addressable version of such functions must be provided. (This is
11661 normally the case anyway, but if you get lucky and the optimizer always
11662 expands the functions inline, you might have gotten away without
11663 providing static copies.)
11665 A function may be given the attribute @code{no_instrument_function}, in
11666 which case this instrumentation is not done. This can be used, for
11667 example, for the profiling functions listed above, high-priority
11668 interrupt routines, and any functions from which the profiling functions
11669 cannot safely be called (perhaps signal handlers, if the profiling
11670 routines generate output or allocate memory).
11672 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11673 @opindex finstrument-functions-exclude-file-list
11675 Set the list of functions that are excluded from instrumentation (see
11676 the description of @option{-finstrument-functions}). If the file that
11677 contains a function definition matches with one of @var{file}, then
11678 that function is not instrumented. The match is done on substrings:
11679 if the @var{file} parameter is a substring of the file name, it is
11680 considered to be a match.
11685 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11689 excludes any inline function defined in files whose pathnames
11690 contain @file{/bits/stl} or @file{include/sys}.
11692 If, for some reason, you want to include letter @samp{,} in one of
11693 @var{sym}, write @samp{\,}. For example,
11694 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11695 (note the single quote surrounding the option).
11697 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11698 @opindex finstrument-functions-exclude-function-list
11700 This is similar to @option{-finstrument-functions-exclude-file-list},
11701 but this option sets the list of function names to be excluded from
11702 instrumentation. The function name to be matched is its user-visible
11703 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11704 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11705 match is done on substrings: if the @var{sym} parameter is a substring
11706 of the function name, it is considered to be a match. For C99 and C++
11707 extended identifiers, the function name must be given in UTF-8, not
11708 using universal character names.
11710 @item -fpatchable-function-entry=@var{N}[,@var{M}]
11711 @opindex fpatchable-function-entry
11712 Generate @var{N} NOPs right at the beginning
11713 of each function, with the function entry point before the @var{M}th NOP.
11714 If @var{M} is omitted, it defaults to @code{0} so the
11715 function entry points to the address just at the first NOP.
11716 The NOP instructions reserve extra space which can be used to patch in
11717 any desired instrumentation at run time, provided that the code segment
11718 is writable. The amount of space is controllable indirectly via
11719 the number of NOPs; the NOP instruction used corresponds to the instruction
11720 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
11721 is target-specific and may also depend on the architecture variant and/or
11722 other compilation options.
11724 For run-time identification, the starting addresses of these areas,
11725 which correspond to their respective function entries minus @var{M},
11726 are additionally collected in the @code{__patchable_function_entries}
11727 section of the resulting binary.
11729 Note that the value of @code{__attribute__ ((patchable_function_entry
11730 (N,M)))} takes precedence over command-line option
11731 @option{-fpatchable-function-entry=N,M}. This can be used to increase
11732 the area size or to remove it completely on a single function.
11733 If @code{N=0}, no pad location is recorded.
11735 The NOP instructions are inserted at---and maybe before, depending on
11736 @var{M}---the function entry address, even before the prologue.
11741 @node Preprocessor Options
11742 @section Options Controlling the Preprocessor
11743 @cindex preprocessor options
11744 @cindex options, preprocessor
11746 These options control the C preprocessor, which is run on each C source
11747 file before actual compilation.
11749 If you use the @option{-E} option, nothing is done except preprocessing.
11750 Some of these options make sense only together with @option{-E} because
11751 they cause the preprocessor output to be unsuitable for actual
11754 In addition to the options listed here, there are a number of options
11755 to control search paths for include files documented in
11756 @ref{Directory Options}.
11757 Options to control preprocessor diagnostics are listed in
11758 @ref{Warning Options}.
11761 @include cppopts.texi
11763 @item -Wp,@var{option}
11765 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11766 and pass @var{option} directly through to the preprocessor. If
11767 @var{option} contains commas, it is split into multiple options at the
11768 commas. However, many options are modified, translated or interpreted
11769 by the compiler driver before being passed to the preprocessor, and
11770 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11771 interface is undocumented and subject to change, so whenever possible
11772 you should avoid using @option{-Wp} and let the driver handle the
11775 @item -Xpreprocessor @var{option}
11776 @opindex Xpreprocessor
11777 Pass @var{option} as an option to the preprocessor. You can use this to
11778 supply system-specific preprocessor options that GCC does not
11781 If you want to pass an option that takes an argument, you must use
11782 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11784 @item -no-integrated-cpp
11785 @opindex no-integrated-cpp
11786 Perform preprocessing as a separate pass before compilation.
11787 By default, GCC performs preprocessing as an integrated part of
11788 input tokenization and parsing.
11789 If this option is provided, the appropriate language front end
11790 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11791 and Objective-C, respectively) is instead invoked twice,
11792 once for preprocessing only and once for actual compilation
11793 of the preprocessed input.
11794 This option may be useful in conjunction with the @option{-B} or
11795 @option{-wrapper} options to specify an alternate preprocessor or
11796 perform additional processing of the program source between
11797 normal preprocessing and compilation.
11801 @node Assembler Options
11802 @section Passing Options to the Assembler
11804 @c prevent bad page break with this line
11805 You can pass options to the assembler.
11808 @item -Wa,@var{option}
11810 Pass @var{option} as an option to the assembler. If @var{option}
11811 contains commas, it is split into multiple options at the commas.
11813 @item -Xassembler @var{option}
11814 @opindex Xassembler
11815 Pass @var{option} as an option to the assembler. You can use this to
11816 supply system-specific assembler options that GCC does not
11819 If you want to pass an option that takes an argument, you must use
11820 @option{-Xassembler} twice, once for the option and once for the argument.
11825 @section Options for Linking
11826 @cindex link options
11827 @cindex options, linking
11829 These options come into play when the compiler links object files into
11830 an executable output file. They are meaningless if the compiler is
11831 not doing a link step.
11835 @item @var{object-file-name}
11836 A file name that does not end in a special recognized suffix is
11837 considered to name an object file or library. (Object files are
11838 distinguished from libraries by the linker according to the file
11839 contents.) If linking is done, these object files are used as input
11848 If any of these options is used, then the linker is not run, and
11849 object file names should not be used as arguments. @xref{Overall
11853 @opindex fuse-ld=bfd
11854 Use the @command{bfd} linker instead of the default linker.
11856 @item -fuse-ld=gold
11857 @opindex fuse-ld=gold
11858 Use the @command{gold} linker instead of the default linker.
11861 @item -l@var{library}
11862 @itemx -l @var{library}
11864 Search the library named @var{library} when linking. (The second
11865 alternative with the library as a separate argument is only for
11866 POSIX compliance and is not recommended.)
11868 It makes a difference where in the command you write this option; the
11869 linker searches and processes libraries and object files in the order they
11870 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11871 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11872 to functions in @samp{z}, those functions may not be loaded.
11874 The linker searches a standard list of directories for the library,
11875 which is actually a file named @file{lib@var{library}.a}. The linker
11876 then uses this file as if it had been specified precisely by name.
11878 The directories searched include several standard system directories
11879 plus any that you specify with @option{-L}.
11881 Normally the files found this way are library files---archive files
11882 whose members are object files. The linker handles an archive file by
11883 scanning through it for members which define symbols that have so far
11884 been referenced but not defined. But if the file that is found is an
11885 ordinary object file, it is linked in the usual fashion. The only
11886 difference between using an @option{-l} option and specifying a file name
11887 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11888 and searches several directories.
11892 You need this special case of the @option{-l} option in order to
11893 link an Objective-C or Objective-C++ program.
11895 @item -nostartfiles
11896 @opindex nostartfiles
11897 Do not use the standard system startup files when linking.
11898 The standard system libraries are used normally, unless @option{-nostdlib}
11899 or @option{-nodefaultlibs} is used.
11901 @item -nodefaultlibs
11902 @opindex nodefaultlibs
11903 Do not use the standard system libraries when linking.
11904 Only the libraries you specify are passed to the linker, and options
11905 specifying linkage of the system libraries, such as @option{-static-libgcc}
11906 or @option{-shared-libgcc}, are ignored.
11907 The standard startup files are used normally, unless @option{-nostartfiles}
11910 The compiler may generate calls to @code{memcmp},
11911 @code{memset}, @code{memcpy} and @code{memmove}.
11912 These entries are usually resolved by entries in
11913 libc. These entry points should be supplied through some other
11914 mechanism when this option is specified.
11918 Do not use the standard system startup files or libraries when linking.
11919 No startup files and only the libraries you specify are passed to
11920 the linker, and options specifying linkage of the system libraries, such as
11921 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11923 The compiler may generate calls to @code{memcmp}, @code{memset},
11924 @code{memcpy} and @code{memmove}.
11925 These entries are usually resolved by entries in
11926 libc. These entry points should be supplied through some other
11927 mechanism when this option is specified.
11929 @cindex @option{-lgcc}, use with @option{-nostdlib}
11930 @cindex @option{-nostdlib} and unresolved references
11931 @cindex unresolved references and @option{-nostdlib}
11932 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11933 @cindex @option{-nodefaultlibs} and unresolved references
11934 @cindex unresolved references and @option{-nodefaultlibs}
11935 One of the standard libraries bypassed by @option{-nostdlib} and
11936 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11937 which GCC uses to overcome shortcomings of particular machines, or special
11938 needs for some languages.
11939 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11940 Collection (GCC) Internals},
11941 for more discussion of @file{libgcc.a}.)
11942 In most cases, you need @file{libgcc.a} even when you want to avoid
11943 other standard libraries. In other words, when you specify @option{-nostdlib}
11944 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11945 This ensures that you have no unresolved references to internal GCC
11946 library subroutines.
11947 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11948 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11949 GNU Compiler Collection (GCC) Internals}.)
11953 Produce a dynamically linked position independent executable on targets
11954 that support it. For predictable results, you must also specify the same
11955 set of options used for compilation (@option{-fpie}, @option{-fPIE},
11956 or model suboptions) when you specify this linker option.
11960 Don't produce a dynamically linked position independent executable.
11963 @opindex static-pie
11964 Produce a static position independent executable on targets that support
11965 it. A static position independent executable is similar to a static
11966 executable, but can be loaded at any address without a dynamic linker.
11967 For predictable results, you must also specify the same set of options
11968 used for compilation (@option{-fpie}, @option{-fPIE}, or model
11969 suboptions) when you specify this linker option.
11973 Link with the POSIX threads library. This option is supported on
11974 GNU/Linux targets, most other Unix derivatives, and also on
11975 x86 Cygwin and MinGW targets. On some targets this option also sets
11976 flags for the preprocessor, so it should be used consistently for both
11977 compilation and linking.
11981 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11982 that support it. This instructs the linker to add all symbols, not
11983 only used ones, to the dynamic symbol table. This option is needed
11984 for some uses of @code{dlopen} or to allow obtaining backtraces
11985 from within a program.
11989 Remove all symbol table and relocation information from the executable.
11993 On systems that support dynamic linking, this overrides @option{-pie}
11994 and prevents linking with the shared libraries. On other systems, this
11995 option has no effect.
11999 Produce a shared object which can then be linked with other objects to
12000 form an executable. Not all systems support this option. For predictable
12001 results, you must also specify the same set of options used for compilation
12002 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
12003 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
12004 needs to build supplementary stub code for constructors to work. On
12005 multi-libbed systems, @samp{gcc -shared} must select the correct support
12006 libraries to link against. Failing to supply the correct flags may lead
12007 to subtle defects. Supplying them in cases where they are not necessary
12010 @item -shared-libgcc
12011 @itemx -static-libgcc
12012 @opindex shared-libgcc
12013 @opindex static-libgcc
12014 On systems that provide @file{libgcc} as a shared library, these options
12015 force the use of either the shared or static version, respectively.
12016 If no shared version of @file{libgcc} was built when the compiler was
12017 configured, these options have no effect.
12019 There are several situations in which an application should use the
12020 shared @file{libgcc} instead of the static version. The most common
12021 of these is when the application wishes to throw and catch exceptions
12022 across different shared libraries. In that case, each of the libraries
12023 as well as the application itself should use the shared @file{libgcc}.
12025 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
12026 whenever you build a shared library or a main executable, because C++
12027 programs typically use exceptions, so this is the right thing to do.
12029 If, instead, you use the GCC driver to create shared libraries, you may
12030 find that they are not always linked with the shared @file{libgcc}.
12031 If GCC finds, at its configuration time, that you have a non-GNU linker
12032 or a GNU linker that does not support option @option{--eh-frame-hdr},
12033 it links the shared version of @file{libgcc} into shared libraries
12034 by default. Otherwise, it takes advantage of the linker and optimizes
12035 away the linking with the shared version of @file{libgcc}, linking with
12036 the static version of libgcc by default. This allows exceptions to
12037 propagate through such shared libraries, without incurring relocation
12038 costs at library load time.
12040 However, if a library or main executable is supposed to throw or catch
12041 exceptions, you must link it using the G++ driver, as appropriate
12042 for the languages used in the program, or using the option
12043 @option{-shared-libgcc}, such that it is linked with the shared
12046 @item -static-libasan
12047 @opindex static-libasan
12048 When the @option{-fsanitize=address} option is used to link a program,
12049 the GCC driver automatically links against @option{libasan}. If
12050 @file{libasan} is available as a shared library, and the @option{-static}
12051 option is not used, then this links against the shared version of
12052 @file{libasan}. The @option{-static-libasan} option directs the GCC
12053 driver to link @file{libasan} statically, without necessarily linking
12054 other libraries statically.
12056 @item -static-libtsan
12057 @opindex static-libtsan
12058 When the @option{-fsanitize=thread} option is used to link a program,
12059 the GCC driver automatically links against @option{libtsan}. If
12060 @file{libtsan} is available as a shared library, and the @option{-static}
12061 option is not used, then this links against the shared version of
12062 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
12063 driver to link @file{libtsan} statically, without necessarily linking
12064 other libraries statically.
12066 @item -static-liblsan
12067 @opindex static-liblsan
12068 When the @option{-fsanitize=leak} option is used to link a program,
12069 the GCC driver automatically links against @option{liblsan}. If
12070 @file{liblsan} is available as a shared library, and the @option{-static}
12071 option is not used, then this links against the shared version of
12072 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
12073 driver to link @file{liblsan} statically, without necessarily linking
12074 other libraries statically.
12076 @item -static-libubsan
12077 @opindex static-libubsan
12078 When the @option{-fsanitize=undefined} option is used to link a program,
12079 the GCC driver automatically links against @option{libubsan}. If
12080 @file{libubsan} is available as a shared library, and the @option{-static}
12081 option is not used, then this links against the shared version of
12082 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
12083 driver to link @file{libubsan} statically, without necessarily linking
12084 other libraries statically.
12086 @item -static-libmpx
12087 @opindex static-libmpx
12088 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
12089 used to link a program, the GCC driver automatically links against
12090 @file{libmpx}. If @file{libmpx} is available as a shared library,
12091 and the @option{-static} option is not used, then this links against
12092 the shared version of @file{libmpx}. The @option{-static-libmpx}
12093 option directs the GCC driver to link @file{libmpx} statically,
12094 without necessarily linking other libraries statically.
12096 @item -static-libmpxwrappers
12097 @opindex static-libmpxwrappers
12098 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
12099 to link a program without also using @option{-fno-chkp-use-wrappers}, the
12100 GCC driver automatically links against @file{libmpxwrappers}. If
12101 @file{libmpxwrappers} is available as a shared library, and the
12102 @option{-static} option is not used, then this links against the shared
12103 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
12104 option directs the GCC driver to link @file{libmpxwrappers} statically,
12105 without necessarily linking other libraries statically.
12107 @item -static-libstdc++
12108 @opindex static-libstdc++
12109 When the @command{g++} program is used to link a C++ program, it
12110 normally automatically links against @option{libstdc++}. If
12111 @file{libstdc++} is available as a shared library, and the
12112 @option{-static} option is not used, then this links against the
12113 shared version of @file{libstdc++}. That is normally fine. However, it
12114 is sometimes useful to freeze the version of @file{libstdc++} used by
12115 the program without going all the way to a fully static link. The
12116 @option{-static-libstdc++} option directs the @command{g++} driver to
12117 link @file{libstdc++} statically, without necessarily linking other
12118 libraries statically.
12122 Bind references to global symbols when building a shared object. Warn
12123 about any unresolved references (unless overridden by the link editor
12124 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
12127 @item -T @var{script}
12129 @cindex linker script
12130 Use @var{script} as the linker script. This option is supported by most
12131 systems using the GNU linker. On some targets, such as bare-board
12132 targets without an operating system, the @option{-T} option may be required
12133 when linking to avoid references to undefined symbols.
12135 @item -Xlinker @var{option}
12137 Pass @var{option} as an option to the linker. You can use this to
12138 supply system-specific linker options that GCC does not recognize.
12140 If you want to pass an option that takes a separate argument, you must use
12141 @option{-Xlinker} twice, once for the option and once for the argument.
12142 For example, to pass @option{-assert definitions}, you must write
12143 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
12144 @option{-Xlinker "-assert definitions"}, because this passes the entire
12145 string as a single argument, which is not what the linker expects.
12147 When using the GNU linker, it is usually more convenient to pass
12148 arguments to linker options using the @option{@var{option}=@var{value}}
12149 syntax than as separate arguments. For example, you can specify
12150 @option{-Xlinker -Map=output.map} rather than
12151 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
12152 this syntax for command-line options.
12154 @item -Wl,@var{option}
12156 Pass @var{option} as an option to the linker. If @var{option} contains
12157 commas, it is split into multiple options at the commas. You can use this
12158 syntax to pass an argument to the option.
12159 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
12160 linker. When using the GNU linker, you can also get the same effect with
12161 @option{-Wl,-Map=output.map}.
12163 @item -u @var{symbol}
12165 Pretend the symbol @var{symbol} is undefined, to force linking of
12166 library modules to define it. You can use @option{-u} multiple times with
12167 different symbols to force loading of additional library modules.
12169 @item -z @var{keyword}
12171 @option{-z} is passed directly on to the linker along with the keyword
12172 @var{keyword}. See the section in the documentation of your linker for
12173 permitted values and their meanings.
12176 @node Directory Options
12177 @section Options for Directory Search
12178 @cindex directory options
12179 @cindex options, directory search
12180 @cindex search path
12182 These options specify directories to search for header files, for
12183 libraries and for parts of the compiler:
12186 @include cppdiropts.texi
12188 @item -iplugindir=@var{dir}
12189 @opindex iplugindir=
12190 Set the directory to search for plugins that are passed
12191 by @option{-fplugin=@var{name}} instead of
12192 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
12193 to be used by the user, but only passed by the driver.
12197 Add directory @var{dir} to the list of directories to be searched
12200 @item -B@var{prefix}
12202 This option specifies where to find the executables, libraries,
12203 include files, and data files of the compiler itself.
12205 The compiler driver program runs one or more of the subprograms
12206 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
12207 @var{prefix} as a prefix for each program it tries to run, both with and
12208 without @samp{@var{machine}/@var{version}/} for the corresponding target
12209 machine and compiler version.
12211 For each subprogram to be run, the compiler driver first tries the
12212 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
12213 is not specified, the driver tries two standard prefixes,
12214 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
12215 those results in a file name that is found, the unmodified program
12216 name is searched for using the directories specified in your
12217 @env{PATH} environment variable.
12219 The compiler checks to see if the path provided by @option{-B}
12220 refers to a directory, and if necessary it adds a directory
12221 separator character at the end of the path.
12223 @option{-B} prefixes that effectively specify directory names also apply
12224 to libraries in the linker, because the compiler translates these
12225 options into @option{-L} options for the linker. They also apply to
12226 include files in the preprocessor, because the compiler translates these
12227 options into @option{-isystem} options for the preprocessor. In this case,
12228 the compiler appends @samp{include} to the prefix.
12230 The runtime support file @file{libgcc.a} can also be searched for using
12231 the @option{-B} prefix, if needed. If it is not found there, the two
12232 standard prefixes above are tried, and that is all. The file is left
12233 out of the link if it is not found by those means.
12235 Another way to specify a prefix much like the @option{-B} prefix is to use
12236 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
12239 As a special kludge, if the path provided by @option{-B} is
12240 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
12241 9, then it is replaced by @file{[dir/]include}. This is to help
12242 with boot-strapping the compiler.
12244 @item -no-canonical-prefixes
12245 @opindex no-canonical-prefixes
12246 Do not expand any symbolic links, resolve references to @samp{/../}
12247 or @samp{/./}, or make the path absolute when generating a relative
12250 @item --sysroot=@var{dir}
12252 Use @var{dir} as the logical root directory for headers and libraries.
12253 For example, if the compiler normally searches for headers in
12254 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
12255 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
12257 If you use both this option and the @option{-isysroot} option, then
12258 the @option{--sysroot} option applies to libraries, but the
12259 @option{-isysroot} option applies to header files.
12261 The GNU linker (beginning with version 2.16) has the necessary support
12262 for this option. If your linker does not support this option, the
12263 header file aspect of @option{--sysroot} still works, but the
12264 library aspect does not.
12266 @item --no-sysroot-suffix
12267 @opindex no-sysroot-suffix
12268 For some targets, a suffix is added to the root directory specified
12269 with @option{--sysroot}, depending on the other options used, so that
12270 headers may for example be found in
12271 @file{@var{dir}/@var{suffix}/usr/include} instead of
12272 @file{@var{dir}/usr/include}. This option disables the addition of
12277 @node Code Gen Options
12278 @section Options for Code Generation Conventions
12279 @cindex code generation conventions
12280 @cindex options, code generation
12281 @cindex run-time options
12283 These machine-independent options control the interface conventions
12284 used in code generation.
12286 Most of them have both positive and negative forms; the negative form
12287 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
12288 one of the forms is listed---the one that is not the default. You
12289 can figure out the other form by either removing @samp{no-} or adding
12293 @item -fstack-reuse=@var{reuse-level}
12294 @opindex fstack_reuse
12295 This option controls stack space reuse for user declared local/auto variables
12296 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
12297 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
12298 local variables and temporaries, @samp{named_vars} enables the reuse only for
12299 user defined local variables with names, and @samp{none} disables stack reuse
12300 completely. The default value is @samp{all}. The option is needed when the
12301 program extends the lifetime of a scoped local variable or a compiler generated
12302 temporary beyond the end point defined by the language. When a lifetime of
12303 a variable ends, and if the variable lives in memory, the optimizing compiler
12304 has the freedom to reuse its stack space with other temporaries or scoped
12305 local variables whose live range does not overlap with it. Legacy code extending
12306 local lifetime is likely to break with the stack reuse optimization.
12325 if (*p == 10) // out of scope use of local1
12336 A(int k) : i(k), j(k) @{ @}
12343 void foo(const A& ar)
12350 foo(A(10)); // temp object's lifetime ends when foo returns
12356 ap->i+= 10; // ap references out of scope temp whose space
12357 // is reused with a. What is the value of ap->i?
12362 The lifetime of a compiler generated temporary is well defined by the C++
12363 standard. When a lifetime of a temporary ends, and if the temporary lives
12364 in memory, the optimizing compiler has the freedom to reuse its stack
12365 space with other temporaries or scoped local variables whose live range
12366 does not overlap with it. However some of the legacy code relies on
12367 the behavior of older compilers in which temporaries' stack space is
12368 not reused, the aggressive stack reuse can lead to runtime errors. This
12369 option is used to control the temporary stack reuse optimization.
12373 This option generates traps for signed overflow on addition, subtraction,
12374 multiplication operations.
12375 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12376 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12377 @option{-fwrapv} being effective. Note that only active options override, so
12378 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12379 results in @option{-ftrapv} being effective.
12383 This option instructs the compiler to assume that signed arithmetic
12384 overflow of addition, subtraction and multiplication wraps around
12385 using twos-complement representation. This flag enables some optimizations
12386 and disables others.
12387 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12388 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12389 @option{-fwrapv} being effective. Note that only active options override, so
12390 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12391 results in @option{-ftrapv} being effective.
12394 @opindex fexceptions
12395 Enable exception handling. Generates extra code needed to propagate
12396 exceptions. For some targets, this implies GCC generates frame
12397 unwind information for all functions, which can produce significant data
12398 size overhead, although it does not affect execution. If you do not
12399 specify this option, GCC enables it by default for languages like
12400 C++ that normally require exception handling, and disables it for
12401 languages like C that do not normally require it. However, you may need
12402 to enable this option when compiling C code that needs to interoperate
12403 properly with exception handlers written in C++. You may also wish to
12404 disable this option if you are compiling older C++ programs that don't
12405 use exception handling.
12407 @item -fnon-call-exceptions
12408 @opindex fnon-call-exceptions
12409 Generate code that allows trapping instructions to throw exceptions.
12410 Note that this requires platform-specific runtime support that does
12411 not exist everywhere. Moreover, it only allows @emph{trapping}
12412 instructions to throw exceptions, i.e.@: memory references or floating-point
12413 instructions. It does not allow exceptions to be thrown from
12414 arbitrary signal handlers such as @code{SIGALRM}.
12416 @item -fdelete-dead-exceptions
12417 @opindex fdelete-dead-exceptions
12418 Consider that instructions that may throw exceptions but don't otherwise
12419 contribute to the execution of the program can be optimized away.
12420 This option is enabled by default for the Ada front end, as permitted by
12421 the Ada language specification.
12422 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12424 @item -funwind-tables
12425 @opindex funwind-tables
12426 Similar to @option{-fexceptions}, except that it just generates any needed
12427 static data, but does not affect the generated code in any other way.
12428 You normally do not need to enable this option; instead, a language processor
12429 that needs this handling enables it on your behalf.
12431 @item -fasynchronous-unwind-tables
12432 @opindex fasynchronous-unwind-tables
12433 Generate unwind table in DWARF format, if supported by target machine. The
12434 table is exact at each instruction boundary, so it can be used for stack
12435 unwinding from asynchronous events (such as debugger or garbage collector).
12437 @item -fno-gnu-unique
12438 @opindex fno-gnu-unique
12439 On systems with recent GNU assembler and C library, the C++ compiler
12440 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12441 of template static data members and static local variables in inline
12442 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12443 is necessary to avoid problems with a library used by two different
12444 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12445 therefore disagreeing with the other one about the binding of the
12446 symbol. But this causes @code{dlclose} to be ignored for affected
12447 DSOs; if your program relies on reinitialization of a DSO via
12448 @code{dlclose} and @code{dlopen}, you can use
12449 @option{-fno-gnu-unique}.
12451 @item -fpcc-struct-return
12452 @opindex fpcc-struct-return
12453 Return ``short'' @code{struct} and @code{union} values in memory like
12454 longer ones, rather than in registers. This convention is less
12455 efficient, but it has the advantage of allowing intercallability between
12456 GCC-compiled files and files compiled with other compilers, particularly
12457 the Portable C Compiler (pcc).
12459 The precise convention for returning structures in memory depends
12460 on the target configuration macros.
12462 Short structures and unions are those whose size and alignment match
12463 that of some integer type.
12465 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12466 switch is not binary compatible with code compiled with the
12467 @option{-freg-struct-return} switch.
12468 Use it to conform to a non-default application binary interface.
12470 @item -freg-struct-return
12471 @opindex freg-struct-return
12472 Return @code{struct} and @code{union} values in registers when possible.
12473 This is more efficient for small structures than
12474 @option{-fpcc-struct-return}.
12476 If you specify neither @option{-fpcc-struct-return} nor
12477 @option{-freg-struct-return}, GCC defaults to whichever convention is
12478 standard for the target. If there is no standard convention, GCC
12479 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12480 the principal compiler. In those cases, we can choose the standard, and
12481 we chose the more efficient register return alternative.
12483 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12484 switch is not binary compatible with code compiled with the
12485 @option{-fpcc-struct-return} switch.
12486 Use it to conform to a non-default application binary interface.
12488 @item -fshort-enums
12489 @opindex fshort-enums
12490 Allocate to an @code{enum} type only as many bytes as it needs for the
12491 declared range of possible values. Specifically, the @code{enum} type
12492 is equivalent to the smallest integer type that has enough room.
12494 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12495 code that is not binary compatible with code generated without that switch.
12496 Use it to conform to a non-default application binary interface.
12498 @item -fshort-wchar
12499 @opindex fshort-wchar
12500 Override the underlying type for @code{wchar_t} to be @code{short
12501 unsigned int} instead of the default for the target. This option is
12502 useful for building programs to run under WINE@.
12504 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12505 code that is not binary compatible with code generated without that switch.
12506 Use it to conform to a non-default application binary interface.
12509 @opindex fno-common
12510 @cindex tentative definitions
12511 In C code, this option controls the placement of global variables
12512 defined without an initializer, known as @dfn{tentative definitions}
12513 in the C standard. Tentative definitions are distinct from declarations
12514 of a variable with the @code{extern} keyword, which do not allocate storage.
12516 Unix C compilers have traditionally allocated storage for
12517 uninitialized global variables in a common block. This allows the
12518 linker to resolve all tentative definitions of the same variable
12519 in different compilation units to the same object, or to a non-tentative
12521 This is the behavior specified by @option{-fcommon}, and is the default for
12522 GCC on most targets.
12523 On the other hand, this behavior is not required by ISO
12524 C, and on some targets may carry a speed or code size penalty on
12525 variable references.
12527 The @option{-fno-common} option specifies that the compiler should instead
12528 place uninitialized global variables in the data section of the object file.
12529 This inhibits the merging of tentative definitions by the linker so
12530 you get a multiple-definition error if the same
12531 variable is defined in more than one compilation unit.
12532 Compiling with @option{-fno-common} is useful on targets for which
12533 it provides better performance, or if you wish to verify that the
12534 program will work on other systems that always treat uninitialized
12535 variable definitions this way.
12539 Ignore the @code{#ident} directive.
12541 @item -finhibit-size-directive
12542 @opindex finhibit-size-directive
12543 Don't output a @code{.size} assembler directive, or anything else that
12544 would cause trouble if the function is split in the middle, and the
12545 two halves are placed at locations far apart in memory. This option is
12546 used when compiling @file{crtstuff.c}; you should not need to use it
12549 @item -fverbose-asm
12550 @opindex fverbose-asm
12551 Put extra commentary information in the generated assembly code to
12552 make it more readable. This option is generally only of use to those
12553 who actually need to read the generated assembly code (perhaps while
12554 debugging the compiler itself).
12556 @option{-fno-verbose-asm}, the default, causes the
12557 extra information to be omitted and is useful when comparing two assembler
12560 The added comments include:
12565 information on the compiler version and command-line options,
12568 the source code lines associated with the assembly instructions,
12569 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12572 hints on which high-level expressions correspond to
12573 the various assembly instruction operands.
12577 For example, given this C source file:
12585 for (i = 0; i < n; i++)
12592 compiling to (x86_64) assembly via @option{-S} and emitting the result
12593 direct to stdout via @option{-o} @option{-}
12596 gcc -S test.c -fverbose-asm -Os -o -
12599 gives output similar to this:
12603 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12610 .type test, @@function
12614 # test.c:4: int total = 0;
12615 xorl %eax, %eax # <retval>
12616 # test.c:6: for (i = 0; i < n; i++)
12617 xorl %edx, %edx # i
12619 # test.c:6: for (i = 0; i < n; i++)
12620 cmpl %edi, %edx # n, i
12622 # test.c:7: total += i * i;
12623 movl %edx, %ecx # i, tmp92
12624 imull %edx, %ecx # i, tmp92
12625 # test.c:6: for (i = 0; i < n; i++)
12627 # test.c:7: total += i * i;
12628 addl %ecx, %eax # tmp92, <retval>
12636 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12637 .section .note.GNU-stack,"",@@progbits
12640 The comments are intended for humans rather than machines and hence the
12641 precise format of the comments is subject to change.
12643 @item -frecord-gcc-switches
12644 @opindex frecord-gcc-switches
12645 This switch causes the command line used to invoke the
12646 compiler to be recorded into the object file that is being created.
12647 This switch is only implemented on some targets and the exact format
12648 of the recording is target and binary file format dependent, but it
12649 usually takes the form of a section containing ASCII text. This
12650 switch is related to the @option{-fverbose-asm} switch, but that
12651 switch only records information in the assembler output file as
12652 comments, so it never reaches the object file.
12653 See also @option{-grecord-gcc-switches} for another
12654 way of storing compiler options into the object file.
12658 @cindex global offset table
12660 Generate position-independent code (PIC) suitable for use in a shared
12661 library, if supported for the target machine. Such code accesses all
12662 constant addresses through a global offset table (GOT)@. The dynamic
12663 loader resolves the GOT entries when the program starts (the dynamic
12664 loader is not part of GCC; it is part of the operating system). If
12665 the GOT size for the linked executable exceeds a machine-specific
12666 maximum size, you get an error message from the linker indicating that
12667 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12668 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12669 on the m68k and RS/6000. The x86 has no such limit.)
12671 Position-independent code requires special support, and therefore works
12672 only on certain machines. For the x86, GCC supports PIC for System V
12673 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12674 position-independent.
12676 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12681 If supported for the target machine, emit position-independent code,
12682 suitable for dynamic linking and avoiding any limit on the size of the
12683 global offset table. This option makes a difference on AArch64, m68k,
12684 PowerPC and SPARC@.
12686 Position-independent code requires special support, and therefore works
12687 only on certain machines.
12689 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12696 These options are similar to @option{-fpic} and @option{-fPIC}, but
12697 generated position independent code can be only linked into executables.
12698 Usually these options are used when @option{-pie} GCC option is
12699 used during linking.
12701 @option{-fpie} and @option{-fPIE} both define the macros
12702 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12703 for @option{-fpie} and 2 for @option{-fPIE}.
12707 Do not use the PLT for external function calls in position-independent code.
12708 Instead, load the callee address at call sites from the GOT and branch to it.
12709 This leads to more efficient code by eliminating PLT stubs and exposing
12710 GOT loads to optimizations. On architectures such as 32-bit x86 where
12711 PLT stubs expect the GOT pointer in a specific register, this gives more
12712 register allocation freedom to the compiler.
12713 Lazy binding requires use of the PLT;
12714 with @option{-fno-plt} all external symbols are resolved at load time.
12716 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12717 through the PLT for specific external functions.
12719 In position-dependent code, a few targets also convert calls to
12720 functions that are marked to not use the PLT to use the GOT instead.
12722 @item -fno-jump-tables
12723 @opindex fno-jump-tables
12724 Do not use jump tables for switch statements even where it would be
12725 more efficient than other code generation strategies. This option is
12726 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12727 building code that forms part of a dynamic linker and cannot
12728 reference the address of a jump table. On some targets, jump tables
12729 do not require a GOT and this option is not needed.
12731 @item -ffixed-@var{reg}
12733 Treat the register named @var{reg} as a fixed register; generated code
12734 should never refer to it (except perhaps as a stack pointer, frame
12735 pointer or in some other fixed role).
12737 @var{reg} must be the name of a register. The register names accepted
12738 are machine-specific and are defined in the @code{REGISTER_NAMES}
12739 macro in the machine description macro file.
12741 This flag does not have a negative form, because it specifies a
12744 @item -fcall-used-@var{reg}
12745 @opindex fcall-used
12746 Treat the register named @var{reg} as an allocable register that is
12747 clobbered by function calls. It may be allocated for temporaries or
12748 variables that do not live across a call. Functions compiled this way
12749 do not save and restore the register @var{reg}.
12751 It is an error to use this flag with the frame pointer or stack pointer.
12752 Use of this flag for other registers that have fixed pervasive roles in
12753 the machine's execution model produces disastrous results.
12755 This flag does not have a negative form, because it specifies a
12758 @item -fcall-saved-@var{reg}
12759 @opindex fcall-saved
12760 Treat the register named @var{reg} as an allocable register saved by
12761 functions. It may be allocated even for temporaries or variables that
12762 live across a call. Functions compiled this way save and restore
12763 the register @var{reg} if they use it.
12765 It is an error to use this flag with the frame pointer or stack pointer.
12766 Use of this flag for other registers that have fixed pervasive roles in
12767 the machine's execution model produces disastrous results.
12769 A different sort of disaster results from the use of this flag for
12770 a register in which function values may be returned.
12772 This flag does not have a negative form, because it specifies a
12775 @item -fpack-struct[=@var{n}]
12776 @opindex fpack-struct
12777 Without a value specified, pack all structure members together without
12778 holes. When a value is specified (which must be a small power of two), pack
12779 structure members according to this value, representing the maximum
12780 alignment (that is, objects with default alignment requirements larger than
12781 this are output potentially unaligned at the next fitting location.
12783 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12784 code that is not binary compatible with code generated without that switch.
12785 Additionally, it makes the code suboptimal.
12786 Use it to conform to a non-default application binary interface.
12788 @item -fleading-underscore
12789 @opindex fleading-underscore
12790 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12791 change the way C symbols are represented in the object file. One use
12792 is to help link with legacy assembly code.
12794 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12795 generate code that is not binary compatible with code generated without that
12796 switch. Use it to conform to a non-default application binary interface.
12797 Not all targets provide complete support for this switch.
12799 @item -ftls-model=@var{model}
12800 @opindex ftls-model
12801 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12802 The @var{model} argument should be one of @samp{global-dynamic},
12803 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12804 Note that the choice is subject to optimization: the compiler may use
12805 a more efficient model for symbols not visible outside of the translation
12806 unit, or if @option{-fpic} is not given on the command line.
12808 The default without @option{-fpic} is @samp{initial-exec}; with
12809 @option{-fpic} the default is @samp{global-dynamic}.
12811 @item -ftrampolines
12812 @opindex ftrampolines
12813 For targets that normally need trampolines for nested functions, always
12814 generate them instead of using descriptors. Otherwise, for targets that
12815 do not need them, like for example HP-PA or IA-64, do nothing.
12817 A trampoline is a small piece of code that is created at run time on the
12818 stack when the address of a nested function is taken, and is used to call
12819 the nested function indirectly. Therefore, it requires the stack to be
12820 made executable in order for the program to work properly.
12822 @option{-fno-trampolines} is enabled by default on a language by language
12823 basis to let the compiler avoid generating them, if it computes that this
12824 is safe, and replace them with descriptors. Descriptors are made up of data
12825 only, but the generated code must be prepared to deal with them. As of this
12826 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12828 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12829 @option{-fno-trampolines} are not binary compatible if nested functions are
12830 present. This option must therefore be used on a program-wide basis and be
12831 manipulated with extreme care.
12833 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12834 @opindex fvisibility
12835 Set the default ELF image symbol visibility to the specified option---all
12836 symbols are marked with this unless overridden within the code.
12837 Using this feature can very substantially improve linking and
12838 load times of shared object libraries, produce more optimized
12839 code, provide near-perfect API export and prevent symbol clashes.
12840 It is @strong{strongly} recommended that you use this in any shared objects
12843 Despite the nomenclature, @samp{default} always means public; i.e.,
12844 available to be linked against from outside the shared object.
12845 @samp{protected} and @samp{internal} are pretty useless in real-world
12846 usage so the only other commonly used option is @samp{hidden}.
12847 The default if @option{-fvisibility} isn't specified is
12848 @samp{default}, i.e., make every symbol public.
12850 A good explanation of the benefits offered by ensuring ELF
12851 symbols have the correct visibility is given by ``How To Write
12852 Shared Libraries'' by Ulrich Drepper (which can be found at
12853 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
12854 solution made possible by this option to marking things hidden when
12855 the default is public is to make the default hidden and mark things
12856 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12857 and @code{__attribute__ ((visibility("default")))} instead of
12858 @code{__declspec(dllexport)} you get almost identical semantics with
12859 identical syntax. This is a great boon to those working with
12860 cross-platform projects.
12862 For those adding visibility support to existing code, you may find
12863 @code{#pragma GCC visibility} of use. This works by you enclosing
12864 the declarations you wish to set visibility for with (for example)
12865 @code{#pragma GCC visibility push(hidden)} and
12866 @code{#pragma GCC visibility pop}.
12867 Bear in mind that symbol visibility should be viewed @strong{as
12868 part of the API interface contract} and thus all new code should
12869 always specify visibility when it is not the default; i.e., declarations
12870 only for use within the local DSO should @strong{always} be marked explicitly
12871 as hidden as so to avoid PLT indirection overheads---making this
12872 abundantly clear also aids readability and self-documentation of the code.
12873 Note that due to ISO C++ specification requirements, @code{operator new} and
12874 @code{operator delete} must always be of default visibility.
12876 Be aware that headers from outside your project, in particular system
12877 headers and headers from any other library you use, may not be
12878 expecting to be compiled with visibility other than the default. You
12879 may need to explicitly say @code{#pragma GCC visibility push(default)}
12880 before including any such headers.
12882 @code{extern} declarations are not affected by @option{-fvisibility}, so
12883 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12884 no modifications. However, this means that calls to @code{extern}
12885 functions with no explicit visibility use the PLT, so it is more
12886 effective to use @code{__attribute ((visibility))} and/or
12887 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12888 declarations should be treated as hidden.
12890 Note that @option{-fvisibility} does affect C++ vague linkage
12891 entities. This means that, for instance, an exception class that is
12892 be thrown between DSOs must be explicitly marked with default
12893 visibility so that the @samp{type_info} nodes are unified between
12896 An overview of these techniques, their benefits and how to use them
12897 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12899 @item -fstrict-volatile-bitfields
12900 @opindex fstrict-volatile-bitfields
12901 This option should be used if accesses to volatile bit-fields (or other
12902 structure fields, although the compiler usually honors those types
12903 anyway) should use a single access of the width of the
12904 field's type, aligned to a natural alignment if possible. For
12905 example, targets with memory-mapped peripheral registers might require
12906 all such accesses to be 16 bits wide; with this flag you can
12907 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12908 is 16 bits on these targets) to force GCC to use 16-bit accesses
12909 instead of, perhaps, a more efficient 32-bit access.
12911 If this option is disabled, the compiler uses the most efficient
12912 instruction. In the previous example, that might be a 32-bit load
12913 instruction, even though that accesses bytes that do not contain
12914 any portion of the bit-field, or memory-mapped registers unrelated to
12915 the one being updated.
12917 In some cases, such as when the @code{packed} attribute is applied to a
12918 structure field, it may not be possible to access the field with a single
12919 read or write that is correctly aligned for the target machine. In this
12920 case GCC falls back to generating multiple accesses rather than code that
12921 will fault or truncate the result at run time.
12923 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12924 not allowed to touch non bit-field members. It is therefore recommended
12925 to define all bits of the field's type as bit-field members.
12927 The default value of this option is determined by the application binary
12928 interface for the target processor.
12930 @item -fsync-libcalls
12931 @opindex fsync-libcalls
12932 This option controls whether any out-of-line instance of the @code{__sync}
12933 family of functions may be used to implement the C++11 @code{__atomic}
12934 family of functions.
12936 The default value of this option is enabled, thus the only useful form
12937 of the option is @option{-fno-sync-libcalls}. This option is used in
12938 the implementation of the @file{libatomic} runtime library.
12942 @node Developer Options
12943 @section GCC Developer Options
12944 @cindex developer options
12945 @cindex debugging GCC
12946 @cindex debug dump options
12947 @cindex dump options
12948 @cindex compilation statistics
12950 This section describes command-line options that are primarily of
12951 interest to GCC developers, including options to support compiler
12952 testing and investigation of compiler bugs and compile-time
12953 performance problems. This includes options that produce debug dumps
12954 at various points in the compilation; that print statistics such as
12955 memory use and execution time; and that print information about GCC's
12956 configuration, such as where it searches for libraries. You should
12957 rarely need to use any of these options for ordinary compilation and
12962 @item -d@var{letters}
12963 @itemx -fdump-rtl-@var{pass}
12964 @itemx -fdump-rtl-@var{pass}=@var{filename}
12966 @opindex fdump-rtl-@var{pass}
12967 Says to make debugging dumps during compilation at times specified by
12968 @var{letters}. This is used for debugging the RTL-based passes of the
12969 compiler. The file names for most of the dumps are made by appending
12970 a pass number and a word to the @var{dumpname}, and the files are
12971 created in the directory of the output file. In case of
12972 @option{=@var{filename}} option, the dump is output on the given file
12973 instead of the pass numbered dump files. Note that the pass number is
12974 assigned as passes are registered into the pass manager. Most passes
12975 are registered in the order that they will execute and for these passes
12976 the number corresponds to the pass execution order. However, passes
12977 registered by plugins, passes specific to compilation targets, or
12978 passes that are otherwise registered after all the other passes are
12979 numbered higher than a pass named "final", even if they are executed
12980 earlier. @var{dumpname} is generated from the name of the output
12981 file if explicitly specified and not an executable, otherwise it is
12982 the basename of the source file.
12984 Some @option{-d@var{letters}} switches have different meaning when
12985 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
12986 for information about preprocessor-specific dump options.
12988 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
12989 @option{-d} option @var{letters}. Here are the possible
12990 letters for use in @var{pass} and @var{letters}, and their meanings:
12994 @item -fdump-rtl-alignments
12995 @opindex fdump-rtl-alignments
12996 Dump after branch alignments have been computed.
12998 @item -fdump-rtl-asmcons
12999 @opindex fdump-rtl-asmcons
13000 Dump after fixing rtl statements that have unsatisfied in/out constraints.
13002 @item -fdump-rtl-auto_inc_dec
13003 @opindex fdump-rtl-auto_inc_dec
13004 Dump after auto-inc-dec discovery. This pass is only run on
13005 architectures that have auto inc or auto dec instructions.
13007 @item -fdump-rtl-barriers
13008 @opindex fdump-rtl-barriers
13009 Dump after cleaning up the barrier instructions.
13011 @item -fdump-rtl-bbpart
13012 @opindex fdump-rtl-bbpart
13013 Dump after partitioning hot and cold basic blocks.
13015 @item -fdump-rtl-bbro
13016 @opindex fdump-rtl-bbro
13017 Dump after block reordering.
13019 @item -fdump-rtl-btl1
13020 @itemx -fdump-rtl-btl2
13021 @opindex fdump-rtl-btl2
13022 @opindex fdump-rtl-btl2
13023 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
13024 after the two branch
13025 target load optimization passes.
13027 @item -fdump-rtl-bypass
13028 @opindex fdump-rtl-bypass
13029 Dump after jump bypassing and control flow optimizations.
13031 @item -fdump-rtl-combine
13032 @opindex fdump-rtl-combine
13033 Dump after the RTL instruction combination pass.
13035 @item -fdump-rtl-compgotos
13036 @opindex fdump-rtl-compgotos
13037 Dump after duplicating the computed gotos.
13039 @item -fdump-rtl-ce1
13040 @itemx -fdump-rtl-ce2
13041 @itemx -fdump-rtl-ce3
13042 @opindex fdump-rtl-ce1
13043 @opindex fdump-rtl-ce2
13044 @opindex fdump-rtl-ce3
13045 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
13046 @option{-fdump-rtl-ce3} enable dumping after the three
13047 if conversion passes.
13049 @item -fdump-rtl-cprop_hardreg
13050 @opindex fdump-rtl-cprop_hardreg
13051 Dump after hard register copy propagation.
13053 @item -fdump-rtl-csa
13054 @opindex fdump-rtl-csa
13055 Dump after combining stack adjustments.
13057 @item -fdump-rtl-cse1
13058 @itemx -fdump-rtl-cse2
13059 @opindex fdump-rtl-cse1
13060 @opindex fdump-rtl-cse2
13061 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
13062 the two common subexpression elimination passes.
13064 @item -fdump-rtl-dce
13065 @opindex fdump-rtl-dce
13066 Dump after the standalone dead code elimination passes.
13068 @item -fdump-rtl-dbr
13069 @opindex fdump-rtl-dbr
13070 Dump after delayed branch scheduling.
13072 @item -fdump-rtl-dce1
13073 @itemx -fdump-rtl-dce2
13074 @opindex fdump-rtl-dce1
13075 @opindex fdump-rtl-dce2
13076 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
13077 the two dead store elimination passes.
13079 @item -fdump-rtl-eh
13080 @opindex fdump-rtl-eh
13081 Dump after finalization of EH handling code.
13083 @item -fdump-rtl-eh_ranges
13084 @opindex fdump-rtl-eh_ranges
13085 Dump after conversion of EH handling range regions.
13087 @item -fdump-rtl-expand
13088 @opindex fdump-rtl-expand
13089 Dump after RTL generation.
13091 @item -fdump-rtl-fwprop1
13092 @itemx -fdump-rtl-fwprop2
13093 @opindex fdump-rtl-fwprop1
13094 @opindex fdump-rtl-fwprop2
13095 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
13096 dumping after the two forward propagation passes.
13098 @item -fdump-rtl-gcse1
13099 @itemx -fdump-rtl-gcse2
13100 @opindex fdump-rtl-gcse1
13101 @opindex fdump-rtl-gcse2
13102 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
13103 after global common subexpression elimination.
13105 @item -fdump-rtl-init-regs
13106 @opindex fdump-rtl-init-regs
13107 Dump after the initialization of the registers.
13109 @item -fdump-rtl-initvals
13110 @opindex fdump-rtl-initvals
13111 Dump after the computation of the initial value sets.
13113 @item -fdump-rtl-into_cfglayout
13114 @opindex fdump-rtl-into_cfglayout
13115 Dump after converting to cfglayout mode.
13117 @item -fdump-rtl-ira
13118 @opindex fdump-rtl-ira
13119 Dump after iterated register allocation.
13121 @item -fdump-rtl-jump
13122 @opindex fdump-rtl-jump
13123 Dump after the second jump optimization.
13125 @item -fdump-rtl-loop2
13126 @opindex fdump-rtl-loop2
13127 @option{-fdump-rtl-loop2} enables dumping after the rtl
13128 loop optimization passes.
13130 @item -fdump-rtl-mach
13131 @opindex fdump-rtl-mach
13132 Dump after performing the machine dependent reorganization pass, if that
13135 @item -fdump-rtl-mode_sw
13136 @opindex fdump-rtl-mode_sw
13137 Dump after removing redundant mode switches.
13139 @item -fdump-rtl-rnreg
13140 @opindex fdump-rtl-rnreg
13141 Dump after register renumbering.
13143 @item -fdump-rtl-outof_cfglayout
13144 @opindex fdump-rtl-outof_cfglayout
13145 Dump after converting from cfglayout mode.
13147 @item -fdump-rtl-peephole2
13148 @opindex fdump-rtl-peephole2
13149 Dump after the peephole pass.
13151 @item -fdump-rtl-postreload
13152 @opindex fdump-rtl-postreload
13153 Dump after post-reload optimizations.
13155 @item -fdump-rtl-pro_and_epilogue
13156 @opindex fdump-rtl-pro_and_epilogue
13157 Dump after generating the function prologues and epilogues.
13159 @item -fdump-rtl-sched1
13160 @itemx -fdump-rtl-sched2
13161 @opindex fdump-rtl-sched1
13162 @opindex fdump-rtl-sched2
13163 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
13164 after the basic block scheduling passes.
13166 @item -fdump-rtl-ree
13167 @opindex fdump-rtl-ree
13168 Dump after sign/zero extension elimination.
13170 @item -fdump-rtl-seqabstr
13171 @opindex fdump-rtl-seqabstr
13172 Dump after common sequence discovery.
13174 @item -fdump-rtl-shorten
13175 @opindex fdump-rtl-shorten
13176 Dump after shortening branches.
13178 @item -fdump-rtl-sibling
13179 @opindex fdump-rtl-sibling
13180 Dump after sibling call optimizations.
13182 @item -fdump-rtl-split1
13183 @itemx -fdump-rtl-split2
13184 @itemx -fdump-rtl-split3
13185 @itemx -fdump-rtl-split4
13186 @itemx -fdump-rtl-split5
13187 @opindex fdump-rtl-split1
13188 @opindex fdump-rtl-split2
13189 @opindex fdump-rtl-split3
13190 @opindex fdump-rtl-split4
13191 @opindex fdump-rtl-split5
13192 These options enable dumping after five rounds of
13193 instruction splitting.
13195 @item -fdump-rtl-sms
13196 @opindex fdump-rtl-sms
13197 Dump after modulo scheduling. This pass is only run on some
13200 @item -fdump-rtl-stack
13201 @opindex fdump-rtl-stack
13202 Dump after conversion from GCC's ``flat register file'' registers to the
13203 x87's stack-like registers. This pass is only run on x86 variants.
13205 @item -fdump-rtl-subreg1
13206 @itemx -fdump-rtl-subreg2
13207 @opindex fdump-rtl-subreg1
13208 @opindex fdump-rtl-subreg2
13209 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
13210 the two subreg expansion passes.
13212 @item -fdump-rtl-unshare
13213 @opindex fdump-rtl-unshare
13214 Dump after all rtl has been unshared.
13216 @item -fdump-rtl-vartrack
13217 @opindex fdump-rtl-vartrack
13218 Dump after variable tracking.
13220 @item -fdump-rtl-vregs
13221 @opindex fdump-rtl-vregs
13222 Dump after converting virtual registers to hard registers.
13224 @item -fdump-rtl-web
13225 @opindex fdump-rtl-web
13226 Dump after live range splitting.
13228 @item -fdump-rtl-regclass
13229 @itemx -fdump-rtl-subregs_of_mode_init
13230 @itemx -fdump-rtl-subregs_of_mode_finish
13231 @itemx -fdump-rtl-dfinit
13232 @itemx -fdump-rtl-dfinish
13233 @opindex fdump-rtl-regclass
13234 @opindex fdump-rtl-subregs_of_mode_init
13235 @opindex fdump-rtl-subregs_of_mode_finish
13236 @opindex fdump-rtl-dfinit
13237 @opindex fdump-rtl-dfinish
13238 These dumps are defined but always produce empty files.
13241 @itemx -fdump-rtl-all
13243 @opindex fdump-rtl-all
13244 Produce all the dumps listed above.
13248 Annotate the assembler output with miscellaneous debugging information.
13252 Dump all macro definitions, at the end of preprocessing, in addition to
13257 Produce a core dump whenever an error occurs.
13261 Annotate the assembler output with a comment indicating which
13262 pattern and alternative is used. The length of each instruction is
13267 Dump the RTL in the assembler output as a comment before each instruction.
13268 Also turns on @option{-dp} annotation.
13272 Just generate RTL for a function instead of compiling it. Usually used
13273 with @option{-fdump-rtl-expand}.
13276 @item -fdump-noaddr
13277 @opindex fdump-noaddr
13278 When doing debugging dumps, suppress address output. This makes it more
13279 feasible to use diff on debugging dumps for compiler invocations with
13280 different compiler binaries and/or different
13281 text / bss / data / heap / stack / dso start locations.
13284 @opindex freport-bug
13285 Collect and dump debug information into a temporary file if an
13286 internal compiler error (ICE) occurs.
13288 @item -fdump-unnumbered
13289 @opindex fdump-unnumbered
13290 When doing debugging dumps, suppress instruction numbers and address output.
13291 This makes it more feasible to use diff on debugging dumps for compiler
13292 invocations with different options, in particular with and without
13295 @item -fdump-unnumbered-links
13296 @opindex fdump-unnumbered-links
13297 When doing debugging dumps (see @option{-d} option above), suppress
13298 instruction numbers for the links to the previous and next instructions
13301 @item -fdump-ipa-@var{switch}
13303 Control the dumping at various stages of inter-procedural analysis
13304 language tree to a file. The file name is generated by appending a
13305 switch specific suffix to the source file name, and the file is created
13306 in the same directory as the output file. The following dumps are
13311 Enables all inter-procedural analysis dumps.
13314 Dumps information about call-graph optimization, unused function removal,
13315 and inlining decisions.
13318 Dump after function inlining.
13322 @item -fdump-lang-all
13323 @itemx -fdump-lang-@var{switch}
13324 @itemx -fdump-lang-@var{switch}-@var{options}
13325 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
13326 @opindex fdump-lang-all
13327 @opindex fdump-lang
13328 Control the dumping of language-specific information. The @var{options}
13329 and @var{filename} portions behave as described in the
13330 @option{-fdump-tree} option. The following @var{switch} values are
13336 Enable all language-specific dumps.
13339 Dump class hierarchy information. Virtual table information is emitted
13340 unless '@option{slim}' is specified. This option is applicable to C++ only.
13343 Dump the raw internal tree data. This option is applicable to C++ only.
13347 @item -fdump-passes
13348 @opindex fdump-passes
13349 Print on @file{stderr} the list of optimization passes that are turned
13350 on and off by the current command-line options.
13352 @item -fdump-statistics-@var{option}
13353 @opindex fdump-statistics
13354 Enable and control dumping of pass statistics in a separate file. The
13355 file name is generated by appending a suffix ending in
13356 @samp{.statistics} to the source file name, and the file is created in
13357 the same directory as the output file. If the @samp{-@var{option}}
13358 form is used, @samp{-stats} causes counters to be summed over the
13359 whole compilation unit while @samp{-details} dumps every event as
13360 the passes generate them. The default with no option is to sum
13361 counters for each function compiled.
13363 @item -fdump-tree-all
13364 @itemx -fdump-tree-@var{switch}
13365 @itemx -fdump-tree-@var{switch}-@var{options}
13366 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
13367 @opindex fdump-tree-all
13368 @opindex fdump-tree
13369 Control the dumping at various stages of processing the intermediate
13370 language tree to a file. The file name is generated by appending a
13371 switch-specific suffix to the source file name, and the file is
13372 created in the same directory as the output file. In case of
13373 @option{=@var{filename}} option, the dump is output on the given file
13374 instead of the auto named dump files. If the @samp{-@var{options}}
13375 form is used, @var{options} is a list of @samp{-} separated options
13376 which control the details of the dump. Not all options are applicable
13377 to all dumps; those that are not meaningful are ignored. The
13378 following options are available
13382 Print the address of each node. Usually this is not meaningful as it
13383 changes according to the environment and source file. Its primary use
13384 is for tying up a dump file with a debug environment.
13386 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
13387 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
13388 use working backward from mangled names in the assembly file.
13390 When dumping front-end intermediate representations, inhibit dumping
13391 of members of a scope or body of a function merely because that scope
13392 has been reached. Only dump such items when they are directly reachable
13393 by some other path.
13395 When dumping pretty-printed trees, this option inhibits dumping the
13396 bodies of control structures.
13398 When dumping RTL, print the RTL in slim (condensed) form instead of
13399 the default LISP-like representation.
13401 Print a raw representation of the tree. By default, trees are
13402 pretty-printed into a C-like representation.
13404 Enable more detailed dumps (not honored by every dump option). Also
13405 include information from the optimization passes.
13407 Enable dumping various statistics about the pass (not honored by every dump
13410 Enable showing basic block boundaries (disabled in raw dumps).
13412 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13413 dump a representation of the control flow graph suitable for viewing with
13414 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13415 the file is pretty-printed as a subgraph, so that GraphViz can render them
13416 all in a single plot.
13418 This option currently only works for RTL dumps, and the RTL is always
13419 dumped in slim form.
13421 Enable showing virtual operands for every statement.
13423 Enable showing line numbers for statements.
13425 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13427 Enable showing the tree dump for each statement.
13429 Enable showing the EH region number holding each statement.
13431 Enable showing scalar evolution analysis details.
13433 Enable showing optimization information (only available in certain
13436 Enable showing missed optimization information (only available in certain
13439 Enable other detailed optimization information (only available in
13441 @item =@var{filename}
13442 Instead of an auto named dump file, output into the given file
13443 name. The file names @file{stdout} and @file{stderr} are treated
13444 specially and are considered already open standard streams. For
13448 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13449 -fdump-tree-pre=/dev/stderr file.c
13452 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13453 output on to @file{stderr}. If two conflicting dump filenames are
13454 given for the same pass, then the latter option overrides the earlier
13458 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13459 and @option{lineno}.
13462 Turn on all optimization options, i.e., @option{optimized},
13463 @option{missed}, and @option{note}.
13466 To determine what tree dumps are available or find the dump for a pass
13467 of interest follow the steps below.
13471 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
13472 look for a code that corresponds to the pass you are interested in.
13473 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
13474 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
13475 The number at the end distinguishes distinct invocations of the same pass.
13477 To enable the creation of the dump file, append the pass code to
13478 the @option{-fdump-} option prefix and invoke GCC with it. For example,
13479 to enable the dump from the Early Value Range Propagation pass, invoke
13480 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
13481 specify the name of the dump file. If you don't specify one, GCC
13482 creates as described below.
13484 Find the pass dump in a file whose name is composed of three components
13485 separated by a period: the name of the source file GCC was invoked to
13486 compile, a numeric suffix indicating the pass number followed by the
13487 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
13488 and finally the pass code. For example, the Early VRP pass dump might
13489 be in a file named @file{myfile.c.038t.evrp} in the current working
13490 directory. Note that the numeric codes are not stable and may change
13491 from one version of GCC to another.
13495 @itemx -fopt-info-@var{options}
13496 @itemx -fopt-info-@var{options}=@var{filename}
13498 Controls optimization dumps from various optimization passes. If the
13499 @samp{-@var{options}} form is used, @var{options} is a list of
13500 @samp{-} separated option keywords to select the dump details and
13503 The @var{options} can be divided into two groups: options describing the
13504 verbosity of the dump, and options describing which optimizations
13505 should be included. The options from both the groups can be freely
13506 mixed as they are non-overlapping. However, in case of any conflicts,
13507 the later options override the earlier options on the command
13510 The following options control the dump verbosity:
13514 Print information when an optimization is successfully applied. It is
13515 up to a pass to decide which information is relevant. For example, the
13516 vectorizer passes print the source location of loops which are
13517 successfully vectorized.
13519 Print information about missed optimizations. Individual passes
13520 control which information to include in the output.
13522 Print verbose information about optimizations, such as certain
13523 transformations, more detailed messages about decisions etc.
13525 Print detailed optimization information. This includes
13526 @samp{optimized}, @samp{missed}, and @samp{note}.
13529 One or more of the following option keywords can be used to describe a
13530 group of optimizations:
13534 Enable dumps from all interprocedural optimizations.
13536 Enable dumps from all loop optimizations.
13538 Enable dumps from all inlining optimizations.
13540 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
13542 Enable dumps from all vectorization optimizations.
13544 Enable dumps from all optimizations. This is a superset of
13545 the optimization groups listed above.
13548 If @var{options} is
13549 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13550 info about successful optimizations from all the passes.
13552 If the @var{filename} is provided, then the dumps from all the
13553 applicable optimizations are concatenated into the @var{filename}.
13554 Otherwise the dump is output onto @file{stderr}. Though multiple
13555 @option{-fopt-info} options are accepted, only one of them can include
13556 a @var{filename}. If other filenames are provided then all but the
13557 first such option are ignored.
13559 Note that the output @var{filename} is overwritten
13560 in case of multiple translation units. If a combined output from
13561 multiple translation units is desired, @file{stderr} should be used
13564 In the following example, the optimization info is output to
13573 gcc -O3 -fopt-info-missed=missed.all
13577 outputs missed optimization report from all the passes into
13578 @file{missed.all}, and this one:
13581 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13585 prints information about missed optimization opportunities from
13586 vectorization passes on @file{stderr}.
13587 Note that @option{-fopt-info-vec-missed} is equivalent to
13588 @option{-fopt-info-missed-vec}. The order of the optimization group
13589 names and message types listed after @option{-fopt-info} does not matter.
13591 As another example,
13593 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13597 outputs information about missed optimizations as well as
13598 optimized locations from all the inlining passes into
13604 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13608 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13609 in conflict since only one output file is allowed. In this case, only
13610 the first option takes effect and the subsequent options are
13611 ignored. Thus only @file{vec.miss} is produced which contains
13612 dumps from the vectorizer about missed opportunities.
13614 @item -fsched-verbose=@var{n}
13615 @opindex fsched-verbose
13616 On targets that use instruction scheduling, this option controls the
13617 amount of debugging output the scheduler prints to the dump files.
13619 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13620 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13621 For @var{n} greater than one, it also output basic block probabilities,
13622 detailed ready list information and unit/insn info. For @var{n} greater
13623 than two, it includes RTL at abort point, control-flow and regions info.
13624 And for @var{n} over four, @option{-fsched-verbose} also includes
13629 @item -fenable-@var{kind}-@var{pass}
13630 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13634 This is a set of options that are used to explicitly disable/enable
13635 optimization passes. These options are intended for use for debugging GCC.
13636 Compiler users should use regular options for enabling/disabling
13641 @item -fdisable-ipa-@var{pass}
13642 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13643 statically invoked in the compiler multiple times, the pass name should be
13644 appended with a sequential number starting from 1.
13646 @item -fdisable-rtl-@var{pass}
13647 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13648 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13649 statically invoked in the compiler multiple times, the pass name should be
13650 appended with a sequential number starting from 1. @var{range-list} is a
13651 comma-separated list of function ranges or assembler names. Each range is a number
13652 pair separated by a colon. The range is inclusive in both ends. If the range
13653 is trivial, the number pair can be simplified as a single number. If the
13654 function's call graph node's @var{uid} falls within one of the specified ranges,
13655 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13656 function header of a dump file, and the pass names can be dumped by using
13657 option @option{-fdump-passes}.
13659 @item -fdisable-tree-@var{pass}
13660 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13661 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13664 @item -fenable-ipa-@var{pass}
13665 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13666 statically invoked in the compiler multiple times, the pass name should be
13667 appended with a sequential number starting from 1.
13669 @item -fenable-rtl-@var{pass}
13670 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13671 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13672 description and examples.
13674 @item -fenable-tree-@var{pass}
13675 @itemx -fenable-tree-@var{pass}=@var{range-list}
13676 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13677 of option arguments.
13681 Here are some examples showing uses of these options.
13685 # disable ccp1 for all functions
13686 -fdisable-tree-ccp1
13687 # disable complete unroll for function whose cgraph node uid is 1
13688 -fenable-tree-cunroll=1
13689 # disable gcse2 for functions at the following ranges [1,1],
13690 # [300,400], and [400,1000]
13691 # disable gcse2 for functions foo and foo2
13692 -fdisable-rtl-gcse2=foo,foo2
13693 # disable early inlining
13694 -fdisable-tree-einline
13695 # disable ipa inlining
13696 -fdisable-ipa-inline
13697 # enable tree full unroll
13698 -fenable-tree-unroll
13703 @itemx -fchecking=@var{n}
13705 @opindex fno-checking
13706 Enable internal consistency checking. The default depends on
13707 the compiler configuration. @option{-fchecking=2} enables further
13708 internal consistency checking that might affect code generation.
13710 @item -frandom-seed=@var{string}
13711 @opindex frandom-seed
13712 This option provides a seed that GCC uses in place of
13713 random numbers in generating certain symbol names
13714 that have to be different in every compiled file. It is also used to
13715 place unique stamps in coverage data files and the object files that
13716 produce them. You can use the @option{-frandom-seed} option to produce
13717 reproducibly identical object files.
13719 The @var{string} can either be a number (decimal, octal or hex) or an
13720 arbitrary string (in which case it's converted to a number by
13723 The @var{string} should be different for every file you compile.
13726 @itemx -save-temps=cwd
13727 @opindex save-temps
13728 Store the usual ``temporary'' intermediate files permanently; place them
13729 in the current directory and name them based on the source file. Thus,
13730 compiling @file{foo.c} with @option{-c -save-temps} produces files
13731 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13732 preprocessed @file{foo.i} output file even though the compiler now
13733 normally uses an integrated preprocessor.
13735 When used in combination with the @option{-x} command-line option,
13736 @option{-save-temps} is sensible enough to avoid over writing an
13737 input source file with the same extension as an intermediate file.
13738 The corresponding intermediate file may be obtained by renaming the
13739 source file before using @option{-save-temps}.
13741 If you invoke GCC in parallel, compiling several different source
13742 files that share a common base name in different subdirectories or the
13743 same source file compiled for multiple output destinations, it is
13744 likely that the different parallel compilers will interfere with each
13745 other, and overwrite the temporary files. For instance:
13748 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13749 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13752 may result in @file{foo.i} and @file{foo.o} being written to
13753 simultaneously by both compilers.
13755 @item -save-temps=obj
13756 @opindex save-temps=obj
13757 Store the usual ``temporary'' intermediate files permanently. If the
13758 @option{-o} option is used, the temporary files are based on the
13759 object file. If the @option{-o} option is not used, the
13760 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13765 gcc -save-temps=obj -c foo.c
13766 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13767 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13771 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13772 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13773 @file{dir2/yfoobar.o}.
13775 @item -time@r{[}=@var{file}@r{]}
13777 Report the CPU time taken by each subprocess in the compilation
13778 sequence. For C source files, this is the compiler proper and assembler
13779 (plus the linker if linking is done).
13781 Without the specification of an output file, the output looks like this:
13788 The first number on each line is the ``user time'', that is time spent
13789 executing the program itself. The second number is ``system time'',
13790 time spent executing operating system routines on behalf of the program.
13791 Both numbers are in seconds.
13793 With the specification of an output file, the output is appended to the
13794 named file, and it looks like this:
13797 0.12 0.01 cc1 @var{options}
13798 0.00 0.01 as @var{options}
13801 The ``user time'' and the ``system time'' are moved before the program
13802 name, and the options passed to the program are displayed, so that one
13803 can later tell what file was being compiled, and with which options.
13805 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13806 @opindex fdump-final-insns
13807 Dump the final internal representation (RTL) to @var{file}. If the
13808 optional argument is omitted (or if @var{file} is @code{.}), the name
13809 of the dump file is determined by appending @code{.gkd} to the
13810 compilation output file name.
13812 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13813 @opindex fcompare-debug
13814 @opindex fno-compare-debug
13815 If no error occurs during compilation, run the compiler a second time,
13816 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13817 passed to the second compilation. Dump the final internal
13818 representation in both compilations, and print an error if they differ.
13820 If the equal sign is omitted, the default @option{-gtoggle} is used.
13822 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13823 and nonzero, implicitly enables @option{-fcompare-debug}. If
13824 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13825 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13828 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13829 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13830 of the final representation and the second compilation, preventing even
13831 @env{GCC_COMPARE_DEBUG} from taking effect.
13833 To verify full coverage during @option{-fcompare-debug} testing, set
13834 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13835 which GCC rejects as an invalid option in any actual compilation
13836 (rather than preprocessing, assembly or linking). To get just a
13837 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13838 not overridden} will do.
13840 @item -fcompare-debug-second
13841 @opindex fcompare-debug-second
13842 This option is implicitly passed to the compiler for the second
13843 compilation requested by @option{-fcompare-debug}, along with options to
13844 silence warnings, and omitting other options that would cause
13845 side-effect compiler outputs to files or to the standard output. Dump
13846 files and preserved temporary files are renamed so as to contain the
13847 @code{.gk} additional extension during the second compilation, to avoid
13848 overwriting those generated by the first.
13850 When this option is passed to the compiler driver, it causes the
13851 @emph{first} compilation to be skipped, which makes it useful for little
13852 other than debugging the compiler proper.
13856 Turn off generation of debug info, if leaving out this option
13857 generates it, or turn it on at level 2 otherwise. The position of this
13858 argument in the command line does not matter; it takes effect after all
13859 other options are processed, and it does so only once, no matter how
13860 many times it is given. This is mainly intended to be used with
13861 @option{-fcompare-debug}.
13863 @item -fvar-tracking-assignments-toggle
13864 @opindex fvar-tracking-assignments-toggle
13865 @opindex fno-var-tracking-assignments-toggle
13866 Toggle @option{-fvar-tracking-assignments}, in the same way that
13867 @option{-gtoggle} toggles @option{-g}.
13871 Makes the compiler print out each function name as it is compiled, and
13872 print some statistics about each pass when it finishes.
13874 @item -ftime-report
13875 @opindex ftime-report
13876 Makes the compiler print some statistics about the time consumed by each
13877 pass when it finishes.
13879 @item -ftime-report-details
13880 @opindex ftime-report-details
13881 Record the time consumed by infrastructure parts separately for each pass.
13883 @item -fira-verbose=@var{n}
13884 @opindex fira-verbose
13885 Control the verbosity of the dump file for the integrated register allocator.
13886 The default value is 5. If the value @var{n} is greater or equal to 10,
13887 the dump output is sent to stderr using the same format as @var{n} minus 10.
13890 @opindex flto-report
13891 Prints a report with internal details on the workings of the link-time
13892 optimizer. The contents of this report vary from version to version.
13893 It is meant to be useful to GCC developers when processing object
13894 files in LTO mode (via @option{-flto}).
13896 Disabled by default.
13898 @item -flto-report-wpa
13899 @opindex flto-report-wpa
13900 Like @option{-flto-report}, but only print for the WPA phase of Link
13904 @opindex fmem-report
13905 Makes the compiler print some statistics about permanent memory
13906 allocation when it finishes.
13908 @item -fmem-report-wpa
13909 @opindex fmem-report-wpa
13910 Makes the compiler print some statistics about permanent memory
13911 allocation for the WPA phase only.
13913 @item -fpre-ipa-mem-report
13914 @opindex fpre-ipa-mem-report
13915 @item -fpost-ipa-mem-report
13916 @opindex fpost-ipa-mem-report
13917 Makes the compiler print some statistics about permanent memory
13918 allocation before or after interprocedural optimization.
13920 @item -fprofile-report
13921 @opindex fprofile-report
13922 Makes the compiler print some statistics about consistency of the
13923 (estimated) profile and effect of individual passes.
13925 @item -fstack-usage
13926 @opindex fstack-usage
13927 Makes the compiler output stack usage information for the program, on a
13928 per-function basis. The filename for the dump is made by appending
13929 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13930 the output file, if explicitly specified and it is not an executable,
13931 otherwise it is the basename of the source file. An entry is made up
13936 The name of the function.
13940 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
13943 The qualifier @code{static} means that the function manipulates the stack
13944 statically: a fixed number of bytes are allocated for the frame on function
13945 entry and released on function exit; no stack adjustments are otherwise made
13946 in the function. The second field is this fixed number of bytes.
13948 The qualifier @code{dynamic} means that the function manipulates the stack
13949 dynamically: in addition to the static allocation described above, stack
13950 adjustments are made in the body of the function, for example to push/pop
13951 arguments around function calls. If the qualifier @code{bounded} is also
13952 present, the amount of these adjustments is bounded at compile time and
13953 the second field is an upper bound of the total amount of stack used by
13954 the function. If it is not present, the amount of these adjustments is
13955 not bounded at compile time and the second field only represents the
13960 Emit statistics about front-end processing at the end of the compilation.
13961 This option is supported only by the C++ front end, and
13962 the information is generally only useful to the G++ development team.
13964 @item -fdbg-cnt-list
13965 @opindex fdbg-cnt-list
13966 Print the name and the counter upper bound for all debug counters.
13969 @item -fdbg-cnt=@var{counter-value-list}
13971 Set the internal debug counter upper bound. @var{counter-value-list}
13972 is a comma-separated list of @var{name}:@var{value} pairs
13973 which sets the upper bound of each debug counter @var{name} to @var{value}.
13974 All debug counters have the initial upper bound of @code{UINT_MAX};
13975 thus @code{dbg_cnt} returns true always unless the upper bound
13976 is set by this option.
13977 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
13978 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
13980 @item -print-file-name=@var{library}
13981 @opindex print-file-name
13982 Print the full absolute name of the library file @var{library} that
13983 would be used when linking---and don't do anything else. With this
13984 option, GCC does not compile or link anything; it just prints the
13987 @item -print-multi-directory
13988 @opindex print-multi-directory
13989 Print the directory name corresponding to the multilib selected by any
13990 other switches present in the command line. This directory is supposed
13991 to exist in @env{GCC_EXEC_PREFIX}.
13993 @item -print-multi-lib
13994 @opindex print-multi-lib
13995 Print the mapping from multilib directory names to compiler switches
13996 that enable them. The directory name is separated from the switches by
13997 @samp{;}, and each switch starts with an @samp{@@} instead of the
13998 @samp{-}, without spaces between multiple switches. This is supposed to
13999 ease shell processing.
14001 @item -print-multi-os-directory
14002 @opindex print-multi-os-directory
14003 Print the path to OS libraries for the selected
14004 multilib, relative to some @file{lib} subdirectory. If OS libraries are
14005 present in the @file{lib} subdirectory and no multilibs are used, this is
14006 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
14007 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
14008 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
14009 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
14011 @item -print-multiarch
14012 @opindex print-multiarch
14013 Print the path to OS libraries for the selected multiarch,
14014 relative to some @file{lib} subdirectory.
14016 @item -print-prog-name=@var{program}
14017 @opindex print-prog-name
14018 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
14020 @item -print-libgcc-file-name
14021 @opindex print-libgcc-file-name
14022 Same as @option{-print-file-name=libgcc.a}.
14024 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
14025 but you do want to link with @file{libgcc.a}. You can do:
14028 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
14031 @item -print-search-dirs
14032 @opindex print-search-dirs
14033 Print the name of the configured installation directory and a list of
14034 program and library directories @command{gcc} searches---and don't do anything else.
14036 This is useful when @command{gcc} prints the error message
14037 @samp{installation problem, cannot exec cpp0: No such file or directory}.
14038 To resolve this you either need to put @file{cpp0} and the other compiler
14039 components where @command{gcc} expects to find them, or you can set the environment
14040 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
14041 Don't forget the trailing @samp{/}.
14042 @xref{Environment Variables}.
14044 @item -print-sysroot
14045 @opindex print-sysroot
14046 Print the target sysroot directory that is used during
14047 compilation. This is the target sysroot specified either at configure
14048 time or using the @option{--sysroot} option, possibly with an extra
14049 suffix that depends on compilation options. If no target sysroot is
14050 specified, the option prints nothing.
14052 @item -print-sysroot-headers-suffix
14053 @opindex print-sysroot-headers-suffix
14054 Print the suffix added to the target sysroot when searching for
14055 headers, or give an error if the compiler is not configured with such
14056 a suffix---and don't do anything else.
14059 @opindex dumpmachine
14060 Print the compiler's target machine (for example,
14061 @samp{i686-pc-linux-gnu})---and don't do anything else.
14064 @opindex dumpversion
14065 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
14066 anything else. This is the compiler version used in filesystem paths,
14067 specs, can be depending on how the compiler has been configured just
14068 a single number (major version), two numbers separated by dot (major and
14069 minor version) or three numbers separated by dots (major, minor and patchlevel
14072 @item -dumpfullversion
14073 @opindex dumpfullversion
14074 Print the full compiler version, always 3 numbers separated by dots,
14075 major, minor and patchlevel version.
14079 Print the compiler's built-in specs---and don't do anything else. (This
14080 is used when GCC itself is being built.) @xref{Spec Files}.
14083 @node Submodel Options
14084 @section Machine-Dependent Options
14085 @cindex submodel options
14086 @cindex specifying hardware config
14087 @cindex hardware models and configurations, specifying
14088 @cindex target-dependent options
14089 @cindex machine-dependent options
14091 Each target machine supported by GCC can have its own options---for
14092 example, to allow you to compile for a particular processor variant or
14093 ABI, or to control optimizations specific to that machine. By
14094 convention, the names of machine-specific options start with
14097 Some configurations of the compiler also support additional target-specific
14098 options, usually for compatibility with other compilers on the same
14101 @c This list is ordered alphanumerically by subsection name.
14102 @c It should be the same order and spelling as these options are listed
14103 @c in Machine Dependent Options
14106 * AArch64 Options::
14107 * Adapteva Epiphany Options::
14111 * Blackfin Options::
14116 * DEC Alpha Options::
14120 * GNU/Linux Options::
14130 * MicroBlaze Options::
14133 * MN10300 Options::
14137 * Nios II Options::
14138 * Nvidia PTX Options::
14140 * picoChip Options::
14141 * PowerPC Options::
14144 * RS/6000 and PowerPC Options::
14146 * S/390 and zSeries Options::
14149 * Solaris 2 Options::
14152 * System V Options::
14153 * TILE-Gx Options::
14154 * TILEPro Options::
14159 * VxWorks Options::
14161 * x86 Windows Options::
14162 * Xstormy16 Options::
14164 * zSeries Options::
14167 @node AArch64 Options
14168 @subsection AArch64 Options
14169 @cindex AArch64 Options
14171 These options are defined for AArch64 implementations:
14175 @item -mabi=@var{name}
14177 Generate code for the specified data model. Permissible values
14178 are @samp{ilp32} for SysV-like data model where int, long int and pointers
14179 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
14180 but long int and pointers are 64 bits.
14182 The default depends on the specific target configuration. Note that
14183 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
14184 entire program with the same ABI, and link with a compatible set of libraries.
14187 @opindex mbig-endian
14188 Generate big-endian code. This is the default when GCC is configured for an
14189 @samp{aarch64_be-*-*} target.
14191 @item -mgeneral-regs-only
14192 @opindex mgeneral-regs-only
14193 Generate code which uses only the general-purpose registers. This will prevent
14194 the compiler from using floating-point and Advanced SIMD registers but will not
14195 impose any restrictions on the assembler.
14197 @item -mlittle-endian
14198 @opindex mlittle-endian
14199 Generate little-endian code. This is the default when GCC is configured for an
14200 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
14202 @item -mcmodel=tiny
14203 @opindex mcmodel=tiny
14204 Generate code for the tiny code model. The program and its statically defined
14205 symbols must be within 1MB of each other. Programs can be statically or
14206 dynamically linked.
14208 @item -mcmodel=small
14209 @opindex mcmodel=small
14210 Generate code for the small code model. The program and its statically defined
14211 symbols must be within 4GB of each other. Programs can be statically or
14212 dynamically linked. This is the default code model.
14214 @item -mcmodel=large
14215 @opindex mcmodel=large
14216 Generate code for the large code model. This makes no assumptions about
14217 addresses and sizes of sections. Programs can be statically linked only.
14219 @item -mstrict-align
14220 @opindex mstrict-align
14221 Avoid generating memory accesses that may not be aligned on a natural object
14222 boundary as described in the architecture specification.
14224 @item -momit-leaf-frame-pointer
14225 @itemx -mno-omit-leaf-frame-pointer
14226 @opindex momit-leaf-frame-pointer
14227 @opindex mno-omit-leaf-frame-pointer
14228 Omit or keep the frame pointer in leaf functions. The former behavior is the
14231 @item -mtls-dialect=desc
14232 @opindex mtls-dialect=desc
14233 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
14234 of TLS variables. This is the default.
14236 @item -mtls-dialect=traditional
14237 @opindex mtls-dialect=traditional
14238 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
14241 @item -mtls-size=@var{size}
14243 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
14244 This option requires binutils 2.26 or newer.
14246 @item -mfix-cortex-a53-835769
14247 @itemx -mno-fix-cortex-a53-835769
14248 @opindex mfix-cortex-a53-835769
14249 @opindex mno-fix-cortex-a53-835769
14250 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
14251 This involves inserting a NOP instruction between memory instructions and
14252 64-bit integer multiply-accumulate instructions.
14254 @item -mfix-cortex-a53-843419
14255 @itemx -mno-fix-cortex-a53-843419
14256 @opindex mfix-cortex-a53-843419
14257 @opindex mno-fix-cortex-a53-843419
14258 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
14259 This erratum workaround is made at link time and this will only pass the
14260 corresponding flag to the linker.
14262 @item -mlow-precision-recip-sqrt
14263 @item -mno-low-precision-recip-sqrt
14264 @opindex mlow-precision-recip-sqrt
14265 @opindex mno-low-precision-recip-sqrt
14266 Enable or disable the reciprocal square root approximation.
14267 This option only has an effect if @option{-ffast-math} or
14268 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14269 precision of reciprocal square root results to about 16 bits for
14270 single precision and to 32 bits for double precision.
14272 @item -mlow-precision-sqrt
14273 @item -mno-low-precision-sqrt
14274 @opindex -mlow-precision-sqrt
14275 @opindex -mno-low-precision-sqrt
14276 Enable or disable the square root approximation.
14277 This option only has an effect if @option{-ffast-math} or
14278 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14279 precision of square root results to about 16 bits for
14280 single precision and to 32 bits for double precision.
14281 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
14283 @item -mlow-precision-div
14284 @item -mno-low-precision-div
14285 @opindex -mlow-precision-div
14286 @opindex -mno-low-precision-div
14287 Enable or disable the division approximation.
14288 This option only has an effect if @option{-ffast-math} or
14289 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14290 precision of division results to about 16 bits for
14291 single precision and to 32 bits for double precision.
14293 @item -march=@var{name}
14295 Specify the name of the target architecture and, optionally, one or
14296 more feature modifiers. This option has the form
14297 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
14299 The permissible values for @var{arch} are @samp{armv8-a},
14300 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @var{native}.
14302 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
14303 support for the ARMv8.3-A architecture extensions.
14305 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
14306 support for the ARMv8.2-A architecture extensions.
14308 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
14309 support for the ARMv8.1-A architecture extension. In particular, it
14310 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
14312 The value @samp{native} is available on native AArch64 GNU/Linux and
14313 causes the compiler to pick the architecture of the host system. This
14314 option has no effect if the compiler is unable to recognize the
14315 architecture of the host system,
14317 The permissible values for @var{feature} are listed in the sub-section
14318 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14319 Feature Modifiers}. Where conflicting feature modifiers are
14320 specified, the right-most feature is used.
14322 GCC uses @var{name} to determine what kind of instructions it can emit
14323 when generating assembly code. If @option{-march} is specified
14324 without either of @option{-mtune} or @option{-mcpu} also being
14325 specified, the code is tuned to perform well across a range of target
14326 processors implementing the target architecture.
14328 @item -mtune=@var{name}
14330 Specify the name of the target processor for which GCC should tune the
14331 performance of the code. Permissible values for this option are:
14332 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
14333 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
14334 @samp{exynos-m1}, @samp{falkor}, @samp{qdf24xx},
14335 @samp{xgene1}, @samp{vulcan}, @samp{thunderx},
14336 @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
14337 @samp{thunderxt83}, @samp{thunderx2t99}, @samp{cortex-a57.cortex-a53},
14338 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
14339 @samp{cortex-a73.cortex-a53}, @samp{cortex-a75.cortex-a55},
14342 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14343 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
14344 @samp{cortex-a75.cortex-a55} specify that GCC should tune for a
14347 Additionally on native AArch64 GNU/Linux systems the value
14348 @samp{native} tunes performance to the host system. This option has no effect
14349 if the compiler is unable to recognize the processor of the host system.
14351 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
14352 are specified, the code is tuned to perform well across a range
14353 of target processors.
14355 This option cannot be suffixed by feature modifiers.
14357 @item -mcpu=@var{name}
14359 Specify the name of the target processor, optionally suffixed by one
14360 or more feature modifiers. This option has the form
14361 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
14362 the permissible values for @var{cpu} are the same as those available
14363 for @option{-mtune}. The permissible values for @var{feature} are
14364 documented in the sub-section on
14365 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14366 Feature Modifiers}. Where conflicting feature modifiers are
14367 specified, the right-most feature is used.
14369 GCC uses @var{name} to determine what kind of instructions it can emit when
14370 generating assembly code (as if by @option{-march}) and to determine
14371 the target processor for which to tune for performance (as if
14372 by @option{-mtune}). Where this option is used in conjunction
14373 with @option{-march} or @option{-mtune}, those options take precedence
14374 over the appropriate part of this option.
14376 @item -moverride=@var{string}
14378 Override tuning decisions made by the back-end in response to a
14379 @option{-mtune=} switch. The syntax, semantics, and accepted values
14380 for @var{string} in this option are not guaranteed to be consistent
14383 This option is only intended to be useful when developing GCC.
14385 @item -mpc-relative-literal-loads
14386 @itemx -mno-pc-relative-literal-loads
14387 @opindex mpc-relative-literal-loads
14388 @opindex mno-pc-relative-literal-loads
14389 Enable or disable PC-relative literal loads. With this option literal pools are
14390 accessed using a single instruction and emitted after each function. This
14391 limits the maximum size of functions to 1MB. This is enabled by default for
14392 @option{-mcmodel=tiny}.
14394 @item -msign-return-address=@var{scope}
14395 @opindex msign-return-address
14396 Select the function scope on which return address signing will be applied.
14397 Permissible values are @samp{none}, which disables return address signing,
14398 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
14399 functions, and @samp{all}, which enables pointer signing for all functions. The
14400 default value is @samp{none}.
14404 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14405 @anchor{aarch64-feature-modifiers}
14406 @cindex @option{-march} feature modifiers
14407 @cindex @option{-mcpu} feature modifiers
14408 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14409 the following and their inverses @option{no@var{feature}}:
14413 Enable CRC extension. This is on by default for
14414 @option{-march=armv8.1-a}.
14416 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14419 Enable floating-point instructions. This is on by default for all possible
14420 values for options @option{-march} and @option{-mcpu}.
14422 Enable Advanced SIMD instructions. This also enables floating-point
14423 instructions. This is on by default for all possible values for options
14424 @option{-march} and @option{-mcpu}.
14426 Enable Large System Extension instructions. This is on by default for
14427 @option{-march=armv8.1-a}.
14429 Enable Round Double Multiply Accumulate instructions. This is on by default
14430 for @option{-march=armv8.1-a}.
14432 Enable FP16 extension. This also enables floating-point instructions.
14434 Enable the RcPc extension. This does not change code generation from GCC,
14435 but is passed on to the assembler, enabling inline asm statements to use
14436 instructions from the RcPc extension.
14438 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
14442 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
14443 Conversely, @option{nofp} implies @option{nosimd}, which implies
14446 @node Adapteva Epiphany Options
14447 @subsection Adapteva Epiphany Options
14449 These @samp{-m} options are defined for Adapteva Epiphany:
14452 @item -mhalf-reg-file
14453 @opindex mhalf-reg-file
14454 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14455 That allows code to run on hardware variants that lack these registers.
14457 @item -mprefer-short-insn-regs
14458 @opindex mprefer-short-insn-regs
14459 Preferentially allocate registers that allow short instruction generation.
14460 This can result in increased instruction count, so this may either reduce or
14461 increase overall code size.
14463 @item -mbranch-cost=@var{num}
14464 @opindex mbranch-cost
14465 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14466 This cost is only a heuristic and is not guaranteed to produce
14467 consistent results across releases.
14471 Enable the generation of conditional moves.
14473 @item -mnops=@var{num}
14475 Emit @var{num} NOPs before every other generated instruction.
14477 @item -mno-soft-cmpsf
14478 @opindex mno-soft-cmpsf
14479 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14480 and test the flags. This is faster than a software comparison, but can
14481 get incorrect results in the presence of NaNs, or when two different small
14482 numbers are compared such that their difference is calculated as zero.
14483 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14484 software comparisons.
14486 @item -mstack-offset=@var{num}
14487 @opindex mstack-offset
14488 Set the offset between the top of the stack and the stack pointer.
14489 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14490 can be used by leaf functions without stack allocation.
14491 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14492 Note also that this option changes the ABI; compiling a program with a
14493 different stack offset than the libraries have been compiled with
14494 generally does not work.
14495 This option can be useful if you want to evaluate if a different stack
14496 offset would give you better code, but to actually use a different stack
14497 offset to build working programs, it is recommended to configure the
14498 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14500 @item -mno-round-nearest
14501 @opindex mno-round-nearest
14502 Make the scheduler assume that the rounding mode has been set to
14503 truncating. The default is @option{-mround-nearest}.
14506 @opindex mlong-calls
14507 If not otherwise specified by an attribute, assume all calls might be beyond
14508 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14509 function address into a register before performing a (otherwise direct) call.
14510 This is the default.
14512 @item -mshort-calls
14513 @opindex short-calls
14514 If not otherwise specified by an attribute, assume all direct calls are
14515 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14516 for direct calls. The default is @option{-mlong-calls}.
14520 Assume addresses can be loaded as 16-bit unsigned values. This does not
14521 apply to function addresses for which @option{-mlong-calls} semantics
14524 @item -mfp-mode=@var{mode}
14526 Set the prevailing mode of the floating-point unit.
14527 This determines the floating-point mode that is provided and expected
14528 at function call and return time. Making this mode match the mode you
14529 predominantly need at function start can make your programs smaller and
14530 faster by avoiding unnecessary mode switches.
14532 @var{mode} can be set to one the following values:
14536 Any mode at function entry is valid, and retained or restored when
14537 the function returns, and when it calls other functions.
14538 This mode is useful for compiling libraries or other compilation units
14539 you might want to incorporate into different programs with different
14540 prevailing FPU modes, and the convenience of being able to use a single
14541 object file outweighs the size and speed overhead for any extra
14542 mode switching that might be needed, compared with what would be needed
14543 with a more specific choice of prevailing FPU mode.
14546 This is the mode used for floating-point calculations with
14547 truncating (i.e.@: round towards zero) rounding mode. That includes
14548 conversion from floating point to integer.
14550 @item round-nearest
14551 This is the mode used for floating-point calculations with
14552 round-to-nearest-or-even rounding mode.
14555 This is the mode used to perform integer calculations in the FPU, e.g.@:
14556 integer multiply, or integer multiply-and-accumulate.
14559 The default is @option{-mfp-mode=caller}
14561 @item -mnosplit-lohi
14562 @itemx -mno-postinc
14563 @itemx -mno-postmodify
14564 @opindex mnosplit-lohi
14565 @opindex mno-postinc
14566 @opindex mno-postmodify
14567 Code generation tweaks that disable, respectively, splitting of 32-bit
14568 loads, generation of post-increment addresses, and generation of
14569 post-modify addresses. The defaults are @option{msplit-lohi},
14570 @option{-mpost-inc}, and @option{-mpost-modify}.
14572 @item -mnovect-double
14573 @opindex mno-vect-double
14574 Change the preferred SIMD mode to SImode. The default is
14575 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14577 @item -max-vect-align=@var{num}
14578 @opindex max-vect-align
14579 The maximum alignment for SIMD vector mode types.
14580 @var{num} may be 4 or 8. The default is 8.
14581 Note that this is an ABI change, even though many library function
14582 interfaces are unaffected if they don't use SIMD vector modes
14583 in places that affect size and/or alignment of relevant types.
14585 @item -msplit-vecmove-early
14586 @opindex msplit-vecmove-early
14587 Split vector moves into single word moves before reload. In theory this
14588 can give better register allocation, but so far the reverse seems to be
14589 generally the case.
14591 @item -m1reg-@var{reg}
14593 Specify a register to hold the constant @minus{}1, which makes loading small negative
14594 constants and certain bitmasks faster.
14595 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14596 which specify use of that register as a fixed register,
14597 and @samp{none}, which means that no register is used for this
14598 purpose. The default is @option{-m1reg-none}.
14603 @subsection ARC Options
14604 @cindex ARC options
14606 The following options control the architecture variant for which code
14609 @c architecture variants
14612 @item -mbarrel-shifter
14613 @opindex mbarrel-shifter
14614 Generate instructions supported by barrel shifter. This is the default
14615 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14617 @item -mcpu=@var{cpu}
14619 Set architecture type, register usage, and instruction scheduling
14620 parameters for @var{cpu}. There are also shortcut alias options
14621 available for backward compatibility and convenience. Supported
14622 values for @var{cpu} are
14628 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14632 Compile for ARC601. Alias: @option{-mARC601}.
14637 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14638 This is the default when configured with @option{--with-cpu=arc700}@.
14641 Compile for ARC EM.
14644 Compile for ARC HS.
14647 Compile for ARC EM CPU with no hardware extensions.
14650 Compile for ARC EM4 CPU.
14653 Compile for ARC EM4 DMIPS CPU.
14656 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
14660 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
14661 double assist instructions.
14664 Compile for ARC HS CPU with no hardware extensions except the atomic
14668 Compile for ARC HS34 CPU.
14671 Compile for ARC HS38 CPU.
14674 Compile for ARC HS38 CPU with all hardware extensions on.
14677 Compile for ARC 600 CPU with @code{norm} instructions enabled.
14679 @item arc600_mul32x16
14680 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
14681 instructions enabled.
14684 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
14685 instructions enabled.
14688 Compile for ARC 601 CPU with @code{norm} instructions enabled.
14690 @item arc601_mul32x16
14691 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
14692 instructions enabled.
14695 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
14696 instructions enabled.
14699 Compile for ARC 700 on NPS400 chip.
14705 @itemx -mdpfp-compact
14706 @opindex mdpfp-compact
14707 Generate double-precision FPX instructions, tuned for the compact
14711 @opindex mdpfp-fast
14712 Generate double-precision FPX instructions, tuned for the fast
14715 @item -mno-dpfp-lrsr
14716 @opindex mno-dpfp-lrsr
14717 Disable @code{lr} and @code{sr} instructions from using FPX extension
14722 Generate extended arithmetic instructions. Currently only
14723 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14724 supported. This is always enabled for @option{-mcpu=ARC700}.
14728 Do not generate @code{mpy}-family instructions for ARC700. This option is
14733 Generate 32x16-bit multiply and multiply-accumulate instructions.
14737 Generate @code{mul64} and @code{mulu64} instructions.
14738 Only valid for @option{-mcpu=ARC600}.
14742 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
14747 @itemx -mspfp-compact
14748 @opindex mspfp-compact
14749 Generate single-precision FPX instructions, tuned for the compact
14753 @opindex mspfp-fast
14754 Generate single-precision FPX instructions, tuned for the fast
14759 Enable generation of ARC SIMD instructions via target-specific
14760 builtins. Only valid for @option{-mcpu=ARC700}.
14763 @opindex msoft-float
14764 This option ignored; it is provided for compatibility purposes only.
14765 Software floating-point code is emitted by default, and this default
14766 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
14767 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
14768 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
14772 Generate @code{swap} instructions.
14776 This enables use of the locked load/store conditional extension to implement
14777 atomic memory built-in functions. Not available for ARC 6xx or ARC
14782 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
14784 @item -mcode-density
14785 @opindex mcode-density
14786 Enable code density instructions for ARC EM.
14787 This option is on by default for ARC HS.
14791 Enable double load/store operations for ARC HS cores.
14793 @item -mtp-regno=@var{regno}
14795 Specify thread pointer register number.
14797 @item -mmpy-option=@var{multo}
14798 @opindex mmpy-option
14799 Compile ARCv2 code with a multiplier design option. You can specify
14800 the option using either a string or numeric value for @var{multo}.
14801 @samp{wlh1} is the default value. The recognized values are:
14806 No multiplier available.
14810 16x16 multiplier, fully pipelined.
14811 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
14815 32x32 multiplier, fully
14816 pipelined (1 stage). The following instructions are additionally
14817 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14821 32x32 multiplier, fully pipelined
14822 (2 stages). The following instructions are additionally enabled: @code{mpy},
14823 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14827 Two 16x16 multipliers, blocking,
14828 sequential. The following instructions are additionally enabled: @code{mpy},
14829 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14833 One 16x16 multiplier, blocking,
14834 sequential. The following instructions are additionally enabled: @code{mpy},
14835 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14839 One 32x4 multiplier, blocking,
14840 sequential. The following instructions are additionally enabled: @code{mpy},
14841 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14845 ARC HS SIMD support.
14849 ARC HS SIMD support.
14853 ARC HS SIMD support.
14857 This option is only available for ARCv2 cores@.
14859 @item -mfpu=@var{fpu}
14861 Enables support for specific floating-point hardware extensions for ARCv2
14862 cores. Supported values for @var{fpu} are:
14867 Enables support for single-precision floating-point hardware
14871 Enables support for double-precision floating-point hardware
14872 extensions. The single-precision floating-point extension is also
14873 enabled. Not available for ARC EM@.
14876 Enables support for double-precision floating-point hardware
14877 extensions using double-precision assist instructions. The single-precision
14878 floating-point extension is also enabled. This option is
14879 only available for ARC EM@.
14882 Enables support for double-precision floating-point hardware
14883 extensions using double-precision assist instructions.
14884 The single-precision floating-point, square-root, and divide
14885 extensions are also enabled. This option is
14886 only available for ARC EM@.
14889 Enables support for double-precision floating-point hardware
14890 extensions using double-precision assist instructions.
14891 The single-precision floating-point and fused multiply and add
14892 hardware extensions are also enabled. This option is
14893 only available for ARC EM@.
14896 Enables support for double-precision floating-point hardware
14897 extensions using double-precision assist instructions.
14898 All single-precision floating-point hardware extensions are also
14899 enabled. This option is only available for ARC EM@.
14902 Enables support for single-precision floating-point, square-root and divide
14903 hardware extensions@.
14906 Enables support for double-precision floating-point, square-root and divide
14907 hardware extensions. This option
14908 includes option @samp{fpus_div}. Not available for ARC EM@.
14911 Enables support for single-precision floating-point and
14912 fused multiply and add hardware extensions@.
14915 Enables support for double-precision floating-point and
14916 fused multiply and add hardware extensions. This option
14917 includes option @samp{fpus_fma}. Not available for ARC EM@.
14920 Enables support for all single-precision floating-point hardware
14924 Enables support for all single- and double-precision floating-point
14925 hardware extensions. Not available for ARC EM@.
14929 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
14930 @opindex mirq-ctrl-saved
14931 Specifies general-purposes registers that the processor automatically
14932 saves/restores on interrupt entry and exit. @var{register-range} is
14933 specified as two registers separated by a dash. The register range
14934 always starts with @code{r0}, the upper limit is @code{fp} register.
14935 @var{blink} and @var{lp_count} are optional. This option is only
14936 valid for ARC EM and ARC HS cores.
14938 @item -mrgf-banked-regs=@var{number}
14939 @opindex mrgf-banked-regs
14940 Specifies the number of registers replicated in second register bank
14941 on entry to fast interrupt. Fast interrupts are interrupts with the
14942 highest priority level P0. These interrupts save only PC and STATUS32
14943 registers to avoid memory transactions during interrupt entry and exit
14944 sequences. Use this option when you are using fast interrupts in an
14945 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
14947 @item -mlpc-width=@var{width}
14948 @opindex mlpc-width
14949 Specify the width of the @code{lp_count} register. Valid values for
14950 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
14951 fixed to 32 bits. If the width is less than 32, the compiler does not
14952 attempt to transform loops in your program to use the zero-delay loop
14953 mechanism unless it is known that the @code{lp_count} register can
14954 hold the required loop-counter value. Depending on the width
14955 specified, the compiler and run-time library might continue to use the
14956 loop mechanism for various needs. This option defines macro
14957 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
14961 The following options are passed through to the assembler, and also
14962 define preprocessor macro symbols.
14964 @c Flags used by the assembler, but for which we define preprocessor
14965 @c macro symbols as well.
14968 @opindex mdsp-packa
14969 Passed down to the assembler to enable the DSP Pack A extensions.
14970 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
14975 Passed down to the assembler to enable the dual Viterbi butterfly
14976 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
14977 option is deprecated.
14979 @c ARC700 4.10 extension instruction
14982 Passed down to the assembler to enable the locked load/store
14983 conditional extension. Also sets the preprocessor symbol
14988 Passed down to the assembler. Also sets the preprocessor symbol
14989 @code{__Xxmac_d16}. This option is deprecated.
14993 Passed down to the assembler. Also sets the preprocessor symbol
14994 @code{__Xxmac_24}. This option is deprecated.
14996 @c ARC700 4.10 extension instruction
14999 Passed down to the assembler to enable the 64-bit time-stamp counter
15000 extension instruction. Also sets the preprocessor symbol
15001 @code{__Xrtsc}. This option is deprecated.
15003 @c ARC700 4.10 extension instruction
15006 Passed down to the assembler to enable the swap byte ordering
15007 extension instruction. Also sets the preprocessor symbol
15011 @opindex mtelephony
15012 Passed down to the assembler to enable dual- and single-operand
15013 instructions for telephony. Also sets the preprocessor symbol
15014 @code{__Xtelephony}. This option is deprecated.
15018 Passed down to the assembler to enable the XY memory extension. Also
15019 sets the preprocessor symbol @code{__Xxy}.
15023 The following options control how the assembly code is annotated:
15025 @c Assembly annotation options
15029 Annotate assembler instructions with estimated addresses.
15031 @item -mannotate-align
15032 @opindex mannotate-align
15033 Explain what alignment considerations lead to the decision to make an
15034 instruction short or long.
15038 The following options are passed through to the linker:
15040 @c options passed through to the linker
15044 Passed through to the linker, to specify use of the @code{arclinux} emulation.
15045 This option is enabled by default in tool chains built for
15046 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
15047 when profiling is not requested.
15049 @item -marclinux_prof
15050 @opindex marclinux_prof
15051 Passed through to the linker, to specify use of the
15052 @code{arclinux_prof} emulation. This option is enabled by default in
15053 tool chains built for @w{@code{arc-linux-uclibc}} and
15054 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
15058 The following options control the semantics of generated code:
15060 @c semantically relevant code generation options
15063 @opindex mlong-calls
15064 Generate calls as register indirect calls, thus providing access
15065 to the full 32-bit address range.
15067 @item -mmedium-calls
15068 @opindex mmedium-calls
15069 Don't use less than 25-bit addressing range for calls, which is the
15070 offset available for an unconditional branch-and-link
15071 instruction. Conditional execution of function calls is suppressed, to
15072 allow use of the 25-bit range, rather than the 21-bit range with
15073 conditional branch-and-link. This is the default for tool chains built
15074 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
15078 Put definitions of externally-visible data in a small data section if
15079 that data is no bigger than @var{num} bytes. The default value of
15080 @var{num} is 4 for any ARC configuration, or 8 when we have double
15081 load/store operations.
15085 Do not generate sdata references. This is the default for tool chains
15086 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
15089 @item -mvolatile-cache
15090 @opindex mvolatile-cache
15091 Use ordinarily cached memory accesses for volatile references. This is the
15094 @item -mno-volatile-cache
15095 @opindex mno-volatile-cache
15096 Enable cache bypass for volatile references.
15100 The following options fine tune code generation:
15101 @c code generation tuning options
15104 @opindex malign-call
15105 Do alignment optimizations for call instructions.
15107 @item -mauto-modify-reg
15108 @opindex mauto-modify-reg
15109 Enable the use of pre/post modify with register displacement.
15111 @item -mbbit-peephole
15112 @opindex mbbit-peephole
15113 Enable bbit peephole2.
15117 This option disables a target-specific pass in @file{arc_reorg} to
15118 generate compare-and-branch (@code{br@var{cc}}) instructions.
15119 It has no effect on
15120 generation of these instructions driven by the combiner pass.
15122 @item -mcase-vector-pcrel
15123 @opindex mcase-vector-pcrel
15124 Use PC-relative switch case tables to enable case table shortening.
15125 This is the default for @option{-Os}.
15127 @item -mcompact-casesi
15128 @opindex mcompact-casesi
15129 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
15130 and only available for ARCv1 cores.
15132 @item -mno-cond-exec
15133 @opindex mno-cond-exec
15134 Disable the ARCompact-specific pass to generate conditional
15135 execution instructions.
15137 Due to delay slot scheduling and interactions between operand numbers,
15138 literal sizes, instruction lengths, and the support for conditional execution,
15139 the target-independent pass to generate conditional execution is often lacking,
15140 so the ARC port has kept a special pass around that tries to find more
15141 conditional execution generation opportunities after register allocation,
15142 branch shortening, and delay slot scheduling have been done. This pass
15143 generally, but not always, improves performance and code size, at the cost of
15144 extra compilation time, which is why there is an option to switch it off.
15145 If you have a problem with call instructions exceeding their allowable
15146 offset range because they are conditionalized, you should consider using
15147 @option{-mmedium-calls} instead.
15149 @item -mearly-cbranchsi
15150 @opindex mearly-cbranchsi
15151 Enable pre-reload use of the @code{cbranchsi} pattern.
15153 @item -mexpand-adddi
15154 @opindex mexpand-adddi
15155 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
15156 @code{add.f}, @code{adc} etc. This option is deprecated.
15158 @item -mindexed-loads
15159 @opindex mindexed-loads
15160 Enable the use of indexed loads. This can be problematic because some
15161 optimizers then assume that indexed stores exist, which is not
15165 Enable Local Register Allocation. This is still experimental for ARC,
15166 so by default the compiler uses standard reload
15167 (i.e. @option{-mno-lra}).
15169 @item -mlra-priority-none
15170 @opindex mlra-priority-none
15171 Don't indicate any priority for target registers.
15173 @item -mlra-priority-compact
15174 @opindex mlra-priority-compact
15175 Indicate target register priority for r0..r3 / r12..r15.
15177 @item -mlra-priority-noncompact
15178 @opindex mlra-priority-noncompact
15179 Reduce target register priority for r0..r3 / r12..r15.
15181 @item -mno-millicode
15182 @opindex mno-millicode
15183 When optimizing for size (using @option{-Os}), prologues and epilogues
15184 that have to save or restore a large number of registers are often
15185 shortened by using call to a special function in libgcc; this is
15186 referred to as a @emph{millicode} call. As these calls can pose
15187 performance issues, and/or cause linking issues when linking in a
15188 nonstandard way, this option is provided to turn off millicode call
15192 @opindex mmixed-code
15193 Tweak register allocation to help 16-bit instruction generation.
15194 This generally has the effect of decreasing the average instruction size
15195 while increasing the instruction count.
15199 Enable @samp{q} instruction alternatives.
15200 This is the default for @option{-Os}.
15204 Enable @samp{Rcq} constraint handling.
15205 Most short code generation depends on this.
15206 This is the default.
15210 Enable @samp{Rcw} constraint handling.
15211 Most ccfsm condexec mostly depends on this.
15212 This is the default.
15214 @item -msize-level=@var{level}
15215 @opindex msize-level
15216 Fine-tune size optimization with regards to instruction lengths and alignment.
15217 The recognized values for @var{level} are:
15220 No size optimization. This level is deprecated and treated like @samp{1}.
15223 Short instructions are used opportunistically.
15226 In addition, alignment of loops and of code after barriers are dropped.
15229 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
15233 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
15234 the behavior when this is not set is equivalent to level @samp{1}.
15236 @item -mtune=@var{cpu}
15238 Set instruction scheduling parameters for @var{cpu}, overriding any implied
15239 by @option{-mcpu=}.
15241 Supported values for @var{cpu} are
15245 Tune for ARC600 CPU.
15248 Tune for ARC601 CPU.
15251 Tune for ARC700 CPU with standard multiplier block.
15254 Tune for ARC700 CPU with XMAC block.
15257 Tune for ARC725D CPU.
15260 Tune for ARC750D CPU.
15264 @item -mmultcost=@var{num}
15266 Cost to assume for a multiply instruction, with @samp{4} being equal to a
15267 normal instruction.
15269 @item -munalign-prob-threshold=@var{probability}
15270 @opindex munalign-prob-threshold
15271 Set probability threshold for unaligning branches.
15272 When tuning for @samp{ARC700} and optimizing for speed, branches without
15273 filled delay slot are preferably emitted unaligned and long, unless
15274 profiling indicates that the probability for the branch to be taken
15275 is below @var{probability}. @xref{Cross-profiling}.
15276 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
15280 The following options are maintained for backward compatibility, but
15281 are now deprecated and will be removed in a future release:
15283 @c Deprecated options
15291 @opindex mbig-endian
15294 Compile code for big-endian targets. Use of these options is now
15295 deprecated. Big-endian code is supported by configuring GCC to build
15296 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
15297 for which big endian is the default.
15299 @item -mlittle-endian
15300 @opindex mlittle-endian
15303 Compile code for little-endian targets. Use of these options is now
15304 deprecated. Little-endian code is supported by configuring GCC to build
15305 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
15306 for which little endian is the default.
15308 @item -mbarrel_shifter
15309 @opindex mbarrel_shifter
15310 Replaced by @option{-mbarrel-shifter}.
15312 @item -mdpfp_compact
15313 @opindex mdpfp_compact
15314 Replaced by @option{-mdpfp-compact}.
15317 @opindex mdpfp_fast
15318 Replaced by @option{-mdpfp-fast}.
15321 @opindex mdsp_packa
15322 Replaced by @option{-mdsp-packa}.
15326 Replaced by @option{-mea}.
15330 Replaced by @option{-mmac-24}.
15334 Replaced by @option{-mmac-d16}.
15336 @item -mspfp_compact
15337 @opindex mspfp_compact
15338 Replaced by @option{-mspfp-compact}.
15341 @opindex mspfp_fast
15342 Replaced by @option{-mspfp-fast}.
15344 @item -mtune=@var{cpu}
15346 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
15347 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
15348 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
15350 @item -multcost=@var{num}
15352 Replaced by @option{-mmultcost}.
15357 @subsection ARM Options
15358 @cindex ARM options
15360 These @samp{-m} options are defined for the ARM port:
15363 @item -mabi=@var{name}
15365 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
15366 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
15369 @opindex mapcs-frame
15370 Generate a stack frame that is compliant with the ARM Procedure Call
15371 Standard for all functions, even if this is not strictly necessary for
15372 correct execution of the code. Specifying @option{-fomit-frame-pointer}
15373 with this option causes the stack frames not to be generated for
15374 leaf functions. The default is @option{-mno-apcs-frame}.
15375 This option is deprecated.
15379 This is a synonym for @option{-mapcs-frame} and is deprecated.
15382 @c not currently implemented
15383 @item -mapcs-stack-check
15384 @opindex mapcs-stack-check
15385 Generate code to check the amount of stack space available upon entry to
15386 every function (that actually uses some stack space). If there is
15387 insufficient space available then either the function
15388 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
15389 called, depending upon the amount of stack space required. The runtime
15390 system is required to provide these functions. The default is
15391 @option{-mno-apcs-stack-check}, since this produces smaller code.
15393 @c not currently implemented
15394 @item -mapcs-reentrant
15395 @opindex mapcs-reentrant
15396 Generate reentrant, position-independent code. The default is
15397 @option{-mno-apcs-reentrant}.
15400 @item -mthumb-interwork
15401 @opindex mthumb-interwork
15402 Generate code that supports calling between the ARM and Thumb
15403 instruction sets. Without this option, on pre-v5 architectures, the
15404 two instruction sets cannot be reliably used inside one program. The
15405 default is @option{-mno-thumb-interwork}, since slightly larger code
15406 is generated when @option{-mthumb-interwork} is specified. In AAPCS
15407 configurations this option is meaningless.
15409 @item -mno-sched-prolog
15410 @opindex mno-sched-prolog
15411 Prevent the reordering of instructions in the function prologue, or the
15412 merging of those instruction with the instructions in the function's
15413 body. This means that all functions start with a recognizable set
15414 of instructions (or in fact one of a choice from a small set of
15415 different function prologues), and this information can be used to
15416 locate the start of functions inside an executable piece of code. The
15417 default is @option{-msched-prolog}.
15419 @item -mfloat-abi=@var{name}
15420 @opindex mfloat-abi
15421 Specifies which floating-point ABI to use. Permissible values
15422 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15424 Specifying @samp{soft} causes GCC to generate output containing
15425 library calls for floating-point operations.
15426 @samp{softfp} allows the generation of code using hardware floating-point
15427 instructions, but still uses the soft-float calling conventions.
15428 @samp{hard} allows generation of floating-point instructions
15429 and uses FPU-specific calling conventions.
15431 The default depends on the specific target configuration. Note that
15432 the hard-float and soft-float ABIs are not link-compatible; you must
15433 compile your entire program with the same ABI, and link with a
15434 compatible set of libraries.
15436 @item -mlittle-endian
15437 @opindex mlittle-endian
15438 Generate code for a processor running in little-endian mode. This is
15439 the default for all standard configurations.
15442 @opindex mbig-endian
15443 Generate code for a processor running in big-endian mode; the default is
15444 to compile code for a little-endian processor.
15449 When linking a big-endian image select between BE8 and BE32 formats.
15450 The option has no effect for little-endian images and is ignored. The
15451 default is dependent on the selected target architecture. For ARMv6
15452 and later architectures the default is BE8, for older architectures
15453 the default is BE32. BE32 format has been deprecated by ARM.
15455 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
15457 This specifies the name of the target ARM architecture. GCC uses this
15458 name to determine what kind of instructions it can emit when generating
15459 assembly code. This option can be used in conjunction with or instead
15460 of the @option{-mcpu=} option.
15462 Permissible names are:
15464 @samp{armv5t}, @samp{armv5te},
15465 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
15466 @samp{armv6z}, @samp{armv6zk},
15467 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
15468 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a},
15471 @samp{armv6-m}, @samp{armv6s-m},
15472 @samp{armv7-m}, @samp{armv7e-m},
15473 @samp{armv8-m.base}, @samp{armv8-m.main},
15474 @samp{iwmmxt} and @samp{iwmmxt2}.
15476 Additionally, the following architectures, which lack support for the
15477 Thumb exection state, are recognized but support is deprecated:
15478 @samp{armv2}, @samp{armv2a}, @samp{armv3}, @samp{armv3m},
15479 @samp{armv4}, @samp{armv5} and @samp{armv5e}.
15481 Many of the architectures support extensions. These can be added by
15482 appending @samp{+@var{extension}} to the architecture name. Extension
15483 options are processed in order and capabilities accumulate. An extension
15484 will also enable any necessary base extensions
15485 upon which it depends. For example, the @samp{+crypto} extension
15486 will always enable the @samp{+simd} extension. The exception to the
15487 additive construction is for extensions that are prefixed with
15488 @samp{+no@dots{}}: these extensions disable the specified option and
15489 any other extensions that may depend on the presence of that
15492 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
15493 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
15494 entirely disabled by the @samp{+nofp} option that follows it.
15496 Most extension names are generically named, but have an effect that is
15497 dependent upon the architecture to which it is applied. For example,
15498 the @samp{+simd} option can be applied to both @samp{armv7-a} and
15499 @samp{armv8-a} architectures, but will enable the original ARMv7
15500 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-a
15501 variant for @samp{armv8-a}.
15503 The table below lists the supported extensions for each architecture.
15504 Architectures not mentioned do not support any extensions.
15518 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
15519 used as an alias for this extension.
15522 Disable the floating-point instructions.
15526 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
15529 The VFPv3 floating-point instructions, with 16 double-precision
15530 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15531 for this extension. Note that floating-point is not supported by the
15532 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
15533 ARMv7-R architectures.
15536 Disable the floating-point instructions.
15542 The VFPv3 floating-point instructions, with 16 double-precision
15543 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15544 for this extension.
15547 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15548 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
15549 for this extension.
15552 The VFPv3 floating-point instructions, with 32 double-precision
15555 @item +vfpv3-d16-fp16
15556 The VFPv3 floating-point instructions, with 16 double-precision
15557 registers and the half-precision floating-point conversion operations.
15560 The VFPv3 floating-point instructions, with 32 double-precision
15561 registers and the half-precision floating-point conversion operations.
15564 The VFPv4 floating-point instructions, with 16 double-precision
15568 The VFPv4 floating-point instructions, with 32 double-precision
15572 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15573 the half-precision floating-point conversion operations.
15576 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
15579 Disable the Advanced SIMD instructions (does not disable floating point).
15582 Disable the floating-point and Advanced SIMD instructions.
15586 The extended version of the ARMv7-A architecture with support for
15590 The VFPv4 floating-point instructions, with 16 double-precision registers.
15591 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
15594 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
15595 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
15598 The VFPv3 floating-point instructions, with 16 double-precision
15602 The VFPv3 floating-point instructions, with 32 double-precision
15605 @item +vfpv3-d16-fp16
15606 The VFPv3 floating-point instructions, with 16 double-precision
15607 registers and the half-precision floating-point conversion operations.
15610 The VFPv3 floating-point instructions, with 32 double-precision
15611 registers and the half-precision floating-point conversion operations.
15614 The VFPv4 floating-point instructions, with 16 double-precision
15618 The VFPv4 floating-point instructions, with 32 double-precision
15622 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15623 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
15626 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15627 the half-precision floating-point conversion operations.
15630 Disable the Advanced SIMD instructions (does not disable floating point).
15633 Disable the floating-point and Advanced SIMD instructions.
15639 The Cyclic Redundancy Check (CRC) instructions.
15641 The ARMv8 Advanced SIMD and floating-point instructions.
15643 The cryptographic instructions.
15645 Disable the cryptographic isntructions.
15647 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15653 The ARMv8.1 Advanced SIMD and floating-point instructions.
15656 The cryptographic instructions. This also enables the Advanced SIMD and
15657 floating-point instructions.
15660 Disable the cryptographic isntructions.
15663 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15669 The half-precision floating-point data processing instructions.
15670 This also enables the Advanced SIMD and floating-point instructions.
15673 The ARMv8.1 Advanced SIMD and floating-point instructions.
15676 The cryptographic instructions. This also enables the Advanced SIMD and
15677 floating-point instructions.
15680 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
15683 Disable the cryptographic extension.
15686 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15692 The single-precision VFPv3 floating-point instructions. The extension
15693 @samp{+vfpv3xd} can be used as an alias for this extension.
15696 The VFPv3 floating-point instructions with 16 double-precision registers.
15697 The extension +vfpv3-d16 can be used as an alias for this extension.
15700 Disable the floating-point extension.
15703 The ARM-state integer division instructions.
15706 Disable the ARM-state integer division extension.
15712 The single-precision VFPv4 floating-point instructions.
15715 The single-precision FPv5 floating-point instructions.
15718 The single- and double-precision FPv5 floating-point instructions.
15721 Disable the floating-point extensions.
15727 The DSP instructions.
15730 Disable the DSP extension.
15733 The single-precision floating-point instructions.
15736 The single- and double-precision floating-point instructions.
15739 Disable the floating-point extension.
15745 The Cyclic Redundancy Check (CRC) instructions.
15747 The single-precision FPv5 floating-point instructions.
15749 The ARMv8 Advanced SIMD and floating-point instructions.
15751 The cryptographic instructions.
15753 Disable the cryptographic isntructions.
15755 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15760 @option{-march=native} causes the compiler to auto-detect the architecture
15761 of the build computer. At present, this feature is only supported on
15762 GNU/Linux, and not all architectures are recognized. If the auto-detect
15763 is unsuccessful the option has no effect.
15765 @item -mtune=@var{name}
15767 This option specifies the name of the target ARM processor for
15768 which GCC should tune the performance of the code.
15769 For some ARM implementations better performance can be obtained by using
15771 Permissible names are: @samp{arm2}, @samp{arm250},
15772 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
15773 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
15774 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
15775 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
15777 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
15778 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
15779 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
15780 @samp{strongarm1110},
15781 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
15782 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
15783 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
15784 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
15785 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
15786 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
15787 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
15788 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
15789 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
15790 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15791 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15792 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7},
15793 @samp{cortex-r8}, @samp{cortex-r52},
15801 @samp{cortex-m0plus},
15802 @samp{cortex-m1.small-multiply},
15803 @samp{cortex-m0.small-multiply},
15804 @samp{cortex-m0plus.small-multiply},
15806 @samp{marvell-pj4},
15807 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
15808 @samp{fa526}, @samp{fa626},
15809 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
15812 Additionally, this option can specify that GCC should tune the performance
15813 of the code for a big.LITTLE system. Permissible names are:
15814 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
15815 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15816 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
15817 @samp{cortex-a75.cortex-a55}.
15819 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
15820 performance for a blend of processors within architecture @var{arch}.
15821 The aim is to generate code that run well on the current most popular
15822 processors, balancing between optimizations that benefit some CPUs in the
15823 range, and avoiding performance pitfalls of other CPUs. The effects of
15824 this option may change in future GCC versions as CPU models come and go.
15826 @option{-mtune} permits the same extension options as @option{-mcpu}, but
15827 the extension options do not affect the tuning of the generated code.
15829 @option{-mtune=native} causes the compiler to auto-detect the CPU
15830 of the build computer. At present, this feature is only supported on
15831 GNU/Linux, and not all architectures are recognized. If the auto-detect is
15832 unsuccessful the option has no effect.
15834 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
15836 This specifies the name of the target ARM processor. GCC uses this name
15837 to derive the name of the target ARM architecture (as if specified
15838 by @option{-march}) and the ARM processor type for which to tune for
15839 performance (as if specified by @option{-mtune}). Where this option
15840 is used in conjunction with @option{-march} or @option{-mtune},
15841 those options take precedence over the appropriate part of this option.
15843 Many of the supported CPUs implement optional architectural
15844 extensions. Where this is so the architectural extensions are
15845 normally enabled by default. If implementations that lack the
15846 extension exist, then the extension syntax can be used to disable
15847 those extensions that have been omitted. For floating-point and
15848 Advanced SIMD (Neon) instructions, the settings of the options
15849 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
15850 floating-point and Advanced SIMD instructions will only be used if
15851 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
15852 @option{-mfpu} other than @samp{auto} will override the available
15853 floating-point and SIMD extension instructions.
15855 For example, @samp{cortex-a9} can be found in three major
15856 configurations: integer only, with just a floating-point unit or with
15857 floating-point and Advanced SIMD. The default is to enable all the
15858 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
15859 be used to disable just the SIMD or both the SIMD and floating-point
15860 instructions respectively.
15862 Permissible names for this option are the same as those for
15865 The following extension options are common to the listed CPUs:
15869 Disable the DSP instructions on @samp{cortex-m33}.
15872 Disables the floating-point instructions on @samp{arm9e},
15873 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
15874 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
15875 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
15876 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
15877 Disables the floating-point and SIMD instructions on
15878 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
15879 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
15880 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
15881 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
15882 @samp{cortex-a53} and @samp{cortex-a55}.
15885 Disables the double-precision component of the floating-point instructions
15886 on @samp{cortex-r5}, @samp{cortex-r52} and @samp{cortex-m7}.
15889 Disables the SIMD (but not floating-point) instructions on
15890 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
15891 and @samp{cortex-a9}.
15894 Enables the cryptographic instructions on @samp{cortex-a32},
15895 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
15896 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
15897 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15898 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
15899 @samp{cortex-a75.cortex-a55}.
15902 Additionally the @samp{generic-armv7-a} pseudo target defaults to
15903 VFPv3 with 16 double-precision registers. It supports the following
15904 extension options: @samp{vfpv3-d16}, @samp{vfpv3},
15905 @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16}, @samp{vfpv4-d16},
15906 @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3}, @samp{neon-fp16},
15907 @samp{neon-vfpv4}. The meanings are the same as for the extensions to
15908 @option{-march=armv7-a}.
15910 @option{-mcpu=generic-@var{arch}} is also permissible, and is
15911 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
15912 See @option{-mtune} for more information.
15914 @option{-mcpu=native} causes the compiler to auto-detect the CPU
15915 of the build computer. At present, this feature is only supported on
15916 GNU/Linux, and not all architectures are recognized. If the auto-detect
15917 is unsuccessful the option has no effect.
15919 @item -mfpu=@var{name}
15921 This specifies what floating-point hardware (or hardware emulation) is
15922 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
15924 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
15925 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
15926 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
15927 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
15928 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
15929 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
15930 is an alias for @samp{vfpv2}.
15932 The setting @samp{auto} is the default and is special. It causes the
15933 compiler to select the floating-point and Advanced SIMD instructions
15934 based on the settings of @option{-mcpu} and @option{-march}.
15936 If the selected floating-point hardware includes the NEON extension
15937 (e.g. @option{-mfpu=neon}), note that floating-point
15938 operations are not generated by GCC's auto-vectorization pass unless
15939 @option{-funsafe-math-optimizations} is also specified. This is
15940 because NEON hardware does not fully implement the IEEE 754 standard for
15941 floating-point arithmetic (in particular denormal values are treated as
15942 zero), so the use of NEON instructions may lead to a loss of precision.
15944 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}).
15946 @item -mfp16-format=@var{name}
15947 @opindex mfp16-format
15948 Specify the format of the @code{__fp16} half-precision floating-point type.
15949 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
15950 the default is @samp{none}, in which case the @code{__fp16} type is not
15951 defined. @xref{Half-Precision}, for more information.
15953 @item -mstructure-size-boundary=@var{n}
15954 @opindex mstructure-size-boundary
15955 The sizes of all structures and unions are rounded up to a multiple
15956 of the number of bits set by this option. Permissible values are 8, 32
15957 and 64. The default value varies for different toolchains. For the COFF
15958 targeted toolchain the default value is 8. A value of 64 is only allowed
15959 if the underlying ABI supports it.
15961 Specifying a larger number can produce faster, more efficient code, but
15962 can also increase the size of the program. Different values are potentially
15963 incompatible. Code compiled with one value cannot necessarily expect to
15964 work with code or libraries compiled with another value, if they exchange
15965 information using structures or unions.
15967 This option is deprecated.
15969 @item -mabort-on-noreturn
15970 @opindex mabort-on-noreturn
15971 Generate a call to the function @code{abort} at the end of a
15972 @code{noreturn} function. It is executed if the function tries to
15976 @itemx -mno-long-calls
15977 @opindex mlong-calls
15978 @opindex mno-long-calls
15979 Tells the compiler to perform function calls by first loading the
15980 address of the function into a register and then performing a subroutine
15981 call on this register. This switch is needed if the target function
15982 lies outside of the 64-megabyte addressing range of the offset-based
15983 version of subroutine call instruction.
15985 Even if this switch is enabled, not all function calls are turned
15986 into long calls. The heuristic is that static functions, functions
15987 that have the @code{short_call} attribute, functions that are inside
15988 the scope of a @code{#pragma no_long_calls} directive, and functions whose
15989 definitions have already been compiled within the current compilation
15990 unit are not turned into long calls. The exceptions to this rule are
15991 that weak function definitions, functions with the @code{long_call}
15992 attribute or the @code{section} attribute, and functions that are within
15993 the scope of a @code{#pragma long_calls} directive are always
15994 turned into long calls.
15996 This feature is not enabled by default. Specifying
15997 @option{-mno-long-calls} restores the default behavior, as does
15998 placing the function calls within the scope of a @code{#pragma
15999 long_calls_off} directive. Note these switches have no effect on how
16000 the compiler generates code to handle function calls via function
16003 @item -msingle-pic-base
16004 @opindex msingle-pic-base
16005 Treat the register used for PIC addressing as read-only, rather than
16006 loading it in the prologue for each function. The runtime system is
16007 responsible for initializing this register with an appropriate value
16008 before execution begins.
16010 @item -mpic-register=@var{reg}
16011 @opindex mpic-register
16012 Specify the register to be used for PIC addressing.
16013 For standard PIC base case, the default is any suitable register
16014 determined by compiler. For single PIC base case, the default is
16015 @samp{R9} if target is EABI based or stack-checking is enabled,
16016 otherwise the default is @samp{R10}.
16018 @item -mpic-data-is-text-relative
16019 @opindex mpic-data-is-text-relative
16020 Assume that the displacement between the text and data segments is fixed
16021 at static link time. This permits using PC-relative addressing
16022 operations to access data known to be in the data segment. For
16023 non-VxWorks RTP targets, this option is enabled by default. When
16024 disabled on such targets, it will enable @option{-msingle-pic-base} by
16027 @item -mpoke-function-name
16028 @opindex mpoke-function-name
16029 Write the name of each function into the text section, directly
16030 preceding the function prologue. The generated code is similar to this:
16034 .ascii "arm_poke_function_name", 0
16037 .word 0xff000000 + (t1 - t0)
16038 arm_poke_function_name
16040 stmfd sp!, @{fp, ip, lr, pc@}
16044 When performing a stack backtrace, code can inspect the value of
16045 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
16046 location @code{pc - 12} and the top 8 bits are set, then we know that
16047 there is a function name embedded immediately preceding this location
16048 and has length @code{((pc[-3]) & 0xff000000)}.
16055 Select between generating code that executes in ARM and Thumb
16056 states. The default for most configurations is to generate code
16057 that executes in ARM state, but the default can be changed by
16058 configuring GCC with the @option{--with-mode=}@var{state}
16061 You can also override the ARM and Thumb mode for each function
16062 by using the @code{target("thumb")} and @code{target("arm")} function attributes
16063 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
16066 @opindex mtpcs-frame
16067 Generate a stack frame that is compliant with the Thumb Procedure Call
16068 Standard for all non-leaf functions. (A leaf function is one that does
16069 not call any other functions.) The default is @option{-mno-tpcs-frame}.
16071 @item -mtpcs-leaf-frame
16072 @opindex mtpcs-leaf-frame
16073 Generate a stack frame that is compliant with the Thumb Procedure Call
16074 Standard for all leaf functions. (A leaf function is one that does
16075 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
16077 @item -mcallee-super-interworking
16078 @opindex mcallee-super-interworking
16079 Gives all externally visible functions in the file being compiled an ARM
16080 instruction set header which switches to Thumb mode before executing the
16081 rest of the function. This allows these functions to be called from
16082 non-interworking code. This option is not valid in AAPCS configurations
16083 because interworking is enabled by default.
16085 @item -mcaller-super-interworking
16086 @opindex mcaller-super-interworking
16087 Allows calls via function pointers (including virtual functions) to
16088 execute correctly regardless of whether the target code has been
16089 compiled for interworking or not. There is a small overhead in the cost
16090 of executing a function pointer if this option is enabled. This option
16091 is not valid in AAPCS configurations because interworking is enabled
16094 @item -mtp=@var{name}
16096 Specify the access model for the thread local storage pointer. The valid
16097 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
16098 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
16099 (supported in the arm6k architecture), and @samp{auto}, which uses the
16100 best available method for the selected processor. The default setting is
16103 @item -mtls-dialect=@var{dialect}
16104 @opindex mtls-dialect
16105 Specify the dialect to use for accessing thread local storage. Two
16106 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
16107 @samp{gnu} dialect selects the original GNU scheme for supporting
16108 local and global dynamic TLS models. The @samp{gnu2} dialect
16109 selects the GNU descriptor scheme, which provides better performance
16110 for shared libraries. The GNU descriptor scheme is compatible with
16111 the original scheme, but does require new assembler, linker and
16112 library support. Initial and local exec TLS models are unaffected by
16113 this option and always use the original scheme.
16115 @item -mword-relocations
16116 @opindex mword-relocations
16117 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
16118 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
16119 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
16122 @item -mfix-cortex-m3-ldrd
16123 @opindex mfix-cortex-m3-ldrd
16124 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
16125 with overlapping destination and base registers are used. This option avoids
16126 generating these instructions. This option is enabled by default when
16127 @option{-mcpu=cortex-m3} is specified.
16129 @item -munaligned-access
16130 @itemx -mno-unaligned-access
16131 @opindex munaligned-access
16132 @opindex mno-unaligned-access
16133 Enables (or disables) reading and writing of 16- and 32- bit values
16134 from addresses that are not 16- or 32- bit aligned. By default
16135 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
16136 ARMv8-M Baseline architectures, and enabled for all other
16137 architectures. If unaligned access is not enabled then words in packed
16138 data structures are accessed a byte at a time.
16140 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
16141 generated object file to either true or false, depending upon the
16142 setting of this option. If unaligned access is enabled then the
16143 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
16146 @item -mneon-for-64bits
16147 @opindex mneon-for-64bits
16148 Enables using Neon to handle scalar 64-bits operations. This is
16149 disabled by default since the cost of moving data from core registers
16152 @item -mslow-flash-data
16153 @opindex mslow-flash-data
16154 Assume loading data from flash is slower than fetching instruction.
16155 Therefore literal load is minimized for better performance.
16156 This option is only supported when compiling for ARMv7 M-profile and
16159 @item -masm-syntax-unified
16160 @opindex masm-syntax-unified
16161 Assume inline assembler is using unified asm syntax. The default is
16162 currently off which implies divided syntax. This option has no impact
16163 on Thumb2. However, this may change in future releases of GCC.
16164 Divided syntax should be considered deprecated.
16166 @item -mrestrict-it
16167 @opindex mrestrict-it
16168 Restricts generation of IT blocks to conform to the rules of ARMv8.
16169 IT blocks can only contain a single 16-bit instruction from a select
16170 set of instructions. This option is on by default for ARMv8 Thumb mode.
16172 @item -mprint-tune-info
16173 @opindex mprint-tune-info
16174 Print CPU tuning information as comment in assembler file. This is
16175 an option used only for regression testing of the compiler and not
16176 intended for ordinary use in compiling code. This option is disabled
16180 @opindex mpure-code
16181 Do not allow constant data to be placed in code sections.
16182 Additionally, when compiling for ELF object format give all text sections the
16183 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
16184 is only available when generating non-pic code for M-profile targets with the
16189 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
16190 Development Tools Engineering Specification", which can be found on
16191 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
16195 @subsection AVR Options
16196 @cindex AVR Options
16198 These options are defined for AVR implementations:
16201 @item -mmcu=@var{mcu}
16203 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
16205 The default for this option is@tie{}@samp{avr2}.
16207 GCC supports the following AVR devices and ISAs:
16209 @include avr-mmcu.texi
16214 Assume that all data in static storage can be accessed by LDS / STS
16215 instructions. This option has only an effect on reduced Tiny devices like
16216 ATtiny40. See also the @code{absdata}
16217 @ref{AVR Variable Attributes,variable attribute}.
16219 @item -maccumulate-args
16220 @opindex maccumulate-args
16221 Accumulate outgoing function arguments and acquire/release the needed
16222 stack space for outgoing function arguments once in function
16223 prologue/epilogue. Without this option, outgoing arguments are pushed
16224 before calling a function and popped afterwards.
16226 Popping the arguments after the function call can be expensive on
16227 AVR so that accumulating the stack space might lead to smaller
16228 executables because arguments need not be removed from the
16229 stack after such a function call.
16231 This option can lead to reduced code size for functions that perform
16232 several calls to functions that get their arguments on the stack like
16233 calls to printf-like functions.
16235 @item -mbranch-cost=@var{cost}
16236 @opindex mbranch-cost
16237 Set the branch costs for conditional branch instructions to
16238 @var{cost}. Reasonable values for @var{cost} are small, non-negative
16239 integers. The default branch cost is 0.
16241 @item -mcall-prologues
16242 @opindex mcall-prologues
16243 Functions prologues/epilogues are expanded as calls to appropriate
16244 subroutines. Code size is smaller.
16246 @item -mgas-isr-prologues
16247 @opindex mgas-isr-prologues
16248 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
16249 instruction supported by GNU Binutils.
16250 If this option is on, the feature can still be disabled for individual
16251 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
16252 function attribute. This feature is activated per default
16253 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
16254 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
16258 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
16259 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
16260 and @code{long long} is 4 bytes. Please note that this option does not
16261 conform to the C standards, but it results in smaller code
16264 @item -mn-flash=@var{num}
16266 Assume that the flash memory has a size of
16267 @var{num} times 64@tie{}KiB.
16269 @item -mno-interrupts
16270 @opindex mno-interrupts
16271 Generated code is not compatible with hardware interrupts.
16272 Code size is smaller.
16276 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
16277 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
16278 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
16279 the assembler's command line and the @option{--relax} option to the
16280 linker's command line.
16282 Jump relaxing is performed by the linker because jump offsets are not
16283 known before code is located. Therefore, the assembler code generated by the
16284 compiler is the same, but the instructions in the executable may
16285 differ from instructions in the assembler code.
16287 Relaxing must be turned on if linker stubs are needed, see the
16288 section on @code{EIND} and linker stubs below.
16292 Assume that the device supports the Read-Modify-Write
16293 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
16295 @item -mshort-calls
16296 @opindex mshort-calls
16298 Assume that @code{RJMP} and @code{RCALL} can target the whole
16301 This option is used internally for multilib selection. It is
16302 not an optimization option, and you don't need to set it by hand.
16306 Treat the stack pointer register as an 8-bit register,
16307 i.e.@: assume the high byte of the stack pointer is zero.
16308 In general, you don't need to set this option by hand.
16310 This option is used internally by the compiler to select and
16311 build multilibs for architectures @code{avr2} and @code{avr25}.
16312 These architectures mix devices with and without @code{SPH}.
16313 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
16314 the compiler driver adds or removes this option from the compiler
16315 proper's command line, because the compiler then knows if the device
16316 or architecture has an 8-bit stack pointer and thus no @code{SPH}
16321 Use address register @code{X} in a way proposed by the hardware. This means
16322 that @code{X} is only used in indirect, post-increment or
16323 pre-decrement addressing.
16325 Without this option, the @code{X} register may be used in the same way
16326 as @code{Y} or @code{Z} which then is emulated by additional
16328 For example, loading a value with @code{X+const} addressing with a
16329 small non-negative @code{const < 64} to a register @var{Rn} is
16333 adiw r26, const ; X += const
16334 ld @var{Rn}, X ; @var{Rn} = *X
16335 sbiw r26, const ; X -= const
16339 @opindex mtiny-stack
16340 Only change the lower 8@tie{}bits of the stack pointer.
16342 @item -mfract-convert-truncate
16343 @opindex mfract-convert-truncate
16344 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
16347 @opindex nodevicelib
16348 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
16350 @item -Waddr-space-convert
16351 @opindex Waddr-space-convert
16352 Warn about conversions between address spaces in the case where the
16353 resulting address space is not contained in the incoming address space.
16355 @item -Wmisspelled-isr
16356 @opindex Wmisspelled-isr
16357 Warn if the ISR is misspelled, i.e. without __vector prefix.
16358 Enabled by default.
16361 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
16362 @cindex @code{EIND}
16363 Pointers in the implementation are 16@tie{}bits wide.
16364 The address of a function or label is represented as word address so
16365 that indirect jumps and calls can target any code address in the
16366 range of 64@tie{}Ki words.
16368 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
16369 bytes of program memory space, there is a special function register called
16370 @code{EIND} that serves as most significant part of the target address
16371 when @code{EICALL} or @code{EIJMP} instructions are used.
16373 Indirect jumps and calls on these devices are handled as follows by
16374 the compiler and are subject to some limitations:
16379 The compiler never sets @code{EIND}.
16382 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
16383 instructions or might read @code{EIND} directly in order to emulate an
16384 indirect call/jump by means of a @code{RET} instruction.
16387 The compiler assumes that @code{EIND} never changes during the startup
16388 code or during the application. In particular, @code{EIND} is not
16389 saved/restored in function or interrupt service routine
16393 For indirect calls to functions and computed goto, the linker
16394 generates @emph{stubs}. Stubs are jump pads sometimes also called
16395 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
16396 The stub contains a direct jump to the desired address.
16399 Linker relaxation must be turned on so that the linker generates
16400 the stubs correctly in all situations. See the compiler option
16401 @option{-mrelax} and the linker option @option{--relax}.
16402 There are corner cases where the linker is supposed to generate stubs
16403 but aborts without relaxation and without a helpful error message.
16406 The default linker script is arranged for code with @code{EIND = 0}.
16407 If code is supposed to work for a setup with @code{EIND != 0}, a custom
16408 linker script has to be used in order to place the sections whose
16409 name start with @code{.trampolines} into the segment where @code{EIND}
16413 The startup code from libgcc never sets @code{EIND}.
16414 Notice that startup code is a blend of code from libgcc and AVR-LibC.
16415 For the impact of AVR-LibC on @code{EIND}, see the
16416 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
16419 It is legitimate for user-specific startup code to set up @code{EIND}
16420 early, for example by means of initialization code located in
16421 section @code{.init3}. Such code runs prior to general startup code
16422 that initializes RAM and calls constructors, but after the bit
16423 of startup code from AVR-LibC that sets @code{EIND} to the segment
16424 where the vector table is located.
16426 #include <avr/io.h>
16429 __attribute__((section(".init3"),naked,used,no_instrument_function))
16430 init3_set_eind (void)
16432 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
16433 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
16438 The @code{__trampolines_start} symbol is defined in the linker script.
16441 Stubs are generated automatically by the linker if
16442 the following two conditions are met:
16445 @item The address of a label is taken by means of the @code{gs} modifier
16446 (short for @emph{generate stubs}) like so:
16448 LDI r24, lo8(gs(@var{func}))
16449 LDI r25, hi8(gs(@var{func}))
16451 @item The final location of that label is in a code segment
16452 @emph{outside} the segment where the stubs are located.
16456 The compiler emits such @code{gs} modifiers for code labels in the
16457 following situations:
16459 @item Taking address of a function or code label.
16460 @item Computed goto.
16461 @item If prologue-save function is used, see @option{-mcall-prologues}
16462 command-line option.
16463 @item Switch/case dispatch tables. If you do not want such dispatch
16464 tables you can specify the @option{-fno-jump-tables} command-line option.
16465 @item C and C++ constructors/destructors called during startup/shutdown.
16466 @item If the tools hit a @code{gs()} modifier explained above.
16470 Jumping to non-symbolic addresses like so is @emph{not} supported:
16475 /* Call function at word address 0x2 */
16476 return ((int(*)(void)) 0x2)();
16480 Instead, a stub has to be set up, i.e.@: the function has to be called
16481 through a symbol (@code{func_4} in the example):
16486 extern int func_4 (void);
16488 /* Call function at byte address 0x4 */
16493 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
16494 Alternatively, @code{func_4} can be defined in the linker script.
16497 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
16498 @cindex @code{RAMPD}
16499 @cindex @code{RAMPX}
16500 @cindex @code{RAMPY}
16501 @cindex @code{RAMPZ}
16502 Some AVR devices support memories larger than the 64@tie{}KiB range
16503 that can be accessed with 16-bit pointers. To access memory locations
16504 outside this 64@tie{}KiB range, the content of a @code{RAMP}
16505 register is used as high part of the address:
16506 The @code{X}, @code{Y}, @code{Z} address register is concatenated
16507 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
16508 register, respectively, to get a wide address. Similarly,
16509 @code{RAMPD} is used together with direct addressing.
16513 The startup code initializes the @code{RAMP} special function
16514 registers with zero.
16517 If a @ref{AVR Named Address Spaces,named address space} other than
16518 generic or @code{__flash} is used, then @code{RAMPZ} is set
16519 as needed before the operation.
16522 If the device supports RAM larger than 64@tie{}KiB and the compiler
16523 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
16524 is reset to zero after the operation.
16527 If the device comes with a specific @code{RAMP} register, the ISR
16528 prologue/epilogue saves/restores that SFR and initializes it with
16529 zero in case the ISR code might (implicitly) use it.
16532 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
16533 If you use inline assembler to read from locations outside the
16534 16-bit address range and change one of the @code{RAMP} registers,
16535 you must reset it to zero after the access.
16539 @subsubsection AVR Built-in Macros
16541 GCC defines several built-in macros so that the user code can test
16542 for the presence or absence of features. Almost any of the following
16543 built-in macros are deduced from device capabilities and thus
16544 triggered by the @option{-mmcu=} command-line option.
16546 For even more AVR-specific built-in macros see
16547 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
16552 Build-in macro that resolves to a decimal number that identifies the
16553 architecture and depends on the @option{-mmcu=@var{mcu}} option.
16554 Possible values are:
16556 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
16557 @code{4}, @code{5}, @code{51}, @code{6}
16559 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
16560 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
16565 @code{102}, @code{103}, @code{104},
16566 @code{105}, @code{106}, @code{107}
16568 for @var{mcu}=@code{avrtiny},
16569 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
16570 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
16571 If @var{mcu} specifies a device, this built-in macro is set
16572 accordingly. For example, with @option{-mmcu=atmega8} the macro is
16573 defined to @code{4}.
16575 @item __AVR_@var{Device}__
16576 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
16577 the device's name. For example, @option{-mmcu=atmega8} defines the
16578 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
16579 @code{__AVR_ATtiny261A__}, etc.
16581 The built-in macros' names follow
16582 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
16583 the device name as from the AVR user manual. The difference between
16584 @var{Device} in the built-in macro and @var{device} in
16585 @option{-mmcu=@var{device}} is that the latter is always lowercase.
16587 If @var{device} is not a device but only a core architecture like
16588 @samp{avr51}, this macro is not defined.
16590 @item __AVR_DEVICE_NAME__
16591 Setting @option{-mmcu=@var{device}} defines this built-in macro to
16592 the device's name. For example, with @option{-mmcu=atmega8} the macro
16593 is defined to @code{atmega8}.
16595 If @var{device} is not a device but only a core architecture like
16596 @samp{avr51}, this macro is not defined.
16598 @item __AVR_XMEGA__
16599 The device / architecture belongs to the XMEGA family of devices.
16601 @item __AVR_HAVE_ELPM__
16602 The device has the @code{ELPM} instruction.
16604 @item __AVR_HAVE_ELPMX__
16605 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
16606 R@var{n},Z+} instructions.
16608 @item __AVR_HAVE_MOVW__
16609 The device has the @code{MOVW} instruction to perform 16-bit
16610 register-register moves.
16612 @item __AVR_HAVE_LPMX__
16613 The device has the @code{LPM R@var{n},Z} and
16614 @code{LPM R@var{n},Z+} instructions.
16616 @item __AVR_HAVE_MUL__
16617 The device has a hardware multiplier.
16619 @item __AVR_HAVE_JMP_CALL__
16620 The device has the @code{JMP} and @code{CALL} instructions.
16621 This is the case for devices with more than 8@tie{}KiB of program
16624 @item __AVR_HAVE_EIJMP_EICALL__
16625 @itemx __AVR_3_BYTE_PC__
16626 The device has the @code{EIJMP} and @code{EICALL} instructions.
16627 This is the case for devices with more than 128@tie{}KiB of program memory.
16628 This also means that the program counter
16629 (PC) is 3@tie{}bytes wide.
16631 @item __AVR_2_BYTE_PC__
16632 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
16633 with up to 128@tie{}KiB of program memory.
16635 @item __AVR_HAVE_8BIT_SP__
16636 @itemx __AVR_HAVE_16BIT_SP__
16637 The stack pointer (SP) register is treated as 8-bit respectively
16638 16-bit register by the compiler.
16639 The definition of these macros is affected by @option{-mtiny-stack}.
16641 @item __AVR_HAVE_SPH__
16643 The device has the SPH (high part of stack pointer) special function
16644 register or has an 8-bit stack pointer, respectively.
16645 The definition of these macros is affected by @option{-mmcu=} and
16646 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
16649 @item __AVR_HAVE_RAMPD__
16650 @itemx __AVR_HAVE_RAMPX__
16651 @itemx __AVR_HAVE_RAMPY__
16652 @itemx __AVR_HAVE_RAMPZ__
16653 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
16654 @code{RAMPZ} special function register, respectively.
16656 @item __NO_INTERRUPTS__
16657 This macro reflects the @option{-mno-interrupts} command-line option.
16659 @item __AVR_ERRATA_SKIP__
16660 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
16661 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
16662 instructions because of a hardware erratum. Skip instructions are
16663 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
16664 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
16667 @item __AVR_ISA_RMW__
16668 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
16670 @item __AVR_SFR_OFFSET__=@var{offset}
16671 Instructions that can address I/O special function registers directly
16672 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
16673 address as if addressed by an instruction to access RAM like @code{LD}
16674 or @code{STS}. This offset depends on the device architecture and has
16675 to be subtracted from the RAM address in order to get the
16676 respective I/O@tie{}address.
16678 @item __AVR_SHORT_CALLS__
16679 The @option{-mshort-calls} command line option is set.
16681 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
16682 Some devices support reading from flash memory by means of @code{LD*}
16683 instructions. The flash memory is seen in the data address space
16684 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
16685 is not defined, this feature is not available. If defined,
16686 the address space is linear and there is no need to put
16687 @code{.rodata} into RAM. This is handled by the default linker
16688 description file, and is currently available for
16689 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
16690 there is no need to use address spaces like @code{__flash} or
16691 features like attribute @code{progmem} and @code{pgm_read_*}.
16693 @item __WITH_AVRLIBC__
16694 The compiler is configured to be used together with AVR-Libc.
16695 See the @option{--with-avrlibc} configure option.
16699 @node Blackfin Options
16700 @subsection Blackfin Options
16701 @cindex Blackfin Options
16704 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
16706 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
16707 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
16708 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
16709 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
16710 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
16711 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
16712 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
16713 @samp{bf561}, @samp{bf592}.
16715 The optional @var{sirevision} specifies the silicon revision of the target
16716 Blackfin processor. Any workarounds available for the targeted silicon revision
16717 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
16718 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
16719 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
16720 hexadecimal digits representing the major and minor numbers in the silicon
16721 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
16722 is not defined. If @var{sirevision} is @samp{any}, the
16723 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
16724 If this optional @var{sirevision} is not used, GCC assumes the latest known
16725 silicon revision of the targeted Blackfin processor.
16727 GCC defines a preprocessor macro for the specified @var{cpu}.
16728 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
16729 provided by libgloss to be linked in if @option{-msim} is not given.
16731 Without this option, @samp{bf532} is used as the processor by default.
16733 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
16734 only the preprocessor macro is defined.
16738 Specifies that the program will be run on the simulator. This causes
16739 the simulator BSP provided by libgloss to be linked in. This option
16740 has effect only for @samp{bfin-elf} toolchain.
16741 Certain other options, such as @option{-mid-shared-library} and
16742 @option{-mfdpic}, imply @option{-msim}.
16744 @item -momit-leaf-frame-pointer
16745 @opindex momit-leaf-frame-pointer
16746 Don't keep the frame pointer in a register for leaf functions. This
16747 avoids the instructions to save, set up and restore frame pointers and
16748 makes an extra register available in leaf functions. The option
16749 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
16750 which might make debugging harder.
16752 @item -mspecld-anomaly
16753 @opindex mspecld-anomaly
16754 When enabled, the compiler ensures that the generated code does not
16755 contain speculative loads after jump instructions. If this option is used,
16756 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
16758 @item -mno-specld-anomaly
16759 @opindex mno-specld-anomaly
16760 Don't generate extra code to prevent speculative loads from occurring.
16762 @item -mcsync-anomaly
16763 @opindex mcsync-anomaly
16764 When enabled, the compiler ensures that the generated code does not
16765 contain CSYNC or SSYNC instructions too soon after conditional branches.
16766 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
16768 @item -mno-csync-anomaly
16769 @opindex mno-csync-anomaly
16770 Don't generate extra code to prevent CSYNC or SSYNC instructions from
16771 occurring too soon after a conditional branch.
16775 When enabled, the compiler is free to take advantage of the knowledge that
16776 the entire program fits into the low 64k of memory.
16779 @opindex mno-low-64k
16780 Assume that the program is arbitrarily large. This is the default.
16782 @item -mstack-check-l1
16783 @opindex mstack-check-l1
16784 Do stack checking using information placed into L1 scratchpad memory by the
16787 @item -mid-shared-library
16788 @opindex mid-shared-library
16789 Generate code that supports shared libraries via the library ID method.
16790 This allows for execute in place and shared libraries in an environment
16791 without virtual memory management. This option implies @option{-fPIC}.
16792 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16794 @item -mno-id-shared-library
16795 @opindex mno-id-shared-library
16796 Generate code that doesn't assume ID-based shared libraries are being used.
16797 This is the default.
16799 @item -mleaf-id-shared-library
16800 @opindex mleaf-id-shared-library
16801 Generate code that supports shared libraries via the library ID method,
16802 but assumes that this library or executable won't link against any other
16803 ID shared libraries. That allows the compiler to use faster code for jumps
16806 @item -mno-leaf-id-shared-library
16807 @opindex mno-leaf-id-shared-library
16808 Do not assume that the code being compiled won't link against any ID shared
16809 libraries. Slower code is generated for jump and call insns.
16811 @item -mshared-library-id=n
16812 @opindex mshared-library-id
16813 Specifies the identification number of the ID-based shared library being
16814 compiled. Specifying a value of 0 generates more compact code; specifying
16815 other values forces the allocation of that number to the current
16816 library but is no more space- or time-efficient than omitting this option.
16820 Generate code that allows the data segment to be located in a different
16821 area of memory from the text segment. This allows for execute in place in
16822 an environment without virtual memory management by eliminating relocations
16823 against the text section.
16825 @item -mno-sep-data
16826 @opindex mno-sep-data
16827 Generate code that assumes that the data segment follows the text segment.
16828 This is the default.
16831 @itemx -mno-long-calls
16832 @opindex mlong-calls
16833 @opindex mno-long-calls
16834 Tells the compiler to perform function calls by first loading the
16835 address of the function into a register and then performing a subroutine
16836 call on this register. This switch is needed if the target function
16837 lies outside of the 24-bit addressing range of the offset-based
16838 version of subroutine call instruction.
16840 This feature is not enabled by default. Specifying
16841 @option{-mno-long-calls} restores the default behavior. Note these
16842 switches have no effect on how the compiler generates code to handle
16843 function calls via function pointers.
16847 Link with the fast floating-point library. This library relaxes some of
16848 the IEEE floating-point standard's rules for checking inputs against
16849 Not-a-Number (NAN), in the interest of performance.
16852 @opindex minline-plt
16853 Enable inlining of PLT entries in function calls to functions that are
16854 not known to bind locally. It has no effect without @option{-mfdpic}.
16857 @opindex mmulticore
16858 Build a standalone application for multicore Blackfin processors.
16859 This option causes proper start files and link scripts supporting
16860 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
16861 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
16863 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
16864 selects the one-application-per-core programming model. Without
16865 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
16866 programming model is used. In this model, the main function of Core B
16867 should be named as @code{coreb_main}.
16869 If this option is not used, the single-core application programming
16874 Build a standalone application for Core A of BF561 when using
16875 the one-application-per-core programming model. Proper start files
16876 and link scripts are used to support Core A, and the macro
16877 @code{__BFIN_COREA} is defined.
16878 This option can only be used in conjunction with @option{-mmulticore}.
16882 Build a standalone application for Core B of BF561 when using
16883 the one-application-per-core programming model. Proper start files
16884 and link scripts are used to support Core B, and the macro
16885 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
16886 should be used instead of @code{main}.
16887 This option can only be used in conjunction with @option{-mmulticore}.
16891 Build a standalone application for SDRAM. Proper start files and
16892 link scripts are used to put the application into SDRAM, and the macro
16893 @code{__BFIN_SDRAM} is defined.
16894 The loader should initialize SDRAM before loading the application.
16898 Assume that ICPLBs are enabled at run time. This has an effect on certain
16899 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
16900 are enabled; for standalone applications the default is off.
16904 @subsection C6X Options
16905 @cindex C6X Options
16908 @item -march=@var{name}
16910 This specifies the name of the target architecture. GCC uses this
16911 name to determine what kind of instructions it can emit when generating
16912 assembly code. Permissible names are: @samp{c62x},
16913 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
16916 @opindex mbig-endian
16917 Generate code for a big-endian target.
16919 @item -mlittle-endian
16920 @opindex mlittle-endian
16921 Generate code for a little-endian target. This is the default.
16925 Choose startup files and linker script suitable for the simulator.
16927 @item -msdata=default
16928 @opindex msdata=default
16929 Put small global and static data in the @code{.neardata} section,
16930 which is pointed to by register @code{B14}. Put small uninitialized
16931 global and static data in the @code{.bss} section, which is adjacent
16932 to the @code{.neardata} section. Put small read-only data into the
16933 @code{.rodata} section. The corresponding sections used for large
16934 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
16937 @opindex msdata=all
16938 Put all data, not just small objects, into the sections reserved for
16939 small data, and use addressing relative to the @code{B14} register to
16943 @opindex msdata=none
16944 Make no use of the sections reserved for small data, and use absolute
16945 addresses to access all data. Put all initialized global and static
16946 data in the @code{.fardata} section, and all uninitialized data in the
16947 @code{.far} section. Put all constant data into the @code{.const}
16952 @subsection CRIS Options
16953 @cindex CRIS Options
16955 These options are defined specifically for the CRIS ports.
16958 @item -march=@var{architecture-type}
16959 @itemx -mcpu=@var{architecture-type}
16962 Generate code for the specified architecture. The choices for
16963 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
16964 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
16965 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
16968 @item -mtune=@var{architecture-type}
16970 Tune to @var{architecture-type} everything applicable about the generated
16971 code, except for the ABI and the set of available instructions. The
16972 choices for @var{architecture-type} are the same as for
16973 @option{-march=@var{architecture-type}}.
16975 @item -mmax-stack-frame=@var{n}
16976 @opindex mmax-stack-frame
16977 Warn when the stack frame of a function exceeds @var{n} bytes.
16983 The options @option{-metrax4} and @option{-metrax100} are synonyms for
16984 @option{-march=v3} and @option{-march=v8} respectively.
16986 @item -mmul-bug-workaround
16987 @itemx -mno-mul-bug-workaround
16988 @opindex mmul-bug-workaround
16989 @opindex mno-mul-bug-workaround
16990 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
16991 models where it applies. This option is active by default.
16995 Enable CRIS-specific verbose debug-related information in the assembly
16996 code. This option also has the effect of turning off the @samp{#NO_APP}
16997 formatted-code indicator to the assembler at the beginning of the
17002 Do not use condition-code results from previous instruction; always emit
17003 compare and test instructions before use of condition codes.
17005 @item -mno-side-effects
17006 @opindex mno-side-effects
17007 Do not emit instructions with side effects in addressing modes other than
17010 @item -mstack-align
17011 @itemx -mno-stack-align
17012 @itemx -mdata-align
17013 @itemx -mno-data-align
17014 @itemx -mconst-align
17015 @itemx -mno-const-align
17016 @opindex mstack-align
17017 @opindex mno-stack-align
17018 @opindex mdata-align
17019 @opindex mno-data-align
17020 @opindex mconst-align
17021 @opindex mno-const-align
17022 These options (@samp{no-} options) arrange (eliminate arrangements) for the
17023 stack frame, individual data and constants to be aligned for the maximum
17024 single data access size for the chosen CPU model. The default is to
17025 arrange for 32-bit alignment. ABI details such as structure layout are
17026 not affected by these options.
17034 Similar to the stack- data- and const-align options above, these options
17035 arrange for stack frame, writable data and constants to all be 32-bit,
17036 16-bit or 8-bit aligned. The default is 32-bit alignment.
17038 @item -mno-prologue-epilogue
17039 @itemx -mprologue-epilogue
17040 @opindex mno-prologue-epilogue
17041 @opindex mprologue-epilogue
17042 With @option{-mno-prologue-epilogue}, the normal function prologue and
17043 epilogue which set up the stack frame are omitted and no return
17044 instructions or return sequences are generated in the code. Use this
17045 option only together with visual inspection of the compiled code: no
17046 warnings or errors are generated when call-saved registers must be saved,
17047 or storage for local variables needs to be allocated.
17051 @opindex mno-gotplt
17053 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
17054 instruction sequences that load addresses for functions from the PLT part
17055 of the GOT rather than (traditional on other architectures) calls to the
17056 PLT@. The default is @option{-mgotplt}.
17060 Legacy no-op option only recognized with the cris-axis-elf and
17061 cris-axis-linux-gnu targets.
17065 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
17069 This option, recognized for the cris-axis-elf, arranges
17070 to link with input-output functions from a simulator library. Code,
17071 initialized data and zero-initialized data are allocated consecutively.
17075 Like @option{-sim}, but pass linker options to locate initialized data at
17076 0x40000000 and zero-initialized data at 0x80000000.
17080 @subsection CR16 Options
17081 @cindex CR16 Options
17083 These options are defined specifically for the CR16 ports.
17089 Enable the use of multiply-accumulate instructions. Disabled by default.
17093 @opindex mcr16cplus
17095 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
17100 Links the library libsim.a which is in compatible with simulator. Applicable
17101 to ELF compiler only.
17105 Choose integer type as 32-bit wide.
17109 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
17111 @item -mdata-model=@var{model}
17112 @opindex mdata-model
17113 Choose a data model. The choices for @var{model} are @samp{near},
17114 @samp{far} or @samp{medium}. @samp{medium} is default.
17115 However, @samp{far} is not valid with @option{-mcr16c}, as the
17116 CR16C architecture does not support the far data model.
17119 @node Darwin Options
17120 @subsection Darwin Options
17121 @cindex Darwin options
17123 These options are defined for all architectures running the Darwin operating
17126 FSF GCC on Darwin does not create ``fat'' object files; it creates
17127 an object file for the single architecture that GCC was built to
17128 target. Apple's GCC on Darwin does create ``fat'' files if multiple
17129 @option{-arch} options are used; it does so by running the compiler or
17130 linker multiple times and joining the results together with
17133 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
17134 @samp{i686}) is determined by the flags that specify the ISA
17135 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
17136 @option{-force_cpusubtype_ALL} option can be used to override this.
17138 The Darwin tools vary in their behavior when presented with an ISA
17139 mismatch. The assembler, @file{as}, only permits instructions to
17140 be used that are valid for the subtype of the file it is generating,
17141 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
17142 The linker for shared libraries, @file{/usr/bin/libtool}, fails
17143 and prints an error if asked to create a shared library with a less
17144 restrictive subtype than its input files (for instance, trying to put
17145 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
17146 for executables, @command{ld}, quietly gives the executable the most
17147 restrictive subtype of any of its input files.
17152 Add the framework directory @var{dir} to the head of the list of
17153 directories to be searched for header files. These directories are
17154 interleaved with those specified by @option{-I} options and are
17155 scanned in a left-to-right order.
17157 A framework directory is a directory with frameworks in it. A
17158 framework is a directory with a @file{Headers} and/or
17159 @file{PrivateHeaders} directory contained directly in it that ends
17160 in @file{.framework}. The name of a framework is the name of this
17161 directory excluding the @file{.framework}. Headers associated with
17162 the framework are found in one of those two directories, with
17163 @file{Headers} being searched first. A subframework is a framework
17164 directory that is in a framework's @file{Frameworks} directory.
17165 Includes of subframework headers can only appear in a header of a
17166 framework that contains the subframework, or in a sibling subframework
17167 header. Two subframeworks are siblings if they occur in the same
17168 framework. A subframework should not have the same name as a
17169 framework; a warning is issued if this is violated. Currently a
17170 subframework cannot have subframeworks; in the future, the mechanism
17171 may be extended to support this. The standard frameworks can be found
17172 in @file{/System/Library/Frameworks} and
17173 @file{/Library/Frameworks}. An example include looks like
17174 @code{#include <Framework/header.h>}, where @file{Framework} denotes
17175 the name of the framework and @file{header.h} is found in the
17176 @file{PrivateHeaders} or @file{Headers} directory.
17178 @item -iframework@var{dir}
17179 @opindex iframework
17180 Like @option{-F} except the directory is a treated as a system
17181 directory. The main difference between this @option{-iframework} and
17182 @option{-F} is that with @option{-iframework} the compiler does not
17183 warn about constructs contained within header files found via
17184 @var{dir}. This option is valid only for the C family of languages.
17188 Emit debugging information for symbols that are used. For stabs
17189 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
17190 This is by default ON@.
17194 Emit debugging information for all symbols and types.
17196 @item -mmacosx-version-min=@var{version}
17197 The earliest version of MacOS X that this executable will run on
17198 is @var{version}. Typical values of @var{version} include @code{10.1},
17199 @code{10.2}, and @code{10.3.9}.
17201 If the compiler was built to use the system's headers by default,
17202 then the default for this option is the system version on which the
17203 compiler is running, otherwise the default is to make choices that
17204 are compatible with as many systems and code bases as possible.
17208 Enable kernel development mode. The @option{-mkernel} option sets
17209 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
17210 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
17211 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
17212 applicable. This mode also sets @option{-mno-altivec},
17213 @option{-msoft-float}, @option{-fno-builtin} and
17214 @option{-mlong-branch} for PowerPC targets.
17216 @item -mone-byte-bool
17217 @opindex mone-byte-bool
17218 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
17219 By default @code{sizeof(bool)} is @code{4} when compiling for
17220 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
17221 option has no effect on x86.
17223 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
17224 to generate code that is not binary compatible with code generated
17225 without that switch. Using this switch may require recompiling all
17226 other modules in a program, including system libraries. Use this
17227 switch to conform to a non-default data model.
17229 @item -mfix-and-continue
17230 @itemx -ffix-and-continue
17231 @itemx -findirect-data
17232 @opindex mfix-and-continue
17233 @opindex ffix-and-continue
17234 @opindex findirect-data
17235 Generate code suitable for fast turnaround development, such as to
17236 allow GDB to dynamically load @file{.o} files into already-running
17237 programs. @option{-findirect-data} and @option{-ffix-and-continue}
17238 are provided for backwards compatibility.
17242 Loads all members of static archive libraries.
17243 See man ld(1) for more information.
17245 @item -arch_errors_fatal
17246 @opindex arch_errors_fatal
17247 Cause the errors having to do with files that have the wrong architecture
17250 @item -bind_at_load
17251 @opindex bind_at_load
17252 Causes the output file to be marked such that the dynamic linker will
17253 bind all undefined references when the file is loaded or launched.
17257 Produce a Mach-o bundle format file.
17258 See man ld(1) for more information.
17260 @item -bundle_loader @var{executable}
17261 @opindex bundle_loader
17262 This option specifies the @var{executable} that will load the build
17263 output file being linked. See man ld(1) for more information.
17266 @opindex dynamiclib
17267 When passed this option, GCC produces a dynamic library instead of
17268 an executable when linking, using the Darwin @file{libtool} command.
17270 @item -force_cpusubtype_ALL
17271 @opindex force_cpusubtype_ALL
17272 This causes GCC's output file to have the @samp{ALL} subtype, instead of
17273 one controlled by the @option{-mcpu} or @option{-march} option.
17275 @item -allowable_client @var{client_name}
17276 @itemx -client_name
17277 @itemx -compatibility_version
17278 @itemx -current_version
17280 @itemx -dependency-file
17282 @itemx -dylinker_install_name
17284 @itemx -exported_symbols_list
17287 @itemx -flat_namespace
17288 @itemx -force_flat_namespace
17289 @itemx -headerpad_max_install_names
17292 @itemx -install_name
17293 @itemx -keep_private_externs
17294 @itemx -multi_module
17295 @itemx -multiply_defined
17296 @itemx -multiply_defined_unused
17299 @itemx -no_dead_strip_inits_and_terms
17300 @itemx -nofixprebinding
17301 @itemx -nomultidefs
17303 @itemx -noseglinkedit
17304 @itemx -pagezero_size
17306 @itemx -prebind_all_twolevel_modules
17307 @itemx -private_bundle
17309 @itemx -read_only_relocs
17311 @itemx -sectobjectsymbols
17315 @itemx -sectobjectsymbols
17318 @itemx -segs_read_only_addr
17320 @itemx -segs_read_write_addr
17321 @itemx -seg_addr_table
17322 @itemx -seg_addr_table_filename
17323 @itemx -seglinkedit
17325 @itemx -segs_read_only_addr
17326 @itemx -segs_read_write_addr
17327 @itemx -single_module
17329 @itemx -sub_library
17331 @itemx -sub_umbrella
17332 @itemx -twolevel_namespace
17335 @itemx -unexported_symbols_list
17336 @itemx -weak_reference_mismatches
17337 @itemx -whatsloaded
17338 @opindex allowable_client
17339 @opindex client_name
17340 @opindex compatibility_version
17341 @opindex current_version
17342 @opindex dead_strip
17343 @opindex dependency-file
17344 @opindex dylib_file
17345 @opindex dylinker_install_name
17347 @opindex exported_symbols_list
17349 @opindex flat_namespace
17350 @opindex force_flat_namespace
17351 @opindex headerpad_max_install_names
17352 @opindex image_base
17354 @opindex install_name
17355 @opindex keep_private_externs
17356 @opindex multi_module
17357 @opindex multiply_defined
17358 @opindex multiply_defined_unused
17359 @opindex noall_load
17360 @opindex no_dead_strip_inits_and_terms
17361 @opindex nofixprebinding
17362 @opindex nomultidefs
17364 @opindex noseglinkedit
17365 @opindex pagezero_size
17367 @opindex prebind_all_twolevel_modules
17368 @opindex private_bundle
17369 @opindex read_only_relocs
17371 @opindex sectobjectsymbols
17374 @opindex sectcreate
17375 @opindex sectobjectsymbols
17378 @opindex segs_read_only_addr
17379 @opindex segs_read_write_addr
17380 @opindex seg_addr_table
17381 @opindex seg_addr_table_filename
17382 @opindex seglinkedit
17384 @opindex segs_read_only_addr
17385 @opindex segs_read_write_addr
17386 @opindex single_module
17388 @opindex sub_library
17389 @opindex sub_umbrella
17390 @opindex twolevel_namespace
17393 @opindex unexported_symbols_list
17394 @opindex weak_reference_mismatches
17395 @opindex whatsloaded
17396 These options are passed to the Darwin linker. The Darwin linker man page
17397 describes them in detail.
17400 @node DEC Alpha Options
17401 @subsection DEC Alpha Options
17403 These @samp{-m} options are defined for the DEC Alpha implementations:
17406 @item -mno-soft-float
17407 @itemx -msoft-float
17408 @opindex mno-soft-float
17409 @opindex msoft-float
17410 Use (do not use) the hardware floating-point instructions for
17411 floating-point operations. When @option{-msoft-float} is specified,
17412 functions in @file{libgcc.a} are used to perform floating-point
17413 operations. Unless they are replaced by routines that emulate the
17414 floating-point operations, or compiled in such a way as to call such
17415 emulations routines, these routines issue floating-point
17416 operations. If you are compiling for an Alpha without floating-point
17417 operations, you must ensure that the library is built so as not to call
17420 Note that Alpha implementations without floating-point operations are
17421 required to have floating-point registers.
17424 @itemx -mno-fp-regs
17426 @opindex mno-fp-regs
17427 Generate code that uses (does not use) the floating-point register set.
17428 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
17429 register set is not used, floating-point operands are passed in integer
17430 registers as if they were integers and floating-point results are passed
17431 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
17432 so any function with a floating-point argument or return value called by code
17433 compiled with @option{-mno-fp-regs} must also be compiled with that
17436 A typical use of this option is building a kernel that does not use,
17437 and hence need not save and restore, any floating-point registers.
17441 The Alpha architecture implements floating-point hardware optimized for
17442 maximum performance. It is mostly compliant with the IEEE floating-point
17443 standard. However, for full compliance, software assistance is
17444 required. This option generates code fully IEEE-compliant code
17445 @emph{except} that the @var{inexact-flag} is not maintained (see below).
17446 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
17447 defined during compilation. The resulting code is less efficient but is
17448 able to correctly support denormalized numbers and exceptional IEEE
17449 values such as not-a-number and plus/minus infinity. Other Alpha
17450 compilers call this option @option{-ieee_with_no_inexact}.
17452 @item -mieee-with-inexact
17453 @opindex mieee-with-inexact
17454 This is like @option{-mieee} except the generated code also maintains
17455 the IEEE @var{inexact-flag}. Turning on this option causes the
17456 generated code to implement fully-compliant IEEE math. In addition to
17457 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
17458 macro. On some Alpha implementations the resulting code may execute
17459 significantly slower than the code generated by default. Since there is
17460 very little code that depends on the @var{inexact-flag}, you should
17461 normally not specify this option. Other Alpha compilers call this
17462 option @option{-ieee_with_inexact}.
17464 @item -mfp-trap-mode=@var{trap-mode}
17465 @opindex mfp-trap-mode
17466 This option controls what floating-point related traps are enabled.
17467 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
17468 The trap mode can be set to one of four values:
17472 This is the default (normal) setting. The only traps that are enabled
17473 are the ones that cannot be disabled in software (e.g., division by zero
17477 In addition to the traps enabled by @samp{n}, underflow traps are enabled
17481 Like @samp{u}, but the instructions are marked to be safe for software
17482 completion (see Alpha architecture manual for details).
17485 Like @samp{su}, but inexact traps are enabled as well.
17488 @item -mfp-rounding-mode=@var{rounding-mode}
17489 @opindex mfp-rounding-mode
17490 Selects the IEEE rounding mode. Other Alpha compilers call this option
17491 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
17496 Normal IEEE rounding mode. Floating-point numbers are rounded towards
17497 the nearest machine number or towards the even machine number in case
17501 Round towards minus infinity.
17504 Chopped rounding mode. Floating-point numbers are rounded towards zero.
17507 Dynamic rounding mode. A field in the floating-point control register
17508 (@var{fpcr}, see Alpha architecture reference manual) controls the
17509 rounding mode in effect. The C library initializes this register for
17510 rounding towards plus infinity. Thus, unless your program modifies the
17511 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
17514 @item -mtrap-precision=@var{trap-precision}
17515 @opindex mtrap-precision
17516 In the Alpha architecture, floating-point traps are imprecise. This
17517 means without software assistance it is impossible to recover from a
17518 floating trap and program execution normally needs to be terminated.
17519 GCC can generate code that can assist operating system trap handlers
17520 in determining the exact location that caused a floating-point trap.
17521 Depending on the requirements of an application, different levels of
17522 precisions can be selected:
17526 Program precision. This option is the default and means a trap handler
17527 can only identify which program caused a floating-point exception.
17530 Function precision. The trap handler can determine the function that
17531 caused a floating-point exception.
17534 Instruction precision. The trap handler can determine the exact
17535 instruction that caused a floating-point exception.
17538 Other Alpha compilers provide the equivalent options called
17539 @option{-scope_safe} and @option{-resumption_safe}.
17541 @item -mieee-conformant
17542 @opindex mieee-conformant
17543 This option marks the generated code as IEEE conformant. You must not
17544 use this option unless you also specify @option{-mtrap-precision=i} and either
17545 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
17546 is to emit the line @samp{.eflag 48} in the function prologue of the
17547 generated assembly file.
17549 @item -mbuild-constants
17550 @opindex mbuild-constants
17551 Normally GCC examines a 32- or 64-bit integer constant to
17552 see if it can construct it from smaller constants in two or three
17553 instructions. If it cannot, it outputs the constant as a literal and
17554 generates code to load it from the data segment at run time.
17556 Use this option to require GCC to construct @emph{all} integer constants
17557 using code, even if it takes more instructions (the maximum is six).
17559 You typically use this option to build a shared library dynamic
17560 loader. Itself a shared library, it must relocate itself in memory
17561 before it can find the variables and constants in its own data segment.
17579 Indicate whether GCC should generate code to use the optional BWX,
17580 CIX, FIX and MAX instruction sets. The default is to use the instruction
17581 sets supported by the CPU type specified via @option{-mcpu=} option or that
17582 of the CPU on which GCC was built if none is specified.
17585 @itemx -mfloat-ieee
17586 @opindex mfloat-vax
17587 @opindex mfloat-ieee
17588 Generate code that uses (does not use) VAX F and G floating-point
17589 arithmetic instead of IEEE single and double precision.
17591 @item -mexplicit-relocs
17592 @itemx -mno-explicit-relocs
17593 @opindex mexplicit-relocs
17594 @opindex mno-explicit-relocs
17595 Older Alpha assemblers provided no way to generate symbol relocations
17596 except via assembler macros. Use of these macros does not allow
17597 optimal instruction scheduling. GNU binutils as of version 2.12
17598 supports a new syntax that allows the compiler to explicitly mark
17599 which relocations should apply to which instructions. This option
17600 is mostly useful for debugging, as GCC detects the capabilities of
17601 the assembler when it is built and sets the default accordingly.
17604 @itemx -mlarge-data
17605 @opindex msmall-data
17606 @opindex mlarge-data
17607 When @option{-mexplicit-relocs} is in effect, static data is
17608 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
17609 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
17610 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
17611 16-bit relocations off of the @code{$gp} register. This limits the
17612 size of the small data area to 64KB, but allows the variables to be
17613 directly accessed via a single instruction.
17615 The default is @option{-mlarge-data}. With this option the data area
17616 is limited to just below 2GB@. Programs that require more than 2GB of
17617 data must use @code{malloc} or @code{mmap} to allocate the data in the
17618 heap instead of in the program's data segment.
17620 When generating code for shared libraries, @option{-fpic} implies
17621 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
17624 @itemx -mlarge-text
17625 @opindex msmall-text
17626 @opindex mlarge-text
17627 When @option{-msmall-text} is used, the compiler assumes that the
17628 code of the entire program (or shared library) fits in 4MB, and is
17629 thus reachable with a branch instruction. When @option{-msmall-data}
17630 is used, the compiler can assume that all local symbols share the
17631 same @code{$gp} value, and thus reduce the number of instructions
17632 required for a function call from 4 to 1.
17634 The default is @option{-mlarge-text}.
17636 @item -mcpu=@var{cpu_type}
17638 Set the instruction set and instruction scheduling parameters for
17639 machine type @var{cpu_type}. You can specify either the @samp{EV}
17640 style name or the corresponding chip number. GCC supports scheduling
17641 parameters for the EV4, EV5 and EV6 family of processors and
17642 chooses the default values for the instruction set from the processor
17643 you specify. If you do not specify a processor type, GCC defaults
17644 to the processor on which the compiler was built.
17646 Supported values for @var{cpu_type} are
17652 Schedules as an EV4 and has no instruction set extensions.
17656 Schedules as an EV5 and has no instruction set extensions.
17660 Schedules as an EV5 and supports the BWX extension.
17665 Schedules as an EV5 and supports the BWX and MAX extensions.
17669 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
17673 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
17676 Native toolchains also support the value @samp{native},
17677 which selects the best architecture option for the host processor.
17678 @option{-mcpu=native} has no effect if GCC does not recognize
17681 @item -mtune=@var{cpu_type}
17683 Set only the instruction scheduling parameters for machine type
17684 @var{cpu_type}. The instruction set is not changed.
17686 Native toolchains also support the value @samp{native},
17687 which selects the best architecture option for the host processor.
17688 @option{-mtune=native} has no effect if GCC does not recognize
17691 @item -mmemory-latency=@var{time}
17692 @opindex mmemory-latency
17693 Sets the latency the scheduler should assume for typical memory
17694 references as seen by the application. This number is highly
17695 dependent on the memory access patterns used by the application
17696 and the size of the external cache on the machine.
17698 Valid options for @var{time} are
17702 A decimal number representing clock cycles.
17708 The compiler contains estimates of the number of clock cycles for
17709 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
17710 (also called Dcache, Scache, and Bcache), as well as to main memory.
17711 Note that L3 is only valid for EV5.
17717 @subsection FR30 Options
17718 @cindex FR30 Options
17720 These options are defined specifically for the FR30 port.
17724 @item -msmall-model
17725 @opindex msmall-model
17726 Use the small address space model. This can produce smaller code, but
17727 it does assume that all symbolic values and addresses fit into a
17732 Assume that runtime support has been provided and so there is no need
17733 to include the simulator library (@file{libsim.a}) on the linker
17739 @subsection FT32 Options
17740 @cindex FT32 Options
17742 These options are defined specifically for the FT32 port.
17748 Specifies that the program will be run on the simulator. This causes
17749 an alternate runtime startup and library to be linked.
17750 You must not use this option when generating programs that will run on
17751 real hardware; you must provide your own runtime library for whatever
17752 I/O functions are needed.
17756 Enable Local Register Allocation. This is still experimental for FT32,
17757 so by default the compiler uses standard reload.
17761 Do not use div and mod instructions.
17766 @subsection FRV Options
17767 @cindex FRV Options
17773 Only use the first 32 general-purpose registers.
17778 Use all 64 general-purpose registers.
17783 Use only the first 32 floating-point registers.
17788 Use all 64 floating-point registers.
17791 @opindex mhard-float
17793 Use hardware instructions for floating-point operations.
17796 @opindex msoft-float
17798 Use library routines for floating-point operations.
17803 Dynamically allocate condition code registers.
17808 Do not try to dynamically allocate condition code registers, only
17809 use @code{icc0} and @code{fcc0}.
17814 Change ABI to use double word insns.
17819 Do not use double word instructions.
17824 Use floating-point double instructions.
17827 @opindex mno-double
17829 Do not use floating-point double instructions.
17834 Use media instructions.
17839 Do not use media instructions.
17844 Use multiply and add/subtract instructions.
17847 @opindex mno-muladd
17849 Do not use multiply and add/subtract instructions.
17854 Select the FDPIC ABI, which uses function descriptors to represent
17855 pointers to functions. Without any PIC/PIE-related options, it
17856 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
17857 assumes GOT entries and small data are within a 12-bit range from the
17858 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
17859 are computed with 32 bits.
17860 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17863 @opindex minline-plt
17865 Enable inlining of PLT entries in function calls to functions that are
17866 not known to bind locally. It has no effect without @option{-mfdpic}.
17867 It's enabled by default if optimizing for speed and compiling for
17868 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
17869 optimization option such as @option{-O3} or above is present in the
17875 Assume a large TLS segment when generating thread-local code.
17880 Do not assume a large TLS segment when generating thread-local code.
17885 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
17886 that is known to be in read-only sections. It's enabled by default,
17887 except for @option{-fpic} or @option{-fpie}: even though it may help
17888 make the global offset table smaller, it trades 1 instruction for 4.
17889 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
17890 one of which may be shared by multiple symbols, and it avoids the need
17891 for a GOT entry for the referenced symbol, so it's more likely to be a
17892 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
17894 @item -multilib-library-pic
17895 @opindex multilib-library-pic
17897 Link with the (library, not FD) pic libraries. It's implied by
17898 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
17899 @option{-fpic} without @option{-mfdpic}. You should never have to use
17903 @opindex mlinked-fp
17905 Follow the EABI requirement of always creating a frame pointer whenever
17906 a stack frame is allocated. This option is enabled by default and can
17907 be disabled with @option{-mno-linked-fp}.
17910 @opindex mlong-calls
17912 Use indirect addressing to call functions outside the current
17913 compilation unit. This allows the functions to be placed anywhere
17914 within the 32-bit address space.
17916 @item -malign-labels
17917 @opindex malign-labels
17919 Try to align labels to an 8-byte boundary by inserting NOPs into the
17920 previous packet. This option only has an effect when VLIW packing
17921 is enabled. It doesn't create new packets; it merely adds NOPs to
17924 @item -mlibrary-pic
17925 @opindex mlibrary-pic
17927 Generate position-independent EABI code.
17932 Use only the first four media accumulator registers.
17937 Use all eight media accumulator registers.
17942 Pack VLIW instructions.
17947 Do not pack VLIW instructions.
17950 @opindex mno-eflags
17952 Do not mark ABI switches in e_flags.
17955 @opindex mcond-move
17957 Enable the use of conditional-move instructions (default).
17959 This switch is mainly for debugging the compiler and will likely be removed
17960 in a future version.
17962 @item -mno-cond-move
17963 @opindex mno-cond-move
17965 Disable the use of conditional-move instructions.
17967 This switch is mainly for debugging the compiler and will likely be removed
17968 in a future version.
17973 Enable the use of conditional set instructions (default).
17975 This switch is mainly for debugging the compiler and will likely be removed
17976 in a future version.
17981 Disable the use of conditional set instructions.
17983 This switch is mainly for debugging the compiler and will likely be removed
17984 in a future version.
17987 @opindex mcond-exec
17989 Enable the use of conditional execution (default).
17991 This switch is mainly for debugging the compiler and will likely be removed
17992 in a future version.
17994 @item -mno-cond-exec
17995 @opindex mno-cond-exec
17997 Disable the use of conditional execution.
17999 This switch is mainly for debugging the compiler and will likely be removed
18000 in a future version.
18002 @item -mvliw-branch
18003 @opindex mvliw-branch
18005 Run a pass to pack branches into VLIW instructions (default).
18007 This switch is mainly for debugging the compiler and will likely be removed
18008 in a future version.
18010 @item -mno-vliw-branch
18011 @opindex mno-vliw-branch
18013 Do not run a pass to pack branches into VLIW instructions.
18015 This switch is mainly for debugging the compiler and will likely be removed
18016 in a future version.
18018 @item -mmulti-cond-exec
18019 @opindex mmulti-cond-exec
18021 Enable optimization of @code{&&} and @code{||} in conditional execution
18024 This switch is mainly for debugging the compiler and will likely be removed
18025 in a future version.
18027 @item -mno-multi-cond-exec
18028 @opindex mno-multi-cond-exec
18030 Disable optimization of @code{&&} and @code{||} in conditional execution.
18032 This switch is mainly for debugging the compiler and will likely be removed
18033 in a future version.
18035 @item -mnested-cond-exec
18036 @opindex mnested-cond-exec
18038 Enable nested conditional execution optimizations (default).
18040 This switch is mainly for debugging the compiler and will likely be removed
18041 in a future version.
18043 @item -mno-nested-cond-exec
18044 @opindex mno-nested-cond-exec
18046 Disable nested conditional execution optimizations.
18048 This switch is mainly for debugging the compiler and will likely be removed
18049 in a future version.
18051 @item -moptimize-membar
18052 @opindex moptimize-membar
18054 This switch removes redundant @code{membar} instructions from the
18055 compiler-generated code. It is enabled by default.
18057 @item -mno-optimize-membar
18058 @opindex mno-optimize-membar
18060 This switch disables the automatic removal of redundant @code{membar}
18061 instructions from the generated code.
18063 @item -mtomcat-stats
18064 @opindex mtomcat-stats
18066 Cause gas to print out tomcat statistics.
18068 @item -mcpu=@var{cpu}
18071 Select the processor type for which to generate code. Possible values are
18072 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
18073 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
18077 @node GNU/Linux Options
18078 @subsection GNU/Linux Options
18080 These @samp{-m} options are defined for GNU/Linux targets:
18085 Use the GNU C library. This is the default except
18086 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
18087 @samp{*-*-linux-*android*} targets.
18091 Use uClibc C library. This is the default on
18092 @samp{*-*-linux-*uclibc*} targets.
18096 Use the musl C library. This is the default on
18097 @samp{*-*-linux-*musl*} targets.
18101 Use Bionic C library. This is the default on
18102 @samp{*-*-linux-*android*} targets.
18106 Compile code compatible with Android platform. This is the default on
18107 @samp{*-*-linux-*android*} targets.
18109 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
18110 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
18111 this option makes the GCC driver pass Android-specific options to the linker.
18112 Finally, this option causes the preprocessor macro @code{__ANDROID__}
18115 @item -tno-android-cc
18116 @opindex tno-android-cc
18117 Disable compilation effects of @option{-mandroid}, i.e., do not enable
18118 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
18119 @option{-fno-rtti} by default.
18121 @item -tno-android-ld
18122 @opindex tno-android-ld
18123 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
18124 linking options to the linker.
18128 @node H8/300 Options
18129 @subsection H8/300 Options
18131 These @samp{-m} options are defined for the H8/300 implementations:
18136 Shorten some address references at link time, when possible; uses the
18137 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
18138 ld, Using ld}, for a fuller description.
18142 Generate code for the H8/300H@.
18146 Generate code for the H8S@.
18150 Generate code for the H8S and H8/300H in the normal mode. This switch
18151 must be used either with @option{-mh} or @option{-ms}.
18155 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
18159 Extended registers are stored on stack before execution of function
18160 with monitor attribute. Default option is @option{-mexr}.
18161 This option is valid only for H8S targets.
18165 Extended registers are not stored on stack before execution of function
18166 with monitor attribute. Default option is @option{-mno-exr}.
18167 This option is valid only for H8S targets.
18171 Make @code{int} data 32 bits by default.
18174 @opindex malign-300
18175 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
18176 The default for the H8/300H and H8S is to align longs and floats on
18178 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
18179 This option has no effect on the H8/300.
18183 @subsection HPPA Options
18184 @cindex HPPA Options
18186 These @samp{-m} options are defined for the HPPA family of computers:
18189 @item -march=@var{architecture-type}
18191 Generate code for the specified architecture. The choices for
18192 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
18193 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
18194 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
18195 architecture option for your machine. Code compiled for lower numbered
18196 architectures runs on higher numbered architectures, but not the
18199 @item -mpa-risc-1-0
18200 @itemx -mpa-risc-1-1
18201 @itemx -mpa-risc-2-0
18202 @opindex mpa-risc-1-0
18203 @opindex mpa-risc-1-1
18204 @opindex mpa-risc-2-0
18205 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
18207 @item -mcaller-copies
18208 @opindex mcaller-copies
18209 The caller copies function arguments passed by hidden reference. This
18210 option should be used with care as it is not compatible with the default
18211 32-bit runtime. However, only aggregates larger than eight bytes are
18212 passed by hidden reference and the option provides better compatibility
18215 @item -mjump-in-delay
18216 @opindex mjump-in-delay
18217 This option is ignored and provided for compatibility purposes only.
18219 @item -mdisable-fpregs
18220 @opindex mdisable-fpregs
18221 Prevent floating-point registers from being used in any manner. This is
18222 necessary for compiling kernels that perform lazy context switching of
18223 floating-point registers. If you use this option and attempt to perform
18224 floating-point operations, the compiler aborts.
18226 @item -mdisable-indexing
18227 @opindex mdisable-indexing
18228 Prevent the compiler from using indexing address modes. This avoids some
18229 rather obscure problems when compiling MIG generated code under MACH@.
18231 @item -mno-space-regs
18232 @opindex mno-space-regs
18233 Generate code that assumes the target has no space registers. This allows
18234 GCC to generate faster indirect calls and use unscaled index address modes.
18236 Such code is suitable for level 0 PA systems and kernels.
18238 @item -mfast-indirect-calls
18239 @opindex mfast-indirect-calls
18240 Generate code that assumes calls never cross space boundaries. This
18241 allows GCC to emit code that performs faster indirect calls.
18243 This option does not work in the presence of shared libraries or nested
18246 @item -mfixed-range=@var{register-range}
18247 @opindex mfixed-range
18248 Generate code treating the given register range as fixed registers.
18249 A fixed register is one that the register allocator cannot use. This is
18250 useful when compiling kernel code. A register range is specified as
18251 two registers separated by a dash. Multiple register ranges can be
18252 specified separated by a comma.
18254 @item -mlong-load-store
18255 @opindex mlong-load-store
18256 Generate 3-instruction load and store sequences as sometimes required by
18257 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
18260 @item -mportable-runtime
18261 @opindex mportable-runtime
18262 Use the portable calling conventions proposed by HP for ELF systems.
18266 Enable the use of assembler directives only GAS understands.
18268 @item -mschedule=@var{cpu-type}
18270 Schedule code according to the constraints for the machine type
18271 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
18272 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
18273 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
18274 proper scheduling option for your machine. The default scheduling is
18278 @opindex mlinker-opt
18279 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
18280 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
18281 linkers in which they give bogus error messages when linking some programs.
18284 @opindex msoft-float
18285 Generate output containing library calls for floating point.
18286 @strong{Warning:} the requisite libraries are not available for all HPPA
18287 targets. Normally the facilities of the machine's usual C compiler are
18288 used, but this cannot be done directly in cross-compilation. You must make
18289 your own arrangements to provide suitable library functions for
18292 @option{-msoft-float} changes the calling convention in the output file;
18293 therefore, it is only useful if you compile @emph{all} of a program with
18294 this option. In particular, you need to compile @file{libgcc.a}, the
18295 library that comes with GCC, with @option{-msoft-float} in order for
18300 Generate the predefine, @code{_SIO}, for server IO@. The default is
18301 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
18302 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
18303 options are available under HP-UX and HI-UX@.
18307 Use options specific to GNU @command{ld}.
18308 This passes @option{-shared} to @command{ld} when
18309 building a shared library. It is the default when GCC is configured,
18310 explicitly or implicitly, with the GNU linker. This option does not
18311 affect which @command{ld} is called; it only changes what parameters
18312 are passed to that @command{ld}.
18313 The @command{ld} that is called is determined by the
18314 @option{--with-ld} configure option, GCC's program search path, and
18315 finally by the user's @env{PATH}. The linker used by GCC can be printed
18316 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
18317 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18321 Use options specific to HP @command{ld}.
18322 This passes @option{-b} to @command{ld} when building
18323 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
18324 links. It is the default when GCC is configured, explicitly or
18325 implicitly, with the HP linker. This option does not affect
18326 which @command{ld} is called; it only changes what parameters are passed to that
18328 The @command{ld} that is called is determined by the @option{--with-ld}
18329 configure option, GCC's program search path, and finally by the user's
18330 @env{PATH}. The linker used by GCC can be printed using @samp{which
18331 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
18332 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18335 @opindex mno-long-calls
18336 Generate code that uses long call sequences. This ensures that a call
18337 is always able to reach linker generated stubs. The default is to generate
18338 long calls only when the distance from the call site to the beginning
18339 of the function or translation unit, as the case may be, exceeds a
18340 predefined limit set by the branch type being used. The limits for
18341 normal calls are 7,600,000 and 240,000 bytes, respectively for the
18342 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
18345 Distances are measured from the beginning of functions when using the
18346 @option{-ffunction-sections} option, or when using the @option{-mgas}
18347 and @option{-mno-portable-runtime} options together under HP-UX with
18350 It is normally not desirable to use this option as it degrades
18351 performance. However, it may be useful in large applications,
18352 particularly when partial linking is used to build the application.
18354 The types of long calls used depends on the capabilities of the
18355 assembler and linker, and the type of code being generated. The
18356 impact on systems that support long absolute calls, and long pic
18357 symbol-difference or pc-relative calls should be relatively small.
18358 However, an indirect call is used on 32-bit ELF systems in pic code
18359 and it is quite long.
18361 @item -munix=@var{unix-std}
18363 Generate compiler predefines and select a startfile for the specified
18364 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
18365 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
18366 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
18367 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
18368 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
18371 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
18372 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
18373 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
18374 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
18375 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
18376 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
18378 It is @emph{important} to note that this option changes the interfaces
18379 for various library routines. It also affects the operational behavior
18380 of the C library. Thus, @emph{extreme} care is needed in using this
18383 Library code that is intended to operate with more than one UNIX
18384 standard must test, set and restore the variable @code{__xpg4_extended_mask}
18385 as appropriate. Most GNU software doesn't provide this capability.
18389 Suppress the generation of link options to search libdld.sl when the
18390 @option{-static} option is specified on HP-UX 10 and later.
18394 The HP-UX implementation of setlocale in libc has a dependency on
18395 libdld.sl. There isn't an archive version of libdld.sl. Thus,
18396 when the @option{-static} option is specified, special link options
18397 are needed to resolve this dependency.
18399 On HP-UX 10 and later, the GCC driver adds the necessary options to
18400 link with libdld.sl when the @option{-static} option is specified.
18401 This causes the resulting binary to be dynamic. On the 64-bit port,
18402 the linkers generate dynamic binaries by default in any case. The
18403 @option{-nolibdld} option can be used to prevent the GCC driver from
18404 adding these link options.
18408 Add support for multithreading with the @dfn{dce thread} library
18409 under HP-UX@. This option sets flags for both the preprocessor and
18413 @node IA-64 Options
18414 @subsection IA-64 Options
18415 @cindex IA-64 Options
18417 These are the @samp{-m} options defined for the Intel IA-64 architecture.
18421 @opindex mbig-endian
18422 Generate code for a big-endian target. This is the default for HP-UX@.
18424 @item -mlittle-endian
18425 @opindex mlittle-endian
18426 Generate code for a little-endian target. This is the default for AIX5
18432 @opindex mno-gnu-as
18433 Generate (or don't) code for the GNU assembler. This is the default.
18434 @c Also, this is the default if the configure option @option{--with-gnu-as}
18440 @opindex mno-gnu-ld
18441 Generate (or don't) code for the GNU linker. This is the default.
18442 @c Also, this is the default if the configure option @option{--with-gnu-ld}
18447 Generate code that does not use a global pointer register. The result
18448 is not position independent code, and violates the IA-64 ABI@.
18450 @item -mvolatile-asm-stop
18451 @itemx -mno-volatile-asm-stop
18452 @opindex mvolatile-asm-stop
18453 @opindex mno-volatile-asm-stop
18454 Generate (or don't) a stop bit immediately before and after volatile asm
18457 @item -mregister-names
18458 @itemx -mno-register-names
18459 @opindex mregister-names
18460 @opindex mno-register-names
18461 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
18462 the stacked registers. This may make assembler output more readable.
18468 Disable (or enable) optimizations that use the small data section. This may
18469 be useful for working around optimizer bugs.
18471 @item -mconstant-gp
18472 @opindex mconstant-gp
18473 Generate code that uses a single constant global pointer value. This is
18474 useful when compiling kernel code.
18478 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
18479 This is useful when compiling firmware code.
18481 @item -minline-float-divide-min-latency
18482 @opindex minline-float-divide-min-latency
18483 Generate code for inline divides of floating-point values
18484 using the minimum latency algorithm.
18486 @item -minline-float-divide-max-throughput
18487 @opindex minline-float-divide-max-throughput
18488 Generate code for inline divides of floating-point values
18489 using the maximum throughput algorithm.
18491 @item -mno-inline-float-divide
18492 @opindex mno-inline-float-divide
18493 Do not generate inline code for divides of floating-point values.
18495 @item -minline-int-divide-min-latency
18496 @opindex minline-int-divide-min-latency
18497 Generate code for inline divides of integer values
18498 using the minimum latency algorithm.
18500 @item -minline-int-divide-max-throughput
18501 @opindex minline-int-divide-max-throughput
18502 Generate code for inline divides of integer values
18503 using the maximum throughput algorithm.
18505 @item -mno-inline-int-divide
18506 @opindex mno-inline-int-divide
18507 Do not generate inline code for divides of integer values.
18509 @item -minline-sqrt-min-latency
18510 @opindex minline-sqrt-min-latency
18511 Generate code for inline square roots
18512 using the minimum latency algorithm.
18514 @item -minline-sqrt-max-throughput
18515 @opindex minline-sqrt-max-throughput
18516 Generate code for inline square roots
18517 using the maximum throughput algorithm.
18519 @item -mno-inline-sqrt
18520 @opindex mno-inline-sqrt
18521 Do not generate inline code for @code{sqrt}.
18524 @itemx -mno-fused-madd
18525 @opindex mfused-madd
18526 @opindex mno-fused-madd
18527 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
18528 instructions. The default is to use these instructions.
18530 @item -mno-dwarf2-asm
18531 @itemx -mdwarf2-asm
18532 @opindex mno-dwarf2-asm
18533 @opindex mdwarf2-asm
18534 Don't (or do) generate assembler code for the DWARF line number debugging
18535 info. This may be useful when not using the GNU assembler.
18537 @item -mearly-stop-bits
18538 @itemx -mno-early-stop-bits
18539 @opindex mearly-stop-bits
18540 @opindex mno-early-stop-bits
18541 Allow stop bits to be placed earlier than immediately preceding the
18542 instruction that triggered the stop bit. This can improve instruction
18543 scheduling, but does not always do so.
18545 @item -mfixed-range=@var{register-range}
18546 @opindex mfixed-range
18547 Generate code treating the given register range as fixed registers.
18548 A fixed register is one that the register allocator cannot use. This is
18549 useful when compiling kernel code. A register range is specified as
18550 two registers separated by a dash. Multiple register ranges can be
18551 specified separated by a comma.
18553 @item -mtls-size=@var{tls-size}
18555 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
18558 @item -mtune=@var{cpu-type}
18560 Tune the instruction scheduling for a particular CPU, Valid values are
18561 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
18562 and @samp{mckinley}.
18568 Generate code for a 32-bit or 64-bit environment.
18569 The 32-bit environment sets int, long and pointer to 32 bits.
18570 The 64-bit environment sets int to 32 bits and long and pointer
18571 to 64 bits. These are HP-UX specific flags.
18573 @item -mno-sched-br-data-spec
18574 @itemx -msched-br-data-spec
18575 @opindex mno-sched-br-data-spec
18576 @opindex msched-br-data-spec
18577 (Dis/En)able data speculative scheduling before reload.
18578 This results in generation of @code{ld.a} instructions and
18579 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18580 The default setting is disabled.
18582 @item -msched-ar-data-spec
18583 @itemx -mno-sched-ar-data-spec
18584 @opindex msched-ar-data-spec
18585 @opindex mno-sched-ar-data-spec
18586 (En/Dis)able data speculative scheduling after reload.
18587 This results in generation of @code{ld.a} instructions and
18588 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18589 The default setting is enabled.
18591 @item -mno-sched-control-spec
18592 @itemx -msched-control-spec
18593 @opindex mno-sched-control-spec
18594 @opindex msched-control-spec
18595 (Dis/En)able control speculative scheduling. This feature is
18596 available only during region scheduling (i.e.@: before reload).
18597 This results in generation of the @code{ld.s} instructions and
18598 the corresponding check instructions @code{chk.s}.
18599 The default setting is disabled.
18601 @item -msched-br-in-data-spec
18602 @itemx -mno-sched-br-in-data-spec
18603 @opindex msched-br-in-data-spec
18604 @opindex mno-sched-br-in-data-spec
18605 (En/Dis)able speculative scheduling of the instructions that
18606 are dependent on the data speculative loads before reload.
18607 This is effective only with @option{-msched-br-data-spec} enabled.
18608 The default setting is enabled.
18610 @item -msched-ar-in-data-spec
18611 @itemx -mno-sched-ar-in-data-spec
18612 @opindex msched-ar-in-data-spec
18613 @opindex mno-sched-ar-in-data-spec
18614 (En/Dis)able speculative scheduling of the instructions that
18615 are dependent on the data speculative loads after reload.
18616 This is effective only with @option{-msched-ar-data-spec} enabled.
18617 The default setting is enabled.
18619 @item -msched-in-control-spec
18620 @itemx -mno-sched-in-control-spec
18621 @opindex msched-in-control-spec
18622 @opindex mno-sched-in-control-spec
18623 (En/Dis)able speculative scheduling of the instructions that
18624 are dependent on the control speculative loads.
18625 This is effective only with @option{-msched-control-spec} enabled.
18626 The default setting is enabled.
18628 @item -mno-sched-prefer-non-data-spec-insns
18629 @itemx -msched-prefer-non-data-spec-insns
18630 @opindex mno-sched-prefer-non-data-spec-insns
18631 @opindex msched-prefer-non-data-spec-insns
18632 If enabled, data-speculative instructions are chosen for schedule
18633 only if there are no other choices at the moment. This makes
18634 the use of the data speculation much more conservative.
18635 The default setting is disabled.
18637 @item -mno-sched-prefer-non-control-spec-insns
18638 @itemx -msched-prefer-non-control-spec-insns
18639 @opindex mno-sched-prefer-non-control-spec-insns
18640 @opindex msched-prefer-non-control-spec-insns
18641 If enabled, control-speculative instructions are chosen for schedule
18642 only if there are no other choices at the moment. This makes
18643 the use of the control speculation much more conservative.
18644 The default setting is disabled.
18646 @item -mno-sched-count-spec-in-critical-path
18647 @itemx -msched-count-spec-in-critical-path
18648 @opindex mno-sched-count-spec-in-critical-path
18649 @opindex msched-count-spec-in-critical-path
18650 If enabled, speculative dependencies are considered during
18651 computation of the instructions priorities. This makes the use of the
18652 speculation a bit more conservative.
18653 The default setting is disabled.
18655 @item -msched-spec-ldc
18656 @opindex msched-spec-ldc
18657 Use a simple data speculation check. This option is on by default.
18659 @item -msched-control-spec-ldc
18660 @opindex msched-spec-ldc
18661 Use a simple check for control speculation. This option is on by default.
18663 @item -msched-stop-bits-after-every-cycle
18664 @opindex msched-stop-bits-after-every-cycle
18665 Place a stop bit after every cycle when scheduling. This option is on
18668 @item -msched-fp-mem-deps-zero-cost
18669 @opindex msched-fp-mem-deps-zero-cost
18670 Assume that floating-point stores and loads are not likely to cause a conflict
18671 when placed into the same instruction group. This option is disabled by
18674 @item -msel-sched-dont-check-control-spec
18675 @opindex msel-sched-dont-check-control-spec
18676 Generate checks for control speculation in selective scheduling.
18677 This flag is disabled by default.
18679 @item -msched-max-memory-insns=@var{max-insns}
18680 @opindex msched-max-memory-insns
18681 Limit on the number of memory insns per instruction group, giving lower
18682 priority to subsequent memory insns attempting to schedule in the same
18683 instruction group. Frequently useful to prevent cache bank conflicts.
18684 The default value is 1.
18686 @item -msched-max-memory-insns-hard-limit
18687 @opindex msched-max-memory-insns-hard-limit
18688 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
18689 disallowing more than that number in an instruction group.
18690 Otherwise, the limit is ``soft'', meaning that non-memory operations
18691 are preferred when the limit is reached, but memory operations may still
18697 @subsection LM32 Options
18698 @cindex LM32 options
18700 These @option{-m} options are defined for the LatticeMico32 architecture:
18703 @item -mbarrel-shift-enabled
18704 @opindex mbarrel-shift-enabled
18705 Enable barrel-shift instructions.
18707 @item -mdivide-enabled
18708 @opindex mdivide-enabled
18709 Enable divide and modulus instructions.
18711 @item -mmultiply-enabled
18712 @opindex multiply-enabled
18713 Enable multiply instructions.
18715 @item -msign-extend-enabled
18716 @opindex msign-extend-enabled
18717 Enable sign extend instructions.
18719 @item -muser-enabled
18720 @opindex muser-enabled
18721 Enable user-defined instructions.
18726 @subsection M32C Options
18727 @cindex M32C options
18730 @item -mcpu=@var{name}
18732 Select the CPU for which code is generated. @var{name} may be one of
18733 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
18734 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
18735 the M32C/80 series.
18739 Specifies that the program will be run on the simulator. This causes
18740 an alternate runtime library to be linked in which supports, for
18741 example, file I/O@. You must not use this option when generating
18742 programs that will run on real hardware; you must provide your own
18743 runtime library for whatever I/O functions are needed.
18745 @item -memregs=@var{number}
18747 Specifies the number of memory-based pseudo-registers GCC uses
18748 during code generation. These pseudo-registers are used like real
18749 registers, so there is a tradeoff between GCC's ability to fit the
18750 code into available registers, and the performance penalty of using
18751 memory instead of registers. Note that all modules in a program must
18752 be compiled with the same value for this option. Because of that, you
18753 must not use this option with GCC's default runtime libraries.
18757 @node M32R/D Options
18758 @subsection M32R/D Options
18759 @cindex M32R/D options
18761 These @option{-m} options are defined for Renesas M32R/D architectures:
18766 Generate code for the M32R/2@.
18770 Generate code for the M32R/X@.
18774 Generate code for the M32R@. This is the default.
18776 @item -mmodel=small
18777 @opindex mmodel=small
18778 Assume all objects live in the lower 16MB of memory (so that their addresses
18779 can be loaded with the @code{ld24} instruction), and assume all subroutines
18780 are reachable with the @code{bl} instruction.
18781 This is the default.
18783 The addressability of a particular object can be set with the
18784 @code{model} attribute.
18786 @item -mmodel=medium
18787 @opindex mmodel=medium
18788 Assume objects may be anywhere in the 32-bit address space (the compiler
18789 generates @code{seth/add3} instructions to load their addresses), and
18790 assume all subroutines are reachable with the @code{bl} instruction.
18792 @item -mmodel=large
18793 @opindex mmodel=large
18794 Assume objects may be anywhere in the 32-bit address space (the compiler
18795 generates @code{seth/add3} instructions to load their addresses), and
18796 assume subroutines may not be reachable with the @code{bl} instruction
18797 (the compiler generates the much slower @code{seth/add3/jl}
18798 instruction sequence).
18801 @opindex msdata=none
18802 Disable use of the small data area. Variables are put into
18803 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
18804 @code{section} attribute has been specified).
18805 This is the default.
18807 The small data area consists of sections @code{.sdata} and @code{.sbss}.
18808 Objects may be explicitly put in the small data area with the
18809 @code{section} attribute using one of these sections.
18811 @item -msdata=sdata
18812 @opindex msdata=sdata
18813 Put small global and static data in the small data area, but do not
18814 generate special code to reference them.
18817 @opindex msdata=use
18818 Put small global and static data in the small data area, and generate
18819 special instructions to reference them.
18823 @cindex smaller data references
18824 Put global and static objects less than or equal to @var{num} bytes
18825 into the small data or BSS sections instead of the normal data or BSS
18826 sections. The default value of @var{num} is 8.
18827 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
18828 for this option to have any effect.
18830 All modules should be compiled with the same @option{-G @var{num}} value.
18831 Compiling with different values of @var{num} may or may not work; if it
18832 doesn't the linker gives an error message---incorrect code is not
18837 Makes the M32R-specific code in the compiler display some statistics
18838 that might help in debugging programs.
18840 @item -malign-loops
18841 @opindex malign-loops
18842 Align all loops to a 32-byte boundary.
18844 @item -mno-align-loops
18845 @opindex mno-align-loops
18846 Do not enforce a 32-byte alignment for loops. This is the default.
18848 @item -missue-rate=@var{number}
18849 @opindex missue-rate=@var{number}
18850 Issue @var{number} instructions per cycle. @var{number} can only be 1
18853 @item -mbranch-cost=@var{number}
18854 @opindex mbranch-cost=@var{number}
18855 @var{number} can only be 1 or 2. If it is 1 then branches are
18856 preferred over conditional code, if it is 2, then the opposite applies.
18858 @item -mflush-trap=@var{number}
18859 @opindex mflush-trap=@var{number}
18860 Specifies the trap number to use to flush the cache. The default is
18861 12. Valid numbers are between 0 and 15 inclusive.
18863 @item -mno-flush-trap
18864 @opindex mno-flush-trap
18865 Specifies that the cache cannot be flushed by using a trap.
18867 @item -mflush-func=@var{name}
18868 @opindex mflush-func=@var{name}
18869 Specifies the name of the operating system function to call to flush
18870 the cache. The default is @samp{_flush_cache}, but a function call
18871 is only used if a trap is not available.
18873 @item -mno-flush-func
18874 @opindex mno-flush-func
18875 Indicates that there is no OS function for flushing the cache.
18879 @node M680x0 Options
18880 @subsection M680x0 Options
18881 @cindex M680x0 options
18883 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
18884 The default settings depend on which architecture was selected when
18885 the compiler was configured; the defaults for the most common choices
18889 @item -march=@var{arch}
18891 Generate code for a specific M680x0 or ColdFire instruction set
18892 architecture. Permissible values of @var{arch} for M680x0
18893 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
18894 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
18895 architectures are selected according to Freescale's ISA classification
18896 and the permissible values are: @samp{isaa}, @samp{isaaplus},
18897 @samp{isab} and @samp{isac}.
18899 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
18900 code for a ColdFire target. The @var{arch} in this macro is one of the
18901 @option{-march} arguments given above.
18903 When used together, @option{-march} and @option{-mtune} select code
18904 that runs on a family of similar processors but that is optimized
18905 for a particular microarchitecture.
18907 @item -mcpu=@var{cpu}
18909 Generate code for a specific M680x0 or ColdFire processor.
18910 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
18911 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
18912 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
18913 below, which also classifies the CPUs into families:
18915 @multitable @columnfractions 0.20 0.80
18916 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
18917 @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}
18918 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
18919 @item @samp{5206e} @tab @samp{5206e}
18920 @item @samp{5208} @tab @samp{5207} @samp{5208}
18921 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
18922 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
18923 @item @samp{5216} @tab @samp{5214} @samp{5216}
18924 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
18925 @item @samp{5225} @tab @samp{5224} @samp{5225}
18926 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
18927 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
18928 @item @samp{5249} @tab @samp{5249}
18929 @item @samp{5250} @tab @samp{5250}
18930 @item @samp{5271} @tab @samp{5270} @samp{5271}
18931 @item @samp{5272} @tab @samp{5272}
18932 @item @samp{5275} @tab @samp{5274} @samp{5275}
18933 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
18934 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
18935 @item @samp{5307} @tab @samp{5307}
18936 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
18937 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
18938 @item @samp{5407} @tab @samp{5407}
18939 @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}
18942 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
18943 @var{arch} is compatible with @var{cpu}. Other combinations of
18944 @option{-mcpu} and @option{-march} are rejected.
18946 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
18947 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
18948 where the value of @var{family} is given by the table above.
18950 @item -mtune=@var{tune}
18952 Tune the code for a particular microarchitecture within the
18953 constraints set by @option{-march} and @option{-mcpu}.
18954 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
18955 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
18956 and @samp{cpu32}. The ColdFire microarchitectures
18957 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
18959 You can also use @option{-mtune=68020-40} for code that needs
18960 to run relatively well on 68020, 68030 and 68040 targets.
18961 @option{-mtune=68020-60} is similar but includes 68060 targets
18962 as well. These two options select the same tuning decisions as
18963 @option{-m68020-40} and @option{-m68020-60} respectively.
18965 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
18966 when tuning for 680x0 architecture @var{arch}. It also defines
18967 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
18968 option is used. If GCC is tuning for a range of architectures,
18969 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
18970 it defines the macros for every architecture in the range.
18972 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
18973 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
18974 of the arguments given above.
18980 Generate output for a 68000. This is the default
18981 when the compiler is configured for 68000-based systems.
18982 It is equivalent to @option{-march=68000}.
18984 Use this option for microcontrollers with a 68000 or EC000 core,
18985 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
18989 Generate output for a 68010. This is the default
18990 when the compiler is configured for 68010-based systems.
18991 It is equivalent to @option{-march=68010}.
18997 Generate output for a 68020. This is the default
18998 when the compiler is configured for 68020-based systems.
18999 It is equivalent to @option{-march=68020}.
19003 Generate output for a 68030. This is the default when the compiler is
19004 configured for 68030-based systems. It is equivalent to
19005 @option{-march=68030}.
19009 Generate output for a 68040. This is the default when the compiler is
19010 configured for 68040-based systems. It is equivalent to
19011 @option{-march=68040}.
19013 This option inhibits the use of 68881/68882 instructions that have to be
19014 emulated by software on the 68040. Use this option if your 68040 does not
19015 have code to emulate those instructions.
19019 Generate output for a 68060. This is the default when the compiler is
19020 configured for 68060-based systems. It is equivalent to
19021 @option{-march=68060}.
19023 This option inhibits the use of 68020 and 68881/68882 instructions that
19024 have to be emulated by software on the 68060. Use this option if your 68060
19025 does not have code to emulate those instructions.
19029 Generate output for a CPU32. This is the default
19030 when the compiler is configured for CPU32-based systems.
19031 It is equivalent to @option{-march=cpu32}.
19033 Use this option for microcontrollers with a
19034 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
19035 68336, 68340, 68341, 68349 and 68360.
19039 Generate output for a 520X ColdFire CPU@. This is the default
19040 when the compiler is configured for 520X-based systems.
19041 It is equivalent to @option{-mcpu=5206}, and is now deprecated
19042 in favor of that option.
19044 Use this option for microcontroller with a 5200 core, including
19045 the MCF5202, MCF5203, MCF5204 and MCF5206.
19049 Generate output for a 5206e ColdFire CPU@. The option is now
19050 deprecated in favor of the equivalent @option{-mcpu=5206e}.
19054 Generate output for a member of the ColdFire 528X family.
19055 The option is now deprecated in favor of the equivalent
19056 @option{-mcpu=528x}.
19060 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
19061 in favor of the equivalent @option{-mcpu=5307}.
19065 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
19066 in favor of the equivalent @option{-mcpu=5407}.
19070 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
19071 This includes use of hardware floating-point instructions.
19072 The option is equivalent to @option{-mcpu=547x}, and is now
19073 deprecated in favor of that option.
19077 Generate output for a 68040, without using any of the new instructions.
19078 This results in code that can run relatively efficiently on either a
19079 68020/68881 or a 68030 or a 68040. The generated code does use the
19080 68881 instructions that are emulated on the 68040.
19082 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
19086 Generate output for a 68060, without using any of the new instructions.
19087 This results in code that can run relatively efficiently on either a
19088 68020/68881 or a 68030 or a 68040. The generated code does use the
19089 68881 instructions that are emulated on the 68060.
19091 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
19095 @opindex mhard-float
19097 Generate floating-point instructions. This is the default for 68020
19098 and above, and for ColdFire devices that have an FPU@. It defines the
19099 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
19100 on ColdFire targets.
19103 @opindex msoft-float
19104 Do not generate floating-point instructions; use library calls instead.
19105 This is the default for 68000, 68010, and 68832 targets. It is also
19106 the default for ColdFire devices that have no FPU.
19112 Generate (do not generate) ColdFire hardware divide and remainder
19113 instructions. If @option{-march} is used without @option{-mcpu},
19114 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
19115 architectures. Otherwise, the default is taken from the target CPU
19116 (either the default CPU, or the one specified by @option{-mcpu}). For
19117 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
19118 @option{-mcpu=5206e}.
19120 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
19124 Consider type @code{int} to be 16 bits wide, like @code{short int}.
19125 Additionally, parameters passed on the stack are also aligned to a
19126 16-bit boundary even on targets whose API mandates promotion to 32-bit.
19130 Do not consider type @code{int} to be 16 bits wide. This is the default.
19133 @itemx -mno-bitfield
19134 @opindex mnobitfield
19135 @opindex mno-bitfield
19136 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
19137 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
19141 Do use the bit-field instructions. The @option{-m68020} option implies
19142 @option{-mbitfield}. This is the default if you use a configuration
19143 designed for a 68020.
19147 Use a different function-calling convention, in which functions
19148 that take a fixed number of arguments return with the @code{rtd}
19149 instruction, which pops their arguments while returning. This
19150 saves one instruction in the caller since there is no need to pop
19151 the arguments there.
19153 This calling convention is incompatible with the one normally
19154 used on Unix, so you cannot use it if you need to call libraries
19155 compiled with the Unix compiler.
19157 Also, you must provide function prototypes for all functions that
19158 take variable numbers of arguments (including @code{printf});
19159 otherwise incorrect code is generated for calls to those
19162 In addition, seriously incorrect code results if you call a
19163 function with too many arguments. (Normally, extra arguments are
19164 harmlessly ignored.)
19166 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
19167 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
19171 Do not use the calling conventions selected by @option{-mrtd}.
19172 This is the default.
19175 @itemx -mno-align-int
19176 @opindex malign-int
19177 @opindex mno-align-int
19178 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
19179 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
19180 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
19181 Aligning variables on 32-bit boundaries produces code that runs somewhat
19182 faster on processors with 32-bit busses at the expense of more memory.
19184 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
19185 aligns structures containing the above types differently than
19186 most published application binary interface specifications for the m68k.
19190 Use the pc-relative addressing mode of the 68000 directly, instead of
19191 using a global offset table. At present, this option implies @option{-fpic},
19192 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
19193 not presently supported with @option{-mpcrel}, though this could be supported for
19194 68020 and higher processors.
19196 @item -mno-strict-align
19197 @itemx -mstrict-align
19198 @opindex mno-strict-align
19199 @opindex mstrict-align
19200 Do not (do) assume that unaligned memory references are handled by
19204 Generate code that allows the data segment to be located in a different
19205 area of memory from the text segment. This allows for execute-in-place in
19206 an environment without virtual memory management. This option implies
19209 @item -mno-sep-data
19210 Generate code that assumes that the data segment follows the text segment.
19211 This is the default.
19213 @item -mid-shared-library
19214 Generate code that supports shared libraries via the library ID method.
19215 This allows for execute-in-place and shared libraries in an environment
19216 without virtual memory management. This option implies @option{-fPIC}.
19218 @item -mno-id-shared-library
19219 Generate code that doesn't assume ID-based shared libraries are being used.
19220 This is the default.
19222 @item -mshared-library-id=n
19223 Specifies the identification number of the ID-based shared library being
19224 compiled. Specifying a value of 0 generates more compact code; specifying
19225 other values forces the allocation of that number to the current
19226 library, but is no more space- or time-efficient than omitting this option.
19232 When generating position-independent code for ColdFire, generate code
19233 that works if the GOT has more than 8192 entries. This code is
19234 larger and slower than code generated without this option. On M680x0
19235 processors, this option is not needed; @option{-fPIC} suffices.
19237 GCC normally uses a single instruction to load values from the GOT@.
19238 While this is relatively efficient, it only works if the GOT
19239 is smaller than about 64k. Anything larger causes the linker
19240 to report an error such as:
19242 @cindex relocation truncated to fit (ColdFire)
19244 relocation truncated to fit: R_68K_GOT16O foobar
19247 If this happens, you should recompile your code with @option{-mxgot}.
19248 It should then work with very large GOTs. However, code generated with
19249 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
19250 the value of a global symbol.
19252 Note that some linkers, including newer versions of the GNU linker,
19253 can create multiple GOTs and sort GOT entries. If you have such a linker,
19254 you should only need to use @option{-mxgot} when compiling a single
19255 object file that accesses more than 8192 GOT entries. Very few do.
19257 These options have no effect unless GCC is generating
19258 position-independent code.
19260 @item -mlong-jump-table-offsets
19261 @opindex mlong-jump-table-offsets
19262 Use 32-bit offsets in @code{switch} tables. The default is to use
19267 @node MCore Options
19268 @subsection MCore Options
19269 @cindex MCore options
19271 These are the @samp{-m} options defined for the Motorola M*Core
19277 @itemx -mno-hardlit
19279 @opindex mno-hardlit
19280 Inline constants into the code stream if it can be done in two
19281 instructions or less.
19287 Use the divide instruction. (Enabled by default).
19289 @item -mrelax-immediate
19290 @itemx -mno-relax-immediate
19291 @opindex mrelax-immediate
19292 @opindex mno-relax-immediate
19293 Allow arbitrary-sized immediates in bit operations.
19295 @item -mwide-bitfields
19296 @itemx -mno-wide-bitfields
19297 @opindex mwide-bitfields
19298 @opindex mno-wide-bitfields
19299 Always treat bit-fields as @code{int}-sized.
19301 @item -m4byte-functions
19302 @itemx -mno-4byte-functions
19303 @opindex m4byte-functions
19304 @opindex mno-4byte-functions
19305 Force all functions to be aligned to a 4-byte boundary.
19307 @item -mcallgraph-data
19308 @itemx -mno-callgraph-data
19309 @opindex mcallgraph-data
19310 @opindex mno-callgraph-data
19311 Emit callgraph information.
19314 @itemx -mno-slow-bytes
19315 @opindex mslow-bytes
19316 @opindex mno-slow-bytes
19317 Prefer word access when reading byte quantities.
19319 @item -mlittle-endian
19320 @itemx -mbig-endian
19321 @opindex mlittle-endian
19322 @opindex mbig-endian
19323 Generate code for a little-endian target.
19329 Generate code for the 210 processor.
19333 Assume that runtime support has been provided and so omit the
19334 simulator library (@file{libsim.a)} from the linker command line.
19336 @item -mstack-increment=@var{size}
19337 @opindex mstack-increment
19338 Set the maximum amount for a single stack increment operation. Large
19339 values can increase the speed of programs that contain functions
19340 that need a large amount of stack space, but they can also trigger a
19341 segmentation fault if the stack is extended too much. The default
19347 @subsection MeP Options
19348 @cindex MeP options
19354 Enables the @code{abs} instruction, which is the absolute difference
19355 between two registers.
19359 Enables all the optional instructions---average, multiply, divide, bit
19360 operations, leading zero, absolute difference, min/max, clip, and
19366 Enables the @code{ave} instruction, which computes the average of two
19369 @item -mbased=@var{n}
19371 Variables of size @var{n} bytes or smaller are placed in the
19372 @code{.based} section by default. Based variables use the @code{$tp}
19373 register as a base register, and there is a 128-byte limit to the
19374 @code{.based} section.
19378 Enables the bit operation instructions---bit test (@code{btstm}), set
19379 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
19380 test-and-set (@code{tas}).
19382 @item -mc=@var{name}
19384 Selects which section constant data is placed in. @var{name} may
19385 be @samp{tiny}, @samp{near}, or @samp{far}.
19389 Enables the @code{clip} instruction. Note that @option{-mclip} is not
19390 useful unless you also provide @option{-mminmax}.
19392 @item -mconfig=@var{name}
19394 Selects one of the built-in core configurations. Each MeP chip has
19395 one or more modules in it; each module has a core CPU and a variety of
19396 coprocessors, optional instructions, and peripherals. The
19397 @code{MeP-Integrator} tool, not part of GCC, provides these
19398 configurations through this option; using this option is the same as
19399 using all the corresponding command-line options. The default
19400 configuration is @samp{default}.
19404 Enables the coprocessor instructions. By default, this is a 32-bit
19405 coprocessor. Note that the coprocessor is normally enabled via the
19406 @option{-mconfig=} option.
19410 Enables the 32-bit coprocessor's instructions.
19414 Enables the 64-bit coprocessor's instructions.
19418 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
19422 Causes constant variables to be placed in the @code{.near} section.
19426 Enables the @code{div} and @code{divu} instructions.
19430 Generate big-endian code.
19434 Generate little-endian code.
19436 @item -mio-volatile
19437 @opindex mio-volatile
19438 Tells the compiler that any variable marked with the @code{io}
19439 attribute is to be considered volatile.
19443 Causes variables to be assigned to the @code{.far} section by default.
19447 Enables the @code{leadz} (leading zero) instruction.
19451 Causes variables to be assigned to the @code{.near} section by default.
19455 Enables the @code{min} and @code{max} instructions.
19459 Enables the multiplication and multiply-accumulate instructions.
19463 Disables all the optional instructions enabled by @option{-mall-opts}.
19467 Enables the @code{repeat} and @code{erepeat} instructions, used for
19468 low-overhead looping.
19472 Causes all variables to default to the @code{.tiny} section. Note
19473 that there is a 65536-byte limit to this section. Accesses to these
19474 variables use the @code{%gp} base register.
19478 Enables the saturation instructions. Note that the compiler does not
19479 currently generate these itself, but this option is included for
19480 compatibility with other tools, like @code{as}.
19484 Link the SDRAM-based runtime instead of the default ROM-based runtime.
19488 Link the simulator run-time libraries.
19492 Link the simulator runtime libraries, excluding built-in support
19493 for reset and exception vectors and tables.
19497 Causes all functions to default to the @code{.far} section. Without
19498 this option, functions default to the @code{.near} section.
19500 @item -mtiny=@var{n}
19502 Variables that are @var{n} bytes or smaller are allocated to the
19503 @code{.tiny} section. These variables use the @code{$gp} base
19504 register. The default for this option is 4, but note that there's a
19505 65536-byte limit to the @code{.tiny} section.
19509 @node MicroBlaze Options
19510 @subsection MicroBlaze Options
19511 @cindex MicroBlaze Options
19516 @opindex msoft-float
19517 Use software emulation for floating point (default).
19520 @opindex mhard-float
19521 Use hardware floating-point instructions.
19525 Do not optimize block moves, use @code{memcpy}.
19527 @item -mno-clearbss
19528 @opindex mno-clearbss
19529 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
19531 @item -mcpu=@var{cpu-type}
19533 Use features of, and schedule code for, the given CPU.
19534 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
19535 where @var{X} is a major version, @var{YY} is the minor version, and
19536 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
19537 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
19539 @item -mxl-soft-mul
19540 @opindex mxl-soft-mul
19541 Use software multiply emulation (default).
19543 @item -mxl-soft-div
19544 @opindex mxl-soft-div
19545 Use software emulation for divides (default).
19547 @item -mxl-barrel-shift
19548 @opindex mxl-barrel-shift
19549 Use the hardware barrel shifter.
19551 @item -mxl-pattern-compare
19552 @opindex mxl-pattern-compare
19553 Use pattern compare instructions.
19555 @item -msmall-divides
19556 @opindex msmall-divides
19557 Use table lookup optimization for small signed integer divisions.
19559 @item -mxl-stack-check
19560 @opindex mxl-stack-check
19561 This option is deprecated. Use @option{-fstack-check} instead.
19564 @opindex mxl-gp-opt
19565 Use GP-relative @code{.sdata}/@code{.sbss} sections.
19567 @item -mxl-multiply-high
19568 @opindex mxl-multiply-high
19569 Use multiply high instructions for high part of 32x32 multiply.
19571 @item -mxl-float-convert
19572 @opindex mxl-float-convert
19573 Use hardware floating-point conversion instructions.
19575 @item -mxl-float-sqrt
19576 @opindex mxl-float-sqrt
19577 Use hardware floating-point square root instruction.
19580 @opindex mbig-endian
19581 Generate code for a big-endian target.
19583 @item -mlittle-endian
19584 @opindex mlittle-endian
19585 Generate code for a little-endian target.
19588 @opindex mxl-reorder
19589 Use reorder instructions (swap and byte reversed load/store).
19591 @item -mxl-mode-@var{app-model}
19592 Select application model @var{app-model}. Valid models are
19595 normal executable (default), uses startup code @file{crt0.o}.
19598 for use with Xilinx Microprocessor Debugger (XMD) based
19599 software intrusive debug agent called xmdstub. This uses startup file
19600 @file{crt1.o} and sets the start address of the program to 0x800.
19603 for applications that are loaded using a bootloader.
19604 This model uses startup file @file{crt2.o} which does not contain a processor
19605 reset vector handler. This is suitable for transferring control on a
19606 processor reset to the bootloader rather than the application.
19609 for applications that do not require any of the
19610 MicroBlaze vectors. This option may be useful for applications running
19611 within a monitoring application. This model uses @file{crt3.o} as a startup file.
19614 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
19615 @option{-mxl-mode-@var{app-model}}.
19620 @subsection MIPS Options
19621 @cindex MIPS options
19627 Generate big-endian code.
19631 Generate little-endian code. This is the default for @samp{mips*el-*-*}
19634 @item -march=@var{arch}
19636 Generate code that runs on @var{arch}, which can be the name of a
19637 generic MIPS ISA, or the name of a particular processor.
19639 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
19640 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
19641 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
19642 @samp{mips64r5} and @samp{mips64r6}.
19643 The processor names are:
19644 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
19645 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
19646 @samp{5kc}, @samp{5kf},
19648 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
19649 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
19650 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
19651 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
19652 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
19655 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
19657 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
19658 @samp{m5100}, @samp{m5101},
19659 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
19662 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
19663 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
19664 @samp{rm7000}, @samp{rm9000},
19665 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
19668 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
19669 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
19670 @samp{xlr} and @samp{xlp}.
19671 The special value @samp{from-abi} selects the
19672 most compatible architecture for the selected ABI (that is,
19673 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
19675 The native Linux/GNU toolchain also supports the value @samp{native},
19676 which selects the best architecture option for the host processor.
19677 @option{-march=native} has no effect if GCC does not recognize
19680 In processor names, a final @samp{000} can be abbreviated as @samp{k}
19681 (for example, @option{-march=r2k}). Prefixes are optional, and
19682 @samp{vr} may be written @samp{r}.
19684 Names of the form @samp{@var{n}f2_1} refer to processors with
19685 FPUs clocked at half the rate of the core, names of the form
19686 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
19687 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
19688 processors with FPUs clocked a ratio of 3:2 with respect to the core.
19689 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
19690 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
19691 accepted as synonyms for @samp{@var{n}f1_1}.
19693 GCC defines two macros based on the value of this option. The first
19694 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
19695 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
19696 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
19697 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
19698 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
19700 Note that the @code{_MIPS_ARCH} macro uses the processor names given
19701 above. In other words, it has the full prefix and does not
19702 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
19703 the macro names the resolved architecture (either @code{"mips1"} or
19704 @code{"mips3"}). It names the default architecture when no
19705 @option{-march} option is given.
19707 @item -mtune=@var{arch}
19709 Optimize for @var{arch}. Among other things, this option controls
19710 the way instructions are scheduled, and the perceived cost of arithmetic
19711 operations. The list of @var{arch} values is the same as for
19714 When this option is not used, GCC optimizes for the processor
19715 specified by @option{-march}. By using @option{-march} and
19716 @option{-mtune} together, it is possible to generate code that
19717 runs on a family of processors, but optimize the code for one
19718 particular member of that family.
19720 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
19721 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
19722 @option{-march} ones described above.
19726 Equivalent to @option{-march=mips1}.
19730 Equivalent to @option{-march=mips2}.
19734 Equivalent to @option{-march=mips3}.
19738 Equivalent to @option{-march=mips4}.
19742 Equivalent to @option{-march=mips32}.
19746 Equivalent to @option{-march=mips32r3}.
19750 Equivalent to @option{-march=mips32r5}.
19754 Equivalent to @option{-march=mips32r6}.
19758 Equivalent to @option{-march=mips64}.
19762 Equivalent to @option{-march=mips64r2}.
19766 Equivalent to @option{-march=mips64r3}.
19770 Equivalent to @option{-march=mips64r5}.
19774 Equivalent to @option{-march=mips64r6}.
19779 @opindex mno-mips16
19780 Generate (do not generate) MIPS16 code. If GCC is targeting a
19781 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
19783 MIPS16 code generation can also be controlled on a per-function basis
19784 by means of @code{mips16} and @code{nomips16} attributes.
19785 @xref{Function Attributes}, for more information.
19787 @item -mflip-mips16
19788 @opindex mflip-mips16
19789 Generate MIPS16 code on alternating functions. This option is provided
19790 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
19791 not intended for ordinary use in compiling user code.
19793 @item -minterlink-compressed
19794 @item -mno-interlink-compressed
19795 @opindex minterlink-compressed
19796 @opindex mno-interlink-compressed
19797 Require (do not require) that code using the standard (uncompressed) MIPS ISA
19798 be link-compatible with MIPS16 and microMIPS code, and vice versa.
19800 For example, code using the standard ISA encoding cannot jump directly
19801 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
19802 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
19803 knows that the target of the jump is not compressed.
19805 @item -minterlink-mips16
19806 @itemx -mno-interlink-mips16
19807 @opindex minterlink-mips16
19808 @opindex mno-interlink-mips16
19809 Aliases of @option{-minterlink-compressed} and
19810 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
19811 and are retained for backwards compatibility.
19823 Generate code for the given ABI@.
19825 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
19826 generates 64-bit code when you select a 64-bit architecture, but you
19827 can use @option{-mgp32} to get 32-bit code instead.
19829 For information about the O64 ABI, see
19830 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
19832 GCC supports a variant of the o32 ABI in which floating-point registers
19833 are 64 rather than 32 bits wide. You can select this combination with
19834 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
19835 and @code{mfhc1} instructions and is therefore only supported for
19836 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
19838 The register assignments for arguments and return values remain the
19839 same, but each scalar value is passed in a single 64-bit register
19840 rather than a pair of 32-bit registers. For example, scalar
19841 floating-point values are returned in @samp{$f0} only, not a
19842 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
19843 remains the same in that the even-numbered double-precision registers
19846 Two additional variants of the o32 ABI are supported to enable
19847 a transition from 32-bit to 64-bit registers. These are FPXX
19848 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
19849 The FPXX extension mandates that all code must execute correctly
19850 when run using 32-bit or 64-bit registers. The code can be interlinked
19851 with either FP32 or FP64, but not both.
19852 The FP64A extension is similar to the FP64 extension but forbids the
19853 use of odd-numbered single-precision registers. This can be used
19854 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
19855 processors and allows both FP32 and FP64A code to interlink and
19856 run in the same process without changing FPU modes.
19859 @itemx -mno-abicalls
19861 @opindex mno-abicalls
19862 Generate (do not generate) code that is suitable for SVR4-style
19863 dynamic objects. @option{-mabicalls} is the default for SVR4-based
19868 Generate (do not generate) code that is fully position-independent,
19869 and that can therefore be linked into shared libraries. This option
19870 only affects @option{-mabicalls}.
19872 All @option{-mabicalls} code has traditionally been position-independent,
19873 regardless of options like @option{-fPIC} and @option{-fpic}. However,
19874 as an extension, the GNU toolchain allows executables to use absolute
19875 accesses for locally-binding symbols. It can also use shorter GP
19876 initialization sequences and generate direct calls to locally-defined
19877 functions. This mode is selected by @option{-mno-shared}.
19879 @option{-mno-shared} depends on binutils 2.16 or higher and generates
19880 objects that can only be linked by the GNU linker. However, the option
19881 does not affect the ABI of the final executable; it only affects the ABI
19882 of relocatable objects. Using @option{-mno-shared} generally makes
19883 executables both smaller and quicker.
19885 @option{-mshared} is the default.
19891 Assume (do not assume) that the static and dynamic linkers
19892 support PLTs and copy relocations. This option only affects
19893 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
19894 has no effect without @option{-msym32}.
19896 You can make @option{-mplt} the default by configuring
19897 GCC with @option{--with-mips-plt}. The default is
19898 @option{-mno-plt} otherwise.
19904 Lift (do not lift) the usual restrictions on the size of the global
19907 GCC normally uses a single instruction to load values from the GOT@.
19908 While this is relatively efficient, it only works if the GOT
19909 is smaller than about 64k. Anything larger causes the linker
19910 to report an error such as:
19912 @cindex relocation truncated to fit (MIPS)
19914 relocation truncated to fit: R_MIPS_GOT16 foobar
19917 If this happens, you should recompile your code with @option{-mxgot}.
19918 This works with very large GOTs, although the code is also
19919 less efficient, since it takes three instructions to fetch the
19920 value of a global symbol.
19922 Note that some linkers can create multiple GOTs. If you have such a
19923 linker, you should only need to use @option{-mxgot} when a single object
19924 file accesses more than 64k's worth of GOT entries. Very few do.
19926 These options have no effect unless GCC is generating position
19931 Assume that general-purpose registers are 32 bits wide.
19935 Assume that general-purpose registers are 64 bits wide.
19939 Assume that floating-point registers are 32 bits wide.
19943 Assume that floating-point registers are 64 bits wide.
19947 Do not assume the width of floating-point registers.
19950 @opindex mhard-float
19951 Use floating-point coprocessor instructions.
19954 @opindex msoft-float
19955 Do not use floating-point coprocessor instructions. Implement
19956 floating-point calculations using library calls instead.
19960 Equivalent to @option{-msoft-float}, but additionally asserts that the
19961 program being compiled does not perform any floating-point operations.
19962 This option is presently supported only by some bare-metal MIPS
19963 configurations, where it may select a special set of libraries
19964 that lack all floating-point support (including, for example, the
19965 floating-point @code{printf} formats).
19966 If code compiled with @option{-mno-float} accidentally contains
19967 floating-point operations, it is likely to suffer a link-time
19968 or run-time failure.
19970 @item -msingle-float
19971 @opindex msingle-float
19972 Assume that the floating-point coprocessor only supports single-precision
19975 @item -mdouble-float
19976 @opindex mdouble-float
19977 Assume that the floating-point coprocessor supports double-precision
19978 operations. This is the default.
19981 @itemx -mno-odd-spreg
19982 @opindex modd-spreg
19983 @opindex mno-odd-spreg
19984 Enable the use of odd-numbered single-precision floating-point registers
19985 for the o32 ABI. This is the default for processors that are known to
19986 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
19990 @itemx -mabs=legacy
19992 @opindex mabs=legacy
19993 These options control the treatment of the special not-a-number (NaN)
19994 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
19995 @code{neg.@i{fmt}} machine instructions.
19997 By default or when @option{-mabs=legacy} is used the legacy
19998 treatment is selected. In this case these instructions are considered
19999 arithmetic and avoided where correct operation is required and the
20000 input operand might be a NaN. A longer sequence of instructions that
20001 manipulate the sign bit of floating-point datum manually is used
20002 instead unless the @option{-ffinite-math-only} option has also been
20005 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
20006 this case these instructions are considered non-arithmetic and therefore
20007 operating correctly in all cases, including in particular where the
20008 input operand is a NaN. These instructions are therefore always used
20009 for the respective operations.
20012 @itemx -mnan=legacy
20014 @opindex mnan=legacy
20015 These options control the encoding of the special not-a-number (NaN)
20016 IEEE 754 floating-point data.
20018 The @option{-mnan=legacy} option selects the legacy encoding. In this
20019 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
20020 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
20021 by the first bit of their trailing significand field being 1.
20023 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
20024 this case qNaNs are denoted by the first bit of their trailing
20025 significand field being 1, whereas sNaNs are denoted by the first bit of
20026 their trailing significand field being 0.
20028 The default is @option{-mnan=legacy} unless GCC has been configured with
20029 @option{--with-nan=2008}.
20035 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
20036 implement atomic memory built-in functions. When neither option is
20037 specified, GCC uses the instructions if the target architecture
20040 @option{-mllsc} is useful if the runtime environment can emulate the
20041 instructions and @option{-mno-llsc} can be useful when compiling for
20042 nonstandard ISAs. You can make either option the default by
20043 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
20044 respectively. @option{--with-llsc} is the default for some
20045 configurations; see the installation documentation for details.
20051 Use (do not use) revision 1 of the MIPS DSP ASE@.
20052 @xref{MIPS DSP Built-in Functions}. This option defines the
20053 preprocessor macro @code{__mips_dsp}. It also defines
20054 @code{__mips_dsp_rev} to 1.
20060 Use (do not use) revision 2 of the MIPS DSP ASE@.
20061 @xref{MIPS DSP Built-in Functions}. This option defines the
20062 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
20063 It also defines @code{__mips_dsp_rev} to 2.
20066 @itemx -mno-smartmips
20067 @opindex msmartmips
20068 @opindex mno-smartmips
20069 Use (do not use) the MIPS SmartMIPS ASE.
20071 @item -mpaired-single
20072 @itemx -mno-paired-single
20073 @opindex mpaired-single
20074 @opindex mno-paired-single
20075 Use (do not use) paired-single floating-point instructions.
20076 @xref{MIPS Paired-Single Support}. This option requires
20077 hardware floating-point support to be enabled.
20083 Use (do not use) MIPS Digital Media Extension instructions.
20084 This option can only be used when generating 64-bit code and requires
20085 hardware floating-point support to be enabled.
20090 @opindex mno-mips3d
20091 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
20092 The option @option{-mips3d} implies @option{-mpaired-single}.
20095 @itemx -mno-micromips
20096 @opindex mmicromips
20097 @opindex mno-mmicromips
20098 Generate (do not generate) microMIPS code.
20100 MicroMIPS code generation can also be controlled on a per-function basis
20101 by means of @code{micromips} and @code{nomicromips} attributes.
20102 @xref{Function Attributes}, for more information.
20108 Use (do not use) MT Multithreading instructions.
20114 Use (do not use) the MIPS MCU ASE instructions.
20120 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
20126 Use (do not use) the MIPS Virtualization (VZ) instructions.
20132 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
20136 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
20137 an explanation of the default and the way that the pointer size is
20142 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
20144 The default size of @code{int}s, @code{long}s and pointers depends on
20145 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
20146 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
20147 32-bit @code{long}s. Pointers are the same size as @code{long}s,
20148 or the same size as integer registers, whichever is smaller.
20154 Assume (do not assume) that all symbols have 32-bit values, regardless
20155 of the selected ABI@. This option is useful in combination with
20156 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
20157 to generate shorter and faster references to symbolic addresses.
20161 Put definitions of externally-visible data in a small data section
20162 if that data is no bigger than @var{num} bytes. GCC can then generate
20163 more efficient accesses to the data; see @option{-mgpopt} for details.
20165 The default @option{-G} option depends on the configuration.
20167 @item -mlocal-sdata
20168 @itemx -mno-local-sdata
20169 @opindex mlocal-sdata
20170 @opindex mno-local-sdata
20171 Extend (do not extend) the @option{-G} behavior to local data too,
20172 such as to static variables in C@. @option{-mlocal-sdata} is the
20173 default for all configurations.
20175 If the linker complains that an application is using too much small data,
20176 you might want to try rebuilding the less performance-critical parts with
20177 @option{-mno-local-sdata}. You might also want to build large
20178 libraries with @option{-mno-local-sdata}, so that the libraries leave
20179 more room for the main program.
20181 @item -mextern-sdata
20182 @itemx -mno-extern-sdata
20183 @opindex mextern-sdata
20184 @opindex mno-extern-sdata
20185 Assume (do not assume) that externally-defined data is in
20186 a small data section if the size of that data is within the @option{-G} limit.
20187 @option{-mextern-sdata} is the default for all configurations.
20189 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
20190 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
20191 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
20192 is placed in a small data section. If @var{Var} is defined by another
20193 module, you must either compile that module with a high-enough
20194 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
20195 definition. If @var{Var} is common, you must link the application
20196 with a high-enough @option{-G} setting.
20198 The easiest way of satisfying these restrictions is to compile
20199 and link every module with the same @option{-G} option. However,
20200 you may wish to build a library that supports several different
20201 small data limits. You can do this by compiling the library with
20202 the highest supported @option{-G} setting and additionally using
20203 @option{-mno-extern-sdata} to stop the library from making assumptions
20204 about externally-defined data.
20210 Use (do not use) GP-relative accesses for symbols that are known to be
20211 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
20212 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
20215 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
20216 might not hold the value of @code{_gp}. For example, if the code is
20217 part of a library that might be used in a boot monitor, programs that
20218 call boot monitor routines pass an unknown value in @code{$gp}.
20219 (In such situations, the boot monitor itself is usually compiled
20220 with @option{-G0}.)
20222 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
20223 @option{-mno-extern-sdata}.
20225 @item -membedded-data
20226 @itemx -mno-embedded-data
20227 @opindex membedded-data
20228 @opindex mno-embedded-data
20229 Allocate variables to the read-only data section first if possible, then
20230 next in the small data section if possible, otherwise in data. This gives
20231 slightly slower code than the default, but reduces the amount of RAM required
20232 when executing, and thus may be preferred for some embedded systems.
20234 @item -muninit-const-in-rodata
20235 @itemx -mno-uninit-const-in-rodata
20236 @opindex muninit-const-in-rodata
20237 @opindex mno-uninit-const-in-rodata
20238 Put uninitialized @code{const} variables in the read-only data section.
20239 This option is only meaningful in conjunction with @option{-membedded-data}.
20241 @item -mcode-readable=@var{setting}
20242 @opindex mcode-readable
20243 Specify whether GCC may generate code that reads from executable sections.
20244 There are three possible settings:
20247 @item -mcode-readable=yes
20248 Instructions may freely access executable sections. This is the
20251 @item -mcode-readable=pcrel
20252 MIPS16 PC-relative load instructions can access executable sections,
20253 but other instructions must not do so. This option is useful on 4KSc
20254 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
20255 It is also useful on processors that can be configured to have a dual
20256 instruction/data SRAM interface and that, like the M4K, automatically
20257 redirect PC-relative loads to the instruction RAM.
20259 @item -mcode-readable=no
20260 Instructions must not access executable sections. This option can be
20261 useful on targets that are configured to have a dual instruction/data
20262 SRAM interface but that (unlike the M4K) do not automatically redirect
20263 PC-relative loads to the instruction RAM.
20266 @item -msplit-addresses
20267 @itemx -mno-split-addresses
20268 @opindex msplit-addresses
20269 @opindex mno-split-addresses
20270 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
20271 relocation operators. This option has been superseded by
20272 @option{-mexplicit-relocs} but is retained for backwards compatibility.
20274 @item -mexplicit-relocs
20275 @itemx -mno-explicit-relocs
20276 @opindex mexplicit-relocs
20277 @opindex mno-explicit-relocs
20278 Use (do not use) assembler relocation operators when dealing with symbolic
20279 addresses. The alternative, selected by @option{-mno-explicit-relocs},
20280 is to use assembler macros instead.
20282 @option{-mexplicit-relocs} is the default if GCC was configured
20283 to use an assembler that supports relocation operators.
20285 @item -mcheck-zero-division
20286 @itemx -mno-check-zero-division
20287 @opindex mcheck-zero-division
20288 @opindex mno-check-zero-division
20289 Trap (do not trap) on integer division by zero.
20291 The default is @option{-mcheck-zero-division}.
20293 @item -mdivide-traps
20294 @itemx -mdivide-breaks
20295 @opindex mdivide-traps
20296 @opindex mdivide-breaks
20297 MIPS systems check for division by zero by generating either a
20298 conditional trap or a break instruction. Using traps results in
20299 smaller code, but is only supported on MIPS II and later. Also, some
20300 versions of the Linux kernel have a bug that prevents trap from
20301 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
20302 allow conditional traps on architectures that support them and
20303 @option{-mdivide-breaks} to force the use of breaks.
20305 The default is usually @option{-mdivide-traps}, but this can be
20306 overridden at configure time using @option{--with-divide=breaks}.
20307 Divide-by-zero checks can be completely disabled using
20308 @option{-mno-check-zero-division}.
20310 @item -mload-store-pairs
20311 @itemx -mno-load-store-pairs
20312 @opindex mload-store-pairs
20313 @opindex mno-load-store-pairs
20314 Enable (disable) an optimization that pairs consecutive load or store
20315 instructions to enable load/store bonding. This option is enabled by
20316 default but only takes effect when the selected architecture is known
20317 to support bonding.
20322 @opindex mno-memcpy
20323 Force (do not force) the use of @code{memcpy} for non-trivial block
20324 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
20325 most constant-sized copies.
20328 @itemx -mno-long-calls
20329 @opindex mlong-calls
20330 @opindex mno-long-calls
20331 Disable (do not disable) use of the @code{jal} instruction. Calling
20332 functions using @code{jal} is more efficient but requires the caller
20333 and callee to be in the same 256 megabyte segment.
20335 This option has no effect on abicalls code. The default is
20336 @option{-mno-long-calls}.
20342 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
20343 instructions, as provided by the R4650 ISA@.
20349 Enable (disable) use of the @code{madd} and @code{msub} integer
20350 instructions. The default is @option{-mimadd} on architectures
20351 that support @code{madd} and @code{msub} except for the 74k
20352 architecture where it was found to generate slower code.
20355 @itemx -mno-fused-madd
20356 @opindex mfused-madd
20357 @opindex mno-fused-madd
20358 Enable (disable) use of the floating-point multiply-accumulate
20359 instructions, when they are available. The default is
20360 @option{-mfused-madd}.
20362 On the R8000 CPU when multiply-accumulate instructions are used,
20363 the intermediate product is calculated to infinite precision
20364 and is not subject to the FCSR Flush to Zero bit. This may be
20365 undesirable in some circumstances. On other processors the result
20366 is numerically identical to the equivalent computation using
20367 separate multiply, add, subtract and negate instructions.
20371 Tell the MIPS assembler to not run its preprocessor over user
20372 assembler files (with a @samp{.s} suffix) when assembling them.
20377 @opindex mno-fix-24k
20378 Work around the 24K E48 (lost data on stores during refill) errata.
20379 The workarounds are implemented by the assembler rather than by GCC@.
20382 @itemx -mno-fix-r4000
20383 @opindex mfix-r4000
20384 @opindex mno-fix-r4000
20385 Work around certain R4000 CPU errata:
20388 A double-word or a variable shift may give an incorrect result if executed
20389 immediately after starting an integer division.
20391 A double-word or a variable shift may give an incorrect result if executed
20392 while an integer multiplication is in progress.
20394 An integer division may give an incorrect result if started in a delay slot
20395 of a taken branch or a jump.
20399 @itemx -mno-fix-r4400
20400 @opindex mfix-r4400
20401 @opindex mno-fix-r4400
20402 Work around certain R4400 CPU errata:
20405 A double-word or a variable shift may give an incorrect result if executed
20406 immediately after starting an integer division.
20410 @itemx -mno-fix-r10000
20411 @opindex mfix-r10000
20412 @opindex mno-fix-r10000
20413 Work around certain R10000 errata:
20416 @code{ll}/@code{sc} sequences may not behave atomically on revisions
20417 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
20420 This option can only be used if the target architecture supports
20421 branch-likely instructions. @option{-mfix-r10000} is the default when
20422 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
20426 @itemx -mno-fix-rm7000
20427 @opindex mfix-rm7000
20428 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
20429 workarounds are implemented by the assembler rather than by GCC@.
20432 @itemx -mno-fix-vr4120
20433 @opindex mfix-vr4120
20434 Work around certain VR4120 errata:
20437 @code{dmultu} does not always produce the correct result.
20439 @code{div} and @code{ddiv} do not always produce the correct result if one
20440 of the operands is negative.
20442 The workarounds for the division errata rely on special functions in
20443 @file{libgcc.a}. At present, these functions are only provided by
20444 the @code{mips64vr*-elf} configurations.
20446 Other VR4120 errata require a NOP to be inserted between certain pairs of
20447 instructions. These errata are handled by the assembler, not by GCC itself.
20450 @opindex mfix-vr4130
20451 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
20452 workarounds are implemented by the assembler rather than by GCC,
20453 although GCC avoids using @code{mflo} and @code{mfhi} if the
20454 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
20455 instructions are available instead.
20458 @itemx -mno-fix-sb1
20460 Work around certain SB-1 CPU core errata.
20461 (This flag currently works around the SB-1 revision 2
20462 ``F1'' and ``F2'' floating-point errata.)
20464 @item -mr10k-cache-barrier=@var{setting}
20465 @opindex mr10k-cache-barrier
20466 Specify whether GCC should insert cache barriers to avoid the
20467 side-effects of speculation on R10K processors.
20469 In common with many processors, the R10K tries to predict the outcome
20470 of a conditional branch and speculatively executes instructions from
20471 the ``taken'' branch. It later aborts these instructions if the
20472 predicted outcome is wrong. However, on the R10K, even aborted
20473 instructions can have side effects.
20475 This problem only affects kernel stores and, depending on the system,
20476 kernel loads. As an example, a speculatively-executed store may load
20477 the target memory into cache and mark the cache line as dirty, even if
20478 the store itself is later aborted. If a DMA operation writes to the
20479 same area of memory before the ``dirty'' line is flushed, the cached
20480 data overwrites the DMA-ed data. See the R10K processor manual
20481 for a full description, including other potential problems.
20483 One workaround is to insert cache barrier instructions before every memory
20484 access that might be speculatively executed and that might have side
20485 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
20486 controls GCC's implementation of this workaround. It assumes that
20487 aborted accesses to any byte in the following regions does not have
20492 the memory occupied by the current function's stack frame;
20495 the memory occupied by an incoming stack argument;
20498 the memory occupied by an object with a link-time-constant address.
20501 It is the kernel's responsibility to ensure that speculative
20502 accesses to these regions are indeed safe.
20504 If the input program contains a function declaration such as:
20510 then the implementation of @code{foo} must allow @code{j foo} and
20511 @code{jal foo} to be executed speculatively. GCC honors this
20512 restriction for functions it compiles itself. It expects non-GCC
20513 functions (such as hand-written assembly code) to do the same.
20515 The option has three forms:
20518 @item -mr10k-cache-barrier=load-store
20519 Insert a cache barrier before a load or store that might be
20520 speculatively executed and that might have side effects even
20523 @item -mr10k-cache-barrier=store
20524 Insert a cache barrier before a store that might be speculatively
20525 executed and that might have side effects even if aborted.
20527 @item -mr10k-cache-barrier=none
20528 Disable the insertion of cache barriers. This is the default setting.
20531 @item -mflush-func=@var{func}
20532 @itemx -mno-flush-func
20533 @opindex mflush-func
20534 Specifies the function to call to flush the I and D caches, or to not
20535 call any such function. If called, the function must take the same
20536 arguments as the common @code{_flush_func}, that is, the address of the
20537 memory range for which the cache is being flushed, the size of the
20538 memory range, and the number 3 (to flush both caches). The default
20539 depends on the target GCC was configured for, but commonly is either
20540 @code{_flush_func} or @code{__cpu_flush}.
20542 @item mbranch-cost=@var{num}
20543 @opindex mbranch-cost
20544 Set the cost of branches to roughly @var{num} ``simple'' instructions.
20545 This cost is only a heuristic and is not guaranteed to produce
20546 consistent results across releases. A zero cost redundantly selects
20547 the default, which is based on the @option{-mtune} setting.
20549 @item -mbranch-likely
20550 @itemx -mno-branch-likely
20551 @opindex mbranch-likely
20552 @opindex mno-branch-likely
20553 Enable or disable use of Branch Likely instructions, regardless of the
20554 default for the selected architecture. By default, Branch Likely
20555 instructions may be generated if they are supported by the selected
20556 architecture. An exception is for the MIPS32 and MIPS64 architectures
20557 and processors that implement those architectures; for those, Branch
20558 Likely instructions are not be generated by default because the MIPS32
20559 and MIPS64 architectures specifically deprecate their use.
20561 @item -mcompact-branches=never
20562 @itemx -mcompact-branches=optimal
20563 @itemx -mcompact-branches=always
20564 @opindex mcompact-branches=never
20565 @opindex mcompact-branches=optimal
20566 @opindex mcompact-branches=always
20567 These options control which form of branches will be generated. The
20568 default is @option{-mcompact-branches=optimal}.
20570 The @option{-mcompact-branches=never} option ensures that compact branch
20571 instructions will never be generated.
20573 The @option{-mcompact-branches=always} option ensures that a compact
20574 branch instruction will be generated if available. If a compact branch
20575 instruction is not available, a delay slot form of the branch will be
20578 This option is supported from MIPS Release 6 onwards.
20580 The @option{-mcompact-branches=optimal} option will cause a delay slot
20581 branch to be used if one is available in the current ISA and the delay
20582 slot is successfully filled. If the delay slot is not filled, a compact
20583 branch will be chosen if one is available.
20585 @item -mfp-exceptions
20586 @itemx -mno-fp-exceptions
20587 @opindex mfp-exceptions
20588 Specifies whether FP exceptions are enabled. This affects how
20589 FP instructions are scheduled for some processors.
20590 The default is that FP exceptions are
20593 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
20594 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
20597 @item -mvr4130-align
20598 @itemx -mno-vr4130-align
20599 @opindex mvr4130-align
20600 The VR4130 pipeline is two-way superscalar, but can only issue two
20601 instructions together if the first one is 8-byte aligned. When this
20602 option is enabled, GCC aligns pairs of instructions that it
20603 thinks should execute in parallel.
20605 This option only has an effect when optimizing for the VR4130.
20606 It normally makes code faster, but at the expense of making it bigger.
20607 It is enabled by default at optimization level @option{-O3}.
20612 Enable (disable) generation of @code{synci} instructions on
20613 architectures that support it. The @code{synci} instructions (if
20614 enabled) are generated when @code{__builtin___clear_cache} is
20617 This option defaults to @option{-mno-synci}, but the default can be
20618 overridden by configuring GCC with @option{--with-synci}.
20620 When compiling code for single processor systems, it is generally safe
20621 to use @code{synci}. However, on many multi-core (SMP) systems, it
20622 does not invalidate the instruction caches on all cores and may lead
20623 to undefined behavior.
20625 @item -mrelax-pic-calls
20626 @itemx -mno-relax-pic-calls
20627 @opindex mrelax-pic-calls
20628 Try to turn PIC calls that are normally dispatched via register
20629 @code{$25} into direct calls. This is only possible if the linker can
20630 resolve the destination at link time and if the destination is within
20631 range for a direct call.
20633 @option{-mrelax-pic-calls} is the default if GCC was configured to use
20634 an assembler and a linker that support the @code{.reloc} assembly
20635 directive and @option{-mexplicit-relocs} is in effect. With
20636 @option{-mno-explicit-relocs}, this optimization can be performed by the
20637 assembler and the linker alone without help from the compiler.
20639 @item -mmcount-ra-address
20640 @itemx -mno-mcount-ra-address
20641 @opindex mmcount-ra-address
20642 @opindex mno-mcount-ra-address
20643 Emit (do not emit) code that allows @code{_mcount} to modify the
20644 calling function's return address. When enabled, this option extends
20645 the usual @code{_mcount} interface with a new @var{ra-address}
20646 parameter, which has type @code{intptr_t *} and is passed in register
20647 @code{$12}. @code{_mcount} can then modify the return address by
20648 doing both of the following:
20651 Returning the new address in register @code{$31}.
20653 Storing the new address in @code{*@var{ra-address}},
20654 if @var{ra-address} is nonnull.
20657 The default is @option{-mno-mcount-ra-address}.
20659 @item -mframe-header-opt
20660 @itemx -mno-frame-header-opt
20661 @opindex mframe-header-opt
20662 Enable (disable) frame header optimization in the o32 ABI. When using the
20663 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
20664 function to write out register arguments. When enabled, this optimization
20665 will suppress the allocation of the frame header if it can be determined that
20668 This optimization is off by default at all optimization levels.
20671 @itemx -mno-lxc1-sxc1
20672 @opindex mlxc1-sxc1
20673 When applicable, enable (disable) the generation of @code{lwxc1},
20674 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
20679 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
20680 @code{madd.d} and related instructions. Enabled by default.
20685 @subsection MMIX Options
20686 @cindex MMIX Options
20688 These options are defined for the MMIX:
20692 @itemx -mno-libfuncs
20694 @opindex mno-libfuncs
20695 Specify that intrinsic library functions are being compiled, passing all
20696 values in registers, no matter the size.
20699 @itemx -mno-epsilon
20701 @opindex mno-epsilon
20702 Generate floating-point comparison instructions that compare with respect
20703 to the @code{rE} epsilon register.
20705 @item -mabi=mmixware
20707 @opindex mabi=mmixware
20709 Generate code that passes function parameters and return values that (in
20710 the called function) are seen as registers @code{$0} and up, as opposed to
20711 the GNU ABI which uses global registers @code{$231} and up.
20713 @item -mzero-extend
20714 @itemx -mno-zero-extend
20715 @opindex mzero-extend
20716 @opindex mno-zero-extend
20717 When reading data from memory in sizes shorter than 64 bits, use (do not
20718 use) zero-extending load instructions by default, rather than
20719 sign-extending ones.
20722 @itemx -mno-knuthdiv
20724 @opindex mno-knuthdiv
20725 Make the result of a division yielding a remainder have the same sign as
20726 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
20727 remainder follows the sign of the dividend. Both methods are
20728 arithmetically valid, the latter being almost exclusively used.
20730 @item -mtoplevel-symbols
20731 @itemx -mno-toplevel-symbols
20732 @opindex mtoplevel-symbols
20733 @opindex mno-toplevel-symbols
20734 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
20735 code can be used with the @code{PREFIX} assembly directive.
20739 Generate an executable in the ELF format, rather than the default
20740 @samp{mmo} format used by the @command{mmix} simulator.
20742 @item -mbranch-predict
20743 @itemx -mno-branch-predict
20744 @opindex mbranch-predict
20745 @opindex mno-branch-predict
20746 Use (do not use) the probable-branch instructions, when static branch
20747 prediction indicates a probable branch.
20749 @item -mbase-addresses
20750 @itemx -mno-base-addresses
20751 @opindex mbase-addresses
20752 @opindex mno-base-addresses
20753 Generate (do not generate) code that uses @emph{base addresses}. Using a
20754 base address automatically generates a request (handled by the assembler
20755 and the linker) for a constant to be set up in a global register. The
20756 register is used for one or more base address requests within the range 0
20757 to 255 from the value held in the register. The generally leads to short
20758 and fast code, but the number of different data items that can be
20759 addressed is limited. This means that a program that uses lots of static
20760 data may require @option{-mno-base-addresses}.
20762 @item -msingle-exit
20763 @itemx -mno-single-exit
20764 @opindex msingle-exit
20765 @opindex mno-single-exit
20766 Force (do not force) generated code to have a single exit point in each
20770 @node MN10300 Options
20771 @subsection MN10300 Options
20772 @cindex MN10300 options
20774 These @option{-m} options are defined for Matsushita MN10300 architectures:
20779 Generate code to avoid bugs in the multiply instructions for the MN10300
20780 processors. This is the default.
20782 @item -mno-mult-bug
20783 @opindex mno-mult-bug
20784 Do not generate code to avoid bugs in the multiply instructions for the
20785 MN10300 processors.
20789 Generate code using features specific to the AM33 processor.
20793 Do not generate code using features specific to the AM33 processor. This
20798 Generate code using features specific to the AM33/2.0 processor.
20802 Generate code using features specific to the AM34 processor.
20804 @item -mtune=@var{cpu-type}
20806 Use the timing characteristics of the indicated CPU type when
20807 scheduling instructions. This does not change the targeted processor
20808 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
20809 @samp{am33-2} or @samp{am34}.
20811 @item -mreturn-pointer-on-d0
20812 @opindex mreturn-pointer-on-d0
20813 When generating a function that returns a pointer, return the pointer
20814 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
20815 only in @code{a0}, and attempts to call such functions without a prototype
20816 result in errors. Note that this option is on by default; use
20817 @option{-mno-return-pointer-on-d0} to disable it.
20821 Do not link in the C run-time initialization object file.
20825 Indicate to the linker that it should perform a relaxation optimization pass
20826 to shorten branches, calls and absolute memory addresses. This option only
20827 has an effect when used on the command line for the final link step.
20829 This option makes symbolic debugging impossible.
20833 Allow the compiler to generate @emph{Long Instruction Word}
20834 instructions if the target is the @samp{AM33} or later. This is the
20835 default. This option defines the preprocessor macro @code{__LIW__}.
20839 Do not allow the compiler to generate @emph{Long Instruction Word}
20840 instructions. This option defines the preprocessor macro
20845 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
20846 instructions if the target is the @samp{AM33} or later. This is the
20847 default. This option defines the preprocessor macro @code{__SETLB__}.
20851 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
20852 instructions. This option defines the preprocessor macro
20853 @code{__NO_SETLB__}.
20857 @node Moxie Options
20858 @subsection Moxie Options
20859 @cindex Moxie Options
20865 Generate big-endian code. This is the default for @samp{moxie-*-*}
20870 Generate little-endian code.
20874 Generate mul.x and umul.x instructions. This is the default for
20875 @samp{moxiebox-*-*} configurations.
20879 Do not link in the C run-time initialization object file.
20883 @node MSP430 Options
20884 @subsection MSP430 Options
20885 @cindex MSP430 Options
20887 These options are defined for the MSP430:
20893 Force assembly output to always use hex constants. Normally such
20894 constants are signed decimals, but this option is available for
20895 testsuite and/or aesthetic purposes.
20899 Select the MCU to target. This is used to create a C preprocessor
20900 symbol based upon the MCU name, converted to upper case and pre- and
20901 post-fixed with @samp{__}. This in turn is used by the
20902 @file{msp430.h} header file to select an MCU-specific supplementary
20905 The option also sets the ISA to use. If the MCU name is one that is
20906 known to only support the 430 ISA then that is selected, otherwise the
20907 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
20908 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
20909 name selects the 430X ISA.
20911 In addition an MCU-specific linker script is added to the linker
20912 command line. The script's name is the name of the MCU with
20913 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
20914 command line defines the C preprocessor symbol @code{__XXX__} and
20915 cause the linker to search for a script called @file{xxx.ld}.
20917 This option is also passed on to the assembler.
20920 @itemx -mno-warn-mcu
20922 @opindex mno-warn-mcu
20923 This option enables or disables warnings about conflicts between the
20924 MCU name specified by the @option{-mmcu} option and the ISA set by the
20925 @option{-mcpu} option and/or the hardware multiply support set by the
20926 @option{-mhwmult} option. It also toggles warnings about unrecognized
20927 MCU names. This option is on by default.
20931 Specifies the ISA to use. Accepted values are @samp{msp430},
20932 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
20933 @option{-mmcu=} option should be used to select the ISA.
20937 Link to the simulator runtime libraries and linker script. Overrides
20938 any scripts that would be selected by the @option{-mmcu=} option.
20942 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
20946 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
20950 This option is passed to the assembler and linker, and allows the
20951 linker to perform certain optimizations that cannot be done until
20956 Describes the type of hardware multiply supported by the target.
20957 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
20958 for the original 16-bit-only multiply supported by early MCUs.
20959 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
20960 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
20961 A value of @samp{auto} can also be given. This tells GCC to deduce
20962 the hardware multiply support based upon the MCU name provided by the
20963 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
20964 the MCU name is not recognized then no hardware multiply support is
20965 assumed. @code{auto} is the default setting.
20967 Hardware multiplies are normally performed by calling a library
20968 routine. This saves space in the generated code. When compiling at
20969 @option{-O3} or higher however the hardware multiplier is invoked
20970 inline. This makes for bigger, but faster code.
20972 The hardware multiply routines disable interrupts whilst running and
20973 restore the previous interrupt state when they finish. This makes
20974 them safe to use inside interrupt handlers as well as in normal code.
20978 Enable the use of a minimum runtime environment - no static
20979 initializers or constructors. This is intended for memory-constrained
20980 devices. The compiler includes special symbols in some objects
20981 that tell the linker and runtime which code fragments are required.
20983 @item -mcode-region=
20984 @itemx -mdata-region=
20985 @opindex mcode-region
20986 @opindex mdata-region
20987 These options tell the compiler where to place functions and data that
20988 do not have one of the @code{lower}, @code{upper}, @code{either} or
20989 @code{section} attributes. Possible values are @code{lower},
20990 @code{upper}, @code{either} or @code{any}. The first three behave
20991 like the corresponding attribute. The fourth possible value -
20992 @code{any} - is the default. It leaves placement entirely up to the
20993 linker script and how it assigns the standard sections
20994 (@code{.text}, @code{.data}, etc) to the memory regions.
20996 @item -msilicon-errata=
20997 @opindex msilicon-errata
20998 This option passes on a request to assembler to enable the fixes for
20999 the named silicon errata.
21001 @item -msilicon-errata-warn=
21002 @opindex msilicon-errata-warn
21003 This option passes on a request to the assembler to enable warning
21004 messages when a silicon errata might need to be applied.
21008 @node NDS32 Options
21009 @subsection NDS32 Options
21010 @cindex NDS32 Options
21012 These options are defined for NDS32 implementations:
21017 @opindex mbig-endian
21018 Generate code in big-endian mode.
21020 @item -mlittle-endian
21021 @opindex mlittle-endian
21022 Generate code in little-endian mode.
21024 @item -mreduced-regs
21025 @opindex mreduced-regs
21026 Use reduced-set registers for register allocation.
21029 @opindex mfull-regs
21030 Use full-set registers for register allocation.
21034 Generate conditional move instructions.
21038 Do not generate conditional move instructions.
21042 Generate performance extension instructions.
21044 @item -mno-perf-ext
21045 @opindex mno-perf-ext
21046 Do not generate performance extension instructions.
21050 Generate v3 push25/pop25 instructions.
21053 @opindex mno-v3push
21054 Do not generate v3 push25/pop25 instructions.
21058 Generate 16-bit instructions.
21061 @opindex mno-16-bit
21062 Do not generate 16-bit instructions.
21064 @item -misr-vector-size=@var{num}
21065 @opindex misr-vector-size
21066 Specify the size of each interrupt vector, which must be 4 or 16.
21068 @item -mcache-block-size=@var{num}
21069 @opindex mcache-block-size
21070 Specify the size of each cache block,
21071 which must be a power of 2 between 4 and 512.
21073 @item -march=@var{arch}
21075 Specify the name of the target architecture.
21077 @item -mcmodel=@var{code-model}
21079 Set the code model to one of
21082 All the data and read-only data segments must be within 512KB addressing space.
21083 The text segment must be within 16MB addressing space.
21084 @item @samp{medium}
21085 The data segment must be within 512KB while the read-only data segment can be
21086 within 4GB addressing space. The text segment should be still within 16MB
21089 All the text and data segments can be within 4GB addressing space.
21093 @opindex mctor-dtor
21094 Enable constructor/destructor feature.
21098 Guide linker to relax instructions.
21102 @node Nios II Options
21103 @subsection Nios II Options
21104 @cindex Nios II options
21105 @cindex Altera Nios II options
21107 These are the options defined for the Altera Nios II processor.
21113 @cindex smaller data references
21114 Put global and static objects less than or equal to @var{num} bytes
21115 into the small data or BSS sections instead of the normal data or BSS
21116 sections. The default value of @var{num} is 8.
21118 @item -mgpopt=@var{option}
21123 Generate (do not generate) GP-relative accesses. The following
21124 @var{option} names are recognized:
21129 Do not generate GP-relative accesses.
21132 Generate GP-relative accesses for small data objects that are not
21133 external, weak, or uninitialized common symbols.
21134 Also use GP-relative addressing for objects that
21135 have been explicitly placed in a small data section via a @code{section}
21139 As for @samp{local}, but also generate GP-relative accesses for
21140 small data objects that are external, weak, or common. If you use this option,
21141 you must ensure that all parts of your program (including libraries) are
21142 compiled with the same @option{-G} setting.
21145 Generate GP-relative accesses for all data objects in the program. If you
21146 use this option, the entire data and BSS segments
21147 of your program must fit in 64K of memory and you must use an appropriate
21148 linker script to allocate them within the addressable range of the
21152 Generate GP-relative addresses for function pointers as well as data
21153 pointers. If you use this option, the entire text, data, and BSS segments
21154 of your program must fit in 64K of memory and you must use an appropriate
21155 linker script to allocate them within the addressable range of the
21160 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
21161 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
21163 The default is @option{-mgpopt} except when @option{-fpic} or
21164 @option{-fPIC} is specified to generate position-independent code.
21165 Note that the Nios II ABI does not permit GP-relative accesses from
21168 You may need to specify @option{-mno-gpopt} explicitly when building
21169 programs that include large amounts of small data, including large
21170 GOT data sections. In this case, the 16-bit offset for GP-relative
21171 addressing may not be large enough to allow access to the entire
21172 small data section.
21178 Generate little-endian (default) or big-endian (experimental) code,
21181 @item -march=@var{arch}
21183 This specifies the name of the target Nios II architecture. GCC uses this
21184 name to determine what kind of instructions it can emit when generating
21185 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
21187 The preprocessor macro @code{__nios2_arch__} is available to programs,
21188 with value 1 or 2, indicating the targeted ISA level.
21190 @item -mbypass-cache
21191 @itemx -mno-bypass-cache
21192 @opindex mno-bypass-cache
21193 @opindex mbypass-cache
21194 Force all load and store instructions to always bypass cache by
21195 using I/O variants of the instructions. The default is not to
21198 @item -mno-cache-volatile
21199 @itemx -mcache-volatile
21200 @opindex mcache-volatile
21201 @opindex mno-cache-volatile
21202 Volatile memory access bypass the cache using the I/O variants of
21203 the load and store instructions. The default is not to bypass the cache.
21205 @item -mno-fast-sw-div
21206 @itemx -mfast-sw-div
21207 @opindex mno-fast-sw-div
21208 @opindex mfast-sw-div
21209 Do not use table-based fast divide for small numbers. The default
21210 is to use the fast divide at @option{-O3} and above.
21214 @itemx -mno-hw-mulx
21218 @opindex mno-hw-mul
21220 @opindex mno-hw-mulx
21222 @opindex mno-hw-div
21224 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
21225 instructions by the compiler. The default is to emit @code{mul}
21226 and not emit @code{div} and @code{mulx}.
21232 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
21233 CDX (code density) instructions. Enabling these instructions also
21234 requires @option{-march=r2}. Since these instructions are optional
21235 extensions to the R2 architecture, the default is not to emit them.
21237 @item -mcustom-@var{insn}=@var{N}
21238 @itemx -mno-custom-@var{insn}
21239 @opindex mcustom-@var{insn}
21240 @opindex mno-custom-@var{insn}
21241 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
21242 custom instruction with encoding @var{N} when generating code that uses
21243 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
21244 instruction 253 for single-precision floating-point add operations instead
21245 of the default behavior of using a library call.
21247 The following values of @var{insn} are supported. Except as otherwise
21248 noted, floating-point operations are expected to be implemented with
21249 normal IEEE 754 semantics and correspond directly to the C operators or the
21250 equivalent GCC built-in functions (@pxref{Other Builtins}).
21252 Single-precision floating point:
21255 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
21256 Binary arithmetic operations.
21262 Unary absolute value.
21264 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
21265 Comparison operations.
21267 @item @samp{fmins}, @samp{fmaxs}
21268 Floating-point minimum and maximum. These instructions are only
21269 generated if @option{-ffinite-math-only} is specified.
21271 @item @samp{fsqrts}
21272 Unary square root operation.
21274 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
21275 Floating-point trigonometric and exponential functions. These instructions
21276 are only generated if @option{-funsafe-math-optimizations} is also specified.
21280 Double-precision floating point:
21283 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
21284 Binary arithmetic operations.
21290 Unary absolute value.
21292 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
21293 Comparison operations.
21295 @item @samp{fmind}, @samp{fmaxd}
21296 Double-precision minimum and maximum. These instructions are only
21297 generated if @option{-ffinite-math-only} is specified.
21299 @item @samp{fsqrtd}
21300 Unary square root operation.
21302 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
21303 Double-precision trigonometric and exponential functions. These instructions
21304 are only generated if @option{-funsafe-math-optimizations} is also specified.
21310 @item @samp{fextsd}
21311 Conversion from single precision to double precision.
21313 @item @samp{ftruncds}
21314 Conversion from double precision to single precision.
21316 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
21317 Conversion from floating point to signed or unsigned integer types, with
21318 truncation towards zero.
21321 Conversion from single-precision floating point to signed integer,
21322 rounding to the nearest integer and ties away from zero.
21323 This corresponds to the @code{__builtin_lroundf} function when
21324 @option{-fno-math-errno} is used.
21326 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
21327 Conversion from signed or unsigned integer types to floating-point types.
21331 In addition, all of the following transfer instructions for internal
21332 registers X and Y must be provided to use any of the double-precision
21333 floating-point instructions. Custom instructions taking two
21334 double-precision source operands expect the first operand in the
21335 64-bit register X. The other operand (or only operand of a unary
21336 operation) is given to the custom arithmetic instruction with the
21337 least significant half in source register @var{src1} and the most
21338 significant half in @var{src2}. A custom instruction that returns a
21339 double-precision result returns the most significant 32 bits in the
21340 destination register and the other half in 32-bit register Y.
21341 GCC automatically generates the necessary code sequences to write
21342 register X and/or read register Y when double-precision floating-point
21343 instructions are used.
21348 Write @var{src1} into the least significant half of X and @var{src2} into
21349 the most significant half of X.
21352 Write @var{src1} into Y.
21354 @item @samp{frdxhi}, @samp{frdxlo}
21355 Read the most or least (respectively) significant half of X and store it in
21359 Read the value of Y and store it into @var{dest}.
21362 Note that you can gain more local control over generation of Nios II custom
21363 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
21364 and @code{target("no-custom-@var{insn}")} function attributes
21365 (@pxref{Function Attributes})
21366 or pragmas (@pxref{Function Specific Option Pragmas}).
21368 @item -mcustom-fpu-cfg=@var{name}
21369 @opindex mcustom-fpu-cfg
21371 This option enables a predefined, named set of custom instruction encodings
21372 (see @option{-mcustom-@var{insn}} above).
21373 Currently, the following sets are defined:
21375 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
21376 @gccoptlist{-mcustom-fmuls=252 @gol
21377 -mcustom-fadds=253 @gol
21378 -mcustom-fsubs=254 @gol
21379 -fsingle-precision-constant}
21381 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
21382 @gccoptlist{-mcustom-fmuls=252 @gol
21383 -mcustom-fadds=253 @gol
21384 -mcustom-fsubs=254 @gol
21385 -mcustom-fdivs=255 @gol
21386 -fsingle-precision-constant}
21388 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
21389 @gccoptlist{-mcustom-floatus=243 @gol
21390 -mcustom-fixsi=244 @gol
21391 -mcustom-floatis=245 @gol
21392 -mcustom-fcmpgts=246 @gol
21393 -mcustom-fcmples=249 @gol
21394 -mcustom-fcmpeqs=250 @gol
21395 -mcustom-fcmpnes=251 @gol
21396 -mcustom-fmuls=252 @gol
21397 -mcustom-fadds=253 @gol
21398 -mcustom-fsubs=254 @gol
21399 -mcustom-fdivs=255 @gol
21400 -fsingle-precision-constant}
21402 Custom instruction assignments given by individual
21403 @option{-mcustom-@var{insn}=} options override those given by
21404 @option{-mcustom-fpu-cfg=}, regardless of the
21405 order of the options on the command line.
21407 Note that you can gain more local control over selection of a FPU
21408 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
21409 function attribute (@pxref{Function Attributes})
21410 or pragma (@pxref{Function Specific Option Pragmas}).
21414 These additional @samp{-m} options are available for the Altera Nios II
21415 ELF (bare-metal) target:
21421 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
21422 startup and termination code, and is typically used in conjunction with
21423 @option{-msys-crt0=} to specify the location of the alternate startup code
21424 provided by the HAL BSP.
21428 Link with a limited version of the C library, @option{-lsmallc}, rather than
21431 @item -msys-crt0=@var{startfile}
21433 @var{startfile} is the file name of the startfile (crt0) to use
21434 when linking. This option is only useful in conjunction with @option{-mhal}.
21436 @item -msys-lib=@var{systemlib}
21438 @var{systemlib} is the library name of the library that provides
21439 low-level system calls required by the C library,
21440 e.g. @code{read} and @code{write}.
21441 This option is typically used to link with a library provided by a HAL BSP.
21445 @node Nvidia PTX Options
21446 @subsection Nvidia PTX Options
21447 @cindex Nvidia PTX options
21448 @cindex nvptx options
21450 These options are defined for Nvidia PTX:
21458 Generate code for 32-bit or 64-bit ABI.
21461 @opindex mmainkernel
21462 Link in code for a __main kernel. This is for stand-alone instead of
21463 offloading execution.
21467 Apply partitioned execution optimizations. This is the default when any
21468 level of optimization is selected.
21471 @opindex msoft-stack
21472 Generate code that does not use @code{.local} memory
21473 directly for stack storage. Instead, a per-warp stack pointer is
21474 maintained explicitly. This enables variable-length stack allocation (with
21475 variable-length arrays or @code{alloca}), and when global memory is used for
21476 underlying storage, makes it possible to access automatic variables from other
21477 threads, or with atomic instructions. This code generation variant is used
21478 for OpenMP offloading, but the option is exposed on its own for the purpose
21479 of testing the compiler; to generate code suitable for linking into programs
21480 using OpenMP offloading, use option @option{-mgomp}.
21482 @item -muniform-simt
21483 @opindex muniform-simt
21484 Switch to code generation variant that allows to execute all threads in each
21485 warp, while maintaining memory state and side effects as if only one thread
21486 in each warp was active outside of OpenMP SIMD regions. All atomic operations
21487 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
21488 current lane index equals the master lane index), and the register being
21489 assigned is copied via a shuffle instruction from the master lane. Outside of
21490 SIMD regions lane 0 is the master; inside, each thread sees itself as the
21491 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
21492 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
21493 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
21494 with current lane index to compute the master lane index.
21498 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
21499 @option{-muniform-simt} options, and selects corresponding multilib variant.
21503 @node PDP-11 Options
21504 @subsection PDP-11 Options
21505 @cindex PDP-11 Options
21507 These options are defined for the PDP-11:
21512 Use hardware FPP floating point. This is the default. (FIS floating
21513 point on the PDP-11/40 is not supported.)
21516 @opindex msoft-float
21517 Do not use hardware floating point.
21521 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
21525 Return floating-point results in memory. This is the default.
21529 Generate code for a PDP-11/40.
21533 Generate code for a PDP-11/45. This is the default.
21537 Generate code for a PDP-11/10.
21539 @item -mbcopy-builtin
21540 @opindex mbcopy-builtin
21541 Use inline @code{movmemhi} patterns for copying memory. This is the
21546 Do not use inline @code{movmemhi} patterns for copying memory.
21552 Use 16-bit @code{int}. This is the default.
21558 Use 32-bit @code{int}.
21561 @itemx -mno-float32
21563 @opindex mno-float32
21564 Use 64-bit @code{float}. This is the default.
21567 @itemx -mno-float64
21569 @opindex mno-float64
21570 Use 32-bit @code{float}.
21574 Use @code{abshi2} pattern. This is the default.
21578 Do not use @code{abshi2} pattern.
21580 @item -mbranch-expensive
21581 @opindex mbranch-expensive
21582 Pretend that branches are expensive. This is for experimenting with
21583 code generation only.
21585 @item -mbranch-cheap
21586 @opindex mbranch-cheap
21587 Do not pretend that branches are expensive. This is the default.
21591 Use Unix assembler syntax. This is the default when configured for
21592 @samp{pdp11-*-bsd}.
21596 Use DEC assembler syntax. This is the default when configured for any
21597 PDP-11 target other than @samp{pdp11-*-bsd}.
21600 @node picoChip Options
21601 @subsection picoChip Options
21602 @cindex picoChip options
21604 These @samp{-m} options are defined for picoChip implementations:
21608 @item -mae=@var{ae_type}
21610 Set the instruction set, register set, and instruction scheduling
21611 parameters for array element type @var{ae_type}. Supported values
21612 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
21614 @option{-mae=ANY} selects a completely generic AE type. Code
21615 generated with this option runs on any of the other AE types. The
21616 code is not as efficient as it would be if compiled for a specific
21617 AE type, and some types of operation (e.g., multiplication) do not
21618 work properly on all types of AE.
21620 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
21621 for compiled code, and is the default.
21623 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
21624 option may suffer from poor performance of byte (char) manipulation,
21625 since the DSP AE does not provide hardware support for byte load/stores.
21627 @item -msymbol-as-address
21628 Enable the compiler to directly use a symbol name as an address in a
21629 load/store instruction, without first loading it into a
21630 register. Typically, the use of this option generates larger
21631 programs, which run faster than when the option isn't used. However, the
21632 results vary from program to program, so it is left as a user option,
21633 rather than being permanently enabled.
21635 @item -mno-inefficient-warnings
21636 Disables warnings about the generation of inefficient code. These
21637 warnings can be generated, for example, when compiling code that
21638 performs byte-level memory operations on the MAC AE type. The MAC AE has
21639 no hardware support for byte-level memory operations, so all byte
21640 load/stores must be synthesized from word load/store operations. This is
21641 inefficient and a warning is generated to indicate
21642 that you should rewrite the code to avoid byte operations, or to target
21643 an AE type that has the necessary hardware support. This option disables
21648 @node PowerPC Options
21649 @subsection PowerPC Options
21650 @cindex PowerPC options
21652 These are listed under @xref{RS/6000 and PowerPC Options}.
21654 @node RISC-V Options
21655 @subsection RISC-V Options
21656 @cindex RISC-V Options
21658 These command-line options are defined for RISC-V targets:
21661 @item -mbranch-cost=@var{n}
21662 @opindex mbranch-cost
21663 Set the cost of branches to roughly @var{n} instructions.
21668 Don't optimize block moves.
21673 When generating PIC code, allow the use of PLTs. Ignored for non-PIC.
21675 @item -mabi=@var{ABI-string}
21677 Specify integer and floating-point calling convention. This defaults to the
21678 natural calling convention: e.g.@ LP64 for RV64I, ILP32 for RV32I, LP64D for
21684 Use hardware floating-point divide and square root instructions. This requires
21685 the F or D extensions for floating-point registers.
21690 Use hardware instructions for integer division. This requires the M extension.
21692 @item -march=@var{ISA-string}
21694 Generate code for given RISC-V ISA (e.g.@ @samp{rv64im}). ISA strings must be
21695 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, and @samp{rv32imaf}.
21697 @item -mtune=@var{processor-string}
21699 Optimize the output for the given processor, specified by microarchitecture
21702 @item -msmall-data-limit=@var{n}
21703 @opindex msmall-data-limit
21704 Put global and static data smaller than @var{n} bytes into a special section
21707 @item -msave-restore
21708 @itemx -mno-save-restore
21709 @opindex msave-restore
21710 Use smaller but slower prologue and epilogue code.
21712 @item -mstrict-align
21713 @itemx -mno-strict-align
21714 @opindex mstrict-align
21715 Do not generate unaligned memory accesses.
21717 @item -mcmodel=@var{code-model}
21719 Specify the code model.
21724 @subsection RL78 Options
21725 @cindex RL78 Options
21731 Links in additional target libraries to support operation within a
21740 Specifies the type of hardware multiplication and division support to
21741 be used. The simplest is @code{none}, which uses software for both
21742 multiplication and division. This is the default. The @code{g13}
21743 value is for the hardware multiply/divide peripheral found on the
21744 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
21745 the multiplication and division instructions supported by the RL78/G14
21746 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
21747 the value @code{mg10} is an alias for @code{none}.
21749 In addition a C preprocessor macro is defined, based upon the setting
21750 of this option. Possible values are: @code{__RL78_MUL_NONE__},
21751 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
21758 Specifies the RL78 core to target. The default is the G14 core, also
21759 known as an S3 core or just RL78. The G13 or S2 core does not have
21760 multiply or divide instructions, instead it uses a hardware peripheral
21761 for these operations. The G10 or S1 core does not have register
21762 banks, so it uses a different calling convention.
21764 If this option is set it also selects the type of hardware multiply
21765 support to use, unless this is overridden by an explicit
21766 @option{-mmul=none} option on the command line. Thus specifying
21767 @option{-mcpu=g13} enables the use of the G13 hardware multiply
21768 peripheral and specifying @option{-mcpu=g10} disables the use of
21769 hardware multiplications altogether.
21771 Note, although the RL78/G14 core is the default target, specifying
21772 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
21773 change the behavior of the toolchain since it also enables G14
21774 hardware multiply support. If these options are not specified on the
21775 command line then software multiplication routines will be used even
21776 though the code targets the RL78 core. This is for backwards
21777 compatibility with older toolchains which did not have hardware
21778 multiply and divide support.
21780 In addition a C preprocessor macro is defined, based upon the setting
21781 of this option. Possible values are: @code{__RL78_G10__},
21782 @code{__RL78_G13__} or @code{__RL78_G14__}.
21792 These are aliases for the corresponding @option{-mcpu=} option. They
21793 are provided for backwards compatibility.
21797 Allow the compiler to use all of the available registers. By default
21798 registers @code{r24..r31} are reserved for use in interrupt handlers.
21799 With this option enabled these registers can be used in ordinary
21802 @item -m64bit-doubles
21803 @itemx -m32bit-doubles
21804 @opindex m64bit-doubles
21805 @opindex m32bit-doubles
21806 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21807 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21808 @option{-m32bit-doubles}.
21810 @item -msave-mduc-in-interrupts
21811 @item -mno-save-mduc-in-interrupts
21812 @opindex msave-mduc-in-interrupts
21813 @opindex mno-save-mduc-in-interrupts
21814 Specifies that interrupt handler functions should preserve the
21815 MDUC registers. This is only necessary if normal code might use
21816 the MDUC registers, for example because it performs multiplication
21817 and division operations. The default is to ignore the MDUC registers
21818 as this makes the interrupt handlers faster. The target option -mg13
21819 needs to be passed for this to work as this feature is only available
21820 on the G13 target (S2 core). The MDUC registers will only be saved
21821 if the interrupt handler performs a multiplication or division
21822 operation or it calls another function.
21826 @node RS/6000 and PowerPC Options
21827 @subsection IBM RS/6000 and PowerPC Options
21828 @cindex RS/6000 and PowerPC Options
21829 @cindex IBM RS/6000 and PowerPC Options
21831 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
21833 @item -mpowerpc-gpopt
21834 @itemx -mno-powerpc-gpopt
21835 @itemx -mpowerpc-gfxopt
21836 @itemx -mno-powerpc-gfxopt
21839 @itemx -mno-powerpc64
21843 @itemx -mno-popcntb
21845 @itemx -mno-popcntd
21854 @itemx -mno-hard-dfp
21855 @opindex mpowerpc-gpopt
21856 @opindex mno-powerpc-gpopt
21857 @opindex mpowerpc-gfxopt
21858 @opindex mno-powerpc-gfxopt
21859 @opindex mpowerpc64
21860 @opindex mno-powerpc64
21864 @opindex mno-popcntb
21866 @opindex mno-popcntd
21872 @opindex mno-mfpgpr
21874 @opindex mno-hard-dfp
21875 You use these options to specify which instructions are available on the
21876 processor you are using. The default value of these options is
21877 determined when configuring GCC@. Specifying the
21878 @option{-mcpu=@var{cpu_type}} overrides the specification of these
21879 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
21880 rather than the options listed above.
21882 Specifying @option{-mpowerpc-gpopt} allows
21883 GCC to use the optional PowerPC architecture instructions in the
21884 General Purpose group, including floating-point square root. Specifying
21885 @option{-mpowerpc-gfxopt} allows GCC to
21886 use the optional PowerPC architecture instructions in the Graphics
21887 group, including floating-point select.
21889 The @option{-mmfcrf} option allows GCC to generate the move from
21890 condition register field instruction implemented on the POWER4
21891 processor and other processors that support the PowerPC V2.01
21893 The @option{-mpopcntb} option allows GCC to generate the popcount and
21894 double-precision FP reciprocal estimate instruction implemented on the
21895 POWER5 processor and other processors that support the PowerPC V2.02
21897 The @option{-mpopcntd} option allows GCC to generate the popcount
21898 instruction implemented on the POWER7 processor and other processors
21899 that support the PowerPC V2.06 architecture.
21900 The @option{-mfprnd} option allows GCC to generate the FP round to
21901 integer instructions implemented on the POWER5+ processor and other
21902 processors that support the PowerPC V2.03 architecture.
21903 The @option{-mcmpb} option allows GCC to generate the compare bytes
21904 instruction implemented on the POWER6 processor and other processors
21905 that support the PowerPC V2.05 architecture.
21906 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
21907 general-purpose register instructions implemented on the POWER6X
21908 processor and other processors that support the extended PowerPC V2.05
21910 The @option{-mhard-dfp} option allows GCC to generate the decimal
21911 floating-point instructions implemented on some POWER processors.
21913 The @option{-mpowerpc64} option allows GCC to generate the additional
21914 64-bit instructions that are found in the full PowerPC64 architecture
21915 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
21916 @option{-mno-powerpc64}.
21918 @item -mcpu=@var{cpu_type}
21920 Set architecture type, register usage, and
21921 instruction scheduling parameters for machine type @var{cpu_type}.
21922 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
21923 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
21924 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
21925 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
21926 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
21927 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
21928 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
21929 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
21930 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
21931 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
21932 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
21935 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
21936 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
21937 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
21938 architecture machine types, with an appropriate, generic processor
21939 model assumed for scheduling purposes.
21941 The other options specify a specific processor. Code generated under
21942 those options runs best on that processor, and may not run at all on
21945 The @option{-mcpu} options automatically enable or disable the
21948 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
21949 -mpopcntb -mpopcntd -mpowerpc64 @gol
21950 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
21951 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
21952 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
21953 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
21955 The particular options set for any particular CPU varies between
21956 compiler versions, depending on what setting seems to produce optimal
21957 code for that CPU; it doesn't necessarily reflect the actual hardware's
21958 capabilities. If you wish to set an individual option to a particular
21959 value, you may specify it after the @option{-mcpu} option, like
21960 @option{-mcpu=970 -mno-altivec}.
21962 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
21963 not enabled or disabled by the @option{-mcpu} option at present because
21964 AIX does not have full support for these options. You may still
21965 enable or disable them individually if you're sure it'll work in your
21968 @item -mtune=@var{cpu_type}
21970 Set the instruction scheduling parameters for machine type
21971 @var{cpu_type}, but do not set the architecture type or register usage,
21972 as @option{-mcpu=@var{cpu_type}} does. The same
21973 values for @var{cpu_type} are used for @option{-mtune} as for
21974 @option{-mcpu}. If both are specified, the code generated uses the
21975 architecture and registers set by @option{-mcpu}, but the
21976 scheduling parameters set by @option{-mtune}.
21978 @item -mcmodel=small
21979 @opindex mcmodel=small
21980 Generate PowerPC64 code for the small model: The TOC is limited to
21983 @item -mcmodel=medium
21984 @opindex mcmodel=medium
21985 Generate PowerPC64 code for the medium model: The TOC and other static
21986 data may be up to a total of 4G in size. This is the default for 64-bit
21989 @item -mcmodel=large
21990 @opindex mcmodel=large
21991 Generate PowerPC64 code for the large model: The TOC may be up to 4G
21992 in size. Other data and code is only limited by the 64-bit address
21996 @itemx -mno-altivec
21998 @opindex mno-altivec
21999 Generate code that uses (does not use) AltiVec instructions, and also
22000 enable the use of built-in functions that allow more direct access to
22001 the AltiVec instruction set. You may also need to set
22002 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
22005 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
22006 @option{-maltivec=be}, the element order for AltiVec intrinsics such
22007 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
22008 match array element order corresponding to the endianness of the
22009 target. That is, element zero identifies the leftmost element in a
22010 vector register when targeting a big-endian platform, and identifies
22011 the rightmost element in a vector register when targeting a
22012 little-endian platform.
22015 @opindex maltivec=be
22016 Generate AltiVec instructions using big-endian element order,
22017 regardless of whether the target is big- or little-endian. This is
22018 the default when targeting a big-endian platform.
22020 The element order is used to interpret element numbers in AltiVec
22021 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22022 @code{vec_insert}. By default, these match array element order
22023 corresponding to the endianness for the target.
22026 @opindex maltivec=le
22027 Generate AltiVec instructions using little-endian element order,
22028 regardless of whether the target is big- or little-endian. This is
22029 the default when targeting a little-endian platform. This option is
22030 currently ignored when targeting a big-endian platform.
22032 The element order is used to interpret element numbers in AltiVec
22033 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22034 @code{vec_insert}. By default, these match array element order
22035 corresponding to the endianness for the target.
22040 @opindex mno-vrsave
22041 Generate VRSAVE instructions when generating AltiVec code.
22044 @opindex msecure-plt
22045 Generate code that allows @command{ld} and @command{ld.so}
22046 to build executables and shared
22047 libraries with non-executable @code{.plt} and @code{.got} sections.
22049 32-bit SYSV ABI option.
22053 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
22055 requires @code{.plt} and @code{.got}
22056 sections that are both writable and executable.
22057 This is a PowerPC 32-bit SYSV ABI option.
22063 This switch enables or disables the generation of ISEL instructions.
22065 @item -misel=@var{yes/no}
22066 This switch has been deprecated. Use @option{-misel} and
22067 @option{-mno-isel} instead.
22073 This switch enables or disables the generation of SPE simd
22079 @opindex mno-paired
22080 This switch enables or disables the generation of PAIRED simd
22083 @item -mspe=@var{yes/no}
22084 This option has been deprecated. Use @option{-mspe} and
22085 @option{-mno-spe} instead.
22091 Generate code that uses (does not use) vector/scalar (VSX)
22092 instructions, and also enable the use of built-in functions that allow
22093 more direct access to the VSX instruction set.
22098 @opindex mno-crypto
22099 Enable the use (disable) of the built-in functions that allow direct
22100 access to the cryptographic instructions that were added in version
22101 2.07 of the PowerPC ISA.
22103 @item -mdirect-move
22104 @itemx -mno-direct-move
22105 @opindex mdirect-move
22106 @opindex mno-direct-move
22107 Generate code that uses (does not use) the instructions to move data
22108 between the general purpose registers and the vector/scalar (VSX)
22109 registers that were added in version 2.07 of the PowerPC ISA.
22115 Enable (disable) the use of the built-in functions that allow direct
22116 access to the Hardware Transactional Memory (HTM) instructions that
22117 were added in version 2.07 of the PowerPC ISA.
22119 @item -mpower8-fusion
22120 @itemx -mno-power8-fusion
22121 @opindex mpower8-fusion
22122 @opindex mno-power8-fusion
22123 Generate code that keeps (does not keeps) some integer operations
22124 adjacent so that the instructions can be fused together on power8 and
22127 @item -mpower8-vector
22128 @itemx -mno-power8-vector
22129 @opindex mpower8-vector
22130 @opindex mno-power8-vector
22131 Generate code that uses (does not use) the vector and scalar
22132 instructions that were added in version 2.07 of the PowerPC ISA. Also
22133 enable the use of built-in functions that allow more direct access to
22134 the vector instructions.
22136 @item -mquad-memory
22137 @itemx -mno-quad-memory
22138 @opindex mquad-memory
22139 @opindex mno-quad-memory
22140 Generate code that uses (does not use) the non-atomic quad word memory
22141 instructions. The @option{-mquad-memory} option requires use of
22144 @item -mquad-memory-atomic
22145 @itemx -mno-quad-memory-atomic
22146 @opindex mquad-memory-atomic
22147 @opindex mno-quad-memory-atomic
22148 Generate code that uses (does not use) the atomic quad word memory
22149 instructions. The @option{-mquad-memory-atomic} option requires use of
22153 @itemx -mno-float128
22155 @opindex mno-float128
22156 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
22157 and use either software emulation for IEEE 128-bit floating point or
22158 hardware instructions.
22160 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
22161 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
22162 use the IEEE 128-bit floating point support. The IEEE 128-bit
22163 floating point support only works on PowerPC Linux systems.
22165 The default for @option{-mfloat128} is enabled on PowerPC Linux
22166 systems using the VSX instruction set, and disabled on other systems.
22168 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
22169 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
22170 point support will also enable the generation of ISA 3.0 IEEE 128-bit
22171 floating point instructions. Otherwise, if you do not specify to
22172 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
22173 system, IEEE 128-bit floating point will be done with software
22176 @item -mfloat128-hardware
22177 @itemx -mno-float128-hardware
22178 @opindex mfloat128-hardware
22179 @opindex mno-float128-hardware
22180 Enable/disable using ISA 3.0 hardware instructions to support the
22181 @var{__float128} data type.
22183 The default for @option{-mfloat128-hardware} is enabled on PowerPC
22184 Linux systems using the ISA 3.0 instruction set, and disabled on other
22187 @item -mfloat-gprs=@var{yes/single/double/no}
22188 @itemx -mfloat-gprs
22189 @opindex mfloat-gprs
22190 This switch enables or disables the generation of floating-point
22191 operations on the general-purpose registers for architectures that
22194 The argument @samp{yes} or @samp{single} enables the use of
22195 single-precision floating-point operations.
22197 The argument @samp{double} enables the use of single and
22198 double-precision floating-point operations.
22200 The argument @samp{no} disables floating-point operations on the
22201 general-purpose registers.
22203 This option is currently only available on the MPC854x.
22209 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
22210 targets (including GNU/Linux). The 32-bit environment sets int, long
22211 and pointer to 32 bits and generates code that runs on any PowerPC
22212 variant. The 64-bit environment sets int to 32 bits and long and
22213 pointer to 64 bits, and generates code for PowerPC64, as for
22214 @option{-mpowerpc64}.
22217 @itemx -mno-fp-in-toc
22218 @itemx -mno-sum-in-toc
22219 @itemx -mminimal-toc
22221 @opindex mno-fp-in-toc
22222 @opindex mno-sum-in-toc
22223 @opindex mminimal-toc
22224 Modify generation of the TOC (Table Of Contents), which is created for
22225 every executable file. The @option{-mfull-toc} option is selected by
22226 default. In that case, GCC allocates at least one TOC entry for
22227 each unique non-automatic variable reference in your program. GCC
22228 also places floating-point constants in the TOC@. However, only
22229 16,384 entries are available in the TOC@.
22231 If you receive a linker error message that saying you have overflowed
22232 the available TOC space, you can reduce the amount of TOC space used
22233 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
22234 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
22235 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
22236 generate code to calculate the sum of an address and a constant at
22237 run time instead of putting that sum into the TOC@. You may specify one
22238 or both of these options. Each causes GCC to produce very slightly
22239 slower and larger code at the expense of conserving TOC space.
22241 If you still run out of space in the TOC even when you specify both of
22242 these options, specify @option{-mminimal-toc} instead. This option causes
22243 GCC to make only one TOC entry for every file. When you specify this
22244 option, GCC produces code that is slower and larger but which
22245 uses extremely little TOC space. You may wish to use this option
22246 only on files that contain less frequently-executed code.
22252 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
22253 @code{long} type, and the infrastructure needed to support them.
22254 Specifying @option{-maix64} implies @option{-mpowerpc64},
22255 while @option{-maix32} disables the 64-bit ABI and
22256 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
22259 @itemx -mno-xl-compat
22260 @opindex mxl-compat
22261 @opindex mno-xl-compat
22262 Produce code that conforms more closely to IBM XL compiler semantics
22263 when using AIX-compatible ABI@. Pass floating-point arguments to
22264 prototyped functions beyond the register save area (RSA) on the stack
22265 in addition to argument FPRs. Do not assume that most significant
22266 double in 128-bit long double value is properly rounded when comparing
22267 values and converting to double. Use XL symbol names for long double
22270 The AIX calling convention was extended but not initially documented to
22271 handle an obscure K&R C case of calling a function that takes the
22272 address of its arguments with fewer arguments than declared. IBM XL
22273 compilers access floating-point arguments that do not fit in the
22274 RSA from the stack when a subroutine is compiled without
22275 optimization. Because always storing floating-point arguments on the
22276 stack is inefficient and rarely needed, this option is not enabled by
22277 default and only is necessary when calling subroutines compiled by IBM
22278 XL compilers without optimization.
22282 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
22283 application written to use message passing with special startup code to
22284 enable the application to run. The system must have PE installed in the
22285 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
22286 must be overridden with the @option{-specs=} option to specify the
22287 appropriate directory location. The Parallel Environment does not
22288 support threads, so the @option{-mpe} option and the @option{-pthread}
22289 option are incompatible.
22291 @item -malign-natural
22292 @itemx -malign-power
22293 @opindex malign-natural
22294 @opindex malign-power
22295 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
22296 @option{-malign-natural} overrides the ABI-defined alignment of larger
22297 types, such as floating-point doubles, on their natural size-based boundary.
22298 The option @option{-malign-power} instructs GCC to follow the ABI-specified
22299 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
22301 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
22305 @itemx -mhard-float
22306 @opindex msoft-float
22307 @opindex mhard-float
22308 Generate code that does not use (uses) the floating-point register set.
22309 Software floating-point emulation is provided if you use the
22310 @option{-msoft-float} option, and pass the option to GCC when linking.
22312 @item -msingle-float
22313 @itemx -mdouble-float
22314 @opindex msingle-float
22315 @opindex mdouble-float
22316 Generate code for single- or double-precision floating-point operations.
22317 @option{-mdouble-float} implies @option{-msingle-float}.
22320 @opindex msimple-fpu
22321 Do not generate @code{sqrt} and @code{div} instructions for hardware
22322 floating-point unit.
22324 @item -mfpu=@var{name}
22326 Specify type of floating-point unit. Valid values for @var{name} are
22327 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
22328 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
22329 @samp{sp_full} (equivalent to @option{-msingle-float}),
22330 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
22333 @opindex mxilinx-fpu
22334 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
22337 @itemx -mno-multiple
22339 @opindex mno-multiple
22340 Generate code that uses (does not use) the load multiple word
22341 instructions and the store multiple word instructions. These
22342 instructions are generated by default on POWER systems, and not
22343 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
22344 PowerPC systems, since those instructions do not work when the
22345 processor is in little-endian mode. The exceptions are PPC740 and
22346 PPC750 which permit these instructions in little-endian mode.
22351 @opindex mno-string
22352 Generate code that uses (does not use) the load string instructions
22353 and the store string word instructions to save multiple registers and
22354 do small block moves. These instructions are generated by default on
22355 POWER systems, and not generated on PowerPC systems. Do not use
22356 @option{-mstring} on little-endian PowerPC systems, since those
22357 instructions do not work when the processor is in little-endian mode.
22358 The exceptions are PPC740 and PPC750 which permit these instructions
22359 in little-endian mode.
22364 @opindex mno-update
22365 Generate code that uses (does not use) the load or store instructions
22366 that update the base register to the address of the calculated memory
22367 location. These instructions are generated by default. If you use
22368 @option{-mno-update}, there is a small window between the time that the
22369 stack pointer is updated and the address of the previous frame is
22370 stored, which means code that walks the stack frame across interrupts or
22371 signals may get corrupted data.
22373 @item -mavoid-indexed-addresses
22374 @itemx -mno-avoid-indexed-addresses
22375 @opindex mavoid-indexed-addresses
22376 @opindex mno-avoid-indexed-addresses
22377 Generate code that tries to avoid (not avoid) the use of indexed load
22378 or store instructions. These instructions can incur a performance
22379 penalty on Power6 processors in certain situations, such as when
22380 stepping through large arrays that cross a 16M boundary. This option
22381 is enabled by default when targeting Power6 and disabled otherwise.
22384 @itemx -mno-fused-madd
22385 @opindex mfused-madd
22386 @opindex mno-fused-madd
22387 Generate code that uses (does not use) the floating-point multiply and
22388 accumulate instructions. These instructions are generated by default
22389 if hardware floating point is used. The machine-dependent
22390 @option{-mfused-madd} option is now mapped to the machine-independent
22391 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22392 mapped to @option{-ffp-contract=off}.
22398 Generate code that uses (does not use) the half-word multiply and
22399 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
22400 These instructions are generated by default when targeting those
22407 Generate code that uses (does not use) the string-search @samp{dlmzb}
22408 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
22409 generated by default when targeting those processors.
22411 @item -mno-bit-align
22413 @opindex mno-bit-align
22414 @opindex mbit-align
22415 On System V.4 and embedded PowerPC systems do not (do) force structures
22416 and unions that contain bit-fields to be aligned to the base type of the
22419 For example, by default a structure containing nothing but 8
22420 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
22421 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
22422 the structure is aligned to a 1-byte boundary and is 1 byte in
22425 @item -mno-strict-align
22426 @itemx -mstrict-align
22427 @opindex mno-strict-align
22428 @opindex mstrict-align
22429 On System V.4 and embedded PowerPC systems do not (do) assume that
22430 unaligned memory references are handled by the system.
22432 @item -mrelocatable
22433 @itemx -mno-relocatable
22434 @opindex mrelocatable
22435 @opindex mno-relocatable
22436 Generate code that allows (does not allow) a static executable to be
22437 relocated to a different address at run time. A simple embedded
22438 PowerPC system loader should relocate the entire contents of
22439 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
22440 a table of 32-bit addresses generated by this option. For this to
22441 work, all objects linked together must be compiled with
22442 @option{-mrelocatable} or @option{-mrelocatable-lib}.
22443 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
22445 @item -mrelocatable-lib
22446 @itemx -mno-relocatable-lib
22447 @opindex mrelocatable-lib
22448 @opindex mno-relocatable-lib
22449 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
22450 @code{.fixup} section to allow static executables to be relocated at
22451 run time, but @option{-mrelocatable-lib} does not use the smaller stack
22452 alignment of @option{-mrelocatable}. Objects compiled with
22453 @option{-mrelocatable-lib} may be linked with objects compiled with
22454 any combination of the @option{-mrelocatable} options.
22460 On System V.4 and embedded PowerPC systems do not (do) assume that
22461 register 2 contains a pointer to a global area pointing to the addresses
22462 used in the program.
22465 @itemx -mlittle-endian
22467 @opindex mlittle-endian
22468 On System V.4 and embedded PowerPC systems compile code for the
22469 processor in little-endian mode. The @option{-mlittle-endian} option is
22470 the same as @option{-mlittle}.
22473 @itemx -mbig-endian
22475 @opindex mbig-endian
22476 On System V.4 and embedded PowerPC systems compile code for the
22477 processor in big-endian mode. The @option{-mbig-endian} option is
22478 the same as @option{-mbig}.
22480 @item -mdynamic-no-pic
22481 @opindex mdynamic-no-pic
22482 On Darwin and Mac OS X systems, compile code so that it is not
22483 relocatable, but that its external references are relocatable. The
22484 resulting code is suitable for applications, but not shared
22487 @item -msingle-pic-base
22488 @opindex msingle-pic-base
22489 Treat the register used for PIC addressing as read-only, rather than
22490 loading it in the prologue for each function. The runtime system is
22491 responsible for initializing this register with an appropriate value
22492 before execution begins.
22494 @item -mprioritize-restricted-insns=@var{priority}
22495 @opindex mprioritize-restricted-insns
22496 This option controls the priority that is assigned to
22497 dispatch-slot restricted instructions during the second scheduling
22498 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
22499 or @samp{2} to assign no, highest, or second-highest (respectively)
22500 priority to dispatch-slot restricted
22503 @item -msched-costly-dep=@var{dependence_type}
22504 @opindex msched-costly-dep
22505 This option controls which dependences are considered costly
22506 by the target during instruction scheduling. The argument
22507 @var{dependence_type} takes one of the following values:
22511 No dependence is costly.
22514 All dependences are costly.
22516 @item @samp{true_store_to_load}
22517 A true dependence from store to load is costly.
22519 @item @samp{store_to_load}
22520 Any dependence from store to load is costly.
22523 Any dependence for which the latency is greater than or equal to
22524 @var{number} is costly.
22527 @item -minsert-sched-nops=@var{scheme}
22528 @opindex minsert-sched-nops
22529 This option controls which NOP insertion scheme is used during
22530 the second scheduling pass. The argument @var{scheme} takes one of the
22538 Pad with NOPs any dispatch group that has vacant issue slots,
22539 according to the scheduler's grouping.
22541 @item @samp{regroup_exact}
22542 Insert NOPs to force costly dependent insns into
22543 separate groups. Insert exactly as many NOPs as needed to force an insn
22544 to a new group, according to the estimated processor grouping.
22547 Insert NOPs to force costly dependent insns into
22548 separate groups. Insert @var{number} NOPs to force an insn to a new group.
22552 @opindex mcall-sysv
22553 On System V.4 and embedded PowerPC systems compile code using calling
22554 conventions that adhere to the March 1995 draft of the System V
22555 Application Binary Interface, PowerPC processor supplement. This is the
22556 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
22558 @item -mcall-sysv-eabi
22560 @opindex mcall-sysv-eabi
22561 @opindex mcall-eabi
22562 Specify both @option{-mcall-sysv} and @option{-meabi} options.
22564 @item -mcall-sysv-noeabi
22565 @opindex mcall-sysv-noeabi
22566 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
22568 @item -mcall-aixdesc
22570 On System V.4 and embedded PowerPC systems compile code for the AIX
22574 @opindex mcall-linux
22575 On System V.4 and embedded PowerPC systems compile code for the
22576 Linux-based GNU system.
22578 @item -mcall-freebsd
22579 @opindex mcall-freebsd
22580 On System V.4 and embedded PowerPC systems compile code for the
22581 FreeBSD operating system.
22583 @item -mcall-netbsd
22584 @opindex mcall-netbsd
22585 On System V.4 and embedded PowerPC systems compile code for the
22586 NetBSD operating system.
22588 @item -mcall-openbsd
22589 @opindex mcall-netbsd
22590 On System V.4 and embedded PowerPC systems compile code for the
22591 OpenBSD operating system.
22593 @item -maix-struct-return
22594 @opindex maix-struct-return
22595 Return all structures in memory (as specified by the AIX ABI)@.
22597 @item -msvr4-struct-return
22598 @opindex msvr4-struct-return
22599 Return structures smaller than 8 bytes in registers (as specified by the
22602 @item -mabi=@var{abi-type}
22604 Extend the current ABI with a particular extension, or remove such extension.
22605 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
22606 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
22607 @samp{elfv1}, @samp{elfv2}@.
22611 Extend the current ABI with SPE ABI extensions. This does not change
22612 the default ABI, instead it adds the SPE ABI extensions to the current
22616 @opindex mabi=no-spe
22617 Disable Book-E SPE ABI extensions for the current ABI@.
22619 @item -mabi=ibmlongdouble
22620 @opindex mabi=ibmlongdouble
22621 Change the current ABI to use IBM extended-precision long double.
22622 This is not likely to work if your system defaults to using IEEE
22623 extended-precision long double. If you change the long double type
22624 from IEEE extended-precision, the compiler will issue a warning unless
22625 you use the @option{-Wno-psabi} option.
22627 @item -mabi=ieeelongdouble
22628 @opindex mabi=ieeelongdouble
22629 Change the current ABI to use IEEE extended-precision long double.
22630 This is not likely to work if your system defaults to using IBM
22631 extended-precision long double. If you change the long double type
22632 from IBM extended-precision, the compiler will issue a warning unless
22633 you use the @option{-Wno-psabi} option.
22636 @opindex mabi=elfv1
22637 Change the current ABI to use the ELFv1 ABI.
22638 This is the default ABI for big-endian PowerPC 64-bit Linux.
22639 Overriding the default ABI requires special system support and is
22640 likely to fail in spectacular ways.
22643 @opindex mabi=elfv2
22644 Change the current ABI to use the ELFv2 ABI.
22645 This is the default ABI for little-endian PowerPC 64-bit Linux.
22646 Overriding the default ABI requires special system support and is
22647 likely to fail in spectacular ways.
22649 @item -mgnu-attribute
22650 @itemx -mno-gnu-attribute
22651 @opindex mgnu-attribute
22652 @opindex mno-gnu-attribute
22653 Emit .gnu_attribute assembly directives to set tag/value pairs in a
22654 .gnu.attributes section that specify ABI variations in function
22655 parameters or return values.
22658 @itemx -mno-prototype
22659 @opindex mprototype
22660 @opindex mno-prototype
22661 On System V.4 and embedded PowerPC systems assume that all calls to
22662 variable argument functions are properly prototyped. Otherwise, the
22663 compiler must insert an instruction before every non-prototyped call to
22664 set or clear bit 6 of the condition code register (@code{CR}) to
22665 indicate whether floating-point values are passed in the floating-point
22666 registers in case the function takes variable arguments. With
22667 @option{-mprototype}, only calls to prototyped variable argument functions
22668 set or clear the bit.
22672 On embedded PowerPC systems, assume that the startup module is called
22673 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
22674 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
22679 On embedded PowerPC systems, assume that the startup module is called
22680 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
22685 On embedded PowerPC systems, assume that the startup module is called
22686 @file{crt0.o} and the standard C libraries are @file{libads.a} and
22689 @item -myellowknife
22690 @opindex myellowknife
22691 On embedded PowerPC systems, assume that the startup module is called
22692 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
22697 On System V.4 and embedded PowerPC systems, specify that you are
22698 compiling for a VxWorks system.
22702 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
22703 header to indicate that @samp{eabi} extended relocations are used.
22709 On System V.4 and embedded PowerPC systems do (do not) adhere to the
22710 Embedded Applications Binary Interface (EABI), which is a set of
22711 modifications to the System V.4 specifications. Selecting @option{-meabi}
22712 means that the stack is aligned to an 8-byte boundary, a function
22713 @code{__eabi} is called from @code{main} to set up the EABI
22714 environment, and the @option{-msdata} option can use both @code{r2} and
22715 @code{r13} to point to two separate small data areas. Selecting
22716 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
22717 no EABI initialization function is called from @code{main}, and the
22718 @option{-msdata} option only uses @code{r13} to point to a single
22719 small data area. The @option{-meabi} option is on by default if you
22720 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
22723 @opindex msdata=eabi
22724 On System V.4 and embedded PowerPC systems, put small initialized
22725 @code{const} global and static data in the @code{.sdata2} section, which
22726 is pointed to by register @code{r2}. Put small initialized
22727 non-@code{const} global and static data in the @code{.sdata} section,
22728 which is pointed to by register @code{r13}. Put small uninitialized
22729 global and static data in the @code{.sbss} section, which is adjacent to
22730 the @code{.sdata} section. The @option{-msdata=eabi} option is
22731 incompatible with the @option{-mrelocatable} option. The
22732 @option{-msdata=eabi} option also sets the @option{-memb} option.
22735 @opindex msdata=sysv
22736 On System V.4 and embedded PowerPC systems, put small global and static
22737 data in the @code{.sdata} section, which is pointed to by register
22738 @code{r13}. Put small uninitialized global and static data in the
22739 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
22740 The @option{-msdata=sysv} option is incompatible with the
22741 @option{-mrelocatable} option.
22743 @item -msdata=default
22745 @opindex msdata=default
22747 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
22748 compile code the same as @option{-msdata=eabi}, otherwise compile code the
22749 same as @option{-msdata=sysv}.
22752 @opindex msdata=data
22753 On System V.4 and embedded PowerPC systems, put small global
22754 data in the @code{.sdata} section. Put small uninitialized global
22755 data in the @code{.sbss} section. Do not use register @code{r13}
22756 to address small data however. This is the default behavior unless
22757 other @option{-msdata} options are used.
22761 @opindex msdata=none
22763 On embedded PowerPC systems, put all initialized global and static data
22764 in the @code{.data} section, and all uninitialized data in the
22765 @code{.bss} section.
22767 @item -mblock-move-inline-limit=@var{num}
22768 @opindex mblock-move-inline-limit
22769 Inline all block moves (such as calls to @code{memcpy} or structure
22770 copies) less than or equal to @var{num} bytes. The minimum value for
22771 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
22772 targets. The default value is target-specific.
22776 @cindex smaller data references (PowerPC)
22777 @cindex .sdata/.sdata2 references (PowerPC)
22778 On embedded PowerPC systems, put global and static items less than or
22779 equal to @var{num} bytes into the small data or BSS sections instead of
22780 the normal data or BSS section. By default, @var{num} is 8. The
22781 @option{-G @var{num}} switch is also passed to the linker.
22782 All modules should be compiled with the same @option{-G @var{num}} value.
22785 @itemx -mno-regnames
22787 @opindex mno-regnames
22788 On System V.4 and embedded PowerPC systems do (do not) emit register
22789 names in the assembly language output using symbolic forms.
22792 @itemx -mno-longcall
22794 @opindex mno-longcall
22795 By default assume that all calls are far away so that a longer and more
22796 expensive calling sequence is required. This is required for calls
22797 farther than 32 megabytes (33,554,432 bytes) from the current location.
22798 A short call is generated if the compiler knows
22799 the call cannot be that far away. This setting can be overridden by
22800 the @code{shortcall} function attribute, or by @code{#pragma
22803 Some linkers are capable of detecting out-of-range calls and generating
22804 glue code on the fly. On these systems, long calls are unnecessary and
22805 generate slower code. As of this writing, the AIX linker can do this,
22806 as can the GNU linker for PowerPC/64. It is planned to add this feature
22807 to the GNU linker for 32-bit PowerPC systems as well.
22809 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
22810 callee, L42}, plus a @dfn{branch island} (glue code). The two target
22811 addresses represent the callee and the branch island. The
22812 Darwin/PPC linker prefers the first address and generates a @code{bl
22813 callee} if the PPC @code{bl} instruction reaches the callee directly;
22814 otherwise, the linker generates @code{bl L42} to call the branch
22815 island. The branch island is appended to the body of the
22816 calling function; it computes the full 32-bit address of the callee
22819 On Mach-O (Darwin) systems, this option directs the compiler emit to
22820 the glue for every direct call, and the Darwin linker decides whether
22821 to use or discard it.
22823 In the future, GCC may ignore all longcall specifications
22824 when the linker is known to generate glue.
22826 @item -mtls-markers
22827 @itemx -mno-tls-markers
22828 @opindex mtls-markers
22829 @opindex mno-tls-markers
22830 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22831 specifying the function argument. The relocation allows the linker to
22832 reliably associate function call with argument setup instructions for
22833 TLS optimization, which in turn allows GCC to better schedule the
22839 This option enables use of the reciprocal estimate and
22840 reciprocal square root estimate instructions with additional
22841 Newton-Raphson steps to increase precision instead of doing a divide or
22842 square root and divide for floating-point arguments. You should use
22843 the @option{-ffast-math} option when using @option{-mrecip} (or at
22844 least @option{-funsafe-math-optimizations},
22845 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22846 @option{-fno-trapping-math}). Note that while the throughput of the
22847 sequence is generally higher than the throughput of the non-reciprocal
22848 instruction, the precision of the sequence can be decreased by up to 2
22849 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
22852 @item -mrecip=@var{opt}
22853 @opindex mrecip=opt
22854 This option controls which reciprocal estimate instructions
22855 may be used. @var{opt} is a comma-separated list of options, which may
22856 be preceded by a @code{!} to invert the option:
22861 Enable all estimate instructions.
22864 Enable the default instructions, equivalent to @option{-mrecip}.
22867 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22870 Enable the reciprocal approximation instructions for both
22871 single and double precision.
22874 Enable the single-precision reciprocal approximation instructions.
22877 Enable the double-precision reciprocal approximation instructions.
22880 Enable the reciprocal square root approximation instructions for both
22881 single and double precision.
22884 Enable the single-precision reciprocal square root approximation instructions.
22887 Enable the double-precision reciprocal square root approximation instructions.
22891 So, for example, @option{-mrecip=all,!rsqrtd} enables
22892 all of the reciprocal estimate instructions, except for the
22893 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
22894 which handle the double-precision reciprocal square root calculations.
22896 @item -mrecip-precision
22897 @itemx -mno-recip-precision
22898 @opindex mrecip-precision
22899 Assume (do not assume) that the reciprocal estimate instructions
22900 provide higher-precision estimates than is mandated by the PowerPC
22901 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
22902 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
22903 The double-precision square root estimate instructions are not generated by
22904 default on low-precision machines, since they do not provide an
22905 estimate that converges after three steps.
22907 @item -mveclibabi=@var{type}
22908 @opindex mveclibabi
22909 Specifies the ABI type to use for vectorizing intrinsics using an
22910 external library. The only type supported at present is @samp{mass},
22911 which specifies to use IBM's Mathematical Acceleration Subsystem
22912 (MASS) libraries for vectorizing intrinsics using external libraries.
22913 GCC currently emits calls to @code{acosd2}, @code{acosf4},
22914 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
22915 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
22916 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
22917 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
22918 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
22919 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
22920 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
22921 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
22922 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
22923 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
22924 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
22925 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
22926 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
22927 for power7. Both @option{-ftree-vectorize} and
22928 @option{-funsafe-math-optimizations} must also be enabled. The MASS
22929 libraries must be specified at link time.
22934 Generate (do not generate) the @code{friz} instruction when the
22935 @option{-funsafe-math-optimizations} option is used to optimize
22936 rounding of floating-point values to 64-bit integer and back to floating
22937 point. The @code{friz} instruction does not return the same value if
22938 the floating-point number is too large to fit in an integer.
22940 @item -mpointers-to-nested-functions
22941 @itemx -mno-pointers-to-nested-functions
22942 @opindex mpointers-to-nested-functions
22943 Generate (do not generate) code to load up the static chain register
22944 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
22945 systems where a function pointer points to a 3-word descriptor giving
22946 the function address, TOC value to be loaded in register @code{r2}, and
22947 static chain value to be loaded in register @code{r11}. The
22948 @option{-mpointers-to-nested-functions} is on by default. You cannot
22949 call through pointers to nested functions or pointers
22950 to functions compiled in other languages that use the static chain if
22951 you use @option{-mno-pointers-to-nested-functions}.
22953 @item -msave-toc-indirect
22954 @itemx -mno-save-toc-indirect
22955 @opindex msave-toc-indirect
22956 Generate (do not generate) code to save the TOC value in the reserved
22957 stack location in the function prologue if the function calls through
22958 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
22959 saved in the prologue, it is saved just before the call through the
22960 pointer. The @option{-mno-save-toc-indirect} option is the default.
22962 @item -mcompat-align-parm
22963 @itemx -mno-compat-align-parm
22964 @opindex mcompat-align-parm
22965 Generate (do not generate) code to pass structure parameters with a
22966 maximum alignment of 64 bits, for compatibility with older versions
22969 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
22970 structure parameter on a 128-bit boundary when that structure contained
22971 a member requiring 128-bit alignment. This is corrected in more
22972 recent versions of GCC. This option may be used to generate code
22973 that is compatible with functions compiled with older versions of
22976 The @option{-mno-compat-align-parm} option is the default.
22978 @item -mstack-protector-guard=@var{guard}
22979 @itemx -mstack-protector-guard-reg=@var{reg}
22980 @itemx -mstack-protector-guard-offset=@var{offset}
22981 @itemx -mstack-protector-guard-symbol=@var{symbol}
22982 @opindex mstack-protector-guard
22983 @opindex mstack-protector-guard-reg
22984 @opindex mstack-protector-guard-offset
22985 @opindex mstack-protector-guard-symbol
22986 Generate stack protection code using canary at @var{guard}. Supported
22987 locations are @samp{global} for global canary or @samp{tls} for per-thread
22988 canary in the TLS block (the default with GNU libc version 2.4 or later).
22990 With the latter choice the options
22991 @option{-mstack-protector-guard-reg=@var{reg}} and
22992 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
22993 which register to use as base register for reading the canary, and from what
22994 offset from that base register. The default for those is as specified in the
22995 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
22996 the offset with a symbol reference to a canary in the TLS block.
23000 @subsection RX Options
23003 These command-line options are defined for RX targets:
23006 @item -m64bit-doubles
23007 @itemx -m32bit-doubles
23008 @opindex m64bit-doubles
23009 @opindex m32bit-doubles
23010 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23011 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23012 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
23013 works on 32-bit values, which is why the default is
23014 @option{-m32bit-doubles}.
23020 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
23021 floating-point hardware. The default is enabled for the RX600
23022 series and disabled for the RX200 series.
23024 Floating-point instructions are only generated for 32-bit floating-point
23025 values, however, so the FPU hardware is not used for doubles if the
23026 @option{-m64bit-doubles} option is used.
23028 @emph{Note} If the @option{-fpu} option is enabled then
23029 @option{-funsafe-math-optimizations} is also enabled automatically.
23030 This is because the RX FPU instructions are themselves unsafe.
23032 @item -mcpu=@var{name}
23034 Selects the type of RX CPU to be targeted. Currently three types are
23035 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
23036 the specific @samp{RX610} CPU. The default is @samp{RX600}.
23038 The only difference between @samp{RX600} and @samp{RX610} is that the
23039 @samp{RX610} does not support the @code{MVTIPL} instruction.
23041 The @samp{RX200} series does not have a hardware floating-point unit
23042 and so @option{-nofpu} is enabled by default when this type is
23045 @item -mbig-endian-data
23046 @itemx -mlittle-endian-data
23047 @opindex mbig-endian-data
23048 @opindex mlittle-endian-data
23049 Store data (but not code) in the big-endian format. The default is
23050 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
23053 @item -msmall-data-limit=@var{N}
23054 @opindex msmall-data-limit
23055 Specifies the maximum size in bytes of global and static variables
23056 which can be placed into the small data area. Using the small data
23057 area can lead to smaller and faster code, but the size of area is
23058 limited and it is up to the programmer to ensure that the area does
23059 not overflow. Also when the small data area is used one of the RX's
23060 registers (usually @code{r13}) is reserved for use pointing to this
23061 area, so it is no longer available for use by the compiler. This
23062 could result in slower and/or larger code if variables are pushed onto
23063 the stack instead of being held in this register.
23065 Note, common variables (variables that have not been initialized) and
23066 constants are not placed into the small data area as they are assigned
23067 to other sections in the output executable.
23069 The default value is zero, which disables this feature. Note, this
23070 feature is not enabled by default with higher optimization levels
23071 (@option{-O2} etc) because of the potentially detrimental effects of
23072 reserving a register. It is up to the programmer to experiment and
23073 discover whether this feature is of benefit to their program. See the
23074 description of the @option{-mpid} option for a description of how the
23075 actual register to hold the small data area pointer is chosen.
23081 Use the simulator runtime. The default is to use the libgloss
23082 board-specific runtime.
23084 @item -mas100-syntax
23085 @itemx -mno-as100-syntax
23086 @opindex mas100-syntax
23087 @opindex mno-as100-syntax
23088 When generating assembler output use a syntax that is compatible with
23089 Renesas's AS100 assembler. This syntax can also be handled by the GAS
23090 assembler, but it has some restrictions so it is not generated by default.
23092 @item -mmax-constant-size=@var{N}
23093 @opindex mmax-constant-size
23094 Specifies the maximum size, in bytes, of a constant that can be used as
23095 an operand in a RX instruction. Although the RX instruction set does
23096 allow constants of up to 4 bytes in length to be used in instructions,
23097 a longer value equates to a longer instruction. Thus in some
23098 circumstances it can be beneficial to restrict the size of constants
23099 that are used in instructions. Constants that are too big are instead
23100 placed into a constant pool and referenced via register indirection.
23102 The value @var{N} can be between 0 and 4. A value of 0 (the default)
23103 or 4 means that constants of any size are allowed.
23107 Enable linker relaxation. Linker relaxation is a process whereby the
23108 linker attempts to reduce the size of a program by finding shorter
23109 versions of various instructions. Disabled by default.
23111 @item -mint-register=@var{N}
23112 @opindex mint-register
23113 Specify the number of registers to reserve for fast interrupt handler
23114 functions. The value @var{N} can be between 0 and 4. A value of 1
23115 means that register @code{r13} is reserved for the exclusive use
23116 of fast interrupt handlers. A value of 2 reserves @code{r13} and
23117 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
23118 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
23119 A value of 0, the default, does not reserve any registers.
23121 @item -msave-acc-in-interrupts
23122 @opindex msave-acc-in-interrupts
23123 Specifies that interrupt handler functions should preserve the
23124 accumulator register. This is only necessary if normal code might use
23125 the accumulator register, for example because it performs 64-bit
23126 multiplications. The default is to ignore the accumulator as this
23127 makes the interrupt handlers faster.
23133 Enables the generation of position independent data. When enabled any
23134 access to constant data is done via an offset from a base address
23135 held in a register. This allows the location of constant data to be
23136 determined at run time without requiring the executable to be
23137 relocated, which is a benefit to embedded applications with tight
23138 memory constraints. Data that can be modified is not affected by this
23141 Note, using this feature reserves a register, usually @code{r13}, for
23142 the constant data base address. This can result in slower and/or
23143 larger code, especially in complicated functions.
23145 The actual register chosen to hold the constant data base address
23146 depends upon whether the @option{-msmall-data-limit} and/or the
23147 @option{-mint-register} command-line options are enabled. Starting
23148 with register @code{r13} and proceeding downwards, registers are
23149 allocated first to satisfy the requirements of @option{-mint-register},
23150 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
23151 is possible for the small data area register to be @code{r8} if both
23152 @option{-mint-register=4} and @option{-mpid} are specified on the
23155 By default this feature is not enabled. The default can be restored
23156 via the @option{-mno-pid} command-line option.
23158 @item -mno-warn-multiple-fast-interrupts
23159 @itemx -mwarn-multiple-fast-interrupts
23160 @opindex mno-warn-multiple-fast-interrupts
23161 @opindex mwarn-multiple-fast-interrupts
23162 Prevents GCC from issuing a warning message if it finds more than one
23163 fast interrupt handler when it is compiling a file. The default is to
23164 issue a warning for each extra fast interrupt handler found, as the RX
23165 only supports one such interrupt.
23167 @item -mallow-string-insns
23168 @itemx -mno-allow-string-insns
23169 @opindex mallow-string-insns
23170 @opindex mno-allow-string-insns
23171 Enables or disables the use of the string manipulation instructions
23172 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
23173 @code{SWHILE} and also the @code{RMPA} instruction. These
23174 instructions may prefetch data, which is not safe to do if accessing
23175 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
23176 for more information).
23178 The default is to allow these instructions, but it is not possible for
23179 GCC to reliably detect all circumstances where a string instruction
23180 might be used to access an I/O register, so their use cannot be
23181 disabled automatically. Instead it is reliant upon the programmer to
23182 use the @option{-mno-allow-string-insns} option if their program
23183 accesses I/O space.
23185 When the instructions are enabled GCC defines the C preprocessor
23186 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
23187 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
23193 Use only (or not only) @code{JSR} instructions to access functions.
23194 This option can be used when code size exceeds the range of @code{BSR}
23195 instructions. Note that @option{-mno-jsr} does not mean to not use
23196 @code{JSR} but instead means that any type of branch may be used.
23199 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
23200 has special significance to the RX port when used with the
23201 @code{interrupt} function attribute. This attribute indicates a
23202 function intended to process fast interrupts. GCC ensures
23203 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
23204 and/or @code{r13} and only provided that the normal use of the
23205 corresponding registers have been restricted via the
23206 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
23209 @node S/390 and zSeries Options
23210 @subsection S/390 and zSeries Options
23211 @cindex S/390 and zSeries Options
23213 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
23217 @itemx -msoft-float
23218 @opindex mhard-float
23219 @opindex msoft-float
23220 Use (do not use) the hardware floating-point instructions and registers
23221 for floating-point operations. When @option{-msoft-float} is specified,
23222 functions in @file{libgcc.a} are used to perform floating-point
23223 operations. When @option{-mhard-float} is specified, the compiler
23224 generates IEEE floating-point instructions. This is the default.
23227 @itemx -mno-hard-dfp
23229 @opindex mno-hard-dfp
23230 Use (do not use) the hardware decimal-floating-point instructions for
23231 decimal-floating-point operations. When @option{-mno-hard-dfp} is
23232 specified, functions in @file{libgcc.a} are used to perform
23233 decimal-floating-point operations. When @option{-mhard-dfp} is
23234 specified, the compiler generates decimal-floating-point hardware
23235 instructions. This is the default for @option{-march=z9-ec} or higher.
23237 @item -mlong-double-64
23238 @itemx -mlong-double-128
23239 @opindex mlong-double-64
23240 @opindex mlong-double-128
23241 These switches control the size of @code{long double} type. A size
23242 of 64 bits makes the @code{long double} type equivalent to the @code{double}
23243 type. This is the default.
23246 @itemx -mno-backchain
23247 @opindex mbackchain
23248 @opindex mno-backchain
23249 Store (do not store) the address of the caller's frame as backchain pointer
23250 into the callee's stack frame.
23251 A backchain may be needed to allow debugging using tools that do not understand
23252 DWARF call frame information.
23253 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
23254 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
23255 the backchain is placed into the topmost word of the 96/160 byte register
23258 In general, code compiled with @option{-mbackchain} is call-compatible with
23259 code compiled with @option{-mmo-backchain}; however, use of the backchain
23260 for debugging purposes usually requires that the whole binary is built with
23261 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
23262 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23263 to build a linux kernel use @option{-msoft-float}.
23265 The default is to not maintain the backchain.
23267 @item -mpacked-stack
23268 @itemx -mno-packed-stack
23269 @opindex mpacked-stack
23270 @opindex mno-packed-stack
23271 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
23272 specified, the compiler uses the all fields of the 96/160 byte register save
23273 area only for their default purpose; unused fields still take up stack space.
23274 When @option{-mpacked-stack} is specified, register save slots are densely
23275 packed at the top of the register save area; unused space is reused for other
23276 purposes, allowing for more efficient use of the available stack space.
23277 However, when @option{-mbackchain} is also in effect, the topmost word of
23278 the save area is always used to store the backchain, and the return address
23279 register is always saved two words below the backchain.
23281 As long as the stack frame backchain is not used, code generated with
23282 @option{-mpacked-stack} is call-compatible with code generated with
23283 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
23284 S/390 or zSeries generated code that uses the stack frame backchain at run
23285 time, not just for debugging purposes. Such code is not call-compatible
23286 with code compiled with @option{-mpacked-stack}. Also, note that the
23287 combination of @option{-mbackchain},
23288 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23289 to build a linux kernel use @option{-msoft-float}.
23291 The default is to not use the packed stack layout.
23294 @itemx -mno-small-exec
23295 @opindex msmall-exec
23296 @opindex mno-small-exec
23297 Generate (or do not generate) code using the @code{bras} instruction
23298 to do subroutine calls.
23299 This only works reliably if the total executable size does not
23300 exceed 64k. The default is to use the @code{basr} instruction instead,
23301 which does not have this limitation.
23307 When @option{-m31} is specified, generate code compliant to the
23308 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
23309 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
23310 particular to generate 64-bit instructions. For the @samp{s390}
23311 targets, the default is @option{-m31}, while the @samp{s390x}
23312 targets default to @option{-m64}.
23318 When @option{-mzarch} is specified, generate code using the
23319 instructions available on z/Architecture.
23320 When @option{-mesa} is specified, generate code using the
23321 instructions available on ESA/390. Note that @option{-mesa} is
23322 not possible with @option{-m64}.
23323 When generating code compliant to the GNU/Linux for S/390 ABI,
23324 the default is @option{-mesa}. When generating code compliant
23325 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
23331 The @option{-mhtm} option enables a set of builtins making use of
23332 instructions available with the transactional execution facility
23333 introduced with the IBM zEnterprise EC12 machine generation
23334 @ref{S/390 System z Built-in Functions}.
23335 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
23341 When @option{-mvx} is specified, generate code using the instructions
23342 available with the vector extension facility introduced with the IBM
23343 z13 machine generation.
23344 This option changes the ABI for some vector type values with regard to
23345 alignment and calling conventions. In case vector type values are
23346 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
23347 command will be added to mark the resulting binary with the ABI used.
23348 @option{-mvx} is enabled by default when using @option{-march=z13}.
23351 @itemx -mno-zvector
23353 @opindex mno-zvector
23354 The @option{-mzvector} option enables vector language extensions and
23355 builtins using instructions available with the vector extension
23356 facility introduced with the IBM z13 machine generation.
23357 This option adds support for @samp{vector} to be used as a keyword to
23358 define vector type variables and arguments. @samp{vector} is only
23359 available when GNU extensions are enabled. It will not be expanded
23360 when requesting strict standard compliance e.g. with @option{-std=c99}.
23361 In addition to the GCC low-level builtins @option{-mzvector} enables
23362 a set of builtins added for compatibility with AltiVec-style
23363 implementations like Power and Cell. In order to make use of these
23364 builtins the header file @file{vecintrin.h} needs to be included.
23365 @option{-mzvector} is disabled by default.
23371 Generate (or do not generate) code using the @code{mvcle} instruction
23372 to perform block moves. When @option{-mno-mvcle} is specified,
23373 use a @code{mvc} loop instead. This is the default unless optimizing for
23380 Print (or do not print) additional debug information when compiling.
23381 The default is to not print debug information.
23383 @item -march=@var{cpu-type}
23385 Generate code that runs on @var{cpu-type}, which is the name of a
23386 system representing a certain processor type. Possible values for
23387 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
23388 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
23389 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
23392 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
23393 @samp{g6} are deprecated and will be removed with future releases.
23395 Specifying @samp{native} as cpu type can be used to select the best
23396 architecture option for the host processor.
23397 @option{-march=native} has no effect if GCC does not recognize the
23400 @item -mtune=@var{cpu-type}
23402 Tune to @var{cpu-type} everything applicable about the generated code,
23403 except for the ABI and the set of available instructions.
23404 The list of @var{cpu-type} values is the same as for @option{-march}.
23405 The default is the value used for @option{-march}.
23408 @itemx -mno-tpf-trace
23409 @opindex mtpf-trace
23410 @opindex mno-tpf-trace
23411 Generate code that adds (does not add) in TPF OS specific branches to trace
23412 routines in the operating system. This option is off by default, even
23413 when compiling for the TPF OS@.
23416 @itemx -mno-fused-madd
23417 @opindex mfused-madd
23418 @opindex mno-fused-madd
23419 Generate code that uses (does not use) the floating-point multiply and
23420 accumulate instructions. These instructions are generated by default if
23421 hardware floating point is used.
23423 @item -mwarn-framesize=@var{framesize}
23424 @opindex mwarn-framesize
23425 Emit a warning if the current function exceeds the given frame size. Because
23426 this is a compile-time check it doesn't need to be a real problem when the program
23427 runs. It is intended to identify functions that most probably cause
23428 a stack overflow. It is useful to be used in an environment with limited stack
23429 size e.g.@: the linux kernel.
23431 @item -mwarn-dynamicstack
23432 @opindex mwarn-dynamicstack
23433 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
23434 arrays. This is generally a bad idea with a limited stack size.
23436 @item -mstack-guard=@var{stack-guard}
23437 @itemx -mstack-size=@var{stack-size}
23438 @opindex mstack-guard
23439 @opindex mstack-size
23440 If these options are provided the S/390 back end emits additional instructions in
23441 the function prologue that trigger a trap if the stack size is @var{stack-guard}
23442 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
23443 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
23444 the frame size of the compiled function is chosen.
23445 These options are intended to be used to help debugging stack overflow problems.
23446 The additionally emitted code causes only little overhead and hence can also be
23447 used in production-like systems without greater performance degradation. The given
23448 values have to be exact powers of 2 and @var{stack-size} has to be greater than
23449 @var{stack-guard} without exceeding 64k.
23450 In order to be efficient the extra code makes the assumption that the stack starts
23451 at an address aligned to the value given by @var{stack-size}.
23452 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
23454 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
23456 If the hotpatch option is enabled, a ``hot-patching'' function
23457 prologue is generated for all functions in the compilation unit.
23458 The funtion label is prepended with the given number of two-byte
23459 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
23460 the label, 2 * @var{post-halfwords} bytes are appended, using the
23461 largest NOP like instructions the architecture allows (maximum
23464 If both arguments are zero, hotpatching is disabled.
23466 This option can be overridden for individual functions with the
23467 @code{hotpatch} attribute.
23470 @node Score Options
23471 @subsection Score Options
23472 @cindex Score Options
23474 These options are defined for Score implementations:
23479 Compile code for big-endian mode. This is the default.
23483 Compile code for little-endian mode.
23487 Disable generation of @code{bcnz} instructions.
23491 Enable generation of unaligned load and store instructions.
23495 Enable the use of multiply-accumulate instructions. Disabled by default.
23499 Specify the SCORE5 as the target architecture.
23503 Specify the SCORE5U of the target architecture.
23507 Specify the SCORE7 as the target architecture. This is the default.
23511 Specify the SCORE7D as the target architecture.
23515 @subsection SH Options
23517 These @samp{-m} options are defined for the SH implementations:
23522 Generate code for the SH1.
23526 Generate code for the SH2.
23529 Generate code for the SH2e.
23533 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
23534 that the floating-point unit is not used.
23536 @item -m2a-single-only
23537 @opindex m2a-single-only
23538 Generate code for the SH2a-FPU, in such a way that no double-precision
23539 floating-point operations are used.
23542 @opindex m2a-single
23543 Generate code for the SH2a-FPU assuming the floating-point unit is in
23544 single-precision mode by default.
23548 Generate code for the SH2a-FPU assuming the floating-point unit is in
23549 double-precision mode by default.
23553 Generate code for the SH3.
23557 Generate code for the SH3e.
23561 Generate code for the SH4 without a floating-point unit.
23563 @item -m4-single-only
23564 @opindex m4-single-only
23565 Generate code for the SH4 with a floating-point unit that only
23566 supports single-precision arithmetic.
23570 Generate code for the SH4 assuming the floating-point unit is in
23571 single-precision mode by default.
23575 Generate code for the SH4.
23579 Generate code for SH4-100.
23581 @item -m4-100-nofpu
23582 @opindex m4-100-nofpu
23583 Generate code for SH4-100 in such a way that the
23584 floating-point unit is not used.
23586 @item -m4-100-single
23587 @opindex m4-100-single
23588 Generate code for SH4-100 assuming the floating-point unit is in
23589 single-precision mode by default.
23591 @item -m4-100-single-only
23592 @opindex m4-100-single-only
23593 Generate code for SH4-100 in such a way that no double-precision
23594 floating-point operations are used.
23598 Generate code for SH4-200.
23600 @item -m4-200-nofpu
23601 @opindex m4-200-nofpu
23602 Generate code for SH4-200 without in such a way that the
23603 floating-point unit is not used.
23605 @item -m4-200-single
23606 @opindex m4-200-single
23607 Generate code for SH4-200 assuming the floating-point unit is in
23608 single-precision mode by default.
23610 @item -m4-200-single-only
23611 @opindex m4-200-single-only
23612 Generate code for SH4-200 in such a way that no double-precision
23613 floating-point operations are used.
23617 Generate code for SH4-300.
23619 @item -m4-300-nofpu
23620 @opindex m4-300-nofpu
23621 Generate code for SH4-300 without in such a way that the
23622 floating-point unit is not used.
23624 @item -m4-300-single
23625 @opindex m4-300-single
23626 Generate code for SH4-300 in such a way that no double-precision
23627 floating-point operations are used.
23629 @item -m4-300-single-only
23630 @opindex m4-300-single-only
23631 Generate code for SH4-300 in such a way that no double-precision
23632 floating-point operations are used.
23636 Generate code for SH4-340 (no MMU, no FPU).
23640 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
23645 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
23646 floating-point unit is not used.
23648 @item -m4a-single-only
23649 @opindex m4a-single-only
23650 Generate code for the SH4a, in such a way that no double-precision
23651 floating-point operations are used.
23654 @opindex m4a-single
23655 Generate code for the SH4a assuming the floating-point unit is in
23656 single-precision mode by default.
23660 Generate code for the SH4a.
23664 Same as @option{-m4a-nofpu}, except that it implicitly passes
23665 @option{-dsp} to the assembler. GCC doesn't generate any DSP
23666 instructions at the moment.
23670 Compile code for the processor in big-endian mode.
23674 Compile code for the processor in little-endian mode.
23678 Align doubles at 64-bit boundaries. Note that this changes the calling
23679 conventions, and thus some functions from the standard C library do
23680 not work unless you recompile it first with @option{-mdalign}.
23684 Shorten some address references at link time, when possible; uses the
23685 linker option @option{-relax}.
23689 Use 32-bit offsets in @code{switch} tables. The default is to use
23694 Enable the use of bit manipulation instructions on SH2A.
23698 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
23699 alignment constraints.
23703 Comply with the calling conventions defined by Renesas.
23706 @opindex mno-renesas
23707 Comply with the calling conventions defined for GCC before the Renesas
23708 conventions were available. This option is the default for all
23709 targets of the SH toolchain.
23712 @opindex mnomacsave
23713 Mark the @code{MAC} register as call-clobbered, even if
23714 @option{-mrenesas} is given.
23720 Control the IEEE compliance of floating-point comparisons, which affects the
23721 handling of cases where the result of a comparison is unordered. By default
23722 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
23723 enabled @option{-mno-ieee} is implicitly set, which results in faster
23724 floating-point greater-equal and less-equal comparisons. The implicit settings
23725 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
23727 @item -minline-ic_invalidate
23728 @opindex minline-ic_invalidate
23729 Inline code to invalidate instruction cache entries after setting up
23730 nested function trampolines.
23731 This option has no effect if @option{-musermode} is in effect and the selected
23732 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
23734 If the selected code generation option does not allow the use of the @code{icbi}
23735 instruction, and @option{-musermode} is not in effect, the inlined code
23736 manipulates the instruction cache address array directly with an associative
23737 write. This not only requires privileged mode at run time, but it also
23738 fails if the cache line had been mapped via the TLB and has become unmapped.
23742 Dump instruction size and location in the assembly code.
23745 @opindex mpadstruct
23746 This option is deprecated. It pads structures to multiple of 4 bytes,
23747 which is incompatible with the SH ABI@.
23749 @item -matomic-model=@var{model}
23750 @opindex matomic-model=@var{model}
23751 Sets the model of atomic operations and additional parameters as a comma
23752 separated list. For details on the atomic built-in functions see
23753 @ref{__atomic Builtins}. The following models and parameters are supported:
23758 Disable compiler generated atomic sequences and emit library calls for atomic
23759 operations. This is the default if the target is not @code{sh*-*-linux*}.
23762 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
23763 built-in functions. The generated atomic sequences require additional support
23764 from the interrupt/exception handling code of the system and are only suitable
23765 for SH3* and SH4* single-core systems. This option is enabled by default when
23766 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
23767 this option also partially utilizes the hardware atomic instructions
23768 @code{movli.l} and @code{movco.l} to create more efficient code, unless
23769 @samp{strict} is specified.
23772 Generate software atomic sequences that use a variable in the thread control
23773 block. This is a variation of the gUSA sequences which can also be used on
23774 SH1* and SH2* targets. The generated atomic sequences require additional
23775 support from the interrupt/exception handling code of the system and are only
23776 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
23777 parameter has to be specified as well.
23780 Generate software atomic sequences that temporarily disable interrupts by
23781 setting @code{SR.IMASK = 1111}. This model works only when the program runs
23782 in privileged mode and is only suitable for single-core systems. Additional
23783 support from the interrupt/exception handling code of the system is not
23784 required. This model is enabled by default when the target is
23785 @code{sh*-*-linux*} and SH1* or SH2*.
23788 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
23789 instructions only. This is only available on SH4A and is suitable for
23790 multi-core systems. Since the hardware instructions support only 32 bit atomic
23791 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
23792 Code compiled with this option is also compatible with other software
23793 atomic model interrupt/exception handling systems if executed on an SH4A
23794 system. Additional support from the interrupt/exception handling code of the
23795 system is not required for this model.
23798 This parameter specifies the offset in bytes of the variable in the thread
23799 control block structure that should be used by the generated atomic sequences
23800 when the @samp{soft-tcb} model has been selected. For other models this
23801 parameter is ignored. The specified value must be an integer multiple of four
23802 and in the range 0-1020.
23805 This parameter prevents mixed usage of multiple atomic models, even if they
23806 are compatible, and makes the compiler generate atomic sequences of the
23807 specified model only.
23813 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
23814 Notice that depending on the particular hardware and software configuration
23815 this can degrade overall performance due to the operand cache line flushes
23816 that are implied by the @code{tas.b} instruction. On multi-core SH4A
23817 processors the @code{tas.b} instruction must be used with caution since it
23818 can result in data corruption for certain cache configurations.
23821 @opindex mprefergot
23822 When generating position-independent code, emit function calls using
23823 the Global Offset Table instead of the Procedure Linkage Table.
23826 @itemx -mno-usermode
23828 @opindex mno-usermode
23829 Don't allow (allow) the compiler generating privileged mode code. Specifying
23830 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
23831 inlined code would not work in user mode. @option{-musermode} is the default
23832 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
23833 @option{-musermode} has no effect, since there is no user mode.
23835 @item -multcost=@var{number}
23836 @opindex multcost=@var{number}
23837 Set the cost to assume for a multiply insn.
23839 @item -mdiv=@var{strategy}
23840 @opindex mdiv=@var{strategy}
23841 Set the division strategy to be used for integer division operations.
23842 @var{strategy} can be one of:
23847 Calls a library function that uses the single-step division instruction
23848 @code{div1} to perform the operation. Division by zero calculates an
23849 unspecified result and does not trap. This is the default except for SH4,
23850 SH2A and SHcompact.
23853 Calls a library function that performs the operation in double precision
23854 floating point. Division by zero causes a floating-point exception. This is
23855 the default for SHcompact with FPU. Specifying this for targets that do not
23856 have a double precision FPU defaults to @code{call-div1}.
23859 Calls a library function that uses a lookup table for small divisors and
23860 the @code{div1} instruction with case distinction for larger divisors. Division
23861 by zero calculates an unspecified result and does not trap. This is the default
23862 for SH4. Specifying this for targets that do not have dynamic shift
23863 instructions defaults to @code{call-div1}.
23867 When a division strategy has not been specified the default strategy is
23868 selected based on the current target. For SH2A the default strategy is to
23869 use the @code{divs} and @code{divu} instructions instead of library function
23872 @item -maccumulate-outgoing-args
23873 @opindex maccumulate-outgoing-args
23874 Reserve space once for outgoing arguments in the function prologue rather
23875 than around each call. Generally beneficial for performance and size. Also
23876 needed for unwinding to avoid changing the stack frame around conditional code.
23878 @item -mdivsi3_libfunc=@var{name}
23879 @opindex mdivsi3_libfunc=@var{name}
23880 Set the name of the library function used for 32-bit signed division to
23882 This only affects the name used in the @samp{call} division strategies, and
23883 the compiler still expects the same sets of input/output/clobbered registers as
23884 if this option were not present.
23886 @item -mfixed-range=@var{register-range}
23887 @opindex mfixed-range
23888 Generate code treating the given register range as fixed registers.
23889 A fixed register is one that the register allocator can not use. This is
23890 useful when compiling kernel code. A register range is specified as
23891 two registers separated by a dash. Multiple register ranges can be
23892 specified separated by a comma.
23894 @item -mbranch-cost=@var{num}
23895 @opindex mbranch-cost=@var{num}
23896 Assume @var{num} to be the cost for a branch instruction. Higher numbers
23897 make the compiler try to generate more branch-free code if possible.
23898 If not specified the value is selected depending on the processor type that
23899 is being compiled for.
23902 @itemx -mno-zdcbranch
23903 @opindex mzdcbranch
23904 @opindex mno-zdcbranch
23905 Assume (do not assume) that zero displacement conditional branch instructions
23906 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
23907 compiler prefers zero displacement branch code sequences. This is
23908 enabled by default when generating code for SH4 and SH4A. It can be explicitly
23909 disabled by specifying @option{-mno-zdcbranch}.
23911 @item -mcbranch-force-delay-slot
23912 @opindex mcbranch-force-delay-slot
23913 Force the usage of delay slots for conditional branches, which stuffs the delay
23914 slot with a @code{nop} if a suitable instruction cannot be found. By default
23915 this option is disabled. It can be enabled to work around hardware bugs as
23916 found in the original SH7055.
23919 @itemx -mno-fused-madd
23920 @opindex mfused-madd
23921 @opindex mno-fused-madd
23922 Generate code that uses (does not use) the floating-point multiply and
23923 accumulate instructions. These instructions are generated by default
23924 if hardware floating point is used. The machine-dependent
23925 @option{-mfused-madd} option is now mapped to the machine-independent
23926 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
23927 mapped to @option{-ffp-contract=off}.
23933 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
23934 and cosine approximations. The option @option{-mfsca} must be used in
23935 combination with @option{-funsafe-math-optimizations}. It is enabled by default
23936 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
23937 approximations even if @option{-funsafe-math-optimizations} is in effect.
23943 Allow or disallow the compiler to emit the @code{fsrra} instruction for
23944 reciprocal square root approximations. The option @option{-mfsrra} must be used
23945 in combination with @option{-funsafe-math-optimizations} and
23946 @option{-ffinite-math-only}. It is enabled by default when generating code for
23947 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
23948 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
23951 @item -mpretend-cmove
23952 @opindex mpretend-cmove
23953 Prefer zero-displacement conditional branches for conditional move instruction
23954 patterns. This can result in faster code on the SH4 processor.
23958 Generate code using the FDPIC ABI.
23962 @node Solaris 2 Options
23963 @subsection Solaris 2 Options
23964 @cindex Solaris 2 options
23966 These @samp{-m} options are supported on Solaris 2:
23969 @item -mclear-hwcap
23970 @opindex mclear-hwcap
23971 @option{-mclear-hwcap} tells the compiler to remove the hardware
23972 capabilities generated by the Solaris assembler. This is only necessary
23973 when object files use ISA extensions not supported by the current
23974 machine, but check at runtime whether or not to use them.
23976 @item -mimpure-text
23977 @opindex mimpure-text
23978 @option{-mimpure-text}, used in addition to @option{-shared}, tells
23979 the compiler to not pass @option{-z text} to the linker when linking a
23980 shared object. Using this option, you can link position-dependent
23981 code into a shared object.
23983 @option{-mimpure-text} suppresses the ``relocations remain against
23984 allocatable but non-writable sections'' linker error message.
23985 However, the necessary relocations trigger copy-on-write, and the
23986 shared object is not actually shared across processes. Instead of
23987 using @option{-mimpure-text}, you should compile all source code with
23988 @option{-fpic} or @option{-fPIC}.
23992 These switches are supported in addition to the above on Solaris 2:
23997 This is a synonym for @option{-pthread}.
24000 @node SPARC Options
24001 @subsection SPARC Options
24002 @cindex SPARC options
24004 These @samp{-m} options are supported on the SPARC:
24007 @item -mno-app-regs
24009 @opindex mno-app-regs
24011 Specify @option{-mapp-regs} to generate output using the global registers
24012 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
24013 global register 1, each global register 2 through 4 is then treated as an
24014 allocable register that is clobbered by function calls. This is the default.
24016 To be fully SVR4 ABI-compliant at the cost of some performance loss,
24017 specify @option{-mno-app-regs}. You should compile libraries and system
24018 software with this option.
24024 With @option{-mflat}, the compiler does not generate save/restore instructions
24025 and uses a ``flat'' or single register window model. This model is compatible
24026 with the regular register window model. The local registers and the input
24027 registers (0--5) are still treated as ``call-saved'' registers and are
24028 saved on the stack as needed.
24030 With @option{-mno-flat} (the default), the compiler generates save/restore
24031 instructions (except for leaf functions). This is the normal operating mode.
24034 @itemx -mhard-float
24036 @opindex mhard-float
24037 Generate output containing floating-point instructions. This is the
24041 @itemx -msoft-float
24043 @opindex msoft-float
24044 Generate output containing library calls for floating point.
24045 @strong{Warning:} the requisite libraries are not available for all SPARC
24046 targets. Normally the facilities of the machine's usual C compiler are
24047 used, but this cannot be done directly in cross-compilation. You must make
24048 your own arrangements to provide suitable library functions for
24049 cross-compilation. The embedded targets @samp{sparc-*-aout} and
24050 @samp{sparclite-*-*} do provide software floating-point support.
24052 @option{-msoft-float} changes the calling convention in the output file;
24053 therefore, it is only useful if you compile @emph{all} of a program with
24054 this option. In particular, you need to compile @file{libgcc.a}, the
24055 library that comes with GCC, with @option{-msoft-float} in order for
24058 @item -mhard-quad-float
24059 @opindex mhard-quad-float
24060 Generate output containing quad-word (long double) floating-point
24063 @item -msoft-quad-float
24064 @opindex msoft-quad-float
24065 Generate output containing library calls for quad-word (long double)
24066 floating-point instructions. The functions called are those specified
24067 in the SPARC ABI@. This is the default.
24069 As of this writing, there are no SPARC implementations that have hardware
24070 support for the quad-word floating-point instructions. They all invoke
24071 a trap handler for one of these instructions, and then the trap handler
24072 emulates the effect of the instruction. Because of the trap handler overhead,
24073 this is much slower than calling the ABI library routines. Thus the
24074 @option{-msoft-quad-float} option is the default.
24076 @item -mno-unaligned-doubles
24077 @itemx -munaligned-doubles
24078 @opindex mno-unaligned-doubles
24079 @opindex munaligned-doubles
24080 Assume that doubles have 8-byte alignment. This is the default.
24082 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
24083 alignment only if they are contained in another type, or if they have an
24084 absolute address. Otherwise, it assumes they have 4-byte alignment.
24085 Specifying this option avoids some rare compatibility problems with code
24086 generated by other compilers. It is not the default because it results
24087 in a performance loss, especially for floating-point code.
24090 @itemx -mno-user-mode
24091 @opindex muser-mode
24092 @opindex mno-user-mode
24093 Do not generate code that can only run in supervisor mode. This is relevant
24094 only for the @code{casa} instruction emitted for the LEON3 processor. This
24097 @item -mfaster-structs
24098 @itemx -mno-faster-structs
24099 @opindex mfaster-structs
24100 @opindex mno-faster-structs
24101 With @option{-mfaster-structs}, the compiler assumes that structures
24102 should have 8-byte alignment. This enables the use of pairs of
24103 @code{ldd} and @code{std} instructions for copies in structure
24104 assignment, in place of twice as many @code{ld} and @code{st} pairs.
24105 However, the use of this changed alignment directly violates the SPARC
24106 ABI@. Thus, it's intended only for use on targets where the developer
24107 acknowledges that their resulting code is not directly in line with
24108 the rules of the ABI@.
24110 @item -mstd-struct-return
24111 @itemx -mno-std-struct-return
24112 @opindex mstd-struct-return
24113 @opindex mno-std-struct-return
24114 With @option{-mstd-struct-return}, the compiler generates checking code
24115 in functions returning structures or unions to detect size mismatches
24116 between the two sides of function calls, as per the 32-bit ABI@.
24118 The default is @option{-mno-std-struct-return}. This option has no effect
24125 Enable Local Register Allocation. This is the default for SPARC since GCC 7
24126 so @option{-mno-lra} needs to be passed to get old Reload.
24128 @item -mcpu=@var{cpu_type}
24130 Set the instruction set, register set, and instruction scheduling parameters
24131 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24132 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
24133 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
24134 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
24135 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
24136 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
24138 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
24139 which selects the best architecture option for the host processor.
24140 @option{-mcpu=native} has no effect if GCC does not recognize
24143 Default instruction scheduling parameters are used for values that select
24144 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
24145 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
24147 Here is a list of each supported architecture and their supported
24155 supersparc, hypersparc, leon, leon3
24158 f930, f934, sparclite86x
24164 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
24168 By default (unless configured otherwise), GCC generates code for the V7
24169 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
24170 additionally optimizes it for the Cypress CY7C602 chip, as used in the
24171 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
24172 SPARCStation 1, 2, IPX etc.
24174 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
24175 architecture. The only difference from V7 code is that the compiler emits
24176 the integer multiply and integer divide instructions which exist in SPARC-V8
24177 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
24178 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
24181 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
24182 the SPARC architecture. This adds the integer multiply, integer divide step
24183 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
24184 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
24185 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
24186 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
24187 MB86934 chip, which is the more recent SPARClite with FPU@.
24189 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
24190 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
24191 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
24192 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
24193 optimizes it for the TEMIC SPARClet chip.
24195 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
24196 architecture. This adds 64-bit integer and floating-point move instructions,
24197 3 additional floating-point condition code registers and conditional move
24198 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
24199 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
24200 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
24201 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
24202 @option{-mcpu=niagara}, the compiler additionally optimizes it for
24203 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
24204 additionally optimizes it for Sun UltraSPARC T2 chips. With
24205 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
24206 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
24207 additionally optimizes it for Sun UltraSPARC T4 chips. With
24208 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
24209 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
24210 additionally optimizes it for Oracle M8 chips.
24212 @item -mtune=@var{cpu_type}
24214 Set the instruction scheduling parameters for machine type
24215 @var{cpu_type}, but do not set the instruction set or register set that the
24216 option @option{-mcpu=@var{cpu_type}} does.
24218 The same values for @option{-mcpu=@var{cpu_type}} can be used for
24219 @option{-mtune=@var{cpu_type}}, but the only useful values are those
24220 that select a particular CPU implementation. Those are
24221 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
24222 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
24223 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
24224 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
24225 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
24226 and GNU/Linux toolchains, @samp{native} can also be used.
24231 @opindex mno-v8plus
24232 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
24233 difference from the V8 ABI is that the global and out registers are
24234 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
24235 mode for all SPARC-V9 processors.
24241 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
24242 Visual Instruction Set extensions. The default is @option{-mno-vis}.
24248 With @option{-mvis2}, GCC generates code that takes advantage of
24249 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
24250 default is @option{-mvis2} when targeting a cpu that supports such
24251 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
24252 also sets @option{-mvis}.
24258 With @option{-mvis3}, GCC generates code that takes advantage of
24259 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
24260 default is @option{-mvis3} when targeting a cpu that supports such
24261 instructions, such as niagara-3 and later. Setting @option{-mvis3}
24262 also sets @option{-mvis2} and @option{-mvis}.
24268 With @option{-mvis4}, GCC generates code that takes advantage of
24269 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
24270 default is @option{-mvis4} when targeting a cpu that supports such
24271 instructions, such as niagara-7 and later. Setting @option{-mvis4}
24272 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
24278 With @option{-mvis4b}, GCC generates code that takes advantage of
24279 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
24280 the additional VIS instructions introduced in the Oracle SPARC
24281 Architecture 2017. The default is @option{-mvis4b} when targeting a
24282 cpu that supports such instructions, such as m8 and later. Setting
24283 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
24284 @option{-mvis2} and @option{-mvis}.
24289 @opindex mno-cbcond
24290 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
24291 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
24292 when targeting a CPU that supports such instructions, such as Niagara-4 and
24299 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
24300 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
24301 when targeting a CPU that supports such instructions, such as Niagara-3 and
24307 @opindex mno-fsmuld
24308 With @option{-mfsmuld}, GCC generates code that takes advantage of the
24309 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
24310 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
24311 or V9 with FPU except @option{-mcpu=leon}.
24317 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
24318 Population Count instruction. The default is @option{-mpopc}
24319 when targeting a CPU that supports such an instruction, such as Niagara-2 and
24326 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
24327 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
24328 when targeting a CPU that supports such an instruction, such as Niagara-7 and
24332 @opindex mfix-at697f
24333 Enable the documented workaround for the single erratum of the Atmel AT697F
24334 processor (which corresponds to erratum #13 of the AT697E processor).
24337 @opindex mfix-ut699
24338 Enable the documented workarounds for the floating-point errata and the data
24339 cache nullify errata of the UT699 processor.
24342 @opindex mfix-ut700
24343 Enable the documented workaround for the back-to-back store errata of
24344 the UT699E/UT700 processor.
24346 @item -mfix-gr712rc
24347 @opindex mfix-gr712rc
24348 Enable the documented workaround for the back-to-back store errata of
24349 the GR712RC processor.
24352 These @samp{-m} options are supported in addition to the above
24353 on SPARC-V9 processors in 64-bit environments:
24360 Generate code for a 32-bit or 64-bit environment.
24361 The 32-bit environment sets int, long and pointer to 32 bits.
24362 The 64-bit environment sets int to 32 bits and long and pointer
24365 @item -mcmodel=@var{which}
24367 Set the code model to one of
24371 The Medium/Low code model: 64-bit addresses, programs
24372 must be linked in the low 32 bits of memory. Programs can be statically
24373 or dynamically linked.
24376 The Medium/Middle code model: 64-bit addresses, programs
24377 must be linked in the low 44 bits of memory, the text and data segments must
24378 be less than 2GB in size and the data segment must be located within 2GB of
24382 The Medium/Anywhere code model: 64-bit addresses, programs
24383 may be linked anywhere in memory, the text and data segments must be less
24384 than 2GB in size and the data segment must be located within 2GB of the
24388 The Medium/Anywhere code model for embedded systems:
24389 64-bit addresses, the text and data segments must be less than 2GB in
24390 size, both starting anywhere in memory (determined at link time). The
24391 global register %g4 points to the base of the data segment. Programs
24392 are statically linked and PIC is not supported.
24395 @item -mmemory-model=@var{mem-model}
24396 @opindex mmemory-model
24397 Set the memory model in force on the processor to one of
24401 The default memory model for the processor and operating system.
24404 Relaxed Memory Order
24407 Partial Store Order
24413 Sequential Consistency
24416 These memory models are formally defined in Appendix D of the SPARC-V9
24417 architecture manual, as set in the processor's @code{PSTATE.MM} field.
24420 @itemx -mno-stack-bias
24421 @opindex mstack-bias
24422 @opindex mno-stack-bias
24423 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
24424 frame pointer if present, are offset by @minus{}2047 which must be added back
24425 when making stack frame references. This is the default in 64-bit mode.
24426 Otherwise, assume no such offset is present.
24430 @subsection SPU Options
24431 @cindex SPU options
24433 These @samp{-m} options are supported on the SPU:
24437 @itemx -merror-reloc
24438 @opindex mwarn-reloc
24439 @opindex merror-reloc
24441 The loader for SPU does not handle dynamic relocations. By default, GCC
24442 gives an error when it generates code that requires a dynamic
24443 relocation. @option{-mno-error-reloc} disables the error,
24444 @option{-mwarn-reloc} generates a warning instead.
24447 @itemx -munsafe-dma
24449 @opindex munsafe-dma
24451 Instructions that initiate or test completion of DMA must not be
24452 reordered with respect to loads and stores of the memory that is being
24454 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
24455 memory accesses, but that can lead to inefficient code in places where the
24456 memory is known to not change. Rather than mark the memory as volatile,
24457 you can use @option{-msafe-dma} to tell the compiler to treat
24458 the DMA instructions as potentially affecting all memory.
24460 @item -mbranch-hints
24461 @opindex mbranch-hints
24463 By default, GCC generates a branch hint instruction to avoid
24464 pipeline stalls for always-taken or probably-taken branches. A hint
24465 is not generated closer than 8 instructions away from its branch.
24466 There is little reason to disable them, except for debugging purposes,
24467 or to make an object a little bit smaller.
24471 @opindex msmall-mem
24472 @opindex mlarge-mem
24474 By default, GCC generates code assuming that addresses are never larger
24475 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
24476 a full 32-bit address.
24481 By default, GCC links against startup code that assumes the SPU-style
24482 main function interface (which has an unconventional parameter list).
24483 With @option{-mstdmain}, GCC links your program against startup
24484 code that assumes a C99-style interface to @code{main}, including a
24485 local copy of @code{argv} strings.
24487 @item -mfixed-range=@var{register-range}
24488 @opindex mfixed-range
24489 Generate code treating the given register range as fixed registers.
24490 A fixed register is one that the register allocator cannot use. This is
24491 useful when compiling kernel code. A register range is specified as
24492 two registers separated by a dash. Multiple register ranges can be
24493 specified separated by a comma.
24499 Compile code assuming that pointers to the PPU address space accessed
24500 via the @code{__ea} named address space qualifier are either 32 or 64
24501 bits wide. The default is 32 bits. As this is an ABI-changing option,
24502 all object code in an executable must be compiled with the same setting.
24504 @item -maddress-space-conversion
24505 @itemx -mno-address-space-conversion
24506 @opindex maddress-space-conversion
24507 @opindex mno-address-space-conversion
24508 Allow/disallow treating the @code{__ea} address space as superset
24509 of the generic address space. This enables explicit type casts
24510 between @code{__ea} and generic pointer as well as implicit
24511 conversions of generic pointers to @code{__ea} pointers. The
24512 default is to allow address space pointer conversions.
24514 @item -mcache-size=@var{cache-size}
24515 @opindex mcache-size
24516 This option controls the version of libgcc that the compiler links to an
24517 executable and selects a software-managed cache for accessing variables
24518 in the @code{__ea} address space with a particular cache size. Possible
24519 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
24520 and @samp{128}. The default cache size is 64KB.
24522 @item -matomic-updates
24523 @itemx -mno-atomic-updates
24524 @opindex matomic-updates
24525 @opindex mno-atomic-updates
24526 This option controls the version of libgcc that the compiler links to an
24527 executable and selects whether atomic updates to the software-managed
24528 cache of PPU-side variables are used. If you use atomic updates, changes
24529 to a PPU variable from SPU code using the @code{__ea} named address space
24530 qualifier do not interfere with changes to other PPU variables residing
24531 in the same cache line from PPU code. If you do not use atomic updates,
24532 such interference may occur; however, writing back cache lines is
24533 more efficient. The default behavior is to use atomic updates.
24536 @itemx -mdual-nops=@var{n}
24537 @opindex mdual-nops
24538 By default, GCC inserts NOPs to increase dual issue when it expects
24539 it to increase performance. @var{n} can be a value from 0 to 10. A
24540 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
24541 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
24543 @item -mhint-max-nops=@var{n}
24544 @opindex mhint-max-nops
24545 Maximum number of NOPs to insert for a branch hint. A branch hint must
24546 be at least 8 instructions away from the branch it is affecting. GCC
24547 inserts up to @var{n} NOPs to enforce this, otherwise it does not
24548 generate the branch hint.
24550 @item -mhint-max-distance=@var{n}
24551 @opindex mhint-max-distance
24552 The encoding of the branch hint instruction limits the hint to be within
24553 256 instructions of the branch it is affecting. By default, GCC makes
24554 sure it is within 125.
24557 @opindex msafe-hints
24558 Work around a hardware bug that causes the SPU to stall indefinitely.
24559 By default, GCC inserts the @code{hbrp} instruction to make sure
24560 this stall won't happen.
24564 @node System V Options
24565 @subsection Options for System V
24567 These additional options are available on System V Release 4 for
24568 compatibility with other compilers on those systems:
24573 Create a shared object.
24574 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
24578 Identify the versions of each tool used by the compiler, in a
24579 @code{.ident} assembler directive in the output.
24583 Refrain from adding @code{.ident} directives to the output file (this is
24586 @item -YP,@var{dirs}
24588 Search the directories @var{dirs}, and no others, for libraries
24589 specified with @option{-l}.
24591 @item -Ym,@var{dir}
24593 Look in the directory @var{dir} to find the M4 preprocessor.
24594 The assembler uses this option.
24595 @c This is supposed to go with a -Yd for predefined M4 macro files, but
24596 @c the generic assembler that comes with Solaris takes just -Ym.
24599 @node TILE-Gx Options
24600 @subsection TILE-Gx Options
24601 @cindex TILE-Gx options
24603 These @samp{-m} options are supported on the TILE-Gx:
24606 @item -mcmodel=small
24607 @opindex mcmodel=small
24608 Generate code for the small model. The distance for direct calls is
24609 limited to 500M in either direction. PC-relative addresses are 32
24610 bits. Absolute addresses support the full address range.
24612 @item -mcmodel=large
24613 @opindex mcmodel=large
24614 Generate code for the large model. There is no limitation on call
24615 distance, pc-relative addresses, or absolute addresses.
24617 @item -mcpu=@var{name}
24619 Selects the type of CPU to be targeted. Currently the only supported
24620 type is @samp{tilegx}.
24626 Generate code for a 32-bit or 64-bit environment. The 32-bit
24627 environment sets int, long, and pointer to 32 bits. The 64-bit
24628 environment sets int to 32 bits and long and pointer to 64 bits.
24631 @itemx -mlittle-endian
24632 @opindex mbig-endian
24633 @opindex mlittle-endian
24634 Generate code in big/little endian mode, respectively.
24637 @node TILEPro Options
24638 @subsection TILEPro Options
24639 @cindex TILEPro options
24641 These @samp{-m} options are supported on the TILEPro:
24644 @item -mcpu=@var{name}
24646 Selects the type of CPU to be targeted. Currently the only supported
24647 type is @samp{tilepro}.
24651 Generate code for a 32-bit environment, which sets int, long, and
24652 pointer to 32 bits. This is the only supported behavior so the flag
24653 is essentially ignored.
24657 @subsection V850 Options
24658 @cindex V850 Options
24660 These @samp{-m} options are defined for V850 implementations:
24664 @itemx -mno-long-calls
24665 @opindex mlong-calls
24666 @opindex mno-long-calls
24667 Treat all calls as being far away (near). If calls are assumed to be
24668 far away, the compiler always loads the function's address into a
24669 register, and calls indirect through the pointer.
24675 Do not optimize (do optimize) basic blocks that use the same index
24676 pointer 4 or more times to copy pointer into the @code{ep} register, and
24677 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
24678 option is on by default if you optimize.
24680 @item -mno-prolog-function
24681 @itemx -mprolog-function
24682 @opindex mno-prolog-function
24683 @opindex mprolog-function
24684 Do not use (do use) external functions to save and restore registers
24685 at the prologue and epilogue of a function. The external functions
24686 are slower, but use less code space if more than one function saves
24687 the same number of registers. The @option{-mprolog-function} option
24688 is on by default if you optimize.
24692 Try to make the code as small as possible. At present, this just turns
24693 on the @option{-mep} and @option{-mprolog-function} options.
24695 @item -mtda=@var{n}
24697 Put static or global variables whose size is @var{n} bytes or less into
24698 the tiny data area that register @code{ep} points to. The tiny data
24699 area can hold up to 256 bytes in total (128 bytes for byte references).
24701 @item -msda=@var{n}
24703 Put static or global variables whose size is @var{n} bytes or less into
24704 the small data area that register @code{gp} points to. The small data
24705 area can hold up to 64 kilobytes.
24707 @item -mzda=@var{n}
24709 Put static or global variables whose size is @var{n} bytes or less into
24710 the first 32 kilobytes of memory.
24714 Specify that the target processor is the V850.
24718 Specify that the target processor is the V850E3V5. The preprocessor
24719 constant @code{__v850e3v5__} is defined if this option is used.
24723 Specify that the target processor is the V850E3V5. This is an alias for
24724 the @option{-mv850e3v5} option.
24728 Specify that the target processor is the V850E2V3. The preprocessor
24729 constant @code{__v850e2v3__} is defined if this option is used.
24733 Specify that the target processor is the V850E2. The preprocessor
24734 constant @code{__v850e2__} is defined if this option is used.
24738 Specify that the target processor is the V850E1. The preprocessor
24739 constants @code{__v850e1__} and @code{__v850e__} are defined if
24740 this option is used.
24744 Specify that the target processor is the V850ES. This is an alias for
24745 the @option{-mv850e1} option.
24749 Specify that the target processor is the V850E@. The preprocessor
24750 constant @code{__v850e__} is defined if this option is used.
24752 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
24753 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
24754 are defined then a default target processor is chosen and the
24755 relevant @samp{__v850*__} preprocessor constant is defined.
24757 The preprocessor constants @code{__v850} and @code{__v851__} are always
24758 defined, regardless of which processor variant is the target.
24760 @item -mdisable-callt
24761 @itemx -mno-disable-callt
24762 @opindex mdisable-callt
24763 @opindex mno-disable-callt
24764 This option suppresses generation of the @code{CALLT} instruction for the
24765 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
24768 This option is enabled by default when the RH850 ABI is
24769 in use (see @option{-mrh850-abi}), and disabled by default when the
24770 GCC ABI is in use. If @code{CALLT} instructions are being generated
24771 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
24777 Pass on (or do not pass on) the @option{-mrelax} command-line option
24781 @itemx -mno-long-jumps
24782 @opindex mlong-jumps
24783 @opindex mno-long-jumps
24784 Disable (or re-enable) the generation of PC-relative jump instructions.
24787 @itemx -mhard-float
24788 @opindex msoft-float
24789 @opindex mhard-float
24790 Disable (or re-enable) the generation of hardware floating point
24791 instructions. This option is only significant when the target
24792 architecture is @samp{V850E2V3} or higher. If hardware floating point
24793 instructions are being generated then the C preprocessor symbol
24794 @code{__FPU_OK__} is defined, otherwise the symbol
24795 @code{__NO_FPU__} is defined.
24799 Enables the use of the e3v5 LOOP instruction. The use of this
24800 instruction is not enabled by default when the e3v5 architecture is
24801 selected because its use is still experimental.
24805 @opindex mrh850-abi
24807 Enables support for the RH850 version of the V850 ABI. This is the
24808 default. With this version of the ABI the following rules apply:
24812 Integer sized structures and unions are returned via a memory pointer
24813 rather than a register.
24816 Large structures and unions (more than 8 bytes in size) are passed by
24820 Functions are aligned to 16-bit boundaries.
24823 The @option{-m8byte-align} command-line option is supported.
24826 The @option{-mdisable-callt} command-line option is enabled by
24827 default. The @option{-mno-disable-callt} command-line option is not
24831 When this version of the ABI is enabled the C preprocessor symbol
24832 @code{__V850_RH850_ABI__} is defined.
24836 Enables support for the old GCC version of the V850 ABI. With this
24837 version of the ABI the following rules apply:
24841 Integer sized structures and unions are returned in register @code{r10}.
24844 Large structures and unions (more than 8 bytes in size) are passed by
24848 Functions are aligned to 32-bit boundaries, unless optimizing for
24852 The @option{-m8byte-align} command-line option is not supported.
24855 The @option{-mdisable-callt} command-line option is supported but not
24856 enabled by default.
24859 When this version of the ABI is enabled the C preprocessor symbol
24860 @code{__V850_GCC_ABI__} is defined.
24862 @item -m8byte-align
24863 @itemx -mno-8byte-align
24864 @opindex m8byte-align
24865 @opindex mno-8byte-align
24866 Enables support for @code{double} and @code{long long} types to be
24867 aligned on 8-byte boundaries. The default is to restrict the
24868 alignment of all objects to at most 4-bytes. When
24869 @option{-m8byte-align} is in effect the C preprocessor symbol
24870 @code{__V850_8BYTE_ALIGN__} is defined.
24873 @opindex mbig-switch
24874 Generate code suitable for big switch tables. Use this option only if
24875 the assembler/linker complain about out of range branches within a switch
24880 This option causes r2 and r5 to be used in the code generated by
24881 the compiler. This setting is the default.
24883 @item -mno-app-regs
24884 @opindex mno-app-regs
24885 This option causes r2 and r5 to be treated as fixed registers.
24890 @subsection VAX Options
24891 @cindex VAX options
24893 These @samp{-m} options are defined for the VAX:
24898 Do not output certain jump instructions (@code{aobleq} and so on)
24899 that the Unix assembler for the VAX cannot handle across long
24904 Do output those jump instructions, on the assumption that the
24905 GNU assembler is being used.
24909 Output code for G-format floating-point numbers instead of D-format.
24912 @node Visium Options
24913 @subsection Visium Options
24914 @cindex Visium options
24920 A program which performs file I/O and is destined to run on an MCM target
24921 should be linked with this option. It causes the libraries libc.a and
24922 libdebug.a to be linked. The program should be run on the target under
24923 the control of the GDB remote debugging stub.
24927 A program which performs file I/O and is destined to run on the simulator
24928 should be linked with option. This causes libraries libc.a and libsim.a to
24932 @itemx -mhard-float
24934 @opindex mhard-float
24935 Generate code containing floating-point instructions. This is the
24939 @itemx -msoft-float
24941 @opindex msoft-float
24942 Generate code containing library calls for floating-point.
24944 @option{-msoft-float} changes the calling convention in the output file;
24945 therefore, it is only useful if you compile @emph{all} of a program with
24946 this option. In particular, you need to compile @file{libgcc.a}, the
24947 library that comes with GCC, with @option{-msoft-float} in order for
24950 @item -mcpu=@var{cpu_type}
24952 Set the instruction set, register set, and instruction scheduling parameters
24953 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24954 @samp{mcm}, @samp{gr5} and @samp{gr6}.
24956 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
24958 By default (unless configured otherwise), GCC generates code for the GR5
24959 variant of the Visium architecture.
24961 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
24962 architecture. The only difference from GR5 code is that the compiler will
24963 generate block move instructions.
24965 @item -mtune=@var{cpu_type}
24967 Set the instruction scheduling parameters for machine type @var{cpu_type},
24968 but do not set the instruction set or register set that the option
24969 @option{-mcpu=@var{cpu_type}} would.
24973 Generate code for the supervisor mode, where there are no restrictions on
24974 the access to general registers. This is the default.
24977 @opindex muser-mode
24978 Generate code for the user mode, where the access to some general registers
24979 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
24980 mode; on the GR6, only registers r29 to r31 are affected.
24984 @subsection VMS Options
24986 These @samp{-m} options are defined for the VMS implementations:
24989 @item -mvms-return-codes
24990 @opindex mvms-return-codes
24991 Return VMS condition codes from @code{main}. The default is to return POSIX-style
24992 condition (e.g.@ error) codes.
24994 @item -mdebug-main=@var{prefix}
24995 @opindex mdebug-main=@var{prefix}
24996 Flag the first routine whose name starts with @var{prefix} as the main
24997 routine for the debugger.
25001 Default to 64-bit memory allocation routines.
25003 @item -mpointer-size=@var{size}
25004 @opindex mpointer-size=@var{size}
25005 Set the default size of pointers. Possible options for @var{size} are
25006 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
25007 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
25008 The later option disables @code{pragma pointer_size}.
25011 @node VxWorks Options
25012 @subsection VxWorks Options
25013 @cindex VxWorks Options
25015 The options in this section are defined for all VxWorks targets.
25016 Options specific to the target hardware are listed with the other
25017 options for that target.
25022 GCC can generate code for both VxWorks kernels and real time processes
25023 (RTPs). This option switches from the former to the latter. It also
25024 defines the preprocessor macro @code{__RTP__}.
25027 @opindex non-static
25028 Link an RTP executable against shared libraries rather than static
25029 libraries. The options @option{-static} and @option{-shared} can
25030 also be used for RTPs (@pxref{Link Options}); @option{-static}
25037 These options are passed down to the linker. They are defined for
25038 compatibility with Diab.
25041 @opindex Xbind-lazy
25042 Enable lazy binding of function calls. This option is equivalent to
25043 @option{-Wl,-z,now} and is defined for compatibility with Diab.
25047 Disable lazy binding of function calls. This option is the default and
25048 is defined for compatibility with Diab.
25052 @subsection x86 Options
25053 @cindex x86 Options
25055 These @samp{-m} options are defined for the x86 family of computers.
25059 @item -march=@var{cpu-type}
25061 Generate instructions for the machine type @var{cpu-type}. In contrast to
25062 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
25063 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
25064 to generate code that may not run at all on processors other than the one
25065 indicated. Specifying @option{-march=@var{cpu-type}} implies
25066 @option{-mtune=@var{cpu-type}}.
25068 The choices for @var{cpu-type} are:
25072 This selects the CPU to generate code for at compilation time by determining
25073 the processor type of the compiling machine. Using @option{-march=native}
25074 enables all instruction subsets supported by the local machine (hence
25075 the result might not run on different machines). Using @option{-mtune=native}
25076 produces code optimized for the local machine under the constraints
25077 of the selected instruction set.
25080 Original Intel i386 CPU@.
25083 Intel i486 CPU@. (No scheduling is implemented for this chip.)
25087 Intel Pentium CPU with no MMX support.
25090 Intel Lakemont MCU, based on Intel Pentium CPU.
25093 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
25096 Intel Pentium Pro CPU@.
25099 When used with @option{-march}, the Pentium Pro
25100 instruction set is used, so the code runs on all i686 family chips.
25101 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
25104 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
25109 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
25113 Intel Pentium M; low-power version of Intel Pentium III CPU
25114 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
25118 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
25121 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
25125 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
25126 SSE2 and SSE3 instruction set support.
25129 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
25130 instruction set support.
25133 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25134 SSE4.1, SSE4.2 and POPCNT instruction set support.
25137 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25138 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
25141 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25142 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
25145 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25146 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
25147 instruction set support.
25150 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25151 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25152 BMI, BMI2 and F16C instruction set support.
25155 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25156 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25157 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
25160 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25161 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25162 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
25163 XSAVES instruction set support.
25166 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
25167 instruction set support.
25170 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25171 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
25174 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25175 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25176 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
25177 AVX512CD instruction set support.
25180 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25181 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25182 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25183 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
25185 @item skylake-avx512
25186 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25187 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25188 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
25189 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
25192 AMD K6 CPU with MMX instruction set support.
25196 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
25199 @itemx athlon-tbird
25200 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
25206 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
25207 instruction set support.
25213 Processors based on the AMD K8 core with x86-64 instruction set support,
25214 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
25215 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
25216 instruction set extensions.)
25219 @itemx opteron-sse3
25220 @itemx athlon64-sse3
25221 Improved versions of AMD K8 cores with SSE3 instruction set support.
25225 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
25226 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
25227 instruction set extensions.)
25230 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
25231 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
25232 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
25234 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25235 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
25236 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
25239 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25240 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
25241 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
25242 64-bit instruction set extensions.
25244 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25245 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
25246 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
25247 SSE4.2, ABM and 64-bit instruction set extensions.
25250 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
25251 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
25252 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
25253 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
25254 instruction set extensions.
25257 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
25258 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
25259 instruction set extensions.)
25262 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
25263 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
25264 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
25267 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
25271 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
25272 instruction set support.
25275 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
25276 (No scheduling is implemented for this chip.)
25279 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
25280 (No scheduling is implemented for this chip.)
25283 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25284 (No scheduling is implemented for this chip.)
25287 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
25288 (No scheduling is implemented for this chip.)
25291 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
25292 (No scheduling is implemented for this chip.)
25295 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25296 (No scheduling is implemented for this chip.)
25299 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
25300 (No scheduling is implemented for this chip.)
25303 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
25304 AVX and AVX2 instruction set support.
25305 (No scheduling is implemented for this chip.)
25308 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25309 instruction set support.
25310 (No scheduling is implemented for this chip.)
25313 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25314 instruction set support.
25315 (No scheduling is implemented for this chip.)
25318 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25319 instruction set support.
25320 (No scheduling is implemented for this chip.)
25323 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25324 instruction set support.
25325 (No scheduling is implemented for this chip.)
25328 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25329 instruction set support.
25330 (No scheduling is implemented for this chip.)
25333 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25334 instruction set support.
25335 (No scheduling is implemented for this chip.)
25338 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
25341 @item -mtune=@var{cpu-type}
25343 Tune to @var{cpu-type} everything applicable about the generated code, except
25344 for the ABI and the set of available instructions.
25345 While picking a specific @var{cpu-type} schedules things appropriately
25346 for that particular chip, the compiler does not generate any code that
25347 cannot run on the default machine type unless you use a
25348 @option{-march=@var{cpu-type}} option.
25349 For example, if GCC is configured for i686-pc-linux-gnu
25350 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
25351 but still runs on i686 machines.
25353 The choices for @var{cpu-type} are the same as for @option{-march}.
25354 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
25358 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
25359 If you know the CPU on which your code will run, then you should use
25360 the corresponding @option{-mtune} or @option{-march} option instead of
25361 @option{-mtune=generic}. But, if you do not know exactly what CPU users
25362 of your application will have, then you should use this option.
25364 As new processors are deployed in the marketplace, the behavior of this
25365 option will change. Therefore, if you upgrade to a newer version of
25366 GCC, code generation controlled by this option will change to reflect
25368 that are most common at the time that version of GCC is released.
25370 There is no @option{-march=generic} option because @option{-march}
25371 indicates the instruction set the compiler can use, and there is no
25372 generic instruction set applicable to all processors. In contrast,
25373 @option{-mtune} indicates the processor (or, in this case, collection of
25374 processors) for which the code is optimized.
25377 Produce code optimized for the most current Intel processors, which are
25378 Haswell and Silvermont for this version of GCC. If you know the CPU
25379 on which your code will run, then you should use the corresponding
25380 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
25381 But, if you want your application performs better on both Haswell and
25382 Silvermont, then you should use this option.
25384 As new Intel processors are deployed in the marketplace, the behavior of
25385 this option will change. Therefore, if you upgrade to a newer version of
25386 GCC, code generation controlled by this option will change to reflect
25387 the most current Intel processors at the time that version of GCC is
25390 There is no @option{-march=intel} option because @option{-march} indicates
25391 the instruction set the compiler can use, and there is no common
25392 instruction set applicable to all processors. In contrast,
25393 @option{-mtune} indicates the processor (or, in this case, collection of
25394 processors) for which the code is optimized.
25397 @item -mcpu=@var{cpu-type}
25399 A deprecated synonym for @option{-mtune}.
25401 @item -mfpmath=@var{unit}
25403 Generate floating-point arithmetic for selected unit @var{unit}. The choices
25404 for @var{unit} are:
25408 Use the standard 387 floating-point coprocessor present on the majority of chips and
25409 emulated otherwise. Code compiled with this option runs almost everywhere.
25410 The temporary results are computed in 80-bit precision instead of the precision
25411 specified by the type, resulting in slightly different results compared to most
25412 of other chips. See @option{-ffloat-store} for more detailed description.
25414 This is the default choice for non-Darwin x86-32 targets.
25417 Use scalar floating-point instructions present in the SSE instruction set.
25418 This instruction set is supported by Pentium III and newer chips,
25419 and in the AMD line
25420 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
25421 instruction set supports only single-precision arithmetic, thus the double and
25422 extended-precision arithmetic are still done using 387. A later version, present
25423 only in Pentium 4 and AMD x86-64 chips, supports double-precision
25426 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
25427 or @option{-msse2} switches to enable SSE extensions and make this option
25428 effective. For the x86-64 compiler, these extensions are enabled by default.
25430 The resulting code should be considerably faster in the majority of cases and avoid
25431 the numerical instability problems of 387 code, but may break some existing
25432 code that expects temporaries to be 80 bits.
25434 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
25435 and the default choice for x86-32 targets with the SSE2 instruction set
25436 when @option{-ffast-math} is enabled.
25441 Attempt to utilize both instruction sets at once. This effectively doubles the
25442 amount of available registers, and on chips with separate execution units for
25443 387 and SSE the execution resources too. Use this option with care, as it is
25444 still experimental, because the GCC register allocator does not model separate
25445 functional units well, resulting in unstable performance.
25448 @item -masm=@var{dialect}
25449 @opindex masm=@var{dialect}
25450 Output assembly instructions using selected @var{dialect}. Also affects
25451 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
25452 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
25453 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
25454 not support @samp{intel}.
25457 @itemx -mno-ieee-fp
25459 @opindex mno-ieee-fp
25460 Control whether or not the compiler uses IEEE floating-point
25461 comparisons. These correctly handle the case where the result of a
25462 comparison is unordered.
25467 @opindex mhard-float
25468 Generate output containing 80387 instructions for floating point.
25473 @opindex msoft-float
25474 Generate output containing library calls for floating point.
25476 @strong{Warning:} the requisite libraries are not part of GCC@.
25477 Normally the facilities of the machine's usual C compiler are used, but
25478 this cannot be done directly in cross-compilation. You must make your
25479 own arrangements to provide suitable library functions for
25482 On machines where a function returns floating-point results in the 80387
25483 register stack, some floating-point opcodes may be emitted even if
25484 @option{-msoft-float} is used.
25486 @item -mno-fp-ret-in-387
25487 @opindex mno-fp-ret-in-387
25488 Do not use the FPU registers for return values of functions.
25490 The usual calling convention has functions return values of types
25491 @code{float} and @code{double} in an FPU register, even if there
25492 is no FPU@. The idea is that the operating system should emulate
25495 The option @option{-mno-fp-ret-in-387} causes such values to be returned
25496 in ordinary CPU registers instead.
25498 @item -mno-fancy-math-387
25499 @opindex mno-fancy-math-387
25500 Some 387 emulators do not support the @code{sin}, @code{cos} and
25501 @code{sqrt} instructions for the 387. Specify this option to avoid
25502 generating those instructions. This option is the default on
25503 OpenBSD and NetBSD@. This option is overridden when @option{-march}
25504 indicates that the target CPU always has an FPU and so the
25505 instruction does not need emulation. These
25506 instructions are not generated unless you also use the
25507 @option{-funsafe-math-optimizations} switch.
25509 @item -malign-double
25510 @itemx -mno-align-double
25511 @opindex malign-double
25512 @opindex mno-align-double
25513 Control whether GCC aligns @code{double}, @code{long double}, and
25514 @code{long long} variables on a two-word boundary or a one-word
25515 boundary. Aligning @code{double} variables on a two-word boundary
25516 produces code that runs somewhat faster on a Pentium at the
25517 expense of more memory.
25519 On x86-64, @option{-malign-double} is enabled by default.
25521 @strong{Warning:} if you use the @option{-malign-double} switch,
25522 structures containing the above types are aligned differently than
25523 the published application binary interface specifications for the x86-32
25524 and are not binary compatible with structures in code compiled
25525 without that switch.
25527 @item -m96bit-long-double
25528 @itemx -m128bit-long-double
25529 @opindex m96bit-long-double
25530 @opindex m128bit-long-double
25531 These switches control the size of @code{long double} type. The x86-32
25532 application binary interface specifies the size to be 96 bits,
25533 so @option{-m96bit-long-double} is the default in 32-bit mode.
25535 Modern architectures (Pentium and newer) prefer @code{long double}
25536 to be aligned to an 8- or 16-byte boundary. In arrays or structures
25537 conforming to the ABI, this is not possible. So specifying
25538 @option{-m128bit-long-double} aligns @code{long double}
25539 to a 16-byte boundary by padding the @code{long double} with an additional
25542 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
25543 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
25545 Notice that neither of these options enable any extra precision over the x87
25546 standard of 80 bits for a @code{long double}.
25548 @strong{Warning:} if you override the default value for your target ABI, this
25549 changes the size of
25550 structures and arrays containing @code{long double} variables,
25551 as well as modifying the function calling convention for functions taking
25552 @code{long double}. Hence they are not binary-compatible
25553 with code compiled without that switch.
25555 @item -mlong-double-64
25556 @itemx -mlong-double-80
25557 @itemx -mlong-double-128
25558 @opindex mlong-double-64
25559 @opindex mlong-double-80
25560 @opindex mlong-double-128
25561 These switches control the size of @code{long double} type. A size
25562 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25563 type. This is the default for 32-bit Bionic C library. A size
25564 of 128 bits makes the @code{long double} type equivalent to the
25565 @code{__float128} type. This is the default for 64-bit Bionic C library.
25567 @strong{Warning:} if you override the default value for your target ABI, this
25568 changes the size of
25569 structures and arrays containing @code{long double} variables,
25570 as well as modifying the function calling convention for functions taking
25571 @code{long double}. Hence they are not binary-compatible
25572 with code compiled without that switch.
25574 @item -malign-data=@var{type}
25575 @opindex malign-data
25576 Control how GCC aligns variables. Supported values for @var{type} are
25577 @samp{compat} uses increased alignment value compatible uses GCC 4.8
25578 and earlier, @samp{abi} uses alignment value as specified by the
25579 psABI, and @samp{cacheline} uses increased alignment value to match
25580 the cache line size. @samp{compat} is the default.
25582 @item -mlarge-data-threshold=@var{threshold}
25583 @opindex mlarge-data-threshold
25584 When @option{-mcmodel=medium} is specified, data objects larger than
25585 @var{threshold} are placed in the large data section. This value must be the
25586 same across all objects linked into the binary, and defaults to 65535.
25590 Use a different function-calling convention, in which functions that
25591 take a fixed number of arguments return with the @code{ret @var{num}}
25592 instruction, which pops their arguments while returning. This saves one
25593 instruction in the caller since there is no need to pop the arguments
25596 You can specify that an individual function is called with this calling
25597 sequence with the function attribute @code{stdcall}. You can also
25598 override the @option{-mrtd} option by using the function attribute
25599 @code{cdecl}. @xref{Function Attributes}.
25601 @strong{Warning:} this calling convention is incompatible with the one
25602 normally used on Unix, so you cannot use it if you need to call
25603 libraries compiled with the Unix compiler.
25605 Also, you must provide function prototypes for all functions that
25606 take variable numbers of arguments (including @code{printf});
25607 otherwise incorrect code is generated for calls to those
25610 In addition, seriously incorrect code results if you call a
25611 function with too many arguments. (Normally, extra arguments are
25612 harmlessly ignored.)
25614 @item -mregparm=@var{num}
25616 Control how many registers are used to pass integer arguments. By
25617 default, no registers are used to pass arguments, and at most 3
25618 registers can be used. You can control this behavior for a specific
25619 function by using the function attribute @code{regparm}.
25620 @xref{Function Attributes}.
25622 @strong{Warning:} if you use this switch, and
25623 @var{num} is nonzero, then you must build all modules with the same
25624 value, including any libraries. This includes the system libraries and
25628 @opindex msseregparm
25629 Use SSE register passing conventions for float and double arguments
25630 and return values. You can control this behavior for a specific
25631 function by using the function attribute @code{sseregparm}.
25632 @xref{Function Attributes}.
25634 @strong{Warning:} if you use this switch then you must build all
25635 modules with the same value, including any libraries. This includes
25636 the system libraries and startup modules.
25638 @item -mvect8-ret-in-mem
25639 @opindex mvect8-ret-in-mem
25640 Return 8-byte vectors in memory instead of MMX registers. This is the
25641 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
25642 Studio compilers until version 12. Later compiler versions (starting
25643 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
25644 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
25645 you need to remain compatible with existing code produced by those
25646 previous compiler versions or older versions of GCC@.
25655 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
25656 is specified, the significands of results of floating-point operations are
25657 rounded to 24 bits (single precision); @option{-mpc64} rounds the
25658 significands of results of floating-point operations to 53 bits (double
25659 precision) and @option{-mpc80} rounds the significands of results of
25660 floating-point operations to 64 bits (extended double precision), which is
25661 the default. When this option is used, floating-point operations in higher
25662 precisions are not available to the programmer without setting the FPU
25663 control word explicitly.
25665 Setting the rounding of floating-point operations to less than the default
25666 80 bits can speed some programs by 2% or more. Note that some mathematical
25667 libraries assume that extended-precision (80-bit) floating-point operations
25668 are enabled by default; routines in such libraries could suffer significant
25669 loss of accuracy, typically through so-called ``catastrophic cancellation'',
25670 when this option is used to set the precision to less than extended precision.
25672 @item -mstackrealign
25673 @opindex mstackrealign
25674 Realign the stack at entry. On the x86, the @option{-mstackrealign}
25675 option generates an alternate prologue and epilogue that realigns the
25676 run-time stack if necessary. This supports mixing legacy codes that keep
25677 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
25678 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
25679 applicable to individual functions.
25681 @item -mpreferred-stack-boundary=@var{num}
25682 @opindex mpreferred-stack-boundary
25683 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
25684 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
25685 the default is 4 (16 bytes or 128 bits).
25687 @strong{Warning:} When generating code for the x86-64 architecture with
25688 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
25689 used to keep the stack boundary aligned to 8 byte boundary. Since
25690 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
25691 intended to be used in controlled environment where stack space is
25692 important limitation. This option leads to wrong code when functions
25693 compiled with 16 byte stack alignment (such as functions from a standard
25694 library) are called with misaligned stack. In this case, SSE
25695 instructions may lead to misaligned memory access traps. In addition,
25696 variable arguments are handled incorrectly for 16 byte aligned
25697 objects (including x87 long double and __int128), leading to wrong
25698 results. You must build all modules with
25699 @option{-mpreferred-stack-boundary=3}, including any libraries. This
25700 includes the system libraries and startup modules.
25702 @item -mincoming-stack-boundary=@var{num}
25703 @opindex mincoming-stack-boundary
25704 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
25705 boundary. If @option{-mincoming-stack-boundary} is not specified,
25706 the one specified by @option{-mpreferred-stack-boundary} is used.
25708 On Pentium and Pentium Pro, @code{double} and @code{long double} values
25709 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
25710 suffer significant run time performance penalties. On Pentium III, the
25711 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
25712 properly if it is not 16-byte aligned.
25714 To ensure proper alignment of this values on the stack, the stack boundary
25715 must be as aligned as that required by any value stored on the stack.
25716 Further, every function must be generated such that it keeps the stack
25717 aligned. Thus calling a function compiled with a higher preferred
25718 stack boundary from a function compiled with a lower preferred stack
25719 boundary most likely misaligns the stack. It is recommended that
25720 libraries that use callbacks always use the default setting.
25722 This extra alignment does consume extra stack space, and generally
25723 increases code size. Code that is sensitive to stack space usage, such
25724 as embedded systems and operating system kernels, may want to reduce the
25725 preferred alignment to @option{-mpreferred-stack-boundary=2}.
25782 @itemx -mavx512ifma
25783 @opindex mavx512ifma
25785 @itemx -mavx512vbmi
25786 @opindex mavx512vbmi
25798 @opindex mclfushopt
25815 @itemx -mprefetchwt1
25816 @opindex mprefetchwt1
25879 These switches enable the use of instructions in the MMX, SSE,
25880 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25881 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
25882 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
25883 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, IBT, SHSTK,
25884 3DNow!@: or enhanced 3DNow!@: extended instruction sets. Each has a
25885 corresponding @option{-mno-} option to disable use of these instructions.
25887 These extensions are also available as built-in functions: see
25888 @ref{x86 Built-in Functions}, for details of the functions enabled and
25889 disabled by these switches.
25891 To generate SSE/SSE2 instructions automatically from floating-point
25892 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
25894 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
25895 generates new AVX instructions or AVX equivalence for all SSEx instructions
25898 These options enable GCC to use these extended instructions in
25899 generated code, even without @option{-mfpmath=sse}. Applications that
25900 perform run-time CPU detection must compile separate files for each
25901 supported architecture, using the appropriate flags. In particular,
25902 the file containing the CPU detection code should be compiled without
25905 The @option{-mcet} option turns on the @option{-mibt} and @option{-mshstk}
25906 options. The @option{-mibt} option enables indirect branch tracking support
25907 and the @option{-mshstk} option enables shadow stack support from
25908 Intel Control-flow Enforcement Technology (CET). The compiler also provides
25909 a number of built-in functions for fine-grained control in a CET-based
25910 application. See @xref{x86 Built-in Functions}, for more information.
25912 @item -mdump-tune-features
25913 @opindex mdump-tune-features
25914 This option instructs GCC to dump the names of the x86 performance
25915 tuning features and default settings. The names can be used in
25916 @option{-mtune-ctrl=@var{feature-list}}.
25918 @item -mtune-ctrl=@var{feature-list}
25919 @opindex mtune-ctrl=@var{feature-list}
25920 This option is used to do fine grain control of x86 code generation features.
25921 @var{feature-list} is a comma separated list of @var{feature} names. See also
25922 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
25923 on if it is not preceded with @samp{^}, otherwise, it is turned off.
25924 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
25925 developers. Using it may lead to code paths not covered by testing and can
25926 potentially result in compiler ICEs or runtime errors.
25929 @opindex mno-default
25930 This option instructs GCC to turn off all tunable features. See also
25931 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
25935 This option instructs GCC to emit a @code{cld} instruction in the prologue
25936 of functions that use string instructions. String instructions depend on
25937 the DF flag to select between autoincrement or autodecrement mode. While the
25938 ABI specifies the DF flag to be cleared on function entry, some operating
25939 systems violate this specification by not clearing the DF flag in their
25940 exception dispatchers. The exception handler can be invoked with the DF flag
25941 set, which leads to wrong direction mode when string instructions are used.
25942 This option can be enabled by default on 32-bit x86 targets by configuring
25943 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
25944 instructions can be suppressed with the @option{-mno-cld} compiler option
25948 @opindex mvzeroupper
25949 This option instructs GCC to emit a @code{vzeroupper} instruction
25950 before a transfer of control flow out of the function to minimize
25951 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
25954 @item -mprefer-avx128
25955 @opindex mprefer-avx128
25956 This option instructs GCC to use 128-bit AVX instructions instead of
25957 256-bit AVX instructions in the auto-vectorizer.
25959 @item -mprefer-avx256
25960 @opindex mprefer-avx256
25961 This option instructs GCC to use 256-bit AVX instructions instead of
25962 512-bit AVX instructions in the auto-vectorizer.
25966 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
25967 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
25968 objects. This is useful for atomic updates of data structures exceeding one
25969 machine word in size. The compiler uses this instruction to implement
25970 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
25971 128-bit integers, a library call is always used.
25975 This option enables generation of @code{SAHF} instructions in 64-bit code.
25976 Early Intel Pentium 4 CPUs with Intel 64 support,
25977 prior to the introduction of Pentium 4 G1 step in December 2005,
25978 lacked the @code{LAHF} and @code{SAHF} instructions
25979 which are supported by AMD64.
25980 These are load and store instructions, respectively, for certain status flags.
25981 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
25982 @code{drem}, and @code{remainder} built-in functions;
25983 see @ref{Other Builtins} for details.
25987 This option enables use of the @code{movbe} instruction to implement
25988 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
25992 This option tells the compiler to use indirect branch tracking support
25993 (for indirect calls and jumps) from x86 Control-flow Enforcement
25994 Technology (CET). The option has effect only if the
25995 @option{-fcf-protection=full} or @option{-fcf-protection=branch} option
25996 is specified. The option @option{-mibt} is on by default when the
25997 @code{-mcet} option is specified.
26001 This option tells the compiler to use shadow stack support (return
26002 address tracking) from x86 Control-flow Enforcement Technology (CET).
26003 The option has effect only if the @option{-fcf-protection=full} or
26004 @option{-fcf-protection=return} option is specified. The option
26005 @option{-mshstk} is on by default when the @option{-mcet} option is
26010 This option enables built-in functions @code{__builtin_ia32_crc32qi},
26011 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
26012 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
26016 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
26017 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
26018 with an additional Newton-Raphson step
26019 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
26020 (and their vectorized
26021 variants) for single-precision floating-point arguments. These instructions
26022 are generated only when @option{-funsafe-math-optimizations} is enabled
26023 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
26024 Note that while the throughput of the sequence is higher than the throughput
26025 of the non-reciprocal instruction, the precision of the sequence can be
26026 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
26028 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
26029 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
26030 combination), and doesn't need @option{-mrecip}.
26032 Also note that GCC emits the above sequence with additional Newton-Raphson step
26033 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
26034 already with @option{-ffast-math} (or the above option combination), and
26035 doesn't need @option{-mrecip}.
26037 @item -mrecip=@var{opt}
26038 @opindex mrecip=opt
26039 This option controls which reciprocal estimate instructions
26040 may be used. @var{opt} is a comma-separated list of options, which may
26041 be preceded by a @samp{!} to invert the option:
26045 Enable all estimate instructions.
26048 Enable the default instructions, equivalent to @option{-mrecip}.
26051 Disable all estimate instructions, equivalent to @option{-mno-recip}.
26054 Enable the approximation for scalar division.
26057 Enable the approximation for vectorized division.
26060 Enable the approximation for scalar square root.
26063 Enable the approximation for vectorized square root.
26066 So, for example, @option{-mrecip=all,!sqrt} enables
26067 all of the reciprocal approximations, except for square root.
26069 @item -mveclibabi=@var{type}
26070 @opindex mveclibabi
26071 Specifies the ABI type to use for vectorizing intrinsics using an
26072 external library. Supported values for @var{type} are @samp{svml}
26073 for the Intel short
26074 vector math library and @samp{acml} for the AMD math core library.
26075 To use this option, both @option{-ftree-vectorize} and
26076 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
26077 ABI-compatible library must be specified at link time.
26079 GCC currently emits calls to @code{vmldExp2},
26080 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
26081 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
26082 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
26083 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
26084 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
26085 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
26086 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
26087 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
26088 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
26089 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
26090 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
26091 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
26092 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
26093 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
26094 when @option{-mveclibabi=acml} is used.
26096 @item -mabi=@var{name}
26098 Generate code for the specified calling convention. Permissible values
26099 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
26100 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
26101 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
26102 You can control this behavior for specific functions by
26103 using the function attributes @code{ms_abi} and @code{sysv_abi}.
26104 @xref{Function Attributes}.
26106 @item -mcall-ms2sysv-xlogues
26107 @opindex mcall-ms2sysv-xlogues
26108 @opindex mno-call-ms2sysv-xlogues
26109 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
26110 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
26111 default, the code for saving and restoring these registers is emitted inline,
26112 resulting in fairly lengthy prologues and epilogues. Using
26113 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
26114 use stubs in the static portion of libgcc to perform these saves and restores,
26115 thus reducing function size at the cost of a few extra instructions.
26117 @item -mtls-dialect=@var{type}
26118 @opindex mtls-dialect
26119 Generate code to access thread-local storage using the @samp{gnu} or
26120 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
26121 @samp{gnu2} is more efficient, but it may add compile- and run-time
26122 requirements that cannot be satisfied on all systems.
26125 @itemx -mno-push-args
26126 @opindex mpush-args
26127 @opindex mno-push-args
26128 Use PUSH operations to store outgoing parameters. This method is shorter
26129 and usually equally fast as method using SUB/MOV operations and is enabled
26130 by default. In some cases disabling it may improve performance because of
26131 improved scheduling and reduced dependencies.
26133 @item -maccumulate-outgoing-args
26134 @opindex maccumulate-outgoing-args
26135 If enabled, the maximum amount of space required for outgoing arguments is
26136 computed in the function prologue. This is faster on most modern CPUs
26137 because of reduced dependencies, improved scheduling and reduced stack usage
26138 when the preferred stack boundary is not equal to 2. The drawback is a notable
26139 increase in code size. This switch implies @option{-mno-push-args}.
26143 Support thread-safe exception handling on MinGW. Programs that rely
26144 on thread-safe exception handling must compile and link all code with the
26145 @option{-mthreads} option. When compiling, @option{-mthreads} defines
26146 @option{-D_MT}; when linking, it links in a special thread helper library
26147 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
26149 @item -mms-bitfields
26150 @itemx -mno-ms-bitfields
26151 @opindex mms-bitfields
26152 @opindex mno-ms-bitfields
26154 Enable/disable bit-field layout compatible with the native Microsoft
26157 If @code{packed} is used on a structure, or if bit-fields are used,
26158 it may be that the Microsoft ABI lays out the structure differently
26159 than the way GCC normally does. Particularly when moving packed
26160 data between functions compiled with GCC and the native Microsoft compiler
26161 (either via function call or as data in a file), it may be necessary to access
26164 This option is enabled by default for Microsoft Windows
26165 targets. This behavior can also be controlled locally by use of variable
26166 or type attributes. For more information, see @ref{x86 Variable Attributes}
26167 and @ref{x86 Type Attributes}.
26169 The Microsoft structure layout algorithm is fairly simple with the exception
26170 of the bit-field packing.
26171 The padding and alignment of members of structures and whether a bit-field
26172 can straddle a storage-unit boundary are determine by these rules:
26175 @item Structure members are stored sequentially in the order in which they are
26176 declared: the first member has the lowest memory address and the last member
26179 @item Every data object has an alignment requirement. The alignment requirement
26180 for all data except structures, unions, and arrays is either the size of the
26181 object or the current packing size (specified with either the
26182 @code{aligned} attribute or the @code{pack} pragma),
26183 whichever is less. For structures, unions, and arrays,
26184 the alignment requirement is the largest alignment requirement of its members.
26185 Every object is allocated an offset so that:
26188 offset % alignment_requirement == 0
26191 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
26192 unit if the integral types are the same size and if the next bit-field fits
26193 into the current allocation unit without crossing the boundary imposed by the
26194 common alignment requirements of the bit-fields.
26197 MSVC interprets zero-length bit-fields in the following ways:
26200 @item If a zero-length bit-field is inserted between two bit-fields that
26201 are normally coalesced, the bit-fields are not coalesced.
26208 unsigned long bf_1 : 12;
26210 unsigned long bf_2 : 12;
26215 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
26216 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
26218 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
26219 alignment of the zero-length bit-field is greater than the member that follows it,
26220 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
26241 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
26242 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
26243 bit-field does not affect the alignment of @code{bar} or, as a result, the size
26246 Taking this into account, it is important to note the following:
26249 @item If a zero-length bit-field follows a normal bit-field, the type of the
26250 zero-length bit-field may affect the alignment of the structure as whole. For
26251 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
26252 normal bit-field, and is of type short.
26254 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
26255 still affect the alignment of the structure:
26266 Here, @code{t4} takes up 4 bytes.
26269 @item Zero-length bit-fields following non-bit-field members are ignored:
26281 Here, @code{t5} takes up 2 bytes.
26285 @item -mno-align-stringops
26286 @opindex mno-align-stringops
26287 Do not align the destination of inlined string operations. This switch reduces
26288 code size and improves performance in case the destination is already aligned,
26289 but GCC doesn't know about it.
26291 @item -minline-all-stringops
26292 @opindex minline-all-stringops
26293 By default GCC inlines string operations only when the destination is
26294 known to be aligned to least a 4-byte boundary.
26295 This enables more inlining and increases code
26296 size, but may improve performance of code that depends on fast
26297 @code{memcpy}, @code{strlen},
26298 and @code{memset} for short lengths.
26300 @item -minline-stringops-dynamically
26301 @opindex minline-stringops-dynamically
26302 For string operations of unknown size, use run-time checks with
26303 inline code for small blocks and a library call for large blocks.
26305 @item -mstringop-strategy=@var{alg}
26306 @opindex mstringop-strategy=@var{alg}
26307 Override the internal decision heuristic for the particular algorithm to use
26308 for inlining string operations. The allowed values for @var{alg} are:
26314 Expand using i386 @code{rep} prefix of the specified size.
26318 @itemx unrolled_loop
26319 Expand into an inline loop.
26322 Always use a library call.
26325 @item -mmemcpy-strategy=@var{strategy}
26326 @opindex mmemcpy-strategy=@var{strategy}
26327 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
26328 should be inlined and what inline algorithm to use when the expected size
26329 of the copy operation is known. @var{strategy}
26330 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
26331 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
26332 the max byte size with which inline algorithm @var{alg} is allowed. For the last
26333 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
26334 in the list must be specified in increasing order. The minimal byte size for
26335 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
26338 @item -mmemset-strategy=@var{strategy}
26339 @opindex mmemset-strategy=@var{strategy}
26340 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
26341 @code{__builtin_memset} expansion.
26343 @item -momit-leaf-frame-pointer
26344 @opindex momit-leaf-frame-pointer
26345 Don't keep the frame pointer in a register for leaf functions. This
26346 avoids the instructions to save, set up, and restore frame pointers and
26347 makes an extra register available in leaf functions. The option
26348 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
26349 which might make debugging harder.
26351 @item -mtls-direct-seg-refs
26352 @itemx -mno-tls-direct-seg-refs
26353 @opindex mtls-direct-seg-refs
26354 Controls whether TLS variables may be accessed with offsets from the
26355 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
26356 or whether the thread base pointer must be added. Whether or not this
26357 is valid depends on the operating system, and whether it maps the
26358 segment to cover the entire TLS area.
26360 For systems that use the GNU C Library, the default is on.
26363 @itemx -mno-sse2avx
26365 Specify that the assembler should encode SSE instructions with VEX
26366 prefix. The option @option{-mavx} turns this on by default.
26371 If profiling is active (@option{-pg}), put the profiling
26372 counter call before the prologue.
26373 Note: On x86 architectures the attribute @code{ms_hook_prologue}
26374 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
26376 @item -mrecord-mcount
26377 @itemx -mno-record-mcount
26378 @opindex mrecord-mcount
26379 If profiling is active (@option{-pg}), generate a __mcount_loc section
26380 that contains pointers to each profiling call. This is useful for
26381 automatically patching and out calls.
26384 @itemx -mno-nop-mcount
26385 @opindex mnop-mcount
26386 If profiling is active (@option{-pg}), generate the calls to
26387 the profiling functions as NOPs. This is useful when they
26388 should be patched in later dynamically. This is likely only
26389 useful together with @option{-mrecord-mcount}.
26391 @item -mskip-rax-setup
26392 @itemx -mno-skip-rax-setup
26393 @opindex mskip-rax-setup
26394 When generating code for the x86-64 architecture with SSE extensions
26395 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
26396 register when there are no variable arguments passed in vector registers.
26398 @strong{Warning:} Since RAX register is used to avoid unnecessarily
26399 saving vector registers on stack when passing variable arguments, the
26400 impacts of this option are callees may waste some stack space,
26401 misbehave or jump to a random location. GCC 4.4 or newer don't have
26402 those issues, regardless the RAX register value.
26405 @itemx -mno-8bit-idiv
26406 @opindex m8bit-idiv
26407 On some processors, like Intel Atom, 8-bit unsigned integer divide is
26408 much faster than 32-bit/64-bit integer divide. This option generates a
26409 run-time check. If both dividend and divisor are within range of 0
26410 to 255, 8-bit unsigned integer divide is used instead of
26411 32-bit/64-bit integer divide.
26413 @item -mavx256-split-unaligned-load
26414 @itemx -mavx256-split-unaligned-store
26415 @opindex mavx256-split-unaligned-load
26416 @opindex mavx256-split-unaligned-store
26417 Split 32-byte AVX unaligned load and store.
26419 @item -mstack-protector-guard=@var{guard}
26420 @itemx -mstack-protector-guard-reg=@var{reg}
26421 @itemx -mstack-protector-guard-offset=@var{offset}
26422 @opindex mstack-protector-guard
26423 @opindex mstack-protector-guard-reg
26424 @opindex mstack-protector-guard-offset
26425 Generate stack protection code using canary at @var{guard}. Supported
26426 locations are @samp{global} for global canary or @samp{tls} for per-thread
26427 canary in the TLS block (the default). This option has effect only when
26428 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
26430 With the latter choice the options
26431 @option{-mstack-protector-guard-reg=@var{reg}} and
26432 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
26433 which segment register (@code{%fs} or @code{%gs}) to use as base register
26434 for reading the canary, and from what offset from that base register.
26435 The default for those is as specified in the relevant ABI.
26437 @item -mmitigate-rop
26438 @opindex mmitigate-rop
26439 Try to avoid generating code sequences that contain unintended return
26440 opcodes, to mitigate against certain forms of attack. At the moment,
26441 this option is limited in what it can do and should not be relied
26442 on to provide serious protection.
26444 @item -mgeneral-regs-only
26445 @opindex mgeneral-regs-only
26446 Generate code that uses only the general-purpose registers. This
26447 prevents the compiler from using floating-point, vector, mask and bound
26452 These @samp{-m} switches are supported in addition to the above
26453 on x86-64 processors in 64-bit environments.
26466 Generate code for a 16-bit, 32-bit or 64-bit environment.
26467 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
26469 generates code that runs on any i386 system.
26471 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
26472 types to 64 bits, and generates code for the x86-64 architecture.
26473 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
26474 and @option{-mdynamic-no-pic} options.
26476 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
26478 generates code for the x86-64 architecture.
26480 The @option{-m16} option is the same as @option{-m32}, except for that
26481 it outputs the @code{.code16gcc} assembly directive at the beginning of
26482 the assembly output so that the binary can run in 16-bit mode.
26484 The @option{-miamcu} option generates code which conforms to Intel MCU
26485 psABI. It requires the @option{-m32} option to be turned on.
26487 @item -mno-red-zone
26488 @opindex mno-red-zone
26489 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
26490 by the x86-64 ABI; it is a 128-byte area beyond the location of the
26491 stack pointer that is not modified by signal or interrupt handlers
26492 and therefore can be used for temporary data without adjusting the stack
26493 pointer. The flag @option{-mno-red-zone} disables this red zone.
26495 @item -mcmodel=small
26496 @opindex mcmodel=small
26497 Generate code for the small code model: the program and its symbols must
26498 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
26499 Programs can be statically or dynamically linked. This is the default
26502 @item -mcmodel=kernel
26503 @opindex mcmodel=kernel
26504 Generate code for the kernel code model. The kernel runs in the
26505 negative 2 GB of the address space.
26506 This model has to be used for Linux kernel code.
26508 @item -mcmodel=medium
26509 @opindex mcmodel=medium
26510 Generate code for the medium model: the program is linked in the lower 2
26511 GB of the address space. Small symbols are also placed there. Symbols
26512 with sizes larger than @option{-mlarge-data-threshold} are put into
26513 large data or BSS sections and can be located above 2GB. Programs can
26514 be statically or dynamically linked.
26516 @item -mcmodel=large
26517 @opindex mcmodel=large
26518 Generate code for the large model. This model makes no assumptions
26519 about addresses and sizes of sections.
26521 @item -maddress-mode=long
26522 @opindex maddress-mode=long
26523 Generate code for long address mode. This is only supported for 64-bit
26524 and x32 environments. It is the default address mode for 64-bit
26527 @item -maddress-mode=short
26528 @opindex maddress-mode=short
26529 Generate code for short address mode. This is only supported for 32-bit
26530 and x32 environments. It is the default address mode for 32-bit and
26534 @node x86 Windows Options
26535 @subsection x86 Windows Options
26536 @cindex x86 Windows Options
26537 @cindex Windows Options for x86
26539 These additional options are available for Microsoft Windows targets:
26545 specifies that a console application is to be generated, by
26546 instructing the linker to set the PE header subsystem type
26547 required for console applications.
26548 This option is available for Cygwin and MinGW targets and is
26549 enabled by default on those targets.
26553 This option is available for Cygwin and MinGW targets. It
26554 specifies that a DLL---a dynamic link library---is to be
26555 generated, enabling the selection of the required runtime
26556 startup object and entry point.
26558 @item -mnop-fun-dllimport
26559 @opindex mnop-fun-dllimport
26560 This option is available for Cygwin and MinGW targets. It
26561 specifies that the @code{dllimport} attribute should be ignored.
26565 This option is available for MinGW targets. It specifies
26566 that MinGW-specific thread support is to be used.
26570 This option is available for MinGW-w64 targets. It causes
26571 the @code{UNICODE} preprocessor macro to be predefined, and
26572 chooses Unicode-capable runtime startup code.
26576 This option is available for Cygwin and MinGW targets. It
26577 specifies that the typical Microsoft Windows predefined macros are to
26578 be set in the pre-processor, but does not influence the choice
26579 of runtime library/startup code.
26583 This option is available for Cygwin and MinGW targets. It
26584 specifies that a GUI application is to be generated by
26585 instructing the linker to set the PE header subsystem type
26588 @item -fno-set-stack-executable
26589 @opindex fno-set-stack-executable
26590 This option is available for MinGW targets. It specifies that
26591 the executable flag for the stack used by nested functions isn't
26592 set. This is necessary for binaries running in kernel mode of
26593 Microsoft Windows, as there the User32 API, which is used to set executable
26594 privileges, isn't available.
26596 @item -fwritable-relocated-rdata
26597 @opindex fno-writable-relocated-rdata
26598 This option is available for MinGW and Cygwin targets. It specifies
26599 that relocated-data in read-only section is put into the @code{.data}
26600 section. This is a necessary for older runtimes not supporting
26601 modification of @code{.rdata} sections for pseudo-relocation.
26603 @item -mpe-aligned-commons
26604 @opindex mpe-aligned-commons
26605 This option is available for Cygwin and MinGW targets. It
26606 specifies that the GNU extension to the PE file format that
26607 permits the correct alignment of COMMON variables should be
26608 used when generating code. It is enabled by default if
26609 GCC detects that the target assembler found during configuration
26610 supports the feature.
26613 See also under @ref{x86 Options} for standard options.
26615 @node Xstormy16 Options
26616 @subsection Xstormy16 Options
26617 @cindex Xstormy16 Options
26619 These options are defined for Xstormy16:
26624 Choose startup files and linker script suitable for the simulator.
26627 @node Xtensa Options
26628 @subsection Xtensa Options
26629 @cindex Xtensa Options
26631 These options are supported for Xtensa targets:
26635 @itemx -mno-const16
26637 @opindex mno-const16
26638 Enable or disable use of @code{CONST16} instructions for loading
26639 constant values. The @code{CONST16} instruction is currently not a
26640 standard option from Tensilica. When enabled, @code{CONST16}
26641 instructions are always used in place of the standard @code{L32R}
26642 instructions. The use of @code{CONST16} is enabled by default only if
26643 the @code{L32R} instruction is not available.
26646 @itemx -mno-fused-madd
26647 @opindex mfused-madd
26648 @opindex mno-fused-madd
26649 Enable or disable use of fused multiply/add and multiply/subtract
26650 instructions in the floating-point option. This has no effect if the
26651 floating-point option is not also enabled. Disabling fused multiply/add
26652 and multiply/subtract instructions forces the compiler to use separate
26653 instructions for the multiply and add/subtract operations. This may be
26654 desirable in some cases where strict IEEE 754-compliant results are
26655 required: the fused multiply add/subtract instructions do not round the
26656 intermediate result, thereby producing results with @emph{more} bits of
26657 precision than specified by the IEEE standard. Disabling fused multiply
26658 add/subtract instructions also ensures that the program output is not
26659 sensitive to the compiler's ability to combine multiply and add/subtract
26662 @item -mserialize-volatile
26663 @itemx -mno-serialize-volatile
26664 @opindex mserialize-volatile
26665 @opindex mno-serialize-volatile
26666 When this option is enabled, GCC inserts @code{MEMW} instructions before
26667 @code{volatile} memory references to guarantee sequential consistency.
26668 The default is @option{-mserialize-volatile}. Use
26669 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
26671 @item -mforce-no-pic
26672 @opindex mforce-no-pic
26673 For targets, like GNU/Linux, where all user-mode Xtensa code must be
26674 position-independent code (PIC), this option disables PIC for compiling
26677 @item -mtext-section-literals
26678 @itemx -mno-text-section-literals
26679 @opindex mtext-section-literals
26680 @opindex mno-text-section-literals
26681 These options control the treatment of literal pools. The default is
26682 @option{-mno-text-section-literals}, which places literals in a separate
26683 section in the output file. This allows the literal pool to be placed
26684 in a data RAM/ROM, and it also allows the linker to combine literal
26685 pools from separate object files to remove redundant literals and
26686 improve code size. With @option{-mtext-section-literals}, the literals
26687 are interspersed in the text section in order to keep them as close as
26688 possible to their references. This may be necessary for large assembly
26689 files. Literals for each function are placed right before that function.
26691 @item -mauto-litpools
26692 @itemx -mno-auto-litpools
26693 @opindex mauto-litpools
26694 @opindex mno-auto-litpools
26695 These options control the treatment of literal pools. The default is
26696 @option{-mno-auto-litpools}, which places literals in a separate
26697 section in the output file unless @option{-mtext-section-literals} is
26698 used. With @option{-mauto-litpools} the literals are interspersed in
26699 the text section by the assembler. Compiler does not produce explicit
26700 @code{.literal} directives and loads literals into registers with
26701 @code{MOVI} instructions instead of @code{L32R} to let the assembler
26702 do relaxation and place literals as necessary. This option allows
26703 assembler to create several literal pools per function and assemble
26704 very big functions, which may not be possible with
26705 @option{-mtext-section-literals}.
26707 @item -mtarget-align
26708 @itemx -mno-target-align
26709 @opindex mtarget-align
26710 @opindex mno-target-align
26711 When this option is enabled, GCC instructs the assembler to
26712 automatically align instructions to reduce branch penalties at the
26713 expense of some code density. The assembler attempts to widen density
26714 instructions to align branch targets and the instructions following call
26715 instructions. If there are not enough preceding safe density
26716 instructions to align a target, no widening is performed. The
26717 default is @option{-mtarget-align}. These options do not affect the
26718 treatment of auto-aligned instructions like @code{LOOP}, which the
26719 assembler always aligns, either by widening density instructions or
26720 by inserting NOP instructions.
26723 @itemx -mno-longcalls
26724 @opindex mlongcalls
26725 @opindex mno-longcalls
26726 When this option is enabled, GCC instructs the assembler to translate
26727 direct calls to indirect calls unless it can determine that the target
26728 of a direct call is in the range allowed by the call instruction. This
26729 translation typically occurs for calls to functions in other source
26730 files. Specifically, the assembler translates a direct @code{CALL}
26731 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
26732 The default is @option{-mno-longcalls}. This option should be used in
26733 programs where the call target can potentially be out of range. This
26734 option is implemented in the assembler, not the compiler, so the
26735 assembly code generated by GCC still shows direct call
26736 instructions---look at the disassembled object code to see the actual
26737 instructions. Note that the assembler uses an indirect call for
26738 every cross-file call, not just those that really are out of range.
26741 @node zSeries Options
26742 @subsection zSeries Options
26743 @cindex zSeries options
26745 These are listed under @xref{S/390 and zSeries Options}.
26751 @section Specifying Subprocesses and the Switches to Pass to Them
26754 @command{gcc} is a driver program. It performs its job by invoking a
26755 sequence of other programs to do the work of compiling, assembling and
26756 linking. GCC interprets its command-line parameters and uses these to
26757 deduce which programs it should invoke, and which command-line options
26758 it ought to place on their command lines. This behavior is controlled
26759 by @dfn{spec strings}. In most cases there is one spec string for each
26760 program that GCC can invoke, but a few programs have multiple spec
26761 strings to control their behavior. The spec strings built into GCC can
26762 be overridden by using the @option{-specs=} command-line switch to specify
26765 @dfn{Spec files} are plain-text files that are used to construct spec
26766 strings. They consist of a sequence of directives separated by blank
26767 lines. The type of directive is determined by the first non-whitespace
26768 character on the line, which can be one of the following:
26771 @item %@var{command}
26772 Issues a @var{command} to the spec file processor. The commands that can
26776 @item %include <@var{file}>
26777 @cindex @code{%include}
26778 Search for @var{file} and insert its text at the current point in the
26781 @item %include_noerr <@var{file}>
26782 @cindex @code{%include_noerr}
26783 Just like @samp{%include}, but do not generate an error message if the include
26784 file cannot be found.
26786 @item %rename @var{old_name} @var{new_name}
26787 @cindex @code{%rename}
26788 Rename the spec string @var{old_name} to @var{new_name}.
26792 @item *[@var{spec_name}]:
26793 This tells the compiler to create, override or delete the named spec
26794 string. All lines after this directive up to the next directive or
26795 blank line are considered to be the text for the spec string. If this
26796 results in an empty string then the spec is deleted. (Or, if the
26797 spec did not exist, then nothing happens.) Otherwise, if the spec
26798 does not currently exist a new spec is created. If the spec does
26799 exist then its contents are overridden by the text of this
26800 directive, unless the first character of that text is the @samp{+}
26801 character, in which case the text is appended to the spec.
26803 @item [@var{suffix}]:
26804 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
26805 and up to the next directive or blank line are considered to make up the
26806 spec string for the indicated suffix. When the compiler encounters an
26807 input file with the named suffix, it processes the spec string in
26808 order to work out how to compile that file. For example:
26812 z-compile -input %i
26815 This says that any input file whose name ends in @samp{.ZZ} should be
26816 passed to the program @samp{z-compile}, which should be invoked with the
26817 command-line switch @option{-input} and with the result of performing the
26818 @samp{%i} substitution. (See below.)
26820 As an alternative to providing a spec string, the text following a
26821 suffix directive can be one of the following:
26824 @item @@@var{language}
26825 This says that the suffix is an alias for a known @var{language}. This is
26826 similar to using the @option{-x} command-line switch to GCC to specify a
26827 language explicitly. For example:
26834 Says that .ZZ files are, in fact, C++ source files.
26837 This causes an error messages saying:
26840 @var{name} compiler not installed on this system.
26844 GCC already has an extensive list of suffixes built into it.
26845 This directive adds an entry to the end of the list of suffixes, but
26846 since the list is searched from the end backwards, it is effectively
26847 possible to override earlier entries using this technique.
26851 GCC has the following spec strings built into it. Spec files can
26852 override these strings or create their own. Note that individual
26853 targets can also add their own spec strings to this list.
26856 asm Options to pass to the assembler
26857 asm_final Options to pass to the assembler post-processor
26858 cpp Options to pass to the C preprocessor
26859 cc1 Options to pass to the C compiler
26860 cc1plus Options to pass to the C++ compiler
26861 endfile Object files to include at the end of the link
26862 link Options to pass to the linker
26863 lib Libraries to include on the command line to the linker
26864 libgcc Decides which GCC support library to pass to the linker
26865 linker Sets the name of the linker
26866 predefines Defines to be passed to the C preprocessor
26867 signed_char Defines to pass to CPP to say whether @code{char} is signed
26869 startfile Object files to include at the start of the link
26872 Here is a small example of a spec file:
26875 %rename lib old_lib
26878 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
26881 This example renames the spec called @samp{lib} to @samp{old_lib} and
26882 then overrides the previous definition of @samp{lib} with a new one.
26883 The new definition adds in some extra command-line options before
26884 including the text of the old definition.
26886 @dfn{Spec strings} are a list of command-line options to be passed to their
26887 corresponding program. In addition, the spec strings can contain
26888 @samp{%}-prefixed sequences to substitute variable text or to
26889 conditionally insert text into the command line. Using these constructs
26890 it is possible to generate quite complex command lines.
26892 Here is a table of all defined @samp{%}-sequences for spec
26893 strings. Note that spaces are not generated automatically around the
26894 results of expanding these sequences. Therefore you can concatenate them
26895 together or combine them with constant text in a single argument.
26899 Substitute one @samp{%} into the program name or argument.
26902 Substitute the name of the input file being processed.
26905 Substitute the basename of the input file being processed.
26906 This is the substring up to (and not including) the last period
26907 and not including the directory.
26910 This is the same as @samp{%b}, but include the file suffix (text after
26914 Marks the argument containing or following the @samp{%d} as a
26915 temporary file name, so that that file is deleted if GCC exits
26916 successfully. Unlike @samp{%g}, this contributes no text to the
26919 @item %g@var{suffix}
26920 Substitute a file name that has suffix @var{suffix} and is chosen
26921 once per compilation, and mark the argument in the same way as
26922 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
26923 name is now chosen in a way that is hard to predict even when previously
26924 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
26925 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
26926 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
26927 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
26928 was simply substituted with a file name chosen once per compilation,
26929 without regard to any appended suffix (which was therefore treated
26930 just like ordinary text), making such attacks more likely to succeed.
26932 @item %u@var{suffix}
26933 Like @samp{%g}, but generates a new temporary file name
26934 each time it appears instead of once per compilation.
26936 @item %U@var{suffix}
26937 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
26938 new one if there is no such last file name. In the absence of any
26939 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
26940 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
26941 involves the generation of two distinct file names, one
26942 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
26943 simply substituted with a file name chosen for the previous @samp{%u},
26944 without regard to any appended suffix.
26946 @item %j@var{suffix}
26947 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
26948 writable, and if @option{-save-temps} is not used;
26949 otherwise, substitute the name
26950 of a temporary file, just like @samp{%u}. This temporary file is not
26951 meant for communication between processes, but rather as a junk
26952 disposal mechanism.
26954 @item %|@var{suffix}
26955 @itemx %m@var{suffix}
26956 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
26957 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
26958 all. These are the two most common ways to instruct a program that it
26959 should read from standard input or write to standard output. If you
26960 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
26961 construct: see for example @file{f/lang-specs.h}.
26963 @item %.@var{SUFFIX}
26964 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
26965 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
26966 terminated by the next space or %.
26969 Marks the argument containing or following the @samp{%w} as the
26970 designated output file of this compilation. This puts the argument
26971 into the sequence of arguments that @samp{%o} substitutes.
26974 Substitutes the names of all the output files, with spaces
26975 automatically placed around them. You should write spaces
26976 around the @samp{%o} as well or the results are undefined.
26977 @samp{%o} is for use in the specs for running the linker.
26978 Input files whose names have no recognized suffix are not compiled
26979 at all, but they are included among the output files, so they are
26983 Substitutes the suffix for object files. Note that this is
26984 handled specially when it immediately follows @samp{%g, %u, or %U},
26985 because of the need for those to form complete file names. The
26986 handling is such that @samp{%O} is treated exactly as if it had already
26987 been substituted, except that @samp{%g, %u, and %U} do not currently
26988 support additional @var{suffix} characters following @samp{%O} as they do
26989 following, for example, @samp{.o}.
26992 Substitutes the standard macro predefinitions for the
26993 current target machine. Use this when running @command{cpp}.
26996 Like @samp{%p}, but puts @samp{__} before and after the name of each
26997 predefined macro, except for macros that start with @samp{__} or with
26998 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
27002 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
27003 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
27004 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
27005 and @option{-imultilib} as necessary.
27008 Current argument is the name of a library or startup file of some sort.
27009 Search for that file in a standard list of directories and substitute
27010 the full name found. The current working directory is included in the
27011 list of directories scanned.
27014 Current argument is the name of a linker script. Search for that file
27015 in the current list of directories to scan for libraries. If the file
27016 is located insert a @option{--script} option into the command line
27017 followed by the full path name found. If the file is not found then
27018 generate an error message. Note: the current working directory is not
27022 Print @var{str} as an error message. @var{str} is terminated by a newline.
27023 Use this when inconsistent options are detected.
27025 @item %(@var{name})
27026 Substitute the contents of spec string @var{name} at this point.
27028 @item %x@{@var{option}@}
27029 Accumulate an option for @samp{%X}.
27032 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
27036 Output the accumulated assembler options specified by @option{-Wa}.
27039 Output the accumulated preprocessor options specified by @option{-Wp}.
27042 Process the @code{asm} spec. This is used to compute the
27043 switches to be passed to the assembler.
27046 Process the @code{asm_final} spec. This is a spec string for
27047 passing switches to an assembler post-processor, if such a program is
27051 Process the @code{link} spec. This is the spec for computing the
27052 command line passed to the linker. Typically it makes use of the
27053 @samp{%L %G %S %D and %E} sequences.
27056 Dump out a @option{-L} option for each directory that GCC believes might
27057 contain startup files. If the target supports multilibs then the
27058 current multilib directory is prepended to each of these paths.
27061 Process the @code{lib} spec. This is a spec string for deciding which
27062 libraries are included on the command line to the linker.
27065 Process the @code{libgcc} spec. This is a spec string for deciding
27066 which GCC support library is included on the command line to the linker.
27069 Process the @code{startfile} spec. This is a spec for deciding which
27070 object files are the first ones passed to the linker. Typically
27071 this might be a file named @file{crt0.o}.
27074 Process the @code{endfile} spec. This is a spec string that specifies
27075 the last object files that are passed to the linker.
27078 Process the @code{cpp} spec. This is used to construct the arguments
27079 to be passed to the C preprocessor.
27082 Process the @code{cc1} spec. This is used to construct the options to be
27083 passed to the actual C compiler (@command{cc1}).
27086 Process the @code{cc1plus} spec. This is used to construct the options to be
27087 passed to the actual C++ compiler (@command{cc1plus}).
27090 Substitute the variable part of a matched option. See below.
27091 Note that each comma in the substituted string is replaced by
27095 Remove all occurrences of @code{-S} from the command line. Note---this
27096 command is position dependent. @samp{%} commands in the spec string
27097 before this one see @code{-S}, @samp{%} commands in the spec string
27098 after this one do not.
27100 @item %:@var{function}(@var{args})
27101 Call the named function @var{function}, passing it @var{args}.
27102 @var{args} is first processed as a nested spec string, then split
27103 into an argument vector in the usual fashion. The function returns
27104 a string which is processed as if it had appeared literally as part
27105 of the current spec.
27107 The following built-in spec functions are provided:
27110 @item @code{getenv}
27111 The @code{getenv} spec function takes two arguments: an environment
27112 variable name and a string. If the environment variable is not
27113 defined, a fatal error is issued. Otherwise, the return value is the
27114 value of the environment variable concatenated with the string. For
27115 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
27118 %:getenv(TOPDIR /include)
27121 expands to @file{/path/to/top/include}.
27123 @item @code{if-exists}
27124 The @code{if-exists} spec function takes one argument, an absolute
27125 pathname to a file. If the file exists, @code{if-exists} returns the
27126 pathname. Here is a small example of its usage:
27130 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
27133 @item @code{if-exists-else}
27134 The @code{if-exists-else} spec function is similar to the @code{if-exists}
27135 spec function, except that it takes two arguments. The first argument is
27136 an absolute pathname to a file. If the file exists, @code{if-exists-else}
27137 returns the pathname. If it does not exist, it returns the second argument.
27138 This way, @code{if-exists-else} can be used to select one file or another,
27139 based on the existence of the first. Here is a small example of its usage:
27143 crt0%O%s %:if-exists(crti%O%s) \
27144 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
27147 @item @code{replace-outfile}
27148 The @code{replace-outfile} spec function takes two arguments. It looks for the
27149 first argument in the outfiles array and replaces it with the second argument. Here
27150 is a small example of its usage:
27153 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
27156 @item @code{remove-outfile}
27157 The @code{remove-outfile} spec function takes one argument. It looks for the
27158 first argument in the outfiles array and removes it. Here is a small example
27162 %:remove-outfile(-lm)
27165 @item @code{pass-through-libs}
27166 The @code{pass-through-libs} spec function takes any number of arguments. It
27167 finds any @option{-l} options and any non-options ending in @file{.a} (which it
27168 assumes are the names of linker input library archive files) and returns a
27169 result containing all the found arguments each prepended by
27170 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
27171 intended to be passed to the LTO linker plugin.
27174 %:pass-through-libs(%G %L %G)
27177 @item @code{print-asm-header}
27178 The @code{print-asm-header} function takes no arguments and simply
27179 prints a banner like:
27185 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
27188 It is used to separate compiler options from assembler options
27189 in the @option{--target-help} output.
27193 Substitutes the @code{-S} switch, if that switch is given to GCC@.
27194 If that switch is not specified, this substitutes nothing. Note that
27195 the leading dash is omitted when specifying this option, and it is
27196 automatically inserted if the substitution is performed. Thus the spec
27197 string @samp{%@{foo@}} matches the command-line option @option{-foo}
27198 and outputs the command-line option @option{-foo}.
27201 Like %@{@code{S}@} but mark last argument supplied within as a file to be
27202 deleted on failure.
27205 Substitutes all the switches specified to GCC whose names start
27206 with @code{-S}, but which also take an argument. This is used for
27207 switches like @option{-o}, @option{-D}, @option{-I}, etc.
27208 GCC considers @option{-o foo} as being
27209 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
27210 text, including the space. Thus two arguments are generated.
27213 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
27214 (the order of @code{S} and @code{T} in the spec is not significant).
27215 There can be any number of ampersand-separated variables; for each the
27216 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
27219 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
27222 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
27225 Substitutes @code{X} if one or more switches whose names start with
27226 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
27227 once, no matter how many such switches appeared. However, if @code{%*}
27228 appears somewhere in @code{X}, then @code{X} is substituted once
27229 for each matching switch, with the @code{%*} replaced by the part of
27230 that switch matching the @code{*}.
27232 If @code{%*} appears as the last part of a spec sequence then a space
27233 is added after the end of the last substitution. If there is more
27234 text in the sequence, however, then a space is not generated. This
27235 allows the @code{%*} substitution to be used as part of a larger
27236 string. For example, a spec string like this:
27239 %@{mcu=*:--script=%*/memory.ld@}
27243 when matching an option like @option{-mcu=newchip} produces:
27246 --script=newchip/memory.ld
27250 Substitutes @code{X}, if processing a file with suffix @code{S}.
27253 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
27256 Substitutes @code{X}, if processing a file for language @code{S}.
27259 Substitutes @code{X}, if not processing a file for language @code{S}.
27262 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
27263 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
27264 @code{*} sequences as well, although they have a stronger binding than
27265 the @samp{|}. If @code{%*} appears in @code{X}, all of the
27266 alternatives must be starred, and only the first matching alternative
27269 For example, a spec string like this:
27272 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
27276 outputs the following command-line options from the following input
27277 command-line options:
27282 -d fred.c -foo -baz -boggle
27283 -d jim.d -bar -baz -boggle
27286 @item %@{S:X; T:Y; :D@}
27288 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
27289 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
27290 be as many clauses as you need. This may be combined with @code{.},
27291 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
27296 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
27297 or similar construct can use a backslash to ignore the special meaning
27298 of the character following it, thus allowing literal matching of a
27299 character that is otherwise specially treated. For example,
27300 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
27301 @option{-std=iso9899:1999} option is given.
27303 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
27304 construct may contain other nested @samp{%} constructs or spaces, or
27305 even newlines. They are processed as usual, as described above.
27306 Trailing white space in @code{X} is ignored. White space may also
27307 appear anywhere on the left side of the colon in these constructs,
27308 except between @code{.} or @code{*} and the corresponding word.
27310 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
27311 handled specifically in these constructs. If another value of
27312 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
27313 @option{-W} switch is found later in the command line, the earlier
27314 switch value is ignored, except with @{@code{S}*@} where @code{S} is
27315 just one letter, which passes all matching options.
27317 The character @samp{|} at the beginning of the predicate text is used to
27318 indicate that a command should be piped to the following command, but
27319 only if @option{-pipe} is specified.
27321 It is built into GCC which switches take arguments and which do not.
27322 (You might think it would be useful to generalize this to allow each
27323 compiler's spec to say which switches take arguments. But this cannot
27324 be done in a consistent fashion. GCC cannot even decide which input
27325 files have been specified without knowing which switches take arguments,
27326 and it must know which input files to compile in order to tell which
27329 GCC also knows implicitly that arguments starting in @option{-l} are to be
27330 treated as compiler output files, and passed to the linker in their
27331 proper position among the other output files.
27333 @node Environment Variables
27334 @section Environment Variables Affecting GCC
27335 @cindex environment variables
27337 @c man begin ENVIRONMENT
27338 This section describes several environment variables that affect how GCC
27339 operates. Some of them work by specifying directories or prefixes to use
27340 when searching for various kinds of files. Some are used to specify other
27341 aspects of the compilation environment.
27343 Note that you can also specify places to search using options such as
27344 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
27345 take precedence over places specified using environment variables, which
27346 in turn take precedence over those specified by the configuration of GCC@.
27347 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
27348 GNU Compiler Collection (GCC) Internals}.
27353 @c @itemx LC_COLLATE
27355 @c @itemx LC_MONETARY
27356 @c @itemx LC_NUMERIC
27361 @c @findex LC_COLLATE
27362 @findex LC_MESSAGES
27363 @c @findex LC_MONETARY
27364 @c @findex LC_NUMERIC
27368 These environment variables control the way that GCC uses
27369 localization information which allows GCC to work with different
27370 national conventions. GCC inspects the locale categories
27371 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
27372 so. These locale categories can be set to any value supported by your
27373 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
27374 Kingdom encoded in UTF-8.
27376 The @env{LC_CTYPE} environment variable specifies character
27377 classification. GCC uses it to determine the character boundaries in
27378 a string; this is needed for some multibyte encodings that contain quote
27379 and escape characters that are otherwise interpreted as a string
27382 The @env{LC_MESSAGES} environment variable specifies the language to
27383 use in diagnostic messages.
27385 If the @env{LC_ALL} environment variable is set, it overrides the value
27386 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
27387 and @env{LC_MESSAGES} default to the value of the @env{LANG}
27388 environment variable. If none of these variables are set, GCC
27389 defaults to traditional C English behavior.
27393 If @env{TMPDIR} is set, it specifies the directory to use for temporary
27394 files. GCC uses temporary files to hold the output of one stage of
27395 compilation which is to be used as input to the next stage: for example,
27396 the output of the preprocessor, which is the input to the compiler
27399 @item GCC_COMPARE_DEBUG
27400 @findex GCC_COMPARE_DEBUG
27401 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
27402 @option{-fcompare-debug} to the compiler driver. See the documentation
27403 of this option for more details.
27405 @item GCC_EXEC_PREFIX
27406 @findex GCC_EXEC_PREFIX
27407 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
27408 names of the subprograms executed by the compiler. No slash is added
27409 when this prefix is combined with the name of a subprogram, but you can
27410 specify a prefix that ends with a slash if you wish.
27412 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
27413 an appropriate prefix to use based on the pathname it is invoked with.
27415 If GCC cannot find the subprogram using the specified prefix, it
27416 tries looking in the usual places for the subprogram.
27418 The default value of @env{GCC_EXEC_PREFIX} is
27419 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
27420 the installed compiler. In many cases @var{prefix} is the value
27421 of @code{prefix} when you ran the @file{configure} script.
27423 Other prefixes specified with @option{-B} take precedence over this prefix.
27425 This prefix is also used for finding files such as @file{crt0.o} that are
27428 In addition, the prefix is used in an unusual way in finding the
27429 directories to search for header files. For each of the standard
27430 directories whose name normally begins with @samp{/usr/local/lib/gcc}
27431 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
27432 replacing that beginning with the specified prefix to produce an
27433 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
27434 @file{foo/bar} just before it searches the standard directory
27435 @file{/usr/local/lib/bar}.
27436 If a standard directory begins with the configured
27437 @var{prefix} then the value of @var{prefix} is replaced by
27438 @env{GCC_EXEC_PREFIX} when looking for header files.
27440 @item COMPILER_PATH
27441 @findex COMPILER_PATH
27442 The value of @env{COMPILER_PATH} is a colon-separated list of
27443 directories, much like @env{PATH}. GCC tries the directories thus
27444 specified when searching for subprograms, if it cannot find the
27445 subprograms using @env{GCC_EXEC_PREFIX}.
27448 @findex LIBRARY_PATH
27449 The value of @env{LIBRARY_PATH} is a colon-separated list of
27450 directories, much like @env{PATH}. When configured as a native compiler,
27451 GCC tries the directories thus specified when searching for special
27452 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
27453 using GCC also uses these directories when searching for ordinary
27454 libraries for the @option{-l} option (but directories specified with
27455 @option{-L} come first).
27459 @cindex locale definition
27460 This variable is used to pass locale information to the compiler. One way in
27461 which this information is used is to determine the character set to be used
27462 when character literals, string literals and comments are parsed in C and C++.
27463 When the compiler is configured to allow multibyte characters,
27464 the following values for @env{LANG} are recognized:
27468 Recognize JIS characters.
27470 Recognize SJIS characters.
27472 Recognize EUCJP characters.
27475 If @env{LANG} is not defined, or if it has some other value, then the
27476 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
27477 recognize and translate multibyte characters.
27481 Some additional environment variables affect the behavior of the
27484 @include cppenv.texi
27488 @node Precompiled Headers
27489 @section Using Precompiled Headers
27490 @cindex precompiled headers
27491 @cindex speed of compilation
27493 Often large projects have many header files that are included in every
27494 source file. The time the compiler takes to process these header files
27495 over and over again can account for nearly all of the time required to
27496 build the project. To make builds faster, GCC allows you to
27497 @dfn{precompile} a header file.
27499 To create a precompiled header file, simply compile it as you would any
27500 other file, if necessary using the @option{-x} option to make the driver
27501 treat it as a C or C++ header file. You may want to use a
27502 tool like @command{make} to keep the precompiled header up-to-date when
27503 the headers it contains change.
27505 A precompiled header file is searched for when @code{#include} is
27506 seen in the compilation. As it searches for the included file
27507 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
27508 compiler looks for a precompiled header in each directory just before it
27509 looks for the include file in that directory. The name searched for is
27510 the name specified in the @code{#include} with @samp{.gch} appended. If
27511 the precompiled header file cannot be used, it is ignored.
27513 For instance, if you have @code{#include "all.h"}, and you have
27514 @file{all.h.gch} in the same directory as @file{all.h}, then the
27515 precompiled header file is used if possible, and the original
27516 header is used otherwise.
27518 Alternatively, you might decide to put the precompiled header file in a
27519 directory and use @option{-I} to ensure that directory is searched
27520 before (or instead of) the directory containing the original header.
27521 Then, if you want to check that the precompiled header file is always
27522 used, you can put a file of the same name as the original header in this
27523 directory containing an @code{#error} command.
27525 This also works with @option{-include}. So yet another way to use
27526 precompiled headers, good for projects not designed with precompiled
27527 header files in mind, is to simply take most of the header files used by
27528 a project, include them from another header file, precompile that header
27529 file, and @option{-include} the precompiled header. If the header files
27530 have guards against multiple inclusion, they are skipped because
27531 they've already been included (in the precompiled header).
27533 If you need to precompile the same header file for different
27534 languages, targets, or compiler options, you can instead make a
27535 @emph{directory} named like @file{all.h.gch}, and put each precompiled
27536 header in the directory, perhaps using @option{-o}. It doesn't matter
27537 what you call the files in the directory; every precompiled header in
27538 the directory is considered. The first precompiled header
27539 encountered in the directory that is valid for this compilation is
27540 used; they're searched in no particular order.
27542 There are many other possibilities, limited only by your imagination,
27543 good sense, and the constraints of your build system.
27545 A precompiled header file can be used only when these conditions apply:
27549 Only one precompiled header can be used in a particular compilation.
27552 A precompiled header cannot be used once the first C token is seen. You
27553 can have preprocessor directives before a precompiled header; you cannot
27554 include a precompiled header from inside another header.
27557 The precompiled header file must be produced for the same language as
27558 the current compilation. You cannot use a C precompiled header for a C++
27562 The precompiled header file must have been produced by the same compiler
27563 binary as the current compilation is using.
27566 Any macros defined before the precompiled header is included must
27567 either be defined in the same way as when the precompiled header was
27568 generated, or must not affect the precompiled header, which usually
27569 means that they don't appear in the precompiled header at all.
27571 The @option{-D} option is one way to define a macro before a
27572 precompiled header is included; using a @code{#define} can also do it.
27573 There are also some options that define macros implicitly, like
27574 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
27577 @item If debugging information is output when using the precompiled
27578 header, using @option{-g} or similar, the same kind of debugging information
27579 must have been output when building the precompiled header. However,
27580 a precompiled header built using @option{-g} can be used in a compilation
27581 when no debugging information is being output.
27583 @item The same @option{-m} options must generally be used when building
27584 and using the precompiled header. @xref{Submodel Options},
27585 for any cases where this rule is relaxed.
27587 @item Each of the following options must be the same when building and using
27588 the precompiled header:
27590 @gccoptlist{-fexceptions}
27593 Some other command-line options starting with @option{-f},
27594 @option{-p}, or @option{-O} must be defined in the same way as when
27595 the precompiled header was generated. At present, it's not clear
27596 which options are safe to change and which are not; the safest choice
27597 is to use exactly the same options when generating and using the
27598 precompiled header. The following are known to be safe:
27600 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
27601 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
27602 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
27607 For all of these except the last, the compiler automatically
27608 ignores the precompiled header if the conditions aren't met. If you
27609 find an option combination that doesn't work and doesn't cause the
27610 precompiled header to be ignored, please consider filing a bug report,
27613 If you do use differing options when generating and using the
27614 precompiled header, the actual behavior is a mixture of the
27615 behavior for the options. For instance, if you use @option{-g} to
27616 generate the precompiled header but not when using it, you may or may
27617 not get debugging information for routines in the precompiled header.