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
951 -mgprel-sec=@var{regexp} @gol
953 -mno-bypass-cache -mbypass-cache @gol
954 -mno-cache-volatile -mcache-volatile @gol
955 -mno-fast-sw-div -mfast-sw-div @gol
956 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
957 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
958 -mcustom-fpu-cfg=@var{name} @gol
959 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
960 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
962 @emph{Nvidia PTX Options}
963 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
965 @emph{PDP-11 Options}
966 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
967 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
968 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
969 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
970 -mbranch-expensive -mbranch-cheap @gol
971 -munix-asm -mdec-asm}
973 @emph{picoChip Options}
974 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
975 -msymbol-as-address -mno-inefficient-warnings}
977 @emph{PowerPC Options}
978 See RS/6000 and PowerPC Options.
980 @emph{RISC-V Options}
981 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
982 -mmemcpy -mno-memcpy @gol
984 -mabi=@var{ABI-string} @gol
985 -mfdiv -mno-fdiv @gol
987 -march=@var{ISA-string} @gol
988 -mtune=@var{processor-string} @gol
989 -msmall-data-limit=@var{N-bytes} @gol
990 -msave-restore -mno-save-restore @gol
991 -mstrict-align -mno-strict-align @gol
992 -mcmodel=@var{code-model} @gol
993 -mexplicit-relocs -mno-explicit-relocs @gol}
996 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
997 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
998 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
1000 @emph{RS/6000 and PowerPC Options}
1001 @gccoptlist{-mcpu=@var{cpu-type} @gol
1002 -mtune=@var{cpu-type} @gol
1003 -mcmodel=@var{code-model} @gol
1005 -maltivec -mno-altivec @gol
1006 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1007 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1008 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1009 -mfprnd -mno-fprnd @gol
1010 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
1011 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1012 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1013 -malign-power -malign-natural @gol
1014 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1015 -msingle-float -mdouble-float -msimple-fpu @gol
1016 -mstring -mno-string -mupdate -mno-update @gol
1017 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1018 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1019 -mstrict-align -mno-strict-align -mrelocatable @gol
1020 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1021 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1022 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
1023 -mprioritize-restricted-insns=@var{priority} @gol
1024 -msched-costly-dep=@var{dependence_type} @gol
1025 -minsert-sched-nops=@var{scheme} @gol
1026 -mcall-sysv -mcall-netbsd @gol
1027 -maix-struct-return -msvr4-struct-return @gol
1028 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1029 -mblock-move-inline-limit=@var{num} @gol
1030 -misel -mno-isel @gol
1031 -misel=yes -misel=no @gol
1033 -mspe=yes -mspe=no @gol
1035 -mvrsave -mno-vrsave @gol
1036 -mmulhw -mno-mulhw @gol
1037 -mdlmzb -mno-dlmzb @gol
1038 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1039 -mprototype -mno-prototype @gol
1040 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1041 -msdata=@var{opt} -mvxworks -G @var{num} @gol
1042 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1043 -mno-recip-precision @gol
1044 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1045 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1046 -msave-toc-indirect -mno-save-toc-indirect @gol
1047 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1048 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1049 -mquad-memory -mno-quad-memory @gol
1050 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1051 -mcompat-align-parm -mno-compat-align-parm @gol
1052 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1053 -mgnu-attribute -mno-gnu-attribute @gol
1054 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1055 -mstack-protector-guard-offset=@var{offset}}
1058 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1060 -mbig-endian-data -mlittle-endian-data @gol
1063 -mas100-syntax -mno-as100-syntax@gol
1065 -mmax-constant-size=@gol
1068 -mallow-string-insns -mno-allow-string-insns@gol
1070 -mno-warn-multiple-fast-interrupts@gol
1071 -msave-acc-in-interrupts}
1073 @emph{S/390 and zSeries Options}
1074 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1075 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1076 -mlong-double-64 -mlong-double-128 @gol
1077 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1078 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1079 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1080 -mhtm -mvx -mzvector @gol
1081 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1082 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1083 -mhotpatch=@var{halfwords},@var{halfwords}}
1085 @emph{Score Options}
1086 @gccoptlist{-meb -mel @gol
1090 -mscore5 -mscore5u -mscore7 -mscore7d}
1093 @gccoptlist{-m1 -m2 -m2e @gol
1094 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1096 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1097 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1098 -mb -ml -mdalign -mrelax @gol
1099 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1100 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1101 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1102 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1103 -maccumulate-outgoing-args @gol
1104 -matomic-model=@var{atomic-model} @gol
1105 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1106 -mcbranch-force-delay-slot @gol
1107 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1108 -mpretend-cmove -mtas}
1110 @emph{Solaris 2 Options}
1111 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1114 @emph{SPARC Options}
1115 @gccoptlist{-mcpu=@var{cpu-type} @gol
1116 -mtune=@var{cpu-type} @gol
1117 -mcmodel=@var{code-model} @gol
1118 -mmemory-model=@var{mem-model} @gol
1119 -m32 -m64 -mapp-regs -mno-app-regs @gol
1120 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1121 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1122 -mhard-quad-float -msoft-quad-float @gol
1123 -mstack-bias -mno-stack-bias @gol
1124 -mstd-struct-return -mno-std-struct-return @gol
1125 -munaligned-doubles -mno-unaligned-doubles @gol
1126 -muser-mode -mno-user-mode @gol
1127 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1128 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1129 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1130 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1131 -mpopc -mno-popc -msubxc -mno-subxc @gol
1132 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1136 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1137 -msafe-dma -munsafe-dma @gol
1139 -msmall-mem -mlarge-mem -mstdmain @gol
1140 -mfixed-range=@var{register-range} @gol
1142 -maddress-space-conversion -mno-address-space-conversion @gol
1143 -mcache-size=@var{cache-size} @gol
1144 -matomic-updates -mno-atomic-updates}
1146 @emph{System V Options}
1147 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1149 @emph{TILE-Gx Options}
1150 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1151 -mcmodel=@var{code-model}}
1153 @emph{TILEPro Options}
1154 @gccoptlist{-mcpu=@var{cpu} -m32}
1157 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1158 -mprolog-function -mno-prolog-function -mspace @gol
1159 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1160 -mapp-regs -mno-app-regs @gol
1161 -mdisable-callt -mno-disable-callt @gol
1162 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1163 -mv850e -mv850 -mv850e3v5 @gol
1174 @gccoptlist{-mg -mgnu -munix}
1176 @emph{Visium Options}
1177 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1178 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1181 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1182 -mpointer-size=@var{size}}
1184 @emph{VxWorks Options}
1185 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1186 -Xbind-lazy -Xbind-now}
1189 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1190 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1191 -mfpmath=@var{unit} @gol
1192 -masm=@var{dialect} -mno-fancy-math-387 @gol
1193 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1194 -mno-wide-multiply -mrtd -malign-double @gol
1195 -mpreferred-stack-boundary=@var{num} @gol
1196 -mincoming-stack-boundary=@var{num} @gol
1197 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1198 -mrecip -mrecip=@var{opt} @gol
1199 -mvzeroupper -mprefer-avx128 -mprefer-avx256 @gol
1200 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1201 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1202 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1203 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1204 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1205 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1206 -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx @gol
1207 -mmwaitx -mclzero -mpku -mthreads @gol
1208 -mcet -mibt -mshstk @gol
1209 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1210 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1211 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1212 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1213 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1214 -mregparm=@var{num} -msseregparm @gol
1215 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1216 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1217 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1218 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1219 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1220 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1221 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1222 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1223 -mstack-protector-guard-reg=@var{reg} @gol
1224 -mstack-protector-guard-offset=@var{offset} @gol
1225 -mstack-protector-guard-symbol=@var{symbol} -mmitigate-rop @gol
1226 -mgeneral-regs-only -mcall-ms2sysv-xlogues}
1228 @emph{x86 Windows Options}
1229 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1230 -mnop-fun-dllimport -mthread @gol
1231 -municode -mwin32 -mwindows -fno-set-stack-executable}
1233 @emph{Xstormy16 Options}
1236 @emph{Xtensa Options}
1237 @gccoptlist{-mconst16 -mno-const16 @gol
1238 -mfused-madd -mno-fused-madd @gol
1240 -mserialize-volatile -mno-serialize-volatile @gol
1241 -mtext-section-literals -mno-text-section-literals @gol
1242 -mauto-litpools -mno-auto-litpools @gol
1243 -mtarget-align -mno-target-align @gol
1244 -mlongcalls -mno-longcalls}
1246 @emph{zSeries Options}
1247 See S/390 and zSeries Options.
1251 @node Overall Options
1252 @section Options Controlling the Kind of Output
1254 Compilation can involve up to four stages: preprocessing, compilation
1255 proper, assembly and linking, always in that order. GCC is capable of
1256 preprocessing and compiling several files either into several
1257 assembler input files, or into one assembler input file; then each
1258 assembler input file produces an object file, and linking combines all
1259 the object files (those newly compiled, and those specified as input)
1260 into an executable file.
1262 @cindex file name suffix
1263 For any given input file, the file name suffix determines what kind of
1264 compilation is done:
1268 C source code that must be preprocessed.
1271 C source code that should not be preprocessed.
1274 C++ source code that should not be preprocessed.
1277 Objective-C source code. Note that you must link with the @file{libobjc}
1278 library to make an Objective-C program work.
1281 Objective-C source code that should not be preprocessed.
1285 Objective-C++ source code. Note that you must link with the @file{libobjc}
1286 library to make an Objective-C++ program work. Note that @samp{.M} refers
1287 to a literal capital M@.
1289 @item @var{file}.mii
1290 Objective-C++ source code that should not be preprocessed.
1293 C, C++, Objective-C or Objective-C++ header file to be turned into a
1294 precompiled header (default), or C, C++ header file to be turned into an
1295 Ada spec (via the @option{-fdump-ada-spec} switch).
1298 @itemx @var{file}.cp
1299 @itemx @var{file}.cxx
1300 @itemx @var{file}.cpp
1301 @itemx @var{file}.CPP
1302 @itemx @var{file}.c++
1304 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1305 the last two letters must both be literally @samp{x}. Likewise,
1306 @samp{.C} refers to a literal capital C@.
1310 Objective-C++ source code that must be preprocessed.
1312 @item @var{file}.mii
1313 Objective-C++ source code that should not be preprocessed.
1317 @itemx @var{file}.hp
1318 @itemx @var{file}.hxx
1319 @itemx @var{file}.hpp
1320 @itemx @var{file}.HPP
1321 @itemx @var{file}.h++
1322 @itemx @var{file}.tcc
1323 C++ header file to be turned into a precompiled header or Ada spec.
1326 @itemx @var{file}.for
1327 @itemx @var{file}.ftn
1328 Fixed form Fortran source code that should not be preprocessed.
1331 @itemx @var{file}.FOR
1332 @itemx @var{file}.fpp
1333 @itemx @var{file}.FPP
1334 @itemx @var{file}.FTN
1335 Fixed form Fortran source code that must be preprocessed (with the traditional
1338 @item @var{file}.f90
1339 @itemx @var{file}.f95
1340 @itemx @var{file}.f03
1341 @itemx @var{file}.f08
1342 Free form Fortran source code that should not be preprocessed.
1344 @item @var{file}.F90
1345 @itemx @var{file}.F95
1346 @itemx @var{file}.F03
1347 @itemx @var{file}.F08
1348 Free form Fortran source code that must be preprocessed (with the
1349 traditional preprocessor).
1354 @item @var{file}.brig
1355 BRIG files (binary representation of HSAIL).
1357 @item @var{file}.ads
1358 Ada source code file that contains a library unit declaration (a
1359 declaration of a package, subprogram, or generic, or a generic
1360 instantiation), or a library unit renaming declaration (a package,
1361 generic, or subprogram renaming declaration). Such files are also
1364 @item @var{file}.adb
1365 Ada source code file containing a library unit body (a subprogram or
1366 package body). Such files are also called @dfn{bodies}.
1368 @c GCC also knows about some suffixes for languages not yet included:
1379 @itemx @var{file}.sx
1380 Assembler code that must be preprocessed.
1383 An object file to be fed straight into linking.
1384 Any file name with no recognized suffix is treated this way.
1388 You can specify the input language explicitly with the @option{-x} option:
1391 @item -x @var{language}
1392 Specify explicitly the @var{language} for the following input files
1393 (rather than letting the compiler choose a default based on the file
1394 name suffix). This option applies to all following input files until
1395 the next @option{-x} option. Possible values for @var{language} are:
1397 c c-header cpp-output
1398 c++ c++-header c++-cpp-output
1399 objective-c objective-c-header objective-c-cpp-output
1400 objective-c++ objective-c++-header objective-c++-cpp-output
1401 assembler assembler-with-cpp
1403 f77 f77-cpp-input f95 f95-cpp-input
1409 Turn off any specification of a language, so that subsequent files are
1410 handled according to their file name suffixes (as they are if @option{-x}
1411 has not been used at all).
1414 If you only want some of the stages of compilation, you can use
1415 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1416 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1417 @command{gcc} is to stop. Note that some combinations (for example,
1418 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1423 Compile or assemble the source files, but do not link. The linking
1424 stage simply is not done. The ultimate output is in the form of an
1425 object file for each source file.
1427 By default, the object file name for a source file is made by replacing
1428 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1430 Unrecognized input files, not requiring compilation or assembly, are
1435 Stop after the stage of compilation proper; do not assemble. The output
1436 is in the form of an assembler code file for each non-assembler input
1439 By default, the assembler file name for a source file is made by
1440 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1442 Input files that don't require compilation are ignored.
1446 Stop after the preprocessing stage; do not run the compiler proper. The
1447 output is in the form of preprocessed source code, which is sent to the
1450 Input files that don't require preprocessing are ignored.
1452 @cindex output file option
1455 Place output in file @var{file}. This applies to whatever
1456 sort of output is being produced, whether it be an executable file,
1457 an object file, an assembler file or preprocessed C code.
1459 If @option{-o} is not specified, the default is to put an executable
1460 file in @file{a.out}, the object file for
1461 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1462 assembler file in @file{@var{source}.s}, a precompiled header file in
1463 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1468 Print (on standard error output) the commands executed to run the stages
1469 of compilation. Also print the version number of the compiler driver
1470 program and of the preprocessor and the compiler proper.
1474 Like @option{-v} except the commands are not executed and arguments
1475 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1476 This is useful for shell scripts to capture the driver-generated command lines.
1480 Print (on the standard output) a description of the command-line options
1481 understood by @command{gcc}. If the @option{-v} option is also specified
1482 then @option{--help} is also passed on to the various processes
1483 invoked by @command{gcc}, so that they can display the command-line options
1484 they accept. If the @option{-Wextra} option has also been specified
1485 (prior to the @option{--help} option), then command-line options that
1486 have no documentation associated with them are also displayed.
1489 @opindex target-help
1490 Print (on the standard output) a description of target-specific command-line
1491 options for each tool. For some targets extra target-specific
1492 information may also be printed.
1494 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1495 Print (on the standard output) a description of the command-line
1496 options understood by the compiler that fit into all specified classes
1497 and qualifiers. These are the supported classes:
1500 @item @samp{optimizers}
1501 Display all of the optimization options supported by the
1504 @item @samp{warnings}
1505 Display all of the options controlling warning messages
1506 produced by the compiler.
1509 Display target-specific options. Unlike the
1510 @option{--target-help} option however, target-specific options of the
1511 linker and assembler are not displayed. This is because those
1512 tools do not currently support the extended @option{--help=} syntax.
1515 Display the values recognized by the @option{--param}
1518 @item @var{language}
1519 Display the options supported for @var{language}, where
1520 @var{language} is the name of one of the languages supported in this
1524 Display the options that are common to all languages.
1527 These are the supported qualifiers:
1530 @item @samp{undocumented}
1531 Display only those options that are undocumented.
1534 Display options taking an argument that appears after an equal
1535 sign in the same continuous piece of text, such as:
1536 @samp{--help=target}.
1538 @item @samp{separate}
1539 Display options taking an argument that appears as a separate word
1540 following the original option, such as: @samp{-o output-file}.
1543 Thus for example to display all the undocumented target-specific
1544 switches supported by the compiler, use:
1547 --help=target,undocumented
1550 The sense of a qualifier can be inverted by prefixing it with the
1551 @samp{^} character, so for example to display all binary warning
1552 options (i.e., ones that are either on or off and that do not take an
1553 argument) that have a description, use:
1556 --help=warnings,^joined,^undocumented
1559 The argument to @option{--help=} should not consist solely of inverted
1562 Combining several classes is possible, although this usually
1563 restricts the output so much that there is nothing to display. One
1564 case where it does work, however, is when one of the classes is
1565 @var{target}. For example, to display all the target-specific
1566 optimization options, use:
1569 --help=target,optimizers
1572 The @option{--help=} option can be repeated on the command line. Each
1573 successive use displays its requested class of options, skipping
1574 those that have already been displayed.
1576 If the @option{-Q} option appears on the command line before the
1577 @option{--help=} option, then the descriptive text displayed by
1578 @option{--help=} is changed. Instead of describing the displayed
1579 options, an indication is given as to whether the option is enabled,
1580 disabled or set to a specific value (assuming that the compiler
1581 knows this at the point where the @option{--help=} option is used).
1583 Here is a truncated example from the ARM port of @command{gcc}:
1586 % gcc -Q -mabi=2 --help=target -c
1587 The following options are target specific:
1589 -mabort-on-noreturn [disabled]
1593 The output is sensitive to the effects of previous command-line
1594 options, so for example it is possible to find out which optimizations
1595 are enabled at @option{-O2} by using:
1598 -Q -O2 --help=optimizers
1601 Alternatively you can discover which binary optimizations are enabled
1602 by @option{-O3} by using:
1605 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1606 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1607 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1612 Display the version number and copyrights of the invoked GCC@.
1614 @item -pass-exit-codes
1615 @opindex pass-exit-codes
1616 Normally the @command{gcc} program exits with the code of 1 if any
1617 phase of the compiler returns a non-success return code. If you specify
1618 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1619 the numerically highest error produced by any phase returning an error
1620 indication. The C, C++, and Fortran front ends return 4 if an internal
1621 compiler error is encountered.
1625 Use pipes rather than temporary files for communication between the
1626 various stages of compilation. This fails to work on some systems where
1627 the assembler is unable to read from a pipe; but the GNU assembler has
1630 @item -specs=@var{file}
1632 Process @var{file} after the compiler reads in the standard @file{specs}
1633 file, in order to override the defaults which the @command{gcc} driver
1634 program uses when determining what switches to pass to @command{cc1},
1635 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1636 @option{-specs=@var{file}} can be specified on the command line, and they
1637 are processed in order, from left to right. @xref{Spec Files}, for
1638 information about the format of the @var{file}.
1642 Invoke all subcommands under a wrapper program. The name of the
1643 wrapper program and its parameters are passed as a comma separated
1647 gcc -c t.c -wrapper gdb,--args
1651 This invokes all subprograms of @command{gcc} under
1652 @samp{gdb --args}, thus the invocation of @command{cc1} is
1653 @samp{gdb --args cc1 @dots{}}.
1655 @item -fplugin=@var{name}.so
1657 Load the plugin code in file @var{name}.so, assumed to be a
1658 shared object to be dlopen'd by the compiler. The base name of
1659 the shared object file is used to identify the plugin for the
1660 purposes of argument parsing (See
1661 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1662 Each plugin should define the callback functions specified in the
1665 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1666 @opindex fplugin-arg
1667 Define an argument called @var{key} with a value of @var{value}
1668 for the plugin called @var{name}.
1670 @item -fdump-ada-spec@r{[}-slim@r{]}
1671 @opindex fdump-ada-spec
1672 For C and C++ source and include files, generate corresponding Ada specs.
1673 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1674 GNAT User's Guide}, which provides detailed documentation on this feature.
1676 @item -fada-spec-parent=@var{unit}
1677 @opindex fada-spec-parent
1678 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1679 Ada specs as child units of parent @var{unit}.
1681 @item -fdump-go-spec=@var{file}
1682 @opindex fdump-go-spec
1683 For input files in any language, generate corresponding Go
1684 declarations in @var{file}. This generates Go @code{const},
1685 @code{type}, @code{var}, and @code{func} declarations which may be a
1686 useful way to start writing a Go interface to code written in some
1689 @include @value{srcdir}/../libiberty/at-file.texi
1693 @section Compiling C++ Programs
1695 @cindex suffixes for C++ source
1696 @cindex C++ source file suffixes
1697 C++ source files conventionally use one of the suffixes @samp{.C},
1698 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1699 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1700 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1701 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1702 files with these names and compiles them as C++ programs even if you
1703 call the compiler the same way as for compiling C programs (usually
1704 with the name @command{gcc}).
1708 However, the use of @command{gcc} does not add the C++ library.
1709 @command{g++} is a program that calls GCC and automatically specifies linking
1710 against the C++ library. It treats @samp{.c},
1711 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1712 files unless @option{-x} is used. This program is also useful when
1713 precompiling a C header file with a @samp{.h} extension for use in C++
1714 compilations. On many systems, @command{g++} is also installed with
1715 the name @command{c++}.
1717 @cindex invoking @command{g++}
1718 When you compile C++ programs, you may specify many of the same
1719 command-line options that you use for compiling programs in any
1720 language; or command-line options meaningful for C and related
1721 languages; or options that are meaningful only for C++ programs.
1722 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1723 explanations of options for languages related to C@.
1724 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1725 explanations of options that are meaningful only for C++ programs.
1727 @node C Dialect Options
1728 @section Options Controlling C Dialect
1729 @cindex dialect options
1730 @cindex language dialect options
1731 @cindex options, dialect
1733 The following options control the dialect of C (or languages derived
1734 from C, such as C++, Objective-C and Objective-C++) that the compiler
1738 @cindex ANSI support
1742 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1743 equivalent to @option{-std=c++98}.
1745 This turns off certain features of GCC that are incompatible with ISO
1746 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1747 such as the @code{asm} and @code{typeof} keywords, and
1748 predefined macros such as @code{unix} and @code{vax} that identify the
1749 type of system you are using. It also enables the undesirable and
1750 rarely used ISO trigraph feature. For the C compiler,
1751 it disables recognition of C++ style @samp{//} comments as well as
1752 the @code{inline} keyword.
1754 The alternate keywords @code{__asm__}, @code{__extension__},
1755 @code{__inline__} and @code{__typeof__} continue to work despite
1756 @option{-ansi}. You would not want to use them in an ISO C program, of
1757 course, but it is useful to put them in header files that might be included
1758 in compilations done with @option{-ansi}. Alternate predefined macros
1759 such as @code{__unix__} and @code{__vax__} are also available, with or
1760 without @option{-ansi}.
1762 The @option{-ansi} option does not cause non-ISO programs to be
1763 rejected gratuitously. For that, @option{-Wpedantic} is required in
1764 addition to @option{-ansi}. @xref{Warning Options}.
1766 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1767 option is used. Some header files may notice this macro and refrain
1768 from declaring certain functions or defining certain macros that the
1769 ISO standard doesn't call for; this is to avoid interfering with any
1770 programs that might use these names for other things.
1772 Functions that are normally built in but do not have semantics
1773 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1774 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1775 built-in functions provided by GCC}, for details of the functions
1780 Determine the language standard. @xref{Standards,,Language Standards
1781 Supported by GCC}, for details of these standard versions. This option
1782 is currently only supported when compiling C or C++.
1784 The compiler can accept several base standards, such as @samp{c90} or
1785 @samp{c++98}, and GNU dialects of those standards, such as
1786 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1787 compiler accepts all programs following that standard plus those
1788 using GNU extensions that do not contradict it. For example,
1789 @option{-std=c90} turns off certain features of GCC that are
1790 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1791 keywords, but not other GNU extensions that do not have a meaning in
1792 ISO C90, such as omitting the middle term of a @code{?:}
1793 expression. On the other hand, when a GNU dialect of a standard is
1794 specified, all features supported by the compiler are enabled, even when
1795 those features change the meaning of the base standard. As a result, some
1796 strict-conforming programs may be rejected. The particular standard
1797 is used by @option{-Wpedantic} to identify which features are GNU
1798 extensions given that version of the standard. For example
1799 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1800 comments, while @option{-std=gnu99 -Wpedantic} does not.
1802 A value for this option must be provided; possible values are
1808 Support all ISO C90 programs (certain GNU extensions that conflict
1809 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1811 @item iso9899:199409
1812 ISO C90 as modified in amendment 1.
1818 ISO C99. This standard is substantially completely supported, modulo
1819 bugs and floating-point issues
1820 (mainly but not entirely relating to optional C99 features from
1821 Annexes F and G). See
1822 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1823 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1828 ISO C11, the 2011 revision of the ISO C standard. This standard is
1829 substantially completely supported, modulo bugs, floating-point issues
1830 (mainly but not entirely relating to optional C11 features from
1831 Annexes F and G) and the optional Annexes K (Bounds-checking
1832 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1836 GNU dialect of ISO C90 (including some C99 features).
1840 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1844 GNU dialect of ISO C11. This is the default for C code.
1845 The name @samp{gnu1x} is deprecated.
1849 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1850 additional defect reports. Same as @option{-ansi} for C++ code.
1854 GNU dialect of @option{-std=c++98}.
1858 The 2011 ISO C++ standard plus amendments.
1859 The name @samp{c++0x} is deprecated.
1863 GNU dialect of @option{-std=c++11}.
1864 The name @samp{gnu++0x} is deprecated.
1868 The 2014 ISO C++ standard plus amendments.
1869 The name @samp{c++1y} is deprecated.
1873 GNU dialect of @option{-std=c++14}.
1874 This is the default for C++ code.
1875 The name @samp{gnu++1y} is deprecated.
1879 The 2017 ISO C++ standard plus amendments.
1880 The name @samp{c++1z} is deprecated.
1884 GNU dialect of @option{-std=c++17}.
1885 The name @samp{gnu++1z} is deprecated.
1888 The next revision of the ISO C++ standard, tentatively planned for
1889 2020. Support is highly experimental, and will almost certainly
1890 change in incompatible ways in future releases.
1893 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
1894 and will almost certainly change in incompatible ways in future
1898 @item -fgnu89-inline
1899 @opindex fgnu89-inline
1900 The option @option{-fgnu89-inline} tells GCC to use the traditional
1901 GNU semantics for @code{inline} functions when in C99 mode.
1902 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1903 Using this option is roughly equivalent to adding the
1904 @code{gnu_inline} function attribute to all inline functions
1905 (@pxref{Function Attributes}).
1907 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1908 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1909 specifies the default behavior).
1910 This option is not supported in @option{-std=c90} or
1911 @option{-std=gnu90} mode.
1913 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1914 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1915 in effect for @code{inline} functions. @xref{Common Predefined
1916 Macros,,,cpp,The C Preprocessor}.
1918 @item -fpermitted-flt-eval-methods=@var{style}
1919 @opindex fpermitted-flt-eval-methods
1920 @opindex fpermitted-flt-eval-methods=c11
1921 @opindex fpermitted-flt-eval-methods=ts-18661-3
1922 ISO/IEC TS 18661-3 defines new permissible values for
1923 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
1924 a semantic type that is an interchange or extended format should be
1925 evaluated to the precision and range of that type. These new values are
1926 a superset of those permitted under C99/C11, which does not specify the
1927 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
1928 conforming to C11 may not have been written expecting the possibility of
1931 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
1932 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
1933 or the extended set of values specified in ISO/IEC TS 18661-3.
1935 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
1937 The default when in a standards compliant mode (@option{-std=c11} or similar)
1938 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
1939 dialect (@option{-std=gnu11} or similar) is
1940 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
1942 @item -aux-info @var{filename}
1944 Output to the given filename prototyped declarations for all functions
1945 declared and/or defined in a translation unit, including those in header
1946 files. This option is silently ignored in any language other than C@.
1948 Besides declarations, the file indicates, in comments, the origin of
1949 each declaration (source file and line), whether the declaration was
1950 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1951 @samp{O} for old, respectively, in the first character after the line
1952 number and the colon), and whether it came from a declaration or a
1953 definition (@samp{C} or @samp{F}, respectively, in the following
1954 character). In the case of function definitions, a K&R-style list of
1955 arguments followed by their declarations is also provided, inside
1956 comments, after the declaration.
1958 @item -fallow-parameterless-variadic-functions
1959 @opindex fallow-parameterless-variadic-functions
1960 Accept variadic functions without named parameters.
1962 Although it is possible to define such a function, this is not very
1963 useful as it is not possible to read the arguments. This is only
1964 supported for C as this construct is allowed by C++.
1968 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1969 keyword, so that code can use these words as identifiers. You can use
1970 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1971 instead. @option{-ansi} implies @option{-fno-asm}.
1973 In C++, this switch only affects the @code{typeof} keyword, since
1974 @code{asm} and @code{inline} are standard keywords. You may want to
1975 use the @option{-fno-gnu-keywords} flag instead, which has the same
1976 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1977 switch only affects the @code{asm} and @code{typeof} keywords, since
1978 @code{inline} is a standard keyword in ISO C99.
1981 @itemx -fno-builtin-@var{function}
1982 @opindex fno-builtin
1983 @cindex built-in functions
1984 Don't recognize built-in functions that do not begin with
1985 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1986 functions provided by GCC}, for details of the functions affected,
1987 including those which are not built-in functions when @option{-ansi} or
1988 @option{-std} options for strict ISO C conformance are used because they
1989 do not have an ISO standard meaning.
1991 GCC normally generates special code to handle certain built-in functions
1992 more efficiently; for instance, calls to @code{alloca} may become single
1993 instructions which adjust the stack directly, and calls to @code{memcpy}
1994 may become inline copy loops. The resulting code is often both smaller
1995 and faster, but since the function calls no longer appear as such, you
1996 cannot set a breakpoint on those calls, nor can you change the behavior
1997 of the functions by linking with a different library. In addition,
1998 when a function is recognized as a built-in function, GCC may use
1999 information about that function to warn about problems with calls to
2000 that function, or to generate more efficient code, even if the
2001 resulting code still contains calls to that function. For example,
2002 warnings are given with @option{-Wformat} for bad calls to
2003 @code{printf} when @code{printf} is built in and @code{strlen} is
2004 known not to modify global memory.
2006 With the @option{-fno-builtin-@var{function}} option
2007 only the built-in function @var{function} is
2008 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2009 function is named that is not built-in in this version of GCC, this
2010 option is ignored. There is no corresponding
2011 @option{-fbuiltin-@var{function}} option; if you wish to enable
2012 built-in functions selectively when using @option{-fno-builtin} or
2013 @option{-ffreestanding}, you may define macros such as:
2016 #define abs(n) __builtin_abs ((n))
2017 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2023 Enable parsing of function definitions marked with @code{__GIMPLE}.
2024 This is an experimental feature that allows unit testing of GIMPLE
2029 @cindex hosted environment
2031 Assert that compilation targets a hosted environment. This implies
2032 @option{-fbuiltin}. A hosted environment is one in which the
2033 entire standard library is available, and in which @code{main} has a return
2034 type of @code{int}. Examples are nearly everything except a kernel.
2035 This is equivalent to @option{-fno-freestanding}.
2037 @item -ffreestanding
2038 @opindex ffreestanding
2039 @cindex hosted environment
2041 Assert that compilation targets a freestanding environment. This
2042 implies @option{-fno-builtin}. A freestanding environment
2043 is one in which the standard library may not exist, and program startup may
2044 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2045 This is equivalent to @option{-fno-hosted}.
2047 @xref{Standards,,Language Standards Supported by GCC}, for details of
2048 freestanding and hosted environments.
2052 @cindex OpenACC accelerator programming
2053 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2054 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2055 compiler generates accelerated code according to the OpenACC Application
2056 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2057 implies @option{-pthread}, and thus is only supported on targets that
2058 have support for @option{-pthread}.
2060 @item -fopenacc-dim=@var{geom}
2061 @opindex fopenacc-dim
2062 @cindex OpenACC accelerator programming
2063 Specify default compute dimensions for parallel offload regions that do
2064 not explicitly specify. The @var{geom} value is a triple of
2065 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2066 can be omitted, to use a target-specific default value.
2070 @cindex OpenMP parallel
2071 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2072 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2073 compiler generates parallel code according to the OpenMP Application
2074 Program Interface v4.5 @w{@uref{http://www.openmp.org/}}. This option
2075 implies @option{-pthread}, and thus is only supported on targets that
2076 have support for @option{-pthread}. @option{-fopenmp} implies
2077 @option{-fopenmp-simd}.
2080 @opindex fopenmp-simd
2083 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2084 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2089 @cindex Enable Cilk Plus
2090 Enable the usage of Cilk Plus language extension features for C/C++.
2091 When the option @option{-fcilkplus} is specified, enable the usage of
2092 the Cilk Plus Language extension features for C/C++. The present
2093 implementation follows ABI version 1.2. This is an experimental
2094 feature that is only partially complete, and whose interface may
2095 change in future versions of GCC as the official specification
2096 changes. Currently, all features but @code{_Cilk_for} have been
2101 When the option @option{-fgnu-tm} is specified, the compiler
2102 generates code for the Linux variant of Intel's current Transactional
2103 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2104 an experimental feature whose interface may change in future versions
2105 of GCC, as the official specification changes. Please note that not
2106 all architectures are supported for this feature.
2108 For more information on GCC's support for transactional memory,
2109 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2110 Transactional Memory Library}.
2112 Note that the transactional memory feature is not supported with
2113 non-call exceptions (@option{-fnon-call-exceptions}).
2115 @item -fms-extensions
2116 @opindex fms-extensions
2117 Accept some non-standard constructs used in Microsoft header files.
2119 In C++ code, this allows member names in structures to be similar
2120 to previous types declarations.
2129 Some cases of unnamed fields in structures and unions are only
2130 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2131 fields within structs/unions}, for details.
2133 Note that this option is off for all targets but x86
2134 targets using ms-abi.
2136 @item -fplan9-extensions
2137 @opindex fplan9-extensions
2138 Accept some non-standard constructs used in Plan 9 code.
2140 This enables @option{-fms-extensions}, permits passing pointers to
2141 structures with anonymous fields to functions that expect pointers to
2142 elements of the type of the field, and permits referring to anonymous
2143 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2144 struct/union fields within structs/unions}, for details. This is only
2145 supported for C, not C++.
2147 @item -fcond-mismatch
2148 @opindex fcond-mismatch
2149 Allow conditional expressions with mismatched types in the second and
2150 third arguments. The value of such an expression is void. This option
2151 is not supported for C++.
2153 @item -flax-vector-conversions
2154 @opindex flax-vector-conversions
2155 Allow implicit conversions between vectors with differing numbers of
2156 elements and/or incompatible element types. This option should not be
2159 @item -funsigned-char
2160 @opindex funsigned-char
2161 Let the type @code{char} be unsigned, like @code{unsigned char}.
2163 Each kind of machine has a default for what @code{char} should
2164 be. It is either like @code{unsigned char} by default or like
2165 @code{signed char} by default.
2167 Ideally, a portable program should always use @code{signed char} or
2168 @code{unsigned char} when it depends on the signedness of an object.
2169 But many programs have been written to use plain @code{char} and
2170 expect it to be signed, or expect it to be unsigned, depending on the
2171 machines they were written for. This option, and its inverse, let you
2172 make such a program work with the opposite default.
2174 The type @code{char} is always a distinct type from each of
2175 @code{signed char} or @code{unsigned char}, even though its behavior
2176 is always just like one of those two.
2179 @opindex fsigned-char
2180 Let the type @code{char} be signed, like @code{signed char}.
2182 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2183 the negative form of @option{-funsigned-char}. Likewise, the option
2184 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2186 @item -fsigned-bitfields
2187 @itemx -funsigned-bitfields
2188 @itemx -fno-signed-bitfields
2189 @itemx -fno-unsigned-bitfields
2190 @opindex fsigned-bitfields
2191 @opindex funsigned-bitfields
2192 @opindex fno-signed-bitfields
2193 @opindex fno-unsigned-bitfields
2194 These options control whether a bit-field is signed or unsigned, when the
2195 declaration does not use either @code{signed} or @code{unsigned}. By
2196 default, such a bit-field is signed, because this is consistent: the
2197 basic integer types such as @code{int} are signed types.
2199 @item -fsso-struct=@var{endianness}
2200 @opindex fsso-struct
2201 Set the default scalar storage order of structures and unions to the
2202 specified endianness. The accepted values are @samp{big-endian},
2203 @samp{little-endian} and @samp{native} for the native endianness of
2204 the target (the default). This option is not supported for C++.
2206 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2207 code that is not binary compatible with code generated without it if the
2208 specified endianness is not the native endianness of the target.
2211 @node C++ Dialect Options
2212 @section Options Controlling C++ Dialect
2214 @cindex compiler options, C++
2215 @cindex C++ options, command-line
2216 @cindex options, C++
2217 This section describes the command-line options that are only meaningful
2218 for C++ programs. You can also use most of the GNU compiler options
2219 regardless of what language your program is in. For example, you
2220 might compile a file @file{firstClass.C} like this:
2223 g++ -g -fstrict-enums -O -c firstClass.C
2227 In this example, only @option{-fstrict-enums} is an option meant
2228 only for C++ programs; you can use the other options with any
2229 language supported by GCC@.
2231 Some options for compiling C programs, such as @option{-std}, are also
2232 relevant for C++ programs.
2233 @xref{C Dialect Options,,Options Controlling C Dialect}.
2235 Here is a list of options that are @emph{only} for compiling C++ programs:
2239 @item -fabi-version=@var{n}
2240 @opindex fabi-version
2241 Use version @var{n} of the C++ ABI@. The default is version 0.
2243 Version 0 refers to the version conforming most closely to
2244 the C++ ABI specification. Therefore, the ABI obtained using version 0
2245 will change in different versions of G++ as ABI bugs are fixed.
2247 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2249 Version 2 is the version of the C++ ABI that first appeared in G++
2250 3.4, and was the default through G++ 4.9.
2252 Version 3 corrects an error in mangling a constant address as a
2255 Version 4, which first appeared in G++ 4.5, implements a standard
2256 mangling for vector types.
2258 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2259 attribute const/volatile on function pointer types, decltype of a
2260 plain decl, and use of a function parameter in the declaration of
2263 Version 6, which first appeared in G++ 4.7, corrects the promotion
2264 behavior of C++11 scoped enums and the mangling of template argument
2265 packs, const/static_cast, prefix ++ and --, and a class scope function
2266 used as a template argument.
2268 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2269 builtin type and corrects the mangling of lambdas in default argument
2272 Version 8, which first appeared in G++ 4.9, corrects the substitution
2273 behavior of function types with function-cv-qualifiers.
2275 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2278 Version 10, which first appeared in G++ 6.1, adds mangling of
2279 attributes that affect type identity, such as ia32 calling convention
2280 attributes (e.g. @samp{stdcall}).
2282 Version 11, which first appeared in G++ 7, corrects the mangling of
2283 sizeof... expressions and operator names. For multiple entities with
2284 the same name within a function, that are declared in different scopes,
2285 the mangling now changes starting with the twelfth occurrence. It also
2286 implies @option{-fnew-inheriting-ctors}.
2288 See also @option{-Wabi}.
2290 @item -fabi-compat-version=@var{n}
2291 @opindex fabi-compat-version
2292 On targets that support strong aliases, G++
2293 works around mangling changes by creating an alias with the correct
2294 mangled name when defining a symbol with an incorrect mangled name.
2295 This switch specifies which ABI version to use for the alias.
2297 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2298 compatibility). If another ABI version is explicitly selected, this
2299 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2300 use @option{-fabi-compat-version=2}.
2302 If this option is not provided but @option{-Wabi=@var{n}} is, that
2303 version is used for compatibility aliases. If this option is provided
2304 along with @option{-Wabi} (without the version), the version from this
2305 option is used for the warning.
2307 @item -fno-access-control
2308 @opindex fno-access-control
2309 Turn off all access checking. This switch is mainly useful for working
2310 around bugs in the access control code.
2313 @opindex faligned-new
2314 Enable support for C++17 @code{new} of types that require more
2315 alignment than @code{void* ::operator new(std::size_t)} provides. A
2316 numeric argument such as @code{-faligned-new=32} can be used to
2317 specify how much alignment (in bytes) is provided by that function,
2318 but few users will need to override the default of
2319 @code{alignof(std::max_align_t)}.
2321 This flag is enabled by default for @option{-std=c++17}.
2325 Check that the pointer returned by @code{operator new} is non-null
2326 before attempting to modify the storage allocated. This check is
2327 normally unnecessary because the C++ standard specifies that
2328 @code{operator new} only returns @code{0} if it is declared
2329 @code{throw()}, in which case the compiler always checks the
2330 return value even without this option. In all other cases, when
2331 @code{operator new} has a non-empty exception specification, memory
2332 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2333 @samp{new (nothrow)}.
2337 Enable support for the C++ Extensions for Concepts Technical
2338 Specification, ISO 19217 (2015), which allows code like
2341 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2342 template <Addable T> T add (T a, T b) @{ return a + b; @}
2345 @item -fconstexpr-depth=@var{n}
2346 @opindex fconstexpr-depth
2347 Set the maximum nested evaluation depth for C++11 constexpr functions
2348 to @var{n}. A limit is needed to detect endless recursion during
2349 constant expression evaluation. The minimum specified by the standard
2352 @item -fconstexpr-loop-limit=@var{n}
2353 @opindex fconstexpr-loop-limit
2354 Set the maximum number of iterations for a loop in C++14 constexpr functions
2355 to @var{n}. A limit is needed to detect infinite loops during
2356 constant expression evaluation. The default is 262144 (1<<18).
2358 @item -fdeduce-init-list
2359 @opindex fdeduce-init-list
2360 Enable deduction of a template type parameter as
2361 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2364 template <class T> auto forward(T t) -> decltype (realfn (t))
2371 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2375 This deduction was implemented as a possible extension to the
2376 originally proposed semantics for the C++11 standard, but was not part
2377 of the final standard, so it is disabled by default. This option is
2378 deprecated, and may be removed in a future version of G++.
2380 @item -ffriend-injection
2381 @opindex ffriend-injection
2382 Inject friend functions into the enclosing namespace, so that they are
2383 visible outside the scope of the class in which they are declared.
2384 Friend functions were documented to work this way in the old Annotated
2385 C++ Reference Manual.
2386 However, in ISO C++ a friend function that is not declared
2387 in an enclosing scope can only be found using argument dependent
2388 lookup. GCC defaults to the standard behavior.
2390 This option is for compatibility, and may be removed in a future
2393 @item -fno-elide-constructors
2394 @opindex fno-elide-constructors
2395 The C++ standard allows an implementation to omit creating a temporary
2396 that is only used to initialize another object of the same type.
2397 Specifying this option disables that optimization, and forces G++ to
2398 call the copy constructor in all cases. This option also causes G++
2399 to call trivial member functions which otherwise would be expanded inline.
2401 In C++17, the compiler is required to omit these temporaries, but this
2402 option still affects trivial member functions.
2404 @item -fno-enforce-eh-specs
2405 @opindex fno-enforce-eh-specs
2406 Don't generate code to check for violation of exception specifications
2407 at run time. This option violates the C++ standard, but may be useful
2408 for reducing code size in production builds, much like defining
2409 @code{NDEBUG}. This does not give user code permission to throw
2410 exceptions in violation of the exception specifications; the compiler
2411 still optimizes based on the specifications, so throwing an
2412 unexpected exception results in undefined behavior at run time.
2414 @item -fextern-tls-init
2415 @itemx -fno-extern-tls-init
2416 @opindex fextern-tls-init
2417 @opindex fno-extern-tls-init
2418 The C++11 and OpenMP standards allow @code{thread_local} and
2419 @code{threadprivate} variables to have dynamic (runtime)
2420 initialization. To support this, any use of such a variable goes
2421 through a wrapper function that performs any necessary initialization.
2422 When the use and definition of the variable are in the same
2423 translation unit, this overhead can be optimized away, but when the
2424 use is in a different translation unit there is significant overhead
2425 even if the variable doesn't actually need dynamic initialization. If
2426 the programmer can be sure that no use of the variable in a
2427 non-defining TU needs to trigger dynamic initialization (either
2428 because the variable is statically initialized, or a use of the
2429 variable in the defining TU will be executed before any uses in
2430 another TU), they can avoid this overhead with the
2431 @option{-fno-extern-tls-init} option.
2433 On targets that support symbol aliases, the default is
2434 @option{-fextern-tls-init}. On targets that do not support symbol
2435 aliases, the default is @option{-fno-extern-tls-init}.
2438 @itemx -fno-for-scope
2440 @opindex fno-for-scope
2441 If @option{-ffor-scope} is specified, the scope of variables declared in
2442 a @i{for-init-statement} is limited to the @code{for} loop itself,
2443 as specified by the C++ standard.
2444 If @option{-fno-for-scope} is specified, the scope of variables declared in
2445 a @i{for-init-statement} extends to the end of the enclosing scope,
2446 as was the case in old versions of G++, and other (traditional)
2447 implementations of C++.
2449 If neither flag is given, the default is to follow the standard,
2450 but to allow and give a warning for old-style code that would
2451 otherwise be invalid, or have different behavior.
2453 @item -fno-gnu-keywords
2454 @opindex fno-gnu-keywords
2455 Do not recognize @code{typeof} as a keyword, so that code can use this
2456 word as an identifier. You can use the keyword @code{__typeof__} instead.
2457 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2458 @option{-std=c++98}, @option{-std=c++11}, etc.
2460 @item -fno-implicit-templates
2461 @opindex fno-implicit-templates
2462 Never emit code for non-inline templates that are instantiated
2463 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2464 @xref{Template Instantiation}, for more information.
2466 @item -fno-implicit-inline-templates
2467 @opindex fno-implicit-inline-templates
2468 Don't emit code for implicit instantiations of inline templates, either.
2469 The default is to handle inlines differently so that compiles with and
2470 without optimization need the same set of explicit instantiations.
2472 @item -fno-implement-inlines
2473 @opindex fno-implement-inlines
2474 To save space, do not emit out-of-line copies of inline functions
2475 controlled by @code{#pragma implementation}. This causes linker
2476 errors if these functions are not inlined everywhere they are called.
2478 @item -fms-extensions
2479 @opindex fms-extensions
2480 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2481 int and getting a pointer to member function via non-standard syntax.
2483 @item -fnew-inheriting-ctors
2484 @opindex fnew-inheriting-ctors
2485 Enable the P0136 adjustment to the semantics of C++11 constructor
2486 inheritance. This is part of C++17 but also considered to be a Defect
2487 Report against C++11 and C++14. This flag is enabled by default
2488 unless @option{-fabi-version=10} or lower is specified.
2490 @item -fnew-ttp-matching
2491 @opindex fnew-ttp-matching
2492 Enable the P0522 resolution to Core issue 150, template template
2493 parameters and default arguments: this allows a template with default
2494 template arguments as an argument for a template template parameter
2495 with fewer template parameters. This flag is enabled by default for
2496 @option{-std=c++17}.
2498 @item -fno-nonansi-builtins
2499 @opindex fno-nonansi-builtins
2500 Disable built-in declarations of functions that are not mandated by
2501 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2502 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2505 @opindex fnothrow-opt
2506 Treat a @code{throw()} exception specification as if it were a
2507 @code{noexcept} specification to reduce or eliminate the text size
2508 overhead relative to a function with no exception specification. If
2509 the function has local variables of types with non-trivial
2510 destructors, the exception specification actually makes the
2511 function smaller because the EH cleanups for those variables can be
2512 optimized away. The semantic effect is that an exception thrown out of
2513 a function with such an exception specification results in a call
2514 to @code{terminate} rather than @code{unexpected}.
2516 @item -fno-operator-names
2517 @opindex fno-operator-names
2518 Do not treat the operator name keywords @code{and}, @code{bitand},
2519 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2520 synonyms as keywords.
2522 @item -fno-optional-diags
2523 @opindex fno-optional-diags
2524 Disable diagnostics that the standard says a compiler does not need to
2525 issue. Currently, the only such diagnostic issued by G++ is the one for
2526 a name having multiple meanings within a class.
2529 @opindex fpermissive
2530 Downgrade some diagnostics about nonconformant code from errors to
2531 warnings. Thus, using @option{-fpermissive} allows some
2532 nonconforming code to compile.
2534 @item -fno-pretty-templates
2535 @opindex fno-pretty-templates
2536 When an error message refers to a specialization of a function
2537 template, the compiler normally prints the signature of the
2538 template followed by the template arguments and any typedefs or
2539 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2540 rather than @code{void f(int)}) so that it's clear which template is
2541 involved. When an error message refers to a specialization of a class
2542 template, the compiler omits any template arguments that match
2543 the default template arguments for that template. If either of these
2544 behaviors make it harder to understand the error message rather than
2545 easier, you can use @option{-fno-pretty-templates} to disable them.
2549 Enable automatic template instantiation at link time. This option also
2550 implies @option{-fno-implicit-templates}. @xref{Template
2551 Instantiation}, for more information.
2555 Disable generation of information about every class with virtual
2556 functions for use by the C++ run-time type identification features
2557 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2558 of the language, you can save some space by using this flag. Note that
2559 exception handling uses the same information, but G++ generates it as
2560 needed. The @code{dynamic_cast} operator can still be used for casts that
2561 do not require run-time type information, i.e.@: casts to @code{void *} or to
2562 unambiguous base classes.
2564 @item -fsized-deallocation
2565 @opindex fsized-deallocation
2566 Enable the built-in global declarations
2568 void operator delete (void *, std::size_t) noexcept;
2569 void operator delete[] (void *, std::size_t) noexcept;
2571 as introduced in C++14. This is useful for user-defined replacement
2572 deallocation functions that, for example, use the size of the object
2573 to make deallocation faster. Enabled by default under
2574 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2575 warns about places that might want to add a definition.
2577 @item -fstrict-enums
2578 @opindex fstrict-enums
2579 Allow the compiler to optimize using the assumption that a value of
2580 enumerated type can only be one of the values of the enumeration (as
2581 defined in the C++ standard; basically, a value that can be
2582 represented in the minimum number of bits needed to represent all the
2583 enumerators). This assumption may not be valid if the program uses a
2584 cast to convert an arbitrary integer value to the enumerated type.
2586 @item -fstrong-eval-order
2587 @opindex fstrong-eval-order
2588 Evaluate member access, array subscripting, and shift expressions in
2589 left-to-right order, and evaluate assignment in right-to-left order,
2590 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2591 @option{-fstrong-eval-order=some} enables just the ordering of member
2592 access and shift expressions, and is the default without
2593 @option{-std=c++17}.
2595 @item -ftemplate-backtrace-limit=@var{n}
2596 @opindex ftemplate-backtrace-limit
2597 Set the maximum number of template instantiation notes for a single
2598 warning or error to @var{n}. The default value is 10.
2600 @item -ftemplate-depth=@var{n}
2601 @opindex ftemplate-depth
2602 Set the maximum instantiation depth for template classes to @var{n}.
2603 A limit on the template instantiation depth is needed to detect
2604 endless recursions during template class instantiation. ANSI/ISO C++
2605 conforming programs must not rely on a maximum depth greater than 17
2606 (changed to 1024 in C++11). The default value is 900, as the compiler
2607 can run out of stack space before hitting 1024 in some situations.
2609 @item -fno-threadsafe-statics
2610 @opindex fno-threadsafe-statics
2611 Do not emit the extra code to use the routines specified in the C++
2612 ABI for thread-safe initialization of local statics. You can use this
2613 option to reduce code size slightly in code that doesn't need to be
2616 @item -fuse-cxa-atexit
2617 @opindex fuse-cxa-atexit
2618 Register destructors for objects with static storage duration with the
2619 @code{__cxa_atexit} function rather than the @code{atexit} function.
2620 This option is required for fully standards-compliant handling of static
2621 destructors, but only works if your C library supports
2622 @code{__cxa_atexit}.
2624 @item -fno-use-cxa-get-exception-ptr
2625 @opindex fno-use-cxa-get-exception-ptr
2626 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2627 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2628 if the runtime routine is not available.
2630 @item -fvisibility-inlines-hidden
2631 @opindex fvisibility-inlines-hidden
2632 This switch declares that the user does not attempt to compare
2633 pointers to inline functions or methods where the addresses of the two functions
2634 are taken in different shared objects.
2636 The effect of this is that GCC may, effectively, mark inline methods with
2637 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2638 appear in the export table of a DSO and do not require a PLT indirection
2639 when used within the DSO@. Enabling this option can have a dramatic effect
2640 on load and link times of a DSO as it massively reduces the size of the
2641 dynamic export table when the library makes heavy use of templates.
2643 The behavior of this switch is not quite the same as marking the
2644 methods as hidden directly, because it does not affect static variables
2645 local to the function or cause the compiler to deduce that
2646 the function is defined in only one shared object.
2648 You may mark a method as having a visibility explicitly to negate the
2649 effect of the switch for that method. For example, if you do want to
2650 compare pointers to a particular inline method, you might mark it as
2651 having default visibility. Marking the enclosing class with explicit
2652 visibility has no effect.
2654 Explicitly instantiated inline methods are unaffected by this option
2655 as their linkage might otherwise cross a shared library boundary.
2656 @xref{Template Instantiation}.
2658 @item -fvisibility-ms-compat
2659 @opindex fvisibility-ms-compat
2660 This flag attempts to use visibility settings to make GCC's C++
2661 linkage model compatible with that of Microsoft Visual Studio.
2663 The flag makes these changes to GCC's linkage model:
2667 It sets the default visibility to @code{hidden}, like
2668 @option{-fvisibility=hidden}.
2671 Types, but not their members, are not hidden by default.
2674 The One Definition Rule is relaxed for types without explicit
2675 visibility specifications that are defined in more than one
2676 shared object: those declarations are permitted if they are
2677 permitted when this option is not used.
2680 In new code it is better to use @option{-fvisibility=hidden} and
2681 export those classes that are intended to be externally visible.
2682 Unfortunately it is possible for code to rely, perhaps accidentally,
2683 on the Visual Studio behavior.
2685 Among the consequences of these changes are that static data members
2686 of the same type with the same name but defined in different shared
2687 objects are different, so changing one does not change the other;
2688 and that pointers to function members defined in different shared
2689 objects may not compare equal. When this flag is given, it is a
2690 violation of the ODR to define types with the same name differently.
2694 Do not use weak symbol support, even if it is provided by the linker.
2695 By default, G++ uses weak symbols if they are available. This
2696 option exists only for testing, and should not be used by end-users;
2697 it results in inferior code and has no benefits. This option may
2698 be removed in a future release of G++.
2702 Do not search for header files in the standard directories specific to
2703 C++, but do still search the other standard directories. (This option
2704 is used when building the C++ library.)
2707 In addition, these optimization, warning, and code generation options
2708 have meanings only for C++ programs:
2711 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2714 Warn when G++ it generates code that is probably not compatible with
2715 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2716 ABI with each major release, normally @option{-Wabi} will warn only if
2717 there is a check added later in a release series for an ABI issue
2718 discovered since the initial release. @option{-Wabi} will warn about
2719 more things if an older ABI version is selected (with
2720 @option{-fabi-version=@var{n}}).
2722 @option{-Wabi} can also be used with an explicit version number to
2723 warn about compatibility with a particular @option{-fabi-version}
2724 level, e.g. @option{-Wabi=2} to warn about changes relative to
2725 @option{-fabi-version=2}.
2727 If an explicit version number is provided and
2728 @option{-fabi-compat-version} is not specified, the version number
2729 from this option is used for compatibility aliases. If no explicit
2730 version number is provided with this option, but
2731 @option{-fabi-compat-version} is specified, that version number is
2732 used for ABI warnings.
2734 Although an effort has been made to warn about
2735 all such cases, there are probably some cases that are not warned about,
2736 even though G++ is generating incompatible code. There may also be
2737 cases where warnings are emitted even though the code that is generated
2740 You should rewrite your code to avoid these warnings if you are
2741 concerned about the fact that code generated by G++ may not be binary
2742 compatible with code generated by other compilers.
2744 Known incompatibilities in @option{-fabi-version=2} (which was the
2745 default from GCC 3.4 to 4.9) include:
2750 A template with a non-type template parameter of reference type was
2751 mangled incorrectly:
2754 template <int &> struct S @{@};
2758 This was fixed in @option{-fabi-version=3}.
2761 SIMD vector types declared using @code{__attribute ((vector_size))} were
2762 mangled in a non-standard way that does not allow for overloading of
2763 functions taking vectors of different sizes.
2765 The mangling was changed in @option{-fabi-version=4}.
2768 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2769 qualifiers, and @code{decltype} of a plain declaration was folded away.
2771 These mangling issues were fixed in @option{-fabi-version=5}.
2774 Scoped enumerators passed as arguments to a variadic function are
2775 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2776 On most targets this does not actually affect the parameter passing
2777 ABI, as there is no way to pass an argument smaller than @code{int}.
2779 Also, the ABI changed the mangling of template argument packs,
2780 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2781 a class scope function used as a template argument.
2783 These issues were corrected in @option{-fabi-version=6}.
2786 Lambdas in default argument scope were mangled incorrectly, and the
2787 ABI changed the mangling of @code{nullptr_t}.
2789 These issues were corrected in @option{-fabi-version=7}.
2792 When mangling a function type with function-cv-qualifiers, the
2793 un-qualified function type was incorrectly treated as a substitution
2796 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2799 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2800 unaligned accesses. Note that this did not affect the ABI of a
2801 function with a @code{nullptr_t} parameter, as parameters have a
2804 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2807 Target-specific attributes that affect the identity of a type, such as
2808 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2809 did not affect the mangled name, leading to name collisions when
2810 function pointers were used as template arguments.
2812 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2816 It also warns about psABI-related changes. The known psABI changes at this
2822 For SysV/x86-64, unions with @code{long double} members are
2823 passed in memory as specified in psABI. For example:
2833 @code{union U} is always passed in memory.
2837 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2840 Warn when a type with an ABI tag is used in a context that does not
2841 have that ABI tag. See @ref{C++ Attributes} for more information
2844 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2845 @opindex Wctor-dtor-privacy
2846 @opindex Wno-ctor-dtor-privacy
2847 Warn when a class seems unusable because all the constructors or
2848 destructors in that class are private, and it has neither friends nor
2849 public static member functions. Also warn if there are no non-private
2850 methods, and there's at least one private member function that isn't
2851 a constructor or destructor.
2853 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2854 @opindex Wdelete-non-virtual-dtor
2855 @opindex Wno-delete-non-virtual-dtor
2856 Warn when @code{delete} is used to destroy an instance of a class that
2857 has virtual functions and non-virtual destructor. It is unsafe to delete
2858 an instance of a derived class through a pointer to a base class if the
2859 base class does not have a virtual destructor. This warning is enabled
2862 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2863 @opindex Wliteral-suffix
2864 @opindex Wno-literal-suffix
2865 Warn when a string or character literal is followed by a ud-suffix which does
2866 not begin with an underscore. As a conforming extension, GCC treats such
2867 suffixes as separate preprocessing tokens in order to maintain backwards
2868 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2872 #define __STDC_FORMAT_MACROS
2873 #include <inttypes.h>
2878 printf("My int64: %" PRId64"\n", i64);
2882 In this case, @code{PRId64} is treated as a separate preprocessing token.
2884 Additionally, warn when a user-defined literal operator is declared with
2885 a literal suffix identifier that doesn't begin with an underscore. Literal
2886 suffix identifiers that don't begin with an underscore are reserved for
2887 future standardization.
2889 This warning is enabled by default.
2891 @item -Wlto-type-mismatch
2892 @opindex Wlto-type-mismatch
2893 @opindex Wno-lto-type-mismatch
2895 During the link-time optimization warn about type mismatches in
2896 global declarations from different compilation units.
2897 Requires @option{-flto} to be enabled. Enabled by default.
2899 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2901 @opindex Wno-narrowing
2902 For C++11 and later standards, narrowing conversions are diagnosed by default,
2903 as required by the standard. A narrowing conversion from a constant produces
2904 an error, and a narrowing conversion from a non-constant produces a warning,
2905 but @option{-Wno-narrowing} suppresses the diagnostic.
2906 Note that this does not affect the meaning of well-formed code;
2907 narrowing conversions are still considered ill-formed in SFINAE contexts.
2909 With @option{-Wnarrowing} in C++98, warn when a narrowing
2910 conversion prohibited by C++11 occurs within
2914 int i = @{ 2.2 @}; // error: narrowing from double to int
2917 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2919 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2921 @opindex Wno-noexcept
2922 Warn when a noexcept-expression evaluates to false because of a call
2923 to a function that does not have a non-throwing exception
2924 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2925 the compiler to never throw an exception.
2927 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
2928 @opindex Wnoexcept-type
2929 @opindex Wno-noexcept-type
2930 Warn if the C++17 feature making @code{noexcept} part of a function
2931 type changes the mangled name of a symbol relative to C++14. Enabled
2932 by @option{-Wabi} and @option{-Wc++17-compat}.
2935 template <class T> void f(T t) @{ t(); @};
2937 void h() @{ f(g); @} // in C++14 calls f<void(*)()>, in C++17 calls f<void(*)()noexcept>
2940 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
2941 @opindex Wclass-memaccess
2942 Warn when the destination of a call to a raw memory function such as
2943 @code{memset} or @code{memcpy} is an object of class type writing into which
2944 might bypass the class non-trivial or deleted constructor or copy assignment,
2945 violate const-correctness or encapsulation, or corrupt the virtual table.
2946 Modifying the representation of such objects may violate invariants maintained
2947 by member functions of the class. For example, the call to @code{memset}
2948 below is undefined becase it modifies a non-trivial class object and is,
2949 therefore, diagnosed. The safe way to either initialize or clear the storage
2950 of objects of such types is by using the appropriate constructor or assignment
2951 operator, if one is available.
2953 std::string str = "abc";
2954 memset (&str, 0, 3);
2956 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
2958 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2959 @opindex Wnon-virtual-dtor
2960 @opindex Wno-non-virtual-dtor
2961 Warn when a class has virtual functions and an accessible non-virtual
2962 destructor itself or in an accessible polymorphic base class, in which
2963 case it is possible but unsafe to delete an instance of a derived
2964 class through a pointer to the class itself or base class. This
2965 warning is automatically enabled if @option{-Weffc++} is specified.
2967 @item -Wregister @r{(C++ and Objective-C++ only)}
2969 @opindex Wno-register
2970 Warn on uses of the @code{register} storage class specifier, except
2971 when it is part of the GNU @ref{Explicit Register Variables} extension.
2972 The use of the @code{register} keyword as storage class specifier has
2973 been deprecated in C++11 and removed in C++17.
2974 Enabled by default with @option{-std=c++17}.
2976 @item -Wreorder @r{(C++ and Objective-C++ only)}
2978 @opindex Wno-reorder
2979 @cindex reordering, warning
2980 @cindex warning for reordering of member initializers
2981 Warn when the order of member initializers given in the code does not
2982 match the order in which they must be executed. For instance:
2988 A(): j (0), i (1) @{ @}
2993 The compiler rearranges the member initializers for @code{i}
2994 and @code{j} to match the declaration order of the members, emitting
2995 a warning to that effect. This warning is enabled by @option{-Wall}.
2997 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2998 @opindex fext-numeric-literals
2999 @opindex fno-ext-numeric-literals
3000 Accept imaginary, fixed-point, or machine-defined
3001 literal number suffixes as GNU extensions.
3002 When this option is turned off these suffixes are treated
3003 as C++11 user-defined literal numeric suffixes.
3004 This is on by default for all pre-C++11 dialects and all GNU dialects:
3005 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3006 @option{-std=gnu++14}.
3007 This option is off by default
3008 for ISO C++11 onwards (@option{-std=c++11}, ...).
3011 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3014 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3017 Warn about violations of the following style guidelines from Scott Meyers'
3018 @cite{Effective C++} series of books:
3022 Define a copy constructor and an assignment operator for classes
3023 with dynamically-allocated memory.
3026 Prefer initialization to assignment in constructors.
3029 Have @code{operator=} return a reference to @code{*this}.
3032 Don't try to return a reference when you must return an object.
3035 Distinguish between prefix and postfix forms of increment and
3036 decrement operators.
3039 Never overload @code{&&}, @code{||}, or @code{,}.
3043 This option also enables @option{-Wnon-virtual-dtor}, which is also
3044 one of the effective C++ recommendations. However, the check is
3045 extended to warn about the lack of virtual destructor in accessible
3046 non-polymorphic bases classes too.
3048 When selecting this option, be aware that the standard library
3049 headers do not obey all of these guidelines; use @samp{grep -v}
3050 to filter out those warnings.
3052 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3053 @opindex Wstrict-null-sentinel
3054 @opindex Wno-strict-null-sentinel
3055 Warn about the use of an uncasted @code{NULL} as sentinel. When
3056 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3057 to @code{__null}. Although it is a null pointer constant rather than a
3058 null pointer, it is guaranteed to be of the same size as a pointer.
3059 But this use is not portable across different compilers.
3061 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3062 @opindex Wno-non-template-friend
3063 @opindex Wnon-template-friend
3064 Disable warnings when non-template friend functions are declared
3065 within a template. In very old versions of GCC that predate implementation
3066 of the ISO standard, declarations such as
3067 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3068 could be interpreted as a particular specialization of a template
3069 function; the warning exists to diagnose compatibility problems,
3070 and is enabled by default.
3072 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3073 @opindex Wold-style-cast
3074 @opindex Wno-old-style-cast
3075 Warn if an old-style (C-style) cast to a non-void type is used within
3076 a C++ program. The new-style casts (@code{dynamic_cast},
3077 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3078 less vulnerable to unintended effects and much easier to search for.
3080 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3081 @opindex Woverloaded-virtual
3082 @opindex Wno-overloaded-virtual
3083 @cindex overloaded virtual function, warning
3084 @cindex warning for overloaded virtual function
3085 Warn when a function declaration hides virtual functions from a
3086 base class. For example, in:
3093 struct B: public A @{
3098 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3109 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3110 @opindex Wno-pmf-conversions
3111 @opindex Wpmf-conversions
3112 Disable the diagnostic for converting a bound pointer to member function
3115 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3116 @opindex Wsign-promo
3117 @opindex Wno-sign-promo
3118 Warn when overload resolution chooses a promotion from unsigned or
3119 enumerated type to a signed type, over a conversion to an unsigned type of
3120 the same size. Previous versions of G++ tried to preserve
3121 unsignedness, but the standard mandates the current behavior.
3123 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3125 Warn when a primary template declaration is encountered. Some coding
3126 rules disallow templates, and this may be used to enforce that rule.
3127 The warning is inactive inside a system header file, such as the STL, so
3128 one can still use the STL. One may also instantiate or specialize
3131 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3132 @opindex Wmultiple-inheritance
3133 Warn when a class is defined with multiple direct base classes. Some
3134 coding rules disallow multiple inheritance, and this may be used to
3135 enforce that rule. The warning is inactive inside a system header file,
3136 such as the STL, so one can still use the STL. One may also define
3137 classes that indirectly use multiple inheritance.
3139 @item -Wvirtual-inheritance
3140 @opindex Wvirtual-inheritance
3141 Warn when a class is defined with a virtual direct base class. Some
3142 coding rules disallow multiple inheritance, and this may be used to
3143 enforce that rule. The warning is inactive inside a system header file,
3144 such as the STL, so one can still use the STL. One may also define
3145 classes that indirectly use virtual inheritance.
3148 @opindex Wnamespaces
3149 Warn when a namespace definition is opened. Some coding rules disallow
3150 namespaces, and this may be used to enforce that rule. The warning is
3151 inactive inside a system header file, such as the STL, so one can still
3152 use the STL. One may also use using directives and qualified names.
3154 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3156 @opindex Wno-terminate
3157 Disable the warning about a throw-expression that will immediately
3158 result in a call to @code{terminate}.
3161 @node Objective-C and Objective-C++ Dialect Options
3162 @section Options Controlling Objective-C and Objective-C++ Dialects
3164 @cindex compiler options, Objective-C and Objective-C++
3165 @cindex Objective-C and Objective-C++ options, command-line
3166 @cindex options, Objective-C and Objective-C++
3167 (NOTE: This manual does not describe the Objective-C and Objective-C++
3168 languages themselves. @xref{Standards,,Language Standards
3169 Supported by GCC}, for references.)
3171 This section describes the command-line options that are only meaningful
3172 for Objective-C and Objective-C++ programs. You can also use most of
3173 the language-independent GNU compiler options.
3174 For example, you might compile a file @file{some_class.m} like this:
3177 gcc -g -fgnu-runtime -O -c some_class.m
3181 In this example, @option{-fgnu-runtime} is an option meant only for
3182 Objective-C and Objective-C++ programs; you can use the other options with
3183 any language supported by GCC@.
3185 Note that since Objective-C is an extension of the C language, Objective-C
3186 compilations may also use options specific to the C front-end (e.g.,
3187 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3188 C++-specific options (e.g., @option{-Wabi}).
3190 Here is a list of options that are @emph{only} for compiling Objective-C
3191 and Objective-C++ programs:
3194 @item -fconstant-string-class=@var{class-name}
3195 @opindex fconstant-string-class
3196 Use @var{class-name} as the name of the class to instantiate for each
3197 literal string specified with the syntax @code{@@"@dots{}"}. The default
3198 class name is @code{NXConstantString} if the GNU runtime is being used, and
3199 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3200 @option{-fconstant-cfstrings} option, if also present, overrides the
3201 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3202 to be laid out as constant CoreFoundation strings.
3205 @opindex fgnu-runtime
3206 Generate object code compatible with the standard GNU Objective-C
3207 runtime. This is the default for most types of systems.
3209 @item -fnext-runtime
3210 @opindex fnext-runtime
3211 Generate output compatible with the NeXT runtime. This is the default
3212 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3213 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3216 @item -fno-nil-receivers
3217 @opindex fno-nil-receivers
3218 Assume that all Objective-C message dispatches (@code{[receiver
3219 message:arg]}) in this translation unit ensure that the receiver is
3220 not @code{nil}. This allows for more efficient entry points in the
3221 runtime to be used. This option is only available in conjunction with
3222 the NeXT runtime and ABI version 0 or 1.
3224 @item -fobjc-abi-version=@var{n}
3225 @opindex fobjc-abi-version
3226 Use version @var{n} of the Objective-C ABI for the selected runtime.
3227 This option is currently supported only for the NeXT runtime. In that
3228 case, Version 0 is the traditional (32-bit) ABI without support for
3229 properties and other Objective-C 2.0 additions. Version 1 is the
3230 traditional (32-bit) ABI with support for properties and other
3231 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3232 nothing is specified, the default is Version 0 on 32-bit target
3233 machines, and Version 2 on 64-bit target machines.
3235 @item -fobjc-call-cxx-cdtors
3236 @opindex fobjc-call-cxx-cdtors
3237 For each Objective-C class, check if any of its instance variables is a
3238 C++ object with a non-trivial default constructor. If so, synthesize a
3239 special @code{- (id) .cxx_construct} instance method which runs
3240 non-trivial default constructors on any such instance variables, in order,
3241 and then return @code{self}. Similarly, check if any instance variable
3242 is a C++ object with a non-trivial destructor, and if so, synthesize a
3243 special @code{- (void) .cxx_destruct} method which runs
3244 all such default destructors, in reverse order.
3246 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3247 methods thusly generated only operate on instance variables
3248 declared in the current Objective-C class, and not those inherited
3249 from superclasses. It is the responsibility of the Objective-C
3250 runtime to invoke all such methods in an object's inheritance
3251 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3252 by the runtime immediately after a new object instance is allocated;
3253 the @code{- (void) .cxx_destruct} methods are invoked immediately
3254 before the runtime deallocates an object instance.
3256 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3257 support for invoking the @code{- (id) .cxx_construct} and
3258 @code{- (void) .cxx_destruct} methods.
3260 @item -fobjc-direct-dispatch
3261 @opindex fobjc-direct-dispatch
3262 Allow fast jumps to the message dispatcher. On Darwin this is
3263 accomplished via the comm page.
3265 @item -fobjc-exceptions
3266 @opindex fobjc-exceptions
3267 Enable syntactic support for structured exception handling in
3268 Objective-C, similar to what is offered by C++. This option
3269 is required to use the Objective-C keywords @code{@@try},
3270 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3271 @code{@@synchronized}. This option is available with both the GNU
3272 runtime and the NeXT runtime (but not available in conjunction with
3273 the NeXT runtime on Mac OS X 10.2 and earlier).
3277 Enable garbage collection (GC) in Objective-C and Objective-C++
3278 programs. This option is only available with the NeXT runtime; the
3279 GNU runtime has a different garbage collection implementation that
3280 does not require special compiler flags.
3282 @item -fobjc-nilcheck
3283 @opindex fobjc-nilcheck
3284 For the NeXT runtime with version 2 of the ABI, check for a nil
3285 receiver in method invocations before doing the actual method call.
3286 This is the default and can be disabled using
3287 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3288 checked for nil in this way no matter what this flag is set to.
3289 Currently this flag does nothing when the GNU runtime, or an older
3290 version of the NeXT runtime ABI, is used.
3292 @item -fobjc-std=objc1
3294 Conform to the language syntax of Objective-C 1.0, the language
3295 recognized by GCC 4.0. This only affects the Objective-C additions to
3296 the C/C++ language; it does not affect conformance to C/C++ standards,
3297 which is controlled by the separate C/C++ dialect option flags. When
3298 this option is used with the Objective-C or Objective-C++ compiler,
3299 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3300 This is useful if you need to make sure that your Objective-C code can
3301 be compiled with older versions of GCC@.
3303 @item -freplace-objc-classes
3304 @opindex freplace-objc-classes
3305 Emit a special marker instructing @command{ld(1)} not to statically link in
3306 the resulting object file, and allow @command{dyld(1)} to load it in at
3307 run time instead. This is used in conjunction with the Fix-and-Continue
3308 debugging mode, where the object file in question may be recompiled and
3309 dynamically reloaded in the course of program execution, without the need
3310 to restart the program itself. Currently, Fix-and-Continue functionality
3311 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3316 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3317 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3318 compile time) with static class references that get initialized at load time,
3319 which improves run-time performance. Specifying the @option{-fzero-link} flag
3320 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3321 to be retained. This is useful in Zero-Link debugging mode, since it allows
3322 for individual class implementations to be modified during program execution.
3323 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3324 regardless of command-line options.
3326 @item -fno-local-ivars
3327 @opindex fno-local-ivars
3328 @opindex flocal-ivars
3329 By default instance variables in Objective-C can be accessed as if
3330 they were local variables from within the methods of the class they're
3331 declared in. This can lead to shadowing between instance variables
3332 and other variables declared either locally inside a class method or
3333 globally with the same name. Specifying the @option{-fno-local-ivars}
3334 flag disables this behavior thus avoiding variable shadowing issues.
3336 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3337 @opindex fivar-visibility
3338 Set the default instance variable visibility to the specified option
3339 so that instance variables declared outside the scope of any access
3340 modifier directives default to the specified visibility.
3344 Dump interface declarations for all classes seen in the source file to a
3345 file named @file{@var{sourcename}.decl}.
3347 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3348 @opindex Wassign-intercept
3349 @opindex Wno-assign-intercept
3350 Warn whenever an Objective-C assignment is being intercepted by the
3353 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3354 @opindex Wno-protocol
3356 If a class is declared to implement a protocol, a warning is issued for
3357 every method in the protocol that is not implemented by the class. The
3358 default behavior is to issue a warning for every method not explicitly
3359 implemented in the class, even if a method implementation is inherited
3360 from the superclass. If you use the @option{-Wno-protocol} option, then
3361 methods inherited from the superclass are considered to be implemented,
3362 and no warning is issued for them.
3364 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3366 @opindex Wno-selector
3367 Warn if multiple methods of different types for the same selector are
3368 found during compilation. The check is performed on the list of methods
3369 in the final stage of compilation. Additionally, a check is performed
3370 for each selector appearing in a @code{@@selector(@dots{})}
3371 expression, and a corresponding method for that selector has been found
3372 during compilation. Because these checks scan the method table only at
3373 the end of compilation, these warnings are not produced if the final
3374 stage of compilation is not reached, for example because an error is
3375 found during compilation, or because the @option{-fsyntax-only} option is
3378 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3379 @opindex Wstrict-selector-match
3380 @opindex Wno-strict-selector-match
3381 Warn if multiple methods with differing argument and/or return types are
3382 found for a given selector when attempting to send a message using this
3383 selector to a receiver of type @code{id} or @code{Class}. When this flag
3384 is off (which is the default behavior), the compiler omits such warnings
3385 if any differences found are confined to types that share the same size
3388 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3389 @opindex Wundeclared-selector
3390 @opindex Wno-undeclared-selector
3391 Warn if a @code{@@selector(@dots{})} expression referring to an
3392 undeclared selector is found. A selector is considered undeclared if no
3393 method with that name has been declared before the
3394 @code{@@selector(@dots{})} expression, either explicitly in an
3395 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3396 an @code{@@implementation} section. This option always performs its
3397 checks as soon as a @code{@@selector(@dots{})} expression is found,
3398 while @option{-Wselector} only performs its checks in the final stage of
3399 compilation. This also enforces the coding style convention
3400 that methods and selectors must be declared before being used.
3402 @item -print-objc-runtime-info
3403 @opindex print-objc-runtime-info
3404 Generate C header describing the largest structure that is passed by
3409 @node Diagnostic Message Formatting Options
3410 @section Options to Control Diagnostic Messages Formatting
3411 @cindex options to control diagnostics formatting
3412 @cindex diagnostic messages
3413 @cindex message formatting
3415 Traditionally, diagnostic messages have been formatted irrespective of
3416 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3417 options described below
3418 to control the formatting algorithm for diagnostic messages,
3419 e.g.@: how many characters per line, how often source location
3420 information should be reported. Note that some language front ends may not
3421 honor these options.
3424 @item -fmessage-length=@var{n}
3425 @opindex fmessage-length
3426 Try to format error messages so that they fit on lines of about
3427 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3428 done; each error message appears on a single line. This is the
3429 default for all front ends.
3431 @item -fdiagnostics-show-location=once
3432 @opindex fdiagnostics-show-location
3433 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3434 reporter to emit source location information @emph{once}; that is, in
3435 case the message is too long to fit on a single physical line and has to
3436 be wrapped, the source location won't be emitted (as prefix) again,
3437 over and over, in subsequent continuation lines. This is the default
3440 @item -fdiagnostics-show-location=every-line
3441 Only meaningful in line-wrapping mode. Instructs the diagnostic
3442 messages reporter to emit the same source location information (as
3443 prefix) for physical lines that result from the process of breaking
3444 a message which is too long to fit on a single line.
3446 @item -fdiagnostics-color[=@var{WHEN}]
3447 @itemx -fno-diagnostics-color
3448 @opindex fdiagnostics-color
3449 @cindex highlight, color
3450 @vindex GCC_COLORS @r{environment variable}
3451 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3452 or @samp{auto}. The default depends on how the compiler has been configured,
3453 it can be any of the above @var{WHEN} options or also @samp{never}
3454 if @env{GCC_COLORS} environment variable isn't present in the environment,
3455 and @samp{auto} otherwise.
3456 @samp{auto} means to use color only when the standard error is a terminal.
3457 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3458 aliases for @option{-fdiagnostics-color=always} and
3459 @option{-fdiagnostics-color=never}, respectively.
3461 The colors are defined by the environment variable @env{GCC_COLORS}.
3462 Its value is a colon-separated list of capabilities and Select Graphic
3463 Rendition (SGR) substrings. SGR commands are interpreted by the
3464 terminal or terminal emulator. (See the section in the documentation
3465 of your text terminal for permitted values and their meanings as
3466 character attributes.) These substring values are integers in decimal
3467 representation and can be concatenated with semicolons.
3468 Common values to concatenate include
3470 @samp{4} for underline,
3472 @samp{7} for inverse,
3473 @samp{39} for default foreground color,
3474 @samp{30} to @samp{37} for foreground colors,
3475 @samp{90} to @samp{97} for 16-color mode foreground colors,
3476 @samp{38;5;0} to @samp{38;5;255}
3477 for 88-color and 256-color modes foreground colors,
3478 @samp{49} for default background color,
3479 @samp{40} to @samp{47} for background colors,
3480 @samp{100} to @samp{107} for 16-color mode background colors,
3481 and @samp{48;5;0} to @samp{48;5;255}
3482 for 88-color and 256-color modes background colors.
3484 The default @env{GCC_COLORS} is
3486 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3487 quote=01:fixit-insert=32:fixit-delete=31:\
3488 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3492 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3493 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3494 @samp{01} is bold, and @samp{31} is red.
3495 Setting @env{GCC_COLORS} to the empty string disables colors.
3496 Supported capabilities are as follows.
3500 @vindex error GCC_COLORS @r{capability}
3501 SGR substring for error: markers.
3504 @vindex warning GCC_COLORS @r{capability}
3505 SGR substring for warning: markers.
3508 @vindex note GCC_COLORS @r{capability}
3509 SGR substring for note: markers.
3512 @vindex range1 GCC_COLORS @r{capability}
3513 SGR substring for first additional range.
3516 @vindex range2 GCC_COLORS @r{capability}
3517 SGR substring for second additional range.
3520 @vindex locus GCC_COLORS @r{capability}
3521 SGR substring for location information, @samp{file:line} or
3522 @samp{file:line:column} etc.
3525 @vindex quote GCC_COLORS @r{capability}
3526 SGR substring for information printed within quotes.
3529 @vindex fixit-insert GCC_COLORS @r{capability}
3530 SGR substring for fix-it hints suggesting text to
3531 be inserted or replaced.
3534 @vindex fixit-delete GCC_COLORS @r{capability}
3535 SGR substring for fix-it hints suggesting text to
3538 @item diff-filename=
3539 @vindex diff-filename GCC_COLORS @r{capability}
3540 SGR substring for filename headers within generated patches.
3543 @vindex diff-hunk GCC_COLORS @r{capability}
3544 SGR substring for the starts of hunks within generated patches.
3547 @vindex diff-delete GCC_COLORS @r{capability}
3548 SGR substring for deleted lines within generated patches.
3551 @vindex diff-insert GCC_COLORS @r{capability}
3552 SGR substring for inserted lines within generated patches.
3555 @vindex type-diff GCC_COLORS @r{capability}
3556 SGR substring for highlighting mismatching types within template
3557 arguments in the C++ frontend.
3560 @item -fno-diagnostics-show-option
3561 @opindex fno-diagnostics-show-option
3562 @opindex fdiagnostics-show-option
3563 By default, each diagnostic emitted includes text indicating the
3564 command-line option that directly controls the diagnostic (if such an
3565 option is known to the diagnostic machinery). Specifying the
3566 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3568 @item -fno-diagnostics-show-caret
3569 @opindex fno-diagnostics-show-caret
3570 @opindex fdiagnostics-show-caret
3571 By default, each diagnostic emitted includes the original source line
3572 and a caret @samp{^} indicating the column. This option suppresses this
3573 information. The source line is truncated to @var{n} characters, if
3574 the @option{-fmessage-length=n} option is given. When the output is done
3575 to the terminal, the width is limited to the width given by the
3576 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3578 @item -fdiagnostics-parseable-fixits
3579 @opindex fdiagnostics-parseable-fixits
3580 Emit fix-it hints in a machine-parseable format, suitable for consumption
3581 by IDEs. For each fix-it, a line will be printed after the relevant
3582 diagnostic, starting with the string ``fix-it:''. For example:
3585 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3588 The location is expressed as a half-open range, expressed as a count of
3589 bytes, starting at byte 1 for the initial column. In the above example,
3590 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3594 00000000011111111112222222222
3595 12345678901234567890123456789
3596 gtk_widget_showall (dlg);
3601 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3602 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3603 (e.g. vertical tab as ``\013'').
3605 An empty replacement string indicates that the given range is to be removed.
3606 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3607 be inserted at the given position.
3609 @item -fdiagnostics-generate-patch
3610 @opindex fdiagnostics-generate-patch
3611 Print fix-it hints to stderr in unified diff format, after any diagnostics
3612 are printed. For example:
3619 void show_cb(GtkDialog *dlg)
3621 - gtk_widget_showall(dlg);
3622 + gtk_widget_show_all(dlg);
3627 The diff may or may not be colorized, following the same rules
3628 as for diagnostics (see @option{-fdiagnostics-color}).
3630 @item -fdiagnostics-show-template-tree
3631 @opindex fdiagnostics-show-template-tree
3633 In the C++ frontend, when printing diagnostics showing mismatching
3634 template types, such as:
3637 could not convert 'std::map<int, std::vector<double> >()'
3638 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3641 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3642 tree-like structure showing the common and differing parts of the types,
3652 The parts that differ are highlighted with color (``double'' and
3653 ``float'' in this case).
3655 @item -fno-elide-type
3656 @opindex fno-elide-type
3657 @opindex felide-type
3658 By default when the C++ frontend prints diagnostics showing mismatching
3659 template types, common parts of the types are printed as ``[...]'' to
3660 simplify the error message. For example:
3663 could not convert 'std::map<int, std::vector<double> >()'
3664 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3667 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
3668 This flag also affects the output of the
3669 @option{-fdiagnostics-show-template-tree} flag.
3671 @item -fno-show-column
3672 @opindex fno-show-column
3673 Do not print column numbers in diagnostics. This may be necessary if
3674 diagnostics are being scanned by a program that does not understand the
3675 column numbers, such as @command{dejagnu}.
3679 @node Warning Options
3680 @section Options to Request or Suppress Warnings
3681 @cindex options to control warnings
3682 @cindex warning messages
3683 @cindex messages, warning
3684 @cindex suppressing warnings
3686 Warnings are diagnostic messages that report constructions that
3687 are not inherently erroneous but that are risky or suggest there
3688 may have been an error.
3690 The following language-independent options do not enable specific
3691 warnings but control the kinds of diagnostics produced by GCC@.
3694 @cindex syntax checking
3696 @opindex fsyntax-only
3697 Check the code for syntax errors, but don't do anything beyond that.
3699 @item -fmax-errors=@var{n}
3700 @opindex fmax-errors
3701 Limits the maximum number of error messages to @var{n}, at which point
3702 GCC bails out rather than attempting to continue processing the source
3703 code. If @var{n} is 0 (the default), there is no limit on the number
3704 of error messages produced. If @option{-Wfatal-errors} is also
3705 specified, then @option{-Wfatal-errors} takes precedence over this
3710 Inhibit all warning messages.
3715 Make all warnings into errors.
3720 Make the specified warning into an error. The specifier for a warning
3721 is appended; for example @option{-Werror=switch} turns the warnings
3722 controlled by @option{-Wswitch} into errors. This switch takes a
3723 negative form, to be used to negate @option{-Werror} for specific
3724 warnings; for example @option{-Wno-error=switch} makes
3725 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3728 The warning message for each controllable warning includes the
3729 option that controls the warning. That option can then be used with
3730 @option{-Werror=} and @option{-Wno-error=} as described above.
3731 (Printing of the option in the warning message can be disabled using the
3732 @option{-fno-diagnostics-show-option} flag.)
3734 Note that specifying @option{-Werror=}@var{foo} automatically implies
3735 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3738 @item -Wfatal-errors
3739 @opindex Wfatal-errors
3740 @opindex Wno-fatal-errors
3741 This option causes the compiler to abort compilation on the first error
3742 occurred rather than trying to keep going and printing further error
3747 You can request many specific warnings with options beginning with
3748 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3749 implicit declarations. Each of these specific warning options also
3750 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3751 example, @option{-Wno-implicit}. This manual lists only one of the
3752 two forms, whichever is not the default. For further
3753 language-specific options also refer to @ref{C++ Dialect Options} and
3754 @ref{Objective-C and Objective-C++ Dialect Options}.
3756 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3757 options, such as @option{-Wunused}, which may turn on further options,
3758 such as @option{-Wunused-value}. The combined effect of positive and
3759 negative forms is that more specific options have priority over less
3760 specific ones, independently of their position in the command-line. For
3761 options of the same specificity, the last one takes effect. Options
3762 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3763 as if they appeared at the end of the command-line.
3765 When an unrecognized warning option is requested (e.g.,
3766 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3767 that the option is not recognized. However, if the @option{-Wno-} form
3768 is used, the behavior is slightly different: no diagnostic is
3769 produced for @option{-Wno-unknown-warning} unless other diagnostics
3770 are being produced. This allows the use of new @option{-Wno-} options
3771 with old compilers, but if something goes wrong, the compiler
3772 warns that an unrecognized option is present.
3779 Issue all the warnings demanded by strict ISO C and ISO C++;
3780 reject all programs that use forbidden extensions, and some other
3781 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3782 version of the ISO C standard specified by any @option{-std} option used.
3784 Valid ISO C and ISO C++ programs should compile properly with or without
3785 this option (though a rare few require @option{-ansi} or a
3786 @option{-std} option specifying the required version of ISO C)@. However,
3787 without this option, certain GNU extensions and traditional C and C++
3788 features are supported as well. With this option, they are rejected.
3790 @option{-Wpedantic} does not cause warning messages for use of the
3791 alternate keywords whose names begin and end with @samp{__}. Pedantic
3792 warnings are also disabled in the expression that follows
3793 @code{__extension__}. However, only system header files should use
3794 these escape routes; application programs should avoid them.
3795 @xref{Alternate Keywords}.
3797 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3798 C conformance. They soon find that it does not do quite what they want:
3799 it finds some non-ISO practices, but not all---only those for which
3800 ISO C @emph{requires} a diagnostic, and some others for which
3801 diagnostics have been added.
3803 A feature to report any failure to conform to ISO C might be useful in
3804 some instances, but would require considerable additional work and would
3805 be quite different from @option{-Wpedantic}. We don't have plans to
3806 support such a feature in the near future.
3808 Where the standard specified with @option{-std} represents a GNU
3809 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3810 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3811 extended dialect is based. Warnings from @option{-Wpedantic} are given
3812 where they are required by the base standard. (It does not make sense
3813 for such warnings to be given only for features not in the specified GNU
3814 C dialect, since by definition the GNU dialects of C include all
3815 features the compiler supports with the given option, and there would be
3816 nothing to warn about.)
3818 @item -pedantic-errors
3819 @opindex pedantic-errors
3820 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3821 requires a diagnostic, in some cases where there is undefined behavior
3822 at compile-time and in some other cases that do not prevent compilation
3823 of programs that are valid according to the standard. This is not
3824 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3825 by this option and not enabled by the latter and vice versa.
3830 This enables all the warnings about constructions that some users
3831 consider questionable, and that are easy to avoid (or modify to
3832 prevent the warning), even in conjunction with macros. This also
3833 enables some language-specific warnings described in @ref{C++ Dialect
3834 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3836 @option{-Wall} turns on the following warning flags:
3838 @gccoptlist{-Waddress @gol
3839 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3841 -Wbool-operation @gol
3842 -Wc++11-compat -Wc++14-compat @gol
3843 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
3844 -Wchar-subscripts @gol
3846 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3847 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3849 -Wint-in-bool-context @gol
3850 -Wimplicit @r{(C and Objective-C only)} @gol
3851 -Wimplicit-int @r{(C and Objective-C only)} @gol
3852 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3853 -Winit-self @r{(only for C++)} @gol
3854 -Wlogical-not-parentheses @gol
3855 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3856 -Wmaybe-uninitialized @gol
3857 -Wmemset-elt-size @gol
3858 -Wmemset-transposed-args @gol
3859 -Wmisleading-indentation @r{(only for C/C++)} @gol
3860 -Wmissing-braces @r{(only for C/ObjC)} @gol
3861 -Wmultistatement-macros @gol
3862 -Wnarrowing @r{(only for C++)} @gol
3864 -Wnonnull-compare @gol
3870 -Wsequence-point @gol
3871 -Wsign-compare @r{(only in C++)} @gol
3872 -Wsizeof-pointer-div @gol
3873 -Wsizeof-pointer-memaccess @gol
3874 -Wstrict-aliasing @gol
3875 -Wstrict-overflow=1 @gol
3877 -Wtautological-compare @gol
3879 -Wuninitialized @gol
3880 -Wunknown-pragmas @gol
3881 -Wunused-function @gol
3884 -Wunused-variable @gol
3885 -Wvolatile-register-var @gol
3888 Note that some warning flags are not implied by @option{-Wall}. Some of
3889 them warn about constructions that users generally do not consider
3890 questionable, but which occasionally you might wish to check for;
3891 others warn about constructions that are necessary or hard to avoid in
3892 some cases, and there is no simple way to modify the code to suppress
3893 the warning. Some of them are enabled by @option{-Wextra} but many of
3894 them must be enabled individually.
3900 This enables some extra warning flags that are not enabled by
3901 @option{-Wall}. (This option used to be called @option{-W}. The older
3902 name is still supported, but the newer name is more descriptive.)
3904 @gccoptlist{-Wclobbered @gol
3906 -Wignored-qualifiers @gol
3907 -Wimplicit-fallthrough=3 @gol
3908 -Wmissing-field-initializers @gol
3909 -Wmissing-parameter-type @r{(C only)} @gol
3910 -Wold-style-declaration @r{(C only)} @gol
3911 -Woverride-init @gol
3912 -Wsign-compare @r{(C only)} @gol
3914 -Wuninitialized @gol
3915 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3916 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3917 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3920 The option @option{-Wextra} also prints warning messages for the
3926 A pointer is compared against integer zero with @code{<}, @code{<=},
3927 @code{>}, or @code{>=}.
3930 (C++ only) An enumerator and a non-enumerator both appear in a
3931 conditional expression.
3934 (C++ only) Ambiguous virtual bases.
3937 (C++ only) Subscripting an array that has been declared @code{register}.
3940 (C++ only) Taking the address of a variable that has been declared
3944 (C++ only) A base class is not initialized in the copy constructor
3949 @item -Wchar-subscripts
3950 @opindex Wchar-subscripts
3951 @opindex Wno-char-subscripts
3952 Warn if an array subscript has type @code{char}. This is a common cause
3953 of error, as programmers often forget that this type is signed on some
3955 This warning is enabled by @option{-Wall}.
3959 Warn about an invalid memory access that is found by Pointer Bounds Checker
3960 (@option{-fcheck-pointer-bounds}).
3962 @item -Wno-coverage-mismatch
3963 @opindex Wno-coverage-mismatch
3964 Warn if feedback profiles do not match when using the
3965 @option{-fprofile-use} option.
3966 If a source file is changed between compiling with @option{-fprofile-gen} and
3967 with @option{-fprofile-use}, the files with the profile feedback can fail
3968 to match the source file and GCC cannot use the profile feedback
3969 information. By default, this warning is enabled and is treated as an
3970 error. @option{-Wno-coverage-mismatch} can be used to disable the
3971 warning or @option{-Wno-error=coverage-mismatch} can be used to
3972 disable the error. Disabling the error for this warning can result in
3973 poorly optimized code and is useful only in the
3974 case of very minor changes such as bug fixes to an existing code-base.
3975 Completely disabling the warning is not recommended.
3978 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3980 Suppress warning messages emitted by @code{#warning} directives.
3982 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3983 @opindex Wdouble-promotion
3984 @opindex Wno-double-promotion
3985 Give a warning when a value of type @code{float} is implicitly
3986 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3987 floating-point unit implement @code{float} in hardware, but emulate
3988 @code{double} in software. On such a machine, doing computations
3989 using @code{double} values is much more expensive because of the
3990 overhead required for software emulation.
3992 It is easy to accidentally do computations with @code{double} because
3993 floating-point literals are implicitly of type @code{double}. For
3997 float area(float radius)
3999 return 3.14159 * radius * radius;
4003 the compiler performs the entire computation with @code{double}
4004 because the floating-point literal is a @code{double}.
4006 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4007 @opindex Wduplicate-decl-specifier
4008 @opindex Wno-duplicate-decl-specifier
4009 Warn if a declaration has duplicate @code{const}, @code{volatile},
4010 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4014 @itemx -Wformat=@var{n}
4017 @opindex ffreestanding
4018 @opindex fno-builtin
4020 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4021 the arguments supplied have types appropriate to the format string
4022 specified, and that the conversions specified in the format string make
4023 sense. This includes standard functions, and others specified by format
4024 attributes (@pxref{Function Attributes}), in the @code{printf},
4025 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4026 not in the C standard) families (or other target-specific families).
4027 Which functions are checked without format attributes having been
4028 specified depends on the standard version selected, and such checks of
4029 functions without the attribute specified are disabled by
4030 @option{-ffreestanding} or @option{-fno-builtin}.
4032 The formats are checked against the format features supported by GNU
4033 libc version 2.2. These include all ISO C90 and C99 features, as well
4034 as features from the Single Unix Specification and some BSD and GNU
4035 extensions. Other library implementations may not support all these
4036 features; GCC does not support warning about features that go beyond a
4037 particular library's limitations. However, if @option{-Wpedantic} is used
4038 with @option{-Wformat}, warnings are given about format features not
4039 in the selected standard version (but not for @code{strfmon} formats,
4040 since those are not in any version of the C standard). @xref{C Dialect
4041 Options,,Options Controlling C Dialect}.
4048 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4049 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4050 @option{-Wformat} also checks for null format arguments for several
4051 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4052 aspects of this level of format checking can be disabled by the
4053 options: @option{-Wno-format-contains-nul},
4054 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4055 @option{-Wformat} is enabled by @option{-Wall}.
4057 @item -Wno-format-contains-nul
4058 @opindex Wno-format-contains-nul
4059 @opindex Wformat-contains-nul
4060 If @option{-Wformat} is specified, do not warn about format strings that
4063 @item -Wno-format-extra-args
4064 @opindex Wno-format-extra-args
4065 @opindex Wformat-extra-args
4066 If @option{-Wformat} is specified, do not warn about excess arguments to a
4067 @code{printf} or @code{scanf} format function. The C standard specifies
4068 that such arguments are ignored.
4070 Where the unused arguments lie between used arguments that are
4071 specified with @samp{$} operand number specifications, normally
4072 warnings are still given, since the implementation could not know what
4073 type to pass to @code{va_arg} to skip the unused arguments. However,
4074 in the case of @code{scanf} formats, this option suppresses the
4075 warning if the unused arguments are all pointers, since the Single
4076 Unix Specification says that such unused arguments are allowed.
4078 @item -Wformat-overflow
4079 @itemx -Wformat-overflow=@var{level}
4080 @opindex Wformat-overflow
4081 @opindex Wno-format-overflow
4082 Warn about calls to formatted input/output functions such as @code{sprintf}
4083 and @code{vsprintf} that might overflow the destination buffer. When the
4084 exact number of bytes written by a format directive cannot be determined
4085 at compile-time it is estimated based on heuristics that depend on the
4086 @var{level} argument and on optimization. While enabling optimization
4087 will in most cases improve the accuracy of the warning, it may also
4088 result in false positives.
4091 @item -Wformat-overflow
4092 @item -Wformat-overflow=1
4093 @opindex Wformat-overflow
4094 @opindex Wno-format-overflow
4095 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4096 employs a conservative approach that warns only about calls that most
4097 likely overflow the buffer. At this level, numeric arguments to format
4098 directives with unknown values are assumed to have the value of one, and
4099 strings of unknown length to be empty. Numeric arguments that are known
4100 to be bounded to a subrange of their type, or string arguments whose output
4101 is bounded either by their directive's precision or by a finite set of
4102 string literals, are assumed to take on the value within the range that
4103 results in the most bytes on output. For example, the call to @code{sprintf}
4104 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4105 the terminating NUL character (@code{'\0'}) appended by the function
4106 to the destination buffer will be written past its end. Increasing
4107 the size of the buffer by a single byte is sufficient to avoid the
4108 warning, though it may not be sufficient to avoid the overflow.
4111 void f (int a, int b)
4114 sprintf (buf, "a = %i, b = %i\n", a, b);
4118 @item -Wformat-overflow=2
4119 Level @var{2} warns also about calls that might overflow the destination
4120 buffer given an argument of sufficient length or magnitude. At level
4121 @var{2}, unknown numeric arguments are assumed to have the minimum
4122 representable value for signed types with a precision greater than 1, and
4123 the maximum representable value otherwise. Unknown string arguments whose
4124 length cannot be assumed to be bounded either by the directive's precision,
4125 or by a finite set of string literals they may evaluate to, or the character
4126 array they may point to, are assumed to be 1 character long.
4128 At level @var{2}, the call in the example above is again diagnosed, but
4129 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4130 @code{%i} directive will write some of its digits beyond the end of
4131 the destination buffer. To make the call safe regardless of the values
4132 of the two variables, the size of the destination buffer must be increased
4133 to at least 34 bytes. GCC includes the minimum size of the buffer in
4134 an informational note following the warning.
4136 An alternative to increasing the size of the destination buffer is to
4137 constrain the range of formatted values. The maximum length of string
4138 arguments can be bounded by specifying the precision in the format
4139 directive. When numeric arguments of format directives can be assumed
4140 to be bounded by less than the precision of their type, choosing
4141 an appropriate length modifier to the format specifier will reduce
4142 the required buffer size. For example, if @var{a} and @var{b} in the
4143 example above can be assumed to be within the precision of
4144 the @code{short int} type then using either the @code{%hi} format
4145 directive or casting the argument to @code{short} reduces the maximum
4146 required size of the buffer to 24 bytes.
4149 void f (int a, int b)
4152 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4157 @item -Wno-format-zero-length
4158 @opindex Wno-format-zero-length
4159 @opindex Wformat-zero-length
4160 If @option{-Wformat} is specified, do not warn about zero-length formats.
4161 The C standard specifies that zero-length formats are allowed.
4166 Enable @option{-Wformat} plus additional format checks. Currently
4167 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4170 @item -Wformat-nonliteral
4171 @opindex Wformat-nonliteral
4172 @opindex Wno-format-nonliteral
4173 If @option{-Wformat} is specified, also warn if the format string is not a
4174 string literal and so cannot be checked, unless the format function
4175 takes its format arguments as a @code{va_list}.
4177 @item -Wformat-security
4178 @opindex Wformat-security
4179 @opindex Wno-format-security
4180 If @option{-Wformat} is specified, also warn about uses of format
4181 functions that represent possible security problems. At present, this
4182 warns about calls to @code{printf} and @code{scanf} functions where the
4183 format string is not a string literal and there are no format arguments,
4184 as in @code{printf (foo);}. This may be a security hole if the format
4185 string came from untrusted input and contains @samp{%n}. (This is
4186 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4187 in future warnings may be added to @option{-Wformat-security} that are not
4188 included in @option{-Wformat-nonliteral}.)
4190 @item -Wformat-signedness
4191 @opindex Wformat-signedness
4192 @opindex Wno-format-signedness
4193 If @option{-Wformat} is specified, also warn if the format string
4194 requires an unsigned argument and the argument is signed and vice versa.
4196 @item -Wformat-truncation
4197 @itemx -Wformat-truncation=@var{level}
4198 @opindex Wformat-truncation
4199 @opindex Wno-format-truncation
4200 Warn about calls to formatted input/output functions such as @code{snprintf}
4201 and @code{vsnprintf} that might result in output truncation. When the exact
4202 number of bytes written by a format directive cannot be determined at
4203 compile-time it is estimated based on heuristics that depend on
4204 the @var{level} argument and on optimization. While enabling optimization
4205 will in most cases improve the accuracy of the warning, it may also result
4206 in false positives. Except as noted otherwise, the option uses the same
4207 logic @option{-Wformat-overflow}.
4210 @item -Wformat-truncation
4211 @item -Wformat-truncation=1
4212 @opindex Wformat-truncation
4213 @opindex Wno-format-overflow
4214 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4215 employs a conservative approach that warns only about calls to bounded
4216 functions whose return value is unused and that will most likely result
4217 in output truncation.
4219 @item -Wformat-truncation=2
4220 Level @var{2} warns also about calls to bounded functions whose return
4221 value is used and that might result in truncation given an argument of
4222 sufficient length or magnitude.
4226 @opindex Wformat-y2k
4227 @opindex Wno-format-y2k
4228 If @option{-Wformat} is specified, also warn about @code{strftime}
4229 formats that may yield only a two-digit year.
4234 @opindex Wno-nonnull
4235 Warn about passing a null pointer for arguments marked as
4236 requiring a non-null value by the @code{nonnull} function attribute.
4238 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4239 can be disabled with the @option{-Wno-nonnull} option.
4241 @item -Wnonnull-compare
4242 @opindex Wnonnull-compare
4243 @opindex Wno-nonnull-compare
4244 Warn when comparing an argument marked with the @code{nonnull}
4245 function attribute against null inside the function.
4247 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4248 can be disabled with the @option{-Wno-nonnull-compare} option.
4250 @item -Wnull-dereference
4251 @opindex Wnull-dereference
4252 @opindex Wno-null-dereference
4253 Warn if the compiler detects paths that trigger erroneous or
4254 undefined behavior due to dereferencing a null pointer. This option
4255 is only active when @option{-fdelete-null-pointer-checks} is active,
4256 which is enabled by optimizations in most targets. The precision of
4257 the warnings depends on the optimization options used.
4259 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4261 @opindex Wno-init-self
4262 Warn about uninitialized variables that are initialized with themselves.
4263 Note this option can only be used with the @option{-Wuninitialized} option.
4265 For example, GCC warns about @code{i} being uninitialized in the
4266 following snippet only when @option{-Winit-self} has been specified:
4277 This warning is enabled by @option{-Wall} in C++.
4279 @item -Wimplicit-int @r{(C and Objective-C only)}
4280 @opindex Wimplicit-int
4281 @opindex Wno-implicit-int
4282 Warn when a declaration does not specify a type.
4283 This warning is enabled by @option{-Wall}.
4285 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4286 @opindex Wimplicit-function-declaration
4287 @opindex Wno-implicit-function-declaration
4288 Give a warning whenever a function is used before being declared. In
4289 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4290 enabled by default and it is made into an error by
4291 @option{-pedantic-errors}. This warning is also enabled by
4294 @item -Wimplicit @r{(C and Objective-C only)}
4296 @opindex Wno-implicit
4297 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4298 This warning is enabled by @option{-Wall}.
4300 @item -Wimplicit-fallthrough
4301 @opindex Wimplicit-fallthrough
4302 @opindex Wno-implicit-fallthrough
4303 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4304 and @option{-Wno-implicit-fallthrough} is the same as
4305 @option{-Wimplicit-fallthrough=0}.
4307 @item -Wimplicit-fallthrough=@var{n}
4308 @opindex Wimplicit-fallthrough=
4309 Warn when a switch case falls through. For example:
4327 This warning does not warn when the last statement of a case cannot
4328 fall through, e.g. when there is a return statement or a call to function
4329 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4330 also takes into account control flow statements, such as ifs, and only
4331 warns when appropriate. E.g.@:
4341 @} else if (i < 1) @{
4351 Since there are occasions where a switch case fall through is desirable,
4352 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4353 to be used along with a null statement to suppress this warning that
4354 would normally occur:
4362 __attribute__ ((fallthrough));
4369 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4370 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4371 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4372 Instead of these attributes, it is also possible to add a fallthrough comment
4373 to silence the warning. The whole body of the C or C++ style comment should
4374 match the given regular expressions listed below. The option argument @var{n}
4375 specifies what kind of comments are accepted:
4379 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4381 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4382 expression, any comment is used as fallthrough comment.
4384 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4385 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4387 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4388 following regular expressions:
4392 @item @code{-fallthrough}
4394 @item @code{@@fallthrough@@}
4396 @item @code{lint -fallthrough[ \t]*}
4398 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4400 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4402 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4406 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4407 following regular expressions:
4411 @item @code{-fallthrough}
4413 @item @code{@@fallthrough@@}
4415 @item @code{lint -fallthrough[ \t]*}
4417 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4421 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4422 fallthrough comments, only attributes disable the warning.
4426 The comment needs to be followed after optional whitespace and other comments
4427 by @code{case} or @code{default} keywords or by a user label that precedes some
4428 @code{case} or @code{default} label.
4443 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4445 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4446 @opindex Wif-not-aligned
4447 @opindex Wno-if-not-aligned
4448 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4449 should be issued. This is is enabled by default.
4450 Use @option{-Wno-if-not-aligned} to disable it.
4452 @item -Wignored-qualifiers @r{(C and C++ only)}
4453 @opindex Wignored-qualifiers
4454 @opindex Wno-ignored-qualifiers
4455 Warn if the return type of a function has a type qualifier
4456 such as @code{const}. For ISO C such a type qualifier has no effect,
4457 since the value returned by a function is not an lvalue.
4458 For C++, the warning is only emitted for scalar types or @code{void}.
4459 ISO C prohibits qualified @code{void} return types on function
4460 definitions, so such return types always receive a warning
4461 even without this option.
4463 This warning is also enabled by @option{-Wextra}.
4465 @item -Wignored-attributes @r{(C and C++ only)}
4466 @opindex Wignored-attributes
4467 @opindex Wno-ignored-attributes
4468 Warn when an attribute is ignored. This is different from the
4469 @option{-Wattributes} option in that it warns whenever the compiler decides
4470 to drop an attribute, not that the attribute is either unknown, used in a
4471 wrong place, etc. This warning is enabled by default.
4476 Warn if the type of @code{main} is suspicious. @code{main} should be
4477 a function with external linkage, returning int, taking either zero
4478 arguments, two, or three arguments of appropriate types. This warning
4479 is enabled by default in C++ and is enabled by either @option{-Wall}
4480 or @option{-Wpedantic}.
4482 @item -Wmisleading-indentation @r{(C and C++ only)}
4483 @opindex Wmisleading-indentation
4484 @opindex Wno-misleading-indentation
4485 Warn when the indentation of the code does not reflect the block structure.
4486 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4487 @code{for} clauses with a guarded statement that does not use braces,
4488 followed by an unguarded statement with the same indentation.
4490 In the following example, the call to ``bar'' is misleadingly indented as
4491 if it were guarded by the ``if'' conditional.
4494 if (some_condition ())
4496 bar (); /* Gotcha: this is not guarded by the "if". */
4499 In the case of mixed tabs and spaces, the warning uses the
4500 @option{-ftabstop=} option to determine if the statements line up
4503 The warning is not issued for code involving multiline preprocessor logic
4504 such as the following example.
4509 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4515 The warning is not issued after a @code{#line} directive, since this
4516 typically indicates autogenerated code, and no assumptions can be made
4517 about the layout of the file that the directive references.
4519 This warning is enabled by @option{-Wall} in C and C++.
4521 @item -Wmissing-braces
4522 @opindex Wmissing-braces
4523 @opindex Wno-missing-braces
4524 Warn if an aggregate or union initializer is not fully bracketed. In
4525 the following example, the initializer for @code{a} is not fully
4526 bracketed, but that for @code{b} is fully bracketed. This warning is
4527 enabled by @option{-Wall} in C.
4530 int a[2][2] = @{ 0, 1, 2, 3 @};
4531 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4534 This warning is enabled by @option{-Wall}.
4536 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4537 @opindex Wmissing-include-dirs
4538 @opindex Wno-missing-include-dirs
4539 Warn if a user-supplied include directory does not exist.
4541 @item -Wmultistatement-macros
4542 @opindex Wmultistatement-macros
4543 @opindex Wno-multistatement-macros
4544 Warn about unsafe multiple statement macros that appear to be guarded
4545 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
4546 @code{while}, in which only the first statement is actually guarded after
4547 the macro is expanded.
4552 #define DOIT x++; y++
4557 will increment @code{y} unconditionally, not just when @code{c} holds.
4558 The can usually be fixed by wrapping the macro in a do-while loop:
4560 #define DOIT do @{ x++; y++; @} while (0)
4565 This warning is enabled by @option{-Wall} in C and C++.
4568 @opindex Wparentheses
4569 @opindex Wno-parentheses
4570 Warn if parentheses are omitted in certain contexts, such
4571 as when there is an assignment in a context where a truth value
4572 is expected, or when operators are nested whose precedence people
4573 often get confused about.
4575 Also warn if a comparison like @code{x<=y<=z} appears; this is
4576 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4577 interpretation from that of ordinary mathematical notation.
4579 Also warn for dangerous uses of the GNU extension to
4580 @code{?:} with omitted middle operand. When the condition
4581 in the @code{?}: operator is a boolean expression, the omitted value is
4582 always 1. Often programmers expect it to be a value computed
4583 inside the conditional expression instead.
4585 For C++ this also warns for some cases of unnecessary parentheses in
4586 declarations, which can indicate an attempt at a function call instead
4590 // Declares a local variable called mymutex.
4591 std::unique_lock<std::mutex> (mymutex);
4592 // User meant std::unique_lock<std::mutex> lock (mymutex);
4596 This warning is enabled by @option{-Wall}.
4598 @item -Wsequence-point
4599 @opindex Wsequence-point
4600 @opindex Wno-sequence-point
4601 Warn about code that may have undefined semantics because of violations
4602 of sequence point rules in the C and C++ standards.
4604 The C and C++ standards define the order in which expressions in a C/C++
4605 program are evaluated in terms of @dfn{sequence points}, which represent
4606 a partial ordering between the execution of parts of the program: those
4607 executed before the sequence point, and those executed after it. These
4608 occur after the evaluation of a full expression (one which is not part
4609 of a larger expression), after the evaluation of the first operand of a
4610 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4611 function is called (but after the evaluation of its arguments and the
4612 expression denoting the called function), and in certain other places.
4613 Other than as expressed by the sequence point rules, the order of
4614 evaluation of subexpressions of an expression is not specified. All
4615 these rules describe only a partial order rather than a total order,
4616 since, for example, if two functions are called within one expression
4617 with no sequence point between them, the order in which the functions
4618 are called is not specified. However, the standards committee have
4619 ruled that function calls do not overlap.
4621 It is not specified when between sequence points modifications to the
4622 values of objects take effect. Programs whose behavior depends on this
4623 have undefined behavior; the C and C++ standards specify that ``Between
4624 the previous and next sequence point an object shall have its stored
4625 value modified at most once by the evaluation of an expression.
4626 Furthermore, the prior value shall be read only to determine the value
4627 to be stored.''. If a program breaks these rules, the results on any
4628 particular implementation are entirely unpredictable.
4630 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4631 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4632 diagnosed by this option, and it may give an occasional false positive
4633 result, but in general it has been found fairly effective at detecting
4634 this sort of problem in programs.
4636 The C++17 standard will define the order of evaluation of operands in
4637 more cases: in particular it requires that the right-hand side of an
4638 assignment be evaluated before the left-hand side, so the above
4639 examples are no longer undefined. But this warning will still warn
4640 about them, to help people avoid writing code that is undefined in C
4641 and earlier revisions of C++.
4643 The standard is worded confusingly, therefore there is some debate
4644 over the precise meaning of the sequence point rules in subtle cases.
4645 Links to discussions of the problem, including proposed formal
4646 definitions, may be found on the GCC readings page, at
4647 @uref{http://gcc.gnu.org/@/readings.html}.
4649 This warning is enabled by @option{-Wall} for C and C++.
4651 @item -Wno-return-local-addr
4652 @opindex Wno-return-local-addr
4653 @opindex Wreturn-local-addr
4654 Do not warn about returning a pointer (or in C++, a reference) to a
4655 variable that goes out of scope after the function returns.
4658 @opindex Wreturn-type
4659 @opindex Wno-return-type
4660 Warn whenever a function is defined with a return type that defaults
4661 to @code{int}. Also warn about any @code{return} statement with no
4662 return value in a function whose return type is not @code{void}
4663 (falling off the end of the function body is considered returning
4666 For C only, warn about a @code{return} statement with an expression in a
4667 function whose return type is @code{void}, unless the expression type is
4668 also @code{void}. As a GNU extension, the latter case is accepted
4669 without a warning unless @option{-Wpedantic} is used.
4671 For C++, a function without return type always produces a diagnostic
4672 message, even when @option{-Wno-return-type} is specified. The only
4673 exceptions are @code{main} and functions defined in system headers.
4675 This warning is enabled by @option{-Wall}.
4677 @item -Wshift-count-negative
4678 @opindex Wshift-count-negative
4679 @opindex Wno-shift-count-negative
4680 Warn if shift count is negative. This warning is enabled by default.
4682 @item -Wshift-count-overflow
4683 @opindex Wshift-count-overflow
4684 @opindex Wno-shift-count-overflow
4685 Warn if shift count >= width of type. This warning is enabled by default.
4687 @item -Wshift-negative-value
4688 @opindex Wshift-negative-value
4689 @opindex Wno-shift-negative-value
4690 Warn if left shifting a negative value. This warning is enabled by
4691 @option{-Wextra} in C99 and C++11 modes (and newer).
4693 @item -Wshift-overflow
4694 @itemx -Wshift-overflow=@var{n}
4695 @opindex Wshift-overflow
4696 @opindex Wno-shift-overflow
4697 Warn about left shift overflows. This warning is enabled by
4698 default in C99 and C++11 modes (and newer).
4701 @item -Wshift-overflow=1
4702 This is the warning level of @option{-Wshift-overflow} and is enabled
4703 by default in C99 and C++11 modes (and newer). This warning level does
4704 not warn about left-shifting 1 into the sign bit. (However, in C, such
4705 an overflow is still rejected in contexts where an integer constant expression
4708 @item -Wshift-overflow=2
4709 This warning level also warns about left-shifting 1 into the sign bit,
4710 unless C++14 mode is active.
4716 Warn whenever a @code{switch} statement has an index of enumerated type
4717 and lacks a @code{case} for one or more of the named codes of that
4718 enumeration. (The presence of a @code{default} label prevents this
4719 warning.) @code{case} labels outside the enumeration range also
4720 provoke warnings when this option is used (even if there is a
4721 @code{default} label).
4722 This warning is enabled by @option{-Wall}.
4724 @item -Wswitch-default
4725 @opindex Wswitch-default
4726 @opindex Wno-switch-default
4727 Warn whenever a @code{switch} statement does not have a @code{default}
4731 @opindex Wswitch-enum
4732 @opindex Wno-switch-enum
4733 Warn whenever a @code{switch} statement has an index of enumerated type
4734 and lacks a @code{case} for one or more of the named codes of that
4735 enumeration. @code{case} labels outside the enumeration range also
4736 provoke warnings when this option is used. The only difference
4737 between @option{-Wswitch} and this option is that this option gives a
4738 warning about an omitted enumeration code even if there is a
4739 @code{default} label.
4742 @opindex Wswitch-bool
4743 @opindex Wno-switch-bool
4744 Warn whenever a @code{switch} statement has an index of boolean type
4745 and the case values are outside the range of a boolean type.
4746 It is possible to suppress this warning by casting the controlling
4747 expression to a type other than @code{bool}. For example:
4750 switch ((int) (a == 4))
4756 This warning is enabled by default for C and C++ programs.
4758 @item -Wswitch-unreachable
4759 @opindex Wswitch-unreachable
4760 @opindex Wno-switch-unreachable
4761 Warn whenever a @code{switch} statement contains statements between the
4762 controlling expression and the first case label, which will never be
4763 executed. For example:
4775 @option{-Wswitch-unreachable} does not warn if the statement between the
4776 controlling expression and the first case label is just a declaration:
4789 This warning is enabled by default for C and C++ programs.
4791 @item -Wsync-nand @r{(C and C++ only)}
4793 @opindex Wno-sync-nand
4794 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4795 built-in functions are used. These functions changed semantics in GCC 4.4.
4797 @item -Wunused-but-set-parameter
4798 @opindex Wunused-but-set-parameter
4799 @opindex Wno-unused-but-set-parameter
4800 Warn whenever a function parameter is assigned to, but otherwise unused
4801 (aside from its declaration).
4803 To suppress this warning use the @code{unused} attribute
4804 (@pxref{Variable Attributes}).
4806 This warning is also enabled by @option{-Wunused} together with
4809 @item -Wunused-but-set-variable
4810 @opindex Wunused-but-set-variable
4811 @opindex Wno-unused-but-set-variable
4812 Warn whenever a local variable is assigned to, but otherwise unused
4813 (aside from its declaration).
4814 This warning is enabled by @option{-Wall}.
4816 To suppress this warning use the @code{unused} attribute
4817 (@pxref{Variable Attributes}).
4819 This warning is also enabled by @option{-Wunused}, which is enabled
4822 @item -Wunused-function
4823 @opindex Wunused-function
4824 @opindex Wno-unused-function
4825 Warn whenever a static function is declared but not defined or a
4826 non-inline static function is unused.
4827 This warning is enabled by @option{-Wall}.
4829 @item -Wunused-label
4830 @opindex Wunused-label
4831 @opindex Wno-unused-label
4832 Warn whenever a label is declared but not used.
4833 This warning is enabled by @option{-Wall}.
4835 To suppress this warning use the @code{unused} attribute
4836 (@pxref{Variable Attributes}).
4838 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4839 @opindex Wunused-local-typedefs
4840 Warn when a typedef locally defined in a function is not used.
4841 This warning is enabled by @option{-Wall}.
4843 @item -Wunused-parameter
4844 @opindex Wunused-parameter
4845 @opindex Wno-unused-parameter
4846 Warn whenever a function parameter is unused aside from its declaration.
4848 To suppress this warning use the @code{unused} attribute
4849 (@pxref{Variable Attributes}).
4851 @item -Wno-unused-result
4852 @opindex Wunused-result
4853 @opindex Wno-unused-result
4854 Do not warn if a caller of a function marked with attribute
4855 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4856 its return value. The default is @option{-Wunused-result}.
4858 @item -Wunused-variable
4859 @opindex Wunused-variable
4860 @opindex Wno-unused-variable
4861 Warn whenever a local or static variable is unused aside from its
4862 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4863 but not for C++. This warning is enabled by @option{-Wall}.
4865 To suppress this warning use the @code{unused} attribute
4866 (@pxref{Variable Attributes}).
4868 @item -Wunused-const-variable
4869 @itemx -Wunused-const-variable=@var{n}
4870 @opindex Wunused-const-variable
4871 @opindex Wno-unused-const-variable
4872 Warn whenever a constant static variable is unused aside from its declaration.
4873 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4874 for C, but not for C++. In C this declares variable storage, but in C++ this
4875 is not an error since const variables take the place of @code{#define}s.
4877 To suppress this warning use the @code{unused} attribute
4878 (@pxref{Variable Attributes}).
4881 @item -Wunused-const-variable=1
4882 This is the warning level that is enabled by @option{-Wunused-variable} for
4883 C. It warns only about unused static const variables defined in the main
4884 compilation unit, but not about static const variables declared in any
4887 @item -Wunused-const-variable=2
4888 This warning level also warns for unused constant static variables in
4889 headers (excluding system headers). This is the warning level of
4890 @option{-Wunused-const-variable} and must be explicitly requested since
4891 in C++ this isn't an error and in C it might be harder to clean up all
4895 @item -Wunused-value
4896 @opindex Wunused-value
4897 @opindex Wno-unused-value
4898 Warn whenever a statement computes a result that is explicitly not
4899 used. To suppress this warning cast the unused expression to
4900 @code{void}. This includes an expression-statement or the left-hand
4901 side of a comma expression that contains no side effects. For example,
4902 an expression such as @code{x[i,j]} causes a warning, while
4903 @code{x[(void)i,j]} does not.
4905 This warning is enabled by @option{-Wall}.
4910 All the above @option{-Wunused} options combined.
4912 In order to get a warning about an unused function parameter, you must
4913 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4914 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4916 @item -Wuninitialized
4917 @opindex Wuninitialized
4918 @opindex Wno-uninitialized
4919 Warn if an automatic variable is used without first being initialized
4920 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4921 warn if a non-static reference or non-static @code{const} member
4922 appears in a class without constructors.
4924 If you want to warn about code that uses the uninitialized value of the
4925 variable in its own initializer, use the @option{-Winit-self} option.
4927 These warnings occur for individual uninitialized or clobbered
4928 elements of structure, union or array variables as well as for
4929 variables that are uninitialized or clobbered as a whole. They do
4930 not occur for variables or elements declared @code{volatile}. Because
4931 these warnings depend on optimization, the exact variables or elements
4932 for which there are warnings depends on the precise optimization
4933 options and version of GCC used.
4935 Note that there may be no warning about a variable that is used only
4936 to compute a value that itself is never used, because such
4937 computations may be deleted by data flow analysis before the warnings
4940 @item -Winvalid-memory-model
4941 @opindex Winvalid-memory-model
4942 @opindex Wno-invalid-memory-model
4943 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4944 and the C11 atomic generic functions with a memory consistency argument
4945 that is either invalid for the operation or outside the range of values
4946 of the @code{memory_order} enumeration. For example, since the
4947 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4948 defined for the relaxed, release, and sequentially consistent memory
4949 orders the following code is diagnosed:
4954 __atomic_store_n (i, 0, memory_order_consume);
4958 @option{-Winvalid-memory-model} is enabled by default.
4960 @item -Wmaybe-uninitialized
4961 @opindex Wmaybe-uninitialized
4962 @opindex Wno-maybe-uninitialized
4963 For an automatic variable, if there exists a path from the function
4964 entry to a use of the variable that is initialized, but there exist
4965 some other paths for which the variable is not initialized, the compiler
4966 emits a warning if it cannot prove the uninitialized paths are not
4967 executed at run time. These warnings are made optional because GCC is
4968 not smart enough to see all the reasons why the code might be correct
4969 in spite of appearing to have an error. Here is one example of how
4990 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4991 always initialized, but GCC doesn't know this. To suppress the
4992 warning, you need to provide a default case with assert(0) or
4995 @cindex @code{longjmp} warnings
4996 This option also warns when a non-volatile automatic variable might be
4997 changed by a call to @code{longjmp}. These warnings as well are possible
4998 only in optimizing compilation.
5000 The compiler sees only the calls to @code{setjmp}. It cannot know
5001 where @code{longjmp} will be called; in fact, a signal handler could
5002 call it at any point in the code. As a result, you may get a warning
5003 even when there is in fact no problem because @code{longjmp} cannot
5004 in fact be called at the place that would cause a problem.
5006 Some spurious warnings can be avoided if you declare all the functions
5007 you use that never return as @code{noreturn}. @xref{Function
5010 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5012 @item -Wunknown-pragmas
5013 @opindex Wunknown-pragmas
5014 @opindex Wno-unknown-pragmas
5015 @cindex warning for unknown pragmas
5016 @cindex unknown pragmas, warning
5017 @cindex pragmas, warning of unknown
5018 Warn when a @code{#pragma} directive is encountered that is not understood by
5019 GCC@. If this command-line option is used, warnings are even issued
5020 for unknown pragmas in system header files. This is not the case if
5021 the warnings are only enabled by the @option{-Wall} command-line option.
5024 @opindex Wno-pragmas
5026 Do not warn about misuses of pragmas, such as incorrect parameters,
5027 invalid syntax, or conflicts between pragmas. See also
5028 @option{-Wunknown-pragmas}.
5030 @item -Wstrict-aliasing
5031 @opindex Wstrict-aliasing
5032 @opindex Wno-strict-aliasing
5033 This option is only active when @option{-fstrict-aliasing} is active.
5034 It warns about code that might break the strict aliasing rules that the
5035 compiler is using for optimization. The warning does not catch all
5036 cases, but does attempt to catch the more common pitfalls. It is
5037 included in @option{-Wall}.
5038 It is equivalent to @option{-Wstrict-aliasing=3}
5040 @item -Wstrict-aliasing=n
5041 @opindex Wstrict-aliasing=n
5042 This option is only active when @option{-fstrict-aliasing} is active.
5043 It warns about code that might break the strict aliasing rules that the
5044 compiler is using for optimization.
5045 Higher levels correspond to higher accuracy (fewer false positives).
5046 Higher levels also correspond to more effort, similar to the way @option{-O}
5048 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5050 Level 1: Most aggressive, quick, least accurate.
5051 Possibly useful when higher levels
5052 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5053 false negatives. However, it has many false positives.
5054 Warns for all pointer conversions between possibly incompatible types,
5055 even if never dereferenced. Runs in the front end only.
5057 Level 2: Aggressive, quick, not too precise.
5058 May still have many false positives (not as many as level 1 though),
5059 and few false negatives (but possibly more than level 1).
5060 Unlike level 1, it only warns when an address is taken. Warns about
5061 incomplete types. Runs in the front end only.
5063 Level 3 (default for @option{-Wstrict-aliasing}):
5064 Should have very few false positives and few false
5065 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5066 Takes care of the common pun+dereference pattern in the front end:
5067 @code{*(int*)&some_float}.
5068 If optimization is enabled, it also runs in the back end, where it deals
5069 with multiple statement cases using flow-sensitive points-to information.
5070 Only warns when the converted pointer is dereferenced.
5071 Does not warn about incomplete types.
5073 @item -Wstrict-overflow
5074 @itemx -Wstrict-overflow=@var{n}
5075 @opindex Wstrict-overflow
5076 @opindex Wno-strict-overflow
5077 This option is only active when signed overflow is undefined.
5078 It warns about cases where the compiler optimizes based on the
5079 assumption that signed overflow does not occur. Note that it does not
5080 warn about all cases where the code might overflow: it only warns
5081 about cases where the compiler implements some optimization. Thus
5082 this warning depends on the optimization level.
5084 An optimization that assumes that signed overflow does not occur is
5085 perfectly safe if the values of the variables involved are such that
5086 overflow never does, in fact, occur. Therefore this warning can
5087 easily give a false positive: a warning about code that is not
5088 actually a problem. To help focus on important issues, several
5089 warning levels are defined. No warnings are issued for the use of
5090 undefined signed overflow when estimating how many iterations a loop
5091 requires, in particular when determining whether a loop will be
5095 @item -Wstrict-overflow=1
5096 Warn about cases that are both questionable and easy to avoid. For
5097 example the compiler simplifies
5098 @code{x + 1 > x} to @code{1}. This level of
5099 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5100 are not, and must be explicitly requested.
5102 @item -Wstrict-overflow=2
5103 Also warn about other cases where a comparison is simplified to a
5104 constant. For example: @code{abs (x) >= 0}. This can only be
5105 simplified when signed integer overflow is undefined, because
5106 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5107 zero. @option{-Wstrict-overflow} (with no level) is the same as
5108 @option{-Wstrict-overflow=2}.
5110 @item -Wstrict-overflow=3
5111 Also warn about other cases where a comparison is simplified. For
5112 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5114 @item -Wstrict-overflow=4
5115 Also warn about other simplifications not covered by the above cases.
5116 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5118 @item -Wstrict-overflow=5
5119 Also warn about cases where the compiler reduces the magnitude of a
5120 constant involved in a comparison. For example: @code{x + 2 > y} is
5121 simplified to @code{x + 1 >= y}. This is reported only at the
5122 highest warning level because this simplification applies to many
5123 comparisons, so this warning level gives a very large number of
5127 @item -Wstringop-overflow
5128 @itemx -Wstringop-overflow=@var{type}
5129 @opindex Wstringop-overflow
5130 @opindex Wno-stringop-overflow
5131 Warn for calls to string manipulation functions such as @code{memcpy} and
5132 @code{strcpy} that are determined to overflow the destination buffer. The
5133 optional argument is one greater than the type of Object Size Checking to
5134 perform to determine the size of the destination. @xref{Object Size Checking}.
5135 The argument is meaningful only for functions that operate on character arrays
5136 but not for raw memory functions like @code{memcpy} which always make use
5137 of Object Size type-0. The option also warns for calls that specify a size
5138 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5139 The option produces the best results with optimization enabled but can detect
5140 a small subset of simple buffer overflows even without optimization in
5141 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5142 correspond to the standard functions. In any case, the option warns about
5143 just a subset of buffer overflows detected by the corresponding overflow
5144 checking built-ins. For example, the option will issue a warning for
5145 the @code{strcpy} call below because it copies at least 5 characters
5146 (the string @code{"blue"} including the terminating NUL) into the buffer
5150 enum Color @{ blue, purple, yellow @};
5151 const char* f (enum Color clr)
5153 static char buf [4];
5157 case blue: str = "blue"; break;
5158 case purple: str = "purple"; break;
5159 case yellow: str = "yellow"; break;
5162 return strcpy (buf, str); // warning here
5166 Option @option{-Wstringop-overflow=2} is enabled by default.
5169 @item -Wstringop-overflow
5170 @item -Wstringop-overflow=1
5171 @opindex Wstringop-overflow
5172 @opindex Wno-stringop-overflow
5173 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5174 to determine the sizes of destination objects. This is the default setting
5175 of the option. At this setting the option will not warn for writes past
5176 the end of subobjects of larger objects accessed by pointers unless the
5177 size of the largest surrounding object is known. When the destination may
5178 be one of several objects it is assumed to be the largest one of them. On
5179 Linux systems, when optimization is enabled at this setting the option warns
5180 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5183 @item -Wstringop-overflow=2
5184 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5185 to determine the sizes of destination objects. At this setting the option
5186 will warn about overflows when writing to members of the largest complete
5187 objects whose exact size is known. It will, however, not warn for excessive
5188 writes to the same members of unknown objects referenced by pointers since
5189 they may point to arrays containing unknown numbers of elements.
5191 @item -Wstringop-overflow=3
5192 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5193 to determine the sizes of destination objects. At this setting the option
5194 warns about overflowing the smallest object or data member. This is the
5195 most restrictive setting of the option that may result in warnings for safe
5198 @item -Wstringop-overflow=4
5199 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5200 to determine the sizes of destination objects. At this setting the option
5201 will warn about overflowing any data members, and when the destination is
5202 one of several objects it uses the size of the largest of them to decide
5203 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5204 setting of the option may result in warnings for benign code.
5207 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{]}
5208 @opindex Wsuggest-attribute=
5209 @opindex Wno-suggest-attribute=
5210 Warn for cases where adding an attribute may be beneficial. The
5211 attributes currently supported are listed below.
5214 @item -Wsuggest-attribute=pure
5215 @itemx -Wsuggest-attribute=const
5216 @itemx -Wsuggest-attribute=noreturn
5217 @opindex Wsuggest-attribute=pure
5218 @opindex Wno-suggest-attribute=pure
5219 @opindex Wsuggest-attribute=const
5220 @opindex Wno-suggest-attribute=const
5221 @opindex Wsuggest-attribute=noreturn
5222 @opindex Wno-suggest-attribute=noreturn
5224 Warn about functions that might be candidates for attributes
5225 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
5226 functions visible in other compilation units or (in the case of @code{pure} and
5227 @code{const}) if it cannot prove that the function returns normally. A function
5228 returns normally if it doesn't contain an infinite loop or return abnormally
5229 by throwing, calling @code{abort} or trapping. This analysis requires option
5230 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
5231 higher. Higher optimization levels improve the accuracy of the analysis.
5233 @item -Wsuggest-attribute=format
5234 @itemx -Wmissing-format-attribute
5235 @opindex Wsuggest-attribute=format
5236 @opindex Wmissing-format-attribute
5237 @opindex Wno-suggest-attribute=format
5238 @opindex Wno-missing-format-attribute
5242 Warn about function pointers that might be candidates for @code{format}
5243 attributes. Note these are only possible candidates, not absolute ones.
5244 GCC guesses that function pointers with @code{format} attributes that
5245 are used in assignment, initialization, parameter passing or return
5246 statements should have a corresponding @code{format} attribute in the
5247 resulting type. I.e.@: the left-hand side of the assignment or
5248 initialization, the type of the parameter variable, or the return type
5249 of the containing function respectively should also have a @code{format}
5250 attribute to avoid the warning.
5252 GCC also warns about function definitions that might be
5253 candidates for @code{format} attributes. Again, these are only
5254 possible candidates. GCC guesses that @code{format} attributes
5255 might be appropriate for any function that calls a function like
5256 @code{vprintf} or @code{vscanf}, but this might not always be the
5257 case, and some functions for which @code{format} attributes are
5258 appropriate may not be detected.
5260 @item -Wsuggest-attribute=cold
5261 @opindex Wsuggest-attribute=cold
5262 @opindex Wno-suggest-attribute=cold
5264 Warn about functions that might be candidates for @code{cold} attribute. This
5265 is based on static detection and generally will only warn about functions which
5266 always leads to a call to another @code{cold} function such as wrappers of
5267 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5270 @item -Wsuggest-final-types
5271 @opindex Wno-suggest-final-types
5272 @opindex Wsuggest-final-types
5273 Warn about types with virtual methods where code quality would be improved
5274 if the type were declared with the C++11 @code{final} specifier,
5276 declared in an anonymous namespace. This allows GCC to more aggressively
5277 devirtualize the polymorphic calls. This warning is more effective with link
5278 time optimization, where the information about the class hierarchy graph is
5281 @item -Wsuggest-final-methods
5282 @opindex Wno-suggest-final-methods
5283 @opindex Wsuggest-final-methods
5284 Warn about virtual methods where code quality would be improved if the method
5285 were declared with the C++11 @code{final} specifier,
5286 or, if possible, its type were
5287 declared in an anonymous namespace or with the @code{final} specifier.
5289 more effective with link-time optimization, where the information about the
5290 class hierarchy graph is more complete. It is recommended to first consider
5291 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5294 @item -Wsuggest-override
5295 Warn about overriding virtual functions that are not marked with the override
5299 @opindex Wno-alloc-zero
5300 @opindex Walloc-zero
5301 Warn about calls to allocation functions decorated with attribute
5302 @code{alloc_size} that specify zero bytes, including those to the built-in
5303 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5304 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5305 when called with a zero size differs among implementations (and in the case
5306 of @code{realloc} has been deprecated) relying on it may result in subtle
5307 portability bugs and should be avoided.
5309 @item -Walloc-size-larger-than=@var{n}
5310 Warn about calls to functions decorated with attribute @code{alloc_size}
5311 that attempt to allocate objects larger than the specified number of bytes,
5312 or where the result of the size computation in an integer type with infinite
5313 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5314 may end in one of the standard suffixes designating a multiple of bytes
5315 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5316 @code{MB} and @code{MiB} for megabyte and mebibyte, and so on.
5317 @xref{Function Attributes}.
5322 This option warns on all uses of @code{alloca} in the source.
5324 @item -Walloca-larger-than=@var{n}
5325 This option warns on calls to @code{alloca} that are not bounded by a
5326 controlling predicate limiting its argument of integer type to at most
5327 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5328 Arguments of non-integer types are considered unbounded even if they
5329 appear to be constrained to the expected range.
5331 For example, a bounded case of @code{alloca} could be:
5334 void func (size_t n)
5345 In the above example, passing @code{-Walloca-larger-than=1000} would not
5346 issue a warning because the call to @code{alloca} is known to be at most
5347 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5348 the compiler would emit a warning.
5350 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5351 controlling predicate constraining its integer argument. For example:
5356 void *p = alloca (n);
5361 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5362 a warning, but this time because of the lack of bounds checking.
5364 Note, that even seemingly correct code involving signed integers could
5368 void func (signed int n)
5378 In the above example, @var{n} could be negative, causing a larger than
5379 expected argument to be implicitly cast into the @code{alloca} call.
5381 This option also warns when @code{alloca} is used in a loop.
5383 This warning is not enabled by @option{-Wall}, and is only active when
5384 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5386 See also @option{-Wvla-larger-than=@var{n}}.
5388 @item -Warray-bounds
5389 @itemx -Warray-bounds=@var{n}
5390 @opindex Wno-array-bounds
5391 @opindex Warray-bounds
5392 This option is only active when @option{-ftree-vrp} is active
5393 (default for @option{-O2} and above). It warns about subscripts to arrays
5394 that are always out of bounds. This warning is enabled by @option{-Wall}.
5397 @item -Warray-bounds=1
5398 This is the warning level of @option{-Warray-bounds} and is enabled
5399 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5401 @item -Warray-bounds=2
5402 This warning level also warns about out of bounds access for
5403 arrays at the end of a struct and for arrays accessed through
5404 pointers. This warning level may give a larger number of
5405 false positives and is deactivated by default.
5408 @item -Wattribute-alias
5409 Warn about declarations using the @code{alias} and similar attributes whose
5410 target is incompatible with the type of the alias. @xref{Function Attributes,
5411 ,Declaring Attributes of Functions}.
5413 @item -Wbool-compare
5414 @opindex Wno-bool-compare
5415 @opindex Wbool-compare
5416 Warn about boolean expression compared with an integer value different from
5417 @code{true}/@code{false}. For instance, the following comparison is
5422 if ((n > 1) == 2) @{ @dots{} @}
5424 This warning is enabled by @option{-Wall}.
5426 @item -Wbool-operation
5427 @opindex Wno-bool-operation
5428 @opindex Wbool-operation
5429 Warn about suspicious operations on expressions of a boolean type. For
5430 instance, bitwise negation of a boolean is very likely a bug in the program.
5431 For C, this warning also warns about incrementing or decrementing a boolean,
5432 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5433 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
5435 This warning is enabled by @option{-Wall}.
5437 @item -Wduplicated-branches
5438 @opindex Wno-duplicated-branches
5439 @opindex Wduplicated-branches
5440 Warn when an if-else has identical branches. This warning detects cases like
5447 It doesn't warn when both branches contain just a null statement. This warning
5448 also warn for conditional operators:
5450 int i = x ? *p : *p;
5453 @item -Wduplicated-cond
5454 @opindex Wno-duplicated-cond
5455 @opindex Wduplicated-cond
5456 Warn about duplicated conditions in an if-else-if chain. For instance,
5457 warn for the following code:
5459 if (p->q != NULL) @{ @dots{} @}
5460 else if (p->q != NULL) @{ @dots{} @}
5463 @item -Wframe-address
5464 @opindex Wno-frame-address
5465 @opindex Wframe-address
5466 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5467 is called with an argument greater than 0. Such calls may return indeterminate
5468 values or crash the program. The warning is included in @option{-Wall}.
5470 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5471 @opindex Wno-discarded-qualifiers
5472 @opindex Wdiscarded-qualifiers
5473 Do not warn if type qualifiers on pointers are being discarded.
5474 Typically, the compiler warns if a @code{const char *} variable is
5475 passed to a function that takes a @code{char *} parameter. This option
5476 can be used to suppress such a warning.
5478 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5479 @opindex Wno-discarded-array-qualifiers
5480 @opindex Wdiscarded-array-qualifiers
5481 Do not warn if type qualifiers on arrays which are pointer targets
5482 are being discarded. Typically, the compiler warns if a
5483 @code{const int (*)[]} variable is passed to a function that
5484 takes a @code{int (*)[]} parameter. This option can be used to
5485 suppress such a warning.
5487 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5488 @opindex Wno-incompatible-pointer-types
5489 @opindex Wincompatible-pointer-types
5490 Do not warn when there is a conversion between pointers that have incompatible
5491 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5492 which warns for pointer argument passing or assignment with different
5495 @item -Wno-int-conversion @r{(C and Objective-C only)}
5496 @opindex Wno-int-conversion
5497 @opindex Wint-conversion
5498 Do not warn about incompatible integer to pointer and pointer to integer
5499 conversions. This warning is about implicit conversions; for explicit
5500 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5501 @option{-Wno-pointer-to-int-cast} may be used.
5503 @item -Wno-div-by-zero
5504 @opindex Wno-div-by-zero
5505 @opindex Wdiv-by-zero
5506 Do not warn about compile-time integer division by zero. Floating-point
5507 division by zero is not warned about, as it can be a legitimate way of
5508 obtaining infinities and NaNs.
5510 @item -Wsystem-headers
5511 @opindex Wsystem-headers
5512 @opindex Wno-system-headers
5513 @cindex warnings from system headers
5514 @cindex system headers, warnings from
5515 Print warning messages for constructs found in system header files.
5516 Warnings from system headers are normally suppressed, on the assumption
5517 that they usually do not indicate real problems and would only make the
5518 compiler output harder to read. Using this command-line option tells
5519 GCC to emit warnings from system headers as if they occurred in user
5520 code. However, note that using @option{-Wall} in conjunction with this
5521 option does @emph{not} warn about unknown pragmas in system
5522 headers---for that, @option{-Wunknown-pragmas} must also be used.
5524 @item -Wtautological-compare
5525 @opindex Wtautological-compare
5526 @opindex Wno-tautological-compare
5527 Warn if a self-comparison always evaluates to true or false. This
5528 warning detects various mistakes such as:
5532 if (i > i) @{ @dots{} @}
5535 This warning also warns about bitwise comparisons that always evaluate
5536 to true or false, for instance:
5538 if ((a & 16) == 10) @{ @dots{} @}
5540 will always be false.
5542 This warning is enabled by @option{-Wall}.
5545 @opindex Wtrampolines
5546 @opindex Wno-trampolines
5547 Warn about trampolines generated for pointers to nested functions.
5548 A trampoline is a small piece of data or code that is created at run
5549 time on the stack when the address of a nested function is taken, and is
5550 used to call the nested function indirectly. For some targets, it is
5551 made up of data only and thus requires no special treatment. But, for
5552 most targets, it is made up of code and thus requires the stack to be
5553 made executable in order for the program to work properly.
5556 @opindex Wfloat-equal
5557 @opindex Wno-float-equal
5558 Warn if floating-point values are used in equality comparisons.
5560 The idea behind this is that sometimes it is convenient (for the
5561 programmer) to consider floating-point values as approximations to
5562 infinitely precise real numbers. If you are doing this, then you need
5563 to compute (by analyzing the code, or in some other way) the maximum or
5564 likely maximum error that the computation introduces, and allow for it
5565 when performing comparisons (and when producing output, but that's a
5566 different problem). In particular, instead of testing for equality, you
5567 should check to see whether the two values have ranges that overlap; and
5568 this is done with the relational operators, so equality comparisons are
5571 @item -Wtraditional @r{(C and Objective-C only)}
5572 @opindex Wtraditional
5573 @opindex Wno-traditional
5574 Warn about certain constructs that behave differently in traditional and
5575 ISO C@. Also warn about ISO C constructs that have no traditional C
5576 equivalent, and/or problematic constructs that should be avoided.
5580 Macro parameters that appear within string literals in the macro body.
5581 In traditional C macro replacement takes place within string literals,
5582 but in ISO C it does not.
5585 In traditional C, some preprocessor directives did not exist.
5586 Traditional preprocessors only considered a line to be a directive
5587 if the @samp{#} appeared in column 1 on the line. Therefore
5588 @option{-Wtraditional} warns about directives that traditional C
5589 understands but ignores because the @samp{#} does not appear as the
5590 first character on the line. It also suggests you hide directives like
5591 @code{#pragma} not understood by traditional C by indenting them. Some
5592 traditional implementations do not recognize @code{#elif}, so this option
5593 suggests avoiding it altogether.
5596 A function-like macro that appears without arguments.
5599 The unary plus operator.
5602 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5603 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5604 constants.) Note, these suffixes appear in macros defined in the system
5605 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5606 Use of these macros in user code might normally lead to spurious
5607 warnings, however GCC's integrated preprocessor has enough context to
5608 avoid warning in these cases.
5611 A function declared external in one block and then used after the end of
5615 A @code{switch} statement has an operand of type @code{long}.
5618 A non-@code{static} function declaration follows a @code{static} one.
5619 This construct is not accepted by some traditional C compilers.
5622 The ISO type of an integer constant has a different width or
5623 signedness from its traditional type. This warning is only issued if
5624 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5625 typically represent bit patterns, are not warned about.
5628 Usage of ISO string concatenation is detected.
5631 Initialization of automatic aggregates.
5634 Identifier conflicts with labels. Traditional C lacks a separate
5635 namespace for labels.
5638 Initialization of unions. If the initializer is zero, the warning is
5639 omitted. This is done under the assumption that the zero initializer in
5640 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5641 initializer warnings and relies on default initialization to zero in the
5645 Conversions by prototypes between fixed/floating-point values and vice
5646 versa. The absence of these prototypes when compiling with traditional
5647 C causes serious problems. This is a subset of the possible
5648 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5651 Use of ISO C style function definitions. This warning intentionally is
5652 @emph{not} issued for prototype declarations or variadic functions
5653 because these ISO C features appear in your code when using
5654 libiberty's traditional C compatibility macros, @code{PARAMS} and
5655 @code{VPARAMS}. This warning is also bypassed for nested functions
5656 because that feature is already a GCC extension and thus not relevant to
5657 traditional C compatibility.
5660 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5661 @opindex Wtraditional-conversion
5662 @opindex Wno-traditional-conversion
5663 Warn if a prototype causes a type conversion that is different from what
5664 would happen to the same argument in the absence of a prototype. This
5665 includes conversions of fixed point to floating and vice versa, and
5666 conversions changing the width or signedness of a fixed-point argument
5667 except when the same as the default promotion.
5669 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5670 @opindex Wdeclaration-after-statement
5671 @opindex Wno-declaration-after-statement
5672 Warn when a declaration is found after a statement in a block. This
5673 construct, known from C++, was introduced with ISO C99 and is by default
5674 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5679 Warn whenever a local variable or type declaration shadows another
5680 variable, parameter, type, class member (in C++), or instance variable
5681 (in Objective-C) or whenever a built-in function is shadowed. Note
5682 that in C++, the compiler warns if a local variable shadows an
5683 explicit typedef, but not if it shadows a struct/class/enum.
5684 Same as @option{-Wshadow=global}.
5686 @item -Wno-shadow-ivar @r{(Objective-C only)}
5687 @opindex Wno-shadow-ivar
5688 @opindex Wshadow-ivar
5689 Do not warn whenever a local variable shadows an instance variable in an
5692 @item -Wshadow=global
5693 @opindex Wshadow=local
5694 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5696 @item -Wshadow=local
5697 @opindex Wshadow=local
5698 Warn when a local variable shadows another local variable or parameter.
5699 This warning is enabled by @option{-Wshadow=global}.
5701 @item -Wshadow=compatible-local
5702 @opindex Wshadow=compatible-local
5703 Warn when a local variable shadows another local variable or parameter
5704 whose type is compatible with that of the shadowing variable. In C++,
5705 type compatibility here means the type of the shadowing variable can be
5706 converted to that of the shadowed variable. The creation of this flag
5707 (in addition to @option{-Wshadow=local}) is based on the idea that when
5708 a local variable shadows another one of incompatible type, it is most
5709 likely intentional, not a bug or typo, as shown in the following example:
5713 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5715 for (int i = 0; i < N; ++i)
5724 Since the two variable @code{i} in the example above have incompatible types,
5725 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5726 Because their types are incompatible, if a programmer accidentally uses one
5727 in place of the other, type checking will catch that and emit an error or
5728 warning. So not warning (about shadowing) in this case will not lead to
5729 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5730 possibly reduce the number of warnings triggered by intentional shadowing.
5732 This warning is enabled by @option{-Wshadow=local}.
5734 @item -Wlarger-than=@var{len}
5735 @opindex Wlarger-than=@var{len}
5736 @opindex Wlarger-than-@var{len}
5737 Warn whenever an object of larger than @var{len} bytes is defined.
5739 @item -Wframe-larger-than=@var{len}
5740 @opindex Wframe-larger-than
5741 Warn if the size of a function frame is larger than @var{len} bytes.
5742 The computation done to determine the stack frame size is approximate
5743 and not conservative.
5744 The actual requirements may be somewhat greater than @var{len}
5745 even if you do not get a warning. In addition, any space allocated
5746 via @code{alloca}, variable-length arrays, or related constructs
5747 is not included by the compiler when determining
5748 whether or not to issue a warning.
5750 @item -Wno-free-nonheap-object
5751 @opindex Wno-free-nonheap-object
5752 @opindex Wfree-nonheap-object
5753 Do not warn when attempting to free an object that was not allocated
5756 @item -Wstack-usage=@var{len}
5757 @opindex Wstack-usage
5758 Warn if the stack usage of a function might be larger than @var{len} bytes.
5759 The computation done to determine the stack usage is conservative.
5760 Any space allocated via @code{alloca}, variable-length arrays, or related
5761 constructs is included by the compiler when determining whether or not to
5764 The message is in keeping with the output of @option{-fstack-usage}.
5768 If the stack usage is fully static but exceeds the specified amount, it's:
5771 warning: stack usage is 1120 bytes
5774 If the stack usage is (partly) dynamic but bounded, it's:
5777 warning: stack usage might be 1648 bytes
5780 If the stack usage is (partly) dynamic and not bounded, it's:
5783 warning: stack usage might be unbounded
5787 @item -Wunsafe-loop-optimizations
5788 @opindex Wunsafe-loop-optimizations
5789 @opindex Wno-unsafe-loop-optimizations
5790 Warn if the loop cannot be optimized because the compiler cannot
5791 assume anything on the bounds of the loop indices. With
5792 @option{-funsafe-loop-optimizations} warn if the compiler makes
5795 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5796 @opindex Wno-pedantic-ms-format
5797 @opindex Wpedantic-ms-format
5798 When used in combination with @option{-Wformat}
5799 and @option{-pedantic} without GNU extensions, this option
5800 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5801 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5802 which depend on the MS runtime.
5805 @opindex Waligned-new
5806 @opindex Wno-aligned-new
5807 Warn about a new-expression of a type that requires greater alignment
5808 than the @code{alignof(std::max_align_t)} but uses an allocation
5809 function without an explicit alignment parameter. This option is
5810 enabled by @option{-Wall}.
5812 Normally this only warns about global allocation functions, but
5813 @option{-Waligned-new=all} also warns about class member allocation
5816 @item -Wplacement-new
5817 @itemx -Wplacement-new=@var{n}
5818 @opindex Wplacement-new
5819 @opindex Wno-placement-new
5820 Warn about placement new expressions with undefined behavior, such as
5821 constructing an object in a buffer that is smaller than the type of
5822 the object. For example, the placement new expression below is diagnosed
5823 because it attempts to construct an array of 64 integers in a buffer only
5829 This warning is enabled by default.
5832 @item -Wplacement-new=1
5833 This is the default warning level of @option{-Wplacement-new}. At this
5834 level the warning is not issued for some strictly undefined constructs that
5835 GCC allows as extensions for compatibility with legacy code. For example,
5836 the following @code{new} expression is not diagnosed at this level even
5837 though it has undefined behavior according to the C++ standard because
5838 it writes past the end of the one-element array.
5840 struct S @{ int n, a[1]; @};
5841 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5842 new (s->a)int [32]();
5845 @item -Wplacement-new=2
5846 At this level, in addition to diagnosing all the same constructs as at level
5847 1, a diagnostic is also issued for placement new expressions that construct
5848 an object in the last member of structure whose type is an array of a single
5849 element and whose size is less than the size of the object being constructed.
5850 While the previous example would be diagnosed, the following construct makes
5851 use of the flexible member array extension to avoid the warning at level 2.
5853 struct S @{ int n, a[]; @};
5854 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5855 new (s->a)int [32]();
5860 @item -Wpointer-arith
5861 @opindex Wpointer-arith
5862 @opindex Wno-pointer-arith
5863 Warn about anything that depends on the ``size of'' a function type or
5864 of @code{void}. GNU C assigns these types a size of 1, for
5865 convenience in calculations with @code{void *} pointers and pointers
5866 to functions. In C++, warn also when an arithmetic operation involves
5867 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5869 @item -Wpointer-compare
5870 @opindex Wpointer-compare
5871 @opindex Wno-pointer-compare
5872 Warn if a pointer is compared with a zero character constant. This usually
5873 means that the pointer was meant to be dereferenced. For example:
5876 const char *p = foo ();
5881 Note that the code above is invalid in C++11.
5883 This warning is enabled by default.
5886 @opindex Wtype-limits
5887 @opindex Wno-type-limits
5888 Warn if a comparison is always true or always false due to the limited
5889 range of the data type, but do not warn for constant expressions. For
5890 example, warn if an unsigned variable is compared against zero with
5891 @code{<} or @code{>=}. This warning is also enabled by
5894 @include cppwarnopts.texi
5896 @item -Wbad-function-cast @r{(C and Objective-C only)}
5897 @opindex Wbad-function-cast
5898 @opindex Wno-bad-function-cast
5899 Warn when a function call is cast to a non-matching type.
5900 For example, warn if a call to a function returning an integer type
5901 is cast to a pointer type.
5903 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5904 @opindex Wc90-c99-compat
5905 @opindex Wno-c90-c99-compat
5906 Warn about features not present in ISO C90, but present in ISO C99.
5907 For instance, warn about use of variable length arrays, @code{long long}
5908 type, @code{bool} type, compound literals, designated initializers, and so
5909 on. This option is independent of the standards mode. Warnings are disabled
5910 in the expression that follows @code{__extension__}.
5912 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5913 @opindex Wc99-c11-compat
5914 @opindex Wno-c99-c11-compat
5915 Warn about features not present in ISO C99, but present in ISO C11.
5916 For instance, warn about use of anonymous structures and unions,
5917 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5918 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5919 and so on. This option is independent of the standards mode. Warnings are
5920 disabled in the expression that follows @code{__extension__}.
5922 @item -Wc++-compat @r{(C and Objective-C only)}
5923 @opindex Wc++-compat
5924 Warn about ISO C constructs that are outside of the common subset of
5925 ISO C and ISO C++, e.g.@: request for implicit conversion from
5926 @code{void *} to a pointer to non-@code{void} type.
5928 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5929 @opindex Wc++11-compat
5930 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5931 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5932 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5933 enabled by @option{-Wall}.
5935 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5936 @opindex Wc++14-compat
5937 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5938 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5940 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
5941 @opindex Wc++17-compat
5942 Warn about C++ constructs whose meaning differs between ISO C++ 2014
5943 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
5947 @opindex Wno-cast-qual
5948 Warn whenever a pointer is cast so as to remove a type qualifier from
5949 the target type. For example, warn if a @code{const char *} is cast
5950 to an ordinary @code{char *}.
5952 Also warn when making a cast that introduces a type qualifier in an
5953 unsafe way. For example, casting @code{char **} to @code{const char **}
5954 is unsafe, as in this example:
5957 /* p is char ** value. */
5958 const char **q = (const char **) p;
5959 /* Assignment of readonly string to const char * is OK. */
5961 /* Now char** pointer points to read-only memory. */
5966 @opindex Wcast-align
5967 @opindex Wno-cast-align
5968 Warn whenever a pointer is cast such that the required alignment of the
5969 target is increased. For example, warn if a @code{char *} is cast to
5970 an @code{int *} on machines where integers can only be accessed at
5971 two- or four-byte boundaries.
5973 @item -Wcast-align=strict
5974 @opindex Wcast-align=strict
5975 Warn whenever a pointer is cast such that the required alignment of the
5976 target is increased. For example, warn if a @code{char *} is cast to
5977 an @code{int *} regardless of the target machine.
5979 @item -Wwrite-strings
5980 @opindex Wwrite-strings
5981 @opindex Wno-write-strings
5982 When compiling C, give string constants the type @code{const
5983 char[@var{length}]} so that copying the address of one into a
5984 non-@code{const} @code{char *} pointer produces a warning. These
5985 warnings help you find at compile time code that can try to write
5986 into a string constant, but only if you have been very careful about
5987 using @code{const} in declarations and prototypes. Otherwise, it is
5988 just a nuisance. This is why we did not make @option{-Wall} request
5991 When compiling C++, warn about the deprecated conversion from string
5992 literals to @code{char *}. This warning is enabled by default for C++
5996 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
5997 @opindex Wcatch-value
5998 @opindex Wno-catch-value
5999 Warn about catch handlers that do not catch via reference.
6000 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6001 warn about polymorphic class types that are caught by value.
6002 With @option{-Wcatch-value=2} warn about all class types that are caught
6003 by value. With @option{-Wcatch-value=3} warn about all types that are
6004 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6008 @opindex Wno-clobbered
6009 Warn for variables that might be changed by @code{longjmp} or
6010 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6012 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6013 @opindex Wconditionally-supported
6014 @opindex Wno-conditionally-supported
6015 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6018 @opindex Wconversion
6019 @opindex Wno-conversion
6020 Warn for implicit conversions that may alter a value. This includes
6021 conversions between real and integer, like @code{abs (x)} when
6022 @code{x} is @code{double}; conversions between signed and unsigned,
6023 like @code{unsigned ui = -1}; and conversions to smaller types, like
6024 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6025 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6026 changed by the conversion like in @code{abs (2.0)}. Warnings about
6027 conversions between signed and unsigned integers can be disabled by
6028 using @option{-Wno-sign-conversion}.
6030 For C++, also warn for confusing overload resolution for user-defined
6031 conversions; and conversions that never use a type conversion
6032 operator: conversions to @code{void}, the same type, a base class or a
6033 reference to them. Warnings about conversions between signed and
6034 unsigned integers are disabled by default in C++ unless
6035 @option{-Wsign-conversion} is explicitly enabled.
6037 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6038 @opindex Wconversion-null
6039 @opindex Wno-conversion-null
6040 Do not warn for conversions between @code{NULL} and non-pointer
6041 types. @option{-Wconversion-null} is enabled by default.
6043 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6044 @opindex Wzero-as-null-pointer-constant
6045 @opindex Wno-zero-as-null-pointer-constant
6046 Warn when a literal @samp{0} is used as null pointer constant. This can
6047 be useful to facilitate the conversion to @code{nullptr} in C++11.
6049 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6050 @opindex Wsubobject-linkage
6051 @opindex Wno-subobject-linkage
6052 Warn if a class type has a base or a field whose type uses the anonymous
6053 namespace or depends on a type with no linkage. If a type A depends on
6054 a type B with no or internal linkage, defining it in multiple
6055 translation units would be an ODR violation because the meaning of B
6056 is different in each translation unit. If A only appears in a single
6057 translation unit, the best way to silence the warning is to give it
6058 internal linkage by putting it in an anonymous namespace as well. The
6059 compiler doesn't give this warning for types defined in the main .C
6060 file, as those are unlikely to have multiple definitions.
6061 @option{-Wsubobject-linkage} is enabled by default.
6063 @item -Wdangling-else
6064 @opindex Wdangling-else
6065 @opindex Wno-dangling-else
6066 Warn about constructions where there may be confusion to which
6067 @code{if} statement an @code{else} branch belongs. Here is an example of
6082 In C/C++, every @code{else} branch belongs to the innermost possible
6083 @code{if} statement, which in this example is @code{if (b)}. This is
6084 often not what the programmer expected, as illustrated in the above
6085 example by indentation the programmer chose. When there is the
6086 potential for this confusion, GCC issues a warning when this flag
6087 is specified. To eliminate the warning, add explicit braces around
6088 the innermost @code{if} statement so there is no way the @code{else}
6089 can belong to the enclosing @code{if}. The resulting code
6106 This warning is enabled by @option{-Wparentheses}.
6110 @opindex Wno-date-time
6111 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6112 are encountered as they might prevent bit-wise-identical reproducible
6115 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6116 @opindex Wdelete-incomplete
6117 @opindex Wno-delete-incomplete
6118 Warn when deleting a pointer to incomplete type, which may cause
6119 undefined behavior at runtime. This warning is enabled by default.
6121 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6122 @opindex Wuseless-cast
6123 @opindex Wno-useless-cast
6124 Warn when an expression is casted to its own type.
6127 @opindex Wempty-body
6128 @opindex Wno-empty-body
6129 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6130 while} statement. This warning is also enabled by @option{-Wextra}.
6132 @item -Wenum-compare
6133 @opindex Wenum-compare
6134 @opindex Wno-enum-compare
6135 Warn about a comparison between values of different enumerated types.
6136 In C++ enumerated type mismatches in conditional expressions are also
6137 diagnosed and the warning is enabled by default. In C this warning is
6138 enabled by @option{-Wall}.
6140 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6141 @opindex Wextra-semi
6142 @opindex Wno-extra-semi
6143 Warn about redundant semicolon after in-class function definition.
6145 @item -Wjump-misses-init @r{(C, Objective-C only)}
6146 @opindex Wjump-misses-init
6147 @opindex Wno-jump-misses-init
6148 Warn if a @code{goto} statement or a @code{switch} statement jumps
6149 forward across the initialization of a variable, or jumps backward to a
6150 label after the variable has been initialized. This only warns about
6151 variables that are initialized when they are declared. This warning is
6152 only supported for C and Objective-C; in C++ this sort of branch is an
6155 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6156 can be disabled with the @option{-Wno-jump-misses-init} option.
6158 @item -Wsign-compare
6159 @opindex Wsign-compare
6160 @opindex Wno-sign-compare
6161 @cindex warning for comparison of signed and unsigned values
6162 @cindex comparison of signed and unsigned values, warning
6163 @cindex signed and unsigned values, comparison warning
6164 Warn when a comparison between signed and unsigned values could produce
6165 an incorrect result when the signed value is converted to unsigned.
6166 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6167 also enabled by @option{-Wextra}.
6169 @item -Wsign-conversion
6170 @opindex Wsign-conversion
6171 @opindex Wno-sign-conversion
6172 Warn for implicit conversions that may change the sign of an integer
6173 value, like assigning a signed integer expression to an unsigned
6174 integer variable. An explicit cast silences the warning. In C, this
6175 option is enabled also by @option{-Wconversion}.
6177 @item -Wfloat-conversion
6178 @opindex Wfloat-conversion
6179 @opindex Wno-float-conversion
6180 Warn for implicit conversions that reduce the precision of a real value.
6181 This includes conversions from real to integer, and from higher precision
6182 real to lower precision real values. This option is also enabled by
6183 @option{-Wconversion}.
6185 @item -Wno-scalar-storage-order
6186 @opindex -Wno-scalar-storage-order
6187 @opindex -Wscalar-storage-order
6188 Do not warn on suspicious constructs involving reverse scalar storage order.
6190 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6191 @opindex Wsized-deallocation
6192 @opindex Wno-sized-deallocation
6193 Warn about a definition of an unsized deallocation function
6195 void operator delete (void *) noexcept;
6196 void operator delete[] (void *) noexcept;
6198 without a definition of the corresponding sized deallocation function
6200 void operator delete (void *, std::size_t) noexcept;
6201 void operator delete[] (void *, std::size_t) noexcept;
6203 or vice versa. Enabled by @option{-Wextra} along with
6204 @option{-fsized-deallocation}.
6206 @item -Wsizeof-pointer-div
6207 @opindex Wsizeof-pointer-div
6208 @opindex Wno-sizeof-pointer-div
6209 Warn for suspicious divisions of two sizeof expressions that divide
6210 the pointer size by the element size, which is the usual way to compute
6211 the array size but won't work out correctly with pointers. This warning
6212 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
6213 not an array, but a pointer. This warning is enabled by @option{-Wall}.
6215 @item -Wsizeof-pointer-memaccess
6216 @opindex Wsizeof-pointer-memaccess
6217 @opindex Wno-sizeof-pointer-memaccess
6218 Warn for suspicious length parameters to certain string and memory built-in
6219 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
6220 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
6221 but a pointer, and suggests a possible fix, or about
6222 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
6225 @item -Wsizeof-array-argument
6226 @opindex Wsizeof-array-argument
6227 @opindex Wno-sizeof-array-argument
6228 Warn when the @code{sizeof} operator is applied to a parameter that is
6229 declared as an array in a function definition. This warning is enabled by
6230 default for C and C++ programs.
6232 @item -Wmemset-elt-size
6233 @opindex Wmemset-elt-size
6234 @opindex Wno-memset-elt-size
6235 Warn for suspicious calls to the @code{memset} built-in function, if the
6236 first argument references an array, and the third argument is a number
6237 equal to the number of elements, but not equal to the size of the array
6238 in memory. This indicates that the user has omitted a multiplication by
6239 the element size. This warning is enabled by @option{-Wall}.
6241 @item -Wmemset-transposed-args
6242 @opindex Wmemset-transposed-args
6243 @opindex Wno-memset-transposed-args
6244 Warn for suspicious calls to the @code{memset} built-in function, if the
6245 second argument is not zero and the third argument is zero. This warns e.g.@
6246 about @code{memset (buf, sizeof buf, 0)} where most probably
6247 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6248 is only emitted if the third argument is literal zero. If it is some
6249 expression that is folded to zero, a cast of zero to some type, etc.,
6250 it is far less likely that the user has mistakenly exchanged the arguments
6251 and no warning is emitted. This warning is enabled by @option{-Wall}.
6255 @opindex Wno-address
6256 Warn about suspicious uses of memory addresses. These include using
6257 the address of a function in a conditional expression, such as
6258 @code{void func(void); if (func)}, and comparisons against the memory
6259 address of a string literal, such as @code{if (x == "abc")}. Such
6260 uses typically indicate a programmer error: the address of a function
6261 always evaluates to true, so their use in a conditional usually
6262 indicate that the programmer forgot the parentheses in a function
6263 call; and comparisons against string literals result in unspecified
6264 behavior and are not portable in C, so they usually indicate that the
6265 programmer intended to use @code{strcmp}. This warning is enabled by
6269 @opindex Wlogical-op
6270 @opindex Wno-logical-op
6271 Warn about suspicious uses of logical operators in expressions.
6272 This includes using logical operators in contexts where a
6273 bit-wise operator is likely to be expected. Also warns when
6274 the operands of a logical operator are the same:
6277 if (a < 0 && a < 0) @{ @dots{} @}
6280 @item -Wlogical-not-parentheses
6281 @opindex Wlogical-not-parentheses
6282 @opindex Wno-logical-not-parentheses
6283 Warn about logical not used on the left hand side operand of a comparison.
6284 This option does not warn if the right operand is considered to be a boolean
6285 expression. Its purpose is to detect suspicious code like the following:
6289 if (!a > 1) @{ @dots{} @}
6292 It is possible to suppress the warning by wrapping the LHS into
6295 if ((!a) > 1) @{ @dots{} @}
6298 This warning is enabled by @option{-Wall}.
6300 @item -Waggregate-return
6301 @opindex Waggregate-return
6302 @opindex Wno-aggregate-return
6303 Warn if any functions that return structures or unions are defined or
6304 called. (In languages where you can return an array, this also elicits
6307 @item -Wno-aggressive-loop-optimizations
6308 @opindex Wno-aggressive-loop-optimizations
6309 @opindex Waggressive-loop-optimizations
6310 Warn if in a loop with constant number of iterations the compiler detects
6311 undefined behavior in some statement during one or more of the iterations.
6313 @item -Wno-attributes
6314 @opindex Wno-attributes
6315 @opindex Wattributes
6316 Do not warn if an unexpected @code{__attribute__} is used, such as
6317 unrecognized attributes, function attributes applied to variables,
6318 etc. This does not stop errors for incorrect use of supported
6321 @item -Wno-builtin-declaration-mismatch
6322 @opindex Wno-builtin-declaration-mismatch
6323 @opindex Wbuiltin-declaration-mismatch
6324 Warn if a built-in function is declared with the wrong signature or
6326 This warning is enabled by default.
6328 @item -Wno-builtin-macro-redefined
6329 @opindex Wno-builtin-macro-redefined
6330 @opindex Wbuiltin-macro-redefined
6331 Do not warn if certain built-in macros are redefined. This suppresses
6332 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6333 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6335 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6336 @opindex Wstrict-prototypes
6337 @opindex Wno-strict-prototypes
6338 Warn if a function is declared or defined without specifying the
6339 argument types. (An old-style function definition is permitted without
6340 a warning if preceded by a declaration that specifies the argument
6343 @item -Wold-style-declaration @r{(C and Objective-C only)}
6344 @opindex Wold-style-declaration
6345 @opindex Wno-old-style-declaration
6346 Warn for obsolescent usages, according to the C Standard, in a
6347 declaration. For example, warn if storage-class specifiers like
6348 @code{static} are not the first things in a declaration. This warning
6349 is also enabled by @option{-Wextra}.
6351 @item -Wold-style-definition @r{(C and Objective-C only)}
6352 @opindex Wold-style-definition
6353 @opindex Wno-old-style-definition
6354 Warn if an old-style function definition is used. A warning is given
6355 even if there is a previous prototype.
6357 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6358 @opindex Wmissing-parameter-type
6359 @opindex Wno-missing-parameter-type
6360 A function parameter is declared without a type specifier in K&R-style
6367 This warning is also enabled by @option{-Wextra}.
6369 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6370 @opindex Wmissing-prototypes
6371 @opindex Wno-missing-prototypes
6372 Warn if a global function is defined without a previous prototype
6373 declaration. This warning is issued even if the definition itself
6374 provides a prototype. Use this option to detect global functions
6375 that do not have a matching prototype declaration in a header file.
6376 This option is not valid for C++ because all function declarations
6377 provide prototypes and a non-matching declaration declares an
6378 overload rather than conflict with an earlier declaration.
6379 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6381 @item -Wmissing-declarations
6382 @opindex Wmissing-declarations
6383 @opindex Wno-missing-declarations
6384 Warn if a global function is defined without a previous declaration.
6385 Do so even if the definition itself provides a prototype.
6386 Use this option to detect global functions that are not declared in
6387 header files. In C, no warnings are issued for functions with previous
6388 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6389 missing prototypes. In C++, no warnings are issued for function templates,
6390 or for inline functions, or for functions in anonymous namespaces.
6392 @item -Wmissing-field-initializers
6393 @opindex Wmissing-field-initializers
6394 @opindex Wno-missing-field-initializers
6398 Warn if a structure's initializer has some fields missing. For
6399 example, the following code causes such a warning, because
6400 @code{x.h} is implicitly zero:
6403 struct s @{ int f, g, h; @};
6404 struct s x = @{ 3, 4 @};
6407 This option does not warn about designated initializers, so the following
6408 modification does not trigger a warning:
6411 struct s @{ int f, g, h; @};
6412 struct s x = @{ .f = 3, .g = 4 @};
6415 In C this option does not warn about the universal zero initializer
6419 struct s @{ int f, g, h; @};
6420 struct s x = @{ 0 @};
6423 Likewise, in C++ this option does not warn about the empty @{ @}
6424 initializer, for example:
6427 struct s @{ int f, g, h; @};
6431 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6432 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6434 @item -Wno-multichar
6435 @opindex Wno-multichar
6437 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6438 Usually they indicate a typo in the user's code, as they have
6439 implementation-defined values, and should not be used in portable code.
6441 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
6442 @opindex Wnormalized=
6443 @opindex Wnormalized
6444 @opindex Wno-normalized
6447 @cindex character set, input normalization
6448 In ISO C and ISO C++, two identifiers are different if they are
6449 different sequences of characters. However, sometimes when characters
6450 outside the basic ASCII character set are used, you can have two
6451 different character sequences that look the same. To avoid confusion,
6452 the ISO 10646 standard sets out some @dfn{normalization rules} which
6453 when applied ensure that two sequences that look the same are turned into
6454 the same sequence. GCC can warn you if you are using identifiers that
6455 have not been normalized; this option controls that warning.
6457 There are four levels of warning supported by GCC@. The default is
6458 @option{-Wnormalized=nfc}, which warns about any identifier that is
6459 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6460 recommended form for most uses. It is equivalent to
6461 @option{-Wnormalized}.
6463 Unfortunately, there are some characters allowed in identifiers by
6464 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6465 identifiers. That is, there's no way to use these symbols in portable
6466 ISO C or C++ and have all your identifiers in NFC@.
6467 @option{-Wnormalized=id} suppresses the warning for these characters.
6468 It is hoped that future versions of the standards involved will correct
6469 this, which is why this option is not the default.
6471 You can switch the warning off for all characters by writing
6472 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6473 only do this if you are using some other normalization scheme (like
6474 ``D''), because otherwise you can easily create bugs that are
6475 literally impossible to see.
6477 Some characters in ISO 10646 have distinct meanings but look identical
6478 in some fonts or display methodologies, especially once formatting has
6479 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6480 LETTER N'', displays just like a regular @code{n} that has been
6481 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6482 normalization scheme to convert all these into a standard form as
6483 well, and GCC warns if your code is not in NFKC if you use
6484 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6485 about every identifier that contains the letter O because it might be
6486 confused with the digit 0, and so is not the default, but may be
6487 useful as a local coding convention if the programming environment
6488 cannot be fixed to display these characters distinctly.
6490 @item -Wno-deprecated
6491 @opindex Wno-deprecated
6492 @opindex Wdeprecated
6493 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6495 @item -Wno-deprecated-declarations
6496 @opindex Wno-deprecated-declarations
6497 @opindex Wdeprecated-declarations
6498 Do not warn about uses of functions (@pxref{Function Attributes}),
6499 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6500 Attributes}) marked as deprecated by using the @code{deprecated}
6504 @opindex Wno-overflow
6506 Do not warn about compile-time overflow in constant expressions.
6511 Warn about One Definition Rule violations during link-time optimization.
6512 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6515 @opindex Wopenm-simd
6516 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6517 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6518 option can be used to relax the cost model.
6520 @item -Woverride-init @r{(C and Objective-C only)}
6521 @opindex Woverride-init
6522 @opindex Wno-override-init
6526 Warn if an initialized field without side effects is overridden when
6527 using designated initializers (@pxref{Designated Inits, , Designated
6530 This warning is included in @option{-Wextra}. To get other
6531 @option{-Wextra} warnings without this one, use @option{-Wextra
6532 -Wno-override-init}.
6534 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6535 @opindex Woverride-init-side-effects
6536 @opindex Wno-override-init-side-effects
6537 Warn if an initialized field with side effects is overridden when
6538 using designated initializers (@pxref{Designated Inits, , Designated
6539 Initializers}). This warning is enabled by default.
6544 Warn if a structure is given the packed attribute, but the packed
6545 attribute has no effect on the layout or size of the structure.
6546 Such structures may be mis-aligned for little benefit. For
6547 instance, in this code, the variable @code{f.x} in @code{struct bar}
6548 is misaligned even though @code{struct bar} does not itself
6549 have the packed attribute:
6556 @} __attribute__((packed));
6564 @item -Wpacked-bitfield-compat
6565 @opindex Wpacked-bitfield-compat
6566 @opindex Wno-packed-bitfield-compat
6567 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6568 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6569 the change can lead to differences in the structure layout. GCC
6570 informs you when the offset of such a field has changed in GCC 4.4.
6571 For example there is no longer a 4-bit padding between field @code{a}
6572 and @code{b} in this structure:
6579 @} __attribute__ ((packed));
6582 This warning is enabled by default. Use
6583 @option{-Wno-packed-bitfield-compat} to disable this warning.
6585 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
6586 @opindex Wpacked-not-aligned
6587 @opindex Wno-packed-not-aligned
6588 Warn if a structure field with explicitly specified alignment in a
6589 packed struct or union is misaligned. For example, a warning will
6590 be issued on @code{struct S}, like, @code{warning: alignment 1 of
6591 'struct S' is less than 8}, in this code:
6595 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
6596 struct __attribute__ ((packed)) S @{
6602 This warning is enabled by @option{-Wall}.
6607 Warn if padding is included in a structure, either to align an element
6608 of the structure or to align the whole structure. Sometimes when this
6609 happens it is possible to rearrange the fields of the structure to
6610 reduce the padding and so make the structure smaller.
6612 @item -Wredundant-decls
6613 @opindex Wredundant-decls
6614 @opindex Wno-redundant-decls
6615 Warn if anything is declared more than once in the same scope, even in
6616 cases where multiple declaration is valid and changes nothing.
6620 @opindex Wno-restrict
6621 Warn when an argument passed to a restrict-qualified parameter
6622 aliases with another argument.
6624 @item -Wnested-externs @r{(C and Objective-C only)}
6625 @opindex Wnested-externs
6626 @opindex Wno-nested-externs
6627 Warn if an @code{extern} declaration is encountered within a function.
6629 @item -Wno-inherited-variadic-ctor
6630 @opindex Winherited-variadic-ctor
6631 @opindex Wno-inherited-variadic-ctor
6632 Suppress warnings about use of C++11 inheriting constructors when the
6633 base class inherited from has a C variadic constructor; the warning is
6634 on by default because the ellipsis is not inherited.
6639 Warn if a function that is declared as inline cannot be inlined.
6640 Even with this option, the compiler does not warn about failures to
6641 inline functions declared in system headers.
6643 The compiler uses a variety of heuristics to determine whether or not
6644 to inline a function. For example, the compiler takes into account
6645 the size of the function being inlined and the amount of inlining
6646 that has already been done in the current function. Therefore,
6647 seemingly insignificant changes in the source program can cause the
6648 warnings produced by @option{-Winline} to appear or disappear.
6650 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6651 @opindex Wno-invalid-offsetof
6652 @opindex Winvalid-offsetof
6653 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6654 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6655 to a non-standard-layout type is undefined. In existing C++ implementations,
6656 however, @code{offsetof} typically gives meaningful results.
6657 This flag is for users who are aware that they are
6658 writing nonportable code and who have deliberately chosen to ignore the
6661 The restrictions on @code{offsetof} may be relaxed in a future version
6662 of the C++ standard.
6664 @item -Wint-in-bool-context
6665 @opindex Wint-in-bool-context
6666 @opindex Wno-int-in-bool-context
6667 Warn for suspicious use of integer values where boolean values are expected,
6668 such as conditional expressions (?:) using non-boolean integer constants in
6669 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6670 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6671 for all kinds of multiplications regardless of the data type.
6672 This warning is enabled by @option{-Wall}.
6674 @item -Wno-int-to-pointer-cast
6675 @opindex Wno-int-to-pointer-cast
6676 @opindex Wint-to-pointer-cast
6677 Suppress warnings from casts to pointer type of an integer of a
6678 different size. In C++, casting to a pointer type of smaller size is
6679 an error. @option{Wint-to-pointer-cast} is enabled by default.
6682 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6683 @opindex Wno-pointer-to-int-cast
6684 @opindex Wpointer-to-int-cast
6685 Suppress warnings from casts from a pointer to an integer type of a
6689 @opindex Winvalid-pch
6690 @opindex Wno-invalid-pch
6691 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6692 the search path but cannot be used.
6696 @opindex Wno-long-long
6697 Warn if @code{long long} type is used. This is enabled by either
6698 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6699 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6701 @item -Wvariadic-macros
6702 @opindex Wvariadic-macros
6703 @opindex Wno-variadic-macros
6704 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6705 alternate syntax is used in ISO C99 mode. This is enabled by either
6706 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6707 messages, use @option{-Wno-variadic-macros}.
6711 @opindex Wno-varargs
6712 Warn upon questionable usage of the macros used to handle variable
6713 arguments like @code{va_start}. This is default. To inhibit the
6714 warning messages, use @option{-Wno-varargs}.
6716 @item -Wvector-operation-performance
6717 @opindex Wvector-operation-performance
6718 @opindex Wno-vector-operation-performance
6719 Warn if vector operation is not implemented via SIMD capabilities of the
6720 architecture. Mainly useful for the performance tuning.
6721 Vector operation can be implemented @code{piecewise}, which means that the
6722 scalar operation is performed on every vector element;
6723 @code{in parallel}, which means that the vector operation is implemented
6724 using scalars of wider type, which normally is more performance efficient;
6725 and @code{as a single scalar}, which means that vector fits into a
6728 @item -Wno-virtual-move-assign
6729 @opindex Wvirtual-move-assign
6730 @opindex Wno-virtual-move-assign
6731 Suppress warnings about inheriting from a virtual base with a
6732 non-trivial C++11 move assignment operator. This is dangerous because
6733 if the virtual base is reachable along more than one path, it is
6734 moved multiple times, which can mean both objects end up in the
6735 moved-from state. If the move assignment operator is written to avoid
6736 moving from a moved-from object, this warning can be disabled.
6741 Warn if a variable-length array is used in the code.
6742 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6743 the variable-length array.
6745 @item -Wvla-larger-than=@var{n}
6746 If this option is used, the compiler will warn on uses of
6747 variable-length arrays where the size is either unbounded, or bounded
6748 by an argument that can be larger than @var{n} bytes. This is similar
6749 to how @option{-Walloca-larger-than=@var{n}} works, but with
6750 variable-length arrays.
6752 Note that GCC may optimize small variable-length arrays of a known
6753 value into plain arrays, so this warning may not get triggered for
6756 This warning is not enabled by @option{-Wall}, and is only active when
6757 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6759 See also @option{-Walloca-larger-than=@var{n}}.
6761 @item -Wvolatile-register-var
6762 @opindex Wvolatile-register-var
6763 @opindex Wno-volatile-register-var
6764 Warn if a register variable is declared volatile. The volatile
6765 modifier does not inhibit all optimizations that may eliminate reads
6766 and/or writes to register variables. This warning is enabled by
6769 @item -Wdisabled-optimization
6770 @opindex Wdisabled-optimization
6771 @opindex Wno-disabled-optimization
6772 Warn if a requested optimization pass is disabled. This warning does
6773 not generally indicate that there is anything wrong with your code; it
6774 merely indicates that GCC's optimizers are unable to handle the code
6775 effectively. Often, the problem is that your code is too big or too
6776 complex; GCC refuses to optimize programs when the optimization
6777 itself is likely to take inordinate amounts of time.
6779 @item -Wpointer-sign @r{(C and Objective-C only)}
6780 @opindex Wpointer-sign
6781 @opindex Wno-pointer-sign
6782 Warn for pointer argument passing or assignment with different signedness.
6783 This option is only supported for C and Objective-C@. It is implied by
6784 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6785 @option{-Wno-pointer-sign}.
6787 @item -Wstack-protector
6788 @opindex Wstack-protector
6789 @opindex Wno-stack-protector
6790 This option is only active when @option{-fstack-protector} is active. It
6791 warns about functions that are not protected against stack smashing.
6793 @item -Woverlength-strings
6794 @opindex Woverlength-strings
6795 @opindex Wno-overlength-strings
6796 Warn about string constants that are longer than the ``minimum
6797 maximum'' length specified in the C standard. Modern compilers
6798 generally allow string constants that are much longer than the
6799 standard's minimum limit, but very portable programs should avoid
6800 using longer strings.
6802 The limit applies @emph{after} string constant concatenation, and does
6803 not count the trailing NUL@. In C90, the limit was 509 characters; in
6804 C99, it was raised to 4095. C++98 does not specify a normative
6805 minimum maximum, so we do not diagnose overlength strings in C++@.
6807 This option is implied by @option{-Wpedantic}, and can be disabled with
6808 @option{-Wno-overlength-strings}.
6810 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6811 @opindex Wunsuffixed-float-constants
6813 Issue a warning for any floating constant that does not have
6814 a suffix. When used together with @option{-Wsystem-headers} it
6815 warns about such constants in system header files. This can be useful
6816 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6817 from the decimal floating-point extension to C99.
6819 @item -Wno-designated-init @r{(C and Objective-C only)}
6820 Suppress warnings when a positional initializer is used to initialize
6821 a structure that has been marked with the @code{designated_init}
6825 Issue a warning when HSAIL cannot be emitted for the compiled function or
6830 @node Debugging Options
6831 @section Options for Debugging Your Program
6832 @cindex options, debugging
6833 @cindex debugging information options
6835 To tell GCC to emit extra information for use by a debugger, in almost
6836 all cases you need only to add @option{-g} to your other options.
6838 GCC allows you to use @option{-g} with
6839 @option{-O}. The shortcuts taken by optimized code may occasionally
6840 be surprising: some variables you declared may not exist
6841 at all; flow of control may briefly move where you did not expect it;
6842 some statements may not be executed because they compute constant
6843 results or their values are already at hand; some statements may
6844 execute in different places because they have been moved out of loops.
6845 Nevertheless it is possible to debug optimized output. This makes
6846 it reasonable to use the optimizer for programs that might have bugs.
6848 If you are not using some other optimization option, consider
6849 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6850 With no @option{-O} option at all, some compiler passes that collect
6851 information useful for debugging do not run at all, so that
6852 @option{-Og} may result in a better debugging experience.
6857 Produce debugging information in the operating system's native format
6858 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6861 On most systems that use stabs format, @option{-g} enables use of extra
6862 debugging information that only GDB can use; this extra information
6863 makes debugging work better in GDB but probably makes other debuggers
6865 refuse to read the program. If you want to control for certain whether
6866 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6867 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6871 Produce debugging information for use by GDB@. This means to use the
6872 most expressive format available (DWARF, stabs, or the native format
6873 if neither of those are supported), including GDB extensions if at all
6877 @itemx -gdwarf-@var{version}
6879 Produce debugging information in DWARF format (if that is supported).
6880 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6881 for most targets is 4. DWARF Version 5 is only experimental.
6883 Note that with DWARF Version 2, some ports require and always
6884 use some non-conflicting DWARF 3 extensions in the unwind tables.
6886 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6887 for maximum benefit.
6889 GCC no longer supports DWARF Version 1, which is substantially
6890 different than Version 2 and later. For historical reasons, some
6891 other DWARF-related options such as
6892 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6893 in their names, but apply to all currently-supported versions of DWARF.
6897 Produce debugging information in stabs format (if that is supported),
6898 without GDB extensions. This is the format used by DBX on most BSD
6899 systems. On MIPS, Alpha and System V Release 4 systems this option
6900 produces stabs debugging output that is not understood by DBX or SDB@.
6901 On System V Release 4 systems this option requires the GNU assembler.
6905 Produce debugging information in stabs format (if that is supported),
6906 using GNU extensions understood only by the GNU debugger (GDB)@. The
6907 use of these extensions is likely to make other debuggers crash or
6908 refuse to read the program.
6912 Produce debugging information in COFF format (if that is supported).
6913 This is the format used by SDB on most System V systems prior to
6918 Produce debugging information in XCOFF format (if that is supported).
6919 This is the format used by the DBX debugger on IBM RS/6000 systems.
6923 Produce debugging information in XCOFF format (if that is supported),
6924 using GNU extensions understood only by the GNU debugger (GDB)@. The
6925 use of these extensions is likely to make other debuggers crash or
6926 refuse to read the program, and may cause assemblers other than the GNU
6927 assembler (GAS) to fail with an error.
6931 Produce debugging information in Alpha/VMS debug format (if that is
6932 supported). This is the format used by DEBUG on Alpha/VMS systems.
6935 @itemx -ggdb@var{level}
6936 @itemx -gstabs@var{level}
6937 @itemx -gcoff@var{level}
6938 @itemx -gxcoff@var{level}
6939 @itemx -gvms@var{level}
6940 Request debugging information and also use @var{level} to specify how
6941 much information. The default level is 2.
6943 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6946 Level 1 produces minimal information, enough for making backtraces in
6947 parts of the program that you don't plan to debug. This includes
6948 descriptions of functions and external variables, and line number
6949 tables, but no information about local variables.
6951 Level 3 includes extra information, such as all the macro definitions
6952 present in the program. Some debuggers support macro expansion when
6953 you use @option{-g3}.
6955 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6956 confusion with @option{-gdwarf-@var{level}}.
6957 Instead use an additional @option{-g@var{level}} option to change the
6958 debug level for DWARF.
6960 @item -feliminate-unused-debug-symbols
6961 @opindex feliminate-unused-debug-symbols
6962 Produce debugging information in stabs format (if that is supported),
6963 for only symbols that are actually used.
6965 @item -femit-class-debug-always
6966 @opindex femit-class-debug-always
6967 Instead of emitting debugging information for a C++ class in only one
6968 object file, emit it in all object files using the class. This option
6969 should be used only with debuggers that are unable to handle the way GCC
6970 normally emits debugging information for classes because using this
6971 option increases the size of debugging information by as much as a
6974 @item -fno-merge-debug-strings
6975 @opindex fmerge-debug-strings
6976 @opindex fno-merge-debug-strings
6977 Direct the linker to not merge together strings in the debugging
6978 information that are identical in different object files. Merging is
6979 not supported by all assemblers or linkers. Merging decreases the size
6980 of the debug information in the output file at the cost of increasing
6981 link processing time. Merging is enabled by default.
6983 @item -fdebug-prefix-map=@var{old}=@var{new}
6984 @opindex fdebug-prefix-map
6985 When compiling files in directory @file{@var{old}}, record debugging
6986 information describing them as in @file{@var{new}} instead. This can be
6987 used to replace a build-time path with an install-time path in the debug info.
6988 It can also be used to change an absolute path to a relative path by using
6989 @file{.} for @var{new}. This can give more reproducible builds, which are
6990 location independent, but may require an extra command to tell GDB where to
6991 find the source files.
6993 @item -fvar-tracking
6994 @opindex fvar-tracking
6995 Run variable tracking pass. It computes where variables are stored at each
6996 position in code. Better debugging information is then generated
6997 (if the debugging information format supports this information).
6999 It is enabled by default when compiling with optimization (@option{-Os},
7000 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7001 the debug info format supports it.
7003 @item -fvar-tracking-assignments
7004 @opindex fvar-tracking-assignments
7005 @opindex fno-var-tracking-assignments
7006 Annotate assignments to user variables early in the compilation and
7007 attempt to carry the annotations over throughout the compilation all the
7008 way to the end, in an attempt to improve debug information while
7009 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7011 It can be enabled even if var-tracking is disabled, in which case
7012 annotations are created and maintained, but discarded at the end.
7013 By default, this flag is enabled together with @option{-fvar-tracking},
7014 except when selective scheduling is enabled.
7017 @opindex gsplit-dwarf
7018 Separate as much DWARF debugging information as possible into a
7019 separate output file with the extension @file{.dwo}. This option allows
7020 the build system to avoid linking files with debug information. To
7021 be useful, this option requires a debugger capable of reading @file{.dwo}
7026 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7028 @item -ggnu-pubnames
7029 @opindex ggnu-pubnames
7030 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7031 suitable for conversion into a GDB@ index. This option is only useful
7032 with a linker that can produce GDB@ index version 7.
7034 @item -fdebug-types-section
7035 @opindex fdebug-types-section
7036 @opindex fno-debug-types-section
7037 When using DWARF Version 4 or higher, type DIEs can be put into
7038 their own @code{.debug_types} section instead of making them part of the
7039 @code{.debug_info} section. It is more efficient to put them in a separate
7040 comdat sections since the linker can then remove duplicates.
7041 But not all DWARF consumers support @code{.debug_types} sections yet
7042 and on some objects @code{.debug_types} produces larger instead of smaller
7043 debugging information.
7045 @item -grecord-gcc-switches
7046 @item -gno-record-gcc-switches
7047 @opindex grecord-gcc-switches
7048 @opindex gno-record-gcc-switches
7049 This switch causes the command-line options used to invoke the
7050 compiler that may affect code generation to be appended to the
7051 DW_AT_producer attribute in DWARF debugging information. The options
7052 are concatenated with spaces separating them from each other and from
7053 the compiler version.
7054 It is enabled by default.
7055 See also @option{-frecord-gcc-switches} for another
7056 way of storing compiler options into the object file.
7058 @item -gstrict-dwarf
7059 @opindex gstrict-dwarf
7060 Disallow using extensions of later DWARF standard version than selected
7061 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7062 DWARF extensions from later standard versions is allowed.
7064 @item -gno-strict-dwarf
7065 @opindex gno-strict-dwarf
7066 Allow using extensions of later DWARF standard version than selected with
7067 @option{-gdwarf-@var{version}}.
7070 @item -gno-column-info
7071 @opindex gcolumn-info
7072 @opindex gno-column-info
7073 Emit location column information into DWARF debugging information, rather
7074 than just file and line.
7075 This option is enabled by default.
7077 @item -gz@r{[}=@var{type}@r{]}
7079 Produce compressed debug sections in DWARF format, if that is supported.
7080 If @var{type} is not given, the default type depends on the capabilities
7081 of the assembler and linker used. @var{type} may be one of
7082 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
7083 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
7084 compression in traditional GNU format). If the linker doesn't support
7085 writing compressed debug sections, the option is rejected. Otherwise,
7086 if the assembler does not support them, @option{-gz} is silently ignored
7087 when producing object files.
7089 @item -femit-struct-debug-baseonly
7090 @opindex femit-struct-debug-baseonly
7091 Emit debug information for struct-like types
7092 only when the base name of the compilation source file
7093 matches the base name of file in which the struct is defined.
7095 This option substantially reduces the size of debugging information,
7096 but at significant potential loss in type information to the debugger.
7097 See @option{-femit-struct-debug-reduced} for a less aggressive option.
7098 See @option{-femit-struct-debug-detailed} for more detailed control.
7100 This option works only with DWARF debug output.
7102 @item -femit-struct-debug-reduced
7103 @opindex femit-struct-debug-reduced
7104 Emit debug information for struct-like types
7105 only when the base name of the compilation source file
7106 matches the base name of file in which the type is defined,
7107 unless the struct is a template or defined in a system header.
7109 This option significantly reduces the size of debugging information,
7110 with some potential loss in type information to the debugger.
7111 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
7112 See @option{-femit-struct-debug-detailed} for more detailed control.
7114 This option works only with DWARF debug output.
7116 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
7117 @opindex femit-struct-debug-detailed
7118 Specify the struct-like types
7119 for which the compiler generates debug information.
7120 The intent is to reduce duplicate struct debug information
7121 between different object files within the same program.
7123 This option is a detailed version of
7124 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
7125 which serves for most needs.
7127 A specification has the syntax@*
7128 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
7130 The optional first word limits the specification to
7131 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
7132 A struct type is used directly when it is the type of a variable, member.
7133 Indirect uses arise through pointers to structs.
7134 That is, when use of an incomplete struct is valid, the use is indirect.
7136 @samp{struct one direct; struct two * indirect;}.
7138 The optional second word limits the specification to
7139 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
7140 Generic structs are a bit complicated to explain.
7141 For C++, these are non-explicit specializations of template classes,
7142 or non-template classes within the above.
7143 Other programming languages have generics,
7144 but @option{-femit-struct-debug-detailed} does not yet implement them.
7146 The third word specifies the source files for those
7147 structs for which the compiler should emit debug information.
7148 The values @samp{none} and @samp{any} have the normal meaning.
7149 The value @samp{base} means that
7150 the base of name of the file in which the type declaration appears
7151 must match the base of the name of the main compilation file.
7152 In practice, this means that when compiling @file{foo.c}, debug information
7153 is generated for types declared in that file and @file{foo.h},
7154 but not other header files.
7155 The value @samp{sys} means those types satisfying @samp{base}
7156 or declared in system or compiler headers.
7158 You may need to experiment to determine the best settings for your application.
7160 The default is @option{-femit-struct-debug-detailed=all}.
7162 This option works only with DWARF debug output.
7164 @item -fno-dwarf2-cfi-asm
7165 @opindex fdwarf2-cfi-asm
7166 @opindex fno-dwarf2-cfi-asm
7167 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
7168 instead of using GAS @code{.cfi_*} directives.
7170 @item -fno-eliminate-unused-debug-types
7171 @opindex feliminate-unused-debug-types
7172 @opindex fno-eliminate-unused-debug-types
7173 Normally, when producing DWARF output, GCC avoids producing debug symbol
7174 output for types that are nowhere used in the source file being compiled.
7175 Sometimes it is useful to have GCC emit debugging
7176 information for all types declared in a compilation
7177 unit, regardless of whether or not they are actually used
7178 in that compilation unit, for example
7179 if, in the debugger, you want to cast a value to a type that is
7180 not actually used in your program (but is declared). More often,
7181 however, this results in a significant amount of wasted space.
7184 @node Optimize Options
7185 @section Options That Control Optimization
7186 @cindex optimize options
7187 @cindex options, optimization
7189 These options control various sorts of optimizations.
7191 Without any optimization option, the compiler's goal is to reduce the
7192 cost of compilation and to make debugging produce the expected
7193 results. Statements are independent: if you stop the program with a
7194 breakpoint between statements, you can then assign a new value to any
7195 variable or change the program counter to any other statement in the
7196 function and get exactly the results you expect from the source
7199 Turning on optimization flags makes the compiler attempt to improve
7200 the performance and/or code size at the expense of compilation time
7201 and possibly the ability to debug the program.
7203 The compiler performs optimization based on the knowledge it has of the
7204 program. Compiling multiple files at once to a single output file mode allows
7205 the compiler to use information gained from all of the files when compiling
7208 Not all optimizations are controlled directly by a flag. Only
7209 optimizations that have a flag are listed in this section.
7211 Most optimizations are only enabled if an @option{-O} level is set on
7212 the command line. Otherwise they are disabled, even if individual
7213 optimization flags are specified.
7215 Depending on the target and how GCC was configured, a slightly different
7216 set of optimizations may be enabled at each @option{-O} level than
7217 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7218 to find out the exact set of optimizations that are enabled at each level.
7219 @xref{Overall Options}, for examples.
7226 Optimize. Optimizing compilation takes somewhat more time, and a lot
7227 more memory for a large function.
7229 With @option{-O}, the compiler tries to reduce code size and execution
7230 time, without performing any optimizations that take a great deal of
7233 @option{-O} turns on the following optimization flags:
7236 -fbranch-count-reg @gol
7237 -fcombine-stack-adjustments @gol
7239 -fcprop-registers @gol
7242 -fdelayed-branch @gol
7244 -fforward-propagate @gol
7245 -fguess-branch-probability @gol
7246 -fif-conversion2 @gol
7247 -fif-conversion @gol
7248 -finline-functions-called-once @gol
7249 -fipa-pure-const @gol
7251 -fipa-reference @gol
7252 -fmerge-constants @gol
7253 -fmove-loop-invariants @gol
7254 -freorder-blocks @gol
7256 -fshrink-wrap-separate @gol
7257 -fsplit-wide-types @gol
7263 -ftree-coalesce-vars @gol
7264 -ftree-copy-prop @gol
7266 -ftree-dominator-opts @gol
7268 -ftree-forwprop @gol
7278 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7279 where doing so does not interfere with debugging.
7283 Optimize even more. GCC performs nearly all supported optimizations
7284 that do not involve a space-speed tradeoff.
7285 As compared to @option{-O}, this option increases both compilation time
7286 and the performance of the generated code.
7288 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7289 also turns on the following optimization flags:
7290 @gccoptlist{-fthread-jumps @gol
7291 -falign-functions -falign-jumps @gol
7292 -falign-loops -falign-labels @gol
7295 -fcse-follow-jumps -fcse-skip-blocks @gol
7296 -fdelete-null-pointer-checks @gol
7297 -fdevirtualize -fdevirtualize-speculatively @gol
7298 -fexpensive-optimizations @gol
7299 -fgcse -fgcse-lm @gol
7300 -fhoist-adjacent-loads @gol
7301 -finline-small-functions @gol
7302 -findirect-inlining @gol
7308 -fisolate-erroneous-paths-dereference @gol
7310 -foptimize-sibling-calls @gol
7311 -foptimize-strlen @gol
7312 -fpartial-inlining @gol
7314 -freorder-blocks-algorithm=stc @gol
7315 -freorder-blocks-and-partition -freorder-functions @gol
7316 -frerun-cse-after-loop @gol
7317 -fsched-interblock -fsched-spec @gol
7318 -fschedule-insns -fschedule-insns2 @gol
7319 -fstore-merging @gol
7320 -fstrict-aliasing @gol
7321 -ftree-builtin-call-dce @gol
7322 -ftree-switch-conversion -ftree-tail-merge @gol
7323 -fcode-hoisting @gol
7328 Please note the warning under @option{-fgcse} about
7329 invoking @option{-O2} on programs that use computed gotos.
7333 Optimize yet more. @option{-O3} turns on all optimizations specified
7334 by @option{-O2} and also turns on the following optimization flags:
7335 @gccoptlist{-finline-functions @gol
7336 -funswitch-loops @gol
7337 -fpredictive-commoning @gol
7338 -fgcse-after-reload @gol
7339 -ftree-loop-vectorize @gol
7340 -ftree-loop-distribution @gol
7341 -ftree-loop-distribute-patterns @gol
7343 -ftree-slp-vectorize @gol
7344 -fvect-cost-model @gol
7345 -ftree-partial-pre @gol
7351 Reduce compilation time and make debugging produce the expected
7352 results. This is the default.
7356 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7357 do not typically increase code size. It also performs further
7358 optimizations designed to reduce code size.
7360 @option{-Os} disables the following optimization flags:
7361 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7362 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7363 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7367 Disregard strict standards compliance. @option{-Ofast} enables all
7368 @option{-O3} optimizations. It also enables optimizations that are not
7369 valid for all standard-compliant programs.
7370 It turns on @option{-ffast-math} and the Fortran-specific
7371 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
7372 specified, and @option{-fno-protect-parens}.
7376 Optimize debugging experience. @option{-Og} enables optimizations
7377 that do not interfere with debugging. It should be the optimization
7378 level of choice for the standard edit-compile-debug cycle, offering
7379 a reasonable level of optimization while maintaining fast compilation
7380 and a good debugging experience.
7383 If you use multiple @option{-O} options, with or without level numbers,
7384 the last such option is the one that is effective.
7386 Options of the form @option{-f@var{flag}} specify machine-independent
7387 flags. Most flags have both positive and negative forms; the negative
7388 form of @option{-ffoo} is @option{-fno-foo}. In the table
7389 below, only one of the forms is listed---the one you typically
7390 use. You can figure out the other form by either removing @samp{no-}
7393 The following options control specific optimizations. They are either
7394 activated by @option{-O} options or are related to ones that are. You
7395 can use the following flags in the rare cases when ``fine-tuning'' of
7396 optimizations to be performed is desired.
7399 @item -fno-defer-pop
7400 @opindex fno-defer-pop
7401 Always pop the arguments to each function call as soon as that function
7402 returns. For machines that must pop arguments after a function call,
7403 the compiler normally lets arguments accumulate on the stack for several
7404 function calls and pops them all at once.
7406 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7408 @item -fforward-propagate
7409 @opindex fforward-propagate
7410 Perform a forward propagation pass on RTL@. The pass tries to combine two
7411 instructions and checks if the result can be simplified. If loop unrolling
7412 is active, two passes are performed and the second is scheduled after
7415 This option is enabled by default at optimization levels @option{-O},
7416 @option{-O2}, @option{-O3}, @option{-Os}.
7418 @item -ffp-contract=@var{style}
7419 @opindex ffp-contract
7420 @option{-ffp-contract=off} disables floating-point expression contraction.
7421 @option{-ffp-contract=fast} enables floating-point expression contraction
7422 such as forming of fused multiply-add operations if the target has
7423 native support for them.
7424 @option{-ffp-contract=on} enables floating-point expression contraction
7425 if allowed by the language standard. This is currently not implemented
7426 and treated equal to @option{-ffp-contract=off}.
7428 The default is @option{-ffp-contract=fast}.
7430 @item -fomit-frame-pointer
7431 @opindex fomit-frame-pointer
7432 Don't keep the frame pointer in a register for functions that
7433 don't need one. This avoids the instructions to save, set up and
7434 restore frame pointers; it also makes an extra register available
7435 in many functions. @strong{It also makes debugging impossible on
7438 On some machines, such as the VAX, this flag has no effect, because
7439 the standard calling sequence automatically handles the frame pointer
7440 and nothing is saved by pretending it doesn't exist. The
7441 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7442 whether a target machine supports this flag. @xref{Registers,,Register
7443 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7445 The default setting (when not optimizing for
7446 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7447 @option{-fomit-frame-pointer}. You can configure GCC with the
7448 @option{--enable-frame-pointer} configure option to change the default.
7450 Note that @option{-fno-omit-frame-pointer} doesn't force a new stack
7451 frame for all functions if it isn't otherwise needed, and hence doesn't
7452 guarantee a new frame pointer for all functions.
7454 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7456 @item -foptimize-sibling-calls
7457 @opindex foptimize-sibling-calls
7458 Optimize sibling and tail recursive calls.
7460 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7462 @item -foptimize-strlen
7463 @opindex foptimize-strlen
7464 Optimize various standard C string functions (e.g. @code{strlen},
7465 @code{strchr} or @code{strcpy}) and
7466 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7468 Enabled at levels @option{-O2}, @option{-O3}.
7472 Do not expand any functions inline apart from those marked with
7473 the @code{always_inline} attribute. This is the default when not
7476 Single functions can be exempted from inlining by marking them
7477 with the @code{noinline} attribute.
7479 @item -finline-small-functions
7480 @opindex finline-small-functions
7481 Integrate functions into their callers when their body is smaller than expected
7482 function call code (so overall size of program gets smaller). The compiler
7483 heuristically decides which functions are simple enough to be worth integrating
7484 in this way. This inlining applies to all functions, even those not declared
7487 Enabled at level @option{-O2}.
7489 @item -findirect-inlining
7490 @opindex findirect-inlining
7491 Inline also indirect calls that are discovered to be known at compile
7492 time thanks to previous inlining. This option has any effect only
7493 when inlining itself is turned on by the @option{-finline-functions}
7494 or @option{-finline-small-functions} options.
7496 Enabled at level @option{-O2}.
7498 @item -finline-functions
7499 @opindex finline-functions
7500 Consider all functions for inlining, even if they are not declared inline.
7501 The compiler heuristically decides which functions are worth integrating
7504 If all calls to a given function are integrated, and the function is
7505 declared @code{static}, then the function is normally not output as
7506 assembler code in its own right.
7508 Enabled at level @option{-O3}.
7510 @item -finline-functions-called-once
7511 @opindex finline-functions-called-once
7512 Consider all @code{static} functions called once for inlining into their
7513 caller even if they are not marked @code{inline}. If a call to a given
7514 function is integrated, then the function is not output as assembler code
7517 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7519 @item -fearly-inlining
7520 @opindex fearly-inlining
7521 Inline functions marked by @code{always_inline} and functions whose body seems
7522 smaller than the function call overhead early before doing
7523 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7524 makes profiling significantly cheaper and usually inlining faster on programs
7525 having large chains of nested wrapper functions.
7531 Perform interprocedural scalar replacement of aggregates, removal of
7532 unused parameters and replacement of parameters passed by reference
7533 by parameters passed by value.
7535 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7537 @item -finline-limit=@var{n}
7538 @opindex finline-limit
7539 By default, GCC limits the size of functions that can be inlined. This flag
7540 allows coarse control of this limit. @var{n} is the size of functions that
7541 can be inlined in number of pseudo instructions.
7543 Inlining is actually controlled by a number of parameters, which may be
7544 specified individually by using @option{--param @var{name}=@var{value}}.
7545 The @option{-finline-limit=@var{n}} option sets some of these parameters
7549 @item max-inline-insns-single
7550 is set to @var{n}/2.
7551 @item max-inline-insns-auto
7552 is set to @var{n}/2.
7555 See below for a documentation of the individual
7556 parameters controlling inlining and for the defaults of these parameters.
7558 @emph{Note:} there may be no value to @option{-finline-limit} that results
7559 in default behavior.
7561 @emph{Note:} pseudo instruction represents, in this particular context, an
7562 abstract measurement of function's size. In no way does it represent a count
7563 of assembly instructions and as such its exact meaning might change from one
7564 release to an another.
7566 @item -fno-keep-inline-dllexport
7567 @opindex fno-keep-inline-dllexport
7568 This is a more fine-grained version of @option{-fkeep-inline-functions},
7569 which applies only to functions that are declared using the @code{dllexport}
7570 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
7573 @item -fkeep-inline-functions
7574 @opindex fkeep-inline-functions
7575 In C, emit @code{static} functions that are declared @code{inline}
7576 into the object file, even if the function has been inlined into all
7577 of its callers. This switch does not affect functions using the
7578 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7579 inline functions into the object file.
7581 @item -fkeep-static-functions
7582 @opindex fkeep-static-functions
7583 Emit @code{static} functions into the object file, even if the function
7586 @item -fkeep-static-consts
7587 @opindex fkeep-static-consts
7588 Emit variables declared @code{static const} when optimization isn't turned
7589 on, even if the variables aren't referenced.
7591 GCC enables this option by default. If you want to force the compiler to
7592 check if a variable is referenced, regardless of whether or not
7593 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7595 @item -fmerge-constants
7596 @opindex fmerge-constants
7597 Attempt to merge identical constants (string constants and floating-point
7598 constants) across compilation units.
7600 This option is the default for optimized compilation if the assembler and
7601 linker support it. Use @option{-fno-merge-constants} to inhibit this
7604 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7606 @item -fmerge-all-constants
7607 @opindex fmerge-all-constants
7608 Attempt to merge identical constants and identical variables.
7610 This option implies @option{-fmerge-constants}. In addition to
7611 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7612 arrays or initialized constant variables with integral or floating-point
7613 types. Languages like C or C++ require each variable, including multiple
7614 instances of the same variable in recursive calls, to have distinct locations,
7615 so using this option results in non-conforming
7618 @item -fmodulo-sched
7619 @opindex fmodulo-sched
7620 Perform swing modulo scheduling immediately before the first scheduling
7621 pass. This pass looks at innermost loops and reorders their
7622 instructions by overlapping different iterations.
7624 @item -fmodulo-sched-allow-regmoves
7625 @opindex fmodulo-sched-allow-regmoves
7626 Perform more aggressive SMS-based modulo scheduling with register moves
7627 allowed. By setting this flag certain anti-dependences edges are
7628 deleted, which triggers the generation of reg-moves based on the
7629 life-range analysis. This option is effective only with
7630 @option{-fmodulo-sched} enabled.
7632 @item -fno-branch-count-reg
7633 @opindex fno-branch-count-reg
7634 Avoid running a pass scanning for opportunities to use ``decrement and
7635 branch'' instructions on a count register instead of generating sequences
7636 of instructions that decrement a register, compare it against zero, and
7637 then branch based upon the result. This option is only meaningful on
7638 architectures that support such instructions, which include x86, PowerPC,
7639 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7640 doesn't remove the decrement and branch instructions from the generated
7641 instruction stream introduced by other optimization passes.
7643 Enabled by default at @option{-O1} and higher.
7645 The default is @option{-fbranch-count-reg}.
7647 @item -fno-function-cse
7648 @opindex fno-function-cse
7649 Do not put function addresses in registers; make each instruction that
7650 calls a constant function contain the function's address explicitly.
7652 This option results in less efficient code, but some strange hacks
7653 that alter the assembler output may be confused by the optimizations
7654 performed when this option is not used.
7656 The default is @option{-ffunction-cse}
7658 @item -fno-zero-initialized-in-bss
7659 @opindex fno-zero-initialized-in-bss
7660 If the target supports a BSS section, GCC by default puts variables that
7661 are initialized to zero into BSS@. This can save space in the resulting
7664 This option turns off this behavior because some programs explicitly
7665 rely on variables going to the data section---e.g., so that the
7666 resulting executable can find the beginning of that section and/or make
7667 assumptions based on that.
7669 The default is @option{-fzero-initialized-in-bss}.
7671 @item -fthread-jumps
7672 @opindex fthread-jumps
7673 Perform optimizations that check to see if a jump branches to a
7674 location where another comparison subsumed by the first is found. If
7675 so, the first branch is redirected to either the destination of the
7676 second branch or a point immediately following it, depending on whether
7677 the condition is known to be true or false.
7679 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7681 @item -fsplit-wide-types
7682 @opindex fsplit-wide-types
7683 When using a type that occupies multiple registers, such as @code{long
7684 long} on a 32-bit system, split the registers apart and allocate them
7685 independently. This normally generates better code for those types,
7686 but may make debugging more difficult.
7688 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7691 @item -fcse-follow-jumps
7692 @opindex fcse-follow-jumps
7693 In common subexpression elimination (CSE), scan through jump instructions
7694 when the target of the jump is not reached by any other path. For
7695 example, when CSE encounters an @code{if} statement with an
7696 @code{else} clause, CSE follows the jump when the condition
7699 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7701 @item -fcse-skip-blocks
7702 @opindex fcse-skip-blocks
7703 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7704 follow jumps that conditionally skip over blocks. When CSE
7705 encounters a simple @code{if} statement with no else clause,
7706 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7707 body of the @code{if}.
7709 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7711 @item -frerun-cse-after-loop
7712 @opindex frerun-cse-after-loop
7713 Re-run common subexpression elimination after loop optimizations are
7716 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7720 Perform a global common subexpression elimination pass.
7721 This pass also performs global constant and copy propagation.
7723 @emph{Note:} When compiling a program using computed gotos, a GCC
7724 extension, you may get better run-time performance if you disable
7725 the global common subexpression elimination pass by adding
7726 @option{-fno-gcse} to the command line.
7728 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7732 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7733 attempts to move loads that are only killed by stores into themselves. This
7734 allows a loop containing a load/store sequence to be changed to a load outside
7735 the loop, and a copy/store within the loop.
7737 Enabled by default when @option{-fgcse} is enabled.
7741 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7742 global common subexpression elimination. This pass attempts to move
7743 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7744 loops containing a load/store sequence can be changed to a load before
7745 the loop and a store after the loop.
7747 Not enabled at any optimization level.
7751 When @option{-fgcse-las} is enabled, the global common subexpression
7752 elimination pass eliminates redundant loads that come after stores to the
7753 same memory location (both partial and full redundancies).
7755 Not enabled at any optimization level.
7757 @item -fgcse-after-reload
7758 @opindex fgcse-after-reload
7759 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7760 pass is performed after reload. The purpose of this pass is to clean up
7763 @item -faggressive-loop-optimizations
7764 @opindex faggressive-loop-optimizations
7765 This option tells the loop optimizer to use language constraints to
7766 derive bounds for the number of iterations of a loop. This assumes that
7767 loop code does not invoke undefined behavior by for example causing signed
7768 integer overflows or out-of-bound array accesses. The bounds for the
7769 number of iterations of a loop are used to guide loop unrolling and peeling
7770 and loop exit test optimizations.
7771 This option is enabled by default.
7773 @item -funconstrained-commons
7774 @opindex funconstrained-commons
7775 This option tells the compiler that variables declared in common blocks
7776 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7777 prevents certain optimizations that depend on knowing the array bounds.
7779 @item -fcrossjumping
7780 @opindex fcrossjumping
7781 Perform cross-jumping transformation.
7782 This transformation unifies equivalent code and saves code size. The
7783 resulting code may or may not perform better than without cross-jumping.
7785 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7787 @item -fauto-inc-dec
7788 @opindex fauto-inc-dec
7789 Combine increments or decrements of addresses with memory accesses.
7790 This pass is always skipped on architectures that do not have
7791 instructions to support this. Enabled by default at @option{-O} and
7792 higher on architectures that support this.
7796 Perform dead code elimination (DCE) on RTL@.
7797 Enabled by default at @option{-O} and higher.
7801 Perform dead store elimination (DSE) on RTL@.
7802 Enabled by default at @option{-O} and higher.
7804 @item -fif-conversion
7805 @opindex fif-conversion
7806 Attempt to transform conditional jumps into branch-less equivalents. This
7807 includes use of conditional moves, min, max, set flags and abs instructions, and
7808 some tricks doable by standard arithmetics. The use of conditional execution
7809 on chips where it is available is controlled by @option{-fif-conversion2}.
7811 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7813 @item -fif-conversion2
7814 @opindex fif-conversion2
7815 Use conditional execution (where available) to transform conditional jumps into
7816 branch-less equivalents.
7818 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7820 @item -fdeclone-ctor-dtor
7821 @opindex fdeclone-ctor-dtor
7822 The C++ ABI requires multiple entry points for constructors and
7823 destructors: one for a base subobject, one for a complete object, and
7824 one for a virtual destructor that calls operator delete afterwards.
7825 For a hierarchy with virtual bases, the base and complete variants are
7826 clones, which means two copies of the function. With this option, the
7827 base and complete variants are changed to be thunks that call a common
7830 Enabled by @option{-Os}.
7832 @item -fdelete-null-pointer-checks
7833 @opindex fdelete-null-pointer-checks
7834 Assume that programs cannot safely dereference null pointers, and that
7835 no code or data element resides at address zero.
7836 This option enables simple constant
7837 folding optimizations at all optimization levels. In addition, other
7838 optimization passes in GCC use this flag to control global dataflow
7839 analyses that eliminate useless checks for null pointers; these assume
7840 that a memory access to address zero always results in a trap, so
7841 that if a pointer is checked after it has already been dereferenced,
7844 Note however that in some environments this assumption is not true.
7845 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7846 for programs that depend on that behavior.
7848 This option is enabled by default on most targets. On Nios II ELF, it
7849 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
7851 Passes that use the dataflow information
7852 are enabled independently at different optimization levels.
7854 @item -fdevirtualize
7855 @opindex fdevirtualize
7856 Attempt to convert calls to virtual functions to direct calls. This
7857 is done both within a procedure and interprocedurally as part of
7858 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7859 propagation (@option{-fipa-cp}).
7860 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7862 @item -fdevirtualize-speculatively
7863 @opindex fdevirtualize-speculatively
7864 Attempt to convert calls to virtual functions to speculative direct calls.
7865 Based on the analysis of the type inheritance graph, determine for a given call
7866 the set of likely targets. If the set is small, preferably of size 1, change
7867 the call into a conditional deciding between direct and indirect calls. The
7868 speculative calls enable more optimizations, such as inlining. When they seem
7869 useless after further optimization, they are converted back into original form.
7871 @item -fdevirtualize-at-ltrans
7872 @opindex fdevirtualize-at-ltrans
7873 Stream extra information needed for aggressive devirtualization when running
7874 the link-time optimizer in local transformation mode.
7875 This option enables more devirtualization but
7876 significantly increases the size of streamed data. For this reason it is
7877 disabled by default.
7879 @item -fexpensive-optimizations
7880 @opindex fexpensive-optimizations
7881 Perform a number of minor optimizations that are relatively expensive.
7883 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7887 Attempt to remove redundant extension instructions. This is especially
7888 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7889 registers after writing to their lower 32-bit half.
7891 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7892 @option{-O3}, @option{-Os}.
7894 @item -fno-lifetime-dse
7895 @opindex fno-lifetime-dse
7896 In C++ the value of an object is only affected by changes within its
7897 lifetime: when the constructor begins, the object has an indeterminate
7898 value, and any changes during the lifetime of the object are dead when
7899 the object is destroyed. Normally dead store elimination will take
7900 advantage of this; if your code relies on the value of the object
7901 storage persisting beyond the lifetime of the object, you can use this
7902 flag to disable this optimization. To preserve stores before the
7903 constructor starts (e.g. because your operator new clears the object
7904 storage) but still treat the object as dead after the destructor you,
7905 can use @option{-flifetime-dse=1}. The default behavior can be
7906 explicitly selected with @option{-flifetime-dse=2}.
7907 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7909 @item -flive-range-shrinkage
7910 @opindex flive-range-shrinkage
7911 Attempt to decrease register pressure through register live range
7912 shrinkage. This is helpful for fast processors with small or moderate
7915 @item -fira-algorithm=@var{algorithm}
7916 @opindex fira-algorithm
7917 Use the specified coloring algorithm for the integrated register
7918 allocator. The @var{algorithm} argument can be @samp{priority}, which
7919 specifies Chow's priority coloring, or @samp{CB}, which specifies
7920 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7921 for all architectures, but for those targets that do support it, it is
7922 the default because it generates better code.
7924 @item -fira-region=@var{region}
7925 @opindex fira-region
7926 Use specified regions for the integrated register allocator. The
7927 @var{region} argument should be one of the following:
7932 Use all loops as register allocation regions.
7933 This can give the best results for machines with a small and/or
7934 irregular register set.
7937 Use all loops except for loops with small register pressure
7938 as the regions. This value usually gives
7939 the best results in most cases and for most architectures,
7940 and is enabled by default when compiling with optimization for speed
7941 (@option{-O}, @option{-O2}, @dots{}).
7944 Use all functions as a single region.
7945 This typically results in the smallest code size, and is enabled by default for
7946 @option{-Os} or @option{-O0}.
7950 @item -fira-hoist-pressure
7951 @opindex fira-hoist-pressure
7952 Use IRA to evaluate register pressure in the code hoisting pass for
7953 decisions to hoist expressions. This option usually results in smaller
7954 code, but it can slow the compiler down.
7956 This option is enabled at level @option{-Os} for all targets.
7958 @item -fira-loop-pressure
7959 @opindex fira-loop-pressure
7960 Use IRA to evaluate register pressure in loops for decisions to move
7961 loop invariants. This option usually results in generation
7962 of faster and smaller code on machines with large register files (>= 32
7963 registers), but it can slow the compiler down.
7965 This option is enabled at level @option{-O3} for some targets.
7967 @item -fno-ira-share-save-slots
7968 @opindex fno-ira-share-save-slots
7969 Disable sharing of stack slots used for saving call-used hard
7970 registers living through a call. Each hard register gets a
7971 separate stack slot, and as a result function stack frames are
7974 @item -fno-ira-share-spill-slots
7975 @opindex fno-ira-share-spill-slots
7976 Disable sharing of stack slots allocated for pseudo-registers. Each
7977 pseudo-register that does not get a hard register gets a separate
7978 stack slot, and as a result function stack frames are larger.
7982 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7983 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7984 values if it is profitable.
7986 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7988 @item -fdelayed-branch
7989 @opindex fdelayed-branch
7990 If supported for the target machine, attempt to reorder instructions
7991 to exploit instruction slots available after delayed branch
7994 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7996 @item -fschedule-insns
7997 @opindex fschedule-insns
7998 If supported for the target machine, attempt to reorder instructions to
7999 eliminate execution stalls due to required data being unavailable. This
8000 helps machines that have slow floating point or memory load instructions
8001 by allowing other instructions to be issued until the result of the load
8002 or floating-point instruction is required.
8004 Enabled at levels @option{-O2}, @option{-O3}.
8006 @item -fschedule-insns2
8007 @opindex fschedule-insns2
8008 Similar to @option{-fschedule-insns}, but requests an additional pass of
8009 instruction scheduling after register allocation has been done. This is
8010 especially useful on machines with a relatively small number of
8011 registers and where memory load instructions take more than one cycle.
8013 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8015 @item -fno-sched-interblock
8016 @opindex fno-sched-interblock
8017 Don't schedule instructions across basic blocks. This is normally
8018 enabled by default when scheduling before register allocation, i.e.@:
8019 with @option{-fschedule-insns} or at @option{-O2} or higher.
8021 @item -fno-sched-spec
8022 @opindex fno-sched-spec
8023 Don't allow speculative motion of non-load instructions. This is normally
8024 enabled by default when scheduling before register allocation, i.e.@:
8025 with @option{-fschedule-insns} or at @option{-O2} or higher.
8027 @item -fsched-pressure
8028 @opindex fsched-pressure
8029 Enable register pressure sensitive insn scheduling before register
8030 allocation. This only makes sense when scheduling before register
8031 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8032 @option{-O2} or higher. Usage of this option can improve the
8033 generated code and decrease its size by preventing register pressure
8034 increase above the number of available hard registers and subsequent
8035 spills in register allocation.
8037 @item -fsched-spec-load
8038 @opindex fsched-spec-load
8039 Allow speculative motion of some load instructions. This only makes
8040 sense when scheduling before register allocation, i.e.@: with
8041 @option{-fschedule-insns} or at @option{-O2} or higher.
8043 @item -fsched-spec-load-dangerous
8044 @opindex fsched-spec-load-dangerous
8045 Allow speculative motion of more load instructions. This only makes
8046 sense when scheduling before register allocation, i.e.@: with
8047 @option{-fschedule-insns} or at @option{-O2} or higher.
8049 @item -fsched-stalled-insns
8050 @itemx -fsched-stalled-insns=@var{n}
8051 @opindex fsched-stalled-insns
8052 Define how many insns (if any) can be moved prematurely from the queue
8053 of stalled insns into the ready list during the second scheduling pass.
8054 @option{-fno-sched-stalled-insns} means that no insns are moved
8055 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8056 on how many queued insns can be moved prematurely.
8057 @option{-fsched-stalled-insns} without a value is equivalent to
8058 @option{-fsched-stalled-insns=1}.
8060 @item -fsched-stalled-insns-dep
8061 @itemx -fsched-stalled-insns-dep=@var{n}
8062 @opindex fsched-stalled-insns-dep
8063 Define how many insn groups (cycles) are examined for a dependency
8064 on a stalled insn that is a candidate for premature removal from the queue
8065 of stalled insns. This has an effect only during the second scheduling pass,
8066 and only if @option{-fsched-stalled-insns} is used.
8067 @option{-fno-sched-stalled-insns-dep} is equivalent to
8068 @option{-fsched-stalled-insns-dep=0}.
8069 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8070 @option{-fsched-stalled-insns-dep=1}.
8072 @item -fsched2-use-superblocks
8073 @opindex fsched2-use-superblocks
8074 When scheduling after register allocation, use superblock scheduling.
8075 This allows motion across basic block boundaries,
8076 resulting in faster schedules. This option is experimental, as not all machine
8077 descriptions used by GCC model the CPU closely enough to avoid unreliable
8078 results from the algorithm.
8080 This only makes sense when scheduling after register allocation, i.e.@: with
8081 @option{-fschedule-insns2} or at @option{-O2} or higher.
8083 @item -fsched-group-heuristic
8084 @opindex fsched-group-heuristic
8085 Enable the group heuristic in the scheduler. This heuristic favors
8086 the instruction that belongs to a schedule group. This is enabled
8087 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8088 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8090 @item -fsched-critical-path-heuristic
8091 @opindex fsched-critical-path-heuristic
8092 Enable the critical-path heuristic in the scheduler. This heuristic favors
8093 instructions on the critical path. This is enabled by default when
8094 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8095 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8097 @item -fsched-spec-insn-heuristic
8098 @opindex fsched-spec-insn-heuristic
8099 Enable the speculative instruction heuristic in the scheduler. This
8100 heuristic favors speculative instructions with greater dependency weakness.
8101 This is enabled by default when scheduling is enabled, i.e.@:
8102 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8103 or at @option{-O2} or higher.
8105 @item -fsched-rank-heuristic
8106 @opindex fsched-rank-heuristic
8107 Enable the rank heuristic in the scheduler. This heuristic favors
8108 the instruction belonging to a basic block with greater size or frequency.
8109 This is enabled by default when scheduling is enabled, i.e.@:
8110 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8111 at @option{-O2} or higher.
8113 @item -fsched-last-insn-heuristic
8114 @opindex fsched-last-insn-heuristic
8115 Enable the last-instruction heuristic in the scheduler. This heuristic
8116 favors the instruction that is less dependent on the last instruction
8117 scheduled. This is enabled by default when scheduling is enabled,
8118 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8119 at @option{-O2} or higher.
8121 @item -fsched-dep-count-heuristic
8122 @opindex fsched-dep-count-heuristic
8123 Enable the dependent-count heuristic in the scheduler. This heuristic
8124 favors the instruction that has more instructions depending on it.
8125 This is enabled by default when scheduling is enabled, i.e.@:
8126 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8127 at @option{-O2} or higher.
8129 @item -freschedule-modulo-scheduled-loops
8130 @opindex freschedule-modulo-scheduled-loops
8131 Modulo scheduling is performed before traditional scheduling. If a loop
8132 is modulo scheduled, later scheduling passes may change its schedule.
8133 Use this option to control that behavior.
8135 @item -fselective-scheduling
8136 @opindex fselective-scheduling
8137 Schedule instructions using selective scheduling algorithm. Selective
8138 scheduling runs instead of the first scheduler pass.
8140 @item -fselective-scheduling2
8141 @opindex fselective-scheduling2
8142 Schedule instructions using selective scheduling algorithm. Selective
8143 scheduling runs instead of the second scheduler pass.
8145 @item -fsel-sched-pipelining
8146 @opindex fsel-sched-pipelining
8147 Enable software pipelining of innermost loops during selective scheduling.
8148 This option has no effect unless one of @option{-fselective-scheduling} or
8149 @option{-fselective-scheduling2} is turned on.
8151 @item -fsel-sched-pipelining-outer-loops
8152 @opindex fsel-sched-pipelining-outer-loops
8153 When pipelining loops during selective scheduling, also pipeline outer loops.
8154 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8156 @item -fsemantic-interposition
8157 @opindex fsemantic-interposition
8158 Some object formats, like ELF, allow interposing of symbols by the
8160 This means that for symbols exported from the DSO, the compiler cannot perform
8161 interprocedural propagation, inlining and other optimizations in anticipation
8162 that the function or variable in question may change. While this feature is
8163 useful, for example, to rewrite memory allocation functions by a debugging
8164 implementation, it is expensive in the terms of code quality.
8165 With @option{-fno-semantic-interposition} the compiler assumes that
8166 if interposition happens for functions the overwriting function will have
8167 precisely the same semantics (and side effects).
8168 Similarly if interposition happens
8169 for variables, the constructor of the variable will be the same. The flag
8170 has no effect for functions explicitly declared inline
8171 (where it is never allowed for interposition to change semantics)
8172 and for symbols explicitly declared weak.
8175 @opindex fshrink-wrap
8176 Emit function prologues only before parts of the function that need it,
8177 rather than at the top of the function. This flag is enabled by default at
8178 @option{-O} and higher.
8180 @item -fshrink-wrap-separate
8181 @opindex fshrink-wrap-separate
8182 Shrink-wrap separate parts of the prologue and epilogue separately, so that
8183 those parts are only executed when needed.
8184 This option is on by default, but has no effect unless @option{-fshrink-wrap}
8185 is also turned on and the target supports this.
8187 @item -fcaller-saves
8188 @opindex fcaller-saves
8189 Enable allocation of values to registers that are clobbered by
8190 function calls, by emitting extra instructions to save and restore the
8191 registers around such calls. Such allocation is done only when it
8192 seems to result in better code.
8194 This option is always enabled by default on certain machines, usually
8195 those which have no call-preserved registers to use instead.
8197 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8199 @item -fcombine-stack-adjustments
8200 @opindex fcombine-stack-adjustments
8201 Tracks stack adjustments (pushes and pops) and stack memory references
8202 and then tries to find ways to combine them.
8204 Enabled by default at @option{-O1} and higher.
8208 Use caller save registers for allocation if those registers are not used by
8209 any called function. In that case it is not necessary to save and restore
8210 them around calls. This is only possible if called functions are part of
8211 same compilation unit as current function and they are compiled before it.
8213 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
8214 is disabled if generated code will be instrumented for profiling
8215 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
8216 exactly (this happens on targets that do not expose prologues
8217 and epilogues in RTL).
8219 @item -fconserve-stack
8220 @opindex fconserve-stack
8221 Attempt to minimize stack usage. The compiler attempts to use less
8222 stack space, even if that makes the program slower. This option
8223 implies setting the @option{large-stack-frame} parameter to 100
8224 and the @option{large-stack-frame-growth} parameter to 400.
8226 @item -ftree-reassoc
8227 @opindex ftree-reassoc
8228 Perform reassociation on trees. This flag is enabled by default
8229 at @option{-O} and higher.
8231 @item -fcode-hoisting
8232 @opindex fcode-hoisting
8233 Perform code hoisting. Code hoisting tries to move the
8234 evaluation of expressions executed on all paths to the function exit
8235 as early as possible. This is especially useful as a code size
8236 optimization, but it often helps for code speed as well.
8237 This flag is enabled by default at @option{-O2} and higher.
8241 Perform partial redundancy elimination (PRE) on trees. This flag is
8242 enabled by default at @option{-O2} and @option{-O3}.
8244 @item -ftree-partial-pre
8245 @opindex ftree-partial-pre
8246 Make partial redundancy elimination (PRE) more aggressive. This flag is
8247 enabled by default at @option{-O3}.
8249 @item -ftree-forwprop
8250 @opindex ftree-forwprop
8251 Perform forward propagation on trees. This flag is enabled by default
8252 at @option{-O} and higher.
8256 Perform full redundancy elimination (FRE) on trees. The difference
8257 between FRE and PRE is that FRE only considers expressions
8258 that are computed on all paths leading to the redundant computation.
8259 This analysis is faster than PRE, though it exposes fewer redundancies.
8260 This flag is enabled by default at @option{-O} and higher.
8262 @item -ftree-phiprop
8263 @opindex ftree-phiprop
8264 Perform hoisting of loads from conditional pointers on trees. This
8265 pass is enabled by default at @option{-O} and higher.
8267 @item -fhoist-adjacent-loads
8268 @opindex fhoist-adjacent-loads
8269 Speculatively hoist loads from both branches of an if-then-else if the
8270 loads are from adjacent locations in the same structure and the target
8271 architecture has a conditional move instruction. This flag is enabled
8272 by default at @option{-O2} and higher.
8274 @item -ftree-copy-prop
8275 @opindex ftree-copy-prop
8276 Perform copy propagation on trees. This pass eliminates unnecessary
8277 copy operations. This flag is enabled by default at @option{-O} and
8280 @item -fipa-pure-const
8281 @opindex fipa-pure-const
8282 Discover which functions are pure or constant.
8283 Enabled by default at @option{-O} and higher.
8285 @item -fipa-reference
8286 @opindex fipa-reference
8287 Discover which static variables do not escape the
8289 Enabled by default at @option{-O} and higher.
8293 Perform interprocedural pointer analysis and interprocedural modification
8294 and reference analysis. This option can cause excessive memory and
8295 compile-time usage on large compilation units. It is not enabled by
8296 default at any optimization level.
8299 @opindex fipa-profile
8300 Perform interprocedural profile propagation. The functions called only from
8301 cold functions are marked as cold. Also functions executed once (such as
8302 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8303 functions and loop less parts of functions executed once are then optimized for
8305 Enabled by default at @option{-O} and higher.
8309 Perform interprocedural constant propagation.
8310 This optimization analyzes the program to determine when values passed
8311 to functions are constants and then optimizes accordingly.
8312 This optimization can substantially increase performance
8313 if the application has constants passed to functions.
8314 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8316 @item -fipa-cp-clone
8317 @opindex fipa-cp-clone
8318 Perform function cloning to make interprocedural constant propagation stronger.
8319 When enabled, interprocedural constant propagation performs function cloning
8320 when externally visible function can be called with constant arguments.
8321 Because this optimization can create multiple copies of functions,
8322 it may significantly increase code size
8323 (see @option{--param ipcp-unit-growth=@var{value}}).
8324 This flag is enabled by default at @option{-O3}.
8327 @opindex -fipa-bit-cp
8328 When enabled, perform interprocedural bitwise constant
8329 propagation. This flag is enabled by default at @option{-O2}. It
8330 requires that @option{-fipa-cp} is enabled.
8334 When enabled, perform interprocedural propagation of value
8335 ranges. This flag is enabled by default at @option{-O2}. It requires
8336 that @option{-fipa-cp} is enabled.
8340 Perform Identical Code Folding for functions and read-only variables.
8341 The optimization reduces code size and may disturb unwind stacks by replacing
8342 a function by equivalent one with a different name. The optimization works
8343 more effectively with link-time optimization enabled.
8345 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8346 works on different levels and thus the optimizations are not same - there are
8347 equivalences that are found only by GCC and equivalences found only by Gold.
8349 This flag is enabled by default at @option{-O2} and @option{-Os}.
8351 @item -fisolate-erroneous-paths-dereference
8352 @opindex fisolate-erroneous-paths-dereference
8353 Detect paths that trigger erroneous or undefined behavior due to
8354 dereferencing a null pointer. Isolate those paths from the main control
8355 flow and turn the statement with erroneous or undefined behavior into a trap.
8356 This flag is enabled by default at @option{-O2} and higher and depends on
8357 @option{-fdelete-null-pointer-checks} also being enabled.
8359 @item -fisolate-erroneous-paths-attribute
8360 @opindex fisolate-erroneous-paths-attribute
8361 Detect paths that trigger erroneous or undefined behavior due a null value
8362 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8363 attribute. Isolate those paths from the main control flow and turn the
8364 statement with erroneous or undefined behavior into a trap. This is not
8365 currently enabled, but may be enabled by @option{-O2} in the future.
8369 Perform forward store motion on trees. This flag is
8370 enabled by default at @option{-O} and higher.
8372 @item -ftree-bit-ccp
8373 @opindex ftree-bit-ccp
8374 Perform sparse conditional bit constant propagation on trees and propagate
8375 pointer alignment information.
8376 This pass only operates on local scalar variables and is enabled by default
8377 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8381 Perform sparse conditional constant propagation (CCP) on trees. This
8382 pass only operates on local scalar variables and is enabled by default
8383 at @option{-O} and higher.
8385 @item -fssa-backprop
8386 @opindex fssa-backprop
8387 Propagate information about uses of a value up the definition chain
8388 in order to simplify the definitions. For example, this pass strips
8389 sign operations if the sign of a value never matters. The flag is
8390 enabled by default at @option{-O} and higher.
8393 @opindex fssa-phiopt
8394 Perform pattern matching on SSA PHI nodes to optimize conditional
8395 code. This pass is enabled by default at @option{-O} and higher.
8397 @item -ftree-switch-conversion
8398 @opindex ftree-switch-conversion
8399 Perform conversion of simple initializations in a switch to
8400 initializations from a scalar array. This flag is enabled by default
8401 at @option{-O2} and higher.
8403 @item -ftree-tail-merge
8404 @opindex ftree-tail-merge
8405 Look for identical code sequences. When found, replace one with a jump to the
8406 other. This optimization is known as tail merging or cross jumping. This flag
8407 is enabled by default at @option{-O2} and higher. The compilation time
8409 be limited using @option{max-tail-merge-comparisons} parameter and
8410 @option{max-tail-merge-iterations} parameter.
8414 Perform dead code elimination (DCE) on trees. This flag is enabled by
8415 default at @option{-O} and higher.
8417 @item -ftree-builtin-call-dce
8418 @opindex ftree-builtin-call-dce
8419 Perform conditional dead code elimination (DCE) for calls to built-in functions
8420 that may set @code{errno} but are otherwise side-effect free. This flag is
8421 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8424 @item -ftree-dominator-opts
8425 @opindex ftree-dominator-opts
8426 Perform a variety of simple scalar cleanups (constant/copy
8427 propagation, redundancy elimination, range propagation and expression
8428 simplification) based on a dominator tree traversal. This also
8429 performs jump threading (to reduce jumps to jumps). This flag is
8430 enabled by default at @option{-O} and higher.
8434 Perform dead store elimination (DSE) on trees. A dead store is a store into
8435 a memory location that is later overwritten by another store without
8436 any intervening loads. In this case the earlier store can be deleted. This
8437 flag is enabled by default at @option{-O} and higher.
8441 Perform loop header copying on trees. This is beneficial since it increases
8442 effectiveness of code motion optimizations. It also saves one jump. This flag
8443 is enabled by default at @option{-O} and higher. It is not enabled
8444 for @option{-Os}, since it usually increases code size.
8446 @item -ftree-loop-optimize
8447 @opindex ftree-loop-optimize
8448 Perform loop optimizations on trees. This flag is enabled by default
8449 at @option{-O} and higher.
8451 @item -ftree-loop-linear
8452 @itemx -floop-interchange
8453 @itemx -floop-strip-mine
8455 @itemx -floop-unroll-and-jam
8456 @opindex ftree-loop-linear
8457 @opindex floop-interchange
8458 @opindex floop-strip-mine
8459 @opindex floop-block
8460 @opindex floop-unroll-and-jam
8461 Perform loop nest optimizations. Same as
8462 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8463 to be configured with @option{--with-isl} to enable the Graphite loop
8464 transformation infrastructure.
8466 @item -fgraphite-identity
8467 @opindex fgraphite-identity
8468 Enable the identity transformation for graphite. For every SCoP we generate
8469 the polyhedral representation and transform it back to gimple. Using
8470 @option{-fgraphite-identity} we can check the costs or benefits of the
8471 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8472 are also performed by the code generator isl, like index splitting and
8473 dead code elimination in loops.
8475 @item -floop-nest-optimize
8476 @opindex floop-nest-optimize
8477 Enable the isl based loop nest optimizer. This is a generic loop nest
8478 optimizer based on the Pluto optimization algorithms. It calculates a loop
8479 structure optimized for data-locality and parallelism. This option
8482 @item -floop-parallelize-all
8483 @opindex floop-parallelize-all
8484 Use the Graphite data dependence analysis to identify loops that can
8485 be parallelized. Parallelize all the loops that can be analyzed to
8486 not contain loop carried dependences without checking that it is
8487 profitable to parallelize the loops.
8489 @item -ftree-coalesce-vars
8490 @opindex ftree-coalesce-vars
8491 While transforming the program out of the SSA representation, attempt to
8492 reduce copying by coalescing versions of different user-defined
8493 variables, instead of just compiler temporaries. This may severely
8494 limit the ability to debug an optimized program compiled with
8495 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8496 prevents SSA coalescing of user variables. This option is enabled by
8497 default if optimization is enabled, and it does very little otherwise.
8499 @item -ftree-loop-if-convert
8500 @opindex ftree-loop-if-convert
8501 Attempt to transform conditional jumps in the innermost loops to
8502 branch-less equivalents. The intent is to remove control-flow from
8503 the innermost loops in order to improve the ability of the
8504 vectorization pass to handle these loops. This is enabled by default
8505 if vectorization is enabled.
8507 @item -ftree-loop-distribution
8508 @opindex ftree-loop-distribution
8509 Perform loop distribution. This flag can improve cache performance on
8510 big loop bodies and allow further loop optimizations, like
8511 parallelization or vectorization, to take place. For example, the loop
8528 @item -ftree-loop-distribute-patterns
8529 @opindex ftree-loop-distribute-patterns
8530 Perform loop distribution of patterns that can be code generated with
8531 calls to a library. This flag is enabled by default at @option{-O3}.
8533 This pass distributes the initialization loops and generates a call to
8534 memset zero. For example, the loop
8550 and the initialization loop is transformed into a call to memset zero.
8552 @item -ftree-loop-im
8553 @opindex ftree-loop-im
8554 Perform loop invariant motion on trees. This pass moves only invariants that
8555 are hard to handle at RTL level (function calls, operations that expand to
8556 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8557 operands of conditions that are invariant out of the loop, so that we can use
8558 just trivial invariantness analysis in loop unswitching. The pass also includes
8561 @item -ftree-loop-ivcanon
8562 @opindex ftree-loop-ivcanon
8563 Create a canonical counter for number of iterations in loops for which
8564 determining number of iterations requires complicated analysis. Later
8565 optimizations then may determine the number easily. Useful especially
8566 in connection with unrolling.
8570 Perform induction variable optimizations (strength reduction, induction
8571 variable merging and induction variable elimination) on trees.
8573 @item -ftree-parallelize-loops=n
8574 @opindex ftree-parallelize-loops
8575 Parallelize loops, i.e., split their iteration space to run in n threads.
8576 This is only possible for loops whose iterations are independent
8577 and can be arbitrarily reordered. The optimization is only
8578 profitable on multiprocessor machines, for loops that are CPU-intensive,
8579 rather than constrained e.g.@: by memory bandwidth. This option
8580 implies @option{-pthread}, and thus is only supported on targets
8581 that have support for @option{-pthread}.
8585 Perform function-local points-to analysis on trees. This flag is
8586 enabled by default at @option{-O} and higher.
8590 Perform scalar replacement of aggregates. This pass replaces structure
8591 references with scalars to prevent committing structures to memory too
8592 early. This flag is enabled by default at @option{-O} and higher.
8594 @item -fstore-merging
8595 @opindex fstore-merging
8596 Perform merging of narrow stores to consecutive memory addresses. This pass
8597 merges contiguous stores of immediate values narrower than a word into fewer
8598 wider stores to reduce the number of instructions. This is enabled by default
8599 at @option{-O2} and higher as well as @option{-Os}.
8603 Perform temporary expression replacement during the SSA->normal phase. Single
8604 use/single def temporaries are replaced at their use location with their
8605 defining expression. This results in non-GIMPLE code, but gives the expanders
8606 much more complex trees to work on resulting in better RTL generation. This is
8607 enabled by default at @option{-O} and higher.
8611 Perform straight-line strength reduction on trees. This recognizes related
8612 expressions involving multiplications and replaces them by less expensive
8613 calculations when possible. This is enabled by default at @option{-O} and
8616 @item -ftree-vectorize
8617 @opindex ftree-vectorize
8618 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8619 and @option{-ftree-slp-vectorize} if not explicitly specified.
8621 @item -ftree-loop-vectorize
8622 @opindex ftree-loop-vectorize
8623 Perform loop vectorization on trees. This flag is enabled by default at
8624 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8626 @item -ftree-slp-vectorize
8627 @opindex ftree-slp-vectorize
8628 Perform basic block vectorization on trees. This flag is enabled by default at
8629 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8631 @item -fvect-cost-model=@var{model}
8632 @opindex fvect-cost-model
8633 Alter the cost model used for vectorization. The @var{model} argument
8634 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8635 With the @samp{unlimited} model the vectorized code-path is assumed
8636 to be profitable while with the @samp{dynamic} model a runtime check
8637 guards the vectorized code-path to enable it only for iteration
8638 counts that will likely execute faster than when executing the original
8639 scalar loop. The @samp{cheap} model disables vectorization of
8640 loops where doing so would be cost prohibitive for example due to
8641 required runtime checks for data dependence or alignment but otherwise
8642 is equal to the @samp{dynamic} model.
8643 The default cost model depends on other optimization flags and is
8644 either @samp{dynamic} or @samp{cheap}.
8646 @item -fsimd-cost-model=@var{model}
8647 @opindex fsimd-cost-model
8648 Alter the cost model used for vectorization of loops marked with the OpenMP
8649 or Cilk Plus simd directive. The @var{model} argument should be one of
8650 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8651 have the same meaning as described in @option{-fvect-cost-model} and by
8652 default a cost model defined with @option{-fvect-cost-model} is used.
8656 Perform Value Range Propagation on trees. This is similar to the
8657 constant propagation pass, but instead of values, ranges of values are
8658 propagated. This allows the optimizers to remove unnecessary range
8659 checks like array bound checks and null pointer checks. This is
8660 enabled by default at @option{-O2} and higher. Null pointer check
8661 elimination is only done if @option{-fdelete-null-pointer-checks} is
8665 @opindex fsplit-paths
8666 Split paths leading to loop backedges. This can improve dead code
8667 elimination and common subexpression elimination. This is enabled by
8668 default at @option{-O2} and above.
8670 @item -fsplit-ivs-in-unroller
8671 @opindex fsplit-ivs-in-unroller
8672 Enables expression of values of induction variables in later iterations
8673 of the unrolled loop using the value in the first iteration. This breaks
8674 long dependency chains, thus improving efficiency of the scheduling passes.
8676 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8677 same effect. However, that is not reliable in cases where the loop body
8678 is more complicated than a single basic block. It also does not work at all
8679 on some architectures due to restrictions in the CSE pass.
8681 This optimization is enabled by default.
8683 @item -fvariable-expansion-in-unroller
8684 @opindex fvariable-expansion-in-unroller
8685 With this option, the compiler creates multiple copies of some
8686 local variables when unrolling a loop, which can result in superior code.
8688 @item -fpartial-inlining
8689 @opindex fpartial-inlining
8690 Inline parts of functions. This option has any effect only
8691 when inlining itself is turned on by the @option{-finline-functions}
8692 or @option{-finline-small-functions} options.
8694 Enabled at level @option{-O2}.
8696 @item -fpredictive-commoning
8697 @opindex fpredictive-commoning
8698 Perform predictive commoning optimization, i.e., reusing computations
8699 (especially memory loads and stores) performed in previous
8700 iterations of loops.
8702 This option is enabled at level @option{-O3}.
8704 @item -fprefetch-loop-arrays
8705 @opindex fprefetch-loop-arrays
8706 If supported by the target machine, generate instructions to prefetch
8707 memory to improve the performance of loops that access large arrays.
8709 This option may generate better or worse code; results are highly
8710 dependent on the structure of loops within the source code.
8712 Disabled at level @option{-Os}.
8714 @item -fno-printf-return-value
8715 @opindex fno-printf-return-value
8716 Do not substitute constants for known return value of formatted output
8717 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
8718 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
8719 transformation allows GCC to optimize or even eliminate branches based
8720 on the known return value of these functions called with arguments that
8721 are either constant, or whose values are known to be in a range that
8722 makes determining the exact return value possible. For example, when
8723 @option{-fprintf-return-value} is in effect, both the branch and the
8724 body of the @code{if} statement (but not the call to @code{snprint})
8725 can be optimized away when @code{i} is a 32-bit or smaller integer
8726 because the return value is guaranteed to be at most 8.
8730 if (snprintf (buf, "%08x", i) >= sizeof buf)
8734 The @option{-fprintf-return-value} option relies on other optimizations
8735 and yields best results with @option{-O2}. It works in tandem with the
8736 @option{-Wformat-overflow} and @option{-Wformat-truncation} options.
8737 The @option{-fprintf-return-value} option is enabled by default.
8740 @itemx -fno-peephole2
8741 @opindex fno-peephole
8742 @opindex fno-peephole2
8743 Disable any machine-specific peephole optimizations. The difference
8744 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8745 are implemented in the compiler; some targets use one, some use the
8746 other, a few use both.
8748 @option{-fpeephole} is enabled by default.
8749 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8751 @item -fno-guess-branch-probability
8752 @opindex fno-guess-branch-probability
8753 Do not guess branch probabilities using heuristics.
8755 GCC uses heuristics to guess branch probabilities if they are
8756 not provided by profiling feedback (@option{-fprofile-arcs}). These
8757 heuristics are based on the control flow graph. If some branch probabilities
8758 are specified by @code{__builtin_expect}, then the heuristics are
8759 used to guess branch probabilities for the rest of the control flow graph,
8760 taking the @code{__builtin_expect} info into account. The interactions
8761 between the heuristics and @code{__builtin_expect} can be complex, and in
8762 some cases, it may be useful to disable the heuristics so that the effects
8763 of @code{__builtin_expect} are easier to understand.
8765 The default is @option{-fguess-branch-probability} at levels
8766 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8768 @item -freorder-blocks
8769 @opindex freorder-blocks
8770 Reorder basic blocks in the compiled function in order to reduce number of
8771 taken branches and improve code locality.
8773 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8775 @item -freorder-blocks-algorithm=@var{algorithm}
8776 @opindex freorder-blocks-algorithm
8777 Use the specified algorithm for basic block reordering. The
8778 @var{algorithm} argument can be @samp{simple}, which does not increase
8779 code size (except sometimes due to secondary effects like alignment),
8780 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8781 put all often executed code together, minimizing the number of branches
8782 executed by making extra copies of code.
8784 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8785 @samp{stc} at levels @option{-O2}, @option{-O3}.
8787 @item -freorder-blocks-and-partition
8788 @opindex freorder-blocks-and-partition
8789 In addition to reordering basic blocks in the compiled function, in order
8790 to reduce number of taken branches, partitions hot and cold basic blocks
8791 into separate sections of the assembly and @file{.o} files, to improve
8792 paging and cache locality performance.
8794 This optimization is automatically turned off in the presence of
8795 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
8796 section attribute and on any architecture that does not support named
8797 sections. When @option{-fsplit-stack} is used this option is not
8798 enabled by default (to avoid linker errors), but may be enabled
8799 explicitly (if using a working linker).
8801 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8803 @item -freorder-functions
8804 @opindex freorder-functions
8805 Reorder functions in the object file in order to
8806 improve code locality. This is implemented by using special
8807 subsections @code{.text.hot} for most frequently executed functions and
8808 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8809 the linker so object file format must support named sections and linker must
8810 place them in a reasonable way.
8812 Also profile feedback must be available to make this option effective. See
8813 @option{-fprofile-arcs} for details.
8815 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8817 @item -fstrict-aliasing
8818 @opindex fstrict-aliasing
8819 Allow the compiler to assume the strictest aliasing rules applicable to
8820 the language being compiled. For C (and C++), this activates
8821 optimizations based on the type of expressions. In particular, an
8822 object of one type is assumed never to reside at the same address as an
8823 object of a different type, unless the types are almost the same. For
8824 example, an @code{unsigned int} can alias an @code{int}, but not a
8825 @code{void*} or a @code{double}. A character type may alias any other
8828 @anchor{Type-punning}Pay special attention to code like this:
8841 The practice of reading from a different union member than the one most
8842 recently written to (called ``type-punning'') is common. Even with
8843 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8844 is accessed through the union type. So, the code above works as
8845 expected. @xref{Structures unions enumerations and bit-fields
8846 implementation}. However, this code might not:
8857 Similarly, access by taking the address, casting the resulting pointer
8858 and dereferencing the result has undefined behavior, even if the cast
8859 uses a union type, e.g.:
8863 return ((union a_union *) &d)->i;
8867 The @option{-fstrict-aliasing} option is enabled at levels
8868 @option{-O2}, @option{-O3}, @option{-Os}.
8870 @item -falign-functions
8871 @itemx -falign-functions=@var{n}
8872 @opindex falign-functions
8873 Align the start of functions to the next power-of-two greater than
8874 @var{n}, skipping up to @var{n} bytes. For instance,
8875 @option{-falign-functions=32} aligns functions to the next 32-byte
8876 boundary, but @option{-falign-functions=24} aligns to the next
8877 32-byte boundary only if this can be done by skipping 23 bytes or less.
8879 @option{-fno-align-functions} and @option{-falign-functions=1} are
8880 equivalent and mean that functions are not aligned.
8882 Some assemblers only support this flag when @var{n} is a power of two;
8883 in that case, it is rounded up.
8885 If @var{n} is not specified or is zero, use a machine-dependent default.
8887 Enabled at levels @option{-O2}, @option{-O3}.
8889 @item -flimit-function-alignment
8890 If this option is enabled, the compiler tries to avoid unnecessarily
8891 overaligning functions. It attempts to instruct the assembler to align
8892 by the amount specified by @option{-falign-functions}, but not to
8893 skip more bytes than the size of the function.
8895 @item -falign-labels
8896 @itemx -falign-labels=@var{n}
8897 @opindex falign-labels
8898 Align all branch targets to a power-of-two boundary, skipping up to
8899 @var{n} bytes like @option{-falign-functions}. This option can easily
8900 make code slower, because it must insert dummy operations for when the
8901 branch target is reached in the usual flow of the code.
8903 @option{-fno-align-labels} and @option{-falign-labels=1} are
8904 equivalent and mean that labels are not aligned.
8906 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8907 are greater than this value, then their values are used instead.
8909 If @var{n} is not specified or is zero, use a machine-dependent default
8910 which is very likely to be @samp{1}, meaning no alignment.
8912 Enabled at levels @option{-O2}, @option{-O3}.
8915 @itemx -falign-loops=@var{n}
8916 @opindex falign-loops
8917 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8918 like @option{-falign-functions}. If the loops are
8919 executed many times, this makes up for any execution of the dummy
8922 @option{-fno-align-loops} and @option{-falign-loops=1} are
8923 equivalent and mean that loops are not aligned.
8925 If @var{n} is not specified or is zero, use a machine-dependent default.
8927 Enabled at levels @option{-O2}, @option{-O3}.
8930 @itemx -falign-jumps=@var{n}
8931 @opindex falign-jumps
8932 Align branch targets to a power-of-two boundary, for branch targets
8933 where the targets can only be reached by jumping, skipping up to @var{n}
8934 bytes like @option{-falign-functions}. In this case, no dummy operations
8937 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8938 equivalent and mean that loops are not aligned.
8940 If @var{n} is not specified or is zero, use a machine-dependent default.
8942 Enabled at levels @option{-O2}, @option{-O3}.
8944 @item -funit-at-a-time
8945 @opindex funit-at-a-time
8946 This option is left for compatibility reasons. @option{-funit-at-a-time}
8947 has no effect, while @option{-fno-unit-at-a-time} implies
8948 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8952 @item -fno-toplevel-reorder
8953 @opindex fno-toplevel-reorder
8954 Do not reorder top-level functions, variables, and @code{asm}
8955 statements. Output them in the same order that they appear in the
8956 input file. When this option is used, unreferenced static variables
8957 are not removed. This option is intended to support existing code
8958 that relies on a particular ordering. For new code, it is better to
8959 use attributes when possible.
8961 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8962 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8967 Constructs webs as commonly used for register allocation purposes and assign
8968 each web individual pseudo register. This allows the register allocation pass
8969 to operate on pseudos directly, but also strengthens several other optimization
8970 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8971 however, make debugging impossible, since variables no longer stay in a
8974 Enabled by default with @option{-funroll-loops}.
8976 @item -fwhole-program
8977 @opindex fwhole-program
8978 Assume that the current compilation unit represents the whole program being
8979 compiled. All public functions and variables with the exception of @code{main}
8980 and those merged by attribute @code{externally_visible} become static functions
8981 and in effect are optimized more aggressively by interprocedural optimizers.
8983 This option should not be used in combination with @option{-flto}.
8984 Instead relying on a linker plugin should provide safer and more precise
8987 @item -flto[=@var{n}]
8989 This option runs the standard link-time optimizer. When invoked
8990 with source code, it generates GIMPLE (one of GCC's internal
8991 representations) and writes it to special ELF sections in the object
8992 file. When the object files are linked together, all the function
8993 bodies are read from these ELF sections and instantiated as if they
8994 had been part of the same translation unit.
8996 To use the link-time optimizer, @option{-flto} and optimization
8997 options should be specified at compile time and during the final link.
8998 It is recommended that you compile all the files participating in the
8999 same link with the same options and also specify those options at
9004 gcc -c -O2 -flto foo.c
9005 gcc -c -O2 -flto bar.c
9006 gcc -o myprog -flto -O2 foo.o bar.o
9009 The first two invocations to GCC save a bytecode representation
9010 of GIMPLE into special ELF sections inside @file{foo.o} and
9011 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9012 @file{foo.o} and @file{bar.o}, merges the two files into a single
9013 internal image, and compiles the result as usual. Since both
9014 @file{foo.o} and @file{bar.o} are merged into a single image, this
9015 causes all the interprocedural analyses and optimizations in GCC to
9016 work across the two files as if they were a single one. This means,
9017 for example, that the inliner is able to inline functions in
9018 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9020 Another (simpler) way to enable link-time optimization is:
9023 gcc -o myprog -flto -O2 foo.c bar.c
9026 The above generates bytecode for @file{foo.c} and @file{bar.c},
9027 merges them together into a single GIMPLE representation and optimizes
9028 them as usual to produce @file{myprog}.
9030 The only important thing to keep in mind is that to enable link-time
9031 optimizations you need to use the GCC driver to perform the link step.
9032 GCC then automatically performs link-time optimization if any of the
9033 objects involved were compiled with the @option{-flto} command-line option.
9035 should specify the optimization options to be used for link-time
9036 optimization though GCC tries to be clever at guessing an
9037 optimization level to use from the options used at compile time
9038 if you fail to specify one at link time. You can always override
9039 the automatic decision to do link-time optimization
9040 by passing @option{-fno-lto} to the link command.
9042 To make whole program optimization effective, it is necessary to make
9043 certain whole program assumptions. The compiler needs to know
9044 what functions and variables can be accessed by libraries and runtime
9045 outside of the link-time optimized unit. When supported by the linker,
9046 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9047 to the compiler about used and externally visible symbols. When
9048 the linker plugin is not available, @option{-fwhole-program} should be
9049 used to allow the compiler to make these assumptions, which leads
9050 to more aggressive optimization decisions.
9052 When @option{-fuse-linker-plugin} is not enabled, when a file is
9053 compiled with @option{-flto}, the generated object file is larger than
9054 a regular object file because it contains GIMPLE bytecodes and the usual
9055 final code (see @option{-ffat-lto-objects}. This means that
9056 object files with LTO information can be linked as normal object
9057 files; if @option{-fno-lto} is passed to the linker, no
9058 interprocedural optimizations are applied. Note that when
9059 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
9060 but you cannot perform a regular, non-LTO link on them.
9062 Additionally, the optimization flags used to compile individual files
9063 are not necessarily related to those used at link time. For instance,
9066 gcc -c -O0 -ffat-lto-objects -flto foo.c
9067 gcc -c -O0 -ffat-lto-objects -flto bar.c
9068 gcc -o myprog -O3 foo.o bar.o
9071 This produces individual object files with unoptimized assembler
9072 code, but the resulting binary @file{myprog} is optimized at
9073 @option{-O3}. If, instead, the final binary is generated with
9074 @option{-fno-lto}, then @file{myprog} is not optimized.
9076 When producing the final binary, GCC only
9077 applies link-time optimizations to those files that contain bytecode.
9078 Therefore, you can mix and match object files and libraries with
9079 GIMPLE bytecodes and final object code. GCC automatically selects
9080 which files to optimize in LTO mode and which files to link without
9083 There are some code generation flags preserved by GCC when
9084 generating bytecodes, as they need to be used during the final link
9085 stage. Generally options specified at link time override those
9086 specified at compile time.
9088 If you do not specify an optimization level option @option{-O} at
9089 link time, then GCC uses the highest optimization level
9090 used when compiling the object files.
9092 Currently, the following options and their settings are taken from
9093 the first object file that explicitly specifies them:
9094 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9095 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9096 and all the @option{-m} target flags.
9098 Certain ABI-changing flags are required to match in all compilation units,
9099 and trying to override this at link time with a conflicting value
9100 is ignored. This includes options such as @option{-freg-struct-return}
9101 and @option{-fpcc-struct-return}.
9103 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9104 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9105 are passed through to the link stage and merged conservatively for
9106 conflicting translation units. Specifically
9107 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9108 precedence; and for example @option{-ffp-contract=off} takes precedence
9109 over @option{-ffp-contract=fast}. You can override them at link time.
9111 If LTO encounters objects with C linkage declared with incompatible
9112 types in separate translation units to be linked together (undefined
9113 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9114 issued. The behavior is still undefined at run time. Similar
9115 diagnostics may be raised for other languages.
9117 Another feature of LTO is that it is possible to apply interprocedural
9118 optimizations on files written in different languages:
9123 gfortran -c -flto baz.f90
9124 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9127 Notice that the final link is done with @command{g++} to get the C++
9128 runtime libraries and @option{-lgfortran} is added to get the Fortran
9129 runtime libraries. In general, when mixing languages in LTO mode, you
9130 should use the same link command options as when mixing languages in a
9131 regular (non-LTO) compilation.
9133 If object files containing GIMPLE bytecode are stored in a library archive, say
9134 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9135 are using a linker with plugin support. To create static libraries suitable
9136 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9137 and @command{ranlib};
9138 to show the symbols of object files with GIMPLE bytecode, use
9139 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9140 and @command{nm} have been compiled with plugin support. At link time, use the the
9141 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9142 the LTO optimization process:
9145 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9148 With the linker plugin enabled, the linker extracts the needed
9149 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9150 to make them part of the aggregated GIMPLE image to be optimized.
9152 If you are not using a linker with plugin support and/or do not
9153 enable the linker plugin, then the objects inside @file{libfoo.a}
9154 are extracted and linked as usual, but they do not participate
9155 in the LTO optimization process. In order to make a static library suitable
9156 for both LTO optimization and usual linkage, compile its object files with
9157 @option{-flto} @option{-ffat-lto-objects}.
9159 Link-time optimizations do not require the presence of the whole program to
9160 operate. If the program does not require any symbols to be exported, it is
9161 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9162 the interprocedural optimizers to use more aggressive assumptions which may
9163 lead to improved optimization opportunities.
9164 Use of @option{-fwhole-program} is not needed when linker plugin is
9165 active (see @option{-fuse-linker-plugin}).
9167 The current implementation of LTO makes no
9168 attempt to generate bytecode that is portable between different
9169 types of hosts. The bytecode files are versioned and there is a
9170 strict version check, so bytecode files generated in one version of
9171 GCC do not work with an older or newer version of GCC.
9173 Link-time optimization does not work well with generation of debugging
9174 information. Combining @option{-flto} with
9175 @option{-g} is currently experimental and expected to produce unexpected
9178 If you specify the optional @var{n}, the optimization and code
9179 generation done at link time is executed in parallel using @var{n}
9180 parallel jobs by utilizing an installed @command{make} program. The
9181 environment variable @env{MAKE} may be used to override the program
9182 used. The default value for @var{n} is 1.
9184 You can also specify @option{-flto=jobserver} to use GNU make's
9185 job server mode to determine the number of parallel jobs. This
9186 is useful when the Makefile calling GCC is already executing in parallel.
9187 You must prepend a @samp{+} to the command recipe in the parent Makefile
9188 for this to work. This option likely only works if @env{MAKE} is
9191 @item -flto-partition=@var{alg}
9192 @opindex flto-partition
9193 Specify the partitioning algorithm used by the link-time optimizer.
9194 The value is either @samp{1to1} to specify a partitioning mirroring
9195 the original source files or @samp{balanced} to specify partitioning
9196 into equally sized chunks (whenever possible) or @samp{max} to create
9197 new partition for every symbol where possible. Specifying @samp{none}
9198 as an algorithm disables partitioning and streaming completely.
9199 The default value is @samp{balanced}. While @samp{1to1} can be used
9200 as an workaround for various code ordering issues, the @samp{max}
9201 partitioning is intended for internal testing only.
9202 The value @samp{one} specifies that exactly one partition should be
9203 used while the value @samp{none} bypasses partitioning and executes
9204 the link-time optimization step directly from the WPA phase.
9206 @item -flto-odr-type-merging
9207 @opindex flto-odr-type-merging
9208 Enable streaming of mangled types names of C++ types and their unification
9209 at link time. This increases size of LTO object files, but enables
9210 diagnostics about One Definition Rule violations.
9212 @item -flto-compression-level=@var{n}
9213 @opindex flto-compression-level
9214 This option specifies the level of compression used for intermediate
9215 language written to LTO object files, and is only meaningful in
9216 conjunction with LTO mode (@option{-flto}). Valid
9217 values are 0 (no compression) to 9 (maximum compression). Values
9218 outside this range are clamped to either 0 or 9. If the option is not
9219 given, a default balanced compression setting is used.
9221 @item -fuse-linker-plugin
9222 @opindex fuse-linker-plugin
9223 Enables the use of a linker plugin during link-time optimization. This
9224 option relies on plugin support in the linker, which is available in gold
9225 or in GNU ld 2.21 or newer.
9227 This option enables the extraction of object files with GIMPLE bytecode out
9228 of library archives. This improves the quality of optimization by exposing
9229 more code to the link-time optimizer. This information specifies what
9230 symbols can be accessed externally (by non-LTO object or during dynamic
9231 linking). Resulting code quality improvements on binaries (and shared
9232 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9233 See @option{-flto} for a description of the effect of this flag and how to
9236 This option is enabled by default when LTO support in GCC is enabled
9237 and GCC was configured for use with
9238 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9240 @item -ffat-lto-objects
9241 @opindex ffat-lto-objects
9242 Fat LTO objects are object files that contain both the intermediate language
9243 and the object code. This makes them usable for both LTO linking and normal
9244 linking. This option is effective only when compiling with @option{-flto}
9245 and is ignored at link time.
9247 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9248 requires the complete toolchain to be aware of LTO. It requires a linker with
9249 linker plugin support for basic functionality. Additionally,
9250 @command{nm}, @command{ar} and @command{ranlib}
9251 need to support linker plugins to allow a full-featured build environment
9252 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9253 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9254 to these tools. With non fat LTO makefiles need to be modified to use them.
9256 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9259 @item -fcompare-elim
9260 @opindex fcompare-elim
9261 After register allocation and post-register allocation instruction splitting,
9262 identify arithmetic instructions that compute processor flags similar to a
9263 comparison operation based on that arithmetic. If possible, eliminate the
9264 explicit comparison operation.
9266 This pass only applies to certain targets that cannot explicitly represent
9267 the comparison operation before register allocation is complete.
9269 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9271 @item -fcprop-registers
9272 @opindex fcprop-registers
9273 After register allocation and post-register allocation instruction splitting,
9274 perform a copy-propagation pass to try to reduce scheduling dependencies
9275 and occasionally eliminate the copy.
9277 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9279 @item -fprofile-correction
9280 @opindex fprofile-correction
9281 Profiles collected using an instrumented binary for multi-threaded programs may
9282 be inconsistent due to missed counter updates. When this option is specified,
9283 GCC uses heuristics to correct or smooth out such inconsistencies. By
9284 default, GCC emits an error message when an inconsistent profile is detected.
9287 @itemx -fprofile-use=@var{path}
9288 @opindex fprofile-use
9289 Enable profile feedback-directed optimizations,
9290 and the following optimizations
9291 which are generally profitable only with profile feedback available:
9292 @option{-fbranch-probabilities}, @option{-fvpt},
9293 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9294 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9296 Before you can use this option, you must first generate profiling information.
9297 @xref{Instrumentation Options}, for information about the
9298 @option{-fprofile-generate} option.
9300 By default, GCC emits an error message if the feedback profiles do not
9301 match the source code. This error can be turned into a warning by using
9302 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9305 If @var{path} is specified, GCC looks at the @var{path} to find
9306 the profile feedback data files. See @option{-fprofile-dir}.
9308 @item -fauto-profile
9309 @itemx -fauto-profile=@var{path}
9310 @opindex fauto-profile
9311 Enable sampling-based feedback-directed optimizations,
9312 and the following optimizations
9313 which are generally profitable only with profile feedback available:
9314 @option{-fbranch-probabilities}, @option{-fvpt},
9315 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9316 @option{-ftree-vectorize},
9317 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9318 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9319 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9321 @var{path} is the name of a file containing AutoFDO profile information.
9322 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9324 Producing an AutoFDO profile data file requires running your program
9325 with the @command{perf} utility on a supported GNU/Linux target system.
9326 For more information, see @uref{https://perf.wiki.kernel.org/}.
9330 perf record -e br_inst_retired:near_taken -b -o perf.data \
9334 Then use the @command{create_gcov} tool to convert the raw profile data
9335 to a format that can be used by GCC.@ You must also supply the
9336 unstripped binary for your program to this tool.
9337 See @uref{https://github.com/google/autofdo}.
9341 create_gcov --binary=your_program.unstripped --profile=perf.data \
9346 The following options control compiler behavior regarding floating-point
9347 arithmetic. These options trade off between speed and
9348 correctness. All must be specifically enabled.
9352 @opindex ffloat-store
9353 Do not store floating-point variables in registers, and inhibit other
9354 options that might change whether a floating-point value is taken from a
9357 @cindex floating-point precision
9358 This option prevents undesirable excess precision on machines such as
9359 the 68000 where the floating registers (of the 68881) keep more
9360 precision than a @code{double} is supposed to have. Similarly for the
9361 x86 architecture. For most programs, the excess precision does only
9362 good, but a few programs rely on the precise definition of IEEE floating
9363 point. Use @option{-ffloat-store} for such programs, after modifying
9364 them to store all pertinent intermediate computations into variables.
9366 @item -fexcess-precision=@var{style}
9367 @opindex fexcess-precision
9368 This option allows further control over excess precision on machines
9369 where floating-point operations occur in a format with more precision or
9370 range than the IEEE standard and interchange floating-point types. By
9371 default, @option{-fexcess-precision=fast} is in effect; this means that
9372 operations may be carried out in a wider precision than the types specified
9373 in the source if that would result in faster code, and it is unpredictable
9374 when rounding to the types specified in the source code takes place.
9375 When compiling C, if @option{-fexcess-precision=standard} is specified then
9376 excess precision follows the rules specified in ISO C99; in particular,
9377 both casts and assignments cause values to be rounded to their
9378 semantic types (whereas @option{-ffloat-store} only affects
9379 assignments). This option is enabled by default for C if a strict
9380 conformance option such as @option{-std=c99} is used.
9381 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
9382 regardless of whether a strict conformance option is used.
9385 @option{-fexcess-precision=standard} is not implemented for languages
9386 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
9387 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9388 semantics apply without excess precision, and in the latter, rounding
9393 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9394 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9395 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
9396 @option{-fexcess-precision=fast}.
9398 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9400 This option is not turned on by any @option{-O} option besides
9401 @option{-Ofast} since it can result in incorrect output for programs
9402 that depend on an exact implementation of IEEE or ISO rules/specifications
9403 for math functions. It may, however, yield faster code for programs
9404 that do not require the guarantees of these specifications.
9406 @item -fno-math-errno
9407 @opindex fno-math-errno
9408 Do not set @code{errno} after calling math functions that are executed
9409 with a single instruction, e.g., @code{sqrt}. A program that relies on
9410 IEEE exceptions for math error handling may want to use this flag
9411 for speed while maintaining IEEE arithmetic compatibility.
9413 This option is not turned on by any @option{-O} option since
9414 it can result in incorrect output for programs that depend on
9415 an exact implementation of IEEE or ISO rules/specifications for
9416 math functions. It may, however, yield faster code for programs
9417 that do not require the guarantees of these specifications.
9419 The default is @option{-fmath-errno}.
9421 On Darwin systems, the math library never sets @code{errno}. There is
9422 therefore no reason for the compiler to consider the possibility that
9423 it might, and @option{-fno-math-errno} is the default.
9425 @item -funsafe-math-optimizations
9426 @opindex funsafe-math-optimizations
9428 Allow optimizations for floating-point arithmetic that (a) assume
9429 that arguments and results are valid and (b) may violate IEEE or
9430 ANSI standards. When used at link time, it may include libraries
9431 or startup files that change the default FPU control word or other
9432 similar optimizations.
9434 This option is not turned on by any @option{-O} option since
9435 it can result in incorrect output for programs that depend on
9436 an exact implementation of IEEE or ISO rules/specifications for
9437 math functions. It may, however, yield faster code for programs
9438 that do not require the guarantees of these specifications.
9439 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9440 @option{-fassociative-math} and @option{-freciprocal-math}.
9442 The default is @option{-fno-unsafe-math-optimizations}.
9444 @item -fassociative-math
9445 @opindex fassociative-math
9447 Allow re-association of operands in series of floating-point operations.
9448 This violates the ISO C and C++ language standard by possibly changing
9449 computation result. NOTE: re-ordering may change the sign of zero as
9450 well as ignore NaNs and inhibit or create underflow or overflow (and
9451 thus cannot be used on code that relies on rounding behavior like
9452 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9453 and thus may not be used when ordered comparisons are required.
9454 This option requires that both @option{-fno-signed-zeros} and
9455 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9456 much sense with @option{-frounding-math}. For Fortran the option
9457 is automatically enabled when both @option{-fno-signed-zeros} and
9458 @option{-fno-trapping-math} are in effect.
9460 The default is @option{-fno-associative-math}.
9462 @item -freciprocal-math
9463 @opindex freciprocal-math
9465 Allow the reciprocal of a value to be used instead of dividing by
9466 the value if this enables optimizations. For example @code{x / y}
9467 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9468 is subject to common subexpression elimination. Note that this loses
9469 precision and increases the number of flops operating on the value.
9471 The default is @option{-fno-reciprocal-math}.
9473 @item -ffinite-math-only
9474 @opindex ffinite-math-only
9475 Allow optimizations for floating-point arithmetic that assume
9476 that arguments and results are not NaNs or +-Infs.
9478 This option is not turned on by any @option{-O} option since
9479 it can result in incorrect output for programs that depend on
9480 an exact implementation of IEEE or ISO rules/specifications for
9481 math functions. It may, however, yield faster code for programs
9482 that do not require the guarantees of these specifications.
9484 The default is @option{-fno-finite-math-only}.
9486 @item -fno-signed-zeros
9487 @opindex fno-signed-zeros
9488 Allow optimizations for floating-point arithmetic that ignore the
9489 signedness of zero. IEEE arithmetic specifies the behavior of
9490 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9491 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9492 This option implies that the sign of a zero result isn't significant.
9494 The default is @option{-fsigned-zeros}.
9496 @item -fno-trapping-math
9497 @opindex fno-trapping-math
9498 Compile code assuming that floating-point operations cannot generate
9499 user-visible traps. These traps include division by zero, overflow,
9500 underflow, inexact result and invalid operation. This option requires
9501 that @option{-fno-signaling-nans} be in effect. Setting this option may
9502 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9504 This option should never be turned on by any @option{-O} option since
9505 it can result in incorrect output for programs that depend on
9506 an exact implementation of IEEE or ISO rules/specifications for
9509 The default is @option{-ftrapping-math}.
9511 @item -frounding-math
9512 @opindex frounding-math
9513 Disable transformations and optimizations that assume default floating-point
9514 rounding behavior. This is round-to-zero for all floating point
9515 to integer conversions, and round-to-nearest for all other arithmetic
9516 truncations. This option should be specified for programs that change
9517 the FP rounding mode dynamically, or that may be executed with a
9518 non-default rounding mode. This option disables constant folding of
9519 floating-point expressions at compile time (which may be affected by
9520 rounding mode) and arithmetic transformations that are unsafe in the
9521 presence of sign-dependent rounding modes.
9523 The default is @option{-fno-rounding-math}.
9525 This option is experimental and does not currently guarantee to
9526 disable all GCC optimizations that are affected by rounding mode.
9527 Future versions of GCC may provide finer control of this setting
9528 using C99's @code{FENV_ACCESS} pragma. This command-line option
9529 will be used to specify the default state for @code{FENV_ACCESS}.
9531 @item -fsignaling-nans
9532 @opindex fsignaling-nans
9533 Compile code assuming that IEEE signaling NaNs may generate user-visible
9534 traps during floating-point operations. Setting this option disables
9535 optimizations that may change the number of exceptions visible with
9536 signaling NaNs. This option implies @option{-ftrapping-math}.
9538 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9541 The default is @option{-fno-signaling-nans}.
9543 This option is experimental and does not currently guarantee to
9544 disable all GCC optimizations that affect signaling NaN behavior.
9546 @item -fno-fp-int-builtin-inexact
9547 @opindex fno-fp-int-builtin-inexact
9548 Do not allow the built-in functions @code{ceil}, @code{floor},
9549 @code{round} and @code{trunc}, and their @code{float} and @code{long
9550 double} variants, to generate code that raises the ``inexact''
9551 floating-point exception for noninteger arguments. ISO C99 and C11
9552 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9553 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9556 The default is @option{-ffp-int-builtin-inexact}, allowing the
9557 exception to be raised. This option does nothing unless
9558 @option{-ftrapping-math} is in effect.
9560 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9561 generate a call to a library function then the ``inexact'' exception
9562 may be raised if the library implementation does not follow TS 18661.
9564 @item -fsingle-precision-constant
9565 @opindex fsingle-precision-constant
9566 Treat floating-point constants as single precision instead of
9567 implicitly converting them to double-precision constants.
9569 @item -fcx-limited-range
9570 @opindex fcx-limited-range
9571 When enabled, this option states that a range reduction step is not
9572 needed when performing complex division. Also, there is no checking
9573 whether the result of a complex multiplication or division is @code{NaN
9574 + I*NaN}, with an attempt to rescue the situation in that case. The
9575 default is @option{-fno-cx-limited-range}, but is enabled by
9576 @option{-ffast-math}.
9578 This option controls the default setting of the ISO C99
9579 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9582 @item -fcx-fortran-rules
9583 @opindex fcx-fortran-rules
9584 Complex multiplication and division follow Fortran rules. Range
9585 reduction is done as part of complex division, but there is no checking
9586 whether the result of a complex multiplication or division is @code{NaN
9587 + I*NaN}, with an attempt to rescue the situation in that case.
9589 The default is @option{-fno-cx-fortran-rules}.
9593 The following options control optimizations that may improve
9594 performance, but are not enabled by any @option{-O} options. This
9595 section includes experimental options that may produce broken code.
9598 @item -fbranch-probabilities
9599 @opindex fbranch-probabilities
9600 After running a program compiled with @option{-fprofile-arcs}
9601 (@pxref{Instrumentation Options}),
9602 you can compile it a second time using
9603 @option{-fbranch-probabilities}, to improve optimizations based on
9604 the number of times each branch was taken. When a program
9605 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9606 counts to a file called @file{@var{sourcename}.gcda} for each source
9607 file. The information in this data file is very dependent on the
9608 structure of the generated code, so you must use the same source code
9609 and the same optimization options for both compilations.
9611 With @option{-fbranch-probabilities}, GCC puts a
9612 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9613 These can be used to improve optimization. Currently, they are only
9614 used in one place: in @file{reorg.c}, instead of guessing which path a
9615 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9616 exactly determine which path is taken more often.
9618 @item -fprofile-values
9619 @opindex fprofile-values
9620 If combined with @option{-fprofile-arcs}, it adds code so that some
9621 data about values of expressions in the program is gathered.
9623 With @option{-fbranch-probabilities}, it reads back the data gathered
9624 from profiling values of expressions for usage in optimizations.
9626 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9628 @item -fprofile-reorder-functions
9629 @opindex fprofile-reorder-functions
9630 Function reordering based on profile instrumentation collects
9631 first time of execution of a function and orders these functions
9634 Enabled with @option{-fprofile-use}.
9638 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9639 to add code to gather information about values of expressions.
9641 With @option{-fbranch-probabilities}, it reads back the data gathered
9642 and actually performs the optimizations based on them.
9643 Currently the optimizations include specialization of division operations
9644 using the knowledge about the value of the denominator.
9646 @item -frename-registers
9647 @opindex frename-registers
9648 Attempt to avoid false dependencies in scheduled code by making use
9649 of registers left over after register allocation. This optimization
9650 most benefits processors with lots of registers. Depending on the
9651 debug information format adopted by the target, however, it can
9652 make debugging impossible, since variables no longer stay in
9653 a ``home register''.
9655 Enabled by default with @option{-funroll-loops}.
9657 @item -fschedule-fusion
9658 @opindex fschedule-fusion
9659 Performs a target dependent pass over the instruction stream to schedule
9660 instructions of same type together because target machine can execute them
9661 more efficiently if they are adjacent to each other in the instruction flow.
9663 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9667 Perform tail duplication to enlarge superblock size. This transformation
9668 simplifies the control flow of the function allowing other optimizations to do
9671 Enabled with @option{-fprofile-use}.
9673 @item -funroll-loops
9674 @opindex funroll-loops
9675 Unroll loops whose number of iterations can be determined at compile time or
9676 upon entry to the loop. @option{-funroll-loops} implies
9677 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9678 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9679 a small constant number of iterations). This option makes code larger, and may
9680 or may not make it run faster.
9682 Enabled with @option{-fprofile-use}.
9684 @item -funroll-all-loops
9685 @opindex funroll-all-loops
9686 Unroll all loops, even if their number of iterations is uncertain when
9687 the loop is entered. This usually makes programs run more slowly.
9688 @option{-funroll-all-loops} implies the same options as
9689 @option{-funroll-loops}.
9692 @opindex fpeel-loops
9693 Peels loops for which there is enough information that they do not
9694 roll much (from profile feedback or static analysis). It also turns on
9695 complete loop peeling (i.e.@: complete removal of loops with small constant
9696 number of iterations).
9698 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9700 @item -fmove-loop-invariants
9701 @opindex fmove-loop-invariants
9702 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9703 at level @option{-O1}
9706 @opindex fsplit-loops
9707 Split a loop into two if it contains a condition that's always true
9708 for one side of the iteration space and false for the other.
9710 @item -funswitch-loops
9711 @opindex funswitch-loops
9712 Move branches with loop invariant conditions out of the loop, with duplicates
9713 of the loop on both branches (modified according to result of the condition).
9715 @item -ffunction-sections
9716 @itemx -fdata-sections
9717 @opindex ffunction-sections
9718 @opindex fdata-sections
9719 Place each function or data item into its own section in the output
9720 file if the target supports arbitrary sections. The name of the
9721 function or the name of the data item determines the section's name
9724 Use these options on systems where the linker can perform optimizations to
9725 improve locality of reference in the instruction space. Most systems using the
9726 ELF object format have linkers with such optimizations. On AIX, the linker
9727 rearranges sections (CSECTs) based on the call graph. The performance impact
9730 Together with a linker garbage collection (linker @option{--gc-sections}
9731 option) these options may lead to smaller statically-linked executables (after
9734 On ELF/DWARF systems these options do not degenerate the quality of the debug
9735 information. There could be issues with other object files/debug info formats.
9737 Only use these options when there are significant benefits from doing so. When
9738 you specify these options, the assembler and linker create larger object and
9739 executable files and are also slower. These options affect code generation.
9740 They prevent optimizations by the compiler and assembler using relative
9741 locations inside a translation unit since the locations are unknown until
9742 link time. An example of such an optimization is relaxing calls to short call
9745 @item -fbranch-target-load-optimize
9746 @opindex fbranch-target-load-optimize
9747 Perform branch target register load optimization before prologue / epilogue
9749 The use of target registers can typically be exposed only during reload,
9750 thus hoisting loads out of loops and doing inter-block scheduling needs
9751 a separate optimization pass.
9753 @item -fbranch-target-load-optimize2
9754 @opindex fbranch-target-load-optimize2
9755 Perform branch target register load optimization after prologue / epilogue
9758 @item -fbtr-bb-exclusive
9759 @opindex fbtr-bb-exclusive
9760 When performing branch target register load optimization, don't reuse
9761 branch target registers within any basic block.
9764 @opindex fstdarg-opt
9765 Optimize the prologue of variadic argument functions with respect to usage of
9768 @item -fsection-anchors
9769 @opindex fsection-anchors
9770 Try to reduce the number of symbolic address calculations by using
9771 shared ``anchor'' symbols to address nearby objects. This transformation
9772 can help to reduce the number of GOT entries and GOT accesses on some
9775 For example, the implementation of the following function @code{foo}:
9779 int foo (void) @{ return a + b + c; @}
9783 usually calculates the addresses of all three variables, but if you
9784 compile it with @option{-fsection-anchors}, it accesses the variables
9785 from a common anchor point instead. The effect is similar to the
9786 following pseudocode (which isn't valid C):
9791 register int *xr = &x;
9792 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9796 Not all targets support this option.
9798 @item --param @var{name}=@var{value}
9800 In some places, GCC uses various constants to control the amount of
9801 optimization that is done. For example, GCC does not inline functions
9802 that contain more than a certain number of instructions. You can
9803 control some of these constants on the command line using the
9804 @option{--param} option.
9806 The names of specific parameters, and the meaning of the values, are
9807 tied to the internals of the compiler, and are subject to change
9808 without notice in future releases.
9810 In each case, the @var{value} is an integer. The allowable choices for
9814 @item predictable-branch-outcome
9815 When branch is predicted to be taken with probability lower than this threshold
9816 (in percent), then it is considered well predictable. The default is 10.
9818 @item max-rtl-if-conversion-insns
9819 RTL if-conversion tries to remove conditional branches around a block and
9820 replace them with conditionally executed instructions. This parameter
9821 gives the maximum number of instructions in a block which should be
9822 considered for if-conversion. The default is 10, though the compiler will
9823 also use other heuristics to decide whether if-conversion is likely to be
9826 @item max-rtl-if-conversion-predictable-cost
9827 @item max-rtl-if-conversion-unpredictable-cost
9828 RTL if-conversion will try to remove conditional branches around a block
9829 and replace them with conditionally executed instructions. These parameters
9830 give the maximum permissible cost for the sequence that would be generated
9831 by if-conversion depending on whether the branch is statically determined
9832 to be predictable or not. The units for this parameter are the same as
9833 those for the GCC internal seq_cost metric. The compiler will try to
9834 provide a reasonable default for this parameter using the BRANCH_COST
9837 @item max-crossjump-edges
9838 The maximum number of incoming edges to consider for cross-jumping.
9839 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9840 the number of edges incoming to each block. Increasing values mean
9841 more aggressive optimization, making the compilation time increase with
9842 probably small improvement in executable size.
9844 @item min-crossjump-insns
9845 The minimum number of instructions that must be matched at the end
9846 of two blocks before cross-jumping is performed on them. This
9847 value is ignored in the case where all instructions in the block being
9848 cross-jumped from are matched. The default value is 5.
9850 @item max-grow-copy-bb-insns
9851 The maximum code size expansion factor when copying basic blocks
9852 instead of jumping. The expansion is relative to a jump instruction.
9853 The default value is 8.
9855 @item max-goto-duplication-insns
9856 The maximum number of instructions to duplicate to a block that jumps
9857 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9858 passes, GCC factors computed gotos early in the compilation process,
9859 and unfactors them as late as possible. Only computed jumps at the
9860 end of a basic blocks with no more than max-goto-duplication-insns are
9861 unfactored. The default value is 8.
9863 @item max-delay-slot-insn-search
9864 The maximum number of instructions to consider when looking for an
9865 instruction to fill a delay slot. If more than this arbitrary number of
9866 instructions are searched, the time savings from filling the delay slot
9867 are minimal, so stop searching. Increasing values mean more
9868 aggressive optimization, making the compilation time increase with probably
9869 small improvement in execution time.
9871 @item max-delay-slot-live-search
9872 When trying to fill delay slots, the maximum number of instructions to
9873 consider when searching for a block with valid live register
9874 information. Increasing this arbitrarily chosen value means more
9875 aggressive optimization, increasing the compilation time. This parameter
9876 should be removed when the delay slot code is rewritten to maintain the
9879 @item max-gcse-memory
9880 The approximate maximum amount of memory that can be allocated in
9881 order to perform the global common subexpression elimination
9882 optimization. If more memory than specified is required, the
9883 optimization is not done.
9885 @item max-gcse-insertion-ratio
9886 If the ratio of expression insertions to deletions is larger than this value
9887 for any expression, then RTL PRE inserts or removes the expression and thus
9888 leaves partially redundant computations in the instruction stream. The default value is 20.
9890 @item max-pending-list-length
9891 The maximum number of pending dependencies scheduling allows
9892 before flushing the current state and starting over. Large functions
9893 with few branches or calls can create excessively large lists which
9894 needlessly consume memory and resources.
9896 @item max-modulo-backtrack-attempts
9897 The maximum number of backtrack attempts the scheduler should make
9898 when modulo scheduling a loop. Larger values can exponentially increase
9901 @item max-inline-insns-single
9902 Several parameters control the tree inliner used in GCC@.
9903 This number sets the maximum number of instructions (counted in GCC's
9904 internal representation) in a single function that the tree inliner
9905 considers for inlining. This only affects functions declared
9906 inline and methods implemented in a class declaration (C++).
9907 The default value is 400.
9909 @item max-inline-insns-auto
9910 When you use @option{-finline-functions} (included in @option{-O3}),
9911 a lot of functions that would otherwise not be considered for inlining
9912 by the compiler are investigated. To those functions, a different
9913 (more restrictive) limit compared to functions declared inline can
9915 The default value is 40.
9917 @item inline-min-speedup
9918 When estimated performance improvement of caller + callee runtime exceeds this
9919 threshold (in percent), the function can be inlined regardless of the limit on
9920 @option{--param max-inline-insns-single} and @option{--param
9921 max-inline-insns-auto}.
9923 @item large-function-insns
9924 The limit specifying really large functions. For functions larger than this
9925 limit after inlining, inlining is constrained by
9926 @option{--param large-function-growth}. This parameter is useful primarily
9927 to avoid extreme compilation time caused by non-linear algorithms used by the
9929 The default value is 2700.
9931 @item large-function-growth
9932 Specifies maximal growth of large function caused by inlining in percents.
9933 The default value is 100 which limits large function growth to 2.0 times
9936 @item large-unit-insns
9937 The limit specifying large translation unit. Growth caused by inlining of
9938 units larger than this limit is limited by @option{--param inline-unit-growth}.
9939 For small units this might be too tight.
9940 For example, consider a unit consisting of function A
9941 that is inline and B that just calls A three times. If B is small relative to
9942 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9943 large units consisting of small inlineable functions, however, the overall unit
9944 growth limit is needed to avoid exponential explosion of code size. Thus for
9945 smaller units, the size is increased to @option{--param large-unit-insns}
9946 before applying @option{--param inline-unit-growth}. The default is 10000.
9948 @item inline-unit-growth
9949 Specifies maximal overall growth of the compilation unit caused by inlining.
9950 The default value is 20 which limits unit growth to 1.2 times the original
9951 size. Cold functions (either marked cold via an attribute or by profile
9952 feedback) are not accounted into the unit size.
9954 @item ipcp-unit-growth
9955 Specifies maximal overall growth of the compilation unit caused by
9956 interprocedural constant propagation. The default value is 10 which limits
9957 unit growth to 1.1 times the original size.
9959 @item large-stack-frame
9960 The limit specifying large stack frames. While inlining the algorithm is trying
9961 to not grow past this limit too much. The default value is 256 bytes.
9963 @item large-stack-frame-growth
9964 Specifies maximal growth of large stack frames caused by inlining in percents.
9965 The default value is 1000 which limits large stack frame growth to 11 times
9968 @item max-inline-insns-recursive
9969 @itemx max-inline-insns-recursive-auto
9970 Specifies the maximum number of instructions an out-of-line copy of a
9971 self-recursive inline
9972 function can grow into by performing recursive inlining.
9974 @option{--param max-inline-insns-recursive} applies to functions
9976 For functions not declared inline, recursive inlining
9977 happens only when @option{-finline-functions} (included in @option{-O3}) is
9978 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
9979 default value is 450.
9981 @item max-inline-recursive-depth
9982 @itemx max-inline-recursive-depth-auto
9983 Specifies the maximum recursion depth used for recursive inlining.
9985 @option{--param max-inline-recursive-depth} applies to functions
9986 declared inline. For functions not declared inline, recursive inlining
9987 happens only when @option{-finline-functions} (included in @option{-O3}) is
9988 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9991 @item min-inline-recursive-probability
9992 Recursive inlining is profitable only for function having deep recursion
9993 in average and can hurt for function having little recursion depth by
9994 increasing the prologue size or complexity of function body to other
9997 When profile feedback is available (see @option{-fprofile-generate}) the actual
9998 recursion depth can be guessed from the probability that function recurses
9999 via a given call expression. This parameter limits inlining only to call
10000 expressions whose probability exceeds the given threshold (in percents).
10001 The default value is 10.
10003 @item early-inlining-insns
10004 Specify growth that the early inliner can make. In effect it increases
10005 the amount of inlining for code having a large abstraction penalty.
10006 The default value is 14.
10008 @item max-early-inliner-iterations
10009 Limit of iterations of the early inliner. This basically bounds
10010 the number of nested indirect calls the early inliner can resolve.
10011 Deeper chains are still handled by late inlining.
10013 @item comdat-sharing-probability
10014 Probability (in percent) that C++ inline function with comdat visibility
10015 are shared across multiple compilation units. The default value is 20.
10017 @item profile-func-internal-id
10018 A parameter to control whether to use function internal id in profile
10019 database lookup. If the value is 0, the compiler uses an id that
10020 is based on function assembler name and filename, which makes old profile
10021 data more tolerant to source changes such as function reordering etc.
10022 The default value is 0.
10024 @item min-vect-loop-bound
10025 The minimum number of iterations under which loops are not vectorized
10026 when @option{-ftree-vectorize} is used. The number of iterations after
10027 vectorization needs to be greater than the value specified by this option
10028 to allow vectorization. The default value is 0.
10030 @item gcse-cost-distance-ratio
10031 Scaling factor in calculation of maximum distance an expression
10032 can be moved by GCSE optimizations. This is currently supported only in the
10033 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10034 is with simple expressions, i.e., the expressions that have cost
10035 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10036 hoisting of simple expressions. The default value is 10.
10038 @item gcse-unrestricted-cost
10039 Cost, roughly measured as the cost of a single typical machine
10040 instruction, at which GCSE optimizations do not constrain
10041 the distance an expression can travel. This is currently
10042 supported only in the code hoisting pass. The lesser the cost,
10043 the more aggressive code hoisting is. Specifying 0
10044 allows all expressions to travel unrestricted distances.
10045 The default value is 3.
10047 @item max-hoist-depth
10048 The depth of search in the dominator tree for expressions to hoist.
10049 This is used to avoid quadratic behavior in hoisting algorithm.
10050 The value of 0 does not limit on the search, but may slow down compilation
10051 of huge functions. The default value is 30.
10053 @item max-tail-merge-comparisons
10054 The maximum amount of similar bbs to compare a bb with. This is used to
10055 avoid quadratic behavior in tree tail merging. The default value is 10.
10057 @item max-tail-merge-iterations
10058 The maximum amount of iterations of the pass over the function. This is used to
10059 limit compilation time in tree tail merging. The default value is 2.
10061 @item store-merging-allow-unaligned
10062 Allow the store merging pass to introduce unaligned stores if it is legal to
10063 do so. The default value is 1.
10065 @item max-stores-to-merge
10066 The maximum number of stores to attempt to merge into wider stores in the store
10067 merging pass. The minimum value is 2 and the default is 64.
10069 @item max-unrolled-insns
10070 The maximum number of instructions that a loop may have to be unrolled.
10071 If a loop is unrolled, this parameter also determines how many times
10072 the loop code is unrolled.
10074 @item max-average-unrolled-insns
10075 The maximum number of instructions biased by probabilities of their execution
10076 that a loop may have to be unrolled. If a loop is unrolled,
10077 this parameter also determines how many times the loop code is unrolled.
10079 @item max-unroll-times
10080 The maximum number of unrollings of a single loop.
10082 @item max-peeled-insns
10083 The maximum number of instructions that a loop may have to be peeled.
10084 If a loop is peeled, this parameter also determines how many times
10085 the loop code is peeled.
10087 @item max-peel-times
10088 The maximum number of peelings of a single loop.
10090 @item max-peel-branches
10091 The maximum number of branches on the hot path through the peeled sequence.
10093 @item max-completely-peeled-insns
10094 The maximum number of insns of a completely peeled loop.
10096 @item max-completely-peel-times
10097 The maximum number of iterations of a loop to be suitable for complete peeling.
10099 @item max-completely-peel-loop-nest-depth
10100 The maximum depth of a loop nest suitable for complete peeling.
10102 @item max-unswitch-insns
10103 The maximum number of insns of an unswitched loop.
10105 @item max-unswitch-level
10106 The maximum number of branches unswitched in a single loop.
10108 @item max-loop-headers-insns
10109 The maximum number of insns in loop header duplicated by the copy loop headers
10112 @item lim-expensive
10113 The minimum cost of an expensive expression in the loop invariant motion.
10115 @item iv-consider-all-candidates-bound
10116 Bound on number of candidates for induction variables, below which
10117 all candidates are considered for each use in induction variable
10118 optimizations. If there are more candidates than this,
10119 only the most relevant ones are considered to avoid quadratic time complexity.
10121 @item iv-max-considered-uses
10122 The induction variable optimizations give up on loops that contain more
10123 induction variable uses.
10125 @item iv-always-prune-cand-set-bound
10126 If the number of candidates in the set is smaller than this value,
10127 always try to remove unnecessary ivs from the set
10128 when adding a new one.
10130 @item avg-loop-niter
10131 Average number of iterations of a loop.
10133 @item dse-max-object-size
10134 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
10135 Larger values may result in larger compilation times.
10137 @item scev-max-expr-size
10138 Bound on size of expressions used in the scalar evolutions analyzer.
10139 Large expressions slow the analyzer.
10141 @item scev-max-expr-complexity
10142 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10143 Complex expressions slow the analyzer.
10145 @item max-tree-if-conversion-phi-args
10146 Maximum number of arguments in a PHI supported by TREE if conversion
10147 unless the loop is marked with simd pragma.
10149 @item vect-max-version-for-alignment-checks
10150 The maximum number of run-time checks that can be performed when
10151 doing loop versioning for alignment in the vectorizer.
10153 @item vect-max-version-for-alias-checks
10154 The maximum number of run-time checks that can be performed when
10155 doing loop versioning for alias in the vectorizer.
10157 @item vect-max-peeling-for-alignment
10158 The maximum number of loop peels to enhance access alignment
10159 for vectorizer. Value -1 means no limit.
10161 @item max-iterations-to-track
10162 The maximum number of iterations of a loop the brute-force algorithm
10163 for analysis of the number of iterations of the loop tries to evaluate.
10165 @item hot-bb-count-ws-permille
10166 A basic block profile count is considered hot if it contributes to
10167 the given permillage (i.e. 0...1000) of the entire profiled execution.
10169 @item hot-bb-frequency-fraction
10170 Select fraction of the entry block frequency of executions of basic block in
10171 function given basic block needs to have to be considered hot.
10173 @item max-predicted-iterations
10174 The maximum number of loop iterations we predict statically. This is useful
10175 in cases where a function contains a single loop with known bound and
10176 another loop with unknown bound.
10177 The known number of iterations is predicted correctly, while
10178 the unknown number of iterations average to roughly 10. This means that the
10179 loop without bounds appears artificially cold relative to the other one.
10181 @item builtin-expect-probability
10182 Control the probability of the expression having the specified value. This
10183 parameter takes a percentage (i.e. 0 ... 100) as input.
10184 The default probability of 90 is obtained empirically.
10186 @item align-threshold
10188 Select fraction of the maximal frequency of executions of a basic block in
10189 a function to align the basic block.
10191 @item align-loop-iterations
10193 A loop expected to iterate at least the selected number of iterations is
10196 @item tracer-dynamic-coverage
10197 @itemx tracer-dynamic-coverage-feedback
10199 This value is used to limit superblock formation once the given percentage of
10200 executed instructions is covered. This limits unnecessary code size
10203 The @option{tracer-dynamic-coverage-feedback} parameter
10204 is used only when profile
10205 feedback is available. The real profiles (as opposed to statically estimated
10206 ones) are much less balanced allowing the threshold to be larger value.
10208 @item tracer-max-code-growth
10209 Stop tail duplication once code growth has reached given percentage. This is
10210 a rather artificial limit, as most of the duplicates are eliminated later in
10211 cross jumping, so it may be set to much higher values than is the desired code
10214 @item tracer-min-branch-ratio
10216 Stop reverse growth when the reverse probability of best edge is less than this
10217 threshold (in percent).
10219 @item tracer-min-branch-probability
10220 @itemx tracer-min-branch-probability-feedback
10222 Stop forward growth if the best edge has probability lower than this
10225 Similarly to @option{tracer-dynamic-coverage} two parameters are
10226 provided. @option{tracer-min-branch-probability-feedback} is used for
10227 compilation with profile feedback and @option{tracer-min-branch-probability}
10228 compilation without. The value for compilation with profile feedback
10229 needs to be more conservative (higher) in order to make tracer
10232 @item stack-clash-protection-guard-size
10233 Specify the size of the operating system provided stack guard as
10234 2 raised to @var{num} bytes. The default value is 12 (4096 bytes).
10235 Acceptable values are between 12 and 30. Higher values may reduce the
10236 number of explicit probes, but a value larger than the operating system
10237 provided guard will leave code vulnerable to stack clash style attacks.
10239 @item stack-clash-protection-probe-interval
10240 Stack clash protection involves probing stack space as it is allocated. This
10241 param controls the maximum distance between probes into the stack as 2 raised
10242 to @var{num} bytes. Acceptable values are between 10 and 16 and defaults to
10243 12. Higher values may reduce the number of explicit probes, but a value
10244 larger than the operating system provided guard will leave code vulnerable to
10245 stack clash style attacks.
10247 @item max-cse-path-length
10249 The maximum number of basic blocks on path that CSE considers.
10252 @item max-cse-insns
10253 The maximum number of instructions CSE processes before flushing.
10254 The default is 1000.
10256 @item ggc-min-expand
10258 GCC uses a garbage collector to manage its own memory allocation. This
10259 parameter specifies the minimum percentage by which the garbage
10260 collector's heap should be allowed to expand between collections.
10261 Tuning this may improve compilation speed; it has no effect on code
10264 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10265 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10266 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10267 GCC is not able to calculate RAM on a particular platform, the lower
10268 bound of 30% is used. Setting this parameter and
10269 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10270 every opportunity. This is extremely slow, but can be useful for
10273 @item ggc-min-heapsize
10275 Minimum size of the garbage collector's heap before it begins bothering
10276 to collect garbage. The first collection occurs after the heap expands
10277 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10278 tuning this may improve compilation speed, and has no effect on code
10281 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10282 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10283 with a lower bound of 4096 (four megabytes) and an upper bound of
10284 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10285 particular platform, the lower bound is used. Setting this parameter
10286 very large effectively disables garbage collection. Setting this
10287 parameter and @option{ggc-min-expand} to zero causes a full collection
10288 to occur at every opportunity.
10290 @item max-reload-search-insns
10291 The maximum number of instruction reload should look backward for equivalent
10292 register. Increasing values mean more aggressive optimization, making the
10293 compilation time increase with probably slightly better performance.
10294 The default value is 100.
10296 @item max-cselib-memory-locations
10297 The maximum number of memory locations cselib should take into account.
10298 Increasing values mean more aggressive optimization, making the compilation time
10299 increase with probably slightly better performance. The default value is 500.
10301 @item max-sched-ready-insns
10302 The maximum number of instructions ready to be issued the scheduler should
10303 consider at any given time during the first scheduling pass. Increasing
10304 values mean more thorough searches, making the compilation time increase
10305 with probably little benefit. The default value is 100.
10307 @item max-sched-region-blocks
10308 The maximum number of blocks in a region to be considered for
10309 interblock scheduling. The default value is 10.
10311 @item max-pipeline-region-blocks
10312 The maximum number of blocks in a region to be considered for
10313 pipelining in the selective scheduler. The default value is 15.
10315 @item max-sched-region-insns
10316 The maximum number of insns in a region to be considered for
10317 interblock scheduling. The default value is 100.
10319 @item max-pipeline-region-insns
10320 The maximum number of insns in a region to be considered for
10321 pipelining in the selective scheduler. The default value is 200.
10323 @item min-spec-prob
10324 The minimum probability (in percents) of reaching a source block
10325 for interblock speculative scheduling. The default value is 40.
10327 @item max-sched-extend-regions-iters
10328 The maximum number of iterations through CFG to extend regions.
10329 A value of 0 (the default) disables region extensions.
10331 @item max-sched-insn-conflict-delay
10332 The maximum conflict delay for an insn to be considered for speculative motion.
10333 The default value is 3.
10335 @item sched-spec-prob-cutoff
10336 The minimal probability of speculation success (in percents), so that
10337 speculative insns are scheduled.
10338 The default value is 40.
10340 @item sched-state-edge-prob-cutoff
10341 The minimum probability an edge must have for the scheduler to save its
10343 The default value is 10.
10345 @item sched-mem-true-dep-cost
10346 Minimal distance (in CPU cycles) between store and load targeting same
10347 memory locations. The default value is 1.
10349 @item selsched-max-lookahead
10350 The maximum size of the lookahead window of selective scheduling. It is a
10351 depth of search for available instructions.
10352 The default value is 50.
10354 @item selsched-max-sched-times
10355 The maximum number of times that an instruction is scheduled during
10356 selective scheduling. This is the limit on the number of iterations
10357 through which the instruction may be pipelined. The default value is 2.
10359 @item selsched-insns-to-rename
10360 The maximum number of best instructions in the ready list that are considered
10361 for renaming in the selective scheduler. The default value is 2.
10364 The minimum value of stage count that swing modulo scheduler
10365 generates. The default value is 2.
10367 @item max-last-value-rtl
10368 The maximum size measured as number of RTLs that can be recorded in an expression
10369 in combiner for a pseudo register as last known value of that register. The default
10372 @item max-combine-insns
10373 The maximum number of instructions the RTL combiner tries to combine.
10374 The default value is 2 at @option{-Og} and 4 otherwise.
10376 @item integer-share-limit
10377 Small integer constants can use a shared data structure, reducing the
10378 compiler's memory usage and increasing its speed. This sets the maximum
10379 value of a shared integer constant. The default value is 256.
10381 @item ssp-buffer-size
10382 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10383 protection when @option{-fstack-protection} is used.
10385 @item min-size-for-stack-sharing
10386 The minimum size of variables taking part in stack slot sharing when not
10387 optimizing. The default value is 32.
10389 @item max-jump-thread-duplication-stmts
10390 Maximum number of statements allowed in a block that needs to be
10391 duplicated when threading jumps.
10393 @item max-fields-for-field-sensitive
10394 Maximum number of fields in a structure treated in
10395 a field sensitive manner during pointer analysis. The default is zero
10396 for @option{-O0} and @option{-O1},
10397 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10399 @item prefetch-latency
10400 Estimate on average number of instructions that are executed before
10401 prefetch finishes. The distance prefetched ahead is proportional
10402 to this constant. Increasing this number may also lead to less
10403 streams being prefetched (see @option{simultaneous-prefetches}).
10405 @item simultaneous-prefetches
10406 Maximum number of prefetches that can run at the same time.
10408 @item l1-cache-line-size
10409 The size of cache line in L1 cache, in bytes.
10411 @item l1-cache-size
10412 The size of L1 cache, in kilobytes.
10414 @item l2-cache-size
10415 The size of L2 cache, in kilobytes.
10417 @item min-insn-to-prefetch-ratio
10418 The minimum ratio between the number of instructions and the
10419 number of prefetches to enable prefetching in a loop.
10421 @item prefetch-min-insn-to-mem-ratio
10422 The minimum ratio between the number of instructions and the
10423 number of memory references to enable prefetching in a loop.
10425 @item use-canonical-types
10426 Whether the compiler should use the ``canonical'' type system. By
10427 default, this should always be 1, which uses a more efficient internal
10428 mechanism for comparing types in C++ and Objective-C++. However, if
10429 bugs in the canonical type system are causing compilation failures,
10430 set this value to 0 to disable canonical types.
10432 @item switch-conversion-max-branch-ratio
10433 Switch initialization conversion refuses to create arrays that are
10434 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10435 branches in the switch.
10437 @item max-partial-antic-length
10438 Maximum length of the partial antic set computed during the tree
10439 partial redundancy elimination optimization (@option{-ftree-pre}) when
10440 optimizing at @option{-O3} and above. For some sorts of source code
10441 the enhanced partial redundancy elimination optimization can run away,
10442 consuming all of the memory available on the host machine. This
10443 parameter sets a limit on the length of the sets that are computed,
10444 which prevents the runaway behavior. Setting a value of 0 for
10445 this parameter allows an unlimited set length.
10447 @item sccvn-max-scc-size
10448 Maximum size of a strongly connected component (SCC) during SCCVN
10449 processing. If this limit is hit, SCCVN processing for the whole
10450 function is not done and optimizations depending on it are
10451 disabled. The default maximum SCC size is 10000.
10453 @item sccvn-max-alias-queries-per-access
10454 Maximum number of alias-oracle queries we perform when looking for
10455 redundancies for loads and stores. If this limit is hit the search
10456 is aborted and the load or store is not considered redundant. The
10457 number of queries is algorithmically limited to the number of
10458 stores on all paths from the load to the function entry.
10459 The default maximum number of queries is 1000.
10461 @item ira-max-loops-num
10462 IRA uses regional register allocation by default. If a function
10463 contains more loops than the number given by this parameter, only at most
10464 the given number of the most frequently-executed loops form regions
10465 for regional register allocation. The default value of the
10468 @item ira-max-conflict-table-size
10469 Although IRA uses a sophisticated algorithm to compress the conflict
10470 table, the table can still require excessive amounts of memory for
10471 huge functions. If the conflict table for a function could be more
10472 than the size in MB given by this parameter, the register allocator
10473 instead uses a faster, simpler, and lower-quality
10474 algorithm that does not require building a pseudo-register conflict table.
10475 The default value of the parameter is 2000.
10477 @item ira-loop-reserved-regs
10478 IRA can be used to evaluate more accurate register pressure in loops
10479 for decisions to move loop invariants (see @option{-O3}). The number
10480 of available registers reserved for some other purposes is given
10481 by this parameter. The default value of the parameter is 2, which is
10482 the minimal number of registers needed by typical instructions.
10483 This value is the best found from numerous experiments.
10485 @item lra-inheritance-ebb-probability-cutoff
10486 LRA tries to reuse values reloaded in registers in subsequent insns.
10487 This optimization is called inheritance. EBB is used as a region to
10488 do this optimization. The parameter defines a minimal fall-through
10489 edge probability in percentage used to add BB to inheritance EBB in
10490 LRA. The default value of the parameter is 40. The value was chosen
10491 from numerous runs of SPEC2000 on x86-64.
10493 @item loop-invariant-max-bbs-in-loop
10494 Loop invariant motion can be very expensive, both in compilation time and
10495 in amount of needed compile-time memory, with very large loops. Loops
10496 with more basic blocks than this parameter won't have loop invariant
10497 motion optimization performed on them. The default value of the
10498 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10500 @item loop-max-datarefs-for-datadeps
10501 Building data dependencies is expensive for very large loops. This
10502 parameter limits the number of data references in loops that are
10503 considered for data dependence analysis. These large loops are no
10504 handled by the optimizations using loop data dependencies.
10505 The default value is 1000.
10507 @item max-vartrack-size
10508 Sets a maximum number of hash table slots to use during variable
10509 tracking dataflow analysis of any function. If this limit is exceeded
10510 with variable tracking at assignments enabled, analysis for that
10511 function is retried without it, after removing all debug insns from
10512 the function. If the limit is exceeded even without debug insns, var
10513 tracking analysis is completely disabled for the function. Setting
10514 the parameter to zero makes it unlimited.
10516 @item max-vartrack-expr-depth
10517 Sets a maximum number of recursion levels when attempting to map
10518 variable names or debug temporaries to value expressions. This trades
10519 compilation time for more complete debug information. If this is set too
10520 low, value expressions that are available and could be represented in
10521 debug information may end up not being used; setting this higher may
10522 enable the compiler to find more complex debug expressions, but compile
10523 time and memory use may grow. The default is 12.
10525 @item min-nondebug-insn-uid
10526 Use uids starting at this parameter for nondebug insns. The range below
10527 the parameter is reserved exclusively for debug insns created by
10528 @option{-fvar-tracking-assignments}, but debug insns may get
10529 (non-overlapping) uids above it if the reserved range is exhausted.
10531 @item ipa-sra-ptr-growth-factor
10532 IPA-SRA replaces a pointer to an aggregate with one or more new
10533 parameters only when their cumulative size is less or equal to
10534 @option{ipa-sra-ptr-growth-factor} times the size of the original
10537 @item sra-max-scalarization-size-Ospeed
10538 @item sra-max-scalarization-size-Osize
10539 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10540 replace scalar parts of aggregates with uses of independent scalar
10541 variables. These parameters control the maximum size, in storage units,
10542 of aggregate which is considered for replacement when compiling for
10544 (@option{sra-max-scalarization-size-Ospeed}) or size
10545 (@option{sra-max-scalarization-size-Osize}) respectively.
10547 @item tm-max-aggregate-size
10548 When making copies of thread-local variables in a transaction, this
10549 parameter specifies the size in bytes after which variables are
10550 saved with the logging functions as opposed to save/restore code
10551 sequence pairs. This option only applies when using
10554 @item graphite-max-nb-scop-params
10555 To avoid exponential effects in the Graphite loop transforms, the
10556 number of parameters in a Static Control Part (SCoP) is bounded. The
10557 default value is 10 parameters, a value of zero can be used to lift
10558 the bound. A variable whose value is unknown at compilation time and
10559 defined outside a SCoP is a parameter of the SCoP.
10561 @item loop-block-tile-size
10562 Loop blocking or strip mining transforms, enabled with
10563 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10564 loop in the loop nest by a given number of iterations. The strip
10565 length can be changed using the @option{loop-block-tile-size}
10566 parameter. The default value is 51 iterations.
10568 @item loop-unroll-jam-size
10569 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10570 default value is 4.
10572 @item loop-unroll-jam-depth
10573 Specify the dimension to be unrolled (counting from the most inner loop)
10574 for the @option{-floop-unroll-and-jam}. The default value is 2.
10576 @item ipa-cp-value-list-size
10577 IPA-CP attempts to track all possible values and types passed to a function's
10578 parameter in order to propagate them and perform devirtualization.
10579 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10580 stores per one formal parameter of a function.
10582 @item ipa-cp-eval-threshold
10583 IPA-CP calculates its own score of cloning profitability heuristics
10584 and performs those cloning opportunities with scores that exceed
10585 @option{ipa-cp-eval-threshold}.
10587 @item ipa-cp-recursion-penalty
10588 Percentage penalty the recursive functions will receive when they
10589 are evaluated for cloning.
10591 @item ipa-cp-single-call-penalty
10592 Percentage penalty functions containing a single call to another
10593 function will receive when they are evaluated for cloning.
10596 @item ipa-max-agg-items
10597 IPA-CP is also capable to propagate a number of scalar values passed
10598 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10599 number of such values per one parameter.
10601 @item ipa-cp-loop-hint-bonus
10602 When IPA-CP determines that a cloning candidate would make the number
10603 of iterations of a loop known, it adds a bonus of
10604 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10607 @item ipa-cp-array-index-hint-bonus
10608 When IPA-CP determines that a cloning candidate would make the index of
10609 an array access known, it adds a bonus of
10610 @option{ipa-cp-array-index-hint-bonus} to the profitability
10611 score of the candidate.
10613 @item ipa-max-aa-steps
10614 During its analysis of function bodies, IPA-CP employs alias analysis
10615 in order to track values pointed to by function parameters. In order
10616 not spend too much time analyzing huge functions, it gives up and
10617 consider all memory clobbered after examining
10618 @option{ipa-max-aa-steps} statements modifying memory.
10620 @item lto-partitions
10621 Specify desired number of partitions produced during WHOPR compilation.
10622 The number of partitions should exceed the number of CPUs used for compilation.
10623 The default value is 32.
10625 @item lto-min-partition
10626 Size of minimal partition for WHOPR (in estimated instructions).
10627 This prevents expenses of splitting very small programs into too many
10630 @item lto-max-partition
10631 Size of max partition for WHOPR (in estimated instructions).
10632 to provide an upper bound for individual size of partition.
10633 Meant to be used only with balanced partitioning.
10635 @item cxx-max-namespaces-for-diagnostic-help
10636 The maximum number of namespaces to consult for suggestions when C++
10637 name lookup fails for an identifier. The default is 1000.
10639 @item sink-frequency-threshold
10640 The maximum relative execution frequency (in percents) of the target block
10641 relative to a statement's original block to allow statement sinking of a
10642 statement. Larger numbers result in more aggressive statement sinking.
10643 The default value is 75. A small positive adjustment is applied for
10644 statements with memory operands as those are even more profitable so sink.
10646 @item max-stores-to-sink
10647 The maximum number of conditional store pairs that can be sunk. Set to 0
10648 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10649 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10651 @item allow-store-data-races
10652 Allow optimizers to introduce new data races on stores.
10653 Set to 1 to allow, otherwise to 0. This option is enabled by default
10654 at optimization level @option{-Ofast}.
10656 @item case-values-threshold
10657 The smallest number of different values for which it is best to use a
10658 jump-table instead of a tree of conditional branches. If the value is
10659 0, use the default for the machine. The default is 0.
10661 @item tree-reassoc-width
10662 Set the maximum number of instructions executed in parallel in
10663 reassociated tree. This parameter overrides target dependent
10664 heuristics used by default if has non zero value.
10666 @item sched-pressure-algorithm
10667 Choose between the two available implementations of
10668 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10669 and is the more likely to prevent instructions from being reordered.
10670 Algorithm 2 was designed to be a compromise between the relatively
10671 conservative approach taken by algorithm 1 and the rather aggressive
10672 approach taken by the default scheduler. It relies more heavily on
10673 having a regular register file and accurate register pressure classes.
10674 See @file{haifa-sched.c} in the GCC sources for more details.
10676 The default choice depends on the target.
10678 @item max-slsr-cand-scan
10679 Set the maximum number of existing candidates that are considered when
10680 seeking a basis for a new straight-line strength reduction candidate.
10683 Enable buffer overflow detection for global objects. This kind
10684 of protection is enabled by default if you are using
10685 @option{-fsanitize=address} option.
10686 To disable global objects protection use @option{--param asan-globals=0}.
10689 Enable buffer overflow detection for stack objects. This kind of
10690 protection is enabled by default when using @option{-fsanitize=address}.
10691 To disable stack protection use @option{--param asan-stack=0} option.
10693 @item asan-instrument-reads
10694 Enable buffer overflow detection for memory reads. This kind of
10695 protection is enabled by default when using @option{-fsanitize=address}.
10696 To disable memory reads protection use
10697 @option{--param asan-instrument-reads=0}.
10699 @item asan-instrument-writes
10700 Enable buffer overflow detection for memory writes. This kind of
10701 protection is enabled by default when using @option{-fsanitize=address}.
10702 To disable memory writes protection use
10703 @option{--param asan-instrument-writes=0} option.
10705 @item asan-memintrin
10706 Enable detection for built-in functions. This kind of protection
10707 is enabled by default when using @option{-fsanitize=address}.
10708 To disable built-in functions protection use
10709 @option{--param asan-memintrin=0}.
10711 @item asan-use-after-return
10712 Enable detection of use-after-return. This kind of protection
10713 is enabled by default when using the @option{-fsanitize=address} option.
10714 To disable it use @option{--param asan-use-after-return=0}.
10716 Note: By default the check is disabled at run time. To enable it,
10717 add @code{detect_stack_use_after_return=1} to the environment variable
10718 @env{ASAN_OPTIONS}.
10720 @item asan-instrumentation-with-call-threshold
10721 If number of memory accesses in function being instrumented
10722 is greater or equal to this number, use callbacks instead of inline checks.
10723 E.g. to disable inline code use
10724 @option{--param asan-instrumentation-with-call-threshold=0}.
10726 @item use-after-scope-direct-emission-threshold
10727 If the size of a local variable in bytes is smaller or equal to this
10728 number, directly poison (or unpoison) shadow memory instead of using
10729 run-time callbacks. The default value is 256.
10731 @item chkp-max-ctor-size
10732 Static constructors generated by Pointer Bounds Checker may become very
10733 large and significantly increase compile time at optimization level
10734 @option{-O1} and higher. This parameter is a maximum number of statements
10735 in a single generated constructor. Default value is 5000.
10737 @item max-fsm-thread-path-insns
10738 Maximum number of instructions to copy when duplicating blocks on a
10739 finite state automaton jump thread path. The default is 100.
10741 @item max-fsm-thread-length
10742 Maximum number of basic blocks on a finite state automaton jump thread
10743 path. The default is 10.
10745 @item max-fsm-thread-paths
10746 Maximum number of new jump thread paths to create for a finite state
10747 automaton. The default is 50.
10749 @item parloops-chunk-size
10750 Chunk size of omp schedule for loops parallelized by parloops. The default
10753 @item parloops-schedule
10754 Schedule type of omp schedule for loops parallelized by parloops (static,
10755 dynamic, guided, auto, runtime). The default is static.
10757 @item max-ssa-name-query-depth
10758 Maximum depth of recursion when querying properties of SSA names in things
10759 like fold routines. One level of recursion corresponds to following a
10762 @item hsa-gen-debug-stores
10763 Enable emission of special debug stores within HSA kernels which are
10764 then read and reported by libgomp plugin. Generation of these stores
10765 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10768 @item max-speculative-devirt-maydefs
10769 The maximum number of may-defs we analyze when looking for a must-def
10770 specifying the dynamic type of an object that invokes a virtual call
10771 we may be able to devirtualize speculatively.
10773 @item max-vrp-switch-assertions
10774 The maximum number of assertions to add along the default edge of a switch
10775 statement during VRP. The default is 10.
10779 @node Instrumentation Options
10780 @section Program Instrumentation Options
10781 @cindex instrumentation options
10782 @cindex program instrumentation options
10783 @cindex run-time error checking options
10784 @cindex profiling options
10785 @cindex options, program instrumentation
10786 @cindex options, run-time error checking
10787 @cindex options, profiling
10789 GCC supports a number of command-line options that control adding
10790 run-time instrumentation to the code it normally generates.
10791 For example, one purpose of instrumentation is collect profiling
10792 statistics for use in finding program hot spots, code coverage
10793 analysis, or profile-guided optimizations.
10794 Another class of program instrumentation is adding run-time checking
10795 to detect programming errors like invalid pointer
10796 dereferences or out-of-bounds array accesses, as well as deliberately
10797 hostile attacks such as stack smashing or C++ vtable hijacking.
10798 There is also a general hook which can be used to implement other
10799 forms of tracing or function-level instrumentation for debug or
10800 program analysis purposes.
10803 @cindex @command{prof}
10806 Generate extra code to write profile information suitable for the
10807 analysis program @command{prof}. You must use this option when compiling
10808 the source files you want data about, and you must also use it when
10811 @cindex @command{gprof}
10814 Generate extra code to write profile information suitable for the
10815 analysis program @command{gprof}. You must use this option when compiling
10816 the source files you want data about, and you must also use it when
10819 @item -fprofile-arcs
10820 @opindex fprofile-arcs
10821 Add code so that program flow @dfn{arcs} are instrumented. During
10822 execution the program records how many times each branch and call is
10823 executed and how many times it is taken or returns. On targets that support
10824 constructors with priority support, profiling properly handles constructors,
10825 destructors and C++ constructors (and destructors) of classes which are used
10826 as a type of a global variable.
10829 program exits it saves this data to a file called
10830 @file{@var{auxname}.gcda} for each source file. The data may be used for
10831 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10832 test coverage analysis (@option{-ftest-coverage}). Each object file's
10833 @var{auxname} is generated from the name of the output file, if
10834 explicitly specified and it is not the final executable, otherwise it is
10835 the basename of the source file. In both cases any suffix is removed
10836 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10837 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10838 @xref{Cross-profiling}.
10840 @cindex @command{gcov}
10844 This option is used to compile and link code instrumented for coverage
10845 analysis. The option is a synonym for @option{-fprofile-arcs}
10846 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10847 linking). See the documentation for those options for more details.
10852 Compile the source files with @option{-fprofile-arcs} plus optimization
10853 and code generation options. For test coverage analysis, use the
10854 additional @option{-ftest-coverage} option. You do not need to profile
10855 every source file in a program.
10858 Compile the source files additionally with @option{-fprofile-abs-path}
10859 to create absolute path names in the @file{.gcno} files. This allows
10860 @command{gcov} to find the correct sources in projects where compilations
10861 occur with different working directories.
10864 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10865 (the latter implies the former).
10868 Run the program on a representative workload to generate the arc profile
10869 information. This may be repeated any number of times. You can run
10870 concurrent instances of your program, and provided that the file system
10871 supports locking, the data files will be correctly updated. Also
10872 @code{fork} calls are detected and correctly handled (double counting
10876 For profile-directed optimizations, compile the source files again with
10877 the same optimization and code generation options plus
10878 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10879 Control Optimization}).
10882 For test coverage analysis, use @command{gcov} to produce human readable
10883 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10884 @command{gcov} documentation for further information.
10888 With @option{-fprofile-arcs}, for each function of your program GCC
10889 creates a program flow graph, then finds a spanning tree for the graph.
10890 Only arcs that are not on the spanning tree have to be instrumented: the
10891 compiler adds code to count the number of times that these arcs are
10892 executed. When an arc is the only exit or only entrance to a block, the
10893 instrumentation code can be added to the block; otherwise, a new basic
10894 block must be created to hold the instrumentation code.
10897 @item -ftest-coverage
10898 @opindex ftest-coverage
10899 Produce a notes file that the @command{gcov} code-coverage utility
10900 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10901 show program coverage. Each source file's note file is called
10902 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10903 above for a description of @var{auxname} and instructions on how to
10904 generate test coverage data. Coverage data matches the source files
10905 more closely if you do not optimize.
10907 @item -fprofile-abs-path
10908 @opindex fprofile-abs-path
10909 Automatically convert relative source file names to absolute path names
10910 in the @file{.gcno} files. This allows @command{gcov} to find the correct
10911 sources in projects where compilations occur with different working
10914 @item -fprofile-dir=@var{path}
10915 @opindex fprofile-dir
10917 Set the directory to search for the profile data files in to @var{path}.
10918 This option affects only the profile data generated by
10919 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10920 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10921 and its related options. Both absolute and relative paths can be used.
10922 By default, GCC uses the current directory as @var{path}, thus the
10923 profile data file appears in the same directory as the object file.
10925 @item -fprofile-generate
10926 @itemx -fprofile-generate=@var{path}
10927 @opindex fprofile-generate
10929 Enable options usually used for instrumenting application to produce
10930 profile useful for later recompilation with profile feedback based
10931 optimization. You must use @option{-fprofile-generate} both when
10932 compiling and when linking your program.
10934 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10936 If @var{path} is specified, GCC looks at the @var{path} to find
10937 the profile feedback data files. See @option{-fprofile-dir}.
10939 To optimize the program based on the collected profile information, use
10940 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10942 @item -fprofile-update=@var{method}
10943 @opindex fprofile-update
10945 Alter the update method for an application instrumented for profile
10946 feedback based optimization. The @var{method} argument should be one of
10947 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
10948 The first one is useful for single-threaded applications,
10949 while the second one prevents profile corruption by emitting thread-safe code.
10951 @strong{Warning:} When an application does not properly join all threads
10952 (or creates an detached thread), a profile file can be still corrupted.
10954 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
10955 when supported by a target, or to @samp{single} otherwise. The GCC driver
10956 automatically selects @samp{prefer-atomic} when @option{-pthread}
10957 is present in the command line.
10959 @item -fsanitize=address
10960 @opindex fsanitize=address
10961 Enable AddressSanitizer, a fast memory error detector.
10962 Memory access instructions are instrumented to detect
10963 out-of-bounds and use-after-free bugs.
10964 The option enables @option{-fsanitize-address-use-after-scope}.
10965 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
10966 more details. The run-time behavior can be influenced using the
10967 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
10968 the available options are shown at startup of the instrumented program. See
10969 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
10970 for a list of supported options.
10971 The option cannot be combined with @option{-fsanitize=thread}
10972 and/or @option{-fcheck-pointer-bounds}.
10974 @item -fsanitize=kernel-address
10975 @opindex fsanitize=kernel-address
10976 Enable AddressSanitizer for Linux kernel.
10977 See @uref{https://github.com/google/kasan/wiki} for more details.
10978 The option cannot be combined with @option{-fcheck-pointer-bounds}.
10980 @item -fsanitize=thread
10981 @opindex fsanitize=thread
10982 Enable ThreadSanitizer, a fast data race detector.
10983 Memory access instructions are instrumented to detect
10984 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
10985 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
10986 environment variable; see
10987 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
10989 The option cannot be combined with @option{-fsanitize=address},
10990 @option{-fsanitize=leak} and/or @option{-fcheck-pointer-bounds}.
10992 Note that sanitized atomic builtins cannot throw exceptions when
10993 operating on invalid memory addresses with non-call exceptions
10994 (@option{-fnon-call-exceptions}).
10996 @item -fsanitize=leak
10997 @opindex fsanitize=leak
10998 Enable LeakSanitizer, a memory leak detector.
10999 This option only matters for linking of executables and
11000 the executable is linked against a library that overrides @code{malloc}
11001 and other allocator functions. See
11002 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
11003 details. The run-time behavior can be influenced using the
11004 @env{LSAN_OPTIONS} environment variable.
11005 The option cannot be combined with @option{-fsanitize=thread}.
11007 @item -fsanitize=undefined
11008 @opindex fsanitize=undefined
11009 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
11010 Various computations are instrumented to detect undefined behavior
11011 at runtime. Current suboptions are:
11015 @item -fsanitize=shift
11016 @opindex fsanitize=shift
11017 This option enables checking that the result of a shift operation is
11018 not undefined. Note that what exactly is considered undefined differs
11019 slightly between C and C++, as well as between ISO C90 and C99, etc.
11020 This option has two suboptions, @option{-fsanitize=shift-base} and
11021 @option{-fsanitize=shift-exponent}.
11023 @item -fsanitize=shift-exponent
11024 @opindex fsanitize=shift-exponent
11025 This option enables checking that the second argument of a shift operation
11026 is not negative and is smaller than the precision of the promoted first
11029 @item -fsanitize=shift-base
11030 @opindex fsanitize=shift-base
11031 If the second argument of a shift operation is within range, check that the
11032 result of a shift operation is not undefined. Note that what exactly is
11033 considered undefined differs slightly between C and C++, as well as between
11034 ISO C90 and C99, etc.
11036 @item -fsanitize=integer-divide-by-zero
11037 @opindex fsanitize=integer-divide-by-zero
11038 Detect integer division by zero as well as @code{INT_MIN / -1} division.
11040 @item -fsanitize=unreachable
11041 @opindex fsanitize=unreachable
11042 With this option, the compiler turns the @code{__builtin_unreachable}
11043 call into a diagnostics message call instead. When reaching the
11044 @code{__builtin_unreachable} call, the behavior is undefined.
11046 @item -fsanitize=vla-bound
11047 @opindex fsanitize=vla-bound
11048 This option instructs the compiler to check that the size of a variable
11049 length array is positive.
11051 @item -fsanitize=null
11052 @opindex fsanitize=null
11053 This option enables pointer checking. Particularly, the application
11054 built with this option turned on will issue an error message when it
11055 tries to dereference a NULL pointer, or if a reference (possibly an
11056 rvalue reference) is bound to a NULL pointer, or if a method is invoked
11057 on an object pointed by a NULL pointer.
11059 @item -fsanitize=return
11060 @opindex fsanitize=return
11061 This option enables return statement checking. Programs
11062 built with this option turned on will issue an error message
11063 when the end of a non-void function is reached without actually
11064 returning a value. This option works in C++ only.
11066 @item -fsanitize=signed-integer-overflow
11067 @opindex fsanitize=signed-integer-overflow
11068 This option enables signed integer overflow checking. We check that
11069 the result of @code{+}, @code{*}, and both unary and binary @code{-}
11070 does not overflow in the signed arithmetics. Note, integer promotion
11071 rules must be taken into account. That is, the following is not an
11074 signed char a = SCHAR_MAX;
11078 @item -fsanitize=bounds
11079 @opindex fsanitize=bounds
11080 This option enables instrumentation of array bounds. Various out of bounds
11081 accesses are detected. Flexible array members, flexible array member-like
11082 arrays, and initializers of variables with static storage are not instrumented.
11083 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11085 @item -fsanitize=bounds-strict
11086 @opindex fsanitize=bounds-strict
11087 This option enables strict instrumentation of array bounds. Most out of bounds
11088 accesses are detected, including flexible array members and flexible array
11089 member-like arrays. Initializers of variables with static storage are not
11090 instrumented. The option cannot be combined
11091 with @option{-fcheck-pointer-bounds}.
11093 @item -fsanitize=alignment
11094 @opindex fsanitize=alignment
11096 This option enables checking of alignment of pointers when they are
11097 dereferenced, or when a reference is bound to insufficiently aligned target,
11098 or when a method or constructor is invoked on insufficiently aligned object.
11100 @item -fsanitize=object-size
11101 @opindex fsanitize=object-size
11102 This option enables instrumentation of memory references using the
11103 @code{__builtin_object_size} function. Various out of bounds pointer
11104 accesses are detected.
11106 @item -fsanitize=float-divide-by-zero
11107 @opindex fsanitize=float-divide-by-zero
11108 Detect floating-point division by zero. Unlike other similar options,
11109 @option{-fsanitize=float-divide-by-zero} is not enabled by
11110 @option{-fsanitize=undefined}, since floating-point division by zero can
11111 be a legitimate way of obtaining infinities and NaNs.
11113 @item -fsanitize=float-cast-overflow
11114 @opindex fsanitize=float-cast-overflow
11115 This option enables floating-point type to integer conversion checking.
11116 We check that the result of the conversion does not overflow.
11117 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
11118 not enabled by @option{-fsanitize=undefined}.
11119 This option does not work well with @code{FE_INVALID} exceptions enabled.
11121 @item -fsanitize=nonnull-attribute
11122 @opindex fsanitize=nonnull-attribute
11124 This option enables instrumentation of calls, checking whether null values
11125 are not passed to arguments marked as requiring a non-null value by the
11126 @code{nonnull} function attribute.
11128 @item -fsanitize=returns-nonnull-attribute
11129 @opindex fsanitize=returns-nonnull-attribute
11131 This option enables instrumentation of return statements in functions
11132 marked with @code{returns_nonnull} function attribute, to detect returning
11133 of null values from such functions.
11135 @item -fsanitize=bool
11136 @opindex fsanitize=bool
11138 This option enables instrumentation of loads from bool. If a value other
11139 than 0/1 is loaded, a run-time error is issued.
11141 @item -fsanitize=enum
11142 @opindex fsanitize=enum
11144 This option enables instrumentation of loads from an enum type. If
11145 a value outside the range of values for the enum type is loaded,
11146 a run-time error is issued.
11148 @item -fsanitize=vptr
11149 @opindex fsanitize=vptr
11151 This option enables instrumentation of C++ member function calls, member
11152 accesses and some conversions between pointers to base and derived classes,
11153 to verify the referenced object has the correct dynamic type.
11155 @item -fsanitize=pointer-overflow
11156 @opindex fsanitize=pointer-overflow
11158 This option enables instrumentation of pointer arithmetics. If the pointer
11159 arithmetics overflows, a run-time error is issued.
11161 @item -fsanitize=builtin
11162 @opindex fsanitize=builtin
11164 This option enables instrumentation of arguments to selected builtin
11165 functions. If an invalid value is passed to such arguments, a run-time
11166 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
11167 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
11172 While @option{-ftrapv} causes traps for signed overflows to be emitted,
11173 @option{-fsanitize=undefined} gives a diagnostic message.
11174 This currently works only for the C family of languages.
11176 @item -fno-sanitize=all
11177 @opindex fno-sanitize=all
11179 This option disables all previously enabled sanitizers.
11180 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
11183 @item -fasan-shadow-offset=@var{number}
11184 @opindex fasan-shadow-offset
11185 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
11186 It is useful for experimenting with different shadow memory layouts in
11187 Kernel AddressSanitizer.
11189 @item -fsanitize-sections=@var{s1},@var{s2},...
11190 @opindex fsanitize-sections
11191 Sanitize global variables in selected user-defined sections. @var{si} may
11194 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
11195 @opindex fsanitize-recover
11196 @opindex fno-sanitize-recover
11197 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
11198 mentioned in comma-separated list of @var{opts}. Enabling this option
11199 for a sanitizer component causes it to attempt to continue
11200 running the program as if no error happened. This means multiple
11201 runtime errors can be reported in a single program run, and the exit
11202 code of the program may indicate success even when errors
11203 have been reported. The @option{-fno-sanitize-recover=} option
11204 can be used to alter
11205 this behavior: only the first detected error is reported
11206 and program then exits with a non-zero exit code.
11208 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
11209 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
11210 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
11211 @option{-fsanitize=bounds-strict},
11212 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
11213 For these sanitizers error recovery is turned on by default,
11214 except @option{-fsanitize=address}, for which this feature is experimental.
11215 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
11216 accepted, the former enables recovery for all sanitizers that support it,
11217 the latter disables recovery for all sanitizers that support it.
11219 Even if a recovery mode is turned on the compiler side, it needs to be also
11220 enabled on the runtime library side, otherwise the failures are still fatal.
11221 The runtime library defaults to @code{halt_on_error=0} for
11222 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
11223 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
11224 setting the @code{halt_on_error} flag in the corresponding environment variable.
11226 Syntax without an explicit @var{opts} parameter is deprecated. It is
11227 equivalent to specifying an @var{opts} list of:
11230 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
11233 @item -fsanitize-address-use-after-scope
11234 @opindex fsanitize-address-use-after-scope
11235 Enable sanitization of local variables to detect use-after-scope bugs.
11236 The option sets @option{-fstack-reuse} to @samp{none}.
11238 @item -fsanitize-undefined-trap-on-error
11239 @opindex fsanitize-undefined-trap-on-error
11240 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
11241 report undefined behavior using @code{__builtin_trap} rather than
11242 a @code{libubsan} library routine. The advantage of this is that the
11243 @code{libubsan} library is not needed and is not linked in, so this
11244 is usable even in freestanding environments.
11246 @item -fsanitize-coverage=trace-pc
11247 @opindex fsanitize-coverage=trace-pc
11248 Enable coverage-guided fuzzing code instrumentation.
11249 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
11251 @item -fsanitize-coverage=trace-cmp
11252 @opindex fsanitize-coverage=trace-cmp
11253 Enable dataflow guided fuzzing code instrumentation.
11254 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
11255 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
11256 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
11257 variable or @code{__sanitizer_cov_trace_const_cmp1},
11258 @code{__sanitizer_cov_trace_const_cmp2},
11259 @code{__sanitizer_cov_trace_const_cmp4} or
11260 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
11261 operand constant, @code{__sanitizer_cov_trace_cmpf} or
11262 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
11263 @code{__sanitizer_cov_trace_switch} for switch statements.
11265 @item -fbounds-check
11266 @opindex fbounds-check
11267 For front ends that support it, generate additional code to check that
11268 indices used to access arrays are within the declared range. This is
11269 currently only supported by the Fortran front end, where this option
11272 @item -fcheck-pointer-bounds
11273 @opindex fcheck-pointer-bounds
11274 @opindex fno-check-pointer-bounds
11275 @cindex Pointer Bounds Checker options
11276 Enable Pointer Bounds Checker instrumentation. Each memory reference
11277 is instrumented with checks of the pointer used for memory access against
11278 bounds associated with that pointer.
11281 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
11282 and @option{-mmpx} are required to enable this feature.
11283 MPX-based instrumentation requires
11284 a runtime library to enable MPX in hardware and handle bounds
11285 violation signals. By default when @option{-fcheck-pointer-bounds}
11286 and @option{-mmpx} options are used to link a program, the GCC driver
11287 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
11288 Bounds checking on calls to dynamic libraries requires a linker
11289 with @option{-z bndplt} support; if GCC was configured with a linker
11290 without support for this option (including the Gold linker and older
11291 versions of ld), a warning is given if you link with @option{-mmpx}
11292 without also specifying @option{-static}, since the overall effectiveness
11293 of the bounds checking protection is reduced.
11294 See also @option{-static-libmpxwrappers}.
11296 MPX-based instrumentation
11297 may be used for debugging and also may be included in production code
11298 to increase program security. Depending on usage, you may
11299 have different requirements for the runtime library. The current version
11300 of the MPX runtime library is more oriented for use as a debugging
11301 tool. MPX runtime library usage implies @option{-lpthread}. See
11302 also @option{-static-libmpx}. The runtime library behavior can be
11303 influenced using various @env{CHKP_RT_*} environment variables. See
11304 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
11307 Generated instrumentation may be controlled by various
11308 @option{-fchkp-*} options and by the @code{bnd_variable_size}
11309 structure field attribute (@pxref{Type Attributes}) and
11310 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
11311 (@pxref{Function Attributes}). GCC also provides a number of built-in
11312 functions for controlling the Pointer Bounds Checker. @xref{Pointer
11313 Bounds Checker builtins}, for more information.
11315 @item -fchkp-check-incomplete-type
11316 @opindex fchkp-check-incomplete-type
11317 @opindex fno-chkp-check-incomplete-type
11318 Generate pointer bounds checks for variables with incomplete type.
11319 Enabled by default.
11321 @item -fchkp-narrow-bounds
11322 @opindex fchkp-narrow-bounds
11323 @opindex fno-chkp-narrow-bounds
11324 Controls bounds used by Pointer Bounds Checker for pointers to object
11325 fields. If narrowing is enabled then field bounds are used. Otherwise
11326 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
11327 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
11329 @item -fchkp-first-field-has-own-bounds
11330 @opindex fchkp-first-field-has-own-bounds
11331 @opindex fno-chkp-first-field-has-own-bounds
11332 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
11333 first field in the structure. By default a pointer to the first field has
11334 the same bounds as a pointer to the whole structure.
11336 @item -fchkp-flexible-struct-trailing-arrays
11337 @opindex fchkp-flexible-struct-trailing-arrays
11338 @opindex fno-chkp-flexible-struct-trailing-arrays
11339 Forces Pointer Bounds Checker to treat all trailing arrays in structures as
11340 possibly flexible. By default only array fields with zero length or that are
11341 marked with attribute bnd_variable_size are treated as flexible.
11343 @item -fchkp-narrow-to-innermost-array
11344 @opindex fchkp-narrow-to-innermost-array
11345 @opindex fno-chkp-narrow-to-innermost-array
11346 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
11347 case of nested static array access. By default this option is disabled and
11348 bounds of the outermost array are used.
11350 @item -fchkp-optimize
11351 @opindex fchkp-optimize
11352 @opindex fno-chkp-optimize
11353 Enables Pointer Bounds Checker optimizations. Enabled by default at
11354 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11356 @item -fchkp-use-fast-string-functions
11357 @opindex fchkp-use-fast-string-functions
11358 @opindex fno-chkp-use-fast-string-functions
11359 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11360 by Pointer Bounds Checker. Disabled by default.
11362 @item -fchkp-use-nochk-string-functions
11363 @opindex fchkp-use-nochk-string-functions
11364 @opindex fno-chkp-use-nochk-string-functions
11365 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11366 by Pointer Bounds Checker. Disabled by default.
11368 @item -fchkp-use-static-bounds
11369 @opindex fchkp-use-static-bounds
11370 @opindex fno-chkp-use-static-bounds
11371 Allow Pointer Bounds Checker to generate static bounds holding
11372 bounds of static variables. Enabled by default.
11374 @item -fchkp-use-static-const-bounds
11375 @opindex fchkp-use-static-const-bounds
11376 @opindex fno-chkp-use-static-const-bounds
11377 Use statically-initialized bounds for constant bounds instead of
11378 generating them each time they are required. By default enabled when
11379 @option{-fchkp-use-static-bounds} is enabled.
11381 @item -fchkp-treat-zero-dynamic-size-as-infinite
11382 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11383 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11384 With this option, objects with incomplete type whose
11385 dynamically-obtained size is zero are treated as having infinite size
11386 instead by Pointer Bounds
11387 Checker. This option may be helpful if a program is linked with a library
11388 missing size information for some symbols. Disabled by default.
11390 @item -fchkp-check-read
11391 @opindex fchkp-check-read
11392 @opindex fno-chkp-check-read
11393 Instructs Pointer Bounds Checker to generate checks for all read
11394 accesses to memory. Enabled by default.
11396 @item -fchkp-check-write
11397 @opindex fchkp-check-write
11398 @opindex fno-chkp-check-write
11399 Instructs Pointer Bounds Checker to generate checks for all write
11400 accesses to memory. Enabled by default.
11402 @item -fchkp-store-bounds
11403 @opindex fchkp-store-bounds
11404 @opindex fno-chkp-store-bounds
11405 Instructs Pointer Bounds Checker to generate bounds stores for
11406 pointer writes. Enabled by default.
11408 @item -fchkp-instrument-calls
11409 @opindex fchkp-instrument-calls
11410 @opindex fno-chkp-instrument-calls
11411 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11412 Enabled by default.
11414 @item -fchkp-instrument-marked-only
11415 @opindex fchkp-instrument-marked-only
11416 @opindex fno-chkp-instrument-marked-only
11417 Instructs Pointer Bounds Checker to instrument only functions
11418 marked with the @code{bnd_instrument} attribute
11419 (@pxref{Function Attributes}). Disabled by default.
11421 @item -fchkp-use-wrappers
11422 @opindex fchkp-use-wrappers
11423 @opindex fno-chkp-use-wrappers
11424 Allows Pointer Bounds Checker to replace calls to built-in functions
11425 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11426 is used to link a program, the GCC driver automatically links
11427 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11428 Enabled by default.
11430 @item -fcf-protection==@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
11431 @opindex fcf-protection
11432 Enable code instrumentation of control-flow transfers to increase
11433 program security by checking that target addresses of control-flow
11434 transfer instructions (such as indirect function call, function return,
11435 indirect jump) are valid. This prevents diverting the flow of control
11436 to an unexpected target. This is intended to protect against such
11437 threats as Return-oriented Programming (ROP), and similarly
11438 call/jmp-oriented programming (COP/JOP).
11440 The value @code{branch} tells the compiler to implement checking of
11441 validity of control-flow transfer at the point of indirect branch
11442 instructions, i.e. call/jmp instructions. The value @code{return}
11443 implements checking of validity at the point of returning from a
11444 function. The value @code{full} is an alias for specifying both
11445 @code{branch} and @code{return}. The value @code{none} turns off
11448 You can also use the @code{nocf_check} attribute to identify
11449 which functions and calls should be skipped from instrumentation
11450 (@pxref{Function Attributes}).
11452 Currently the x86 GNU/Linux target provides an implementation based
11453 on Intel Control-flow Enforcement Technology (CET). Instrumentation
11454 for x86 is controlled by target-specific options @option{-mcet},
11455 @option{-mibt} and @option{-mshstk} (@pxref{x86 Options}).
11457 @item -fstack-protector
11458 @opindex fstack-protector
11459 Emit extra code to check for buffer overflows, such as stack smashing
11460 attacks. This is done by adding a guard variable to functions with
11461 vulnerable objects. This includes functions that call @code{alloca}, and
11462 functions with buffers larger than 8 bytes. The guards are initialized
11463 when a function is entered and then checked when the function exits.
11464 If a guard check fails, an error message is printed and the program exits.
11466 @item -fstack-protector-all
11467 @opindex fstack-protector-all
11468 Like @option{-fstack-protector} except that all functions are protected.
11470 @item -fstack-protector-strong
11471 @opindex fstack-protector-strong
11472 Like @option{-fstack-protector} but includes additional functions to
11473 be protected --- those that have local array definitions, or have
11474 references to local frame addresses.
11476 @item -fstack-protector-explicit
11477 @opindex fstack-protector-explicit
11478 Like @option{-fstack-protector} but only protects those functions which
11479 have the @code{stack_protect} attribute.
11481 @item -fstack-check
11482 @opindex fstack-check
11483 Generate code to verify that you do not go beyond the boundary of the
11484 stack. You should specify this flag if you are running in an
11485 environment with multiple threads, but you only rarely need to specify it in
11486 a single-threaded environment since stack overflow is automatically
11487 detected on nearly all systems if there is only one stack.
11489 Note that this switch does not actually cause checking to be done; the
11490 operating system or the language runtime must do that. The switch causes
11491 generation of code to ensure that they see the stack being extended.
11493 You can additionally specify a string parameter: @samp{no} means no
11494 checking, @samp{generic} means force the use of old-style checking,
11495 @samp{specific} means use the best checking method and is equivalent
11496 to bare @option{-fstack-check}.
11498 Old-style checking is a generic mechanism that requires no specific
11499 target support in the compiler but comes with the following drawbacks:
11503 Modified allocation strategy for large objects: they are always
11504 allocated dynamically if their size exceeds a fixed threshold. Note this
11505 may change the semantics of some code.
11508 Fixed limit on the size of the static frame of functions: when it is
11509 topped by a particular function, stack checking is not reliable and
11510 a warning is issued by the compiler.
11513 Inefficiency: because of both the modified allocation strategy and the
11514 generic implementation, code performance is hampered.
11517 Note that old-style stack checking is also the fallback method for
11518 @samp{specific} if no target support has been added in the compiler.
11520 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
11521 and stack overflows. @samp{specific} is an excellent choice when compiling
11522 Ada code. It is not generally sufficient to protect against stack-clash
11523 attacks. To protect against those you want @samp{-fstack-clash-protection}.
11525 @item -fstack-clash-protection
11526 @opindex fstack-clash-protection
11527 Generate code to prevent stack clash style attacks. When this option is
11528 enabled, the compiler will only allocate one page of stack space at a time
11529 and each page is accessed immediately after allocation. Thus, it prevents
11530 allocations from jumping over any stack guard page provided by the
11533 Most targets do not fully support stack clash protection. However, on
11534 those targets @option{-fstack-clash-protection} will protect dynamic stack
11535 allocations. @option{-fstack-clash-protection} may also provide limited
11536 protection for static stack allocations if the target supports
11537 @option{-fstack-check=specific}.
11539 @item -fstack-limit-register=@var{reg}
11540 @itemx -fstack-limit-symbol=@var{sym}
11541 @itemx -fno-stack-limit
11542 @opindex fstack-limit-register
11543 @opindex fstack-limit-symbol
11544 @opindex fno-stack-limit
11545 Generate code to ensure that the stack does not grow beyond a certain value,
11546 either the value of a register or the address of a symbol. If a larger
11547 stack is required, a signal is raised at run time. For most targets,
11548 the signal is raised before the stack overruns the boundary, so
11549 it is possible to catch the signal without taking special precautions.
11551 For instance, if the stack starts at absolute address @samp{0x80000000}
11552 and grows downwards, you can use the flags
11553 @option{-fstack-limit-symbol=__stack_limit} and
11554 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11555 of 128KB@. Note that this may only work with the GNU linker.
11557 You can locally override stack limit checking by using the
11558 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11560 @item -fsplit-stack
11561 @opindex fsplit-stack
11562 Generate code to automatically split the stack before it overflows.
11563 The resulting program has a discontiguous stack which can only
11564 overflow if the program is unable to allocate any more memory. This
11565 is most useful when running threaded programs, as it is no longer
11566 necessary to calculate a good stack size to use for each thread. This
11567 is currently only implemented for the x86 targets running
11570 When code compiled with @option{-fsplit-stack} calls code compiled
11571 without @option{-fsplit-stack}, there may not be much stack space
11572 available for the latter code to run. If compiling all code,
11573 including library code, with @option{-fsplit-stack} is not an option,
11574 then the linker can fix up these calls so that the code compiled
11575 without @option{-fsplit-stack} always has a large stack. Support for
11576 this is implemented in the gold linker in GNU binutils release 2.21
11579 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11580 @opindex fvtable-verify
11581 This option is only available when compiling C++ code.
11582 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11583 feature that verifies at run time, for every virtual call, that
11584 the vtable pointer through which the call is made is valid for the type of
11585 the object, and has not been corrupted or overwritten. If an invalid vtable
11586 pointer is detected at run time, an error is reported and execution of the
11587 program is immediately halted.
11589 This option causes run-time data structures to be built at program startup,
11590 which are used for verifying the vtable pointers.
11591 The options @samp{std} and @samp{preinit}
11592 control the timing of when these data structures are built. In both cases the
11593 data structures are built before execution reaches @code{main}. Using
11594 @option{-fvtable-verify=std} causes the data structures to be built after
11595 shared libraries have been loaded and initialized.
11596 @option{-fvtable-verify=preinit} causes them to be built before shared
11597 libraries have been loaded and initialized.
11599 If this option appears multiple times in the command line with different
11600 values specified, @samp{none} takes highest priority over both @samp{std} and
11601 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11604 @opindex fvtv-debug
11605 When used in conjunction with @option{-fvtable-verify=std} or
11606 @option{-fvtable-verify=preinit}, causes debug versions of the
11607 runtime functions for the vtable verification feature to be called.
11608 This flag also causes the compiler to log information about which
11609 vtable pointers it finds for each class.
11610 This information is written to a file named @file{vtv_set_ptr_data.log}
11611 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11612 if that is defined or the current working directory otherwise.
11614 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11615 file, be sure to delete any existing one.
11618 @opindex fvtv-counts
11619 This is a debugging flag. When used in conjunction with
11620 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11621 causes the compiler to keep track of the total number of virtual calls
11622 it encounters and the number of verifications it inserts. It also
11623 counts the number of calls to certain run-time library functions
11624 that it inserts and logs this information for each compilation unit.
11625 The compiler writes this information to a file named
11626 @file{vtv_count_data.log} in the directory named by the environment
11627 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11628 directory otherwise. It also counts the size of the vtable pointer sets
11629 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11630 in the same directory.
11632 Note: This feature @emph{appends} data to the log files. To get fresh log
11633 files, be sure to delete any existing ones.
11635 @item -finstrument-functions
11636 @opindex finstrument-functions
11637 Generate instrumentation calls for entry and exit to functions. Just
11638 after function entry and just before function exit, the following
11639 profiling functions are called with the address of the current
11640 function and its call site. (On some platforms,
11641 @code{__builtin_return_address} does not work beyond the current
11642 function, so the call site information may not be available to the
11643 profiling functions otherwise.)
11646 void __cyg_profile_func_enter (void *this_fn,
11648 void __cyg_profile_func_exit (void *this_fn,
11652 The first argument is the address of the start of the current function,
11653 which may be looked up exactly in the symbol table.
11655 This instrumentation is also done for functions expanded inline in other
11656 functions. The profiling calls indicate where, conceptually, the
11657 inline function is entered and exited. This means that addressable
11658 versions of such functions must be available. If all your uses of a
11659 function are expanded inline, this may mean an additional expansion of
11660 code size. If you use @code{extern inline} in your C code, an
11661 addressable version of such functions must be provided. (This is
11662 normally the case anyway, but if you get lucky and the optimizer always
11663 expands the functions inline, you might have gotten away without
11664 providing static copies.)
11666 A function may be given the attribute @code{no_instrument_function}, in
11667 which case this instrumentation is not done. This can be used, for
11668 example, for the profiling functions listed above, high-priority
11669 interrupt routines, and any functions from which the profiling functions
11670 cannot safely be called (perhaps signal handlers, if the profiling
11671 routines generate output or allocate memory).
11673 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11674 @opindex finstrument-functions-exclude-file-list
11676 Set the list of functions that are excluded from instrumentation (see
11677 the description of @option{-finstrument-functions}). If the file that
11678 contains a function definition matches with one of @var{file}, then
11679 that function is not instrumented. The match is done on substrings:
11680 if the @var{file} parameter is a substring of the file name, it is
11681 considered to be a match.
11686 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11690 excludes any inline function defined in files whose pathnames
11691 contain @file{/bits/stl} or @file{include/sys}.
11693 If, for some reason, you want to include letter @samp{,} in one of
11694 @var{sym}, write @samp{\,}. For example,
11695 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11696 (note the single quote surrounding the option).
11698 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11699 @opindex finstrument-functions-exclude-function-list
11701 This is similar to @option{-finstrument-functions-exclude-file-list},
11702 but this option sets the list of function names to be excluded from
11703 instrumentation. The function name to be matched is its user-visible
11704 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11705 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11706 match is done on substrings: if the @var{sym} parameter is a substring
11707 of the function name, it is considered to be a match. For C99 and C++
11708 extended identifiers, the function name must be given in UTF-8, not
11709 using universal character names.
11711 @item -fpatchable-function-entry=@var{N}[,@var{M}]
11712 @opindex fpatchable-function-entry
11713 Generate @var{N} NOPs right at the beginning
11714 of each function, with the function entry point before the @var{M}th NOP.
11715 If @var{M} is omitted, it defaults to @code{0} so the
11716 function entry points to the address just at the first NOP.
11717 The NOP instructions reserve extra space which can be used to patch in
11718 any desired instrumentation at run time, provided that the code segment
11719 is writable. The amount of space is controllable indirectly via
11720 the number of NOPs; the NOP instruction used corresponds to the instruction
11721 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
11722 is target-specific and may also depend on the architecture variant and/or
11723 other compilation options.
11725 For run-time identification, the starting addresses of these areas,
11726 which correspond to their respective function entries minus @var{M},
11727 are additionally collected in the @code{__patchable_function_entries}
11728 section of the resulting binary.
11730 Note that the value of @code{__attribute__ ((patchable_function_entry
11731 (N,M)))} takes precedence over command-line option
11732 @option{-fpatchable-function-entry=N,M}. This can be used to increase
11733 the area size or to remove it completely on a single function.
11734 If @code{N=0}, no pad location is recorded.
11736 The NOP instructions are inserted at---and maybe before, depending on
11737 @var{M}---the function entry address, even before the prologue.
11742 @node Preprocessor Options
11743 @section Options Controlling the Preprocessor
11744 @cindex preprocessor options
11745 @cindex options, preprocessor
11747 These options control the C preprocessor, which is run on each C source
11748 file before actual compilation.
11750 If you use the @option{-E} option, nothing is done except preprocessing.
11751 Some of these options make sense only together with @option{-E} because
11752 they cause the preprocessor output to be unsuitable for actual
11755 In addition to the options listed here, there are a number of options
11756 to control search paths for include files documented in
11757 @ref{Directory Options}.
11758 Options to control preprocessor diagnostics are listed in
11759 @ref{Warning Options}.
11762 @include cppopts.texi
11764 @item -Wp,@var{option}
11766 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11767 and pass @var{option} directly through to the preprocessor. If
11768 @var{option} contains commas, it is split into multiple options at the
11769 commas. However, many options are modified, translated or interpreted
11770 by the compiler driver before being passed to the preprocessor, and
11771 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11772 interface is undocumented and subject to change, so whenever possible
11773 you should avoid using @option{-Wp} and let the driver handle the
11776 @item -Xpreprocessor @var{option}
11777 @opindex Xpreprocessor
11778 Pass @var{option} as an option to the preprocessor. You can use this to
11779 supply system-specific preprocessor options that GCC does not
11782 If you want to pass an option that takes an argument, you must use
11783 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11785 @item -no-integrated-cpp
11786 @opindex no-integrated-cpp
11787 Perform preprocessing as a separate pass before compilation.
11788 By default, GCC performs preprocessing as an integrated part of
11789 input tokenization and parsing.
11790 If this option is provided, the appropriate language front end
11791 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11792 and Objective-C, respectively) is instead invoked twice,
11793 once for preprocessing only and once for actual compilation
11794 of the preprocessed input.
11795 This option may be useful in conjunction with the @option{-B} or
11796 @option{-wrapper} options to specify an alternate preprocessor or
11797 perform additional processing of the program source between
11798 normal preprocessing and compilation.
11802 @node Assembler Options
11803 @section Passing Options to the Assembler
11805 @c prevent bad page break with this line
11806 You can pass options to the assembler.
11809 @item -Wa,@var{option}
11811 Pass @var{option} as an option to the assembler. If @var{option}
11812 contains commas, it is split into multiple options at the commas.
11814 @item -Xassembler @var{option}
11815 @opindex Xassembler
11816 Pass @var{option} as an option to the assembler. You can use this to
11817 supply system-specific assembler options that GCC does not
11820 If you want to pass an option that takes an argument, you must use
11821 @option{-Xassembler} twice, once for the option and once for the argument.
11826 @section Options for Linking
11827 @cindex link options
11828 @cindex options, linking
11830 These options come into play when the compiler links object files into
11831 an executable output file. They are meaningless if the compiler is
11832 not doing a link step.
11836 @item @var{object-file-name}
11837 A file name that does not end in a special recognized suffix is
11838 considered to name an object file or library. (Object files are
11839 distinguished from libraries by the linker according to the file
11840 contents.) If linking is done, these object files are used as input
11849 If any of these options is used, then the linker is not run, and
11850 object file names should not be used as arguments. @xref{Overall
11854 @opindex fuse-ld=bfd
11855 Use the @command{bfd} linker instead of the default linker.
11857 @item -fuse-ld=gold
11858 @opindex fuse-ld=gold
11859 Use the @command{gold} linker instead of the default linker.
11862 @item -l@var{library}
11863 @itemx -l @var{library}
11865 Search the library named @var{library} when linking. (The second
11866 alternative with the library as a separate argument is only for
11867 POSIX compliance and is not recommended.)
11869 It makes a difference where in the command you write this option; the
11870 linker searches and processes libraries and object files in the order they
11871 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11872 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11873 to functions in @samp{z}, those functions may not be loaded.
11875 The linker searches a standard list of directories for the library,
11876 which is actually a file named @file{lib@var{library}.a}. The linker
11877 then uses this file as if it had been specified precisely by name.
11879 The directories searched include several standard system directories
11880 plus any that you specify with @option{-L}.
11882 Normally the files found this way are library files---archive files
11883 whose members are object files. The linker handles an archive file by
11884 scanning through it for members which define symbols that have so far
11885 been referenced but not defined. But if the file that is found is an
11886 ordinary object file, it is linked in the usual fashion. The only
11887 difference between using an @option{-l} option and specifying a file name
11888 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11889 and searches several directories.
11893 You need this special case of the @option{-l} option in order to
11894 link an Objective-C or Objective-C++ program.
11896 @item -nostartfiles
11897 @opindex nostartfiles
11898 Do not use the standard system startup files when linking.
11899 The standard system libraries are used normally, unless @option{-nostdlib}
11900 or @option{-nodefaultlibs} is used.
11902 @item -nodefaultlibs
11903 @opindex nodefaultlibs
11904 Do not use the standard system libraries when linking.
11905 Only the libraries you specify are passed to the linker, and options
11906 specifying linkage of the system libraries, such as @option{-static-libgcc}
11907 or @option{-shared-libgcc}, are ignored.
11908 The standard startup files are used normally, unless @option{-nostartfiles}
11911 The compiler may generate calls to @code{memcmp},
11912 @code{memset}, @code{memcpy} and @code{memmove}.
11913 These entries are usually resolved by entries in
11914 libc. These entry points should be supplied through some other
11915 mechanism when this option is specified.
11919 Do not use the standard system startup files or libraries when linking.
11920 No startup files and only the libraries you specify are passed to
11921 the linker, and options specifying linkage of the system libraries, such as
11922 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11924 The compiler may generate calls to @code{memcmp}, @code{memset},
11925 @code{memcpy} and @code{memmove}.
11926 These entries are usually resolved by entries in
11927 libc. These entry points should be supplied through some other
11928 mechanism when this option is specified.
11930 @cindex @option{-lgcc}, use with @option{-nostdlib}
11931 @cindex @option{-nostdlib} and unresolved references
11932 @cindex unresolved references and @option{-nostdlib}
11933 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11934 @cindex @option{-nodefaultlibs} and unresolved references
11935 @cindex unresolved references and @option{-nodefaultlibs}
11936 One of the standard libraries bypassed by @option{-nostdlib} and
11937 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11938 which GCC uses to overcome shortcomings of particular machines, or special
11939 needs for some languages.
11940 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11941 Collection (GCC) Internals},
11942 for more discussion of @file{libgcc.a}.)
11943 In most cases, you need @file{libgcc.a} even when you want to avoid
11944 other standard libraries. In other words, when you specify @option{-nostdlib}
11945 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11946 This ensures that you have no unresolved references to internal GCC
11947 library subroutines.
11948 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11949 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11950 GNU Compiler Collection (GCC) Internals}.)
11954 Produce a dynamically linked position independent executable on targets
11955 that support it. For predictable results, you must also specify the same
11956 set of options used for compilation (@option{-fpie}, @option{-fPIE},
11957 or model suboptions) when you specify this linker option.
11961 Don't produce a dynamically linked position independent executable.
11964 @opindex static-pie
11965 Produce a static position independent executable on targets that support
11966 it. A static position independent executable is similar to a static
11967 executable, but can be loaded at any address without a dynamic linker.
11968 For predictable results, you must also specify the same set of options
11969 used for compilation (@option{-fpie}, @option{-fPIE}, or model
11970 suboptions) when you specify this linker option.
11974 Link with the POSIX threads library. This option is supported on
11975 GNU/Linux targets, most other Unix derivatives, and also on
11976 x86 Cygwin and MinGW targets. On some targets this option also sets
11977 flags for the preprocessor, so it should be used consistently for both
11978 compilation and linking.
11982 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11983 that support it. This instructs the linker to add all symbols, not
11984 only used ones, to the dynamic symbol table. This option is needed
11985 for some uses of @code{dlopen} or to allow obtaining backtraces
11986 from within a program.
11990 Remove all symbol table and relocation information from the executable.
11994 On systems that support dynamic linking, this overrides @option{-pie}
11995 and prevents linking with the shared libraries. On other systems, this
11996 option has no effect.
12000 Produce a shared object which can then be linked with other objects to
12001 form an executable. Not all systems support this option. For predictable
12002 results, you must also specify the same set of options used for compilation
12003 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
12004 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
12005 needs to build supplementary stub code for constructors to work. On
12006 multi-libbed systems, @samp{gcc -shared} must select the correct support
12007 libraries to link against. Failing to supply the correct flags may lead
12008 to subtle defects. Supplying them in cases where they are not necessary
12011 @item -shared-libgcc
12012 @itemx -static-libgcc
12013 @opindex shared-libgcc
12014 @opindex static-libgcc
12015 On systems that provide @file{libgcc} as a shared library, these options
12016 force the use of either the shared or static version, respectively.
12017 If no shared version of @file{libgcc} was built when the compiler was
12018 configured, these options have no effect.
12020 There are several situations in which an application should use the
12021 shared @file{libgcc} instead of the static version. The most common
12022 of these is when the application wishes to throw and catch exceptions
12023 across different shared libraries. In that case, each of the libraries
12024 as well as the application itself should use the shared @file{libgcc}.
12026 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
12027 whenever you build a shared library or a main executable, because C++
12028 programs typically use exceptions, so this is the right thing to do.
12030 If, instead, you use the GCC driver to create shared libraries, you may
12031 find that they are not always linked with the shared @file{libgcc}.
12032 If GCC finds, at its configuration time, that you have a non-GNU linker
12033 or a GNU linker that does not support option @option{--eh-frame-hdr},
12034 it links the shared version of @file{libgcc} into shared libraries
12035 by default. Otherwise, it takes advantage of the linker and optimizes
12036 away the linking with the shared version of @file{libgcc}, linking with
12037 the static version of libgcc by default. This allows exceptions to
12038 propagate through such shared libraries, without incurring relocation
12039 costs at library load time.
12041 However, if a library or main executable is supposed to throw or catch
12042 exceptions, you must link it using the G++ driver, as appropriate
12043 for the languages used in the program, or using the option
12044 @option{-shared-libgcc}, such that it is linked with the shared
12047 @item -static-libasan
12048 @opindex static-libasan
12049 When the @option{-fsanitize=address} option is used to link a program,
12050 the GCC driver automatically links against @option{libasan}. If
12051 @file{libasan} is available as a shared library, and the @option{-static}
12052 option is not used, then this links against the shared version of
12053 @file{libasan}. The @option{-static-libasan} option directs the GCC
12054 driver to link @file{libasan} statically, without necessarily linking
12055 other libraries statically.
12057 @item -static-libtsan
12058 @opindex static-libtsan
12059 When the @option{-fsanitize=thread} option is used to link a program,
12060 the GCC driver automatically links against @option{libtsan}. If
12061 @file{libtsan} is available as a shared library, and the @option{-static}
12062 option is not used, then this links against the shared version of
12063 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
12064 driver to link @file{libtsan} statically, without necessarily linking
12065 other libraries statically.
12067 @item -static-liblsan
12068 @opindex static-liblsan
12069 When the @option{-fsanitize=leak} option is used to link a program,
12070 the GCC driver automatically links against @option{liblsan}. If
12071 @file{liblsan} is available as a shared library, and the @option{-static}
12072 option is not used, then this links against the shared version of
12073 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
12074 driver to link @file{liblsan} statically, without necessarily linking
12075 other libraries statically.
12077 @item -static-libubsan
12078 @opindex static-libubsan
12079 When the @option{-fsanitize=undefined} option is used to link a program,
12080 the GCC driver automatically links against @option{libubsan}. If
12081 @file{libubsan} is available as a shared library, and the @option{-static}
12082 option is not used, then this links against the shared version of
12083 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
12084 driver to link @file{libubsan} statically, without necessarily linking
12085 other libraries statically.
12087 @item -static-libmpx
12088 @opindex static-libmpx
12089 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
12090 used to link a program, the GCC driver automatically links against
12091 @file{libmpx}. If @file{libmpx} is available as a shared library,
12092 and the @option{-static} option is not used, then this links against
12093 the shared version of @file{libmpx}. The @option{-static-libmpx}
12094 option directs the GCC driver to link @file{libmpx} statically,
12095 without necessarily linking other libraries statically.
12097 @item -static-libmpxwrappers
12098 @opindex static-libmpxwrappers
12099 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
12100 to link a program without also using @option{-fno-chkp-use-wrappers}, the
12101 GCC driver automatically links against @file{libmpxwrappers}. If
12102 @file{libmpxwrappers} is available as a shared library, and the
12103 @option{-static} option is not used, then this links against the shared
12104 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
12105 option directs the GCC driver to link @file{libmpxwrappers} statically,
12106 without necessarily linking other libraries statically.
12108 @item -static-libstdc++
12109 @opindex static-libstdc++
12110 When the @command{g++} program is used to link a C++ program, it
12111 normally automatically links against @option{libstdc++}. If
12112 @file{libstdc++} is available as a shared library, and the
12113 @option{-static} option is not used, then this links against the
12114 shared version of @file{libstdc++}. That is normally fine. However, it
12115 is sometimes useful to freeze the version of @file{libstdc++} used by
12116 the program without going all the way to a fully static link. The
12117 @option{-static-libstdc++} option directs the @command{g++} driver to
12118 link @file{libstdc++} statically, without necessarily linking other
12119 libraries statically.
12123 Bind references to global symbols when building a shared object. Warn
12124 about any unresolved references (unless overridden by the link editor
12125 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
12128 @item -T @var{script}
12130 @cindex linker script
12131 Use @var{script} as the linker script. This option is supported by most
12132 systems using the GNU linker. On some targets, such as bare-board
12133 targets without an operating system, the @option{-T} option may be required
12134 when linking to avoid references to undefined symbols.
12136 @item -Xlinker @var{option}
12138 Pass @var{option} as an option to the linker. You can use this to
12139 supply system-specific linker options that GCC does not recognize.
12141 If you want to pass an option that takes a separate argument, you must use
12142 @option{-Xlinker} twice, once for the option and once for the argument.
12143 For example, to pass @option{-assert definitions}, you must write
12144 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
12145 @option{-Xlinker "-assert definitions"}, because this passes the entire
12146 string as a single argument, which is not what the linker expects.
12148 When using the GNU linker, it is usually more convenient to pass
12149 arguments to linker options using the @option{@var{option}=@var{value}}
12150 syntax than as separate arguments. For example, you can specify
12151 @option{-Xlinker -Map=output.map} rather than
12152 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
12153 this syntax for command-line options.
12155 @item -Wl,@var{option}
12157 Pass @var{option} as an option to the linker. If @var{option} contains
12158 commas, it is split into multiple options at the commas. You can use this
12159 syntax to pass an argument to the option.
12160 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
12161 linker. When using the GNU linker, you can also get the same effect with
12162 @option{-Wl,-Map=output.map}.
12164 @item -u @var{symbol}
12166 Pretend the symbol @var{symbol} is undefined, to force linking of
12167 library modules to define it. You can use @option{-u} multiple times with
12168 different symbols to force loading of additional library modules.
12170 @item -z @var{keyword}
12172 @option{-z} is passed directly on to the linker along with the keyword
12173 @var{keyword}. See the section in the documentation of your linker for
12174 permitted values and their meanings.
12177 @node Directory Options
12178 @section Options for Directory Search
12179 @cindex directory options
12180 @cindex options, directory search
12181 @cindex search path
12183 These options specify directories to search for header files, for
12184 libraries and for parts of the compiler:
12187 @include cppdiropts.texi
12189 @item -iplugindir=@var{dir}
12190 @opindex iplugindir=
12191 Set the directory to search for plugins that are passed
12192 by @option{-fplugin=@var{name}} instead of
12193 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
12194 to be used by the user, but only passed by the driver.
12198 Add directory @var{dir} to the list of directories to be searched
12201 @item -B@var{prefix}
12203 This option specifies where to find the executables, libraries,
12204 include files, and data files of the compiler itself.
12206 The compiler driver program runs one or more of the subprograms
12207 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
12208 @var{prefix} as a prefix for each program it tries to run, both with and
12209 without @samp{@var{machine}/@var{version}/} for the corresponding target
12210 machine and compiler version.
12212 For each subprogram to be run, the compiler driver first tries the
12213 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
12214 is not specified, the driver tries two standard prefixes,
12215 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
12216 those results in a file name that is found, the unmodified program
12217 name is searched for using the directories specified in your
12218 @env{PATH} environment variable.
12220 The compiler checks to see if the path provided by @option{-B}
12221 refers to a directory, and if necessary it adds a directory
12222 separator character at the end of the path.
12224 @option{-B} prefixes that effectively specify directory names also apply
12225 to libraries in the linker, because the compiler translates these
12226 options into @option{-L} options for the linker. They also apply to
12227 include files in the preprocessor, because the compiler translates these
12228 options into @option{-isystem} options for the preprocessor. In this case,
12229 the compiler appends @samp{include} to the prefix.
12231 The runtime support file @file{libgcc.a} can also be searched for using
12232 the @option{-B} prefix, if needed. If it is not found there, the two
12233 standard prefixes above are tried, and that is all. The file is left
12234 out of the link if it is not found by those means.
12236 Another way to specify a prefix much like the @option{-B} prefix is to use
12237 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
12240 As a special kludge, if the path provided by @option{-B} is
12241 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
12242 9, then it is replaced by @file{[dir/]include}. This is to help
12243 with boot-strapping the compiler.
12245 @item -no-canonical-prefixes
12246 @opindex no-canonical-prefixes
12247 Do not expand any symbolic links, resolve references to @samp{/../}
12248 or @samp{/./}, or make the path absolute when generating a relative
12251 @item --sysroot=@var{dir}
12253 Use @var{dir} as the logical root directory for headers and libraries.
12254 For example, if the compiler normally searches for headers in
12255 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
12256 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
12258 If you use both this option and the @option{-isysroot} option, then
12259 the @option{--sysroot} option applies to libraries, but the
12260 @option{-isysroot} option applies to header files.
12262 The GNU linker (beginning with version 2.16) has the necessary support
12263 for this option. If your linker does not support this option, the
12264 header file aspect of @option{--sysroot} still works, but the
12265 library aspect does not.
12267 @item --no-sysroot-suffix
12268 @opindex no-sysroot-suffix
12269 For some targets, a suffix is added to the root directory specified
12270 with @option{--sysroot}, depending on the other options used, so that
12271 headers may for example be found in
12272 @file{@var{dir}/@var{suffix}/usr/include} instead of
12273 @file{@var{dir}/usr/include}. This option disables the addition of
12278 @node Code Gen Options
12279 @section Options for Code Generation Conventions
12280 @cindex code generation conventions
12281 @cindex options, code generation
12282 @cindex run-time options
12284 These machine-independent options control the interface conventions
12285 used in code generation.
12287 Most of them have both positive and negative forms; the negative form
12288 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
12289 one of the forms is listed---the one that is not the default. You
12290 can figure out the other form by either removing @samp{no-} or adding
12294 @item -fstack-reuse=@var{reuse-level}
12295 @opindex fstack_reuse
12296 This option controls stack space reuse for user declared local/auto variables
12297 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
12298 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
12299 local variables and temporaries, @samp{named_vars} enables the reuse only for
12300 user defined local variables with names, and @samp{none} disables stack reuse
12301 completely. The default value is @samp{all}. The option is needed when the
12302 program extends the lifetime of a scoped local variable or a compiler generated
12303 temporary beyond the end point defined by the language. When a lifetime of
12304 a variable ends, and if the variable lives in memory, the optimizing compiler
12305 has the freedom to reuse its stack space with other temporaries or scoped
12306 local variables whose live range does not overlap with it. Legacy code extending
12307 local lifetime is likely to break with the stack reuse optimization.
12326 if (*p == 10) // out of scope use of local1
12337 A(int k) : i(k), j(k) @{ @}
12344 void foo(const A& ar)
12351 foo(A(10)); // temp object's lifetime ends when foo returns
12357 ap->i+= 10; // ap references out of scope temp whose space
12358 // is reused with a. What is the value of ap->i?
12363 The lifetime of a compiler generated temporary is well defined by the C++
12364 standard. When a lifetime of a temporary ends, and if the temporary lives
12365 in memory, the optimizing compiler has the freedom to reuse its stack
12366 space with other temporaries or scoped local variables whose live range
12367 does not overlap with it. However some of the legacy code relies on
12368 the behavior of older compilers in which temporaries' stack space is
12369 not reused, the aggressive stack reuse can lead to runtime errors. This
12370 option is used to control the temporary stack reuse optimization.
12374 This option generates traps for signed overflow on addition, subtraction,
12375 multiplication operations.
12376 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12377 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12378 @option{-fwrapv} being effective. Note that only active options override, so
12379 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12380 results in @option{-ftrapv} being effective.
12384 This option instructs the compiler to assume that signed arithmetic
12385 overflow of addition, subtraction and multiplication wraps around
12386 using twos-complement representation. This flag enables some optimizations
12387 and disables others.
12388 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12389 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12390 @option{-fwrapv} being effective. Note that only active options override, so
12391 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12392 results in @option{-ftrapv} being effective.
12395 @opindex fexceptions
12396 Enable exception handling. Generates extra code needed to propagate
12397 exceptions. For some targets, this implies GCC generates frame
12398 unwind information for all functions, which can produce significant data
12399 size overhead, although it does not affect execution. If you do not
12400 specify this option, GCC enables it by default for languages like
12401 C++ that normally require exception handling, and disables it for
12402 languages like C that do not normally require it. However, you may need
12403 to enable this option when compiling C code that needs to interoperate
12404 properly with exception handlers written in C++. You may also wish to
12405 disable this option if you are compiling older C++ programs that don't
12406 use exception handling.
12408 @item -fnon-call-exceptions
12409 @opindex fnon-call-exceptions
12410 Generate code that allows trapping instructions to throw exceptions.
12411 Note that this requires platform-specific runtime support that does
12412 not exist everywhere. Moreover, it only allows @emph{trapping}
12413 instructions to throw exceptions, i.e.@: memory references or floating-point
12414 instructions. It does not allow exceptions to be thrown from
12415 arbitrary signal handlers such as @code{SIGALRM}.
12417 @item -fdelete-dead-exceptions
12418 @opindex fdelete-dead-exceptions
12419 Consider that instructions that may throw exceptions but don't otherwise
12420 contribute to the execution of the program can be optimized away.
12421 This option is enabled by default for the Ada front end, as permitted by
12422 the Ada language specification.
12423 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12425 @item -funwind-tables
12426 @opindex funwind-tables
12427 Similar to @option{-fexceptions}, except that it just generates any needed
12428 static data, but does not affect the generated code in any other way.
12429 You normally do not need to enable this option; instead, a language processor
12430 that needs this handling enables it on your behalf.
12432 @item -fasynchronous-unwind-tables
12433 @opindex fasynchronous-unwind-tables
12434 Generate unwind table in DWARF format, if supported by target machine. The
12435 table is exact at each instruction boundary, so it can be used for stack
12436 unwinding from asynchronous events (such as debugger or garbage collector).
12438 @item -fno-gnu-unique
12439 @opindex fno-gnu-unique
12440 On systems with recent GNU assembler and C library, the C++ compiler
12441 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12442 of template static data members and static local variables in inline
12443 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12444 is necessary to avoid problems with a library used by two different
12445 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12446 therefore disagreeing with the other one about the binding of the
12447 symbol. But this causes @code{dlclose} to be ignored for affected
12448 DSOs; if your program relies on reinitialization of a DSO via
12449 @code{dlclose} and @code{dlopen}, you can use
12450 @option{-fno-gnu-unique}.
12452 @item -fpcc-struct-return
12453 @opindex fpcc-struct-return
12454 Return ``short'' @code{struct} and @code{union} values in memory like
12455 longer ones, rather than in registers. This convention is less
12456 efficient, but it has the advantage of allowing intercallability between
12457 GCC-compiled files and files compiled with other compilers, particularly
12458 the Portable C Compiler (pcc).
12460 The precise convention for returning structures in memory depends
12461 on the target configuration macros.
12463 Short structures and unions are those whose size and alignment match
12464 that of some integer type.
12466 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12467 switch is not binary compatible with code compiled with the
12468 @option{-freg-struct-return} switch.
12469 Use it to conform to a non-default application binary interface.
12471 @item -freg-struct-return
12472 @opindex freg-struct-return
12473 Return @code{struct} and @code{union} values in registers when possible.
12474 This is more efficient for small structures than
12475 @option{-fpcc-struct-return}.
12477 If you specify neither @option{-fpcc-struct-return} nor
12478 @option{-freg-struct-return}, GCC defaults to whichever convention is
12479 standard for the target. If there is no standard convention, GCC
12480 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12481 the principal compiler. In those cases, we can choose the standard, and
12482 we chose the more efficient register return alternative.
12484 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12485 switch is not binary compatible with code compiled with the
12486 @option{-fpcc-struct-return} switch.
12487 Use it to conform to a non-default application binary interface.
12489 @item -fshort-enums
12490 @opindex fshort-enums
12491 Allocate to an @code{enum} type only as many bytes as it needs for the
12492 declared range of possible values. Specifically, the @code{enum} type
12493 is equivalent to the smallest integer type that has enough room.
12495 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12496 code that is not binary compatible with code generated without that switch.
12497 Use it to conform to a non-default application binary interface.
12499 @item -fshort-wchar
12500 @opindex fshort-wchar
12501 Override the underlying type for @code{wchar_t} to be @code{short
12502 unsigned int} instead of the default for the target. This option is
12503 useful for building programs to run under WINE@.
12505 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12506 code that is not binary compatible with code generated without that switch.
12507 Use it to conform to a non-default application binary interface.
12510 @opindex fno-common
12511 @cindex tentative definitions
12512 In C code, this option controls the placement of global variables
12513 defined without an initializer, known as @dfn{tentative definitions}
12514 in the C standard. Tentative definitions are distinct from declarations
12515 of a variable with the @code{extern} keyword, which do not allocate storage.
12517 Unix C compilers have traditionally allocated storage for
12518 uninitialized global variables in a common block. This allows the
12519 linker to resolve all tentative definitions of the same variable
12520 in different compilation units to the same object, or to a non-tentative
12522 This is the behavior specified by @option{-fcommon}, and is the default for
12523 GCC on most targets.
12524 On the other hand, this behavior is not required by ISO
12525 C, and on some targets may carry a speed or code size penalty on
12526 variable references.
12528 The @option{-fno-common} option specifies that the compiler should instead
12529 place uninitialized global variables in the data section of the object file.
12530 This inhibits the merging of tentative definitions by the linker so
12531 you get a multiple-definition error if the same
12532 variable is defined in more than one compilation unit.
12533 Compiling with @option{-fno-common} is useful on targets for which
12534 it provides better performance, or if you wish to verify that the
12535 program will work on other systems that always treat uninitialized
12536 variable definitions this way.
12540 Ignore the @code{#ident} directive.
12542 @item -finhibit-size-directive
12543 @opindex finhibit-size-directive
12544 Don't output a @code{.size} assembler directive, or anything else that
12545 would cause trouble if the function is split in the middle, and the
12546 two halves are placed at locations far apart in memory. This option is
12547 used when compiling @file{crtstuff.c}; you should not need to use it
12550 @item -fverbose-asm
12551 @opindex fverbose-asm
12552 Put extra commentary information in the generated assembly code to
12553 make it more readable. This option is generally only of use to those
12554 who actually need to read the generated assembly code (perhaps while
12555 debugging the compiler itself).
12557 @option{-fno-verbose-asm}, the default, causes the
12558 extra information to be omitted and is useful when comparing two assembler
12561 The added comments include:
12566 information on the compiler version and command-line options,
12569 the source code lines associated with the assembly instructions,
12570 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12573 hints on which high-level expressions correspond to
12574 the various assembly instruction operands.
12578 For example, given this C source file:
12586 for (i = 0; i < n; i++)
12593 compiling to (x86_64) assembly via @option{-S} and emitting the result
12594 direct to stdout via @option{-o} @option{-}
12597 gcc -S test.c -fverbose-asm -Os -o -
12600 gives output similar to this:
12604 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12611 .type test, @@function
12615 # test.c:4: int total = 0;
12616 xorl %eax, %eax # <retval>
12617 # test.c:6: for (i = 0; i < n; i++)
12618 xorl %edx, %edx # i
12620 # test.c:6: for (i = 0; i < n; i++)
12621 cmpl %edi, %edx # n, i
12623 # test.c:7: total += i * i;
12624 movl %edx, %ecx # i, tmp92
12625 imull %edx, %ecx # i, tmp92
12626 # test.c:6: for (i = 0; i < n; i++)
12628 # test.c:7: total += i * i;
12629 addl %ecx, %eax # tmp92, <retval>
12637 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12638 .section .note.GNU-stack,"",@@progbits
12641 The comments are intended for humans rather than machines and hence the
12642 precise format of the comments is subject to change.
12644 @item -frecord-gcc-switches
12645 @opindex frecord-gcc-switches
12646 This switch causes the command line used to invoke the
12647 compiler to be recorded into the object file that is being created.
12648 This switch is only implemented on some targets and the exact format
12649 of the recording is target and binary file format dependent, but it
12650 usually takes the form of a section containing ASCII text. This
12651 switch is related to the @option{-fverbose-asm} switch, but that
12652 switch only records information in the assembler output file as
12653 comments, so it never reaches the object file.
12654 See also @option{-grecord-gcc-switches} for another
12655 way of storing compiler options into the object file.
12659 @cindex global offset table
12661 Generate position-independent code (PIC) suitable for use in a shared
12662 library, if supported for the target machine. Such code accesses all
12663 constant addresses through a global offset table (GOT)@. The dynamic
12664 loader resolves the GOT entries when the program starts (the dynamic
12665 loader is not part of GCC; it is part of the operating system). If
12666 the GOT size for the linked executable exceeds a machine-specific
12667 maximum size, you get an error message from the linker indicating that
12668 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12669 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12670 on the m68k and RS/6000. The x86 has no such limit.)
12672 Position-independent code requires special support, and therefore works
12673 only on certain machines. For the x86, GCC supports PIC for System V
12674 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12675 position-independent.
12677 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12682 If supported for the target machine, emit position-independent code,
12683 suitable for dynamic linking and avoiding any limit on the size of the
12684 global offset table. This option makes a difference on AArch64, m68k,
12685 PowerPC and SPARC@.
12687 Position-independent code requires special support, and therefore works
12688 only on certain machines.
12690 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12697 These options are similar to @option{-fpic} and @option{-fPIC}, but
12698 generated position independent code can be only linked into executables.
12699 Usually these options are used when @option{-pie} GCC option is
12700 used during linking.
12702 @option{-fpie} and @option{-fPIE} both define the macros
12703 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12704 for @option{-fpie} and 2 for @option{-fPIE}.
12708 Do not use the PLT for external function calls in position-independent code.
12709 Instead, load the callee address at call sites from the GOT and branch to it.
12710 This leads to more efficient code by eliminating PLT stubs and exposing
12711 GOT loads to optimizations. On architectures such as 32-bit x86 where
12712 PLT stubs expect the GOT pointer in a specific register, this gives more
12713 register allocation freedom to the compiler.
12714 Lazy binding requires use of the PLT;
12715 with @option{-fno-plt} all external symbols are resolved at load time.
12717 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12718 through the PLT for specific external functions.
12720 In position-dependent code, a few targets also convert calls to
12721 functions that are marked to not use the PLT to use the GOT instead.
12723 @item -fno-jump-tables
12724 @opindex fno-jump-tables
12725 Do not use jump tables for switch statements even where it would be
12726 more efficient than other code generation strategies. This option is
12727 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12728 building code that forms part of a dynamic linker and cannot
12729 reference the address of a jump table. On some targets, jump tables
12730 do not require a GOT and this option is not needed.
12732 @item -ffixed-@var{reg}
12734 Treat the register named @var{reg} as a fixed register; generated code
12735 should never refer to it (except perhaps as a stack pointer, frame
12736 pointer or in some other fixed role).
12738 @var{reg} must be the name of a register. The register names accepted
12739 are machine-specific and are defined in the @code{REGISTER_NAMES}
12740 macro in the machine description macro file.
12742 This flag does not have a negative form, because it specifies a
12745 @item -fcall-used-@var{reg}
12746 @opindex fcall-used
12747 Treat the register named @var{reg} as an allocable register that is
12748 clobbered by function calls. It may be allocated for temporaries or
12749 variables that do not live across a call. Functions compiled this way
12750 do not save and restore the register @var{reg}.
12752 It is an error to use this flag with the frame pointer or stack pointer.
12753 Use of this flag for other registers that have fixed pervasive roles in
12754 the machine's execution model produces disastrous results.
12756 This flag does not have a negative form, because it specifies a
12759 @item -fcall-saved-@var{reg}
12760 @opindex fcall-saved
12761 Treat the register named @var{reg} as an allocable register saved by
12762 functions. It may be allocated even for temporaries or variables that
12763 live across a call. Functions compiled this way save and restore
12764 the register @var{reg} if they use it.
12766 It is an error to use this flag with the frame pointer or stack pointer.
12767 Use of this flag for other registers that have fixed pervasive roles in
12768 the machine's execution model produces disastrous results.
12770 A different sort of disaster results from the use of this flag for
12771 a register in which function values may be returned.
12773 This flag does not have a negative form, because it specifies a
12776 @item -fpack-struct[=@var{n}]
12777 @opindex fpack-struct
12778 Without a value specified, pack all structure members together without
12779 holes. When a value is specified (which must be a small power of two), pack
12780 structure members according to this value, representing the maximum
12781 alignment (that is, objects with default alignment requirements larger than
12782 this are output potentially unaligned at the next fitting location.
12784 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12785 code that is not binary compatible with code generated without that switch.
12786 Additionally, it makes the code suboptimal.
12787 Use it to conform to a non-default application binary interface.
12789 @item -fleading-underscore
12790 @opindex fleading-underscore
12791 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12792 change the way C symbols are represented in the object file. One use
12793 is to help link with legacy assembly code.
12795 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12796 generate code that is not binary compatible with code generated without that
12797 switch. Use it to conform to a non-default application binary interface.
12798 Not all targets provide complete support for this switch.
12800 @item -ftls-model=@var{model}
12801 @opindex ftls-model
12802 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12803 The @var{model} argument should be one of @samp{global-dynamic},
12804 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12805 Note that the choice is subject to optimization: the compiler may use
12806 a more efficient model for symbols not visible outside of the translation
12807 unit, or if @option{-fpic} is not given on the command line.
12809 The default without @option{-fpic} is @samp{initial-exec}; with
12810 @option{-fpic} the default is @samp{global-dynamic}.
12812 @item -ftrampolines
12813 @opindex ftrampolines
12814 For targets that normally need trampolines for nested functions, always
12815 generate them instead of using descriptors. Otherwise, for targets that
12816 do not need them, like for example HP-PA or IA-64, do nothing.
12818 A trampoline is a small piece of code that is created at run time on the
12819 stack when the address of a nested function is taken, and is used to call
12820 the nested function indirectly. Therefore, it requires the stack to be
12821 made executable in order for the program to work properly.
12823 @option{-fno-trampolines} is enabled by default on a language by language
12824 basis to let the compiler avoid generating them, if it computes that this
12825 is safe, and replace them with descriptors. Descriptors are made up of data
12826 only, but the generated code must be prepared to deal with them. As of this
12827 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12829 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12830 @option{-fno-trampolines} are not binary compatible if nested functions are
12831 present. This option must therefore be used on a program-wide basis and be
12832 manipulated with extreme care.
12834 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12835 @opindex fvisibility
12836 Set the default ELF image symbol visibility to the specified option---all
12837 symbols are marked with this unless overridden within the code.
12838 Using this feature can very substantially improve linking and
12839 load times of shared object libraries, produce more optimized
12840 code, provide near-perfect API export and prevent symbol clashes.
12841 It is @strong{strongly} recommended that you use this in any shared objects
12844 Despite the nomenclature, @samp{default} always means public; i.e.,
12845 available to be linked against from outside the shared object.
12846 @samp{protected} and @samp{internal} are pretty useless in real-world
12847 usage so the only other commonly used option is @samp{hidden}.
12848 The default if @option{-fvisibility} isn't specified is
12849 @samp{default}, i.e., make every symbol public.
12851 A good explanation of the benefits offered by ensuring ELF
12852 symbols have the correct visibility is given by ``How To Write
12853 Shared Libraries'' by Ulrich Drepper (which can be found at
12854 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
12855 solution made possible by this option to marking things hidden when
12856 the default is public is to make the default hidden and mark things
12857 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12858 and @code{__attribute__ ((visibility("default")))} instead of
12859 @code{__declspec(dllexport)} you get almost identical semantics with
12860 identical syntax. This is a great boon to those working with
12861 cross-platform projects.
12863 For those adding visibility support to existing code, you may find
12864 @code{#pragma GCC visibility} of use. This works by you enclosing
12865 the declarations you wish to set visibility for with (for example)
12866 @code{#pragma GCC visibility push(hidden)} and
12867 @code{#pragma GCC visibility pop}.
12868 Bear in mind that symbol visibility should be viewed @strong{as
12869 part of the API interface contract} and thus all new code should
12870 always specify visibility when it is not the default; i.e., declarations
12871 only for use within the local DSO should @strong{always} be marked explicitly
12872 as hidden as so to avoid PLT indirection overheads---making this
12873 abundantly clear also aids readability and self-documentation of the code.
12874 Note that due to ISO C++ specification requirements, @code{operator new} and
12875 @code{operator delete} must always be of default visibility.
12877 Be aware that headers from outside your project, in particular system
12878 headers and headers from any other library you use, may not be
12879 expecting to be compiled with visibility other than the default. You
12880 may need to explicitly say @code{#pragma GCC visibility push(default)}
12881 before including any such headers.
12883 @code{extern} declarations are not affected by @option{-fvisibility}, so
12884 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12885 no modifications. However, this means that calls to @code{extern}
12886 functions with no explicit visibility use the PLT, so it is more
12887 effective to use @code{__attribute ((visibility))} and/or
12888 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12889 declarations should be treated as hidden.
12891 Note that @option{-fvisibility} does affect C++ vague linkage
12892 entities. This means that, for instance, an exception class that is
12893 be thrown between DSOs must be explicitly marked with default
12894 visibility so that the @samp{type_info} nodes are unified between
12897 An overview of these techniques, their benefits and how to use them
12898 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12900 @item -fstrict-volatile-bitfields
12901 @opindex fstrict-volatile-bitfields
12902 This option should be used if accesses to volatile bit-fields (or other
12903 structure fields, although the compiler usually honors those types
12904 anyway) should use a single access of the width of the
12905 field's type, aligned to a natural alignment if possible. For
12906 example, targets with memory-mapped peripheral registers might require
12907 all such accesses to be 16 bits wide; with this flag you can
12908 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12909 is 16 bits on these targets) to force GCC to use 16-bit accesses
12910 instead of, perhaps, a more efficient 32-bit access.
12912 If this option is disabled, the compiler uses the most efficient
12913 instruction. In the previous example, that might be a 32-bit load
12914 instruction, even though that accesses bytes that do not contain
12915 any portion of the bit-field, or memory-mapped registers unrelated to
12916 the one being updated.
12918 In some cases, such as when the @code{packed} attribute is applied to a
12919 structure field, it may not be possible to access the field with a single
12920 read or write that is correctly aligned for the target machine. In this
12921 case GCC falls back to generating multiple accesses rather than code that
12922 will fault or truncate the result at run time.
12924 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12925 not allowed to touch non bit-field members. It is therefore recommended
12926 to define all bits of the field's type as bit-field members.
12928 The default value of this option is determined by the application binary
12929 interface for the target processor.
12931 @item -fsync-libcalls
12932 @opindex fsync-libcalls
12933 This option controls whether any out-of-line instance of the @code{__sync}
12934 family of functions may be used to implement the C++11 @code{__atomic}
12935 family of functions.
12937 The default value of this option is enabled, thus the only useful form
12938 of the option is @option{-fno-sync-libcalls}. This option is used in
12939 the implementation of the @file{libatomic} runtime library.
12943 @node Developer Options
12944 @section GCC Developer Options
12945 @cindex developer options
12946 @cindex debugging GCC
12947 @cindex debug dump options
12948 @cindex dump options
12949 @cindex compilation statistics
12951 This section describes command-line options that are primarily of
12952 interest to GCC developers, including options to support compiler
12953 testing and investigation of compiler bugs and compile-time
12954 performance problems. This includes options that produce debug dumps
12955 at various points in the compilation; that print statistics such as
12956 memory use and execution time; and that print information about GCC's
12957 configuration, such as where it searches for libraries. You should
12958 rarely need to use any of these options for ordinary compilation and
12963 @item -d@var{letters}
12964 @itemx -fdump-rtl-@var{pass}
12965 @itemx -fdump-rtl-@var{pass}=@var{filename}
12967 @opindex fdump-rtl-@var{pass}
12968 Says to make debugging dumps during compilation at times specified by
12969 @var{letters}. This is used for debugging the RTL-based passes of the
12970 compiler. The file names for most of the dumps are made by appending
12971 a pass number and a word to the @var{dumpname}, and the files are
12972 created in the directory of the output file. In case of
12973 @option{=@var{filename}} option, the dump is output on the given file
12974 instead of the pass numbered dump files. Note that the pass number is
12975 assigned as passes are registered into the pass manager. Most passes
12976 are registered in the order that they will execute and for these passes
12977 the number corresponds to the pass execution order. However, passes
12978 registered by plugins, passes specific to compilation targets, or
12979 passes that are otherwise registered after all the other passes are
12980 numbered higher than a pass named "final", even if they are executed
12981 earlier. @var{dumpname} is generated from the name of the output
12982 file if explicitly specified and not an executable, otherwise it is
12983 the basename of the source file.
12985 Some @option{-d@var{letters}} switches have different meaning when
12986 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
12987 for information about preprocessor-specific dump options.
12989 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
12990 @option{-d} option @var{letters}. Here are the possible
12991 letters for use in @var{pass} and @var{letters}, and their meanings:
12995 @item -fdump-rtl-alignments
12996 @opindex fdump-rtl-alignments
12997 Dump after branch alignments have been computed.
12999 @item -fdump-rtl-asmcons
13000 @opindex fdump-rtl-asmcons
13001 Dump after fixing rtl statements that have unsatisfied in/out constraints.
13003 @item -fdump-rtl-auto_inc_dec
13004 @opindex fdump-rtl-auto_inc_dec
13005 Dump after auto-inc-dec discovery. This pass is only run on
13006 architectures that have auto inc or auto dec instructions.
13008 @item -fdump-rtl-barriers
13009 @opindex fdump-rtl-barriers
13010 Dump after cleaning up the barrier instructions.
13012 @item -fdump-rtl-bbpart
13013 @opindex fdump-rtl-bbpart
13014 Dump after partitioning hot and cold basic blocks.
13016 @item -fdump-rtl-bbro
13017 @opindex fdump-rtl-bbro
13018 Dump after block reordering.
13020 @item -fdump-rtl-btl1
13021 @itemx -fdump-rtl-btl2
13022 @opindex fdump-rtl-btl2
13023 @opindex fdump-rtl-btl2
13024 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
13025 after the two branch
13026 target load optimization passes.
13028 @item -fdump-rtl-bypass
13029 @opindex fdump-rtl-bypass
13030 Dump after jump bypassing and control flow optimizations.
13032 @item -fdump-rtl-combine
13033 @opindex fdump-rtl-combine
13034 Dump after the RTL instruction combination pass.
13036 @item -fdump-rtl-compgotos
13037 @opindex fdump-rtl-compgotos
13038 Dump after duplicating the computed gotos.
13040 @item -fdump-rtl-ce1
13041 @itemx -fdump-rtl-ce2
13042 @itemx -fdump-rtl-ce3
13043 @opindex fdump-rtl-ce1
13044 @opindex fdump-rtl-ce2
13045 @opindex fdump-rtl-ce3
13046 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
13047 @option{-fdump-rtl-ce3} enable dumping after the three
13048 if conversion passes.
13050 @item -fdump-rtl-cprop_hardreg
13051 @opindex fdump-rtl-cprop_hardreg
13052 Dump after hard register copy propagation.
13054 @item -fdump-rtl-csa
13055 @opindex fdump-rtl-csa
13056 Dump after combining stack adjustments.
13058 @item -fdump-rtl-cse1
13059 @itemx -fdump-rtl-cse2
13060 @opindex fdump-rtl-cse1
13061 @opindex fdump-rtl-cse2
13062 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
13063 the two common subexpression elimination passes.
13065 @item -fdump-rtl-dce
13066 @opindex fdump-rtl-dce
13067 Dump after the standalone dead code elimination passes.
13069 @item -fdump-rtl-dbr
13070 @opindex fdump-rtl-dbr
13071 Dump after delayed branch scheduling.
13073 @item -fdump-rtl-dce1
13074 @itemx -fdump-rtl-dce2
13075 @opindex fdump-rtl-dce1
13076 @opindex fdump-rtl-dce2
13077 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
13078 the two dead store elimination passes.
13080 @item -fdump-rtl-eh
13081 @opindex fdump-rtl-eh
13082 Dump after finalization of EH handling code.
13084 @item -fdump-rtl-eh_ranges
13085 @opindex fdump-rtl-eh_ranges
13086 Dump after conversion of EH handling range regions.
13088 @item -fdump-rtl-expand
13089 @opindex fdump-rtl-expand
13090 Dump after RTL generation.
13092 @item -fdump-rtl-fwprop1
13093 @itemx -fdump-rtl-fwprop2
13094 @opindex fdump-rtl-fwprop1
13095 @opindex fdump-rtl-fwprop2
13096 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
13097 dumping after the two forward propagation passes.
13099 @item -fdump-rtl-gcse1
13100 @itemx -fdump-rtl-gcse2
13101 @opindex fdump-rtl-gcse1
13102 @opindex fdump-rtl-gcse2
13103 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
13104 after global common subexpression elimination.
13106 @item -fdump-rtl-init-regs
13107 @opindex fdump-rtl-init-regs
13108 Dump after the initialization of the registers.
13110 @item -fdump-rtl-initvals
13111 @opindex fdump-rtl-initvals
13112 Dump after the computation of the initial value sets.
13114 @item -fdump-rtl-into_cfglayout
13115 @opindex fdump-rtl-into_cfglayout
13116 Dump after converting to cfglayout mode.
13118 @item -fdump-rtl-ira
13119 @opindex fdump-rtl-ira
13120 Dump after iterated register allocation.
13122 @item -fdump-rtl-jump
13123 @opindex fdump-rtl-jump
13124 Dump after the second jump optimization.
13126 @item -fdump-rtl-loop2
13127 @opindex fdump-rtl-loop2
13128 @option{-fdump-rtl-loop2} enables dumping after the rtl
13129 loop optimization passes.
13131 @item -fdump-rtl-mach
13132 @opindex fdump-rtl-mach
13133 Dump after performing the machine dependent reorganization pass, if that
13136 @item -fdump-rtl-mode_sw
13137 @opindex fdump-rtl-mode_sw
13138 Dump after removing redundant mode switches.
13140 @item -fdump-rtl-rnreg
13141 @opindex fdump-rtl-rnreg
13142 Dump after register renumbering.
13144 @item -fdump-rtl-outof_cfglayout
13145 @opindex fdump-rtl-outof_cfglayout
13146 Dump after converting from cfglayout mode.
13148 @item -fdump-rtl-peephole2
13149 @opindex fdump-rtl-peephole2
13150 Dump after the peephole pass.
13152 @item -fdump-rtl-postreload
13153 @opindex fdump-rtl-postreload
13154 Dump after post-reload optimizations.
13156 @item -fdump-rtl-pro_and_epilogue
13157 @opindex fdump-rtl-pro_and_epilogue
13158 Dump after generating the function prologues and epilogues.
13160 @item -fdump-rtl-sched1
13161 @itemx -fdump-rtl-sched2
13162 @opindex fdump-rtl-sched1
13163 @opindex fdump-rtl-sched2
13164 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
13165 after the basic block scheduling passes.
13167 @item -fdump-rtl-ree
13168 @opindex fdump-rtl-ree
13169 Dump after sign/zero extension elimination.
13171 @item -fdump-rtl-seqabstr
13172 @opindex fdump-rtl-seqabstr
13173 Dump after common sequence discovery.
13175 @item -fdump-rtl-shorten
13176 @opindex fdump-rtl-shorten
13177 Dump after shortening branches.
13179 @item -fdump-rtl-sibling
13180 @opindex fdump-rtl-sibling
13181 Dump after sibling call optimizations.
13183 @item -fdump-rtl-split1
13184 @itemx -fdump-rtl-split2
13185 @itemx -fdump-rtl-split3
13186 @itemx -fdump-rtl-split4
13187 @itemx -fdump-rtl-split5
13188 @opindex fdump-rtl-split1
13189 @opindex fdump-rtl-split2
13190 @opindex fdump-rtl-split3
13191 @opindex fdump-rtl-split4
13192 @opindex fdump-rtl-split5
13193 These options enable dumping after five rounds of
13194 instruction splitting.
13196 @item -fdump-rtl-sms
13197 @opindex fdump-rtl-sms
13198 Dump after modulo scheduling. This pass is only run on some
13201 @item -fdump-rtl-stack
13202 @opindex fdump-rtl-stack
13203 Dump after conversion from GCC's ``flat register file'' registers to the
13204 x87's stack-like registers. This pass is only run on x86 variants.
13206 @item -fdump-rtl-subreg1
13207 @itemx -fdump-rtl-subreg2
13208 @opindex fdump-rtl-subreg1
13209 @opindex fdump-rtl-subreg2
13210 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
13211 the two subreg expansion passes.
13213 @item -fdump-rtl-unshare
13214 @opindex fdump-rtl-unshare
13215 Dump after all rtl has been unshared.
13217 @item -fdump-rtl-vartrack
13218 @opindex fdump-rtl-vartrack
13219 Dump after variable tracking.
13221 @item -fdump-rtl-vregs
13222 @opindex fdump-rtl-vregs
13223 Dump after converting virtual registers to hard registers.
13225 @item -fdump-rtl-web
13226 @opindex fdump-rtl-web
13227 Dump after live range splitting.
13229 @item -fdump-rtl-regclass
13230 @itemx -fdump-rtl-subregs_of_mode_init
13231 @itemx -fdump-rtl-subregs_of_mode_finish
13232 @itemx -fdump-rtl-dfinit
13233 @itemx -fdump-rtl-dfinish
13234 @opindex fdump-rtl-regclass
13235 @opindex fdump-rtl-subregs_of_mode_init
13236 @opindex fdump-rtl-subregs_of_mode_finish
13237 @opindex fdump-rtl-dfinit
13238 @opindex fdump-rtl-dfinish
13239 These dumps are defined but always produce empty files.
13242 @itemx -fdump-rtl-all
13244 @opindex fdump-rtl-all
13245 Produce all the dumps listed above.
13249 Annotate the assembler output with miscellaneous debugging information.
13253 Dump all macro definitions, at the end of preprocessing, in addition to
13258 Produce a core dump whenever an error occurs.
13262 Annotate the assembler output with a comment indicating which
13263 pattern and alternative is used. The length of each instruction is
13268 Dump the RTL in the assembler output as a comment before each instruction.
13269 Also turns on @option{-dp} annotation.
13273 Just generate RTL for a function instead of compiling it. Usually used
13274 with @option{-fdump-rtl-expand}.
13277 @item -fdump-noaddr
13278 @opindex fdump-noaddr
13279 When doing debugging dumps, suppress address output. This makes it more
13280 feasible to use diff on debugging dumps for compiler invocations with
13281 different compiler binaries and/or different
13282 text / bss / data / heap / stack / dso start locations.
13285 @opindex freport-bug
13286 Collect and dump debug information into a temporary file if an
13287 internal compiler error (ICE) occurs.
13289 @item -fdump-unnumbered
13290 @opindex fdump-unnumbered
13291 When doing debugging dumps, suppress instruction numbers and address output.
13292 This makes it more feasible to use diff on debugging dumps for compiler
13293 invocations with different options, in particular with and without
13296 @item -fdump-unnumbered-links
13297 @opindex fdump-unnumbered-links
13298 When doing debugging dumps (see @option{-d} option above), suppress
13299 instruction numbers for the links to the previous and next instructions
13302 @item -fdump-ipa-@var{switch}
13304 Control the dumping at various stages of inter-procedural analysis
13305 language tree to a file. The file name is generated by appending a
13306 switch specific suffix to the source file name, and the file is created
13307 in the same directory as the output file. The following dumps are
13312 Enables all inter-procedural analysis dumps.
13315 Dumps information about call-graph optimization, unused function removal,
13316 and inlining decisions.
13319 Dump after function inlining.
13323 @item -fdump-lang-all
13324 @itemx -fdump-lang-@var{switch}
13325 @itemx -fdump-lang-@var{switch}-@var{options}
13326 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
13327 @opindex fdump-lang-all
13328 @opindex fdump-lang
13329 Control the dumping of language-specific information. The @var{options}
13330 and @var{filename} portions behave as described in the
13331 @option{-fdump-tree} option. The following @var{switch} values are
13337 Enable all language-specific dumps.
13340 Dump class hierarchy information. Virtual table information is emitted
13341 unless '@option{slim}' is specified. This option is applicable to C++ only.
13344 Dump the raw internal tree data. This option is applicable to C++ only.
13348 @item -fdump-passes
13349 @opindex fdump-passes
13350 Print on @file{stderr} the list of optimization passes that are turned
13351 on and off by the current command-line options.
13353 @item -fdump-statistics-@var{option}
13354 @opindex fdump-statistics
13355 Enable and control dumping of pass statistics in a separate file. The
13356 file name is generated by appending a suffix ending in
13357 @samp{.statistics} to the source file name, and the file is created in
13358 the same directory as the output file. If the @samp{-@var{option}}
13359 form is used, @samp{-stats} causes counters to be summed over the
13360 whole compilation unit while @samp{-details} dumps every event as
13361 the passes generate them. The default with no option is to sum
13362 counters for each function compiled.
13364 @item -fdump-tree-all
13365 @itemx -fdump-tree-@var{switch}
13366 @itemx -fdump-tree-@var{switch}-@var{options}
13367 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
13368 @opindex fdump-tree-all
13369 @opindex fdump-tree
13370 Control the dumping at various stages of processing the intermediate
13371 language tree to a file. The file name is generated by appending a
13372 switch-specific suffix to the source file name, and the file is
13373 created in the same directory as the output file. In case of
13374 @option{=@var{filename}} option, the dump is output on the given file
13375 instead of the auto named dump files. If the @samp{-@var{options}}
13376 form is used, @var{options} is a list of @samp{-} separated options
13377 which control the details of the dump. Not all options are applicable
13378 to all dumps; those that are not meaningful are ignored. The
13379 following options are available
13383 Print the address of each node. Usually this is not meaningful as it
13384 changes according to the environment and source file. Its primary use
13385 is for tying up a dump file with a debug environment.
13387 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
13388 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
13389 use working backward from mangled names in the assembly file.
13391 When dumping front-end intermediate representations, inhibit dumping
13392 of members of a scope or body of a function merely because that scope
13393 has been reached. Only dump such items when they are directly reachable
13394 by some other path.
13396 When dumping pretty-printed trees, this option inhibits dumping the
13397 bodies of control structures.
13399 When dumping RTL, print the RTL in slim (condensed) form instead of
13400 the default LISP-like representation.
13402 Print a raw representation of the tree. By default, trees are
13403 pretty-printed into a C-like representation.
13405 Enable more detailed dumps (not honored by every dump option). Also
13406 include information from the optimization passes.
13408 Enable dumping various statistics about the pass (not honored by every dump
13411 Enable showing basic block boundaries (disabled in raw dumps).
13413 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13414 dump a representation of the control flow graph suitable for viewing with
13415 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13416 the file is pretty-printed as a subgraph, so that GraphViz can render them
13417 all in a single plot.
13419 This option currently only works for RTL dumps, and the RTL is always
13420 dumped in slim form.
13422 Enable showing virtual operands for every statement.
13424 Enable showing line numbers for statements.
13426 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13428 Enable showing the tree dump for each statement.
13430 Enable showing the EH region number holding each statement.
13432 Enable showing scalar evolution analysis details.
13434 Enable showing optimization information (only available in certain
13437 Enable showing missed optimization information (only available in certain
13440 Enable other detailed optimization information (only available in
13442 @item =@var{filename}
13443 Instead of an auto named dump file, output into the given file
13444 name. The file names @file{stdout} and @file{stderr} are treated
13445 specially and are considered already open standard streams. For
13449 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13450 -fdump-tree-pre=/dev/stderr file.c
13453 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13454 output on to @file{stderr}. If two conflicting dump filenames are
13455 given for the same pass, then the latter option overrides the earlier
13459 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13460 and @option{lineno}.
13463 Turn on all optimization options, i.e., @option{optimized},
13464 @option{missed}, and @option{note}.
13467 To determine what tree dumps are available or find the dump for a pass
13468 of interest follow the steps below.
13472 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
13473 look for a code that corresponds to the pass you are interested in.
13474 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
13475 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
13476 The number at the end distinguishes distinct invocations of the same pass.
13478 To enable the creation of the dump file, append the pass code to
13479 the @option{-fdump-} option prefix and invoke GCC with it. For example,
13480 to enable the dump from the Early Value Range Propagation pass, invoke
13481 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
13482 specify the name of the dump file. If you don't specify one, GCC
13483 creates as described below.
13485 Find the pass dump in a file whose name is composed of three components
13486 separated by a period: the name of the source file GCC was invoked to
13487 compile, a numeric suffix indicating the pass number followed by the
13488 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
13489 and finally the pass code. For example, the Early VRP pass dump might
13490 be in a file named @file{myfile.c.038t.evrp} in the current working
13491 directory. Note that the numeric codes are not stable and may change
13492 from one version of GCC to another.
13496 @itemx -fopt-info-@var{options}
13497 @itemx -fopt-info-@var{options}=@var{filename}
13499 Controls optimization dumps from various optimization passes. If the
13500 @samp{-@var{options}} form is used, @var{options} is a list of
13501 @samp{-} separated option keywords to select the dump details and
13504 The @var{options} can be divided into two groups: options describing the
13505 verbosity of the dump, and options describing which optimizations
13506 should be included. The options from both the groups can be freely
13507 mixed as they are non-overlapping. However, in case of any conflicts,
13508 the later options override the earlier options on the command
13511 The following options control the dump verbosity:
13515 Print information when an optimization is successfully applied. It is
13516 up to a pass to decide which information is relevant. For example, the
13517 vectorizer passes print the source location of loops which are
13518 successfully vectorized.
13520 Print information about missed optimizations. Individual passes
13521 control which information to include in the output.
13523 Print verbose information about optimizations, such as certain
13524 transformations, more detailed messages about decisions etc.
13526 Print detailed optimization information. This includes
13527 @samp{optimized}, @samp{missed}, and @samp{note}.
13530 One or more of the following option keywords can be used to describe a
13531 group of optimizations:
13535 Enable dumps from all interprocedural optimizations.
13537 Enable dumps from all loop optimizations.
13539 Enable dumps from all inlining optimizations.
13541 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
13543 Enable dumps from all vectorization optimizations.
13545 Enable dumps from all optimizations. This is a superset of
13546 the optimization groups listed above.
13549 If @var{options} is
13550 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13551 info about successful optimizations from all the passes.
13553 If the @var{filename} is provided, then the dumps from all the
13554 applicable optimizations are concatenated into the @var{filename}.
13555 Otherwise the dump is output onto @file{stderr}. Though multiple
13556 @option{-fopt-info} options are accepted, only one of them can include
13557 a @var{filename}. If other filenames are provided then all but the
13558 first such option are ignored.
13560 Note that the output @var{filename} is overwritten
13561 in case of multiple translation units. If a combined output from
13562 multiple translation units is desired, @file{stderr} should be used
13565 In the following example, the optimization info is output to
13574 gcc -O3 -fopt-info-missed=missed.all
13578 outputs missed optimization report from all the passes into
13579 @file{missed.all}, and this one:
13582 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13586 prints information about missed optimization opportunities from
13587 vectorization passes on @file{stderr}.
13588 Note that @option{-fopt-info-vec-missed} is equivalent to
13589 @option{-fopt-info-missed-vec}. The order of the optimization group
13590 names and message types listed after @option{-fopt-info} does not matter.
13592 As another example,
13594 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13598 outputs information about missed optimizations as well as
13599 optimized locations from all the inlining passes into
13605 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13609 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13610 in conflict since only one output file is allowed. In this case, only
13611 the first option takes effect and the subsequent options are
13612 ignored. Thus only @file{vec.miss} is produced which contains
13613 dumps from the vectorizer about missed opportunities.
13615 @item -fsched-verbose=@var{n}
13616 @opindex fsched-verbose
13617 On targets that use instruction scheduling, this option controls the
13618 amount of debugging output the scheduler prints to the dump files.
13620 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13621 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13622 For @var{n} greater than one, it also output basic block probabilities,
13623 detailed ready list information and unit/insn info. For @var{n} greater
13624 than two, it includes RTL at abort point, control-flow and regions info.
13625 And for @var{n} over four, @option{-fsched-verbose} also includes
13630 @item -fenable-@var{kind}-@var{pass}
13631 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13635 This is a set of options that are used to explicitly disable/enable
13636 optimization passes. These options are intended for use for debugging GCC.
13637 Compiler users should use regular options for enabling/disabling
13642 @item -fdisable-ipa-@var{pass}
13643 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13644 statically invoked in the compiler multiple times, the pass name should be
13645 appended with a sequential number starting from 1.
13647 @item -fdisable-rtl-@var{pass}
13648 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13649 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13650 statically invoked in the compiler multiple times, the pass name should be
13651 appended with a sequential number starting from 1. @var{range-list} is a
13652 comma-separated list of function ranges or assembler names. Each range is a number
13653 pair separated by a colon. The range is inclusive in both ends. If the range
13654 is trivial, the number pair can be simplified as a single number. If the
13655 function's call graph node's @var{uid} falls within one of the specified ranges,
13656 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13657 function header of a dump file, and the pass names can be dumped by using
13658 option @option{-fdump-passes}.
13660 @item -fdisable-tree-@var{pass}
13661 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13662 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13665 @item -fenable-ipa-@var{pass}
13666 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13667 statically invoked in the compiler multiple times, the pass name should be
13668 appended with a sequential number starting from 1.
13670 @item -fenable-rtl-@var{pass}
13671 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13672 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13673 description and examples.
13675 @item -fenable-tree-@var{pass}
13676 @itemx -fenable-tree-@var{pass}=@var{range-list}
13677 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13678 of option arguments.
13682 Here are some examples showing uses of these options.
13686 # disable ccp1 for all functions
13687 -fdisable-tree-ccp1
13688 # disable complete unroll for function whose cgraph node uid is 1
13689 -fenable-tree-cunroll=1
13690 # disable gcse2 for functions at the following ranges [1,1],
13691 # [300,400], and [400,1000]
13692 # disable gcse2 for functions foo and foo2
13693 -fdisable-rtl-gcse2=foo,foo2
13694 # disable early inlining
13695 -fdisable-tree-einline
13696 # disable ipa inlining
13697 -fdisable-ipa-inline
13698 # enable tree full unroll
13699 -fenable-tree-unroll
13704 @itemx -fchecking=@var{n}
13706 @opindex fno-checking
13707 Enable internal consistency checking. The default depends on
13708 the compiler configuration. @option{-fchecking=2} enables further
13709 internal consistency checking that might affect code generation.
13711 @item -frandom-seed=@var{string}
13712 @opindex frandom-seed
13713 This option provides a seed that GCC uses in place of
13714 random numbers in generating certain symbol names
13715 that have to be different in every compiled file. It is also used to
13716 place unique stamps in coverage data files and the object files that
13717 produce them. You can use the @option{-frandom-seed} option to produce
13718 reproducibly identical object files.
13720 The @var{string} can either be a number (decimal, octal or hex) or an
13721 arbitrary string (in which case it's converted to a number by
13724 The @var{string} should be different for every file you compile.
13727 @itemx -save-temps=cwd
13728 @opindex save-temps
13729 Store the usual ``temporary'' intermediate files permanently; place them
13730 in the current directory and name them based on the source file. Thus,
13731 compiling @file{foo.c} with @option{-c -save-temps} produces files
13732 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13733 preprocessed @file{foo.i} output file even though the compiler now
13734 normally uses an integrated preprocessor.
13736 When used in combination with the @option{-x} command-line option,
13737 @option{-save-temps} is sensible enough to avoid over writing an
13738 input source file with the same extension as an intermediate file.
13739 The corresponding intermediate file may be obtained by renaming the
13740 source file before using @option{-save-temps}.
13742 If you invoke GCC in parallel, compiling several different source
13743 files that share a common base name in different subdirectories or the
13744 same source file compiled for multiple output destinations, it is
13745 likely that the different parallel compilers will interfere with each
13746 other, and overwrite the temporary files. For instance:
13749 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13750 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13753 may result in @file{foo.i} and @file{foo.o} being written to
13754 simultaneously by both compilers.
13756 @item -save-temps=obj
13757 @opindex save-temps=obj
13758 Store the usual ``temporary'' intermediate files permanently. If the
13759 @option{-o} option is used, the temporary files are based on the
13760 object file. If the @option{-o} option is not used, the
13761 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13766 gcc -save-temps=obj -c foo.c
13767 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13768 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13772 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13773 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13774 @file{dir2/yfoobar.o}.
13776 @item -time@r{[}=@var{file}@r{]}
13778 Report the CPU time taken by each subprocess in the compilation
13779 sequence. For C source files, this is the compiler proper and assembler
13780 (plus the linker if linking is done).
13782 Without the specification of an output file, the output looks like this:
13789 The first number on each line is the ``user time'', that is time spent
13790 executing the program itself. The second number is ``system time'',
13791 time spent executing operating system routines on behalf of the program.
13792 Both numbers are in seconds.
13794 With the specification of an output file, the output is appended to the
13795 named file, and it looks like this:
13798 0.12 0.01 cc1 @var{options}
13799 0.00 0.01 as @var{options}
13802 The ``user time'' and the ``system time'' are moved before the program
13803 name, and the options passed to the program are displayed, so that one
13804 can later tell what file was being compiled, and with which options.
13806 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13807 @opindex fdump-final-insns
13808 Dump the final internal representation (RTL) to @var{file}. If the
13809 optional argument is omitted (or if @var{file} is @code{.}), the name
13810 of the dump file is determined by appending @code{.gkd} to the
13811 compilation output file name.
13813 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13814 @opindex fcompare-debug
13815 @opindex fno-compare-debug
13816 If no error occurs during compilation, run the compiler a second time,
13817 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13818 passed to the second compilation. Dump the final internal
13819 representation in both compilations, and print an error if they differ.
13821 If the equal sign is omitted, the default @option{-gtoggle} is used.
13823 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13824 and nonzero, implicitly enables @option{-fcompare-debug}. If
13825 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13826 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13829 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13830 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13831 of the final representation and the second compilation, preventing even
13832 @env{GCC_COMPARE_DEBUG} from taking effect.
13834 To verify full coverage during @option{-fcompare-debug} testing, set
13835 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13836 which GCC rejects as an invalid option in any actual compilation
13837 (rather than preprocessing, assembly or linking). To get just a
13838 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13839 not overridden} will do.
13841 @item -fcompare-debug-second
13842 @opindex fcompare-debug-second
13843 This option is implicitly passed to the compiler for the second
13844 compilation requested by @option{-fcompare-debug}, along with options to
13845 silence warnings, and omitting other options that would cause
13846 side-effect compiler outputs to files or to the standard output. Dump
13847 files and preserved temporary files are renamed so as to contain the
13848 @code{.gk} additional extension during the second compilation, to avoid
13849 overwriting those generated by the first.
13851 When this option is passed to the compiler driver, it causes the
13852 @emph{first} compilation to be skipped, which makes it useful for little
13853 other than debugging the compiler proper.
13857 Turn off generation of debug info, if leaving out this option
13858 generates it, or turn it on at level 2 otherwise. The position of this
13859 argument in the command line does not matter; it takes effect after all
13860 other options are processed, and it does so only once, no matter how
13861 many times it is given. This is mainly intended to be used with
13862 @option{-fcompare-debug}.
13864 @item -fvar-tracking-assignments-toggle
13865 @opindex fvar-tracking-assignments-toggle
13866 @opindex fno-var-tracking-assignments-toggle
13867 Toggle @option{-fvar-tracking-assignments}, in the same way that
13868 @option{-gtoggle} toggles @option{-g}.
13872 Makes the compiler print out each function name as it is compiled, and
13873 print some statistics about each pass when it finishes.
13875 @item -ftime-report
13876 @opindex ftime-report
13877 Makes the compiler print some statistics about the time consumed by each
13878 pass when it finishes.
13880 @item -ftime-report-details
13881 @opindex ftime-report-details
13882 Record the time consumed by infrastructure parts separately for each pass.
13884 @item -fira-verbose=@var{n}
13885 @opindex fira-verbose
13886 Control the verbosity of the dump file for the integrated register allocator.
13887 The default value is 5. If the value @var{n} is greater or equal to 10,
13888 the dump output is sent to stderr using the same format as @var{n} minus 10.
13891 @opindex flto-report
13892 Prints a report with internal details on the workings of the link-time
13893 optimizer. The contents of this report vary from version to version.
13894 It is meant to be useful to GCC developers when processing object
13895 files in LTO mode (via @option{-flto}).
13897 Disabled by default.
13899 @item -flto-report-wpa
13900 @opindex flto-report-wpa
13901 Like @option{-flto-report}, but only print for the WPA phase of Link
13905 @opindex fmem-report
13906 Makes the compiler print some statistics about permanent memory
13907 allocation when it finishes.
13909 @item -fmem-report-wpa
13910 @opindex fmem-report-wpa
13911 Makes the compiler print some statistics about permanent memory
13912 allocation for the WPA phase only.
13914 @item -fpre-ipa-mem-report
13915 @opindex fpre-ipa-mem-report
13916 @item -fpost-ipa-mem-report
13917 @opindex fpost-ipa-mem-report
13918 Makes the compiler print some statistics about permanent memory
13919 allocation before or after interprocedural optimization.
13921 @item -fprofile-report
13922 @opindex fprofile-report
13923 Makes the compiler print some statistics about consistency of the
13924 (estimated) profile and effect of individual passes.
13926 @item -fstack-usage
13927 @opindex fstack-usage
13928 Makes the compiler output stack usage information for the program, on a
13929 per-function basis. The filename for the dump is made by appending
13930 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13931 the output file, if explicitly specified and it is not an executable,
13932 otherwise it is the basename of the source file. An entry is made up
13937 The name of the function.
13941 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
13944 The qualifier @code{static} means that the function manipulates the stack
13945 statically: a fixed number of bytes are allocated for the frame on function
13946 entry and released on function exit; no stack adjustments are otherwise made
13947 in the function. The second field is this fixed number of bytes.
13949 The qualifier @code{dynamic} means that the function manipulates the stack
13950 dynamically: in addition to the static allocation described above, stack
13951 adjustments are made in the body of the function, for example to push/pop
13952 arguments around function calls. If the qualifier @code{bounded} is also
13953 present, the amount of these adjustments is bounded at compile time and
13954 the second field is an upper bound of the total amount of stack used by
13955 the function. If it is not present, the amount of these adjustments is
13956 not bounded at compile time and the second field only represents the
13961 Emit statistics about front-end processing at the end of the compilation.
13962 This option is supported only by the C++ front end, and
13963 the information is generally only useful to the G++ development team.
13965 @item -fdbg-cnt-list
13966 @opindex fdbg-cnt-list
13967 Print the name and the counter upper bound for all debug counters.
13970 @item -fdbg-cnt=@var{counter-value-list}
13972 Set the internal debug counter upper bound. @var{counter-value-list}
13973 is a comma-separated list of @var{name}:@var{value} pairs
13974 which sets the upper bound of each debug counter @var{name} to @var{value}.
13975 All debug counters have the initial upper bound of @code{UINT_MAX};
13976 thus @code{dbg_cnt} returns true always unless the upper bound
13977 is set by this option.
13978 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
13979 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
13981 @item -print-file-name=@var{library}
13982 @opindex print-file-name
13983 Print the full absolute name of the library file @var{library} that
13984 would be used when linking---and don't do anything else. With this
13985 option, GCC does not compile or link anything; it just prints the
13988 @item -print-multi-directory
13989 @opindex print-multi-directory
13990 Print the directory name corresponding to the multilib selected by any
13991 other switches present in the command line. This directory is supposed
13992 to exist in @env{GCC_EXEC_PREFIX}.
13994 @item -print-multi-lib
13995 @opindex print-multi-lib
13996 Print the mapping from multilib directory names to compiler switches
13997 that enable them. The directory name is separated from the switches by
13998 @samp{;}, and each switch starts with an @samp{@@} instead of the
13999 @samp{-}, without spaces between multiple switches. This is supposed to
14000 ease shell processing.
14002 @item -print-multi-os-directory
14003 @opindex print-multi-os-directory
14004 Print the path to OS libraries for the selected
14005 multilib, relative to some @file{lib} subdirectory. If OS libraries are
14006 present in the @file{lib} subdirectory and no multilibs are used, this is
14007 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
14008 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
14009 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
14010 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
14012 @item -print-multiarch
14013 @opindex print-multiarch
14014 Print the path to OS libraries for the selected multiarch,
14015 relative to some @file{lib} subdirectory.
14017 @item -print-prog-name=@var{program}
14018 @opindex print-prog-name
14019 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
14021 @item -print-libgcc-file-name
14022 @opindex print-libgcc-file-name
14023 Same as @option{-print-file-name=libgcc.a}.
14025 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
14026 but you do want to link with @file{libgcc.a}. You can do:
14029 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
14032 @item -print-search-dirs
14033 @opindex print-search-dirs
14034 Print the name of the configured installation directory and a list of
14035 program and library directories @command{gcc} searches---and don't do anything else.
14037 This is useful when @command{gcc} prints the error message
14038 @samp{installation problem, cannot exec cpp0: No such file or directory}.
14039 To resolve this you either need to put @file{cpp0} and the other compiler
14040 components where @command{gcc} expects to find them, or you can set the environment
14041 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
14042 Don't forget the trailing @samp{/}.
14043 @xref{Environment Variables}.
14045 @item -print-sysroot
14046 @opindex print-sysroot
14047 Print the target sysroot directory that is used during
14048 compilation. This is the target sysroot specified either at configure
14049 time or using the @option{--sysroot} option, possibly with an extra
14050 suffix that depends on compilation options. If no target sysroot is
14051 specified, the option prints nothing.
14053 @item -print-sysroot-headers-suffix
14054 @opindex print-sysroot-headers-suffix
14055 Print the suffix added to the target sysroot when searching for
14056 headers, or give an error if the compiler is not configured with such
14057 a suffix---and don't do anything else.
14060 @opindex dumpmachine
14061 Print the compiler's target machine (for example,
14062 @samp{i686-pc-linux-gnu})---and don't do anything else.
14065 @opindex dumpversion
14066 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
14067 anything else. This is the compiler version used in filesystem paths,
14068 specs, can be depending on how the compiler has been configured just
14069 a single number (major version), two numbers separated by dot (major and
14070 minor version) or three numbers separated by dots (major, minor and patchlevel
14073 @item -dumpfullversion
14074 @opindex dumpfullversion
14075 Print the full compiler version, always 3 numbers separated by dots,
14076 major, minor and patchlevel version.
14080 Print the compiler's built-in specs---and don't do anything else. (This
14081 is used when GCC itself is being built.) @xref{Spec Files}.
14084 @node Submodel Options
14085 @section Machine-Dependent Options
14086 @cindex submodel options
14087 @cindex specifying hardware config
14088 @cindex hardware models and configurations, specifying
14089 @cindex target-dependent options
14090 @cindex machine-dependent options
14092 Each target machine supported by GCC can have its own options---for
14093 example, to allow you to compile for a particular processor variant or
14094 ABI, or to control optimizations specific to that machine. By
14095 convention, the names of machine-specific options start with
14098 Some configurations of the compiler also support additional target-specific
14099 options, usually for compatibility with other compilers on the same
14102 @c This list is ordered alphanumerically by subsection name.
14103 @c It should be the same order and spelling as these options are listed
14104 @c in Machine Dependent Options
14107 * AArch64 Options::
14108 * Adapteva Epiphany Options::
14112 * Blackfin Options::
14117 * DEC Alpha Options::
14121 * GNU/Linux Options::
14131 * MicroBlaze Options::
14134 * MN10300 Options::
14138 * Nios II Options::
14139 * Nvidia PTX Options::
14141 * picoChip Options::
14142 * PowerPC Options::
14145 * RS/6000 and PowerPC Options::
14147 * S/390 and zSeries Options::
14150 * Solaris 2 Options::
14153 * System V Options::
14154 * TILE-Gx Options::
14155 * TILEPro Options::
14160 * VxWorks Options::
14162 * x86 Windows Options::
14163 * Xstormy16 Options::
14165 * zSeries Options::
14168 @node AArch64 Options
14169 @subsection AArch64 Options
14170 @cindex AArch64 Options
14172 These options are defined for AArch64 implementations:
14176 @item -mabi=@var{name}
14178 Generate code for the specified data model. Permissible values
14179 are @samp{ilp32} for SysV-like data model where int, long int and pointers
14180 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
14181 but long int and pointers are 64 bits.
14183 The default depends on the specific target configuration. Note that
14184 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
14185 entire program with the same ABI, and link with a compatible set of libraries.
14188 @opindex mbig-endian
14189 Generate big-endian code. This is the default when GCC is configured for an
14190 @samp{aarch64_be-*-*} target.
14192 @item -mgeneral-regs-only
14193 @opindex mgeneral-regs-only
14194 Generate code which uses only the general-purpose registers. This will prevent
14195 the compiler from using floating-point and Advanced SIMD registers but will not
14196 impose any restrictions on the assembler.
14198 @item -mlittle-endian
14199 @opindex mlittle-endian
14200 Generate little-endian code. This is the default when GCC is configured for an
14201 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
14203 @item -mcmodel=tiny
14204 @opindex mcmodel=tiny
14205 Generate code for the tiny code model. The program and its statically defined
14206 symbols must be within 1MB of each other. Programs can be statically or
14207 dynamically linked.
14209 @item -mcmodel=small
14210 @opindex mcmodel=small
14211 Generate code for the small code model. The program and its statically defined
14212 symbols must be within 4GB of each other. Programs can be statically or
14213 dynamically linked. This is the default code model.
14215 @item -mcmodel=large
14216 @opindex mcmodel=large
14217 Generate code for the large code model. This makes no assumptions about
14218 addresses and sizes of sections. Programs can be statically linked only.
14220 @item -mstrict-align
14221 @opindex mstrict-align
14222 Avoid generating memory accesses that may not be aligned on a natural object
14223 boundary as described in the architecture specification.
14225 @item -momit-leaf-frame-pointer
14226 @itemx -mno-omit-leaf-frame-pointer
14227 @opindex momit-leaf-frame-pointer
14228 @opindex mno-omit-leaf-frame-pointer
14229 Omit or keep the frame pointer in leaf functions. The former behavior is the
14232 @item -mtls-dialect=desc
14233 @opindex mtls-dialect=desc
14234 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
14235 of TLS variables. This is the default.
14237 @item -mtls-dialect=traditional
14238 @opindex mtls-dialect=traditional
14239 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
14242 @item -mtls-size=@var{size}
14244 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
14245 This option requires binutils 2.26 or newer.
14247 @item -mfix-cortex-a53-835769
14248 @itemx -mno-fix-cortex-a53-835769
14249 @opindex mfix-cortex-a53-835769
14250 @opindex mno-fix-cortex-a53-835769
14251 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
14252 This involves inserting a NOP instruction between memory instructions and
14253 64-bit integer multiply-accumulate instructions.
14255 @item -mfix-cortex-a53-843419
14256 @itemx -mno-fix-cortex-a53-843419
14257 @opindex mfix-cortex-a53-843419
14258 @opindex mno-fix-cortex-a53-843419
14259 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
14260 This erratum workaround is made at link time and this will only pass the
14261 corresponding flag to the linker.
14263 @item -mlow-precision-recip-sqrt
14264 @item -mno-low-precision-recip-sqrt
14265 @opindex mlow-precision-recip-sqrt
14266 @opindex mno-low-precision-recip-sqrt
14267 Enable or disable the reciprocal square root approximation.
14268 This option only has an effect if @option{-ffast-math} or
14269 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14270 precision of reciprocal square root results to about 16 bits for
14271 single precision and to 32 bits for double precision.
14273 @item -mlow-precision-sqrt
14274 @item -mno-low-precision-sqrt
14275 @opindex -mlow-precision-sqrt
14276 @opindex -mno-low-precision-sqrt
14277 Enable or disable the square root approximation.
14278 This option only has an effect if @option{-ffast-math} or
14279 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14280 precision of square root results to about 16 bits for
14281 single precision and to 32 bits for double precision.
14282 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
14284 @item -mlow-precision-div
14285 @item -mno-low-precision-div
14286 @opindex -mlow-precision-div
14287 @opindex -mno-low-precision-div
14288 Enable or disable the division approximation.
14289 This option only has an effect if @option{-ffast-math} or
14290 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14291 precision of division results to about 16 bits for
14292 single precision and to 32 bits for double precision.
14294 @item -march=@var{name}
14296 Specify the name of the target architecture and, optionally, one or
14297 more feature modifiers. This option has the form
14298 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
14300 The permissible values for @var{arch} are @samp{armv8-a},
14301 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @var{native}.
14303 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
14304 support for the ARMv8.3-A architecture extensions.
14306 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
14307 support for the ARMv8.2-A architecture extensions.
14309 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
14310 support for the ARMv8.1-A architecture extension. In particular, it
14311 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
14313 The value @samp{native} is available on native AArch64 GNU/Linux and
14314 causes the compiler to pick the architecture of the host system. This
14315 option has no effect if the compiler is unable to recognize the
14316 architecture of the host system,
14318 The permissible values for @var{feature} are listed in the sub-section
14319 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14320 Feature Modifiers}. Where conflicting feature modifiers are
14321 specified, the right-most feature is used.
14323 GCC uses @var{name} to determine what kind of instructions it can emit
14324 when generating assembly code. If @option{-march} is specified
14325 without either of @option{-mtune} or @option{-mcpu} also being
14326 specified, the code is tuned to perform well across a range of target
14327 processors implementing the target architecture.
14329 @item -mtune=@var{name}
14331 Specify the name of the target processor for which GCC should tune the
14332 performance of the code. Permissible values for this option are:
14333 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
14334 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
14335 @samp{exynos-m1}, @samp{falkor}, @samp{qdf24xx},
14336 @samp{xgene1}, @samp{vulcan}, @samp{thunderx},
14337 @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
14338 @samp{thunderxt83}, @samp{thunderx2t99}, @samp{cortex-a57.cortex-a53},
14339 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
14340 @samp{cortex-a73.cortex-a53}, @samp{cortex-a75.cortex-a55},
14343 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14344 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
14345 @samp{cortex-a75.cortex-a55} specify that GCC should tune for a
14348 Additionally on native AArch64 GNU/Linux systems the value
14349 @samp{native} tunes performance to the host system. This option has no effect
14350 if the compiler is unable to recognize the processor of the host system.
14352 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
14353 are specified, the code is tuned to perform well across a range
14354 of target processors.
14356 This option cannot be suffixed by feature modifiers.
14358 @item -mcpu=@var{name}
14360 Specify the name of the target processor, optionally suffixed by one
14361 or more feature modifiers. This option has the form
14362 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
14363 the permissible values for @var{cpu} are the same as those available
14364 for @option{-mtune}. The permissible values for @var{feature} are
14365 documented in the sub-section on
14366 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14367 Feature Modifiers}. Where conflicting feature modifiers are
14368 specified, the right-most feature is used.
14370 GCC uses @var{name} to determine what kind of instructions it can emit when
14371 generating assembly code (as if by @option{-march}) and to determine
14372 the target processor for which to tune for performance (as if
14373 by @option{-mtune}). Where this option is used in conjunction
14374 with @option{-march} or @option{-mtune}, those options take precedence
14375 over the appropriate part of this option.
14377 @item -moverride=@var{string}
14379 Override tuning decisions made by the back-end in response to a
14380 @option{-mtune=} switch. The syntax, semantics, and accepted values
14381 for @var{string} in this option are not guaranteed to be consistent
14384 This option is only intended to be useful when developing GCC.
14386 @item -mpc-relative-literal-loads
14387 @itemx -mno-pc-relative-literal-loads
14388 @opindex mpc-relative-literal-loads
14389 @opindex mno-pc-relative-literal-loads
14390 Enable or disable PC-relative literal loads. With this option literal pools are
14391 accessed using a single instruction and emitted after each function. This
14392 limits the maximum size of functions to 1MB. This is enabled by default for
14393 @option{-mcmodel=tiny}.
14395 @item -msign-return-address=@var{scope}
14396 @opindex msign-return-address
14397 Select the function scope on which return address signing will be applied.
14398 Permissible values are @samp{none}, which disables return address signing,
14399 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
14400 functions, and @samp{all}, which enables pointer signing for all functions. The
14401 default value is @samp{none}.
14405 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14406 @anchor{aarch64-feature-modifiers}
14407 @cindex @option{-march} feature modifiers
14408 @cindex @option{-mcpu} feature modifiers
14409 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14410 the following and their inverses @option{no@var{feature}}:
14414 Enable CRC extension. This is on by default for
14415 @option{-march=armv8.1-a}.
14417 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14420 Enable floating-point instructions. This is on by default for all possible
14421 values for options @option{-march} and @option{-mcpu}.
14423 Enable Advanced SIMD instructions. This also enables floating-point
14424 instructions. This is on by default for all possible values for options
14425 @option{-march} and @option{-mcpu}.
14427 Enable Large System Extension instructions. This is on by default for
14428 @option{-march=armv8.1-a}.
14430 Enable Round Double Multiply Accumulate instructions. This is on by default
14431 for @option{-march=armv8.1-a}.
14433 Enable FP16 extension. This also enables floating-point instructions.
14435 Enable the RcPc extension. This does not change code generation from GCC,
14436 but is passed on to the assembler, enabling inline asm statements to use
14437 instructions from the RcPc extension.
14439 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
14443 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
14444 Conversely, @option{nofp} implies @option{nosimd}, which implies
14447 @node Adapteva Epiphany Options
14448 @subsection Adapteva Epiphany Options
14450 These @samp{-m} options are defined for Adapteva Epiphany:
14453 @item -mhalf-reg-file
14454 @opindex mhalf-reg-file
14455 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14456 That allows code to run on hardware variants that lack these registers.
14458 @item -mprefer-short-insn-regs
14459 @opindex mprefer-short-insn-regs
14460 Preferentially allocate registers that allow short instruction generation.
14461 This can result in increased instruction count, so this may either reduce or
14462 increase overall code size.
14464 @item -mbranch-cost=@var{num}
14465 @opindex mbranch-cost
14466 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14467 This cost is only a heuristic and is not guaranteed to produce
14468 consistent results across releases.
14472 Enable the generation of conditional moves.
14474 @item -mnops=@var{num}
14476 Emit @var{num} NOPs before every other generated instruction.
14478 @item -mno-soft-cmpsf
14479 @opindex mno-soft-cmpsf
14480 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14481 and test the flags. This is faster than a software comparison, but can
14482 get incorrect results in the presence of NaNs, or when two different small
14483 numbers are compared such that their difference is calculated as zero.
14484 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14485 software comparisons.
14487 @item -mstack-offset=@var{num}
14488 @opindex mstack-offset
14489 Set the offset between the top of the stack and the stack pointer.
14490 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14491 can be used by leaf functions without stack allocation.
14492 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14493 Note also that this option changes the ABI; compiling a program with a
14494 different stack offset than the libraries have been compiled with
14495 generally does not work.
14496 This option can be useful if you want to evaluate if a different stack
14497 offset would give you better code, but to actually use a different stack
14498 offset to build working programs, it is recommended to configure the
14499 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14501 @item -mno-round-nearest
14502 @opindex mno-round-nearest
14503 Make the scheduler assume that the rounding mode has been set to
14504 truncating. The default is @option{-mround-nearest}.
14507 @opindex mlong-calls
14508 If not otherwise specified by an attribute, assume all calls might be beyond
14509 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14510 function address into a register before performing a (otherwise direct) call.
14511 This is the default.
14513 @item -mshort-calls
14514 @opindex short-calls
14515 If not otherwise specified by an attribute, assume all direct calls are
14516 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14517 for direct calls. The default is @option{-mlong-calls}.
14521 Assume addresses can be loaded as 16-bit unsigned values. This does not
14522 apply to function addresses for which @option{-mlong-calls} semantics
14525 @item -mfp-mode=@var{mode}
14527 Set the prevailing mode of the floating-point unit.
14528 This determines the floating-point mode that is provided and expected
14529 at function call and return time. Making this mode match the mode you
14530 predominantly need at function start can make your programs smaller and
14531 faster by avoiding unnecessary mode switches.
14533 @var{mode} can be set to one the following values:
14537 Any mode at function entry is valid, and retained or restored when
14538 the function returns, and when it calls other functions.
14539 This mode is useful for compiling libraries or other compilation units
14540 you might want to incorporate into different programs with different
14541 prevailing FPU modes, and the convenience of being able to use a single
14542 object file outweighs the size and speed overhead for any extra
14543 mode switching that might be needed, compared with what would be needed
14544 with a more specific choice of prevailing FPU mode.
14547 This is the mode used for floating-point calculations with
14548 truncating (i.e.@: round towards zero) rounding mode. That includes
14549 conversion from floating point to integer.
14551 @item round-nearest
14552 This is the mode used for floating-point calculations with
14553 round-to-nearest-or-even rounding mode.
14556 This is the mode used to perform integer calculations in the FPU, e.g.@:
14557 integer multiply, or integer multiply-and-accumulate.
14560 The default is @option{-mfp-mode=caller}
14562 @item -mnosplit-lohi
14563 @itemx -mno-postinc
14564 @itemx -mno-postmodify
14565 @opindex mnosplit-lohi
14566 @opindex mno-postinc
14567 @opindex mno-postmodify
14568 Code generation tweaks that disable, respectively, splitting of 32-bit
14569 loads, generation of post-increment addresses, and generation of
14570 post-modify addresses. The defaults are @option{msplit-lohi},
14571 @option{-mpost-inc}, and @option{-mpost-modify}.
14573 @item -mnovect-double
14574 @opindex mno-vect-double
14575 Change the preferred SIMD mode to SImode. The default is
14576 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14578 @item -max-vect-align=@var{num}
14579 @opindex max-vect-align
14580 The maximum alignment for SIMD vector mode types.
14581 @var{num} may be 4 or 8. The default is 8.
14582 Note that this is an ABI change, even though many library function
14583 interfaces are unaffected if they don't use SIMD vector modes
14584 in places that affect size and/or alignment of relevant types.
14586 @item -msplit-vecmove-early
14587 @opindex msplit-vecmove-early
14588 Split vector moves into single word moves before reload. In theory this
14589 can give better register allocation, but so far the reverse seems to be
14590 generally the case.
14592 @item -m1reg-@var{reg}
14594 Specify a register to hold the constant @minus{}1, which makes loading small negative
14595 constants and certain bitmasks faster.
14596 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14597 which specify use of that register as a fixed register,
14598 and @samp{none}, which means that no register is used for this
14599 purpose. The default is @option{-m1reg-none}.
14604 @subsection ARC Options
14605 @cindex ARC options
14607 The following options control the architecture variant for which code
14610 @c architecture variants
14613 @item -mbarrel-shifter
14614 @opindex mbarrel-shifter
14615 Generate instructions supported by barrel shifter. This is the default
14616 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14618 @item -mcpu=@var{cpu}
14620 Set architecture type, register usage, and instruction scheduling
14621 parameters for @var{cpu}. There are also shortcut alias options
14622 available for backward compatibility and convenience. Supported
14623 values for @var{cpu} are
14629 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14633 Compile for ARC601. Alias: @option{-mARC601}.
14638 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14639 This is the default when configured with @option{--with-cpu=arc700}@.
14642 Compile for ARC EM.
14645 Compile for ARC HS.
14648 Compile for ARC EM CPU with no hardware extensions.
14651 Compile for ARC EM4 CPU.
14654 Compile for ARC EM4 DMIPS CPU.
14657 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
14661 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
14662 double assist instructions.
14665 Compile for ARC HS CPU with no hardware extensions except the atomic
14669 Compile for ARC HS34 CPU.
14672 Compile for ARC HS38 CPU.
14675 Compile for ARC HS38 CPU with all hardware extensions on.
14678 Compile for ARC 600 CPU with @code{norm} instructions enabled.
14680 @item arc600_mul32x16
14681 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
14682 instructions enabled.
14685 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
14686 instructions enabled.
14689 Compile for ARC 601 CPU with @code{norm} instructions enabled.
14691 @item arc601_mul32x16
14692 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
14693 instructions enabled.
14696 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
14697 instructions enabled.
14700 Compile for ARC 700 on NPS400 chip.
14706 @itemx -mdpfp-compact
14707 @opindex mdpfp-compact
14708 Generate double-precision FPX instructions, tuned for the compact
14712 @opindex mdpfp-fast
14713 Generate double-precision FPX instructions, tuned for the fast
14716 @item -mno-dpfp-lrsr
14717 @opindex mno-dpfp-lrsr
14718 Disable @code{lr} and @code{sr} instructions from using FPX extension
14723 Generate extended arithmetic instructions. Currently only
14724 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14725 supported. This is always enabled for @option{-mcpu=ARC700}.
14729 Do not generate @code{mpy}-family instructions for ARC700. This option is
14734 Generate 32x16-bit multiply and multiply-accumulate instructions.
14738 Generate @code{mul64} and @code{mulu64} instructions.
14739 Only valid for @option{-mcpu=ARC600}.
14743 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
14748 @itemx -mspfp-compact
14749 @opindex mspfp-compact
14750 Generate single-precision FPX instructions, tuned for the compact
14754 @opindex mspfp-fast
14755 Generate single-precision FPX instructions, tuned for the fast
14760 Enable generation of ARC SIMD instructions via target-specific
14761 builtins. Only valid for @option{-mcpu=ARC700}.
14764 @opindex msoft-float
14765 This option ignored; it is provided for compatibility purposes only.
14766 Software floating-point code is emitted by default, and this default
14767 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
14768 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
14769 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
14773 Generate @code{swap} instructions.
14777 This enables use of the locked load/store conditional extension to implement
14778 atomic memory built-in functions. Not available for ARC 6xx or ARC
14783 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
14785 @item -mcode-density
14786 @opindex mcode-density
14787 Enable code density instructions for ARC EM.
14788 This option is on by default for ARC HS.
14792 Enable double load/store operations for ARC HS cores.
14794 @item -mtp-regno=@var{regno}
14796 Specify thread pointer register number.
14798 @item -mmpy-option=@var{multo}
14799 @opindex mmpy-option
14800 Compile ARCv2 code with a multiplier design option. You can specify
14801 the option using either a string or numeric value for @var{multo}.
14802 @samp{wlh1} is the default value. The recognized values are:
14807 No multiplier available.
14811 16x16 multiplier, fully pipelined.
14812 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
14816 32x32 multiplier, fully
14817 pipelined (1 stage). The following instructions are additionally
14818 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14822 32x32 multiplier, fully pipelined
14823 (2 stages). The following instructions are additionally enabled: @code{mpy},
14824 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14828 Two 16x16 multipliers, blocking,
14829 sequential. The following instructions are additionally enabled: @code{mpy},
14830 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14834 One 16x16 multiplier, blocking,
14835 sequential. The following instructions are additionally enabled: @code{mpy},
14836 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14840 One 32x4 multiplier, blocking,
14841 sequential. The following instructions are additionally enabled: @code{mpy},
14842 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14846 ARC HS SIMD support.
14850 ARC HS SIMD support.
14854 ARC HS SIMD support.
14858 This option is only available for ARCv2 cores@.
14860 @item -mfpu=@var{fpu}
14862 Enables support for specific floating-point hardware extensions for ARCv2
14863 cores. Supported values for @var{fpu} are:
14868 Enables support for single-precision floating-point hardware
14872 Enables support for double-precision floating-point hardware
14873 extensions. The single-precision floating-point extension is also
14874 enabled. Not available for ARC EM@.
14877 Enables support for double-precision floating-point hardware
14878 extensions using double-precision assist instructions. The single-precision
14879 floating-point extension is also enabled. This option is
14880 only available for ARC EM@.
14883 Enables support for double-precision floating-point hardware
14884 extensions using double-precision assist instructions.
14885 The single-precision floating-point, square-root, and divide
14886 extensions are also enabled. This option is
14887 only available for ARC EM@.
14890 Enables support for double-precision floating-point hardware
14891 extensions using double-precision assist instructions.
14892 The single-precision floating-point and fused multiply and add
14893 hardware extensions are also enabled. This option is
14894 only available for ARC EM@.
14897 Enables support for double-precision floating-point hardware
14898 extensions using double-precision assist instructions.
14899 All single-precision floating-point hardware extensions are also
14900 enabled. This option is only available for ARC EM@.
14903 Enables support for single-precision floating-point, square-root and divide
14904 hardware extensions@.
14907 Enables support for double-precision floating-point, square-root and divide
14908 hardware extensions. This option
14909 includes option @samp{fpus_div}. Not available for ARC EM@.
14912 Enables support for single-precision floating-point and
14913 fused multiply and add hardware extensions@.
14916 Enables support for double-precision floating-point and
14917 fused multiply and add hardware extensions. This option
14918 includes option @samp{fpus_fma}. Not available for ARC EM@.
14921 Enables support for all single-precision floating-point hardware
14925 Enables support for all single- and double-precision floating-point
14926 hardware extensions. Not available for ARC EM@.
14930 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
14931 @opindex mirq-ctrl-saved
14932 Specifies general-purposes registers that the processor automatically
14933 saves/restores on interrupt entry and exit. @var{register-range} is
14934 specified as two registers separated by a dash. The register range
14935 always starts with @code{r0}, the upper limit is @code{fp} register.
14936 @var{blink} and @var{lp_count} are optional. This option is only
14937 valid for ARC EM and ARC HS cores.
14939 @item -mrgf-banked-regs=@var{number}
14940 @opindex mrgf-banked-regs
14941 Specifies the number of registers replicated in second register bank
14942 on entry to fast interrupt. Fast interrupts are interrupts with the
14943 highest priority level P0. These interrupts save only PC and STATUS32
14944 registers to avoid memory transactions during interrupt entry and exit
14945 sequences. Use this option when you are using fast interrupts in an
14946 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
14948 @item -mlpc-width=@var{width}
14949 @opindex mlpc-width
14950 Specify the width of the @code{lp_count} register. Valid values for
14951 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
14952 fixed to 32 bits. If the width is less than 32, the compiler does not
14953 attempt to transform loops in your program to use the zero-delay loop
14954 mechanism unless it is known that the @code{lp_count} register can
14955 hold the required loop-counter value. Depending on the width
14956 specified, the compiler and run-time library might continue to use the
14957 loop mechanism for various needs. This option defines macro
14958 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
14962 The following options are passed through to the assembler, and also
14963 define preprocessor macro symbols.
14965 @c Flags used by the assembler, but for which we define preprocessor
14966 @c macro symbols as well.
14969 @opindex mdsp-packa
14970 Passed down to the assembler to enable the DSP Pack A extensions.
14971 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
14976 Passed down to the assembler to enable the dual Viterbi butterfly
14977 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
14978 option is deprecated.
14980 @c ARC700 4.10 extension instruction
14983 Passed down to the assembler to enable the locked load/store
14984 conditional extension. Also sets the preprocessor symbol
14989 Passed down to the assembler. Also sets the preprocessor symbol
14990 @code{__Xxmac_d16}. This option is deprecated.
14994 Passed down to the assembler. Also sets the preprocessor symbol
14995 @code{__Xxmac_24}. This option is deprecated.
14997 @c ARC700 4.10 extension instruction
15000 Passed down to the assembler to enable the 64-bit time-stamp counter
15001 extension instruction. Also sets the preprocessor symbol
15002 @code{__Xrtsc}. This option is deprecated.
15004 @c ARC700 4.10 extension instruction
15007 Passed down to the assembler to enable the swap byte ordering
15008 extension instruction. Also sets the preprocessor symbol
15012 @opindex mtelephony
15013 Passed down to the assembler to enable dual- and single-operand
15014 instructions for telephony. Also sets the preprocessor symbol
15015 @code{__Xtelephony}. This option is deprecated.
15019 Passed down to the assembler to enable the XY memory extension. Also
15020 sets the preprocessor symbol @code{__Xxy}.
15024 The following options control how the assembly code is annotated:
15026 @c Assembly annotation options
15030 Annotate assembler instructions with estimated addresses.
15032 @item -mannotate-align
15033 @opindex mannotate-align
15034 Explain what alignment considerations lead to the decision to make an
15035 instruction short or long.
15039 The following options are passed through to the linker:
15041 @c options passed through to the linker
15045 Passed through to the linker, to specify use of the @code{arclinux} emulation.
15046 This option is enabled by default in tool chains built for
15047 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
15048 when profiling is not requested.
15050 @item -marclinux_prof
15051 @opindex marclinux_prof
15052 Passed through to the linker, to specify use of the
15053 @code{arclinux_prof} emulation. This option is enabled by default in
15054 tool chains built for @w{@code{arc-linux-uclibc}} and
15055 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
15059 The following options control the semantics of generated code:
15061 @c semantically relevant code generation options
15064 @opindex mlong-calls
15065 Generate calls as register indirect calls, thus providing access
15066 to the full 32-bit address range.
15068 @item -mmedium-calls
15069 @opindex mmedium-calls
15070 Don't use less than 25-bit addressing range for calls, which is the
15071 offset available for an unconditional branch-and-link
15072 instruction. Conditional execution of function calls is suppressed, to
15073 allow use of the 25-bit range, rather than the 21-bit range with
15074 conditional branch-and-link. This is the default for tool chains built
15075 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
15079 Put definitions of externally-visible data in a small data section if
15080 that data is no bigger than @var{num} bytes. The default value of
15081 @var{num} is 4 for any ARC configuration, or 8 when we have double
15082 load/store operations.
15086 Do not generate sdata references. This is the default for tool chains
15087 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
15090 @item -mvolatile-cache
15091 @opindex mvolatile-cache
15092 Use ordinarily cached memory accesses for volatile references. This is the
15095 @item -mno-volatile-cache
15096 @opindex mno-volatile-cache
15097 Enable cache bypass for volatile references.
15101 The following options fine tune code generation:
15102 @c code generation tuning options
15105 @opindex malign-call
15106 Do alignment optimizations for call instructions.
15108 @item -mauto-modify-reg
15109 @opindex mauto-modify-reg
15110 Enable the use of pre/post modify with register displacement.
15112 @item -mbbit-peephole
15113 @opindex mbbit-peephole
15114 Enable bbit peephole2.
15118 This option disables a target-specific pass in @file{arc_reorg} to
15119 generate compare-and-branch (@code{br@var{cc}}) instructions.
15120 It has no effect on
15121 generation of these instructions driven by the combiner pass.
15123 @item -mcase-vector-pcrel
15124 @opindex mcase-vector-pcrel
15125 Use PC-relative switch case tables to enable case table shortening.
15126 This is the default for @option{-Os}.
15128 @item -mcompact-casesi
15129 @opindex mcompact-casesi
15130 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
15131 and only available for ARCv1 cores.
15133 @item -mno-cond-exec
15134 @opindex mno-cond-exec
15135 Disable the ARCompact-specific pass to generate conditional
15136 execution instructions.
15138 Due to delay slot scheduling and interactions between operand numbers,
15139 literal sizes, instruction lengths, and the support for conditional execution,
15140 the target-independent pass to generate conditional execution is often lacking,
15141 so the ARC port has kept a special pass around that tries to find more
15142 conditional execution generation opportunities after register allocation,
15143 branch shortening, and delay slot scheduling have been done. This pass
15144 generally, but not always, improves performance and code size, at the cost of
15145 extra compilation time, which is why there is an option to switch it off.
15146 If you have a problem with call instructions exceeding their allowable
15147 offset range because they are conditionalized, you should consider using
15148 @option{-mmedium-calls} instead.
15150 @item -mearly-cbranchsi
15151 @opindex mearly-cbranchsi
15152 Enable pre-reload use of the @code{cbranchsi} pattern.
15154 @item -mexpand-adddi
15155 @opindex mexpand-adddi
15156 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
15157 @code{add.f}, @code{adc} etc. This option is deprecated.
15159 @item -mindexed-loads
15160 @opindex mindexed-loads
15161 Enable the use of indexed loads. This can be problematic because some
15162 optimizers then assume that indexed stores exist, which is not
15166 Enable Local Register Allocation. This is still experimental for ARC,
15167 so by default the compiler uses standard reload
15168 (i.e. @option{-mno-lra}).
15170 @item -mlra-priority-none
15171 @opindex mlra-priority-none
15172 Don't indicate any priority for target registers.
15174 @item -mlra-priority-compact
15175 @opindex mlra-priority-compact
15176 Indicate target register priority for r0..r3 / r12..r15.
15178 @item -mlra-priority-noncompact
15179 @opindex mlra-priority-noncompact
15180 Reduce target register priority for r0..r3 / r12..r15.
15182 @item -mno-millicode
15183 @opindex mno-millicode
15184 When optimizing for size (using @option{-Os}), prologues and epilogues
15185 that have to save or restore a large number of registers are often
15186 shortened by using call to a special function in libgcc; this is
15187 referred to as a @emph{millicode} call. As these calls can pose
15188 performance issues, and/or cause linking issues when linking in a
15189 nonstandard way, this option is provided to turn off millicode call
15193 @opindex mmixed-code
15194 Tweak register allocation to help 16-bit instruction generation.
15195 This generally has the effect of decreasing the average instruction size
15196 while increasing the instruction count.
15200 Enable @samp{q} instruction alternatives.
15201 This is the default for @option{-Os}.
15205 Enable @samp{Rcq} constraint handling.
15206 Most short code generation depends on this.
15207 This is the default.
15211 Enable @samp{Rcw} constraint handling.
15212 Most ccfsm condexec mostly depends on this.
15213 This is the default.
15215 @item -msize-level=@var{level}
15216 @opindex msize-level
15217 Fine-tune size optimization with regards to instruction lengths and alignment.
15218 The recognized values for @var{level} are:
15221 No size optimization. This level is deprecated and treated like @samp{1}.
15224 Short instructions are used opportunistically.
15227 In addition, alignment of loops and of code after barriers are dropped.
15230 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
15234 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
15235 the behavior when this is not set is equivalent to level @samp{1}.
15237 @item -mtune=@var{cpu}
15239 Set instruction scheduling parameters for @var{cpu}, overriding any implied
15240 by @option{-mcpu=}.
15242 Supported values for @var{cpu} are
15246 Tune for ARC600 CPU.
15249 Tune for ARC601 CPU.
15252 Tune for ARC700 CPU with standard multiplier block.
15255 Tune for ARC700 CPU with XMAC block.
15258 Tune for ARC725D CPU.
15261 Tune for ARC750D CPU.
15265 @item -mmultcost=@var{num}
15267 Cost to assume for a multiply instruction, with @samp{4} being equal to a
15268 normal instruction.
15270 @item -munalign-prob-threshold=@var{probability}
15271 @opindex munalign-prob-threshold
15272 Set probability threshold for unaligning branches.
15273 When tuning for @samp{ARC700} and optimizing for speed, branches without
15274 filled delay slot are preferably emitted unaligned and long, unless
15275 profiling indicates that the probability for the branch to be taken
15276 is below @var{probability}. @xref{Cross-profiling}.
15277 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
15281 The following options are maintained for backward compatibility, but
15282 are now deprecated and will be removed in a future release:
15284 @c Deprecated options
15292 @opindex mbig-endian
15295 Compile code for big-endian targets. Use of these options is now
15296 deprecated. Big-endian code is supported by configuring GCC to build
15297 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
15298 for which big endian is the default.
15300 @item -mlittle-endian
15301 @opindex mlittle-endian
15304 Compile code for little-endian targets. Use of these options is now
15305 deprecated. Little-endian code is supported by configuring GCC to build
15306 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
15307 for which little endian is the default.
15309 @item -mbarrel_shifter
15310 @opindex mbarrel_shifter
15311 Replaced by @option{-mbarrel-shifter}.
15313 @item -mdpfp_compact
15314 @opindex mdpfp_compact
15315 Replaced by @option{-mdpfp-compact}.
15318 @opindex mdpfp_fast
15319 Replaced by @option{-mdpfp-fast}.
15322 @opindex mdsp_packa
15323 Replaced by @option{-mdsp-packa}.
15327 Replaced by @option{-mea}.
15331 Replaced by @option{-mmac-24}.
15335 Replaced by @option{-mmac-d16}.
15337 @item -mspfp_compact
15338 @opindex mspfp_compact
15339 Replaced by @option{-mspfp-compact}.
15342 @opindex mspfp_fast
15343 Replaced by @option{-mspfp-fast}.
15345 @item -mtune=@var{cpu}
15347 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
15348 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
15349 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
15351 @item -multcost=@var{num}
15353 Replaced by @option{-mmultcost}.
15358 @subsection ARM Options
15359 @cindex ARM options
15361 These @samp{-m} options are defined for the ARM port:
15364 @item -mabi=@var{name}
15366 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
15367 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
15370 @opindex mapcs-frame
15371 Generate a stack frame that is compliant with the ARM Procedure Call
15372 Standard for all functions, even if this is not strictly necessary for
15373 correct execution of the code. Specifying @option{-fomit-frame-pointer}
15374 with this option causes the stack frames not to be generated for
15375 leaf functions. The default is @option{-mno-apcs-frame}.
15376 This option is deprecated.
15380 This is a synonym for @option{-mapcs-frame} and is deprecated.
15383 @c not currently implemented
15384 @item -mapcs-stack-check
15385 @opindex mapcs-stack-check
15386 Generate code to check the amount of stack space available upon entry to
15387 every function (that actually uses some stack space). If there is
15388 insufficient space available then either the function
15389 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
15390 called, depending upon the amount of stack space required. The runtime
15391 system is required to provide these functions. The default is
15392 @option{-mno-apcs-stack-check}, since this produces smaller code.
15394 @c not currently implemented
15395 @item -mapcs-reentrant
15396 @opindex mapcs-reentrant
15397 Generate reentrant, position-independent code. The default is
15398 @option{-mno-apcs-reentrant}.
15401 @item -mthumb-interwork
15402 @opindex mthumb-interwork
15403 Generate code that supports calling between the ARM and Thumb
15404 instruction sets. Without this option, on pre-v5 architectures, the
15405 two instruction sets cannot be reliably used inside one program. The
15406 default is @option{-mno-thumb-interwork}, since slightly larger code
15407 is generated when @option{-mthumb-interwork} is specified. In AAPCS
15408 configurations this option is meaningless.
15410 @item -mno-sched-prolog
15411 @opindex mno-sched-prolog
15412 Prevent the reordering of instructions in the function prologue, or the
15413 merging of those instruction with the instructions in the function's
15414 body. This means that all functions start with a recognizable set
15415 of instructions (or in fact one of a choice from a small set of
15416 different function prologues), and this information can be used to
15417 locate the start of functions inside an executable piece of code. The
15418 default is @option{-msched-prolog}.
15420 @item -mfloat-abi=@var{name}
15421 @opindex mfloat-abi
15422 Specifies which floating-point ABI to use. Permissible values
15423 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15425 Specifying @samp{soft} causes GCC to generate output containing
15426 library calls for floating-point operations.
15427 @samp{softfp} allows the generation of code using hardware floating-point
15428 instructions, but still uses the soft-float calling conventions.
15429 @samp{hard} allows generation of floating-point instructions
15430 and uses FPU-specific calling conventions.
15432 The default depends on the specific target configuration. Note that
15433 the hard-float and soft-float ABIs are not link-compatible; you must
15434 compile your entire program with the same ABI, and link with a
15435 compatible set of libraries.
15437 @item -mlittle-endian
15438 @opindex mlittle-endian
15439 Generate code for a processor running in little-endian mode. This is
15440 the default for all standard configurations.
15443 @opindex mbig-endian
15444 Generate code for a processor running in big-endian mode; the default is
15445 to compile code for a little-endian processor.
15450 When linking a big-endian image select between BE8 and BE32 formats.
15451 The option has no effect for little-endian images and is ignored. The
15452 default is dependent on the selected target architecture. For ARMv6
15453 and later architectures the default is BE8, for older architectures
15454 the default is BE32. BE32 format has been deprecated by ARM.
15456 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
15458 This specifies the name of the target ARM architecture. GCC uses this
15459 name to determine what kind of instructions it can emit when generating
15460 assembly code. This option can be used in conjunction with or instead
15461 of the @option{-mcpu=} option.
15463 Permissible names are:
15465 @samp{armv5t}, @samp{armv5te},
15466 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
15467 @samp{armv6z}, @samp{armv6zk},
15468 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
15469 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a},
15472 @samp{armv6-m}, @samp{armv6s-m},
15473 @samp{armv7-m}, @samp{armv7e-m},
15474 @samp{armv8-m.base}, @samp{armv8-m.main},
15475 @samp{iwmmxt} and @samp{iwmmxt2}.
15477 Additionally, the following architectures, which lack support for the
15478 Thumb exection state, are recognized but support is deprecated:
15479 @samp{armv2}, @samp{armv2a}, @samp{armv3}, @samp{armv3m},
15480 @samp{armv4}, @samp{armv5} and @samp{armv5e}.
15482 Many of the architectures support extensions. These can be added by
15483 appending @samp{+@var{extension}} to the architecture name. Extension
15484 options are processed in order and capabilities accumulate. An extension
15485 will also enable any necessary base extensions
15486 upon which it depends. For example, the @samp{+crypto} extension
15487 will always enable the @samp{+simd} extension. The exception to the
15488 additive construction is for extensions that are prefixed with
15489 @samp{+no@dots{}}: these extensions disable the specified option and
15490 any other extensions that may depend on the presence of that
15493 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
15494 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
15495 entirely disabled by the @samp{+nofp} option that follows it.
15497 Most extension names are generically named, but have an effect that is
15498 dependent upon the architecture to which it is applied. For example,
15499 the @samp{+simd} option can be applied to both @samp{armv7-a} and
15500 @samp{armv8-a} architectures, but will enable the original ARMv7
15501 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-a
15502 variant for @samp{armv8-a}.
15504 The table below lists the supported extensions for each architecture.
15505 Architectures not mentioned do not support any extensions.
15519 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
15520 used as an alias for this extension.
15523 Disable the floating-point instructions.
15527 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
15530 The VFPv3 floating-point instructions, with 16 double-precision
15531 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15532 for this extension. Note that floating-point is not supported by the
15533 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
15534 ARMv7-R architectures.
15537 Disable the floating-point instructions.
15543 The VFPv3 floating-point instructions, with 16 double-precision
15544 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15545 for this extension.
15548 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15549 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
15550 for this extension.
15553 The VFPv3 floating-point instructions, with 32 double-precision
15556 @item +vfpv3-d16-fp16
15557 The VFPv3 floating-point instructions, with 16 double-precision
15558 registers and the half-precision floating-point conversion operations.
15561 The VFPv3 floating-point instructions, with 32 double-precision
15562 registers and the half-precision floating-point conversion operations.
15565 The VFPv4 floating-point instructions, with 16 double-precision
15569 The VFPv4 floating-point instructions, with 32 double-precision
15573 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15574 the half-precision floating-point conversion operations.
15577 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
15580 Disable the Advanced SIMD instructions (does not disable floating point).
15583 Disable the floating-point and Advanced SIMD instructions.
15587 The extended version of the ARMv7-A architecture with support for
15591 The VFPv4 floating-point instructions, with 16 double-precision registers.
15592 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
15595 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
15596 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
15599 The VFPv3 floating-point instructions, with 16 double-precision
15603 The VFPv3 floating-point instructions, with 32 double-precision
15606 @item +vfpv3-d16-fp16
15607 The VFPv3 floating-point instructions, with 16 double-precision
15608 registers and the half-precision floating-point conversion operations.
15611 The VFPv3 floating-point instructions, with 32 double-precision
15612 registers and the half-precision floating-point conversion operations.
15615 The VFPv4 floating-point instructions, with 16 double-precision
15619 The VFPv4 floating-point instructions, with 32 double-precision
15623 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15624 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
15627 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15628 the half-precision floating-point conversion operations.
15631 Disable the Advanced SIMD instructions (does not disable floating point).
15634 Disable the floating-point and Advanced SIMD instructions.
15640 The Cyclic Redundancy Check (CRC) instructions.
15642 The ARMv8 Advanced SIMD and floating-point instructions.
15644 The cryptographic instructions.
15646 Disable the cryptographic isntructions.
15648 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15654 The ARMv8.1 Advanced SIMD and floating-point instructions.
15657 The cryptographic instructions. This also enables the Advanced SIMD and
15658 floating-point instructions.
15661 Disable the cryptographic isntructions.
15664 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15670 The half-precision floating-point data processing instructions.
15671 This also enables the Advanced SIMD and floating-point instructions.
15674 The ARMv8.1 Advanced SIMD and floating-point instructions.
15677 The cryptographic instructions. This also enables the Advanced SIMD and
15678 floating-point instructions.
15681 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
15684 Disable the cryptographic extension.
15687 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15693 The single-precision VFPv3 floating-point instructions. The extension
15694 @samp{+vfpv3xd} can be used as an alias for this extension.
15697 The VFPv3 floating-point instructions with 16 double-precision registers.
15698 The extension +vfpv3-d16 can be used as an alias for this extension.
15701 Disable the floating-point extension.
15704 The ARM-state integer division instructions.
15707 Disable the ARM-state integer division extension.
15713 The single-precision VFPv4 floating-point instructions.
15716 The single-precision FPv5 floating-point instructions.
15719 The single- and double-precision FPv5 floating-point instructions.
15722 Disable the floating-point extensions.
15728 The DSP instructions.
15731 Disable the DSP extension.
15734 The single-precision floating-point instructions.
15737 The single- and double-precision floating-point instructions.
15740 Disable the floating-point extension.
15746 The Cyclic Redundancy Check (CRC) instructions.
15748 The single-precision FPv5 floating-point instructions.
15750 The ARMv8 Advanced SIMD and floating-point instructions.
15752 The cryptographic instructions.
15754 Disable the cryptographic isntructions.
15756 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15761 @option{-march=native} causes the compiler to auto-detect the architecture
15762 of the build computer. At present, this feature is only supported on
15763 GNU/Linux, and not all architectures are recognized. If the auto-detect
15764 is unsuccessful the option has no effect.
15766 @item -mtune=@var{name}
15768 This option specifies the name of the target ARM processor for
15769 which GCC should tune the performance of the code.
15770 For some ARM implementations better performance can be obtained by using
15772 Permissible names are: @samp{arm2}, @samp{arm250},
15773 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
15774 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
15775 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
15776 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
15778 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
15779 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
15780 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
15781 @samp{strongarm1110},
15782 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
15783 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
15784 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
15785 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
15786 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
15787 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
15788 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
15789 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
15790 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
15791 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15792 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15793 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7},
15794 @samp{cortex-r8}, @samp{cortex-r52},
15802 @samp{cortex-m0plus},
15803 @samp{cortex-m1.small-multiply},
15804 @samp{cortex-m0.small-multiply},
15805 @samp{cortex-m0plus.small-multiply},
15807 @samp{marvell-pj4},
15808 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
15809 @samp{fa526}, @samp{fa626},
15810 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
15813 Additionally, this option can specify that GCC should tune the performance
15814 of the code for a big.LITTLE system. Permissible names are:
15815 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
15816 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15817 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
15818 @samp{cortex-a75.cortex-a55}.
15820 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
15821 performance for a blend of processors within architecture @var{arch}.
15822 The aim is to generate code that run well on the current most popular
15823 processors, balancing between optimizations that benefit some CPUs in the
15824 range, and avoiding performance pitfalls of other CPUs. The effects of
15825 this option may change in future GCC versions as CPU models come and go.
15827 @option{-mtune} permits the same extension options as @option{-mcpu}, but
15828 the extension options do not affect the tuning of the generated code.
15830 @option{-mtune=native} causes the compiler to auto-detect the CPU
15831 of the build computer. At present, this feature is only supported on
15832 GNU/Linux, and not all architectures are recognized. If the auto-detect is
15833 unsuccessful the option has no effect.
15835 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
15837 This specifies the name of the target ARM processor. GCC uses this name
15838 to derive the name of the target ARM architecture (as if specified
15839 by @option{-march}) and the ARM processor type for which to tune for
15840 performance (as if specified by @option{-mtune}). Where this option
15841 is used in conjunction with @option{-march} or @option{-mtune},
15842 those options take precedence over the appropriate part of this option.
15844 Many of the supported CPUs implement optional architectural
15845 extensions. Where this is so the architectural extensions are
15846 normally enabled by default. If implementations that lack the
15847 extension exist, then the extension syntax can be used to disable
15848 those extensions that have been omitted. For floating-point and
15849 Advanced SIMD (Neon) instructions, the settings of the options
15850 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
15851 floating-point and Advanced SIMD instructions will only be used if
15852 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
15853 @option{-mfpu} other than @samp{auto} will override the available
15854 floating-point and SIMD extension instructions.
15856 For example, @samp{cortex-a9} can be found in three major
15857 configurations: integer only, with just a floating-point unit or with
15858 floating-point and Advanced SIMD. The default is to enable all the
15859 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
15860 be used to disable just the SIMD or both the SIMD and floating-point
15861 instructions respectively.
15863 Permissible names for this option are the same as those for
15866 The following extension options are common to the listed CPUs:
15870 Disable the DSP instructions on @samp{cortex-m33}.
15873 Disables the floating-point instructions on @samp{arm9e},
15874 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
15875 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
15876 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
15877 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
15878 Disables the floating-point and SIMD instructions on
15879 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
15880 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
15881 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
15882 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
15883 @samp{cortex-a53} and @samp{cortex-a55}.
15886 Disables the double-precision component of the floating-point instructions
15887 on @samp{cortex-r5}, @samp{cortex-r52} and @samp{cortex-m7}.
15890 Disables the SIMD (but not floating-point) instructions on
15891 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
15892 and @samp{cortex-a9}.
15895 Enables the cryptographic instructions on @samp{cortex-a32},
15896 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
15897 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
15898 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15899 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
15900 @samp{cortex-a75.cortex-a55}.
15903 Additionally the @samp{generic-armv7-a} pseudo target defaults to
15904 VFPv3 with 16 double-precision registers. It supports the following
15905 extension options: @samp{vfpv3-d16}, @samp{vfpv3},
15906 @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16}, @samp{vfpv4-d16},
15907 @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3}, @samp{neon-fp16},
15908 @samp{neon-vfpv4}. The meanings are the same as for the extensions to
15909 @option{-march=armv7-a}.
15911 @option{-mcpu=generic-@var{arch}} is also permissible, and is
15912 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
15913 See @option{-mtune} for more information.
15915 @option{-mcpu=native} causes the compiler to auto-detect the CPU
15916 of the build computer. At present, this feature is only supported on
15917 GNU/Linux, and not all architectures are recognized. If the auto-detect
15918 is unsuccessful the option has no effect.
15920 @item -mfpu=@var{name}
15922 This specifies what floating-point hardware (or hardware emulation) is
15923 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
15925 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
15926 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
15927 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
15928 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
15929 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
15930 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
15931 is an alias for @samp{vfpv2}.
15933 The setting @samp{auto} is the default and is special. It causes the
15934 compiler to select the floating-point and Advanced SIMD instructions
15935 based on the settings of @option{-mcpu} and @option{-march}.
15937 If the selected floating-point hardware includes the NEON extension
15938 (e.g. @option{-mfpu=neon}), note that floating-point
15939 operations are not generated by GCC's auto-vectorization pass unless
15940 @option{-funsafe-math-optimizations} is also specified. This is
15941 because NEON hardware does not fully implement the IEEE 754 standard for
15942 floating-point arithmetic (in particular denormal values are treated as
15943 zero), so the use of NEON instructions may lead to a loss of precision.
15945 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}).
15947 @item -mfp16-format=@var{name}
15948 @opindex mfp16-format
15949 Specify the format of the @code{__fp16} half-precision floating-point type.
15950 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
15951 the default is @samp{none}, in which case the @code{__fp16} type is not
15952 defined. @xref{Half-Precision}, for more information.
15954 @item -mstructure-size-boundary=@var{n}
15955 @opindex mstructure-size-boundary
15956 The sizes of all structures and unions are rounded up to a multiple
15957 of the number of bits set by this option. Permissible values are 8, 32
15958 and 64. The default value varies for different toolchains. For the COFF
15959 targeted toolchain the default value is 8. A value of 64 is only allowed
15960 if the underlying ABI supports it.
15962 Specifying a larger number can produce faster, more efficient code, but
15963 can also increase the size of the program. Different values are potentially
15964 incompatible. Code compiled with one value cannot necessarily expect to
15965 work with code or libraries compiled with another value, if they exchange
15966 information using structures or unions.
15968 This option is deprecated.
15970 @item -mabort-on-noreturn
15971 @opindex mabort-on-noreturn
15972 Generate a call to the function @code{abort} at the end of a
15973 @code{noreturn} function. It is executed if the function tries to
15977 @itemx -mno-long-calls
15978 @opindex mlong-calls
15979 @opindex mno-long-calls
15980 Tells the compiler to perform function calls by first loading the
15981 address of the function into a register and then performing a subroutine
15982 call on this register. This switch is needed if the target function
15983 lies outside of the 64-megabyte addressing range of the offset-based
15984 version of subroutine call instruction.
15986 Even if this switch is enabled, not all function calls are turned
15987 into long calls. The heuristic is that static functions, functions
15988 that have the @code{short_call} attribute, functions that are inside
15989 the scope of a @code{#pragma no_long_calls} directive, and functions whose
15990 definitions have already been compiled within the current compilation
15991 unit are not turned into long calls. The exceptions to this rule are
15992 that weak function definitions, functions with the @code{long_call}
15993 attribute or the @code{section} attribute, and functions that are within
15994 the scope of a @code{#pragma long_calls} directive are always
15995 turned into long calls.
15997 This feature is not enabled by default. Specifying
15998 @option{-mno-long-calls} restores the default behavior, as does
15999 placing the function calls within the scope of a @code{#pragma
16000 long_calls_off} directive. Note these switches have no effect on how
16001 the compiler generates code to handle function calls via function
16004 @item -msingle-pic-base
16005 @opindex msingle-pic-base
16006 Treat the register used for PIC addressing as read-only, rather than
16007 loading it in the prologue for each function. The runtime system is
16008 responsible for initializing this register with an appropriate value
16009 before execution begins.
16011 @item -mpic-register=@var{reg}
16012 @opindex mpic-register
16013 Specify the register to be used for PIC addressing.
16014 For standard PIC base case, the default is any suitable register
16015 determined by compiler. For single PIC base case, the default is
16016 @samp{R9} if target is EABI based or stack-checking is enabled,
16017 otherwise the default is @samp{R10}.
16019 @item -mpic-data-is-text-relative
16020 @opindex mpic-data-is-text-relative
16021 Assume that the displacement between the text and data segments is fixed
16022 at static link time. This permits using PC-relative addressing
16023 operations to access data known to be in the data segment. For
16024 non-VxWorks RTP targets, this option is enabled by default. When
16025 disabled on such targets, it will enable @option{-msingle-pic-base} by
16028 @item -mpoke-function-name
16029 @opindex mpoke-function-name
16030 Write the name of each function into the text section, directly
16031 preceding the function prologue. The generated code is similar to this:
16035 .ascii "arm_poke_function_name", 0
16038 .word 0xff000000 + (t1 - t0)
16039 arm_poke_function_name
16041 stmfd sp!, @{fp, ip, lr, pc@}
16045 When performing a stack backtrace, code can inspect the value of
16046 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
16047 location @code{pc - 12} and the top 8 bits are set, then we know that
16048 there is a function name embedded immediately preceding this location
16049 and has length @code{((pc[-3]) & 0xff000000)}.
16056 Select between generating code that executes in ARM and Thumb
16057 states. The default for most configurations is to generate code
16058 that executes in ARM state, but the default can be changed by
16059 configuring GCC with the @option{--with-mode=}@var{state}
16062 You can also override the ARM and Thumb mode for each function
16063 by using the @code{target("thumb")} and @code{target("arm")} function attributes
16064 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
16067 @opindex mtpcs-frame
16068 Generate a stack frame that is compliant with the Thumb Procedure Call
16069 Standard for all non-leaf functions. (A leaf function is one that does
16070 not call any other functions.) The default is @option{-mno-tpcs-frame}.
16072 @item -mtpcs-leaf-frame
16073 @opindex mtpcs-leaf-frame
16074 Generate a stack frame that is compliant with the Thumb Procedure Call
16075 Standard for all leaf functions. (A leaf function is one that does
16076 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
16078 @item -mcallee-super-interworking
16079 @opindex mcallee-super-interworking
16080 Gives all externally visible functions in the file being compiled an ARM
16081 instruction set header which switches to Thumb mode before executing the
16082 rest of the function. This allows these functions to be called from
16083 non-interworking code. This option is not valid in AAPCS configurations
16084 because interworking is enabled by default.
16086 @item -mcaller-super-interworking
16087 @opindex mcaller-super-interworking
16088 Allows calls via function pointers (including virtual functions) to
16089 execute correctly regardless of whether the target code has been
16090 compiled for interworking or not. There is a small overhead in the cost
16091 of executing a function pointer if this option is enabled. This option
16092 is not valid in AAPCS configurations because interworking is enabled
16095 @item -mtp=@var{name}
16097 Specify the access model for the thread local storage pointer. The valid
16098 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
16099 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
16100 (supported in the arm6k architecture), and @samp{auto}, which uses the
16101 best available method for the selected processor. The default setting is
16104 @item -mtls-dialect=@var{dialect}
16105 @opindex mtls-dialect
16106 Specify the dialect to use for accessing thread local storage. Two
16107 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
16108 @samp{gnu} dialect selects the original GNU scheme for supporting
16109 local and global dynamic TLS models. The @samp{gnu2} dialect
16110 selects the GNU descriptor scheme, which provides better performance
16111 for shared libraries. The GNU descriptor scheme is compatible with
16112 the original scheme, but does require new assembler, linker and
16113 library support. Initial and local exec TLS models are unaffected by
16114 this option and always use the original scheme.
16116 @item -mword-relocations
16117 @opindex mword-relocations
16118 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
16119 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
16120 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
16123 @item -mfix-cortex-m3-ldrd
16124 @opindex mfix-cortex-m3-ldrd
16125 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
16126 with overlapping destination and base registers are used. This option avoids
16127 generating these instructions. This option is enabled by default when
16128 @option{-mcpu=cortex-m3} is specified.
16130 @item -munaligned-access
16131 @itemx -mno-unaligned-access
16132 @opindex munaligned-access
16133 @opindex mno-unaligned-access
16134 Enables (or disables) reading and writing of 16- and 32- bit values
16135 from addresses that are not 16- or 32- bit aligned. By default
16136 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
16137 ARMv8-M Baseline architectures, and enabled for all other
16138 architectures. If unaligned access is not enabled then words in packed
16139 data structures are accessed a byte at a time.
16141 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
16142 generated object file to either true or false, depending upon the
16143 setting of this option. If unaligned access is enabled then the
16144 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
16147 @item -mneon-for-64bits
16148 @opindex mneon-for-64bits
16149 Enables using Neon to handle scalar 64-bits operations. This is
16150 disabled by default since the cost of moving data from core registers
16153 @item -mslow-flash-data
16154 @opindex mslow-flash-data
16155 Assume loading data from flash is slower than fetching instruction.
16156 Therefore literal load is minimized for better performance.
16157 This option is only supported when compiling for ARMv7 M-profile and
16160 @item -masm-syntax-unified
16161 @opindex masm-syntax-unified
16162 Assume inline assembler is using unified asm syntax. The default is
16163 currently off which implies divided syntax. This option has no impact
16164 on Thumb2. However, this may change in future releases of GCC.
16165 Divided syntax should be considered deprecated.
16167 @item -mrestrict-it
16168 @opindex mrestrict-it
16169 Restricts generation of IT blocks to conform to the rules of ARMv8.
16170 IT blocks can only contain a single 16-bit instruction from a select
16171 set of instructions. This option is on by default for ARMv8 Thumb mode.
16173 @item -mprint-tune-info
16174 @opindex mprint-tune-info
16175 Print CPU tuning information as comment in assembler file. This is
16176 an option used only for regression testing of the compiler and not
16177 intended for ordinary use in compiling code. This option is disabled
16181 @opindex mpure-code
16182 Do not allow constant data to be placed in code sections.
16183 Additionally, when compiling for ELF object format give all text sections the
16184 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
16185 is only available when generating non-pic code for M-profile targets with the
16190 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
16191 Development Tools Engineering Specification", which can be found on
16192 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
16196 @subsection AVR Options
16197 @cindex AVR Options
16199 These options are defined for AVR implementations:
16202 @item -mmcu=@var{mcu}
16204 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
16206 The default for this option is@tie{}@samp{avr2}.
16208 GCC supports the following AVR devices and ISAs:
16210 @include avr-mmcu.texi
16215 Assume that all data in static storage can be accessed by LDS / STS
16216 instructions. This option has only an effect on reduced Tiny devices like
16217 ATtiny40. See also the @code{absdata}
16218 @ref{AVR Variable Attributes,variable attribute}.
16220 @item -maccumulate-args
16221 @opindex maccumulate-args
16222 Accumulate outgoing function arguments and acquire/release the needed
16223 stack space for outgoing function arguments once in function
16224 prologue/epilogue. Without this option, outgoing arguments are pushed
16225 before calling a function and popped afterwards.
16227 Popping the arguments after the function call can be expensive on
16228 AVR so that accumulating the stack space might lead to smaller
16229 executables because arguments need not be removed from the
16230 stack after such a function call.
16232 This option can lead to reduced code size for functions that perform
16233 several calls to functions that get their arguments on the stack like
16234 calls to printf-like functions.
16236 @item -mbranch-cost=@var{cost}
16237 @opindex mbranch-cost
16238 Set the branch costs for conditional branch instructions to
16239 @var{cost}. Reasonable values for @var{cost} are small, non-negative
16240 integers. The default branch cost is 0.
16242 @item -mcall-prologues
16243 @opindex mcall-prologues
16244 Functions prologues/epilogues are expanded as calls to appropriate
16245 subroutines. Code size is smaller.
16247 @item -mgas-isr-prologues
16248 @opindex mgas-isr-prologues
16249 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
16250 instruction supported by GNU Binutils.
16251 If this option is on, the feature can still be disabled for individual
16252 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
16253 function attribute. This feature is activated per default
16254 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
16255 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
16259 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
16260 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
16261 and @code{long long} is 4 bytes. Please note that this option does not
16262 conform to the C standards, but it results in smaller code
16265 @item -mn-flash=@var{num}
16267 Assume that the flash memory has a size of
16268 @var{num} times 64@tie{}KiB.
16270 @item -mno-interrupts
16271 @opindex mno-interrupts
16272 Generated code is not compatible with hardware interrupts.
16273 Code size is smaller.
16277 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
16278 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
16279 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
16280 the assembler's command line and the @option{--relax} option to the
16281 linker's command line.
16283 Jump relaxing is performed by the linker because jump offsets are not
16284 known before code is located. Therefore, the assembler code generated by the
16285 compiler is the same, but the instructions in the executable may
16286 differ from instructions in the assembler code.
16288 Relaxing must be turned on if linker stubs are needed, see the
16289 section on @code{EIND} and linker stubs below.
16293 Assume that the device supports the Read-Modify-Write
16294 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
16296 @item -mshort-calls
16297 @opindex mshort-calls
16299 Assume that @code{RJMP} and @code{RCALL} can target the whole
16302 This option is used internally for multilib selection. It is
16303 not an optimization option, and you don't need to set it by hand.
16307 Treat the stack pointer register as an 8-bit register,
16308 i.e.@: assume the high byte of the stack pointer is zero.
16309 In general, you don't need to set this option by hand.
16311 This option is used internally by the compiler to select and
16312 build multilibs for architectures @code{avr2} and @code{avr25}.
16313 These architectures mix devices with and without @code{SPH}.
16314 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
16315 the compiler driver adds or removes this option from the compiler
16316 proper's command line, because the compiler then knows if the device
16317 or architecture has an 8-bit stack pointer and thus no @code{SPH}
16322 Use address register @code{X} in a way proposed by the hardware. This means
16323 that @code{X} is only used in indirect, post-increment or
16324 pre-decrement addressing.
16326 Without this option, the @code{X} register may be used in the same way
16327 as @code{Y} or @code{Z} which then is emulated by additional
16329 For example, loading a value with @code{X+const} addressing with a
16330 small non-negative @code{const < 64} to a register @var{Rn} is
16334 adiw r26, const ; X += const
16335 ld @var{Rn}, X ; @var{Rn} = *X
16336 sbiw r26, const ; X -= const
16340 @opindex mtiny-stack
16341 Only change the lower 8@tie{}bits of the stack pointer.
16343 @item -mfract-convert-truncate
16344 @opindex mfract-convert-truncate
16345 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
16348 @opindex nodevicelib
16349 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
16351 @item -Waddr-space-convert
16352 @opindex Waddr-space-convert
16353 Warn about conversions between address spaces in the case where the
16354 resulting address space is not contained in the incoming address space.
16356 @item -Wmisspelled-isr
16357 @opindex Wmisspelled-isr
16358 Warn if the ISR is misspelled, i.e. without __vector prefix.
16359 Enabled by default.
16362 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
16363 @cindex @code{EIND}
16364 Pointers in the implementation are 16@tie{}bits wide.
16365 The address of a function or label is represented as word address so
16366 that indirect jumps and calls can target any code address in the
16367 range of 64@tie{}Ki words.
16369 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
16370 bytes of program memory space, there is a special function register called
16371 @code{EIND} that serves as most significant part of the target address
16372 when @code{EICALL} or @code{EIJMP} instructions are used.
16374 Indirect jumps and calls on these devices are handled as follows by
16375 the compiler and are subject to some limitations:
16380 The compiler never sets @code{EIND}.
16383 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
16384 instructions or might read @code{EIND} directly in order to emulate an
16385 indirect call/jump by means of a @code{RET} instruction.
16388 The compiler assumes that @code{EIND} never changes during the startup
16389 code or during the application. In particular, @code{EIND} is not
16390 saved/restored in function or interrupt service routine
16394 For indirect calls to functions and computed goto, the linker
16395 generates @emph{stubs}. Stubs are jump pads sometimes also called
16396 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
16397 The stub contains a direct jump to the desired address.
16400 Linker relaxation must be turned on so that the linker generates
16401 the stubs correctly in all situations. See the compiler option
16402 @option{-mrelax} and the linker option @option{--relax}.
16403 There are corner cases where the linker is supposed to generate stubs
16404 but aborts without relaxation and without a helpful error message.
16407 The default linker script is arranged for code with @code{EIND = 0}.
16408 If code is supposed to work for a setup with @code{EIND != 0}, a custom
16409 linker script has to be used in order to place the sections whose
16410 name start with @code{.trampolines} into the segment where @code{EIND}
16414 The startup code from libgcc never sets @code{EIND}.
16415 Notice that startup code is a blend of code from libgcc and AVR-LibC.
16416 For the impact of AVR-LibC on @code{EIND}, see the
16417 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
16420 It is legitimate for user-specific startup code to set up @code{EIND}
16421 early, for example by means of initialization code located in
16422 section @code{.init3}. Such code runs prior to general startup code
16423 that initializes RAM and calls constructors, but after the bit
16424 of startup code from AVR-LibC that sets @code{EIND} to the segment
16425 where the vector table is located.
16427 #include <avr/io.h>
16430 __attribute__((section(".init3"),naked,used,no_instrument_function))
16431 init3_set_eind (void)
16433 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
16434 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
16439 The @code{__trampolines_start} symbol is defined in the linker script.
16442 Stubs are generated automatically by the linker if
16443 the following two conditions are met:
16446 @item The address of a label is taken by means of the @code{gs} modifier
16447 (short for @emph{generate stubs}) like so:
16449 LDI r24, lo8(gs(@var{func}))
16450 LDI r25, hi8(gs(@var{func}))
16452 @item The final location of that label is in a code segment
16453 @emph{outside} the segment where the stubs are located.
16457 The compiler emits such @code{gs} modifiers for code labels in the
16458 following situations:
16460 @item Taking address of a function or code label.
16461 @item Computed goto.
16462 @item If prologue-save function is used, see @option{-mcall-prologues}
16463 command-line option.
16464 @item Switch/case dispatch tables. If you do not want such dispatch
16465 tables you can specify the @option{-fno-jump-tables} command-line option.
16466 @item C and C++ constructors/destructors called during startup/shutdown.
16467 @item If the tools hit a @code{gs()} modifier explained above.
16471 Jumping to non-symbolic addresses like so is @emph{not} supported:
16476 /* Call function at word address 0x2 */
16477 return ((int(*)(void)) 0x2)();
16481 Instead, a stub has to be set up, i.e.@: the function has to be called
16482 through a symbol (@code{func_4} in the example):
16487 extern int func_4 (void);
16489 /* Call function at byte address 0x4 */
16494 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
16495 Alternatively, @code{func_4} can be defined in the linker script.
16498 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
16499 @cindex @code{RAMPD}
16500 @cindex @code{RAMPX}
16501 @cindex @code{RAMPY}
16502 @cindex @code{RAMPZ}
16503 Some AVR devices support memories larger than the 64@tie{}KiB range
16504 that can be accessed with 16-bit pointers. To access memory locations
16505 outside this 64@tie{}KiB range, the content of a @code{RAMP}
16506 register is used as high part of the address:
16507 The @code{X}, @code{Y}, @code{Z} address register is concatenated
16508 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
16509 register, respectively, to get a wide address. Similarly,
16510 @code{RAMPD} is used together with direct addressing.
16514 The startup code initializes the @code{RAMP} special function
16515 registers with zero.
16518 If a @ref{AVR Named Address Spaces,named address space} other than
16519 generic or @code{__flash} is used, then @code{RAMPZ} is set
16520 as needed before the operation.
16523 If the device supports RAM larger than 64@tie{}KiB and the compiler
16524 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
16525 is reset to zero after the operation.
16528 If the device comes with a specific @code{RAMP} register, the ISR
16529 prologue/epilogue saves/restores that SFR and initializes it with
16530 zero in case the ISR code might (implicitly) use it.
16533 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
16534 If you use inline assembler to read from locations outside the
16535 16-bit address range and change one of the @code{RAMP} registers,
16536 you must reset it to zero after the access.
16540 @subsubsection AVR Built-in Macros
16542 GCC defines several built-in macros so that the user code can test
16543 for the presence or absence of features. Almost any of the following
16544 built-in macros are deduced from device capabilities and thus
16545 triggered by the @option{-mmcu=} command-line option.
16547 For even more AVR-specific built-in macros see
16548 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
16553 Build-in macro that resolves to a decimal number that identifies the
16554 architecture and depends on the @option{-mmcu=@var{mcu}} option.
16555 Possible values are:
16557 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
16558 @code{4}, @code{5}, @code{51}, @code{6}
16560 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
16561 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
16566 @code{102}, @code{103}, @code{104},
16567 @code{105}, @code{106}, @code{107}
16569 for @var{mcu}=@code{avrtiny},
16570 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
16571 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
16572 If @var{mcu} specifies a device, this built-in macro is set
16573 accordingly. For example, with @option{-mmcu=atmega8} the macro is
16574 defined to @code{4}.
16576 @item __AVR_@var{Device}__
16577 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
16578 the device's name. For example, @option{-mmcu=atmega8} defines the
16579 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
16580 @code{__AVR_ATtiny261A__}, etc.
16582 The built-in macros' names follow
16583 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
16584 the device name as from the AVR user manual. The difference between
16585 @var{Device} in the built-in macro and @var{device} in
16586 @option{-mmcu=@var{device}} is that the latter is always lowercase.
16588 If @var{device} is not a device but only a core architecture like
16589 @samp{avr51}, this macro is not defined.
16591 @item __AVR_DEVICE_NAME__
16592 Setting @option{-mmcu=@var{device}} defines this built-in macro to
16593 the device's name. For example, with @option{-mmcu=atmega8} the macro
16594 is defined to @code{atmega8}.
16596 If @var{device} is not a device but only a core architecture like
16597 @samp{avr51}, this macro is not defined.
16599 @item __AVR_XMEGA__
16600 The device / architecture belongs to the XMEGA family of devices.
16602 @item __AVR_HAVE_ELPM__
16603 The device has the @code{ELPM} instruction.
16605 @item __AVR_HAVE_ELPMX__
16606 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
16607 R@var{n},Z+} instructions.
16609 @item __AVR_HAVE_MOVW__
16610 The device has the @code{MOVW} instruction to perform 16-bit
16611 register-register moves.
16613 @item __AVR_HAVE_LPMX__
16614 The device has the @code{LPM R@var{n},Z} and
16615 @code{LPM R@var{n},Z+} instructions.
16617 @item __AVR_HAVE_MUL__
16618 The device has a hardware multiplier.
16620 @item __AVR_HAVE_JMP_CALL__
16621 The device has the @code{JMP} and @code{CALL} instructions.
16622 This is the case for devices with more than 8@tie{}KiB of program
16625 @item __AVR_HAVE_EIJMP_EICALL__
16626 @itemx __AVR_3_BYTE_PC__
16627 The device has the @code{EIJMP} and @code{EICALL} instructions.
16628 This is the case for devices with more than 128@tie{}KiB of program memory.
16629 This also means that the program counter
16630 (PC) is 3@tie{}bytes wide.
16632 @item __AVR_2_BYTE_PC__
16633 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
16634 with up to 128@tie{}KiB of program memory.
16636 @item __AVR_HAVE_8BIT_SP__
16637 @itemx __AVR_HAVE_16BIT_SP__
16638 The stack pointer (SP) register is treated as 8-bit respectively
16639 16-bit register by the compiler.
16640 The definition of these macros is affected by @option{-mtiny-stack}.
16642 @item __AVR_HAVE_SPH__
16644 The device has the SPH (high part of stack pointer) special function
16645 register or has an 8-bit stack pointer, respectively.
16646 The definition of these macros is affected by @option{-mmcu=} and
16647 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
16650 @item __AVR_HAVE_RAMPD__
16651 @itemx __AVR_HAVE_RAMPX__
16652 @itemx __AVR_HAVE_RAMPY__
16653 @itemx __AVR_HAVE_RAMPZ__
16654 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
16655 @code{RAMPZ} special function register, respectively.
16657 @item __NO_INTERRUPTS__
16658 This macro reflects the @option{-mno-interrupts} command-line option.
16660 @item __AVR_ERRATA_SKIP__
16661 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
16662 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
16663 instructions because of a hardware erratum. Skip instructions are
16664 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
16665 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
16668 @item __AVR_ISA_RMW__
16669 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
16671 @item __AVR_SFR_OFFSET__=@var{offset}
16672 Instructions that can address I/O special function registers directly
16673 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
16674 address as if addressed by an instruction to access RAM like @code{LD}
16675 or @code{STS}. This offset depends on the device architecture and has
16676 to be subtracted from the RAM address in order to get the
16677 respective I/O@tie{}address.
16679 @item __AVR_SHORT_CALLS__
16680 The @option{-mshort-calls} command line option is set.
16682 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
16683 Some devices support reading from flash memory by means of @code{LD*}
16684 instructions. The flash memory is seen in the data address space
16685 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
16686 is not defined, this feature is not available. If defined,
16687 the address space is linear and there is no need to put
16688 @code{.rodata} into RAM. This is handled by the default linker
16689 description file, and is currently available for
16690 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
16691 there is no need to use address spaces like @code{__flash} or
16692 features like attribute @code{progmem} and @code{pgm_read_*}.
16694 @item __WITH_AVRLIBC__
16695 The compiler is configured to be used together with AVR-Libc.
16696 See the @option{--with-avrlibc} configure option.
16700 @node Blackfin Options
16701 @subsection Blackfin Options
16702 @cindex Blackfin Options
16705 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
16707 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
16708 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
16709 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
16710 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
16711 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
16712 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
16713 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
16714 @samp{bf561}, @samp{bf592}.
16716 The optional @var{sirevision} specifies the silicon revision of the target
16717 Blackfin processor. Any workarounds available for the targeted silicon revision
16718 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
16719 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
16720 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
16721 hexadecimal digits representing the major and minor numbers in the silicon
16722 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
16723 is not defined. If @var{sirevision} is @samp{any}, the
16724 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
16725 If this optional @var{sirevision} is not used, GCC assumes the latest known
16726 silicon revision of the targeted Blackfin processor.
16728 GCC defines a preprocessor macro for the specified @var{cpu}.
16729 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
16730 provided by libgloss to be linked in if @option{-msim} is not given.
16732 Without this option, @samp{bf532} is used as the processor by default.
16734 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
16735 only the preprocessor macro is defined.
16739 Specifies that the program will be run on the simulator. This causes
16740 the simulator BSP provided by libgloss to be linked in. This option
16741 has effect only for @samp{bfin-elf} toolchain.
16742 Certain other options, such as @option{-mid-shared-library} and
16743 @option{-mfdpic}, imply @option{-msim}.
16745 @item -momit-leaf-frame-pointer
16746 @opindex momit-leaf-frame-pointer
16747 Don't keep the frame pointer in a register for leaf functions. This
16748 avoids the instructions to save, set up and restore frame pointers and
16749 makes an extra register available in leaf functions. The option
16750 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
16751 which might make debugging harder.
16753 @item -mspecld-anomaly
16754 @opindex mspecld-anomaly
16755 When enabled, the compiler ensures that the generated code does not
16756 contain speculative loads after jump instructions. If this option is used,
16757 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
16759 @item -mno-specld-anomaly
16760 @opindex mno-specld-anomaly
16761 Don't generate extra code to prevent speculative loads from occurring.
16763 @item -mcsync-anomaly
16764 @opindex mcsync-anomaly
16765 When enabled, the compiler ensures that the generated code does not
16766 contain CSYNC or SSYNC instructions too soon after conditional branches.
16767 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
16769 @item -mno-csync-anomaly
16770 @opindex mno-csync-anomaly
16771 Don't generate extra code to prevent CSYNC or SSYNC instructions from
16772 occurring too soon after a conditional branch.
16776 When enabled, the compiler is free to take advantage of the knowledge that
16777 the entire program fits into the low 64k of memory.
16780 @opindex mno-low-64k
16781 Assume that the program is arbitrarily large. This is the default.
16783 @item -mstack-check-l1
16784 @opindex mstack-check-l1
16785 Do stack checking using information placed into L1 scratchpad memory by the
16788 @item -mid-shared-library
16789 @opindex mid-shared-library
16790 Generate code that supports shared libraries via the library ID method.
16791 This allows for execute in place and shared libraries in an environment
16792 without virtual memory management. This option implies @option{-fPIC}.
16793 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16795 @item -mno-id-shared-library
16796 @opindex mno-id-shared-library
16797 Generate code that doesn't assume ID-based shared libraries are being used.
16798 This is the default.
16800 @item -mleaf-id-shared-library
16801 @opindex mleaf-id-shared-library
16802 Generate code that supports shared libraries via the library ID method,
16803 but assumes that this library or executable won't link against any other
16804 ID shared libraries. That allows the compiler to use faster code for jumps
16807 @item -mno-leaf-id-shared-library
16808 @opindex mno-leaf-id-shared-library
16809 Do not assume that the code being compiled won't link against any ID shared
16810 libraries. Slower code is generated for jump and call insns.
16812 @item -mshared-library-id=n
16813 @opindex mshared-library-id
16814 Specifies the identification number of the ID-based shared library being
16815 compiled. Specifying a value of 0 generates more compact code; specifying
16816 other values forces the allocation of that number to the current
16817 library but is no more space- or time-efficient than omitting this option.
16821 Generate code that allows the data segment to be located in a different
16822 area of memory from the text segment. This allows for execute in place in
16823 an environment without virtual memory management by eliminating relocations
16824 against the text section.
16826 @item -mno-sep-data
16827 @opindex mno-sep-data
16828 Generate code that assumes that the data segment follows the text segment.
16829 This is the default.
16832 @itemx -mno-long-calls
16833 @opindex mlong-calls
16834 @opindex mno-long-calls
16835 Tells the compiler to perform function calls by first loading the
16836 address of the function into a register and then performing a subroutine
16837 call on this register. This switch is needed if the target function
16838 lies outside of the 24-bit addressing range of the offset-based
16839 version of subroutine call instruction.
16841 This feature is not enabled by default. Specifying
16842 @option{-mno-long-calls} restores the default behavior. Note these
16843 switches have no effect on how the compiler generates code to handle
16844 function calls via function pointers.
16848 Link with the fast floating-point library. This library relaxes some of
16849 the IEEE floating-point standard's rules for checking inputs against
16850 Not-a-Number (NAN), in the interest of performance.
16853 @opindex minline-plt
16854 Enable inlining of PLT entries in function calls to functions that are
16855 not known to bind locally. It has no effect without @option{-mfdpic}.
16858 @opindex mmulticore
16859 Build a standalone application for multicore Blackfin processors.
16860 This option causes proper start files and link scripts supporting
16861 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
16862 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
16864 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
16865 selects the one-application-per-core programming model. Without
16866 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
16867 programming model is used. In this model, the main function of Core B
16868 should be named as @code{coreb_main}.
16870 If this option is not used, the single-core application programming
16875 Build a standalone application for Core A of BF561 when using
16876 the one-application-per-core programming model. Proper start files
16877 and link scripts are used to support Core A, and the macro
16878 @code{__BFIN_COREA} is defined.
16879 This option can only be used in conjunction with @option{-mmulticore}.
16883 Build a standalone application for Core B of BF561 when using
16884 the one-application-per-core programming model. Proper start files
16885 and link scripts are used to support Core B, and the macro
16886 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
16887 should be used instead of @code{main}.
16888 This option can only be used in conjunction with @option{-mmulticore}.
16892 Build a standalone application for SDRAM. Proper start files and
16893 link scripts are used to put the application into SDRAM, and the macro
16894 @code{__BFIN_SDRAM} is defined.
16895 The loader should initialize SDRAM before loading the application.
16899 Assume that ICPLBs are enabled at run time. This has an effect on certain
16900 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
16901 are enabled; for standalone applications the default is off.
16905 @subsection C6X Options
16906 @cindex C6X Options
16909 @item -march=@var{name}
16911 This specifies the name of the target architecture. GCC uses this
16912 name to determine what kind of instructions it can emit when generating
16913 assembly code. Permissible names are: @samp{c62x},
16914 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
16917 @opindex mbig-endian
16918 Generate code for a big-endian target.
16920 @item -mlittle-endian
16921 @opindex mlittle-endian
16922 Generate code for a little-endian target. This is the default.
16926 Choose startup files and linker script suitable for the simulator.
16928 @item -msdata=default
16929 @opindex msdata=default
16930 Put small global and static data in the @code{.neardata} section,
16931 which is pointed to by register @code{B14}. Put small uninitialized
16932 global and static data in the @code{.bss} section, which is adjacent
16933 to the @code{.neardata} section. Put small read-only data into the
16934 @code{.rodata} section. The corresponding sections used for large
16935 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
16938 @opindex msdata=all
16939 Put all data, not just small objects, into the sections reserved for
16940 small data, and use addressing relative to the @code{B14} register to
16944 @opindex msdata=none
16945 Make no use of the sections reserved for small data, and use absolute
16946 addresses to access all data. Put all initialized global and static
16947 data in the @code{.fardata} section, and all uninitialized data in the
16948 @code{.far} section. Put all constant data into the @code{.const}
16953 @subsection CRIS Options
16954 @cindex CRIS Options
16956 These options are defined specifically for the CRIS ports.
16959 @item -march=@var{architecture-type}
16960 @itemx -mcpu=@var{architecture-type}
16963 Generate code for the specified architecture. The choices for
16964 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
16965 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
16966 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
16969 @item -mtune=@var{architecture-type}
16971 Tune to @var{architecture-type} everything applicable about the generated
16972 code, except for the ABI and the set of available instructions. The
16973 choices for @var{architecture-type} are the same as for
16974 @option{-march=@var{architecture-type}}.
16976 @item -mmax-stack-frame=@var{n}
16977 @opindex mmax-stack-frame
16978 Warn when the stack frame of a function exceeds @var{n} bytes.
16984 The options @option{-metrax4} and @option{-metrax100} are synonyms for
16985 @option{-march=v3} and @option{-march=v8} respectively.
16987 @item -mmul-bug-workaround
16988 @itemx -mno-mul-bug-workaround
16989 @opindex mmul-bug-workaround
16990 @opindex mno-mul-bug-workaround
16991 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
16992 models where it applies. This option is active by default.
16996 Enable CRIS-specific verbose debug-related information in the assembly
16997 code. This option also has the effect of turning off the @samp{#NO_APP}
16998 formatted-code indicator to the assembler at the beginning of the
17003 Do not use condition-code results from previous instruction; always emit
17004 compare and test instructions before use of condition codes.
17006 @item -mno-side-effects
17007 @opindex mno-side-effects
17008 Do not emit instructions with side effects in addressing modes other than
17011 @item -mstack-align
17012 @itemx -mno-stack-align
17013 @itemx -mdata-align
17014 @itemx -mno-data-align
17015 @itemx -mconst-align
17016 @itemx -mno-const-align
17017 @opindex mstack-align
17018 @opindex mno-stack-align
17019 @opindex mdata-align
17020 @opindex mno-data-align
17021 @opindex mconst-align
17022 @opindex mno-const-align
17023 These options (@samp{no-} options) arrange (eliminate arrangements) for the
17024 stack frame, individual data and constants to be aligned for the maximum
17025 single data access size for the chosen CPU model. The default is to
17026 arrange for 32-bit alignment. ABI details such as structure layout are
17027 not affected by these options.
17035 Similar to the stack- data- and const-align options above, these options
17036 arrange for stack frame, writable data and constants to all be 32-bit,
17037 16-bit or 8-bit aligned. The default is 32-bit alignment.
17039 @item -mno-prologue-epilogue
17040 @itemx -mprologue-epilogue
17041 @opindex mno-prologue-epilogue
17042 @opindex mprologue-epilogue
17043 With @option{-mno-prologue-epilogue}, the normal function prologue and
17044 epilogue which set up the stack frame are omitted and no return
17045 instructions or return sequences are generated in the code. Use this
17046 option only together with visual inspection of the compiled code: no
17047 warnings or errors are generated when call-saved registers must be saved,
17048 or storage for local variables needs to be allocated.
17052 @opindex mno-gotplt
17054 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
17055 instruction sequences that load addresses for functions from the PLT part
17056 of the GOT rather than (traditional on other architectures) calls to the
17057 PLT@. The default is @option{-mgotplt}.
17061 Legacy no-op option only recognized with the cris-axis-elf and
17062 cris-axis-linux-gnu targets.
17066 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
17070 This option, recognized for the cris-axis-elf, arranges
17071 to link with input-output functions from a simulator library. Code,
17072 initialized data and zero-initialized data are allocated consecutively.
17076 Like @option{-sim}, but pass linker options to locate initialized data at
17077 0x40000000 and zero-initialized data at 0x80000000.
17081 @subsection CR16 Options
17082 @cindex CR16 Options
17084 These options are defined specifically for the CR16 ports.
17090 Enable the use of multiply-accumulate instructions. Disabled by default.
17094 @opindex mcr16cplus
17096 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
17101 Links the library libsim.a which is in compatible with simulator. Applicable
17102 to ELF compiler only.
17106 Choose integer type as 32-bit wide.
17110 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
17112 @item -mdata-model=@var{model}
17113 @opindex mdata-model
17114 Choose a data model. The choices for @var{model} are @samp{near},
17115 @samp{far} or @samp{medium}. @samp{medium} is default.
17116 However, @samp{far} is not valid with @option{-mcr16c}, as the
17117 CR16C architecture does not support the far data model.
17120 @node Darwin Options
17121 @subsection Darwin Options
17122 @cindex Darwin options
17124 These options are defined for all architectures running the Darwin operating
17127 FSF GCC on Darwin does not create ``fat'' object files; it creates
17128 an object file for the single architecture that GCC was built to
17129 target. Apple's GCC on Darwin does create ``fat'' files if multiple
17130 @option{-arch} options are used; it does so by running the compiler or
17131 linker multiple times and joining the results together with
17134 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
17135 @samp{i686}) is determined by the flags that specify the ISA
17136 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
17137 @option{-force_cpusubtype_ALL} option can be used to override this.
17139 The Darwin tools vary in their behavior when presented with an ISA
17140 mismatch. The assembler, @file{as}, only permits instructions to
17141 be used that are valid for the subtype of the file it is generating,
17142 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
17143 The linker for shared libraries, @file{/usr/bin/libtool}, fails
17144 and prints an error if asked to create a shared library with a less
17145 restrictive subtype than its input files (for instance, trying to put
17146 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
17147 for executables, @command{ld}, quietly gives the executable the most
17148 restrictive subtype of any of its input files.
17153 Add the framework directory @var{dir} to the head of the list of
17154 directories to be searched for header files. These directories are
17155 interleaved with those specified by @option{-I} options and are
17156 scanned in a left-to-right order.
17158 A framework directory is a directory with frameworks in it. A
17159 framework is a directory with a @file{Headers} and/or
17160 @file{PrivateHeaders} directory contained directly in it that ends
17161 in @file{.framework}. The name of a framework is the name of this
17162 directory excluding the @file{.framework}. Headers associated with
17163 the framework are found in one of those two directories, with
17164 @file{Headers} being searched first. A subframework is a framework
17165 directory that is in a framework's @file{Frameworks} directory.
17166 Includes of subframework headers can only appear in a header of a
17167 framework that contains the subframework, or in a sibling subframework
17168 header. Two subframeworks are siblings if they occur in the same
17169 framework. A subframework should not have the same name as a
17170 framework; a warning is issued if this is violated. Currently a
17171 subframework cannot have subframeworks; in the future, the mechanism
17172 may be extended to support this. The standard frameworks can be found
17173 in @file{/System/Library/Frameworks} and
17174 @file{/Library/Frameworks}. An example include looks like
17175 @code{#include <Framework/header.h>}, where @file{Framework} denotes
17176 the name of the framework and @file{header.h} is found in the
17177 @file{PrivateHeaders} or @file{Headers} directory.
17179 @item -iframework@var{dir}
17180 @opindex iframework
17181 Like @option{-F} except the directory is a treated as a system
17182 directory. The main difference between this @option{-iframework} and
17183 @option{-F} is that with @option{-iframework} the compiler does not
17184 warn about constructs contained within header files found via
17185 @var{dir}. This option is valid only for the C family of languages.
17189 Emit debugging information for symbols that are used. For stabs
17190 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
17191 This is by default ON@.
17195 Emit debugging information for all symbols and types.
17197 @item -mmacosx-version-min=@var{version}
17198 The earliest version of MacOS X that this executable will run on
17199 is @var{version}. Typical values of @var{version} include @code{10.1},
17200 @code{10.2}, and @code{10.3.9}.
17202 If the compiler was built to use the system's headers by default,
17203 then the default for this option is the system version on which the
17204 compiler is running, otherwise the default is to make choices that
17205 are compatible with as many systems and code bases as possible.
17209 Enable kernel development mode. The @option{-mkernel} option sets
17210 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
17211 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
17212 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
17213 applicable. This mode also sets @option{-mno-altivec},
17214 @option{-msoft-float}, @option{-fno-builtin} and
17215 @option{-mlong-branch} for PowerPC targets.
17217 @item -mone-byte-bool
17218 @opindex mone-byte-bool
17219 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
17220 By default @code{sizeof(bool)} is @code{4} when compiling for
17221 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
17222 option has no effect on x86.
17224 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
17225 to generate code that is not binary compatible with code generated
17226 without that switch. Using this switch may require recompiling all
17227 other modules in a program, including system libraries. Use this
17228 switch to conform to a non-default data model.
17230 @item -mfix-and-continue
17231 @itemx -ffix-and-continue
17232 @itemx -findirect-data
17233 @opindex mfix-and-continue
17234 @opindex ffix-and-continue
17235 @opindex findirect-data
17236 Generate code suitable for fast turnaround development, such as to
17237 allow GDB to dynamically load @file{.o} files into already-running
17238 programs. @option{-findirect-data} and @option{-ffix-and-continue}
17239 are provided for backwards compatibility.
17243 Loads all members of static archive libraries.
17244 See man ld(1) for more information.
17246 @item -arch_errors_fatal
17247 @opindex arch_errors_fatal
17248 Cause the errors having to do with files that have the wrong architecture
17251 @item -bind_at_load
17252 @opindex bind_at_load
17253 Causes the output file to be marked such that the dynamic linker will
17254 bind all undefined references when the file is loaded or launched.
17258 Produce a Mach-o bundle format file.
17259 See man ld(1) for more information.
17261 @item -bundle_loader @var{executable}
17262 @opindex bundle_loader
17263 This option specifies the @var{executable} that will load the build
17264 output file being linked. See man ld(1) for more information.
17267 @opindex dynamiclib
17268 When passed this option, GCC produces a dynamic library instead of
17269 an executable when linking, using the Darwin @file{libtool} command.
17271 @item -force_cpusubtype_ALL
17272 @opindex force_cpusubtype_ALL
17273 This causes GCC's output file to have the @samp{ALL} subtype, instead of
17274 one controlled by the @option{-mcpu} or @option{-march} option.
17276 @item -allowable_client @var{client_name}
17277 @itemx -client_name
17278 @itemx -compatibility_version
17279 @itemx -current_version
17281 @itemx -dependency-file
17283 @itemx -dylinker_install_name
17285 @itemx -exported_symbols_list
17288 @itemx -flat_namespace
17289 @itemx -force_flat_namespace
17290 @itemx -headerpad_max_install_names
17293 @itemx -install_name
17294 @itemx -keep_private_externs
17295 @itemx -multi_module
17296 @itemx -multiply_defined
17297 @itemx -multiply_defined_unused
17300 @itemx -no_dead_strip_inits_and_terms
17301 @itemx -nofixprebinding
17302 @itemx -nomultidefs
17304 @itemx -noseglinkedit
17305 @itemx -pagezero_size
17307 @itemx -prebind_all_twolevel_modules
17308 @itemx -private_bundle
17310 @itemx -read_only_relocs
17312 @itemx -sectobjectsymbols
17316 @itemx -sectobjectsymbols
17319 @itemx -segs_read_only_addr
17321 @itemx -segs_read_write_addr
17322 @itemx -seg_addr_table
17323 @itemx -seg_addr_table_filename
17324 @itemx -seglinkedit
17326 @itemx -segs_read_only_addr
17327 @itemx -segs_read_write_addr
17328 @itemx -single_module
17330 @itemx -sub_library
17332 @itemx -sub_umbrella
17333 @itemx -twolevel_namespace
17336 @itemx -unexported_symbols_list
17337 @itemx -weak_reference_mismatches
17338 @itemx -whatsloaded
17339 @opindex allowable_client
17340 @opindex client_name
17341 @opindex compatibility_version
17342 @opindex current_version
17343 @opindex dead_strip
17344 @opindex dependency-file
17345 @opindex dylib_file
17346 @opindex dylinker_install_name
17348 @opindex exported_symbols_list
17350 @opindex flat_namespace
17351 @opindex force_flat_namespace
17352 @opindex headerpad_max_install_names
17353 @opindex image_base
17355 @opindex install_name
17356 @opindex keep_private_externs
17357 @opindex multi_module
17358 @opindex multiply_defined
17359 @opindex multiply_defined_unused
17360 @opindex noall_load
17361 @opindex no_dead_strip_inits_and_terms
17362 @opindex nofixprebinding
17363 @opindex nomultidefs
17365 @opindex noseglinkedit
17366 @opindex pagezero_size
17368 @opindex prebind_all_twolevel_modules
17369 @opindex private_bundle
17370 @opindex read_only_relocs
17372 @opindex sectobjectsymbols
17375 @opindex sectcreate
17376 @opindex sectobjectsymbols
17379 @opindex segs_read_only_addr
17380 @opindex segs_read_write_addr
17381 @opindex seg_addr_table
17382 @opindex seg_addr_table_filename
17383 @opindex seglinkedit
17385 @opindex segs_read_only_addr
17386 @opindex segs_read_write_addr
17387 @opindex single_module
17389 @opindex sub_library
17390 @opindex sub_umbrella
17391 @opindex twolevel_namespace
17394 @opindex unexported_symbols_list
17395 @opindex weak_reference_mismatches
17396 @opindex whatsloaded
17397 These options are passed to the Darwin linker. The Darwin linker man page
17398 describes them in detail.
17401 @node DEC Alpha Options
17402 @subsection DEC Alpha Options
17404 These @samp{-m} options are defined for the DEC Alpha implementations:
17407 @item -mno-soft-float
17408 @itemx -msoft-float
17409 @opindex mno-soft-float
17410 @opindex msoft-float
17411 Use (do not use) the hardware floating-point instructions for
17412 floating-point operations. When @option{-msoft-float} is specified,
17413 functions in @file{libgcc.a} are used to perform floating-point
17414 operations. Unless they are replaced by routines that emulate the
17415 floating-point operations, or compiled in such a way as to call such
17416 emulations routines, these routines issue floating-point
17417 operations. If you are compiling for an Alpha without floating-point
17418 operations, you must ensure that the library is built so as not to call
17421 Note that Alpha implementations without floating-point operations are
17422 required to have floating-point registers.
17425 @itemx -mno-fp-regs
17427 @opindex mno-fp-regs
17428 Generate code that uses (does not use) the floating-point register set.
17429 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
17430 register set is not used, floating-point operands are passed in integer
17431 registers as if they were integers and floating-point results are passed
17432 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
17433 so any function with a floating-point argument or return value called by code
17434 compiled with @option{-mno-fp-regs} must also be compiled with that
17437 A typical use of this option is building a kernel that does not use,
17438 and hence need not save and restore, any floating-point registers.
17442 The Alpha architecture implements floating-point hardware optimized for
17443 maximum performance. It is mostly compliant with the IEEE floating-point
17444 standard. However, for full compliance, software assistance is
17445 required. This option generates code fully IEEE-compliant code
17446 @emph{except} that the @var{inexact-flag} is not maintained (see below).
17447 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
17448 defined during compilation. The resulting code is less efficient but is
17449 able to correctly support denormalized numbers and exceptional IEEE
17450 values such as not-a-number and plus/minus infinity. Other Alpha
17451 compilers call this option @option{-ieee_with_no_inexact}.
17453 @item -mieee-with-inexact
17454 @opindex mieee-with-inexact
17455 This is like @option{-mieee} except the generated code also maintains
17456 the IEEE @var{inexact-flag}. Turning on this option causes the
17457 generated code to implement fully-compliant IEEE math. In addition to
17458 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
17459 macro. On some Alpha implementations the resulting code may execute
17460 significantly slower than the code generated by default. Since there is
17461 very little code that depends on the @var{inexact-flag}, you should
17462 normally not specify this option. Other Alpha compilers call this
17463 option @option{-ieee_with_inexact}.
17465 @item -mfp-trap-mode=@var{trap-mode}
17466 @opindex mfp-trap-mode
17467 This option controls what floating-point related traps are enabled.
17468 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
17469 The trap mode can be set to one of four values:
17473 This is the default (normal) setting. The only traps that are enabled
17474 are the ones that cannot be disabled in software (e.g., division by zero
17478 In addition to the traps enabled by @samp{n}, underflow traps are enabled
17482 Like @samp{u}, but the instructions are marked to be safe for software
17483 completion (see Alpha architecture manual for details).
17486 Like @samp{su}, but inexact traps are enabled as well.
17489 @item -mfp-rounding-mode=@var{rounding-mode}
17490 @opindex mfp-rounding-mode
17491 Selects the IEEE rounding mode. Other Alpha compilers call this option
17492 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
17497 Normal IEEE rounding mode. Floating-point numbers are rounded towards
17498 the nearest machine number or towards the even machine number in case
17502 Round towards minus infinity.
17505 Chopped rounding mode. Floating-point numbers are rounded towards zero.
17508 Dynamic rounding mode. A field in the floating-point control register
17509 (@var{fpcr}, see Alpha architecture reference manual) controls the
17510 rounding mode in effect. The C library initializes this register for
17511 rounding towards plus infinity. Thus, unless your program modifies the
17512 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
17515 @item -mtrap-precision=@var{trap-precision}
17516 @opindex mtrap-precision
17517 In the Alpha architecture, floating-point traps are imprecise. This
17518 means without software assistance it is impossible to recover from a
17519 floating trap and program execution normally needs to be terminated.
17520 GCC can generate code that can assist operating system trap handlers
17521 in determining the exact location that caused a floating-point trap.
17522 Depending on the requirements of an application, different levels of
17523 precisions can be selected:
17527 Program precision. This option is the default and means a trap handler
17528 can only identify which program caused a floating-point exception.
17531 Function precision. The trap handler can determine the function that
17532 caused a floating-point exception.
17535 Instruction precision. The trap handler can determine the exact
17536 instruction that caused a floating-point exception.
17539 Other Alpha compilers provide the equivalent options called
17540 @option{-scope_safe} and @option{-resumption_safe}.
17542 @item -mieee-conformant
17543 @opindex mieee-conformant
17544 This option marks the generated code as IEEE conformant. You must not
17545 use this option unless you also specify @option{-mtrap-precision=i} and either
17546 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
17547 is to emit the line @samp{.eflag 48} in the function prologue of the
17548 generated assembly file.
17550 @item -mbuild-constants
17551 @opindex mbuild-constants
17552 Normally GCC examines a 32- or 64-bit integer constant to
17553 see if it can construct it from smaller constants in two or three
17554 instructions. If it cannot, it outputs the constant as a literal and
17555 generates code to load it from the data segment at run time.
17557 Use this option to require GCC to construct @emph{all} integer constants
17558 using code, even if it takes more instructions (the maximum is six).
17560 You typically use this option to build a shared library dynamic
17561 loader. Itself a shared library, it must relocate itself in memory
17562 before it can find the variables and constants in its own data segment.
17580 Indicate whether GCC should generate code to use the optional BWX,
17581 CIX, FIX and MAX instruction sets. The default is to use the instruction
17582 sets supported by the CPU type specified via @option{-mcpu=} option or that
17583 of the CPU on which GCC was built if none is specified.
17586 @itemx -mfloat-ieee
17587 @opindex mfloat-vax
17588 @opindex mfloat-ieee
17589 Generate code that uses (does not use) VAX F and G floating-point
17590 arithmetic instead of IEEE single and double precision.
17592 @item -mexplicit-relocs
17593 @itemx -mno-explicit-relocs
17594 @opindex mexplicit-relocs
17595 @opindex mno-explicit-relocs
17596 Older Alpha assemblers provided no way to generate symbol relocations
17597 except via assembler macros. Use of these macros does not allow
17598 optimal instruction scheduling. GNU binutils as of version 2.12
17599 supports a new syntax that allows the compiler to explicitly mark
17600 which relocations should apply to which instructions. This option
17601 is mostly useful for debugging, as GCC detects the capabilities of
17602 the assembler when it is built and sets the default accordingly.
17605 @itemx -mlarge-data
17606 @opindex msmall-data
17607 @opindex mlarge-data
17608 When @option{-mexplicit-relocs} is in effect, static data is
17609 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
17610 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
17611 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
17612 16-bit relocations off of the @code{$gp} register. This limits the
17613 size of the small data area to 64KB, but allows the variables to be
17614 directly accessed via a single instruction.
17616 The default is @option{-mlarge-data}. With this option the data area
17617 is limited to just below 2GB@. Programs that require more than 2GB of
17618 data must use @code{malloc} or @code{mmap} to allocate the data in the
17619 heap instead of in the program's data segment.
17621 When generating code for shared libraries, @option{-fpic} implies
17622 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
17625 @itemx -mlarge-text
17626 @opindex msmall-text
17627 @opindex mlarge-text
17628 When @option{-msmall-text} is used, the compiler assumes that the
17629 code of the entire program (or shared library) fits in 4MB, and is
17630 thus reachable with a branch instruction. When @option{-msmall-data}
17631 is used, the compiler can assume that all local symbols share the
17632 same @code{$gp} value, and thus reduce the number of instructions
17633 required for a function call from 4 to 1.
17635 The default is @option{-mlarge-text}.
17637 @item -mcpu=@var{cpu_type}
17639 Set the instruction set and instruction scheduling parameters for
17640 machine type @var{cpu_type}. You can specify either the @samp{EV}
17641 style name or the corresponding chip number. GCC supports scheduling
17642 parameters for the EV4, EV5 and EV6 family of processors and
17643 chooses the default values for the instruction set from the processor
17644 you specify. If you do not specify a processor type, GCC defaults
17645 to the processor on which the compiler was built.
17647 Supported values for @var{cpu_type} are
17653 Schedules as an EV4 and has no instruction set extensions.
17657 Schedules as an EV5 and has no instruction set extensions.
17661 Schedules as an EV5 and supports the BWX extension.
17666 Schedules as an EV5 and supports the BWX and MAX extensions.
17670 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
17674 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
17677 Native toolchains also support the value @samp{native},
17678 which selects the best architecture option for the host processor.
17679 @option{-mcpu=native} has no effect if GCC does not recognize
17682 @item -mtune=@var{cpu_type}
17684 Set only the instruction scheduling parameters for machine type
17685 @var{cpu_type}. The instruction set is not changed.
17687 Native toolchains also support the value @samp{native},
17688 which selects the best architecture option for the host processor.
17689 @option{-mtune=native} has no effect if GCC does not recognize
17692 @item -mmemory-latency=@var{time}
17693 @opindex mmemory-latency
17694 Sets the latency the scheduler should assume for typical memory
17695 references as seen by the application. This number is highly
17696 dependent on the memory access patterns used by the application
17697 and the size of the external cache on the machine.
17699 Valid options for @var{time} are
17703 A decimal number representing clock cycles.
17709 The compiler contains estimates of the number of clock cycles for
17710 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
17711 (also called Dcache, Scache, and Bcache), as well as to main memory.
17712 Note that L3 is only valid for EV5.
17718 @subsection FR30 Options
17719 @cindex FR30 Options
17721 These options are defined specifically for the FR30 port.
17725 @item -msmall-model
17726 @opindex msmall-model
17727 Use the small address space model. This can produce smaller code, but
17728 it does assume that all symbolic values and addresses fit into a
17733 Assume that runtime support has been provided and so there is no need
17734 to include the simulator library (@file{libsim.a}) on the linker
17740 @subsection FT32 Options
17741 @cindex FT32 Options
17743 These options are defined specifically for the FT32 port.
17749 Specifies that the program will be run on the simulator. This causes
17750 an alternate runtime startup and library to be linked.
17751 You must not use this option when generating programs that will run on
17752 real hardware; you must provide your own runtime library for whatever
17753 I/O functions are needed.
17757 Enable Local Register Allocation. This is still experimental for FT32,
17758 so by default the compiler uses standard reload.
17762 Do not use div and mod instructions.
17767 @subsection FRV Options
17768 @cindex FRV Options
17774 Only use the first 32 general-purpose registers.
17779 Use all 64 general-purpose registers.
17784 Use only the first 32 floating-point registers.
17789 Use all 64 floating-point registers.
17792 @opindex mhard-float
17794 Use hardware instructions for floating-point operations.
17797 @opindex msoft-float
17799 Use library routines for floating-point operations.
17804 Dynamically allocate condition code registers.
17809 Do not try to dynamically allocate condition code registers, only
17810 use @code{icc0} and @code{fcc0}.
17815 Change ABI to use double word insns.
17820 Do not use double word instructions.
17825 Use floating-point double instructions.
17828 @opindex mno-double
17830 Do not use floating-point double instructions.
17835 Use media instructions.
17840 Do not use media instructions.
17845 Use multiply and add/subtract instructions.
17848 @opindex mno-muladd
17850 Do not use multiply and add/subtract instructions.
17855 Select the FDPIC ABI, which uses function descriptors to represent
17856 pointers to functions. Without any PIC/PIE-related options, it
17857 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
17858 assumes GOT entries and small data are within a 12-bit range from the
17859 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
17860 are computed with 32 bits.
17861 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17864 @opindex minline-plt
17866 Enable inlining of PLT entries in function calls to functions that are
17867 not known to bind locally. It has no effect without @option{-mfdpic}.
17868 It's enabled by default if optimizing for speed and compiling for
17869 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
17870 optimization option such as @option{-O3} or above is present in the
17876 Assume a large TLS segment when generating thread-local code.
17881 Do not assume a large TLS segment when generating thread-local code.
17886 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
17887 that is known to be in read-only sections. It's enabled by default,
17888 except for @option{-fpic} or @option{-fpie}: even though it may help
17889 make the global offset table smaller, it trades 1 instruction for 4.
17890 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
17891 one of which may be shared by multiple symbols, and it avoids the need
17892 for a GOT entry for the referenced symbol, so it's more likely to be a
17893 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
17895 @item -multilib-library-pic
17896 @opindex multilib-library-pic
17898 Link with the (library, not FD) pic libraries. It's implied by
17899 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
17900 @option{-fpic} without @option{-mfdpic}. You should never have to use
17904 @opindex mlinked-fp
17906 Follow the EABI requirement of always creating a frame pointer whenever
17907 a stack frame is allocated. This option is enabled by default and can
17908 be disabled with @option{-mno-linked-fp}.
17911 @opindex mlong-calls
17913 Use indirect addressing to call functions outside the current
17914 compilation unit. This allows the functions to be placed anywhere
17915 within the 32-bit address space.
17917 @item -malign-labels
17918 @opindex malign-labels
17920 Try to align labels to an 8-byte boundary by inserting NOPs into the
17921 previous packet. This option only has an effect when VLIW packing
17922 is enabled. It doesn't create new packets; it merely adds NOPs to
17925 @item -mlibrary-pic
17926 @opindex mlibrary-pic
17928 Generate position-independent EABI code.
17933 Use only the first four media accumulator registers.
17938 Use all eight media accumulator registers.
17943 Pack VLIW instructions.
17948 Do not pack VLIW instructions.
17951 @opindex mno-eflags
17953 Do not mark ABI switches in e_flags.
17956 @opindex mcond-move
17958 Enable the use of conditional-move instructions (default).
17960 This switch is mainly for debugging the compiler and will likely be removed
17961 in a future version.
17963 @item -mno-cond-move
17964 @opindex mno-cond-move
17966 Disable the use of conditional-move instructions.
17968 This switch is mainly for debugging the compiler and will likely be removed
17969 in a future version.
17974 Enable the use of conditional set instructions (default).
17976 This switch is mainly for debugging the compiler and will likely be removed
17977 in a future version.
17982 Disable the use of conditional set instructions.
17984 This switch is mainly for debugging the compiler and will likely be removed
17985 in a future version.
17988 @opindex mcond-exec
17990 Enable the use of conditional execution (default).
17992 This switch is mainly for debugging the compiler and will likely be removed
17993 in a future version.
17995 @item -mno-cond-exec
17996 @opindex mno-cond-exec
17998 Disable the use of conditional execution.
18000 This switch is mainly for debugging the compiler and will likely be removed
18001 in a future version.
18003 @item -mvliw-branch
18004 @opindex mvliw-branch
18006 Run a pass to pack branches into VLIW instructions (default).
18008 This switch is mainly for debugging the compiler and will likely be removed
18009 in a future version.
18011 @item -mno-vliw-branch
18012 @opindex mno-vliw-branch
18014 Do not run a pass to pack branches into VLIW instructions.
18016 This switch is mainly for debugging the compiler and will likely be removed
18017 in a future version.
18019 @item -mmulti-cond-exec
18020 @opindex mmulti-cond-exec
18022 Enable optimization of @code{&&} and @code{||} in conditional execution
18025 This switch is mainly for debugging the compiler and will likely be removed
18026 in a future version.
18028 @item -mno-multi-cond-exec
18029 @opindex mno-multi-cond-exec
18031 Disable optimization of @code{&&} and @code{||} in conditional execution.
18033 This switch is mainly for debugging the compiler and will likely be removed
18034 in a future version.
18036 @item -mnested-cond-exec
18037 @opindex mnested-cond-exec
18039 Enable nested conditional execution optimizations (default).
18041 This switch is mainly for debugging the compiler and will likely be removed
18042 in a future version.
18044 @item -mno-nested-cond-exec
18045 @opindex mno-nested-cond-exec
18047 Disable nested conditional execution optimizations.
18049 This switch is mainly for debugging the compiler and will likely be removed
18050 in a future version.
18052 @item -moptimize-membar
18053 @opindex moptimize-membar
18055 This switch removes redundant @code{membar} instructions from the
18056 compiler-generated code. It is enabled by default.
18058 @item -mno-optimize-membar
18059 @opindex mno-optimize-membar
18061 This switch disables the automatic removal of redundant @code{membar}
18062 instructions from the generated code.
18064 @item -mtomcat-stats
18065 @opindex mtomcat-stats
18067 Cause gas to print out tomcat statistics.
18069 @item -mcpu=@var{cpu}
18072 Select the processor type for which to generate code. Possible values are
18073 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
18074 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
18078 @node GNU/Linux Options
18079 @subsection GNU/Linux Options
18081 These @samp{-m} options are defined for GNU/Linux targets:
18086 Use the GNU C library. This is the default except
18087 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
18088 @samp{*-*-linux-*android*} targets.
18092 Use uClibc C library. This is the default on
18093 @samp{*-*-linux-*uclibc*} targets.
18097 Use the musl C library. This is the default on
18098 @samp{*-*-linux-*musl*} targets.
18102 Use Bionic C library. This is the default on
18103 @samp{*-*-linux-*android*} targets.
18107 Compile code compatible with Android platform. This is the default on
18108 @samp{*-*-linux-*android*} targets.
18110 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
18111 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
18112 this option makes the GCC driver pass Android-specific options to the linker.
18113 Finally, this option causes the preprocessor macro @code{__ANDROID__}
18116 @item -tno-android-cc
18117 @opindex tno-android-cc
18118 Disable compilation effects of @option{-mandroid}, i.e., do not enable
18119 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
18120 @option{-fno-rtti} by default.
18122 @item -tno-android-ld
18123 @opindex tno-android-ld
18124 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
18125 linking options to the linker.
18129 @node H8/300 Options
18130 @subsection H8/300 Options
18132 These @samp{-m} options are defined for the H8/300 implementations:
18137 Shorten some address references at link time, when possible; uses the
18138 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
18139 ld, Using ld}, for a fuller description.
18143 Generate code for the H8/300H@.
18147 Generate code for the H8S@.
18151 Generate code for the H8S and H8/300H in the normal mode. This switch
18152 must be used either with @option{-mh} or @option{-ms}.
18156 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
18160 Extended registers are stored on stack before execution of function
18161 with monitor attribute. Default option is @option{-mexr}.
18162 This option is valid only for H8S targets.
18166 Extended registers are not stored on stack before execution of function
18167 with monitor attribute. Default option is @option{-mno-exr}.
18168 This option is valid only for H8S targets.
18172 Make @code{int} data 32 bits by default.
18175 @opindex malign-300
18176 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
18177 The default for the H8/300H and H8S is to align longs and floats on
18179 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
18180 This option has no effect on the H8/300.
18184 @subsection HPPA Options
18185 @cindex HPPA Options
18187 These @samp{-m} options are defined for the HPPA family of computers:
18190 @item -march=@var{architecture-type}
18192 Generate code for the specified architecture. The choices for
18193 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
18194 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
18195 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
18196 architecture option for your machine. Code compiled for lower numbered
18197 architectures runs on higher numbered architectures, but not the
18200 @item -mpa-risc-1-0
18201 @itemx -mpa-risc-1-1
18202 @itemx -mpa-risc-2-0
18203 @opindex mpa-risc-1-0
18204 @opindex mpa-risc-1-1
18205 @opindex mpa-risc-2-0
18206 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
18208 @item -mcaller-copies
18209 @opindex mcaller-copies
18210 The caller copies function arguments passed by hidden reference. This
18211 option should be used with care as it is not compatible with the default
18212 32-bit runtime. However, only aggregates larger than eight bytes are
18213 passed by hidden reference and the option provides better compatibility
18216 @item -mjump-in-delay
18217 @opindex mjump-in-delay
18218 This option is ignored and provided for compatibility purposes only.
18220 @item -mdisable-fpregs
18221 @opindex mdisable-fpregs
18222 Prevent floating-point registers from being used in any manner. This is
18223 necessary for compiling kernels that perform lazy context switching of
18224 floating-point registers. If you use this option and attempt to perform
18225 floating-point operations, the compiler aborts.
18227 @item -mdisable-indexing
18228 @opindex mdisable-indexing
18229 Prevent the compiler from using indexing address modes. This avoids some
18230 rather obscure problems when compiling MIG generated code under MACH@.
18232 @item -mno-space-regs
18233 @opindex mno-space-regs
18234 Generate code that assumes the target has no space registers. This allows
18235 GCC to generate faster indirect calls and use unscaled index address modes.
18237 Such code is suitable for level 0 PA systems and kernels.
18239 @item -mfast-indirect-calls
18240 @opindex mfast-indirect-calls
18241 Generate code that assumes calls never cross space boundaries. This
18242 allows GCC to emit code that performs faster indirect calls.
18244 This option does not work in the presence of shared libraries or nested
18247 @item -mfixed-range=@var{register-range}
18248 @opindex mfixed-range
18249 Generate code treating the given register range as fixed registers.
18250 A fixed register is one that the register allocator cannot use. This is
18251 useful when compiling kernel code. A register range is specified as
18252 two registers separated by a dash. Multiple register ranges can be
18253 specified separated by a comma.
18255 @item -mlong-load-store
18256 @opindex mlong-load-store
18257 Generate 3-instruction load and store sequences as sometimes required by
18258 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
18261 @item -mportable-runtime
18262 @opindex mportable-runtime
18263 Use the portable calling conventions proposed by HP for ELF systems.
18267 Enable the use of assembler directives only GAS understands.
18269 @item -mschedule=@var{cpu-type}
18271 Schedule code according to the constraints for the machine type
18272 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
18273 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
18274 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
18275 proper scheduling option for your machine. The default scheduling is
18279 @opindex mlinker-opt
18280 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
18281 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
18282 linkers in which they give bogus error messages when linking some programs.
18285 @opindex msoft-float
18286 Generate output containing library calls for floating point.
18287 @strong{Warning:} the requisite libraries are not available for all HPPA
18288 targets. Normally the facilities of the machine's usual C compiler are
18289 used, but this cannot be done directly in cross-compilation. You must make
18290 your own arrangements to provide suitable library functions for
18293 @option{-msoft-float} changes the calling convention in the output file;
18294 therefore, it is only useful if you compile @emph{all} of a program with
18295 this option. In particular, you need to compile @file{libgcc.a}, the
18296 library that comes with GCC, with @option{-msoft-float} in order for
18301 Generate the predefine, @code{_SIO}, for server IO@. The default is
18302 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
18303 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
18304 options are available under HP-UX and HI-UX@.
18308 Use options specific to GNU @command{ld}.
18309 This passes @option{-shared} to @command{ld} when
18310 building a shared library. It is the default when GCC is configured,
18311 explicitly or implicitly, with the GNU linker. This option does not
18312 affect which @command{ld} is called; it only changes what parameters
18313 are passed to that @command{ld}.
18314 The @command{ld} that is called is determined by the
18315 @option{--with-ld} configure option, GCC's program search path, and
18316 finally by the user's @env{PATH}. The linker used by GCC can be printed
18317 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
18318 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18322 Use options specific to HP @command{ld}.
18323 This passes @option{-b} to @command{ld} when building
18324 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
18325 links. It is the default when GCC is configured, explicitly or
18326 implicitly, with the HP linker. This option does not affect
18327 which @command{ld} is called; it only changes what parameters are passed to that
18329 The @command{ld} that is called is determined by the @option{--with-ld}
18330 configure option, GCC's program search path, and finally by the user's
18331 @env{PATH}. The linker used by GCC can be printed using @samp{which
18332 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
18333 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18336 @opindex mno-long-calls
18337 Generate code that uses long call sequences. This ensures that a call
18338 is always able to reach linker generated stubs. The default is to generate
18339 long calls only when the distance from the call site to the beginning
18340 of the function or translation unit, as the case may be, exceeds a
18341 predefined limit set by the branch type being used. The limits for
18342 normal calls are 7,600,000 and 240,000 bytes, respectively for the
18343 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
18346 Distances are measured from the beginning of functions when using the
18347 @option{-ffunction-sections} option, or when using the @option{-mgas}
18348 and @option{-mno-portable-runtime} options together under HP-UX with
18351 It is normally not desirable to use this option as it degrades
18352 performance. However, it may be useful in large applications,
18353 particularly when partial linking is used to build the application.
18355 The types of long calls used depends on the capabilities of the
18356 assembler and linker, and the type of code being generated. The
18357 impact on systems that support long absolute calls, and long pic
18358 symbol-difference or pc-relative calls should be relatively small.
18359 However, an indirect call is used on 32-bit ELF systems in pic code
18360 and it is quite long.
18362 @item -munix=@var{unix-std}
18364 Generate compiler predefines and select a startfile for the specified
18365 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
18366 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
18367 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
18368 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
18369 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
18372 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
18373 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
18374 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
18375 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
18376 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
18377 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
18379 It is @emph{important} to note that this option changes the interfaces
18380 for various library routines. It also affects the operational behavior
18381 of the C library. Thus, @emph{extreme} care is needed in using this
18384 Library code that is intended to operate with more than one UNIX
18385 standard must test, set and restore the variable @code{__xpg4_extended_mask}
18386 as appropriate. Most GNU software doesn't provide this capability.
18390 Suppress the generation of link options to search libdld.sl when the
18391 @option{-static} option is specified on HP-UX 10 and later.
18395 The HP-UX implementation of setlocale in libc has a dependency on
18396 libdld.sl. There isn't an archive version of libdld.sl. Thus,
18397 when the @option{-static} option is specified, special link options
18398 are needed to resolve this dependency.
18400 On HP-UX 10 and later, the GCC driver adds the necessary options to
18401 link with libdld.sl when the @option{-static} option is specified.
18402 This causes the resulting binary to be dynamic. On the 64-bit port,
18403 the linkers generate dynamic binaries by default in any case. The
18404 @option{-nolibdld} option can be used to prevent the GCC driver from
18405 adding these link options.
18409 Add support for multithreading with the @dfn{dce thread} library
18410 under HP-UX@. This option sets flags for both the preprocessor and
18414 @node IA-64 Options
18415 @subsection IA-64 Options
18416 @cindex IA-64 Options
18418 These are the @samp{-m} options defined for the Intel IA-64 architecture.
18422 @opindex mbig-endian
18423 Generate code for a big-endian target. This is the default for HP-UX@.
18425 @item -mlittle-endian
18426 @opindex mlittle-endian
18427 Generate code for a little-endian target. This is the default for AIX5
18433 @opindex mno-gnu-as
18434 Generate (or don't) code for the GNU assembler. This is the default.
18435 @c Also, this is the default if the configure option @option{--with-gnu-as}
18441 @opindex mno-gnu-ld
18442 Generate (or don't) code for the GNU linker. This is the default.
18443 @c Also, this is the default if the configure option @option{--with-gnu-ld}
18448 Generate code that does not use a global pointer register. The result
18449 is not position independent code, and violates the IA-64 ABI@.
18451 @item -mvolatile-asm-stop
18452 @itemx -mno-volatile-asm-stop
18453 @opindex mvolatile-asm-stop
18454 @opindex mno-volatile-asm-stop
18455 Generate (or don't) a stop bit immediately before and after volatile asm
18458 @item -mregister-names
18459 @itemx -mno-register-names
18460 @opindex mregister-names
18461 @opindex mno-register-names
18462 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
18463 the stacked registers. This may make assembler output more readable.
18469 Disable (or enable) optimizations that use the small data section. This may
18470 be useful for working around optimizer bugs.
18472 @item -mconstant-gp
18473 @opindex mconstant-gp
18474 Generate code that uses a single constant global pointer value. This is
18475 useful when compiling kernel code.
18479 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
18480 This is useful when compiling firmware code.
18482 @item -minline-float-divide-min-latency
18483 @opindex minline-float-divide-min-latency
18484 Generate code for inline divides of floating-point values
18485 using the minimum latency algorithm.
18487 @item -minline-float-divide-max-throughput
18488 @opindex minline-float-divide-max-throughput
18489 Generate code for inline divides of floating-point values
18490 using the maximum throughput algorithm.
18492 @item -mno-inline-float-divide
18493 @opindex mno-inline-float-divide
18494 Do not generate inline code for divides of floating-point values.
18496 @item -minline-int-divide-min-latency
18497 @opindex minline-int-divide-min-latency
18498 Generate code for inline divides of integer values
18499 using the minimum latency algorithm.
18501 @item -minline-int-divide-max-throughput
18502 @opindex minline-int-divide-max-throughput
18503 Generate code for inline divides of integer values
18504 using the maximum throughput algorithm.
18506 @item -mno-inline-int-divide
18507 @opindex mno-inline-int-divide
18508 Do not generate inline code for divides of integer values.
18510 @item -minline-sqrt-min-latency
18511 @opindex minline-sqrt-min-latency
18512 Generate code for inline square roots
18513 using the minimum latency algorithm.
18515 @item -minline-sqrt-max-throughput
18516 @opindex minline-sqrt-max-throughput
18517 Generate code for inline square roots
18518 using the maximum throughput algorithm.
18520 @item -mno-inline-sqrt
18521 @opindex mno-inline-sqrt
18522 Do not generate inline code for @code{sqrt}.
18525 @itemx -mno-fused-madd
18526 @opindex mfused-madd
18527 @opindex mno-fused-madd
18528 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
18529 instructions. The default is to use these instructions.
18531 @item -mno-dwarf2-asm
18532 @itemx -mdwarf2-asm
18533 @opindex mno-dwarf2-asm
18534 @opindex mdwarf2-asm
18535 Don't (or do) generate assembler code for the DWARF line number debugging
18536 info. This may be useful when not using the GNU assembler.
18538 @item -mearly-stop-bits
18539 @itemx -mno-early-stop-bits
18540 @opindex mearly-stop-bits
18541 @opindex mno-early-stop-bits
18542 Allow stop bits to be placed earlier than immediately preceding the
18543 instruction that triggered the stop bit. This can improve instruction
18544 scheduling, but does not always do so.
18546 @item -mfixed-range=@var{register-range}
18547 @opindex mfixed-range
18548 Generate code treating the given register range as fixed registers.
18549 A fixed register is one that the register allocator cannot use. This is
18550 useful when compiling kernel code. A register range is specified as
18551 two registers separated by a dash. Multiple register ranges can be
18552 specified separated by a comma.
18554 @item -mtls-size=@var{tls-size}
18556 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
18559 @item -mtune=@var{cpu-type}
18561 Tune the instruction scheduling for a particular CPU, Valid values are
18562 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
18563 and @samp{mckinley}.
18569 Generate code for a 32-bit or 64-bit environment.
18570 The 32-bit environment sets int, long and pointer to 32 bits.
18571 The 64-bit environment sets int to 32 bits and long and pointer
18572 to 64 bits. These are HP-UX specific flags.
18574 @item -mno-sched-br-data-spec
18575 @itemx -msched-br-data-spec
18576 @opindex mno-sched-br-data-spec
18577 @opindex msched-br-data-spec
18578 (Dis/En)able data speculative scheduling before reload.
18579 This results in generation of @code{ld.a} instructions and
18580 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18581 The default setting is disabled.
18583 @item -msched-ar-data-spec
18584 @itemx -mno-sched-ar-data-spec
18585 @opindex msched-ar-data-spec
18586 @opindex mno-sched-ar-data-spec
18587 (En/Dis)able data speculative scheduling after reload.
18588 This results in generation of @code{ld.a} instructions and
18589 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18590 The default setting is enabled.
18592 @item -mno-sched-control-spec
18593 @itemx -msched-control-spec
18594 @opindex mno-sched-control-spec
18595 @opindex msched-control-spec
18596 (Dis/En)able control speculative scheduling. This feature is
18597 available only during region scheduling (i.e.@: before reload).
18598 This results in generation of the @code{ld.s} instructions and
18599 the corresponding check instructions @code{chk.s}.
18600 The default setting is disabled.
18602 @item -msched-br-in-data-spec
18603 @itemx -mno-sched-br-in-data-spec
18604 @opindex msched-br-in-data-spec
18605 @opindex mno-sched-br-in-data-spec
18606 (En/Dis)able speculative scheduling of the instructions that
18607 are dependent on the data speculative loads before reload.
18608 This is effective only with @option{-msched-br-data-spec} enabled.
18609 The default setting is enabled.
18611 @item -msched-ar-in-data-spec
18612 @itemx -mno-sched-ar-in-data-spec
18613 @opindex msched-ar-in-data-spec
18614 @opindex mno-sched-ar-in-data-spec
18615 (En/Dis)able speculative scheduling of the instructions that
18616 are dependent on the data speculative loads after reload.
18617 This is effective only with @option{-msched-ar-data-spec} enabled.
18618 The default setting is enabled.
18620 @item -msched-in-control-spec
18621 @itemx -mno-sched-in-control-spec
18622 @opindex msched-in-control-spec
18623 @opindex mno-sched-in-control-spec
18624 (En/Dis)able speculative scheduling of the instructions that
18625 are dependent on the control speculative loads.
18626 This is effective only with @option{-msched-control-spec} enabled.
18627 The default setting is enabled.
18629 @item -mno-sched-prefer-non-data-spec-insns
18630 @itemx -msched-prefer-non-data-spec-insns
18631 @opindex mno-sched-prefer-non-data-spec-insns
18632 @opindex msched-prefer-non-data-spec-insns
18633 If enabled, data-speculative instructions are chosen for schedule
18634 only if there are no other choices at the moment. This makes
18635 the use of the data speculation much more conservative.
18636 The default setting is disabled.
18638 @item -mno-sched-prefer-non-control-spec-insns
18639 @itemx -msched-prefer-non-control-spec-insns
18640 @opindex mno-sched-prefer-non-control-spec-insns
18641 @opindex msched-prefer-non-control-spec-insns
18642 If enabled, control-speculative instructions are chosen for schedule
18643 only if there are no other choices at the moment. This makes
18644 the use of the control speculation much more conservative.
18645 The default setting is disabled.
18647 @item -mno-sched-count-spec-in-critical-path
18648 @itemx -msched-count-spec-in-critical-path
18649 @opindex mno-sched-count-spec-in-critical-path
18650 @opindex msched-count-spec-in-critical-path
18651 If enabled, speculative dependencies are considered during
18652 computation of the instructions priorities. This makes the use of the
18653 speculation a bit more conservative.
18654 The default setting is disabled.
18656 @item -msched-spec-ldc
18657 @opindex msched-spec-ldc
18658 Use a simple data speculation check. This option is on by default.
18660 @item -msched-control-spec-ldc
18661 @opindex msched-spec-ldc
18662 Use a simple check for control speculation. This option is on by default.
18664 @item -msched-stop-bits-after-every-cycle
18665 @opindex msched-stop-bits-after-every-cycle
18666 Place a stop bit after every cycle when scheduling. This option is on
18669 @item -msched-fp-mem-deps-zero-cost
18670 @opindex msched-fp-mem-deps-zero-cost
18671 Assume that floating-point stores and loads are not likely to cause a conflict
18672 when placed into the same instruction group. This option is disabled by
18675 @item -msel-sched-dont-check-control-spec
18676 @opindex msel-sched-dont-check-control-spec
18677 Generate checks for control speculation in selective scheduling.
18678 This flag is disabled by default.
18680 @item -msched-max-memory-insns=@var{max-insns}
18681 @opindex msched-max-memory-insns
18682 Limit on the number of memory insns per instruction group, giving lower
18683 priority to subsequent memory insns attempting to schedule in the same
18684 instruction group. Frequently useful to prevent cache bank conflicts.
18685 The default value is 1.
18687 @item -msched-max-memory-insns-hard-limit
18688 @opindex msched-max-memory-insns-hard-limit
18689 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
18690 disallowing more than that number in an instruction group.
18691 Otherwise, the limit is ``soft'', meaning that non-memory operations
18692 are preferred when the limit is reached, but memory operations may still
18698 @subsection LM32 Options
18699 @cindex LM32 options
18701 These @option{-m} options are defined for the LatticeMico32 architecture:
18704 @item -mbarrel-shift-enabled
18705 @opindex mbarrel-shift-enabled
18706 Enable barrel-shift instructions.
18708 @item -mdivide-enabled
18709 @opindex mdivide-enabled
18710 Enable divide and modulus instructions.
18712 @item -mmultiply-enabled
18713 @opindex multiply-enabled
18714 Enable multiply instructions.
18716 @item -msign-extend-enabled
18717 @opindex msign-extend-enabled
18718 Enable sign extend instructions.
18720 @item -muser-enabled
18721 @opindex muser-enabled
18722 Enable user-defined instructions.
18727 @subsection M32C Options
18728 @cindex M32C options
18731 @item -mcpu=@var{name}
18733 Select the CPU for which code is generated. @var{name} may be one of
18734 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
18735 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
18736 the M32C/80 series.
18740 Specifies that the program will be run on the simulator. This causes
18741 an alternate runtime library to be linked in which supports, for
18742 example, file I/O@. You must not use this option when generating
18743 programs that will run on real hardware; you must provide your own
18744 runtime library for whatever I/O functions are needed.
18746 @item -memregs=@var{number}
18748 Specifies the number of memory-based pseudo-registers GCC uses
18749 during code generation. These pseudo-registers are used like real
18750 registers, so there is a tradeoff between GCC's ability to fit the
18751 code into available registers, and the performance penalty of using
18752 memory instead of registers. Note that all modules in a program must
18753 be compiled with the same value for this option. Because of that, you
18754 must not use this option with GCC's default runtime libraries.
18758 @node M32R/D Options
18759 @subsection M32R/D Options
18760 @cindex M32R/D options
18762 These @option{-m} options are defined for Renesas M32R/D architectures:
18767 Generate code for the M32R/2@.
18771 Generate code for the M32R/X@.
18775 Generate code for the M32R@. This is the default.
18777 @item -mmodel=small
18778 @opindex mmodel=small
18779 Assume all objects live in the lower 16MB of memory (so that their addresses
18780 can be loaded with the @code{ld24} instruction), and assume all subroutines
18781 are reachable with the @code{bl} instruction.
18782 This is the default.
18784 The addressability of a particular object can be set with the
18785 @code{model} attribute.
18787 @item -mmodel=medium
18788 @opindex mmodel=medium
18789 Assume objects may be anywhere in the 32-bit address space (the compiler
18790 generates @code{seth/add3} instructions to load their addresses), and
18791 assume all subroutines are reachable with the @code{bl} instruction.
18793 @item -mmodel=large
18794 @opindex mmodel=large
18795 Assume objects may be anywhere in the 32-bit address space (the compiler
18796 generates @code{seth/add3} instructions to load their addresses), and
18797 assume subroutines may not be reachable with the @code{bl} instruction
18798 (the compiler generates the much slower @code{seth/add3/jl}
18799 instruction sequence).
18802 @opindex msdata=none
18803 Disable use of the small data area. Variables are put into
18804 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
18805 @code{section} attribute has been specified).
18806 This is the default.
18808 The small data area consists of sections @code{.sdata} and @code{.sbss}.
18809 Objects may be explicitly put in the small data area with the
18810 @code{section} attribute using one of these sections.
18812 @item -msdata=sdata
18813 @opindex msdata=sdata
18814 Put small global and static data in the small data area, but do not
18815 generate special code to reference them.
18818 @opindex msdata=use
18819 Put small global and static data in the small data area, and generate
18820 special instructions to reference them.
18824 @cindex smaller data references
18825 Put global and static objects less than or equal to @var{num} bytes
18826 into the small data or BSS sections instead of the normal data or BSS
18827 sections. The default value of @var{num} is 8.
18828 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
18829 for this option to have any effect.
18831 All modules should be compiled with the same @option{-G @var{num}} value.
18832 Compiling with different values of @var{num} may or may not work; if it
18833 doesn't the linker gives an error message---incorrect code is not
18838 Makes the M32R-specific code in the compiler display some statistics
18839 that might help in debugging programs.
18841 @item -malign-loops
18842 @opindex malign-loops
18843 Align all loops to a 32-byte boundary.
18845 @item -mno-align-loops
18846 @opindex mno-align-loops
18847 Do not enforce a 32-byte alignment for loops. This is the default.
18849 @item -missue-rate=@var{number}
18850 @opindex missue-rate=@var{number}
18851 Issue @var{number} instructions per cycle. @var{number} can only be 1
18854 @item -mbranch-cost=@var{number}
18855 @opindex mbranch-cost=@var{number}
18856 @var{number} can only be 1 or 2. If it is 1 then branches are
18857 preferred over conditional code, if it is 2, then the opposite applies.
18859 @item -mflush-trap=@var{number}
18860 @opindex mflush-trap=@var{number}
18861 Specifies the trap number to use to flush the cache. The default is
18862 12. Valid numbers are between 0 and 15 inclusive.
18864 @item -mno-flush-trap
18865 @opindex mno-flush-trap
18866 Specifies that the cache cannot be flushed by using a trap.
18868 @item -mflush-func=@var{name}
18869 @opindex mflush-func=@var{name}
18870 Specifies the name of the operating system function to call to flush
18871 the cache. The default is @samp{_flush_cache}, but a function call
18872 is only used if a trap is not available.
18874 @item -mno-flush-func
18875 @opindex mno-flush-func
18876 Indicates that there is no OS function for flushing the cache.
18880 @node M680x0 Options
18881 @subsection M680x0 Options
18882 @cindex M680x0 options
18884 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
18885 The default settings depend on which architecture was selected when
18886 the compiler was configured; the defaults for the most common choices
18890 @item -march=@var{arch}
18892 Generate code for a specific M680x0 or ColdFire instruction set
18893 architecture. Permissible values of @var{arch} for M680x0
18894 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
18895 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
18896 architectures are selected according to Freescale's ISA classification
18897 and the permissible values are: @samp{isaa}, @samp{isaaplus},
18898 @samp{isab} and @samp{isac}.
18900 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
18901 code for a ColdFire target. The @var{arch} in this macro is one of the
18902 @option{-march} arguments given above.
18904 When used together, @option{-march} and @option{-mtune} select code
18905 that runs on a family of similar processors but that is optimized
18906 for a particular microarchitecture.
18908 @item -mcpu=@var{cpu}
18910 Generate code for a specific M680x0 or ColdFire processor.
18911 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
18912 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
18913 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
18914 below, which also classifies the CPUs into families:
18916 @multitable @columnfractions 0.20 0.80
18917 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
18918 @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}
18919 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
18920 @item @samp{5206e} @tab @samp{5206e}
18921 @item @samp{5208} @tab @samp{5207} @samp{5208}
18922 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
18923 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
18924 @item @samp{5216} @tab @samp{5214} @samp{5216}
18925 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
18926 @item @samp{5225} @tab @samp{5224} @samp{5225}
18927 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
18928 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
18929 @item @samp{5249} @tab @samp{5249}
18930 @item @samp{5250} @tab @samp{5250}
18931 @item @samp{5271} @tab @samp{5270} @samp{5271}
18932 @item @samp{5272} @tab @samp{5272}
18933 @item @samp{5275} @tab @samp{5274} @samp{5275}
18934 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
18935 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
18936 @item @samp{5307} @tab @samp{5307}
18937 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
18938 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
18939 @item @samp{5407} @tab @samp{5407}
18940 @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}
18943 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
18944 @var{arch} is compatible with @var{cpu}. Other combinations of
18945 @option{-mcpu} and @option{-march} are rejected.
18947 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
18948 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
18949 where the value of @var{family} is given by the table above.
18951 @item -mtune=@var{tune}
18953 Tune the code for a particular microarchitecture within the
18954 constraints set by @option{-march} and @option{-mcpu}.
18955 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
18956 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
18957 and @samp{cpu32}. The ColdFire microarchitectures
18958 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
18960 You can also use @option{-mtune=68020-40} for code that needs
18961 to run relatively well on 68020, 68030 and 68040 targets.
18962 @option{-mtune=68020-60} is similar but includes 68060 targets
18963 as well. These two options select the same tuning decisions as
18964 @option{-m68020-40} and @option{-m68020-60} respectively.
18966 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
18967 when tuning for 680x0 architecture @var{arch}. It also defines
18968 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
18969 option is used. If GCC is tuning for a range of architectures,
18970 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
18971 it defines the macros for every architecture in the range.
18973 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
18974 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
18975 of the arguments given above.
18981 Generate output for a 68000. This is the default
18982 when the compiler is configured for 68000-based systems.
18983 It is equivalent to @option{-march=68000}.
18985 Use this option for microcontrollers with a 68000 or EC000 core,
18986 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
18990 Generate output for a 68010. This is the default
18991 when the compiler is configured for 68010-based systems.
18992 It is equivalent to @option{-march=68010}.
18998 Generate output for a 68020. This is the default
18999 when the compiler is configured for 68020-based systems.
19000 It is equivalent to @option{-march=68020}.
19004 Generate output for a 68030. This is the default when the compiler is
19005 configured for 68030-based systems. It is equivalent to
19006 @option{-march=68030}.
19010 Generate output for a 68040. This is the default when the compiler is
19011 configured for 68040-based systems. It is equivalent to
19012 @option{-march=68040}.
19014 This option inhibits the use of 68881/68882 instructions that have to be
19015 emulated by software on the 68040. Use this option if your 68040 does not
19016 have code to emulate those instructions.
19020 Generate output for a 68060. This is the default when the compiler is
19021 configured for 68060-based systems. It is equivalent to
19022 @option{-march=68060}.
19024 This option inhibits the use of 68020 and 68881/68882 instructions that
19025 have to be emulated by software on the 68060. Use this option if your 68060
19026 does not have code to emulate those instructions.
19030 Generate output for a CPU32. This is the default
19031 when the compiler is configured for CPU32-based systems.
19032 It is equivalent to @option{-march=cpu32}.
19034 Use this option for microcontrollers with a
19035 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
19036 68336, 68340, 68341, 68349 and 68360.
19040 Generate output for a 520X ColdFire CPU@. This is the default
19041 when the compiler is configured for 520X-based systems.
19042 It is equivalent to @option{-mcpu=5206}, and is now deprecated
19043 in favor of that option.
19045 Use this option for microcontroller with a 5200 core, including
19046 the MCF5202, MCF5203, MCF5204 and MCF5206.
19050 Generate output for a 5206e ColdFire CPU@. The option is now
19051 deprecated in favor of the equivalent @option{-mcpu=5206e}.
19055 Generate output for a member of the ColdFire 528X family.
19056 The option is now deprecated in favor of the equivalent
19057 @option{-mcpu=528x}.
19061 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
19062 in favor of the equivalent @option{-mcpu=5307}.
19066 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
19067 in favor of the equivalent @option{-mcpu=5407}.
19071 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
19072 This includes use of hardware floating-point instructions.
19073 The option is equivalent to @option{-mcpu=547x}, and is now
19074 deprecated in favor of that option.
19078 Generate output for a 68040, without using any of the new instructions.
19079 This results in code that can run relatively efficiently on either a
19080 68020/68881 or a 68030 or a 68040. The generated code does use the
19081 68881 instructions that are emulated on the 68040.
19083 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
19087 Generate output for a 68060, without using any of the new instructions.
19088 This results in code that can run relatively efficiently on either a
19089 68020/68881 or a 68030 or a 68040. The generated code does use the
19090 68881 instructions that are emulated on the 68060.
19092 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
19096 @opindex mhard-float
19098 Generate floating-point instructions. This is the default for 68020
19099 and above, and for ColdFire devices that have an FPU@. It defines the
19100 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
19101 on ColdFire targets.
19104 @opindex msoft-float
19105 Do not generate floating-point instructions; use library calls instead.
19106 This is the default for 68000, 68010, and 68832 targets. It is also
19107 the default for ColdFire devices that have no FPU.
19113 Generate (do not generate) ColdFire hardware divide and remainder
19114 instructions. If @option{-march} is used without @option{-mcpu},
19115 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
19116 architectures. Otherwise, the default is taken from the target CPU
19117 (either the default CPU, or the one specified by @option{-mcpu}). For
19118 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
19119 @option{-mcpu=5206e}.
19121 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
19125 Consider type @code{int} to be 16 bits wide, like @code{short int}.
19126 Additionally, parameters passed on the stack are also aligned to a
19127 16-bit boundary even on targets whose API mandates promotion to 32-bit.
19131 Do not consider type @code{int} to be 16 bits wide. This is the default.
19134 @itemx -mno-bitfield
19135 @opindex mnobitfield
19136 @opindex mno-bitfield
19137 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
19138 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
19142 Do use the bit-field instructions. The @option{-m68020} option implies
19143 @option{-mbitfield}. This is the default if you use a configuration
19144 designed for a 68020.
19148 Use a different function-calling convention, in which functions
19149 that take a fixed number of arguments return with the @code{rtd}
19150 instruction, which pops their arguments while returning. This
19151 saves one instruction in the caller since there is no need to pop
19152 the arguments there.
19154 This calling convention is incompatible with the one normally
19155 used on Unix, so you cannot use it if you need to call libraries
19156 compiled with the Unix compiler.
19158 Also, you must provide function prototypes for all functions that
19159 take variable numbers of arguments (including @code{printf});
19160 otherwise incorrect code is generated for calls to those
19163 In addition, seriously incorrect code results if you call a
19164 function with too many arguments. (Normally, extra arguments are
19165 harmlessly ignored.)
19167 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
19168 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
19172 Do not use the calling conventions selected by @option{-mrtd}.
19173 This is the default.
19176 @itemx -mno-align-int
19177 @opindex malign-int
19178 @opindex mno-align-int
19179 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
19180 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
19181 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
19182 Aligning variables on 32-bit boundaries produces code that runs somewhat
19183 faster on processors with 32-bit busses at the expense of more memory.
19185 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
19186 aligns structures containing the above types differently than
19187 most published application binary interface specifications for the m68k.
19191 Use the pc-relative addressing mode of the 68000 directly, instead of
19192 using a global offset table. At present, this option implies @option{-fpic},
19193 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
19194 not presently supported with @option{-mpcrel}, though this could be supported for
19195 68020 and higher processors.
19197 @item -mno-strict-align
19198 @itemx -mstrict-align
19199 @opindex mno-strict-align
19200 @opindex mstrict-align
19201 Do not (do) assume that unaligned memory references are handled by
19205 Generate code that allows the data segment to be located in a different
19206 area of memory from the text segment. This allows for execute-in-place in
19207 an environment without virtual memory management. This option implies
19210 @item -mno-sep-data
19211 Generate code that assumes that the data segment follows the text segment.
19212 This is the default.
19214 @item -mid-shared-library
19215 Generate code that supports shared libraries via the library ID method.
19216 This allows for execute-in-place and shared libraries in an environment
19217 without virtual memory management. This option implies @option{-fPIC}.
19219 @item -mno-id-shared-library
19220 Generate code that doesn't assume ID-based shared libraries are being used.
19221 This is the default.
19223 @item -mshared-library-id=n
19224 Specifies the identification number of the ID-based shared library being
19225 compiled. Specifying a value of 0 generates more compact code; specifying
19226 other values forces the allocation of that number to the current
19227 library, but is no more space- or time-efficient than omitting this option.
19233 When generating position-independent code for ColdFire, generate code
19234 that works if the GOT has more than 8192 entries. This code is
19235 larger and slower than code generated without this option. On M680x0
19236 processors, this option is not needed; @option{-fPIC} suffices.
19238 GCC normally uses a single instruction to load values from the GOT@.
19239 While this is relatively efficient, it only works if the GOT
19240 is smaller than about 64k. Anything larger causes the linker
19241 to report an error such as:
19243 @cindex relocation truncated to fit (ColdFire)
19245 relocation truncated to fit: R_68K_GOT16O foobar
19248 If this happens, you should recompile your code with @option{-mxgot}.
19249 It should then work with very large GOTs. However, code generated with
19250 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
19251 the value of a global symbol.
19253 Note that some linkers, including newer versions of the GNU linker,
19254 can create multiple GOTs and sort GOT entries. If you have such a linker,
19255 you should only need to use @option{-mxgot} when compiling a single
19256 object file that accesses more than 8192 GOT entries. Very few do.
19258 These options have no effect unless GCC is generating
19259 position-independent code.
19261 @item -mlong-jump-table-offsets
19262 @opindex mlong-jump-table-offsets
19263 Use 32-bit offsets in @code{switch} tables. The default is to use
19268 @node MCore Options
19269 @subsection MCore Options
19270 @cindex MCore options
19272 These are the @samp{-m} options defined for the Motorola M*Core
19278 @itemx -mno-hardlit
19280 @opindex mno-hardlit
19281 Inline constants into the code stream if it can be done in two
19282 instructions or less.
19288 Use the divide instruction. (Enabled by default).
19290 @item -mrelax-immediate
19291 @itemx -mno-relax-immediate
19292 @opindex mrelax-immediate
19293 @opindex mno-relax-immediate
19294 Allow arbitrary-sized immediates in bit operations.
19296 @item -mwide-bitfields
19297 @itemx -mno-wide-bitfields
19298 @opindex mwide-bitfields
19299 @opindex mno-wide-bitfields
19300 Always treat bit-fields as @code{int}-sized.
19302 @item -m4byte-functions
19303 @itemx -mno-4byte-functions
19304 @opindex m4byte-functions
19305 @opindex mno-4byte-functions
19306 Force all functions to be aligned to a 4-byte boundary.
19308 @item -mcallgraph-data
19309 @itemx -mno-callgraph-data
19310 @opindex mcallgraph-data
19311 @opindex mno-callgraph-data
19312 Emit callgraph information.
19315 @itemx -mno-slow-bytes
19316 @opindex mslow-bytes
19317 @opindex mno-slow-bytes
19318 Prefer word access when reading byte quantities.
19320 @item -mlittle-endian
19321 @itemx -mbig-endian
19322 @opindex mlittle-endian
19323 @opindex mbig-endian
19324 Generate code for a little-endian target.
19330 Generate code for the 210 processor.
19334 Assume that runtime support has been provided and so omit the
19335 simulator library (@file{libsim.a)} from the linker command line.
19337 @item -mstack-increment=@var{size}
19338 @opindex mstack-increment
19339 Set the maximum amount for a single stack increment operation. Large
19340 values can increase the speed of programs that contain functions
19341 that need a large amount of stack space, but they can also trigger a
19342 segmentation fault if the stack is extended too much. The default
19348 @subsection MeP Options
19349 @cindex MeP options
19355 Enables the @code{abs} instruction, which is the absolute difference
19356 between two registers.
19360 Enables all the optional instructions---average, multiply, divide, bit
19361 operations, leading zero, absolute difference, min/max, clip, and
19367 Enables the @code{ave} instruction, which computes the average of two
19370 @item -mbased=@var{n}
19372 Variables of size @var{n} bytes or smaller are placed in the
19373 @code{.based} section by default. Based variables use the @code{$tp}
19374 register as a base register, and there is a 128-byte limit to the
19375 @code{.based} section.
19379 Enables the bit operation instructions---bit test (@code{btstm}), set
19380 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
19381 test-and-set (@code{tas}).
19383 @item -mc=@var{name}
19385 Selects which section constant data is placed in. @var{name} may
19386 be @samp{tiny}, @samp{near}, or @samp{far}.
19390 Enables the @code{clip} instruction. Note that @option{-mclip} is not
19391 useful unless you also provide @option{-mminmax}.
19393 @item -mconfig=@var{name}
19395 Selects one of the built-in core configurations. Each MeP chip has
19396 one or more modules in it; each module has a core CPU and a variety of
19397 coprocessors, optional instructions, and peripherals. The
19398 @code{MeP-Integrator} tool, not part of GCC, provides these
19399 configurations through this option; using this option is the same as
19400 using all the corresponding command-line options. The default
19401 configuration is @samp{default}.
19405 Enables the coprocessor instructions. By default, this is a 32-bit
19406 coprocessor. Note that the coprocessor is normally enabled via the
19407 @option{-mconfig=} option.
19411 Enables the 32-bit coprocessor's instructions.
19415 Enables the 64-bit coprocessor's instructions.
19419 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
19423 Causes constant variables to be placed in the @code{.near} section.
19427 Enables the @code{div} and @code{divu} instructions.
19431 Generate big-endian code.
19435 Generate little-endian code.
19437 @item -mio-volatile
19438 @opindex mio-volatile
19439 Tells the compiler that any variable marked with the @code{io}
19440 attribute is to be considered volatile.
19444 Causes variables to be assigned to the @code{.far} section by default.
19448 Enables the @code{leadz} (leading zero) instruction.
19452 Causes variables to be assigned to the @code{.near} section by default.
19456 Enables the @code{min} and @code{max} instructions.
19460 Enables the multiplication and multiply-accumulate instructions.
19464 Disables all the optional instructions enabled by @option{-mall-opts}.
19468 Enables the @code{repeat} and @code{erepeat} instructions, used for
19469 low-overhead looping.
19473 Causes all variables to default to the @code{.tiny} section. Note
19474 that there is a 65536-byte limit to this section. Accesses to these
19475 variables use the @code{%gp} base register.
19479 Enables the saturation instructions. Note that the compiler does not
19480 currently generate these itself, but this option is included for
19481 compatibility with other tools, like @code{as}.
19485 Link the SDRAM-based runtime instead of the default ROM-based runtime.
19489 Link the simulator run-time libraries.
19493 Link the simulator runtime libraries, excluding built-in support
19494 for reset and exception vectors and tables.
19498 Causes all functions to default to the @code{.far} section. Without
19499 this option, functions default to the @code{.near} section.
19501 @item -mtiny=@var{n}
19503 Variables that are @var{n} bytes or smaller are allocated to the
19504 @code{.tiny} section. These variables use the @code{$gp} base
19505 register. The default for this option is 4, but note that there's a
19506 65536-byte limit to the @code{.tiny} section.
19510 @node MicroBlaze Options
19511 @subsection MicroBlaze Options
19512 @cindex MicroBlaze Options
19517 @opindex msoft-float
19518 Use software emulation for floating point (default).
19521 @opindex mhard-float
19522 Use hardware floating-point instructions.
19526 Do not optimize block moves, use @code{memcpy}.
19528 @item -mno-clearbss
19529 @opindex mno-clearbss
19530 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
19532 @item -mcpu=@var{cpu-type}
19534 Use features of, and schedule code for, the given CPU.
19535 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
19536 where @var{X} is a major version, @var{YY} is the minor version, and
19537 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
19538 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
19540 @item -mxl-soft-mul
19541 @opindex mxl-soft-mul
19542 Use software multiply emulation (default).
19544 @item -mxl-soft-div
19545 @opindex mxl-soft-div
19546 Use software emulation for divides (default).
19548 @item -mxl-barrel-shift
19549 @opindex mxl-barrel-shift
19550 Use the hardware barrel shifter.
19552 @item -mxl-pattern-compare
19553 @opindex mxl-pattern-compare
19554 Use pattern compare instructions.
19556 @item -msmall-divides
19557 @opindex msmall-divides
19558 Use table lookup optimization for small signed integer divisions.
19560 @item -mxl-stack-check
19561 @opindex mxl-stack-check
19562 This option is deprecated. Use @option{-fstack-check} instead.
19565 @opindex mxl-gp-opt
19566 Use GP-relative @code{.sdata}/@code{.sbss} sections.
19568 @item -mxl-multiply-high
19569 @opindex mxl-multiply-high
19570 Use multiply high instructions for high part of 32x32 multiply.
19572 @item -mxl-float-convert
19573 @opindex mxl-float-convert
19574 Use hardware floating-point conversion instructions.
19576 @item -mxl-float-sqrt
19577 @opindex mxl-float-sqrt
19578 Use hardware floating-point square root instruction.
19581 @opindex mbig-endian
19582 Generate code for a big-endian target.
19584 @item -mlittle-endian
19585 @opindex mlittle-endian
19586 Generate code for a little-endian target.
19589 @opindex mxl-reorder
19590 Use reorder instructions (swap and byte reversed load/store).
19592 @item -mxl-mode-@var{app-model}
19593 Select application model @var{app-model}. Valid models are
19596 normal executable (default), uses startup code @file{crt0.o}.
19599 for use with Xilinx Microprocessor Debugger (XMD) based
19600 software intrusive debug agent called xmdstub. This uses startup file
19601 @file{crt1.o} and sets the start address of the program to 0x800.
19604 for applications that are loaded using a bootloader.
19605 This model uses startup file @file{crt2.o} which does not contain a processor
19606 reset vector handler. This is suitable for transferring control on a
19607 processor reset to the bootloader rather than the application.
19610 for applications that do not require any of the
19611 MicroBlaze vectors. This option may be useful for applications running
19612 within a monitoring application. This model uses @file{crt3.o} as a startup file.
19615 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
19616 @option{-mxl-mode-@var{app-model}}.
19621 @subsection MIPS Options
19622 @cindex MIPS options
19628 Generate big-endian code.
19632 Generate little-endian code. This is the default for @samp{mips*el-*-*}
19635 @item -march=@var{arch}
19637 Generate code that runs on @var{arch}, which can be the name of a
19638 generic MIPS ISA, or the name of a particular processor.
19640 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
19641 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
19642 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
19643 @samp{mips64r5} and @samp{mips64r6}.
19644 The processor names are:
19645 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
19646 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
19647 @samp{5kc}, @samp{5kf},
19649 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
19650 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
19651 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
19652 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
19653 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
19656 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
19658 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
19659 @samp{m5100}, @samp{m5101},
19660 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
19663 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
19664 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
19665 @samp{rm7000}, @samp{rm9000},
19666 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
19669 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
19670 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
19671 @samp{xlr} and @samp{xlp}.
19672 The special value @samp{from-abi} selects the
19673 most compatible architecture for the selected ABI (that is,
19674 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
19676 The native Linux/GNU toolchain also supports the value @samp{native},
19677 which selects the best architecture option for the host processor.
19678 @option{-march=native} has no effect if GCC does not recognize
19681 In processor names, a final @samp{000} can be abbreviated as @samp{k}
19682 (for example, @option{-march=r2k}). Prefixes are optional, and
19683 @samp{vr} may be written @samp{r}.
19685 Names of the form @samp{@var{n}f2_1} refer to processors with
19686 FPUs clocked at half the rate of the core, names of the form
19687 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
19688 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
19689 processors with FPUs clocked a ratio of 3:2 with respect to the core.
19690 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
19691 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
19692 accepted as synonyms for @samp{@var{n}f1_1}.
19694 GCC defines two macros based on the value of this option. The first
19695 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
19696 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
19697 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
19698 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
19699 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
19701 Note that the @code{_MIPS_ARCH} macro uses the processor names given
19702 above. In other words, it has the full prefix and does not
19703 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
19704 the macro names the resolved architecture (either @code{"mips1"} or
19705 @code{"mips3"}). It names the default architecture when no
19706 @option{-march} option is given.
19708 @item -mtune=@var{arch}
19710 Optimize for @var{arch}. Among other things, this option controls
19711 the way instructions are scheduled, and the perceived cost of arithmetic
19712 operations. The list of @var{arch} values is the same as for
19715 When this option is not used, GCC optimizes for the processor
19716 specified by @option{-march}. By using @option{-march} and
19717 @option{-mtune} together, it is possible to generate code that
19718 runs on a family of processors, but optimize the code for one
19719 particular member of that family.
19721 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
19722 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
19723 @option{-march} ones described above.
19727 Equivalent to @option{-march=mips1}.
19731 Equivalent to @option{-march=mips2}.
19735 Equivalent to @option{-march=mips3}.
19739 Equivalent to @option{-march=mips4}.
19743 Equivalent to @option{-march=mips32}.
19747 Equivalent to @option{-march=mips32r3}.
19751 Equivalent to @option{-march=mips32r5}.
19755 Equivalent to @option{-march=mips32r6}.
19759 Equivalent to @option{-march=mips64}.
19763 Equivalent to @option{-march=mips64r2}.
19767 Equivalent to @option{-march=mips64r3}.
19771 Equivalent to @option{-march=mips64r5}.
19775 Equivalent to @option{-march=mips64r6}.
19780 @opindex mno-mips16
19781 Generate (do not generate) MIPS16 code. If GCC is targeting a
19782 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
19784 MIPS16 code generation can also be controlled on a per-function basis
19785 by means of @code{mips16} and @code{nomips16} attributes.
19786 @xref{Function Attributes}, for more information.
19788 @item -mflip-mips16
19789 @opindex mflip-mips16
19790 Generate MIPS16 code on alternating functions. This option is provided
19791 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
19792 not intended for ordinary use in compiling user code.
19794 @item -minterlink-compressed
19795 @item -mno-interlink-compressed
19796 @opindex minterlink-compressed
19797 @opindex mno-interlink-compressed
19798 Require (do not require) that code using the standard (uncompressed) MIPS ISA
19799 be link-compatible with MIPS16 and microMIPS code, and vice versa.
19801 For example, code using the standard ISA encoding cannot jump directly
19802 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
19803 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
19804 knows that the target of the jump is not compressed.
19806 @item -minterlink-mips16
19807 @itemx -mno-interlink-mips16
19808 @opindex minterlink-mips16
19809 @opindex mno-interlink-mips16
19810 Aliases of @option{-minterlink-compressed} and
19811 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
19812 and are retained for backwards compatibility.
19824 Generate code for the given ABI@.
19826 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
19827 generates 64-bit code when you select a 64-bit architecture, but you
19828 can use @option{-mgp32} to get 32-bit code instead.
19830 For information about the O64 ABI, see
19831 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
19833 GCC supports a variant of the o32 ABI in which floating-point registers
19834 are 64 rather than 32 bits wide. You can select this combination with
19835 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
19836 and @code{mfhc1} instructions and is therefore only supported for
19837 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
19839 The register assignments for arguments and return values remain the
19840 same, but each scalar value is passed in a single 64-bit register
19841 rather than a pair of 32-bit registers. For example, scalar
19842 floating-point values are returned in @samp{$f0} only, not a
19843 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
19844 remains the same in that the even-numbered double-precision registers
19847 Two additional variants of the o32 ABI are supported to enable
19848 a transition from 32-bit to 64-bit registers. These are FPXX
19849 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
19850 The FPXX extension mandates that all code must execute correctly
19851 when run using 32-bit or 64-bit registers. The code can be interlinked
19852 with either FP32 or FP64, but not both.
19853 The FP64A extension is similar to the FP64 extension but forbids the
19854 use of odd-numbered single-precision registers. This can be used
19855 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
19856 processors and allows both FP32 and FP64A code to interlink and
19857 run in the same process without changing FPU modes.
19860 @itemx -mno-abicalls
19862 @opindex mno-abicalls
19863 Generate (do not generate) code that is suitable for SVR4-style
19864 dynamic objects. @option{-mabicalls} is the default for SVR4-based
19869 Generate (do not generate) code that is fully position-independent,
19870 and that can therefore be linked into shared libraries. This option
19871 only affects @option{-mabicalls}.
19873 All @option{-mabicalls} code has traditionally been position-independent,
19874 regardless of options like @option{-fPIC} and @option{-fpic}. However,
19875 as an extension, the GNU toolchain allows executables to use absolute
19876 accesses for locally-binding symbols. It can also use shorter GP
19877 initialization sequences and generate direct calls to locally-defined
19878 functions. This mode is selected by @option{-mno-shared}.
19880 @option{-mno-shared} depends on binutils 2.16 or higher and generates
19881 objects that can only be linked by the GNU linker. However, the option
19882 does not affect the ABI of the final executable; it only affects the ABI
19883 of relocatable objects. Using @option{-mno-shared} generally makes
19884 executables both smaller and quicker.
19886 @option{-mshared} is the default.
19892 Assume (do not assume) that the static and dynamic linkers
19893 support PLTs and copy relocations. This option only affects
19894 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
19895 has no effect without @option{-msym32}.
19897 You can make @option{-mplt} the default by configuring
19898 GCC with @option{--with-mips-plt}. The default is
19899 @option{-mno-plt} otherwise.
19905 Lift (do not lift) the usual restrictions on the size of the global
19908 GCC normally uses a single instruction to load values from the GOT@.
19909 While this is relatively efficient, it only works if the GOT
19910 is smaller than about 64k. Anything larger causes the linker
19911 to report an error such as:
19913 @cindex relocation truncated to fit (MIPS)
19915 relocation truncated to fit: R_MIPS_GOT16 foobar
19918 If this happens, you should recompile your code with @option{-mxgot}.
19919 This works with very large GOTs, although the code is also
19920 less efficient, since it takes three instructions to fetch the
19921 value of a global symbol.
19923 Note that some linkers can create multiple GOTs. If you have such a
19924 linker, you should only need to use @option{-mxgot} when a single object
19925 file accesses more than 64k's worth of GOT entries. Very few do.
19927 These options have no effect unless GCC is generating position
19932 Assume that general-purpose registers are 32 bits wide.
19936 Assume that general-purpose registers are 64 bits wide.
19940 Assume that floating-point registers are 32 bits wide.
19944 Assume that floating-point registers are 64 bits wide.
19948 Do not assume the width of floating-point registers.
19951 @opindex mhard-float
19952 Use floating-point coprocessor instructions.
19955 @opindex msoft-float
19956 Do not use floating-point coprocessor instructions. Implement
19957 floating-point calculations using library calls instead.
19961 Equivalent to @option{-msoft-float}, but additionally asserts that the
19962 program being compiled does not perform any floating-point operations.
19963 This option is presently supported only by some bare-metal MIPS
19964 configurations, where it may select a special set of libraries
19965 that lack all floating-point support (including, for example, the
19966 floating-point @code{printf} formats).
19967 If code compiled with @option{-mno-float} accidentally contains
19968 floating-point operations, it is likely to suffer a link-time
19969 or run-time failure.
19971 @item -msingle-float
19972 @opindex msingle-float
19973 Assume that the floating-point coprocessor only supports single-precision
19976 @item -mdouble-float
19977 @opindex mdouble-float
19978 Assume that the floating-point coprocessor supports double-precision
19979 operations. This is the default.
19982 @itemx -mno-odd-spreg
19983 @opindex modd-spreg
19984 @opindex mno-odd-spreg
19985 Enable the use of odd-numbered single-precision floating-point registers
19986 for the o32 ABI. This is the default for processors that are known to
19987 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
19991 @itemx -mabs=legacy
19993 @opindex mabs=legacy
19994 These options control the treatment of the special not-a-number (NaN)
19995 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
19996 @code{neg.@i{fmt}} machine instructions.
19998 By default or when @option{-mabs=legacy} is used the legacy
19999 treatment is selected. In this case these instructions are considered
20000 arithmetic and avoided where correct operation is required and the
20001 input operand might be a NaN. A longer sequence of instructions that
20002 manipulate the sign bit of floating-point datum manually is used
20003 instead unless the @option{-ffinite-math-only} option has also been
20006 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
20007 this case these instructions are considered non-arithmetic and therefore
20008 operating correctly in all cases, including in particular where the
20009 input operand is a NaN. These instructions are therefore always used
20010 for the respective operations.
20013 @itemx -mnan=legacy
20015 @opindex mnan=legacy
20016 These options control the encoding of the special not-a-number (NaN)
20017 IEEE 754 floating-point data.
20019 The @option{-mnan=legacy} option selects the legacy encoding. In this
20020 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
20021 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
20022 by the first bit of their trailing significand field being 1.
20024 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
20025 this case qNaNs are denoted by the first bit of their trailing
20026 significand field being 1, whereas sNaNs are denoted by the first bit of
20027 their trailing significand field being 0.
20029 The default is @option{-mnan=legacy} unless GCC has been configured with
20030 @option{--with-nan=2008}.
20036 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
20037 implement atomic memory built-in functions. When neither option is
20038 specified, GCC uses the instructions if the target architecture
20041 @option{-mllsc} is useful if the runtime environment can emulate the
20042 instructions and @option{-mno-llsc} can be useful when compiling for
20043 nonstandard ISAs. You can make either option the default by
20044 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
20045 respectively. @option{--with-llsc} is the default for some
20046 configurations; see the installation documentation for details.
20052 Use (do not use) revision 1 of the MIPS DSP ASE@.
20053 @xref{MIPS DSP Built-in Functions}. This option defines the
20054 preprocessor macro @code{__mips_dsp}. It also defines
20055 @code{__mips_dsp_rev} to 1.
20061 Use (do not use) revision 2 of the MIPS DSP ASE@.
20062 @xref{MIPS DSP Built-in Functions}. This option defines the
20063 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
20064 It also defines @code{__mips_dsp_rev} to 2.
20067 @itemx -mno-smartmips
20068 @opindex msmartmips
20069 @opindex mno-smartmips
20070 Use (do not use) the MIPS SmartMIPS ASE.
20072 @item -mpaired-single
20073 @itemx -mno-paired-single
20074 @opindex mpaired-single
20075 @opindex mno-paired-single
20076 Use (do not use) paired-single floating-point instructions.
20077 @xref{MIPS Paired-Single Support}. This option requires
20078 hardware floating-point support to be enabled.
20084 Use (do not use) MIPS Digital Media Extension instructions.
20085 This option can only be used when generating 64-bit code and requires
20086 hardware floating-point support to be enabled.
20091 @opindex mno-mips3d
20092 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
20093 The option @option{-mips3d} implies @option{-mpaired-single}.
20096 @itemx -mno-micromips
20097 @opindex mmicromips
20098 @opindex mno-mmicromips
20099 Generate (do not generate) microMIPS code.
20101 MicroMIPS code generation can also be controlled on a per-function basis
20102 by means of @code{micromips} and @code{nomicromips} attributes.
20103 @xref{Function Attributes}, for more information.
20109 Use (do not use) MT Multithreading instructions.
20115 Use (do not use) the MIPS MCU ASE instructions.
20121 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
20127 Use (do not use) the MIPS Virtualization (VZ) instructions.
20133 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
20137 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
20138 an explanation of the default and the way that the pointer size is
20143 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
20145 The default size of @code{int}s, @code{long}s and pointers depends on
20146 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
20147 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
20148 32-bit @code{long}s. Pointers are the same size as @code{long}s,
20149 or the same size as integer registers, whichever is smaller.
20155 Assume (do not assume) that all symbols have 32-bit values, regardless
20156 of the selected ABI@. This option is useful in combination with
20157 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
20158 to generate shorter and faster references to symbolic addresses.
20162 Put definitions of externally-visible data in a small data section
20163 if that data is no bigger than @var{num} bytes. GCC can then generate
20164 more efficient accesses to the data; see @option{-mgpopt} for details.
20166 The default @option{-G} option depends on the configuration.
20168 @item -mlocal-sdata
20169 @itemx -mno-local-sdata
20170 @opindex mlocal-sdata
20171 @opindex mno-local-sdata
20172 Extend (do not extend) the @option{-G} behavior to local data too,
20173 such as to static variables in C@. @option{-mlocal-sdata} is the
20174 default for all configurations.
20176 If the linker complains that an application is using too much small data,
20177 you might want to try rebuilding the less performance-critical parts with
20178 @option{-mno-local-sdata}. You might also want to build large
20179 libraries with @option{-mno-local-sdata}, so that the libraries leave
20180 more room for the main program.
20182 @item -mextern-sdata
20183 @itemx -mno-extern-sdata
20184 @opindex mextern-sdata
20185 @opindex mno-extern-sdata
20186 Assume (do not assume) that externally-defined data is in
20187 a small data section if the size of that data is within the @option{-G} limit.
20188 @option{-mextern-sdata} is the default for all configurations.
20190 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
20191 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
20192 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
20193 is placed in a small data section. If @var{Var} is defined by another
20194 module, you must either compile that module with a high-enough
20195 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
20196 definition. If @var{Var} is common, you must link the application
20197 with a high-enough @option{-G} setting.
20199 The easiest way of satisfying these restrictions is to compile
20200 and link every module with the same @option{-G} option. However,
20201 you may wish to build a library that supports several different
20202 small data limits. You can do this by compiling the library with
20203 the highest supported @option{-G} setting and additionally using
20204 @option{-mno-extern-sdata} to stop the library from making assumptions
20205 about externally-defined data.
20211 Use (do not use) GP-relative accesses for symbols that are known to be
20212 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
20213 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
20216 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
20217 might not hold the value of @code{_gp}. For example, if the code is
20218 part of a library that might be used in a boot monitor, programs that
20219 call boot monitor routines pass an unknown value in @code{$gp}.
20220 (In such situations, the boot monitor itself is usually compiled
20221 with @option{-G0}.)
20223 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
20224 @option{-mno-extern-sdata}.
20226 @item -membedded-data
20227 @itemx -mno-embedded-data
20228 @opindex membedded-data
20229 @opindex mno-embedded-data
20230 Allocate variables to the read-only data section first if possible, then
20231 next in the small data section if possible, otherwise in data. This gives
20232 slightly slower code than the default, but reduces the amount of RAM required
20233 when executing, and thus may be preferred for some embedded systems.
20235 @item -muninit-const-in-rodata
20236 @itemx -mno-uninit-const-in-rodata
20237 @opindex muninit-const-in-rodata
20238 @opindex mno-uninit-const-in-rodata
20239 Put uninitialized @code{const} variables in the read-only data section.
20240 This option is only meaningful in conjunction with @option{-membedded-data}.
20242 @item -mcode-readable=@var{setting}
20243 @opindex mcode-readable
20244 Specify whether GCC may generate code that reads from executable sections.
20245 There are three possible settings:
20248 @item -mcode-readable=yes
20249 Instructions may freely access executable sections. This is the
20252 @item -mcode-readable=pcrel
20253 MIPS16 PC-relative load instructions can access executable sections,
20254 but other instructions must not do so. This option is useful on 4KSc
20255 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
20256 It is also useful on processors that can be configured to have a dual
20257 instruction/data SRAM interface and that, like the M4K, automatically
20258 redirect PC-relative loads to the instruction RAM.
20260 @item -mcode-readable=no
20261 Instructions must not access executable sections. This option can be
20262 useful on targets that are configured to have a dual instruction/data
20263 SRAM interface but that (unlike the M4K) do not automatically redirect
20264 PC-relative loads to the instruction RAM.
20267 @item -msplit-addresses
20268 @itemx -mno-split-addresses
20269 @opindex msplit-addresses
20270 @opindex mno-split-addresses
20271 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
20272 relocation operators. This option has been superseded by
20273 @option{-mexplicit-relocs} but is retained for backwards compatibility.
20275 @item -mexplicit-relocs
20276 @itemx -mno-explicit-relocs
20277 @opindex mexplicit-relocs
20278 @opindex mno-explicit-relocs
20279 Use (do not use) assembler relocation operators when dealing with symbolic
20280 addresses. The alternative, selected by @option{-mno-explicit-relocs},
20281 is to use assembler macros instead.
20283 @option{-mexplicit-relocs} is the default if GCC was configured
20284 to use an assembler that supports relocation operators.
20286 @item -mcheck-zero-division
20287 @itemx -mno-check-zero-division
20288 @opindex mcheck-zero-division
20289 @opindex mno-check-zero-division
20290 Trap (do not trap) on integer division by zero.
20292 The default is @option{-mcheck-zero-division}.
20294 @item -mdivide-traps
20295 @itemx -mdivide-breaks
20296 @opindex mdivide-traps
20297 @opindex mdivide-breaks
20298 MIPS systems check for division by zero by generating either a
20299 conditional trap or a break instruction. Using traps results in
20300 smaller code, but is only supported on MIPS II and later. Also, some
20301 versions of the Linux kernel have a bug that prevents trap from
20302 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
20303 allow conditional traps on architectures that support them and
20304 @option{-mdivide-breaks} to force the use of breaks.
20306 The default is usually @option{-mdivide-traps}, but this can be
20307 overridden at configure time using @option{--with-divide=breaks}.
20308 Divide-by-zero checks can be completely disabled using
20309 @option{-mno-check-zero-division}.
20311 @item -mload-store-pairs
20312 @itemx -mno-load-store-pairs
20313 @opindex mload-store-pairs
20314 @opindex mno-load-store-pairs
20315 Enable (disable) an optimization that pairs consecutive load or store
20316 instructions to enable load/store bonding. This option is enabled by
20317 default but only takes effect when the selected architecture is known
20318 to support bonding.
20323 @opindex mno-memcpy
20324 Force (do not force) the use of @code{memcpy} for non-trivial block
20325 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
20326 most constant-sized copies.
20329 @itemx -mno-long-calls
20330 @opindex mlong-calls
20331 @opindex mno-long-calls
20332 Disable (do not disable) use of the @code{jal} instruction. Calling
20333 functions using @code{jal} is more efficient but requires the caller
20334 and callee to be in the same 256 megabyte segment.
20336 This option has no effect on abicalls code. The default is
20337 @option{-mno-long-calls}.
20343 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
20344 instructions, as provided by the R4650 ISA@.
20350 Enable (disable) use of the @code{madd} and @code{msub} integer
20351 instructions. The default is @option{-mimadd} on architectures
20352 that support @code{madd} and @code{msub} except for the 74k
20353 architecture where it was found to generate slower code.
20356 @itemx -mno-fused-madd
20357 @opindex mfused-madd
20358 @opindex mno-fused-madd
20359 Enable (disable) use of the floating-point multiply-accumulate
20360 instructions, when they are available. The default is
20361 @option{-mfused-madd}.
20363 On the R8000 CPU when multiply-accumulate instructions are used,
20364 the intermediate product is calculated to infinite precision
20365 and is not subject to the FCSR Flush to Zero bit. This may be
20366 undesirable in some circumstances. On other processors the result
20367 is numerically identical to the equivalent computation using
20368 separate multiply, add, subtract and negate instructions.
20372 Tell the MIPS assembler to not run its preprocessor over user
20373 assembler files (with a @samp{.s} suffix) when assembling them.
20378 @opindex mno-fix-24k
20379 Work around the 24K E48 (lost data on stores during refill) errata.
20380 The workarounds are implemented by the assembler rather than by GCC@.
20383 @itemx -mno-fix-r4000
20384 @opindex mfix-r4000
20385 @opindex mno-fix-r4000
20386 Work around certain R4000 CPU errata:
20389 A double-word or a variable shift may give an incorrect result if executed
20390 immediately after starting an integer division.
20392 A double-word or a variable shift may give an incorrect result if executed
20393 while an integer multiplication is in progress.
20395 An integer division may give an incorrect result if started in a delay slot
20396 of a taken branch or a jump.
20400 @itemx -mno-fix-r4400
20401 @opindex mfix-r4400
20402 @opindex mno-fix-r4400
20403 Work around certain R4400 CPU errata:
20406 A double-word or a variable shift may give an incorrect result if executed
20407 immediately after starting an integer division.
20411 @itemx -mno-fix-r10000
20412 @opindex mfix-r10000
20413 @opindex mno-fix-r10000
20414 Work around certain R10000 errata:
20417 @code{ll}/@code{sc} sequences may not behave atomically on revisions
20418 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
20421 This option can only be used if the target architecture supports
20422 branch-likely instructions. @option{-mfix-r10000} is the default when
20423 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
20427 @itemx -mno-fix-rm7000
20428 @opindex mfix-rm7000
20429 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
20430 workarounds are implemented by the assembler rather than by GCC@.
20433 @itemx -mno-fix-vr4120
20434 @opindex mfix-vr4120
20435 Work around certain VR4120 errata:
20438 @code{dmultu} does not always produce the correct result.
20440 @code{div} and @code{ddiv} do not always produce the correct result if one
20441 of the operands is negative.
20443 The workarounds for the division errata rely on special functions in
20444 @file{libgcc.a}. At present, these functions are only provided by
20445 the @code{mips64vr*-elf} configurations.
20447 Other VR4120 errata require a NOP to be inserted between certain pairs of
20448 instructions. These errata are handled by the assembler, not by GCC itself.
20451 @opindex mfix-vr4130
20452 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
20453 workarounds are implemented by the assembler rather than by GCC,
20454 although GCC avoids using @code{mflo} and @code{mfhi} if the
20455 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
20456 instructions are available instead.
20459 @itemx -mno-fix-sb1
20461 Work around certain SB-1 CPU core errata.
20462 (This flag currently works around the SB-1 revision 2
20463 ``F1'' and ``F2'' floating-point errata.)
20465 @item -mr10k-cache-barrier=@var{setting}
20466 @opindex mr10k-cache-barrier
20467 Specify whether GCC should insert cache barriers to avoid the
20468 side-effects of speculation on R10K processors.
20470 In common with many processors, the R10K tries to predict the outcome
20471 of a conditional branch and speculatively executes instructions from
20472 the ``taken'' branch. It later aborts these instructions if the
20473 predicted outcome is wrong. However, on the R10K, even aborted
20474 instructions can have side effects.
20476 This problem only affects kernel stores and, depending on the system,
20477 kernel loads. As an example, a speculatively-executed store may load
20478 the target memory into cache and mark the cache line as dirty, even if
20479 the store itself is later aborted. If a DMA operation writes to the
20480 same area of memory before the ``dirty'' line is flushed, the cached
20481 data overwrites the DMA-ed data. See the R10K processor manual
20482 for a full description, including other potential problems.
20484 One workaround is to insert cache barrier instructions before every memory
20485 access that might be speculatively executed and that might have side
20486 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
20487 controls GCC's implementation of this workaround. It assumes that
20488 aborted accesses to any byte in the following regions does not have
20493 the memory occupied by the current function's stack frame;
20496 the memory occupied by an incoming stack argument;
20499 the memory occupied by an object with a link-time-constant address.
20502 It is the kernel's responsibility to ensure that speculative
20503 accesses to these regions are indeed safe.
20505 If the input program contains a function declaration such as:
20511 then the implementation of @code{foo} must allow @code{j foo} and
20512 @code{jal foo} to be executed speculatively. GCC honors this
20513 restriction for functions it compiles itself. It expects non-GCC
20514 functions (such as hand-written assembly code) to do the same.
20516 The option has three forms:
20519 @item -mr10k-cache-barrier=load-store
20520 Insert a cache barrier before a load or store that might be
20521 speculatively executed and that might have side effects even
20524 @item -mr10k-cache-barrier=store
20525 Insert a cache barrier before a store that might be speculatively
20526 executed and that might have side effects even if aborted.
20528 @item -mr10k-cache-barrier=none
20529 Disable the insertion of cache barriers. This is the default setting.
20532 @item -mflush-func=@var{func}
20533 @itemx -mno-flush-func
20534 @opindex mflush-func
20535 Specifies the function to call to flush the I and D caches, or to not
20536 call any such function. If called, the function must take the same
20537 arguments as the common @code{_flush_func}, that is, the address of the
20538 memory range for which the cache is being flushed, the size of the
20539 memory range, and the number 3 (to flush both caches). The default
20540 depends on the target GCC was configured for, but commonly is either
20541 @code{_flush_func} or @code{__cpu_flush}.
20543 @item mbranch-cost=@var{num}
20544 @opindex mbranch-cost
20545 Set the cost of branches to roughly @var{num} ``simple'' instructions.
20546 This cost is only a heuristic and is not guaranteed to produce
20547 consistent results across releases. A zero cost redundantly selects
20548 the default, which is based on the @option{-mtune} setting.
20550 @item -mbranch-likely
20551 @itemx -mno-branch-likely
20552 @opindex mbranch-likely
20553 @opindex mno-branch-likely
20554 Enable or disable use of Branch Likely instructions, regardless of the
20555 default for the selected architecture. By default, Branch Likely
20556 instructions may be generated if they are supported by the selected
20557 architecture. An exception is for the MIPS32 and MIPS64 architectures
20558 and processors that implement those architectures; for those, Branch
20559 Likely instructions are not be generated by default because the MIPS32
20560 and MIPS64 architectures specifically deprecate their use.
20562 @item -mcompact-branches=never
20563 @itemx -mcompact-branches=optimal
20564 @itemx -mcompact-branches=always
20565 @opindex mcompact-branches=never
20566 @opindex mcompact-branches=optimal
20567 @opindex mcompact-branches=always
20568 These options control which form of branches will be generated. The
20569 default is @option{-mcompact-branches=optimal}.
20571 The @option{-mcompact-branches=never} option ensures that compact branch
20572 instructions will never be generated.
20574 The @option{-mcompact-branches=always} option ensures that a compact
20575 branch instruction will be generated if available. If a compact branch
20576 instruction is not available, a delay slot form of the branch will be
20579 This option is supported from MIPS Release 6 onwards.
20581 The @option{-mcompact-branches=optimal} option will cause a delay slot
20582 branch to be used if one is available in the current ISA and the delay
20583 slot is successfully filled. If the delay slot is not filled, a compact
20584 branch will be chosen if one is available.
20586 @item -mfp-exceptions
20587 @itemx -mno-fp-exceptions
20588 @opindex mfp-exceptions
20589 Specifies whether FP exceptions are enabled. This affects how
20590 FP instructions are scheduled for some processors.
20591 The default is that FP exceptions are
20594 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
20595 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
20598 @item -mvr4130-align
20599 @itemx -mno-vr4130-align
20600 @opindex mvr4130-align
20601 The VR4130 pipeline is two-way superscalar, but can only issue two
20602 instructions together if the first one is 8-byte aligned. When this
20603 option is enabled, GCC aligns pairs of instructions that it
20604 thinks should execute in parallel.
20606 This option only has an effect when optimizing for the VR4130.
20607 It normally makes code faster, but at the expense of making it bigger.
20608 It is enabled by default at optimization level @option{-O3}.
20613 Enable (disable) generation of @code{synci} instructions on
20614 architectures that support it. The @code{synci} instructions (if
20615 enabled) are generated when @code{__builtin___clear_cache} is
20618 This option defaults to @option{-mno-synci}, but the default can be
20619 overridden by configuring GCC with @option{--with-synci}.
20621 When compiling code for single processor systems, it is generally safe
20622 to use @code{synci}. However, on many multi-core (SMP) systems, it
20623 does not invalidate the instruction caches on all cores and may lead
20624 to undefined behavior.
20626 @item -mrelax-pic-calls
20627 @itemx -mno-relax-pic-calls
20628 @opindex mrelax-pic-calls
20629 Try to turn PIC calls that are normally dispatched via register
20630 @code{$25} into direct calls. This is only possible if the linker can
20631 resolve the destination at link time and if the destination is within
20632 range for a direct call.
20634 @option{-mrelax-pic-calls} is the default if GCC was configured to use
20635 an assembler and a linker that support the @code{.reloc} assembly
20636 directive and @option{-mexplicit-relocs} is in effect. With
20637 @option{-mno-explicit-relocs}, this optimization can be performed by the
20638 assembler and the linker alone without help from the compiler.
20640 @item -mmcount-ra-address
20641 @itemx -mno-mcount-ra-address
20642 @opindex mmcount-ra-address
20643 @opindex mno-mcount-ra-address
20644 Emit (do not emit) code that allows @code{_mcount} to modify the
20645 calling function's return address. When enabled, this option extends
20646 the usual @code{_mcount} interface with a new @var{ra-address}
20647 parameter, which has type @code{intptr_t *} and is passed in register
20648 @code{$12}. @code{_mcount} can then modify the return address by
20649 doing both of the following:
20652 Returning the new address in register @code{$31}.
20654 Storing the new address in @code{*@var{ra-address}},
20655 if @var{ra-address} is nonnull.
20658 The default is @option{-mno-mcount-ra-address}.
20660 @item -mframe-header-opt
20661 @itemx -mno-frame-header-opt
20662 @opindex mframe-header-opt
20663 Enable (disable) frame header optimization in the o32 ABI. When using the
20664 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
20665 function to write out register arguments. When enabled, this optimization
20666 will suppress the allocation of the frame header if it can be determined that
20669 This optimization is off by default at all optimization levels.
20672 @itemx -mno-lxc1-sxc1
20673 @opindex mlxc1-sxc1
20674 When applicable, enable (disable) the generation of @code{lwxc1},
20675 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
20680 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
20681 @code{madd.d} and related instructions. Enabled by default.
20686 @subsection MMIX Options
20687 @cindex MMIX Options
20689 These options are defined for the MMIX:
20693 @itemx -mno-libfuncs
20695 @opindex mno-libfuncs
20696 Specify that intrinsic library functions are being compiled, passing all
20697 values in registers, no matter the size.
20700 @itemx -mno-epsilon
20702 @opindex mno-epsilon
20703 Generate floating-point comparison instructions that compare with respect
20704 to the @code{rE} epsilon register.
20706 @item -mabi=mmixware
20708 @opindex mabi=mmixware
20710 Generate code that passes function parameters and return values that (in
20711 the called function) are seen as registers @code{$0} and up, as opposed to
20712 the GNU ABI which uses global registers @code{$231} and up.
20714 @item -mzero-extend
20715 @itemx -mno-zero-extend
20716 @opindex mzero-extend
20717 @opindex mno-zero-extend
20718 When reading data from memory in sizes shorter than 64 bits, use (do not
20719 use) zero-extending load instructions by default, rather than
20720 sign-extending ones.
20723 @itemx -mno-knuthdiv
20725 @opindex mno-knuthdiv
20726 Make the result of a division yielding a remainder have the same sign as
20727 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
20728 remainder follows the sign of the dividend. Both methods are
20729 arithmetically valid, the latter being almost exclusively used.
20731 @item -mtoplevel-symbols
20732 @itemx -mno-toplevel-symbols
20733 @opindex mtoplevel-symbols
20734 @opindex mno-toplevel-symbols
20735 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
20736 code can be used with the @code{PREFIX} assembly directive.
20740 Generate an executable in the ELF format, rather than the default
20741 @samp{mmo} format used by the @command{mmix} simulator.
20743 @item -mbranch-predict
20744 @itemx -mno-branch-predict
20745 @opindex mbranch-predict
20746 @opindex mno-branch-predict
20747 Use (do not use) the probable-branch instructions, when static branch
20748 prediction indicates a probable branch.
20750 @item -mbase-addresses
20751 @itemx -mno-base-addresses
20752 @opindex mbase-addresses
20753 @opindex mno-base-addresses
20754 Generate (do not generate) code that uses @emph{base addresses}. Using a
20755 base address automatically generates a request (handled by the assembler
20756 and the linker) for a constant to be set up in a global register. The
20757 register is used for one or more base address requests within the range 0
20758 to 255 from the value held in the register. The generally leads to short
20759 and fast code, but the number of different data items that can be
20760 addressed is limited. This means that a program that uses lots of static
20761 data may require @option{-mno-base-addresses}.
20763 @item -msingle-exit
20764 @itemx -mno-single-exit
20765 @opindex msingle-exit
20766 @opindex mno-single-exit
20767 Force (do not force) generated code to have a single exit point in each
20771 @node MN10300 Options
20772 @subsection MN10300 Options
20773 @cindex MN10300 options
20775 These @option{-m} options are defined for Matsushita MN10300 architectures:
20780 Generate code to avoid bugs in the multiply instructions for the MN10300
20781 processors. This is the default.
20783 @item -mno-mult-bug
20784 @opindex mno-mult-bug
20785 Do not generate code to avoid bugs in the multiply instructions for the
20786 MN10300 processors.
20790 Generate code using features specific to the AM33 processor.
20794 Do not generate code using features specific to the AM33 processor. This
20799 Generate code using features specific to the AM33/2.0 processor.
20803 Generate code using features specific to the AM34 processor.
20805 @item -mtune=@var{cpu-type}
20807 Use the timing characteristics of the indicated CPU type when
20808 scheduling instructions. This does not change the targeted processor
20809 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
20810 @samp{am33-2} or @samp{am34}.
20812 @item -mreturn-pointer-on-d0
20813 @opindex mreturn-pointer-on-d0
20814 When generating a function that returns a pointer, return the pointer
20815 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
20816 only in @code{a0}, and attempts to call such functions without a prototype
20817 result in errors. Note that this option is on by default; use
20818 @option{-mno-return-pointer-on-d0} to disable it.
20822 Do not link in the C run-time initialization object file.
20826 Indicate to the linker that it should perform a relaxation optimization pass
20827 to shorten branches, calls and absolute memory addresses. This option only
20828 has an effect when used on the command line for the final link step.
20830 This option makes symbolic debugging impossible.
20834 Allow the compiler to generate @emph{Long Instruction Word}
20835 instructions if the target is the @samp{AM33} or later. This is the
20836 default. This option defines the preprocessor macro @code{__LIW__}.
20840 Do not allow the compiler to generate @emph{Long Instruction Word}
20841 instructions. This option defines the preprocessor macro
20846 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
20847 instructions if the target is the @samp{AM33} or later. This is the
20848 default. This option defines the preprocessor macro @code{__SETLB__}.
20852 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
20853 instructions. This option defines the preprocessor macro
20854 @code{__NO_SETLB__}.
20858 @node Moxie Options
20859 @subsection Moxie Options
20860 @cindex Moxie Options
20866 Generate big-endian code. This is the default for @samp{moxie-*-*}
20871 Generate little-endian code.
20875 Generate mul.x and umul.x instructions. This is the default for
20876 @samp{moxiebox-*-*} configurations.
20880 Do not link in the C run-time initialization object file.
20884 @node MSP430 Options
20885 @subsection MSP430 Options
20886 @cindex MSP430 Options
20888 These options are defined for the MSP430:
20894 Force assembly output to always use hex constants. Normally such
20895 constants are signed decimals, but this option is available for
20896 testsuite and/or aesthetic purposes.
20900 Select the MCU to target. This is used to create a C preprocessor
20901 symbol based upon the MCU name, converted to upper case and pre- and
20902 post-fixed with @samp{__}. This in turn is used by the
20903 @file{msp430.h} header file to select an MCU-specific supplementary
20906 The option also sets the ISA to use. If the MCU name is one that is
20907 known to only support the 430 ISA then that is selected, otherwise the
20908 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
20909 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
20910 name selects the 430X ISA.
20912 In addition an MCU-specific linker script is added to the linker
20913 command line. The script's name is the name of the MCU with
20914 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
20915 command line defines the C preprocessor symbol @code{__XXX__} and
20916 cause the linker to search for a script called @file{xxx.ld}.
20918 This option is also passed on to the assembler.
20921 @itemx -mno-warn-mcu
20923 @opindex mno-warn-mcu
20924 This option enables or disables warnings about conflicts between the
20925 MCU name specified by the @option{-mmcu} option and the ISA set by the
20926 @option{-mcpu} option and/or the hardware multiply support set by the
20927 @option{-mhwmult} option. It also toggles warnings about unrecognized
20928 MCU names. This option is on by default.
20932 Specifies the ISA to use. Accepted values are @samp{msp430},
20933 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
20934 @option{-mmcu=} option should be used to select the ISA.
20938 Link to the simulator runtime libraries and linker script. Overrides
20939 any scripts that would be selected by the @option{-mmcu=} option.
20943 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
20947 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
20951 This option is passed to the assembler and linker, and allows the
20952 linker to perform certain optimizations that cannot be done until
20957 Describes the type of hardware multiply supported by the target.
20958 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
20959 for the original 16-bit-only multiply supported by early MCUs.
20960 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
20961 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
20962 A value of @samp{auto} can also be given. This tells GCC to deduce
20963 the hardware multiply support based upon the MCU name provided by the
20964 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
20965 the MCU name is not recognized then no hardware multiply support is
20966 assumed. @code{auto} is the default setting.
20968 Hardware multiplies are normally performed by calling a library
20969 routine. This saves space in the generated code. When compiling at
20970 @option{-O3} or higher however the hardware multiplier is invoked
20971 inline. This makes for bigger, but faster code.
20973 The hardware multiply routines disable interrupts whilst running and
20974 restore the previous interrupt state when they finish. This makes
20975 them safe to use inside interrupt handlers as well as in normal code.
20979 Enable the use of a minimum runtime environment - no static
20980 initializers or constructors. This is intended for memory-constrained
20981 devices. The compiler includes special symbols in some objects
20982 that tell the linker and runtime which code fragments are required.
20984 @item -mcode-region=
20985 @itemx -mdata-region=
20986 @opindex mcode-region
20987 @opindex mdata-region
20988 These options tell the compiler where to place functions and data that
20989 do not have one of the @code{lower}, @code{upper}, @code{either} or
20990 @code{section} attributes. Possible values are @code{lower},
20991 @code{upper}, @code{either} or @code{any}. The first three behave
20992 like the corresponding attribute. The fourth possible value -
20993 @code{any} - is the default. It leaves placement entirely up to the
20994 linker script and how it assigns the standard sections
20995 (@code{.text}, @code{.data}, etc) to the memory regions.
20997 @item -msilicon-errata=
20998 @opindex msilicon-errata
20999 This option passes on a request to assembler to enable the fixes for
21000 the named silicon errata.
21002 @item -msilicon-errata-warn=
21003 @opindex msilicon-errata-warn
21004 This option passes on a request to the assembler to enable warning
21005 messages when a silicon errata might need to be applied.
21009 @node NDS32 Options
21010 @subsection NDS32 Options
21011 @cindex NDS32 Options
21013 These options are defined for NDS32 implementations:
21018 @opindex mbig-endian
21019 Generate code in big-endian mode.
21021 @item -mlittle-endian
21022 @opindex mlittle-endian
21023 Generate code in little-endian mode.
21025 @item -mreduced-regs
21026 @opindex mreduced-regs
21027 Use reduced-set registers for register allocation.
21030 @opindex mfull-regs
21031 Use full-set registers for register allocation.
21035 Generate conditional move instructions.
21039 Do not generate conditional move instructions.
21043 Generate performance extension instructions.
21045 @item -mno-perf-ext
21046 @opindex mno-perf-ext
21047 Do not generate performance extension instructions.
21051 Generate v3 push25/pop25 instructions.
21054 @opindex mno-v3push
21055 Do not generate v3 push25/pop25 instructions.
21059 Generate 16-bit instructions.
21062 @opindex mno-16-bit
21063 Do not generate 16-bit instructions.
21065 @item -misr-vector-size=@var{num}
21066 @opindex misr-vector-size
21067 Specify the size of each interrupt vector, which must be 4 or 16.
21069 @item -mcache-block-size=@var{num}
21070 @opindex mcache-block-size
21071 Specify the size of each cache block,
21072 which must be a power of 2 between 4 and 512.
21074 @item -march=@var{arch}
21076 Specify the name of the target architecture.
21078 @item -mcmodel=@var{code-model}
21080 Set the code model to one of
21083 All the data and read-only data segments must be within 512KB addressing space.
21084 The text segment must be within 16MB addressing space.
21085 @item @samp{medium}
21086 The data segment must be within 512KB while the read-only data segment can be
21087 within 4GB addressing space. The text segment should be still within 16MB
21090 All the text and data segments can be within 4GB addressing space.
21094 @opindex mctor-dtor
21095 Enable constructor/destructor feature.
21099 Guide linker to relax instructions.
21103 @node Nios II Options
21104 @subsection Nios II Options
21105 @cindex Nios II options
21106 @cindex Altera Nios II options
21108 These are the options defined for the Altera Nios II processor.
21114 @cindex smaller data references
21115 Put global and static objects less than or equal to @var{num} bytes
21116 into the small data or BSS sections instead of the normal data or BSS
21117 sections. The default value of @var{num} is 8.
21119 @item -mgpopt=@var{option}
21124 Generate (do not generate) GP-relative accesses. The following
21125 @var{option} names are recognized:
21130 Do not generate GP-relative accesses.
21133 Generate GP-relative accesses for small data objects that are not
21134 external, weak, or uninitialized common symbols.
21135 Also use GP-relative addressing for objects that
21136 have been explicitly placed in a small data section via a @code{section}
21140 As for @samp{local}, but also generate GP-relative accesses for
21141 small data objects that are external, weak, or common. If you use this option,
21142 you must ensure that all parts of your program (including libraries) are
21143 compiled with the same @option{-G} setting.
21146 Generate GP-relative accesses for all data objects in the program. If you
21147 use this option, the entire data and BSS segments
21148 of your program must fit in 64K of memory and you must use an appropriate
21149 linker script to allocate them within the addressable range of the
21153 Generate GP-relative addresses for function pointers as well as data
21154 pointers. If you use this option, the entire text, data, and BSS segments
21155 of your program must fit in 64K of memory and you must use an appropriate
21156 linker script to allocate them within the addressable range of the
21161 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
21162 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
21164 The default is @option{-mgpopt} except when @option{-fpic} or
21165 @option{-fPIC} is specified to generate position-independent code.
21166 Note that the Nios II ABI does not permit GP-relative accesses from
21169 You may need to specify @option{-mno-gpopt} explicitly when building
21170 programs that include large amounts of small data, including large
21171 GOT data sections. In this case, the 16-bit offset for GP-relative
21172 addressing may not be large enough to allow access to the entire
21173 small data section.
21175 @item -mgprel-sec=@var{regexp}
21176 @opindex mgprel-sec
21177 This option specifies additional section names that can be accessed via
21178 GP-relative addressing. It is most useful in conjunction with
21179 @code{section} attributes on variable declarations
21180 (@pxref{Common Variable Attributes}) and a custom linker script.
21181 The @var{regexp} is a POSIX Extended Regular Expression.
21183 This option does not affect the behavior of the @option{-G} option, and
21184 and the specified sections are in addition to the standard @code{.sdata}
21185 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
21191 Generate little-endian (default) or big-endian (experimental) code,
21194 @item -march=@var{arch}
21196 This specifies the name of the target Nios II architecture. GCC uses this
21197 name to determine what kind of instructions it can emit when generating
21198 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
21200 The preprocessor macro @code{__nios2_arch__} is available to programs,
21201 with value 1 or 2, indicating the targeted ISA level.
21203 @item -mbypass-cache
21204 @itemx -mno-bypass-cache
21205 @opindex mno-bypass-cache
21206 @opindex mbypass-cache
21207 Force all load and store instructions to always bypass cache by
21208 using I/O variants of the instructions. The default is not to
21211 @item -mno-cache-volatile
21212 @itemx -mcache-volatile
21213 @opindex mcache-volatile
21214 @opindex mno-cache-volatile
21215 Volatile memory access bypass the cache using the I/O variants of
21216 the load and store instructions. The default is not to bypass the cache.
21218 @item -mno-fast-sw-div
21219 @itemx -mfast-sw-div
21220 @opindex mno-fast-sw-div
21221 @opindex mfast-sw-div
21222 Do not use table-based fast divide for small numbers. The default
21223 is to use the fast divide at @option{-O3} and above.
21227 @itemx -mno-hw-mulx
21231 @opindex mno-hw-mul
21233 @opindex mno-hw-mulx
21235 @opindex mno-hw-div
21237 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
21238 instructions by the compiler. The default is to emit @code{mul}
21239 and not emit @code{div} and @code{mulx}.
21245 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
21246 CDX (code density) instructions. Enabling these instructions also
21247 requires @option{-march=r2}. Since these instructions are optional
21248 extensions to the R2 architecture, the default is not to emit them.
21250 @item -mcustom-@var{insn}=@var{N}
21251 @itemx -mno-custom-@var{insn}
21252 @opindex mcustom-@var{insn}
21253 @opindex mno-custom-@var{insn}
21254 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
21255 custom instruction with encoding @var{N} when generating code that uses
21256 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
21257 instruction 253 for single-precision floating-point add operations instead
21258 of the default behavior of using a library call.
21260 The following values of @var{insn} are supported. Except as otherwise
21261 noted, floating-point operations are expected to be implemented with
21262 normal IEEE 754 semantics and correspond directly to the C operators or the
21263 equivalent GCC built-in functions (@pxref{Other Builtins}).
21265 Single-precision floating point:
21268 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
21269 Binary arithmetic operations.
21275 Unary absolute value.
21277 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
21278 Comparison operations.
21280 @item @samp{fmins}, @samp{fmaxs}
21281 Floating-point minimum and maximum. These instructions are only
21282 generated if @option{-ffinite-math-only} is specified.
21284 @item @samp{fsqrts}
21285 Unary square root operation.
21287 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
21288 Floating-point trigonometric and exponential functions. These instructions
21289 are only generated if @option{-funsafe-math-optimizations} is also specified.
21293 Double-precision floating point:
21296 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
21297 Binary arithmetic operations.
21303 Unary absolute value.
21305 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
21306 Comparison operations.
21308 @item @samp{fmind}, @samp{fmaxd}
21309 Double-precision minimum and maximum. These instructions are only
21310 generated if @option{-ffinite-math-only} is specified.
21312 @item @samp{fsqrtd}
21313 Unary square root operation.
21315 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
21316 Double-precision trigonometric and exponential functions. These instructions
21317 are only generated if @option{-funsafe-math-optimizations} is also specified.
21323 @item @samp{fextsd}
21324 Conversion from single precision to double precision.
21326 @item @samp{ftruncds}
21327 Conversion from double precision to single precision.
21329 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
21330 Conversion from floating point to signed or unsigned integer types, with
21331 truncation towards zero.
21334 Conversion from single-precision floating point to signed integer,
21335 rounding to the nearest integer and ties away from zero.
21336 This corresponds to the @code{__builtin_lroundf} function when
21337 @option{-fno-math-errno} is used.
21339 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
21340 Conversion from signed or unsigned integer types to floating-point types.
21344 In addition, all of the following transfer instructions for internal
21345 registers X and Y must be provided to use any of the double-precision
21346 floating-point instructions. Custom instructions taking two
21347 double-precision source operands expect the first operand in the
21348 64-bit register X. The other operand (or only operand of a unary
21349 operation) is given to the custom arithmetic instruction with the
21350 least significant half in source register @var{src1} and the most
21351 significant half in @var{src2}. A custom instruction that returns a
21352 double-precision result returns the most significant 32 bits in the
21353 destination register and the other half in 32-bit register Y.
21354 GCC automatically generates the necessary code sequences to write
21355 register X and/or read register Y when double-precision floating-point
21356 instructions are used.
21361 Write @var{src1} into the least significant half of X and @var{src2} into
21362 the most significant half of X.
21365 Write @var{src1} into Y.
21367 @item @samp{frdxhi}, @samp{frdxlo}
21368 Read the most or least (respectively) significant half of X and store it in
21372 Read the value of Y and store it into @var{dest}.
21375 Note that you can gain more local control over generation of Nios II custom
21376 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
21377 and @code{target("no-custom-@var{insn}")} function attributes
21378 (@pxref{Function Attributes})
21379 or pragmas (@pxref{Function Specific Option Pragmas}).
21381 @item -mcustom-fpu-cfg=@var{name}
21382 @opindex mcustom-fpu-cfg
21384 This option enables a predefined, named set of custom instruction encodings
21385 (see @option{-mcustom-@var{insn}} above).
21386 Currently, the following sets are defined:
21388 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
21389 @gccoptlist{-mcustom-fmuls=252 @gol
21390 -mcustom-fadds=253 @gol
21391 -mcustom-fsubs=254 @gol
21392 -fsingle-precision-constant}
21394 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
21395 @gccoptlist{-mcustom-fmuls=252 @gol
21396 -mcustom-fadds=253 @gol
21397 -mcustom-fsubs=254 @gol
21398 -mcustom-fdivs=255 @gol
21399 -fsingle-precision-constant}
21401 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
21402 @gccoptlist{-mcustom-floatus=243 @gol
21403 -mcustom-fixsi=244 @gol
21404 -mcustom-floatis=245 @gol
21405 -mcustom-fcmpgts=246 @gol
21406 -mcustom-fcmples=249 @gol
21407 -mcustom-fcmpeqs=250 @gol
21408 -mcustom-fcmpnes=251 @gol
21409 -mcustom-fmuls=252 @gol
21410 -mcustom-fadds=253 @gol
21411 -mcustom-fsubs=254 @gol
21412 -mcustom-fdivs=255 @gol
21413 -fsingle-precision-constant}
21415 Custom instruction assignments given by individual
21416 @option{-mcustom-@var{insn}=} options override those given by
21417 @option{-mcustom-fpu-cfg=}, regardless of the
21418 order of the options on the command line.
21420 Note that you can gain more local control over selection of a FPU
21421 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
21422 function attribute (@pxref{Function Attributes})
21423 or pragma (@pxref{Function Specific Option Pragmas}).
21427 These additional @samp{-m} options are available for the Altera Nios II
21428 ELF (bare-metal) target:
21434 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
21435 startup and termination code, and is typically used in conjunction with
21436 @option{-msys-crt0=} to specify the location of the alternate startup code
21437 provided by the HAL BSP.
21441 Link with a limited version of the C library, @option{-lsmallc}, rather than
21444 @item -msys-crt0=@var{startfile}
21446 @var{startfile} is the file name of the startfile (crt0) to use
21447 when linking. This option is only useful in conjunction with @option{-mhal}.
21449 @item -msys-lib=@var{systemlib}
21451 @var{systemlib} is the library name of the library that provides
21452 low-level system calls required by the C library,
21453 e.g. @code{read} and @code{write}.
21454 This option is typically used to link with a library provided by a HAL BSP.
21458 @node Nvidia PTX Options
21459 @subsection Nvidia PTX Options
21460 @cindex Nvidia PTX options
21461 @cindex nvptx options
21463 These options are defined for Nvidia PTX:
21471 Generate code for 32-bit or 64-bit ABI.
21474 @opindex mmainkernel
21475 Link in code for a __main kernel. This is for stand-alone instead of
21476 offloading execution.
21480 Apply partitioned execution optimizations. This is the default when any
21481 level of optimization is selected.
21484 @opindex msoft-stack
21485 Generate code that does not use @code{.local} memory
21486 directly for stack storage. Instead, a per-warp stack pointer is
21487 maintained explicitly. This enables variable-length stack allocation (with
21488 variable-length arrays or @code{alloca}), and when global memory is used for
21489 underlying storage, makes it possible to access automatic variables from other
21490 threads, or with atomic instructions. This code generation variant is used
21491 for OpenMP offloading, but the option is exposed on its own for the purpose
21492 of testing the compiler; to generate code suitable for linking into programs
21493 using OpenMP offloading, use option @option{-mgomp}.
21495 @item -muniform-simt
21496 @opindex muniform-simt
21497 Switch to code generation variant that allows to execute all threads in each
21498 warp, while maintaining memory state and side effects as if only one thread
21499 in each warp was active outside of OpenMP SIMD regions. All atomic operations
21500 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
21501 current lane index equals the master lane index), and the register being
21502 assigned is copied via a shuffle instruction from the master lane. Outside of
21503 SIMD regions lane 0 is the master; inside, each thread sees itself as the
21504 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
21505 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
21506 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
21507 with current lane index to compute the master lane index.
21511 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
21512 @option{-muniform-simt} options, and selects corresponding multilib variant.
21516 @node PDP-11 Options
21517 @subsection PDP-11 Options
21518 @cindex PDP-11 Options
21520 These options are defined for the PDP-11:
21525 Use hardware FPP floating point. This is the default. (FIS floating
21526 point on the PDP-11/40 is not supported.)
21529 @opindex msoft-float
21530 Do not use hardware floating point.
21534 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
21538 Return floating-point results in memory. This is the default.
21542 Generate code for a PDP-11/40.
21546 Generate code for a PDP-11/45. This is the default.
21550 Generate code for a PDP-11/10.
21552 @item -mbcopy-builtin
21553 @opindex mbcopy-builtin
21554 Use inline @code{movmemhi} patterns for copying memory. This is the
21559 Do not use inline @code{movmemhi} patterns for copying memory.
21565 Use 16-bit @code{int}. This is the default.
21571 Use 32-bit @code{int}.
21574 @itemx -mno-float32
21576 @opindex mno-float32
21577 Use 64-bit @code{float}. This is the default.
21580 @itemx -mno-float64
21582 @opindex mno-float64
21583 Use 32-bit @code{float}.
21587 Use @code{abshi2} pattern. This is the default.
21591 Do not use @code{abshi2} pattern.
21593 @item -mbranch-expensive
21594 @opindex mbranch-expensive
21595 Pretend that branches are expensive. This is for experimenting with
21596 code generation only.
21598 @item -mbranch-cheap
21599 @opindex mbranch-cheap
21600 Do not pretend that branches are expensive. This is the default.
21604 Use Unix assembler syntax. This is the default when configured for
21605 @samp{pdp11-*-bsd}.
21609 Use DEC assembler syntax. This is the default when configured for any
21610 PDP-11 target other than @samp{pdp11-*-bsd}.
21613 @node picoChip Options
21614 @subsection picoChip Options
21615 @cindex picoChip options
21617 These @samp{-m} options are defined for picoChip implementations:
21621 @item -mae=@var{ae_type}
21623 Set the instruction set, register set, and instruction scheduling
21624 parameters for array element type @var{ae_type}. Supported values
21625 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
21627 @option{-mae=ANY} selects a completely generic AE type. Code
21628 generated with this option runs on any of the other AE types. The
21629 code is not as efficient as it would be if compiled for a specific
21630 AE type, and some types of operation (e.g., multiplication) do not
21631 work properly on all types of AE.
21633 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
21634 for compiled code, and is the default.
21636 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
21637 option may suffer from poor performance of byte (char) manipulation,
21638 since the DSP AE does not provide hardware support for byte load/stores.
21640 @item -msymbol-as-address
21641 Enable the compiler to directly use a symbol name as an address in a
21642 load/store instruction, without first loading it into a
21643 register. Typically, the use of this option generates larger
21644 programs, which run faster than when the option isn't used. However, the
21645 results vary from program to program, so it is left as a user option,
21646 rather than being permanently enabled.
21648 @item -mno-inefficient-warnings
21649 Disables warnings about the generation of inefficient code. These
21650 warnings can be generated, for example, when compiling code that
21651 performs byte-level memory operations on the MAC AE type. The MAC AE has
21652 no hardware support for byte-level memory operations, so all byte
21653 load/stores must be synthesized from word load/store operations. This is
21654 inefficient and a warning is generated to indicate
21655 that you should rewrite the code to avoid byte operations, or to target
21656 an AE type that has the necessary hardware support. This option disables
21661 @node PowerPC Options
21662 @subsection PowerPC Options
21663 @cindex PowerPC options
21665 These are listed under @xref{RS/6000 and PowerPC Options}.
21667 @node RISC-V Options
21668 @subsection RISC-V Options
21669 @cindex RISC-V Options
21671 These command-line options are defined for RISC-V targets:
21674 @item -mbranch-cost=@var{n}
21675 @opindex mbranch-cost
21676 Set the cost of branches to roughly @var{n} instructions.
21681 Don't optimize block moves.
21686 When generating PIC code, allow the use of PLTs. Ignored for non-PIC.
21688 @item -mabi=@var{ABI-string}
21690 Specify integer and floating-point calling convention. This defaults to the
21691 natural calling convention: e.g.@ LP64 for RV64I, ILP32 for RV32I, LP64D for
21697 Use hardware floating-point divide and square root instructions. This requires
21698 the F or D extensions for floating-point registers.
21703 Use hardware instructions for integer division. This requires the M extension.
21705 @item -march=@var{ISA-string}
21707 Generate code for given RISC-V ISA (e.g.@ @samp{rv64im}). ISA strings must be
21708 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, and @samp{rv32imaf}.
21710 @item -mtune=@var{processor-string}
21712 Optimize the output for the given processor, specified by microarchitecture
21715 @item -msmall-data-limit=@var{n}
21716 @opindex msmall-data-limit
21717 Put global and static data smaller than @var{n} bytes into a special section
21720 @item -msave-restore
21721 @itemx -mno-save-restore
21722 @opindex msave-restore
21723 Use smaller but slower prologue and epilogue code.
21725 @item -mstrict-align
21726 @itemx -mno-strict-align
21727 @opindex mstrict-align
21728 Do not generate unaligned memory accesses.
21730 @item -mcmodel=@var{code-model}
21732 Specify the code model.
21737 @subsection RL78 Options
21738 @cindex RL78 Options
21744 Links in additional target libraries to support operation within a
21753 Specifies the type of hardware multiplication and division support to
21754 be used. The simplest is @code{none}, which uses software for both
21755 multiplication and division. This is the default. The @code{g13}
21756 value is for the hardware multiply/divide peripheral found on the
21757 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
21758 the multiplication and division instructions supported by the RL78/G14
21759 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
21760 the value @code{mg10} is an alias for @code{none}.
21762 In addition a C preprocessor macro is defined, based upon the setting
21763 of this option. Possible values are: @code{__RL78_MUL_NONE__},
21764 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
21771 Specifies the RL78 core to target. The default is the G14 core, also
21772 known as an S3 core or just RL78. The G13 or S2 core does not have
21773 multiply or divide instructions, instead it uses a hardware peripheral
21774 for these operations. The G10 or S1 core does not have register
21775 banks, so it uses a different calling convention.
21777 If this option is set it also selects the type of hardware multiply
21778 support to use, unless this is overridden by an explicit
21779 @option{-mmul=none} option on the command line. Thus specifying
21780 @option{-mcpu=g13} enables the use of the G13 hardware multiply
21781 peripheral and specifying @option{-mcpu=g10} disables the use of
21782 hardware multiplications altogether.
21784 Note, although the RL78/G14 core is the default target, specifying
21785 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
21786 change the behavior of the toolchain since it also enables G14
21787 hardware multiply support. If these options are not specified on the
21788 command line then software multiplication routines will be used even
21789 though the code targets the RL78 core. This is for backwards
21790 compatibility with older toolchains which did not have hardware
21791 multiply and divide support.
21793 In addition a C preprocessor macro is defined, based upon the setting
21794 of this option. Possible values are: @code{__RL78_G10__},
21795 @code{__RL78_G13__} or @code{__RL78_G14__}.
21805 These are aliases for the corresponding @option{-mcpu=} option. They
21806 are provided for backwards compatibility.
21810 Allow the compiler to use all of the available registers. By default
21811 registers @code{r24..r31} are reserved for use in interrupt handlers.
21812 With this option enabled these registers can be used in ordinary
21815 @item -m64bit-doubles
21816 @itemx -m32bit-doubles
21817 @opindex m64bit-doubles
21818 @opindex m32bit-doubles
21819 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21820 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21821 @option{-m32bit-doubles}.
21823 @item -msave-mduc-in-interrupts
21824 @item -mno-save-mduc-in-interrupts
21825 @opindex msave-mduc-in-interrupts
21826 @opindex mno-save-mduc-in-interrupts
21827 Specifies that interrupt handler functions should preserve the
21828 MDUC registers. This is only necessary if normal code might use
21829 the MDUC registers, for example because it performs multiplication
21830 and division operations. The default is to ignore the MDUC registers
21831 as this makes the interrupt handlers faster. The target option -mg13
21832 needs to be passed for this to work as this feature is only available
21833 on the G13 target (S2 core). The MDUC registers will only be saved
21834 if the interrupt handler performs a multiplication or division
21835 operation or it calls another function.
21839 @node RS/6000 and PowerPC Options
21840 @subsection IBM RS/6000 and PowerPC Options
21841 @cindex RS/6000 and PowerPC Options
21842 @cindex IBM RS/6000 and PowerPC Options
21844 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
21846 @item -mpowerpc-gpopt
21847 @itemx -mno-powerpc-gpopt
21848 @itemx -mpowerpc-gfxopt
21849 @itemx -mno-powerpc-gfxopt
21852 @itemx -mno-powerpc64
21856 @itemx -mno-popcntb
21858 @itemx -mno-popcntd
21867 @itemx -mno-hard-dfp
21868 @opindex mpowerpc-gpopt
21869 @opindex mno-powerpc-gpopt
21870 @opindex mpowerpc-gfxopt
21871 @opindex mno-powerpc-gfxopt
21872 @opindex mpowerpc64
21873 @opindex mno-powerpc64
21877 @opindex mno-popcntb
21879 @opindex mno-popcntd
21885 @opindex mno-mfpgpr
21887 @opindex mno-hard-dfp
21888 You use these options to specify which instructions are available on the
21889 processor you are using. The default value of these options is
21890 determined when configuring GCC@. Specifying the
21891 @option{-mcpu=@var{cpu_type}} overrides the specification of these
21892 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
21893 rather than the options listed above.
21895 Specifying @option{-mpowerpc-gpopt} allows
21896 GCC to use the optional PowerPC architecture instructions in the
21897 General Purpose group, including floating-point square root. Specifying
21898 @option{-mpowerpc-gfxopt} allows GCC to
21899 use the optional PowerPC architecture instructions in the Graphics
21900 group, including floating-point select.
21902 The @option{-mmfcrf} option allows GCC to generate the move from
21903 condition register field instruction implemented on the POWER4
21904 processor and other processors that support the PowerPC V2.01
21906 The @option{-mpopcntb} option allows GCC to generate the popcount and
21907 double-precision FP reciprocal estimate instruction implemented on the
21908 POWER5 processor and other processors that support the PowerPC V2.02
21910 The @option{-mpopcntd} option allows GCC to generate the popcount
21911 instruction implemented on the POWER7 processor and other processors
21912 that support the PowerPC V2.06 architecture.
21913 The @option{-mfprnd} option allows GCC to generate the FP round to
21914 integer instructions implemented on the POWER5+ processor and other
21915 processors that support the PowerPC V2.03 architecture.
21916 The @option{-mcmpb} option allows GCC to generate the compare bytes
21917 instruction implemented on the POWER6 processor and other processors
21918 that support the PowerPC V2.05 architecture.
21919 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
21920 general-purpose register instructions implemented on the POWER6X
21921 processor and other processors that support the extended PowerPC V2.05
21923 The @option{-mhard-dfp} option allows GCC to generate the decimal
21924 floating-point instructions implemented on some POWER processors.
21926 The @option{-mpowerpc64} option allows GCC to generate the additional
21927 64-bit instructions that are found in the full PowerPC64 architecture
21928 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
21929 @option{-mno-powerpc64}.
21931 @item -mcpu=@var{cpu_type}
21933 Set architecture type, register usage, and
21934 instruction scheduling parameters for machine type @var{cpu_type}.
21935 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
21936 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
21937 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
21938 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
21939 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
21940 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
21941 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
21942 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
21943 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
21944 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
21945 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
21948 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
21949 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
21950 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
21951 architecture machine types, with an appropriate, generic processor
21952 model assumed for scheduling purposes.
21954 The other options specify a specific processor. Code generated under
21955 those options runs best on that processor, and may not run at all on
21958 The @option{-mcpu} options automatically enable or disable the
21961 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
21962 -mpopcntb -mpopcntd -mpowerpc64 @gol
21963 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
21964 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
21965 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
21966 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
21968 The particular options set for any particular CPU varies between
21969 compiler versions, depending on what setting seems to produce optimal
21970 code for that CPU; it doesn't necessarily reflect the actual hardware's
21971 capabilities. If you wish to set an individual option to a particular
21972 value, you may specify it after the @option{-mcpu} option, like
21973 @option{-mcpu=970 -mno-altivec}.
21975 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
21976 not enabled or disabled by the @option{-mcpu} option at present because
21977 AIX does not have full support for these options. You may still
21978 enable or disable them individually if you're sure it'll work in your
21981 @item -mtune=@var{cpu_type}
21983 Set the instruction scheduling parameters for machine type
21984 @var{cpu_type}, but do not set the architecture type or register usage,
21985 as @option{-mcpu=@var{cpu_type}} does. The same
21986 values for @var{cpu_type} are used for @option{-mtune} as for
21987 @option{-mcpu}. If both are specified, the code generated uses the
21988 architecture and registers set by @option{-mcpu}, but the
21989 scheduling parameters set by @option{-mtune}.
21991 @item -mcmodel=small
21992 @opindex mcmodel=small
21993 Generate PowerPC64 code for the small model: The TOC is limited to
21996 @item -mcmodel=medium
21997 @opindex mcmodel=medium
21998 Generate PowerPC64 code for the medium model: The TOC and other static
21999 data may be up to a total of 4G in size. This is the default for 64-bit
22002 @item -mcmodel=large
22003 @opindex mcmodel=large
22004 Generate PowerPC64 code for the large model: The TOC may be up to 4G
22005 in size. Other data and code is only limited by the 64-bit address
22009 @itemx -mno-altivec
22011 @opindex mno-altivec
22012 Generate code that uses (does not use) AltiVec instructions, and also
22013 enable the use of built-in functions that allow more direct access to
22014 the AltiVec instruction set. You may also need to set
22015 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
22018 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
22019 @option{-maltivec=be}, the element order for AltiVec intrinsics such
22020 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
22021 match array element order corresponding to the endianness of the
22022 target. That is, element zero identifies the leftmost element in a
22023 vector register when targeting a big-endian platform, and identifies
22024 the rightmost element in a vector register when targeting a
22025 little-endian platform.
22028 @opindex maltivec=be
22029 Generate AltiVec instructions using big-endian element order,
22030 regardless of whether the target is big- or little-endian. This is
22031 the default when targeting a big-endian platform.
22033 The element order is used to interpret element numbers in AltiVec
22034 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22035 @code{vec_insert}. By default, these match array element order
22036 corresponding to the endianness for the target.
22039 @opindex maltivec=le
22040 Generate AltiVec instructions using little-endian element order,
22041 regardless of whether the target is big- or little-endian. This is
22042 the default when targeting a little-endian platform. This option is
22043 currently ignored when targeting a big-endian platform.
22045 The element order is used to interpret element numbers in AltiVec
22046 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22047 @code{vec_insert}. By default, these match array element order
22048 corresponding to the endianness for the target.
22053 @opindex mno-vrsave
22054 Generate VRSAVE instructions when generating AltiVec code.
22057 @opindex msecure-plt
22058 Generate code that allows @command{ld} and @command{ld.so}
22059 to build executables and shared
22060 libraries with non-executable @code{.plt} and @code{.got} sections.
22062 32-bit SYSV ABI option.
22066 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
22068 requires @code{.plt} and @code{.got}
22069 sections that are both writable and executable.
22070 This is a PowerPC 32-bit SYSV ABI option.
22076 This switch enables or disables the generation of ISEL instructions.
22078 @item -misel=@var{yes/no}
22079 This switch has been deprecated. Use @option{-misel} and
22080 @option{-mno-isel} instead.
22086 This switch enables or disables the generation of SPE simd
22092 @opindex mno-paired
22093 This switch enables or disables the generation of PAIRED simd
22096 @item -mspe=@var{yes/no}
22097 This option has been deprecated. Use @option{-mspe} and
22098 @option{-mno-spe} instead.
22104 Generate code that uses (does not use) vector/scalar (VSX)
22105 instructions, and also enable the use of built-in functions that allow
22106 more direct access to the VSX instruction set.
22111 @opindex mno-crypto
22112 Enable the use (disable) of the built-in functions that allow direct
22113 access to the cryptographic instructions that were added in version
22114 2.07 of the PowerPC ISA.
22116 @item -mdirect-move
22117 @itemx -mno-direct-move
22118 @opindex mdirect-move
22119 @opindex mno-direct-move
22120 Generate code that uses (does not use) the instructions to move data
22121 between the general purpose registers and the vector/scalar (VSX)
22122 registers that were added in version 2.07 of the PowerPC ISA.
22128 Enable (disable) the use of the built-in functions that allow direct
22129 access to the Hardware Transactional Memory (HTM) instructions that
22130 were added in version 2.07 of the PowerPC ISA.
22132 @item -mpower8-fusion
22133 @itemx -mno-power8-fusion
22134 @opindex mpower8-fusion
22135 @opindex mno-power8-fusion
22136 Generate code that keeps (does not keeps) some integer operations
22137 adjacent so that the instructions can be fused together on power8 and
22140 @item -mpower8-vector
22141 @itemx -mno-power8-vector
22142 @opindex mpower8-vector
22143 @opindex mno-power8-vector
22144 Generate code that uses (does not use) the vector and scalar
22145 instructions that were added in version 2.07 of the PowerPC ISA. Also
22146 enable the use of built-in functions that allow more direct access to
22147 the vector instructions.
22149 @item -mquad-memory
22150 @itemx -mno-quad-memory
22151 @opindex mquad-memory
22152 @opindex mno-quad-memory
22153 Generate code that uses (does not use) the non-atomic quad word memory
22154 instructions. The @option{-mquad-memory} option requires use of
22157 @item -mquad-memory-atomic
22158 @itemx -mno-quad-memory-atomic
22159 @opindex mquad-memory-atomic
22160 @opindex mno-quad-memory-atomic
22161 Generate code that uses (does not use) the atomic quad word memory
22162 instructions. The @option{-mquad-memory-atomic} option requires use of
22166 @itemx -mno-float128
22168 @opindex mno-float128
22169 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
22170 and use either software emulation for IEEE 128-bit floating point or
22171 hardware instructions.
22173 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
22174 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
22175 use the IEEE 128-bit floating point support. The IEEE 128-bit
22176 floating point support only works on PowerPC Linux systems.
22178 The default for @option{-mfloat128} is enabled on PowerPC Linux
22179 systems using the VSX instruction set, and disabled on other systems.
22181 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
22182 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
22183 point support will also enable the generation of ISA 3.0 IEEE 128-bit
22184 floating point instructions. Otherwise, if you do not specify to
22185 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
22186 system, IEEE 128-bit floating point will be done with software
22189 @item -mfloat128-hardware
22190 @itemx -mno-float128-hardware
22191 @opindex mfloat128-hardware
22192 @opindex mno-float128-hardware
22193 Enable/disable using ISA 3.0 hardware instructions to support the
22194 @var{__float128} data type.
22196 The default for @option{-mfloat128-hardware} is enabled on PowerPC
22197 Linux systems using the ISA 3.0 instruction set, and disabled on other
22200 @item -mfloat-gprs=@var{yes/single/double/no}
22201 @itemx -mfloat-gprs
22202 @opindex mfloat-gprs
22203 This switch enables or disables the generation of floating-point
22204 operations on the general-purpose registers for architectures that
22207 The argument @samp{yes} or @samp{single} enables the use of
22208 single-precision floating-point operations.
22210 The argument @samp{double} enables the use of single and
22211 double-precision floating-point operations.
22213 The argument @samp{no} disables floating-point operations on the
22214 general-purpose registers.
22216 This option is currently only available on the MPC854x.
22222 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
22223 targets (including GNU/Linux). The 32-bit environment sets int, long
22224 and pointer to 32 bits and generates code that runs on any PowerPC
22225 variant. The 64-bit environment sets int to 32 bits and long and
22226 pointer to 64 bits, and generates code for PowerPC64, as for
22227 @option{-mpowerpc64}.
22230 @itemx -mno-fp-in-toc
22231 @itemx -mno-sum-in-toc
22232 @itemx -mminimal-toc
22234 @opindex mno-fp-in-toc
22235 @opindex mno-sum-in-toc
22236 @opindex mminimal-toc
22237 Modify generation of the TOC (Table Of Contents), which is created for
22238 every executable file. The @option{-mfull-toc} option is selected by
22239 default. In that case, GCC allocates at least one TOC entry for
22240 each unique non-automatic variable reference in your program. GCC
22241 also places floating-point constants in the TOC@. However, only
22242 16,384 entries are available in the TOC@.
22244 If you receive a linker error message that saying you have overflowed
22245 the available TOC space, you can reduce the amount of TOC space used
22246 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
22247 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
22248 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
22249 generate code to calculate the sum of an address and a constant at
22250 run time instead of putting that sum into the TOC@. You may specify one
22251 or both of these options. Each causes GCC to produce very slightly
22252 slower and larger code at the expense of conserving TOC space.
22254 If you still run out of space in the TOC even when you specify both of
22255 these options, specify @option{-mminimal-toc} instead. This option causes
22256 GCC to make only one TOC entry for every file. When you specify this
22257 option, GCC produces code that is slower and larger but which
22258 uses extremely little TOC space. You may wish to use this option
22259 only on files that contain less frequently-executed code.
22265 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
22266 @code{long} type, and the infrastructure needed to support them.
22267 Specifying @option{-maix64} implies @option{-mpowerpc64},
22268 while @option{-maix32} disables the 64-bit ABI and
22269 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
22272 @itemx -mno-xl-compat
22273 @opindex mxl-compat
22274 @opindex mno-xl-compat
22275 Produce code that conforms more closely to IBM XL compiler semantics
22276 when using AIX-compatible ABI@. Pass floating-point arguments to
22277 prototyped functions beyond the register save area (RSA) on the stack
22278 in addition to argument FPRs. Do not assume that most significant
22279 double in 128-bit long double value is properly rounded when comparing
22280 values and converting to double. Use XL symbol names for long double
22283 The AIX calling convention was extended but not initially documented to
22284 handle an obscure K&R C case of calling a function that takes the
22285 address of its arguments with fewer arguments than declared. IBM XL
22286 compilers access floating-point arguments that do not fit in the
22287 RSA from the stack when a subroutine is compiled without
22288 optimization. Because always storing floating-point arguments on the
22289 stack is inefficient and rarely needed, this option is not enabled by
22290 default and only is necessary when calling subroutines compiled by IBM
22291 XL compilers without optimization.
22295 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
22296 application written to use message passing with special startup code to
22297 enable the application to run. The system must have PE installed in the
22298 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
22299 must be overridden with the @option{-specs=} option to specify the
22300 appropriate directory location. The Parallel Environment does not
22301 support threads, so the @option{-mpe} option and the @option{-pthread}
22302 option are incompatible.
22304 @item -malign-natural
22305 @itemx -malign-power
22306 @opindex malign-natural
22307 @opindex malign-power
22308 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
22309 @option{-malign-natural} overrides the ABI-defined alignment of larger
22310 types, such as floating-point doubles, on their natural size-based boundary.
22311 The option @option{-malign-power} instructs GCC to follow the ABI-specified
22312 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
22314 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
22318 @itemx -mhard-float
22319 @opindex msoft-float
22320 @opindex mhard-float
22321 Generate code that does not use (uses) the floating-point register set.
22322 Software floating-point emulation is provided if you use the
22323 @option{-msoft-float} option, and pass the option to GCC when linking.
22325 @item -msingle-float
22326 @itemx -mdouble-float
22327 @opindex msingle-float
22328 @opindex mdouble-float
22329 Generate code for single- or double-precision floating-point operations.
22330 @option{-mdouble-float} implies @option{-msingle-float}.
22333 @opindex msimple-fpu
22334 Do not generate @code{sqrt} and @code{div} instructions for hardware
22335 floating-point unit.
22337 @item -mfpu=@var{name}
22339 Specify type of floating-point unit. Valid values for @var{name} are
22340 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
22341 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
22342 @samp{sp_full} (equivalent to @option{-msingle-float}),
22343 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
22346 @opindex mxilinx-fpu
22347 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
22350 @itemx -mno-multiple
22352 @opindex mno-multiple
22353 Generate code that uses (does not use) the load multiple word
22354 instructions and the store multiple word instructions. These
22355 instructions are generated by default on POWER systems, and not
22356 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
22357 PowerPC systems, since those instructions do not work when the
22358 processor is in little-endian mode. The exceptions are PPC740 and
22359 PPC750 which permit these instructions in little-endian mode.
22364 @opindex mno-string
22365 Generate code that uses (does not use) the load string instructions
22366 and the store string word instructions to save multiple registers and
22367 do small block moves. These instructions are generated by default on
22368 POWER systems, and not generated on PowerPC systems. Do not use
22369 @option{-mstring} on little-endian PowerPC systems, since those
22370 instructions do not work when the processor is in little-endian mode.
22371 The exceptions are PPC740 and PPC750 which permit these instructions
22372 in little-endian mode.
22377 @opindex mno-update
22378 Generate code that uses (does not use) the load or store instructions
22379 that update the base register to the address of the calculated memory
22380 location. These instructions are generated by default. If you use
22381 @option{-mno-update}, there is a small window between the time that the
22382 stack pointer is updated and the address of the previous frame is
22383 stored, which means code that walks the stack frame across interrupts or
22384 signals may get corrupted data.
22386 @item -mavoid-indexed-addresses
22387 @itemx -mno-avoid-indexed-addresses
22388 @opindex mavoid-indexed-addresses
22389 @opindex mno-avoid-indexed-addresses
22390 Generate code that tries to avoid (not avoid) the use of indexed load
22391 or store instructions. These instructions can incur a performance
22392 penalty on Power6 processors in certain situations, such as when
22393 stepping through large arrays that cross a 16M boundary. This option
22394 is enabled by default when targeting Power6 and disabled otherwise.
22397 @itemx -mno-fused-madd
22398 @opindex mfused-madd
22399 @opindex mno-fused-madd
22400 Generate code that uses (does not use) the floating-point multiply and
22401 accumulate instructions. These instructions are generated by default
22402 if hardware floating point is used. The machine-dependent
22403 @option{-mfused-madd} option is now mapped to the machine-independent
22404 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22405 mapped to @option{-ffp-contract=off}.
22411 Generate code that uses (does not use) the half-word multiply and
22412 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
22413 These instructions are generated by default when targeting those
22420 Generate code that uses (does not use) the string-search @samp{dlmzb}
22421 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
22422 generated by default when targeting those processors.
22424 @item -mno-bit-align
22426 @opindex mno-bit-align
22427 @opindex mbit-align
22428 On System V.4 and embedded PowerPC systems do not (do) force structures
22429 and unions that contain bit-fields to be aligned to the base type of the
22432 For example, by default a structure containing nothing but 8
22433 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
22434 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
22435 the structure is aligned to a 1-byte boundary and is 1 byte in
22438 @item -mno-strict-align
22439 @itemx -mstrict-align
22440 @opindex mno-strict-align
22441 @opindex mstrict-align
22442 On System V.4 and embedded PowerPC systems do not (do) assume that
22443 unaligned memory references are handled by the system.
22445 @item -mrelocatable
22446 @itemx -mno-relocatable
22447 @opindex mrelocatable
22448 @opindex mno-relocatable
22449 Generate code that allows (does not allow) a static executable to be
22450 relocated to a different address at run time. A simple embedded
22451 PowerPC system loader should relocate the entire contents of
22452 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
22453 a table of 32-bit addresses generated by this option. For this to
22454 work, all objects linked together must be compiled with
22455 @option{-mrelocatable} or @option{-mrelocatable-lib}.
22456 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
22458 @item -mrelocatable-lib
22459 @itemx -mno-relocatable-lib
22460 @opindex mrelocatable-lib
22461 @opindex mno-relocatable-lib
22462 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
22463 @code{.fixup} section to allow static executables to be relocated at
22464 run time, but @option{-mrelocatable-lib} does not use the smaller stack
22465 alignment of @option{-mrelocatable}. Objects compiled with
22466 @option{-mrelocatable-lib} may be linked with objects compiled with
22467 any combination of the @option{-mrelocatable} options.
22473 On System V.4 and embedded PowerPC systems do not (do) assume that
22474 register 2 contains a pointer to a global area pointing to the addresses
22475 used in the program.
22478 @itemx -mlittle-endian
22480 @opindex mlittle-endian
22481 On System V.4 and embedded PowerPC systems compile code for the
22482 processor in little-endian mode. The @option{-mlittle-endian} option is
22483 the same as @option{-mlittle}.
22486 @itemx -mbig-endian
22488 @opindex mbig-endian
22489 On System V.4 and embedded PowerPC systems compile code for the
22490 processor in big-endian mode. The @option{-mbig-endian} option is
22491 the same as @option{-mbig}.
22493 @item -mdynamic-no-pic
22494 @opindex mdynamic-no-pic
22495 On Darwin and Mac OS X systems, compile code so that it is not
22496 relocatable, but that its external references are relocatable. The
22497 resulting code is suitable for applications, but not shared
22500 @item -msingle-pic-base
22501 @opindex msingle-pic-base
22502 Treat the register used for PIC addressing as read-only, rather than
22503 loading it in the prologue for each function. The runtime system is
22504 responsible for initializing this register with an appropriate value
22505 before execution begins.
22507 @item -mprioritize-restricted-insns=@var{priority}
22508 @opindex mprioritize-restricted-insns
22509 This option controls the priority that is assigned to
22510 dispatch-slot restricted instructions during the second scheduling
22511 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
22512 or @samp{2} to assign no, highest, or second-highest (respectively)
22513 priority to dispatch-slot restricted
22516 @item -msched-costly-dep=@var{dependence_type}
22517 @opindex msched-costly-dep
22518 This option controls which dependences are considered costly
22519 by the target during instruction scheduling. The argument
22520 @var{dependence_type} takes one of the following values:
22524 No dependence is costly.
22527 All dependences are costly.
22529 @item @samp{true_store_to_load}
22530 A true dependence from store to load is costly.
22532 @item @samp{store_to_load}
22533 Any dependence from store to load is costly.
22536 Any dependence for which the latency is greater than or equal to
22537 @var{number} is costly.
22540 @item -minsert-sched-nops=@var{scheme}
22541 @opindex minsert-sched-nops
22542 This option controls which NOP insertion scheme is used during
22543 the second scheduling pass. The argument @var{scheme} takes one of the
22551 Pad with NOPs any dispatch group that has vacant issue slots,
22552 according to the scheduler's grouping.
22554 @item @samp{regroup_exact}
22555 Insert NOPs to force costly dependent insns into
22556 separate groups. Insert exactly as many NOPs as needed to force an insn
22557 to a new group, according to the estimated processor grouping.
22560 Insert NOPs to force costly dependent insns into
22561 separate groups. Insert @var{number} NOPs to force an insn to a new group.
22565 @opindex mcall-sysv
22566 On System V.4 and embedded PowerPC systems compile code using calling
22567 conventions that adhere to the March 1995 draft of the System V
22568 Application Binary Interface, PowerPC processor supplement. This is the
22569 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
22571 @item -mcall-sysv-eabi
22573 @opindex mcall-sysv-eabi
22574 @opindex mcall-eabi
22575 Specify both @option{-mcall-sysv} and @option{-meabi} options.
22577 @item -mcall-sysv-noeabi
22578 @opindex mcall-sysv-noeabi
22579 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
22581 @item -mcall-aixdesc
22583 On System V.4 and embedded PowerPC systems compile code for the AIX
22587 @opindex mcall-linux
22588 On System V.4 and embedded PowerPC systems compile code for the
22589 Linux-based GNU system.
22591 @item -mcall-freebsd
22592 @opindex mcall-freebsd
22593 On System V.4 and embedded PowerPC systems compile code for the
22594 FreeBSD operating system.
22596 @item -mcall-netbsd
22597 @opindex mcall-netbsd
22598 On System V.4 and embedded PowerPC systems compile code for the
22599 NetBSD operating system.
22601 @item -mcall-openbsd
22602 @opindex mcall-netbsd
22603 On System V.4 and embedded PowerPC systems compile code for the
22604 OpenBSD operating system.
22606 @item -maix-struct-return
22607 @opindex maix-struct-return
22608 Return all structures in memory (as specified by the AIX ABI)@.
22610 @item -msvr4-struct-return
22611 @opindex msvr4-struct-return
22612 Return structures smaller than 8 bytes in registers (as specified by the
22615 @item -mabi=@var{abi-type}
22617 Extend the current ABI with a particular extension, or remove such extension.
22618 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
22619 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
22620 @samp{elfv1}, @samp{elfv2}@.
22624 Extend the current ABI with SPE ABI extensions. This does not change
22625 the default ABI, instead it adds the SPE ABI extensions to the current
22629 @opindex mabi=no-spe
22630 Disable Book-E SPE ABI extensions for the current ABI@.
22632 @item -mabi=ibmlongdouble
22633 @opindex mabi=ibmlongdouble
22634 Change the current ABI to use IBM extended-precision long double.
22635 This is not likely to work if your system defaults to using IEEE
22636 extended-precision long double. If you change the long double type
22637 from IEEE extended-precision, the compiler will issue a warning unless
22638 you use the @option{-Wno-psabi} option.
22640 @item -mabi=ieeelongdouble
22641 @opindex mabi=ieeelongdouble
22642 Change the current ABI to use IEEE extended-precision long double.
22643 This is not likely to work if your system defaults to using IBM
22644 extended-precision long double. If you change the long double type
22645 from IBM extended-precision, the compiler will issue a warning unless
22646 you use the @option{-Wno-psabi} option.
22649 @opindex mabi=elfv1
22650 Change the current ABI to use the ELFv1 ABI.
22651 This is the default ABI for big-endian PowerPC 64-bit Linux.
22652 Overriding the default ABI requires special system support and is
22653 likely to fail in spectacular ways.
22656 @opindex mabi=elfv2
22657 Change the current ABI to use the ELFv2 ABI.
22658 This is the default ABI for little-endian PowerPC 64-bit Linux.
22659 Overriding the default ABI requires special system support and is
22660 likely to fail in spectacular ways.
22662 @item -mgnu-attribute
22663 @itemx -mno-gnu-attribute
22664 @opindex mgnu-attribute
22665 @opindex mno-gnu-attribute
22666 Emit .gnu_attribute assembly directives to set tag/value pairs in a
22667 .gnu.attributes section that specify ABI variations in function
22668 parameters or return values.
22671 @itemx -mno-prototype
22672 @opindex mprototype
22673 @opindex mno-prototype
22674 On System V.4 and embedded PowerPC systems assume that all calls to
22675 variable argument functions are properly prototyped. Otherwise, the
22676 compiler must insert an instruction before every non-prototyped call to
22677 set or clear bit 6 of the condition code register (@code{CR}) to
22678 indicate whether floating-point values are passed in the floating-point
22679 registers in case the function takes variable arguments. With
22680 @option{-mprototype}, only calls to prototyped variable argument functions
22681 set or clear the bit.
22685 On embedded PowerPC systems, assume that the startup module is called
22686 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
22687 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
22692 On embedded PowerPC systems, assume that the startup module is called
22693 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
22698 On embedded PowerPC systems, assume that the startup module is called
22699 @file{crt0.o} and the standard C libraries are @file{libads.a} and
22702 @item -myellowknife
22703 @opindex myellowknife
22704 On embedded PowerPC systems, assume that the startup module is called
22705 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
22710 On System V.4 and embedded PowerPC systems, specify that you are
22711 compiling for a VxWorks system.
22715 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
22716 header to indicate that @samp{eabi} extended relocations are used.
22722 On System V.4 and embedded PowerPC systems do (do not) adhere to the
22723 Embedded Applications Binary Interface (EABI), which is a set of
22724 modifications to the System V.4 specifications. Selecting @option{-meabi}
22725 means that the stack is aligned to an 8-byte boundary, a function
22726 @code{__eabi} is called from @code{main} to set up the EABI
22727 environment, and the @option{-msdata} option can use both @code{r2} and
22728 @code{r13} to point to two separate small data areas. Selecting
22729 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
22730 no EABI initialization function is called from @code{main}, and the
22731 @option{-msdata} option only uses @code{r13} to point to a single
22732 small data area. The @option{-meabi} option is on by default if you
22733 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
22736 @opindex msdata=eabi
22737 On System V.4 and embedded PowerPC systems, put small initialized
22738 @code{const} global and static data in the @code{.sdata2} section, which
22739 is pointed to by register @code{r2}. Put small initialized
22740 non-@code{const} global and static data in the @code{.sdata} section,
22741 which is pointed to by register @code{r13}. Put small uninitialized
22742 global and static data in the @code{.sbss} section, which is adjacent to
22743 the @code{.sdata} section. The @option{-msdata=eabi} option is
22744 incompatible with the @option{-mrelocatable} option. The
22745 @option{-msdata=eabi} option also sets the @option{-memb} option.
22748 @opindex msdata=sysv
22749 On System V.4 and embedded PowerPC systems, put small global and static
22750 data in the @code{.sdata} section, which is pointed to by register
22751 @code{r13}. Put small uninitialized global and static data in the
22752 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
22753 The @option{-msdata=sysv} option is incompatible with the
22754 @option{-mrelocatable} option.
22756 @item -msdata=default
22758 @opindex msdata=default
22760 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
22761 compile code the same as @option{-msdata=eabi}, otherwise compile code the
22762 same as @option{-msdata=sysv}.
22765 @opindex msdata=data
22766 On System V.4 and embedded PowerPC systems, put small global
22767 data in the @code{.sdata} section. Put small uninitialized global
22768 data in the @code{.sbss} section. Do not use register @code{r13}
22769 to address small data however. This is the default behavior unless
22770 other @option{-msdata} options are used.
22774 @opindex msdata=none
22776 On embedded PowerPC systems, put all initialized global and static data
22777 in the @code{.data} section, and all uninitialized data in the
22778 @code{.bss} section.
22780 @item -mblock-move-inline-limit=@var{num}
22781 @opindex mblock-move-inline-limit
22782 Inline all block moves (such as calls to @code{memcpy} or structure
22783 copies) less than or equal to @var{num} bytes. The minimum value for
22784 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
22785 targets. The default value is target-specific.
22789 @cindex smaller data references (PowerPC)
22790 @cindex .sdata/.sdata2 references (PowerPC)
22791 On embedded PowerPC systems, put global and static items less than or
22792 equal to @var{num} bytes into the small data or BSS sections instead of
22793 the normal data or BSS section. By default, @var{num} is 8. The
22794 @option{-G @var{num}} switch is also passed to the linker.
22795 All modules should be compiled with the same @option{-G @var{num}} value.
22798 @itemx -mno-regnames
22800 @opindex mno-regnames
22801 On System V.4 and embedded PowerPC systems do (do not) emit register
22802 names in the assembly language output using symbolic forms.
22805 @itemx -mno-longcall
22807 @opindex mno-longcall
22808 By default assume that all calls are far away so that a longer and more
22809 expensive calling sequence is required. This is required for calls
22810 farther than 32 megabytes (33,554,432 bytes) from the current location.
22811 A short call is generated if the compiler knows
22812 the call cannot be that far away. This setting can be overridden by
22813 the @code{shortcall} function attribute, or by @code{#pragma
22816 Some linkers are capable of detecting out-of-range calls and generating
22817 glue code on the fly. On these systems, long calls are unnecessary and
22818 generate slower code. As of this writing, the AIX linker can do this,
22819 as can the GNU linker for PowerPC/64. It is planned to add this feature
22820 to the GNU linker for 32-bit PowerPC systems as well.
22822 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
22823 callee, L42}, plus a @dfn{branch island} (glue code). The two target
22824 addresses represent the callee and the branch island. The
22825 Darwin/PPC linker prefers the first address and generates a @code{bl
22826 callee} if the PPC @code{bl} instruction reaches the callee directly;
22827 otherwise, the linker generates @code{bl L42} to call the branch
22828 island. The branch island is appended to the body of the
22829 calling function; it computes the full 32-bit address of the callee
22832 On Mach-O (Darwin) systems, this option directs the compiler emit to
22833 the glue for every direct call, and the Darwin linker decides whether
22834 to use or discard it.
22836 In the future, GCC may ignore all longcall specifications
22837 when the linker is known to generate glue.
22839 @item -mtls-markers
22840 @itemx -mno-tls-markers
22841 @opindex mtls-markers
22842 @opindex mno-tls-markers
22843 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22844 specifying the function argument. The relocation allows the linker to
22845 reliably associate function call with argument setup instructions for
22846 TLS optimization, which in turn allows GCC to better schedule the
22852 This option enables use of the reciprocal estimate and
22853 reciprocal square root estimate instructions with additional
22854 Newton-Raphson steps to increase precision instead of doing a divide or
22855 square root and divide for floating-point arguments. You should use
22856 the @option{-ffast-math} option when using @option{-mrecip} (or at
22857 least @option{-funsafe-math-optimizations},
22858 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22859 @option{-fno-trapping-math}). Note that while the throughput of the
22860 sequence is generally higher than the throughput of the non-reciprocal
22861 instruction, the precision of the sequence can be decreased by up to 2
22862 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
22865 @item -mrecip=@var{opt}
22866 @opindex mrecip=opt
22867 This option controls which reciprocal estimate instructions
22868 may be used. @var{opt} is a comma-separated list of options, which may
22869 be preceded by a @code{!} to invert the option:
22874 Enable all estimate instructions.
22877 Enable the default instructions, equivalent to @option{-mrecip}.
22880 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22883 Enable the reciprocal approximation instructions for both
22884 single and double precision.
22887 Enable the single-precision reciprocal approximation instructions.
22890 Enable the double-precision reciprocal approximation instructions.
22893 Enable the reciprocal square root approximation instructions for both
22894 single and double precision.
22897 Enable the single-precision reciprocal square root approximation instructions.
22900 Enable the double-precision reciprocal square root approximation instructions.
22904 So, for example, @option{-mrecip=all,!rsqrtd} enables
22905 all of the reciprocal estimate instructions, except for the
22906 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
22907 which handle the double-precision reciprocal square root calculations.
22909 @item -mrecip-precision
22910 @itemx -mno-recip-precision
22911 @opindex mrecip-precision
22912 Assume (do not assume) that the reciprocal estimate instructions
22913 provide higher-precision estimates than is mandated by the PowerPC
22914 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
22915 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
22916 The double-precision square root estimate instructions are not generated by
22917 default on low-precision machines, since they do not provide an
22918 estimate that converges after three steps.
22920 @item -mveclibabi=@var{type}
22921 @opindex mveclibabi
22922 Specifies the ABI type to use for vectorizing intrinsics using an
22923 external library. The only type supported at present is @samp{mass},
22924 which specifies to use IBM's Mathematical Acceleration Subsystem
22925 (MASS) libraries for vectorizing intrinsics using external libraries.
22926 GCC currently emits calls to @code{acosd2}, @code{acosf4},
22927 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
22928 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
22929 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
22930 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
22931 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
22932 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
22933 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
22934 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
22935 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
22936 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
22937 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
22938 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
22939 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
22940 for power7. Both @option{-ftree-vectorize} and
22941 @option{-funsafe-math-optimizations} must also be enabled. The MASS
22942 libraries must be specified at link time.
22947 Generate (do not generate) the @code{friz} instruction when the
22948 @option{-funsafe-math-optimizations} option is used to optimize
22949 rounding of floating-point values to 64-bit integer and back to floating
22950 point. The @code{friz} instruction does not return the same value if
22951 the floating-point number is too large to fit in an integer.
22953 @item -mpointers-to-nested-functions
22954 @itemx -mno-pointers-to-nested-functions
22955 @opindex mpointers-to-nested-functions
22956 Generate (do not generate) code to load up the static chain register
22957 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
22958 systems where a function pointer points to a 3-word descriptor giving
22959 the function address, TOC value to be loaded in register @code{r2}, and
22960 static chain value to be loaded in register @code{r11}. The
22961 @option{-mpointers-to-nested-functions} is on by default. You cannot
22962 call through pointers to nested functions or pointers
22963 to functions compiled in other languages that use the static chain if
22964 you use @option{-mno-pointers-to-nested-functions}.
22966 @item -msave-toc-indirect
22967 @itemx -mno-save-toc-indirect
22968 @opindex msave-toc-indirect
22969 Generate (do not generate) code to save the TOC value in the reserved
22970 stack location in the function prologue if the function calls through
22971 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
22972 saved in the prologue, it is saved just before the call through the
22973 pointer. The @option{-mno-save-toc-indirect} option is the default.
22975 @item -mcompat-align-parm
22976 @itemx -mno-compat-align-parm
22977 @opindex mcompat-align-parm
22978 Generate (do not generate) code to pass structure parameters with a
22979 maximum alignment of 64 bits, for compatibility with older versions
22982 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
22983 structure parameter on a 128-bit boundary when that structure contained
22984 a member requiring 128-bit alignment. This is corrected in more
22985 recent versions of GCC. This option may be used to generate code
22986 that is compatible with functions compiled with older versions of
22989 The @option{-mno-compat-align-parm} option is the default.
22991 @item -mstack-protector-guard=@var{guard}
22992 @itemx -mstack-protector-guard-reg=@var{reg}
22993 @itemx -mstack-protector-guard-offset=@var{offset}
22994 @itemx -mstack-protector-guard-symbol=@var{symbol}
22995 @opindex mstack-protector-guard
22996 @opindex mstack-protector-guard-reg
22997 @opindex mstack-protector-guard-offset
22998 @opindex mstack-protector-guard-symbol
22999 Generate stack protection code using canary at @var{guard}. Supported
23000 locations are @samp{global} for global canary or @samp{tls} for per-thread
23001 canary in the TLS block (the default with GNU libc version 2.4 or later).
23003 With the latter choice the options
23004 @option{-mstack-protector-guard-reg=@var{reg}} and
23005 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
23006 which register to use as base register for reading the canary, and from what
23007 offset from that base register. The default for those is as specified in the
23008 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
23009 the offset with a symbol reference to a canary in the TLS block.
23013 @subsection RX Options
23016 These command-line options are defined for RX targets:
23019 @item -m64bit-doubles
23020 @itemx -m32bit-doubles
23021 @opindex m64bit-doubles
23022 @opindex m32bit-doubles
23023 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23024 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23025 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
23026 works on 32-bit values, which is why the default is
23027 @option{-m32bit-doubles}.
23033 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
23034 floating-point hardware. The default is enabled for the RX600
23035 series and disabled for the RX200 series.
23037 Floating-point instructions are only generated for 32-bit floating-point
23038 values, however, so the FPU hardware is not used for doubles if the
23039 @option{-m64bit-doubles} option is used.
23041 @emph{Note} If the @option{-fpu} option is enabled then
23042 @option{-funsafe-math-optimizations} is also enabled automatically.
23043 This is because the RX FPU instructions are themselves unsafe.
23045 @item -mcpu=@var{name}
23047 Selects the type of RX CPU to be targeted. Currently three types are
23048 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
23049 the specific @samp{RX610} CPU. The default is @samp{RX600}.
23051 The only difference between @samp{RX600} and @samp{RX610} is that the
23052 @samp{RX610} does not support the @code{MVTIPL} instruction.
23054 The @samp{RX200} series does not have a hardware floating-point unit
23055 and so @option{-nofpu} is enabled by default when this type is
23058 @item -mbig-endian-data
23059 @itemx -mlittle-endian-data
23060 @opindex mbig-endian-data
23061 @opindex mlittle-endian-data
23062 Store data (but not code) in the big-endian format. The default is
23063 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
23066 @item -msmall-data-limit=@var{N}
23067 @opindex msmall-data-limit
23068 Specifies the maximum size in bytes of global and static variables
23069 which can be placed into the small data area. Using the small data
23070 area can lead to smaller and faster code, but the size of area is
23071 limited and it is up to the programmer to ensure that the area does
23072 not overflow. Also when the small data area is used one of the RX's
23073 registers (usually @code{r13}) is reserved for use pointing to this
23074 area, so it is no longer available for use by the compiler. This
23075 could result in slower and/or larger code if variables are pushed onto
23076 the stack instead of being held in this register.
23078 Note, common variables (variables that have not been initialized) and
23079 constants are not placed into the small data area as they are assigned
23080 to other sections in the output executable.
23082 The default value is zero, which disables this feature. Note, this
23083 feature is not enabled by default with higher optimization levels
23084 (@option{-O2} etc) because of the potentially detrimental effects of
23085 reserving a register. It is up to the programmer to experiment and
23086 discover whether this feature is of benefit to their program. See the
23087 description of the @option{-mpid} option for a description of how the
23088 actual register to hold the small data area pointer is chosen.
23094 Use the simulator runtime. The default is to use the libgloss
23095 board-specific runtime.
23097 @item -mas100-syntax
23098 @itemx -mno-as100-syntax
23099 @opindex mas100-syntax
23100 @opindex mno-as100-syntax
23101 When generating assembler output use a syntax that is compatible with
23102 Renesas's AS100 assembler. This syntax can also be handled by the GAS
23103 assembler, but it has some restrictions so it is not generated by default.
23105 @item -mmax-constant-size=@var{N}
23106 @opindex mmax-constant-size
23107 Specifies the maximum size, in bytes, of a constant that can be used as
23108 an operand in a RX instruction. Although the RX instruction set does
23109 allow constants of up to 4 bytes in length to be used in instructions,
23110 a longer value equates to a longer instruction. Thus in some
23111 circumstances it can be beneficial to restrict the size of constants
23112 that are used in instructions. Constants that are too big are instead
23113 placed into a constant pool and referenced via register indirection.
23115 The value @var{N} can be between 0 and 4. A value of 0 (the default)
23116 or 4 means that constants of any size are allowed.
23120 Enable linker relaxation. Linker relaxation is a process whereby the
23121 linker attempts to reduce the size of a program by finding shorter
23122 versions of various instructions. Disabled by default.
23124 @item -mint-register=@var{N}
23125 @opindex mint-register
23126 Specify the number of registers to reserve for fast interrupt handler
23127 functions. The value @var{N} can be between 0 and 4. A value of 1
23128 means that register @code{r13} is reserved for the exclusive use
23129 of fast interrupt handlers. A value of 2 reserves @code{r13} and
23130 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
23131 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
23132 A value of 0, the default, does not reserve any registers.
23134 @item -msave-acc-in-interrupts
23135 @opindex msave-acc-in-interrupts
23136 Specifies that interrupt handler functions should preserve the
23137 accumulator register. This is only necessary if normal code might use
23138 the accumulator register, for example because it performs 64-bit
23139 multiplications. The default is to ignore the accumulator as this
23140 makes the interrupt handlers faster.
23146 Enables the generation of position independent data. When enabled any
23147 access to constant data is done via an offset from a base address
23148 held in a register. This allows the location of constant data to be
23149 determined at run time without requiring the executable to be
23150 relocated, which is a benefit to embedded applications with tight
23151 memory constraints. Data that can be modified is not affected by this
23154 Note, using this feature reserves a register, usually @code{r13}, for
23155 the constant data base address. This can result in slower and/or
23156 larger code, especially in complicated functions.
23158 The actual register chosen to hold the constant data base address
23159 depends upon whether the @option{-msmall-data-limit} and/or the
23160 @option{-mint-register} command-line options are enabled. Starting
23161 with register @code{r13} and proceeding downwards, registers are
23162 allocated first to satisfy the requirements of @option{-mint-register},
23163 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
23164 is possible for the small data area register to be @code{r8} if both
23165 @option{-mint-register=4} and @option{-mpid} are specified on the
23168 By default this feature is not enabled. The default can be restored
23169 via the @option{-mno-pid} command-line option.
23171 @item -mno-warn-multiple-fast-interrupts
23172 @itemx -mwarn-multiple-fast-interrupts
23173 @opindex mno-warn-multiple-fast-interrupts
23174 @opindex mwarn-multiple-fast-interrupts
23175 Prevents GCC from issuing a warning message if it finds more than one
23176 fast interrupt handler when it is compiling a file. The default is to
23177 issue a warning for each extra fast interrupt handler found, as the RX
23178 only supports one such interrupt.
23180 @item -mallow-string-insns
23181 @itemx -mno-allow-string-insns
23182 @opindex mallow-string-insns
23183 @opindex mno-allow-string-insns
23184 Enables or disables the use of the string manipulation instructions
23185 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
23186 @code{SWHILE} and also the @code{RMPA} instruction. These
23187 instructions may prefetch data, which is not safe to do if accessing
23188 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
23189 for more information).
23191 The default is to allow these instructions, but it is not possible for
23192 GCC to reliably detect all circumstances where a string instruction
23193 might be used to access an I/O register, so their use cannot be
23194 disabled automatically. Instead it is reliant upon the programmer to
23195 use the @option{-mno-allow-string-insns} option if their program
23196 accesses I/O space.
23198 When the instructions are enabled GCC defines the C preprocessor
23199 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
23200 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
23206 Use only (or not only) @code{JSR} instructions to access functions.
23207 This option can be used when code size exceeds the range of @code{BSR}
23208 instructions. Note that @option{-mno-jsr} does not mean to not use
23209 @code{JSR} but instead means that any type of branch may be used.
23212 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
23213 has special significance to the RX port when used with the
23214 @code{interrupt} function attribute. This attribute indicates a
23215 function intended to process fast interrupts. GCC ensures
23216 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
23217 and/or @code{r13} and only provided that the normal use of the
23218 corresponding registers have been restricted via the
23219 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
23222 @node S/390 and zSeries Options
23223 @subsection S/390 and zSeries Options
23224 @cindex S/390 and zSeries Options
23226 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
23230 @itemx -msoft-float
23231 @opindex mhard-float
23232 @opindex msoft-float
23233 Use (do not use) the hardware floating-point instructions and registers
23234 for floating-point operations. When @option{-msoft-float} is specified,
23235 functions in @file{libgcc.a} are used to perform floating-point
23236 operations. When @option{-mhard-float} is specified, the compiler
23237 generates IEEE floating-point instructions. This is the default.
23240 @itemx -mno-hard-dfp
23242 @opindex mno-hard-dfp
23243 Use (do not use) the hardware decimal-floating-point instructions for
23244 decimal-floating-point operations. When @option{-mno-hard-dfp} is
23245 specified, functions in @file{libgcc.a} are used to perform
23246 decimal-floating-point operations. When @option{-mhard-dfp} is
23247 specified, the compiler generates decimal-floating-point hardware
23248 instructions. This is the default for @option{-march=z9-ec} or higher.
23250 @item -mlong-double-64
23251 @itemx -mlong-double-128
23252 @opindex mlong-double-64
23253 @opindex mlong-double-128
23254 These switches control the size of @code{long double} type. A size
23255 of 64 bits makes the @code{long double} type equivalent to the @code{double}
23256 type. This is the default.
23259 @itemx -mno-backchain
23260 @opindex mbackchain
23261 @opindex mno-backchain
23262 Store (do not store) the address of the caller's frame as backchain pointer
23263 into the callee's stack frame.
23264 A backchain may be needed to allow debugging using tools that do not understand
23265 DWARF call frame information.
23266 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
23267 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
23268 the backchain is placed into the topmost word of the 96/160 byte register
23271 In general, code compiled with @option{-mbackchain} is call-compatible with
23272 code compiled with @option{-mmo-backchain}; however, use of the backchain
23273 for debugging purposes usually requires that the whole binary is built with
23274 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
23275 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23276 to build a linux kernel use @option{-msoft-float}.
23278 The default is to not maintain the backchain.
23280 @item -mpacked-stack
23281 @itemx -mno-packed-stack
23282 @opindex mpacked-stack
23283 @opindex mno-packed-stack
23284 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
23285 specified, the compiler uses the all fields of the 96/160 byte register save
23286 area only for their default purpose; unused fields still take up stack space.
23287 When @option{-mpacked-stack} is specified, register save slots are densely
23288 packed at the top of the register save area; unused space is reused for other
23289 purposes, allowing for more efficient use of the available stack space.
23290 However, when @option{-mbackchain} is also in effect, the topmost word of
23291 the save area is always used to store the backchain, and the return address
23292 register is always saved two words below the backchain.
23294 As long as the stack frame backchain is not used, code generated with
23295 @option{-mpacked-stack} is call-compatible with code generated with
23296 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
23297 S/390 or zSeries generated code that uses the stack frame backchain at run
23298 time, not just for debugging purposes. Such code is not call-compatible
23299 with code compiled with @option{-mpacked-stack}. Also, note that the
23300 combination of @option{-mbackchain},
23301 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23302 to build a linux kernel use @option{-msoft-float}.
23304 The default is to not use the packed stack layout.
23307 @itemx -mno-small-exec
23308 @opindex msmall-exec
23309 @opindex mno-small-exec
23310 Generate (or do not generate) code using the @code{bras} instruction
23311 to do subroutine calls.
23312 This only works reliably if the total executable size does not
23313 exceed 64k. The default is to use the @code{basr} instruction instead,
23314 which does not have this limitation.
23320 When @option{-m31} is specified, generate code compliant to the
23321 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
23322 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
23323 particular to generate 64-bit instructions. For the @samp{s390}
23324 targets, the default is @option{-m31}, while the @samp{s390x}
23325 targets default to @option{-m64}.
23331 When @option{-mzarch} is specified, generate code using the
23332 instructions available on z/Architecture.
23333 When @option{-mesa} is specified, generate code using the
23334 instructions available on ESA/390. Note that @option{-mesa} is
23335 not possible with @option{-m64}.
23336 When generating code compliant to the GNU/Linux for S/390 ABI,
23337 the default is @option{-mesa}. When generating code compliant
23338 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
23344 The @option{-mhtm} option enables a set of builtins making use of
23345 instructions available with the transactional execution facility
23346 introduced with the IBM zEnterprise EC12 machine generation
23347 @ref{S/390 System z Built-in Functions}.
23348 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
23354 When @option{-mvx} is specified, generate code using the instructions
23355 available with the vector extension facility introduced with the IBM
23356 z13 machine generation.
23357 This option changes the ABI for some vector type values with regard to
23358 alignment and calling conventions. In case vector type values are
23359 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
23360 command will be added to mark the resulting binary with the ABI used.
23361 @option{-mvx} is enabled by default when using @option{-march=z13}.
23364 @itemx -mno-zvector
23366 @opindex mno-zvector
23367 The @option{-mzvector} option enables vector language extensions and
23368 builtins using instructions available with the vector extension
23369 facility introduced with the IBM z13 machine generation.
23370 This option adds support for @samp{vector} to be used as a keyword to
23371 define vector type variables and arguments. @samp{vector} is only
23372 available when GNU extensions are enabled. It will not be expanded
23373 when requesting strict standard compliance e.g. with @option{-std=c99}.
23374 In addition to the GCC low-level builtins @option{-mzvector} enables
23375 a set of builtins added for compatibility with AltiVec-style
23376 implementations like Power and Cell. In order to make use of these
23377 builtins the header file @file{vecintrin.h} needs to be included.
23378 @option{-mzvector} is disabled by default.
23384 Generate (or do not generate) code using the @code{mvcle} instruction
23385 to perform block moves. When @option{-mno-mvcle} is specified,
23386 use a @code{mvc} loop instead. This is the default unless optimizing for
23393 Print (or do not print) additional debug information when compiling.
23394 The default is to not print debug information.
23396 @item -march=@var{cpu-type}
23398 Generate code that runs on @var{cpu-type}, which is the name of a
23399 system representing a certain processor type. Possible values for
23400 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
23401 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
23402 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
23405 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
23406 @samp{g6} are deprecated and will be removed with future releases.
23408 Specifying @samp{native} as cpu type can be used to select the best
23409 architecture option for the host processor.
23410 @option{-march=native} has no effect if GCC does not recognize the
23413 @item -mtune=@var{cpu-type}
23415 Tune to @var{cpu-type} everything applicable about the generated code,
23416 except for the ABI and the set of available instructions.
23417 The list of @var{cpu-type} values is the same as for @option{-march}.
23418 The default is the value used for @option{-march}.
23421 @itemx -mno-tpf-trace
23422 @opindex mtpf-trace
23423 @opindex mno-tpf-trace
23424 Generate code that adds (does not add) in TPF OS specific branches to trace
23425 routines in the operating system. This option is off by default, even
23426 when compiling for the TPF OS@.
23429 @itemx -mno-fused-madd
23430 @opindex mfused-madd
23431 @opindex mno-fused-madd
23432 Generate code that uses (does not use) the floating-point multiply and
23433 accumulate instructions. These instructions are generated by default if
23434 hardware floating point is used.
23436 @item -mwarn-framesize=@var{framesize}
23437 @opindex mwarn-framesize
23438 Emit a warning if the current function exceeds the given frame size. Because
23439 this is a compile-time check it doesn't need to be a real problem when the program
23440 runs. It is intended to identify functions that most probably cause
23441 a stack overflow. It is useful to be used in an environment with limited stack
23442 size e.g.@: the linux kernel.
23444 @item -mwarn-dynamicstack
23445 @opindex mwarn-dynamicstack
23446 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
23447 arrays. This is generally a bad idea with a limited stack size.
23449 @item -mstack-guard=@var{stack-guard}
23450 @itemx -mstack-size=@var{stack-size}
23451 @opindex mstack-guard
23452 @opindex mstack-size
23453 If these options are provided the S/390 back end emits additional instructions in
23454 the function prologue that trigger a trap if the stack size is @var{stack-guard}
23455 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
23456 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
23457 the frame size of the compiled function is chosen.
23458 These options are intended to be used to help debugging stack overflow problems.
23459 The additionally emitted code causes only little overhead and hence can also be
23460 used in production-like systems without greater performance degradation. The given
23461 values have to be exact powers of 2 and @var{stack-size} has to be greater than
23462 @var{stack-guard} without exceeding 64k.
23463 In order to be efficient the extra code makes the assumption that the stack starts
23464 at an address aligned to the value given by @var{stack-size}.
23465 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
23467 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
23469 If the hotpatch option is enabled, a ``hot-patching'' function
23470 prologue is generated for all functions in the compilation unit.
23471 The funtion label is prepended with the given number of two-byte
23472 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
23473 the label, 2 * @var{post-halfwords} bytes are appended, using the
23474 largest NOP like instructions the architecture allows (maximum
23477 If both arguments are zero, hotpatching is disabled.
23479 This option can be overridden for individual functions with the
23480 @code{hotpatch} attribute.
23483 @node Score Options
23484 @subsection Score Options
23485 @cindex Score Options
23487 These options are defined for Score implementations:
23492 Compile code for big-endian mode. This is the default.
23496 Compile code for little-endian mode.
23500 Disable generation of @code{bcnz} instructions.
23504 Enable generation of unaligned load and store instructions.
23508 Enable the use of multiply-accumulate instructions. Disabled by default.
23512 Specify the SCORE5 as the target architecture.
23516 Specify the SCORE5U of the target architecture.
23520 Specify the SCORE7 as the target architecture. This is the default.
23524 Specify the SCORE7D as the target architecture.
23528 @subsection SH Options
23530 These @samp{-m} options are defined for the SH implementations:
23535 Generate code for the SH1.
23539 Generate code for the SH2.
23542 Generate code for the SH2e.
23546 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
23547 that the floating-point unit is not used.
23549 @item -m2a-single-only
23550 @opindex m2a-single-only
23551 Generate code for the SH2a-FPU, in such a way that no double-precision
23552 floating-point operations are used.
23555 @opindex m2a-single
23556 Generate code for the SH2a-FPU assuming the floating-point unit is in
23557 single-precision mode by default.
23561 Generate code for the SH2a-FPU assuming the floating-point unit is in
23562 double-precision mode by default.
23566 Generate code for the SH3.
23570 Generate code for the SH3e.
23574 Generate code for the SH4 without a floating-point unit.
23576 @item -m4-single-only
23577 @opindex m4-single-only
23578 Generate code for the SH4 with a floating-point unit that only
23579 supports single-precision arithmetic.
23583 Generate code for the SH4 assuming the floating-point unit is in
23584 single-precision mode by default.
23588 Generate code for the SH4.
23592 Generate code for SH4-100.
23594 @item -m4-100-nofpu
23595 @opindex m4-100-nofpu
23596 Generate code for SH4-100 in such a way that the
23597 floating-point unit is not used.
23599 @item -m4-100-single
23600 @opindex m4-100-single
23601 Generate code for SH4-100 assuming the floating-point unit is in
23602 single-precision mode by default.
23604 @item -m4-100-single-only
23605 @opindex m4-100-single-only
23606 Generate code for SH4-100 in such a way that no double-precision
23607 floating-point operations are used.
23611 Generate code for SH4-200.
23613 @item -m4-200-nofpu
23614 @opindex m4-200-nofpu
23615 Generate code for SH4-200 without in such a way that the
23616 floating-point unit is not used.
23618 @item -m4-200-single
23619 @opindex m4-200-single
23620 Generate code for SH4-200 assuming the floating-point unit is in
23621 single-precision mode by default.
23623 @item -m4-200-single-only
23624 @opindex m4-200-single-only
23625 Generate code for SH4-200 in such a way that no double-precision
23626 floating-point operations are used.
23630 Generate code for SH4-300.
23632 @item -m4-300-nofpu
23633 @opindex m4-300-nofpu
23634 Generate code for SH4-300 without in such a way that the
23635 floating-point unit is not used.
23637 @item -m4-300-single
23638 @opindex m4-300-single
23639 Generate code for SH4-300 in such a way that no double-precision
23640 floating-point operations are used.
23642 @item -m4-300-single-only
23643 @opindex m4-300-single-only
23644 Generate code for SH4-300 in such a way that no double-precision
23645 floating-point operations are used.
23649 Generate code for SH4-340 (no MMU, no FPU).
23653 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
23658 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
23659 floating-point unit is not used.
23661 @item -m4a-single-only
23662 @opindex m4a-single-only
23663 Generate code for the SH4a, in such a way that no double-precision
23664 floating-point operations are used.
23667 @opindex m4a-single
23668 Generate code for the SH4a assuming the floating-point unit is in
23669 single-precision mode by default.
23673 Generate code for the SH4a.
23677 Same as @option{-m4a-nofpu}, except that it implicitly passes
23678 @option{-dsp} to the assembler. GCC doesn't generate any DSP
23679 instructions at the moment.
23683 Compile code for the processor in big-endian mode.
23687 Compile code for the processor in little-endian mode.
23691 Align doubles at 64-bit boundaries. Note that this changes the calling
23692 conventions, and thus some functions from the standard C library do
23693 not work unless you recompile it first with @option{-mdalign}.
23697 Shorten some address references at link time, when possible; uses the
23698 linker option @option{-relax}.
23702 Use 32-bit offsets in @code{switch} tables. The default is to use
23707 Enable the use of bit manipulation instructions on SH2A.
23711 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
23712 alignment constraints.
23716 Comply with the calling conventions defined by Renesas.
23719 @opindex mno-renesas
23720 Comply with the calling conventions defined for GCC before the Renesas
23721 conventions were available. This option is the default for all
23722 targets of the SH toolchain.
23725 @opindex mnomacsave
23726 Mark the @code{MAC} register as call-clobbered, even if
23727 @option{-mrenesas} is given.
23733 Control the IEEE compliance of floating-point comparisons, which affects the
23734 handling of cases where the result of a comparison is unordered. By default
23735 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
23736 enabled @option{-mno-ieee} is implicitly set, which results in faster
23737 floating-point greater-equal and less-equal comparisons. The implicit settings
23738 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
23740 @item -minline-ic_invalidate
23741 @opindex minline-ic_invalidate
23742 Inline code to invalidate instruction cache entries after setting up
23743 nested function trampolines.
23744 This option has no effect if @option{-musermode} is in effect and the selected
23745 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
23747 If the selected code generation option does not allow the use of the @code{icbi}
23748 instruction, and @option{-musermode} is not in effect, the inlined code
23749 manipulates the instruction cache address array directly with an associative
23750 write. This not only requires privileged mode at run time, but it also
23751 fails if the cache line had been mapped via the TLB and has become unmapped.
23755 Dump instruction size and location in the assembly code.
23758 @opindex mpadstruct
23759 This option is deprecated. It pads structures to multiple of 4 bytes,
23760 which is incompatible with the SH ABI@.
23762 @item -matomic-model=@var{model}
23763 @opindex matomic-model=@var{model}
23764 Sets the model of atomic operations and additional parameters as a comma
23765 separated list. For details on the atomic built-in functions see
23766 @ref{__atomic Builtins}. The following models and parameters are supported:
23771 Disable compiler generated atomic sequences and emit library calls for atomic
23772 operations. This is the default if the target is not @code{sh*-*-linux*}.
23775 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
23776 built-in functions. The generated atomic sequences require additional support
23777 from the interrupt/exception handling code of the system and are only suitable
23778 for SH3* and SH4* single-core systems. This option is enabled by default when
23779 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
23780 this option also partially utilizes the hardware atomic instructions
23781 @code{movli.l} and @code{movco.l} to create more efficient code, unless
23782 @samp{strict} is specified.
23785 Generate software atomic sequences that use a variable in the thread control
23786 block. This is a variation of the gUSA sequences which can also be used on
23787 SH1* and SH2* targets. The generated atomic sequences require additional
23788 support from the interrupt/exception handling code of the system and are only
23789 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
23790 parameter has to be specified as well.
23793 Generate software atomic sequences that temporarily disable interrupts by
23794 setting @code{SR.IMASK = 1111}. This model works only when the program runs
23795 in privileged mode and is only suitable for single-core systems. Additional
23796 support from the interrupt/exception handling code of the system is not
23797 required. This model is enabled by default when the target is
23798 @code{sh*-*-linux*} and SH1* or SH2*.
23801 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
23802 instructions only. This is only available on SH4A and is suitable for
23803 multi-core systems. Since the hardware instructions support only 32 bit atomic
23804 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
23805 Code compiled with this option is also compatible with other software
23806 atomic model interrupt/exception handling systems if executed on an SH4A
23807 system. Additional support from the interrupt/exception handling code of the
23808 system is not required for this model.
23811 This parameter specifies the offset in bytes of the variable in the thread
23812 control block structure that should be used by the generated atomic sequences
23813 when the @samp{soft-tcb} model has been selected. For other models this
23814 parameter is ignored. The specified value must be an integer multiple of four
23815 and in the range 0-1020.
23818 This parameter prevents mixed usage of multiple atomic models, even if they
23819 are compatible, and makes the compiler generate atomic sequences of the
23820 specified model only.
23826 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
23827 Notice that depending on the particular hardware and software configuration
23828 this can degrade overall performance due to the operand cache line flushes
23829 that are implied by the @code{tas.b} instruction. On multi-core SH4A
23830 processors the @code{tas.b} instruction must be used with caution since it
23831 can result in data corruption for certain cache configurations.
23834 @opindex mprefergot
23835 When generating position-independent code, emit function calls using
23836 the Global Offset Table instead of the Procedure Linkage Table.
23839 @itemx -mno-usermode
23841 @opindex mno-usermode
23842 Don't allow (allow) the compiler generating privileged mode code. Specifying
23843 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
23844 inlined code would not work in user mode. @option{-musermode} is the default
23845 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
23846 @option{-musermode} has no effect, since there is no user mode.
23848 @item -multcost=@var{number}
23849 @opindex multcost=@var{number}
23850 Set the cost to assume for a multiply insn.
23852 @item -mdiv=@var{strategy}
23853 @opindex mdiv=@var{strategy}
23854 Set the division strategy to be used for integer division operations.
23855 @var{strategy} can be one of:
23860 Calls a library function that uses the single-step division instruction
23861 @code{div1} to perform the operation. Division by zero calculates an
23862 unspecified result and does not trap. This is the default except for SH4,
23863 SH2A and SHcompact.
23866 Calls a library function that performs the operation in double precision
23867 floating point. Division by zero causes a floating-point exception. This is
23868 the default for SHcompact with FPU. Specifying this for targets that do not
23869 have a double precision FPU defaults to @code{call-div1}.
23872 Calls a library function that uses a lookup table for small divisors and
23873 the @code{div1} instruction with case distinction for larger divisors. Division
23874 by zero calculates an unspecified result and does not trap. This is the default
23875 for SH4. Specifying this for targets that do not have dynamic shift
23876 instructions defaults to @code{call-div1}.
23880 When a division strategy has not been specified the default strategy is
23881 selected based on the current target. For SH2A the default strategy is to
23882 use the @code{divs} and @code{divu} instructions instead of library function
23885 @item -maccumulate-outgoing-args
23886 @opindex maccumulate-outgoing-args
23887 Reserve space once for outgoing arguments in the function prologue rather
23888 than around each call. Generally beneficial for performance and size. Also
23889 needed for unwinding to avoid changing the stack frame around conditional code.
23891 @item -mdivsi3_libfunc=@var{name}
23892 @opindex mdivsi3_libfunc=@var{name}
23893 Set the name of the library function used for 32-bit signed division to
23895 This only affects the name used in the @samp{call} division strategies, and
23896 the compiler still expects the same sets of input/output/clobbered registers as
23897 if this option were not present.
23899 @item -mfixed-range=@var{register-range}
23900 @opindex mfixed-range
23901 Generate code treating the given register range as fixed registers.
23902 A fixed register is one that the register allocator can not use. This is
23903 useful when compiling kernel code. A register range is specified as
23904 two registers separated by a dash. Multiple register ranges can be
23905 specified separated by a comma.
23907 @item -mbranch-cost=@var{num}
23908 @opindex mbranch-cost=@var{num}
23909 Assume @var{num} to be the cost for a branch instruction. Higher numbers
23910 make the compiler try to generate more branch-free code if possible.
23911 If not specified the value is selected depending on the processor type that
23912 is being compiled for.
23915 @itemx -mno-zdcbranch
23916 @opindex mzdcbranch
23917 @opindex mno-zdcbranch
23918 Assume (do not assume) that zero displacement conditional branch instructions
23919 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
23920 compiler prefers zero displacement branch code sequences. This is
23921 enabled by default when generating code for SH4 and SH4A. It can be explicitly
23922 disabled by specifying @option{-mno-zdcbranch}.
23924 @item -mcbranch-force-delay-slot
23925 @opindex mcbranch-force-delay-slot
23926 Force the usage of delay slots for conditional branches, which stuffs the delay
23927 slot with a @code{nop} if a suitable instruction cannot be found. By default
23928 this option is disabled. It can be enabled to work around hardware bugs as
23929 found in the original SH7055.
23932 @itemx -mno-fused-madd
23933 @opindex mfused-madd
23934 @opindex mno-fused-madd
23935 Generate code that uses (does not use) the floating-point multiply and
23936 accumulate instructions. These instructions are generated by default
23937 if hardware floating point is used. The machine-dependent
23938 @option{-mfused-madd} option is now mapped to the machine-independent
23939 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
23940 mapped to @option{-ffp-contract=off}.
23946 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
23947 and cosine approximations. The option @option{-mfsca} must be used in
23948 combination with @option{-funsafe-math-optimizations}. It is enabled by default
23949 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
23950 approximations even if @option{-funsafe-math-optimizations} is in effect.
23956 Allow or disallow the compiler to emit the @code{fsrra} instruction for
23957 reciprocal square root approximations. The option @option{-mfsrra} must be used
23958 in combination with @option{-funsafe-math-optimizations} and
23959 @option{-ffinite-math-only}. It is enabled by default when generating code for
23960 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
23961 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
23964 @item -mpretend-cmove
23965 @opindex mpretend-cmove
23966 Prefer zero-displacement conditional branches for conditional move instruction
23967 patterns. This can result in faster code on the SH4 processor.
23971 Generate code using the FDPIC ABI.
23975 @node Solaris 2 Options
23976 @subsection Solaris 2 Options
23977 @cindex Solaris 2 options
23979 These @samp{-m} options are supported on Solaris 2:
23982 @item -mclear-hwcap
23983 @opindex mclear-hwcap
23984 @option{-mclear-hwcap} tells the compiler to remove the hardware
23985 capabilities generated by the Solaris assembler. This is only necessary
23986 when object files use ISA extensions not supported by the current
23987 machine, but check at runtime whether or not to use them.
23989 @item -mimpure-text
23990 @opindex mimpure-text
23991 @option{-mimpure-text}, used in addition to @option{-shared}, tells
23992 the compiler to not pass @option{-z text} to the linker when linking a
23993 shared object. Using this option, you can link position-dependent
23994 code into a shared object.
23996 @option{-mimpure-text} suppresses the ``relocations remain against
23997 allocatable but non-writable sections'' linker error message.
23998 However, the necessary relocations trigger copy-on-write, and the
23999 shared object is not actually shared across processes. Instead of
24000 using @option{-mimpure-text}, you should compile all source code with
24001 @option{-fpic} or @option{-fPIC}.
24005 These switches are supported in addition to the above on Solaris 2:
24010 This is a synonym for @option{-pthread}.
24013 @node SPARC Options
24014 @subsection SPARC Options
24015 @cindex SPARC options
24017 These @samp{-m} options are supported on the SPARC:
24020 @item -mno-app-regs
24022 @opindex mno-app-regs
24024 Specify @option{-mapp-regs} to generate output using the global registers
24025 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
24026 global register 1, each global register 2 through 4 is then treated as an
24027 allocable register that is clobbered by function calls. This is the default.
24029 To be fully SVR4 ABI-compliant at the cost of some performance loss,
24030 specify @option{-mno-app-regs}. You should compile libraries and system
24031 software with this option.
24037 With @option{-mflat}, the compiler does not generate save/restore instructions
24038 and uses a ``flat'' or single register window model. This model is compatible
24039 with the regular register window model. The local registers and the input
24040 registers (0--5) are still treated as ``call-saved'' registers and are
24041 saved on the stack as needed.
24043 With @option{-mno-flat} (the default), the compiler generates save/restore
24044 instructions (except for leaf functions). This is the normal operating mode.
24047 @itemx -mhard-float
24049 @opindex mhard-float
24050 Generate output containing floating-point instructions. This is the
24054 @itemx -msoft-float
24056 @opindex msoft-float
24057 Generate output containing library calls for floating point.
24058 @strong{Warning:} the requisite libraries are not available for all SPARC
24059 targets. Normally the facilities of the machine's usual C compiler are
24060 used, but this cannot be done directly in cross-compilation. You must make
24061 your own arrangements to provide suitable library functions for
24062 cross-compilation. The embedded targets @samp{sparc-*-aout} and
24063 @samp{sparclite-*-*} do provide software floating-point support.
24065 @option{-msoft-float} changes the calling convention in the output file;
24066 therefore, it is only useful if you compile @emph{all} of a program with
24067 this option. In particular, you need to compile @file{libgcc.a}, the
24068 library that comes with GCC, with @option{-msoft-float} in order for
24071 @item -mhard-quad-float
24072 @opindex mhard-quad-float
24073 Generate output containing quad-word (long double) floating-point
24076 @item -msoft-quad-float
24077 @opindex msoft-quad-float
24078 Generate output containing library calls for quad-word (long double)
24079 floating-point instructions. The functions called are those specified
24080 in the SPARC ABI@. This is the default.
24082 As of this writing, there are no SPARC implementations that have hardware
24083 support for the quad-word floating-point instructions. They all invoke
24084 a trap handler for one of these instructions, and then the trap handler
24085 emulates the effect of the instruction. Because of the trap handler overhead,
24086 this is much slower than calling the ABI library routines. Thus the
24087 @option{-msoft-quad-float} option is the default.
24089 @item -mno-unaligned-doubles
24090 @itemx -munaligned-doubles
24091 @opindex mno-unaligned-doubles
24092 @opindex munaligned-doubles
24093 Assume that doubles have 8-byte alignment. This is the default.
24095 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
24096 alignment only if they are contained in another type, or if they have an
24097 absolute address. Otherwise, it assumes they have 4-byte alignment.
24098 Specifying this option avoids some rare compatibility problems with code
24099 generated by other compilers. It is not the default because it results
24100 in a performance loss, especially for floating-point code.
24103 @itemx -mno-user-mode
24104 @opindex muser-mode
24105 @opindex mno-user-mode
24106 Do not generate code that can only run in supervisor mode. This is relevant
24107 only for the @code{casa} instruction emitted for the LEON3 processor. This
24110 @item -mfaster-structs
24111 @itemx -mno-faster-structs
24112 @opindex mfaster-structs
24113 @opindex mno-faster-structs
24114 With @option{-mfaster-structs}, the compiler assumes that structures
24115 should have 8-byte alignment. This enables the use of pairs of
24116 @code{ldd} and @code{std} instructions for copies in structure
24117 assignment, in place of twice as many @code{ld} and @code{st} pairs.
24118 However, the use of this changed alignment directly violates the SPARC
24119 ABI@. Thus, it's intended only for use on targets where the developer
24120 acknowledges that their resulting code is not directly in line with
24121 the rules of the ABI@.
24123 @item -mstd-struct-return
24124 @itemx -mno-std-struct-return
24125 @opindex mstd-struct-return
24126 @opindex mno-std-struct-return
24127 With @option{-mstd-struct-return}, the compiler generates checking code
24128 in functions returning structures or unions to detect size mismatches
24129 between the two sides of function calls, as per the 32-bit ABI@.
24131 The default is @option{-mno-std-struct-return}. This option has no effect
24138 Enable Local Register Allocation. This is the default for SPARC since GCC 7
24139 so @option{-mno-lra} needs to be passed to get old Reload.
24141 @item -mcpu=@var{cpu_type}
24143 Set the instruction set, register set, and instruction scheduling parameters
24144 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24145 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
24146 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
24147 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
24148 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
24149 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
24151 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
24152 which selects the best architecture option for the host processor.
24153 @option{-mcpu=native} has no effect if GCC does not recognize
24156 Default instruction scheduling parameters are used for values that select
24157 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
24158 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
24160 Here is a list of each supported architecture and their supported
24168 supersparc, hypersparc, leon, leon3
24171 f930, f934, sparclite86x
24177 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
24181 By default (unless configured otherwise), GCC generates code for the V7
24182 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
24183 additionally optimizes it for the Cypress CY7C602 chip, as used in the
24184 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
24185 SPARCStation 1, 2, IPX etc.
24187 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
24188 architecture. The only difference from V7 code is that the compiler emits
24189 the integer multiply and integer divide instructions which exist in SPARC-V8
24190 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
24191 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
24194 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
24195 the SPARC architecture. This adds the integer multiply, integer divide step
24196 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
24197 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
24198 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
24199 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
24200 MB86934 chip, which is the more recent SPARClite with FPU@.
24202 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
24203 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
24204 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
24205 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
24206 optimizes it for the TEMIC SPARClet chip.
24208 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
24209 architecture. This adds 64-bit integer and floating-point move instructions,
24210 3 additional floating-point condition code registers and conditional move
24211 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
24212 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
24213 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
24214 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
24215 @option{-mcpu=niagara}, the compiler additionally optimizes it for
24216 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
24217 additionally optimizes it for Sun UltraSPARC T2 chips. With
24218 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
24219 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
24220 additionally optimizes it for Sun UltraSPARC T4 chips. With
24221 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
24222 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
24223 additionally optimizes it for Oracle M8 chips.
24225 @item -mtune=@var{cpu_type}
24227 Set the instruction scheduling parameters for machine type
24228 @var{cpu_type}, but do not set the instruction set or register set that the
24229 option @option{-mcpu=@var{cpu_type}} does.
24231 The same values for @option{-mcpu=@var{cpu_type}} can be used for
24232 @option{-mtune=@var{cpu_type}}, but the only useful values are those
24233 that select a particular CPU implementation. Those are
24234 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
24235 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
24236 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
24237 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
24238 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
24239 and GNU/Linux toolchains, @samp{native} can also be used.
24244 @opindex mno-v8plus
24245 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
24246 difference from the V8 ABI is that the global and out registers are
24247 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
24248 mode for all SPARC-V9 processors.
24254 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
24255 Visual Instruction Set extensions. The default is @option{-mno-vis}.
24261 With @option{-mvis2}, GCC generates code that takes advantage of
24262 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
24263 default is @option{-mvis2} when targeting a cpu that supports such
24264 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
24265 also sets @option{-mvis}.
24271 With @option{-mvis3}, GCC generates code that takes advantage of
24272 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
24273 default is @option{-mvis3} when targeting a cpu that supports such
24274 instructions, such as niagara-3 and later. Setting @option{-mvis3}
24275 also sets @option{-mvis2} and @option{-mvis}.
24281 With @option{-mvis4}, GCC generates code that takes advantage of
24282 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
24283 default is @option{-mvis4} when targeting a cpu that supports such
24284 instructions, such as niagara-7 and later. Setting @option{-mvis4}
24285 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
24291 With @option{-mvis4b}, GCC generates code that takes advantage of
24292 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
24293 the additional VIS instructions introduced in the Oracle SPARC
24294 Architecture 2017. The default is @option{-mvis4b} when targeting a
24295 cpu that supports such instructions, such as m8 and later. Setting
24296 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
24297 @option{-mvis2} and @option{-mvis}.
24302 @opindex mno-cbcond
24303 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
24304 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
24305 when targeting a CPU that supports such instructions, such as Niagara-4 and
24312 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
24313 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
24314 when targeting a CPU that supports such instructions, such as Niagara-3 and
24320 @opindex mno-fsmuld
24321 With @option{-mfsmuld}, GCC generates code that takes advantage of the
24322 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
24323 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
24324 or V9 with FPU except @option{-mcpu=leon}.
24330 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
24331 Population Count instruction. The default is @option{-mpopc}
24332 when targeting a CPU that supports such an instruction, such as Niagara-2 and
24339 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
24340 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
24341 when targeting a CPU that supports such an instruction, such as Niagara-7 and
24345 @opindex mfix-at697f
24346 Enable the documented workaround for the single erratum of the Atmel AT697F
24347 processor (which corresponds to erratum #13 of the AT697E processor).
24350 @opindex mfix-ut699
24351 Enable the documented workarounds for the floating-point errata and the data
24352 cache nullify errata of the UT699 processor.
24355 @opindex mfix-ut700
24356 Enable the documented workaround for the back-to-back store errata of
24357 the UT699E/UT700 processor.
24359 @item -mfix-gr712rc
24360 @opindex mfix-gr712rc
24361 Enable the documented workaround for the back-to-back store errata of
24362 the GR712RC processor.
24365 These @samp{-m} options are supported in addition to the above
24366 on SPARC-V9 processors in 64-bit environments:
24373 Generate code for a 32-bit or 64-bit environment.
24374 The 32-bit environment sets int, long and pointer to 32 bits.
24375 The 64-bit environment sets int to 32 bits and long and pointer
24378 @item -mcmodel=@var{which}
24380 Set the code model to one of
24384 The Medium/Low code model: 64-bit addresses, programs
24385 must be linked in the low 32 bits of memory. Programs can be statically
24386 or dynamically linked.
24389 The Medium/Middle code model: 64-bit addresses, programs
24390 must be linked in the low 44 bits of memory, the text and data segments must
24391 be less than 2GB in size and the data segment must be located within 2GB of
24395 The Medium/Anywhere code model: 64-bit addresses, programs
24396 may be linked anywhere in memory, the text and data segments must be less
24397 than 2GB in size and the data segment must be located within 2GB of the
24401 The Medium/Anywhere code model for embedded systems:
24402 64-bit addresses, the text and data segments must be less than 2GB in
24403 size, both starting anywhere in memory (determined at link time). The
24404 global register %g4 points to the base of the data segment. Programs
24405 are statically linked and PIC is not supported.
24408 @item -mmemory-model=@var{mem-model}
24409 @opindex mmemory-model
24410 Set the memory model in force on the processor to one of
24414 The default memory model for the processor and operating system.
24417 Relaxed Memory Order
24420 Partial Store Order
24426 Sequential Consistency
24429 These memory models are formally defined in Appendix D of the SPARC-V9
24430 architecture manual, as set in the processor's @code{PSTATE.MM} field.
24433 @itemx -mno-stack-bias
24434 @opindex mstack-bias
24435 @opindex mno-stack-bias
24436 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
24437 frame pointer if present, are offset by @minus{}2047 which must be added back
24438 when making stack frame references. This is the default in 64-bit mode.
24439 Otherwise, assume no such offset is present.
24443 @subsection SPU Options
24444 @cindex SPU options
24446 These @samp{-m} options are supported on the SPU:
24450 @itemx -merror-reloc
24451 @opindex mwarn-reloc
24452 @opindex merror-reloc
24454 The loader for SPU does not handle dynamic relocations. By default, GCC
24455 gives an error when it generates code that requires a dynamic
24456 relocation. @option{-mno-error-reloc} disables the error,
24457 @option{-mwarn-reloc} generates a warning instead.
24460 @itemx -munsafe-dma
24462 @opindex munsafe-dma
24464 Instructions that initiate or test completion of DMA must not be
24465 reordered with respect to loads and stores of the memory that is being
24467 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
24468 memory accesses, but that can lead to inefficient code in places where the
24469 memory is known to not change. Rather than mark the memory as volatile,
24470 you can use @option{-msafe-dma} to tell the compiler to treat
24471 the DMA instructions as potentially affecting all memory.
24473 @item -mbranch-hints
24474 @opindex mbranch-hints
24476 By default, GCC generates a branch hint instruction to avoid
24477 pipeline stalls for always-taken or probably-taken branches. A hint
24478 is not generated closer than 8 instructions away from its branch.
24479 There is little reason to disable them, except for debugging purposes,
24480 or to make an object a little bit smaller.
24484 @opindex msmall-mem
24485 @opindex mlarge-mem
24487 By default, GCC generates code assuming that addresses are never larger
24488 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
24489 a full 32-bit address.
24494 By default, GCC links against startup code that assumes the SPU-style
24495 main function interface (which has an unconventional parameter list).
24496 With @option{-mstdmain}, GCC links your program against startup
24497 code that assumes a C99-style interface to @code{main}, including a
24498 local copy of @code{argv} strings.
24500 @item -mfixed-range=@var{register-range}
24501 @opindex mfixed-range
24502 Generate code treating the given register range as fixed registers.
24503 A fixed register is one that the register allocator cannot use. This is
24504 useful when compiling kernel code. A register range is specified as
24505 two registers separated by a dash. Multiple register ranges can be
24506 specified separated by a comma.
24512 Compile code assuming that pointers to the PPU address space accessed
24513 via the @code{__ea} named address space qualifier are either 32 or 64
24514 bits wide. The default is 32 bits. As this is an ABI-changing option,
24515 all object code in an executable must be compiled with the same setting.
24517 @item -maddress-space-conversion
24518 @itemx -mno-address-space-conversion
24519 @opindex maddress-space-conversion
24520 @opindex mno-address-space-conversion
24521 Allow/disallow treating the @code{__ea} address space as superset
24522 of the generic address space. This enables explicit type casts
24523 between @code{__ea} and generic pointer as well as implicit
24524 conversions of generic pointers to @code{__ea} pointers. The
24525 default is to allow address space pointer conversions.
24527 @item -mcache-size=@var{cache-size}
24528 @opindex mcache-size
24529 This option controls the version of libgcc that the compiler links to an
24530 executable and selects a software-managed cache for accessing variables
24531 in the @code{__ea} address space with a particular cache size. Possible
24532 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
24533 and @samp{128}. The default cache size is 64KB.
24535 @item -matomic-updates
24536 @itemx -mno-atomic-updates
24537 @opindex matomic-updates
24538 @opindex mno-atomic-updates
24539 This option controls the version of libgcc that the compiler links to an
24540 executable and selects whether atomic updates to the software-managed
24541 cache of PPU-side variables are used. If you use atomic updates, changes
24542 to a PPU variable from SPU code using the @code{__ea} named address space
24543 qualifier do not interfere with changes to other PPU variables residing
24544 in the same cache line from PPU code. If you do not use atomic updates,
24545 such interference may occur; however, writing back cache lines is
24546 more efficient. The default behavior is to use atomic updates.
24549 @itemx -mdual-nops=@var{n}
24550 @opindex mdual-nops
24551 By default, GCC inserts NOPs to increase dual issue when it expects
24552 it to increase performance. @var{n} can be a value from 0 to 10. A
24553 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
24554 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
24556 @item -mhint-max-nops=@var{n}
24557 @opindex mhint-max-nops
24558 Maximum number of NOPs to insert for a branch hint. A branch hint must
24559 be at least 8 instructions away from the branch it is affecting. GCC
24560 inserts up to @var{n} NOPs to enforce this, otherwise it does not
24561 generate the branch hint.
24563 @item -mhint-max-distance=@var{n}
24564 @opindex mhint-max-distance
24565 The encoding of the branch hint instruction limits the hint to be within
24566 256 instructions of the branch it is affecting. By default, GCC makes
24567 sure it is within 125.
24570 @opindex msafe-hints
24571 Work around a hardware bug that causes the SPU to stall indefinitely.
24572 By default, GCC inserts the @code{hbrp} instruction to make sure
24573 this stall won't happen.
24577 @node System V Options
24578 @subsection Options for System V
24580 These additional options are available on System V Release 4 for
24581 compatibility with other compilers on those systems:
24586 Create a shared object.
24587 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
24591 Identify the versions of each tool used by the compiler, in a
24592 @code{.ident} assembler directive in the output.
24596 Refrain from adding @code{.ident} directives to the output file (this is
24599 @item -YP,@var{dirs}
24601 Search the directories @var{dirs}, and no others, for libraries
24602 specified with @option{-l}.
24604 @item -Ym,@var{dir}
24606 Look in the directory @var{dir} to find the M4 preprocessor.
24607 The assembler uses this option.
24608 @c This is supposed to go with a -Yd for predefined M4 macro files, but
24609 @c the generic assembler that comes with Solaris takes just -Ym.
24612 @node TILE-Gx Options
24613 @subsection TILE-Gx Options
24614 @cindex TILE-Gx options
24616 These @samp{-m} options are supported on the TILE-Gx:
24619 @item -mcmodel=small
24620 @opindex mcmodel=small
24621 Generate code for the small model. The distance for direct calls is
24622 limited to 500M in either direction. PC-relative addresses are 32
24623 bits. Absolute addresses support the full address range.
24625 @item -mcmodel=large
24626 @opindex mcmodel=large
24627 Generate code for the large model. There is no limitation on call
24628 distance, pc-relative addresses, or absolute addresses.
24630 @item -mcpu=@var{name}
24632 Selects the type of CPU to be targeted. Currently the only supported
24633 type is @samp{tilegx}.
24639 Generate code for a 32-bit or 64-bit environment. The 32-bit
24640 environment sets int, long, and pointer to 32 bits. The 64-bit
24641 environment sets int to 32 bits and long and pointer to 64 bits.
24644 @itemx -mlittle-endian
24645 @opindex mbig-endian
24646 @opindex mlittle-endian
24647 Generate code in big/little endian mode, respectively.
24650 @node TILEPro Options
24651 @subsection TILEPro Options
24652 @cindex TILEPro options
24654 These @samp{-m} options are supported on the TILEPro:
24657 @item -mcpu=@var{name}
24659 Selects the type of CPU to be targeted. Currently the only supported
24660 type is @samp{tilepro}.
24664 Generate code for a 32-bit environment, which sets int, long, and
24665 pointer to 32 bits. This is the only supported behavior so the flag
24666 is essentially ignored.
24670 @subsection V850 Options
24671 @cindex V850 Options
24673 These @samp{-m} options are defined for V850 implementations:
24677 @itemx -mno-long-calls
24678 @opindex mlong-calls
24679 @opindex mno-long-calls
24680 Treat all calls as being far away (near). If calls are assumed to be
24681 far away, the compiler always loads the function's address into a
24682 register, and calls indirect through the pointer.
24688 Do not optimize (do optimize) basic blocks that use the same index
24689 pointer 4 or more times to copy pointer into the @code{ep} register, and
24690 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
24691 option is on by default if you optimize.
24693 @item -mno-prolog-function
24694 @itemx -mprolog-function
24695 @opindex mno-prolog-function
24696 @opindex mprolog-function
24697 Do not use (do use) external functions to save and restore registers
24698 at the prologue and epilogue of a function. The external functions
24699 are slower, but use less code space if more than one function saves
24700 the same number of registers. The @option{-mprolog-function} option
24701 is on by default if you optimize.
24705 Try to make the code as small as possible. At present, this just turns
24706 on the @option{-mep} and @option{-mprolog-function} options.
24708 @item -mtda=@var{n}
24710 Put static or global variables whose size is @var{n} bytes or less into
24711 the tiny data area that register @code{ep} points to. The tiny data
24712 area can hold up to 256 bytes in total (128 bytes for byte references).
24714 @item -msda=@var{n}
24716 Put static or global variables whose size is @var{n} bytes or less into
24717 the small data area that register @code{gp} points to. The small data
24718 area can hold up to 64 kilobytes.
24720 @item -mzda=@var{n}
24722 Put static or global variables whose size is @var{n} bytes or less into
24723 the first 32 kilobytes of memory.
24727 Specify that the target processor is the V850.
24731 Specify that the target processor is the V850E3V5. The preprocessor
24732 constant @code{__v850e3v5__} is defined if this option is used.
24736 Specify that the target processor is the V850E3V5. This is an alias for
24737 the @option{-mv850e3v5} option.
24741 Specify that the target processor is the V850E2V3. The preprocessor
24742 constant @code{__v850e2v3__} is defined if this option is used.
24746 Specify that the target processor is the V850E2. The preprocessor
24747 constant @code{__v850e2__} is defined if this option is used.
24751 Specify that the target processor is the V850E1. The preprocessor
24752 constants @code{__v850e1__} and @code{__v850e__} are defined if
24753 this option is used.
24757 Specify that the target processor is the V850ES. This is an alias for
24758 the @option{-mv850e1} option.
24762 Specify that the target processor is the V850E@. The preprocessor
24763 constant @code{__v850e__} is defined if this option is used.
24765 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
24766 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
24767 are defined then a default target processor is chosen and the
24768 relevant @samp{__v850*__} preprocessor constant is defined.
24770 The preprocessor constants @code{__v850} and @code{__v851__} are always
24771 defined, regardless of which processor variant is the target.
24773 @item -mdisable-callt
24774 @itemx -mno-disable-callt
24775 @opindex mdisable-callt
24776 @opindex mno-disable-callt
24777 This option suppresses generation of the @code{CALLT} instruction for the
24778 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
24781 This option is enabled by default when the RH850 ABI is
24782 in use (see @option{-mrh850-abi}), and disabled by default when the
24783 GCC ABI is in use. If @code{CALLT} instructions are being generated
24784 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
24790 Pass on (or do not pass on) the @option{-mrelax} command-line option
24794 @itemx -mno-long-jumps
24795 @opindex mlong-jumps
24796 @opindex mno-long-jumps
24797 Disable (or re-enable) the generation of PC-relative jump instructions.
24800 @itemx -mhard-float
24801 @opindex msoft-float
24802 @opindex mhard-float
24803 Disable (or re-enable) the generation of hardware floating point
24804 instructions. This option is only significant when the target
24805 architecture is @samp{V850E2V3} or higher. If hardware floating point
24806 instructions are being generated then the C preprocessor symbol
24807 @code{__FPU_OK__} is defined, otherwise the symbol
24808 @code{__NO_FPU__} is defined.
24812 Enables the use of the e3v5 LOOP instruction. The use of this
24813 instruction is not enabled by default when the e3v5 architecture is
24814 selected because its use is still experimental.
24818 @opindex mrh850-abi
24820 Enables support for the RH850 version of the V850 ABI. This is the
24821 default. With this version of the ABI the following rules apply:
24825 Integer sized structures and unions are returned via a memory pointer
24826 rather than a register.
24829 Large structures and unions (more than 8 bytes in size) are passed by
24833 Functions are aligned to 16-bit boundaries.
24836 The @option{-m8byte-align} command-line option is supported.
24839 The @option{-mdisable-callt} command-line option is enabled by
24840 default. The @option{-mno-disable-callt} command-line option is not
24844 When this version of the ABI is enabled the C preprocessor symbol
24845 @code{__V850_RH850_ABI__} is defined.
24849 Enables support for the old GCC version of the V850 ABI. With this
24850 version of the ABI the following rules apply:
24854 Integer sized structures and unions are returned in register @code{r10}.
24857 Large structures and unions (more than 8 bytes in size) are passed by
24861 Functions are aligned to 32-bit boundaries, unless optimizing for
24865 The @option{-m8byte-align} command-line option is not supported.
24868 The @option{-mdisable-callt} command-line option is supported but not
24869 enabled by default.
24872 When this version of the ABI is enabled the C preprocessor symbol
24873 @code{__V850_GCC_ABI__} is defined.
24875 @item -m8byte-align
24876 @itemx -mno-8byte-align
24877 @opindex m8byte-align
24878 @opindex mno-8byte-align
24879 Enables support for @code{double} and @code{long long} types to be
24880 aligned on 8-byte boundaries. The default is to restrict the
24881 alignment of all objects to at most 4-bytes. When
24882 @option{-m8byte-align} is in effect the C preprocessor symbol
24883 @code{__V850_8BYTE_ALIGN__} is defined.
24886 @opindex mbig-switch
24887 Generate code suitable for big switch tables. Use this option only if
24888 the assembler/linker complain about out of range branches within a switch
24893 This option causes r2 and r5 to be used in the code generated by
24894 the compiler. This setting is the default.
24896 @item -mno-app-regs
24897 @opindex mno-app-regs
24898 This option causes r2 and r5 to be treated as fixed registers.
24903 @subsection VAX Options
24904 @cindex VAX options
24906 These @samp{-m} options are defined for the VAX:
24911 Do not output certain jump instructions (@code{aobleq} and so on)
24912 that the Unix assembler for the VAX cannot handle across long
24917 Do output those jump instructions, on the assumption that the
24918 GNU assembler is being used.
24922 Output code for G-format floating-point numbers instead of D-format.
24925 @node Visium Options
24926 @subsection Visium Options
24927 @cindex Visium options
24933 A program which performs file I/O and is destined to run on an MCM target
24934 should be linked with this option. It causes the libraries libc.a and
24935 libdebug.a to be linked. The program should be run on the target under
24936 the control of the GDB remote debugging stub.
24940 A program which performs file I/O and is destined to run on the simulator
24941 should be linked with option. This causes libraries libc.a and libsim.a to
24945 @itemx -mhard-float
24947 @opindex mhard-float
24948 Generate code containing floating-point instructions. This is the
24952 @itemx -msoft-float
24954 @opindex msoft-float
24955 Generate code containing library calls for floating-point.
24957 @option{-msoft-float} changes the calling convention in the output file;
24958 therefore, it is only useful if you compile @emph{all} of a program with
24959 this option. In particular, you need to compile @file{libgcc.a}, the
24960 library that comes with GCC, with @option{-msoft-float} in order for
24963 @item -mcpu=@var{cpu_type}
24965 Set the instruction set, register set, and instruction scheduling parameters
24966 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24967 @samp{mcm}, @samp{gr5} and @samp{gr6}.
24969 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
24971 By default (unless configured otherwise), GCC generates code for the GR5
24972 variant of the Visium architecture.
24974 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
24975 architecture. The only difference from GR5 code is that the compiler will
24976 generate block move instructions.
24978 @item -mtune=@var{cpu_type}
24980 Set the instruction scheduling parameters for machine type @var{cpu_type},
24981 but do not set the instruction set or register set that the option
24982 @option{-mcpu=@var{cpu_type}} would.
24986 Generate code for the supervisor mode, where there are no restrictions on
24987 the access to general registers. This is the default.
24990 @opindex muser-mode
24991 Generate code for the user mode, where the access to some general registers
24992 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
24993 mode; on the GR6, only registers r29 to r31 are affected.
24997 @subsection VMS Options
24999 These @samp{-m} options are defined for the VMS implementations:
25002 @item -mvms-return-codes
25003 @opindex mvms-return-codes
25004 Return VMS condition codes from @code{main}. The default is to return POSIX-style
25005 condition (e.g.@ error) codes.
25007 @item -mdebug-main=@var{prefix}
25008 @opindex mdebug-main=@var{prefix}
25009 Flag the first routine whose name starts with @var{prefix} as the main
25010 routine for the debugger.
25014 Default to 64-bit memory allocation routines.
25016 @item -mpointer-size=@var{size}
25017 @opindex mpointer-size=@var{size}
25018 Set the default size of pointers. Possible options for @var{size} are
25019 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
25020 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
25021 The later option disables @code{pragma pointer_size}.
25024 @node VxWorks Options
25025 @subsection VxWorks Options
25026 @cindex VxWorks Options
25028 The options in this section are defined for all VxWorks targets.
25029 Options specific to the target hardware are listed with the other
25030 options for that target.
25035 GCC can generate code for both VxWorks kernels and real time processes
25036 (RTPs). This option switches from the former to the latter. It also
25037 defines the preprocessor macro @code{__RTP__}.
25040 @opindex non-static
25041 Link an RTP executable against shared libraries rather than static
25042 libraries. The options @option{-static} and @option{-shared} can
25043 also be used for RTPs (@pxref{Link Options}); @option{-static}
25050 These options are passed down to the linker. They are defined for
25051 compatibility with Diab.
25054 @opindex Xbind-lazy
25055 Enable lazy binding of function calls. This option is equivalent to
25056 @option{-Wl,-z,now} and is defined for compatibility with Diab.
25060 Disable lazy binding of function calls. This option is the default and
25061 is defined for compatibility with Diab.
25065 @subsection x86 Options
25066 @cindex x86 Options
25068 These @samp{-m} options are defined for the x86 family of computers.
25072 @item -march=@var{cpu-type}
25074 Generate instructions for the machine type @var{cpu-type}. In contrast to
25075 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
25076 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
25077 to generate code that may not run at all on processors other than the one
25078 indicated. Specifying @option{-march=@var{cpu-type}} implies
25079 @option{-mtune=@var{cpu-type}}.
25081 The choices for @var{cpu-type} are:
25085 This selects the CPU to generate code for at compilation time by determining
25086 the processor type of the compiling machine. Using @option{-march=native}
25087 enables all instruction subsets supported by the local machine (hence
25088 the result might not run on different machines). Using @option{-mtune=native}
25089 produces code optimized for the local machine under the constraints
25090 of the selected instruction set.
25093 Original Intel i386 CPU@.
25096 Intel i486 CPU@. (No scheduling is implemented for this chip.)
25100 Intel Pentium CPU with no MMX support.
25103 Intel Lakemont MCU, based on Intel Pentium CPU.
25106 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
25109 Intel Pentium Pro CPU@.
25112 When used with @option{-march}, the Pentium Pro
25113 instruction set is used, so the code runs on all i686 family chips.
25114 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
25117 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
25122 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
25126 Intel Pentium M; low-power version of Intel Pentium III CPU
25127 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
25131 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
25134 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
25138 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
25139 SSE2 and SSE3 instruction set support.
25142 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
25143 instruction set support.
25146 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25147 SSE4.1, SSE4.2 and POPCNT instruction set support.
25150 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25151 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
25154 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25155 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
25158 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25159 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
25160 instruction set support.
25163 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25164 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25165 BMI, BMI2 and F16C instruction set support.
25168 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25169 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25170 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
25173 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25174 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25175 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
25176 XSAVES instruction set support.
25179 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
25180 instruction set support.
25183 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25184 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
25187 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25188 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25189 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
25190 AVX512CD instruction set support.
25193 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25194 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25195 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25196 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
25198 @item skylake-avx512
25199 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25200 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25201 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
25202 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
25205 AMD K6 CPU with MMX instruction set support.
25209 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
25212 @itemx athlon-tbird
25213 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
25219 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
25220 instruction set support.
25226 Processors based on the AMD K8 core with x86-64 instruction set support,
25227 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
25228 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
25229 instruction set extensions.)
25232 @itemx opteron-sse3
25233 @itemx athlon64-sse3
25234 Improved versions of AMD K8 cores with SSE3 instruction set support.
25238 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
25239 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
25240 instruction set extensions.)
25243 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
25244 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
25245 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
25247 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25248 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
25249 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
25252 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25253 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
25254 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
25255 64-bit instruction set extensions.
25257 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25258 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
25259 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
25260 SSE4.2, ABM and 64-bit instruction set extensions.
25263 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
25264 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
25265 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
25266 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
25267 instruction set extensions.
25270 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
25271 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
25272 instruction set extensions.)
25275 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
25276 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
25277 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
25280 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
25284 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
25285 instruction set support.
25288 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
25289 (No scheduling is implemented for this chip.)
25292 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
25293 (No scheduling is implemented for this chip.)
25296 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25297 (No scheduling is implemented for this chip.)
25300 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
25301 (No scheduling is implemented for this chip.)
25304 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
25305 (No scheduling is implemented for this chip.)
25308 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25309 (No scheduling is implemented for this chip.)
25312 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
25313 (No scheduling is implemented for this chip.)
25316 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
25317 AVX and AVX2 instruction set support.
25318 (No scheduling is implemented for this chip.)
25321 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25322 instruction set support.
25323 (No scheduling is implemented for this chip.)
25326 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25327 instruction set support.
25328 (No scheduling is implemented for this chip.)
25331 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25332 instruction set support.
25333 (No scheduling is implemented for this chip.)
25336 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25337 instruction set support.
25338 (No scheduling is implemented for this chip.)
25341 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25342 instruction set support.
25343 (No scheduling is implemented for this chip.)
25346 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25347 instruction set support.
25348 (No scheduling is implemented for this chip.)
25351 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
25354 @item -mtune=@var{cpu-type}
25356 Tune to @var{cpu-type} everything applicable about the generated code, except
25357 for the ABI and the set of available instructions.
25358 While picking a specific @var{cpu-type} schedules things appropriately
25359 for that particular chip, the compiler does not generate any code that
25360 cannot run on the default machine type unless you use a
25361 @option{-march=@var{cpu-type}} option.
25362 For example, if GCC is configured for i686-pc-linux-gnu
25363 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
25364 but still runs on i686 machines.
25366 The choices for @var{cpu-type} are the same as for @option{-march}.
25367 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
25371 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
25372 If you know the CPU on which your code will run, then you should use
25373 the corresponding @option{-mtune} or @option{-march} option instead of
25374 @option{-mtune=generic}. But, if you do not know exactly what CPU users
25375 of your application will have, then you should use this option.
25377 As new processors are deployed in the marketplace, the behavior of this
25378 option will change. Therefore, if you upgrade to a newer version of
25379 GCC, code generation controlled by this option will change to reflect
25381 that are most common at the time that version of GCC is released.
25383 There is no @option{-march=generic} option because @option{-march}
25384 indicates the instruction set the compiler can use, and there is no
25385 generic instruction set applicable to all processors. In contrast,
25386 @option{-mtune} indicates the processor (or, in this case, collection of
25387 processors) for which the code is optimized.
25390 Produce code optimized for the most current Intel processors, which are
25391 Haswell and Silvermont for this version of GCC. If you know the CPU
25392 on which your code will run, then you should use the corresponding
25393 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
25394 But, if you want your application performs better on both Haswell and
25395 Silvermont, then you should use this option.
25397 As new Intel processors are deployed in the marketplace, the behavior of
25398 this option will change. Therefore, if you upgrade to a newer version of
25399 GCC, code generation controlled by this option will change to reflect
25400 the most current Intel processors at the time that version of GCC is
25403 There is no @option{-march=intel} option because @option{-march} indicates
25404 the instruction set the compiler can use, and there is no common
25405 instruction set applicable to all processors. In contrast,
25406 @option{-mtune} indicates the processor (or, in this case, collection of
25407 processors) for which the code is optimized.
25410 @item -mcpu=@var{cpu-type}
25412 A deprecated synonym for @option{-mtune}.
25414 @item -mfpmath=@var{unit}
25416 Generate floating-point arithmetic for selected unit @var{unit}. The choices
25417 for @var{unit} are:
25421 Use the standard 387 floating-point coprocessor present on the majority of chips and
25422 emulated otherwise. Code compiled with this option runs almost everywhere.
25423 The temporary results are computed in 80-bit precision instead of the precision
25424 specified by the type, resulting in slightly different results compared to most
25425 of other chips. See @option{-ffloat-store} for more detailed description.
25427 This is the default choice for non-Darwin x86-32 targets.
25430 Use scalar floating-point instructions present in the SSE instruction set.
25431 This instruction set is supported by Pentium III and newer chips,
25432 and in the AMD line
25433 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
25434 instruction set supports only single-precision arithmetic, thus the double and
25435 extended-precision arithmetic are still done using 387. A later version, present
25436 only in Pentium 4 and AMD x86-64 chips, supports double-precision
25439 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
25440 or @option{-msse2} switches to enable SSE extensions and make this option
25441 effective. For the x86-64 compiler, these extensions are enabled by default.
25443 The resulting code should be considerably faster in the majority of cases and avoid
25444 the numerical instability problems of 387 code, but may break some existing
25445 code that expects temporaries to be 80 bits.
25447 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
25448 and the default choice for x86-32 targets with the SSE2 instruction set
25449 when @option{-ffast-math} is enabled.
25454 Attempt to utilize both instruction sets at once. This effectively doubles the
25455 amount of available registers, and on chips with separate execution units for
25456 387 and SSE the execution resources too. Use this option with care, as it is
25457 still experimental, because the GCC register allocator does not model separate
25458 functional units well, resulting in unstable performance.
25461 @item -masm=@var{dialect}
25462 @opindex masm=@var{dialect}
25463 Output assembly instructions using selected @var{dialect}. Also affects
25464 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
25465 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
25466 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
25467 not support @samp{intel}.
25470 @itemx -mno-ieee-fp
25472 @opindex mno-ieee-fp
25473 Control whether or not the compiler uses IEEE floating-point
25474 comparisons. These correctly handle the case where the result of a
25475 comparison is unordered.
25480 @opindex mhard-float
25481 Generate output containing 80387 instructions for floating point.
25486 @opindex msoft-float
25487 Generate output containing library calls for floating point.
25489 @strong{Warning:} the requisite libraries are not part of GCC@.
25490 Normally the facilities of the machine's usual C compiler are used, but
25491 this cannot be done directly in cross-compilation. You must make your
25492 own arrangements to provide suitable library functions for
25495 On machines where a function returns floating-point results in the 80387
25496 register stack, some floating-point opcodes may be emitted even if
25497 @option{-msoft-float} is used.
25499 @item -mno-fp-ret-in-387
25500 @opindex mno-fp-ret-in-387
25501 Do not use the FPU registers for return values of functions.
25503 The usual calling convention has functions return values of types
25504 @code{float} and @code{double} in an FPU register, even if there
25505 is no FPU@. The idea is that the operating system should emulate
25508 The option @option{-mno-fp-ret-in-387} causes such values to be returned
25509 in ordinary CPU registers instead.
25511 @item -mno-fancy-math-387
25512 @opindex mno-fancy-math-387
25513 Some 387 emulators do not support the @code{sin}, @code{cos} and
25514 @code{sqrt} instructions for the 387. Specify this option to avoid
25515 generating those instructions. This option is the default on
25516 OpenBSD and NetBSD@. This option is overridden when @option{-march}
25517 indicates that the target CPU always has an FPU and so the
25518 instruction does not need emulation. These
25519 instructions are not generated unless you also use the
25520 @option{-funsafe-math-optimizations} switch.
25522 @item -malign-double
25523 @itemx -mno-align-double
25524 @opindex malign-double
25525 @opindex mno-align-double
25526 Control whether GCC aligns @code{double}, @code{long double}, and
25527 @code{long long} variables on a two-word boundary or a one-word
25528 boundary. Aligning @code{double} variables on a two-word boundary
25529 produces code that runs somewhat faster on a Pentium at the
25530 expense of more memory.
25532 On x86-64, @option{-malign-double} is enabled by default.
25534 @strong{Warning:} if you use the @option{-malign-double} switch,
25535 structures containing the above types are aligned differently than
25536 the published application binary interface specifications for the x86-32
25537 and are not binary compatible with structures in code compiled
25538 without that switch.
25540 @item -m96bit-long-double
25541 @itemx -m128bit-long-double
25542 @opindex m96bit-long-double
25543 @opindex m128bit-long-double
25544 These switches control the size of @code{long double} type. The x86-32
25545 application binary interface specifies the size to be 96 bits,
25546 so @option{-m96bit-long-double} is the default in 32-bit mode.
25548 Modern architectures (Pentium and newer) prefer @code{long double}
25549 to be aligned to an 8- or 16-byte boundary. In arrays or structures
25550 conforming to the ABI, this is not possible. So specifying
25551 @option{-m128bit-long-double} aligns @code{long double}
25552 to a 16-byte boundary by padding the @code{long double} with an additional
25555 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
25556 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
25558 Notice that neither of these options enable any extra precision over the x87
25559 standard of 80 bits for a @code{long double}.
25561 @strong{Warning:} if you override the default value for your target ABI, this
25562 changes the size of
25563 structures and arrays containing @code{long double} variables,
25564 as well as modifying the function calling convention for functions taking
25565 @code{long double}. Hence they are not binary-compatible
25566 with code compiled without that switch.
25568 @item -mlong-double-64
25569 @itemx -mlong-double-80
25570 @itemx -mlong-double-128
25571 @opindex mlong-double-64
25572 @opindex mlong-double-80
25573 @opindex mlong-double-128
25574 These switches control the size of @code{long double} type. A size
25575 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25576 type. This is the default for 32-bit Bionic C library. A size
25577 of 128 bits makes the @code{long double} type equivalent to the
25578 @code{__float128} type. This is the default for 64-bit Bionic C library.
25580 @strong{Warning:} if you override the default value for your target ABI, this
25581 changes the size of
25582 structures and arrays containing @code{long double} variables,
25583 as well as modifying the function calling convention for functions taking
25584 @code{long double}. Hence they are not binary-compatible
25585 with code compiled without that switch.
25587 @item -malign-data=@var{type}
25588 @opindex malign-data
25589 Control how GCC aligns variables. Supported values for @var{type} are
25590 @samp{compat} uses increased alignment value compatible uses GCC 4.8
25591 and earlier, @samp{abi} uses alignment value as specified by the
25592 psABI, and @samp{cacheline} uses increased alignment value to match
25593 the cache line size. @samp{compat} is the default.
25595 @item -mlarge-data-threshold=@var{threshold}
25596 @opindex mlarge-data-threshold
25597 When @option{-mcmodel=medium} is specified, data objects larger than
25598 @var{threshold} are placed in the large data section. This value must be the
25599 same across all objects linked into the binary, and defaults to 65535.
25603 Use a different function-calling convention, in which functions that
25604 take a fixed number of arguments return with the @code{ret @var{num}}
25605 instruction, which pops their arguments while returning. This saves one
25606 instruction in the caller since there is no need to pop the arguments
25609 You can specify that an individual function is called with this calling
25610 sequence with the function attribute @code{stdcall}. You can also
25611 override the @option{-mrtd} option by using the function attribute
25612 @code{cdecl}. @xref{Function Attributes}.
25614 @strong{Warning:} this calling convention is incompatible with the one
25615 normally used on Unix, so you cannot use it if you need to call
25616 libraries compiled with the Unix compiler.
25618 Also, you must provide function prototypes for all functions that
25619 take variable numbers of arguments (including @code{printf});
25620 otherwise incorrect code is generated for calls to those
25623 In addition, seriously incorrect code results if you call a
25624 function with too many arguments. (Normally, extra arguments are
25625 harmlessly ignored.)
25627 @item -mregparm=@var{num}
25629 Control how many registers are used to pass integer arguments. By
25630 default, no registers are used to pass arguments, and at most 3
25631 registers can be used. You can control this behavior for a specific
25632 function by using the function attribute @code{regparm}.
25633 @xref{Function Attributes}.
25635 @strong{Warning:} if you use this switch, and
25636 @var{num} is nonzero, then you must build all modules with the same
25637 value, including any libraries. This includes the system libraries and
25641 @opindex msseregparm
25642 Use SSE register passing conventions for float and double arguments
25643 and return values. You can control this behavior for a specific
25644 function by using the function attribute @code{sseregparm}.
25645 @xref{Function Attributes}.
25647 @strong{Warning:} if you use this switch then you must build all
25648 modules with the same value, including any libraries. This includes
25649 the system libraries and startup modules.
25651 @item -mvect8-ret-in-mem
25652 @opindex mvect8-ret-in-mem
25653 Return 8-byte vectors in memory instead of MMX registers. This is the
25654 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
25655 Studio compilers until version 12. Later compiler versions (starting
25656 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
25657 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
25658 you need to remain compatible with existing code produced by those
25659 previous compiler versions or older versions of GCC@.
25668 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
25669 is specified, the significands of results of floating-point operations are
25670 rounded to 24 bits (single precision); @option{-mpc64} rounds the
25671 significands of results of floating-point operations to 53 bits (double
25672 precision) and @option{-mpc80} rounds the significands of results of
25673 floating-point operations to 64 bits (extended double precision), which is
25674 the default. When this option is used, floating-point operations in higher
25675 precisions are not available to the programmer without setting the FPU
25676 control word explicitly.
25678 Setting the rounding of floating-point operations to less than the default
25679 80 bits can speed some programs by 2% or more. Note that some mathematical
25680 libraries assume that extended-precision (80-bit) floating-point operations
25681 are enabled by default; routines in such libraries could suffer significant
25682 loss of accuracy, typically through so-called ``catastrophic cancellation'',
25683 when this option is used to set the precision to less than extended precision.
25685 @item -mstackrealign
25686 @opindex mstackrealign
25687 Realign the stack at entry. On the x86, the @option{-mstackrealign}
25688 option generates an alternate prologue and epilogue that realigns the
25689 run-time stack if necessary. This supports mixing legacy codes that keep
25690 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
25691 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
25692 applicable to individual functions.
25694 @item -mpreferred-stack-boundary=@var{num}
25695 @opindex mpreferred-stack-boundary
25696 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
25697 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
25698 the default is 4 (16 bytes or 128 bits).
25700 @strong{Warning:} When generating code for the x86-64 architecture with
25701 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
25702 used to keep the stack boundary aligned to 8 byte boundary. Since
25703 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
25704 intended to be used in controlled environment where stack space is
25705 important limitation. This option leads to wrong code when functions
25706 compiled with 16 byte stack alignment (such as functions from a standard
25707 library) are called with misaligned stack. In this case, SSE
25708 instructions may lead to misaligned memory access traps. In addition,
25709 variable arguments are handled incorrectly for 16 byte aligned
25710 objects (including x87 long double and __int128), leading to wrong
25711 results. You must build all modules with
25712 @option{-mpreferred-stack-boundary=3}, including any libraries. This
25713 includes the system libraries and startup modules.
25715 @item -mincoming-stack-boundary=@var{num}
25716 @opindex mincoming-stack-boundary
25717 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
25718 boundary. If @option{-mincoming-stack-boundary} is not specified,
25719 the one specified by @option{-mpreferred-stack-boundary} is used.
25721 On Pentium and Pentium Pro, @code{double} and @code{long double} values
25722 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
25723 suffer significant run time performance penalties. On Pentium III, the
25724 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
25725 properly if it is not 16-byte aligned.
25727 To ensure proper alignment of this values on the stack, the stack boundary
25728 must be as aligned as that required by any value stored on the stack.
25729 Further, every function must be generated such that it keeps the stack
25730 aligned. Thus calling a function compiled with a higher preferred
25731 stack boundary from a function compiled with a lower preferred stack
25732 boundary most likely misaligns the stack. It is recommended that
25733 libraries that use callbacks always use the default setting.
25735 This extra alignment does consume extra stack space, and generally
25736 increases code size. Code that is sensitive to stack space usage, such
25737 as embedded systems and operating system kernels, may want to reduce the
25738 preferred alignment to @option{-mpreferred-stack-boundary=2}.
25795 @itemx -mavx512ifma
25796 @opindex mavx512ifma
25798 @itemx -mavx512vbmi
25799 @opindex mavx512vbmi
25811 @opindex mclfushopt
25828 @itemx -mprefetchwt1
25829 @opindex mprefetchwt1
25892 These switches enable the use of instructions in the MMX, SSE,
25893 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25894 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
25895 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
25896 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, IBT, SHSTK,
25897 3DNow!@: or enhanced 3DNow!@: extended instruction sets. Each has a
25898 corresponding @option{-mno-} option to disable use of these instructions.
25900 These extensions are also available as built-in functions: see
25901 @ref{x86 Built-in Functions}, for details of the functions enabled and
25902 disabled by these switches.
25904 To generate SSE/SSE2 instructions automatically from floating-point
25905 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
25907 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
25908 generates new AVX instructions or AVX equivalence for all SSEx instructions
25911 These options enable GCC to use these extended instructions in
25912 generated code, even without @option{-mfpmath=sse}. Applications that
25913 perform run-time CPU detection must compile separate files for each
25914 supported architecture, using the appropriate flags. In particular,
25915 the file containing the CPU detection code should be compiled without
25918 The @option{-mcet} option turns on the @option{-mibt} and @option{-mshstk}
25919 options. The @option{-mibt} option enables indirect branch tracking support
25920 and the @option{-mshstk} option enables shadow stack support from
25921 Intel Control-flow Enforcement Technology (CET). The compiler also provides
25922 a number of built-in functions for fine-grained control in a CET-based
25923 application. See @xref{x86 Built-in Functions}, for more information.
25925 @item -mdump-tune-features
25926 @opindex mdump-tune-features
25927 This option instructs GCC to dump the names of the x86 performance
25928 tuning features and default settings. The names can be used in
25929 @option{-mtune-ctrl=@var{feature-list}}.
25931 @item -mtune-ctrl=@var{feature-list}
25932 @opindex mtune-ctrl=@var{feature-list}
25933 This option is used to do fine grain control of x86 code generation features.
25934 @var{feature-list} is a comma separated list of @var{feature} names. See also
25935 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
25936 on if it is not preceded with @samp{^}, otherwise, it is turned off.
25937 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
25938 developers. Using it may lead to code paths not covered by testing and can
25939 potentially result in compiler ICEs or runtime errors.
25942 @opindex mno-default
25943 This option instructs GCC to turn off all tunable features. See also
25944 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
25948 This option instructs GCC to emit a @code{cld} instruction in the prologue
25949 of functions that use string instructions. String instructions depend on
25950 the DF flag to select between autoincrement or autodecrement mode. While the
25951 ABI specifies the DF flag to be cleared on function entry, some operating
25952 systems violate this specification by not clearing the DF flag in their
25953 exception dispatchers. The exception handler can be invoked with the DF flag
25954 set, which leads to wrong direction mode when string instructions are used.
25955 This option can be enabled by default on 32-bit x86 targets by configuring
25956 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
25957 instructions can be suppressed with the @option{-mno-cld} compiler option
25961 @opindex mvzeroupper
25962 This option instructs GCC to emit a @code{vzeroupper} instruction
25963 before a transfer of control flow out of the function to minimize
25964 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
25967 @item -mprefer-avx128
25968 @opindex mprefer-avx128
25969 This option instructs GCC to use 128-bit AVX instructions instead of
25970 256-bit AVX instructions in the auto-vectorizer.
25972 @item -mprefer-avx256
25973 @opindex mprefer-avx256
25974 This option instructs GCC to use 256-bit AVX instructions instead of
25975 512-bit AVX instructions in the auto-vectorizer.
25979 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
25980 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
25981 objects. This is useful for atomic updates of data structures exceeding one
25982 machine word in size. The compiler uses this instruction to implement
25983 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
25984 128-bit integers, a library call is always used.
25988 This option enables generation of @code{SAHF} instructions in 64-bit code.
25989 Early Intel Pentium 4 CPUs with Intel 64 support,
25990 prior to the introduction of Pentium 4 G1 step in December 2005,
25991 lacked the @code{LAHF} and @code{SAHF} instructions
25992 which are supported by AMD64.
25993 These are load and store instructions, respectively, for certain status flags.
25994 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
25995 @code{drem}, and @code{remainder} built-in functions;
25996 see @ref{Other Builtins} for details.
26000 This option enables use of the @code{movbe} instruction to implement
26001 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
26005 This option tells the compiler to use indirect branch tracking support
26006 (for indirect calls and jumps) from x86 Control-flow Enforcement
26007 Technology (CET). The option has effect only if the
26008 @option{-fcf-protection=full} or @option{-fcf-protection=branch} option
26009 is specified. The option @option{-mibt} is on by default when the
26010 @code{-mcet} option is specified.
26014 This option tells the compiler to use shadow stack support (return
26015 address tracking) from x86 Control-flow Enforcement Technology (CET).
26016 The option has effect only if the @option{-fcf-protection=full} or
26017 @option{-fcf-protection=return} option is specified. The option
26018 @option{-mshstk} is on by default when the @option{-mcet} option is
26023 This option enables built-in functions @code{__builtin_ia32_crc32qi},
26024 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
26025 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
26029 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
26030 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
26031 with an additional Newton-Raphson step
26032 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
26033 (and their vectorized
26034 variants) for single-precision floating-point arguments. These instructions
26035 are generated only when @option{-funsafe-math-optimizations} is enabled
26036 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
26037 Note that while the throughput of the sequence is higher than the throughput
26038 of the non-reciprocal instruction, the precision of the sequence can be
26039 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
26041 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
26042 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
26043 combination), and doesn't need @option{-mrecip}.
26045 Also note that GCC emits the above sequence with additional Newton-Raphson step
26046 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
26047 already with @option{-ffast-math} (or the above option combination), and
26048 doesn't need @option{-mrecip}.
26050 @item -mrecip=@var{opt}
26051 @opindex mrecip=opt
26052 This option controls which reciprocal estimate instructions
26053 may be used. @var{opt} is a comma-separated list of options, which may
26054 be preceded by a @samp{!} to invert the option:
26058 Enable all estimate instructions.
26061 Enable the default instructions, equivalent to @option{-mrecip}.
26064 Disable all estimate instructions, equivalent to @option{-mno-recip}.
26067 Enable the approximation for scalar division.
26070 Enable the approximation for vectorized division.
26073 Enable the approximation for scalar square root.
26076 Enable the approximation for vectorized square root.
26079 So, for example, @option{-mrecip=all,!sqrt} enables
26080 all of the reciprocal approximations, except for square root.
26082 @item -mveclibabi=@var{type}
26083 @opindex mveclibabi
26084 Specifies the ABI type to use for vectorizing intrinsics using an
26085 external library. Supported values for @var{type} are @samp{svml}
26086 for the Intel short
26087 vector math library and @samp{acml} for the AMD math core library.
26088 To use this option, both @option{-ftree-vectorize} and
26089 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
26090 ABI-compatible library must be specified at link time.
26092 GCC currently emits calls to @code{vmldExp2},
26093 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
26094 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
26095 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
26096 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
26097 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
26098 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
26099 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
26100 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
26101 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
26102 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
26103 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
26104 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
26105 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
26106 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
26107 when @option{-mveclibabi=acml} is used.
26109 @item -mabi=@var{name}
26111 Generate code for the specified calling convention. Permissible values
26112 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
26113 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
26114 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
26115 You can control this behavior for specific functions by
26116 using the function attributes @code{ms_abi} and @code{sysv_abi}.
26117 @xref{Function Attributes}.
26119 @item -mcall-ms2sysv-xlogues
26120 @opindex mcall-ms2sysv-xlogues
26121 @opindex mno-call-ms2sysv-xlogues
26122 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
26123 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
26124 default, the code for saving and restoring these registers is emitted inline,
26125 resulting in fairly lengthy prologues and epilogues. Using
26126 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
26127 use stubs in the static portion of libgcc to perform these saves and restores,
26128 thus reducing function size at the cost of a few extra instructions.
26130 @item -mtls-dialect=@var{type}
26131 @opindex mtls-dialect
26132 Generate code to access thread-local storage using the @samp{gnu} or
26133 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
26134 @samp{gnu2} is more efficient, but it may add compile- and run-time
26135 requirements that cannot be satisfied on all systems.
26138 @itemx -mno-push-args
26139 @opindex mpush-args
26140 @opindex mno-push-args
26141 Use PUSH operations to store outgoing parameters. This method is shorter
26142 and usually equally fast as method using SUB/MOV operations and is enabled
26143 by default. In some cases disabling it may improve performance because of
26144 improved scheduling and reduced dependencies.
26146 @item -maccumulate-outgoing-args
26147 @opindex maccumulate-outgoing-args
26148 If enabled, the maximum amount of space required for outgoing arguments is
26149 computed in the function prologue. This is faster on most modern CPUs
26150 because of reduced dependencies, improved scheduling and reduced stack usage
26151 when the preferred stack boundary is not equal to 2. The drawback is a notable
26152 increase in code size. This switch implies @option{-mno-push-args}.
26156 Support thread-safe exception handling on MinGW. Programs that rely
26157 on thread-safe exception handling must compile and link all code with the
26158 @option{-mthreads} option. When compiling, @option{-mthreads} defines
26159 @option{-D_MT}; when linking, it links in a special thread helper library
26160 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
26162 @item -mms-bitfields
26163 @itemx -mno-ms-bitfields
26164 @opindex mms-bitfields
26165 @opindex mno-ms-bitfields
26167 Enable/disable bit-field layout compatible with the native Microsoft
26170 If @code{packed} is used on a structure, or if bit-fields are used,
26171 it may be that the Microsoft ABI lays out the structure differently
26172 than the way GCC normally does. Particularly when moving packed
26173 data between functions compiled with GCC and the native Microsoft compiler
26174 (either via function call or as data in a file), it may be necessary to access
26177 This option is enabled by default for Microsoft Windows
26178 targets. This behavior can also be controlled locally by use of variable
26179 or type attributes. For more information, see @ref{x86 Variable Attributes}
26180 and @ref{x86 Type Attributes}.
26182 The Microsoft structure layout algorithm is fairly simple with the exception
26183 of the bit-field packing.
26184 The padding and alignment of members of structures and whether a bit-field
26185 can straddle a storage-unit boundary are determine by these rules:
26188 @item Structure members are stored sequentially in the order in which they are
26189 declared: the first member has the lowest memory address and the last member
26192 @item Every data object has an alignment requirement. The alignment requirement
26193 for all data except structures, unions, and arrays is either the size of the
26194 object or the current packing size (specified with either the
26195 @code{aligned} attribute or the @code{pack} pragma),
26196 whichever is less. For structures, unions, and arrays,
26197 the alignment requirement is the largest alignment requirement of its members.
26198 Every object is allocated an offset so that:
26201 offset % alignment_requirement == 0
26204 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
26205 unit if the integral types are the same size and if the next bit-field fits
26206 into the current allocation unit without crossing the boundary imposed by the
26207 common alignment requirements of the bit-fields.
26210 MSVC interprets zero-length bit-fields in the following ways:
26213 @item If a zero-length bit-field is inserted between two bit-fields that
26214 are normally coalesced, the bit-fields are not coalesced.
26221 unsigned long bf_1 : 12;
26223 unsigned long bf_2 : 12;
26228 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
26229 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
26231 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
26232 alignment of the zero-length bit-field is greater than the member that follows it,
26233 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
26254 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
26255 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
26256 bit-field does not affect the alignment of @code{bar} or, as a result, the size
26259 Taking this into account, it is important to note the following:
26262 @item If a zero-length bit-field follows a normal bit-field, the type of the
26263 zero-length bit-field may affect the alignment of the structure as whole. For
26264 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
26265 normal bit-field, and is of type short.
26267 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
26268 still affect the alignment of the structure:
26279 Here, @code{t4} takes up 4 bytes.
26282 @item Zero-length bit-fields following non-bit-field members are ignored:
26294 Here, @code{t5} takes up 2 bytes.
26298 @item -mno-align-stringops
26299 @opindex mno-align-stringops
26300 Do not align the destination of inlined string operations. This switch reduces
26301 code size and improves performance in case the destination is already aligned,
26302 but GCC doesn't know about it.
26304 @item -minline-all-stringops
26305 @opindex minline-all-stringops
26306 By default GCC inlines string operations only when the destination is
26307 known to be aligned to least a 4-byte boundary.
26308 This enables more inlining and increases code
26309 size, but may improve performance of code that depends on fast
26310 @code{memcpy}, @code{strlen},
26311 and @code{memset} for short lengths.
26313 @item -minline-stringops-dynamically
26314 @opindex minline-stringops-dynamically
26315 For string operations of unknown size, use run-time checks with
26316 inline code for small blocks and a library call for large blocks.
26318 @item -mstringop-strategy=@var{alg}
26319 @opindex mstringop-strategy=@var{alg}
26320 Override the internal decision heuristic for the particular algorithm to use
26321 for inlining string operations. The allowed values for @var{alg} are:
26327 Expand using i386 @code{rep} prefix of the specified size.
26331 @itemx unrolled_loop
26332 Expand into an inline loop.
26335 Always use a library call.
26338 @item -mmemcpy-strategy=@var{strategy}
26339 @opindex mmemcpy-strategy=@var{strategy}
26340 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
26341 should be inlined and what inline algorithm to use when the expected size
26342 of the copy operation is known. @var{strategy}
26343 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
26344 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
26345 the max byte size with which inline algorithm @var{alg} is allowed. For the last
26346 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
26347 in the list must be specified in increasing order. The minimal byte size for
26348 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
26351 @item -mmemset-strategy=@var{strategy}
26352 @opindex mmemset-strategy=@var{strategy}
26353 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
26354 @code{__builtin_memset} expansion.
26356 @item -momit-leaf-frame-pointer
26357 @opindex momit-leaf-frame-pointer
26358 Don't keep the frame pointer in a register for leaf functions. This
26359 avoids the instructions to save, set up, and restore frame pointers and
26360 makes an extra register available in leaf functions. The option
26361 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
26362 which might make debugging harder.
26364 @item -mtls-direct-seg-refs
26365 @itemx -mno-tls-direct-seg-refs
26366 @opindex mtls-direct-seg-refs
26367 Controls whether TLS variables may be accessed with offsets from the
26368 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
26369 or whether the thread base pointer must be added. Whether or not this
26370 is valid depends on the operating system, and whether it maps the
26371 segment to cover the entire TLS area.
26373 For systems that use the GNU C Library, the default is on.
26376 @itemx -mno-sse2avx
26378 Specify that the assembler should encode SSE instructions with VEX
26379 prefix. The option @option{-mavx} turns this on by default.
26384 If profiling is active (@option{-pg}), put the profiling
26385 counter call before the prologue.
26386 Note: On x86 architectures the attribute @code{ms_hook_prologue}
26387 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
26389 @item -mrecord-mcount
26390 @itemx -mno-record-mcount
26391 @opindex mrecord-mcount
26392 If profiling is active (@option{-pg}), generate a __mcount_loc section
26393 that contains pointers to each profiling call. This is useful for
26394 automatically patching and out calls.
26397 @itemx -mno-nop-mcount
26398 @opindex mnop-mcount
26399 If profiling is active (@option{-pg}), generate the calls to
26400 the profiling functions as NOPs. This is useful when they
26401 should be patched in later dynamically. This is likely only
26402 useful together with @option{-mrecord-mcount}.
26404 @item -mskip-rax-setup
26405 @itemx -mno-skip-rax-setup
26406 @opindex mskip-rax-setup
26407 When generating code for the x86-64 architecture with SSE extensions
26408 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
26409 register when there are no variable arguments passed in vector registers.
26411 @strong{Warning:} Since RAX register is used to avoid unnecessarily
26412 saving vector registers on stack when passing variable arguments, the
26413 impacts of this option are callees may waste some stack space,
26414 misbehave or jump to a random location. GCC 4.4 or newer don't have
26415 those issues, regardless the RAX register value.
26418 @itemx -mno-8bit-idiv
26419 @opindex m8bit-idiv
26420 On some processors, like Intel Atom, 8-bit unsigned integer divide is
26421 much faster than 32-bit/64-bit integer divide. This option generates a
26422 run-time check. If both dividend and divisor are within range of 0
26423 to 255, 8-bit unsigned integer divide is used instead of
26424 32-bit/64-bit integer divide.
26426 @item -mavx256-split-unaligned-load
26427 @itemx -mavx256-split-unaligned-store
26428 @opindex mavx256-split-unaligned-load
26429 @opindex mavx256-split-unaligned-store
26430 Split 32-byte AVX unaligned load and store.
26432 @item -mstack-protector-guard=@var{guard}
26433 @itemx -mstack-protector-guard-reg=@var{reg}
26434 @itemx -mstack-protector-guard-offset=@var{offset}
26435 @opindex mstack-protector-guard
26436 @opindex mstack-protector-guard-reg
26437 @opindex mstack-protector-guard-offset
26438 Generate stack protection code using canary at @var{guard}. Supported
26439 locations are @samp{global} for global canary or @samp{tls} for per-thread
26440 canary in the TLS block (the default). This option has effect only when
26441 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
26443 With the latter choice the options
26444 @option{-mstack-protector-guard-reg=@var{reg}} and
26445 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
26446 which segment register (@code{%fs} or @code{%gs}) to use as base register
26447 for reading the canary, and from what offset from that base register.
26448 The default for those is as specified in the relevant ABI.
26450 @item -mmitigate-rop
26451 @opindex mmitigate-rop
26452 Try to avoid generating code sequences that contain unintended return
26453 opcodes, to mitigate against certain forms of attack. At the moment,
26454 this option is limited in what it can do and should not be relied
26455 on to provide serious protection.
26457 @item -mgeneral-regs-only
26458 @opindex mgeneral-regs-only
26459 Generate code that uses only the general-purpose registers. This
26460 prevents the compiler from using floating-point, vector, mask and bound
26465 These @samp{-m} switches are supported in addition to the above
26466 on x86-64 processors in 64-bit environments.
26479 Generate code for a 16-bit, 32-bit or 64-bit environment.
26480 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
26482 generates code that runs on any i386 system.
26484 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
26485 types to 64 bits, and generates code for the x86-64 architecture.
26486 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
26487 and @option{-mdynamic-no-pic} options.
26489 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
26491 generates code for the x86-64 architecture.
26493 The @option{-m16} option is the same as @option{-m32}, except for that
26494 it outputs the @code{.code16gcc} assembly directive at the beginning of
26495 the assembly output so that the binary can run in 16-bit mode.
26497 The @option{-miamcu} option generates code which conforms to Intel MCU
26498 psABI. It requires the @option{-m32} option to be turned on.
26500 @item -mno-red-zone
26501 @opindex mno-red-zone
26502 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
26503 by the x86-64 ABI; it is a 128-byte area beyond the location of the
26504 stack pointer that is not modified by signal or interrupt handlers
26505 and therefore can be used for temporary data without adjusting the stack
26506 pointer. The flag @option{-mno-red-zone} disables this red zone.
26508 @item -mcmodel=small
26509 @opindex mcmodel=small
26510 Generate code for the small code model: the program and its symbols must
26511 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
26512 Programs can be statically or dynamically linked. This is the default
26515 @item -mcmodel=kernel
26516 @opindex mcmodel=kernel
26517 Generate code for the kernel code model. The kernel runs in the
26518 negative 2 GB of the address space.
26519 This model has to be used for Linux kernel code.
26521 @item -mcmodel=medium
26522 @opindex mcmodel=medium
26523 Generate code for the medium model: the program is linked in the lower 2
26524 GB of the address space. Small symbols are also placed there. Symbols
26525 with sizes larger than @option{-mlarge-data-threshold} are put into
26526 large data or BSS sections and can be located above 2GB. Programs can
26527 be statically or dynamically linked.
26529 @item -mcmodel=large
26530 @opindex mcmodel=large
26531 Generate code for the large model. This model makes no assumptions
26532 about addresses and sizes of sections.
26534 @item -maddress-mode=long
26535 @opindex maddress-mode=long
26536 Generate code for long address mode. This is only supported for 64-bit
26537 and x32 environments. It is the default address mode for 64-bit
26540 @item -maddress-mode=short
26541 @opindex maddress-mode=short
26542 Generate code for short address mode. This is only supported for 32-bit
26543 and x32 environments. It is the default address mode for 32-bit and
26547 @node x86 Windows Options
26548 @subsection x86 Windows Options
26549 @cindex x86 Windows Options
26550 @cindex Windows Options for x86
26552 These additional options are available for Microsoft Windows targets:
26558 specifies that a console application is to be generated, by
26559 instructing the linker to set the PE header subsystem type
26560 required for console applications.
26561 This option is available for Cygwin and MinGW targets and is
26562 enabled by default on those targets.
26566 This option is available for Cygwin and MinGW targets. It
26567 specifies that a DLL---a dynamic link library---is to be
26568 generated, enabling the selection of the required runtime
26569 startup object and entry point.
26571 @item -mnop-fun-dllimport
26572 @opindex mnop-fun-dllimport
26573 This option is available for Cygwin and MinGW targets. It
26574 specifies that the @code{dllimport} attribute should be ignored.
26578 This option is available for MinGW targets. It specifies
26579 that MinGW-specific thread support is to be used.
26583 This option is available for MinGW-w64 targets. It causes
26584 the @code{UNICODE} preprocessor macro to be predefined, and
26585 chooses Unicode-capable runtime startup code.
26589 This option is available for Cygwin and MinGW targets. It
26590 specifies that the typical Microsoft Windows predefined macros are to
26591 be set in the pre-processor, but does not influence the choice
26592 of runtime library/startup code.
26596 This option is available for Cygwin and MinGW targets. It
26597 specifies that a GUI application is to be generated by
26598 instructing the linker to set the PE header subsystem type
26601 @item -fno-set-stack-executable
26602 @opindex fno-set-stack-executable
26603 This option is available for MinGW targets. It specifies that
26604 the executable flag for the stack used by nested functions isn't
26605 set. This is necessary for binaries running in kernel mode of
26606 Microsoft Windows, as there the User32 API, which is used to set executable
26607 privileges, isn't available.
26609 @item -fwritable-relocated-rdata
26610 @opindex fno-writable-relocated-rdata
26611 This option is available for MinGW and Cygwin targets. It specifies
26612 that relocated-data in read-only section is put into the @code{.data}
26613 section. This is a necessary for older runtimes not supporting
26614 modification of @code{.rdata} sections for pseudo-relocation.
26616 @item -mpe-aligned-commons
26617 @opindex mpe-aligned-commons
26618 This option is available for Cygwin and MinGW targets. It
26619 specifies that the GNU extension to the PE file format that
26620 permits the correct alignment of COMMON variables should be
26621 used when generating code. It is enabled by default if
26622 GCC detects that the target assembler found during configuration
26623 supports the feature.
26626 See also under @ref{x86 Options} for standard options.
26628 @node Xstormy16 Options
26629 @subsection Xstormy16 Options
26630 @cindex Xstormy16 Options
26632 These options are defined for Xstormy16:
26637 Choose startup files and linker script suitable for the simulator.
26640 @node Xtensa Options
26641 @subsection Xtensa Options
26642 @cindex Xtensa Options
26644 These options are supported for Xtensa targets:
26648 @itemx -mno-const16
26650 @opindex mno-const16
26651 Enable or disable use of @code{CONST16} instructions for loading
26652 constant values. The @code{CONST16} instruction is currently not a
26653 standard option from Tensilica. When enabled, @code{CONST16}
26654 instructions are always used in place of the standard @code{L32R}
26655 instructions. The use of @code{CONST16} is enabled by default only if
26656 the @code{L32R} instruction is not available.
26659 @itemx -mno-fused-madd
26660 @opindex mfused-madd
26661 @opindex mno-fused-madd
26662 Enable or disable use of fused multiply/add and multiply/subtract
26663 instructions in the floating-point option. This has no effect if the
26664 floating-point option is not also enabled. Disabling fused multiply/add
26665 and multiply/subtract instructions forces the compiler to use separate
26666 instructions for the multiply and add/subtract operations. This may be
26667 desirable in some cases where strict IEEE 754-compliant results are
26668 required: the fused multiply add/subtract instructions do not round the
26669 intermediate result, thereby producing results with @emph{more} bits of
26670 precision than specified by the IEEE standard. Disabling fused multiply
26671 add/subtract instructions also ensures that the program output is not
26672 sensitive to the compiler's ability to combine multiply and add/subtract
26675 @item -mserialize-volatile
26676 @itemx -mno-serialize-volatile
26677 @opindex mserialize-volatile
26678 @opindex mno-serialize-volatile
26679 When this option is enabled, GCC inserts @code{MEMW} instructions before
26680 @code{volatile} memory references to guarantee sequential consistency.
26681 The default is @option{-mserialize-volatile}. Use
26682 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
26684 @item -mforce-no-pic
26685 @opindex mforce-no-pic
26686 For targets, like GNU/Linux, where all user-mode Xtensa code must be
26687 position-independent code (PIC), this option disables PIC for compiling
26690 @item -mtext-section-literals
26691 @itemx -mno-text-section-literals
26692 @opindex mtext-section-literals
26693 @opindex mno-text-section-literals
26694 These options control the treatment of literal pools. The default is
26695 @option{-mno-text-section-literals}, which places literals in a separate
26696 section in the output file. This allows the literal pool to be placed
26697 in a data RAM/ROM, and it also allows the linker to combine literal
26698 pools from separate object files to remove redundant literals and
26699 improve code size. With @option{-mtext-section-literals}, the literals
26700 are interspersed in the text section in order to keep them as close as
26701 possible to their references. This may be necessary for large assembly
26702 files. Literals for each function are placed right before that function.
26704 @item -mauto-litpools
26705 @itemx -mno-auto-litpools
26706 @opindex mauto-litpools
26707 @opindex mno-auto-litpools
26708 These options control the treatment of literal pools. The default is
26709 @option{-mno-auto-litpools}, which places literals in a separate
26710 section in the output file unless @option{-mtext-section-literals} is
26711 used. With @option{-mauto-litpools} the literals are interspersed in
26712 the text section by the assembler. Compiler does not produce explicit
26713 @code{.literal} directives and loads literals into registers with
26714 @code{MOVI} instructions instead of @code{L32R} to let the assembler
26715 do relaxation and place literals as necessary. This option allows
26716 assembler to create several literal pools per function and assemble
26717 very big functions, which may not be possible with
26718 @option{-mtext-section-literals}.
26720 @item -mtarget-align
26721 @itemx -mno-target-align
26722 @opindex mtarget-align
26723 @opindex mno-target-align
26724 When this option is enabled, GCC instructs the assembler to
26725 automatically align instructions to reduce branch penalties at the
26726 expense of some code density. The assembler attempts to widen density
26727 instructions to align branch targets and the instructions following call
26728 instructions. If there are not enough preceding safe density
26729 instructions to align a target, no widening is performed. The
26730 default is @option{-mtarget-align}. These options do not affect the
26731 treatment of auto-aligned instructions like @code{LOOP}, which the
26732 assembler always aligns, either by widening density instructions or
26733 by inserting NOP instructions.
26736 @itemx -mno-longcalls
26737 @opindex mlongcalls
26738 @opindex mno-longcalls
26739 When this option is enabled, GCC instructs the assembler to translate
26740 direct calls to indirect calls unless it can determine that the target
26741 of a direct call is in the range allowed by the call instruction. This
26742 translation typically occurs for calls to functions in other source
26743 files. Specifically, the assembler translates a direct @code{CALL}
26744 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
26745 The default is @option{-mno-longcalls}. This option should be used in
26746 programs where the call target can potentially be out of range. This
26747 option is implemented in the assembler, not the compiler, so the
26748 assembly code generated by GCC still shows direct call
26749 instructions---look at the disassembled object code to see the actual
26750 instructions. Note that the assembler uses an indirect call for
26751 every cross-file call, not just those that really are out of range.
26754 @node zSeries Options
26755 @subsection zSeries Options
26756 @cindex zSeries options
26758 These are listed under @xref{S/390 and zSeries Options}.
26764 @section Specifying Subprocesses and the Switches to Pass to Them
26767 @command{gcc} is a driver program. It performs its job by invoking a
26768 sequence of other programs to do the work of compiling, assembling and
26769 linking. GCC interprets its command-line parameters and uses these to
26770 deduce which programs it should invoke, and which command-line options
26771 it ought to place on their command lines. This behavior is controlled
26772 by @dfn{spec strings}. In most cases there is one spec string for each
26773 program that GCC can invoke, but a few programs have multiple spec
26774 strings to control their behavior. The spec strings built into GCC can
26775 be overridden by using the @option{-specs=} command-line switch to specify
26778 @dfn{Spec files} are plain-text files that are used to construct spec
26779 strings. They consist of a sequence of directives separated by blank
26780 lines. The type of directive is determined by the first non-whitespace
26781 character on the line, which can be one of the following:
26784 @item %@var{command}
26785 Issues a @var{command} to the spec file processor. The commands that can
26789 @item %include <@var{file}>
26790 @cindex @code{%include}
26791 Search for @var{file} and insert its text at the current point in the
26794 @item %include_noerr <@var{file}>
26795 @cindex @code{%include_noerr}
26796 Just like @samp{%include}, but do not generate an error message if the include
26797 file cannot be found.
26799 @item %rename @var{old_name} @var{new_name}
26800 @cindex @code{%rename}
26801 Rename the spec string @var{old_name} to @var{new_name}.
26805 @item *[@var{spec_name}]:
26806 This tells the compiler to create, override or delete the named spec
26807 string. All lines after this directive up to the next directive or
26808 blank line are considered to be the text for the spec string. If this
26809 results in an empty string then the spec is deleted. (Or, if the
26810 spec did not exist, then nothing happens.) Otherwise, if the spec
26811 does not currently exist a new spec is created. If the spec does
26812 exist then its contents are overridden by the text of this
26813 directive, unless the first character of that text is the @samp{+}
26814 character, in which case the text is appended to the spec.
26816 @item [@var{suffix}]:
26817 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
26818 and up to the next directive or blank line are considered to make up the
26819 spec string for the indicated suffix. When the compiler encounters an
26820 input file with the named suffix, it processes the spec string in
26821 order to work out how to compile that file. For example:
26825 z-compile -input %i
26828 This says that any input file whose name ends in @samp{.ZZ} should be
26829 passed to the program @samp{z-compile}, which should be invoked with the
26830 command-line switch @option{-input} and with the result of performing the
26831 @samp{%i} substitution. (See below.)
26833 As an alternative to providing a spec string, the text following a
26834 suffix directive can be one of the following:
26837 @item @@@var{language}
26838 This says that the suffix is an alias for a known @var{language}. This is
26839 similar to using the @option{-x} command-line switch to GCC to specify a
26840 language explicitly. For example:
26847 Says that .ZZ files are, in fact, C++ source files.
26850 This causes an error messages saying:
26853 @var{name} compiler not installed on this system.
26857 GCC already has an extensive list of suffixes built into it.
26858 This directive adds an entry to the end of the list of suffixes, but
26859 since the list is searched from the end backwards, it is effectively
26860 possible to override earlier entries using this technique.
26864 GCC has the following spec strings built into it. Spec files can
26865 override these strings or create their own. Note that individual
26866 targets can also add their own spec strings to this list.
26869 asm Options to pass to the assembler
26870 asm_final Options to pass to the assembler post-processor
26871 cpp Options to pass to the C preprocessor
26872 cc1 Options to pass to the C compiler
26873 cc1plus Options to pass to the C++ compiler
26874 endfile Object files to include at the end of the link
26875 link Options to pass to the linker
26876 lib Libraries to include on the command line to the linker
26877 libgcc Decides which GCC support library to pass to the linker
26878 linker Sets the name of the linker
26879 predefines Defines to be passed to the C preprocessor
26880 signed_char Defines to pass to CPP to say whether @code{char} is signed
26882 startfile Object files to include at the start of the link
26885 Here is a small example of a spec file:
26888 %rename lib old_lib
26891 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
26894 This example renames the spec called @samp{lib} to @samp{old_lib} and
26895 then overrides the previous definition of @samp{lib} with a new one.
26896 The new definition adds in some extra command-line options before
26897 including the text of the old definition.
26899 @dfn{Spec strings} are a list of command-line options to be passed to their
26900 corresponding program. In addition, the spec strings can contain
26901 @samp{%}-prefixed sequences to substitute variable text or to
26902 conditionally insert text into the command line. Using these constructs
26903 it is possible to generate quite complex command lines.
26905 Here is a table of all defined @samp{%}-sequences for spec
26906 strings. Note that spaces are not generated automatically around the
26907 results of expanding these sequences. Therefore you can concatenate them
26908 together or combine them with constant text in a single argument.
26912 Substitute one @samp{%} into the program name or argument.
26915 Substitute the name of the input file being processed.
26918 Substitute the basename of the input file being processed.
26919 This is the substring up to (and not including) the last period
26920 and not including the directory.
26923 This is the same as @samp{%b}, but include the file suffix (text after
26927 Marks the argument containing or following the @samp{%d} as a
26928 temporary file name, so that that file is deleted if GCC exits
26929 successfully. Unlike @samp{%g}, this contributes no text to the
26932 @item %g@var{suffix}
26933 Substitute a file name that has suffix @var{suffix} and is chosen
26934 once per compilation, and mark the argument in the same way as
26935 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
26936 name is now chosen in a way that is hard to predict even when previously
26937 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
26938 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
26939 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
26940 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
26941 was simply substituted with a file name chosen once per compilation,
26942 without regard to any appended suffix (which was therefore treated
26943 just like ordinary text), making such attacks more likely to succeed.
26945 @item %u@var{suffix}
26946 Like @samp{%g}, but generates a new temporary file name
26947 each time it appears instead of once per compilation.
26949 @item %U@var{suffix}
26950 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
26951 new one if there is no such last file name. In the absence of any
26952 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
26953 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
26954 involves the generation of two distinct file names, one
26955 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
26956 simply substituted with a file name chosen for the previous @samp{%u},
26957 without regard to any appended suffix.
26959 @item %j@var{suffix}
26960 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
26961 writable, and if @option{-save-temps} is not used;
26962 otherwise, substitute the name
26963 of a temporary file, just like @samp{%u}. This temporary file is not
26964 meant for communication between processes, but rather as a junk
26965 disposal mechanism.
26967 @item %|@var{suffix}
26968 @itemx %m@var{suffix}
26969 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
26970 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
26971 all. These are the two most common ways to instruct a program that it
26972 should read from standard input or write to standard output. If you
26973 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
26974 construct: see for example @file{f/lang-specs.h}.
26976 @item %.@var{SUFFIX}
26977 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
26978 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
26979 terminated by the next space or %.
26982 Marks the argument containing or following the @samp{%w} as the
26983 designated output file of this compilation. This puts the argument
26984 into the sequence of arguments that @samp{%o} substitutes.
26987 Substitutes the names of all the output files, with spaces
26988 automatically placed around them. You should write spaces
26989 around the @samp{%o} as well or the results are undefined.
26990 @samp{%o} is for use in the specs for running the linker.
26991 Input files whose names have no recognized suffix are not compiled
26992 at all, but they are included among the output files, so they are
26996 Substitutes the suffix for object files. Note that this is
26997 handled specially when it immediately follows @samp{%g, %u, or %U},
26998 because of the need for those to form complete file names. The
26999 handling is such that @samp{%O} is treated exactly as if it had already
27000 been substituted, except that @samp{%g, %u, and %U} do not currently
27001 support additional @var{suffix} characters following @samp{%O} as they do
27002 following, for example, @samp{.o}.
27005 Substitutes the standard macro predefinitions for the
27006 current target machine. Use this when running @command{cpp}.
27009 Like @samp{%p}, but puts @samp{__} before and after the name of each
27010 predefined macro, except for macros that start with @samp{__} or with
27011 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
27015 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
27016 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
27017 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
27018 and @option{-imultilib} as necessary.
27021 Current argument is the name of a library or startup file of some sort.
27022 Search for that file in a standard list of directories and substitute
27023 the full name found. The current working directory is included in the
27024 list of directories scanned.
27027 Current argument is the name of a linker script. Search for that file
27028 in the current list of directories to scan for libraries. If the file
27029 is located insert a @option{--script} option into the command line
27030 followed by the full path name found. If the file is not found then
27031 generate an error message. Note: the current working directory is not
27035 Print @var{str} as an error message. @var{str} is terminated by a newline.
27036 Use this when inconsistent options are detected.
27038 @item %(@var{name})
27039 Substitute the contents of spec string @var{name} at this point.
27041 @item %x@{@var{option}@}
27042 Accumulate an option for @samp{%X}.
27045 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
27049 Output the accumulated assembler options specified by @option{-Wa}.
27052 Output the accumulated preprocessor options specified by @option{-Wp}.
27055 Process the @code{asm} spec. This is used to compute the
27056 switches to be passed to the assembler.
27059 Process the @code{asm_final} spec. This is a spec string for
27060 passing switches to an assembler post-processor, if such a program is
27064 Process the @code{link} spec. This is the spec for computing the
27065 command line passed to the linker. Typically it makes use of the
27066 @samp{%L %G %S %D and %E} sequences.
27069 Dump out a @option{-L} option for each directory that GCC believes might
27070 contain startup files. If the target supports multilibs then the
27071 current multilib directory is prepended to each of these paths.
27074 Process the @code{lib} spec. This is a spec string for deciding which
27075 libraries are included on the command line to the linker.
27078 Process the @code{libgcc} spec. This is a spec string for deciding
27079 which GCC support library is included on the command line to the linker.
27082 Process the @code{startfile} spec. This is a spec for deciding which
27083 object files are the first ones passed to the linker. Typically
27084 this might be a file named @file{crt0.o}.
27087 Process the @code{endfile} spec. This is a spec string that specifies
27088 the last object files that are passed to the linker.
27091 Process the @code{cpp} spec. This is used to construct the arguments
27092 to be passed to the C preprocessor.
27095 Process the @code{cc1} spec. This is used to construct the options to be
27096 passed to the actual C compiler (@command{cc1}).
27099 Process the @code{cc1plus} spec. This is used to construct the options to be
27100 passed to the actual C++ compiler (@command{cc1plus}).
27103 Substitute the variable part of a matched option. See below.
27104 Note that each comma in the substituted string is replaced by
27108 Remove all occurrences of @code{-S} from the command line. Note---this
27109 command is position dependent. @samp{%} commands in the spec string
27110 before this one see @code{-S}, @samp{%} commands in the spec string
27111 after this one do not.
27113 @item %:@var{function}(@var{args})
27114 Call the named function @var{function}, passing it @var{args}.
27115 @var{args} is first processed as a nested spec string, then split
27116 into an argument vector in the usual fashion. The function returns
27117 a string which is processed as if it had appeared literally as part
27118 of the current spec.
27120 The following built-in spec functions are provided:
27123 @item @code{getenv}
27124 The @code{getenv} spec function takes two arguments: an environment
27125 variable name and a string. If the environment variable is not
27126 defined, a fatal error is issued. Otherwise, the return value is the
27127 value of the environment variable concatenated with the string. For
27128 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
27131 %:getenv(TOPDIR /include)
27134 expands to @file{/path/to/top/include}.
27136 @item @code{if-exists}
27137 The @code{if-exists} spec function takes one argument, an absolute
27138 pathname to a file. If the file exists, @code{if-exists} returns the
27139 pathname. Here is a small example of its usage:
27143 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
27146 @item @code{if-exists-else}
27147 The @code{if-exists-else} spec function is similar to the @code{if-exists}
27148 spec function, except that it takes two arguments. The first argument is
27149 an absolute pathname to a file. If the file exists, @code{if-exists-else}
27150 returns the pathname. If it does not exist, it returns the second argument.
27151 This way, @code{if-exists-else} can be used to select one file or another,
27152 based on the existence of the first. Here is a small example of its usage:
27156 crt0%O%s %:if-exists(crti%O%s) \
27157 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
27160 @item @code{replace-outfile}
27161 The @code{replace-outfile} spec function takes two arguments. It looks for the
27162 first argument in the outfiles array and replaces it with the second argument. Here
27163 is a small example of its usage:
27166 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
27169 @item @code{remove-outfile}
27170 The @code{remove-outfile} spec function takes one argument. It looks for the
27171 first argument in the outfiles array and removes it. Here is a small example
27175 %:remove-outfile(-lm)
27178 @item @code{pass-through-libs}
27179 The @code{pass-through-libs} spec function takes any number of arguments. It
27180 finds any @option{-l} options and any non-options ending in @file{.a} (which it
27181 assumes are the names of linker input library archive files) and returns a
27182 result containing all the found arguments each prepended by
27183 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
27184 intended to be passed to the LTO linker plugin.
27187 %:pass-through-libs(%G %L %G)
27190 @item @code{print-asm-header}
27191 The @code{print-asm-header} function takes no arguments and simply
27192 prints a banner like:
27198 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
27201 It is used to separate compiler options from assembler options
27202 in the @option{--target-help} output.
27206 Substitutes the @code{-S} switch, if that switch is given to GCC@.
27207 If that switch is not specified, this substitutes nothing. Note that
27208 the leading dash is omitted when specifying this option, and it is
27209 automatically inserted if the substitution is performed. Thus the spec
27210 string @samp{%@{foo@}} matches the command-line option @option{-foo}
27211 and outputs the command-line option @option{-foo}.
27214 Like %@{@code{S}@} but mark last argument supplied within as a file to be
27215 deleted on failure.
27218 Substitutes all the switches specified to GCC whose names start
27219 with @code{-S}, but which also take an argument. This is used for
27220 switches like @option{-o}, @option{-D}, @option{-I}, etc.
27221 GCC considers @option{-o foo} as being
27222 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
27223 text, including the space. Thus two arguments are generated.
27226 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
27227 (the order of @code{S} and @code{T} in the spec is not significant).
27228 There can be any number of ampersand-separated variables; for each the
27229 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
27232 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
27235 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
27238 Substitutes @code{X} if one or more switches whose names start with
27239 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
27240 once, no matter how many such switches appeared. However, if @code{%*}
27241 appears somewhere in @code{X}, then @code{X} is substituted once
27242 for each matching switch, with the @code{%*} replaced by the part of
27243 that switch matching the @code{*}.
27245 If @code{%*} appears as the last part of a spec sequence then a space
27246 is added after the end of the last substitution. If there is more
27247 text in the sequence, however, then a space is not generated. This
27248 allows the @code{%*} substitution to be used as part of a larger
27249 string. For example, a spec string like this:
27252 %@{mcu=*:--script=%*/memory.ld@}
27256 when matching an option like @option{-mcu=newchip} produces:
27259 --script=newchip/memory.ld
27263 Substitutes @code{X}, if processing a file with suffix @code{S}.
27266 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
27269 Substitutes @code{X}, if processing a file for language @code{S}.
27272 Substitutes @code{X}, if not processing a file for language @code{S}.
27275 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
27276 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
27277 @code{*} sequences as well, although they have a stronger binding than
27278 the @samp{|}. If @code{%*} appears in @code{X}, all of the
27279 alternatives must be starred, and only the first matching alternative
27282 For example, a spec string like this:
27285 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
27289 outputs the following command-line options from the following input
27290 command-line options:
27295 -d fred.c -foo -baz -boggle
27296 -d jim.d -bar -baz -boggle
27299 @item %@{S:X; T:Y; :D@}
27301 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
27302 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
27303 be as many clauses as you need. This may be combined with @code{.},
27304 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
27309 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
27310 or similar construct can use a backslash to ignore the special meaning
27311 of the character following it, thus allowing literal matching of a
27312 character that is otherwise specially treated. For example,
27313 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
27314 @option{-std=iso9899:1999} option is given.
27316 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
27317 construct may contain other nested @samp{%} constructs or spaces, or
27318 even newlines. They are processed as usual, as described above.
27319 Trailing white space in @code{X} is ignored. White space may also
27320 appear anywhere on the left side of the colon in these constructs,
27321 except between @code{.} or @code{*} and the corresponding word.
27323 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
27324 handled specifically in these constructs. If another value of
27325 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
27326 @option{-W} switch is found later in the command line, the earlier
27327 switch value is ignored, except with @{@code{S}*@} where @code{S} is
27328 just one letter, which passes all matching options.
27330 The character @samp{|} at the beginning of the predicate text is used to
27331 indicate that a command should be piped to the following command, but
27332 only if @option{-pipe} is specified.
27334 It is built into GCC which switches take arguments and which do not.
27335 (You might think it would be useful to generalize this to allow each
27336 compiler's spec to say which switches take arguments. But this cannot
27337 be done in a consistent fashion. GCC cannot even decide which input
27338 files have been specified without knowing which switches take arguments,
27339 and it must know which input files to compile in order to tell which
27342 GCC also knows implicitly that arguments starting in @option{-l} are to be
27343 treated as compiler output files, and passed to the linker in their
27344 proper position among the other output files.
27346 @node Environment Variables
27347 @section Environment Variables Affecting GCC
27348 @cindex environment variables
27350 @c man begin ENVIRONMENT
27351 This section describes several environment variables that affect how GCC
27352 operates. Some of them work by specifying directories or prefixes to use
27353 when searching for various kinds of files. Some are used to specify other
27354 aspects of the compilation environment.
27356 Note that you can also specify places to search using options such as
27357 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
27358 take precedence over places specified using environment variables, which
27359 in turn take precedence over those specified by the configuration of GCC@.
27360 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
27361 GNU Compiler Collection (GCC) Internals}.
27366 @c @itemx LC_COLLATE
27368 @c @itemx LC_MONETARY
27369 @c @itemx LC_NUMERIC
27374 @c @findex LC_COLLATE
27375 @findex LC_MESSAGES
27376 @c @findex LC_MONETARY
27377 @c @findex LC_NUMERIC
27381 These environment variables control the way that GCC uses
27382 localization information which allows GCC to work with different
27383 national conventions. GCC inspects the locale categories
27384 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
27385 so. These locale categories can be set to any value supported by your
27386 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
27387 Kingdom encoded in UTF-8.
27389 The @env{LC_CTYPE} environment variable specifies character
27390 classification. GCC uses it to determine the character boundaries in
27391 a string; this is needed for some multibyte encodings that contain quote
27392 and escape characters that are otherwise interpreted as a string
27395 The @env{LC_MESSAGES} environment variable specifies the language to
27396 use in diagnostic messages.
27398 If the @env{LC_ALL} environment variable is set, it overrides the value
27399 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
27400 and @env{LC_MESSAGES} default to the value of the @env{LANG}
27401 environment variable. If none of these variables are set, GCC
27402 defaults to traditional C English behavior.
27406 If @env{TMPDIR} is set, it specifies the directory to use for temporary
27407 files. GCC uses temporary files to hold the output of one stage of
27408 compilation which is to be used as input to the next stage: for example,
27409 the output of the preprocessor, which is the input to the compiler
27412 @item GCC_COMPARE_DEBUG
27413 @findex GCC_COMPARE_DEBUG
27414 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
27415 @option{-fcompare-debug} to the compiler driver. See the documentation
27416 of this option for more details.
27418 @item GCC_EXEC_PREFIX
27419 @findex GCC_EXEC_PREFIX
27420 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
27421 names of the subprograms executed by the compiler. No slash is added
27422 when this prefix is combined with the name of a subprogram, but you can
27423 specify a prefix that ends with a slash if you wish.
27425 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
27426 an appropriate prefix to use based on the pathname it is invoked with.
27428 If GCC cannot find the subprogram using the specified prefix, it
27429 tries looking in the usual places for the subprogram.
27431 The default value of @env{GCC_EXEC_PREFIX} is
27432 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
27433 the installed compiler. In many cases @var{prefix} is the value
27434 of @code{prefix} when you ran the @file{configure} script.
27436 Other prefixes specified with @option{-B} take precedence over this prefix.
27438 This prefix is also used for finding files such as @file{crt0.o} that are
27441 In addition, the prefix is used in an unusual way in finding the
27442 directories to search for header files. For each of the standard
27443 directories whose name normally begins with @samp{/usr/local/lib/gcc}
27444 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
27445 replacing that beginning with the specified prefix to produce an
27446 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
27447 @file{foo/bar} just before it searches the standard directory
27448 @file{/usr/local/lib/bar}.
27449 If a standard directory begins with the configured
27450 @var{prefix} then the value of @var{prefix} is replaced by
27451 @env{GCC_EXEC_PREFIX} when looking for header files.
27453 @item COMPILER_PATH
27454 @findex COMPILER_PATH
27455 The value of @env{COMPILER_PATH} is a colon-separated list of
27456 directories, much like @env{PATH}. GCC tries the directories thus
27457 specified when searching for subprograms, if it cannot find the
27458 subprograms using @env{GCC_EXEC_PREFIX}.
27461 @findex LIBRARY_PATH
27462 The value of @env{LIBRARY_PATH} is a colon-separated list of
27463 directories, much like @env{PATH}. When configured as a native compiler,
27464 GCC tries the directories thus specified when searching for special
27465 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
27466 using GCC also uses these directories when searching for ordinary
27467 libraries for the @option{-l} option (but directories specified with
27468 @option{-L} come first).
27472 @cindex locale definition
27473 This variable is used to pass locale information to the compiler. One way in
27474 which this information is used is to determine the character set to be used
27475 when character literals, string literals and comments are parsed in C and C++.
27476 When the compiler is configured to allow multibyte characters,
27477 the following values for @env{LANG} are recognized:
27481 Recognize JIS characters.
27483 Recognize SJIS characters.
27485 Recognize EUCJP characters.
27488 If @env{LANG} is not defined, or if it has some other value, then the
27489 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
27490 recognize and translate multibyte characters.
27494 Some additional environment variables affect the behavior of the
27497 @include cppenv.texi
27501 @node Precompiled Headers
27502 @section Using Precompiled Headers
27503 @cindex precompiled headers
27504 @cindex speed of compilation
27506 Often large projects have many header files that are included in every
27507 source file. The time the compiler takes to process these header files
27508 over and over again can account for nearly all of the time required to
27509 build the project. To make builds faster, GCC allows you to
27510 @dfn{precompile} a header file.
27512 To create a precompiled header file, simply compile it as you would any
27513 other file, if necessary using the @option{-x} option to make the driver
27514 treat it as a C or C++ header file. You may want to use a
27515 tool like @command{make} to keep the precompiled header up-to-date when
27516 the headers it contains change.
27518 A precompiled header file is searched for when @code{#include} is
27519 seen in the compilation. As it searches for the included file
27520 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
27521 compiler looks for a precompiled header in each directory just before it
27522 looks for the include file in that directory. The name searched for is
27523 the name specified in the @code{#include} with @samp{.gch} appended. If
27524 the precompiled header file cannot be used, it is ignored.
27526 For instance, if you have @code{#include "all.h"}, and you have
27527 @file{all.h.gch} in the same directory as @file{all.h}, then the
27528 precompiled header file is used if possible, and the original
27529 header is used otherwise.
27531 Alternatively, you might decide to put the precompiled header file in a
27532 directory and use @option{-I} to ensure that directory is searched
27533 before (or instead of) the directory containing the original header.
27534 Then, if you want to check that the precompiled header file is always
27535 used, you can put a file of the same name as the original header in this
27536 directory containing an @code{#error} command.
27538 This also works with @option{-include}. So yet another way to use
27539 precompiled headers, good for projects not designed with precompiled
27540 header files in mind, is to simply take most of the header files used by
27541 a project, include them from another header file, precompile that header
27542 file, and @option{-include} the precompiled header. If the header files
27543 have guards against multiple inclusion, they are skipped because
27544 they've already been included (in the precompiled header).
27546 If you need to precompile the same header file for different
27547 languages, targets, or compiler options, you can instead make a
27548 @emph{directory} named like @file{all.h.gch}, and put each precompiled
27549 header in the directory, perhaps using @option{-o}. It doesn't matter
27550 what you call the files in the directory; every precompiled header in
27551 the directory is considered. The first precompiled header
27552 encountered in the directory that is valid for this compilation is
27553 used; they're searched in no particular order.
27555 There are many other possibilities, limited only by your imagination,
27556 good sense, and the constraints of your build system.
27558 A precompiled header file can be used only when these conditions apply:
27562 Only one precompiled header can be used in a particular compilation.
27565 A precompiled header cannot be used once the first C token is seen. You
27566 can have preprocessor directives before a precompiled header; you cannot
27567 include a precompiled header from inside another header.
27570 The precompiled header file must be produced for the same language as
27571 the current compilation. You cannot use a C precompiled header for a C++
27575 The precompiled header file must have been produced by the same compiler
27576 binary as the current compilation is using.
27579 Any macros defined before the precompiled header is included must
27580 either be defined in the same way as when the precompiled header was
27581 generated, or must not affect the precompiled header, which usually
27582 means that they don't appear in the precompiled header at all.
27584 The @option{-D} option is one way to define a macro before a
27585 precompiled header is included; using a @code{#define} can also do it.
27586 There are also some options that define macros implicitly, like
27587 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
27590 @item If debugging information is output when using the precompiled
27591 header, using @option{-g} or similar, the same kind of debugging information
27592 must have been output when building the precompiled header. However,
27593 a precompiled header built using @option{-g} can be used in a compilation
27594 when no debugging information is being output.
27596 @item The same @option{-m} options must generally be used when building
27597 and using the precompiled header. @xref{Submodel Options},
27598 for any cases where this rule is relaxed.
27600 @item Each of the following options must be the same when building and using
27601 the precompiled header:
27603 @gccoptlist{-fexceptions}
27606 Some other command-line options starting with @option{-f},
27607 @option{-p}, or @option{-O} must be defined in the same way as when
27608 the precompiled header was generated. At present, it's not clear
27609 which options are safe to change and which are not; the safest choice
27610 is to use exactly the same options when generating and using the
27611 precompiled header. The following are known to be safe:
27613 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
27614 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
27615 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
27620 For all of these except the last, the compiler automatically
27621 ignores the precompiled header if the conditions aren't met. If you
27622 find an option combination that doesn't work and doesn't cause the
27623 precompiled header to be ignored, please consider filing a bug report,
27626 If you do use differing options when generating and using the
27627 precompiled header, the actual behavior is a mixture of the
27628 behavior for the options. For instance, if you use @option{-g} to
27629 generate the precompiled header but not when using it, you may or may
27630 not get debugging information for routines in the precompiled header.