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), dbx(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
126 @xref{Option Index}, for an index to GCC's options.
129 * Option Summary:: Brief list of all options, without explanations.
130 * Overall Options:: Controlling the kind of output:
131 an executable, object files, assembler files,
132 or preprocessed source.
133 * Invoking G++:: Compiling C++ programs.
134 * C Dialect Options:: Controlling the variant of C language compiled.
135 * C++ Dialect Options:: Variations on C++.
136 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
138 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
140 * Warning Options:: How picky should the compiler be?
141 * Debugging Options:: Producing debuggable code.
142 * Optimize Options:: How much optimization?
143 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
144 * Preprocessor Options:: Controlling header files and macro definitions.
145 Also, getting dependency information for Make.
146 * Assembler Options:: Passing options to the assembler.
147 * Link Options:: Specifying libraries and so on.
148 * Directory Options:: Where to find header files and libraries.
149 Where to find the compiler executable files.
150 * Code Gen Options:: Specifying conventions for function calls, data layout
152 * Developer Options:: Printing GCC configuration info, statistics, and
154 * Submodel Options:: Target-specific options, such as compiling for a
155 specific processor variant.
156 * Spec Files:: How to pass switches to sub-processes.
157 * Environment Variables:: Env vars that affect GCC.
158 * Precompiled Headers:: Compiling a header once, and using it many times.
164 @section Option Summary
166 Here is a summary of all the options, grouped by type. Explanations are
167 in the following sections.
170 @item Overall Options
171 @xref{Overall Options,,Options Controlling the Kind of Output}.
172 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
173 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
174 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
175 @@@var{file} -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} -Wstringop-truncation @gol
318 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@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} -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 -mft32b -mcompress -mnopm}
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} -mr0rel-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=medlow -mcmodel=medany @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-vector-width=@var{opt} @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 -mgfni @gol
1208 -mcet -mibt -mshstk -mforce-indirect-call -mavx512vbmi2 @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.
1838 ISO C17, the 2017 revision of the ISO C standard (expected to be
1839 published in 2018). This standard is
1840 same as C11 except for corrections of defects (all of which are also
1841 applied with @option{-std=c11}) and a new value of
1842 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1846 GNU dialect of ISO C90 (including some C99 features).
1850 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1854 GNU dialect of ISO C11.
1855 The name @samp{gnu1x} is deprecated.
1859 GNU dialect of ISO C17. This is the default for C code.
1863 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1864 additional defect reports. Same as @option{-ansi} for C++ code.
1868 GNU dialect of @option{-std=c++98}.
1872 The 2011 ISO C++ standard plus amendments.
1873 The name @samp{c++0x} is deprecated.
1877 GNU dialect of @option{-std=c++11}.
1878 The name @samp{gnu++0x} is deprecated.
1882 The 2014 ISO C++ standard plus amendments.
1883 The name @samp{c++1y} is deprecated.
1887 GNU dialect of @option{-std=c++14}.
1888 This is the default for C++ code.
1889 The name @samp{gnu++1y} is deprecated.
1893 The 2017 ISO C++ standard plus amendments.
1894 The name @samp{c++1z} is deprecated.
1898 GNU dialect of @option{-std=c++17}.
1899 The name @samp{gnu++1z} is deprecated.
1902 The next revision of the ISO C++ standard, tentatively planned for
1903 2020. Support is highly experimental, and will almost certainly
1904 change in incompatible ways in future releases.
1907 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
1908 and will almost certainly change in incompatible ways in future
1912 @item -fgnu89-inline
1913 @opindex fgnu89-inline
1914 The option @option{-fgnu89-inline} tells GCC to use the traditional
1915 GNU semantics for @code{inline} functions when in C99 mode.
1916 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1917 Using this option is roughly equivalent to adding the
1918 @code{gnu_inline} function attribute to all inline functions
1919 (@pxref{Function Attributes}).
1921 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1922 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1923 specifies the default behavior).
1924 This option is not supported in @option{-std=c90} or
1925 @option{-std=gnu90} mode.
1927 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1928 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1929 in effect for @code{inline} functions. @xref{Common Predefined
1930 Macros,,,cpp,The C Preprocessor}.
1932 @item -fpermitted-flt-eval-methods=@var{style}
1933 @opindex fpermitted-flt-eval-methods
1934 @opindex fpermitted-flt-eval-methods=c11
1935 @opindex fpermitted-flt-eval-methods=ts-18661-3
1936 ISO/IEC TS 18661-3 defines new permissible values for
1937 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
1938 a semantic type that is an interchange or extended format should be
1939 evaluated to the precision and range of that type. These new values are
1940 a superset of those permitted under C99/C11, which does not specify the
1941 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
1942 conforming to C11 may not have been written expecting the possibility of
1945 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
1946 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
1947 or the extended set of values specified in ISO/IEC TS 18661-3.
1949 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
1951 The default when in a standards compliant mode (@option{-std=c11} or similar)
1952 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
1953 dialect (@option{-std=gnu11} or similar) is
1954 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
1956 @item -aux-info @var{filename}
1958 Output to the given filename prototyped declarations for all functions
1959 declared and/or defined in a translation unit, including those in header
1960 files. This option is silently ignored in any language other than C@.
1962 Besides declarations, the file indicates, in comments, the origin of
1963 each declaration (source file and line), whether the declaration was
1964 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1965 @samp{O} for old, respectively, in the first character after the line
1966 number and the colon), and whether it came from a declaration or a
1967 definition (@samp{C} or @samp{F}, respectively, in the following
1968 character). In the case of function definitions, a K&R-style list of
1969 arguments followed by their declarations is also provided, inside
1970 comments, after the declaration.
1972 @item -fallow-parameterless-variadic-functions
1973 @opindex fallow-parameterless-variadic-functions
1974 Accept variadic functions without named parameters.
1976 Although it is possible to define such a function, this is not very
1977 useful as it is not possible to read the arguments. This is only
1978 supported for C as this construct is allowed by C++.
1982 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1983 keyword, so that code can use these words as identifiers. You can use
1984 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1985 instead. @option{-ansi} implies @option{-fno-asm}.
1987 In C++, this switch only affects the @code{typeof} keyword, since
1988 @code{asm} and @code{inline} are standard keywords. You may want to
1989 use the @option{-fno-gnu-keywords} flag instead, which has the same
1990 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1991 switch only affects the @code{asm} and @code{typeof} keywords, since
1992 @code{inline} is a standard keyword in ISO C99.
1995 @itemx -fno-builtin-@var{function}
1996 @opindex fno-builtin
1997 @cindex built-in functions
1998 Don't recognize built-in functions that do not begin with
1999 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
2000 functions provided by GCC}, for details of the functions affected,
2001 including those which are not built-in functions when @option{-ansi} or
2002 @option{-std} options for strict ISO C conformance are used because they
2003 do not have an ISO standard meaning.
2005 GCC normally generates special code to handle certain built-in functions
2006 more efficiently; for instance, calls to @code{alloca} may become single
2007 instructions which adjust the stack directly, and calls to @code{memcpy}
2008 may become inline copy loops. The resulting code is often both smaller
2009 and faster, but since the function calls no longer appear as such, you
2010 cannot set a breakpoint on those calls, nor can you change the behavior
2011 of the functions by linking with a different library. In addition,
2012 when a function is recognized as a built-in function, GCC may use
2013 information about that function to warn about problems with calls to
2014 that function, or to generate more efficient code, even if the
2015 resulting code still contains calls to that function. For example,
2016 warnings are given with @option{-Wformat} for bad calls to
2017 @code{printf} when @code{printf} is built in and @code{strlen} is
2018 known not to modify global memory.
2020 With the @option{-fno-builtin-@var{function}} option
2021 only the built-in function @var{function} is
2022 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2023 function is named that is not built-in in this version of GCC, this
2024 option is ignored. There is no corresponding
2025 @option{-fbuiltin-@var{function}} option; if you wish to enable
2026 built-in functions selectively when using @option{-fno-builtin} or
2027 @option{-ffreestanding}, you may define macros such as:
2030 #define abs(n) __builtin_abs ((n))
2031 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2037 Enable parsing of function definitions marked with @code{__GIMPLE}.
2038 This is an experimental feature that allows unit testing of GIMPLE
2043 @cindex hosted environment
2045 Assert that compilation targets a hosted environment. This implies
2046 @option{-fbuiltin}. A hosted environment is one in which the
2047 entire standard library is available, and in which @code{main} has a return
2048 type of @code{int}. Examples are nearly everything except a kernel.
2049 This is equivalent to @option{-fno-freestanding}.
2051 @item -ffreestanding
2052 @opindex ffreestanding
2053 @cindex hosted environment
2055 Assert that compilation targets a freestanding environment. This
2056 implies @option{-fno-builtin}. A freestanding environment
2057 is one in which the standard library may not exist, and program startup may
2058 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2059 This is equivalent to @option{-fno-hosted}.
2061 @xref{Standards,,Language Standards Supported by GCC}, for details of
2062 freestanding and hosted environments.
2066 @cindex OpenACC accelerator programming
2067 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2068 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2069 compiler generates accelerated code according to the OpenACC Application
2070 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2071 implies @option{-pthread}, and thus is only supported on targets that
2072 have support for @option{-pthread}.
2074 @item -fopenacc-dim=@var{geom}
2075 @opindex fopenacc-dim
2076 @cindex OpenACC accelerator programming
2077 Specify default compute dimensions for parallel offload regions that do
2078 not explicitly specify. The @var{geom} value is a triple of
2079 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2080 can be omitted, to use a target-specific default value.
2084 @cindex OpenMP parallel
2085 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2086 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2087 compiler generates parallel code according to the OpenMP Application
2088 Program Interface v4.5 @w{@uref{http://www.openmp.org/}}. This option
2089 implies @option{-pthread}, and thus is only supported on targets that
2090 have support for @option{-pthread}. @option{-fopenmp} implies
2091 @option{-fopenmp-simd}.
2094 @opindex fopenmp-simd
2097 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2098 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2103 When the option @option{-fgnu-tm} is specified, the compiler
2104 generates code for the Linux variant of Intel's current Transactional
2105 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2106 an experimental feature whose interface may change in future versions
2107 of GCC, as the official specification changes. Please note that not
2108 all architectures are supported for this feature.
2110 For more information on GCC's support for transactional memory,
2111 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2112 Transactional Memory Library}.
2114 Note that the transactional memory feature is not supported with
2115 non-call exceptions (@option{-fnon-call-exceptions}).
2117 @item -fms-extensions
2118 @opindex fms-extensions
2119 Accept some non-standard constructs used in Microsoft header files.
2121 In C++ code, this allows member names in structures to be similar
2122 to previous types declarations.
2131 Some cases of unnamed fields in structures and unions are only
2132 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2133 fields within structs/unions}, for details.
2135 Note that this option is off for all targets but x86
2136 targets using ms-abi.
2138 @item -fplan9-extensions
2139 @opindex fplan9-extensions
2140 Accept some non-standard constructs used in Plan 9 code.
2142 This enables @option{-fms-extensions}, permits passing pointers to
2143 structures with anonymous fields to functions that expect pointers to
2144 elements of the type of the field, and permits referring to anonymous
2145 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2146 struct/union fields within structs/unions}, for details. This is only
2147 supported for C, not C++.
2149 @item -fcond-mismatch
2150 @opindex fcond-mismatch
2151 Allow conditional expressions with mismatched types in the second and
2152 third arguments. The value of such an expression is void. This option
2153 is not supported for C++.
2155 @item -flax-vector-conversions
2156 @opindex flax-vector-conversions
2157 Allow implicit conversions between vectors with differing numbers of
2158 elements and/or incompatible element types. This option should not be
2161 @item -funsigned-char
2162 @opindex funsigned-char
2163 Let the type @code{char} be unsigned, like @code{unsigned char}.
2165 Each kind of machine has a default for what @code{char} should
2166 be. It is either like @code{unsigned char} by default or like
2167 @code{signed char} by default.
2169 Ideally, a portable program should always use @code{signed char} or
2170 @code{unsigned char} when it depends on the signedness of an object.
2171 But many programs have been written to use plain @code{char} and
2172 expect it to be signed, or expect it to be unsigned, depending on the
2173 machines they were written for. This option, and its inverse, let you
2174 make such a program work with the opposite default.
2176 The type @code{char} is always a distinct type from each of
2177 @code{signed char} or @code{unsigned char}, even though its behavior
2178 is always just like one of those two.
2181 @opindex fsigned-char
2182 Let the type @code{char} be signed, like @code{signed char}.
2184 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2185 the negative form of @option{-funsigned-char}. Likewise, the option
2186 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2188 @item -fsigned-bitfields
2189 @itemx -funsigned-bitfields
2190 @itemx -fno-signed-bitfields
2191 @itemx -fno-unsigned-bitfields
2192 @opindex fsigned-bitfields
2193 @opindex funsigned-bitfields
2194 @opindex fno-signed-bitfields
2195 @opindex fno-unsigned-bitfields
2196 These options control whether a bit-field is signed or unsigned, when the
2197 declaration does not use either @code{signed} or @code{unsigned}. By
2198 default, such a bit-field is signed, because this is consistent: the
2199 basic integer types such as @code{int} are signed types.
2201 @item -fsso-struct=@var{endianness}
2202 @opindex fsso-struct
2203 Set the default scalar storage order of structures and unions to the
2204 specified endianness. The accepted values are @samp{big-endian},
2205 @samp{little-endian} and @samp{native} for the native endianness of
2206 the target (the default). This option is not supported for C++.
2208 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2209 code that is not binary compatible with code generated without it if the
2210 specified endianness is not the native endianness of the target.
2213 @node C++ Dialect Options
2214 @section Options Controlling C++ Dialect
2216 @cindex compiler options, C++
2217 @cindex C++ options, command-line
2218 @cindex options, C++
2219 This section describes the command-line options that are only meaningful
2220 for C++ programs. You can also use most of the GNU compiler options
2221 regardless of what language your program is in. For example, you
2222 might compile a file @file{firstClass.C} like this:
2225 g++ -g -fstrict-enums -O -c firstClass.C
2229 In this example, only @option{-fstrict-enums} is an option meant
2230 only for C++ programs; you can use the other options with any
2231 language supported by GCC@.
2233 Some options for compiling C programs, such as @option{-std}, are also
2234 relevant for C++ programs.
2235 @xref{C Dialect Options,,Options Controlling C Dialect}.
2237 Here is a list of options that are @emph{only} for compiling C++ programs:
2241 @item -fabi-version=@var{n}
2242 @opindex fabi-version
2243 Use version @var{n} of the C++ ABI@. The default is version 0.
2245 Version 0 refers to the version conforming most closely to
2246 the C++ ABI specification. Therefore, the ABI obtained using version 0
2247 will change in different versions of G++ as ABI bugs are fixed.
2249 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2251 Version 2 is the version of the C++ ABI that first appeared in G++
2252 3.4, and was the default through G++ 4.9.
2254 Version 3 corrects an error in mangling a constant address as a
2257 Version 4, which first appeared in G++ 4.5, implements a standard
2258 mangling for vector types.
2260 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2261 attribute const/volatile on function pointer types, decltype of a
2262 plain decl, and use of a function parameter in the declaration of
2265 Version 6, which first appeared in G++ 4.7, corrects the promotion
2266 behavior of C++11 scoped enums and the mangling of template argument
2267 packs, const/static_cast, prefix ++ and --, and a class scope function
2268 used as a template argument.
2270 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2271 builtin type and corrects the mangling of lambdas in default argument
2274 Version 8, which first appeared in G++ 4.9, corrects the substitution
2275 behavior of function types with function-cv-qualifiers.
2277 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2280 Version 10, which first appeared in G++ 6.1, adds mangling of
2281 attributes that affect type identity, such as ia32 calling convention
2282 attributes (e.g. @samp{stdcall}).
2284 Version 11, which first appeared in G++ 7, corrects the mangling of
2285 sizeof... expressions and operator names. For multiple entities with
2286 the same name within a function, that are declared in different scopes,
2287 the mangling now changes starting with the twelfth occurrence. It also
2288 implies @option{-fnew-inheriting-ctors}.
2290 See also @option{-Wabi}.
2292 @item -fabi-compat-version=@var{n}
2293 @opindex fabi-compat-version
2294 On targets that support strong aliases, G++
2295 works around mangling changes by creating an alias with the correct
2296 mangled name when defining a symbol with an incorrect mangled name.
2297 This switch specifies which ABI version to use for the alias.
2299 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2300 compatibility). If another ABI version is explicitly selected, this
2301 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2302 use @option{-fabi-compat-version=2}.
2304 If this option is not provided but @option{-Wabi=@var{n}} is, that
2305 version is used for compatibility aliases. If this option is provided
2306 along with @option{-Wabi} (without the version), the version from this
2307 option is used for the warning.
2309 @item -fno-access-control
2310 @opindex fno-access-control
2311 Turn off all access checking. This switch is mainly useful for working
2312 around bugs in the access control code.
2315 @opindex faligned-new
2316 Enable support for C++17 @code{new} of types that require more
2317 alignment than @code{void* ::operator new(std::size_t)} provides. A
2318 numeric argument such as @code{-faligned-new=32} can be used to
2319 specify how much alignment (in bytes) is provided by that function,
2320 but few users will need to override the default of
2321 @code{alignof(std::max_align_t)}.
2323 This flag is enabled by default for @option{-std=c++17}.
2327 Check that the pointer returned by @code{operator new} is non-null
2328 before attempting to modify the storage allocated. This check is
2329 normally unnecessary because the C++ standard specifies that
2330 @code{operator new} only returns @code{0} if it is declared
2331 @code{throw()}, in which case the compiler always checks the
2332 return value even without this option. In all other cases, when
2333 @code{operator new} has a non-empty exception specification, memory
2334 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2335 @samp{new (nothrow)}.
2339 Enable support for the C++ Extensions for Concepts Technical
2340 Specification, ISO 19217 (2015), which allows code like
2343 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2344 template <Addable T> T add (T a, T b) @{ return a + b; @}
2347 @item -fconstexpr-depth=@var{n}
2348 @opindex fconstexpr-depth
2349 Set the maximum nested evaluation depth for C++11 constexpr functions
2350 to @var{n}. A limit is needed to detect endless recursion during
2351 constant expression evaluation. The minimum specified by the standard
2354 @item -fconstexpr-loop-limit=@var{n}
2355 @opindex fconstexpr-loop-limit
2356 Set the maximum number of iterations for a loop in C++14 constexpr functions
2357 to @var{n}. A limit is needed to detect infinite loops during
2358 constant expression evaluation. The default is 262144 (1<<18).
2360 @item -fdeduce-init-list
2361 @opindex fdeduce-init-list
2362 Enable deduction of a template type parameter as
2363 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2366 template <class T> auto forward(T t) -> decltype (realfn (t))
2373 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2377 This deduction was implemented as a possible extension to the
2378 originally proposed semantics for the C++11 standard, but was not part
2379 of the final standard, so it is disabled by default. This option is
2380 deprecated, and may be removed in a future version of G++.
2382 @item -ffriend-injection
2383 @opindex ffriend-injection
2384 Inject friend functions into the enclosing namespace, so that they are
2385 visible outside the scope of the class in which they are declared.
2386 Friend functions were documented to work this way in the old Annotated
2387 C++ Reference Manual.
2388 However, in ISO C++ a friend function that is not declared
2389 in an enclosing scope can only be found using argument dependent
2390 lookup. GCC defaults to the standard behavior.
2392 This option is for compatibility, and may be removed in a future
2395 @item -fno-elide-constructors
2396 @opindex fno-elide-constructors
2397 The C++ standard allows an implementation to omit creating a temporary
2398 that is only used to initialize another object of the same type.
2399 Specifying this option disables that optimization, and forces G++ to
2400 call the copy constructor in all cases. This option also causes G++
2401 to call trivial member functions which otherwise would be expanded inline.
2403 In C++17, the compiler is required to omit these temporaries, but this
2404 option still affects trivial member functions.
2406 @item -fno-enforce-eh-specs
2407 @opindex fno-enforce-eh-specs
2408 Don't generate code to check for violation of exception specifications
2409 at run time. This option violates the C++ standard, but may be useful
2410 for reducing code size in production builds, much like defining
2411 @code{NDEBUG}. This does not give user code permission to throw
2412 exceptions in violation of the exception specifications; the compiler
2413 still optimizes based on the specifications, so throwing an
2414 unexpected exception results in undefined behavior at run time.
2416 @item -fextern-tls-init
2417 @itemx -fno-extern-tls-init
2418 @opindex fextern-tls-init
2419 @opindex fno-extern-tls-init
2420 The C++11 and OpenMP standards allow @code{thread_local} and
2421 @code{threadprivate} variables to have dynamic (runtime)
2422 initialization. To support this, any use of such a variable goes
2423 through a wrapper function that performs any necessary initialization.
2424 When the use and definition of the variable are in the same
2425 translation unit, this overhead can be optimized away, but when the
2426 use is in a different translation unit there is significant overhead
2427 even if the variable doesn't actually need dynamic initialization. If
2428 the programmer can be sure that no use of the variable in a
2429 non-defining TU needs to trigger dynamic initialization (either
2430 because the variable is statically initialized, or a use of the
2431 variable in the defining TU will be executed before any uses in
2432 another TU), they can avoid this overhead with the
2433 @option{-fno-extern-tls-init} option.
2435 On targets that support symbol aliases, the default is
2436 @option{-fextern-tls-init}. On targets that do not support symbol
2437 aliases, the default is @option{-fno-extern-tls-init}.
2440 @itemx -fno-for-scope
2442 @opindex fno-for-scope
2443 If @option{-ffor-scope} is specified, the scope of variables declared in
2444 a @i{for-init-statement} is limited to the @code{for} loop itself,
2445 as specified by the C++ standard.
2446 If @option{-fno-for-scope} is specified, the scope of variables declared in
2447 a @i{for-init-statement} extends to the end of the enclosing scope,
2448 as was the case in old versions of G++, and other (traditional)
2449 implementations of C++.
2451 If neither flag is given, the default is to follow the standard,
2452 but to allow and give a warning for old-style code that would
2453 otherwise be invalid, or have different behavior.
2455 @item -fno-gnu-keywords
2456 @opindex fno-gnu-keywords
2457 Do not recognize @code{typeof} as a keyword, so that code can use this
2458 word as an identifier. You can use the keyword @code{__typeof__} instead.
2459 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2460 @option{-std=c++98}, @option{-std=c++11}, etc.
2462 @item -fno-implicit-templates
2463 @opindex fno-implicit-templates
2464 Never emit code for non-inline templates that are instantiated
2465 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2466 @xref{Template Instantiation}, for more information.
2468 @item -fno-implicit-inline-templates
2469 @opindex fno-implicit-inline-templates
2470 Don't emit code for implicit instantiations of inline templates, either.
2471 The default is to handle inlines differently so that compiles with and
2472 without optimization need the same set of explicit instantiations.
2474 @item -fno-implement-inlines
2475 @opindex fno-implement-inlines
2476 To save space, do not emit out-of-line copies of inline functions
2477 controlled by @code{#pragma implementation}. This causes linker
2478 errors if these functions are not inlined everywhere they are called.
2480 @item -fms-extensions
2481 @opindex fms-extensions
2482 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2483 int and getting a pointer to member function via non-standard syntax.
2485 @item -fnew-inheriting-ctors
2486 @opindex fnew-inheriting-ctors
2487 Enable the P0136 adjustment to the semantics of C++11 constructor
2488 inheritance. This is part of C++17 but also considered to be a Defect
2489 Report against C++11 and C++14. This flag is enabled by default
2490 unless @option{-fabi-version=10} or lower is specified.
2492 @item -fnew-ttp-matching
2493 @opindex fnew-ttp-matching
2494 Enable the P0522 resolution to Core issue 150, template template
2495 parameters and default arguments: this allows a template with default
2496 template arguments as an argument for a template template parameter
2497 with fewer template parameters. This flag is enabled by default for
2498 @option{-std=c++17}.
2500 @item -fno-nonansi-builtins
2501 @opindex fno-nonansi-builtins
2502 Disable built-in declarations of functions that are not mandated by
2503 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2504 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2507 @opindex fnothrow-opt
2508 Treat a @code{throw()} exception specification as if it were a
2509 @code{noexcept} specification to reduce or eliminate the text size
2510 overhead relative to a function with no exception specification. If
2511 the function has local variables of types with non-trivial
2512 destructors, the exception specification actually makes the
2513 function smaller because the EH cleanups for those variables can be
2514 optimized away. The semantic effect is that an exception thrown out of
2515 a function with such an exception specification results in a call
2516 to @code{terminate} rather than @code{unexpected}.
2518 @item -fno-operator-names
2519 @opindex fno-operator-names
2520 Do not treat the operator name keywords @code{and}, @code{bitand},
2521 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2522 synonyms as keywords.
2524 @item -fno-optional-diags
2525 @opindex fno-optional-diags
2526 Disable diagnostics that the standard says a compiler does not need to
2527 issue. Currently, the only such diagnostic issued by G++ is the one for
2528 a name having multiple meanings within a class.
2531 @opindex fpermissive
2532 Downgrade some diagnostics about nonconformant code from errors to
2533 warnings. Thus, using @option{-fpermissive} allows some
2534 nonconforming code to compile.
2536 @item -fno-pretty-templates
2537 @opindex fno-pretty-templates
2538 When an error message refers to a specialization of a function
2539 template, the compiler normally prints the signature of the
2540 template followed by the template arguments and any typedefs or
2541 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2542 rather than @code{void f(int)}) so that it's clear which template is
2543 involved. When an error message refers to a specialization of a class
2544 template, the compiler omits any template arguments that match
2545 the default template arguments for that template. If either of these
2546 behaviors make it harder to understand the error message rather than
2547 easier, you can use @option{-fno-pretty-templates} to disable them.
2551 Enable automatic template instantiation at link time. This option also
2552 implies @option{-fno-implicit-templates}. @xref{Template
2553 Instantiation}, for more information.
2557 Disable generation of information about every class with virtual
2558 functions for use by the C++ run-time type identification features
2559 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2560 of the language, you can save some space by using this flag. Note that
2561 exception handling uses the same information, but G++ generates it as
2562 needed. The @code{dynamic_cast} operator can still be used for casts that
2563 do not require run-time type information, i.e.@: casts to @code{void *} or to
2564 unambiguous base classes.
2566 @item -fsized-deallocation
2567 @opindex fsized-deallocation
2568 Enable the built-in global declarations
2570 void operator delete (void *, std::size_t) noexcept;
2571 void operator delete[] (void *, std::size_t) noexcept;
2573 as introduced in C++14. This is useful for user-defined replacement
2574 deallocation functions that, for example, use the size of the object
2575 to make deallocation faster. Enabled by default under
2576 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2577 warns about places that might want to add a definition.
2579 @item -fstrict-enums
2580 @opindex fstrict-enums
2581 Allow the compiler to optimize using the assumption that a value of
2582 enumerated type can only be one of the values of the enumeration (as
2583 defined in the C++ standard; basically, a value that can be
2584 represented in the minimum number of bits needed to represent all the
2585 enumerators). This assumption may not be valid if the program uses a
2586 cast to convert an arbitrary integer value to the enumerated type.
2588 @item -fstrong-eval-order
2589 @opindex fstrong-eval-order
2590 Evaluate member access, array subscripting, and shift expressions in
2591 left-to-right order, and evaluate assignment in right-to-left order,
2592 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2593 @option{-fstrong-eval-order=some} enables just the ordering of member
2594 access and shift expressions, and is the default without
2595 @option{-std=c++17}.
2597 @item -ftemplate-backtrace-limit=@var{n}
2598 @opindex ftemplate-backtrace-limit
2599 Set the maximum number of template instantiation notes for a single
2600 warning or error to @var{n}. The default value is 10.
2602 @item -ftemplate-depth=@var{n}
2603 @opindex ftemplate-depth
2604 Set the maximum instantiation depth for template classes to @var{n}.
2605 A limit on the template instantiation depth is needed to detect
2606 endless recursions during template class instantiation. ANSI/ISO C++
2607 conforming programs must not rely on a maximum depth greater than 17
2608 (changed to 1024 in C++11). The default value is 900, as the compiler
2609 can run out of stack space before hitting 1024 in some situations.
2611 @item -fno-threadsafe-statics
2612 @opindex fno-threadsafe-statics
2613 Do not emit the extra code to use the routines specified in the C++
2614 ABI for thread-safe initialization of local statics. You can use this
2615 option to reduce code size slightly in code that doesn't need to be
2618 @item -fuse-cxa-atexit
2619 @opindex fuse-cxa-atexit
2620 Register destructors for objects with static storage duration with the
2621 @code{__cxa_atexit} function rather than the @code{atexit} function.
2622 This option is required for fully standards-compliant handling of static
2623 destructors, but only works if your C library supports
2624 @code{__cxa_atexit}.
2626 @item -fno-use-cxa-get-exception-ptr
2627 @opindex fno-use-cxa-get-exception-ptr
2628 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2629 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2630 if the runtime routine is not available.
2632 @item -fvisibility-inlines-hidden
2633 @opindex fvisibility-inlines-hidden
2634 This switch declares that the user does not attempt to compare
2635 pointers to inline functions or methods where the addresses of the two functions
2636 are taken in different shared objects.
2638 The effect of this is that GCC may, effectively, mark inline methods with
2639 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2640 appear in the export table of a DSO and do not require a PLT indirection
2641 when used within the DSO@. Enabling this option can have a dramatic effect
2642 on load and link times of a DSO as it massively reduces the size of the
2643 dynamic export table when the library makes heavy use of templates.
2645 The behavior of this switch is not quite the same as marking the
2646 methods as hidden directly, because it does not affect static variables
2647 local to the function or cause the compiler to deduce that
2648 the function is defined in only one shared object.
2650 You may mark a method as having a visibility explicitly to negate the
2651 effect of the switch for that method. For example, if you do want to
2652 compare pointers to a particular inline method, you might mark it as
2653 having default visibility. Marking the enclosing class with explicit
2654 visibility has no effect.
2656 Explicitly instantiated inline methods are unaffected by this option
2657 as their linkage might otherwise cross a shared library boundary.
2658 @xref{Template Instantiation}.
2660 @item -fvisibility-ms-compat
2661 @opindex fvisibility-ms-compat
2662 This flag attempts to use visibility settings to make GCC's C++
2663 linkage model compatible with that of Microsoft Visual Studio.
2665 The flag makes these changes to GCC's linkage model:
2669 It sets the default visibility to @code{hidden}, like
2670 @option{-fvisibility=hidden}.
2673 Types, but not their members, are not hidden by default.
2676 The One Definition Rule is relaxed for types without explicit
2677 visibility specifications that are defined in more than one
2678 shared object: those declarations are permitted if they are
2679 permitted when this option is not used.
2682 In new code it is better to use @option{-fvisibility=hidden} and
2683 export those classes that are intended to be externally visible.
2684 Unfortunately it is possible for code to rely, perhaps accidentally,
2685 on the Visual Studio behavior.
2687 Among the consequences of these changes are that static data members
2688 of the same type with the same name but defined in different shared
2689 objects are different, so changing one does not change the other;
2690 and that pointers to function members defined in different shared
2691 objects may not compare equal. When this flag is given, it is a
2692 violation of the ODR to define types with the same name differently.
2696 Do not use weak symbol support, even if it is provided by the linker.
2697 By default, G++ uses weak symbols if they are available. This
2698 option exists only for testing, and should not be used by end-users;
2699 it results in inferior code and has no benefits. This option may
2700 be removed in a future release of G++.
2704 Do not search for header files in the standard directories specific to
2705 C++, but do still search the other standard directories. (This option
2706 is used when building the C++ library.)
2709 In addition, these optimization, warning, and code generation options
2710 have meanings only for C++ programs:
2713 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2716 Warn when G++ it generates code that is probably not compatible with
2717 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2718 ABI with each major release, normally @option{-Wabi} will warn only if
2719 there is a check added later in a release series for an ABI issue
2720 discovered since the initial release. @option{-Wabi} will warn about
2721 more things if an older ABI version is selected (with
2722 @option{-fabi-version=@var{n}}).
2724 @option{-Wabi} can also be used with an explicit version number to
2725 warn about compatibility with a particular @option{-fabi-version}
2726 level, e.g. @option{-Wabi=2} to warn about changes relative to
2727 @option{-fabi-version=2}.
2729 If an explicit version number is provided and
2730 @option{-fabi-compat-version} is not specified, the version number
2731 from this option is used for compatibility aliases. If no explicit
2732 version number is provided with this option, but
2733 @option{-fabi-compat-version} is specified, that version number is
2734 used for ABI warnings.
2736 Although an effort has been made to warn about
2737 all such cases, there are probably some cases that are not warned about,
2738 even though G++ is generating incompatible code. There may also be
2739 cases where warnings are emitted even though the code that is generated
2742 You should rewrite your code to avoid these warnings if you are
2743 concerned about the fact that code generated by G++ may not be binary
2744 compatible with code generated by other compilers.
2746 Known incompatibilities in @option{-fabi-version=2} (which was the
2747 default from GCC 3.4 to 4.9) include:
2752 A template with a non-type template parameter of reference type was
2753 mangled incorrectly:
2756 template <int &> struct S @{@};
2760 This was fixed in @option{-fabi-version=3}.
2763 SIMD vector types declared using @code{__attribute ((vector_size))} were
2764 mangled in a non-standard way that does not allow for overloading of
2765 functions taking vectors of different sizes.
2767 The mangling was changed in @option{-fabi-version=4}.
2770 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2771 qualifiers, and @code{decltype} of a plain declaration was folded away.
2773 These mangling issues were fixed in @option{-fabi-version=5}.
2776 Scoped enumerators passed as arguments to a variadic function are
2777 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2778 On most targets this does not actually affect the parameter passing
2779 ABI, as there is no way to pass an argument smaller than @code{int}.
2781 Also, the ABI changed the mangling of template argument packs,
2782 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2783 a class scope function used as a template argument.
2785 These issues were corrected in @option{-fabi-version=6}.
2788 Lambdas in default argument scope were mangled incorrectly, and the
2789 ABI changed the mangling of @code{nullptr_t}.
2791 These issues were corrected in @option{-fabi-version=7}.
2794 When mangling a function type with function-cv-qualifiers, the
2795 un-qualified function type was incorrectly treated as a substitution
2798 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2801 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2802 unaligned accesses. Note that this did not affect the ABI of a
2803 function with a @code{nullptr_t} parameter, as parameters have a
2806 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2809 Target-specific attributes that affect the identity of a type, such as
2810 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2811 did not affect the mangled name, leading to name collisions when
2812 function pointers were used as template arguments.
2814 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2818 It also warns about psABI-related changes. The known psABI changes at this
2824 For SysV/x86-64, unions with @code{long double} members are
2825 passed in memory as specified in psABI. For example:
2835 @code{union U} is always passed in memory.
2839 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2842 Warn when a type with an ABI tag is used in a context that does not
2843 have that ABI tag. See @ref{C++ Attributes} for more information
2846 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2847 @opindex Wctor-dtor-privacy
2848 @opindex Wno-ctor-dtor-privacy
2849 Warn when a class seems unusable because all the constructors or
2850 destructors in that class are private, and it has neither friends nor
2851 public static member functions. Also warn if there are no non-private
2852 methods, and there's at least one private member function that isn't
2853 a constructor or destructor.
2855 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2856 @opindex Wdelete-non-virtual-dtor
2857 @opindex Wno-delete-non-virtual-dtor
2858 Warn when @code{delete} is used to destroy an instance of a class that
2859 has virtual functions and non-virtual destructor. It is unsafe to delete
2860 an instance of a derived class through a pointer to a base class if the
2861 base class does not have a virtual destructor. This warning is enabled
2864 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2865 @opindex Wliteral-suffix
2866 @opindex Wno-literal-suffix
2867 Warn when a string or character literal is followed by a ud-suffix which does
2868 not begin with an underscore. As a conforming extension, GCC treats such
2869 suffixes as separate preprocessing tokens in order to maintain backwards
2870 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2874 #define __STDC_FORMAT_MACROS
2875 #include <inttypes.h>
2880 printf("My int64: %" PRId64"\n", i64);
2884 In this case, @code{PRId64} is treated as a separate preprocessing token.
2886 Additionally, warn when a user-defined literal operator is declared with
2887 a literal suffix identifier that doesn't begin with an underscore. Literal
2888 suffix identifiers that don't begin with an underscore are reserved for
2889 future standardization.
2891 This warning is enabled by default.
2893 @item -Wlto-type-mismatch
2894 @opindex Wlto-type-mismatch
2895 @opindex Wno-lto-type-mismatch
2897 During the link-time optimization warn about type mismatches in
2898 global declarations from different compilation units.
2899 Requires @option{-flto} to be enabled. Enabled by default.
2901 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2903 @opindex Wno-narrowing
2904 For C++11 and later standards, narrowing conversions are diagnosed by default,
2905 as required by the standard. A narrowing conversion from a constant produces
2906 an error, and a narrowing conversion from a non-constant produces a warning,
2907 but @option{-Wno-narrowing} suppresses the diagnostic.
2908 Note that this does not affect the meaning of well-formed code;
2909 narrowing conversions are still considered ill-formed in SFINAE contexts.
2911 With @option{-Wnarrowing} in C++98, warn when a narrowing
2912 conversion prohibited by C++11 occurs within
2916 int i = @{ 2.2 @}; // error: narrowing from double to int
2919 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2921 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2923 @opindex Wno-noexcept
2924 Warn when a noexcept-expression evaluates to false because of a call
2925 to a function that does not have a non-throwing exception
2926 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2927 the compiler to never throw an exception.
2929 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
2930 @opindex Wnoexcept-type
2931 @opindex Wno-noexcept-type
2932 Warn if the C++17 feature making @code{noexcept} part of a function
2933 type changes the mangled name of a symbol relative to C++14. Enabled
2934 by @option{-Wabi} and @option{-Wc++17-compat}.
2937 template <class T> void f(T t) @{ t(); @};
2939 void h() @{ f(g); @} // in C++14 calls f<void(*)()>, in C++17 calls f<void(*)()noexcept>
2942 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
2943 @opindex Wclass-memaccess
2944 Warn when the destination of a call to a raw memory function such as
2945 @code{memset} or @code{memcpy} is an object of class type writing into which
2946 might bypass the class non-trivial or deleted constructor or copy assignment,
2947 violate const-correctness or encapsulation, or corrupt the virtual table.
2948 Modifying the representation of such objects may violate invariants maintained
2949 by member functions of the class. For example, the call to @code{memset}
2950 below is undefined becase it modifies a non-trivial class object and is,
2951 therefore, diagnosed. The safe way to either initialize or clear the storage
2952 of objects of such types is by using the appropriate constructor or assignment
2953 operator, if one is available.
2955 std::string str = "abc";
2956 memset (&str, 0, 3);
2958 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
2960 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2961 @opindex Wnon-virtual-dtor
2962 @opindex Wno-non-virtual-dtor
2963 Warn when a class has virtual functions and an accessible non-virtual
2964 destructor itself or in an accessible polymorphic base class, in which
2965 case it is possible but unsafe to delete an instance of a derived
2966 class through a pointer to the class itself or base class. This
2967 warning is automatically enabled if @option{-Weffc++} is specified.
2969 @item -Wregister @r{(C++ and Objective-C++ only)}
2971 @opindex Wno-register
2972 Warn on uses of the @code{register} storage class specifier, except
2973 when it is part of the GNU @ref{Explicit Register Variables} extension.
2974 The use of the @code{register} keyword as storage class specifier has
2975 been deprecated in C++11 and removed in C++17.
2976 Enabled by default with @option{-std=c++17}.
2978 @item -Wreorder @r{(C++ and Objective-C++ only)}
2980 @opindex Wno-reorder
2981 @cindex reordering, warning
2982 @cindex warning for reordering of member initializers
2983 Warn when the order of member initializers given in the code does not
2984 match the order in which they must be executed. For instance:
2990 A(): j (0), i (1) @{ @}
2995 The compiler rearranges the member initializers for @code{i}
2996 and @code{j} to match the declaration order of the members, emitting
2997 a warning to that effect. This warning is enabled by @option{-Wall}.
2999 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
3000 @opindex fext-numeric-literals
3001 @opindex fno-ext-numeric-literals
3002 Accept imaginary, fixed-point, or machine-defined
3003 literal number suffixes as GNU extensions.
3004 When this option is turned off these suffixes are treated
3005 as C++11 user-defined literal numeric suffixes.
3006 This is on by default for all pre-C++11 dialects and all GNU dialects:
3007 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3008 @option{-std=gnu++14}.
3009 This option is off by default
3010 for ISO C++11 onwards (@option{-std=c++11}, ...).
3013 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3016 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3019 Warn about violations of the following style guidelines from Scott Meyers'
3020 @cite{Effective C++} series of books:
3024 Define a copy constructor and an assignment operator for classes
3025 with dynamically-allocated memory.
3028 Prefer initialization to assignment in constructors.
3031 Have @code{operator=} return a reference to @code{*this}.
3034 Don't try to return a reference when you must return an object.
3037 Distinguish between prefix and postfix forms of increment and
3038 decrement operators.
3041 Never overload @code{&&}, @code{||}, or @code{,}.
3045 This option also enables @option{-Wnon-virtual-dtor}, which is also
3046 one of the effective C++ recommendations. However, the check is
3047 extended to warn about the lack of virtual destructor in accessible
3048 non-polymorphic bases classes too.
3050 When selecting this option, be aware that the standard library
3051 headers do not obey all of these guidelines; use @samp{grep -v}
3052 to filter out those warnings.
3054 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3055 @opindex Wstrict-null-sentinel
3056 @opindex Wno-strict-null-sentinel
3057 Warn about the use of an uncasted @code{NULL} as sentinel. When
3058 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3059 to @code{__null}. Although it is a null pointer constant rather than a
3060 null pointer, it is guaranteed to be of the same size as a pointer.
3061 But this use is not portable across different compilers.
3063 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3064 @opindex Wno-non-template-friend
3065 @opindex Wnon-template-friend
3066 Disable warnings when non-template friend functions are declared
3067 within a template. In very old versions of GCC that predate implementation
3068 of the ISO standard, declarations such as
3069 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3070 could be interpreted as a particular specialization of a template
3071 function; the warning exists to diagnose compatibility problems,
3072 and is enabled by default.
3074 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3075 @opindex Wold-style-cast
3076 @opindex Wno-old-style-cast
3077 Warn if an old-style (C-style) cast to a non-void type is used within
3078 a C++ program. The new-style casts (@code{dynamic_cast},
3079 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3080 less vulnerable to unintended effects and much easier to search for.
3082 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3083 @opindex Woverloaded-virtual
3084 @opindex Wno-overloaded-virtual
3085 @cindex overloaded virtual function, warning
3086 @cindex warning for overloaded virtual function
3087 Warn when a function declaration hides virtual functions from a
3088 base class. For example, in:
3095 struct B: public A @{
3100 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3111 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3112 @opindex Wno-pmf-conversions
3113 @opindex Wpmf-conversions
3114 Disable the diagnostic for converting a bound pointer to member function
3117 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3118 @opindex Wsign-promo
3119 @opindex Wno-sign-promo
3120 Warn when overload resolution chooses a promotion from unsigned or
3121 enumerated type to a signed type, over a conversion to an unsigned type of
3122 the same size. Previous versions of G++ tried to preserve
3123 unsignedness, but the standard mandates the current behavior.
3125 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3127 Warn when a primary template declaration is encountered. Some coding
3128 rules disallow templates, and this may be used to enforce that rule.
3129 The warning is inactive inside a system header file, such as the STL, so
3130 one can still use the STL. One may also instantiate or specialize
3133 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3134 @opindex Wmultiple-inheritance
3135 Warn when a class is defined with multiple direct base classes. Some
3136 coding rules disallow multiple inheritance, and this may be used to
3137 enforce that rule. The warning is inactive inside a system header file,
3138 such as the STL, so one can still use the STL. One may also define
3139 classes that indirectly use multiple inheritance.
3141 @item -Wvirtual-inheritance
3142 @opindex Wvirtual-inheritance
3143 Warn when a class is defined with a virtual direct base class. Some
3144 coding rules disallow multiple inheritance, and this may be used to
3145 enforce that rule. The warning is inactive inside a system header file,
3146 such as the STL, so one can still use the STL. One may also define
3147 classes that indirectly use virtual inheritance.
3150 @opindex Wnamespaces
3151 Warn when a namespace definition is opened. Some coding rules disallow
3152 namespaces, and this may be used to enforce that rule. The warning is
3153 inactive inside a system header file, such as the STL, so one can still
3154 use the STL. One may also use using directives and qualified names.
3156 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3158 @opindex Wno-terminate
3159 Disable the warning about a throw-expression that will immediately
3160 result in a call to @code{terminate}.
3163 @node Objective-C and Objective-C++ Dialect Options
3164 @section Options Controlling Objective-C and Objective-C++ Dialects
3166 @cindex compiler options, Objective-C and Objective-C++
3167 @cindex Objective-C and Objective-C++ options, command-line
3168 @cindex options, Objective-C and Objective-C++
3169 (NOTE: This manual does not describe the Objective-C and Objective-C++
3170 languages themselves. @xref{Standards,,Language Standards
3171 Supported by GCC}, for references.)
3173 This section describes the command-line options that are only meaningful
3174 for Objective-C and Objective-C++ programs. You can also use most of
3175 the language-independent GNU compiler options.
3176 For example, you might compile a file @file{some_class.m} like this:
3179 gcc -g -fgnu-runtime -O -c some_class.m
3183 In this example, @option{-fgnu-runtime} is an option meant only for
3184 Objective-C and Objective-C++ programs; you can use the other options with
3185 any language supported by GCC@.
3187 Note that since Objective-C is an extension of the C language, Objective-C
3188 compilations may also use options specific to the C front-end (e.g.,
3189 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3190 C++-specific options (e.g., @option{-Wabi}).
3192 Here is a list of options that are @emph{only} for compiling Objective-C
3193 and Objective-C++ programs:
3196 @item -fconstant-string-class=@var{class-name}
3197 @opindex fconstant-string-class
3198 Use @var{class-name} as the name of the class to instantiate for each
3199 literal string specified with the syntax @code{@@"@dots{}"}. The default
3200 class name is @code{NXConstantString} if the GNU runtime is being used, and
3201 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3202 @option{-fconstant-cfstrings} option, if also present, overrides the
3203 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3204 to be laid out as constant CoreFoundation strings.
3207 @opindex fgnu-runtime
3208 Generate object code compatible with the standard GNU Objective-C
3209 runtime. This is the default for most types of systems.
3211 @item -fnext-runtime
3212 @opindex fnext-runtime
3213 Generate output compatible with the NeXT runtime. This is the default
3214 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3215 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3218 @item -fno-nil-receivers
3219 @opindex fno-nil-receivers
3220 Assume that all Objective-C message dispatches (@code{[receiver
3221 message:arg]}) in this translation unit ensure that the receiver is
3222 not @code{nil}. This allows for more efficient entry points in the
3223 runtime to be used. This option is only available in conjunction with
3224 the NeXT runtime and ABI version 0 or 1.
3226 @item -fobjc-abi-version=@var{n}
3227 @opindex fobjc-abi-version
3228 Use version @var{n} of the Objective-C ABI for the selected runtime.
3229 This option is currently supported only for the NeXT runtime. In that
3230 case, Version 0 is the traditional (32-bit) ABI without support for
3231 properties and other Objective-C 2.0 additions. Version 1 is the
3232 traditional (32-bit) ABI with support for properties and other
3233 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3234 nothing is specified, the default is Version 0 on 32-bit target
3235 machines, and Version 2 on 64-bit target machines.
3237 @item -fobjc-call-cxx-cdtors
3238 @opindex fobjc-call-cxx-cdtors
3239 For each Objective-C class, check if any of its instance variables is a
3240 C++ object with a non-trivial default constructor. If so, synthesize a
3241 special @code{- (id) .cxx_construct} instance method which runs
3242 non-trivial default constructors on any such instance variables, in order,
3243 and then return @code{self}. Similarly, check if any instance variable
3244 is a C++ object with a non-trivial destructor, and if so, synthesize a
3245 special @code{- (void) .cxx_destruct} method which runs
3246 all such default destructors, in reverse order.
3248 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3249 methods thusly generated only operate on instance variables
3250 declared in the current Objective-C class, and not those inherited
3251 from superclasses. It is the responsibility of the Objective-C
3252 runtime to invoke all such methods in an object's inheritance
3253 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3254 by the runtime immediately after a new object instance is allocated;
3255 the @code{- (void) .cxx_destruct} methods are invoked immediately
3256 before the runtime deallocates an object instance.
3258 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3259 support for invoking the @code{- (id) .cxx_construct} and
3260 @code{- (void) .cxx_destruct} methods.
3262 @item -fobjc-direct-dispatch
3263 @opindex fobjc-direct-dispatch
3264 Allow fast jumps to the message dispatcher. On Darwin this is
3265 accomplished via the comm page.
3267 @item -fobjc-exceptions
3268 @opindex fobjc-exceptions
3269 Enable syntactic support for structured exception handling in
3270 Objective-C, similar to what is offered by C++. This option
3271 is required to use the Objective-C keywords @code{@@try},
3272 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3273 @code{@@synchronized}. This option is available with both the GNU
3274 runtime and the NeXT runtime (but not available in conjunction with
3275 the NeXT runtime on Mac OS X 10.2 and earlier).
3279 Enable garbage collection (GC) in Objective-C and Objective-C++
3280 programs. This option is only available with the NeXT runtime; the
3281 GNU runtime has a different garbage collection implementation that
3282 does not require special compiler flags.
3284 @item -fobjc-nilcheck
3285 @opindex fobjc-nilcheck
3286 For the NeXT runtime with version 2 of the ABI, check for a nil
3287 receiver in method invocations before doing the actual method call.
3288 This is the default and can be disabled using
3289 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3290 checked for nil in this way no matter what this flag is set to.
3291 Currently this flag does nothing when the GNU runtime, or an older
3292 version of the NeXT runtime ABI, is used.
3294 @item -fobjc-std=objc1
3296 Conform to the language syntax of Objective-C 1.0, the language
3297 recognized by GCC 4.0. This only affects the Objective-C additions to
3298 the C/C++ language; it does not affect conformance to C/C++ standards,
3299 which is controlled by the separate C/C++ dialect option flags. When
3300 this option is used with the Objective-C or Objective-C++ compiler,
3301 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3302 This is useful if you need to make sure that your Objective-C code can
3303 be compiled with older versions of GCC@.
3305 @item -freplace-objc-classes
3306 @opindex freplace-objc-classes
3307 Emit a special marker instructing @command{ld(1)} not to statically link in
3308 the resulting object file, and allow @command{dyld(1)} to load it in at
3309 run time instead. This is used in conjunction with the Fix-and-Continue
3310 debugging mode, where the object file in question may be recompiled and
3311 dynamically reloaded in the course of program execution, without the need
3312 to restart the program itself. Currently, Fix-and-Continue functionality
3313 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3318 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3319 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3320 compile time) with static class references that get initialized at load time,
3321 which improves run-time performance. Specifying the @option{-fzero-link} flag
3322 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3323 to be retained. This is useful in Zero-Link debugging mode, since it allows
3324 for individual class implementations to be modified during program execution.
3325 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3326 regardless of command-line options.
3328 @item -fno-local-ivars
3329 @opindex fno-local-ivars
3330 @opindex flocal-ivars
3331 By default instance variables in Objective-C can be accessed as if
3332 they were local variables from within the methods of the class they're
3333 declared in. This can lead to shadowing between instance variables
3334 and other variables declared either locally inside a class method or
3335 globally with the same name. Specifying the @option{-fno-local-ivars}
3336 flag disables this behavior thus avoiding variable shadowing issues.
3338 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3339 @opindex fivar-visibility
3340 Set the default instance variable visibility to the specified option
3341 so that instance variables declared outside the scope of any access
3342 modifier directives default to the specified visibility.
3346 Dump interface declarations for all classes seen in the source file to a
3347 file named @file{@var{sourcename}.decl}.
3349 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3350 @opindex Wassign-intercept
3351 @opindex Wno-assign-intercept
3352 Warn whenever an Objective-C assignment is being intercepted by the
3355 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3356 @opindex Wno-protocol
3358 If a class is declared to implement a protocol, a warning is issued for
3359 every method in the protocol that is not implemented by the class. The
3360 default behavior is to issue a warning for every method not explicitly
3361 implemented in the class, even if a method implementation is inherited
3362 from the superclass. If you use the @option{-Wno-protocol} option, then
3363 methods inherited from the superclass are considered to be implemented,
3364 and no warning is issued for them.
3366 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3368 @opindex Wno-selector
3369 Warn if multiple methods of different types for the same selector are
3370 found during compilation. The check is performed on the list of methods
3371 in the final stage of compilation. Additionally, a check is performed
3372 for each selector appearing in a @code{@@selector(@dots{})}
3373 expression, and a corresponding method for that selector has been found
3374 during compilation. Because these checks scan the method table only at
3375 the end of compilation, these warnings are not produced if the final
3376 stage of compilation is not reached, for example because an error is
3377 found during compilation, or because the @option{-fsyntax-only} option is
3380 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3381 @opindex Wstrict-selector-match
3382 @opindex Wno-strict-selector-match
3383 Warn if multiple methods with differing argument and/or return types are
3384 found for a given selector when attempting to send a message using this
3385 selector to a receiver of type @code{id} or @code{Class}. When this flag
3386 is off (which is the default behavior), the compiler omits such warnings
3387 if any differences found are confined to types that share the same size
3390 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3391 @opindex Wundeclared-selector
3392 @opindex Wno-undeclared-selector
3393 Warn if a @code{@@selector(@dots{})} expression referring to an
3394 undeclared selector is found. A selector is considered undeclared if no
3395 method with that name has been declared before the
3396 @code{@@selector(@dots{})} expression, either explicitly in an
3397 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3398 an @code{@@implementation} section. This option always performs its
3399 checks as soon as a @code{@@selector(@dots{})} expression is found,
3400 while @option{-Wselector} only performs its checks in the final stage of
3401 compilation. This also enforces the coding style convention
3402 that methods and selectors must be declared before being used.
3404 @item -print-objc-runtime-info
3405 @opindex print-objc-runtime-info
3406 Generate C header describing the largest structure that is passed by
3411 @node Diagnostic Message Formatting Options
3412 @section Options to Control Diagnostic Messages Formatting
3413 @cindex options to control diagnostics formatting
3414 @cindex diagnostic messages
3415 @cindex message formatting
3417 Traditionally, diagnostic messages have been formatted irrespective of
3418 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3419 options described below
3420 to control the formatting algorithm for diagnostic messages,
3421 e.g.@: how many characters per line, how often source location
3422 information should be reported. Note that some language front ends may not
3423 honor these options.
3426 @item -fmessage-length=@var{n}
3427 @opindex fmessage-length
3428 Try to format error messages so that they fit on lines of about
3429 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3430 done; each error message appears on a single line. This is the
3431 default for all front ends.
3433 @item -fdiagnostics-show-location=once
3434 @opindex fdiagnostics-show-location
3435 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3436 reporter to emit source location information @emph{once}; that is, in
3437 case the message is too long to fit on a single physical line and has to
3438 be wrapped, the source location won't be emitted (as prefix) again,
3439 over and over, in subsequent continuation lines. This is the default
3442 @item -fdiagnostics-show-location=every-line
3443 Only meaningful in line-wrapping mode. Instructs the diagnostic
3444 messages reporter to emit the same source location information (as
3445 prefix) for physical lines that result from the process of breaking
3446 a message which is too long to fit on a single line.
3448 @item -fdiagnostics-color[=@var{WHEN}]
3449 @itemx -fno-diagnostics-color
3450 @opindex fdiagnostics-color
3451 @cindex highlight, color
3452 @vindex GCC_COLORS @r{environment variable}
3453 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3454 or @samp{auto}. The default depends on how the compiler has been configured,
3455 it can be any of the above @var{WHEN} options or also @samp{never}
3456 if @env{GCC_COLORS} environment variable isn't present in the environment,
3457 and @samp{auto} otherwise.
3458 @samp{auto} means to use color only when the standard error is a terminal.
3459 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3460 aliases for @option{-fdiagnostics-color=always} and
3461 @option{-fdiagnostics-color=never}, respectively.
3463 The colors are defined by the environment variable @env{GCC_COLORS}.
3464 Its value is a colon-separated list of capabilities and Select Graphic
3465 Rendition (SGR) substrings. SGR commands are interpreted by the
3466 terminal or terminal emulator. (See the section in the documentation
3467 of your text terminal for permitted values and their meanings as
3468 character attributes.) These substring values are integers in decimal
3469 representation and can be concatenated with semicolons.
3470 Common values to concatenate include
3472 @samp{4} for underline,
3474 @samp{7} for inverse,
3475 @samp{39} for default foreground color,
3476 @samp{30} to @samp{37} for foreground colors,
3477 @samp{90} to @samp{97} for 16-color mode foreground colors,
3478 @samp{38;5;0} to @samp{38;5;255}
3479 for 88-color and 256-color modes foreground colors,
3480 @samp{49} for default background color,
3481 @samp{40} to @samp{47} for background colors,
3482 @samp{100} to @samp{107} for 16-color mode background colors,
3483 and @samp{48;5;0} to @samp{48;5;255}
3484 for 88-color and 256-color modes background colors.
3486 The default @env{GCC_COLORS} is
3488 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3489 quote=01:fixit-insert=32:fixit-delete=31:\
3490 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3494 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3495 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3496 @samp{01} is bold, and @samp{31} is red.
3497 Setting @env{GCC_COLORS} to the empty string disables colors.
3498 Supported capabilities are as follows.
3502 @vindex error GCC_COLORS @r{capability}
3503 SGR substring for error: markers.
3506 @vindex warning GCC_COLORS @r{capability}
3507 SGR substring for warning: markers.
3510 @vindex note GCC_COLORS @r{capability}
3511 SGR substring for note: markers.
3514 @vindex range1 GCC_COLORS @r{capability}
3515 SGR substring for first additional range.
3518 @vindex range2 GCC_COLORS @r{capability}
3519 SGR substring for second additional range.
3522 @vindex locus GCC_COLORS @r{capability}
3523 SGR substring for location information, @samp{file:line} or
3524 @samp{file:line:column} etc.
3527 @vindex quote GCC_COLORS @r{capability}
3528 SGR substring for information printed within quotes.
3531 @vindex fixit-insert GCC_COLORS @r{capability}
3532 SGR substring for fix-it hints suggesting text to
3533 be inserted or replaced.
3536 @vindex fixit-delete GCC_COLORS @r{capability}
3537 SGR substring for fix-it hints suggesting text to
3540 @item diff-filename=
3541 @vindex diff-filename GCC_COLORS @r{capability}
3542 SGR substring for filename headers within generated patches.
3545 @vindex diff-hunk GCC_COLORS @r{capability}
3546 SGR substring for the starts of hunks within generated patches.
3549 @vindex diff-delete GCC_COLORS @r{capability}
3550 SGR substring for deleted lines within generated patches.
3553 @vindex diff-insert GCC_COLORS @r{capability}
3554 SGR substring for inserted lines within generated patches.
3557 @vindex type-diff GCC_COLORS @r{capability}
3558 SGR substring for highlighting mismatching types within template
3559 arguments in the C++ frontend.
3562 @item -fno-diagnostics-show-option
3563 @opindex fno-diagnostics-show-option
3564 @opindex fdiagnostics-show-option
3565 By default, each diagnostic emitted includes text indicating the
3566 command-line option that directly controls the diagnostic (if such an
3567 option is known to the diagnostic machinery). Specifying the
3568 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3570 @item -fno-diagnostics-show-caret
3571 @opindex fno-diagnostics-show-caret
3572 @opindex fdiagnostics-show-caret
3573 By default, each diagnostic emitted includes the original source line
3574 and a caret @samp{^} indicating the column. This option suppresses this
3575 information. The source line is truncated to @var{n} characters, if
3576 the @option{-fmessage-length=n} option is given. When the output is done
3577 to the terminal, the width is limited to the width given by the
3578 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3580 @item -fdiagnostics-parseable-fixits
3581 @opindex fdiagnostics-parseable-fixits
3582 Emit fix-it hints in a machine-parseable format, suitable for consumption
3583 by IDEs. For each fix-it, a line will be printed after the relevant
3584 diagnostic, starting with the string ``fix-it:''. For example:
3587 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3590 The location is expressed as a half-open range, expressed as a count of
3591 bytes, starting at byte 1 for the initial column. In the above example,
3592 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3596 00000000011111111112222222222
3597 12345678901234567890123456789
3598 gtk_widget_showall (dlg);
3603 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3604 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3605 (e.g. vertical tab as ``\013'').
3607 An empty replacement string indicates that the given range is to be removed.
3608 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3609 be inserted at the given position.
3611 @item -fdiagnostics-generate-patch
3612 @opindex fdiagnostics-generate-patch
3613 Print fix-it hints to stderr in unified diff format, after any diagnostics
3614 are printed. For example:
3621 void show_cb(GtkDialog *dlg)
3623 - gtk_widget_showall(dlg);
3624 + gtk_widget_show_all(dlg);
3629 The diff may or may not be colorized, following the same rules
3630 as for diagnostics (see @option{-fdiagnostics-color}).
3632 @item -fdiagnostics-show-template-tree
3633 @opindex fdiagnostics-show-template-tree
3635 In the C++ frontend, when printing diagnostics showing mismatching
3636 template types, such as:
3639 could not convert 'std::map<int, std::vector<double> >()'
3640 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3643 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3644 tree-like structure showing the common and differing parts of the types,
3654 The parts that differ are highlighted with color (``double'' and
3655 ``float'' in this case).
3657 @item -fno-elide-type
3658 @opindex fno-elide-type
3659 @opindex felide-type
3660 By default when the C++ frontend prints diagnostics showing mismatching
3661 template types, common parts of the types are printed as ``[...]'' to
3662 simplify the error message. For example:
3665 could not convert 'std::map<int, std::vector<double> >()'
3666 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3669 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
3670 This flag also affects the output of the
3671 @option{-fdiagnostics-show-template-tree} flag.
3673 @item -fno-show-column
3674 @opindex fno-show-column
3675 Do not print column numbers in diagnostics. This may be necessary if
3676 diagnostics are being scanned by a program that does not understand the
3677 column numbers, such as @command{dejagnu}.
3681 @node Warning Options
3682 @section Options to Request or Suppress Warnings
3683 @cindex options to control warnings
3684 @cindex warning messages
3685 @cindex messages, warning
3686 @cindex suppressing warnings
3688 Warnings are diagnostic messages that report constructions that
3689 are not inherently erroneous but that are risky or suggest there
3690 may have been an error.
3692 The following language-independent options do not enable specific
3693 warnings but control the kinds of diagnostics produced by GCC@.
3696 @cindex syntax checking
3698 @opindex fsyntax-only
3699 Check the code for syntax errors, but don't do anything beyond that.
3701 @item -fmax-errors=@var{n}
3702 @opindex fmax-errors
3703 Limits the maximum number of error messages to @var{n}, at which point
3704 GCC bails out rather than attempting to continue processing the source
3705 code. If @var{n} is 0 (the default), there is no limit on the number
3706 of error messages produced. If @option{-Wfatal-errors} is also
3707 specified, then @option{-Wfatal-errors} takes precedence over this
3712 Inhibit all warning messages.
3717 Make all warnings into errors.
3722 Make the specified warning into an error. The specifier for a warning
3723 is appended; for example @option{-Werror=switch} turns the warnings
3724 controlled by @option{-Wswitch} into errors. This switch takes a
3725 negative form, to be used to negate @option{-Werror} for specific
3726 warnings; for example @option{-Wno-error=switch} makes
3727 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3730 The warning message for each controllable warning includes the
3731 option that controls the warning. That option can then be used with
3732 @option{-Werror=} and @option{-Wno-error=} as described above.
3733 (Printing of the option in the warning message can be disabled using the
3734 @option{-fno-diagnostics-show-option} flag.)
3736 Note that specifying @option{-Werror=}@var{foo} automatically implies
3737 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3740 @item -Wfatal-errors
3741 @opindex Wfatal-errors
3742 @opindex Wno-fatal-errors
3743 This option causes the compiler to abort compilation on the first error
3744 occurred rather than trying to keep going and printing further error
3749 You can request many specific warnings with options beginning with
3750 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3751 implicit declarations. Each of these specific warning options also
3752 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3753 example, @option{-Wno-implicit}. This manual lists only one of the
3754 two forms, whichever is not the default. For further
3755 language-specific options also refer to @ref{C++ Dialect Options} and
3756 @ref{Objective-C and Objective-C++ Dialect Options}.
3758 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3759 options, such as @option{-Wunused}, which may turn on further options,
3760 such as @option{-Wunused-value}. The combined effect of positive and
3761 negative forms is that more specific options have priority over less
3762 specific ones, independently of their position in the command-line. For
3763 options of the same specificity, the last one takes effect. Options
3764 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3765 as if they appeared at the end of the command-line.
3767 When an unrecognized warning option is requested (e.g.,
3768 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3769 that the option is not recognized. However, if the @option{-Wno-} form
3770 is used, the behavior is slightly different: no diagnostic is
3771 produced for @option{-Wno-unknown-warning} unless other diagnostics
3772 are being produced. This allows the use of new @option{-Wno-} options
3773 with old compilers, but if something goes wrong, the compiler
3774 warns that an unrecognized option is present.
3781 Issue all the warnings demanded by strict ISO C and ISO C++;
3782 reject all programs that use forbidden extensions, and some other
3783 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3784 version of the ISO C standard specified by any @option{-std} option used.
3786 Valid ISO C and ISO C++ programs should compile properly with or without
3787 this option (though a rare few require @option{-ansi} or a
3788 @option{-std} option specifying the required version of ISO C)@. However,
3789 without this option, certain GNU extensions and traditional C and C++
3790 features are supported as well. With this option, they are rejected.
3792 @option{-Wpedantic} does not cause warning messages for use of the
3793 alternate keywords whose names begin and end with @samp{__}. Pedantic
3794 warnings are also disabled in the expression that follows
3795 @code{__extension__}. However, only system header files should use
3796 these escape routes; application programs should avoid them.
3797 @xref{Alternate Keywords}.
3799 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3800 C conformance. They soon find that it does not do quite what they want:
3801 it finds some non-ISO practices, but not all---only those for which
3802 ISO C @emph{requires} a diagnostic, and some others for which
3803 diagnostics have been added.
3805 A feature to report any failure to conform to ISO C might be useful in
3806 some instances, but would require considerable additional work and would
3807 be quite different from @option{-Wpedantic}. We don't have plans to
3808 support such a feature in the near future.
3810 Where the standard specified with @option{-std} represents a GNU
3811 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3812 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3813 extended dialect is based. Warnings from @option{-Wpedantic} are given
3814 where they are required by the base standard. (It does not make sense
3815 for such warnings to be given only for features not in the specified GNU
3816 C dialect, since by definition the GNU dialects of C include all
3817 features the compiler supports with the given option, and there would be
3818 nothing to warn about.)
3820 @item -pedantic-errors
3821 @opindex pedantic-errors
3822 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3823 requires a diagnostic, in some cases where there is undefined behavior
3824 at compile-time and in some other cases that do not prevent compilation
3825 of programs that are valid according to the standard. This is not
3826 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3827 by this option and not enabled by the latter and vice versa.
3832 This enables all the warnings about constructions that some users
3833 consider questionable, and that are easy to avoid (or modify to
3834 prevent the warning), even in conjunction with macros. This also
3835 enables some language-specific warnings described in @ref{C++ Dialect
3836 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3838 @option{-Wall} turns on the following warning flags:
3840 @gccoptlist{-Waddress @gol
3841 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3843 -Wbool-operation @gol
3844 -Wc++11-compat -Wc++14-compat @gol
3845 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
3846 -Wchar-subscripts @gol
3848 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3849 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3851 -Wint-in-bool-context @gol
3852 -Wimplicit @r{(C and Objective-C only)} @gol
3853 -Wimplicit-int @r{(C and Objective-C only)} @gol
3854 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3855 -Winit-self @r{(only for C++)} @gol
3856 -Wlogical-not-parentheses @gol
3857 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3858 -Wmaybe-uninitialized @gol
3859 -Wmemset-elt-size @gol
3860 -Wmemset-transposed-args @gol
3861 -Wmisleading-indentation @r{(only for C/C++)} @gol
3862 -Wmissing-braces @r{(only for C/ObjC)} @gol
3863 -Wmultistatement-macros @gol
3864 -Wnarrowing @r{(only for C++)} @gol
3866 -Wnonnull-compare @gol
3872 -Wsequence-point @gol
3873 -Wsign-compare @r{(only in C++)} @gol
3874 -Wsizeof-pointer-div @gol
3875 -Wsizeof-pointer-memaccess @gol
3876 -Wstrict-aliasing @gol
3877 -Wstrict-overflow=1 @gol
3879 -Wtautological-compare @gol
3881 -Wuninitialized @gol
3882 -Wunknown-pragmas @gol
3883 -Wunused-function @gol
3886 -Wunused-variable @gol
3887 -Wvolatile-register-var @gol
3890 Note that some warning flags are not implied by @option{-Wall}. Some of
3891 them warn about constructions that users generally do not consider
3892 questionable, but which occasionally you might wish to check for;
3893 others warn about constructions that are necessary or hard to avoid in
3894 some cases, and there is no simple way to modify the code to suppress
3895 the warning. Some of them are enabled by @option{-Wextra} but many of
3896 them must be enabled individually.
3902 This enables some extra warning flags that are not enabled by
3903 @option{-Wall}. (This option used to be called @option{-W}. The older
3904 name is still supported, but the newer name is more descriptive.)
3906 @gccoptlist{-Wclobbered @gol
3908 -Wignored-qualifiers @gol
3909 -Wimplicit-fallthrough=3 @gol
3910 -Wmissing-field-initializers @gol
3911 -Wmissing-parameter-type @r{(C only)} @gol
3912 -Wold-style-declaration @r{(C only)} @gol
3913 -Woverride-init @gol
3914 -Wsign-compare @r{(C only)} @gol
3916 -Wuninitialized @gol
3917 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3918 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3919 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3922 The option @option{-Wextra} also prints warning messages for the
3928 A pointer is compared against integer zero with @code{<}, @code{<=},
3929 @code{>}, or @code{>=}.
3932 (C++ only) An enumerator and a non-enumerator both appear in a
3933 conditional expression.
3936 (C++ only) Ambiguous virtual bases.
3939 (C++ only) Subscripting an array that has been declared @code{register}.
3942 (C++ only) Taking the address of a variable that has been declared
3946 (C++ only) A base class is not initialized in the copy constructor
3951 @item -Wchar-subscripts
3952 @opindex Wchar-subscripts
3953 @opindex Wno-char-subscripts
3954 Warn if an array subscript has type @code{char}. This is a common cause
3955 of error, as programmers often forget that this type is signed on some
3957 This warning is enabled by @option{-Wall}.
3961 Warn about an invalid memory access that is found by Pointer Bounds Checker
3962 (@option{-fcheck-pointer-bounds}).
3964 @item -Wno-coverage-mismatch
3965 @opindex Wno-coverage-mismatch
3966 Warn if feedback profiles do not match when using the
3967 @option{-fprofile-use} option.
3968 If a source file is changed between compiling with @option{-fprofile-gen} and
3969 with @option{-fprofile-use}, the files with the profile feedback can fail
3970 to match the source file and GCC cannot use the profile feedback
3971 information. By default, this warning is enabled and is treated as an
3972 error. @option{-Wno-coverage-mismatch} can be used to disable the
3973 warning or @option{-Wno-error=coverage-mismatch} can be used to
3974 disable the error. Disabling the error for this warning can result in
3975 poorly optimized code and is useful only in the
3976 case of very minor changes such as bug fixes to an existing code-base.
3977 Completely disabling the warning is not recommended.
3980 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3982 Suppress warning messages emitted by @code{#warning} directives.
3984 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3985 @opindex Wdouble-promotion
3986 @opindex Wno-double-promotion
3987 Give a warning when a value of type @code{float} is implicitly
3988 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3989 floating-point unit implement @code{float} in hardware, but emulate
3990 @code{double} in software. On such a machine, doing computations
3991 using @code{double} values is much more expensive because of the
3992 overhead required for software emulation.
3994 It is easy to accidentally do computations with @code{double} because
3995 floating-point literals are implicitly of type @code{double}. For
3999 float area(float radius)
4001 return 3.14159 * radius * radius;
4005 the compiler performs the entire computation with @code{double}
4006 because the floating-point literal is a @code{double}.
4008 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4009 @opindex Wduplicate-decl-specifier
4010 @opindex Wno-duplicate-decl-specifier
4011 Warn if a declaration has duplicate @code{const}, @code{volatile},
4012 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4016 @itemx -Wformat=@var{n}
4019 @opindex ffreestanding
4020 @opindex fno-builtin
4022 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4023 the arguments supplied have types appropriate to the format string
4024 specified, and that the conversions specified in the format string make
4025 sense. This includes standard functions, and others specified by format
4026 attributes (@pxref{Function Attributes}), in the @code{printf},
4027 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4028 not in the C standard) families (or other target-specific families).
4029 Which functions are checked without format attributes having been
4030 specified depends on the standard version selected, and such checks of
4031 functions without the attribute specified are disabled by
4032 @option{-ffreestanding} or @option{-fno-builtin}.
4034 The formats are checked against the format features supported by GNU
4035 libc version 2.2. These include all ISO C90 and C99 features, as well
4036 as features from the Single Unix Specification and some BSD and GNU
4037 extensions. Other library implementations may not support all these
4038 features; GCC does not support warning about features that go beyond a
4039 particular library's limitations. However, if @option{-Wpedantic} is used
4040 with @option{-Wformat}, warnings are given about format features not
4041 in the selected standard version (but not for @code{strfmon} formats,
4042 since those are not in any version of the C standard). @xref{C Dialect
4043 Options,,Options Controlling C Dialect}.
4050 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4051 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4052 @option{-Wformat} also checks for null format arguments for several
4053 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4054 aspects of this level of format checking can be disabled by the
4055 options: @option{-Wno-format-contains-nul},
4056 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4057 @option{-Wformat} is enabled by @option{-Wall}.
4059 @item -Wno-format-contains-nul
4060 @opindex Wno-format-contains-nul
4061 @opindex Wformat-contains-nul
4062 If @option{-Wformat} is specified, do not warn about format strings that
4065 @item -Wno-format-extra-args
4066 @opindex Wno-format-extra-args
4067 @opindex Wformat-extra-args
4068 If @option{-Wformat} is specified, do not warn about excess arguments to a
4069 @code{printf} or @code{scanf} format function. The C standard specifies
4070 that such arguments are ignored.
4072 Where the unused arguments lie between used arguments that are
4073 specified with @samp{$} operand number specifications, normally
4074 warnings are still given, since the implementation could not know what
4075 type to pass to @code{va_arg} to skip the unused arguments. However,
4076 in the case of @code{scanf} formats, this option suppresses the
4077 warning if the unused arguments are all pointers, since the Single
4078 Unix Specification says that such unused arguments are allowed.
4080 @item -Wformat-overflow
4081 @itemx -Wformat-overflow=@var{level}
4082 @opindex Wformat-overflow
4083 @opindex Wno-format-overflow
4084 Warn about calls to formatted input/output functions such as @code{sprintf}
4085 and @code{vsprintf} that might overflow the destination buffer. When the
4086 exact number of bytes written by a format directive cannot be determined
4087 at compile-time it is estimated based on heuristics that depend on the
4088 @var{level} argument and on optimization. While enabling optimization
4089 will in most cases improve the accuracy of the warning, it may also
4090 result in false positives.
4093 @item -Wformat-overflow
4094 @item -Wformat-overflow=1
4095 @opindex Wformat-overflow
4096 @opindex Wno-format-overflow
4097 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4098 employs a conservative approach that warns only about calls that most
4099 likely overflow the buffer. At this level, numeric arguments to format
4100 directives with unknown values are assumed to have the value of one, and
4101 strings of unknown length to be empty. Numeric arguments that are known
4102 to be bounded to a subrange of their type, or string arguments whose output
4103 is bounded either by their directive's precision or by a finite set of
4104 string literals, are assumed to take on the value within the range that
4105 results in the most bytes on output. For example, the call to @code{sprintf}
4106 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4107 the terminating NUL character (@code{'\0'}) appended by the function
4108 to the destination buffer will be written past its end. Increasing
4109 the size of the buffer by a single byte is sufficient to avoid the
4110 warning, though it may not be sufficient to avoid the overflow.
4113 void f (int a, int b)
4116 sprintf (buf, "a = %i, b = %i\n", a, b);
4120 @item -Wformat-overflow=2
4121 Level @var{2} warns also about calls that might overflow the destination
4122 buffer given an argument of sufficient length or magnitude. At level
4123 @var{2}, unknown numeric arguments are assumed to have the minimum
4124 representable value for signed types with a precision greater than 1, and
4125 the maximum representable value otherwise. Unknown string arguments whose
4126 length cannot be assumed to be bounded either by the directive's precision,
4127 or by a finite set of string literals they may evaluate to, or the character
4128 array they may point to, are assumed to be 1 character long.
4130 At level @var{2}, the call in the example above is again diagnosed, but
4131 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4132 @code{%i} directive will write some of its digits beyond the end of
4133 the destination buffer. To make the call safe regardless of the values
4134 of the two variables, the size of the destination buffer must be increased
4135 to at least 34 bytes. GCC includes the minimum size of the buffer in
4136 an informational note following the warning.
4138 An alternative to increasing the size of the destination buffer is to
4139 constrain the range of formatted values. The maximum length of string
4140 arguments can be bounded by specifying the precision in the format
4141 directive. When numeric arguments of format directives can be assumed
4142 to be bounded by less than the precision of their type, choosing
4143 an appropriate length modifier to the format specifier will reduce
4144 the required buffer size. For example, if @var{a} and @var{b} in the
4145 example above can be assumed to be within the precision of
4146 the @code{short int} type then using either the @code{%hi} format
4147 directive or casting the argument to @code{short} reduces the maximum
4148 required size of the buffer to 24 bytes.
4151 void f (int a, int b)
4154 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4159 @item -Wno-format-zero-length
4160 @opindex Wno-format-zero-length
4161 @opindex Wformat-zero-length
4162 If @option{-Wformat} is specified, do not warn about zero-length formats.
4163 The C standard specifies that zero-length formats are allowed.
4168 Enable @option{-Wformat} plus additional format checks. Currently
4169 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4172 @item -Wformat-nonliteral
4173 @opindex Wformat-nonliteral
4174 @opindex Wno-format-nonliteral
4175 If @option{-Wformat} is specified, also warn if the format string is not a
4176 string literal and so cannot be checked, unless the format function
4177 takes its format arguments as a @code{va_list}.
4179 @item -Wformat-security
4180 @opindex Wformat-security
4181 @opindex Wno-format-security
4182 If @option{-Wformat} is specified, also warn about uses of format
4183 functions that represent possible security problems. At present, this
4184 warns about calls to @code{printf} and @code{scanf} functions where the
4185 format string is not a string literal and there are no format arguments,
4186 as in @code{printf (foo);}. This may be a security hole if the format
4187 string came from untrusted input and contains @samp{%n}. (This is
4188 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4189 in future warnings may be added to @option{-Wformat-security} that are not
4190 included in @option{-Wformat-nonliteral}.)
4192 @item -Wformat-signedness
4193 @opindex Wformat-signedness
4194 @opindex Wno-format-signedness
4195 If @option{-Wformat} is specified, also warn if the format string
4196 requires an unsigned argument and the argument is signed and vice versa.
4198 @item -Wformat-truncation
4199 @itemx -Wformat-truncation=@var{level}
4200 @opindex Wformat-truncation
4201 @opindex Wno-format-truncation
4202 Warn about calls to formatted input/output functions such as @code{snprintf}
4203 and @code{vsnprintf} that might result in output truncation. When the exact
4204 number of bytes written by a format directive cannot be determined at
4205 compile-time it is estimated based on heuristics that depend on
4206 the @var{level} argument and on optimization. While enabling optimization
4207 will in most cases improve the accuracy of the warning, it may also result
4208 in false positives. Except as noted otherwise, the option uses the same
4209 logic @option{-Wformat-overflow}.
4212 @item -Wformat-truncation
4213 @item -Wformat-truncation=1
4214 @opindex Wformat-truncation
4215 @opindex Wno-format-overflow
4216 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4217 employs a conservative approach that warns only about calls to bounded
4218 functions whose return value is unused and that will most likely result
4219 in output truncation.
4221 @item -Wformat-truncation=2
4222 Level @var{2} warns also about calls to bounded functions whose return
4223 value is used and that might result in truncation given an argument of
4224 sufficient length or magnitude.
4228 @opindex Wformat-y2k
4229 @opindex Wno-format-y2k
4230 If @option{-Wformat} is specified, also warn about @code{strftime}
4231 formats that may yield only a two-digit year.
4236 @opindex Wno-nonnull
4237 Warn about passing a null pointer for arguments marked as
4238 requiring a non-null value by the @code{nonnull} function attribute.
4240 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4241 can be disabled with the @option{-Wno-nonnull} option.
4243 @item -Wnonnull-compare
4244 @opindex Wnonnull-compare
4245 @opindex Wno-nonnull-compare
4246 Warn when comparing an argument marked with the @code{nonnull}
4247 function attribute against null inside the function.
4249 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4250 can be disabled with the @option{-Wno-nonnull-compare} option.
4252 @item -Wnull-dereference
4253 @opindex Wnull-dereference
4254 @opindex Wno-null-dereference
4255 Warn if the compiler detects paths that trigger erroneous or
4256 undefined behavior due to dereferencing a null pointer. This option
4257 is only active when @option{-fdelete-null-pointer-checks} is active,
4258 which is enabled by optimizations in most targets. The precision of
4259 the warnings depends on the optimization options used.
4261 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4263 @opindex Wno-init-self
4264 Warn about uninitialized variables that are initialized with themselves.
4265 Note this option can only be used with the @option{-Wuninitialized} option.
4267 For example, GCC warns about @code{i} being uninitialized in the
4268 following snippet only when @option{-Winit-self} has been specified:
4279 This warning is enabled by @option{-Wall} in C++.
4281 @item -Wimplicit-int @r{(C and Objective-C only)}
4282 @opindex Wimplicit-int
4283 @opindex Wno-implicit-int
4284 Warn when a declaration does not specify a type.
4285 This warning is enabled by @option{-Wall}.
4287 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4288 @opindex Wimplicit-function-declaration
4289 @opindex Wno-implicit-function-declaration
4290 Give a warning whenever a function is used before being declared. In
4291 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4292 enabled by default and it is made into an error by
4293 @option{-pedantic-errors}. This warning is also enabled by
4296 @item -Wimplicit @r{(C and Objective-C only)}
4298 @opindex Wno-implicit
4299 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4300 This warning is enabled by @option{-Wall}.
4302 @item -Wimplicit-fallthrough
4303 @opindex Wimplicit-fallthrough
4304 @opindex Wno-implicit-fallthrough
4305 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4306 and @option{-Wno-implicit-fallthrough} is the same as
4307 @option{-Wimplicit-fallthrough=0}.
4309 @item -Wimplicit-fallthrough=@var{n}
4310 @opindex Wimplicit-fallthrough=
4311 Warn when a switch case falls through. For example:
4329 This warning does not warn when the last statement of a case cannot
4330 fall through, e.g. when there is a return statement or a call to function
4331 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4332 also takes into account control flow statements, such as ifs, and only
4333 warns when appropriate. E.g.@:
4343 @} else if (i < 1) @{
4353 Since there are occasions where a switch case fall through is desirable,
4354 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4355 to be used along with a null statement to suppress this warning that
4356 would normally occur:
4364 __attribute__ ((fallthrough));
4371 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4372 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4373 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4374 Instead of these attributes, it is also possible to add a fallthrough comment
4375 to silence the warning. The whole body of the C or C++ style comment should
4376 match the given regular expressions listed below. The option argument @var{n}
4377 specifies what kind of comments are accepted:
4381 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4383 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4384 expression, any comment is used as fallthrough comment.
4386 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4387 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4389 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4390 following regular expressions:
4394 @item @code{-fallthrough}
4396 @item @code{@@fallthrough@@}
4398 @item @code{lint -fallthrough[ \t]*}
4400 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4402 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4404 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4408 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4409 following regular expressions:
4413 @item @code{-fallthrough}
4415 @item @code{@@fallthrough@@}
4417 @item @code{lint -fallthrough[ \t]*}
4419 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4423 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4424 fallthrough comments, only attributes disable the warning.
4428 The comment needs to be followed after optional whitespace and other comments
4429 by @code{case} or @code{default} keywords or by a user label that precedes some
4430 @code{case} or @code{default} label.
4445 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4447 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4448 @opindex Wif-not-aligned
4449 @opindex Wno-if-not-aligned
4450 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4451 should be issued. This is is enabled by default.
4452 Use @option{-Wno-if-not-aligned} to disable it.
4454 @item -Wignored-qualifiers @r{(C and C++ only)}
4455 @opindex Wignored-qualifiers
4456 @opindex Wno-ignored-qualifiers
4457 Warn if the return type of a function has a type qualifier
4458 such as @code{const}. For ISO C such a type qualifier has no effect,
4459 since the value returned by a function is not an lvalue.
4460 For C++, the warning is only emitted for scalar types or @code{void}.
4461 ISO C prohibits qualified @code{void} return types on function
4462 definitions, so such return types always receive a warning
4463 even without this option.
4465 This warning is also enabled by @option{-Wextra}.
4467 @item -Wignored-attributes @r{(C and C++ only)}
4468 @opindex Wignored-attributes
4469 @opindex Wno-ignored-attributes
4470 Warn when an attribute is ignored. This is different from the
4471 @option{-Wattributes} option in that it warns whenever the compiler decides
4472 to drop an attribute, not that the attribute is either unknown, used in a
4473 wrong place, etc. This warning is enabled by default.
4478 Warn if the type of @code{main} is suspicious. @code{main} should be
4479 a function with external linkage, returning int, taking either zero
4480 arguments, two, or three arguments of appropriate types. This warning
4481 is enabled by default in C++ and is enabled by either @option{-Wall}
4482 or @option{-Wpedantic}.
4484 @item -Wmisleading-indentation @r{(C and C++ only)}
4485 @opindex Wmisleading-indentation
4486 @opindex Wno-misleading-indentation
4487 Warn when the indentation of the code does not reflect the block structure.
4488 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4489 @code{for} clauses with a guarded statement that does not use braces,
4490 followed by an unguarded statement with the same indentation.
4492 In the following example, the call to ``bar'' is misleadingly indented as
4493 if it were guarded by the ``if'' conditional.
4496 if (some_condition ())
4498 bar (); /* Gotcha: this is not guarded by the "if". */
4501 In the case of mixed tabs and spaces, the warning uses the
4502 @option{-ftabstop=} option to determine if the statements line up
4505 The warning is not issued for code involving multiline preprocessor logic
4506 such as the following example.
4511 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4517 The warning is not issued after a @code{#line} directive, since this
4518 typically indicates autogenerated code, and no assumptions can be made
4519 about the layout of the file that the directive references.
4521 This warning is enabled by @option{-Wall} in C and C++.
4523 @item -Wmissing-braces
4524 @opindex Wmissing-braces
4525 @opindex Wno-missing-braces
4526 Warn if an aggregate or union initializer is not fully bracketed. In
4527 the following example, the initializer for @code{a} is not fully
4528 bracketed, but that for @code{b} is fully bracketed. This warning is
4529 enabled by @option{-Wall} in C.
4532 int a[2][2] = @{ 0, 1, 2, 3 @};
4533 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4536 This warning is enabled by @option{-Wall}.
4538 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4539 @opindex Wmissing-include-dirs
4540 @opindex Wno-missing-include-dirs
4541 Warn if a user-supplied include directory does not exist.
4543 @item -Wmultistatement-macros
4544 @opindex Wmultistatement-macros
4545 @opindex Wno-multistatement-macros
4546 Warn about unsafe multiple statement macros that appear to be guarded
4547 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
4548 @code{while}, in which only the first statement is actually guarded after
4549 the macro is expanded.
4554 #define DOIT x++; y++
4559 will increment @code{y} unconditionally, not just when @code{c} holds.
4560 The can usually be fixed by wrapping the macro in a do-while loop:
4562 #define DOIT do @{ x++; y++; @} while (0)
4567 This warning is enabled by @option{-Wall} in C and C++.
4570 @opindex Wparentheses
4571 @opindex Wno-parentheses
4572 Warn if parentheses are omitted in certain contexts, such
4573 as when there is an assignment in a context where a truth value
4574 is expected, or when operators are nested whose precedence people
4575 often get confused about.
4577 Also warn if a comparison like @code{x<=y<=z} appears; this is
4578 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4579 interpretation from that of ordinary mathematical notation.
4581 Also warn for dangerous uses of the GNU extension to
4582 @code{?:} with omitted middle operand. When the condition
4583 in the @code{?}: operator is a boolean expression, the omitted value is
4584 always 1. Often programmers expect it to be a value computed
4585 inside the conditional expression instead.
4587 For C++ this also warns for some cases of unnecessary parentheses in
4588 declarations, which can indicate an attempt at a function call instead
4592 // Declares a local variable called mymutex.
4593 std::unique_lock<std::mutex> (mymutex);
4594 // User meant std::unique_lock<std::mutex> lock (mymutex);
4598 This warning is enabled by @option{-Wall}.
4600 @item -Wsequence-point
4601 @opindex Wsequence-point
4602 @opindex Wno-sequence-point
4603 Warn about code that may have undefined semantics because of violations
4604 of sequence point rules in the C and C++ standards.
4606 The C and C++ standards define the order in which expressions in a C/C++
4607 program are evaluated in terms of @dfn{sequence points}, which represent
4608 a partial ordering between the execution of parts of the program: those
4609 executed before the sequence point, and those executed after it. These
4610 occur after the evaluation of a full expression (one which is not part
4611 of a larger expression), after the evaluation of the first operand of a
4612 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4613 function is called (but after the evaluation of its arguments and the
4614 expression denoting the called function), and in certain other places.
4615 Other than as expressed by the sequence point rules, the order of
4616 evaluation of subexpressions of an expression is not specified. All
4617 these rules describe only a partial order rather than a total order,
4618 since, for example, if two functions are called within one expression
4619 with no sequence point between them, the order in which the functions
4620 are called is not specified. However, the standards committee have
4621 ruled that function calls do not overlap.
4623 It is not specified when between sequence points modifications to the
4624 values of objects take effect. Programs whose behavior depends on this
4625 have undefined behavior; the C and C++ standards specify that ``Between
4626 the previous and next sequence point an object shall have its stored
4627 value modified at most once by the evaluation of an expression.
4628 Furthermore, the prior value shall be read only to determine the value
4629 to be stored.''. If a program breaks these rules, the results on any
4630 particular implementation are entirely unpredictable.
4632 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4633 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4634 diagnosed by this option, and it may give an occasional false positive
4635 result, but in general it has been found fairly effective at detecting
4636 this sort of problem in programs.
4638 The C++17 standard will define the order of evaluation of operands in
4639 more cases: in particular it requires that the right-hand side of an
4640 assignment be evaluated before the left-hand side, so the above
4641 examples are no longer undefined. But this warning will still warn
4642 about them, to help people avoid writing code that is undefined in C
4643 and earlier revisions of C++.
4645 The standard is worded confusingly, therefore there is some debate
4646 over the precise meaning of the sequence point rules in subtle cases.
4647 Links to discussions of the problem, including proposed formal
4648 definitions, may be found on the GCC readings page, at
4649 @uref{http://gcc.gnu.org/@/readings.html}.
4651 This warning is enabled by @option{-Wall} for C and C++.
4653 @item -Wno-return-local-addr
4654 @opindex Wno-return-local-addr
4655 @opindex Wreturn-local-addr
4656 Do not warn about returning a pointer (or in C++, a reference) to a
4657 variable that goes out of scope after the function returns.
4660 @opindex Wreturn-type
4661 @opindex Wno-return-type
4662 Warn whenever a function is defined with a return type that defaults
4663 to @code{int}. Also warn about any @code{return} statement with no
4664 return value in a function whose return type is not @code{void}
4665 (falling off the end of the function body is considered returning
4668 For C only, warn about a @code{return} statement with an expression in a
4669 function whose return type is @code{void}, unless the expression type is
4670 also @code{void}. As a GNU extension, the latter case is accepted
4671 without a warning unless @option{-Wpedantic} is used.
4673 For C++, a function without return type always produces a diagnostic
4674 message, even when @option{-Wno-return-type} is specified. The only
4675 exceptions are @code{main} and functions defined in system headers.
4677 This warning is enabled by @option{-Wall}.
4679 @item -Wshift-count-negative
4680 @opindex Wshift-count-negative
4681 @opindex Wno-shift-count-negative
4682 Warn if shift count is negative. This warning is enabled by default.
4684 @item -Wshift-count-overflow
4685 @opindex Wshift-count-overflow
4686 @opindex Wno-shift-count-overflow
4687 Warn if shift count >= width of type. This warning is enabled by default.
4689 @item -Wshift-negative-value
4690 @opindex Wshift-negative-value
4691 @opindex Wno-shift-negative-value
4692 Warn if left shifting a negative value. This warning is enabled by
4693 @option{-Wextra} in C99 and C++11 modes (and newer).
4695 @item -Wshift-overflow
4696 @itemx -Wshift-overflow=@var{n}
4697 @opindex Wshift-overflow
4698 @opindex Wno-shift-overflow
4699 Warn about left shift overflows. This warning is enabled by
4700 default in C99 and C++11 modes (and newer).
4703 @item -Wshift-overflow=1
4704 This is the warning level of @option{-Wshift-overflow} and is enabled
4705 by default in C99 and C++11 modes (and newer). This warning level does
4706 not warn about left-shifting 1 into the sign bit. (However, in C, such
4707 an overflow is still rejected in contexts where an integer constant expression
4710 @item -Wshift-overflow=2
4711 This warning level also warns about left-shifting 1 into the sign bit,
4712 unless C++14 mode is active.
4718 Warn whenever a @code{switch} statement has an index of enumerated type
4719 and lacks a @code{case} for one or more of the named codes of that
4720 enumeration. (The presence of a @code{default} label prevents this
4721 warning.) @code{case} labels outside the enumeration range also
4722 provoke warnings when this option is used (even if there is a
4723 @code{default} label).
4724 This warning is enabled by @option{-Wall}.
4726 @item -Wswitch-default
4727 @opindex Wswitch-default
4728 @opindex Wno-switch-default
4729 Warn whenever a @code{switch} statement does not have a @code{default}
4733 @opindex Wswitch-enum
4734 @opindex Wno-switch-enum
4735 Warn whenever a @code{switch} statement has an index of enumerated type
4736 and lacks a @code{case} for one or more of the named codes of that
4737 enumeration. @code{case} labels outside the enumeration range also
4738 provoke warnings when this option is used. The only difference
4739 between @option{-Wswitch} and this option is that this option gives a
4740 warning about an omitted enumeration code even if there is a
4741 @code{default} label.
4744 @opindex Wswitch-bool
4745 @opindex Wno-switch-bool
4746 Warn whenever a @code{switch} statement has an index of boolean type
4747 and the case values are outside the range of a boolean type.
4748 It is possible to suppress this warning by casting the controlling
4749 expression to a type other than @code{bool}. For example:
4752 switch ((int) (a == 4))
4758 This warning is enabled by default for C and C++ programs.
4760 @item -Wswitch-unreachable
4761 @opindex Wswitch-unreachable
4762 @opindex Wno-switch-unreachable
4763 Warn whenever a @code{switch} statement contains statements between the
4764 controlling expression and the first case label, which will never be
4765 executed. For example:
4777 @option{-Wswitch-unreachable} does not warn if the statement between the
4778 controlling expression and the first case label is just a declaration:
4791 This warning is enabled by default for C and C++ programs.
4793 @item -Wsync-nand @r{(C and C++ only)}
4795 @opindex Wno-sync-nand
4796 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4797 built-in functions are used. These functions changed semantics in GCC 4.4.
4799 @item -Wunused-but-set-parameter
4800 @opindex Wunused-but-set-parameter
4801 @opindex Wno-unused-but-set-parameter
4802 Warn whenever a function parameter is assigned to, but otherwise unused
4803 (aside from its declaration).
4805 To suppress this warning use the @code{unused} attribute
4806 (@pxref{Variable Attributes}).
4808 This warning is also enabled by @option{-Wunused} together with
4811 @item -Wunused-but-set-variable
4812 @opindex Wunused-but-set-variable
4813 @opindex Wno-unused-but-set-variable
4814 Warn whenever a local variable is assigned to, but otherwise unused
4815 (aside from its declaration).
4816 This warning is enabled by @option{-Wall}.
4818 To suppress this warning use the @code{unused} attribute
4819 (@pxref{Variable Attributes}).
4821 This warning is also enabled by @option{-Wunused}, which is enabled
4824 @item -Wunused-function
4825 @opindex Wunused-function
4826 @opindex Wno-unused-function
4827 Warn whenever a static function is declared but not defined or a
4828 non-inline static function is unused.
4829 This warning is enabled by @option{-Wall}.
4831 @item -Wunused-label
4832 @opindex Wunused-label
4833 @opindex Wno-unused-label
4834 Warn whenever a label is declared but not used.
4835 This warning is enabled by @option{-Wall}.
4837 To suppress this warning use the @code{unused} attribute
4838 (@pxref{Variable Attributes}).
4840 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4841 @opindex Wunused-local-typedefs
4842 Warn when a typedef locally defined in a function is not used.
4843 This warning is enabled by @option{-Wall}.
4845 @item -Wunused-parameter
4846 @opindex Wunused-parameter
4847 @opindex Wno-unused-parameter
4848 Warn whenever a function parameter is unused aside from its declaration.
4850 To suppress this warning use the @code{unused} attribute
4851 (@pxref{Variable Attributes}).
4853 @item -Wno-unused-result
4854 @opindex Wunused-result
4855 @opindex Wno-unused-result
4856 Do not warn if a caller of a function marked with attribute
4857 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4858 its return value. The default is @option{-Wunused-result}.
4860 @item -Wunused-variable
4861 @opindex Wunused-variable
4862 @opindex Wno-unused-variable
4863 Warn whenever a local or static variable is unused aside from its
4864 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4865 but not for C++. This warning is enabled by @option{-Wall}.
4867 To suppress this warning use the @code{unused} attribute
4868 (@pxref{Variable Attributes}).
4870 @item -Wunused-const-variable
4871 @itemx -Wunused-const-variable=@var{n}
4872 @opindex Wunused-const-variable
4873 @opindex Wno-unused-const-variable
4874 Warn whenever a constant static variable is unused aside from its declaration.
4875 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4876 for C, but not for C++. In C this declares variable storage, but in C++ this
4877 is not an error since const variables take the place of @code{#define}s.
4879 To suppress this warning use the @code{unused} attribute
4880 (@pxref{Variable Attributes}).
4883 @item -Wunused-const-variable=1
4884 This is the warning level that is enabled by @option{-Wunused-variable} for
4885 C. It warns only about unused static const variables defined in the main
4886 compilation unit, but not about static const variables declared in any
4889 @item -Wunused-const-variable=2
4890 This warning level also warns for unused constant static variables in
4891 headers (excluding system headers). This is the warning level of
4892 @option{-Wunused-const-variable} and must be explicitly requested since
4893 in C++ this isn't an error and in C it might be harder to clean up all
4897 @item -Wunused-value
4898 @opindex Wunused-value
4899 @opindex Wno-unused-value
4900 Warn whenever a statement computes a result that is explicitly not
4901 used. To suppress this warning cast the unused expression to
4902 @code{void}. This includes an expression-statement or the left-hand
4903 side of a comma expression that contains no side effects. For example,
4904 an expression such as @code{x[i,j]} causes a warning, while
4905 @code{x[(void)i,j]} does not.
4907 This warning is enabled by @option{-Wall}.
4912 All the above @option{-Wunused} options combined.
4914 In order to get a warning about an unused function parameter, you must
4915 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4916 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4918 @item -Wuninitialized
4919 @opindex Wuninitialized
4920 @opindex Wno-uninitialized
4921 Warn if an automatic variable is used without first being initialized
4922 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4923 warn if a non-static reference or non-static @code{const} member
4924 appears in a class without constructors.
4926 If you want to warn about code that uses the uninitialized value of the
4927 variable in its own initializer, use the @option{-Winit-self} option.
4929 These warnings occur for individual uninitialized or clobbered
4930 elements of structure, union or array variables as well as for
4931 variables that are uninitialized or clobbered as a whole. They do
4932 not occur for variables or elements declared @code{volatile}. Because
4933 these warnings depend on optimization, the exact variables or elements
4934 for which there are warnings depends on the precise optimization
4935 options and version of GCC used.
4937 Note that there may be no warning about a variable that is used only
4938 to compute a value that itself is never used, because such
4939 computations may be deleted by data flow analysis before the warnings
4942 @item -Winvalid-memory-model
4943 @opindex Winvalid-memory-model
4944 @opindex Wno-invalid-memory-model
4945 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4946 and the C11 atomic generic functions with a memory consistency argument
4947 that is either invalid for the operation or outside the range of values
4948 of the @code{memory_order} enumeration. For example, since the
4949 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4950 defined for the relaxed, release, and sequentially consistent memory
4951 orders the following code is diagnosed:
4956 __atomic_store_n (i, 0, memory_order_consume);
4960 @option{-Winvalid-memory-model} is enabled by default.
4962 @item -Wmaybe-uninitialized
4963 @opindex Wmaybe-uninitialized
4964 @opindex Wno-maybe-uninitialized
4965 For an automatic (i.e.@ local) variable, if there exists a path from the
4966 function entry to a use of the variable that is initialized, but there exist
4967 some other paths for which the variable is not initialized, the compiler
4968 emits a warning if it cannot prove the uninitialized paths are not
4969 executed at run time.
4971 These warnings are only possible in optimizing compilation, because otherwise
4972 GCC does not keep track of the state of variables.
4974 These warnings are made optional because GCC may not be able to determine when
4975 the code is correct in spite of appearing to have an error. Here is one
4976 example of how this can happen:
4996 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4997 always initialized, but GCC doesn't know this. To suppress the
4998 warning, you need to provide a default case with assert(0) or
5001 @cindex @code{longjmp} warnings
5002 This option also warns when a non-volatile automatic variable might be
5003 changed by a call to @code{longjmp}.
5004 The compiler sees only the calls to @code{setjmp}. It cannot know
5005 where @code{longjmp} will be called; in fact, a signal handler could
5006 call it at any point in the code. As a result, you may get a warning
5007 even when there is in fact no problem because @code{longjmp} cannot
5008 in fact be called at the place that would cause a problem.
5010 Some spurious warnings can be avoided if you declare all the functions
5011 you use that never return as @code{noreturn}. @xref{Function
5014 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5016 @item -Wunknown-pragmas
5017 @opindex Wunknown-pragmas
5018 @opindex Wno-unknown-pragmas
5019 @cindex warning for unknown pragmas
5020 @cindex unknown pragmas, warning
5021 @cindex pragmas, warning of unknown
5022 Warn when a @code{#pragma} directive is encountered that is not understood by
5023 GCC@. If this command-line option is used, warnings are even issued
5024 for unknown pragmas in system header files. This is not the case if
5025 the warnings are only enabled by the @option{-Wall} command-line option.
5028 @opindex Wno-pragmas
5030 Do not warn about misuses of pragmas, such as incorrect parameters,
5031 invalid syntax, or conflicts between pragmas. See also
5032 @option{-Wunknown-pragmas}.
5034 @item -Wstrict-aliasing
5035 @opindex Wstrict-aliasing
5036 @opindex Wno-strict-aliasing
5037 This option is only active when @option{-fstrict-aliasing} is active.
5038 It warns about code that might break the strict aliasing rules that the
5039 compiler is using for optimization. The warning does not catch all
5040 cases, but does attempt to catch the more common pitfalls. It is
5041 included in @option{-Wall}.
5042 It is equivalent to @option{-Wstrict-aliasing=3}
5044 @item -Wstrict-aliasing=n
5045 @opindex Wstrict-aliasing=n
5046 This option is only active when @option{-fstrict-aliasing} is active.
5047 It warns about code that might break the strict aliasing rules that the
5048 compiler is using for optimization.
5049 Higher levels correspond to higher accuracy (fewer false positives).
5050 Higher levels also correspond to more effort, similar to the way @option{-O}
5052 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5054 Level 1: Most aggressive, quick, least accurate.
5055 Possibly useful when higher levels
5056 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5057 false negatives. However, it has many false positives.
5058 Warns for all pointer conversions between possibly incompatible types,
5059 even if never dereferenced. Runs in the front end only.
5061 Level 2: Aggressive, quick, not too precise.
5062 May still have many false positives (not as many as level 1 though),
5063 and few false negatives (but possibly more than level 1).
5064 Unlike level 1, it only warns when an address is taken. Warns about
5065 incomplete types. Runs in the front end only.
5067 Level 3 (default for @option{-Wstrict-aliasing}):
5068 Should have very few false positives and few false
5069 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5070 Takes care of the common pun+dereference pattern in the front end:
5071 @code{*(int*)&some_float}.
5072 If optimization is enabled, it also runs in the back end, where it deals
5073 with multiple statement cases using flow-sensitive points-to information.
5074 Only warns when the converted pointer is dereferenced.
5075 Does not warn about incomplete types.
5077 @item -Wstrict-overflow
5078 @itemx -Wstrict-overflow=@var{n}
5079 @opindex Wstrict-overflow
5080 @opindex Wno-strict-overflow
5081 This option is only active when signed overflow is undefined.
5082 It warns about cases where the compiler optimizes based on the
5083 assumption that signed overflow does not occur. Note that it does not
5084 warn about all cases where the code might overflow: it only warns
5085 about cases where the compiler implements some optimization. Thus
5086 this warning depends on the optimization level.
5088 An optimization that assumes that signed overflow does not occur is
5089 perfectly safe if the values of the variables involved are such that
5090 overflow never does, in fact, occur. Therefore this warning can
5091 easily give a false positive: a warning about code that is not
5092 actually a problem. To help focus on important issues, several
5093 warning levels are defined. No warnings are issued for the use of
5094 undefined signed overflow when estimating how many iterations a loop
5095 requires, in particular when determining whether a loop will be
5099 @item -Wstrict-overflow=1
5100 Warn about cases that are both questionable and easy to avoid. For
5101 example the compiler simplifies
5102 @code{x + 1 > x} to @code{1}. This level of
5103 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5104 are not, and must be explicitly requested.
5106 @item -Wstrict-overflow=2
5107 Also warn about other cases where a comparison is simplified to a
5108 constant. For example: @code{abs (x) >= 0}. This can only be
5109 simplified when signed integer overflow is undefined, because
5110 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5111 zero. @option{-Wstrict-overflow} (with no level) is the same as
5112 @option{-Wstrict-overflow=2}.
5114 @item -Wstrict-overflow=3
5115 Also warn about other cases where a comparison is simplified. For
5116 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5118 @item -Wstrict-overflow=4
5119 Also warn about other simplifications not covered by the above cases.
5120 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5122 @item -Wstrict-overflow=5
5123 Also warn about cases where the compiler reduces the magnitude of a
5124 constant involved in a comparison. For example: @code{x + 2 > y} is
5125 simplified to @code{x + 1 >= y}. This is reported only at the
5126 highest warning level because this simplification applies to many
5127 comparisons, so this warning level gives a very large number of
5131 @item -Wstringop-overflow
5132 @itemx -Wstringop-overflow=@var{type}
5133 @opindex Wstringop-overflow
5134 @opindex Wno-stringop-overflow
5135 Warn for calls to string manipulation functions such as @code{memcpy} and
5136 @code{strcpy} that are determined to overflow the destination buffer. The
5137 optional argument is one greater than the type of Object Size Checking to
5138 perform to determine the size of the destination. @xref{Object Size Checking}.
5139 The argument is meaningful only for functions that operate on character arrays
5140 but not for raw memory functions like @code{memcpy} which always make use
5141 of Object Size type-0. The option also warns for calls that specify a size
5142 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5143 The option produces the best results with optimization enabled but can detect
5144 a small subset of simple buffer overflows even without optimization in
5145 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5146 correspond to the standard functions. In any case, the option warns about
5147 just a subset of buffer overflows detected by the corresponding overflow
5148 checking built-ins. For example, the option will issue a warning for
5149 the @code{strcpy} call below because it copies at least 5 characters
5150 (the string @code{"blue"} including the terminating NUL) into the buffer
5154 enum Color @{ blue, purple, yellow @};
5155 const char* f (enum Color clr)
5157 static char buf [4];
5161 case blue: str = "blue"; break;
5162 case purple: str = "purple"; break;
5163 case yellow: str = "yellow"; break;
5166 return strcpy (buf, str); // warning here
5170 Option @option{-Wstringop-overflow=2} is enabled by default.
5173 @item -Wstringop-overflow
5174 @item -Wstringop-overflow=1
5175 @opindex Wstringop-overflow
5176 @opindex Wno-stringop-overflow
5177 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5178 to determine the sizes of destination objects. This is the default setting
5179 of the option. At this setting the option will not warn for writes past
5180 the end of subobjects of larger objects accessed by pointers unless the
5181 size of the largest surrounding object is known. When the destination may
5182 be one of several objects it is assumed to be the largest one of them. On
5183 Linux systems, when optimization is enabled at this setting the option warns
5184 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5187 @item -Wstringop-overflow=2
5188 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5189 to determine the sizes of destination objects. At this setting the option
5190 will warn about overflows when writing to members of the largest complete
5191 objects whose exact size is known. It will, however, not warn for excessive
5192 writes to the same members of unknown objects referenced by pointers since
5193 they may point to arrays containing unknown numbers of elements.
5195 @item -Wstringop-overflow=3
5196 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5197 to determine the sizes of destination objects. At this setting the option
5198 warns about overflowing the smallest object or data member. This is the
5199 most restrictive setting of the option that may result in warnings for safe
5202 @item -Wstringop-overflow=4
5203 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5204 to determine the sizes of destination objects. At this setting the option
5205 will warn about overflowing any data members, and when the destination is
5206 one of several objects it uses the size of the largest of them to decide
5207 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5208 setting of the option may result in warnings for benign code.
5211 @item -Wstringop-truncation
5212 @opindex Wstringop-truncation
5213 @opindex Wno-stringop-truncation
5214 Warn for calls to bounded string manipulation functions such as @code{strncat},
5215 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5216 or leave the destination unchanged.
5218 In the following example, the call to @code{strncat} specifies a bound that
5219 is less than the length of the source string. As a result, the copy of
5220 the source will be truncated and so the call is diagnosed. To avoid the
5221 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5224 void append (char *buf, size_t bufsize)
5226 strncat (buf, ".txt", 3);
5230 As another example, the following call to @code{strncpy} results in copying
5231 to @code{d} just the characters preceding the terminating NUL, without
5232 appending the NUL to the end. Assuming the result of @code{strncpy} is
5233 necessarily a NUL-terminated string is a common mistake, and so the call
5234 is diagnosed. To avoid the warning when the result is not expected to be
5235 NUL-terminated, call @code{memcpy} instead.
5238 void copy (char *d, const char *s)
5240 strncpy (d, s, strlen (s));
5244 In the following example, the call to @code{strncpy} specifies the size
5245 of the destination buffer as the bound. If the length of the source
5246 string is equal to or greater than this size the result of the copy will
5247 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5248 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5249 element of the buffer to @code{NUL}.
5252 void copy (const char *s)
5255 strncpy (buf, s, sizeof buf);
5260 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5261 @opindex Wsuggest-attribute=
5262 @opindex Wno-suggest-attribute=
5263 Warn for cases where adding an attribute may be beneficial. The
5264 attributes currently supported are listed below.
5267 @item -Wsuggest-attribute=pure
5268 @itemx -Wsuggest-attribute=const
5269 @itemx -Wsuggest-attribute=noreturn
5270 @itemx -Wsuggest-attribute=malloc
5271 @opindex Wsuggest-attribute=pure
5272 @opindex Wno-suggest-attribute=pure
5273 @opindex Wsuggest-attribute=const
5274 @opindex Wno-suggest-attribute=const
5275 @opindex Wsuggest-attribute=noreturn
5276 @opindex Wno-suggest-attribute=noreturn
5277 @opindex Wsuggest-attribute=malloc
5278 @opindex Wno-suggest-attribute=malloc
5280 Warn about functions that might be candidates for attributes
5281 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5282 only warns for functions visible in other compilation units or (in the case of
5283 @code{pure} and @code{const}) if it cannot prove that the function returns
5284 normally. A function returns normally if it doesn't contain an infinite loop or
5285 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5286 requires option @option{-fipa-pure-const}, which is enabled by default at
5287 @option{-O} and higher. Higher optimization levels improve the accuracy
5290 @item -Wsuggest-attribute=format
5291 @itemx -Wmissing-format-attribute
5292 @opindex Wsuggest-attribute=format
5293 @opindex Wmissing-format-attribute
5294 @opindex Wno-suggest-attribute=format
5295 @opindex Wno-missing-format-attribute
5299 Warn about function pointers that might be candidates for @code{format}
5300 attributes. Note these are only possible candidates, not absolute ones.
5301 GCC guesses that function pointers with @code{format} attributes that
5302 are used in assignment, initialization, parameter passing or return
5303 statements should have a corresponding @code{format} attribute in the
5304 resulting type. I.e.@: the left-hand side of the assignment or
5305 initialization, the type of the parameter variable, or the return type
5306 of the containing function respectively should also have a @code{format}
5307 attribute to avoid the warning.
5309 GCC also warns about function definitions that might be
5310 candidates for @code{format} attributes. Again, these are only
5311 possible candidates. GCC guesses that @code{format} attributes
5312 might be appropriate for any function that calls a function like
5313 @code{vprintf} or @code{vscanf}, but this might not always be the
5314 case, and some functions for which @code{format} attributes are
5315 appropriate may not be detected.
5317 @item -Wsuggest-attribute=cold
5318 @opindex Wsuggest-attribute=cold
5319 @opindex Wno-suggest-attribute=cold
5321 Warn about functions that might be candidates for @code{cold} attribute. This
5322 is based on static detection and generally will only warn about functions which
5323 always leads to a call to another @code{cold} function such as wrappers of
5324 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5327 @item -Wsuggest-final-types
5328 @opindex Wno-suggest-final-types
5329 @opindex Wsuggest-final-types
5330 Warn about types with virtual methods where code quality would be improved
5331 if the type were declared with the C++11 @code{final} specifier,
5333 declared in an anonymous namespace. This allows GCC to more aggressively
5334 devirtualize the polymorphic calls. This warning is more effective with link
5335 time optimization, where the information about the class hierarchy graph is
5338 @item -Wsuggest-final-methods
5339 @opindex Wno-suggest-final-methods
5340 @opindex Wsuggest-final-methods
5341 Warn about virtual methods where code quality would be improved if the method
5342 were declared with the C++11 @code{final} specifier,
5343 or, if possible, its type were
5344 declared in an anonymous namespace or with the @code{final} specifier.
5346 more effective with link-time optimization, where the information about the
5347 class hierarchy graph is more complete. It is recommended to first consider
5348 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5351 @item -Wsuggest-override
5352 Warn about overriding virtual functions that are not marked with the override
5356 @opindex Wno-alloc-zero
5357 @opindex Walloc-zero
5358 Warn about calls to allocation functions decorated with attribute
5359 @code{alloc_size} that specify zero bytes, including those to the built-in
5360 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5361 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5362 when called with a zero size differs among implementations (and in the case
5363 of @code{realloc} has been deprecated) relying on it may result in subtle
5364 portability bugs and should be avoided.
5366 @item -Walloc-size-larger-than=@var{n}
5367 Warn about calls to functions decorated with attribute @code{alloc_size}
5368 that attempt to allocate objects larger than the specified number of bytes,
5369 or where the result of the size computation in an integer type with infinite
5370 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5371 may end in one of the standard suffixes designating a multiple of bytes
5372 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5373 @code{MB} and @code{MiB} for megabyte and mebibyte, and so on.
5374 @xref{Function Attributes}.
5379 This option warns on all uses of @code{alloca} in the source.
5381 @item -Walloca-larger-than=@var{n}
5382 This option warns on calls to @code{alloca} that are not bounded by a
5383 controlling predicate limiting its argument of integer type to at most
5384 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5385 Arguments of non-integer types are considered unbounded even if they
5386 appear to be constrained to the expected range.
5388 For example, a bounded case of @code{alloca} could be:
5391 void func (size_t n)
5402 In the above example, passing @code{-Walloca-larger-than=1000} would not
5403 issue a warning because the call to @code{alloca} is known to be at most
5404 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5405 the compiler would emit a warning.
5407 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5408 controlling predicate constraining its integer argument. For example:
5413 void *p = alloca (n);
5418 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5419 a warning, but this time because of the lack of bounds checking.
5421 Note, that even seemingly correct code involving signed integers could
5425 void func (signed int n)
5435 In the above example, @var{n} could be negative, causing a larger than
5436 expected argument to be implicitly cast into the @code{alloca} call.
5438 This option also warns when @code{alloca} is used in a loop.
5440 This warning is not enabled by @option{-Wall}, and is only active when
5441 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5443 See also @option{-Wvla-larger-than=@var{n}}.
5445 @item -Warray-bounds
5446 @itemx -Warray-bounds=@var{n}
5447 @opindex Wno-array-bounds
5448 @opindex Warray-bounds
5449 This option is only active when @option{-ftree-vrp} is active
5450 (default for @option{-O2} and above). It warns about subscripts to arrays
5451 that are always out of bounds. This warning is enabled by @option{-Wall}.
5454 @item -Warray-bounds=1
5455 This is the warning level of @option{-Warray-bounds} and is enabled
5456 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5458 @item -Warray-bounds=2
5459 This warning level also warns about out of bounds access for
5460 arrays at the end of a struct and for arrays accessed through
5461 pointers. This warning level may give a larger number of
5462 false positives and is deactivated by default.
5465 @item -Wattribute-alias
5466 Warn about declarations using the @code{alias} and similar attributes whose
5467 target is incompatible with the type of the alias. @xref{Function Attributes,
5468 ,Declaring Attributes of Functions}.
5470 @item -Wbool-compare
5471 @opindex Wno-bool-compare
5472 @opindex Wbool-compare
5473 Warn about boolean expression compared with an integer value different from
5474 @code{true}/@code{false}. For instance, the following comparison is
5479 if ((n > 1) == 2) @{ @dots{} @}
5481 This warning is enabled by @option{-Wall}.
5483 @item -Wbool-operation
5484 @opindex Wno-bool-operation
5485 @opindex Wbool-operation
5486 Warn about suspicious operations on expressions of a boolean type. For
5487 instance, bitwise negation of a boolean is very likely a bug in the program.
5488 For C, this warning also warns about incrementing or decrementing a boolean,
5489 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5490 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
5492 This warning is enabled by @option{-Wall}.
5494 @item -Wduplicated-branches
5495 @opindex Wno-duplicated-branches
5496 @opindex Wduplicated-branches
5497 Warn when an if-else has identical branches. This warning detects cases like
5504 It doesn't warn when both branches contain just a null statement. This warning
5505 also warn for conditional operators:
5507 int i = x ? *p : *p;
5510 @item -Wduplicated-cond
5511 @opindex Wno-duplicated-cond
5512 @opindex Wduplicated-cond
5513 Warn about duplicated conditions in an if-else-if chain. For instance,
5514 warn for the following code:
5516 if (p->q != NULL) @{ @dots{} @}
5517 else if (p->q != NULL) @{ @dots{} @}
5520 @item -Wframe-address
5521 @opindex Wno-frame-address
5522 @opindex Wframe-address
5523 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5524 is called with an argument greater than 0. Such calls may return indeterminate
5525 values or crash the program. The warning is included in @option{-Wall}.
5527 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5528 @opindex Wno-discarded-qualifiers
5529 @opindex Wdiscarded-qualifiers
5530 Do not warn if type qualifiers on pointers are being discarded.
5531 Typically, the compiler warns if a @code{const char *} variable is
5532 passed to a function that takes a @code{char *} parameter. This option
5533 can be used to suppress such a warning.
5535 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5536 @opindex Wno-discarded-array-qualifiers
5537 @opindex Wdiscarded-array-qualifiers
5538 Do not warn if type qualifiers on arrays which are pointer targets
5539 are being discarded. Typically, the compiler warns if a
5540 @code{const int (*)[]} variable is passed to a function that
5541 takes a @code{int (*)[]} parameter. This option can be used to
5542 suppress such a warning.
5544 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5545 @opindex Wno-incompatible-pointer-types
5546 @opindex Wincompatible-pointer-types
5547 Do not warn when there is a conversion between pointers that have incompatible
5548 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5549 which warns for pointer argument passing or assignment with different
5552 @item -Wno-int-conversion @r{(C and Objective-C only)}
5553 @opindex Wno-int-conversion
5554 @opindex Wint-conversion
5555 Do not warn about incompatible integer to pointer and pointer to integer
5556 conversions. This warning is about implicit conversions; for explicit
5557 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5558 @option{-Wno-pointer-to-int-cast} may be used.
5560 @item -Wno-div-by-zero
5561 @opindex Wno-div-by-zero
5562 @opindex Wdiv-by-zero
5563 Do not warn about compile-time integer division by zero. Floating-point
5564 division by zero is not warned about, as it can be a legitimate way of
5565 obtaining infinities and NaNs.
5567 @item -Wsystem-headers
5568 @opindex Wsystem-headers
5569 @opindex Wno-system-headers
5570 @cindex warnings from system headers
5571 @cindex system headers, warnings from
5572 Print warning messages for constructs found in system header files.
5573 Warnings from system headers are normally suppressed, on the assumption
5574 that they usually do not indicate real problems and would only make the
5575 compiler output harder to read. Using this command-line option tells
5576 GCC to emit warnings from system headers as if they occurred in user
5577 code. However, note that using @option{-Wall} in conjunction with this
5578 option does @emph{not} warn about unknown pragmas in system
5579 headers---for that, @option{-Wunknown-pragmas} must also be used.
5581 @item -Wtautological-compare
5582 @opindex Wtautological-compare
5583 @opindex Wno-tautological-compare
5584 Warn if a self-comparison always evaluates to true or false. This
5585 warning detects various mistakes such as:
5589 if (i > i) @{ @dots{} @}
5592 This warning also warns about bitwise comparisons that always evaluate
5593 to true or false, for instance:
5595 if ((a & 16) == 10) @{ @dots{} @}
5597 will always be false.
5599 This warning is enabled by @option{-Wall}.
5602 @opindex Wtrampolines
5603 @opindex Wno-trampolines
5604 Warn about trampolines generated for pointers to nested functions.
5605 A trampoline is a small piece of data or code that is created at run
5606 time on the stack when the address of a nested function is taken, and is
5607 used to call the nested function indirectly. For some targets, it is
5608 made up of data only and thus requires no special treatment. But, for
5609 most targets, it is made up of code and thus requires the stack to be
5610 made executable in order for the program to work properly.
5613 @opindex Wfloat-equal
5614 @opindex Wno-float-equal
5615 Warn if floating-point values are used in equality comparisons.
5617 The idea behind this is that sometimes it is convenient (for the
5618 programmer) to consider floating-point values as approximations to
5619 infinitely precise real numbers. If you are doing this, then you need
5620 to compute (by analyzing the code, or in some other way) the maximum or
5621 likely maximum error that the computation introduces, and allow for it
5622 when performing comparisons (and when producing output, but that's a
5623 different problem). In particular, instead of testing for equality, you
5624 should check to see whether the two values have ranges that overlap; and
5625 this is done with the relational operators, so equality comparisons are
5628 @item -Wtraditional @r{(C and Objective-C only)}
5629 @opindex Wtraditional
5630 @opindex Wno-traditional
5631 Warn about certain constructs that behave differently in traditional and
5632 ISO C@. Also warn about ISO C constructs that have no traditional C
5633 equivalent, and/or problematic constructs that should be avoided.
5637 Macro parameters that appear within string literals in the macro body.
5638 In traditional C macro replacement takes place within string literals,
5639 but in ISO C it does not.
5642 In traditional C, some preprocessor directives did not exist.
5643 Traditional preprocessors only considered a line to be a directive
5644 if the @samp{#} appeared in column 1 on the line. Therefore
5645 @option{-Wtraditional} warns about directives that traditional C
5646 understands but ignores because the @samp{#} does not appear as the
5647 first character on the line. It also suggests you hide directives like
5648 @code{#pragma} not understood by traditional C by indenting them. Some
5649 traditional implementations do not recognize @code{#elif}, so this option
5650 suggests avoiding it altogether.
5653 A function-like macro that appears without arguments.
5656 The unary plus operator.
5659 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5660 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5661 constants.) Note, these suffixes appear in macros defined in the system
5662 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5663 Use of these macros in user code might normally lead to spurious
5664 warnings, however GCC's integrated preprocessor has enough context to
5665 avoid warning in these cases.
5668 A function declared external in one block and then used after the end of
5672 A @code{switch} statement has an operand of type @code{long}.
5675 A non-@code{static} function declaration follows a @code{static} one.
5676 This construct is not accepted by some traditional C compilers.
5679 The ISO type of an integer constant has a different width or
5680 signedness from its traditional type. This warning is only issued if
5681 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5682 typically represent bit patterns, are not warned about.
5685 Usage of ISO string concatenation is detected.
5688 Initialization of automatic aggregates.
5691 Identifier conflicts with labels. Traditional C lacks a separate
5692 namespace for labels.
5695 Initialization of unions. If the initializer is zero, the warning is
5696 omitted. This is done under the assumption that the zero initializer in
5697 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5698 initializer warnings and relies on default initialization to zero in the
5702 Conversions by prototypes between fixed/floating-point values and vice
5703 versa. The absence of these prototypes when compiling with traditional
5704 C causes serious problems. This is a subset of the possible
5705 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5708 Use of ISO C style function definitions. This warning intentionally is
5709 @emph{not} issued for prototype declarations or variadic functions
5710 because these ISO C features appear in your code when using
5711 libiberty's traditional C compatibility macros, @code{PARAMS} and
5712 @code{VPARAMS}. This warning is also bypassed for nested functions
5713 because that feature is already a GCC extension and thus not relevant to
5714 traditional C compatibility.
5717 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5718 @opindex Wtraditional-conversion
5719 @opindex Wno-traditional-conversion
5720 Warn if a prototype causes a type conversion that is different from what
5721 would happen to the same argument in the absence of a prototype. This
5722 includes conversions of fixed point to floating and vice versa, and
5723 conversions changing the width or signedness of a fixed-point argument
5724 except when the same as the default promotion.
5726 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5727 @opindex Wdeclaration-after-statement
5728 @opindex Wno-declaration-after-statement
5729 Warn when a declaration is found after a statement in a block. This
5730 construct, known from C++, was introduced with ISO C99 and is by default
5731 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5736 Warn whenever a local variable or type declaration shadows another
5737 variable, parameter, type, class member (in C++), or instance variable
5738 (in Objective-C) or whenever a built-in function is shadowed. Note
5739 that in C++, the compiler warns if a local variable shadows an
5740 explicit typedef, but not if it shadows a struct/class/enum.
5741 Same as @option{-Wshadow=global}.
5743 @item -Wno-shadow-ivar @r{(Objective-C only)}
5744 @opindex Wno-shadow-ivar
5745 @opindex Wshadow-ivar
5746 Do not warn whenever a local variable shadows an instance variable in an
5749 @item -Wshadow=global
5750 @opindex Wshadow=local
5751 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5753 @item -Wshadow=local
5754 @opindex Wshadow=local
5755 Warn when a local variable shadows another local variable or parameter.
5756 This warning is enabled by @option{-Wshadow=global}.
5758 @item -Wshadow=compatible-local
5759 @opindex Wshadow=compatible-local
5760 Warn when a local variable shadows another local variable or parameter
5761 whose type is compatible with that of the shadowing variable. In C++,
5762 type compatibility here means the type of the shadowing variable can be
5763 converted to that of the shadowed variable. The creation of this flag
5764 (in addition to @option{-Wshadow=local}) is based on the idea that when
5765 a local variable shadows another one of incompatible type, it is most
5766 likely intentional, not a bug or typo, as shown in the following example:
5770 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5772 for (int i = 0; i < N; ++i)
5781 Since the two variable @code{i} in the example above have incompatible types,
5782 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5783 Because their types are incompatible, if a programmer accidentally uses one
5784 in place of the other, type checking will catch that and emit an error or
5785 warning. So not warning (about shadowing) in this case will not lead to
5786 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5787 possibly reduce the number of warnings triggered by intentional shadowing.
5789 This warning is enabled by @option{-Wshadow=local}.
5791 @item -Wlarger-than=@var{len}
5792 @opindex Wlarger-than=@var{len}
5793 @opindex Wlarger-than-@var{len}
5794 Warn whenever an object of larger than @var{len} bytes is defined.
5796 @item -Wframe-larger-than=@var{len}
5797 @opindex Wframe-larger-than
5798 Warn if the size of a function frame is larger than @var{len} bytes.
5799 The computation done to determine the stack frame size is approximate
5800 and not conservative.
5801 The actual requirements may be somewhat greater than @var{len}
5802 even if you do not get a warning. In addition, any space allocated
5803 via @code{alloca}, variable-length arrays, or related constructs
5804 is not included by the compiler when determining
5805 whether or not to issue a warning.
5807 @item -Wno-free-nonheap-object
5808 @opindex Wno-free-nonheap-object
5809 @opindex Wfree-nonheap-object
5810 Do not warn when attempting to free an object that was not allocated
5813 @item -Wstack-usage=@var{len}
5814 @opindex Wstack-usage
5815 Warn if the stack usage of a function might be larger than @var{len} bytes.
5816 The computation done to determine the stack usage is conservative.
5817 Any space allocated via @code{alloca}, variable-length arrays, or related
5818 constructs is included by the compiler when determining whether or not to
5821 The message is in keeping with the output of @option{-fstack-usage}.
5825 If the stack usage is fully static but exceeds the specified amount, it's:
5828 warning: stack usage is 1120 bytes
5831 If the stack usage is (partly) dynamic but bounded, it's:
5834 warning: stack usage might be 1648 bytes
5837 If the stack usage is (partly) dynamic and not bounded, it's:
5840 warning: stack usage might be unbounded
5844 @item -Wunsafe-loop-optimizations
5845 @opindex Wunsafe-loop-optimizations
5846 @opindex Wno-unsafe-loop-optimizations
5847 Warn if the loop cannot be optimized because the compiler cannot
5848 assume anything on the bounds of the loop indices. With
5849 @option{-funsafe-loop-optimizations} warn if the compiler makes
5852 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5853 @opindex Wno-pedantic-ms-format
5854 @opindex Wpedantic-ms-format
5855 When used in combination with @option{-Wformat}
5856 and @option{-pedantic} without GNU extensions, this option
5857 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5858 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5859 which depend on the MS runtime.
5862 @opindex Waligned-new
5863 @opindex Wno-aligned-new
5864 Warn about a new-expression of a type that requires greater alignment
5865 than the @code{alignof(std::max_align_t)} but uses an allocation
5866 function without an explicit alignment parameter. This option is
5867 enabled by @option{-Wall}.
5869 Normally this only warns about global allocation functions, but
5870 @option{-Waligned-new=all} also warns about class member allocation
5873 @item -Wplacement-new
5874 @itemx -Wplacement-new=@var{n}
5875 @opindex Wplacement-new
5876 @opindex Wno-placement-new
5877 Warn about placement new expressions with undefined behavior, such as
5878 constructing an object in a buffer that is smaller than the type of
5879 the object. For example, the placement new expression below is diagnosed
5880 because it attempts to construct an array of 64 integers in a buffer only
5886 This warning is enabled by default.
5889 @item -Wplacement-new=1
5890 This is the default warning level of @option{-Wplacement-new}. At this
5891 level the warning is not issued for some strictly undefined constructs that
5892 GCC allows as extensions for compatibility with legacy code. For example,
5893 the following @code{new} expression is not diagnosed at this level even
5894 though it has undefined behavior according to the C++ standard because
5895 it writes past the end of the one-element array.
5897 struct S @{ int n, a[1]; @};
5898 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5899 new (s->a)int [32]();
5902 @item -Wplacement-new=2
5903 At this level, in addition to diagnosing all the same constructs as at level
5904 1, a diagnostic is also issued for placement new expressions that construct
5905 an object in the last member of structure whose type is an array of a single
5906 element and whose size is less than the size of the object being constructed.
5907 While the previous example would be diagnosed, the following construct makes
5908 use of the flexible member array extension to avoid the warning at level 2.
5910 struct S @{ int n, a[]; @};
5911 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5912 new (s->a)int [32]();
5917 @item -Wpointer-arith
5918 @opindex Wpointer-arith
5919 @opindex Wno-pointer-arith
5920 Warn about anything that depends on the ``size of'' a function type or
5921 of @code{void}. GNU C assigns these types a size of 1, for
5922 convenience in calculations with @code{void *} pointers and pointers
5923 to functions. In C++, warn also when an arithmetic operation involves
5924 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5926 @item -Wpointer-compare
5927 @opindex Wpointer-compare
5928 @opindex Wno-pointer-compare
5929 Warn if a pointer is compared with a zero character constant. This usually
5930 means that the pointer was meant to be dereferenced. For example:
5933 const char *p = foo ();
5938 Note that the code above is invalid in C++11.
5940 This warning is enabled by default.
5943 @opindex Wtype-limits
5944 @opindex Wno-type-limits
5945 Warn if a comparison is always true or always false due to the limited
5946 range of the data type, but do not warn for constant expressions. For
5947 example, warn if an unsigned variable is compared against zero with
5948 @code{<} or @code{>=}. This warning is also enabled by
5951 @include cppwarnopts.texi
5953 @item -Wbad-function-cast @r{(C and Objective-C only)}
5954 @opindex Wbad-function-cast
5955 @opindex Wno-bad-function-cast
5956 Warn when a function call is cast to a non-matching type.
5957 For example, warn if a call to a function returning an integer type
5958 is cast to a pointer type.
5960 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5961 @opindex Wc90-c99-compat
5962 @opindex Wno-c90-c99-compat
5963 Warn about features not present in ISO C90, but present in ISO C99.
5964 For instance, warn about use of variable length arrays, @code{long long}
5965 type, @code{bool} type, compound literals, designated initializers, and so
5966 on. This option is independent of the standards mode. Warnings are disabled
5967 in the expression that follows @code{__extension__}.
5969 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5970 @opindex Wc99-c11-compat
5971 @opindex Wno-c99-c11-compat
5972 Warn about features not present in ISO C99, but present in ISO C11.
5973 For instance, warn about use of anonymous structures and unions,
5974 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5975 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5976 and so on. This option is independent of the standards mode. Warnings are
5977 disabled in the expression that follows @code{__extension__}.
5979 @item -Wc++-compat @r{(C and Objective-C only)}
5980 @opindex Wc++-compat
5981 Warn about ISO C constructs that are outside of the common subset of
5982 ISO C and ISO C++, e.g.@: request for implicit conversion from
5983 @code{void *} to a pointer to non-@code{void} type.
5985 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5986 @opindex Wc++11-compat
5987 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5988 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5989 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5990 enabled by @option{-Wall}.
5992 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5993 @opindex Wc++14-compat
5994 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5995 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5997 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
5998 @opindex Wc++17-compat
5999 Warn about C++ constructs whose meaning differs between ISO C++ 2014
6000 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6004 @opindex Wno-cast-qual
6005 Warn whenever a pointer is cast so as to remove a type qualifier from
6006 the target type. For example, warn if a @code{const char *} is cast
6007 to an ordinary @code{char *}.
6009 Also warn when making a cast that introduces a type qualifier in an
6010 unsafe way. For example, casting @code{char **} to @code{const char **}
6011 is unsafe, as in this example:
6014 /* p is char ** value. */
6015 const char **q = (const char **) p;
6016 /* Assignment of readonly string to const char * is OK. */
6018 /* Now char** pointer points to read-only memory. */
6023 @opindex Wcast-align
6024 @opindex Wno-cast-align
6025 Warn whenever a pointer is cast such that the required alignment of the
6026 target is increased. For example, warn if a @code{char *} is cast to
6027 an @code{int *} on machines where integers can only be accessed at
6028 two- or four-byte boundaries.
6030 @item -Wcast-align=strict
6031 @opindex Wcast-align=strict
6032 Warn whenever a pointer is cast such that the required alignment of the
6033 target is increased. For example, warn if a @code{char *} is cast to
6034 an @code{int *} regardless of the target machine.
6036 @item -Wwrite-strings
6037 @opindex Wwrite-strings
6038 @opindex Wno-write-strings
6039 When compiling C, give string constants the type @code{const
6040 char[@var{length}]} so that copying the address of one into a
6041 non-@code{const} @code{char *} pointer produces a warning. These
6042 warnings help you find at compile time code that can try to write
6043 into a string constant, but only if you have been very careful about
6044 using @code{const} in declarations and prototypes. Otherwise, it is
6045 just a nuisance. This is why we did not make @option{-Wall} request
6048 When compiling C++, warn about the deprecated conversion from string
6049 literals to @code{char *}. This warning is enabled by default for C++
6053 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6054 @opindex Wcatch-value
6055 @opindex Wno-catch-value
6056 Warn about catch handlers that do not catch via reference.
6057 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6058 warn about polymorphic class types that are caught by value.
6059 With @option{-Wcatch-value=2} warn about all class types that are caught
6060 by value. With @option{-Wcatch-value=3} warn about all types that are
6061 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6065 @opindex Wno-clobbered
6066 Warn for variables that might be changed by @code{longjmp} or
6067 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6069 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6070 @opindex Wconditionally-supported
6071 @opindex Wno-conditionally-supported
6072 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6075 @opindex Wconversion
6076 @opindex Wno-conversion
6077 Warn for implicit conversions that may alter a value. This includes
6078 conversions between real and integer, like @code{abs (x)} when
6079 @code{x} is @code{double}; conversions between signed and unsigned,
6080 like @code{unsigned ui = -1}; and conversions to smaller types, like
6081 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6082 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6083 changed by the conversion like in @code{abs (2.0)}. Warnings about
6084 conversions between signed and unsigned integers can be disabled by
6085 using @option{-Wno-sign-conversion}.
6087 For C++, also warn for confusing overload resolution for user-defined
6088 conversions; and conversions that never use a type conversion
6089 operator: conversions to @code{void}, the same type, a base class or a
6090 reference to them. Warnings about conversions between signed and
6091 unsigned integers are disabled by default in C++ unless
6092 @option{-Wsign-conversion} is explicitly enabled.
6094 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6095 @opindex Wconversion-null
6096 @opindex Wno-conversion-null
6097 Do not warn for conversions between @code{NULL} and non-pointer
6098 types. @option{-Wconversion-null} is enabled by default.
6100 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6101 @opindex Wzero-as-null-pointer-constant
6102 @opindex Wno-zero-as-null-pointer-constant
6103 Warn when a literal @samp{0} is used as null pointer constant. This can
6104 be useful to facilitate the conversion to @code{nullptr} in C++11.
6106 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6107 @opindex Wsubobject-linkage
6108 @opindex Wno-subobject-linkage
6109 Warn if a class type has a base or a field whose type uses the anonymous
6110 namespace or depends on a type with no linkage. If a type A depends on
6111 a type B with no or internal linkage, defining it in multiple
6112 translation units would be an ODR violation because the meaning of B
6113 is different in each translation unit. If A only appears in a single
6114 translation unit, the best way to silence the warning is to give it
6115 internal linkage by putting it in an anonymous namespace as well. The
6116 compiler doesn't give this warning for types defined in the main .C
6117 file, as those are unlikely to have multiple definitions.
6118 @option{-Wsubobject-linkage} is enabled by default.
6120 @item -Wdangling-else
6121 @opindex Wdangling-else
6122 @opindex Wno-dangling-else
6123 Warn about constructions where there may be confusion to which
6124 @code{if} statement an @code{else} branch belongs. Here is an example of
6139 In C/C++, every @code{else} branch belongs to the innermost possible
6140 @code{if} statement, which in this example is @code{if (b)}. This is
6141 often not what the programmer expected, as illustrated in the above
6142 example by indentation the programmer chose. When there is the
6143 potential for this confusion, GCC issues a warning when this flag
6144 is specified. To eliminate the warning, add explicit braces around
6145 the innermost @code{if} statement so there is no way the @code{else}
6146 can belong to the enclosing @code{if}. The resulting code
6163 This warning is enabled by @option{-Wparentheses}.
6167 @opindex Wno-date-time
6168 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6169 are encountered as they might prevent bit-wise-identical reproducible
6172 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6173 @opindex Wdelete-incomplete
6174 @opindex Wno-delete-incomplete
6175 Warn when deleting a pointer to incomplete type, which may cause
6176 undefined behavior at runtime. This warning is enabled by default.
6178 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6179 @opindex Wuseless-cast
6180 @opindex Wno-useless-cast
6181 Warn when an expression is casted to its own type.
6184 @opindex Wempty-body
6185 @opindex Wno-empty-body
6186 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6187 while} statement. This warning is also enabled by @option{-Wextra}.
6189 @item -Wenum-compare
6190 @opindex Wenum-compare
6191 @opindex Wno-enum-compare
6192 Warn about a comparison between values of different enumerated types.
6193 In C++ enumerated type mismatches in conditional expressions are also
6194 diagnosed and the warning is enabled by default. In C this warning is
6195 enabled by @option{-Wall}.
6197 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6198 @opindex Wextra-semi
6199 @opindex Wno-extra-semi
6200 Warn about redundant semicolon after in-class function definition.
6202 @item -Wjump-misses-init @r{(C, Objective-C only)}
6203 @opindex Wjump-misses-init
6204 @opindex Wno-jump-misses-init
6205 Warn if a @code{goto} statement or a @code{switch} statement jumps
6206 forward across the initialization of a variable, or jumps backward to a
6207 label after the variable has been initialized. This only warns about
6208 variables that are initialized when they are declared. This warning is
6209 only supported for C and Objective-C; in C++ this sort of branch is an
6212 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6213 can be disabled with the @option{-Wno-jump-misses-init} option.
6215 @item -Wsign-compare
6216 @opindex Wsign-compare
6217 @opindex Wno-sign-compare
6218 @cindex warning for comparison of signed and unsigned values
6219 @cindex comparison of signed and unsigned values, warning
6220 @cindex signed and unsigned values, comparison warning
6221 Warn when a comparison between signed and unsigned values could produce
6222 an incorrect result when the signed value is converted to unsigned.
6223 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6224 also enabled by @option{-Wextra}.
6226 @item -Wsign-conversion
6227 @opindex Wsign-conversion
6228 @opindex Wno-sign-conversion
6229 Warn for implicit conversions that may change the sign of an integer
6230 value, like assigning a signed integer expression to an unsigned
6231 integer variable. An explicit cast silences the warning. In C, this
6232 option is enabled also by @option{-Wconversion}.
6234 @item -Wfloat-conversion
6235 @opindex Wfloat-conversion
6236 @opindex Wno-float-conversion
6237 Warn for implicit conversions that reduce the precision of a real value.
6238 This includes conversions from real to integer, and from higher precision
6239 real to lower precision real values. This option is also enabled by
6240 @option{-Wconversion}.
6242 @item -Wno-scalar-storage-order
6243 @opindex -Wno-scalar-storage-order
6244 @opindex -Wscalar-storage-order
6245 Do not warn on suspicious constructs involving reverse scalar storage order.
6247 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6248 @opindex Wsized-deallocation
6249 @opindex Wno-sized-deallocation
6250 Warn about a definition of an unsized deallocation function
6252 void operator delete (void *) noexcept;
6253 void operator delete[] (void *) noexcept;
6255 without a definition of the corresponding sized deallocation function
6257 void operator delete (void *, std::size_t) noexcept;
6258 void operator delete[] (void *, std::size_t) noexcept;
6260 or vice versa. Enabled by @option{-Wextra} along with
6261 @option{-fsized-deallocation}.
6263 @item -Wsizeof-pointer-div
6264 @opindex Wsizeof-pointer-div
6265 @opindex Wno-sizeof-pointer-div
6266 Warn for suspicious divisions of two sizeof expressions that divide
6267 the pointer size by the element size, which is the usual way to compute
6268 the array size but won't work out correctly with pointers. This warning
6269 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
6270 not an array, but a pointer. This warning is enabled by @option{-Wall}.
6272 @item -Wsizeof-pointer-memaccess
6273 @opindex Wsizeof-pointer-memaccess
6274 @opindex Wno-sizeof-pointer-memaccess
6275 Warn for suspicious length parameters to certain string and memory built-in
6276 functions if the argument uses @code{sizeof}. This warning triggers for
6277 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
6278 an array, but a pointer, and suggests a possible fix, or about
6279 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
6280 also warns about calls to bounded string copy functions like @code{strncat}
6281 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
6282 the source array. For example, in the following function the call to
6283 @code{strncat} specifies the size of the source string as the bound. That
6284 is almost certainly a mistake and so the call is diagnosed.
6286 void make_file (const char *name)
6288 char path[PATH_MAX];
6289 strncpy (path, name, sizeof path - 1);
6290 strncat (path, ".text", sizeof ".text");
6295 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
6297 @item -Wsizeof-array-argument
6298 @opindex Wsizeof-array-argument
6299 @opindex Wno-sizeof-array-argument
6300 Warn when the @code{sizeof} operator is applied to a parameter that is
6301 declared as an array in a function definition. This warning is enabled by
6302 default for C and C++ programs.
6304 @item -Wmemset-elt-size
6305 @opindex Wmemset-elt-size
6306 @opindex Wno-memset-elt-size
6307 Warn for suspicious calls to the @code{memset} built-in function, if the
6308 first argument references an array, and the third argument is a number
6309 equal to the number of elements, but not equal to the size of the array
6310 in memory. This indicates that the user has omitted a multiplication by
6311 the element size. This warning is enabled by @option{-Wall}.
6313 @item -Wmemset-transposed-args
6314 @opindex Wmemset-transposed-args
6315 @opindex Wno-memset-transposed-args
6316 Warn for suspicious calls to the @code{memset} built-in function, if the
6317 second argument is not zero and the third argument is zero. This warns e.g.@
6318 about @code{memset (buf, sizeof buf, 0)} where most probably
6319 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6320 is only emitted if the third argument is literal zero. If it is some
6321 expression that is folded to zero, a cast of zero to some type, etc.,
6322 it is far less likely that the user has mistakenly exchanged the arguments
6323 and no warning is emitted. This warning is enabled by @option{-Wall}.
6327 @opindex Wno-address
6328 Warn about suspicious uses of memory addresses. These include using
6329 the address of a function in a conditional expression, such as
6330 @code{void func(void); if (func)}, and comparisons against the memory
6331 address of a string literal, such as @code{if (x == "abc")}. Such
6332 uses typically indicate a programmer error: the address of a function
6333 always evaluates to true, so their use in a conditional usually
6334 indicate that the programmer forgot the parentheses in a function
6335 call; and comparisons against string literals result in unspecified
6336 behavior and are not portable in C, so they usually indicate that the
6337 programmer intended to use @code{strcmp}. This warning is enabled by
6341 @opindex Wlogical-op
6342 @opindex Wno-logical-op
6343 Warn about suspicious uses of logical operators in expressions.
6344 This includes using logical operators in contexts where a
6345 bit-wise operator is likely to be expected. Also warns when
6346 the operands of a logical operator are the same:
6349 if (a < 0 && a < 0) @{ @dots{} @}
6352 @item -Wlogical-not-parentheses
6353 @opindex Wlogical-not-parentheses
6354 @opindex Wno-logical-not-parentheses
6355 Warn about logical not used on the left hand side operand of a comparison.
6356 This option does not warn if the right operand is considered to be a boolean
6357 expression. Its purpose is to detect suspicious code like the following:
6361 if (!a > 1) @{ @dots{} @}
6364 It is possible to suppress the warning by wrapping the LHS into
6367 if ((!a) > 1) @{ @dots{} @}
6370 This warning is enabled by @option{-Wall}.
6372 @item -Waggregate-return
6373 @opindex Waggregate-return
6374 @opindex Wno-aggregate-return
6375 Warn if any functions that return structures or unions are defined or
6376 called. (In languages where you can return an array, this also elicits
6379 @item -Wno-aggressive-loop-optimizations
6380 @opindex Wno-aggressive-loop-optimizations
6381 @opindex Waggressive-loop-optimizations
6382 Warn if in a loop with constant number of iterations the compiler detects
6383 undefined behavior in some statement during one or more of the iterations.
6385 @item -Wno-attributes
6386 @opindex Wno-attributes
6387 @opindex Wattributes
6388 Do not warn if an unexpected @code{__attribute__} is used, such as
6389 unrecognized attributes, function attributes applied to variables,
6390 etc. This does not stop errors for incorrect use of supported
6393 @item -Wno-builtin-declaration-mismatch
6394 @opindex Wno-builtin-declaration-mismatch
6395 @opindex Wbuiltin-declaration-mismatch
6396 Warn if a built-in function is declared with the wrong signature or
6398 This warning is enabled by default.
6400 @item -Wno-builtin-macro-redefined
6401 @opindex Wno-builtin-macro-redefined
6402 @opindex Wbuiltin-macro-redefined
6403 Do not warn if certain built-in macros are redefined. This suppresses
6404 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6405 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6407 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6408 @opindex Wstrict-prototypes
6409 @opindex Wno-strict-prototypes
6410 Warn if a function is declared or defined without specifying the
6411 argument types. (An old-style function definition is permitted without
6412 a warning if preceded by a declaration that specifies the argument
6415 @item -Wold-style-declaration @r{(C and Objective-C only)}
6416 @opindex Wold-style-declaration
6417 @opindex Wno-old-style-declaration
6418 Warn for obsolescent usages, according to the C Standard, in a
6419 declaration. For example, warn if storage-class specifiers like
6420 @code{static} are not the first things in a declaration. This warning
6421 is also enabled by @option{-Wextra}.
6423 @item -Wold-style-definition @r{(C and Objective-C only)}
6424 @opindex Wold-style-definition
6425 @opindex Wno-old-style-definition
6426 Warn if an old-style function definition is used. A warning is given
6427 even if there is a previous prototype.
6429 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6430 @opindex Wmissing-parameter-type
6431 @opindex Wno-missing-parameter-type
6432 A function parameter is declared without a type specifier in K&R-style
6439 This warning is also enabled by @option{-Wextra}.
6441 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6442 @opindex Wmissing-prototypes
6443 @opindex Wno-missing-prototypes
6444 Warn if a global function is defined without a previous prototype
6445 declaration. This warning is issued even if the definition itself
6446 provides a prototype. Use this option to detect global functions
6447 that do not have a matching prototype declaration in a header file.
6448 This option is not valid for C++ because all function declarations
6449 provide prototypes and a non-matching declaration declares an
6450 overload rather than conflict with an earlier declaration.
6451 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6453 @item -Wmissing-declarations
6454 @opindex Wmissing-declarations
6455 @opindex Wno-missing-declarations
6456 Warn if a global function is defined without a previous declaration.
6457 Do so even if the definition itself provides a prototype.
6458 Use this option to detect global functions that are not declared in
6459 header files. In C, no warnings are issued for functions with previous
6460 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6461 missing prototypes. In C++, no warnings are issued for function templates,
6462 or for inline functions, or for functions in anonymous namespaces.
6464 @item -Wmissing-field-initializers
6465 @opindex Wmissing-field-initializers
6466 @opindex Wno-missing-field-initializers
6470 Warn if a structure's initializer has some fields missing. For
6471 example, the following code causes such a warning, because
6472 @code{x.h} is implicitly zero:
6475 struct s @{ int f, g, h; @};
6476 struct s x = @{ 3, 4 @};
6479 This option does not warn about designated initializers, so the following
6480 modification does not trigger a warning:
6483 struct s @{ int f, g, h; @};
6484 struct s x = @{ .f = 3, .g = 4 @};
6487 In C this option does not warn about the universal zero initializer
6491 struct s @{ int f, g, h; @};
6492 struct s x = @{ 0 @};
6495 Likewise, in C++ this option does not warn about the empty @{ @}
6496 initializer, for example:
6499 struct s @{ int f, g, h; @};
6503 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6504 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6506 @item -Wno-multichar
6507 @opindex Wno-multichar
6509 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6510 Usually they indicate a typo in the user's code, as they have
6511 implementation-defined values, and should not be used in portable code.
6513 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
6514 @opindex Wnormalized=
6515 @opindex Wnormalized
6516 @opindex Wno-normalized
6519 @cindex character set, input normalization
6520 In ISO C and ISO C++, two identifiers are different if they are
6521 different sequences of characters. However, sometimes when characters
6522 outside the basic ASCII character set are used, you can have two
6523 different character sequences that look the same. To avoid confusion,
6524 the ISO 10646 standard sets out some @dfn{normalization rules} which
6525 when applied ensure that two sequences that look the same are turned into
6526 the same sequence. GCC can warn you if you are using identifiers that
6527 have not been normalized; this option controls that warning.
6529 There are four levels of warning supported by GCC@. The default is
6530 @option{-Wnormalized=nfc}, which warns about any identifier that is
6531 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6532 recommended form for most uses. It is equivalent to
6533 @option{-Wnormalized}.
6535 Unfortunately, there are some characters allowed in identifiers by
6536 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6537 identifiers. That is, there's no way to use these symbols in portable
6538 ISO C or C++ and have all your identifiers in NFC@.
6539 @option{-Wnormalized=id} suppresses the warning for these characters.
6540 It is hoped that future versions of the standards involved will correct
6541 this, which is why this option is not the default.
6543 You can switch the warning off for all characters by writing
6544 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6545 only do this if you are using some other normalization scheme (like
6546 ``D''), because otherwise you can easily create bugs that are
6547 literally impossible to see.
6549 Some characters in ISO 10646 have distinct meanings but look identical
6550 in some fonts or display methodologies, especially once formatting has
6551 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6552 LETTER N'', displays just like a regular @code{n} that has been
6553 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6554 normalization scheme to convert all these into a standard form as
6555 well, and GCC warns if your code is not in NFKC if you use
6556 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6557 about every identifier that contains the letter O because it might be
6558 confused with the digit 0, and so is not the default, but may be
6559 useful as a local coding convention if the programming environment
6560 cannot be fixed to display these characters distinctly.
6562 @item -Wno-deprecated
6563 @opindex Wno-deprecated
6564 @opindex Wdeprecated
6565 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6567 @item -Wno-deprecated-declarations
6568 @opindex Wno-deprecated-declarations
6569 @opindex Wdeprecated-declarations
6570 Do not warn about uses of functions (@pxref{Function Attributes}),
6571 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6572 Attributes}) marked as deprecated by using the @code{deprecated}
6576 @opindex Wno-overflow
6578 Do not warn about compile-time overflow in constant expressions.
6583 Warn about One Definition Rule violations during link-time optimization.
6584 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6587 @opindex Wopenm-simd
6588 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6589 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6590 option can be used to relax the cost model.
6592 @item -Woverride-init @r{(C and Objective-C only)}
6593 @opindex Woverride-init
6594 @opindex Wno-override-init
6598 Warn if an initialized field without side effects is overridden when
6599 using designated initializers (@pxref{Designated Inits, , Designated
6602 This warning is included in @option{-Wextra}. To get other
6603 @option{-Wextra} warnings without this one, use @option{-Wextra
6604 -Wno-override-init}.
6606 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6607 @opindex Woverride-init-side-effects
6608 @opindex Wno-override-init-side-effects
6609 Warn if an initialized field with side effects is overridden when
6610 using designated initializers (@pxref{Designated Inits, , Designated
6611 Initializers}). This warning is enabled by default.
6616 Warn if a structure is given the packed attribute, but the packed
6617 attribute has no effect on the layout or size of the structure.
6618 Such structures may be mis-aligned for little benefit. For
6619 instance, in this code, the variable @code{f.x} in @code{struct bar}
6620 is misaligned even though @code{struct bar} does not itself
6621 have the packed attribute:
6628 @} __attribute__((packed));
6636 @item -Wpacked-bitfield-compat
6637 @opindex Wpacked-bitfield-compat
6638 @opindex Wno-packed-bitfield-compat
6639 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6640 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6641 the change can lead to differences in the structure layout. GCC
6642 informs you when the offset of such a field has changed in GCC 4.4.
6643 For example there is no longer a 4-bit padding between field @code{a}
6644 and @code{b} in this structure:
6651 @} __attribute__ ((packed));
6654 This warning is enabled by default. Use
6655 @option{-Wno-packed-bitfield-compat} to disable this warning.
6657 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
6658 @opindex Wpacked-not-aligned
6659 @opindex Wno-packed-not-aligned
6660 Warn if a structure field with explicitly specified alignment in a
6661 packed struct or union is misaligned. For example, a warning will
6662 be issued on @code{struct S}, like, @code{warning: alignment 1 of
6663 'struct S' is less than 8}, in this code:
6667 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
6668 struct __attribute__ ((packed)) S @{
6674 This warning is enabled by @option{-Wall}.
6679 Warn if padding is included in a structure, either to align an element
6680 of the structure or to align the whole structure. Sometimes when this
6681 happens it is possible to rearrange the fields of the structure to
6682 reduce the padding and so make the structure smaller.
6684 @item -Wredundant-decls
6685 @opindex Wredundant-decls
6686 @opindex Wno-redundant-decls
6687 Warn if anything is declared more than once in the same scope, even in
6688 cases where multiple declaration is valid and changes nothing.
6692 @opindex Wno-restrict
6693 Warn when an argument passed to a restrict-qualified parameter
6694 aliases with another argument.
6696 @item -Wnested-externs @r{(C and Objective-C only)}
6697 @opindex Wnested-externs
6698 @opindex Wno-nested-externs
6699 Warn if an @code{extern} declaration is encountered within a function.
6701 @item -Wno-inherited-variadic-ctor
6702 @opindex Winherited-variadic-ctor
6703 @opindex Wno-inherited-variadic-ctor
6704 Suppress warnings about use of C++11 inheriting constructors when the
6705 base class inherited from has a C variadic constructor; the warning is
6706 on by default because the ellipsis is not inherited.
6711 Warn if a function that is declared as inline cannot be inlined.
6712 Even with this option, the compiler does not warn about failures to
6713 inline functions declared in system headers.
6715 The compiler uses a variety of heuristics to determine whether or not
6716 to inline a function. For example, the compiler takes into account
6717 the size of the function being inlined and the amount of inlining
6718 that has already been done in the current function. Therefore,
6719 seemingly insignificant changes in the source program can cause the
6720 warnings produced by @option{-Winline} to appear or disappear.
6722 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6723 @opindex Wno-invalid-offsetof
6724 @opindex Winvalid-offsetof
6725 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6726 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6727 to a non-standard-layout type is undefined. In existing C++ implementations,
6728 however, @code{offsetof} typically gives meaningful results.
6729 This flag is for users who are aware that they are
6730 writing nonportable code and who have deliberately chosen to ignore the
6733 The restrictions on @code{offsetof} may be relaxed in a future version
6734 of the C++ standard.
6736 @item -Wint-in-bool-context
6737 @opindex Wint-in-bool-context
6738 @opindex Wno-int-in-bool-context
6739 Warn for suspicious use of integer values where boolean values are expected,
6740 such as conditional expressions (?:) using non-boolean integer constants in
6741 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6742 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6743 for all kinds of multiplications regardless of the data type.
6744 This warning is enabled by @option{-Wall}.
6746 @item -Wno-int-to-pointer-cast
6747 @opindex Wno-int-to-pointer-cast
6748 @opindex Wint-to-pointer-cast
6749 Suppress warnings from casts to pointer type of an integer of a
6750 different size. In C++, casting to a pointer type of smaller size is
6751 an error. @option{Wint-to-pointer-cast} is enabled by default.
6754 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6755 @opindex Wno-pointer-to-int-cast
6756 @opindex Wpointer-to-int-cast
6757 Suppress warnings from casts from a pointer to an integer type of a
6761 @opindex Winvalid-pch
6762 @opindex Wno-invalid-pch
6763 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6764 the search path but cannot be used.
6768 @opindex Wno-long-long
6769 Warn if @code{long long} type is used. This is enabled by either
6770 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6771 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6773 @item -Wvariadic-macros
6774 @opindex Wvariadic-macros
6775 @opindex Wno-variadic-macros
6776 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6777 alternate syntax is used in ISO C99 mode. This is enabled by either
6778 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6779 messages, use @option{-Wno-variadic-macros}.
6783 @opindex Wno-varargs
6784 Warn upon questionable usage of the macros used to handle variable
6785 arguments like @code{va_start}. This is default. To inhibit the
6786 warning messages, use @option{-Wno-varargs}.
6788 @item -Wvector-operation-performance
6789 @opindex Wvector-operation-performance
6790 @opindex Wno-vector-operation-performance
6791 Warn if vector operation is not implemented via SIMD capabilities of the
6792 architecture. Mainly useful for the performance tuning.
6793 Vector operation can be implemented @code{piecewise}, which means that the
6794 scalar operation is performed on every vector element;
6795 @code{in parallel}, which means that the vector operation is implemented
6796 using scalars of wider type, which normally is more performance efficient;
6797 and @code{as a single scalar}, which means that vector fits into a
6800 @item -Wno-virtual-move-assign
6801 @opindex Wvirtual-move-assign
6802 @opindex Wno-virtual-move-assign
6803 Suppress warnings about inheriting from a virtual base with a
6804 non-trivial C++11 move assignment operator. This is dangerous because
6805 if the virtual base is reachable along more than one path, it is
6806 moved multiple times, which can mean both objects end up in the
6807 moved-from state. If the move assignment operator is written to avoid
6808 moving from a moved-from object, this warning can be disabled.
6813 Warn if a variable-length array is used in the code.
6814 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6815 the variable-length array.
6817 @item -Wvla-larger-than=@var{n}
6818 If this option is used, the compiler will warn on uses of
6819 variable-length arrays where the size is either unbounded, or bounded
6820 by an argument that can be larger than @var{n} bytes. This is similar
6821 to how @option{-Walloca-larger-than=@var{n}} works, but with
6822 variable-length arrays.
6824 Note that GCC may optimize small variable-length arrays of a known
6825 value into plain arrays, so this warning may not get triggered for
6828 This warning is not enabled by @option{-Wall}, and is only active when
6829 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6831 See also @option{-Walloca-larger-than=@var{n}}.
6833 @item -Wvolatile-register-var
6834 @opindex Wvolatile-register-var
6835 @opindex Wno-volatile-register-var
6836 Warn if a register variable is declared volatile. The volatile
6837 modifier does not inhibit all optimizations that may eliminate reads
6838 and/or writes to register variables. This warning is enabled by
6841 @item -Wdisabled-optimization
6842 @opindex Wdisabled-optimization
6843 @opindex Wno-disabled-optimization
6844 Warn if a requested optimization pass is disabled. This warning does
6845 not generally indicate that there is anything wrong with your code; it
6846 merely indicates that GCC's optimizers are unable to handle the code
6847 effectively. Often, the problem is that your code is too big or too
6848 complex; GCC refuses to optimize programs when the optimization
6849 itself is likely to take inordinate amounts of time.
6851 @item -Wpointer-sign @r{(C and Objective-C only)}
6852 @opindex Wpointer-sign
6853 @opindex Wno-pointer-sign
6854 Warn for pointer argument passing or assignment with different signedness.
6855 This option is only supported for C and Objective-C@. It is implied by
6856 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6857 @option{-Wno-pointer-sign}.
6859 @item -Wstack-protector
6860 @opindex Wstack-protector
6861 @opindex Wno-stack-protector
6862 This option is only active when @option{-fstack-protector} is active. It
6863 warns about functions that are not protected against stack smashing.
6865 @item -Woverlength-strings
6866 @opindex Woverlength-strings
6867 @opindex Wno-overlength-strings
6868 Warn about string constants that are longer than the ``minimum
6869 maximum'' length specified in the C standard. Modern compilers
6870 generally allow string constants that are much longer than the
6871 standard's minimum limit, but very portable programs should avoid
6872 using longer strings.
6874 The limit applies @emph{after} string constant concatenation, and does
6875 not count the trailing NUL@. In C90, the limit was 509 characters; in
6876 C99, it was raised to 4095. C++98 does not specify a normative
6877 minimum maximum, so we do not diagnose overlength strings in C++@.
6879 This option is implied by @option{-Wpedantic}, and can be disabled with
6880 @option{-Wno-overlength-strings}.
6882 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6883 @opindex Wunsuffixed-float-constants
6885 Issue a warning for any floating constant that does not have
6886 a suffix. When used together with @option{-Wsystem-headers} it
6887 warns about such constants in system header files. This can be useful
6888 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6889 from the decimal floating-point extension to C99.
6891 @item -Wno-designated-init @r{(C and Objective-C only)}
6892 Suppress warnings when a positional initializer is used to initialize
6893 a structure that has been marked with the @code{designated_init}
6897 Issue a warning when HSAIL cannot be emitted for the compiled function or
6902 @node Debugging Options
6903 @section Options for Debugging Your Program
6904 @cindex options, debugging
6905 @cindex debugging information options
6907 To tell GCC to emit extra information for use by a debugger, in almost
6908 all cases you need only to add @option{-g} to your other options.
6910 GCC allows you to use @option{-g} with
6911 @option{-O}. The shortcuts taken by optimized code may occasionally
6912 be surprising: some variables you declared may not exist
6913 at all; flow of control may briefly move where you did not expect it;
6914 some statements may not be executed because they compute constant
6915 results or their values are already at hand; some statements may
6916 execute in different places because they have been moved out of loops.
6917 Nevertheless it is possible to debug optimized output. This makes
6918 it reasonable to use the optimizer for programs that might have bugs.
6920 If you are not using some other optimization option, consider
6921 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6922 With no @option{-O} option at all, some compiler passes that collect
6923 information useful for debugging do not run at all, so that
6924 @option{-Og} may result in a better debugging experience.
6929 Produce debugging information in the operating system's native format
6930 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6933 On most systems that use stabs format, @option{-g} enables use of extra
6934 debugging information that only GDB can use; this extra information
6935 makes debugging work better in GDB but probably makes other debuggers
6937 refuse to read the program. If you want to control for certain whether
6938 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6939 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6943 Produce debugging information for use by GDB@. This means to use the
6944 most expressive format available (DWARF, stabs, or the native format
6945 if neither of those are supported), including GDB extensions if at all
6949 @itemx -gdwarf-@var{version}
6951 Produce debugging information in DWARF format (if that is supported).
6952 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6953 for most targets is 4. DWARF Version 5 is only experimental.
6955 Note that with DWARF Version 2, some ports require and always
6956 use some non-conflicting DWARF 3 extensions in the unwind tables.
6958 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6959 for maximum benefit.
6961 GCC no longer supports DWARF Version 1, which is substantially
6962 different than Version 2 and later. For historical reasons, some
6963 other DWARF-related options such as
6964 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6965 in their names, but apply to all currently-supported versions of DWARF.
6969 Produce debugging information in stabs format (if that is supported),
6970 without GDB extensions. This is the format used by DBX on most BSD
6971 systems. On MIPS, Alpha and System V Release 4 systems this option
6972 produces stabs debugging output that is not understood by DBX@.
6973 On System V Release 4 systems this option requires the GNU assembler.
6977 Produce debugging information in stabs format (if that is supported),
6978 using GNU extensions understood only by the GNU debugger (GDB)@. The
6979 use of these extensions is likely to make other debuggers crash or
6980 refuse to read the program.
6984 Produce debugging information in XCOFF format (if that is supported).
6985 This is the format used by the DBX debugger on IBM RS/6000 systems.
6989 Produce debugging information in XCOFF format (if that is supported),
6990 using GNU extensions understood only by the GNU debugger (GDB)@. The
6991 use of these extensions is likely to make other debuggers crash or
6992 refuse to read the program, and may cause assemblers other than the GNU
6993 assembler (GAS) to fail with an error.
6997 Produce debugging information in Alpha/VMS debug format (if that is
6998 supported). This is the format used by DEBUG on Alpha/VMS systems.
7001 @itemx -ggdb@var{level}
7002 @itemx -gstabs@var{level}
7003 @itemx -gxcoff@var{level}
7004 @itemx -gvms@var{level}
7005 Request debugging information and also use @var{level} to specify how
7006 much information. The default level is 2.
7008 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7011 Level 1 produces minimal information, enough for making backtraces in
7012 parts of the program that you don't plan to debug. This includes
7013 descriptions of functions and external variables, and line number
7014 tables, but no information about local variables.
7016 Level 3 includes extra information, such as all the macro definitions
7017 present in the program. Some debuggers support macro expansion when
7018 you use @option{-g3}.
7020 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7021 confusion with @option{-gdwarf-@var{level}}.
7022 Instead use an additional @option{-g@var{level}} option to change the
7023 debug level for DWARF.
7025 @item -feliminate-unused-debug-symbols
7026 @opindex feliminate-unused-debug-symbols
7027 Produce debugging information in stabs format (if that is supported),
7028 for only symbols that are actually used.
7030 @item -femit-class-debug-always
7031 @opindex femit-class-debug-always
7032 Instead of emitting debugging information for a C++ class in only one
7033 object file, emit it in all object files using the class. This option
7034 should be used only with debuggers that are unable to handle the way GCC
7035 normally emits debugging information for classes because using this
7036 option increases the size of debugging information by as much as a
7039 @item -fno-merge-debug-strings
7040 @opindex fmerge-debug-strings
7041 @opindex fno-merge-debug-strings
7042 Direct the linker to not merge together strings in the debugging
7043 information that are identical in different object files. Merging is
7044 not supported by all assemblers or linkers. Merging decreases the size
7045 of the debug information in the output file at the cost of increasing
7046 link processing time. Merging is enabled by default.
7048 @item -fdebug-prefix-map=@var{old}=@var{new}
7049 @opindex fdebug-prefix-map
7050 When compiling files in directory @file{@var{old}}, record debugging
7051 information describing them as in @file{@var{new}} instead. This can be
7052 used to replace a build-time path with an install-time path in the debug info.
7053 It can also be used to change an absolute path to a relative path by using
7054 @file{.} for @var{new}. This can give more reproducible builds, which are
7055 location independent, but may require an extra command to tell GDB where to
7056 find the source files.
7058 @item -fvar-tracking
7059 @opindex fvar-tracking
7060 Run variable tracking pass. It computes where variables are stored at each
7061 position in code. Better debugging information is then generated
7062 (if the debugging information format supports this information).
7064 It is enabled by default when compiling with optimization (@option{-Os},
7065 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7066 the debug info format supports it.
7068 @item -fvar-tracking-assignments
7069 @opindex fvar-tracking-assignments
7070 @opindex fno-var-tracking-assignments
7071 Annotate assignments to user variables early in the compilation and
7072 attempt to carry the annotations over throughout the compilation all the
7073 way to the end, in an attempt to improve debug information while
7074 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7076 It can be enabled even if var-tracking is disabled, in which case
7077 annotations are created and maintained, but discarded at the end.
7078 By default, this flag is enabled together with @option{-fvar-tracking},
7079 except when selective scheduling is enabled.
7082 @opindex gsplit-dwarf
7083 Separate as much DWARF debugging information as possible into a
7084 separate output file with the extension @file{.dwo}. This option allows
7085 the build system to avoid linking files with debug information. To
7086 be useful, this option requires a debugger capable of reading @file{.dwo}
7091 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7093 @item -ggnu-pubnames
7094 @opindex ggnu-pubnames
7095 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7096 suitable for conversion into a GDB@ index. This option is only useful
7097 with a linker that can produce GDB@ index version 7.
7099 @item -fdebug-types-section
7100 @opindex fdebug-types-section
7101 @opindex fno-debug-types-section
7102 When using DWARF Version 4 or higher, type DIEs can be put into
7103 their own @code{.debug_types} section instead of making them part of the
7104 @code{.debug_info} section. It is more efficient to put them in a separate
7105 comdat sections since the linker can then remove duplicates.
7106 But not all DWARF consumers support @code{.debug_types} sections yet
7107 and on some objects @code{.debug_types} produces larger instead of smaller
7108 debugging information.
7110 @item -grecord-gcc-switches
7111 @item -gno-record-gcc-switches
7112 @opindex grecord-gcc-switches
7113 @opindex gno-record-gcc-switches
7114 This switch causes the command-line options used to invoke the
7115 compiler that may affect code generation to be appended to the
7116 DW_AT_producer attribute in DWARF debugging information. The options
7117 are concatenated with spaces separating them from each other and from
7118 the compiler version.
7119 It is enabled by default.
7120 See also @option{-frecord-gcc-switches} for another
7121 way of storing compiler options into the object file.
7123 @item -gstrict-dwarf
7124 @opindex gstrict-dwarf
7125 Disallow using extensions of later DWARF standard version than selected
7126 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7127 DWARF extensions from later standard versions is allowed.
7129 @item -gno-strict-dwarf
7130 @opindex gno-strict-dwarf
7131 Allow using extensions of later DWARF standard version than selected with
7132 @option{-gdwarf-@var{version}}.
7135 @item -gno-column-info
7136 @opindex gcolumn-info
7137 @opindex gno-column-info
7138 Emit location column information into DWARF debugging information, rather
7139 than just file and line.
7140 This option is enabled by default.
7142 @item -gz@r{[}=@var{type}@r{]}
7144 Produce compressed debug sections in DWARF format, if that is supported.
7145 If @var{type} is not given, the default type depends on the capabilities
7146 of the assembler and linker used. @var{type} may be one of
7147 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
7148 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
7149 compression in traditional GNU format). If the linker doesn't support
7150 writing compressed debug sections, the option is rejected. Otherwise,
7151 if the assembler does not support them, @option{-gz} is silently ignored
7152 when producing object files.
7154 @item -femit-struct-debug-baseonly
7155 @opindex femit-struct-debug-baseonly
7156 Emit debug information for struct-like types
7157 only when the base name of the compilation source file
7158 matches the base name of file in which the struct is defined.
7160 This option substantially reduces the size of debugging information,
7161 but at significant potential loss in type information to the debugger.
7162 See @option{-femit-struct-debug-reduced} for a less aggressive option.
7163 See @option{-femit-struct-debug-detailed} for more detailed control.
7165 This option works only with DWARF debug output.
7167 @item -femit-struct-debug-reduced
7168 @opindex femit-struct-debug-reduced
7169 Emit debug information for struct-like types
7170 only when the base name of the compilation source file
7171 matches the base name of file in which the type is defined,
7172 unless the struct is a template or defined in a system header.
7174 This option significantly reduces the size of debugging information,
7175 with some potential loss in type information to the debugger.
7176 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
7177 See @option{-femit-struct-debug-detailed} for more detailed control.
7179 This option works only with DWARF debug output.
7181 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
7182 @opindex femit-struct-debug-detailed
7183 Specify the struct-like types
7184 for which the compiler generates debug information.
7185 The intent is to reduce duplicate struct debug information
7186 between different object files within the same program.
7188 This option is a detailed version of
7189 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
7190 which serves for most needs.
7192 A specification has the syntax@*
7193 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
7195 The optional first word limits the specification to
7196 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
7197 A struct type is used directly when it is the type of a variable, member.
7198 Indirect uses arise through pointers to structs.
7199 That is, when use of an incomplete struct is valid, the use is indirect.
7201 @samp{struct one direct; struct two * indirect;}.
7203 The optional second word limits the specification to
7204 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
7205 Generic structs are a bit complicated to explain.
7206 For C++, these are non-explicit specializations of template classes,
7207 or non-template classes within the above.
7208 Other programming languages have generics,
7209 but @option{-femit-struct-debug-detailed} does not yet implement them.
7211 The third word specifies the source files for those
7212 structs for which the compiler should emit debug information.
7213 The values @samp{none} and @samp{any} have the normal meaning.
7214 The value @samp{base} means that
7215 the base of name of the file in which the type declaration appears
7216 must match the base of the name of the main compilation file.
7217 In practice, this means that when compiling @file{foo.c}, debug information
7218 is generated for types declared in that file and @file{foo.h},
7219 but not other header files.
7220 The value @samp{sys} means those types satisfying @samp{base}
7221 or declared in system or compiler headers.
7223 You may need to experiment to determine the best settings for your application.
7225 The default is @option{-femit-struct-debug-detailed=all}.
7227 This option works only with DWARF debug output.
7229 @item -fno-dwarf2-cfi-asm
7230 @opindex fdwarf2-cfi-asm
7231 @opindex fno-dwarf2-cfi-asm
7232 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
7233 instead of using GAS @code{.cfi_*} directives.
7235 @item -fno-eliminate-unused-debug-types
7236 @opindex feliminate-unused-debug-types
7237 @opindex fno-eliminate-unused-debug-types
7238 Normally, when producing DWARF output, GCC avoids producing debug symbol
7239 output for types that are nowhere used in the source file being compiled.
7240 Sometimes it is useful to have GCC emit debugging
7241 information for all types declared in a compilation
7242 unit, regardless of whether or not they are actually used
7243 in that compilation unit, for example
7244 if, in the debugger, you want to cast a value to a type that is
7245 not actually used in your program (but is declared). More often,
7246 however, this results in a significant amount of wasted space.
7249 @node Optimize Options
7250 @section Options That Control Optimization
7251 @cindex optimize options
7252 @cindex options, optimization
7254 These options control various sorts of optimizations.
7256 Without any optimization option, the compiler's goal is to reduce the
7257 cost of compilation and to make debugging produce the expected
7258 results. Statements are independent: if you stop the program with a
7259 breakpoint between statements, you can then assign a new value to any
7260 variable or change the program counter to any other statement in the
7261 function and get exactly the results you expect from the source
7264 Turning on optimization flags makes the compiler attempt to improve
7265 the performance and/or code size at the expense of compilation time
7266 and possibly the ability to debug the program.
7268 The compiler performs optimization based on the knowledge it has of the
7269 program. Compiling multiple files at once to a single output file mode allows
7270 the compiler to use information gained from all of the files when compiling
7273 Not all optimizations are controlled directly by a flag. Only
7274 optimizations that have a flag are listed in this section.
7276 Most optimizations are only enabled if an @option{-O} level is set on
7277 the command line. Otherwise they are disabled, even if individual
7278 optimization flags are specified.
7280 Depending on the target and how GCC was configured, a slightly different
7281 set of optimizations may be enabled at each @option{-O} level than
7282 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7283 to find out the exact set of optimizations that are enabled at each level.
7284 @xref{Overall Options}, for examples.
7291 Optimize. Optimizing compilation takes somewhat more time, and a lot
7292 more memory for a large function.
7294 With @option{-O}, the compiler tries to reduce code size and execution
7295 time, without performing any optimizations that take a great deal of
7298 @option{-O} turns on the following optimization flags:
7301 -fbranch-count-reg @gol
7302 -fcombine-stack-adjustments @gol
7304 -fcprop-registers @gol
7307 -fdelayed-branch @gol
7309 -fforward-propagate @gol
7310 -fguess-branch-probability @gol
7311 -fif-conversion2 @gol
7312 -fif-conversion @gol
7313 -finline-functions-called-once @gol
7314 -fipa-pure-const @gol
7316 -fipa-reference @gol
7317 -fmerge-constants @gol
7318 -fmove-loop-invariants @gol
7319 -fomit-frame-pointer @gol
7320 -freorder-blocks @gol
7322 -fshrink-wrap-separate @gol
7323 -fsplit-wide-types @gol
7329 -ftree-coalesce-vars @gol
7330 -ftree-copy-prop @gol
7332 -ftree-dominator-opts @gol
7334 -ftree-forwprop @gol
7346 Optimize even more. GCC performs nearly all supported optimizations
7347 that do not involve a space-speed tradeoff.
7348 As compared to @option{-O}, this option increases both compilation time
7349 and the performance of the generated code.
7351 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7352 also turns on the following optimization flags:
7353 @gccoptlist{-fthread-jumps @gol
7354 -falign-functions -falign-jumps @gol
7355 -falign-loops -falign-labels @gol
7358 -fcse-follow-jumps -fcse-skip-blocks @gol
7359 -fdelete-null-pointer-checks @gol
7360 -fdevirtualize -fdevirtualize-speculatively @gol
7361 -fexpensive-optimizations @gol
7362 -fgcse -fgcse-lm @gol
7363 -fhoist-adjacent-loads @gol
7364 -finline-small-functions @gol
7365 -findirect-inlining @gol
7371 -fisolate-erroneous-paths-dereference @gol
7373 -foptimize-sibling-calls @gol
7374 -foptimize-strlen @gol
7375 -fpartial-inlining @gol
7377 -freorder-blocks-algorithm=stc @gol
7378 -freorder-blocks-and-partition -freorder-functions @gol
7379 -frerun-cse-after-loop @gol
7380 -fsched-interblock -fsched-spec @gol
7381 -fschedule-insns -fschedule-insns2 @gol
7382 -fstore-merging @gol
7383 -fstrict-aliasing @gol
7384 -ftree-builtin-call-dce @gol
7385 -ftree-switch-conversion -ftree-tail-merge @gol
7386 -fcode-hoisting @gol
7391 Please note the warning under @option{-fgcse} about
7392 invoking @option{-O2} on programs that use computed gotos.
7396 Optimize yet more. @option{-O3} turns on all optimizations specified
7397 by @option{-O2} and also turns on the following optimization flags:
7398 @gccoptlist{-finline-functions @gol
7399 -funswitch-loops @gol
7400 -fpredictive-commoning @gol
7401 -fgcse-after-reload @gol
7402 -ftree-loop-vectorize @gol
7403 -ftree-loop-distribution @gol
7404 -ftree-loop-distribute-patterns @gol
7406 -ftree-slp-vectorize @gol
7407 -fvect-cost-model @gol
7408 -ftree-partial-pre @gol
7414 Reduce compilation time and make debugging produce the expected
7415 results. This is the default.
7419 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7420 do not typically increase code size. It also performs further
7421 optimizations designed to reduce code size.
7423 @option{-Os} disables the following optimization flags:
7424 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7425 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7426 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7430 Disregard strict standards compliance. @option{-Ofast} enables all
7431 @option{-O3} optimizations. It also enables optimizations that are not
7432 valid for all standard-compliant programs.
7433 It turns on @option{-ffast-math} and the Fortran-specific
7434 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
7435 specified, and @option{-fno-protect-parens}.
7439 Optimize debugging experience. @option{-Og} enables optimizations
7440 that do not interfere with debugging. It should be the optimization
7441 level of choice for the standard edit-compile-debug cycle, offering
7442 a reasonable level of optimization while maintaining fast compilation
7443 and a good debugging experience.
7446 If you use multiple @option{-O} options, with or without level numbers,
7447 the last such option is the one that is effective.
7449 Options of the form @option{-f@var{flag}} specify machine-independent
7450 flags. Most flags have both positive and negative forms; the negative
7451 form of @option{-ffoo} is @option{-fno-foo}. In the table
7452 below, only one of the forms is listed---the one you typically
7453 use. You can figure out the other form by either removing @samp{no-}
7456 The following options control specific optimizations. They are either
7457 activated by @option{-O} options or are related to ones that are. You
7458 can use the following flags in the rare cases when ``fine-tuning'' of
7459 optimizations to be performed is desired.
7462 @item -fno-defer-pop
7463 @opindex fno-defer-pop
7464 Always pop the arguments to each function call as soon as that function
7465 returns. For machines that must pop arguments after a function call,
7466 the compiler normally lets arguments accumulate on the stack for several
7467 function calls and pops them all at once.
7469 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7471 @item -fforward-propagate
7472 @opindex fforward-propagate
7473 Perform a forward propagation pass on RTL@. The pass tries to combine two
7474 instructions and checks if the result can be simplified. If loop unrolling
7475 is active, two passes are performed and the second is scheduled after
7478 This option is enabled by default at optimization levels @option{-O},
7479 @option{-O2}, @option{-O3}, @option{-Os}.
7481 @item -ffp-contract=@var{style}
7482 @opindex ffp-contract
7483 @option{-ffp-contract=off} disables floating-point expression contraction.
7484 @option{-ffp-contract=fast} enables floating-point expression contraction
7485 such as forming of fused multiply-add operations if the target has
7486 native support for them.
7487 @option{-ffp-contract=on} enables floating-point expression contraction
7488 if allowed by the language standard. This is currently not implemented
7489 and treated equal to @option{-ffp-contract=off}.
7491 The default is @option{-ffp-contract=fast}.
7493 @item -fomit-frame-pointer
7494 @opindex fomit-frame-pointer
7495 Omit the frame pointer in functions that don't need one. This avoids the
7496 instructions to save, set up and restore the frame pointer; on many targets
7497 it also makes an extra register available.
7499 On some targets this flag has no effect because the standard calling sequence
7500 always uses a frame pointer, so it cannot be omitted.
7502 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
7503 is used in all functions. Several targets always omit the frame pointer in
7506 Enabled by default at @option{-O} and higher.
7508 @item -foptimize-sibling-calls
7509 @opindex foptimize-sibling-calls
7510 Optimize sibling and tail recursive calls.
7512 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7514 @item -foptimize-strlen
7515 @opindex foptimize-strlen
7516 Optimize various standard C string functions (e.g. @code{strlen},
7517 @code{strchr} or @code{strcpy}) and
7518 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7520 Enabled at levels @option{-O2}, @option{-O3}.
7524 Do not expand any functions inline apart from those marked with
7525 the @code{always_inline} attribute. This is the default when not
7528 Single functions can be exempted from inlining by marking them
7529 with the @code{noinline} attribute.
7531 @item -finline-small-functions
7532 @opindex finline-small-functions
7533 Integrate functions into their callers when their body is smaller than expected
7534 function call code (so overall size of program gets smaller). The compiler
7535 heuristically decides which functions are simple enough to be worth integrating
7536 in this way. This inlining applies to all functions, even those not declared
7539 Enabled at level @option{-O2}.
7541 @item -findirect-inlining
7542 @opindex findirect-inlining
7543 Inline also indirect calls that are discovered to be known at compile
7544 time thanks to previous inlining. This option has any effect only
7545 when inlining itself is turned on by the @option{-finline-functions}
7546 or @option{-finline-small-functions} options.
7548 Enabled at level @option{-O2}.
7550 @item -finline-functions
7551 @opindex finline-functions
7552 Consider all functions for inlining, even if they are not declared inline.
7553 The compiler heuristically decides which functions are worth integrating
7556 If all calls to a given function are integrated, and the function is
7557 declared @code{static}, then the function is normally not output as
7558 assembler code in its own right.
7560 Enabled at level @option{-O3}.
7562 @item -finline-functions-called-once
7563 @opindex finline-functions-called-once
7564 Consider all @code{static} functions called once for inlining into their
7565 caller even if they are not marked @code{inline}. If a call to a given
7566 function is integrated, then the function is not output as assembler code
7569 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7571 @item -fearly-inlining
7572 @opindex fearly-inlining
7573 Inline functions marked by @code{always_inline} and functions whose body seems
7574 smaller than the function call overhead early before doing
7575 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7576 makes profiling significantly cheaper and usually inlining faster on programs
7577 having large chains of nested wrapper functions.
7583 Perform interprocedural scalar replacement of aggregates, removal of
7584 unused parameters and replacement of parameters passed by reference
7585 by parameters passed by value.
7587 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7589 @item -finline-limit=@var{n}
7590 @opindex finline-limit
7591 By default, GCC limits the size of functions that can be inlined. This flag
7592 allows coarse control of this limit. @var{n} is the size of functions that
7593 can be inlined in number of pseudo instructions.
7595 Inlining is actually controlled by a number of parameters, which may be
7596 specified individually by using @option{--param @var{name}=@var{value}}.
7597 The @option{-finline-limit=@var{n}} option sets some of these parameters
7601 @item max-inline-insns-single
7602 is set to @var{n}/2.
7603 @item max-inline-insns-auto
7604 is set to @var{n}/2.
7607 See below for a documentation of the individual
7608 parameters controlling inlining and for the defaults of these parameters.
7610 @emph{Note:} there may be no value to @option{-finline-limit} that results
7611 in default behavior.
7613 @emph{Note:} pseudo instruction represents, in this particular context, an
7614 abstract measurement of function's size. In no way does it represent a count
7615 of assembly instructions and as such its exact meaning might change from one
7616 release to an another.
7618 @item -fno-keep-inline-dllexport
7619 @opindex fno-keep-inline-dllexport
7620 This is a more fine-grained version of @option{-fkeep-inline-functions},
7621 which applies only to functions that are declared using the @code{dllexport}
7622 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
7625 @item -fkeep-inline-functions
7626 @opindex fkeep-inline-functions
7627 In C, emit @code{static} functions that are declared @code{inline}
7628 into the object file, even if the function has been inlined into all
7629 of its callers. This switch does not affect functions using the
7630 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7631 inline functions into the object file.
7633 @item -fkeep-static-functions
7634 @opindex fkeep-static-functions
7635 Emit @code{static} functions into the object file, even if the function
7638 @item -fkeep-static-consts
7639 @opindex fkeep-static-consts
7640 Emit variables declared @code{static const} when optimization isn't turned
7641 on, even if the variables aren't referenced.
7643 GCC enables this option by default. If you want to force the compiler to
7644 check if a variable is referenced, regardless of whether or not
7645 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7647 @item -fmerge-constants
7648 @opindex fmerge-constants
7649 Attempt to merge identical constants (string constants and floating-point
7650 constants) across compilation units.
7652 This option is the default for optimized compilation if the assembler and
7653 linker support it. Use @option{-fno-merge-constants} to inhibit this
7656 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7658 @item -fmerge-all-constants
7659 @opindex fmerge-all-constants
7660 Attempt to merge identical constants and identical variables.
7662 This option implies @option{-fmerge-constants}. In addition to
7663 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7664 arrays or initialized constant variables with integral or floating-point
7665 types. Languages like C or C++ require each variable, including multiple
7666 instances of the same variable in recursive calls, to have distinct locations,
7667 so using this option results in non-conforming
7670 @item -fmodulo-sched
7671 @opindex fmodulo-sched
7672 Perform swing modulo scheduling immediately before the first scheduling
7673 pass. This pass looks at innermost loops and reorders their
7674 instructions by overlapping different iterations.
7676 @item -fmodulo-sched-allow-regmoves
7677 @opindex fmodulo-sched-allow-regmoves
7678 Perform more aggressive SMS-based modulo scheduling with register moves
7679 allowed. By setting this flag certain anti-dependences edges are
7680 deleted, which triggers the generation of reg-moves based on the
7681 life-range analysis. This option is effective only with
7682 @option{-fmodulo-sched} enabled.
7684 @item -fno-branch-count-reg
7685 @opindex fno-branch-count-reg
7686 Avoid running a pass scanning for opportunities to use ``decrement and
7687 branch'' instructions on a count register instead of generating sequences
7688 of instructions that decrement a register, compare it against zero, and
7689 then branch based upon the result. This option is only meaningful on
7690 architectures that support such instructions, which include x86, PowerPC,
7691 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7692 doesn't remove the decrement and branch instructions from the generated
7693 instruction stream introduced by other optimization passes.
7695 Enabled by default at @option{-O1} and higher.
7697 The default is @option{-fbranch-count-reg}.
7699 @item -fno-function-cse
7700 @opindex fno-function-cse
7701 Do not put function addresses in registers; make each instruction that
7702 calls a constant function contain the function's address explicitly.
7704 This option results in less efficient code, but some strange hacks
7705 that alter the assembler output may be confused by the optimizations
7706 performed when this option is not used.
7708 The default is @option{-ffunction-cse}
7710 @item -fno-zero-initialized-in-bss
7711 @opindex fno-zero-initialized-in-bss
7712 If the target supports a BSS section, GCC by default puts variables that
7713 are initialized to zero into BSS@. This can save space in the resulting
7716 This option turns off this behavior because some programs explicitly
7717 rely on variables going to the data section---e.g., so that the
7718 resulting executable can find the beginning of that section and/or make
7719 assumptions based on that.
7721 The default is @option{-fzero-initialized-in-bss}.
7723 @item -fthread-jumps
7724 @opindex fthread-jumps
7725 Perform optimizations that check to see if a jump branches to a
7726 location where another comparison subsumed by the first is found. If
7727 so, the first branch is redirected to either the destination of the
7728 second branch or a point immediately following it, depending on whether
7729 the condition is known to be true or false.
7731 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7733 @item -fsplit-wide-types
7734 @opindex fsplit-wide-types
7735 When using a type that occupies multiple registers, such as @code{long
7736 long} on a 32-bit system, split the registers apart and allocate them
7737 independently. This normally generates better code for those types,
7738 but may make debugging more difficult.
7740 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7743 @item -fcse-follow-jumps
7744 @opindex fcse-follow-jumps
7745 In common subexpression elimination (CSE), scan through jump instructions
7746 when the target of the jump is not reached by any other path. For
7747 example, when CSE encounters an @code{if} statement with an
7748 @code{else} clause, CSE follows the jump when the condition
7751 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7753 @item -fcse-skip-blocks
7754 @opindex fcse-skip-blocks
7755 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7756 follow jumps that conditionally skip over blocks. When CSE
7757 encounters a simple @code{if} statement with no else clause,
7758 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7759 body of the @code{if}.
7761 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7763 @item -frerun-cse-after-loop
7764 @opindex frerun-cse-after-loop
7765 Re-run common subexpression elimination after loop optimizations are
7768 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7772 Perform a global common subexpression elimination pass.
7773 This pass also performs global constant and copy propagation.
7775 @emph{Note:} When compiling a program using computed gotos, a GCC
7776 extension, you may get better run-time performance if you disable
7777 the global common subexpression elimination pass by adding
7778 @option{-fno-gcse} to the command line.
7780 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7784 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7785 attempts to move loads that are only killed by stores into themselves. This
7786 allows a loop containing a load/store sequence to be changed to a load outside
7787 the loop, and a copy/store within the loop.
7789 Enabled by default when @option{-fgcse} is enabled.
7793 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7794 global common subexpression elimination. This pass attempts to move
7795 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7796 loops containing a load/store sequence can be changed to a load before
7797 the loop and a store after the loop.
7799 Not enabled at any optimization level.
7803 When @option{-fgcse-las} is enabled, the global common subexpression
7804 elimination pass eliminates redundant loads that come after stores to the
7805 same memory location (both partial and full redundancies).
7807 Not enabled at any optimization level.
7809 @item -fgcse-after-reload
7810 @opindex fgcse-after-reload
7811 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7812 pass is performed after reload. The purpose of this pass is to clean up
7815 @item -faggressive-loop-optimizations
7816 @opindex faggressive-loop-optimizations
7817 This option tells the loop optimizer to use language constraints to
7818 derive bounds for the number of iterations of a loop. This assumes that
7819 loop code does not invoke undefined behavior by for example causing signed
7820 integer overflows or out-of-bound array accesses. The bounds for the
7821 number of iterations of a loop are used to guide loop unrolling and peeling
7822 and loop exit test optimizations.
7823 This option is enabled by default.
7825 @item -funconstrained-commons
7826 @opindex funconstrained-commons
7827 This option tells the compiler that variables declared in common blocks
7828 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7829 prevents certain optimizations that depend on knowing the array bounds.
7831 @item -fcrossjumping
7832 @opindex fcrossjumping
7833 Perform cross-jumping transformation.
7834 This transformation unifies equivalent code and saves code size. The
7835 resulting code may or may not perform better than without cross-jumping.
7837 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7839 @item -fauto-inc-dec
7840 @opindex fauto-inc-dec
7841 Combine increments or decrements of addresses with memory accesses.
7842 This pass is always skipped on architectures that do not have
7843 instructions to support this. Enabled by default at @option{-O} and
7844 higher on architectures that support this.
7848 Perform dead code elimination (DCE) on RTL@.
7849 Enabled by default at @option{-O} and higher.
7853 Perform dead store elimination (DSE) on RTL@.
7854 Enabled by default at @option{-O} and higher.
7856 @item -fif-conversion
7857 @opindex fif-conversion
7858 Attempt to transform conditional jumps into branch-less equivalents. This
7859 includes use of conditional moves, min, max, set flags and abs instructions, and
7860 some tricks doable by standard arithmetics. The use of conditional execution
7861 on chips where it is available is controlled by @option{-fif-conversion2}.
7863 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7865 @item -fif-conversion2
7866 @opindex fif-conversion2
7867 Use conditional execution (where available) to transform conditional jumps into
7868 branch-less equivalents.
7870 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7872 @item -fdeclone-ctor-dtor
7873 @opindex fdeclone-ctor-dtor
7874 The C++ ABI requires multiple entry points for constructors and
7875 destructors: one for a base subobject, one for a complete object, and
7876 one for a virtual destructor that calls operator delete afterwards.
7877 For a hierarchy with virtual bases, the base and complete variants are
7878 clones, which means two copies of the function. With this option, the
7879 base and complete variants are changed to be thunks that call a common
7882 Enabled by @option{-Os}.
7884 @item -fdelete-null-pointer-checks
7885 @opindex fdelete-null-pointer-checks
7886 Assume that programs cannot safely dereference null pointers, and that
7887 no code or data element resides at address zero.
7888 This option enables simple constant
7889 folding optimizations at all optimization levels. In addition, other
7890 optimization passes in GCC use this flag to control global dataflow
7891 analyses that eliminate useless checks for null pointers; these assume
7892 that a memory access to address zero always results in a trap, so
7893 that if a pointer is checked after it has already been dereferenced,
7896 Note however that in some environments this assumption is not true.
7897 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7898 for programs that depend on that behavior.
7900 This option is enabled by default on most targets. On Nios II ELF, it
7901 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
7903 Passes that use the dataflow information
7904 are enabled independently at different optimization levels.
7906 @item -fdevirtualize
7907 @opindex fdevirtualize
7908 Attempt to convert calls to virtual functions to direct calls. This
7909 is done both within a procedure and interprocedurally as part of
7910 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7911 propagation (@option{-fipa-cp}).
7912 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7914 @item -fdevirtualize-speculatively
7915 @opindex fdevirtualize-speculatively
7916 Attempt to convert calls to virtual functions to speculative direct calls.
7917 Based on the analysis of the type inheritance graph, determine for a given call
7918 the set of likely targets. If the set is small, preferably of size 1, change
7919 the call into a conditional deciding between direct and indirect calls. The
7920 speculative calls enable more optimizations, such as inlining. When they seem
7921 useless after further optimization, they are converted back into original form.
7923 @item -fdevirtualize-at-ltrans
7924 @opindex fdevirtualize-at-ltrans
7925 Stream extra information needed for aggressive devirtualization when running
7926 the link-time optimizer in local transformation mode.
7927 This option enables more devirtualization but
7928 significantly increases the size of streamed data. For this reason it is
7929 disabled by default.
7931 @item -fexpensive-optimizations
7932 @opindex fexpensive-optimizations
7933 Perform a number of minor optimizations that are relatively expensive.
7935 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7939 Attempt to remove redundant extension instructions. This is especially
7940 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7941 registers after writing to their lower 32-bit half.
7943 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7944 @option{-O3}, @option{-Os}.
7946 @item -fno-lifetime-dse
7947 @opindex fno-lifetime-dse
7948 In C++ the value of an object is only affected by changes within its
7949 lifetime: when the constructor begins, the object has an indeterminate
7950 value, and any changes during the lifetime of the object are dead when
7951 the object is destroyed. Normally dead store elimination will take
7952 advantage of this; if your code relies on the value of the object
7953 storage persisting beyond the lifetime of the object, you can use this
7954 flag to disable this optimization. To preserve stores before the
7955 constructor starts (e.g. because your operator new clears the object
7956 storage) but still treat the object as dead after the destructor you,
7957 can use @option{-flifetime-dse=1}. The default behavior can be
7958 explicitly selected with @option{-flifetime-dse=2}.
7959 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7961 @item -flive-range-shrinkage
7962 @opindex flive-range-shrinkage
7963 Attempt to decrease register pressure through register live range
7964 shrinkage. This is helpful for fast processors with small or moderate
7967 @item -fira-algorithm=@var{algorithm}
7968 @opindex fira-algorithm
7969 Use the specified coloring algorithm for the integrated register
7970 allocator. The @var{algorithm} argument can be @samp{priority}, which
7971 specifies Chow's priority coloring, or @samp{CB}, which specifies
7972 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7973 for all architectures, but for those targets that do support it, it is
7974 the default because it generates better code.
7976 @item -fira-region=@var{region}
7977 @opindex fira-region
7978 Use specified regions for the integrated register allocator. The
7979 @var{region} argument should be one of the following:
7984 Use all loops as register allocation regions.
7985 This can give the best results for machines with a small and/or
7986 irregular register set.
7989 Use all loops except for loops with small register pressure
7990 as the regions. This value usually gives
7991 the best results in most cases and for most architectures,
7992 and is enabled by default when compiling with optimization for speed
7993 (@option{-O}, @option{-O2}, @dots{}).
7996 Use all functions as a single region.
7997 This typically results in the smallest code size, and is enabled by default for
7998 @option{-Os} or @option{-O0}.
8002 @item -fira-hoist-pressure
8003 @opindex fira-hoist-pressure
8004 Use IRA to evaluate register pressure in the code hoisting pass for
8005 decisions to hoist expressions. This option usually results in smaller
8006 code, but it can slow the compiler down.
8008 This option is enabled at level @option{-Os} for all targets.
8010 @item -fira-loop-pressure
8011 @opindex fira-loop-pressure
8012 Use IRA to evaluate register pressure in loops for decisions to move
8013 loop invariants. This option usually results in generation
8014 of faster and smaller code on machines with large register files (>= 32
8015 registers), but it can slow the compiler down.
8017 This option is enabled at level @option{-O3} for some targets.
8019 @item -fno-ira-share-save-slots
8020 @opindex fno-ira-share-save-slots
8021 Disable sharing of stack slots used for saving call-used hard
8022 registers living through a call. Each hard register gets a
8023 separate stack slot, and as a result function stack frames are
8026 @item -fno-ira-share-spill-slots
8027 @opindex fno-ira-share-spill-slots
8028 Disable sharing of stack slots allocated for pseudo-registers. Each
8029 pseudo-register that does not get a hard register gets a separate
8030 stack slot, and as a result function stack frames are larger.
8034 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8035 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8036 values if it is profitable.
8038 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8040 @item -fdelayed-branch
8041 @opindex fdelayed-branch
8042 If supported for the target machine, attempt to reorder instructions
8043 to exploit instruction slots available after delayed branch
8046 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8048 @item -fschedule-insns
8049 @opindex fschedule-insns
8050 If supported for the target machine, attempt to reorder instructions to
8051 eliminate execution stalls due to required data being unavailable. This
8052 helps machines that have slow floating point or memory load instructions
8053 by allowing other instructions to be issued until the result of the load
8054 or floating-point instruction is required.
8056 Enabled at levels @option{-O2}, @option{-O3}.
8058 @item -fschedule-insns2
8059 @opindex fschedule-insns2
8060 Similar to @option{-fschedule-insns}, but requests an additional pass of
8061 instruction scheduling after register allocation has been done. This is
8062 especially useful on machines with a relatively small number of
8063 registers and where memory load instructions take more than one cycle.
8065 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8067 @item -fno-sched-interblock
8068 @opindex fno-sched-interblock
8069 Don't schedule instructions across basic blocks. This is normally
8070 enabled by default when scheduling before register allocation, i.e.@:
8071 with @option{-fschedule-insns} or at @option{-O2} or higher.
8073 @item -fno-sched-spec
8074 @opindex fno-sched-spec
8075 Don't allow speculative motion of non-load instructions. This is normally
8076 enabled by default when scheduling before register allocation, i.e.@:
8077 with @option{-fschedule-insns} or at @option{-O2} or higher.
8079 @item -fsched-pressure
8080 @opindex fsched-pressure
8081 Enable register pressure sensitive insn scheduling before register
8082 allocation. This only makes sense when scheduling before register
8083 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8084 @option{-O2} or higher. Usage of this option can improve the
8085 generated code and decrease its size by preventing register pressure
8086 increase above the number of available hard registers and subsequent
8087 spills in register allocation.
8089 @item -fsched-spec-load
8090 @opindex fsched-spec-load
8091 Allow speculative motion of some load instructions. This only makes
8092 sense when scheduling before register allocation, i.e.@: with
8093 @option{-fschedule-insns} or at @option{-O2} or higher.
8095 @item -fsched-spec-load-dangerous
8096 @opindex fsched-spec-load-dangerous
8097 Allow speculative motion of more load instructions. This only makes
8098 sense when scheduling before register allocation, i.e.@: with
8099 @option{-fschedule-insns} or at @option{-O2} or higher.
8101 @item -fsched-stalled-insns
8102 @itemx -fsched-stalled-insns=@var{n}
8103 @opindex fsched-stalled-insns
8104 Define how many insns (if any) can be moved prematurely from the queue
8105 of stalled insns into the ready list during the second scheduling pass.
8106 @option{-fno-sched-stalled-insns} means that no insns are moved
8107 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8108 on how many queued insns can be moved prematurely.
8109 @option{-fsched-stalled-insns} without a value is equivalent to
8110 @option{-fsched-stalled-insns=1}.
8112 @item -fsched-stalled-insns-dep
8113 @itemx -fsched-stalled-insns-dep=@var{n}
8114 @opindex fsched-stalled-insns-dep
8115 Define how many insn groups (cycles) are examined for a dependency
8116 on a stalled insn that is a candidate for premature removal from the queue
8117 of stalled insns. This has an effect only during the second scheduling pass,
8118 and only if @option{-fsched-stalled-insns} is used.
8119 @option{-fno-sched-stalled-insns-dep} is equivalent to
8120 @option{-fsched-stalled-insns-dep=0}.
8121 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8122 @option{-fsched-stalled-insns-dep=1}.
8124 @item -fsched2-use-superblocks
8125 @opindex fsched2-use-superblocks
8126 When scheduling after register allocation, use superblock scheduling.
8127 This allows motion across basic block boundaries,
8128 resulting in faster schedules. This option is experimental, as not all machine
8129 descriptions used by GCC model the CPU closely enough to avoid unreliable
8130 results from the algorithm.
8132 This only makes sense when scheduling after register allocation, i.e.@: with
8133 @option{-fschedule-insns2} or at @option{-O2} or higher.
8135 @item -fsched-group-heuristic
8136 @opindex fsched-group-heuristic
8137 Enable the group heuristic in the scheduler. This heuristic favors
8138 the instruction that belongs to a schedule group. This is enabled
8139 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8140 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8142 @item -fsched-critical-path-heuristic
8143 @opindex fsched-critical-path-heuristic
8144 Enable the critical-path heuristic in the scheduler. This heuristic favors
8145 instructions on the critical path. This is enabled by default when
8146 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8147 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8149 @item -fsched-spec-insn-heuristic
8150 @opindex fsched-spec-insn-heuristic
8151 Enable the speculative instruction heuristic in the scheduler. This
8152 heuristic favors speculative instructions with greater dependency weakness.
8153 This is enabled by default when scheduling is enabled, i.e.@:
8154 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8155 or at @option{-O2} or higher.
8157 @item -fsched-rank-heuristic
8158 @opindex fsched-rank-heuristic
8159 Enable the rank heuristic in the scheduler. This heuristic favors
8160 the instruction belonging to a basic block with greater size or frequency.
8161 This is enabled by default when scheduling is enabled, i.e.@:
8162 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8163 at @option{-O2} or higher.
8165 @item -fsched-last-insn-heuristic
8166 @opindex fsched-last-insn-heuristic
8167 Enable the last-instruction heuristic in the scheduler. This heuristic
8168 favors the instruction that is less dependent on the last instruction
8169 scheduled. This is enabled by default when scheduling is enabled,
8170 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8171 at @option{-O2} or higher.
8173 @item -fsched-dep-count-heuristic
8174 @opindex fsched-dep-count-heuristic
8175 Enable the dependent-count heuristic in the scheduler. This heuristic
8176 favors the instruction that has more instructions depending on it.
8177 This is enabled by default when scheduling is enabled, i.e.@:
8178 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8179 at @option{-O2} or higher.
8181 @item -freschedule-modulo-scheduled-loops
8182 @opindex freschedule-modulo-scheduled-loops
8183 Modulo scheduling is performed before traditional scheduling. If a loop
8184 is modulo scheduled, later scheduling passes may change its schedule.
8185 Use this option to control that behavior.
8187 @item -fselective-scheduling
8188 @opindex fselective-scheduling
8189 Schedule instructions using selective scheduling algorithm. Selective
8190 scheduling runs instead of the first scheduler pass.
8192 @item -fselective-scheduling2
8193 @opindex fselective-scheduling2
8194 Schedule instructions using selective scheduling algorithm. Selective
8195 scheduling runs instead of the second scheduler pass.
8197 @item -fsel-sched-pipelining
8198 @opindex fsel-sched-pipelining
8199 Enable software pipelining of innermost loops during selective scheduling.
8200 This option has no effect unless one of @option{-fselective-scheduling} or
8201 @option{-fselective-scheduling2} is turned on.
8203 @item -fsel-sched-pipelining-outer-loops
8204 @opindex fsel-sched-pipelining-outer-loops
8205 When pipelining loops during selective scheduling, also pipeline outer loops.
8206 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8208 @item -fsemantic-interposition
8209 @opindex fsemantic-interposition
8210 Some object formats, like ELF, allow interposing of symbols by the
8212 This means that for symbols exported from the DSO, the compiler cannot perform
8213 interprocedural propagation, inlining and other optimizations in anticipation
8214 that the function or variable in question may change. While this feature is
8215 useful, for example, to rewrite memory allocation functions by a debugging
8216 implementation, it is expensive in the terms of code quality.
8217 With @option{-fno-semantic-interposition} the compiler assumes that
8218 if interposition happens for functions the overwriting function will have
8219 precisely the same semantics (and side effects).
8220 Similarly if interposition happens
8221 for variables, the constructor of the variable will be the same. The flag
8222 has no effect for functions explicitly declared inline
8223 (where it is never allowed for interposition to change semantics)
8224 and for symbols explicitly declared weak.
8227 @opindex fshrink-wrap
8228 Emit function prologues only before parts of the function that need it,
8229 rather than at the top of the function. This flag is enabled by default at
8230 @option{-O} and higher.
8232 @item -fshrink-wrap-separate
8233 @opindex fshrink-wrap-separate
8234 Shrink-wrap separate parts of the prologue and epilogue separately, so that
8235 those parts are only executed when needed.
8236 This option is on by default, but has no effect unless @option{-fshrink-wrap}
8237 is also turned on and the target supports this.
8239 @item -fcaller-saves
8240 @opindex fcaller-saves
8241 Enable allocation of values to registers that are clobbered by
8242 function calls, by emitting extra instructions to save and restore the
8243 registers around such calls. Such allocation is done only when it
8244 seems to result in better code.
8246 This option is always enabled by default on certain machines, usually
8247 those which have no call-preserved registers to use instead.
8249 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8251 @item -fcombine-stack-adjustments
8252 @opindex fcombine-stack-adjustments
8253 Tracks stack adjustments (pushes and pops) and stack memory references
8254 and then tries to find ways to combine them.
8256 Enabled by default at @option{-O1} and higher.
8260 Use caller save registers for allocation if those registers are not used by
8261 any called function. In that case it is not necessary to save and restore
8262 them around calls. This is only possible if called functions are part of
8263 same compilation unit as current function and they are compiled before it.
8265 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
8266 is disabled if generated code will be instrumented for profiling
8267 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
8268 exactly (this happens on targets that do not expose prologues
8269 and epilogues in RTL).
8271 @item -fconserve-stack
8272 @opindex fconserve-stack
8273 Attempt to minimize stack usage. The compiler attempts to use less
8274 stack space, even if that makes the program slower. This option
8275 implies setting the @option{large-stack-frame} parameter to 100
8276 and the @option{large-stack-frame-growth} parameter to 400.
8278 @item -ftree-reassoc
8279 @opindex ftree-reassoc
8280 Perform reassociation on trees. This flag is enabled by default
8281 at @option{-O} and higher.
8283 @item -fcode-hoisting
8284 @opindex fcode-hoisting
8285 Perform code hoisting. Code hoisting tries to move the
8286 evaluation of expressions executed on all paths to the function exit
8287 as early as possible. This is especially useful as a code size
8288 optimization, but it often helps for code speed as well.
8289 This flag is enabled by default at @option{-O2} and higher.
8293 Perform partial redundancy elimination (PRE) on trees. This flag is
8294 enabled by default at @option{-O2} and @option{-O3}.
8296 @item -ftree-partial-pre
8297 @opindex ftree-partial-pre
8298 Make partial redundancy elimination (PRE) more aggressive. This flag is
8299 enabled by default at @option{-O3}.
8301 @item -ftree-forwprop
8302 @opindex ftree-forwprop
8303 Perform forward propagation on trees. This flag is enabled by default
8304 at @option{-O} and higher.
8308 Perform full redundancy elimination (FRE) on trees. The difference
8309 between FRE and PRE is that FRE only considers expressions
8310 that are computed on all paths leading to the redundant computation.
8311 This analysis is faster than PRE, though it exposes fewer redundancies.
8312 This flag is enabled by default at @option{-O} and higher.
8314 @item -ftree-phiprop
8315 @opindex ftree-phiprop
8316 Perform hoisting of loads from conditional pointers on trees. This
8317 pass is enabled by default at @option{-O} and higher.
8319 @item -fhoist-adjacent-loads
8320 @opindex fhoist-adjacent-loads
8321 Speculatively hoist loads from both branches of an if-then-else if the
8322 loads are from adjacent locations in the same structure and the target
8323 architecture has a conditional move instruction. This flag is enabled
8324 by default at @option{-O2} and higher.
8326 @item -ftree-copy-prop
8327 @opindex ftree-copy-prop
8328 Perform copy propagation on trees. This pass eliminates unnecessary
8329 copy operations. This flag is enabled by default at @option{-O} and
8332 @item -fipa-pure-const
8333 @opindex fipa-pure-const
8334 Discover which functions are pure or constant.
8335 Enabled by default at @option{-O} and higher.
8337 @item -fipa-reference
8338 @opindex fipa-reference
8339 Discover which static variables do not escape the
8341 Enabled by default at @option{-O} and higher.
8345 Perform interprocedural pointer analysis and interprocedural modification
8346 and reference analysis. This option can cause excessive memory and
8347 compile-time usage on large compilation units. It is not enabled by
8348 default at any optimization level.
8351 @opindex fipa-profile
8352 Perform interprocedural profile propagation. The functions called only from
8353 cold functions are marked as cold. Also functions executed once (such as
8354 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8355 functions and loop less parts of functions executed once are then optimized for
8357 Enabled by default at @option{-O} and higher.
8361 Perform interprocedural constant propagation.
8362 This optimization analyzes the program to determine when values passed
8363 to functions are constants and then optimizes accordingly.
8364 This optimization can substantially increase performance
8365 if the application has constants passed to functions.
8366 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8368 @item -fipa-cp-clone
8369 @opindex fipa-cp-clone
8370 Perform function cloning to make interprocedural constant propagation stronger.
8371 When enabled, interprocedural constant propagation performs function cloning
8372 when externally visible function can be called with constant arguments.
8373 Because this optimization can create multiple copies of functions,
8374 it may significantly increase code size
8375 (see @option{--param ipcp-unit-growth=@var{value}}).
8376 This flag is enabled by default at @option{-O3}.
8379 @opindex -fipa-bit-cp
8380 When enabled, perform interprocedural bitwise constant
8381 propagation. This flag is enabled by default at @option{-O2}. It
8382 requires that @option{-fipa-cp} is enabled.
8386 When enabled, perform interprocedural propagation of value
8387 ranges. This flag is enabled by default at @option{-O2}. It requires
8388 that @option{-fipa-cp} is enabled.
8392 Perform Identical Code Folding for functions and read-only variables.
8393 The optimization reduces code size and may disturb unwind stacks by replacing
8394 a function by equivalent one with a different name. The optimization works
8395 more effectively with link-time optimization enabled.
8397 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8398 works on different levels and thus the optimizations are not same - there are
8399 equivalences that are found only by GCC and equivalences found only by Gold.
8401 This flag is enabled by default at @option{-O2} and @option{-Os}.
8403 @item -fisolate-erroneous-paths-dereference
8404 @opindex fisolate-erroneous-paths-dereference
8405 Detect paths that trigger erroneous or undefined behavior due to
8406 dereferencing a null pointer. Isolate those paths from the main control
8407 flow and turn the statement with erroneous or undefined behavior into a trap.
8408 This flag is enabled by default at @option{-O2} and higher and depends on
8409 @option{-fdelete-null-pointer-checks} also being enabled.
8411 @item -fisolate-erroneous-paths-attribute
8412 @opindex fisolate-erroneous-paths-attribute
8413 Detect paths that trigger erroneous or undefined behavior due a null value
8414 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8415 attribute. Isolate those paths from the main control flow and turn the
8416 statement with erroneous or undefined behavior into a trap. This is not
8417 currently enabled, but may be enabled by @option{-O2} in the future.
8421 Perform forward store motion on trees. This flag is
8422 enabled by default at @option{-O} and higher.
8424 @item -ftree-bit-ccp
8425 @opindex ftree-bit-ccp
8426 Perform sparse conditional bit constant propagation on trees and propagate
8427 pointer alignment information.
8428 This pass only operates on local scalar variables and is enabled by default
8429 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8433 Perform sparse conditional constant propagation (CCP) on trees. This
8434 pass only operates on local scalar variables and is enabled by default
8435 at @option{-O} and higher.
8437 @item -fssa-backprop
8438 @opindex fssa-backprop
8439 Propagate information about uses of a value up the definition chain
8440 in order to simplify the definitions. For example, this pass strips
8441 sign operations if the sign of a value never matters. The flag is
8442 enabled by default at @option{-O} and higher.
8445 @opindex fssa-phiopt
8446 Perform pattern matching on SSA PHI nodes to optimize conditional
8447 code. This pass is enabled by default at @option{-O} and higher.
8449 @item -ftree-switch-conversion
8450 @opindex ftree-switch-conversion
8451 Perform conversion of simple initializations in a switch to
8452 initializations from a scalar array. This flag is enabled by default
8453 at @option{-O2} and higher.
8455 @item -ftree-tail-merge
8456 @opindex ftree-tail-merge
8457 Look for identical code sequences. When found, replace one with a jump to the
8458 other. This optimization is known as tail merging or cross jumping. This flag
8459 is enabled by default at @option{-O2} and higher. The compilation time
8461 be limited using @option{max-tail-merge-comparisons} parameter and
8462 @option{max-tail-merge-iterations} parameter.
8466 Perform dead code elimination (DCE) on trees. This flag is enabled by
8467 default at @option{-O} and higher.
8469 @item -ftree-builtin-call-dce
8470 @opindex ftree-builtin-call-dce
8471 Perform conditional dead code elimination (DCE) for calls to built-in functions
8472 that may set @code{errno} but are otherwise side-effect free. This flag is
8473 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8476 @item -ftree-dominator-opts
8477 @opindex ftree-dominator-opts
8478 Perform a variety of simple scalar cleanups (constant/copy
8479 propagation, redundancy elimination, range propagation and expression
8480 simplification) based on a dominator tree traversal. This also
8481 performs jump threading (to reduce jumps to jumps). This flag is
8482 enabled by default at @option{-O} and higher.
8486 Perform dead store elimination (DSE) on trees. A dead store is a store into
8487 a memory location that is later overwritten by another store without
8488 any intervening loads. In this case the earlier store can be deleted. This
8489 flag is enabled by default at @option{-O} and higher.
8493 Perform loop header copying on trees. This is beneficial since it increases
8494 effectiveness of code motion optimizations. It also saves one jump. This flag
8495 is enabled by default at @option{-O} and higher. It is not enabled
8496 for @option{-Os}, since it usually increases code size.
8498 @item -ftree-loop-optimize
8499 @opindex ftree-loop-optimize
8500 Perform loop optimizations on trees. This flag is enabled by default
8501 at @option{-O} and higher.
8503 @item -ftree-loop-linear
8504 @itemx -floop-interchange
8505 @itemx -floop-strip-mine
8507 @itemx -floop-unroll-and-jam
8508 @opindex ftree-loop-linear
8509 @opindex floop-interchange
8510 @opindex floop-strip-mine
8511 @opindex floop-block
8512 @opindex floop-unroll-and-jam
8513 Perform loop nest optimizations. Same as
8514 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8515 to be configured with @option{--with-isl} to enable the Graphite loop
8516 transformation infrastructure.
8518 @item -fgraphite-identity
8519 @opindex fgraphite-identity
8520 Enable the identity transformation for graphite. For every SCoP we generate
8521 the polyhedral representation and transform it back to gimple. Using
8522 @option{-fgraphite-identity} we can check the costs or benefits of the
8523 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8524 are also performed by the code generator isl, like index splitting and
8525 dead code elimination in loops.
8527 @item -floop-nest-optimize
8528 @opindex floop-nest-optimize
8529 Enable the isl based loop nest optimizer. This is a generic loop nest
8530 optimizer based on the Pluto optimization algorithms. It calculates a loop
8531 structure optimized for data-locality and parallelism. This option
8534 @item -floop-parallelize-all
8535 @opindex floop-parallelize-all
8536 Use the Graphite data dependence analysis to identify loops that can
8537 be parallelized. Parallelize all the loops that can be analyzed to
8538 not contain loop carried dependences without checking that it is
8539 profitable to parallelize the loops.
8541 @item -ftree-coalesce-vars
8542 @opindex ftree-coalesce-vars
8543 While transforming the program out of the SSA representation, attempt to
8544 reduce copying by coalescing versions of different user-defined
8545 variables, instead of just compiler temporaries. This may severely
8546 limit the ability to debug an optimized program compiled with
8547 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8548 prevents SSA coalescing of user variables. This option is enabled by
8549 default if optimization is enabled, and it does very little otherwise.
8551 @item -ftree-loop-if-convert
8552 @opindex ftree-loop-if-convert
8553 Attempt to transform conditional jumps in the innermost loops to
8554 branch-less equivalents. The intent is to remove control-flow from
8555 the innermost loops in order to improve the ability of the
8556 vectorization pass to handle these loops. This is enabled by default
8557 if vectorization is enabled.
8559 @item -ftree-loop-distribution
8560 @opindex ftree-loop-distribution
8561 Perform loop distribution. This flag can improve cache performance on
8562 big loop bodies and allow further loop optimizations, like
8563 parallelization or vectorization, to take place. For example, the loop
8580 @item -ftree-loop-distribute-patterns
8581 @opindex ftree-loop-distribute-patterns
8582 Perform loop distribution of patterns that can be code generated with
8583 calls to a library. This flag is enabled by default at @option{-O3}.
8585 This pass distributes the initialization loops and generates a call to
8586 memset zero. For example, the loop
8602 and the initialization loop is transformed into a call to memset zero.
8604 @item -ftree-loop-im
8605 @opindex ftree-loop-im
8606 Perform loop invariant motion on trees. This pass moves only invariants that
8607 are hard to handle at RTL level (function calls, operations that expand to
8608 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8609 operands of conditions that are invariant out of the loop, so that we can use
8610 just trivial invariantness analysis in loop unswitching. The pass also includes
8613 @item -ftree-loop-ivcanon
8614 @opindex ftree-loop-ivcanon
8615 Create a canonical counter for number of iterations in loops for which
8616 determining number of iterations requires complicated analysis. Later
8617 optimizations then may determine the number easily. Useful especially
8618 in connection with unrolling.
8622 Perform induction variable optimizations (strength reduction, induction
8623 variable merging and induction variable elimination) on trees.
8625 @item -ftree-parallelize-loops=n
8626 @opindex ftree-parallelize-loops
8627 Parallelize loops, i.e., split their iteration space to run in n threads.
8628 This is only possible for loops whose iterations are independent
8629 and can be arbitrarily reordered. The optimization is only
8630 profitable on multiprocessor machines, for loops that are CPU-intensive,
8631 rather than constrained e.g.@: by memory bandwidth. This option
8632 implies @option{-pthread}, and thus is only supported on targets
8633 that have support for @option{-pthread}.
8637 Perform function-local points-to analysis on trees. This flag is
8638 enabled by default at @option{-O} and higher.
8642 Perform scalar replacement of aggregates. This pass replaces structure
8643 references with scalars to prevent committing structures to memory too
8644 early. This flag is enabled by default at @option{-O} and higher.
8646 @item -fstore-merging
8647 @opindex fstore-merging
8648 Perform merging of narrow stores to consecutive memory addresses. This pass
8649 merges contiguous stores of immediate values narrower than a word into fewer
8650 wider stores to reduce the number of instructions. This is enabled by default
8651 at @option{-O2} and higher as well as @option{-Os}.
8655 Perform temporary expression replacement during the SSA->normal phase. Single
8656 use/single def temporaries are replaced at their use location with their
8657 defining expression. This results in non-GIMPLE code, but gives the expanders
8658 much more complex trees to work on resulting in better RTL generation. This is
8659 enabled by default at @option{-O} and higher.
8663 Perform straight-line strength reduction on trees. This recognizes related
8664 expressions involving multiplications and replaces them by less expensive
8665 calculations when possible. This is enabled by default at @option{-O} and
8668 @item -ftree-vectorize
8669 @opindex ftree-vectorize
8670 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8671 and @option{-ftree-slp-vectorize} if not explicitly specified.
8673 @item -ftree-loop-vectorize
8674 @opindex ftree-loop-vectorize
8675 Perform loop vectorization on trees. This flag is enabled by default at
8676 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8678 @item -ftree-slp-vectorize
8679 @opindex ftree-slp-vectorize
8680 Perform basic block vectorization on trees. This flag is enabled by default at
8681 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8683 @item -fvect-cost-model=@var{model}
8684 @opindex fvect-cost-model
8685 Alter the cost model used for vectorization. The @var{model} argument
8686 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8687 With the @samp{unlimited} model the vectorized code-path is assumed
8688 to be profitable while with the @samp{dynamic} model a runtime check
8689 guards the vectorized code-path to enable it only for iteration
8690 counts that will likely execute faster than when executing the original
8691 scalar loop. The @samp{cheap} model disables vectorization of
8692 loops where doing so would be cost prohibitive for example due to
8693 required runtime checks for data dependence or alignment but otherwise
8694 is equal to the @samp{dynamic} model.
8695 The default cost model depends on other optimization flags and is
8696 either @samp{dynamic} or @samp{cheap}.
8698 @item -fsimd-cost-model=@var{model}
8699 @opindex fsimd-cost-model
8700 Alter the cost model used for vectorization of loops marked with the OpenMP
8701 or Cilk Plus simd directive. The @var{model} argument should be one of
8702 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8703 have the same meaning as described in @option{-fvect-cost-model} and by
8704 default a cost model defined with @option{-fvect-cost-model} is used.
8708 Perform Value Range Propagation on trees. This is similar to the
8709 constant propagation pass, but instead of values, ranges of values are
8710 propagated. This allows the optimizers to remove unnecessary range
8711 checks like array bound checks and null pointer checks. This is
8712 enabled by default at @option{-O2} and higher. Null pointer check
8713 elimination is only done if @option{-fdelete-null-pointer-checks} is
8717 @opindex fsplit-paths
8718 Split paths leading to loop backedges. This can improve dead code
8719 elimination and common subexpression elimination. This is enabled by
8720 default at @option{-O2} and above.
8722 @item -fsplit-ivs-in-unroller
8723 @opindex fsplit-ivs-in-unroller
8724 Enables expression of values of induction variables in later iterations
8725 of the unrolled loop using the value in the first iteration. This breaks
8726 long dependency chains, thus improving efficiency of the scheduling passes.
8728 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8729 same effect. However, that is not reliable in cases where the loop body
8730 is more complicated than a single basic block. It also does not work at all
8731 on some architectures due to restrictions in the CSE pass.
8733 This optimization is enabled by default.
8735 @item -fvariable-expansion-in-unroller
8736 @opindex fvariable-expansion-in-unroller
8737 With this option, the compiler creates multiple copies of some
8738 local variables when unrolling a loop, which can result in superior code.
8740 @item -fpartial-inlining
8741 @opindex fpartial-inlining
8742 Inline parts of functions. This option has any effect only
8743 when inlining itself is turned on by the @option{-finline-functions}
8744 or @option{-finline-small-functions} options.
8746 Enabled at level @option{-O2}.
8748 @item -fpredictive-commoning
8749 @opindex fpredictive-commoning
8750 Perform predictive commoning optimization, i.e., reusing computations
8751 (especially memory loads and stores) performed in previous
8752 iterations of loops.
8754 This option is enabled at level @option{-O3}.
8756 @item -fprefetch-loop-arrays
8757 @opindex fprefetch-loop-arrays
8758 If supported by the target machine, generate instructions to prefetch
8759 memory to improve the performance of loops that access large arrays.
8761 This option may generate better or worse code; results are highly
8762 dependent on the structure of loops within the source code.
8764 Disabled at level @option{-Os}.
8766 @item -fno-printf-return-value
8767 @opindex fno-printf-return-value
8768 Do not substitute constants for known return value of formatted output
8769 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
8770 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
8771 transformation allows GCC to optimize or even eliminate branches based
8772 on the known return value of these functions called with arguments that
8773 are either constant, or whose values are known to be in a range that
8774 makes determining the exact return value possible. For example, when
8775 @option{-fprintf-return-value} is in effect, both the branch and the
8776 body of the @code{if} statement (but not the call to @code{snprint})
8777 can be optimized away when @code{i} is a 32-bit or smaller integer
8778 because the return value is guaranteed to be at most 8.
8782 if (snprintf (buf, "%08x", i) >= sizeof buf)
8786 The @option{-fprintf-return-value} option relies on other optimizations
8787 and yields best results with @option{-O2}. It works in tandem with the
8788 @option{-Wformat-overflow} and @option{-Wformat-truncation} options.
8789 The @option{-fprintf-return-value} option is enabled by default.
8792 @itemx -fno-peephole2
8793 @opindex fno-peephole
8794 @opindex fno-peephole2
8795 Disable any machine-specific peephole optimizations. The difference
8796 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8797 are implemented in the compiler; some targets use one, some use the
8798 other, a few use both.
8800 @option{-fpeephole} is enabled by default.
8801 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8803 @item -fno-guess-branch-probability
8804 @opindex fno-guess-branch-probability
8805 Do not guess branch probabilities using heuristics.
8807 GCC uses heuristics to guess branch probabilities if they are
8808 not provided by profiling feedback (@option{-fprofile-arcs}). These
8809 heuristics are based on the control flow graph. If some branch probabilities
8810 are specified by @code{__builtin_expect}, then the heuristics are
8811 used to guess branch probabilities for the rest of the control flow graph,
8812 taking the @code{__builtin_expect} info into account. The interactions
8813 between the heuristics and @code{__builtin_expect} can be complex, and in
8814 some cases, it may be useful to disable the heuristics so that the effects
8815 of @code{__builtin_expect} are easier to understand.
8817 The default is @option{-fguess-branch-probability} at levels
8818 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8820 @item -freorder-blocks
8821 @opindex freorder-blocks
8822 Reorder basic blocks in the compiled function in order to reduce number of
8823 taken branches and improve code locality.
8825 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8827 @item -freorder-blocks-algorithm=@var{algorithm}
8828 @opindex freorder-blocks-algorithm
8829 Use the specified algorithm for basic block reordering. The
8830 @var{algorithm} argument can be @samp{simple}, which does not increase
8831 code size (except sometimes due to secondary effects like alignment),
8832 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8833 put all often executed code together, minimizing the number of branches
8834 executed by making extra copies of code.
8836 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8837 @samp{stc} at levels @option{-O2}, @option{-O3}.
8839 @item -freorder-blocks-and-partition
8840 @opindex freorder-blocks-and-partition
8841 In addition to reordering basic blocks in the compiled function, in order
8842 to reduce number of taken branches, partitions hot and cold basic blocks
8843 into separate sections of the assembly and @file{.o} files, to improve
8844 paging and cache locality performance.
8846 This optimization is automatically turned off in the presence of
8847 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
8848 section attribute and on any architecture that does not support named
8849 sections. When @option{-fsplit-stack} is used this option is not
8850 enabled by default (to avoid linker errors), but may be enabled
8851 explicitly (if using a working linker).
8853 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8855 @item -freorder-functions
8856 @opindex freorder-functions
8857 Reorder functions in the object file in order to
8858 improve code locality. This is implemented by using special
8859 subsections @code{.text.hot} for most frequently executed functions and
8860 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8861 the linker so object file format must support named sections and linker must
8862 place them in a reasonable way.
8864 Also profile feedback must be available to make this option effective. See
8865 @option{-fprofile-arcs} for details.
8867 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8869 @item -fstrict-aliasing
8870 @opindex fstrict-aliasing
8871 Allow the compiler to assume the strictest aliasing rules applicable to
8872 the language being compiled. For C (and C++), this activates
8873 optimizations based on the type of expressions. In particular, an
8874 object of one type is assumed never to reside at the same address as an
8875 object of a different type, unless the types are almost the same. For
8876 example, an @code{unsigned int} can alias an @code{int}, but not a
8877 @code{void*} or a @code{double}. A character type may alias any other
8880 @anchor{Type-punning}Pay special attention to code like this:
8893 The practice of reading from a different union member than the one most
8894 recently written to (called ``type-punning'') is common. Even with
8895 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8896 is accessed through the union type. So, the code above works as
8897 expected. @xref{Structures unions enumerations and bit-fields
8898 implementation}. However, this code might not:
8909 Similarly, access by taking the address, casting the resulting pointer
8910 and dereferencing the result has undefined behavior, even if the cast
8911 uses a union type, e.g.:
8915 return ((union a_union *) &d)->i;
8919 The @option{-fstrict-aliasing} option is enabled at levels
8920 @option{-O2}, @option{-O3}, @option{-Os}.
8922 @item -falign-functions
8923 @itemx -falign-functions=@var{n}
8924 @opindex falign-functions
8925 Align the start of functions to the next power-of-two greater than
8926 @var{n}, skipping up to @var{n} bytes. For instance,
8927 @option{-falign-functions=32} aligns functions to the next 32-byte
8928 boundary, but @option{-falign-functions=24} aligns to the next
8929 32-byte boundary only if this can be done by skipping 23 bytes or less.
8931 @option{-fno-align-functions} and @option{-falign-functions=1} are
8932 equivalent and mean that functions are not aligned.
8934 Some assemblers only support this flag when @var{n} is a power of two;
8935 in that case, it is rounded up.
8937 If @var{n} is not specified or is zero, use a machine-dependent default.
8939 Enabled at levels @option{-O2}, @option{-O3}.
8941 @item -flimit-function-alignment
8942 If this option is enabled, the compiler tries to avoid unnecessarily
8943 overaligning functions. It attempts to instruct the assembler to align
8944 by the amount specified by @option{-falign-functions}, but not to
8945 skip more bytes than the size of the function.
8947 @item -falign-labels
8948 @itemx -falign-labels=@var{n}
8949 @opindex falign-labels
8950 Align all branch targets to a power-of-two boundary, skipping up to
8951 @var{n} bytes like @option{-falign-functions}. This option can easily
8952 make code slower, because it must insert dummy operations for when the
8953 branch target is reached in the usual flow of the code.
8955 @option{-fno-align-labels} and @option{-falign-labels=1} are
8956 equivalent and mean that labels are not aligned.
8958 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8959 are greater than this value, then their values are used instead.
8961 If @var{n} is not specified or is zero, use a machine-dependent default
8962 which is very likely to be @samp{1}, meaning no alignment.
8964 Enabled at levels @option{-O2}, @option{-O3}.
8967 @itemx -falign-loops=@var{n}
8968 @opindex falign-loops
8969 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8970 like @option{-falign-functions}. If the loops are
8971 executed many times, this makes up for any execution of the dummy
8974 @option{-fno-align-loops} and @option{-falign-loops=1} are
8975 equivalent and mean that loops are not aligned.
8977 If @var{n} is not specified or is zero, use a machine-dependent default.
8979 Enabled at levels @option{-O2}, @option{-O3}.
8982 @itemx -falign-jumps=@var{n}
8983 @opindex falign-jumps
8984 Align branch targets to a power-of-two boundary, for branch targets
8985 where the targets can only be reached by jumping, skipping up to @var{n}
8986 bytes like @option{-falign-functions}. In this case, no dummy operations
8989 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8990 equivalent and mean that loops are not aligned.
8992 If @var{n} is not specified or is zero, use a machine-dependent default.
8994 Enabled at levels @option{-O2}, @option{-O3}.
8996 @item -funit-at-a-time
8997 @opindex funit-at-a-time
8998 This option is left for compatibility reasons. @option{-funit-at-a-time}
8999 has no effect, while @option{-fno-unit-at-a-time} implies
9000 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9004 @item -fno-toplevel-reorder
9005 @opindex fno-toplevel-reorder
9006 Do not reorder top-level functions, variables, and @code{asm}
9007 statements. Output them in the same order that they appear in the
9008 input file. When this option is used, unreferenced static variables
9009 are not removed. This option is intended to support existing code
9010 that relies on a particular ordering. For new code, it is better to
9011 use attributes when possible.
9013 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9014 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9019 Constructs webs as commonly used for register allocation purposes and assign
9020 each web individual pseudo register. This allows the register allocation pass
9021 to operate on pseudos directly, but also strengthens several other optimization
9022 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9023 however, make debugging impossible, since variables no longer stay in a
9026 Enabled by default with @option{-funroll-loops}.
9028 @item -fwhole-program
9029 @opindex fwhole-program
9030 Assume that the current compilation unit represents the whole program being
9031 compiled. All public functions and variables with the exception of @code{main}
9032 and those merged by attribute @code{externally_visible} become static functions
9033 and in effect are optimized more aggressively by interprocedural optimizers.
9035 This option should not be used in combination with @option{-flto}.
9036 Instead relying on a linker plugin should provide safer and more precise
9039 @item -flto[=@var{n}]
9041 This option runs the standard link-time optimizer. When invoked
9042 with source code, it generates GIMPLE (one of GCC's internal
9043 representations) and writes it to special ELF sections in the object
9044 file. When the object files are linked together, all the function
9045 bodies are read from these ELF sections and instantiated as if they
9046 had been part of the same translation unit.
9048 To use the link-time optimizer, @option{-flto} and optimization
9049 options should be specified at compile time and during the final link.
9050 It is recommended that you compile all the files participating in the
9051 same link with the same options and also specify those options at
9056 gcc -c -O2 -flto foo.c
9057 gcc -c -O2 -flto bar.c
9058 gcc -o myprog -flto -O2 foo.o bar.o
9061 The first two invocations to GCC save a bytecode representation
9062 of GIMPLE into special ELF sections inside @file{foo.o} and
9063 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9064 @file{foo.o} and @file{bar.o}, merges the two files into a single
9065 internal image, and compiles the result as usual. Since both
9066 @file{foo.o} and @file{bar.o} are merged into a single image, this
9067 causes all the interprocedural analyses and optimizations in GCC to
9068 work across the two files as if they were a single one. This means,
9069 for example, that the inliner is able to inline functions in
9070 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9072 Another (simpler) way to enable link-time optimization is:
9075 gcc -o myprog -flto -O2 foo.c bar.c
9078 The above generates bytecode for @file{foo.c} and @file{bar.c},
9079 merges them together into a single GIMPLE representation and optimizes
9080 them as usual to produce @file{myprog}.
9082 The only important thing to keep in mind is that to enable link-time
9083 optimizations you need to use the GCC driver to perform the link step.
9084 GCC then automatically performs link-time optimization if any of the
9085 objects involved were compiled with the @option{-flto} command-line option.
9087 should specify the optimization options to be used for link-time
9088 optimization though GCC tries to be clever at guessing an
9089 optimization level to use from the options used at compile time
9090 if you fail to specify one at link time. You can always override
9091 the automatic decision to do link-time optimization
9092 by passing @option{-fno-lto} to the link command.
9094 To make whole program optimization effective, it is necessary to make
9095 certain whole program assumptions. The compiler needs to know
9096 what functions and variables can be accessed by libraries and runtime
9097 outside of the link-time optimized unit. When supported by the linker,
9098 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9099 to the compiler about used and externally visible symbols. When
9100 the linker plugin is not available, @option{-fwhole-program} should be
9101 used to allow the compiler to make these assumptions, which leads
9102 to more aggressive optimization decisions.
9104 When @option{-fuse-linker-plugin} is not enabled, when a file is
9105 compiled with @option{-flto}, the generated object file is larger than
9106 a regular object file because it contains GIMPLE bytecodes and the usual
9107 final code (see @option{-ffat-lto-objects}. This means that
9108 object files with LTO information can be linked as normal object
9109 files; if @option{-fno-lto} is passed to the linker, no
9110 interprocedural optimizations are applied. Note that when
9111 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
9112 but you cannot perform a regular, non-LTO link on them.
9114 Additionally, the optimization flags used to compile individual files
9115 are not necessarily related to those used at link time. For instance,
9118 gcc -c -O0 -ffat-lto-objects -flto foo.c
9119 gcc -c -O0 -ffat-lto-objects -flto bar.c
9120 gcc -o myprog -O3 foo.o bar.o
9123 This produces individual object files with unoptimized assembler
9124 code, but the resulting binary @file{myprog} is optimized at
9125 @option{-O3}. If, instead, the final binary is generated with
9126 @option{-fno-lto}, then @file{myprog} is not optimized.
9128 When producing the final binary, GCC only
9129 applies link-time optimizations to those files that contain bytecode.
9130 Therefore, you can mix and match object files and libraries with
9131 GIMPLE bytecodes and final object code. GCC automatically selects
9132 which files to optimize in LTO mode and which files to link without
9135 There are some code generation flags preserved by GCC when
9136 generating bytecodes, as they need to be used during the final link
9137 stage. Generally options specified at link time override those
9138 specified at compile time.
9140 If you do not specify an optimization level option @option{-O} at
9141 link time, then GCC uses the highest optimization level
9142 used when compiling the object files.
9144 Currently, the following options and their settings are taken from
9145 the first object file that explicitly specifies them:
9146 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9147 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9148 and all the @option{-m} target flags.
9150 Certain ABI-changing flags are required to match in all compilation units,
9151 and trying to override this at link time with a conflicting value
9152 is ignored. This includes options such as @option{-freg-struct-return}
9153 and @option{-fpcc-struct-return}.
9155 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9156 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9157 are passed through to the link stage and merged conservatively for
9158 conflicting translation units. Specifically
9159 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9160 precedence; and for example @option{-ffp-contract=off} takes precedence
9161 over @option{-ffp-contract=fast}. You can override them at link time.
9163 If LTO encounters objects with C linkage declared with incompatible
9164 types in separate translation units to be linked together (undefined
9165 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9166 issued. The behavior is still undefined at run time. Similar
9167 diagnostics may be raised for other languages.
9169 Another feature of LTO is that it is possible to apply interprocedural
9170 optimizations on files written in different languages:
9175 gfortran -c -flto baz.f90
9176 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9179 Notice that the final link is done with @command{g++} to get the C++
9180 runtime libraries and @option{-lgfortran} is added to get the Fortran
9181 runtime libraries. In general, when mixing languages in LTO mode, you
9182 should use the same link command options as when mixing languages in a
9183 regular (non-LTO) compilation.
9185 If object files containing GIMPLE bytecode are stored in a library archive, say
9186 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9187 are using a linker with plugin support. To create static libraries suitable
9188 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9189 and @command{ranlib};
9190 to show the symbols of object files with GIMPLE bytecode, use
9191 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9192 and @command{nm} have been compiled with plugin support. At link time, use the the
9193 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9194 the LTO optimization process:
9197 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9200 With the linker plugin enabled, the linker extracts the needed
9201 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9202 to make them part of the aggregated GIMPLE image to be optimized.
9204 If you are not using a linker with plugin support and/or do not
9205 enable the linker plugin, then the objects inside @file{libfoo.a}
9206 are extracted and linked as usual, but they do not participate
9207 in the LTO optimization process. In order to make a static library suitable
9208 for both LTO optimization and usual linkage, compile its object files with
9209 @option{-flto} @option{-ffat-lto-objects}.
9211 Link-time optimizations do not require the presence of the whole program to
9212 operate. If the program does not require any symbols to be exported, it is
9213 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9214 the interprocedural optimizers to use more aggressive assumptions which may
9215 lead to improved optimization opportunities.
9216 Use of @option{-fwhole-program} is not needed when linker plugin is
9217 active (see @option{-fuse-linker-plugin}).
9219 The current implementation of LTO makes no
9220 attempt to generate bytecode that is portable between different
9221 types of hosts. The bytecode files are versioned and there is a
9222 strict version check, so bytecode files generated in one version of
9223 GCC do not work with an older or newer version of GCC.
9225 Link-time optimization does not work well with generation of debugging
9226 information. Combining @option{-flto} with
9227 @option{-g} is currently experimental and expected to produce unexpected
9230 If you specify the optional @var{n}, the optimization and code
9231 generation done at link time is executed in parallel using @var{n}
9232 parallel jobs by utilizing an installed @command{make} program. The
9233 environment variable @env{MAKE} may be used to override the program
9234 used. The default value for @var{n} is 1.
9236 You can also specify @option{-flto=jobserver} to use GNU make's
9237 job server mode to determine the number of parallel jobs. This
9238 is useful when the Makefile calling GCC is already executing in parallel.
9239 You must prepend a @samp{+} to the command recipe in the parent Makefile
9240 for this to work. This option likely only works if @env{MAKE} is
9243 @item -flto-partition=@var{alg}
9244 @opindex flto-partition
9245 Specify the partitioning algorithm used by the link-time optimizer.
9246 The value is either @samp{1to1} to specify a partitioning mirroring
9247 the original source files or @samp{balanced} to specify partitioning
9248 into equally sized chunks (whenever possible) or @samp{max} to create
9249 new partition for every symbol where possible. Specifying @samp{none}
9250 as an algorithm disables partitioning and streaming completely.
9251 The default value is @samp{balanced}. While @samp{1to1} can be used
9252 as an workaround for various code ordering issues, the @samp{max}
9253 partitioning is intended for internal testing only.
9254 The value @samp{one} specifies that exactly one partition should be
9255 used while the value @samp{none} bypasses partitioning and executes
9256 the link-time optimization step directly from the WPA phase.
9258 @item -flto-odr-type-merging
9259 @opindex flto-odr-type-merging
9260 Enable streaming of mangled types names of C++ types and their unification
9261 at link time. This increases size of LTO object files, but enables
9262 diagnostics about One Definition Rule violations.
9264 @item -flto-compression-level=@var{n}
9265 @opindex flto-compression-level
9266 This option specifies the level of compression used for intermediate
9267 language written to LTO object files, and is only meaningful in
9268 conjunction with LTO mode (@option{-flto}). Valid
9269 values are 0 (no compression) to 9 (maximum compression). Values
9270 outside this range are clamped to either 0 or 9. If the option is not
9271 given, a default balanced compression setting is used.
9273 @item -fuse-linker-plugin
9274 @opindex fuse-linker-plugin
9275 Enables the use of a linker plugin during link-time optimization. This
9276 option relies on plugin support in the linker, which is available in gold
9277 or in GNU ld 2.21 or newer.
9279 This option enables the extraction of object files with GIMPLE bytecode out
9280 of library archives. This improves the quality of optimization by exposing
9281 more code to the link-time optimizer. This information specifies what
9282 symbols can be accessed externally (by non-LTO object or during dynamic
9283 linking). Resulting code quality improvements on binaries (and shared
9284 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9285 See @option{-flto} for a description of the effect of this flag and how to
9288 This option is enabled by default when LTO support in GCC is enabled
9289 and GCC was configured for use with
9290 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9292 @item -ffat-lto-objects
9293 @opindex ffat-lto-objects
9294 Fat LTO objects are object files that contain both the intermediate language
9295 and the object code. This makes them usable for both LTO linking and normal
9296 linking. This option is effective only when compiling with @option{-flto}
9297 and is ignored at link time.
9299 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9300 requires the complete toolchain to be aware of LTO. It requires a linker with
9301 linker plugin support for basic functionality. Additionally,
9302 @command{nm}, @command{ar} and @command{ranlib}
9303 need to support linker plugins to allow a full-featured build environment
9304 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9305 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9306 to these tools. With non fat LTO makefiles need to be modified to use them.
9308 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9311 @item -fcompare-elim
9312 @opindex fcompare-elim
9313 After register allocation and post-register allocation instruction splitting,
9314 identify arithmetic instructions that compute processor flags similar to a
9315 comparison operation based on that arithmetic. If possible, eliminate the
9316 explicit comparison operation.
9318 This pass only applies to certain targets that cannot explicitly represent
9319 the comparison operation before register allocation is complete.
9321 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9323 @item -fcprop-registers
9324 @opindex fcprop-registers
9325 After register allocation and post-register allocation instruction splitting,
9326 perform a copy-propagation pass to try to reduce scheduling dependencies
9327 and occasionally eliminate the copy.
9329 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9331 @item -fprofile-correction
9332 @opindex fprofile-correction
9333 Profiles collected using an instrumented binary for multi-threaded programs may
9334 be inconsistent due to missed counter updates. When this option is specified,
9335 GCC uses heuristics to correct or smooth out such inconsistencies. By
9336 default, GCC emits an error message when an inconsistent profile is detected.
9339 @itemx -fprofile-use=@var{path}
9340 @opindex fprofile-use
9341 Enable profile feedback-directed optimizations,
9342 and the following optimizations
9343 which are generally profitable only with profile feedback available:
9344 @option{-fbranch-probabilities}, @option{-fvpt},
9345 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9346 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9348 Before you can use this option, you must first generate profiling information.
9349 @xref{Instrumentation Options}, for information about the
9350 @option{-fprofile-generate} option.
9352 By default, GCC emits an error message if the feedback profiles do not
9353 match the source code. This error can be turned into a warning by using
9354 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9357 If @var{path} is specified, GCC looks at the @var{path} to find
9358 the profile feedback data files. See @option{-fprofile-dir}.
9360 @item -fauto-profile
9361 @itemx -fauto-profile=@var{path}
9362 @opindex fauto-profile
9363 Enable sampling-based feedback-directed optimizations,
9364 and the following optimizations
9365 which are generally profitable only with profile feedback available:
9366 @option{-fbranch-probabilities}, @option{-fvpt},
9367 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9368 @option{-ftree-vectorize},
9369 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9370 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9371 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9373 @var{path} is the name of a file containing AutoFDO profile information.
9374 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9376 Producing an AutoFDO profile data file requires running your program
9377 with the @command{perf} utility on a supported GNU/Linux target system.
9378 For more information, see @uref{https://perf.wiki.kernel.org/}.
9382 perf record -e br_inst_retired:near_taken -b -o perf.data \
9386 Then use the @command{create_gcov} tool to convert the raw profile data
9387 to a format that can be used by GCC.@ You must also supply the
9388 unstripped binary for your program to this tool.
9389 See @uref{https://github.com/google/autofdo}.
9393 create_gcov --binary=your_program.unstripped --profile=perf.data \
9398 The following options control compiler behavior regarding floating-point
9399 arithmetic. These options trade off between speed and
9400 correctness. All must be specifically enabled.
9404 @opindex ffloat-store
9405 Do not store floating-point variables in registers, and inhibit other
9406 options that might change whether a floating-point value is taken from a
9409 @cindex floating-point precision
9410 This option prevents undesirable excess precision on machines such as
9411 the 68000 where the floating registers (of the 68881) keep more
9412 precision than a @code{double} is supposed to have. Similarly for the
9413 x86 architecture. For most programs, the excess precision does only
9414 good, but a few programs rely on the precise definition of IEEE floating
9415 point. Use @option{-ffloat-store} for such programs, after modifying
9416 them to store all pertinent intermediate computations into variables.
9418 @item -fexcess-precision=@var{style}
9419 @opindex fexcess-precision
9420 This option allows further control over excess precision on machines
9421 where floating-point operations occur in a format with more precision or
9422 range than the IEEE standard and interchange floating-point types. By
9423 default, @option{-fexcess-precision=fast} is in effect; this means that
9424 operations may be carried out in a wider precision than the types specified
9425 in the source if that would result in faster code, and it is unpredictable
9426 when rounding to the types specified in the source code takes place.
9427 When compiling C, if @option{-fexcess-precision=standard} is specified then
9428 excess precision follows the rules specified in ISO C99; in particular,
9429 both casts and assignments cause values to be rounded to their
9430 semantic types (whereas @option{-ffloat-store} only affects
9431 assignments). This option is enabled by default for C if a strict
9432 conformance option such as @option{-std=c99} is used.
9433 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
9434 regardless of whether a strict conformance option is used.
9437 @option{-fexcess-precision=standard} is not implemented for languages
9438 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
9439 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9440 semantics apply without excess precision, and in the latter, rounding
9445 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9446 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9447 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
9448 @option{-fexcess-precision=fast}.
9450 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9452 This option is not turned on by any @option{-O} option besides
9453 @option{-Ofast} since it can result in incorrect output for programs
9454 that depend on an exact implementation of IEEE or ISO rules/specifications
9455 for math functions. It may, however, yield faster code for programs
9456 that do not require the guarantees of these specifications.
9458 @item -fno-math-errno
9459 @opindex fno-math-errno
9460 Do not set @code{errno} after calling math functions that are executed
9461 with a single instruction, e.g., @code{sqrt}. A program that relies on
9462 IEEE exceptions for math error handling may want to use this flag
9463 for speed while maintaining IEEE arithmetic compatibility.
9465 This option is not turned on by any @option{-O} option since
9466 it can result in incorrect output for programs that depend on
9467 an exact implementation of IEEE or ISO rules/specifications for
9468 math functions. It may, however, yield faster code for programs
9469 that do not require the guarantees of these specifications.
9471 The default is @option{-fmath-errno}.
9473 On Darwin systems, the math library never sets @code{errno}. There is
9474 therefore no reason for the compiler to consider the possibility that
9475 it might, and @option{-fno-math-errno} is the default.
9477 @item -funsafe-math-optimizations
9478 @opindex funsafe-math-optimizations
9480 Allow optimizations for floating-point arithmetic that (a) assume
9481 that arguments and results are valid and (b) may violate IEEE or
9482 ANSI standards. When used at link time, it may include libraries
9483 or startup files that change the default FPU control word or other
9484 similar optimizations.
9486 This option is not turned on by any @option{-O} option since
9487 it can result in incorrect output for programs that depend on
9488 an exact implementation of IEEE or ISO rules/specifications for
9489 math functions. It may, however, yield faster code for programs
9490 that do not require the guarantees of these specifications.
9491 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9492 @option{-fassociative-math} and @option{-freciprocal-math}.
9494 The default is @option{-fno-unsafe-math-optimizations}.
9496 @item -fassociative-math
9497 @opindex fassociative-math
9499 Allow re-association of operands in series of floating-point operations.
9500 This violates the ISO C and C++ language standard by possibly changing
9501 computation result. NOTE: re-ordering may change the sign of zero as
9502 well as ignore NaNs and inhibit or create underflow or overflow (and
9503 thus cannot be used on code that relies on rounding behavior like
9504 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9505 and thus may not be used when ordered comparisons are required.
9506 This option requires that both @option{-fno-signed-zeros} and
9507 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9508 much sense with @option{-frounding-math}. For Fortran the option
9509 is automatically enabled when both @option{-fno-signed-zeros} and
9510 @option{-fno-trapping-math} are in effect.
9512 The default is @option{-fno-associative-math}.
9514 @item -freciprocal-math
9515 @opindex freciprocal-math
9517 Allow the reciprocal of a value to be used instead of dividing by
9518 the value if this enables optimizations. For example @code{x / y}
9519 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9520 is subject to common subexpression elimination. Note that this loses
9521 precision and increases the number of flops operating on the value.
9523 The default is @option{-fno-reciprocal-math}.
9525 @item -ffinite-math-only
9526 @opindex ffinite-math-only
9527 Allow optimizations for floating-point arithmetic that assume
9528 that arguments and results are not NaNs or +-Infs.
9530 This option is not turned on by any @option{-O} option since
9531 it can result in incorrect output for programs that depend on
9532 an exact implementation of IEEE or ISO rules/specifications for
9533 math functions. It may, however, yield faster code for programs
9534 that do not require the guarantees of these specifications.
9536 The default is @option{-fno-finite-math-only}.
9538 @item -fno-signed-zeros
9539 @opindex fno-signed-zeros
9540 Allow optimizations for floating-point arithmetic that ignore the
9541 signedness of zero. IEEE arithmetic specifies the behavior of
9542 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9543 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9544 This option implies that the sign of a zero result isn't significant.
9546 The default is @option{-fsigned-zeros}.
9548 @item -fno-trapping-math
9549 @opindex fno-trapping-math
9550 Compile code assuming that floating-point operations cannot generate
9551 user-visible traps. These traps include division by zero, overflow,
9552 underflow, inexact result and invalid operation. This option requires
9553 that @option{-fno-signaling-nans} be in effect. Setting this option may
9554 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9556 This option should never be turned on by any @option{-O} option since
9557 it can result in incorrect output for programs that depend on
9558 an exact implementation of IEEE or ISO rules/specifications for
9561 The default is @option{-ftrapping-math}.
9563 @item -frounding-math
9564 @opindex frounding-math
9565 Disable transformations and optimizations that assume default floating-point
9566 rounding behavior. This is round-to-zero for all floating point
9567 to integer conversions, and round-to-nearest for all other arithmetic
9568 truncations. This option should be specified for programs that change
9569 the FP rounding mode dynamically, or that may be executed with a
9570 non-default rounding mode. This option disables constant folding of
9571 floating-point expressions at compile time (which may be affected by
9572 rounding mode) and arithmetic transformations that are unsafe in the
9573 presence of sign-dependent rounding modes.
9575 The default is @option{-fno-rounding-math}.
9577 This option is experimental and does not currently guarantee to
9578 disable all GCC optimizations that are affected by rounding mode.
9579 Future versions of GCC may provide finer control of this setting
9580 using C99's @code{FENV_ACCESS} pragma. This command-line option
9581 will be used to specify the default state for @code{FENV_ACCESS}.
9583 @item -fsignaling-nans
9584 @opindex fsignaling-nans
9585 Compile code assuming that IEEE signaling NaNs may generate user-visible
9586 traps during floating-point operations. Setting this option disables
9587 optimizations that may change the number of exceptions visible with
9588 signaling NaNs. This option implies @option{-ftrapping-math}.
9590 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9593 The default is @option{-fno-signaling-nans}.
9595 This option is experimental and does not currently guarantee to
9596 disable all GCC optimizations that affect signaling NaN behavior.
9598 @item -fno-fp-int-builtin-inexact
9599 @opindex fno-fp-int-builtin-inexact
9600 Do not allow the built-in functions @code{ceil}, @code{floor},
9601 @code{round} and @code{trunc}, and their @code{float} and @code{long
9602 double} variants, to generate code that raises the ``inexact''
9603 floating-point exception for noninteger arguments. ISO C99 and C11
9604 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9605 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9608 The default is @option{-ffp-int-builtin-inexact}, allowing the
9609 exception to be raised. This option does nothing unless
9610 @option{-ftrapping-math} is in effect.
9612 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9613 generate a call to a library function then the ``inexact'' exception
9614 may be raised if the library implementation does not follow TS 18661.
9616 @item -fsingle-precision-constant
9617 @opindex fsingle-precision-constant
9618 Treat floating-point constants as single precision instead of
9619 implicitly converting them to double-precision constants.
9621 @item -fcx-limited-range
9622 @opindex fcx-limited-range
9623 When enabled, this option states that a range reduction step is not
9624 needed when performing complex division. Also, there is no checking
9625 whether the result of a complex multiplication or division is @code{NaN
9626 + I*NaN}, with an attempt to rescue the situation in that case. The
9627 default is @option{-fno-cx-limited-range}, but is enabled by
9628 @option{-ffast-math}.
9630 This option controls the default setting of the ISO C99
9631 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9634 @item -fcx-fortran-rules
9635 @opindex fcx-fortran-rules
9636 Complex multiplication and division follow Fortran rules. Range
9637 reduction is done as part of complex division, but there is no checking
9638 whether the result of a complex multiplication or division is @code{NaN
9639 + I*NaN}, with an attempt to rescue the situation in that case.
9641 The default is @option{-fno-cx-fortran-rules}.
9645 The following options control optimizations that may improve
9646 performance, but are not enabled by any @option{-O} options. This
9647 section includes experimental options that may produce broken code.
9650 @item -fbranch-probabilities
9651 @opindex fbranch-probabilities
9652 After running a program compiled with @option{-fprofile-arcs}
9653 (@pxref{Instrumentation Options}),
9654 you can compile it a second time using
9655 @option{-fbranch-probabilities}, to improve optimizations based on
9656 the number of times each branch was taken. When a program
9657 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9658 counts to a file called @file{@var{sourcename}.gcda} for each source
9659 file. The information in this data file is very dependent on the
9660 structure of the generated code, so you must use the same source code
9661 and the same optimization options for both compilations.
9663 With @option{-fbranch-probabilities}, GCC puts a
9664 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9665 These can be used to improve optimization. Currently, they are only
9666 used in one place: in @file{reorg.c}, instead of guessing which path a
9667 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9668 exactly determine which path is taken more often.
9670 @item -fprofile-values
9671 @opindex fprofile-values
9672 If combined with @option{-fprofile-arcs}, it adds code so that some
9673 data about values of expressions in the program is gathered.
9675 With @option{-fbranch-probabilities}, it reads back the data gathered
9676 from profiling values of expressions for usage in optimizations.
9678 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9680 @item -fprofile-reorder-functions
9681 @opindex fprofile-reorder-functions
9682 Function reordering based on profile instrumentation collects
9683 first time of execution of a function and orders these functions
9686 Enabled with @option{-fprofile-use}.
9690 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9691 to add code to gather information about values of expressions.
9693 With @option{-fbranch-probabilities}, it reads back the data gathered
9694 and actually performs the optimizations based on them.
9695 Currently the optimizations include specialization of division operations
9696 using the knowledge about the value of the denominator.
9698 @item -frename-registers
9699 @opindex frename-registers
9700 Attempt to avoid false dependencies in scheduled code by making use
9701 of registers left over after register allocation. This optimization
9702 most benefits processors with lots of registers. Depending on the
9703 debug information format adopted by the target, however, it can
9704 make debugging impossible, since variables no longer stay in
9705 a ``home register''.
9707 Enabled by default with @option{-funroll-loops}.
9709 @item -fschedule-fusion
9710 @opindex fschedule-fusion
9711 Performs a target dependent pass over the instruction stream to schedule
9712 instructions of same type together because target machine can execute them
9713 more efficiently if they are adjacent to each other in the instruction flow.
9715 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9719 Perform tail duplication to enlarge superblock size. This transformation
9720 simplifies the control flow of the function allowing other optimizations to do
9723 Enabled with @option{-fprofile-use}.
9725 @item -funroll-loops
9726 @opindex funroll-loops
9727 Unroll loops whose number of iterations can be determined at compile time or
9728 upon entry to the loop. @option{-funroll-loops} implies
9729 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9730 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9731 a small constant number of iterations). This option makes code larger, and may
9732 or may not make it run faster.
9734 Enabled with @option{-fprofile-use}.
9736 @item -funroll-all-loops
9737 @opindex funroll-all-loops
9738 Unroll all loops, even if their number of iterations is uncertain when
9739 the loop is entered. This usually makes programs run more slowly.
9740 @option{-funroll-all-loops} implies the same options as
9741 @option{-funroll-loops}.
9744 @opindex fpeel-loops
9745 Peels loops for which there is enough information that they do not
9746 roll much (from profile feedback or static analysis). It also turns on
9747 complete loop peeling (i.e.@: complete removal of loops with small constant
9748 number of iterations).
9750 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9752 @item -fmove-loop-invariants
9753 @opindex fmove-loop-invariants
9754 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9755 at level @option{-O1}
9758 @opindex fsplit-loops
9759 Split a loop into two if it contains a condition that's always true
9760 for one side of the iteration space and false for the other.
9762 @item -funswitch-loops
9763 @opindex funswitch-loops
9764 Move branches with loop invariant conditions out of the loop, with duplicates
9765 of the loop on both branches (modified according to result of the condition).
9767 @item -ffunction-sections
9768 @itemx -fdata-sections
9769 @opindex ffunction-sections
9770 @opindex fdata-sections
9771 Place each function or data item into its own section in the output
9772 file if the target supports arbitrary sections. The name of the
9773 function or the name of the data item determines the section's name
9776 Use these options on systems where the linker can perform optimizations to
9777 improve locality of reference in the instruction space. Most systems using the
9778 ELF object format have linkers with such optimizations. On AIX, the linker
9779 rearranges sections (CSECTs) based on the call graph. The performance impact
9782 Together with a linker garbage collection (linker @option{--gc-sections}
9783 option) these options may lead to smaller statically-linked executables (after
9786 On ELF/DWARF systems these options do not degenerate the quality of the debug
9787 information. There could be issues with other object files/debug info formats.
9789 Only use these options when there are significant benefits from doing so. When
9790 you specify these options, the assembler and linker create larger object and
9791 executable files and are also slower. These options affect code generation.
9792 They prevent optimizations by the compiler and assembler using relative
9793 locations inside a translation unit since the locations are unknown until
9794 link time. An example of such an optimization is relaxing calls to short call
9797 @item -fbranch-target-load-optimize
9798 @opindex fbranch-target-load-optimize
9799 Perform branch target register load optimization before prologue / epilogue
9801 The use of target registers can typically be exposed only during reload,
9802 thus hoisting loads out of loops and doing inter-block scheduling needs
9803 a separate optimization pass.
9805 @item -fbranch-target-load-optimize2
9806 @opindex fbranch-target-load-optimize2
9807 Perform branch target register load optimization after prologue / epilogue
9810 @item -fbtr-bb-exclusive
9811 @opindex fbtr-bb-exclusive
9812 When performing branch target register load optimization, don't reuse
9813 branch target registers within any basic block.
9816 @opindex fstdarg-opt
9817 Optimize the prologue of variadic argument functions with respect to usage of
9820 @item -fsection-anchors
9821 @opindex fsection-anchors
9822 Try to reduce the number of symbolic address calculations by using
9823 shared ``anchor'' symbols to address nearby objects. This transformation
9824 can help to reduce the number of GOT entries and GOT accesses on some
9827 For example, the implementation of the following function @code{foo}:
9831 int foo (void) @{ return a + b + c; @}
9835 usually calculates the addresses of all three variables, but if you
9836 compile it with @option{-fsection-anchors}, it accesses the variables
9837 from a common anchor point instead. The effect is similar to the
9838 following pseudocode (which isn't valid C):
9843 register int *xr = &x;
9844 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9848 Not all targets support this option.
9850 @item --param @var{name}=@var{value}
9852 In some places, GCC uses various constants to control the amount of
9853 optimization that is done. For example, GCC does not inline functions
9854 that contain more than a certain number of instructions. You can
9855 control some of these constants on the command line using the
9856 @option{--param} option.
9858 The names of specific parameters, and the meaning of the values, are
9859 tied to the internals of the compiler, and are subject to change
9860 without notice in future releases.
9862 In each case, the @var{value} is an integer. The allowable choices for
9866 @item predictable-branch-outcome
9867 When branch is predicted to be taken with probability lower than this threshold
9868 (in percent), then it is considered well predictable. The default is 10.
9870 @item max-rtl-if-conversion-insns
9871 RTL if-conversion tries to remove conditional branches around a block and
9872 replace them with conditionally executed instructions. This parameter
9873 gives the maximum number of instructions in a block which should be
9874 considered for if-conversion. The default is 10, though the compiler will
9875 also use other heuristics to decide whether if-conversion is likely to be
9878 @item max-rtl-if-conversion-predictable-cost
9879 @item max-rtl-if-conversion-unpredictable-cost
9880 RTL if-conversion will try to remove conditional branches around a block
9881 and replace them with conditionally executed instructions. These parameters
9882 give the maximum permissible cost for the sequence that would be generated
9883 by if-conversion depending on whether the branch is statically determined
9884 to be predictable or not. The units for this parameter are the same as
9885 those for the GCC internal seq_cost metric. The compiler will try to
9886 provide a reasonable default for this parameter using the BRANCH_COST
9889 @item max-crossjump-edges
9890 The maximum number of incoming edges to consider for cross-jumping.
9891 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9892 the number of edges incoming to each block. Increasing values mean
9893 more aggressive optimization, making the compilation time increase with
9894 probably small improvement in executable size.
9896 @item min-crossjump-insns
9897 The minimum number of instructions that must be matched at the end
9898 of two blocks before cross-jumping is performed on them. This
9899 value is ignored in the case where all instructions in the block being
9900 cross-jumped from are matched. The default value is 5.
9902 @item max-grow-copy-bb-insns
9903 The maximum code size expansion factor when copying basic blocks
9904 instead of jumping. The expansion is relative to a jump instruction.
9905 The default value is 8.
9907 @item max-goto-duplication-insns
9908 The maximum number of instructions to duplicate to a block that jumps
9909 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9910 passes, GCC factors computed gotos early in the compilation process,
9911 and unfactors them as late as possible. Only computed jumps at the
9912 end of a basic blocks with no more than max-goto-duplication-insns are
9913 unfactored. The default value is 8.
9915 @item max-delay-slot-insn-search
9916 The maximum number of instructions to consider when looking for an
9917 instruction to fill a delay slot. If more than this arbitrary number of
9918 instructions are searched, the time savings from filling the delay slot
9919 are minimal, so stop searching. Increasing values mean more
9920 aggressive optimization, making the compilation time increase with probably
9921 small improvement in execution time.
9923 @item max-delay-slot-live-search
9924 When trying to fill delay slots, the maximum number of instructions to
9925 consider when searching for a block with valid live register
9926 information. Increasing this arbitrarily chosen value means more
9927 aggressive optimization, increasing the compilation time. This parameter
9928 should be removed when the delay slot code is rewritten to maintain the
9931 @item max-gcse-memory
9932 The approximate maximum amount of memory that can be allocated in
9933 order to perform the global common subexpression elimination
9934 optimization. If more memory than specified is required, the
9935 optimization is not done.
9937 @item max-gcse-insertion-ratio
9938 If the ratio of expression insertions to deletions is larger than this value
9939 for any expression, then RTL PRE inserts or removes the expression and thus
9940 leaves partially redundant computations in the instruction stream. The default value is 20.
9942 @item max-pending-list-length
9943 The maximum number of pending dependencies scheduling allows
9944 before flushing the current state and starting over. Large functions
9945 with few branches or calls can create excessively large lists which
9946 needlessly consume memory and resources.
9948 @item max-modulo-backtrack-attempts
9949 The maximum number of backtrack attempts the scheduler should make
9950 when modulo scheduling a loop. Larger values can exponentially increase
9953 @item max-inline-insns-single
9954 Several parameters control the tree inliner used in GCC@.
9955 This number sets the maximum number of instructions (counted in GCC's
9956 internal representation) in a single function that the tree inliner
9957 considers for inlining. This only affects functions declared
9958 inline and methods implemented in a class declaration (C++).
9959 The default value is 400.
9961 @item max-inline-insns-auto
9962 When you use @option{-finline-functions} (included in @option{-O3}),
9963 a lot of functions that would otherwise not be considered for inlining
9964 by the compiler are investigated. To those functions, a different
9965 (more restrictive) limit compared to functions declared inline can
9967 The default value is 40.
9969 @item inline-min-speedup
9970 When estimated performance improvement of caller + callee runtime exceeds this
9971 threshold (in percent), the function can be inlined regardless of the limit on
9972 @option{--param max-inline-insns-single} and @option{--param
9973 max-inline-insns-auto}.
9975 @item large-function-insns
9976 The limit specifying really large functions. For functions larger than this
9977 limit after inlining, inlining is constrained by
9978 @option{--param large-function-growth}. This parameter is useful primarily
9979 to avoid extreme compilation time caused by non-linear algorithms used by the
9981 The default value is 2700.
9983 @item large-function-growth
9984 Specifies maximal growth of large function caused by inlining in percents.
9985 The default value is 100 which limits large function growth to 2.0 times
9988 @item large-unit-insns
9989 The limit specifying large translation unit. Growth caused by inlining of
9990 units larger than this limit is limited by @option{--param inline-unit-growth}.
9991 For small units this might be too tight.
9992 For example, consider a unit consisting of function A
9993 that is inline and B that just calls A three times. If B is small relative to
9994 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9995 large units consisting of small inlineable functions, however, the overall unit
9996 growth limit is needed to avoid exponential explosion of code size. Thus for
9997 smaller units, the size is increased to @option{--param large-unit-insns}
9998 before applying @option{--param inline-unit-growth}. The default is 10000.
10000 @item inline-unit-growth
10001 Specifies maximal overall growth of the compilation unit caused by inlining.
10002 The default value is 20 which limits unit growth to 1.2 times the original
10003 size. Cold functions (either marked cold via an attribute or by profile
10004 feedback) are not accounted into the unit size.
10006 @item ipcp-unit-growth
10007 Specifies maximal overall growth of the compilation unit caused by
10008 interprocedural constant propagation. The default value is 10 which limits
10009 unit growth to 1.1 times the original size.
10011 @item large-stack-frame
10012 The limit specifying large stack frames. While inlining the algorithm is trying
10013 to not grow past this limit too much. The default value is 256 bytes.
10015 @item large-stack-frame-growth
10016 Specifies maximal growth of large stack frames caused by inlining in percents.
10017 The default value is 1000 which limits large stack frame growth to 11 times
10020 @item max-inline-insns-recursive
10021 @itemx max-inline-insns-recursive-auto
10022 Specifies the maximum number of instructions an out-of-line copy of a
10023 self-recursive inline
10024 function can grow into by performing recursive inlining.
10026 @option{--param max-inline-insns-recursive} applies to functions
10028 For functions not declared inline, recursive inlining
10029 happens only when @option{-finline-functions} (included in @option{-O3}) is
10030 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10031 default value is 450.
10033 @item max-inline-recursive-depth
10034 @itemx max-inline-recursive-depth-auto
10035 Specifies the maximum recursion depth used for recursive inlining.
10037 @option{--param max-inline-recursive-depth} applies to functions
10038 declared inline. For functions not declared inline, recursive inlining
10039 happens only when @option{-finline-functions} (included in @option{-O3}) is
10040 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10041 default value is 8.
10043 @item min-inline-recursive-probability
10044 Recursive inlining is profitable only for function having deep recursion
10045 in average and can hurt for function having little recursion depth by
10046 increasing the prologue size or complexity of function body to other
10049 When profile feedback is available (see @option{-fprofile-generate}) the actual
10050 recursion depth can be guessed from the probability that function recurses
10051 via a given call expression. This parameter limits inlining only to call
10052 expressions whose probability exceeds the given threshold (in percents).
10053 The default value is 10.
10055 @item early-inlining-insns
10056 Specify growth that the early inliner can make. In effect it increases
10057 the amount of inlining for code having a large abstraction penalty.
10058 The default value is 14.
10060 @item max-early-inliner-iterations
10061 Limit of iterations of the early inliner. This basically bounds
10062 the number of nested indirect calls the early inliner can resolve.
10063 Deeper chains are still handled by late inlining.
10065 @item comdat-sharing-probability
10066 Probability (in percent) that C++ inline function with comdat visibility
10067 are shared across multiple compilation units. The default value is 20.
10069 @item profile-func-internal-id
10070 A parameter to control whether to use function internal id in profile
10071 database lookup. If the value is 0, the compiler uses an id that
10072 is based on function assembler name and filename, which makes old profile
10073 data more tolerant to source changes such as function reordering etc.
10074 The default value is 0.
10076 @item min-vect-loop-bound
10077 The minimum number of iterations under which loops are not vectorized
10078 when @option{-ftree-vectorize} is used. The number of iterations after
10079 vectorization needs to be greater than the value specified by this option
10080 to allow vectorization. The default value is 0.
10082 @item gcse-cost-distance-ratio
10083 Scaling factor in calculation of maximum distance an expression
10084 can be moved by GCSE optimizations. This is currently supported only in the
10085 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10086 is with simple expressions, i.e., the expressions that have cost
10087 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10088 hoisting of simple expressions. The default value is 10.
10090 @item gcse-unrestricted-cost
10091 Cost, roughly measured as the cost of a single typical machine
10092 instruction, at which GCSE optimizations do not constrain
10093 the distance an expression can travel. This is currently
10094 supported only in the code hoisting pass. The lesser the cost,
10095 the more aggressive code hoisting is. Specifying 0
10096 allows all expressions to travel unrestricted distances.
10097 The default value is 3.
10099 @item max-hoist-depth
10100 The depth of search in the dominator tree for expressions to hoist.
10101 This is used to avoid quadratic behavior in hoisting algorithm.
10102 The value of 0 does not limit on the search, but may slow down compilation
10103 of huge functions. The default value is 30.
10105 @item max-tail-merge-comparisons
10106 The maximum amount of similar bbs to compare a bb with. This is used to
10107 avoid quadratic behavior in tree tail merging. The default value is 10.
10109 @item max-tail-merge-iterations
10110 The maximum amount of iterations of the pass over the function. This is used to
10111 limit compilation time in tree tail merging. The default value is 2.
10113 @item store-merging-allow-unaligned
10114 Allow the store merging pass to introduce unaligned stores if it is legal to
10115 do so. The default value is 1.
10117 @item max-stores-to-merge
10118 The maximum number of stores to attempt to merge into wider stores in the store
10119 merging pass. The minimum value is 2 and the default is 64.
10121 @item max-unrolled-insns
10122 The maximum number of instructions that a loop may have to be unrolled.
10123 If a loop is unrolled, this parameter also determines how many times
10124 the loop code is unrolled.
10126 @item max-average-unrolled-insns
10127 The maximum number of instructions biased by probabilities of their execution
10128 that a loop may have to be unrolled. If a loop is unrolled,
10129 this parameter also determines how many times the loop code is unrolled.
10131 @item max-unroll-times
10132 The maximum number of unrollings of a single loop.
10134 @item max-peeled-insns
10135 The maximum number of instructions that a loop may have to be peeled.
10136 If a loop is peeled, this parameter also determines how many times
10137 the loop code is peeled.
10139 @item max-peel-times
10140 The maximum number of peelings of a single loop.
10142 @item max-peel-branches
10143 The maximum number of branches on the hot path through the peeled sequence.
10145 @item max-completely-peeled-insns
10146 The maximum number of insns of a completely peeled loop.
10148 @item max-completely-peel-times
10149 The maximum number of iterations of a loop to be suitable for complete peeling.
10151 @item max-completely-peel-loop-nest-depth
10152 The maximum depth of a loop nest suitable for complete peeling.
10154 @item max-unswitch-insns
10155 The maximum number of insns of an unswitched loop.
10157 @item max-unswitch-level
10158 The maximum number of branches unswitched in a single loop.
10160 @item max-loop-headers-insns
10161 The maximum number of insns in loop header duplicated by the copy loop headers
10164 @item lim-expensive
10165 The minimum cost of an expensive expression in the loop invariant motion.
10167 @item iv-consider-all-candidates-bound
10168 Bound on number of candidates for induction variables, below which
10169 all candidates are considered for each use in induction variable
10170 optimizations. If there are more candidates than this,
10171 only the most relevant ones are considered to avoid quadratic time complexity.
10173 @item iv-max-considered-uses
10174 The induction variable optimizations give up on loops that contain more
10175 induction variable uses.
10177 @item iv-always-prune-cand-set-bound
10178 If the number of candidates in the set is smaller than this value,
10179 always try to remove unnecessary ivs from the set
10180 when adding a new one.
10182 @item avg-loop-niter
10183 Average number of iterations of a loop.
10185 @item dse-max-object-size
10186 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
10187 Larger values may result in larger compilation times.
10189 @item scev-max-expr-size
10190 Bound on size of expressions used in the scalar evolutions analyzer.
10191 Large expressions slow the analyzer.
10193 @item scev-max-expr-complexity
10194 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10195 Complex expressions slow the analyzer.
10197 @item max-tree-if-conversion-phi-args
10198 Maximum number of arguments in a PHI supported by TREE if conversion
10199 unless the loop is marked with simd pragma.
10201 @item vect-max-version-for-alignment-checks
10202 The maximum number of run-time checks that can be performed when
10203 doing loop versioning for alignment in the vectorizer.
10205 @item vect-max-version-for-alias-checks
10206 The maximum number of run-time checks that can be performed when
10207 doing loop versioning for alias in the vectorizer.
10209 @item vect-max-peeling-for-alignment
10210 The maximum number of loop peels to enhance access alignment
10211 for vectorizer. Value -1 means no limit.
10213 @item max-iterations-to-track
10214 The maximum number of iterations of a loop the brute-force algorithm
10215 for analysis of the number of iterations of the loop tries to evaluate.
10217 @item hot-bb-count-ws-permille
10218 A basic block profile count is considered hot if it contributes to
10219 the given permillage (i.e. 0...1000) of the entire profiled execution.
10221 @item hot-bb-frequency-fraction
10222 Select fraction of the entry block frequency of executions of basic block in
10223 function given basic block needs to have to be considered hot.
10225 @item max-predicted-iterations
10226 The maximum number of loop iterations we predict statically. This is useful
10227 in cases where a function contains a single loop with known bound and
10228 another loop with unknown bound.
10229 The known number of iterations is predicted correctly, while
10230 the unknown number of iterations average to roughly 10. This means that the
10231 loop without bounds appears artificially cold relative to the other one.
10233 @item builtin-expect-probability
10234 Control the probability of the expression having the specified value. This
10235 parameter takes a percentage (i.e. 0 ... 100) as input.
10236 The default probability of 90 is obtained empirically.
10238 @item align-threshold
10240 Select fraction of the maximal frequency of executions of a basic block in
10241 a function to align the basic block.
10243 @item align-loop-iterations
10245 A loop expected to iterate at least the selected number of iterations is
10248 @item tracer-dynamic-coverage
10249 @itemx tracer-dynamic-coverage-feedback
10251 This value is used to limit superblock formation once the given percentage of
10252 executed instructions is covered. This limits unnecessary code size
10255 The @option{tracer-dynamic-coverage-feedback} parameter
10256 is used only when profile
10257 feedback is available. The real profiles (as opposed to statically estimated
10258 ones) are much less balanced allowing the threshold to be larger value.
10260 @item tracer-max-code-growth
10261 Stop tail duplication once code growth has reached given percentage. This is
10262 a rather artificial limit, as most of the duplicates are eliminated later in
10263 cross jumping, so it may be set to much higher values than is the desired code
10266 @item tracer-min-branch-ratio
10268 Stop reverse growth when the reverse probability of best edge is less than this
10269 threshold (in percent).
10271 @item tracer-min-branch-probability
10272 @itemx tracer-min-branch-probability-feedback
10274 Stop forward growth if the best edge has probability lower than this
10277 Similarly to @option{tracer-dynamic-coverage} two parameters are
10278 provided. @option{tracer-min-branch-probability-feedback} is used for
10279 compilation with profile feedback and @option{tracer-min-branch-probability}
10280 compilation without. The value for compilation with profile feedback
10281 needs to be more conservative (higher) in order to make tracer
10284 @item stack-clash-protection-guard-size
10285 Specify the size of the operating system provided stack guard as
10286 2 raised to @var{num} bytes. The default value is 12 (4096 bytes).
10287 Acceptable values are between 12 and 30. Higher values may reduce the
10288 number of explicit probes, but a value larger than the operating system
10289 provided guard will leave code vulnerable to stack clash style attacks.
10291 @item stack-clash-protection-probe-interval
10292 Stack clash protection involves probing stack space as it is allocated. This
10293 param controls the maximum distance between probes into the stack as 2 raised
10294 to @var{num} bytes. Acceptable values are between 10 and 16 and defaults to
10295 12. Higher values may reduce the number of explicit probes, but a value
10296 larger than the operating system provided guard will leave code vulnerable to
10297 stack clash style attacks.
10299 @item max-cse-path-length
10301 The maximum number of basic blocks on path that CSE considers.
10304 @item max-cse-insns
10305 The maximum number of instructions CSE processes before flushing.
10306 The default is 1000.
10308 @item ggc-min-expand
10310 GCC uses a garbage collector to manage its own memory allocation. This
10311 parameter specifies the minimum percentage by which the garbage
10312 collector's heap should be allowed to expand between collections.
10313 Tuning this may improve compilation speed; it has no effect on code
10316 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10317 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10318 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10319 GCC is not able to calculate RAM on a particular platform, the lower
10320 bound of 30% is used. Setting this parameter and
10321 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10322 every opportunity. This is extremely slow, but can be useful for
10325 @item ggc-min-heapsize
10327 Minimum size of the garbage collector's heap before it begins bothering
10328 to collect garbage. The first collection occurs after the heap expands
10329 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10330 tuning this may improve compilation speed, and has no effect on code
10333 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10334 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10335 with a lower bound of 4096 (four megabytes) and an upper bound of
10336 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10337 particular platform, the lower bound is used. Setting this parameter
10338 very large effectively disables garbage collection. Setting this
10339 parameter and @option{ggc-min-expand} to zero causes a full collection
10340 to occur at every opportunity.
10342 @item max-reload-search-insns
10343 The maximum number of instruction reload should look backward for equivalent
10344 register. Increasing values mean more aggressive optimization, making the
10345 compilation time increase with probably slightly better performance.
10346 The default value is 100.
10348 @item max-cselib-memory-locations
10349 The maximum number of memory locations cselib should take into account.
10350 Increasing values mean more aggressive optimization, making the compilation time
10351 increase with probably slightly better performance. The default value is 500.
10353 @item max-sched-ready-insns
10354 The maximum number of instructions ready to be issued the scheduler should
10355 consider at any given time during the first scheduling pass. Increasing
10356 values mean more thorough searches, making the compilation time increase
10357 with probably little benefit. The default value is 100.
10359 @item max-sched-region-blocks
10360 The maximum number of blocks in a region to be considered for
10361 interblock scheduling. The default value is 10.
10363 @item max-pipeline-region-blocks
10364 The maximum number of blocks in a region to be considered for
10365 pipelining in the selective scheduler. The default value is 15.
10367 @item max-sched-region-insns
10368 The maximum number of insns in a region to be considered for
10369 interblock scheduling. The default value is 100.
10371 @item max-pipeline-region-insns
10372 The maximum number of insns in a region to be considered for
10373 pipelining in the selective scheduler. The default value is 200.
10375 @item min-spec-prob
10376 The minimum probability (in percents) of reaching a source block
10377 for interblock speculative scheduling. The default value is 40.
10379 @item max-sched-extend-regions-iters
10380 The maximum number of iterations through CFG to extend regions.
10381 A value of 0 (the default) disables region extensions.
10383 @item max-sched-insn-conflict-delay
10384 The maximum conflict delay for an insn to be considered for speculative motion.
10385 The default value is 3.
10387 @item sched-spec-prob-cutoff
10388 The minimal probability of speculation success (in percents), so that
10389 speculative insns are scheduled.
10390 The default value is 40.
10392 @item sched-state-edge-prob-cutoff
10393 The minimum probability an edge must have for the scheduler to save its
10395 The default value is 10.
10397 @item sched-mem-true-dep-cost
10398 Minimal distance (in CPU cycles) between store and load targeting same
10399 memory locations. The default value is 1.
10401 @item selsched-max-lookahead
10402 The maximum size of the lookahead window of selective scheduling. It is a
10403 depth of search for available instructions.
10404 The default value is 50.
10406 @item selsched-max-sched-times
10407 The maximum number of times that an instruction is scheduled during
10408 selective scheduling. This is the limit on the number of iterations
10409 through which the instruction may be pipelined. The default value is 2.
10411 @item selsched-insns-to-rename
10412 The maximum number of best instructions in the ready list that are considered
10413 for renaming in the selective scheduler. The default value is 2.
10416 The minimum value of stage count that swing modulo scheduler
10417 generates. The default value is 2.
10419 @item max-last-value-rtl
10420 The maximum size measured as number of RTLs that can be recorded in an expression
10421 in combiner for a pseudo register as last known value of that register. The default
10424 @item max-combine-insns
10425 The maximum number of instructions the RTL combiner tries to combine.
10426 The default value is 2 at @option{-Og} and 4 otherwise.
10428 @item integer-share-limit
10429 Small integer constants can use a shared data structure, reducing the
10430 compiler's memory usage and increasing its speed. This sets the maximum
10431 value of a shared integer constant. The default value is 256.
10433 @item ssp-buffer-size
10434 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10435 protection when @option{-fstack-protection} is used.
10437 @item min-size-for-stack-sharing
10438 The minimum size of variables taking part in stack slot sharing when not
10439 optimizing. The default value is 32.
10441 @item max-jump-thread-duplication-stmts
10442 Maximum number of statements allowed in a block that needs to be
10443 duplicated when threading jumps.
10445 @item max-fields-for-field-sensitive
10446 Maximum number of fields in a structure treated in
10447 a field sensitive manner during pointer analysis. The default is zero
10448 for @option{-O0} and @option{-O1},
10449 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10451 @item prefetch-latency
10452 Estimate on average number of instructions that are executed before
10453 prefetch finishes. The distance prefetched ahead is proportional
10454 to this constant. Increasing this number may also lead to less
10455 streams being prefetched (see @option{simultaneous-prefetches}).
10457 @item simultaneous-prefetches
10458 Maximum number of prefetches that can run at the same time.
10460 @item l1-cache-line-size
10461 The size of cache line in L1 cache, in bytes.
10463 @item l1-cache-size
10464 The size of L1 cache, in kilobytes.
10466 @item l2-cache-size
10467 The size of L2 cache, in kilobytes.
10469 @item min-insn-to-prefetch-ratio
10470 The minimum ratio between the number of instructions and the
10471 number of prefetches to enable prefetching in a loop.
10473 @item prefetch-min-insn-to-mem-ratio
10474 The minimum ratio between the number of instructions and the
10475 number of memory references to enable prefetching in a loop.
10477 @item use-canonical-types
10478 Whether the compiler should use the ``canonical'' type system. By
10479 default, this should always be 1, which uses a more efficient internal
10480 mechanism for comparing types in C++ and Objective-C++. However, if
10481 bugs in the canonical type system are causing compilation failures,
10482 set this value to 0 to disable canonical types.
10484 @item switch-conversion-max-branch-ratio
10485 Switch initialization conversion refuses to create arrays that are
10486 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10487 branches in the switch.
10489 @item max-partial-antic-length
10490 Maximum length of the partial antic set computed during the tree
10491 partial redundancy elimination optimization (@option{-ftree-pre}) when
10492 optimizing at @option{-O3} and above. For some sorts of source code
10493 the enhanced partial redundancy elimination optimization can run away,
10494 consuming all of the memory available on the host machine. This
10495 parameter sets a limit on the length of the sets that are computed,
10496 which prevents the runaway behavior. Setting a value of 0 for
10497 this parameter allows an unlimited set length.
10499 @item sccvn-max-scc-size
10500 Maximum size of a strongly connected component (SCC) during SCCVN
10501 processing. If this limit is hit, SCCVN processing for the whole
10502 function is not done and optimizations depending on it are
10503 disabled. The default maximum SCC size is 10000.
10505 @item sccvn-max-alias-queries-per-access
10506 Maximum number of alias-oracle queries we perform when looking for
10507 redundancies for loads and stores. If this limit is hit the search
10508 is aborted and the load or store is not considered redundant. The
10509 number of queries is algorithmically limited to the number of
10510 stores on all paths from the load to the function entry.
10511 The default maximum number of queries is 1000.
10513 @item ira-max-loops-num
10514 IRA uses regional register allocation by default. If a function
10515 contains more loops than the number given by this parameter, only at most
10516 the given number of the most frequently-executed loops form regions
10517 for regional register allocation. The default value of the
10520 @item ira-max-conflict-table-size
10521 Although IRA uses a sophisticated algorithm to compress the conflict
10522 table, the table can still require excessive amounts of memory for
10523 huge functions. If the conflict table for a function could be more
10524 than the size in MB given by this parameter, the register allocator
10525 instead uses a faster, simpler, and lower-quality
10526 algorithm that does not require building a pseudo-register conflict table.
10527 The default value of the parameter is 2000.
10529 @item ira-loop-reserved-regs
10530 IRA can be used to evaluate more accurate register pressure in loops
10531 for decisions to move loop invariants (see @option{-O3}). The number
10532 of available registers reserved for some other purposes is given
10533 by this parameter. The default value of the parameter is 2, which is
10534 the minimal number of registers needed by typical instructions.
10535 This value is the best found from numerous experiments.
10537 @item lra-inheritance-ebb-probability-cutoff
10538 LRA tries to reuse values reloaded in registers in subsequent insns.
10539 This optimization is called inheritance. EBB is used as a region to
10540 do this optimization. The parameter defines a minimal fall-through
10541 edge probability in percentage used to add BB to inheritance EBB in
10542 LRA. The default value of the parameter is 40. The value was chosen
10543 from numerous runs of SPEC2000 on x86-64.
10545 @item loop-invariant-max-bbs-in-loop
10546 Loop invariant motion can be very expensive, both in compilation time and
10547 in amount of needed compile-time memory, with very large loops. Loops
10548 with more basic blocks than this parameter won't have loop invariant
10549 motion optimization performed on them. The default value of the
10550 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10552 @item loop-max-datarefs-for-datadeps
10553 Building data dependencies is expensive for very large loops. This
10554 parameter limits the number of data references in loops that are
10555 considered for data dependence analysis. These large loops are no
10556 handled by the optimizations using loop data dependencies.
10557 The default value is 1000.
10559 @item max-vartrack-size
10560 Sets a maximum number of hash table slots to use during variable
10561 tracking dataflow analysis of any function. If this limit is exceeded
10562 with variable tracking at assignments enabled, analysis for that
10563 function is retried without it, after removing all debug insns from
10564 the function. If the limit is exceeded even without debug insns, var
10565 tracking analysis is completely disabled for the function. Setting
10566 the parameter to zero makes it unlimited.
10568 @item max-vartrack-expr-depth
10569 Sets a maximum number of recursion levels when attempting to map
10570 variable names or debug temporaries to value expressions. This trades
10571 compilation time for more complete debug information. If this is set too
10572 low, value expressions that are available and could be represented in
10573 debug information may end up not being used; setting this higher may
10574 enable the compiler to find more complex debug expressions, but compile
10575 time and memory use may grow. The default is 12.
10577 @item min-nondebug-insn-uid
10578 Use uids starting at this parameter for nondebug insns. The range below
10579 the parameter is reserved exclusively for debug insns created by
10580 @option{-fvar-tracking-assignments}, but debug insns may get
10581 (non-overlapping) uids above it if the reserved range is exhausted.
10583 @item ipa-sra-ptr-growth-factor
10584 IPA-SRA replaces a pointer to an aggregate with one or more new
10585 parameters only when their cumulative size is less or equal to
10586 @option{ipa-sra-ptr-growth-factor} times the size of the original
10589 @item sra-max-scalarization-size-Ospeed
10590 @item sra-max-scalarization-size-Osize
10591 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10592 replace scalar parts of aggregates with uses of independent scalar
10593 variables. These parameters control the maximum size, in storage units,
10594 of aggregate which is considered for replacement when compiling for
10596 (@option{sra-max-scalarization-size-Ospeed}) or size
10597 (@option{sra-max-scalarization-size-Osize}) respectively.
10599 @item tm-max-aggregate-size
10600 When making copies of thread-local variables in a transaction, this
10601 parameter specifies the size in bytes after which variables are
10602 saved with the logging functions as opposed to save/restore code
10603 sequence pairs. This option only applies when using
10606 @item graphite-max-nb-scop-params
10607 To avoid exponential effects in the Graphite loop transforms, the
10608 number of parameters in a Static Control Part (SCoP) is bounded. The
10609 default value is 10 parameters, a value of zero can be used to lift
10610 the bound. A variable whose value is unknown at compilation time and
10611 defined outside a SCoP is a parameter of the SCoP.
10613 @item loop-block-tile-size
10614 Loop blocking or strip mining transforms, enabled with
10615 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10616 loop in the loop nest by a given number of iterations. The strip
10617 length can be changed using the @option{loop-block-tile-size}
10618 parameter. The default value is 51 iterations.
10620 @item loop-unroll-jam-size
10621 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10622 default value is 4.
10624 @item loop-unroll-jam-depth
10625 Specify the dimension to be unrolled (counting from the most inner loop)
10626 for the @option{-floop-unroll-and-jam}. The default value is 2.
10628 @item ipa-cp-value-list-size
10629 IPA-CP attempts to track all possible values and types passed to a function's
10630 parameter in order to propagate them and perform devirtualization.
10631 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10632 stores per one formal parameter of a function.
10634 @item ipa-cp-eval-threshold
10635 IPA-CP calculates its own score of cloning profitability heuristics
10636 and performs those cloning opportunities with scores that exceed
10637 @option{ipa-cp-eval-threshold}.
10639 @item ipa-cp-recursion-penalty
10640 Percentage penalty the recursive functions will receive when they
10641 are evaluated for cloning.
10643 @item ipa-cp-single-call-penalty
10644 Percentage penalty functions containing a single call to another
10645 function will receive when they are evaluated for cloning.
10648 @item ipa-max-agg-items
10649 IPA-CP is also capable to propagate a number of scalar values passed
10650 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10651 number of such values per one parameter.
10653 @item ipa-cp-loop-hint-bonus
10654 When IPA-CP determines that a cloning candidate would make the number
10655 of iterations of a loop known, it adds a bonus of
10656 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10659 @item ipa-cp-array-index-hint-bonus
10660 When IPA-CP determines that a cloning candidate would make the index of
10661 an array access known, it adds a bonus of
10662 @option{ipa-cp-array-index-hint-bonus} to the profitability
10663 score of the candidate.
10665 @item ipa-max-aa-steps
10666 During its analysis of function bodies, IPA-CP employs alias analysis
10667 in order to track values pointed to by function parameters. In order
10668 not spend too much time analyzing huge functions, it gives up and
10669 consider all memory clobbered after examining
10670 @option{ipa-max-aa-steps} statements modifying memory.
10672 @item lto-partitions
10673 Specify desired number of partitions produced during WHOPR compilation.
10674 The number of partitions should exceed the number of CPUs used for compilation.
10675 The default value is 32.
10677 @item lto-min-partition
10678 Size of minimal partition for WHOPR (in estimated instructions).
10679 This prevents expenses of splitting very small programs into too many
10682 @item lto-max-partition
10683 Size of max partition for WHOPR (in estimated instructions).
10684 to provide an upper bound for individual size of partition.
10685 Meant to be used only with balanced partitioning.
10687 @item cxx-max-namespaces-for-diagnostic-help
10688 The maximum number of namespaces to consult for suggestions when C++
10689 name lookup fails for an identifier. The default is 1000.
10691 @item sink-frequency-threshold
10692 The maximum relative execution frequency (in percents) of the target block
10693 relative to a statement's original block to allow statement sinking of a
10694 statement. Larger numbers result in more aggressive statement sinking.
10695 The default value is 75. A small positive adjustment is applied for
10696 statements with memory operands as those are even more profitable so sink.
10698 @item max-stores-to-sink
10699 The maximum number of conditional store pairs that can be sunk. Set to 0
10700 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10701 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10703 @item allow-store-data-races
10704 Allow optimizers to introduce new data races on stores.
10705 Set to 1 to allow, otherwise to 0. This option is enabled by default
10706 at optimization level @option{-Ofast}.
10708 @item case-values-threshold
10709 The smallest number of different values for which it is best to use a
10710 jump-table instead of a tree of conditional branches. If the value is
10711 0, use the default for the machine. The default is 0.
10713 @item tree-reassoc-width
10714 Set the maximum number of instructions executed in parallel in
10715 reassociated tree. This parameter overrides target dependent
10716 heuristics used by default if has non zero value.
10718 @item sched-pressure-algorithm
10719 Choose between the two available implementations of
10720 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10721 and is the more likely to prevent instructions from being reordered.
10722 Algorithm 2 was designed to be a compromise between the relatively
10723 conservative approach taken by algorithm 1 and the rather aggressive
10724 approach taken by the default scheduler. It relies more heavily on
10725 having a regular register file and accurate register pressure classes.
10726 See @file{haifa-sched.c} in the GCC sources for more details.
10728 The default choice depends on the target.
10730 @item max-slsr-cand-scan
10731 Set the maximum number of existing candidates that are considered when
10732 seeking a basis for a new straight-line strength reduction candidate.
10735 Enable buffer overflow detection for global objects. This kind
10736 of protection is enabled by default if you are using
10737 @option{-fsanitize=address} option.
10738 To disable global objects protection use @option{--param asan-globals=0}.
10741 Enable buffer overflow detection for stack objects. This kind of
10742 protection is enabled by default when using @option{-fsanitize=address}.
10743 To disable stack protection use @option{--param asan-stack=0} option.
10745 @item asan-instrument-reads
10746 Enable buffer overflow detection for memory reads. This kind of
10747 protection is enabled by default when using @option{-fsanitize=address}.
10748 To disable memory reads protection use
10749 @option{--param asan-instrument-reads=0}.
10751 @item asan-instrument-writes
10752 Enable buffer overflow detection for memory writes. This kind of
10753 protection is enabled by default when using @option{-fsanitize=address}.
10754 To disable memory writes protection use
10755 @option{--param asan-instrument-writes=0} option.
10757 @item asan-memintrin
10758 Enable detection for built-in functions. This kind of protection
10759 is enabled by default when using @option{-fsanitize=address}.
10760 To disable built-in functions protection use
10761 @option{--param asan-memintrin=0}.
10763 @item asan-use-after-return
10764 Enable detection of use-after-return. This kind of protection
10765 is enabled by default when using the @option{-fsanitize=address} option.
10766 To disable it use @option{--param asan-use-after-return=0}.
10768 Note: By default the check is disabled at run time. To enable it,
10769 add @code{detect_stack_use_after_return=1} to the environment variable
10770 @env{ASAN_OPTIONS}.
10772 @item asan-instrumentation-with-call-threshold
10773 If number of memory accesses in function being instrumented
10774 is greater or equal to this number, use callbacks instead of inline checks.
10775 E.g. to disable inline code use
10776 @option{--param asan-instrumentation-with-call-threshold=0}.
10778 @item use-after-scope-direct-emission-threshold
10779 If the size of a local variable in bytes is smaller or equal to this
10780 number, directly poison (or unpoison) shadow memory instead of using
10781 run-time callbacks. The default value is 256.
10783 @item chkp-max-ctor-size
10784 Static constructors generated by Pointer Bounds Checker may become very
10785 large and significantly increase compile time at optimization level
10786 @option{-O1} and higher. This parameter is a maximum number of statements
10787 in a single generated constructor. Default value is 5000.
10789 @item max-fsm-thread-path-insns
10790 Maximum number of instructions to copy when duplicating blocks on a
10791 finite state automaton jump thread path. The default is 100.
10793 @item max-fsm-thread-length
10794 Maximum number of basic blocks on a finite state automaton jump thread
10795 path. The default is 10.
10797 @item max-fsm-thread-paths
10798 Maximum number of new jump thread paths to create for a finite state
10799 automaton. The default is 50.
10801 @item parloops-chunk-size
10802 Chunk size of omp schedule for loops parallelized by parloops. The default
10805 @item parloops-schedule
10806 Schedule type of omp schedule for loops parallelized by parloops (static,
10807 dynamic, guided, auto, runtime). The default is static.
10809 @item parloops-min-per-thread
10810 The minimum number of iterations per thread of an innermost parallelized
10811 loop for which the parallelized variant is prefered over the single threaded
10812 one. The default is 100. Note that for a parallelized loop nest the
10813 minimum number of iterations of the outermost loop per thread is two.
10815 @item max-ssa-name-query-depth
10816 Maximum depth of recursion when querying properties of SSA names in things
10817 like fold routines. One level of recursion corresponds to following a
10820 @item hsa-gen-debug-stores
10821 Enable emission of special debug stores within HSA kernels which are
10822 then read and reported by libgomp plugin. Generation of these stores
10823 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10826 @item max-speculative-devirt-maydefs
10827 The maximum number of may-defs we analyze when looking for a must-def
10828 specifying the dynamic type of an object that invokes a virtual call
10829 we may be able to devirtualize speculatively.
10831 @item max-vrp-switch-assertions
10832 The maximum number of assertions to add along the default edge of a switch
10833 statement during VRP. The default is 10.
10837 @node Instrumentation Options
10838 @section Program Instrumentation Options
10839 @cindex instrumentation options
10840 @cindex program instrumentation options
10841 @cindex run-time error checking options
10842 @cindex profiling options
10843 @cindex options, program instrumentation
10844 @cindex options, run-time error checking
10845 @cindex options, profiling
10847 GCC supports a number of command-line options that control adding
10848 run-time instrumentation to the code it normally generates.
10849 For example, one purpose of instrumentation is collect profiling
10850 statistics for use in finding program hot spots, code coverage
10851 analysis, or profile-guided optimizations.
10852 Another class of program instrumentation is adding run-time checking
10853 to detect programming errors like invalid pointer
10854 dereferences or out-of-bounds array accesses, as well as deliberately
10855 hostile attacks such as stack smashing or C++ vtable hijacking.
10856 There is also a general hook which can be used to implement other
10857 forms of tracing or function-level instrumentation for debug or
10858 program analysis purposes.
10861 @cindex @command{prof}
10864 Generate extra code to write profile information suitable for the
10865 analysis program @command{prof}. You must use this option when compiling
10866 the source files you want data about, and you must also use it when
10869 @cindex @command{gprof}
10872 Generate extra code to write profile information suitable for the
10873 analysis program @command{gprof}. You must use this option when compiling
10874 the source files you want data about, and you must also use it when
10877 @item -fprofile-arcs
10878 @opindex fprofile-arcs
10879 Add code so that program flow @dfn{arcs} are instrumented. During
10880 execution the program records how many times each branch and call is
10881 executed and how many times it is taken or returns. On targets that support
10882 constructors with priority support, profiling properly handles constructors,
10883 destructors and C++ constructors (and destructors) of classes which are used
10884 as a type of a global variable.
10887 program exits it saves this data to a file called
10888 @file{@var{auxname}.gcda} for each source file. The data may be used for
10889 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10890 test coverage analysis (@option{-ftest-coverage}). Each object file's
10891 @var{auxname} is generated from the name of the output file, if
10892 explicitly specified and it is not the final executable, otherwise it is
10893 the basename of the source file. In both cases any suffix is removed
10894 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10895 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10896 @xref{Cross-profiling}.
10898 @cindex @command{gcov}
10902 This option is used to compile and link code instrumented for coverage
10903 analysis. The option is a synonym for @option{-fprofile-arcs}
10904 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10905 linking). See the documentation for those options for more details.
10910 Compile the source files with @option{-fprofile-arcs} plus optimization
10911 and code generation options. For test coverage analysis, use the
10912 additional @option{-ftest-coverage} option. You do not need to profile
10913 every source file in a program.
10916 Compile the source files additionally with @option{-fprofile-abs-path}
10917 to create absolute path names in the @file{.gcno} files. This allows
10918 @command{gcov} to find the correct sources in projects where compilations
10919 occur with different working directories.
10922 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10923 (the latter implies the former).
10926 Run the program on a representative workload to generate the arc profile
10927 information. This may be repeated any number of times. You can run
10928 concurrent instances of your program, and provided that the file system
10929 supports locking, the data files will be correctly updated. Unless
10930 a strict ISO C dialect option is in effect, @code{fork} calls are
10931 detected and correctly handled without double counting.
10934 For profile-directed optimizations, compile the source files again with
10935 the same optimization and code generation options plus
10936 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10937 Control Optimization}).
10940 For test coverage analysis, use @command{gcov} to produce human readable
10941 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10942 @command{gcov} documentation for further information.
10946 With @option{-fprofile-arcs}, for each function of your program GCC
10947 creates a program flow graph, then finds a spanning tree for the graph.
10948 Only arcs that are not on the spanning tree have to be instrumented: the
10949 compiler adds code to count the number of times that these arcs are
10950 executed. When an arc is the only exit or only entrance to a block, the
10951 instrumentation code can be added to the block; otherwise, a new basic
10952 block must be created to hold the instrumentation code.
10955 @item -ftest-coverage
10956 @opindex ftest-coverage
10957 Produce a notes file that the @command{gcov} code-coverage utility
10958 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10959 show program coverage. Each source file's note file is called
10960 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10961 above for a description of @var{auxname} and instructions on how to
10962 generate test coverage data. Coverage data matches the source files
10963 more closely if you do not optimize.
10965 @item -fprofile-abs-path
10966 @opindex fprofile-abs-path
10967 Automatically convert relative source file names to absolute path names
10968 in the @file{.gcno} files. This allows @command{gcov} to find the correct
10969 sources in projects where compilations occur with different working
10972 @item -fprofile-dir=@var{path}
10973 @opindex fprofile-dir
10975 Set the directory to search for the profile data files in to @var{path}.
10976 This option affects only the profile data generated by
10977 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10978 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10979 and its related options. Both absolute and relative paths can be used.
10980 By default, GCC uses the current directory as @var{path}, thus the
10981 profile data file appears in the same directory as the object file.
10983 @item -fprofile-generate
10984 @itemx -fprofile-generate=@var{path}
10985 @opindex fprofile-generate
10987 Enable options usually used for instrumenting application to produce
10988 profile useful for later recompilation with profile feedback based
10989 optimization. You must use @option{-fprofile-generate} both when
10990 compiling and when linking your program.
10992 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10994 If @var{path} is specified, GCC looks at the @var{path} to find
10995 the profile feedback data files. See @option{-fprofile-dir}.
10997 To optimize the program based on the collected profile information, use
10998 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
11000 @item -fprofile-update=@var{method}
11001 @opindex fprofile-update
11003 Alter the update method for an application instrumented for profile
11004 feedback based optimization. The @var{method} argument should be one of
11005 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
11006 The first one is useful for single-threaded applications,
11007 while the second one prevents profile corruption by emitting thread-safe code.
11009 @strong{Warning:} When an application does not properly join all threads
11010 (or creates an detached thread), a profile file can be still corrupted.
11012 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
11013 when supported by a target, or to @samp{single} otherwise. The GCC driver
11014 automatically selects @samp{prefer-atomic} when @option{-pthread}
11015 is present in the command line.
11017 @item -fsanitize=address
11018 @opindex fsanitize=address
11019 Enable AddressSanitizer, a fast memory error detector.
11020 Memory access instructions are instrumented to detect
11021 out-of-bounds and use-after-free bugs.
11022 The option enables @option{-fsanitize-address-use-after-scope}.
11023 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
11024 more details. The run-time behavior can be influenced using the
11025 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
11026 the available options are shown at startup of the instrumented program. See
11027 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
11028 for a list of supported options.
11029 The option cannot be combined with @option{-fsanitize=thread}
11030 and/or @option{-fcheck-pointer-bounds}.
11032 @item -fsanitize=kernel-address
11033 @opindex fsanitize=kernel-address
11034 Enable AddressSanitizer for Linux kernel.
11035 See @uref{https://github.com/google/kasan/wiki} for more details.
11036 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11038 @item -fsanitize=thread
11039 @opindex fsanitize=thread
11040 Enable ThreadSanitizer, a fast data race detector.
11041 Memory access instructions are instrumented to detect
11042 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
11043 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
11044 environment variable; see
11045 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
11047 The option cannot be combined with @option{-fsanitize=address},
11048 @option{-fsanitize=leak} and/or @option{-fcheck-pointer-bounds}.
11050 Note that sanitized atomic builtins cannot throw exceptions when
11051 operating on invalid memory addresses with non-call exceptions
11052 (@option{-fnon-call-exceptions}).
11054 @item -fsanitize=leak
11055 @opindex fsanitize=leak
11056 Enable LeakSanitizer, a memory leak detector.
11057 This option only matters for linking of executables and
11058 the executable is linked against a library that overrides @code{malloc}
11059 and other allocator functions. See
11060 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
11061 details. The run-time behavior can be influenced using the
11062 @env{LSAN_OPTIONS} environment variable.
11063 The option cannot be combined with @option{-fsanitize=thread}.
11065 @item -fsanitize=undefined
11066 @opindex fsanitize=undefined
11067 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
11068 Various computations are instrumented to detect undefined behavior
11069 at runtime. Current suboptions are:
11073 @item -fsanitize=shift
11074 @opindex fsanitize=shift
11075 This option enables checking that the result of a shift operation is
11076 not undefined. Note that what exactly is considered undefined differs
11077 slightly between C and C++, as well as between ISO C90 and C99, etc.
11078 This option has two suboptions, @option{-fsanitize=shift-base} and
11079 @option{-fsanitize=shift-exponent}.
11081 @item -fsanitize=shift-exponent
11082 @opindex fsanitize=shift-exponent
11083 This option enables checking that the second argument of a shift operation
11084 is not negative and is smaller than the precision of the promoted first
11087 @item -fsanitize=shift-base
11088 @opindex fsanitize=shift-base
11089 If the second argument of a shift operation is within range, check that the
11090 result of a shift operation is not undefined. Note that what exactly is
11091 considered undefined differs slightly between C and C++, as well as between
11092 ISO C90 and C99, etc.
11094 @item -fsanitize=integer-divide-by-zero
11095 @opindex fsanitize=integer-divide-by-zero
11096 Detect integer division by zero as well as @code{INT_MIN / -1} division.
11098 @item -fsanitize=unreachable
11099 @opindex fsanitize=unreachable
11100 With this option, the compiler turns the @code{__builtin_unreachable}
11101 call into a diagnostics message call instead. When reaching the
11102 @code{__builtin_unreachable} call, the behavior is undefined.
11104 @item -fsanitize=vla-bound
11105 @opindex fsanitize=vla-bound
11106 This option instructs the compiler to check that the size of a variable
11107 length array is positive.
11109 @item -fsanitize=null
11110 @opindex fsanitize=null
11111 This option enables pointer checking. Particularly, the application
11112 built with this option turned on will issue an error message when it
11113 tries to dereference a NULL pointer, or if a reference (possibly an
11114 rvalue reference) is bound to a NULL pointer, or if a method is invoked
11115 on an object pointed by a NULL pointer.
11117 @item -fsanitize=return
11118 @opindex fsanitize=return
11119 This option enables return statement checking. Programs
11120 built with this option turned on will issue an error message
11121 when the end of a non-void function is reached without actually
11122 returning a value. This option works in C++ only.
11124 @item -fsanitize=signed-integer-overflow
11125 @opindex fsanitize=signed-integer-overflow
11126 This option enables signed integer overflow checking. We check that
11127 the result of @code{+}, @code{*}, and both unary and binary @code{-}
11128 does not overflow in the signed arithmetics. Note, integer promotion
11129 rules must be taken into account. That is, the following is not an
11132 signed char a = SCHAR_MAX;
11136 @item -fsanitize=bounds
11137 @opindex fsanitize=bounds
11138 This option enables instrumentation of array bounds. Various out of bounds
11139 accesses are detected. Flexible array members, flexible array member-like
11140 arrays, and initializers of variables with static storage are not instrumented.
11141 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11143 @item -fsanitize=bounds-strict
11144 @opindex fsanitize=bounds-strict
11145 This option enables strict instrumentation of array bounds. Most out of bounds
11146 accesses are detected, including flexible array members and flexible array
11147 member-like arrays. Initializers of variables with static storage are not
11148 instrumented. The option cannot be combined
11149 with @option{-fcheck-pointer-bounds}.
11151 @item -fsanitize=alignment
11152 @opindex fsanitize=alignment
11154 This option enables checking of alignment of pointers when they are
11155 dereferenced, or when a reference is bound to insufficiently aligned target,
11156 or when a method or constructor is invoked on insufficiently aligned object.
11158 @item -fsanitize=object-size
11159 @opindex fsanitize=object-size
11160 This option enables instrumentation of memory references using the
11161 @code{__builtin_object_size} function. Various out of bounds pointer
11162 accesses are detected.
11164 @item -fsanitize=float-divide-by-zero
11165 @opindex fsanitize=float-divide-by-zero
11166 Detect floating-point division by zero. Unlike other similar options,
11167 @option{-fsanitize=float-divide-by-zero} is not enabled by
11168 @option{-fsanitize=undefined}, since floating-point division by zero can
11169 be a legitimate way of obtaining infinities and NaNs.
11171 @item -fsanitize=float-cast-overflow
11172 @opindex fsanitize=float-cast-overflow
11173 This option enables floating-point type to integer conversion checking.
11174 We check that the result of the conversion does not overflow.
11175 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
11176 not enabled by @option{-fsanitize=undefined}.
11177 This option does not work well with @code{FE_INVALID} exceptions enabled.
11179 @item -fsanitize=nonnull-attribute
11180 @opindex fsanitize=nonnull-attribute
11182 This option enables instrumentation of calls, checking whether null values
11183 are not passed to arguments marked as requiring a non-null value by the
11184 @code{nonnull} function attribute.
11186 @item -fsanitize=returns-nonnull-attribute
11187 @opindex fsanitize=returns-nonnull-attribute
11189 This option enables instrumentation of return statements in functions
11190 marked with @code{returns_nonnull} function attribute, to detect returning
11191 of null values from such functions.
11193 @item -fsanitize=bool
11194 @opindex fsanitize=bool
11196 This option enables instrumentation of loads from bool. If a value other
11197 than 0/1 is loaded, a run-time error is issued.
11199 @item -fsanitize=enum
11200 @opindex fsanitize=enum
11202 This option enables instrumentation of loads from an enum type. If
11203 a value outside the range of values for the enum type is loaded,
11204 a run-time error is issued.
11206 @item -fsanitize=vptr
11207 @opindex fsanitize=vptr
11209 This option enables instrumentation of C++ member function calls, member
11210 accesses and some conversions between pointers to base and derived classes,
11211 to verify the referenced object has the correct dynamic type.
11213 @item -fsanitize=pointer-overflow
11214 @opindex fsanitize=pointer-overflow
11216 This option enables instrumentation of pointer arithmetics. If the pointer
11217 arithmetics overflows, a run-time error is issued.
11219 @item -fsanitize=builtin
11220 @opindex fsanitize=builtin
11222 This option enables instrumentation of arguments to selected builtin
11223 functions. If an invalid value is passed to such arguments, a run-time
11224 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
11225 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
11230 While @option{-ftrapv} causes traps for signed overflows to be emitted,
11231 @option{-fsanitize=undefined} gives a diagnostic message.
11232 This currently works only for the C family of languages.
11234 @item -fno-sanitize=all
11235 @opindex fno-sanitize=all
11237 This option disables all previously enabled sanitizers.
11238 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
11241 @item -fasan-shadow-offset=@var{number}
11242 @opindex fasan-shadow-offset
11243 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
11244 It is useful for experimenting with different shadow memory layouts in
11245 Kernel AddressSanitizer.
11247 @item -fsanitize-sections=@var{s1},@var{s2},...
11248 @opindex fsanitize-sections
11249 Sanitize global variables in selected user-defined sections. @var{si} may
11252 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
11253 @opindex fsanitize-recover
11254 @opindex fno-sanitize-recover
11255 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
11256 mentioned in comma-separated list of @var{opts}. Enabling this option
11257 for a sanitizer component causes it to attempt to continue
11258 running the program as if no error happened. This means multiple
11259 runtime errors can be reported in a single program run, and the exit
11260 code of the program may indicate success even when errors
11261 have been reported. The @option{-fno-sanitize-recover=} option
11262 can be used to alter
11263 this behavior: only the first detected error is reported
11264 and program then exits with a non-zero exit code.
11266 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
11267 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
11268 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
11269 @option{-fsanitize=bounds-strict},
11270 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
11271 For these sanitizers error recovery is turned on by default,
11272 except @option{-fsanitize=address}, for which this feature is experimental.
11273 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
11274 accepted, the former enables recovery for all sanitizers that support it,
11275 the latter disables recovery for all sanitizers that support it.
11277 Even if a recovery mode is turned on the compiler side, it needs to be also
11278 enabled on the runtime library side, otherwise the failures are still fatal.
11279 The runtime library defaults to @code{halt_on_error=0} for
11280 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
11281 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
11282 setting the @code{halt_on_error} flag in the corresponding environment variable.
11284 Syntax without an explicit @var{opts} parameter is deprecated. It is
11285 equivalent to specifying an @var{opts} list of:
11288 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
11291 @item -fsanitize-address-use-after-scope
11292 @opindex fsanitize-address-use-after-scope
11293 Enable sanitization of local variables to detect use-after-scope bugs.
11294 The option sets @option{-fstack-reuse} to @samp{none}.
11296 @item -fsanitize-undefined-trap-on-error
11297 @opindex fsanitize-undefined-trap-on-error
11298 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
11299 report undefined behavior using @code{__builtin_trap} rather than
11300 a @code{libubsan} library routine. The advantage of this is that the
11301 @code{libubsan} library is not needed and is not linked in, so this
11302 is usable even in freestanding environments.
11304 @item -fsanitize-coverage=trace-pc
11305 @opindex fsanitize-coverage=trace-pc
11306 Enable coverage-guided fuzzing code instrumentation.
11307 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
11309 @item -fsanitize-coverage=trace-cmp
11310 @opindex fsanitize-coverage=trace-cmp
11311 Enable dataflow guided fuzzing code instrumentation.
11312 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
11313 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
11314 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
11315 variable or @code{__sanitizer_cov_trace_const_cmp1},
11316 @code{__sanitizer_cov_trace_const_cmp2},
11317 @code{__sanitizer_cov_trace_const_cmp4} or
11318 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
11319 operand constant, @code{__sanitizer_cov_trace_cmpf} or
11320 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
11321 @code{__sanitizer_cov_trace_switch} for switch statements.
11323 @item -fbounds-check
11324 @opindex fbounds-check
11325 For front ends that support it, generate additional code to check that
11326 indices used to access arrays are within the declared range. This is
11327 currently only supported by the Fortran front end, where this option
11330 @item -fcheck-pointer-bounds
11331 @opindex fcheck-pointer-bounds
11332 @opindex fno-check-pointer-bounds
11333 @cindex Pointer Bounds Checker options
11334 Enable Pointer Bounds Checker instrumentation. Each memory reference
11335 is instrumented with checks of the pointer used for memory access against
11336 bounds associated with that pointer.
11339 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
11340 and @option{-mmpx} are required to enable this feature.
11341 MPX-based instrumentation requires
11342 a runtime library to enable MPX in hardware and handle bounds
11343 violation signals. By default when @option{-fcheck-pointer-bounds}
11344 and @option{-mmpx} options are used to link a program, the GCC driver
11345 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
11346 Bounds checking on calls to dynamic libraries requires a linker
11347 with @option{-z bndplt} support; if GCC was configured with a linker
11348 without support for this option (including the Gold linker and older
11349 versions of ld), a warning is given if you link with @option{-mmpx}
11350 without also specifying @option{-static}, since the overall effectiveness
11351 of the bounds checking protection is reduced.
11352 See also @option{-static-libmpxwrappers}.
11354 MPX-based instrumentation
11355 may be used for debugging and also may be included in production code
11356 to increase program security. Depending on usage, you may
11357 have different requirements for the runtime library. The current version
11358 of the MPX runtime library is more oriented for use as a debugging
11359 tool. MPX runtime library usage implies @option{-lpthread}. See
11360 also @option{-static-libmpx}. The runtime library behavior can be
11361 influenced using various @env{CHKP_RT_*} environment variables. See
11362 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
11365 Generated instrumentation may be controlled by various
11366 @option{-fchkp-*} options and by the @code{bnd_variable_size}
11367 structure field attribute (@pxref{Type Attributes}) and
11368 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
11369 (@pxref{Function Attributes}). GCC also provides a number of built-in
11370 functions for controlling the Pointer Bounds Checker. @xref{Pointer
11371 Bounds Checker builtins}, for more information.
11373 @item -fchkp-check-incomplete-type
11374 @opindex fchkp-check-incomplete-type
11375 @opindex fno-chkp-check-incomplete-type
11376 Generate pointer bounds checks for variables with incomplete type.
11377 Enabled by default.
11379 @item -fchkp-narrow-bounds
11380 @opindex fchkp-narrow-bounds
11381 @opindex fno-chkp-narrow-bounds
11382 Controls bounds used by Pointer Bounds Checker for pointers to object
11383 fields. If narrowing is enabled then field bounds are used. Otherwise
11384 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
11385 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
11387 @item -fchkp-first-field-has-own-bounds
11388 @opindex fchkp-first-field-has-own-bounds
11389 @opindex fno-chkp-first-field-has-own-bounds
11390 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
11391 first field in the structure. By default a pointer to the first field has
11392 the same bounds as a pointer to the whole structure.
11394 @item -fchkp-flexible-struct-trailing-arrays
11395 @opindex fchkp-flexible-struct-trailing-arrays
11396 @opindex fno-chkp-flexible-struct-trailing-arrays
11397 Forces Pointer Bounds Checker to treat all trailing arrays in structures as
11398 possibly flexible. By default only array fields with zero length or that are
11399 marked with attribute bnd_variable_size are treated as flexible.
11401 @item -fchkp-narrow-to-innermost-array
11402 @opindex fchkp-narrow-to-innermost-array
11403 @opindex fno-chkp-narrow-to-innermost-array
11404 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
11405 case of nested static array access. By default this option is disabled and
11406 bounds of the outermost array are used.
11408 @item -fchkp-optimize
11409 @opindex fchkp-optimize
11410 @opindex fno-chkp-optimize
11411 Enables Pointer Bounds Checker optimizations. Enabled by default at
11412 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11414 @item -fchkp-use-fast-string-functions
11415 @opindex fchkp-use-fast-string-functions
11416 @opindex fno-chkp-use-fast-string-functions
11417 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11418 by Pointer Bounds Checker. Disabled by default.
11420 @item -fchkp-use-nochk-string-functions
11421 @opindex fchkp-use-nochk-string-functions
11422 @opindex fno-chkp-use-nochk-string-functions
11423 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11424 by Pointer Bounds Checker. Disabled by default.
11426 @item -fchkp-use-static-bounds
11427 @opindex fchkp-use-static-bounds
11428 @opindex fno-chkp-use-static-bounds
11429 Allow Pointer Bounds Checker to generate static bounds holding
11430 bounds of static variables. Enabled by default.
11432 @item -fchkp-use-static-const-bounds
11433 @opindex fchkp-use-static-const-bounds
11434 @opindex fno-chkp-use-static-const-bounds
11435 Use statically-initialized bounds for constant bounds instead of
11436 generating them each time they are required. By default enabled when
11437 @option{-fchkp-use-static-bounds} is enabled.
11439 @item -fchkp-treat-zero-dynamic-size-as-infinite
11440 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11441 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11442 With this option, objects with incomplete type whose
11443 dynamically-obtained size is zero are treated as having infinite size
11444 instead by Pointer Bounds
11445 Checker. This option may be helpful if a program is linked with a library
11446 missing size information for some symbols. Disabled by default.
11448 @item -fchkp-check-read
11449 @opindex fchkp-check-read
11450 @opindex fno-chkp-check-read
11451 Instructs Pointer Bounds Checker to generate checks for all read
11452 accesses to memory. Enabled by default.
11454 @item -fchkp-check-write
11455 @opindex fchkp-check-write
11456 @opindex fno-chkp-check-write
11457 Instructs Pointer Bounds Checker to generate checks for all write
11458 accesses to memory. Enabled by default.
11460 @item -fchkp-store-bounds
11461 @opindex fchkp-store-bounds
11462 @opindex fno-chkp-store-bounds
11463 Instructs Pointer Bounds Checker to generate bounds stores for
11464 pointer writes. Enabled by default.
11466 @item -fchkp-instrument-calls
11467 @opindex fchkp-instrument-calls
11468 @opindex fno-chkp-instrument-calls
11469 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11470 Enabled by default.
11472 @item -fchkp-instrument-marked-only
11473 @opindex fchkp-instrument-marked-only
11474 @opindex fno-chkp-instrument-marked-only
11475 Instructs Pointer Bounds Checker to instrument only functions
11476 marked with the @code{bnd_instrument} attribute
11477 (@pxref{Function Attributes}). Disabled by default.
11479 @item -fchkp-use-wrappers
11480 @opindex fchkp-use-wrappers
11481 @opindex fno-chkp-use-wrappers
11482 Allows Pointer Bounds Checker to replace calls to built-in functions
11483 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11484 is used to link a program, the GCC driver automatically links
11485 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11486 Enabled by default.
11488 @item -fcf-protection==@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
11489 @opindex fcf-protection
11490 Enable code instrumentation of control-flow transfers to increase
11491 program security by checking that target addresses of control-flow
11492 transfer instructions (such as indirect function call, function return,
11493 indirect jump) are valid. This prevents diverting the flow of control
11494 to an unexpected target. This is intended to protect against such
11495 threats as Return-oriented Programming (ROP), and similarly
11496 call/jmp-oriented programming (COP/JOP).
11498 The value @code{branch} tells the compiler to implement checking of
11499 validity of control-flow transfer at the point of indirect branch
11500 instructions, i.e. call/jmp instructions. The value @code{return}
11501 implements checking of validity at the point of returning from a
11502 function. The value @code{full} is an alias for specifying both
11503 @code{branch} and @code{return}. The value @code{none} turns off
11506 You can also use the @code{nocf_check} attribute to identify
11507 which functions and calls should be skipped from instrumentation
11508 (@pxref{Function Attributes}).
11510 Currently the x86 GNU/Linux target provides an implementation based
11511 on Intel Control-flow Enforcement Technology (CET). Instrumentation
11512 for x86 is controlled by target-specific options @option{-mcet},
11513 @option{-mibt} and @option{-mshstk} (@pxref{x86 Options}).
11515 @item -fstack-protector
11516 @opindex fstack-protector
11517 Emit extra code to check for buffer overflows, such as stack smashing
11518 attacks. This is done by adding a guard variable to functions with
11519 vulnerable objects. This includes functions that call @code{alloca}, and
11520 functions with buffers larger than 8 bytes. The guards are initialized
11521 when a function is entered and then checked when the function exits.
11522 If a guard check fails, an error message is printed and the program exits.
11524 @item -fstack-protector-all
11525 @opindex fstack-protector-all
11526 Like @option{-fstack-protector} except that all functions are protected.
11528 @item -fstack-protector-strong
11529 @opindex fstack-protector-strong
11530 Like @option{-fstack-protector} but includes additional functions to
11531 be protected --- those that have local array definitions, or have
11532 references to local frame addresses.
11534 @item -fstack-protector-explicit
11535 @opindex fstack-protector-explicit
11536 Like @option{-fstack-protector} but only protects those functions which
11537 have the @code{stack_protect} attribute.
11539 @item -fstack-check
11540 @opindex fstack-check
11541 Generate code to verify that you do not go beyond the boundary of the
11542 stack. You should specify this flag if you are running in an
11543 environment with multiple threads, but you only rarely need to specify it in
11544 a single-threaded environment since stack overflow is automatically
11545 detected on nearly all systems if there is only one stack.
11547 Note that this switch does not actually cause checking to be done; the
11548 operating system or the language runtime must do that. The switch causes
11549 generation of code to ensure that they see the stack being extended.
11551 You can additionally specify a string parameter: @samp{no} means no
11552 checking, @samp{generic} means force the use of old-style checking,
11553 @samp{specific} means use the best checking method and is equivalent
11554 to bare @option{-fstack-check}.
11556 Old-style checking is a generic mechanism that requires no specific
11557 target support in the compiler but comes with the following drawbacks:
11561 Modified allocation strategy for large objects: they are always
11562 allocated dynamically if their size exceeds a fixed threshold. Note this
11563 may change the semantics of some code.
11566 Fixed limit on the size of the static frame of functions: when it is
11567 topped by a particular function, stack checking is not reliable and
11568 a warning is issued by the compiler.
11571 Inefficiency: because of both the modified allocation strategy and the
11572 generic implementation, code performance is hampered.
11575 Note that old-style stack checking is also the fallback method for
11576 @samp{specific} if no target support has been added in the compiler.
11578 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
11579 and stack overflows. @samp{specific} is an excellent choice when compiling
11580 Ada code. It is not generally sufficient to protect against stack-clash
11581 attacks. To protect against those you want @samp{-fstack-clash-protection}.
11583 @item -fstack-clash-protection
11584 @opindex fstack-clash-protection
11585 Generate code to prevent stack clash style attacks. When this option is
11586 enabled, the compiler will only allocate one page of stack space at a time
11587 and each page is accessed immediately after allocation. Thus, it prevents
11588 allocations from jumping over any stack guard page provided by the
11591 Most targets do not fully support stack clash protection. However, on
11592 those targets @option{-fstack-clash-protection} will protect dynamic stack
11593 allocations. @option{-fstack-clash-protection} may also provide limited
11594 protection for static stack allocations if the target supports
11595 @option{-fstack-check=specific}.
11597 @item -fstack-limit-register=@var{reg}
11598 @itemx -fstack-limit-symbol=@var{sym}
11599 @itemx -fno-stack-limit
11600 @opindex fstack-limit-register
11601 @opindex fstack-limit-symbol
11602 @opindex fno-stack-limit
11603 Generate code to ensure that the stack does not grow beyond a certain value,
11604 either the value of a register or the address of a symbol. If a larger
11605 stack is required, a signal is raised at run time. For most targets,
11606 the signal is raised before the stack overruns the boundary, so
11607 it is possible to catch the signal without taking special precautions.
11609 For instance, if the stack starts at absolute address @samp{0x80000000}
11610 and grows downwards, you can use the flags
11611 @option{-fstack-limit-symbol=__stack_limit} and
11612 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11613 of 128KB@. Note that this may only work with the GNU linker.
11615 You can locally override stack limit checking by using the
11616 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11618 @item -fsplit-stack
11619 @opindex fsplit-stack
11620 Generate code to automatically split the stack before it overflows.
11621 The resulting program has a discontiguous stack which can only
11622 overflow if the program is unable to allocate any more memory. This
11623 is most useful when running threaded programs, as it is no longer
11624 necessary to calculate a good stack size to use for each thread. This
11625 is currently only implemented for the x86 targets running
11628 When code compiled with @option{-fsplit-stack} calls code compiled
11629 without @option{-fsplit-stack}, there may not be much stack space
11630 available for the latter code to run. If compiling all code,
11631 including library code, with @option{-fsplit-stack} is not an option,
11632 then the linker can fix up these calls so that the code compiled
11633 without @option{-fsplit-stack} always has a large stack. Support for
11634 this is implemented in the gold linker in GNU binutils release 2.21
11637 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11638 @opindex fvtable-verify
11639 This option is only available when compiling C++ code.
11640 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11641 feature that verifies at run time, for every virtual call, that
11642 the vtable pointer through which the call is made is valid for the type of
11643 the object, and has not been corrupted or overwritten. If an invalid vtable
11644 pointer is detected at run time, an error is reported and execution of the
11645 program is immediately halted.
11647 This option causes run-time data structures to be built at program startup,
11648 which are used for verifying the vtable pointers.
11649 The options @samp{std} and @samp{preinit}
11650 control the timing of when these data structures are built. In both cases the
11651 data structures are built before execution reaches @code{main}. Using
11652 @option{-fvtable-verify=std} causes the data structures to be built after
11653 shared libraries have been loaded and initialized.
11654 @option{-fvtable-verify=preinit} causes them to be built before shared
11655 libraries have been loaded and initialized.
11657 If this option appears multiple times in the command line with different
11658 values specified, @samp{none} takes highest priority over both @samp{std} and
11659 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11662 @opindex fvtv-debug
11663 When used in conjunction with @option{-fvtable-verify=std} or
11664 @option{-fvtable-verify=preinit}, causes debug versions of the
11665 runtime functions for the vtable verification feature to be called.
11666 This flag also causes the compiler to log information about which
11667 vtable pointers it finds for each class.
11668 This information is written to a file named @file{vtv_set_ptr_data.log}
11669 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11670 if that is defined or the current working directory otherwise.
11672 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11673 file, be sure to delete any existing one.
11676 @opindex fvtv-counts
11677 This is a debugging flag. When used in conjunction with
11678 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11679 causes the compiler to keep track of the total number of virtual calls
11680 it encounters and the number of verifications it inserts. It also
11681 counts the number of calls to certain run-time library functions
11682 that it inserts and logs this information for each compilation unit.
11683 The compiler writes this information to a file named
11684 @file{vtv_count_data.log} in the directory named by the environment
11685 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11686 directory otherwise. It also counts the size of the vtable pointer sets
11687 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11688 in the same directory.
11690 Note: This feature @emph{appends} data to the log files. To get fresh log
11691 files, be sure to delete any existing ones.
11693 @item -finstrument-functions
11694 @opindex finstrument-functions
11695 Generate instrumentation calls for entry and exit to functions. Just
11696 after function entry and just before function exit, the following
11697 profiling functions are called with the address of the current
11698 function and its call site. (On some platforms,
11699 @code{__builtin_return_address} does not work beyond the current
11700 function, so the call site information may not be available to the
11701 profiling functions otherwise.)
11704 void __cyg_profile_func_enter (void *this_fn,
11706 void __cyg_profile_func_exit (void *this_fn,
11710 The first argument is the address of the start of the current function,
11711 which may be looked up exactly in the symbol table.
11713 This instrumentation is also done for functions expanded inline in other
11714 functions. The profiling calls indicate where, conceptually, the
11715 inline function is entered and exited. This means that addressable
11716 versions of such functions must be available. If all your uses of a
11717 function are expanded inline, this may mean an additional expansion of
11718 code size. If you use @code{extern inline} in your C code, an
11719 addressable version of such functions must be provided. (This is
11720 normally the case anyway, but if you get lucky and the optimizer always
11721 expands the functions inline, you might have gotten away without
11722 providing static copies.)
11724 A function may be given the attribute @code{no_instrument_function}, in
11725 which case this instrumentation is not done. This can be used, for
11726 example, for the profiling functions listed above, high-priority
11727 interrupt routines, and any functions from which the profiling functions
11728 cannot safely be called (perhaps signal handlers, if the profiling
11729 routines generate output or allocate memory).
11731 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11732 @opindex finstrument-functions-exclude-file-list
11734 Set the list of functions that are excluded from instrumentation (see
11735 the description of @option{-finstrument-functions}). If the file that
11736 contains a function definition matches with one of @var{file}, then
11737 that function is not instrumented. The match is done on substrings:
11738 if the @var{file} parameter is a substring of the file name, it is
11739 considered to be a match.
11744 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11748 excludes any inline function defined in files whose pathnames
11749 contain @file{/bits/stl} or @file{include/sys}.
11751 If, for some reason, you want to include letter @samp{,} in one of
11752 @var{sym}, write @samp{\,}. For example,
11753 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11754 (note the single quote surrounding the option).
11756 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11757 @opindex finstrument-functions-exclude-function-list
11759 This is similar to @option{-finstrument-functions-exclude-file-list},
11760 but this option sets the list of function names to be excluded from
11761 instrumentation. The function name to be matched is its user-visible
11762 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11763 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11764 match is done on substrings: if the @var{sym} parameter is a substring
11765 of the function name, it is considered to be a match. For C99 and C++
11766 extended identifiers, the function name must be given in UTF-8, not
11767 using universal character names.
11769 @item -fpatchable-function-entry=@var{N}[,@var{M}]
11770 @opindex fpatchable-function-entry
11771 Generate @var{N} NOPs right at the beginning
11772 of each function, with the function entry point before the @var{M}th NOP.
11773 If @var{M} is omitted, it defaults to @code{0} so the
11774 function entry points to the address just at the first NOP.
11775 The NOP instructions reserve extra space which can be used to patch in
11776 any desired instrumentation at run time, provided that the code segment
11777 is writable. The amount of space is controllable indirectly via
11778 the number of NOPs; the NOP instruction used corresponds to the instruction
11779 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
11780 is target-specific and may also depend on the architecture variant and/or
11781 other compilation options.
11783 For run-time identification, the starting addresses of these areas,
11784 which correspond to their respective function entries minus @var{M},
11785 are additionally collected in the @code{__patchable_function_entries}
11786 section of the resulting binary.
11788 Note that the value of @code{__attribute__ ((patchable_function_entry
11789 (N,M)))} takes precedence over command-line option
11790 @option{-fpatchable-function-entry=N,M}. This can be used to increase
11791 the area size or to remove it completely on a single function.
11792 If @code{N=0}, no pad location is recorded.
11794 The NOP instructions are inserted at---and maybe before, depending on
11795 @var{M}---the function entry address, even before the prologue.
11800 @node Preprocessor Options
11801 @section Options Controlling the Preprocessor
11802 @cindex preprocessor options
11803 @cindex options, preprocessor
11805 These options control the C preprocessor, which is run on each C source
11806 file before actual compilation.
11808 If you use the @option{-E} option, nothing is done except preprocessing.
11809 Some of these options make sense only together with @option{-E} because
11810 they cause the preprocessor output to be unsuitable for actual
11813 In addition to the options listed here, there are a number of options
11814 to control search paths for include files documented in
11815 @ref{Directory Options}.
11816 Options to control preprocessor diagnostics are listed in
11817 @ref{Warning Options}.
11820 @include cppopts.texi
11822 @item -Wp,@var{option}
11824 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11825 and pass @var{option} directly through to the preprocessor. If
11826 @var{option} contains commas, it is split into multiple options at the
11827 commas. However, many options are modified, translated or interpreted
11828 by the compiler driver before being passed to the preprocessor, and
11829 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11830 interface is undocumented and subject to change, so whenever possible
11831 you should avoid using @option{-Wp} and let the driver handle the
11834 @item -Xpreprocessor @var{option}
11835 @opindex Xpreprocessor
11836 Pass @var{option} as an option to the preprocessor. You can use this to
11837 supply system-specific preprocessor options that GCC does not
11840 If you want to pass an option that takes an argument, you must use
11841 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11843 @item -no-integrated-cpp
11844 @opindex no-integrated-cpp
11845 Perform preprocessing as a separate pass before compilation.
11846 By default, GCC performs preprocessing as an integrated part of
11847 input tokenization and parsing.
11848 If this option is provided, the appropriate language front end
11849 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11850 and Objective-C, respectively) is instead invoked twice,
11851 once for preprocessing only and once for actual compilation
11852 of the preprocessed input.
11853 This option may be useful in conjunction with the @option{-B} or
11854 @option{-wrapper} options to specify an alternate preprocessor or
11855 perform additional processing of the program source between
11856 normal preprocessing and compilation.
11860 @node Assembler Options
11861 @section Passing Options to the Assembler
11863 @c prevent bad page break with this line
11864 You can pass options to the assembler.
11867 @item -Wa,@var{option}
11869 Pass @var{option} as an option to the assembler. If @var{option}
11870 contains commas, it is split into multiple options at the commas.
11872 @item -Xassembler @var{option}
11873 @opindex Xassembler
11874 Pass @var{option} as an option to the assembler. You can use this to
11875 supply system-specific assembler options that GCC does not
11878 If you want to pass an option that takes an argument, you must use
11879 @option{-Xassembler} twice, once for the option and once for the argument.
11884 @section Options for Linking
11885 @cindex link options
11886 @cindex options, linking
11888 These options come into play when the compiler links object files into
11889 an executable output file. They are meaningless if the compiler is
11890 not doing a link step.
11894 @item @var{object-file-name}
11895 A file name that does not end in a special recognized suffix is
11896 considered to name an object file or library. (Object files are
11897 distinguished from libraries by the linker according to the file
11898 contents.) If linking is done, these object files are used as input
11907 If any of these options is used, then the linker is not run, and
11908 object file names should not be used as arguments. @xref{Overall
11912 @opindex fuse-ld=bfd
11913 Use the @command{bfd} linker instead of the default linker.
11915 @item -fuse-ld=gold
11916 @opindex fuse-ld=gold
11917 Use the @command{gold} linker instead of the default linker.
11920 @item -l@var{library}
11921 @itemx -l @var{library}
11923 Search the library named @var{library} when linking. (The second
11924 alternative with the library as a separate argument is only for
11925 POSIX compliance and is not recommended.)
11927 It makes a difference where in the command you write this option; the
11928 linker searches and processes libraries and object files in the order they
11929 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11930 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11931 to functions in @samp{z}, those functions may not be loaded.
11933 The linker searches a standard list of directories for the library,
11934 which is actually a file named @file{lib@var{library}.a}. The linker
11935 then uses this file as if it had been specified precisely by name.
11937 The directories searched include several standard system directories
11938 plus any that you specify with @option{-L}.
11940 Normally the files found this way are library files---archive files
11941 whose members are object files. The linker handles an archive file by
11942 scanning through it for members which define symbols that have so far
11943 been referenced but not defined. But if the file that is found is an
11944 ordinary object file, it is linked in the usual fashion. The only
11945 difference between using an @option{-l} option and specifying a file name
11946 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11947 and searches several directories.
11951 You need this special case of the @option{-l} option in order to
11952 link an Objective-C or Objective-C++ program.
11954 @item -nostartfiles
11955 @opindex nostartfiles
11956 Do not use the standard system startup files when linking.
11957 The standard system libraries are used normally, unless @option{-nostdlib}
11958 or @option{-nodefaultlibs} is used.
11960 @item -nodefaultlibs
11961 @opindex nodefaultlibs
11962 Do not use the standard system libraries when linking.
11963 Only the libraries you specify are passed to the linker, and options
11964 specifying linkage of the system libraries, such as @option{-static-libgcc}
11965 or @option{-shared-libgcc}, are ignored.
11966 The standard startup files are used normally, unless @option{-nostartfiles}
11969 The compiler may generate calls to @code{memcmp},
11970 @code{memset}, @code{memcpy} and @code{memmove}.
11971 These entries are usually resolved by entries in
11972 libc. These entry points should be supplied through some other
11973 mechanism when this option is specified.
11977 Do not use the standard system startup files or libraries when linking.
11978 No startup files and only the libraries you specify are passed to
11979 the linker, and options specifying linkage of the system libraries, such as
11980 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11982 The compiler may generate calls to @code{memcmp}, @code{memset},
11983 @code{memcpy} and @code{memmove}.
11984 These entries are usually resolved by entries in
11985 libc. These entry points should be supplied through some other
11986 mechanism when this option is specified.
11988 @cindex @option{-lgcc}, use with @option{-nostdlib}
11989 @cindex @option{-nostdlib} and unresolved references
11990 @cindex unresolved references and @option{-nostdlib}
11991 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11992 @cindex @option{-nodefaultlibs} and unresolved references
11993 @cindex unresolved references and @option{-nodefaultlibs}
11994 One of the standard libraries bypassed by @option{-nostdlib} and
11995 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11996 which GCC uses to overcome shortcomings of particular machines, or special
11997 needs for some languages.
11998 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11999 Collection (GCC) Internals},
12000 for more discussion of @file{libgcc.a}.)
12001 In most cases, you need @file{libgcc.a} even when you want to avoid
12002 other standard libraries. In other words, when you specify @option{-nostdlib}
12003 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
12004 This ensures that you have no unresolved references to internal GCC
12005 library subroutines.
12006 (An example of such an internal subroutine is @code{__main}, used to ensure C++
12007 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
12008 GNU Compiler Collection (GCC) Internals}.)
12012 Produce a dynamically linked position independent executable on targets
12013 that support it. For predictable results, you must also specify the same
12014 set of options used for compilation (@option{-fpie}, @option{-fPIE},
12015 or model suboptions) when you specify this linker option.
12019 Don't produce a dynamically linked position independent executable.
12022 @opindex static-pie
12023 Produce a static position independent executable on targets that support
12024 it. A static position independent executable is similar to a static
12025 executable, but can be loaded at any address without a dynamic linker.
12026 For predictable results, you must also specify the same set of options
12027 used for compilation (@option{-fpie}, @option{-fPIE}, or model
12028 suboptions) when you specify this linker option.
12032 Link with the POSIX threads library. This option is supported on
12033 GNU/Linux targets, most other Unix derivatives, and also on
12034 x86 Cygwin and MinGW targets. On some targets this option also sets
12035 flags for the preprocessor, so it should be used consistently for both
12036 compilation and linking.
12040 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
12041 that support it. This instructs the linker to add all symbols, not
12042 only used ones, to the dynamic symbol table. This option is needed
12043 for some uses of @code{dlopen} or to allow obtaining backtraces
12044 from within a program.
12048 Remove all symbol table and relocation information from the executable.
12052 On systems that support dynamic linking, this overrides @option{-pie}
12053 and prevents linking with the shared libraries. On other systems, this
12054 option has no effect.
12058 Produce a shared object which can then be linked with other objects to
12059 form an executable. Not all systems support this option. For predictable
12060 results, you must also specify the same set of options used for compilation
12061 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
12062 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
12063 needs to build supplementary stub code for constructors to work. On
12064 multi-libbed systems, @samp{gcc -shared} must select the correct support
12065 libraries to link against. Failing to supply the correct flags may lead
12066 to subtle defects. Supplying them in cases where they are not necessary
12069 @item -shared-libgcc
12070 @itemx -static-libgcc
12071 @opindex shared-libgcc
12072 @opindex static-libgcc
12073 On systems that provide @file{libgcc} as a shared library, these options
12074 force the use of either the shared or static version, respectively.
12075 If no shared version of @file{libgcc} was built when the compiler was
12076 configured, these options have no effect.
12078 There are several situations in which an application should use the
12079 shared @file{libgcc} instead of the static version. The most common
12080 of these is when the application wishes to throw and catch exceptions
12081 across different shared libraries. In that case, each of the libraries
12082 as well as the application itself should use the shared @file{libgcc}.
12084 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
12085 whenever you build a shared library or a main executable, because C++
12086 programs typically use exceptions, so this is the right thing to do.
12088 If, instead, you use the GCC driver to create shared libraries, you may
12089 find that they are not always linked with the shared @file{libgcc}.
12090 If GCC finds, at its configuration time, that you have a non-GNU linker
12091 or a GNU linker that does not support option @option{--eh-frame-hdr},
12092 it links the shared version of @file{libgcc} into shared libraries
12093 by default. Otherwise, it takes advantage of the linker and optimizes
12094 away the linking with the shared version of @file{libgcc}, linking with
12095 the static version of libgcc by default. This allows exceptions to
12096 propagate through such shared libraries, without incurring relocation
12097 costs at library load time.
12099 However, if a library or main executable is supposed to throw or catch
12100 exceptions, you must link it using the G++ driver, as appropriate
12101 for the languages used in the program, or using the option
12102 @option{-shared-libgcc}, such that it is linked with the shared
12105 @item -static-libasan
12106 @opindex static-libasan
12107 When the @option{-fsanitize=address} option is used to link a program,
12108 the GCC driver automatically links against @option{libasan}. If
12109 @file{libasan} is available as a shared library, and the @option{-static}
12110 option is not used, then this links against the shared version of
12111 @file{libasan}. The @option{-static-libasan} option directs the GCC
12112 driver to link @file{libasan} statically, without necessarily linking
12113 other libraries statically.
12115 @item -static-libtsan
12116 @opindex static-libtsan
12117 When the @option{-fsanitize=thread} option is used to link a program,
12118 the GCC driver automatically links against @option{libtsan}. If
12119 @file{libtsan} is available as a shared library, and the @option{-static}
12120 option is not used, then this links against the shared version of
12121 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
12122 driver to link @file{libtsan} statically, without necessarily linking
12123 other libraries statically.
12125 @item -static-liblsan
12126 @opindex static-liblsan
12127 When the @option{-fsanitize=leak} option is used to link a program,
12128 the GCC driver automatically links against @option{liblsan}. If
12129 @file{liblsan} is available as a shared library, and the @option{-static}
12130 option is not used, then this links against the shared version of
12131 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
12132 driver to link @file{liblsan} statically, without necessarily linking
12133 other libraries statically.
12135 @item -static-libubsan
12136 @opindex static-libubsan
12137 When the @option{-fsanitize=undefined} option is used to link a program,
12138 the GCC driver automatically links against @option{libubsan}. If
12139 @file{libubsan} is available as a shared library, and the @option{-static}
12140 option is not used, then this links against the shared version of
12141 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
12142 driver to link @file{libubsan} statically, without necessarily linking
12143 other libraries statically.
12145 @item -static-libmpx
12146 @opindex static-libmpx
12147 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
12148 used to link a program, the GCC driver automatically links against
12149 @file{libmpx}. If @file{libmpx} is available as a shared library,
12150 and the @option{-static} option is not used, then this links against
12151 the shared version of @file{libmpx}. The @option{-static-libmpx}
12152 option directs the GCC driver to link @file{libmpx} statically,
12153 without necessarily linking other libraries statically.
12155 @item -static-libmpxwrappers
12156 @opindex static-libmpxwrappers
12157 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
12158 to link a program without also using @option{-fno-chkp-use-wrappers}, the
12159 GCC driver automatically links against @file{libmpxwrappers}. If
12160 @file{libmpxwrappers} is available as a shared library, and the
12161 @option{-static} option is not used, then this links against the shared
12162 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
12163 option directs the GCC driver to link @file{libmpxwrappers} statically,
12164 without necessarily linking other libraries statically.
12166 @item -static-libstdc++
12167 @opindex static-libstdc++
12168 When the @command{g++} program is used to link a C++ program, it
12169 normally automatically links against @option{libstdc++}. If
12170 @file{libstdc++} is available as a shared library, and the
12171 @option{-static} option is not used, then this links against the
12172 shared version of @file{libstdc++}. That is normally fine. However, it
12173 is sometimes useful to freeze the version of @file{libstdc++} used by
12174 the program without going all the way to a fully static link. The
12175 @option{-static-libstdc++} option directs the @command{g++} driver to
12176 link @file{libstdc++} statically, without necessarily linking other
12177 libraries statically.
12181 Bind references to global symbols when building a shared object. Warn
12182 about any unresolved references (unless overridden by the link editor
12183 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
12186 @item -T @var{script}
12188 @cindex linker script
12189 Use @var{script} as the linker script. This option is supported by most
12190 systems using the GNU linker. On some targets, such as bare-board
12191 targets without an operating system, the @option{-T} option may be required
12192 when linking to avoid references to undefined symbols.
12194 @item -Xlinker @var{option}
12196 Pass @var{option} as an option to the linker. You can use this to
12197 supply system-specific linker options that GCC does not recognize.
12199 If you want to pass an option that takes a separate argument, you must use
12200 @option{-Xlinker} twice, once for the option and once for the argument.
12201 For example, to pass @option{-assert definitions}, you must write
12202 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
12203 @option{-Xlinker "-assert definitions"}, because this passes the entire
12204 string as a single argument, which is not what the linker expects.
12206 When using the GNU linker, it is usually more convenient to pass
12207 arguments to linker options using the @option{@var{option}=@var{value}}
12208 syntax than as separate arguments. For example, you can specify
12209 @option{-Xlinker -Map=output.map} rather than
12210 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
12211 this syntax for command-line options.
12213 @item -Wl,@var{option}
12215 Pass @var{option} as an option to the linker. If @var{option} contains
12216 commas, it is split into multiple options at the commas. You can use this
12217 syntax to pass an argument to the option.
12218 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
12219 linker. When using the GNU linker, you can also get the same effect with
12220 @option{-Wl,-Map=output.map}.
12222 @item -u @var{symbol}
12224 Pretend the symbol @var{symbol} is undefined, to force linking of
12225 library modules to define it. You can use @option{-u} multiple times with
12226 different symbols to force loading of additional library modules.
12228 @item -z @var{keyword}
12230 @option{-z} is passed directly on to the linker along with the keyword
12231 @var{keyword}. See the section in the documentation of your linker for
12232 permitted values and their meanings.
12235 @node Directory Options
12236 @section Options for Directory Search
12237 @cindex directory options
12238 @cindex options, directory search
12239 @cindex search path
12241 These options specify directories to search for header files, for
12242 libraries and for parts of the compiler:
12245 @include cppdiropts.texi
12247 @item -iplugindir=@var{dir}
12248 @opindex iplugindir=
12249 Set the directory to search for plugins that are passed
12250 by @option{-fplugin=@var{name}} instead of
12251 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
12252 to be used by the user, but only passed by the driver.
12256 Add directory @var{dir} to the list of directories to be searched
12259 @item -B@var{prefix}
12261 This option specifies where to find the executables, libraries,
12262 include files, and data files of the compiler itself.
12264 The compiler driver program runs one or more of the subprograms
12265 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
12266 @var{prefix} as a prefix for each program it tries to run, both with and
12267 without @samp{@var{machine}/@var{version}/} for the corresponding target
12268 machine and compiler version.
12270 For each subprogram to be run, the compiler driver first tries the
12271 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
12272 is not specified, the driver tries two standard prefixes,
12273 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
12274 those results in a file name that is found, the unmodified program
12275 name is searched for using the directories specified in your
12276 @env{PATH} environment variable.
12278 The compiler checks to see if the path provided by @option{-B}
12279 refers to a directory, and if necessary it adds a directory
12280 separator character at the end of the path.
12282 @option{-B} prefixes that effectively specify directory names also apply
12283 to libraries in the linker, because the compiler translates these
12284 options into @option{-L} options for the linker. They also apply to
12285 include files in the preprocessor, because the compiler translates these
12286 options into @option{-isystem} options for the preprocessor. In this case,
12287 the compiler appends @samp{include} to the prefix.
12289 The runtime support file @file{libgcc.a} can also be searched for using
12290 the @option{-B} prefix, if needed. If it is not found there, the two
12291 standard prefixes above are tried, and that is all. The file is left
12292 out of the link if it is not found by those means.
12294 Another way to specify a prefix much like the @option{-B} prefix is to use
12295 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
12298 As a special kludge, if the path provided by @option{-B} is
12299 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
12300 9, then it is replaced by @file{[dir/]include}. This is to help
12301 with boot-strapping the compiler.
12303 @item -no-canonical-prefixes
12304 @opindex no-canonical-prefixes
12305 Do not expand any symbolic links, resolve references to @samp{/../}
12306 or @samp{/./}, or make the path absolute when generating a relative
12309 @item --sysroot=@var{dir}
12311 Use @var{dir} as the logical root directory for headers and libraries.
12312 For example, if the compiler normally searches for headers in
12313 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
12314 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
12316 If you use both this option and the @option{-isysroot} option, then
12317 the @option{--sysroot} option applies to libraries, but the
12318 @option{-isysroot} option applies to header files.
12320 The GNU linker (beginning with version 2.16) has the necessary support
12321 for this option. If your linker does not support this option, the
12322 header file aspect of @option{--sysroot} still works, but the
12323 library aspect does not.
12325 @item --no-sysroot-suffix
12326 @opindex no-sysroot-suffix
12327 For some targets, a suffix is added to the root directory specified
12328 with @option{--sysroot}, depending on the other options used, so that
12329 headers may for example be found in
12330 @file{@var{dir}/@var{suffix}/usr/include} instead of
12331 @file{@var{dir}/usr/include}. This option disables the addition of
12336 @node Code Gen Options
12337 @section Options for Code Generation Conventions
12338 @cindex code generation conventions
12339 @cindex options, code generation
12340 @cindex run-time options
12342 These machine-independent options control the interface conventions
12343 used in code generation.
12345 Most of them have both positive and negative forms; the negative form
12346 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
12347 one of the forms is listed---the one that is not the default. You
12348 can figure out the other form by either removing @samp{no-} or adding
12352 @item -fstack-reuse=@var{reuse-level}
12353 @opindex fstack_reuse
12354 This option controls stack space reuse for user declared local/auto variables
12355 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
12356 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
12357 local variables and temporaries, @samp{named_vars} enables the reuse only for
12358 user defined local variables with names, and @samp{none} disables stack reuse
12359 completely. The default value is @samp{all}. The option is needed when the
12360 program extends the lifetime of a scoped local variable or a compiler generated
12361 temporary beyond the end point defined by the language. When a lifetime of
12362 a variable ends, and if the variable lives in memory, the optimizing compiler
12363 has the freedom to reuse its stack space with other temporaries or scoped
12364 local variables whose live range does not overlap with it. Legacy code extending
12365 local lifetime is likely to break with the stack reuse optimization.
12384 if (*p == 10) // out of scope use of local1
12395 A(int k) : i(k), j(k) @{ @}
12402 void foo(const A& ar)
12409 foo(A(10)); // temp object's lifetime ends when foo returns
12415 ap->i+= 10; // ap references out of scope temp whose space
12416 // is reused with a. What is the value of ap->i?
12421 The lifetime of a compiler generated temporary is well defined by the C++
12422 standard. When a lifetime of a temporary ends, and if the temporary lives
12423 in memory, the optimizing compiler has the freedom to reuse its stack
12424 space with other temporaries or scoped local variables whose live range
12425 does not overlap with it. However some of the legacy code relies on
12426 the behavior of older compilers in which temporaries' stack space is
12427 not reused, the aggressive stack reuse can lead to runtime errors. This
12428 option is used to control the temporary stack reuse optimization.
12432 This option generates traps for signed overflow on addition, subtraction,
12433 multiplication operations.
12434 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12435 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12436 @option{-fwrapv} being effective. Note that only active options override, so
12437 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12438 results in @option{-ftrapv} being effective.
12442 This option instructs the compiler to assume that signed arithmetic
12443 overflow of addition, subtraction and multiplication wraps around
12444 using twos-complement representation. This flag enables some optimizations
12445 and disables others.
12446 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12447 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12448 @option{-fwrapv} being effective. Note that only active options override, so
12449 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12450 results in @option{-ftrapv} being effective.
12453 @opindex fexceptions
12454 Enable exception handling. Generates extra code needed to propagate
12455 exceptions. For some targets, this implies GCC generates frame
12456 unwind information for all functions, which can produce significant data
12457 size overhead, although it does not affect execution. If you do not
12458 specify this option, GCC enables it by default for languages like
12459 C++ that normally require exception handling, and disables it for
12460 languages like C that do not normally require it. However, you may need
12461 to enable this option when compiling C code that needs to interoperate
12462 properly with exception handlers written in C++. You may also wish to
12463 disable this option if you are compiling older C++ programs that don't
12464 use exception handling.
12466 @item -fnon-call-exceptions
12467 @opindex fnon-call-exceptions
12468 Generate code that allows trapping instructions to throw exceptions.
12469 Note that this requires platform-specific runtime support that does
12470 not exist everywhere. Moreover, it only allows @emph{trapping}
12471 instructions to throw exceptions, i.e.@: memory references or floating-point
12472 instructions. It does not allow exceptions to be thrown from
12473 arbitrary signal handlers such as @code{SIGALRM}.
12475 @item -fdelete-dead-exceptions
12476 @opindex fdelete-dead-exceptions
12477 Consider that instructions that may throw exceptions but don't otherwise
12478 contribute to the execution of the program can be optimized away.
12479 This option is enabled by default for the Ada front end, as permitted by
12480 the Ada language specification.
12481 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12483 @item -funwind-tables
12484 @opindex funwind-tables
12485 Similar to @option{-fexceptions}, except that it just generates any needed
12486 static data, but does not affect the generated code in any other way.
12487 You normally do not need to enable this option; instead, a language processor
12488 that needs this handling enables it on your behalf.
12490 @item -fasynchronous-unwind-tables
12491 @opindex fasynchronous-unwind-tables
12492 Generate unwind table in DWARF format, if supported by target machine. The
12493 table is exact at each instruction boundary, so it can be used for stack
12494 unwinding from asynchronous events (such as debugger or garbage collector).
12496 @item -fno-gnu-unique
12497 @opindex fno-gnu-unique
12498 On systems with recent GNU assembler and C library, the C++ compiler
12499 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12500 of template static data members and static local variables in inline
12501 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12502 is necessary to avoid problems with a library used by two different
12503 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12504 therefore disagreeing with the other one about the binding of the
12505 symbol. But this causes @code{dlclose} to be ignored for affected
12506 DSOs; if your program relies on reinitialization of a DSO via
12507 @code{dlclose} and @code{dlopen}, you can use
12508 @option{-fno-gnu-unique}.
12510 @item -fpcc-struct-return
12511 @opindex fpcc-struct-return
12512 Return ``short'' @code{struct} and @code{union} values in memory like
12513 longer ones, rather than in registers. This convention is less
12514 efficient, but it has the advantage of allowing intercallability between
12515 GCC-compiled files and files compiled with other compilers, particularly
12516 the Portable C Compiler (pcc).
12518 The precise convention for returning structures in memory depends
12519 on the target configuration macros.
12521 Short structures and unions are those whose size and alignment match
12522 that of some integer type.
12524 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12525 switch is not binary compatible with code compiled with the
12526 @option{-freg-struct-return} switch.
12527 Use it to conform to a non-default application binary interface.
12529 @item -freg-struct-return
12530 @opindex freg-struct-return
12531 Return @code{struct} and @code{union} values in registers when possible.
12532 This is more efficient for small structures than
12533 @option{-fpcc-struct-return}.
12535 If you specify neither @option{-fpcc-struct-return} nor
12536 @option{-freg-struct-return}, GCC defaults to whichever convention is
12537 standard for the target. If there is no standard convention, GCC
12538 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12539 the principal compiler. In those cases, we can choose the standard, and
12540 we chose the more efficient register return alternative.
12542 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12543 switch is not binary compatible with code compiled with the
12544 @option{-fpcc-struct-return} switch.
12545 Use it to conform to a non-default application binary interface.
12547 @item -fshort-enums
12548 @opindex fshort-enums
12549 Allocate to an @code{enum} type only as many bytes as it needs for the
12550 declared range of possible values. Specifically, the @code{enum} type
12551 is equivalent to the smallest integer type that has enough room.
12553 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12554 code that is not binary compatible with code generated without that switch.
12555 Use it to conform to a non-default application binary interface.
12557 @item -fshort-wchar
12558 @opindex fshort-wchar
12559 Override the underlying type for @code{wchar_t} to be @code{short
12560 unsigned int} instead of the default for the target. This option is
12561 useful for building programs to run under WINE@.
12563 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12564 code that is not binary compatible with code generated without that switch.
12565 Use it to conform to a non-default application binary interface.
12568 @opindex fno-common
12569 @cindex tentative definitions
12570 In C code, this option controls the placement of global variables
12571 defined without an initializer, known as @dfn{tentative definitions}
12572 in the C standard. Tentative definitions are distinct from declarations
12573 of a variable with the @code{extern} keyword, which do not allocate storage.
12575 Unix C compilers have traditionally allocated storage for
12576 uninitialized global variables in a common block. This allows the
12577 linker to resolve all tentative definitions of the same variable
12578 in different compilation units to the same object, or to a non-tentative
12580 This is the behavior specified by @option{-fcommon}, and is the default for
12581 GCC on most targets.
12582 On the other hand, this behavior is not required by ISO
12583 C, and on some targets may carry a speed or code size penalty on
12584 variable references.
12586 The @option{-fno-common} option specifies that the compiler should instead
12587 place uninitialized global variables in the data section of the object file.
12588 This inhibits the merging of tentative definitions by the linker so
12589 you get a multiple-definition error if the same
12590 variable is defined in more than one compilation unit.
12591 Compiling with @option{-fno-common} is useful on targets for which
12592 it provides better performance, or if you wish to verify that the
12593 program will work on other systems that always treat uninitialized
12594 variable definitions this way.
12598 Ignore the @code{#ident} directive.
12600 @item -finhibit-size-directive
12601 @opindex finhibit-size-directive
12602 Don't output a @code{.size} assembler directive, or anything else that
12603 would cause trouble if the function is split in the middle, and the
12604 two halves are placed at locations far apart in memory. This option is
12605 used when compiling @file{crtstuff.c}; you should not need to use it
12608 @item -fverbose-asm
12609 @opindex fverbose-asm
12610 Put extra commentary information in the generated assembly code to
12611 make it more readable. This option is generally only of use to those
12612 who actually need to read the generated assembly code (perhaps while
12613 debugging the compiler itself).
12615 @option{-fno-verbose-asm}, the default, causes the
12616 extra information to be omitted and is useful when comparing two assembler
12619 The added comments include:
12624 information on the compiler version and command-line options,
12627 the source code lines associated with the assembly instructions,
12628 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12631 hints on which high-level expressions correspond to
12632 the various assembly instruction operands.
12636 For example, given this C source file:
12644 for (i = 0; i < n; i++)
12651 compiling to (x86_64) assembly via @option{-S} and emitting the result
12652 direct to stdout via @option{-o} @option{-}
12655 gcc -S test.c -fverbose-asm -Os -o -
12658 gives output similar to this:
12662 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12669 .type test, @@function
12673 # test.c:4: int total = 0;
12674 xorl %eax, %eax # <retval>
12675 # test.c:6: for (i = 0; i < n; i++)
12676 xorl %edx, %edx # i
12678 # test.c:6: for (i = 0; i < n; i++)
12679 cmpl %edi, %edx # n, i
12681 # test.c:7: total += i * i;
12682 movl %edx, %ecx # i, tmp92
12683 imull %edx, %ecx # i, tmp92
12684 # test.c:6: for (i = 0; i < n; i++)
12686 # test.c:7: total += i * i;
12687 addl %ecx, %eax # tmp92, <retval>
12695 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12696 .section .note.GNU-stack,"",@@progbits
12699 The comments are intended for humans rather than machines and hence the
12700 precise format of the comments is subject to change.
12702 @item -frecord-gcc-switches
12703 @opindex frecord-gcc-switches
12704 This switch causes the command line used to invoke the
12705 compiler to be recorded into the object file that is being created.
12706 This switch is only implemented on some targets and the exact format
12707 of the recording is target and binary file format dependent, but it
12708 usually takes the form of a section containing ASCII text. This
12709 switch is related to the @option{-fverbose-asm} switch, but that
12710 switch only records information in the assembler output file as
12711 comments, so it never reaches the object file.
12712 See also @option{-grecord-gcc-switches} for another
12713 way of storing compiler options into the object file.
12717 @cindex global offset table
12719 Generate position-independent code (PIC) suitable for use in a shared
12720 library, if supported for the target machine. Such code accesses all
12721 constant addresses through a global offset table (GOT)@. The dynamic
12722 loader resolves the GOT entries when the program starts (the dynamic
12723 loader is not part of GCC; it is part of the operating system). If
12724 the GOT size for the linked executable exceeds a machine-specific
12725 maximum size, you get an error message from the linker indicating that
12726 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12727 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12728 on the m68k and RS/6000. The x86 has no such limit.)
12730 Position-independent code requires special support, and therefore works
12731 only on certain machines. For the x86, GCC supports PIC for System V
12732 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12733 position-independent.
12735 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12740 If supported for the target machine, emit position-independent code,
12741 suitable for dynamic linking and avoiding any limit on the size of the
12742 global offset table. This option makes a difference on AArch64, m68k,
12743 PowerPC and SPARC@.
12745 Position-independent code requires special support, and therefore works
12746 only on certain machines.
12748 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12755 These options are similar to @option{-fpic} and @option{-fPIC}, but
12756 generated position independent code can be only linked into executables.
12757 Usually these options are used when @option{-pie} GCC option is
12758 used during linking.
12760 @option{-fpie} and @option{-fPIE} both define the macros
12761 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12762 for @option{-fpie} and 2 for @option{-fPIE}.
12766 Do not use the PLT for external function calls in position-independent code.
12767 Instead, load the callee address at call sites from the GOT and branch to it.
12768 This leads to more efficient code by eliminating PLT stubs and exposing
12769 GOT loads to optimizations. On architectures such as 32-bit x86 where
12770 PLT stubs expect the GOT pointer in a specific register, this gives more
12771 register allocation freedom to the compiler.
12772 Lazy binding requires use of the PLT;
12773 with @option{-fno-plt} all external symbols are resolved at load time.
12775 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12776 through the PLT for specific external functions.
12778 In position-dependent code, a few targets also convert calls to
12779 functions that are marked to not use the PLT to use the GOT instead.
12781 @item -fno-jump-tables
12782 @opindex fno-jump-tables
12783 Do not use jump tables for switch statements even where it would be
12784 more efficient than other code generation strategies. This option is
12785 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12786 building code that forms part of a dynamic linker and cannot
12787 reference the address of a jump table. On some targets, jump tables
12788 do not require a GOT and this option is not needed.
12790 @item -ffixed-@var{reg}
12792 Treat the register named @var{reg} as a fixed register; generated code
12793 should never refer to it (except perhaps as a stack pointer, frame
12794 pointer or in some other fixed role).
12796 @var{reg} must be the name of a register. The register names accepted
12797 are machine-specific and are defined in the @code{REGISTER_NAMES}
12798 macro in the machine description macro file.
12800 This flag does not have a negative form, because it specifies a
12803 @item -fcall-used-@var{reg}
12804 @opindex fcall-used
12805 Treat the register named @var{reg} as an allocable register that is
12806 clobbered by function calls. It may be allocated for temporaries or
12807 variables that do not live across a call. Functions compiled this way
12808 do not save and restore the register @var{reg}.
12810 It is an error to use this flag with the frame pointer or stack pointer.
12811 Use of this flag for other registers that have fixed pervasive roles in
12812 the machine's execution model produces disastrous results.
12814 This flag does not have a negative form, because it specifies a
12817 @item -fcall-saved-@var{reg}
12818 @opindex fcall-saved
12819 Treat the register named @var{reg} as an allocable register saved by
12820 functions. It may be allocated even for temporaries or variables that
12821 live across a call. Functions compiled this way save and restore
12822 the register @var{reg} if they use it.
12824 It is an error to use this flag with the frame pointer or stack pointer.
12825 Use of this flag for other registers that have fixed pervasive roles in
12826 the machine's execution model produces disastrous results.
12828 A different sort of disaster results from the use of this flag for
12829 a register in which function values may be returned.
12831 This flag does not have a negative form, because it specifies a
12834 @item -fpack-struct[=@var{n}]
12835 @opindex fpack-struct
12836 Without a value specified, pack all structure members together without
12837 holes. When a value is specified (which must be a small power of two), pack
12838 structure members according to this value, representing the maximum
12839 alignment (that is, objects with default alignment requirements larger than
12840 this are output potentially unaligned at the next fitting location.
12842 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12843 code that is not binary compatible with code generated without that switch.
12844 Additionally, it makes the code suboptimal.
12845 Use it to conform to a non-default application binary interface.
12847 @item -fleading-underscore
12848 @opindex fleading-underscore
12849 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12850 change the way C symbols are represented in the object file. One use
12851 is to help link with legacy assembly code.
12853 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12854 generate code that is not binary compatible with code generated without that
12855 switch. Use it to conform to a non-default application binary interface.
12856 Not all targets provide complete support for this switch.
12858 @item -ftls-model=@var{model}
12859 @opindex ftls-model
12860 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12861 The @var{model} argument should be one of @samp{global-dynamic},
12862 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12863 Note that the choice is subject to optimization: the compiler may use
12864 a more efficient model for symbols not visible outside of the translation
12865 unit, or if @option{-fpic} is not given on the command line.
12867 The default without @option{-fpic} is @samp{initial-exec}; with
12868 @option{-fpic} the default is @samp{global-dynamic}.
12870 @item -ftrampolines
12871 @opindex ftrampolines
12872 For targets that normally need trampolines for nested functions, always
12873 generate them instead of using descriptors. Otherwise, for targets that
12874 do not need them, like for example HP-PA or IA-64, do nothing.
12876 A trampoline is a small piece of code that is created at run time on the
12877 stack when the address of a nested function is taken, and is used to call
12878 the nested function indirectly. Therefore, it requires the stack to be
12879 made executable in order for the program to work properly.
12881 @option{-fno-trampolines} is enabled by default on a language by language
12882 basis to let the compiler avoid generating them, if it computes that this
12883 is safe, and replace them with descriptors. Descriptors are made up of data
12884 only, but the generated code must be prepared to deal with them. As of this
12885 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12887 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12888 @option{-fno-trampolines} are not binary compatible if nested functions are
12889 present. This option must therefore be used on a program-wide basis and be
12890 manipulated with extreme care.
12892 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12893 @opindex fvisibility
12894 Set the default ELF image symbol visibility to the specified option---all
12895 symbols are marked with this unless overridden within the code.
12896 Using this feature can very substantially improve linking and
12897 load times of shared object libraries, produce more optimized
12898 code, provide near-perfect API export and prevent symbol clashes.
12899 It is @strong{strongly} recommended that you use this in any shared objects
12902 Despite the nomenclature, @samp{default} always means public; i.e.,
12903 available to be linked against from outside the shared object.
12904 @samp{protected} and @samp{internal} are pretty useless in real-world
12905 usage so the only other commonly used option is @samp{hidden}.
12906 The default if @option{-fvisibility} isn't specified is
12907 @samp{default}, i.e., make every symbol public.
12909 A good explanation of the benefits offered by ensuring ELF
12910 symbols have the correct visibility is given by ``How To Write
12911 Shared Libraries'' by Ulrich Drepper (which can be found at
12912 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
12913 solution made possible by this option to marking things hidden when
12914 the default is public is to make the default hidden and mark things
12915 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12916 and @code{__attribute__ ((visibility("default")))} instead of
12917 @code{__declspec(dllexport)} you get almost identical semantics with
12918 identical syntax. This is a great boon to those working with
12919 cross-platform projects.
12921 For those adding visibility support to existing code, you may find
12922 @code{#pragma GCC visibility} of use. This works by you enclosing
12923 the declarations you wish to set visibility for with (for example)
12924 @code{#pragma GCC visibility push(hidden)} and
12925 @code{#pragma GCC visibility pop}.
12926 Bear in mind that symbol visibility should be viewed @strong{as
12927 part of the API interface contract} and thus all new code should
12928 always specify visibility when it is not the default; i.e., declarations
12929 only for use within the local DSO should @strong{always} be marked explicitly
12930 as hidden as so to avoid PLT indirection overheads---making this
12931 abundantly clear also aids readability and self-documentation of the code.
12932 Note that due to ISO C++ specification requirements, @code{operator new} and
12933 @code{operator delete} must always be of default visibility.
12935 Be aware that headers from outside your project, in particular system
12936 headers and headers from any other library you use, may not be
12937 expecting to be compiled with visibility other than the default. You
12938 may need to explicitly say @code{#pragma GCC visibility push(default)}
12939 before including any such headers.
12941 @code{extern} declarations are not affected by @option{-fvisibility}, so
12942 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12943 no modifications. However, this means that calls to @code{extern}
12944 functions with no explicit visibility use the PLT, so it is more
12945 effective to use @code{__attribute ((visibility))} and/or
12946 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12947 declarations should be treated as hidden.
12949 Note that @option{-fvisibility} does affect C++ vague linkage
12950 entities. This means that, for instance, an exception class that is
12951 be thrown between DSOs must be explicitly marked with default
12952 visibility so that the @samp{type_info} nodes are unified between
12955 An overview of these techniques, their benefits and how to use them
12956 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12958 @item -fstrict-volatile-bitfields
12959 @opindex fstrict-volatile-bitfields
12960 This option should be used if accesses to volatile bit-fields (or other
12961 structure fields, although the compiler usually honors those types
12962 anyway) should use a single access of the width of the
12963 field's type, aligned to a natural alignment if possible. For
12964 example, targets with memory-mapped peripheral registers might require
12965 all such accesses to be 16 bits wide; with this flag you can
12966 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12967 is 16 bits on these targets) to force GCC to use 16-bit accesses
12968 instead of, perhaps, a more efficient 32-bit access.
12970 If this option is disabled, the compiler uses the most efficient
12971 instruction. In the previous example, that might be a 32-bit load
12972 instruction, even though that accesses bytes that do not contain
12973 any portion of the bit-field, or memory-mapped registers unrelated to
12974 the one being updated.
12976 In some cases, such as when the @code{packed} attribute is applied to a
12977 structure field, it may not be possible to access the field with a single
12978 read or write that is correctly aligned for the target machine. In this
12979 case GCC falls back to generating multiple accesses rather than code that
12980 will fault or truncate the result at run time.
12982 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12983 not allowed to touch non bit-field members. It is therefore recommended
12984 to define all bits of the field's type as bit-field members.
12986 The default value of this option is determined by the application binary
12987 interface for the target processor.
12989 @item -fsync-libcalls
12990 @opindex fsync-libcalls
12991 This option controls whether any out-of-line instance of the @code{__sync}
12992 family of functions may be used to implement the C++11 @code{__atomic}
12993 family of functions.
12995 The default value of this option is enabled, thus the only useful form
12996 of the option is @option{-fno-sync-libcalls}. This option is used in
12997 the implementation of the @file{libatomic} runtime library.
13001 @node Developer Options
13002 @section GCC Developer Options
13003 @cindex developer options
13004 @cindex debugging GCC
13005 @cindex debug dump options
13006 @cindex dump options
13007 @cindex compilation statistics
13009 This section describes command-line options that are primarily of
13010 interest to GCC developers, including options to support compiler
13011 testing and investigation of compiler bugs and compile-time
13012 performance problems. This includes options that produce debug dumps
13013 at various points in the compilation; that print statistics such as
13014 memory use and execution time; and that print information about GCC's
13015 configuration, such as where it searches for libraries. You should
13016 rarely need to use any of these options for ordinary compilation and
13021 @item -d@var{letters}
13022 @itemx -fdump-rtl-@var{pass}
13023 @itemx -fdump-rtl-@var{pass}=@var{filename}
13025 @opindex fdump-rtl-@var{pass}
13026 Says to make debugging dumps during compilation at times specified by
13027 @var{letters}. This is used for debugging the RTL-based passes of the
13028 compiler. The file names for most of the dumps are made by appending
13029 a pass number and a word to the @var{dumpname}, and the files are
13030 created in the directory of the output file. In case of
13031 @option{=@var{filename}} option, the dump is output on the given file
13032 instead of the pass numbered dump files. Note that the pass number is
13033 assigned as passes are registered into the pass manager. Most passes
13034 are registered in the order that they will execute and for these passes
13035 the number corresponds to the pass execution order. However, passes
13036 registered by plugins, passes specific to compilation targets, or
13037 passes that are otherwise registered after all the other passes are
13038 numbered higher than a pass named "final", even if they are executed
13039 earlier. @var{dumpname} is generated from the name of the output
13040 file if explicitly specified and not an executable, otherwise it is
13041 the basename of the source file.
13043 Some @option{-d@var{letters}} switches have different meaning when
13044 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
13045 for information about preprocessor-specific dump options.
13047 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
13048 @option{-d} option @var{letters}. Here are the possible
13049 letters for use in @var{pass} and @var{letters}, and their meanings:
13053 @item -fdump-rtl-alignments
13054 @opindex fdump-rtl-alignments
13055 Dump after branch alignments have been computed.
13057 @item -fdump-rtl-asmcons
13058 @opindex fdump-rtl-asmcons
13059 Dump after fixing rtl statements that have unsatisfied in/out constraints.
13061 @item -fdump-rtl-auto_inc_dec
13062 @opindex fdump-rtl-auto_inc_dec
13063 Dump after auto-inc-dec discovery. This pass is only run on
13064 architectures that have auto inc or auto dec instructions.
13066 @item -fdump-rtl-barriers
13067 @opindex fdump-rtl-barriers
13068 Dump after cleaning up the barrier instructions.
13070 @item -fdump-rtl-bbpart
13071 @opindex fdump-rtl-bbpart
13072 Dump after partitioning hot and cold basic blocks.
13074 @item -fdump-rtl-bbro
13075 @opindex fdump-rtl-bbro
13076 Dump after block reordering.
13078 @item -fdump-rtl-btl1
13079 @itemx -fdump-rtl-btl2
13080 @opindex fdump-rtl-btl2
13081 @opindex fdump-rtl-btl2
13082 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
13083 after the two branch
13084 target load optimization passes.
13086 @item -fdump-rtl-bypass
13087 @opindex fdump-rtl-bypass
13088 Dump after jump bypassing and control flow optimizations.
13090 @item -fdump-rtl-combine
13091 @opindex fdump-rtl-combine
13092 Dump after the RTL instruction combination pass.
13094 @item -fdump-rtl-compgotos
13095 @opindex fdump-rtl-compgotos
13096 Dump after duplicating the computed gotos.
13098 @item -fdump-rtl-ce1
13099 @itemx -fdump-rtl-ce2
13100 @itemx -fdump-rtl-ce3
13101 @opindex fdump-rtl-ce1
13102 @opindex fdump-rtl-ce2
13103 @opindex fdump-rtl-ce3
13104 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
13105 @option{-fdump-rtl-ce3} enable dumping after the three
13106 if conversion passes.
13108 @item -fdump-rtl-cprop_hardreg
13109 @opindex fdump-rtl-cprop_hardreg
13110 Dump after hard register copy propagation.
13112 @item -fdump-rtl-csa
13113 @opindex fdump-rtl-csa
13114 Dump after combining stack adjustments.
13116 @item -fdump-rtl-cse1
13117 @itemx -fdump-rtl-cse2
13118 @opindex fdump-rtl-cse1
13119 @opindex fdump-rtl-cse2
13120 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
13121 the two common subexpression elimination passes.
13123 @item -fdump-rtl-dce
13124 @opindex fdump-rtl-dce
13125 Dump after the standalone dead code elimination passes.
13127 @item -fdump-rtl-dbr
13128 @opindex fdump-rtl-dbr
13129 Dump after delayed branch scheduling.
13131 @item -fdump-rtl-dce1
13132 @itemx -fdump-rtl-dce2
13133 @opindex fdump-rtl-dce1
13134 @opindex fdump-rtl-dce2
13135 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
13136 the two dead store elimination passes.
13138 @item -fdump-rtl-eh
13139 @opindex fdump-rtl-eh
13140 Dump after finalization of EH handling code.
13142 @item -fdump-rtl-eh_ranges
13143 @opindex fdump-rtl-eh_ranges
13144 Dump after conversion of EH handling range regions.
13146 @item -fdump-rtl-expand
13147 @opindex fdump-rtl-expand
13148 Dump after RTL generation.
13150 @item -fdump-rtl-fwprop1
13151 @itemx -fdump-rtl-fwprop2
13152 @opindex fdump-rtl-fwprop1
13153 @opindex fdump-rtl-fwprop2
13154 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
13155 dumping after the two forward propagation passes.
13157 @item -fdump-rtl-gcse1
13158 @itemx -fdump-rtl-gcse2
13159 @opindex fdump-rtl-gcse1
13160 @opindex fdump-rtl-gcse2
13161 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
13162 after global common subexpression elimination.
13164 @item -fdump-rtl-init-regs
13165 @opindex fdump-rtl-init-regs
13166 Dump after the initialization of the registers.
13168 @item -fdump-rtl-initvals
13169 @opindex fdump-rtl-initvals
13170 Dump after the computation of the initial value sets.
13172 @item -fdump-rtl-into_cfglayout
13173 @opindex fdump-rtl-into_cfglayout
13174 Dump after converting to cfglayout mode.
13176 @item -fdump-rtl-ira
13177 @opindex fdump-rtl-ira
13178 Dump after iterated register allocation.
13180 @item -fdump-rtl-jump
13181 @opindex fdump-rtl-jump
13182 Dump after the second jump optimization.
13184 @item -fdump-rtl-loop2
13185 @opindex fdump-rtl-loop2
13186 @option{-fdump-rtl-loop2} enables dumping after the rtl
13187 loop optimization passes.
13189 @item -fdump-rtl-mach
13190 @opindex fdump-rtl-mach
13191 Dump after performing the machine dependent reorganization pass, if that
13194 @item -fdump-rtl-mode_sw
13195 @opindex fdump-rtl-mode_sw
13196 Dump after removing redundant mode switches.
13198 @item -fdump-rtl-rnreg
13199 @opindex fdump-rtl-rnreg
13200 Dump after register renumbering.
13202 @item -fdump-rtl-outof_cfglayout
13203 @opindex fdump-rtl-outof_cfglayout
13204 Dump after converting from cfglayout mode.
13206 @item -fdump-rtl-peephole2
13207 @opindex fdump-rtl-peephole2
13208 Dump after the peephole pass.
13210 @item -fdump-rtl-postreload
13211 @opindex fdump-rtl-postreload
13212 Dump after post-reload optimizations.
13214 @item -fdump-rtl-pro_and_epilogue
13215 @opindex fdump-rtl-pro_and_epilogue
13216 Dump after generating the function prologues and epilogues.
13218 @item -fdump-rtl-sched1
13219 @itemx -fdump-rtl-sched2
13220 @opindex fdump-rtl-sched1
13221 @opindex fdump-rtl-sched2
13222 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
13223 after the basic block scheduling passes.
13225 @item -fdump-rtl-ree
13226 @opindex fdump-rtl-ree
13227 Dump after sign/zero extension elimination.
13229 @item -fdump-rtl-seqabstr
13230 @opindex fdump-rtl-seqabstr
13231 Dump after common sequence discovery.
13233 @item -fdump-rtl-shorten
13234 @opindex fdump-rtl-shorten
13235 Dump after shortening branches.
13237 @item -fdump-rtl-sibling
13238 @opindex fdump-rtl-sibling
13239 Dump after sibling call optimizations.
13241 @item -fdump-rtl-split1
13242 @itemx -fdump-rtl-split2
13243 @itemx -fdump-rtl-split3
13244 @itemx -fdump-rtl-split4
13245 @itemx -fdump-rtl-split5
13246 @opindex fdump-rtl-split1
13247 @opindex fdump-rtl-split2
13248 @opindex fdump-rtl-split3
13249 @opindex fdump-rtl-split4
13250 @opindex fdump-rtl-split5
13251 These options enable dumping after five rounds of
13252 instruction splitting.
13254 @item -fdump-rtl-sms
13255 @opindex fdump-rtl-sms
13256 Dump after modulo scheduling. This pass is only run on some
13259 @item -fdump-rtl-stack
13260 @opindex fdump-rtl-stack
13261 Dump after conversion from GCC's ``flat register file'' registers to the
13262 x87's stack-like registers. This pass is only run on x86 variants.
13264 @item -fdump-rtl-subreg1
13265 @itemx -fdump-rtl-subreg2
13266 @opindex fdump-rtl-subreg1
13267 @opindex fdump-rtl-subreg2
13268 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
13269 the two subreg expansion passes.
13271 @item -fdump-rtl-unshare
13272 @opindex fdump-rtl-unshare
13273 Dump after all rtl has been unshared.
13275 @item -fdump-rtl-vartrack
13276 @opindex fdump-rtl-vartrack
13277 Dump after variable tracking.
13279 @item -fdump-rtl-vregs
13280 @opindex fdump-rtl-vregs
13281 Dump after converting virtual registers to hard registers.
13283 @item -fdump-rtl-web
13284 @opindex fdump-rtl-web
13285 Dump after live range splitting.
13287 @item -fdump-rtl-regclass
13288 @itemx -fdump-rtl-subregs_of_mode_init
13289 @itemx -fdump-rtl-subregs_of_mode_finish
13290 @itemx -fdump-rtl-dfinit
13291 @itemx -fdump-rtl-dfinish
13292 @opindex fdump-rtl-regclass
13293 @opindex fdump-rtl-subregs_of_mode_init
13294 @opindex fdump-rtl-subregs_of_mode_finish
13295 @opindex fdump-rtl-dfinit
13296 @opindex fdump-rtl-dfinish
13297 These dumps are defined but always produce empty files.
13300 @itemx -fdump-rtl-all
13302 @opindex fdump-rtl-all
13303 Produce all the dumps listed above.
13307 Annotate the assembler output with miscellaneous debugging information.
13311 Dump all macro definitions, at the end of preprocessing, in addition to
13316 Produce a core dump whenever an error occurs.
13320 Annotate the assembler output with a comment indicating which
13321 pattern and alternative is used. The length of each instruction is
13326 Dump the RTL in the assembler output as a comment before each instruction.
13327 Also turns on @option{-dp} annotation.
13331 Just generate RTL for a function instead of compiling it. Usually used
13332 with @option{-fdump-rtl-expand}.
13335 @item -fdump-noaddr
13336 @opindex fdump-noaddr
13337 When doing debugging dumps, suppress address output. This makes it more
13338 feasible to use diff on debugging dumps for compiler invocations with
13339 different compiler binaries and/or different
13340 text / bss / data / heap / stack / dso start locations.
13343 @opindex freport-bug
13344 Collect and dump debug information into a temporary file if an
13345 internal compiler error (ICE) occurs.
13347 @item -fdump-unnumbered
13348 @opindex fdump-unnumbered
13349 When doing debugging dumps, suppress instruction numbers and address output.
13350 This makes it more feasible to use diff on debugging dumps for compiler
13351 invocations with different options, in particular with and without
13354 @item -fdump-unnumbered-links
13355 @opindex fdump-unnumbered-links
13356 When doing debugging dumps (see @option{-d} option above), suppress
13357 instruction numbers for the links to the previous and next instructions
13360 @item -fdump-ipa-@var{switch}
13362 Control the dumping at various stages of inter-procedural analysis
13363 language tree to a file. The file name is generated by appending a
13364 switch specific suffix to the source file name, and the file is created
13365 in the same directory as the output file. The following dumps are
13370 Enables all inter-procedural analysis dumps.
13373 Dumps information about call-graph optimization, unused function removal,
13374 and inlining decisions.
13377 Dump after function inlining.
13381 @item -fdump-lang-all
13382 @itemx -fdump-lang-@var{switch}
13383 @itemx -fdump-lang-@var{switch}-@var{options}
13384 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
13385 @opindex fdump-lang-all
13386 @opindex fdump-lang
13387 Control the dumping of language-specific information. The @var{options}
13388 and @var{filename} portions behave as described in the
13389 @option{-fdump-tree} option. The following @var{switch} values are
13395 Enable all language-specific dumps.
13398 Dump class hierarchy information. Virtual table information is emitted
13399 unless '@option{slim}' is specified. This option is applicable to C++ only.
13402 Dump the raw internal tree data. This option is applicable to C++ only.
13406 @item -fdump-passes
13407 @opindex fdump-passes
13408 Print on @file{stderr} the list of optimization passes that are turned
13409 on and off by the current command-line options.
13411 @item -fdump-statistics-@var{option}
13412 @opindex fdump-statistics
13413 Enable and control dumping of pass statistics in a separate file. The
13414 file name is generated by appending a suffix ending in
13415 @samp{.statistics} to the source file name, and the file is created in
13416 the same directory as the output file. If the @samp{-@var{option}}
13417 form is used, @samp{-stats} causes counters to be summed over the
13418 whole compilation unit while @samp{-details} dumps every event as
13419 the passes generate them. The default with no option is to sum
13420 counters for each function compiled.
13422 @item -fdump-tree-all
13423 @itemx -fdump-tree-@var{switch}
13424 @itemx -fdump-tree-@var{switch}-@var{options}
13425 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
13426 @opindex fdump-tree-all
13427 @opindex fdump-tree
13428 Control the dumping at various stages of processing the intermediate
13429 language tree to a file. The file name is generated by appending a
13430 switch-specific suffix to the source file name, and the file is
13431 created in the same directory as the output file. In case of
13432 @option{=@var{filename}} option, the dump is output on the given file
13433 instead of the auto named dump files. If the @samp{-@var{options}}
13434 form is used, @var{options} is a list of @samp{-} separated options
13435 which control the details of the dump. Not all options are applicable
13436 to all dumps; those that are not meaningful are ignored. The
13437 following options are available
13441 Print the address of each node. Usually this is not meaningful as it
13442 changes according to the environment and source file. Its primary use
13443 is for tying up a dump file with a debug environment.
13445 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
13446 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
13447 use working backward from mangled names in the assembly file.
13449 When dumping front-end intermediate representations, inhibit dumping
13450 of members of a scope or body of a function merely because that scope
13451 has been reached. Only dump such items when they are directly reachable
13452 by some other path.
13454 When dumping pretty-printed trees, this option inhibits dumping the
13455 bodies of control structures.
13457 When dumping RTL, print the RTL in slim (condensed) form instead of
13458 the default LISP-like representation.
13460 Print a raw representation of the tree. By default, trees are
13461 pretty-printed into a C-like representation.
13463 Enable more detailed dumps (not honored by every dump option). Also
13464 include information from the optimization passes.
13466 Enable dumping various statistics about the pass (not honored by every dump
13469 Enable showing basic block boundaries (disabled in raw dumps).
13471 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13472 dump a representation of the control flow graph suitable for viewing with
13473 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13474 the file is pretty-printed as a subgraph, so that GraphViz can render them
13475 all in a single plot.
13477 This option currently only works for RTL dumps, and the RTL is always
13478 dumped in slim form.
13480 Enable showing virtual operands for every statement.
13482 Enable showing line numbers for statements.
13484 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13486 Enable showing the tree dump for each statement.
13488 Enable showing the EH region number holding each statement.
13490 Enable showing scalar evolution analysis details.
13492 Enable showing optimization information (only available in certain
13495 Enable showing missed optimization information (only available in certain
13498 Enable other detailed optimization information (only available in
13500 @item =@var{filename}
13501 Instead of an auto named dump file, output into the given file
13502 name. The file names @file{stdout} and @file{stderr} are treated
13503 specially and are considered already open standard streams. For
13507 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13508 -fdump-tree-pre=/dev/stderr file.c
13511 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13512 output on to @file{stderr}. If two conflicting dump filenames are
13513 given for the same pass, then the latter option overrides the earlier
13517 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13518 and @option{lineno}.
13521 Turn on all optimization options, i.e., @option{optimized},
13522 @option{missed}, and @option{note}.
13525 To determine what tree dumps are available or find the dump for a pass
13526 of interest follow the steps below.
13530 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
13531 look for a code that corresponds to the pass you are interested in.
13532 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
13533 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
13534 The number at the end distinguishes distinct invocations of the same pass.
13536 To enable the creation of the dump file, append the pass code to
13537 the @option{-fdump-} option prefix and invoke GCC with it. For example,
13538 to enable the dump from the Early Value Range Propagation pass, invoke
13539 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
13540 specify the name of the dump file. If you don't specify one, GCC
13541 creates as described below.
13543 Find the pass dump in a file whose name is composed of three components
13544 separated by a period: the name of the source file GCC was invoked to
13545 compile, a numeric suffix indicating the pass number followed by the
13546 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
13547 and finally the pass code. For example, the Early VRP pass dump might
13548 be in a file named @file{myfile.c.038t.evrp} in the current working
13549 directory. Note that the numeric codes are not stable and may change
13550 from one version of GCC to another.
13554 @itemx -fopt-info-@var{options}
13555 @itemx -fopt-info-@var{options}=@var{filename}
13557 Controls optimization dumps from various optimization passes. If the
13558 @samp{-@var{options}} form is used, @var{options} is a list of
13559 @samp{-} separated option keywords to select the dump details and
13562 The @var{options} can be divided into two groups: options describing the
13563 verbosity of the dump, and options describing which optimizations
13564 should be included. The options from both the groups can be freely
13565 mixed as they are non-overlapping. However, in case of any conflicts,
13566 the later options override the earlier options on the command
13569 The following options control the dump verbosity:
13573 Print information when an optimization is successfully applied. It is
13574 up to a pass to decide which information is relevant. For example, the
13575 vectorizer passes print the source location of loops which are
13576 successfully vectorized.
13578 Print information about missed optimizations. Individual passes
13579 control which information to include in the output.
13581 Print verbose information about optimizations, such as certain
13582 transformations, more detailed messages about decisions etc.
13584 Print detailed optimization information. This includes
13585 @samp{optimized}, @samp{missed}, and @samp{note}.
13588 One or more of the following option keywords can be used to describe a
13589 group of optimizations:
13593 Enable dumps from all interprocedural optimizations.
13595 Enable dumps from all loop optimizations.
13597 Enable dumps from all inlining optimizations.
13599 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
13601 Enable dumps from all vectorization optimizations.
13603 Enable dumps from all optimizations. This is a superset of
13604 the optimization groups listed above.
13607 If @var{options} is
13608 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13609 info about successful optimizations from all the passes.
13611 If the @var{filename} is provided, then the dumps from all the
13612 applicable optimizations are concatenated into the @var{filename}.
13613 Otherwise the dump is output onto @file{stderr}. Though multiple
13614 @option{-fopt-info} options are accepted, only one of them can include
13615 a @var{filename}. If other filenames are provided then all but the
13616 first such option are ignored.
13618 Note that the output @var{filename} is overwritten
13619 in case of multiple translation units. If a combined output from
13620 multiple translation units is desired, @file{stderr} should be used
13623 In the following example, the optimization info is output to
13632 gcc -O3 -fopt-info-missed=missed.all
13636 outputs missed optimization report from all the passes into
13637 @file{missed.all}, and this one:
13640 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13644 prints information about missed optimization opportunities from
13645 vectorization passes on @file{stderr}.
13646 Note that @option{-fopt-info-vec-missed} is equivalent to
13647 @option{-fopt-info-missed-vec}. The order of the optimization group
13648 names and message types listed after @option{-fopt-info} does not matter.
13650 As another example,
13652 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13656 outputs information about missed optimizations as well as
13657 optimized locations from all the inlining passes into
13663 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13667 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13668 in conflict since only one output file is allowed. In this case, only
13669 the first option takes effect and the subsequent options are
13670 ignored. Thus only @file{vec.miss} is produced which contains
13671 dumps from the vectorizer about missed opportunities.
13673 @item -fsched-verbose=@var{n}
13674 @opindex fsched-verbose
13675 On targets that use instruction scheduling, this option controls the
13676 amount of debugging output the scheduler prints to the dump files.
13678 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13679 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13680 For @var{n} greater than one, it also output basic block probabilities,
13681 detailed ready list information and unit/insn info. For @var{n} greater
13682 than two, it includes RTL at abort point, control-flow and regions info.
13683 And for @var{n} over four, @option{-fsched-verbose} also includes
13688 @item -fenable-@var{kind}-@var{pass}
13689 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13693 This is a set of options that are used to explicitly disable/enable
13694 optimization passes. These options are intended for use for debugging GCC.
13695 Compiler users should use regular options for enabling/disabling
13700 @item -fdisable-ipa-@var{pass}
13701 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13702 statically invoked in the compiler multiple times, the pass name should be
13703 appended with a sequential number starting from 1.
13705 @item -fdisable-rtl-@var{pass}
13706 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13707 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13708 statically invoked in the compiler multiple times, the pass name should be
13709 appended with a sequential number starting from 1. @var{range-list} is a
13710 comma-separated list of function ranges or assembler names. Each range is a number
13711 pair separated by a colon. The range is inclusive in both ends. If the range
13712 is trivial, the number pair can be simplified as a single number. If the
13713 function's call graph node's @var{uid} falls within one of the specified ranges,
13714 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13715 function header of a dump file, and the pass names can be dumped by using
13716 option @option{-fdump-passes}.
13718 @item -fdisable-tree-@var{pass}
13719 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13720 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13723 @item -fenable-ipa-@var{pass}
13724 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13725 statically invoked in the compiler multiple times, the pass name should be
13726 appended with a sequential number starting from 1.
13728 @item -fenable-rtl-@var{pass}
13729 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13730 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13731 description and examples.
13733 @item -fenable-tree-@var{pass}
13734 @itemx -fenable-tree-@var{pass}=@var{range-list}
13735 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13736 of option arguments.
13740 Here are some examples showing uses of these options.
13744 # disable ccp1 for all functions
13745 -fdisable-tree-ccp1
13746 # disable complete unroll for function whose cgraph node uid is 1
13747 -fenable-tree-cunroll=1
13748 # disable gcse2 for functions at the following ranges [1,1],
13749 # [300,400], and [400,1000]
13750 # disable gcse2 for functions foo and foo2
13751 -fdisable-rtl-gcse2=foo,foo2
13752 # disable early inlining
13753 -fdisable-tree-einline
13754 # disable ipa inlining
13755 -fdisable-ipa-inline
13756 # enable tree full unroll
13757 -fenable-tree-unroll
13762 @itemx -fchecking=@var{n}
13764 @opindex fno-checking
13765 Enable internal consistency checking. The default depends on
13766 the compiler configuration. @option{-fchecking=2} enables further
13767 internal consistency checking that might affect code generation.
13769 @item -frandom-seed=@var{string}
13770 @opindex frandom-seed
13771 This option provides a seed that GCC uses in place of
13772 random numbers in generating certain symbol names
13773 that have to be different in every compiled file. It is also used to
13774 place unique stamps in coverage data files and the object files that
13775 produce them. You can use the @option{-frandom-seed} option to produce
13776 reproducibly identical object files.
13778 The @var{string} can either be a number (decimal, octal or hex) or an
13779 arbitrary string (in which case it's converted to a number by
13782 The @var{string} should be different for every file you compile.
13785 @itemx -save-temps=cwd
13786 @opindex save-temps
13787 Store the usual ``temporary'' intermediate files permanently; place them
13788 in the current directory and name them based on the source file. Thus,
13789 compiling @file{foo.c} with @option{-c -save-temps} produces files
13790 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13791 preprocessed @file{foo.i} output file even though the compiler now
13792 normally uses an integrated preprocessor.
13794 When used in combination with the @option{-x} command-line option,
13795 @option{-save-temps} is sensible enough to avoid over writing an
13796 input source file with the same extension as an intermediate file.
13797 The corresponding intermediate file may be obtained by renaming the
13798 source file before using @option{-save-temps}.
13800 If you invoke GCC in parallel, compiling several different source
13801 files that share a common base name in different subdirectories or the
13802 same source file compiled for multiple output destinations, it is
13803 likely that the different parallel compilers will interfere with each
13804 other, and overwrite the temporary files. For instance:
13807 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13808 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13811 may result in @file{foo.i} and @file{foo.o} being written to
13812 simultaneously by both compilers.
13814 @item -save-temps=obj
13815 @opindex save-temps=obj
13816 Store the usual ``temporary'' intermediate files permanently. If the
13817 @option{-o} option is used, the temporary files are based on the
13818 object file. If the @option{-o} option is not used, the
13819 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13824 gcc -save-temps=obj -c foo.c
13825 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13826 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13830 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13831 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13832 @file{dir2/yfoobar.o}.
13834 @item -time@r{[}=@var{file}@r{]}
13836 Report the CPU time taken by each subprocess in the compilation
13837 sequence. For C source files, this is the compiler proper and assembler
13838 (plus the linker if linking is done).
13840 Without the specification of an output file, the output looks like this:
13847 The first number on each line is the ``user time'', that is time spent
13848 executing the program itself. The second number is ``system time'',
13849 time spent executing operating system routines on behalf of the program.
13850 Both numbers are in seconds.
13852 With the specification of an output file, the output is appended to the
13853 named file, and it looks like this:
13856 0.12 0.01 cc1 @var{options}
13857 0.00 0.01 as @var{options}
13860 The ``user time'' and the ``system time'' are moved before the program
13861 name, and the options passed to the program are displayed, so that one
13862 can later tell what file was being compiled, and with which options.
13864 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13865 @opindex fdump-final-insns
13866 Dump the final internal representation (RTL) to @var{file}. If the
13867 optional argument is omitted (or if @var{file} is @code{.}), the name
13868 of the dump file is determined by appending @code{.gkd} to the
13869 compilation output file name.
13871 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13872 @opindex fcompare-debug
13873 @opindex fno-compare-debug
13874 If no error occurs during compilation, run the compiler a second time,
13875 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13876 passed to the second compilation. Dump the final internal
13877 representation in both compilations, and print an error if they differ.
13879 If the equal sign is omitted, the default @option{-gtoggle} is used.
13881 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13882 and nonzero, implicitly enables @option{-fcompare-debug}. If
13883 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13884 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13887 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13888 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13889 of the final representation and the second compilation, preventing even
13890 @env{GCC_COMPARE_DEBUG} from taking effect.
13892 To verify full coverage during @option{-fcompare-debug} testing, set
13893 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13894 which GCC rejects as an invalid option in any actual compilation
13895 (rather than preprocessing, assembly or linking). To get just a
13896 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13897 not overridden} will do.
13899 @item -fcompare-debug-second
13900 @opindex fcompare-debug-second
13901 This option is implicitly passed to the compiler for the second
13902 compilation requested by @option{-fcompare-debug}, along with options to
13903 silence warnings, and omitting other options that would cause
13904 side-effect compiler outputs to files or to the standard output. Dump
13905 files and preserved temporary files are renamed so as to contain the
13906 @code{.gk} additional extension during the second compilation, to avoid
13907 overwriting those generated by the first.
13909 When this option is passed to the compiler driver, it causes the
13910 @emph{first} compilation to be skipped, which makes it useful for little
13911 other than debugging the compiler proper.
13915 Turn off generation of debug info, if leaving out this option
13916 generates it, or turn it on at level 2 otherwise. The position of this
13917 argument in the command line does not matter; it takes effect after all
13918 other options are processed, and it does so only once, no matter how
13919 many times it is given. This is mainly intended to be used with
13920 @option{-fcompare-debug}.
13922 @item -fvar-tracking-assignments-toggle
13923 @opindex fvar-tracking-assignments-toggle
13924 @opindex fno-var-tracking-assignments-toggle
13925 Toggle @option{-fvar-tracking-assignments}, in the same way that
13926 @option{-gtoggle} toggles @option{-g}.
13930 Makes the compiler print out each function name as it is compiled, and
13931 print some statistics about each pass when it finishes.
13933 @item -ftime-report
13934 @opindex ftime-report
13935 Makes the compiler print some statistics about the time consumed by each
13936 pass when it finishes.
13938 @item -ftime-report-details
13939 @opindex ftime-report-details
13940 Record the time consumed by infrastructure parts separately for each pass.
13942 @item -fira-verbose=@var{n}
13943 @opindex fira-verbose
13944 Control the verbosity of the dump file for the integrated register allocator.
13945 The default value is 5. If the value @var{n} is greater or equal to 10,
13946 the dump output is sent to stderr using the same format as @var{n} minus 10.
13949 @opindex flto-report
13950 Prints a report with internal details on the workings of the link-time
13951 optimizer. The contents of this report vary from version to version.
13952 It is meant to be useful to GCC developers when processing object
13953 files in LTO mode (via @option{-flto}).
13955 Disabled by default.
13957 @item -flto-report-wpa
13958 @opindex flto-report-wpa
13959 Like @option{-flto-report}, but only print for the WPA phase of Link
13963 @opindex fmem-report
13964 Makes the compiler print some statistics about permanent memory
13965 allocation when it finishes.
13967 @item -fmem-report-wpa
13968 @opindex fmem-report-wpa
13969 Makes the compiler print some statistics about permanent memory
13970 allocation for the WPA phase only.
13972 @item -fpre-ipa-mem-report
13973 @opindex fpre-ipa-mem-report
13974 @item -fpost-ipa-mem-report
13975 @opindex fpost-ipa-mem-report
13976 Makes the compiler print some statistics about permanent memory
13977 allocation before or after interprocedural optimization.
13979 @item -fprofile-report
13980 @opindex fprofile-report
13981 Makes the compiler print some statistics about consistency of the
13982 (estimated) profile and effect of individual passes.
13984 @item -fstack-usage
13985 @opindex fstack-usage
13986 Makes the compiler output stack usage information for the program, on a
13987 per-function basis. The filename for the dump is made by appending
13988 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13989 the output file, if explicitly specified and it is not an executable,
13990 otherwise it is the basename of the source file. An entry is made up
13995 The name of the function.
13999 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
14002 The qualifier @code{static} means that the function manipulates the stack
14003 statically: a fixed number of bytes are allocated for the frame on function
14004 entry and released on function exit; no stack adjustments are otherwise made
14005 in the function. The second field is this fixed number of bytes.
14007 The qualifier @code{dynamic} means that the function manipulates the stack
14008 dynamically: in addition to the static allocation described above, stack
14009 adjustments are made in the body of the function, for example to push/pop
14010 arguments around function calls. If the qualifier @code{bounded} is also
14011 present, the amount of these adjustments is bounded at compile time and
14012 the second field is an upper bound of the total amount of stack used by
14013 the function. If it is not present, the amount of these adjustments is
14014 not bounded at compile time and the second field only represents the
14019 Emit statistics about front-end processing at the end of the compilation.
14020 This option is supported only by the C++ front end, and
14021 the information is generally only useful to the G++ development team.
14023 @item -fdbg-cnt-list
14024 @opindex fdbg-cnt-list
14025 Print the name and the counter upper bound for all debug counters.
14028 @item -fdbg-cnt=@var{counter-value-list}
14030 Set the internal debug counter upper bound. @var{counter-value-list}
14031 is a comma-separated list of @var{name}:@var{value} pairs
14032 which sets the upper bound of each debug counter @var{name} to @var{value}.
14033 All debug counters have the initial upper bound of @code{UINT_MAX};
14034 thus @code{dbg_cnt} returns true always unless the upper bound
14035 is set by this option.
14036 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
14037 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
14039 @item -print-file-name=@var{library}
14040 @opindex print-file-name
14041 Print the full absolute name of the library file @var{library} that
14042 would be used when linking---and don't do anything else. With this
14043 option, GCC does not compile or link anything; it just prints the
14046 @item -print-multi-directory
14047 @opindex print-multi-directory
14048 Print the directory name corresponding to the multilib selected by any
14049 other switches present in the command line. This directory is supposed
14050 to exist in @env{GCC_EXEC_PREFIX}.
14052 @item -print-multi-lib
14053 @opindex print-multi-lib
14054 Print the mapping from multilib directory names to compiler switches
14055 that enable them. The directory name is separated from the switches by
14056 @samp{;}, and each switch starts with an @samp{@@} instead of the
14057 @samp{-}, without spaces between multiple switches. This is supposed to
14058 ease shell processing.
14060 @item -print-multi-os-directory
14061 @opindex print-multi-os-directory
14062 Print the path to OS libraries for the selected
14063 multilib, relative to some @file{lib} subdirectory. If OS libraries are
14064 present in the @file{lib} subdirectory and no multilibs are used, this is
14065 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
14066 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
14067 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
14068 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
14070 @item -print-multiarch
14071 @opindex print-multiarch
14072 Print the path to OS libraries for the selected multiarch,
14073 relative to some @file{lib} subdirectory.
14075 @item -print-prog-name=@var{program}
14076 @opindex print-prog-name
14077 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
14079 @item -print-libgcc-file-name
14080 @opindex print-libgcc-file-name
14081 Same as @option{-print-file-name=libgcc.a}.
14083 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
14084 but you do want to link with @file{libgcc.a}. You can do:
14087 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
14090 @item -print-search-dirs
14091 @opindex print-search-dirs
14092 Print the name of the configured installation directory and a list of
14093 program and library directories @command{gcc} searches---and don't do anything else.
14095 This is useful when @command{gcc} prints the error message
14096 @samp{installation problem, cannot exec cpp0: No such file or directory}.
14097 To resolve this you either need to put @file{cpp0} and the other compiler
14098 components where @command{gcc} expects to find them, or you can set the environment
14099 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
14100 Don't forget the trailing @samp{/}.
14101 @xref{Environment Variables}.
14103 @item -print-sysroot
14104 @opindex print-sysroot
14105 Print the target sysroot directory that is used during
14106 compilation. This is the target sysroot specified either at configure
14107 time or using the @option{--sysroot} option, possibly with an extra
14108 suffix that depends on compilation options. If no target sysroot is
14109 specified, the option prints nothing.
14111 @item -print-sysroot-headers-suffix
14112 @opindex print-sysroot-headers-suffix
14113 Print the suffix added to the target sysroot when searching for
14114 headers, or give an error if the compiler is not configured with such
14115 a suffix---and don't do anything else.
14118 @opindex dumpmachine
14119 Print the compiler's target machine (for example,
14120 @samp{i686-pc-linux-gnu})---and don't do anything else.
14123 @opindex dumpversion
14124 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
14125 anything else. This is the compiler version used in filesystem paths,
14126 specs, can be depending on how the compiler has been configured just
14127 a single number (major version), two numbers separated by dot (major and
14128 minor version) or three numbers separated by dots (major, minor and patchlevel
14131 @item -dumpfullversion
14132 @opindex dumpfullversion
14133 Print the full compiler version, always 3 numbers separated by dots,
14134 major, minor and patchlevel version.
14138 Print the compiler's built-in specs---and don't do anything else. (This
14139 is used when GCC itself is being built.) @xref{Spec Files}.
14142 @node Submodel Options
14143 @section Machine-Dependent Options
14144 @cindex submodel options
14145 @cindex specifying hardware config
14146 @cindex hardware models and configurations, specifying
14147 @cindex target-dependent options
14148 @cindex machine-dependent options
14150 Each target machine supported by GCC can have its own options---for
14151 example, to allow you to compile for a particular processor variant or
14152 ABI, or to control optimizations specific to that machine. By
14153 convention, the names of machine-specific options start with
14156 Some configurations of the compiler also support additional target-specific
14157 options, usually for compatibility with other compilers on the same
14160 @c This list is ordered alphanumerically by subsection name.
14161 @c It should be the same order and spelling as these options are listed
14162 @c in Machine Dependent Options
14165 * AArch64 Options::
14166 * Adapteva Epiphany Options::
14170 * Blackfin Options::
14175 * DEC Alpha Options::
14179 * GNU/Linux Options::
14189 * MicroBlaze Options::
14192 * MN10300 Options::
14196 * Nios II Options::
14197 * Nvidia PTX Options::
14199 * picoChip Options::
14200 * PowerPC Options::
14203 * RS/6000 and PowerPC Options::
14205 * S/390 and zSeries Options::
14208 * Solaris 2 Options::
14211 * System V Options::
14212 * TILE-Gx Options::
14213 * TILEPro Options::
14218 * VxWorks Options::
14220 * x86 Windows Options::
14221 * Xstormy16 Options::
14223 * zSeries Options::
14226 @node AArch64 Options
14227 @subsection AArch64 Options
14228 @cindex AArch64 Options
14230 These options are defined for AArch64 implementations:
14234 @item -mabi=@var{name}
14236 Generate code for the specified data model. Permissible values
14237 are @samp{ilp32} for SysV-like data model where int, long int and pointers
14238 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
14239 but long int and pointers are 64 bits.
14241 The default depends on the specific target configuration. Note that
14242 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
14243 entire program with the same ABI, and link with a compatible set of libraries.
14246 @opindex mbig-endian
14247 Generate big-endian code. This is the default when GCC is configured for an
14248 @samp{aarch64_be-*-*} target.
14250 @item -mgeneral-regs-only
14251 @opindex mgeneral-regs-only
14252 Generate code which uses only the general-purpose registers. This will prevent
14253 the compiler from using floating-point and Advanced SIMD registers but will not
14254 impose any restrictions on the assembler.
14256 @item -mlittle-endian
14257 @opindex mlittle-endian
14258 Generate little-endian code. This is the default when GCC is configured for an
14259 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
14261 @item -mcmodel=tiny
14262 @opindex mcmodel=tiny
14263 Generate code for the tiny code model. The program and its statically defined
14264 symbols must be within 1MB of each other. Programs can be statically or
14265 dynamically linked.
14267 @item -mcmodel=small
14268 @opindex mcmodel=small
14269 Generate code for the small code model. The program and its statically defined
14270 symbols must be within 4GB of each other. Programs can be statically or
14271 dynamically linked. This is the default code model.
14273 @item -mcmodel=large
14274 @opindex mcmodel=large
14275 Generate code for the large code model. This makes no assumptions about
14276 addresses and sizes of sections. Programs can be statically linked only.
14278 @item -mstrict-align
14279 @opindex mstrict-align
14280 Avoid generating memory accesses that may not be aligned on a natural object
14281 boundary as described in the architecture specification.
14283 @item -momit-leaf-frame-pointer
14284 @itemx -mno-omit-leaf-frame-pointer
14285 @opindex momit-leaf-frame-pointer
14286 @opindex mno-omit-leaf-frame-pointer
14287 Omit or keep the frame pointer in leaf functions. The former behavior is the
14290 @item -mtls-dialect=desc
14291 @opindex mtls-dialect=desc
14292 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
14293 of TLS variables. This is the default.
14295 @item -mtls-dialect=traditional
14296 @opindex mtls-dialect=traditional
14297 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
14300 @item -mtls-size=@var{size}
14302 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
14303 This option requires binutils 2.26 or newer.
14305 @item -mfix-cortex-a53-835769
14306 @itemx -mno-fix-cortex-a53-835769
14307 @opindex mfix-cortex-a53-835769
14308 @opindex mno-fix-cortex-a53-835769
14309 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
14310 This involves inserting a NOP instruction between memory instructions and
14311 64-bit integer multiply-accumulate instructions.
14313 @item -mfix-cortex-a53-843419
14314 @itemx -mno-fix-cortex-a53-843419
14315 @opindex mfix-cortex-a53-843419
14316 @opindex mno-fix-cortex-a53-843419
14317 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
14318 This erratum workaround is made at link time and this will only pass the
14319 corresponding flag to the linker.
14321 @item -mlow-precision-recip-sqrt
14322 @item -mno-low-precision-recip-sqrt
14323 @opindex mlow-precision-recip-sqrt
14324 @opindex mno-low-precision-recip-sqrt
14325 Enable or disable the reciprocal square root approximation.
14326 This option only has an effect if @option{-ffast-math} or
14327 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14328 precision of reciprocal square root results to about 16 bits for
14329 single precision and to 32 bits for double precision.
14331 @item -mlow-precision-sqrt
14332 @item -mno-low-precision-sqrt
14333 @opindex -mlow-precision-sqrt
14334 @opindex -mno-low-precision-sqrt
14335 Enable or disable the square root approximation.
14336 This option only has an effect if @option{-ffast-math} or
14337 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14338 precision of square root results to about 16 bits for
14339 single precision and to 32 bits for double precision.
14340 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
14342 @item -mlow-precision-div
14343 @item -mno-low-precision-div
14344 @opindex -mlow-precision-div
14345 @opindex -mno-low-precision-div
14346 Enable or disable the division approximation.
14347 This option only has an effect if @option{-ffast-math} or
14348 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14349 precision of division results to about 16 bits for
14350 single precision and to 32 bits for double precision.
14352 @item -march=@var{name}
14354 Specify the name of the target architecture and, optionally, one or
14355 more feature modifiers. This option has the form
14356 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
14358 The permissible values for @var{arch} are @samp{armv8-a},
14359 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @var{native}.
14361 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
14362 support for the ARMv8.3-A architecture extensions.
14364 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
14365 support for the ARMv8.2-A architecture extensions.
14367 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
14368 support for the ARMv8.1-A architecture extension. In particular, it
14369 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
14371 The value @samp{native} is available on native AArch64 GNU/Linux and
14372 causes the compiler to pick the architecture of the host system. This
14373 option has no effect if the compiler is unable to recognize the
14374 architecture of the host system,
14376 The permissible values for @var{feature} are listed in the sub-section
14377 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14378 Feature Modifiers}. Where conflicting feature modifiers are
14379 specified, the right-most feature is used.
14381 GCC uses @var{name} to determine what kind of instructions it can emit
14382 when generating assembly code. If @option{-march} is specified
14383 without either of @option{-mtune} or @option{-mcpu} also being
14384 specified, the code is tuned to perform well across a range of target
14385 processors implementing the target architecture.
14387 @item -mtune=@var{name}
14389 Specify the name of the target processor for which GCC should tune the
14390 performance of the code. Permissible values for this option are:
14391 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
14392 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
14393 @samp{exynos-m1}, @samp{falkor}, @samp{qdf24xx}, @samp{saphira},
14394 @samp{xgene1}, @samp{vulcan}, @samp{thunderx},
14395 @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
14396 @samp{thunderxt83}, @samp{thunderx2t99}, @samp{cortex-a57.cortex-a53},
14397 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
14398 @samp{cortex-a73.cortex-a53}, @samp{cortex-a75.cortex-a55},
14401 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14402 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
14403 @samp{cortex-a75.cortex-a55} specify that GCC should tune for a
14406 Additionally on native AArch64 GNU/Linux systems the value
14407 @samp{native} tunes performance to the host system. This option has no effect
14408 if the compiler is unable to recognize the processor of the host system.
14410 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
14411 are specified, the code is tuned to perform well across a range
14412 of target processors.
14414 This option cannot be suffixed by feature modifiers.
14416 @item -mcpu=@var{name}
14418 Specify the name of the target processor, optionally suffixed by one
14419 or more feature modifiers. This option has the form
14420 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
14421 the permissible values for @var{cpu} are the same as those available
14422 for @option{-mtune}. The permissible values for @var{feature} are
14423 documented in the sub-section on
14424 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14425 Feature Modifiers}. Where conflicting feature modifiers are
14426 specified, the right-most feature is used.
14428 GCC uses @var{name} to determine what kind of instructions it can emit when
14429 generating assembly code (as if by @option{-march}) and to determine
14430 the target processor for which to tune for performance (as if
14431 by @option{-mtune}). Where this option is used in conjunction
14432 with @option{-march} or @option{-mtune}, those options take precedence
14433 over the appropriate part of this option.
14435 @item -moverride=@var{string}
14437 Override tuning decisions made by the back-end in response to a
14438 @option{-mtune=} switch. The syntax, semantics, and accepted values
14439 for @var{string} in this option are not guaranteed to be consistent
14442 This option is only intended to be useful when developing GCC.
14444 @item -mpc-relative-literal-loads
14445 @itemx -mno-pc-relative-literal-loads
14446 @opindex mpc-relative-literal-loads
14447 @opindex mno-pc-relative-literal-loads
14448 Enable or disable PC-relative literal loads. With this option literal pools are
14449 accessed using a single instruction and emitted after each function. This
14450 limits the maximum size of functions to 1MB. This is enabled by default for
14451 @option{-mcmodel=tiny}.
14453 @item -msign-return-address=@var{scope}
14454 @opindex msign-return-address
14455 Select the function scope on which return address signing will be applied.
14456 Permissible values are @samp{none}, which disables return address signing,
14457 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
14458 functions, and @samp{all}, which enables pointer signing for all functions. The
14459 default value is @samp{none}.
14463 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14464 @anchor{aarch64-feature-modifiers}
14465 @cindex @option{-march} feature modifiers
14466 @cindex @option{-mcpu} feature modifiers
14467 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14468 the following and their inverses @option{no@var{feature}}:
14472 Enable CRC extension. This is on by default for
14473 @option{-march=armv8.1-a}.
14475 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14478 Enable floating-point instructions. This is on by default for all possible
14479 values for options @option{-march} and @option{-mcpu}.
14481 Enable Advanced SIMD instructions. This also enables floating-point
14482 instructions. This is on by default for all possible values for options
14483 @option{-march} and @option{-mcpu}.
14485 Enable Large System Extension instructions. This is on by default for
14486 @option{-march=armv8.1-a}.
14488 Enable Round Double Multiply Accumulate instructions. This is on by default
14489 for @option{-march=armv8.1-a}.
14491 Enable FP16 extension. This also enables floating-point instructions.
14493 Enable the RcPc extension. This does not change code generation from GCC,
14494 but is passed on to the assembler, enabling inline asm statements to use
14495 instructions from the RcPc extension.
14497 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
14501 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
14502 Conversely, @option{nofp} implies @option{nosimd}, which implies
14505 @node Adapteva Epiphany Options
14506 @subsection Adapteva Epiphany Options
14508 These @samp{-m} options are defined for Adapteva Epiphany:
14511 @item -mhalf-reg-file
14512 @opindex mhalf-reg-file
14513 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14514 That allows code to run on hardware variants that lack these registers.
14516 @item -mprefer-short-insn-regs
14517 @opindex mprefer-short-insn-regs
14518 Preferentially allocate registers that allow short instruction generation.
14519 This can result in increased instruction count, so this may either reduce or
14520 increase overall code size.
14522 @item -mbranch-cost=@var{num}
14523 @opindex mbranch-cost
14524 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14525 This cost is only a heuristic and is not guaranteed to produce
14526 consistent results across releases.
14530 Enable the generation of conditional moves.
14532 @item -mnops=@var{num}
14534 Emit @var{num} NOPs before every other generated instruction.
14536 @item -mno-soft-cmpsf
14537 @opindex mno-soft-cmpsf
14538 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14539 and test the flags. This is faster than a software comparison, but can
14540 get incorrect results in the presence of NaNs, or when two different small
14541 numbers are compared such that their difference is calculated as zero.
14542 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14543 software comparisons.
14545 @item -mstack-offset=@var{num}
14546 @opindex mstack-offset
14547 Set the offset between the top of the stack and the stack pointer.
14548 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14549 can be used by leaf functions without stack allocation.
14550 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14551 Note also that this option changes the ABI; compiling a program with a
14552 different stack offset than the libraries have been compiled with
14553 generally does not work.
14554 This option can be useful if you want to evaluate if a different stack
14555 offset would give you better code, but to actually use a different stack
14556 offset to build working programs, it is recommended to configure the
14557 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14559 @item -mno-round-nearest
14560 @opindex mno-round-nearest
14561 Make the scheduler assume that the rounding mode has been set to
14562 truncating. The default is @option{-mround-nearest}.
14565 @opindex mlong-calls
14566 If not otherwise specified by an attribute, assume all calls might be beyond
14567 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14568 function address into a register before performing a (otherwise direct) call.
14569 This is the default.
14571 @item -mshort-calls
14572 @opindex short-calls
14573 If not otherwise specified by an attribute, assume all direct calls are
14574 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14575 for direct calls. The default is @option{-mlong-calls}.
14579 Assume addresses can be loaded as 16-bit unsigned values. This does not
14580 apply to function addresses for which @option{-mlong-calls} semantics
14583 @item -mfp-mode=@var{mode}
14585 Set the prevailing mode of the floating-point unit.
14586 This determines the floating-point mode that is provided and expected
14587 at function call and return time. Making this mode match the mode you
14588 predominantly need at function start can make your programs smaller and
14589 faster by avoiding unnecessary mode switches.
14591 @var{mode} can be set to one the following values:
14595 Any mode at function entry is valid, and retained or restored when
14596 the function returns, and when it calls other functions.
14597 This mode is useful for compiling libraries or other compilation units
14598 you might want to incorporate into different programs with different
14599 prevailing FPU modes, and the convenience of being able to use a single
14600 object file outweighs the size and speed overhead for any extra
14601 mode switching that might be needed, compared with what would be needed
14602 with a more specific choice of prevailing FPU mode.
14605 This is the mode used for floating-point calculations with
14606 truncating (i.e.@: round towards zero) rounding mode. That includes
14607 conversion from floating point to integer.
14609 @item round-nearest
14610 This is the mode used for floating-point calculations with
14611 round-to-nearest-or-even rounding mode.
14614 This is the mode used to perform integer calculations in the FPU, e.g.@:
14615 integer multiply, or integer multiply-and-accumulate.
14618 The default is @option{-mfp-mode=caller}
14620 @item -mnosplit-lohi
14621 @itemx -mno-postinc
14622 @itemx -mno-postmodify
14623 @opindex mnosplit-lohi
14624 @opindex mno-postinc
14625 @opindex mno-postmodify
14626 Code generation tweaks that disable, respectively, splitting of 32-bit
14627 loads, generation of post-increment addresses, and generation of
14628 post-modify addresses. The defaults are @option{msplit-lohi},
14629 @option{-mpost-inc}, and @option{-mpost-modify}.
14631 @item -mnovect-double
14632 @opindex mno-vect-double
14633 Change the preferred SIMD mode to SImode. The default is
14634 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14636 @item -max-vect-align=@var{num}
14637 @opindex max-vect-align
14638 The maximum alignment for SIMD vector mode types.
14639 @var{num} may be 4 or 8. The default is 8.
14640 Note that this is an ABI change, even though many library function
14641 interfaces are unaffected if they don't use SIMD vector modes
14642 in places that affect size and/or alignment of relevant types.
14644 @item -msplit-vecmove-early
14645 @opindex msplit-vecmove-early
14646 Split vector moves into single word moves before reload. In theory this
14647 can give better register allocation, but so far the reverse seems to be
14648 generally the case.
14650 @item -m1reg-@var{reg}
14652 Specify a register to hold the constant @minus{}1, which makes loading small negative
14653 constants and certain bitmasks faster.
14654 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14655 which specify use of that register as a fixed register,
14656 and @samp{none}, which means that no register is used for this
14657 purpose. The default is @option{-m1reg-none}.
14662 @subsection ARC Options
14663 @cindex ARC options
14665 The following options control the architecture variant for which code
14668 @c architecture variants
14671 @item -mbarrel-shifter
14672 @opindex mbarrel-shifter
14673 Generate instructions supported by barrel shifter. This is the default
14674 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14676 @item -mcpu=@var{cpu}
14678 Set architecture type, register usage, and instruction scheduling
14679 parameters for @var{cpu}. There are also shortcut alias options
14680 available for backward compatibility and convenience. Supported
14681 values for @var{cpu} are
14687 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14691 Compile for ARC601. Alias: @option{-mARC601}.
14696 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14697 This is the default when configured with @option{--with-cpu=arc700}@.
14700 Compile for ARC EM.
14703 Compile for ARC HS.
14706 Compile for ARC EM CPU with no hardware extensions.
14709 Compile for ARC EM4 CPU.
14712 Compile for ARC EM4 DMIPS CPU.
14715 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
14719 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
14720 double assist instructions.
14723 Compile for ARC HS CPU with no hardware extensions except the atomic
14727 Compile for ARC HS34 CPU.
14730 Compile for ARC HS38 CPU.
14733 Compile for ARC HS38 CPU with all hardware extensions on.
14736 Compile for ARC 600 CPU with @code{norm} instructions enabled.
14738 @item arc600_mul32x16
14739 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
14740 instructions enabled.
14743 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
14744 instructions enabled.
14747 Compile for ARC 601 CPU with @code{norm} instructions enabled.
14749 @item arc601_mul32x16
14750 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
14751 instructions enabled.
14754 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
14755 instructions enabled.
14758 Compile for ARC 700 on NPS400 chip.
14764 @itemx -mdpfp-compact
14765 @opindex mdpfp-compact
14766 Generate double-precision FPX instructions, tuned for the compact
14770 @opindex mdpfp-fast
14771 Generate double-precision FPX instructions, tuned for the fast
14774 @item -mno-dpfp-lrsr
14775 @opindex mno-dpfp-lrsr
14776 Disable @code{lr} and @code{sr} instructions from using FPX extension
14781 Generate extended arithmetic instructions. Currently only
14782 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14783 supported. This is always enabled for @option{-mcpu=ARC700}.
14787 Do not generate @code{mpy}-family instructions for ARC700. This option is
14792 Generate 32x16-bit multiply and multiply-accumulate instructions.
14796 Generate @code{mul64} and @code{mulu64} instructions.
14797 Only valid for @option{-mcpu=ARC600}.
14801 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
14806 @itemx -mspfp-compact
14807 @opindex mspfp-compact
14808 Generate single-precision FPX instructions, tuned for the compact
14812 @opindex mspfp-fast
14813 Generate single-precision FPX instructions, tuned for the fast
14818 Enable generation of ARC SIMD instructions via target-specific
14819 builtins. Only valid for @option{-mcpu=ARC700}.
14822 @opindex msoft-float
14823 This option ignored; it is provided for compatibility purposes only.
14824 Software floating-point code is emitted by default, and this default
14825 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
14826 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
14827 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
14831 Generate @code{swap} instructions.
14835 This enables use of the locked load/store conditional extension to implement
14836 atomic memory built-in functions. Not available for ARC 6xx or ARC
14841 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
14843 @item -mcode-density
14844 @opindex mcode-density
14845 Enable code density instructions for ARC EM.
14846 This option is on by default for ARC HS.
14850 Enable double load/store operations for ARC HS cores.
14852 @item -mtp-regno=@var{regno}
14854 Specify thread pointer register number.
14856 @item -mmpy-option=@var{multo}
14857 @opindex mmpy-option
14858 Compile ARCv2 code with a multiplier design option. You can specify
14859 the option using either a string or numeric value for @var{multo}.
14860 @samp{wlh1} is the default value. The recognized values are:
14865 No multiplier available.
14869 16x16 multiplier, fully pipelined.
14870 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
14874 32x32 multiplier, fully
14875 pipelined (1 stage). The following instructions are additionally
14876 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14880 32x32 multiplier, fully pipelined
14881 (2 stages). The following instructions are additionally enabled: @code{mpy},
14882 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14886 Two 16x16 multipliers, blocking,
14887 sequential. The following instructions are additionally enabled: @code{mpy},
14888 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14892 One 16x16 multiplier, blocking,
14893 sequential. The following instructions are additionally enabled: @code{mpy},
14894 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14898 One 32x4 multiplier, blocking,
14899 sequential. The following instructions are additionally enabled: @code{mpy},
14900 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14904 ARC HS SIMD support.
14908 ARC HS SIMD support.
14912 ARC HS SIMD support.
14916 This option is only available for ARCv2 cores@.
14918 @item -mfpu=@var{fpu}
14920 Enables support for specific floating-point hardware extensions for ARCv2
14921 cores. Supported values for @var{fpu} are:
14926 Enables support for single-precision floating-point hardware
14930 Enables support for double-precision floating-point hardware
14931 extensions. The single-precision floating-point extension is also
14932 enabled. Not available for ARC EM@.
14935 Enables support for double-precision floating-point hardware
14936 extensions using double-precision assist instructions. The single-precision
14937 floating-point extension is also enabled. This option is
14938 only available for ARC EM@.
14941 Enables support for double-precision floating-point hardware
14942 extensions using double-precision assist instructions.
14943 The single-precision floating-point, square-root, and divide
14944 extensions are also enabled. This option is
14945 only available for ARC EM@.
14948 Enables support for double-precision floating-point hardware
14949 extensions using double-precision assist instructions.
14950 The single-precision floating-point and fused multiply and add
14951 hardware extensions are also enabled. This option is
14952 only available for ARC EM@.
14955 Enables support for double-precision floating-point hardware
14956 extensions using double-precision assist instructions.
14957 All single-precision floating-point hardware extensions are also
14958 enabled. This option is only available for ARC EM@.
14961 Enables support for single-precision floating-point, square-root and divide
14962 hardware extensions@.
14965 Enables support for double-precision floating-point, square-root and divide
14966 hardware extensions. This option
14967 includes option @samp{fpus_div}. Not available for ARC EM@.
14970 Enables support for single-precision floating-point and
14971 fused multiply and add hardware extensions@.
14974 Enables support for double-precision floating-point and
14975 fused multiply and add hardware extensions. This option
14976 includes option @samp{fpus_fma}. Not available for ARC EM@.
14979 Enables support for all single-precision floating-point hardware
14983 Enables support for all single- and double-precision floating-point
14984 hardware extensions. Not available for ARC EM@.
14988 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
14989 @opindex mirq-ctrl-saved
14990 Specifies general-purposes registers that the processor automatically
14991 saves/restores on interrupt entry and exit. @var{register-range} is
14992 specified as two registers separated by a dash. The register range
14993 always starts with @code{r0}, the upper limit is @code{fp} register.
14994 @var{blink} and @var{lp_count} are optional. This option is only
14995 valid for ARC EM and ARC HS cores.
14997 @item -mrgf-banked-regs=@var{number}
14998 @opindex mrgf-banked-regs
14999 Specifies the number of registers replicated in second register bank
15000 on entry to fast interrupt. Fast interrupts are interrupts with the
15001 highest priority level P0. These interrupts save only PC and STATUS32
15002 registers to avoid memory transactions during interrupt entry and exit
15003 sequences. Use this option when you are using fast interrupts in an
15004 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
15006 @item -mlpc-width=@var{width}
15007 @opindex mlpc-width
15008 Specify the width of the @code{lp_count} register. Valid values for
15009 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
15010 fixed to 32 bits. If the width is less than 32, the compiler does not
15011 attempt to transform loops in your program to use the zero-delay loop
15012 mechanism unless it is known that the @code{lp_count} register can
15013 hold the required loop-counter value. Depending on the width
15014 specified, the compiler and run-time library might continue to use the
15015 loop mechanism for various needs. This option defines macro
15016 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
15020 The following options are passed through to the assembler, and also
15021 define preprocessor macro symbols.
15023 @c Flags used by the assembler, but for which we define preprocessor
15024 @c macro symbols as well.
15027 @opindex mdsp-packa
15028 Passed down to the assembler to enable the DSP Pack A extensions.
15029 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
15034 Passed down to the assembler to enable the dual Viterbi butterfly
15035 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
15036 option is deprecated.
15038 @c ARC700 4.10 extension instruction
15041 Passed down to the assembler to enable the locked load/store
15042 conditional extension. Also sets the preprocessor symbol
15047 Passed down to the assembler. Also sets the preprocessor symbol
15048 @code{__Xxmac_d16}. This option is deprecated.
15052 Passed down to the assembler. Also sets the preprocessor symbol
15053 @code{__Xxmac_24}. This option is deprecated.
15055 @c ARC700 4.10 extension instruction
15058 Passed down to the assembler to enable the 64-bit time-stamp counter
15059 extension instruction. Also sets the preprocessor symbol
15060 @code{__Xrtsc}. This option is deprecated.
15062 @c ARC700 4.10 extension instruction
15065 Passed down to the assembler to enable the swap byte ordering
15066 extension instruction. Also sets the preprocessor symbol
15070 @opindex mtelephony
15071 Passed down to the assembler to enable dual- and single-operand
15072 instructions for telephony. Also sets the preprocessor symbol
15073 @code{__Xtelephony}. This option is deprecated.
15077 Passed down to the assembler to enable the XY memory extension. Also
15078 sets the preprocessor symbol @code{__Xxy}.
15082 The following options control how the assembly code is annotated:
15084 @c Assembly annotation options
15088 Annotate assembler instructions with estimated addresses.
15090 @item -mannotate-align
15091 @opindex mannotate-align
15092 Explain what alignment considerations lead to the decision to make an
15093 instruction short or long.
15097 The following options are passed through to the linker:
15099 @c options passed through to the linker
15103 Passed through to the linker, to specify use of the @code{arclinux} emulation.
15104 This option is enabled by default in tool chains built for
15105 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
15106 when profiling is not requested.
15108 @item -marclinux_prof
15109 @opindex marclinux_prof
15110 Passed through to the linker, to specify use of the
15111 @code{arclinux_prof} emulation. This option is enabled by default in
15112 tool chains built for @w{@code{arc-linux-uclibc}} and
15113 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
15117 The following options control the semantics of generated code:
15119 @c semantically relevant code generation options
15122 @opindex mlong-calls
15123 Generate calls as register indirect calls, thus providing access
15124 to the full 32-bit address range.
15126 @item -mmedium-calls
15127 @opindex mmedium-calls
15128 Don't use less than 25-bit addressing range for calls, which is the
15129 offset available for an unconditional branch-and-link
15130 instruction. Conditional execution of function calls is suppressed, to
15131 allow use of the 25-bit range, rather than the 21-bit range with
15132 conditional branch-and-link. This is the default for tool chains built
15133 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
15137 Put definitions of externally-visible data in a small data section if
15138 that data is no bigger than @var{num} bytes. The default value of
15139 @var{num} is 4 for any ARC configuration, or 8 when we have double
15140 load/store operations.
15144 Do not generate sdata references. This is the default for tool chains
15145 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
15148 @item -mvolatile-cache
15149 @opindex mvolatile-cache
15150 Use ordinarily cached memory accesses for volatile references. This is the
15153 @item -mno-volatile-cache
15154 @opindex mno-volatile-cache
15155 Enable cache bypass for volatile references.
15159 The following options fine tune code generation:
15160 @c code generation tuning options
15163 @opindex malign-call
15164 Do alignment optimizations for call instructions.
15166 @item -mauto-modify-reg
15167 @opindex mauto-modify-reg
15168 Enable the use of pre/post modify with register displacement.
15170 @item -mbbit-peephole
15171 @opindex mbbit-peephole
15172 Enable bbit peephole2.
15176 This option disables a target-specific pass in @file{arc_reorg} to
15177 generate compare-and-branch (@code{br@var{cc}}) instructions.
15178 It has no effect on
15179 generation of these instructions driven by the combiner pass.
15181 @item -mcase-vector-pcrel
15182 @opindex mcase-vector-pcrel
15183 Use PC-relative switch case tables to enable case table shortening.
15184 This is the default for @option{-Os}.
15186 @item -mcompact-casesi
15187 @opindex mcompact-casesi
15188 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
15189 and only available for ARCv1 cores.
15191 @item -mno-cond-exec
15192 @opindex mno-cond-exec
15193 Disable the ARCompact-specific pass to generate conditional
15194 execution instructions.
15196 Due to delay slot scheduling and interactions between operand numbers,
15197 literal sizes, instruction lengths, and the support for conditional execution,
15198 the target-independent pass to generate conditional execution is often lacking,
15199 so the ARC port has kept a special pass around that tries to find more
15200 conditional execution generation opportunities after register allocation,
15201 branch shortening, and delay slot scheduling have been done. This pass
15202 generally, but not always, improves performance and code size, at the cost of
15203 extra compilation time, which is why there is an option to switch it off.
15204 If you have a problem with call instructions exceeding their allowable
15205 offset range because they are conditionalized, you should consider using
15206 @option{-mmedium-calls} instead.
15208 @item -mearly-cbranchsi
15209 @opindex mearly-cbranchsi
15210 Enable pre-reload use of the @code{cbranchsi} pattern.
15212 @item -mexpand-adddi
15213 @opindex mexpand-adddi
15214 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
15215 @code{add.f}, @code{adc} etc. This option is deprecated.
15217 @item -mindexed-loads
15218 @opindex mindexed-loads
15219 Enable the use of indexed loads. This can be problematic because some
15220 optimizers then assume that indexed stores exist, which is not
15224 Enable Local Register Allocation. This is still experimental for ARC,
15225 so by default the compiler uses standard reload
15226 (i.e. @option{-mno-lra}).
15228 @item -mlra-priority-none
15229 @opindex mlra-priority-none
15230 Don't indicate any priority for target registers.
15232 @item -mlra-priority-compact
15233 @opindex mlra-priority-compact
15234 Indicate target register priority for r0..r3 / r12..r15.
15236 @item -mlra-priority-noncompact
15237 @opindex mlra-priority-noncompact
15238 Reduce target register priority for r0..r3 / r12..r15.
15240 @item -mno-millicode
15241 @opindex mno-millicode
15242 When optimizing for size (using @option{-Os}), prologues and epilogues
15243 that have to save or restore a large number of registers are often
15244 shortened by using call to a special function in libgcc; this is
15245 referred to as a @emph{millicode} call. As these calls can pose
15246 performance issues, and/or cause linking issues when linking in a
15247 nonstandard way, this option is provided to turn off millicode call
15251 @opindex mmixed-code
15252 Tweak register allocation to help 16-bit instruction generation.
15253 This generally has the effect of decreasing the average instruction size
15254 while increasing the instruction count.
15258 Enable @samp{q} instruction alternatives.
15259 This is the default for @option{-Os}.
15263 Enable @samp{Rcq} constraint handling.
15264 Most short code generation depends on this.
15265 This is the default.
15269 Enable @samp{Rcw} constraint handling.
15270 Most ccfsm condexec mostly depends on this.
15271 This is the default.
15273 @item -msize-level=@var{level}
15274 @opindex msize-level
15275 Fine-tune size optimization with regards to instruction lengths and alignment.
15276 The recognized values for @var{level} are:
15279 No size optimization. This level is deprecated and treated like @samp{1}.
15282 Short instructions are used opportunistically.
15285 In addition, alignment of loops and of code after barriers are dropped.
15288 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
15292 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
15293 the behavior when this is not set is equivalent to level @samp{1}.
15295 @item -mtune=@var{cpu}
15297 Set instruction scheduling parameters for @var{cpu}, overriding any implied
15298 by @option{-mcpu=}.
15300 Supported values for @var{cpu} are
15304 Tune for ARC600 CPU.
15307 Tune for ARC601 CPU.
15310 Tune for ARC700 CPU with standard multiplier block.
15313 Tune for ARC700 CPU with XMAC block.
15316 Tune for ARC725D CPU.
15319 Tune for ARC750D CPU.
15323 @item -mmultcost=@var{num}
15325 Cost to assume for a multiply instruction, with @samp{4} being equal to a
15326 normal instruction.
15328 @item -munalign-prob-threshold=@var{probability}
15329 @opindex munalign-prob-threshold
15330 Set probability threshold for unaligning branches.
15331 When tuning for @samp{ARC700} and optimizing for speed, branches without
15332 filled delay slot are preferably emitted unaligned and long, unless
15333 profiling indicates that the probability for the branch to be taken
15334 is below @var{probability}. @xref{Cross-profiling}.
15335 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
15339 The following options are maintained for backward compatibility, but
15340 are now deprecated and will be removed in a future release:
15342 @c Deprecated options
15350 @opindex mbig-endian
15353 Compile code for big-endian targets. Use of these options is now
15354 deprecated. Big-endian code is supported by configuring GCC to build
15355 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
15356 for which big endian is the default.
15358 @item -mlittle-endian
15359 @opindex mlittle-endian
15362 Compile code for little-endian targets. Use of these options is now
15363 deprecated. Little-endian code is supported by configuring GCC to build
15364 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
15365 for which little endian is the default.
15367 @item -mbarrel_shifter
15368 @opindex mbarrel_shifter
15369 Replaced by @option{-mbarrel-shifter}.
15371 @item -mdpfp_compact
15372 @opindex mdpfp_compact
15373 Replaced by @option{-mdpfp-compact}.
15376 @opindex mdpfp_fast
15377 Replaced by @option{-mdpfp-fast}.
15380 @opindex mdsp_packa
15381 Replaced by @option{-mdsp-packa}.
15385 Replaced by @option{-mea}.
15389 Replaced by @option{-mmac-24}.
15393 Replaced by @option{-mmac-d16}.
15395 @item -mspfp_compact
15396 @opindex mspfp_compact
15397 Replaced by @option{-mspfp-compact}.
15400 @opindex mspfp_fast
15401 Replaced by @option{-mspfp-fast}.
15403 @item -mtune=@var{cpu}
15405 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
15406 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
15407 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
15409 @item -multcost=@var{num}
15411 Replaced by @option{-mmultcost}.
15416 @subsection ARM Options
15417 @cindex ARM options
15419 These @samp{-m} options are defined for the ARM port:
15422 @item -mabi=@var{name}
15424 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
15425 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
15428 @opindex mapcs-frame
15429 Generate a stack frame that is compliant with the ARM Procedure Call
15430 Standard for all functions, even if this is not strictly necessary for
15431 correct execution of the code. Specifying @option{-fomit-frame-pointer}
15432 with this option causes the stack frames not to be generated for
15433 leaf functions. The default is @option{-mno-apcs-frame}.
15434 This option is deprecated.
15438 This is a synonym for @option{-mapcs-frame} and is deprecated.
15441 @c not currently implemented
15442 @item -mapcs-stack-check
15443 @opindex mapcs-stack-check
15444 Generate code to check the amount of stack space available upon entry to
15445 every function (that actually uses some stack space). If there is
15446 insufficient space available then either the function
15447 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
15448 called, depending upon the amount of stack space required. The runtime
15449 system is required to provide these functions. The default is
15450 @option{-mno-apcs-stack-check}, since this produces smaller code.
15452 @c not currently implemented
15453 @item -mapcs-reentrant
15454 @opindex mapcs-reentrant
15455 Generate reentrant, position-independent code. The default is
15456 @option{-mno-apcs-reentrant}.
15459 @item -mthumb-interwork
15460 @opindex mthumb-interwork
15461 Generate code that supports calling between the ARM and Thumb
15462 instruction sets. Without this option, on pre-v5 architectures, the
15463 two instruction sets cannot be reliably used inside one program. The
15464 default is @option{-mno-thumb-interwork}, since slightly larger code
15465 is generated when @option{-mthumb-interwork} is specified. In AAPCS
15466 configurations this option is meaningless.
15468 @item -mno-sched-prolog
15469 @opindex mno-sched-prolog
15470 Prevent the reordering of instructions in the function prologue, or the
15471 merging of those instruction with the instructions in the function's
15472 body. This means that all functions start with a recognizable set
15473 of instructions (or in fact one of a choice from a small set of
15474 different function prologues), and this information can be used to
15475 locate the start of functions inside an executable piece of code. The
15476 default is @option{-msched-prolog}.
15478 @item -mfloat-abi=@var{name}
15479 @opindex mfloat-abi
15480 Specifies which floating-point ABI to use. Permissible values
15481 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15483 Specifying @samp{soft} causes GCC to generate output containing
15484 library calls for floating-point operations.
15485 @samp{softfp} allows the generation of code using hardware floating-point
15486 instructions, but still uses the soft-float calling conventions.
15487 @samp{hard} allows generation of floating-point instructions
15488 and uses FPU-specific calling conventions.
15490 The default depends on the specific target configuration. Note that
15491 the hard-float and soft-float ABIs are not link-compatible; you must
15492 compile your entire program with the same ABI, and link with a
15493 compatible set of libraries.
15495 @item -mlittle-endian
15496 @opindex mlittle-endian
15497 Generate code for a processor running in little-endian mode. This is
15498 the default for all standard configurations.
15501 @opindex mbig-endian
15502 Generate code for a processor running in big-endian mode; the default is
15503 to compile code for a little-endian processor.
15508 When linking a big-endian image select between BE8 and BE32 formats.
15509 The option has no effect for little-endian images and is ignored. The
15510 default is dependent on the selected target architecture. For ARMv6
15511 and later architectures the default is BE8, for older architectures
15512 the default is BE32. BE32 format has been deprecated by ARM.
15514 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
15516 This specifies the name of the target ARM architecture. GCC uses this
15517 name to determine what kind of instructions it can emit when generating
15518 assembly code. This option can be used in conjunction with or instead
15519 of the @option{-mcpu=} option.
15521 Permissible names are:
15523 @samp{armv5t}, @samp{armv5te},
15524 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
15525 @samp{armv6z}, @samp{armv6zk},
15526 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
15527 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
15530 @samp{armv6-m}, @samp{armv6s-m},
15531 @samp{armv7-m}, @samp{armv7e-m},
15532 @samp{armv8-m.base}, @samp{armv8-m.main},
15533 @samp{iwmmxt} and @samp{iwmmxt2}.
15535 Additionally, the following architectures, which lack support for the
15536 Thumb exection state, are recognized but support is deprecated:
15537 @samp{armv2}, @samp{armv2a}, @samp{armv3}, @samp{armv3m},
15538 @samp{armv4}, @samp{armv5} and @samp{armv5e}.
15540 Many of the architectures support extensions. These can be added by
15541 appending @samp{+@var{extension}} to the architecture name. Extension
15542 options are processed in order and capabilities accumulate. An extension
15543 will also enable any necessary base extensions
15544 upon which it depends. For example, the @samp{+crypto} extension
15545 will always enable the @samp{+simd} extension. The exception to the
15546 additive construction is for extensions that are prefixed with
15547 @samp{+no@dots{}}: these extensions disable the specified option and
15548 any other extensions that may depend on the presence of that
15551 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
15552 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
15553 entirely disabled by the @samp{+nofp} option that follows it.
15555 Most extension names are generically named, but have an effect that is
15556 dependent upon the architecture to which it is applied. For example,
15557 the @samp{+simd} option can be applied to both @samp{armv7-a} and
15558 @samp{armv8-a} architectures, but will enable the original ARMv7-A
15559 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
15560 variant for @samp{armv8-a}.
15562 The table below lists the supported extensions for each architecture.
15563 Architectures not mentioned do not support any extensions.
15577 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
15578 used as an alias for this extension.
15581 Disable the floating-point instructions.
15585 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
15588 The VFPv3 floating-point instructions, with 16 double-precision
15589 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15590 for this extension. Note that floating-point is not supported by the
15591 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
15592 ARMv7-R architectures.
15595 Disable the floating-point instructions.
15601 The VFPv3 floating-point instructions, with 16 double-precision
15602 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15603 for this extension.
15606 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15607 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
15608 for this extension.
15611 The VFPv3 floating-point instructions, with 32 double-precision
15614 @item +vfpv3-d16-fp16
15615 The VFPv3 floating-point instructions, with 16 double-precision
15616 registers and the half-precision floating-point conversion operations.
15619 The VFPv3 floating-point instructions, with 32 double-precision
15620 registers and the half-precision floating-point conversion operations.
15623 The VFPv4 floating-point instructions, with 16 double-precision
15627 The VFPv4 floating-point instructions, with 32 double-precision
15631 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15632 the half-precision floating-point conversion operations.
15635 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
15638 Disable the Advanced SIMD instructions (does not disable floating point).
15641 Disable the floating-point and Advanced SIMD instructions.
15645 The extended version of the ARMv7-A architecture with support for
15649 The VFPv4 floating-point instructions, with 16 double-precision registers.
15650 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
15653 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
15654 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
15657 The VFPv3 floating-point instructions, with 16 double-precision
15661 The VFPv3 floating-point instructions, with 32 double-precision
15664 @item +vfpv3-d16-fp16
15665 The VFPv3 floating-point instructions, with 16 double-precision
15666 registers and the half-precision floating-point conversion operations.
15669 The VFPv3 floating-point instructions, with 32 double-precision
15670 registers and the half-precision floating-point conversion operations.
15673 The VFPv4 floating-point instructions, with 16 double-precision
15677 The VFPv4 floating-point instructions, with 32 double-precision
15681 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15682 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
15685 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15686 the half-precision floating-point conversion operations.
15689 Disable the Advanced SIMD instructions (does not disable floating point).
15692 Disable the floating-point and Advanced SIMD instructions.
15698 The Cyclic Redundancy Check (CRC) instructions.
15700 The ARMv8-A Advanced SIMD and floating-point instructions.
15702 The cryptographic instructions.
15704 Disable the cryptographic isntructions.
15706 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15712 The ARMv8.1-A Advanced SIMD and floating-point instructions.
15715 The cryptographic instructions. This also enables the Advanced SIMD and
15716 floating-point instructions.
15719 Disable the cryptographic isntructions.
15722 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15728 The half-precision floating-point data processing instructions.
15729 This also enables the Advanced SIMD and floating-point instructions.
15732 The ARMv8.1-A Advanced SIMD and floating-point instructions.
15735 The cryptographic instructions. This also enables the Advanced SIMD and
15736 floating-point instructions.
15739 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
15742 Disable the cryptographic extension.
15745 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15751 The single-precision VFPv3 floating-point instructions. The extension
15752 @samp{+vfpv3xd} can be used as an alias for this extension.
15755 The VFPv3 floating-point instructions with 16 double-precision registers.
15756 The extension +vfpv3-d16 can be used as an alias for this extension.
15759 Disable the floating-point extension.
15762 The ARM-state integer division instructions.
15765 Disable the ARM-state integer division extension.
15771 The single-precision VFPv4 floating-point instructions.
15774 The single-precision FPv5 floating-point instructions.
15777 The single- and double-precision FPv5 floating-point instructions.
15780 Disable the floating-point extensions.
15786 The DSP instructions.
15789 Disable the DSP extension.
15792 The single-precision floating-point instructions.
15795 The single- and double-precision floating-point instructions.
15798 Disable the floating-point extension.
15804 The Cyclic Redundancy Check (CRC) instructions.
15806 The single-precision FPv5 floating-point instructions.
15808 The ARMv8-A Advanced SIMD and floating-point instructions.
15810 The cryptographic instructions.
15812 Disable the cryptographic isntructions.
15814 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15819 @option{-march=native} causes the compiler to auto-detect the architecture
15820 of the build computer. At present, this feature is only supported on
15821 GNU/Linux, and not all architectures are recognized. If the auto-detect
15822 is unsuccessful the option has no effect.
15824 @item -mtune=@var{name}
15826 This option specifies the name of the target ARM processor for
15827 which GCC should tune the performance of the code.
15828 For some ARM implementations better performance can be obtained by using
15830 Permissible names are: @samp{arm2}, @samp{arm250},
15831 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
15832 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
15833 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
15834 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
15836 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
15837 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
15838 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
15839 @samp{strongarm1110},
15840 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
15841 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
15842 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
15843 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
15844 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
15845 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
15846 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
15847 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
15848 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
15849 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15850 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15851 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7},
15852 @samp{cortex-r8}, @samp{cortex-r52},
15860 @samp{cortex-m0plus},
15861 @samp{cortex-m1.small-multiply},
15862 @samp{cortex-m0.small-multiply},
15863 @samp{cortex-m0plus.small-multiply},
15865 @samp{marvell-pj4},
15866 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
15867 @samp{fa526}, @samp{fa626},
15868 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
15871 Additionally, this option can specify that GCC should tune the performance
15872 of the code for a big.LITTLE system. Permissible names are:
15873 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
15874 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15875 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
15876 @samp{cortex-a75.cortex-a55}.
15878 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
15879 performance for a blend of processors within architecture @var{arch}.
15880 The aim is to generate code that run well on the current most popular
15881 processors, balancing between optimizations that benefit some CPUs in the
15882 range, and avoiding performance pitfalls of other CPUs. The effects of
15883 this option may change in future GCC versions as CPU models come and go.
15885 @option{-mtune} permits the same extension options as @option{-mcpu}, but
15886 the extension options do not affect the tuning of the generated code.
15888 @option{-mtune=native} causes the compiler to auto-detect the CPU
15889 of the build computer. At present, this feature is only supported on
15890 GNU/Linux, and not all architectures are recognized. If the auto-detect is
15891 unsuccessful the option has no effect.
15893 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
15895 This specifies the name of the target ARM processor. GCC uses this name
15896 to derive the name of the target ARM architecture (as if specified
15897 by @option{-march}) and the ARM processor type for which to tune for
15898 performance (as if specified by @option{-mtune}). Where this option
15899 is used in conjunction with @option{-march} or @option{-mtune},
15900 those options take precedence over the appropriate part of this option.
15902 Many of the supported CPUs implement optional architectural
15903 extensions. Where this is so the architectural extensions are
15904 normally enabled by default. If implementations that lack the
15905 extension exist, then the extension syntax can be used to disable
15906 those extensions that have been omitted. For floating-point and
15907 Advanced SIMD (Neon) instructions, the settings of the options
15908 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
15909 floating-point and Advanced SIMD instructions will only be used if
15910 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
15911 @option{-mfpu} other than @samp{auto} will override the available
15912 floating-point and SIMD extension instructions.
15914 For example, @samp{cortex-a9} can be found in three major
15915 configurations: integer only, with just a floating-point unit or with
15916 floating-point and Advanced SIMD. The default is to enable all the
15917 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
15918 be used to disable just the SIMD or both the SIMD and floating-point
15919 instructions respectively.
15921 Permissible names for this option are the same as those for
15924 The following extension options are common to the listed CPUs:
15928 Disable the DSP instructions on @samp{cortex-m33}.
15931 Disables the floating-point instructions on @samp{arm9e},
15932 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
15933 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
15934 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
15935 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
15936 Disables the floating-point and SIMD instructions on
15937 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
15938 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
15939 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
15940 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
15941 @samp{cortex-a53} and @samp{cortex-a55}.
15944 Disables the double-precision component of the floating-point instructions
15945 on @samp{cortex-r5}, @samp{cortex-r52} and @samp{cortex-m7}.
15948 Disables the SIMD (but not floating-point) instructions on
15949 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
15950 and @samp{cortex-a9}.
15953 Enables the cryptographic instructions on @samp{cortex-a32},
15954 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
15955 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
15956 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15957 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
15958 @samp{cortex-a75.cortex-a55}.
15961 Additionally the @samp{generic-armv7-a} pseudo target defaults to
15962 VFPv3 with 16 double-precision registers. It supports the following
15963 extension options: @samp{vfpv3-d16}, @samp{vfpv3},
15964 @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16}, @samp{vfpv4-d16},
15965 @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3}, @samp{neon-fp16},
15966 @samp{neon-vfpv4}. The meanings are the same as for the extensions to
15967 @option{-march=armv7-a}.
15969 @option{-mcpu=generic-@var{arch}} is also permissible, and is
15970 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
15971 See @option{-mtune} for more information.
15973 @option{-mcpu=native} causes the compiler to auto-detect the CPU
15974 of the build computer. At present, this feature is only supported on
15975 GNU/Linux, and not all architectures are recognized. If the auto-detect
15976 is unsuccessful the option has no effect.
15978 @item -mfpu=@var{name}
15980 This specifies what floating-point hardware (or hardware emulation) is
15981 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
15983 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
15984 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
15985 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
15986 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
15987 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
15988 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
15989 is an alias for @samp{vfpv2}.
15991 The setting @samp{auto} is the default and is special. It causes the
15992 compiler to select the floating-point and Advanced SIMD instructions
15993 based on the settings of @option{-mcpu} and @option{-march}.
15995 If the selected floating-point hardware includes the NEON extension
15996 (e.g. @option{-mfpu=neon}), note that floating-point
15997 operations are not generated by GCC's auto-vectorization pass unless
15998 @option{-funsafe-math-optimizations} is also specified. This is
15999 because NEON hardware does not fully implement the IEEE 754 standard for
16000 floating-point arithmetic (in particular denormal values are treated as
16001 zero), so the use of NEON instructions may lead to a loss of precision.
16003 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}).
16005 @item -mfp16-format=@var{name}
16006 @opindex mfp16-format
16007 Specify the format of the @code{__fp16} half-precision floating-point type.
16008 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
16009 the default is @samp{none}, in which case the @code{__fp16} type is not
16010 defined. @xref{Half-Precision}, for more information.
16012 @item -mstructure-size-boundary=@var{n}
16013 @opindex mstructure-size-boundary
16014 The sizes of all structures and unions are rounded up to a multiple
16015 of the number of bits set by this option. Permissible values are 8, 32
16016 and 64. The default value varies for different toolchains. For the COFF
16017 targeted toolchain the default value is 8. A value of 64 is only allowed
16018 if the underlying ABI supports it.
16020 Specifying a larger number can produce faster, more efficient code, but
16021 can also increase the size of the program. Different values are potentially
16022 incompatible. Code compiled with one value cannot necessarily expect to
16023 work with code or libraries compiled with another value, if they exchange
16024 information using structures or unions.
16026 This option is deprecated.
16028 @item -mabort-on-noreturn
16029 @opindex mabort-on-noreturn
16030 Generate a call to the function @code{abort} at the end of a
16031 @code{noreturn} function. It is executed if the function tries to
16035 @itemx -mno-long-calls
16036 @opindex mlong-calls
16037 @opindex mno-long-calls
16038 Tells the compiler to perform function calls by first loading the
16039 address of the function into a register and then performing a subroutine
16040 call on this register. This switch is needed if the target function
16041 lies outside of the 64-megabyte addressing range of the offset-based
16042 version of subroutine call instruction.
16044 Even if this switch is enabled, not all function calls are turned
16045 into long calls. The heuristic is that static functions, functions
16046 that have the @code{short_call} attribute, functions that are inside
16047 the scope of a @code{#pragma no_long_calls} directive, and functions whose
16048 definitions have already been compiled within the current compilation
16049 unit are not turned into long calls. The exceptions to this rule are
16050 that weak function definitions, functions with the @code{long_call}
16051 attribute or the @code{section} attribute, and functions that are within
16052 the scope of a @code{#pragma long_calls} directive are always
16053 turned into long calls.
16055 This feature is not enabled by default. Specifying
16056 @option{-mno-long-calls} restores the default behavior, as does
16057 placing the function calls within the scope of a @code{#pragma
16058 long_calls_off} directive. Note these switches have no effect on how
16059 the compiler generates code to handle function calls via function
16062 @item -msingle-pic-base
16063 @opindex msingle-pic-base
16064 Treat the register used for PIC addressing as read-only, rather than
16065 loading it in the prologue for each function. The runtime system is
16066 responsible for initializing this register with an appropriate value
16067 before execution begins.
16069 @item -mpic-register=@var{reg}
16070 @opindex mpic-register
16071 Specify the register to be used for PIC addressing.
16072 For standard PIC base case, the default is any suitable register
16073 determined by compiler. For single PIC base case, the default is
16074 @samp{R9} if target is EABI based or stack-checking is enabled,
16075 otherwise the default is @samp{R10}.
16077 @item -mpic-data-is-text-relative
16078 @opindex mpic-data-is-text-relative
16079 Assume that the displacement between the text and data segments is fixed
16080 at static link time. This permits using PC-relative addressing
16081 operations to access data known to be in the data segment. For
16082 non-VxWorks RTP targets, this option is enabled by default. When
16083 disabled on such targets, it will enable @option{-msingle-pic-base} by
16086 @item -mpoke-function-name
16087 @opindex mpoke-function-name
16088 Write the name of each function into the text section, directly
16089 preceding the function prologue. The generated code is similar to this:
16093 .ascii "arm_poke_function_name", 0
16096 .word 0xff000000 + (t1 - t0)
16097 arm_poke_function_name
16099 stmfd sp!, @{fp, ip, lr, pc@}
16103 When performing a stack backtrace, code can inspect the value of
16104 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
16105 location @code{pc - 12} and the top 8 bits are set, then we know that
16106 there is a function name embedded immediately preceding this location
16107 and has length @code{((pc[-3]) & 0xff000000)}.
16114 Select between generating code that executes in ARM and Thumb
16115 states. The default for most configurations is to generate code
16116 that executes in ARM state, but the default can be changed by
16117 configuring GCC with the @option{--with-mode=}@var{state}
16120 You can also override the ARM and Thumb mode for each function
16121 by using the @code{target("thumb")} and @code{target("arm")} function attributes
16122 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
16125 @opindex mtpcs-frame
16126 Generate a stack frame that is compliant with the Thumb Procedure Call
16127 Standard for all non-leaf functions. (A leaf function is one that does
16128 not call any other functions.) The default is @option{-mno-tpcs-frame}.
16130 @item -mtpcs-leaf-frame
16131 @opindex mtpcs-leaf-frame
16132 Generate a stack frame that is compliant with the Thumb Procedure Call
16133 Standard for all leaf functions. (A leaf function is one that does
16134 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
16136 @item -mcallee-super-interworking
16137 @opindex mcallee-super-interworking
16138 Gives all externally visible functions in the file being compiled an ARM
16139 instruction set header which switches to Thumb mode before executing the
16140 rest of the function. This allows these functions to be called from
16141 non-interworking code. This option is not valid in AAPCS configurations
16142 because interworking is enabled by default.
16144 @item -mcaller-super-interworking
16145 @opindex mcaller-super-interworking
16146 Allows calls via function pointers (including virtual functions) to
16147 execute correctly regardless of whether the target code has been
16148 compiled for interworking or not. There is a small overhead in the cost
16149 of executing a function pointer if this option is enabled. This option
16150 is not valid in AAPCS configurations because interworking is enabled
16153 @item -mtp=@var{name}
16155 Specify the access model for the thread local storage pointer. The valid
16156 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
16157 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
16158 (supported in the arm6k architecture), and @samp{auto}, which uses the
16159 best available method for the selected processor. The default setting is
16162 @item -mtls-dialect=@var{dialect}
16163 @opindex mtls-dialect
16164 Specify the dialect to use for accessing thread local storage. Two
16165 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
16166 @samp{gnu} dialect selects the original GNU scheme for supporting
16167 local and global dynamic TLS models. The @samp{gnu2} dialect
16168 selects the GNU descriptor scheme, which provides better performance
16169 for shared libraries. The GNU descriptor scheme is compatible with
16170 the original scheme, but does require new assembler, linker and
16171 library support. Initial and local exec TLS models are unaffected by
16172 this option and always use the original scheme.
16174 @item -mword-relocations
16175 @opindex mword-relocations
16176 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
16177 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
16178 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
16181 @item -mfix-cortex-m3-ldrd
16182 @opindex mfix-cortex-m3-ldrd
16183 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
16184 with overlapping destination and base registers are used. This option avoids
16185 generating these instructions. This option is enabled by default when
16186 @option{-mcpu=cortex-m3} is specified.
16188 @item -munaligned-access
16189 @itemx -mno-unaligned-access
16190 @opindex munaligned-access
16191 @opindex mno-unaligned-access
16192 Enables (or disables) reading and writing of 16- and 32- bit values
16193 from addresses that are not 16- or 32- bit aligned. By default
16194 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
16195 ARMv8-M Baseline architectures, and enabled for all other
16196 architectures. If unaligned access is not enabled then words in packed
16197 data structures are accessed a byte at a time.
16199 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
16200 generated object file to either true or false, depending upon the
16201 setting of this option. If unaligned access is enabled then the
16202 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
16205 @item -mneon-for-64bits
16206 @opindex mneon-for-64bits
16207 Enables using Neon to handle scalar 64-bits operations. This is
16208 disabled by default since the cost of moving data from core registers
16211 @item -mslow-flash-data
16212 @opindex mslow-flash-data
16213 Assume loading data from flash is slower than fetching instruction.
16214 Therefore literal load is minimized for better performance.
16215 This option is only supported when compiling for ARMv7 M-profile and
16218 @item -masm-syntax-unified
16219 @opindex masm-syntax-unified
16220 Assume inline assembler is using unified asm syntax. The default is
16221 currently off which implies divided syntax. This option has no impact
16222 on Thumb2. However, this may change in future releases of GCC.
16223 Divided syntax should be considered deprecated.
16225 @item -mrestrict-it
16226 @opindex mrestrict-it
16227 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
16228 IT blocks can only contain a single 16-bit instruction from a select
16229 set of instructions. This option is on by default for ARMv8-A Thumb mode.
16231 @item -mprint-tune-info
16232 @opindex mprint-tune-info
16233 Print CPU tuning information as comment in assembler file. This is
16234 an option used only for regression testing of the compiler and not
16235 intended for ordinary use in compiling code. This option is disabled
16239 @opindex mpure-code
16240 Do not allow constant data to be placed in code sections.
16241 Additionally, when compiling for ELF object format give all text sections the
16242 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
16243 is only available when generating non-pic code for M-profile targets with the
16248 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
16249 Development Tools Engineering Specification", which can be found on
16250 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
16254 @subsection AVR Options
16255 @cindex AVR Options
16257 These options are defined for AVR implementations:
16260 @item -mmcu=@var{mcu}
16262 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
16264 The default for this option is@tie{}@samp{avr2}.
16266 GCC supports the following AVR devices and ISAs:
16268 @include avr-mmcu.texi
16273 Assume that all data in static storage can be accessed by LDS / STS
16274 instructions. This option has only an effect on reduced Tiny devices like
16275 ATtiny40. See also the @code{absdata}
16276 @ref{AVR Variable Attributes,variable attribute}.
16278 @item -maccumulate-args
16279 @opindex maccumulate-args
16280 Accumulate outgoing function arguments and acquire/release the needed
16281 stack space for outgoing function arguments once in function
16282 prologue/epilogue. Without this option, outgoing arguments are pushed
16283 before calling a function and popped afterwards.
16285 Popping the arguments after the function call can be expensive on
16286 AVR so that accumulating the stack space might lead to smaller
16287 executables because arguments need not be removed from the
16288 stack after such a function call.
16290 This option can lead to reduced code size for functions that perform
16291 several calls to functions that get their arguments on the stack like
16292 calls to printf-like functions.
16294 @item -mbranch-cost=@var{cost}
16295 @opindex mbranch-cost
16296 Set the branch costs for conditional branch instructions to
16297 @var{cost}. Reasonable values for @var{cost} are small, non-negative
16298 integers. The default branch cost is 0.
16300 @item -mcall-prologues
16301 @opindex mcall-prologues
16302 Functions prologues/epilogues are expanded as calls to appropriate
16303 subroutines. Code size is smaller.
16305 @item -mgas-isr-prologues
16306 @opindex mgas-isr-prologues
16307 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
16308 instruction supported by GNU Binutils.
16309 If this option is on, the feature can still be disabled for individual
16310 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
16311 function attribute. This feature is activated per default
16312 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
16313 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
16317 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
16318 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
16319 and @code{long long} is 4 bytes. Please note that this option does not
16320 conform to the C standards, but it results in smaller code
16323 @item -mn-flash=@var{num}
16325 Assume that the flash memory has a size of
16326 @var{num} times 64@tie{}KiB.
16328 @item -mno-interrupts
16329 @opindex mno-interrupts
16330 Generated code is not compatible with hardware interrupts.
16331 Code size is smaller.
16335 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
16336 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
16337 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
16338 the assembler's command line and the @option{--relax} option to the
16339 linker's command line.
16341 Jump relaxing is performed by the linker because jump offsets are not
16342 known before code is located. Therefore, the assembler code generated by the
16343 compiler is the same, but the instructions in the executable may
16344 differ from instructions in the assembler code.
16346 Relaxing must be turned on if linker stubs are needed, see the
16347 section on @code{EIND} and linker stubs below.
16351 Assume that the device supports the Read-Modify-Write
16352 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
16354 @item -mshort-calls
16355 @opindex mshort-calls
16357 Assume that @code{RJMP} and @code{RCALL} can target the whole
16360 This option is used internally for multilib selection. It is
16361 not an optimization option, and you don't need to set it by hand.
16365 Treat the stack pointer register as an 8-bit register,
16366 i.e.@: assume the high byte of the stack pointer is zero.
16367 In general, you don't need to set this option by hand.
16369 This option is used internally by the compiler to select and
16370 build multilibs for architectures @code{avr2} and @code{avr25}.
16371 These architectures mix devices with and without @code{SPH}.
16372 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
16373 the compiler driver adds or removes this option from the compiler
16374 proper's command line, because the compiler then knows if the device
16375 or architecture has an 8-bit stack pointer and thus no @code{SPH}
16380 Use address register @code{X} in a way proposed by the hardware. This means
16381 that @code{X} is only used in indirect, post-increment or
16382 pre-decrement addressing.
16384 Without this option, the @code{X} register may be used in the same way
16385 as @code{Y} or @code{Z} which then is emulated by additional
16387 For example, loading a value with @code{X+const} addressing with a
16388 small non-negative @code{const < 64} to a register @var{Rn} is
16392 adiw r26, const ; X += const
16393 ld @var{Rn}, X ; @var{Rn} = *X
16394 sbiw r26, const ; X -= const
16398 @opindex mtiny-stack
16399 Only change the lower 8@tie{}bits of the stack pointer.
16401 @item -mfract-convert-truncate
16402 @opindex mfract-convert-truncate
16403 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
16406 @opindex nodevicelib
16407 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
16409 @item -Waddr-space-convert
16410 @opindex Waddr-space-convert
16411 Warn about conversions between address spaces in the case where the
16412 resulting address space is not contained in the incoming address space.
16414 @item -Wmisspelled-isr
16415 @opindex Wmisspelled-isr
16416 Warn if the ISR is misspelled, i.e. without __vector prefix.
16417 Enabled by default.
16420 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
16421 @cindex @code{EIND}
16422 Pointers in the implementation are 16@tie{}bits wide.
16423 The address of a function or label is represented as word address so
16424 that indirect jumps and calls can target any code address in the
16425 range of 64@tie{}Ki words.
16427 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
16428 bytes of program memory space, there is a special function register called
16429 @code{EIND} that serves as most significant part of the target address
16430 when @code{EICALL} or @code{EIJMP} instructions are used.
16432 Indirect jumps and calls on these devices are handled as follows by
16433 the compiler and are subject to some limitations:
16438 The compiler never sets @code{EIND}.
16441 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
16442 instructions or might read @code{EIND} directly in order to emulate an
16443 indirect call/jump by means of a @code{RET} instruction.
16446 The compiler assumes that @code{EIND} never changes during the startup
16447 code or during the application. In particular, @code{EIND} is not
16448 saved/restored in function or interrupt service routine
16452 For indirect calls to functions and computed goto, the linker
16453 generates @emph{stubs}. Stubs are jump pads sometimes also called
16454 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
16455 The stub contains a direct jump to the desired address.
16458 Linker relaxation must be turned on so that the linker generates
16459 the stubs correctly in all situations. See the compiler option
16460 @option{-mrelax} and the linker option @option{--relax}.
16461 There are corner cases where the linker is supposed to generate stubs
16462 but aborts without relaxation and without a helpful error message.
16465 The default linker script is arranged for code with @code{EIND = 0}.
16466 If code is supposed to work for a setup with @code{EIND != 0}, a custom
16467 linker script has to be used in order to place the sections whose
16468 name start with @code{.trampolines} into the segment where @code{EIND}
16472 The startup code from libgcc never sets @code{EIND}.
16473 Notice that startup code is a blend of code from libgcc and AVR-LibC.
16474 For the impact of AVR-LibC on @code{EIND}, see the
16475 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
16478 It is legitimate for user-specific startup code to set up @code{EIND}
16479 early, for example by means of initialization code located in
16480 section @code{.init3}. Such code runs prior to general startup code
16481 that initializes RAM and calls constructors, but after the bit
16482 of startup code from AVR-LibC that sets @code{EIND} to the segment
16483 where the vector table is located.
16485 #include <avr/io.h>
16488 __attribute__((section(".init3"),naked,used,no_instrument_function))
16489 init3_set_eind (void)
16491 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
16492 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
16497 The @code{__trampolines_start} symbol is defined in the linker script.
16500 Stubs are generated automatically by the linker if
16501 the following two conditions are met:
16504 @item The address of a label is taken by means of the @code{gs} modifier
16505 (short for @emph{generate stubs}) like so:
16507 LDI r24, lo8(gs(@var{func}))
16508 LDI r25, hi8(gs(@var{func}))
16510 @item The final location of that label is in a code segment
16511 @emph{outside} the segment where the stubs are located.
16515 The compiler emits such @code{gs} modifiers for code labels in the
16516 following situations:
16518 @item Taking address of a function or code label.
16519 @item Computed goto.
16520 @item If prologue-save function is used, see @option{-mcall-prologues}
16521 command-line option.
16522 @item Switch/case dispatch tables. If you do not want such dispatch
16523 tables you can specify the @option{-fno-jump-tables} command-line option.
16524 @item C and C++ constructors/destructors called during startup/shutdown.
16525 @item If the tools hit a @code{gs()} modifier explained above.
16529 Jumping to non-symbolic addresses like so is @emph{not} supported:
16534 /* Call function at word address 0x2 */
16535 return ((int(*)(void)) 0x2)();
16539 Instead, a stub has to be set up, i.e.@: the function has to be called
16540 through a symbol (@code{func_4} in the example):
16545 extern int func_4 (void);
16547 /* Call function at byte address 0x4 */
16552 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
16553 Alternatively, @code{func_4} can be defined in the linker script.
16556 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
16557 @cindex @code{RAMPD}
16558 @cindex @code{RAMPX}
16559 @cindex @code{RAMPY}
16560 @cindex @code{RAMPZ}
16561 Some AVR devices support memories larger than the 64@tie{}KiB range
16562 that can be accessed with 16-bit pointers. To access memory locations
16563 outside this 64@tie{}KiB range, the content of a @code{RAMP}
16564 register is used as high part of the address:
16565 The @code{X}, @code{Y}, @code{Z} address register is concatenated
16566 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
16567 register, respectively, to get a wide address. Similarly,
16568 @code{RAMPD} is used together with direct addressing.
16572 The startup code initializes the @code{RAMP} special function
16573 registers with zero.
16576 If a @ref{AVR Named Address Spaces,named address space} other than
16577 generic or @code{__flash} is used, then @code{RAMPZ} is set
16578 as needed before the operation.
16581 If the device supports RAM larger than 64@tie{}KiB and the compiler
16582 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
16583 is reset to zero after the operation.
16586 If the device comes with a specific @code{RAMP} register, the ISR
16587 prologue/epilogue saves/restores that SFR and initializes it with
16588 zero in case the ISR code might (implicitly) use it.
16591 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
16592 If you use inline assembler to read from locations outside the
16593 16-bit address range and change one of the @code{RAMP} registers,
16594 you must reset it to zero after the access.
16598 @subsubsection AVR Built-in Macros
16600 GCC defines several built-in macros so that the user code can test
16601 for the presence or absence of features. Almost any of the following
16602 built-in macros are deduced from device capabilities and thus
16603 triggered by the @option{-mmcu=} command-line option.
16605 For even more AVR-specific built-in macros see
16606 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
16611 Build-in macro that resolves to a decimal number that identifies the
16612 architecture and depends on the @option{-mmcu=@var{mcu}} option.
16613 Possible values are:
16615 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
16616 @code{4}, @code{5}, @code{51}, @code{6}
16618 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
16619 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
16624 @code{102}, @code{103}, @code{104},
16625 @code{105}, @code{106}, @code{107}
16627 for @var{mcu}=@code{avrtiny},
16628 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
16629 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
16630 If @var{mcu} specifies a device, this built-in macro is set
16631 accordingly. For example, with @option{-mmcu=atmega8} the macro is
16632 defined to @code{4}.
16634 @item __AVR_@var{Device}__
16635 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
16636 the device's name. For example, @option{-mmcu=atmega8} defines the
16637 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
16638 @code{__AVR_ATtiny261A__}, etc.
16640 The built-in macros' names follow
16641 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
16642 the device name as from the AVR user manual. The difference between
16643 @var{Device} in the built-in macro and @var{device} in
16644 @option{-mmcu=@var{device}} is that the latter is always lowercase.
16646 If @var{device} is not a device but only a core architecture like
16647 @samp{avr51}, this macro is not defined.
16649 @item __AVR_DEVICE_NAME__
16650 Setting @option{-mmcu=@var{device}} defines this built-in macro to
16651 the device's name. For example, with @option{-mmcu=atmega8} the macro
16652 is defined to @code{atmega8}.
16654 If @var{device} is not a device but only a core architecture like
16655 @samp{avr51}, this macro is not defined.
16657 @item __AVR_XMEGA__
16658 The device / architecture belongs to the XMEGA family of devices.
16660 @item __AVR_HAVE_ELPM__
16661 The device has the @code{ELPM} instruction.
16663 @item __AVR_HAVE_ELPMX__
16664 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
16665 R@var{n},Z+} instructions.
16667 @item __AVR_HAVE_MOVW__
16668 The device has the @code{MOVW} instruction to perform 16-bit
16669 register-register moves.
16671 @item __AVR_HAVE_LPMX__
16672 The device has the @code{LPM R@var{n},Z} and
16673 @code{LPM R@var{n},Z+} instructions.
16675 @item __AVR_HAVE_MUL__
16676 The device has a hardware multiplier.
16678 @item __AVR_HAVE_JMP_CALL__
16679 The device has the @code{JMP} and @code{CALL} instructions.
16680 This is the case for devices with more than 8@tie{}KiB of program
16683 @item __AVR_HAVE_EIJMP_EICALL__
16684 @itemx __AVR_3_BYTE_PC__
16685 The device has the @code{EIJMP} and @code{EICALL} instructions.
16686 This is the case for devices with more than 128@tie{}KiB of program memory.
16687 This also means that the program counter
16688 (PC) is 3@tie{}bytes wide.
16690 @item __AVR_2_BYTE_PC__
16691 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
16692 with up to 128@tie{}KiB of program memory.
16694 @item __AVR_HAVE_8BIT_SP__
16695 @itemx __AVR_HAVE_16BIT_SP__
16696 The stack pointer (SP) register is treated as 8-bit respectively
16697 16-bit register by the compiler.
16698 The definition of these macros is affected by @option{-mtiny-stack}.
16700 @item __AVR_HAVE_SPH__
16702 The device has the SPH (high part of stack pointer) special function
16703 register or has an 8-bit stack pointer, respectively.
16704 The definition of these macros is affected by @option{-mmcu=} and
16705 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
16708 @item __AVR_HAVE_RAMPD__
16709 @itemx __AVR_HAVE_RAMPX__
16710 @itemx __AVR_HAVE_RAMPY__
16711 @itemx __AVR_HAVE_RAMPZ__
16712 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
16713 @code{RAMPZ} special function register, respectively.
16715 @item __NO_INTERRUPTS__
16716 This macro reflects the @option{-mno-interrupts} command-line option.
16718 @item __AVR_ERRATA_SKIP__
16719 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
16720 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
16721 instructions because of a hardware erratum. Skip instructions are
16722 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
16723 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
16726 @item __AVR_ISA_RMW__
16727 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
16729 @item __AVR_SFR_OFFSET__=@var{offset}
16730 Instructions that can address I/O special function registers directly
16731 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
16732 address as if addressed by an instruction to access RAM like @code{LD}
16733 or @code{STS}. This offset depends on the device architecture and has
16734 to be subtracted from the RAM address in order to get the
16735 respective I/O@tie{}address.
16737 @item __AVR_SHORT_CALLS__
16738 The @option{-mshort-calls} command line option is set.
16740 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
16741 Some devices support reading from flash memory by means of @code{LD*}
16742 instructions. The flash memory is seen in the data address space
16743 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
16744 is not defined, this feature is not available. If defined,
16745 the address space is linear and there is no need to put
16746 @code{.rodata} into RAM. This is handled by the default linker
16747 description file, and is currently available for
16748 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
16749 there is no need to use address spaces like @code{__flash} or
16750 features like attribute @code{progmem} and @code{pgm_read_*}.
16752 @item __WITH_AVRLIBC__
16753 The compiler is configured to be used together with AVR-Libc.
16754 See the @option{--with-avrlibc} configure option.
16758 @node Blackfin Options
16759 @subsection Blackfin Options
16760 @cindex Blackfin Options
16763 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
16765 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
16766 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
16767 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
16768 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
16769 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
16770 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
16771 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
16772 @samp{bf561}, @samp{bf592}.
16774 The optional @var{sirevision} specifies the silicon revision of the target
16775 Blackfin processor. Any workarounds available for the targeted silicon revision
16776 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
16777 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
16778 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
16779 hexadecimal digits representing the major and minor numbers in the silicon
16780 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
16781 is not defined. If @var{sirevision} is @samp{any}, the
16782 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
16783 If this optional @var{sirevision} is not used, GCC assumes the latest known
16784 silicon revision of the targeted Blackfin processor.
16786 GCC defines a preprocessor macro for the specified @var{cpu}.
16787 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
16788 provided by libgloss to be linked in if @option{-msim} is not given.
16790 Without this option, @samp{bf532} is used as the processor by default.
16792 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
16793 only the preprocessor macro is defined.
16797 Specifies that the program will be run on the simulator. This causes
16798 the simulator BSP provided by libgloss to be linked in. This option
16799 has effect only for @samp{bfin-elf} toolchain.
16800 Certain other options, such as @option{-mid-shared-library} and
16801 @option{-mfdpic}, imply @option{-msim}.
16803 @item -momit-leaf-frame-pointer
16804 @opindex momit-leaf-frame-pointer
16805 Don't keep the frame pointer in a register for leaf functions. This
16806 avoids the instructions to save, set up and restore frame pointers and
16807 makes an extra register available in leaf functions.
16809 @item -mspecld-anomaly
16810 @opindex mspecld-anomaly
16811 When enabled, the compiler ensures that the generated code does not
16812 contain speculative loads after jump instructions. If this option is used,
16813 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
16815 @item -mno-specld-anomaly
16816 @opindex mno-specld-anomaly
16817 Don't generate extra code to prevent speculative loads from occurring.
16819 @item -mcsync-anomaly
16820 @opindex mcsync-anomaly
16821 When enabled, the compiler ensures that the generated code does not
16822 contain CSYNC or SSYNC instructions too soon after conditional branches.
16823 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
16825 @item -mno-csync-anomaly
16826 @opindex mno-csync-anomaly
16827 Don't generate extra code to prevent CSYNC or SSYNC instructions from
16828 occurring too soon after a conditional branch.
16832 When enabled, the compiler is free to take advantage of the knowledge that
16833 the entire program fits into the low 64k of memory.
16836 @opindex mno-low-64k
16837 Assume that the program is arbitrarily large. This is the default.
16839 @item -mstack-check-l1
16840 @opindex mstack-check-l1
16841 Do stack checking using information placed into L1 scratchpad memory by the
16844 @item -mid-shared-library
16845 @opindex mid-shared-library
16846 Generate code that supports shared libraries via the library ID method.
16847 This allows for execute in place and shared libraries in an environment
16848 without virtual memory management. This option implies @option{-fPIC}.
16849 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16851 @item -mno-id-shared-library
16852 @opindex mno-id-shared-library
16853 Generate code that doesn't assume ID-based shared libraries are being used.
16854 This is the default.
16856 @item -mleaf-id-shared-library
16857 @opindex mleaf-id-shared-library
16858 Generate code that supports shared libraries via the library ID method,
16859 but assumes that this library or executable won't link against any other
16860 ID shared libraries. That allows the compiler to use faster code for jumps
16863 @item -mno-leaf-id-shared-library
16864 @opindex mno-leaf-id-shared-library
16865 Do not assume that the code being compiled won't link against any ID shared
16866 libraries. Slower code is generated for jump and call insns.
16868 @item -mshared-library-id=n
16869 @opindex mshared-library-id
16870 Specifies the identification number of the ID-based shared library being
16871 compiled. Specifying a value of 0 generates more compact code; specifying
16872 other values forces the allocation of that number to the current
16873 library but is no more space- or time-efficient than omitting this option.
16877 Generate code that allows the data segment to be located in a different
16878 area of memory from the text segment. This allows for execute in place in
16879 an environment without virtual memory management by eliminating relocations
16880 against the text section.
16882 @item -mno-sep-data
16883 @opindex mno-sep-data
16884 Generate code that assumes that the data segment follows the text segment.
16885 This is the default.
16888 @itemx -mno-long-calls
16889 @opindex mlong-calls
16890 @opindex mno-long-calls
16891 Tells the compiler to perform function calls by first loading the
16892 address of the function into a register and then performing a subroutine
16893 call on this register. This switch is needed if the target function
16894 lies outside of the 24-bit addressing range of the offset-based
16895 version of subroutine call instruction.
16897 This feature is not enabled by default. Specifying
16898 @option{-mno-long-calls} restores the default behavior. Note these
16899 switches have no effect on how the compiler generates code to handle
16900 function calls via function pointers.
16904 Link with the fast floating-point library. This library relaxes some of
16905 the IEEE floating-point standard's rules for checking inputs against
16906 Not-a-Number (NAN), in the interest of performance.
16909 @opindex minline-plt
16910 Enable inlining of PLT entries in function calls to functions that are
16911 not known to bind locally. It has no effect without @option{-mfdpic}.
16914 @opindex mmulticore
16915 Build a standalone application for multicore Blackfin processors.
16916 This option causes proper start files and link scripts supporting
16917 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
16918 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
16920 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
16921 selects the one-application-per-core programming model. Without
16922 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
16923 programming model is used. In this model, the main function of Core B
16924 should be named as @code{coreb_main}.
16926 If this option is not used, the single-core application programming
16931 Build a standalone application for Core A of BF561 when using
16932 the one-application-per-core programming model. Proper start files
16933 and link scripts are used to support Core A, and the macro
16934 @code{__BFIN_COREA} is defined.
16935 This option can only be used in conjunction with @option{-mmulticore}.
16939 Build a standalone application for Core B of BF561 when using
16940 the one-application-per-core programming model. Proper start files
16941 and link scripts are used to support Core B, and the macro
16942 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
16943 should be used instead of @code{main}.
16944 This option can only be used in conjunction with @option{-mmulticore}.
16948 Build a standalone application for SDRAM. Proper start files and
16949 link scripts are used to put the application into SDRAM, and the macro
16950 @code{__BFIN_SDRAM} is defined.
16951 The loader should initialize SDRAM before loading the application.
16955 Assume that ICPLBs are enabled at run time. This has an effect on certain
16956 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
16957 are enabled; for standalone applications the default is off.
16961 @subsection C6X Options
16962 @cindex C6X Options
16965 @item -march=@var{name}
16967 This specifies the name of the target architecture. GCC uses this
16968 name to determine what kind of instructions it can emit when generating
16969 assembly code. Permissible names are: @samp{c62x},
16970 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
16973 @opindex mbig-endian
16974 Generate code for a big-endian target.
16976 @item -mlittle-endian
16977 @opindex mlittle-endian
16978 Generate code for a little-endian target. This is the default.
16982 Choose startup files and linker script suitable for the simulator.
16984 @item -msdata=default
16985 @opindex msdata=default
16986 Put small global and static data in the @code{.neardata} section,
16987 which is pointed to by register @code{B14}. Put small uninitialized
16988 global and static data in the @code{.bss} section, which is adjacent
16989 to the @code{.neardata} section. Put small read-only data into the
16990 @code{.rodata} section. The corresponding sections used for large
16991 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
16994 @opindex msdata=all
16995 Put all data, not just small objects, into the sections reserved for
16996 small data, and use addressing relative to the @code{B14} register to
17000 @opindex msdata=none
17001 Make no use of the sections reserved for small data, and use absolute
17002 addresses to access all data. Put all initialized global and static
17003 data in the @code{.fardata} section, and all uninitialized data in the
17004 @code{.far} section. Put all constant data into the @code{.const}
17009 @subsection CRIS Options
17010 @cindex CRIS Options
17012 These options are defined specifically for the CRIS ports.
17015 @item -march=@var{architecture-type}
17016 @itemx -mcpu=@var{architecture-type}
17019 Generate code for the specified architecture. The choices for
17020 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
17021 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
17022 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
17025 @item -mtune=@var{architecture-type}
17027 Tune to @var{architecture-type} everything applicable about the generated
17028 code, except for the ABI and the set of available instructions. The
17029 choices for @var{architecture-type} are the same as for
17030 @option{-march=@var{architecture-type}}.
17032 @item -mmax-stack-frame=@var{n}
17033 @opindex mmax-stack-frame
17034 Warn when the stack frame of a function exceeds @var{n} bytes.
17040 The options @option{-metrax4} and @option{-metrax100} are synonyms for
17041 @option{-march=v3} and @option{-march=v8} respectively.
17043 @item -mmul-bug-workaround
17044 @itemx -mno-mul-bug-workaround
17045 @opindex mmul-bug-workaround
17046 @opindex mno-mul-bug-workaround
17047 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
17048 models where it applies. This option is active by default.
17052 Enable CRIS-specific verbose debug-related information in the assembly
17053 code. This option also has the effect of turning off the @samp{#NO_APP}
17054 formatted-code indicator to the assembler at the beginning of the
17059 Do not use condition-code results from previous instruction; always emit
17060 compare and test instructions before use of condition codes.
17062 @item -mno-side-effects
17063 @opindex mno-side-effects
17064 Do not emit instructions with side effects in addressing modes other than
17067 @item -mstack-align
17068 @itemx -mno-stack-align
17069 @itemx -mdata-align
17070 @itemx -mno-data-align
17071 @itemx -mconst-align
17072 @itemx -mno-const-align
17073 @opindex mstack-align
17074 @opindex mno-stack-align
17075 @opindex mdata-align
17076 @opindex mno-data-align
17077 @opindex mconst-align
17078 @opindex mno-const-align
17079 These options (@samp{no-} options) arrange (eliminate arrangements) for the
17080 stack frame, individual data and constants to be aligned for the maximum
17081 single data access size for the chosen CPU model. The default is to
17082 arrange for 32-bit alignment. ABI details such as structure layout are
17083 not affected by these options.
17091 Similar to the stack- data- and const-align options above, these options
17092 arrange for stack frame, writable data and constants to all be 32-bit,
17093 16-bit or 8-bit aligned. The default is 32-bit alignment.
17095 @item -mno-prologue-epilogue
17096 @itemx -mprologue-epilogue
17097 @opindex mno-prologue-epilogue
17098 @opindex mprologue-epilogue
17099 With @option{-mno-prologue-epilogue}, the normal function prologue and
17100 epilogue which set up the stack frame are omitted and no return
17101 instructions or return sequences are generated in the code. Use this
17102 option only together with visual inspection of the compiled code: no
17103 warnings or errors are generated when call-saved registers must be saved,
17104 or storage for local variables needs to be allocated.
17108 @opindex mno-gotplt
17110 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
17111 instruction sequences that load addresses for functions from the PLT part
17112 of the GOT rather than (traditional on other architectures) calls to the
17113 PLT@. The default is @option{-mgotplt}.
17117 Legacy no-op option only recognized with the cris-axis-elf and
17118 cris-axis-linux-gnu targets.
17122 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
17126 This option, recognized for the cris-axis-elf, arranges
17127 to link with input-output functions from a simulator library. Code,
17128 initialized data and zero-initialized data are allocated consecutively.
17132 Like @option{-sim}, but pass linker options to locate initialized data at
17133 0x40000000 and zero-initialized data at 0x80000000.
17137 @subsection CR16 Options
17138 @cindex CR16 Options
17140 These options are defined specifically for the CR16 ports.
17146 Enable the use of multiply-accumulate instructions. Disabled by default.
17150 @opindex mcr16cplus
17152 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
17157 Links the library libsim.a which is in compatible with simulator. Applicable
17158 to ELF compiler only.
17162 Choose integer type as 32-bit wide.
17166 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
17168 @item -mdata-model=@var{model}
17169 @opindex mdata-model
17170 Choose a data model. The choices for @var{model} are @samp{near},
17171 @samp{far} or @samp{medium}. @samp{medium} is default.
17172 However, @samp{far} is not valid with @option{-mcr16c}, as the
17173 CR16C architecture does not support the far data model.
17176 @node Darwin Options
17177 @subsection Darwin Options
17178 @cindex Darwin options
17180 These options are defined for all architectures running the Darwin operating
17183 FSF GCC on Darwin does not create ``fat'' object files; it creates
17184 an object file for the single architecture that GCC was built to
17185 target. Apple's GCC on Darwin does create ``fat'' files if multiple
17186 @option{-arch} options are used; it does so by running the compiler or
17187 linker multiple times and joining the results together with
17190 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
17191 @samp{i686}) is determined by the flags that specify the ISA
17192 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
17193 @option{-force_cpusubtype_ALL} option can be used to override this.
17195 The Darwin tools vary in their behavior when presented with an ISA
17196 mismatch. The assembler, @file{as}, only permits instructions to
17197 be used that are valid for the subtype of the file it is generating,
17198 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
17199 The linker for shared libraries, @file{/usr/bin/libtool}, fails
17200 and prints an error if asked to create a shared library with a less
17201 restrictive subtype than its input files (for instance, trying to put
17202 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
17203 for executables, @command{ld}, quietly gives the executable the most
17204 restrictive subtype of any of its input files.
17209 Add the framework directory @var{dir} to the head of the list of
17210 directories to be searched for header files. These directories are
17211 interleaved with those specified by @option{-I} options and are
17212 scanned in a left-to-right order.
17214 A framework directory is a directory with frameworks in it. A
17215 framework is a directory with a @file{Headers} and/or
17216 @file{PrivateHeaders} directory contained directly in it that ends
17217 in @file{.framework}. The name of a framework is the name of this
17218 directory excluding the @file{.framework}. Headers associated with
17219 the framework are found in one of those two directories, with
17220 @file{Headers} being searched first. A subframework is a framework
17221 directory that is in a framework's @file{Frameworks} directory.
17222 Includes of subframework headers can only appear in a header of a
17223 framework that contains the subframework, or in a sibling subframework
17224 header. Two subframeworks are siblings if they occur in the same
17225 framework. A subframework should not have the same name as a
17226 framework; a warning is issued if this is violated. Currently a
17227 subframework cannot have subframeworks; in the future, the mechanism
17228 may be extended to support this. The standard frameworks can be found
17229 in @file{/System/Library/Frameworks} and
17230 @file{/Library/Frameworks}. An example include looks like
17231 @code{#include <Framework/header.h>}, where @file{Framework} denotes
17232 the name of the framework and @file{header.h} is found in the
17233 @file{PrivateHeaders} or @file{Headers} directory.
17235 @item -iframework@var{dir}
17236 @opindex iframework
17237 Like @option{-F} except the directory is a treated as a system
17238 directory. The main difference between this @option{-iframework} and
17239 @option{-F} is that with @option{-iframework} the compiler does not
17240 warn about constructs contained within header files found via
17241 @var{dir}. This option is valid only for the C family of languages.
17245 Emit debugging information for symbols that are used. For stabs
17246 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
17247 This is by default ON@.
17251 Emit debugging information for all symbols and types.
17253 @item -mmacosx-version-min=@var{version}
17254 The earliest version of MacOS X that this executable will run on
17255 is @var{version}. Typical values of @var{version} include @code{10.1},
17256 @code{10.2}, and @code{10.3.9}.
17258 If the compiler was built to use the system's headers by default,
17259 then the default for this option is the system version on which the
17260 compiler is running, otherwise the default is to make choices that
17261 are compatible with as many systems and code bases as possible.
17265 Enable kernel development mode. The @option{-mkernel} option sets
17266 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
17267 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
17268 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
17269 applicable. This mode also sets @option{-mno-altivec},
17270 @option{-msoft-float}, @option{-fno-builtin} and
17271 @option{-mlong-branch} for PowerPC targets.
17273 @item -mone-byte-bool
17274 @opindex mone-byte-bool
17275 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
17276 By default @code{sizeof(bool)} is @code{4} when compiling for
17277 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
17278 option has no effect on x86.
17280 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
17281 to generate code that is not binary compatible with code generated
17282 without that switch. Using this switch may require recompiling all
17283 other modules in a program, including system libraries. Use this
17284 switch to conform to a non-default data model.
17286 @item -mfix-and-continue
17287 @itemx -ffix-and-continue
17288 @itemx -findirect-data
17289 @opindex mfix-and-continue
17290 @opindex ffix-and-continue
17291 @opindex findirect-data
17292 Generate code suitable for fast turnaround development, such as to
17293 allow GDB to dynamically load @file{.o} files into already-running
17294 programs. @option{-findirect-data} and @option{-ffix-and-continue}
17295 are provided for backwards compatibility.
17299 Loads all members of static archive libraries.
17300 See man ld(1) for more information.
17302 @item -arch_errors_fatal
17303 @opindex arch_errors_fatal
17304 Cause the errors having to do with files that have the wrong architecture
17307 @item -bind_at_load
17308 @opindex bind_at_load
17309 Causes the output file to be marked such that the dynamic linker will
17310 bind all undefined references when the file is loaded or launched.
17314 Produce a Mach-o bundle format file.
17315 See man ld(1) for more information.
17317 @item -bundle_loader @var{executable}
17318 @opindex bundle_loader
17319 This option specifies the @var{executable} that will load the build
17320 output file being linked. See man ld(1) for more information.
17323 @opindex dynamiclib
17324 When passed this option, GCC produces a dynamic library instead of
17325 an executable when linking, using the Darwin @file{libtool} command.
17327 @item -force_cpusubtype_ALL
17328 @opindex force_cpusubtype_ALL
17329 This causes GCC's output file to have the @samp{ALL} subtype, instead of
17330 one controlled by the @option{-mcpu} or @option{-march} option.
17332 @item -allowable_client @var{client_name}
17333 @itemx -client_name
17334 @itemx -compatibility_version
17335 @itemx -current_version
17337 @itemx -dependency-file
17339 @itemx -dylinker_install_name
17341 @itemx -exported_symbols_list
17344 @itemx -flat_namespace
17345 @itemx -force_flat_namespace
17346 @itemx -headerpad_max_install_names
17349 @itemx -install_name
17350 @itemx -keep_private_externs
17351 @itemx -multi_module
17352 @itemx -multiply_defined
17353 @itemx -multiply_defined_unused
17356 @itemx -no_dead_strip_inits_and_terms
17357 @itemx -nofixprebinding
17358 @itemx -nomultidefs
17360 @itemx -noseglinkedit
17361 @itemx -pagezero_size
17363 @itemx -prebind_all_twolevel_modules
17364 @itemx -private_bundle
17366 @itemx -read_only_relocs
17368 @itemx -sectobjectsymbols
17372 @itemx -sectobjectsymbols
17375 @itemx -segs_read_only_addr
17377 @itemx -segs_read_write_addr
17378 @itemx -seg_addr_table
17379 @itemx -seg_addr_table_filename
17380 @itemx -seglinkedit
17382 @itemx -segs_read_only_addr
17383 @itemx -segs_read_write_addr
17384 @itemx -single_module
17386 @itemx -sub_library
17388 @itemx -sub_umbrella
17389 @itemx -twolevel_namespace
17392 @itemx -unexported_symbols_list
17393 @itemx -weak_reference_mismatches
17394 @itemx -whatsloaded
17395 @opindex allowable_client
17396 @opindex client_name
17397 @opindex compatibility_version
17398 @opindex current_version
17399 @opindex dead_strip
17400 @opindex dependency-file
17401 @opindex dylib_file
17402 @opindex dylinker_install_name
17404 @opindex exported_symbols_list
17406 @opindex flat_namespace
17407 @opindex force_flat_namespace
17408 @opindex headerpad_max_install_names
17409 @opindex image_base
17411 @opindex install_name
17412 @opindex keep_private_externs
17413 @opindex multi_module
17414 @opindex multiply_defined
17415 @opindex multiply_defined_unused
17416 @opindex noall_load
17417 @opindex no_dead_strip_inits_and_terms
17418 @opindex nofixprebinding
17419 @opindex nomultidefs
17421 @opindex noseglinkedit
17422 @opindex pagezero_size
17424 @opindex prebind_all_twolevel_modules
17425 @opindex private_bundle
17426 @opindex read_only_relocs
17428 @opindex sectobjectsymbols
17431 @opindex sectcreate
17432 @opindex sectobjectsymbols
17435 @opindex segs_read_only_addr
17436 @opindex segs_read_write_addr
17437 @opindex seg_addr_table
17438 @opindex seg_addr_table_filename
17439 @opindex seglinkedit
17441 @opindex segs_read_only_addr
17442 @opindex segs_read_write_addr
17443 @opindex single_module
17445 @opindex sub_library
17446 @opindex sub_umbrella
17447 @opindex twolevel_namespace
17450 @opindex unexported_symbols_list
17451 @opindex weak_reference_mismatches
17452 @opindex whatsloaded
17453 These options are passed to the Darwin linker. The Darwin linker man page
17454 describes them in detail.
17457 @node DEC Alpha Options
17458 @subsection DEC Alpha Options
17460 These @samp{-m} options are defined for the DEC Alpha implementations:
17463 @item -mno-soft-float
17464 @itemx -msoft-float
17465 @opindex mno-soft-float
17466 @opindex msoft-float
17467 Use (do not use) the hardware floating-point instructions for
17468 floating-point operations. When @option{-msoft-float} is specified,
17469 functions in @file{libgcc.a} are used to perform floating-point
17470 operations. Unless they are replaced by routines that emulate the
17471 floating-point operations, or compiled in such a way as to call such
17472 emulations routines, these routines issue floating-point
17473 operations. If you are compiling for an Alpha without floating-point
17474 operations, you must ensure that the library is built so as not to call
17477 Note that Alpha implementations without floating-point operations are
17478 required to have floating-point registers.
17481 @itemx -mno-fp-regs
17483 @opindex mno-fp-regs
17484 Generate code that uses (does not use) the floating-point register set.
17485 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
17486 register set is not used, floating-point operands are passed in integer
17487 registers as if they were integers and floating-point results are passed
17488 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
17489 so any function with a floating-point argument or return value called by code
17490 compiled with @option{-mno-fp-regs} must also be compiled with that
17493 A typical use of this option is building a kernel that does not use,
17494 and hence need not save and restore, any floating-point registers.
17498 The Alpha architecture implements floating-point hardware optimized for
17499 maximum performance. It is mostly compliant with the IEEE floating-point
17500 standard. However, for full compliance, software assistance is
17501 required. This option generates code fully IEEE-compliant code
17502 @emph{except} that the @var{inexact-flag} is not maintained (see below).
17503 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
17504 defined during compilation. The resulting code is less efficient but is
17505 able to correctly support denormalized numbers and exceptional IEEE
17506 values such as not-a-number and plus/minus infinity. Other Alpha
17507 compilers call this option @option{-ieee_with_no_inexact}.
17509 @item -mieee-with-inexact
17510 @opindex mieee-with-inexact
17511 This is like @option{-mieee} except the generated code also maintains
17512 the IEEE @var{inexact-flag}. Turning on this option causes the
17513 generated code to implement fully-compliant IEEE math. In addition to
17514 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
17515 macro. On some Alpha implementations the resulting code may execute
17516 significantly slower than the code generated by default. Since there is
17517 very little code that depends on the @var{inexact-flag}, you should
17518 normally not specify this option. Other Alpha compilers call this
17519 option @option{-ieee_with_inexact}.
17521 @item -mfp-trap-mode=@var{trap-mode}
17522 @opindex mfp-trap-mode
17523 This option controls what floating-point related traps are enabled.
17524 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
17525 The trap mode can be set to one of four values:
17529 This is the default (normal) setting. The only traps that are enabled
17530 are the ones that cannot be disabled in software (e.g., division by zero
17534 In addition to the traps enabled by @samp{n}, underflow traps are enabled
17538 Like @samp{u}, but the instructions are marked to be safe for software
17539 completion (see Alpha architecture manual for details).
17542 Like @samp{su}, but inexact traps are enabled as well.
17545 @item -mfp-rounding-mode=@var{rounding-mode}
17546 @opindex mfp-rounding-mode
17547 Selects the IEEE rounding mode. Other Alpha compilers call this option
17548 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
17553 Normal IEEE rounding mode. Floating-point numbers are rounded towards
17554 the nearest machine number or towards the even machine number in case
17558 Round towards minus infinity.
17561 Chopped rounding mode. Floating-point numbers are rounded towards zero.
17564 Dynamic rounding mode. A field in the floating-point control register
17565 (@var{fpcr}, see Alpha architecture reference manual) controls the
17566 rounding mode in effect. The C library initializes this register for
17567 rounding towards plus infinity. Thus, unless your program modifies the
17568 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
17571 @item -mtrap-precision=@var{trap-precision}
17572 @opindex mtrap-precision
17573 In the Alpha architecture, floating-point traps are imprecise. This
17574 means without software assistance it is impossible to recover from a
17575 floating trap and program execution normally needs to be terminated.
17576 GCC can generate code that can assist operating system trap handlers
17577 in determining the exact location that caused a floating-point trap.
17578 Depending on the requirements of an application, different levels of
17579 precisions can be selected:
17583 Program precision. This option is the default and means a trap handler
17584 can only identify which program caused a floating-point exception.
17587 Function precision. The trap handler can determine the function that
17588 caused a floating-point exception.
17591 Instruction precision. The trap handler can determine the exact
17592 instruction that caused a floating-point exception.
17595 Other Alpha compilers provide the equivalent options called
17596 @option{-scope_safe} and @option{-resumption_safe}.
17598 @item -mieee-conformant
17599 @opindex mieee-conformant
17600 This option marks the generated code as IEEE conformant. You must not
17601 use this option unless you also specify @option{-mtrap-precision=i} and either
17602 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
17603 is to emit the line @samp{.eflag 48} in the function prologue of the
17604 generated assembly file.
17606 @item -mbuild-constants
17607 @opindex mbuild-constants
17608 Normally GCC examines a 32- or 64-bit integer constant to
17609 see if it can construct it from smaller constants in two or three
17610 instructions. If it cannot, it outputs the constant as a literal and
17611 generates code to load it from the data segment at run time.
17613 Use this option to require GCC to construct @emph{all} integer constants
17614 using code, even if it takes more instructions (the maximum is six).
17616 You typically use this option to build a shared library dynamic
17617 loader. Itself a shared library, it must relocate itself in memory
17618 before it can find the variables and constants in its own data segment.
17636 Indicate whether GCC should generate code to use the optional BWX,
17637 CIX, FIX and MAX instruction sets. The default is to use the instruction
17638 sets supported by the CPU type specified via @option{-mcpu=} option or that
17639 of the CPU on which GCC was built if none is specified.
17642 @itemx -mfloat-ieee
17643 @opindex mfloat-vax
17644 @opindex mfloat-ieee
17645 Generate code that uses (does not use) VAX F and G floating-point
17646 arithmetic instead of IEEE single and double precision.
17648 @item -mexplicit-relocs
17649 @itemx -mno-explicit-relocs
17650 @opindex mexplicit-relocs
17651 @opindex mno-explicit-relocs
17652 Older Alpha assemblers provided no way to generate symbol relocations
17653 except via assembler macros. Use of these macros does not allow
17654 optimal instruction scheduling. GNU binutils as of version 2.12
17655 supports a new syntax that allows the compiler to explicitly mark
17656 which relocations should apply to which instructions. This option
17657 is mostly useful for debugging, as GCC detects the capabilities of
17658 the assembler when it is built and sets the default accordingly.
17661 @itemx -mlarge-data
17662 @opindex msmall-data
17663 @opindex mlarge-data
17664 When @option{-mexplicit-relocs} is in effect, static data is
17665 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
17666 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
17667 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
17668 16-bit relocations off of the @code{$gp} register. This limits the
17669 size of the small data area to 64KB, but allows the variables to be
17670 directly accessed via a single instruction.
17672 The default is @option{-mlarge-data}. With this option the data area
17673 is limited to just below 2GB@. Programs that require more than 2GB of
17674 data must use @code{malloc} or @code{mmap} to allocate the data in the
17675 heap instead of in the program's data segment.
17677 When generating code for shared libraries, @option{-fpic} implies
17678 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
17681 @itemx -mlarge-text
17682 @opindex msmall-text
17683 @opindex mlarge-text
17684 When @option{-msmall-text} is used, the compiler assumes that the
17685 code of the entire program (or shared library) fits in 4MB, and is
17686 thus reachable with a branch instruction. When @option{-msmall-data}
17687 is used, the compiler can assume that all local symbols share the
17688 same @code{$gp} value, and thus reduce the number of instructions
17689 required for a function call from 4 to 1.
17691 The default is @option{-mlarge-text}.
17693 @item -mcpu=@var{cpu_type}
17695 Set the instruction set and instruction scheduling parameters for
17696 machine type @var{cpu_type}. You can specify either the @samp{EV}
17697 style name or the corresponding chip number. GCC supports scheduling
17698 parameters for the EV4, EV5 and EV6 family of processors and
17699 chooses the default values for the instruction set from the processor
17700 you specify. If you do not specify a processor type, GCC defaults
17701 to the processor on which the compiler was built.
17703 Supported values for @var{cpu_type} are
17709 Schedules as an EV4 and has no instruction set extensions.
17713 Schedules as an EV5 and has no instruction set extensions.
17717 Schedules as an EV5 and supports the BWX extension.
17722 Schedules as an EV5 and supports the BWX and MAX extensions.
17726 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
17730 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
17733 Native toolchains also support the value @samp{native},
17734 which selects the best architecture option for the host processor.
17735 @option{-mcpu=native} has no effect if GCC does not recognize
17738 @item -mtune=@var{cpu_type}
17740 Set only the instruction scheduling parameters for machine type
17741 @var{cpu_type}. The instruction set is not changed.
17743 Native toolchains also support the value @samp{native},
17744 which selects the best architecture option for the host processor.
17745 @option{-mtune=native} has no effect if GCC does not recognize
17748 @item -mmemory-latency=@var{time}
17749 @opindex mmemory-latency
17750 Sets the latency the scheduler should assume for typical memory
17751 references as seen by the application. This number is highly
17752 dependent on the memory access patterns used by the application
17753 and the size of the external cache on the machine.
17755 Valid options for @var{time} are
17759 A decimal number representing clock cycles.
17765 The compiler contains estimates of the number of clock cycles for
17766 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
17767 (also called Dcache, Scache, and Bcache), as well as to main memory.
17768 Note that L3 is only valid for EV5.
17774 @subsection FR30 Options
17775 @cindex FR30 Options
17777 These options are defined specifically for the FR30 port.
17781 @item -msmall-model
17782 @opindex msmall-model
17783 Use the small address space model. This can produce smaller code, but
17784 it does assume that all symbolic values and addresses fit into a
17789 Assume that runtime support has been provided and so there is no need
17790 to include the simulator library (@file{libsim.a}) on the linker
17796 @subsection FT32 Options
17797 @cindex FT32 Options
17799 These options are defined specifically for the FT32 port.
17805 Specifies that the program will be run on the simulator. This causes
17806 an alternate runtime startup and library to be linked.
17807 You must not use this option when generating programs that will run on
17808 real hardware; you must provide your own runtime library for whatever
17809 I/O functions are needed.
17813 Enable Local Register Allocation. This is still experimental for FT32,
17814 so by default the compiler uses standard reload.
17818 Do not use div and mod instructions.
17822 Enable use of the extended instructions of the FT32B processor.
17826 Compress all code using the Ft32B code compression scheme.
17830 Do not generate code that reads program memory.
17835 @subsection FRV Options
17836 @cindex FRV Options
17842 Only use the first 32 general-purpose registers.
17847 Use all 64 general-purpose registers.
17852 Use only the first 32 floating-point registers.
17857 Use all 64 floating-point registers.
17860 @opindex mhard-float
17862 Use hardware instructions for floating-point operations.
17865 @opindex msoft-float
17867 Use library routines for floating-point operations.
17872 Dynamically allocate condition code registers.
17877 Do not try to dynamically allocate condition code registers, only
17878 use @code{icc0} and @code{fcc0}.
17883 Change ABI to use double word insns.
17888 Do not use double word instructions.
17893 Use floating-point double instructions.
17896 @opindex mno-double
17898 Do not use floating-point double instructions.
17903 Use media instructions.
17908 Do not use media instructions.
17913 Use multiply and add/subtract instructions.
17916 @opindex mno-muladd
17918 Do not use multiply and add/subtract instructions.
17923 Select the FDPIC ABI, which uses function descriptors to represent
17924 pointers to functions. Without any PIC/PIE-related options, it
17925 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
17926 assumes GOT entries and small data are within a 12-bit range from the
17927 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
17928 are computed with 32 bits.
17929 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17932 @opindex minline-plt
17934 Enable inlining of PLT entries in function calls to functions that are
17935 not known to bind locally. It has no effect without @option{-mfdpic}.
17936 It's enabled by default if optimizing for speed and compiling for
17937 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
17938 optimization option such as @option{-O3} or above is present in the
17944 Assume a large TLS segment when generating thread-local code.
17949 Do not assume a large TLS segment when generating thread-local code.
17954 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
17955 that is known to be in read-only sections. It's enabled by default,
17956 except for @option{-fpic} or @option{-fpie}: even though it may help
17957 make the global offset table smaller, it trades 1 instruction for 4.
17958 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
17959 one of which may be shared by multiple symbols, and it avoids the need
17960 for a GOT entry for the referenced symbol, so it's more likely to be a
17961 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
17963 @item -multilib-library-pic
17964 @opindex multilib-library-pic
17966 Link with the (library, not FD) pic libraries. It's implied by
17967 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
17968 @option{-fpic} without @option{-mfdpic}. You should never have to use
17972 @opindex mlinked-fp
17974 Follow the EABI requirement of always creating a frame pointer whenever
17975 a stack frame is allocated. This option is enabled by default and can
17976 be disabled with @option{-mno-linked-fp}.
17979 @opindex mlong-calls
17981 Use indirect addressing to call functions outside the current
17982 compilation unit. This allows the functions to be placed anywhere
17983 within the 32-bit address space.
17985 @item -malign-labels
17986 @opindex malign-labels
17988 Try to align labels to an 8-byte boundary by inserting NOPs into the
17989 previous packet. This option only has an effect when VLIW packing
17990 is enabled. It doesn't create new packets; it merely adds NOPs to
17993 @item -mlibrary-pic
17994 @opindex mlibrary-pic
17996 Generate position-independent EABI code.
18001 Use only the first four media accumulator registers.
18006 Use all eight media accumulator registers.
18011 Pack VLIW instructions.
18016 Do not pack VLIW instructions.
18019 @opindex mno-eflags
18021 Do not mark ABI switches in e_flags.
18024 @opindex mcond-move
18026 Enable the use of conditional-move instructions (default).
18028 This switch is mainly for debugging the compiler and will likely be removed
18029 in a future version.
18031 @item -mno-cond-move
18032 @opindex mno-cond-move
18034 Disable the use of conditional-move instructions.
18036 This switch is mainly for debugging the compiler and will likely be removed
18037 in a future version.
18042 Enable the use of conditional set instructions (default).
18044 This switch is mainly for debugging the compiler and will likely be removed
18045 in a future version.
18050 Disable the use of conditional set instructions.
18052 This switch is mainly for debugging the compiler and will likely be removed
18053 in a future version.
18056 @opindex mcond-exec
18058 Enable the use of conditional execution (default).
18060 This switch is mainly for debugging the compiler and will likely be removed
18061 in a future version.
18063 @item -mno-cond-exec
18064 @opindex mno-cond-exec
18066 Disable the use of conditional execution.
18068 This switch is mainly for debugging the compiler and will likely be removed
18069 in a future version.
18071 @item -mvliw-branch
18072 @opindex mvliw-branch
18074 Run a pass to pack branches into VLIW instructions (default).
18076 This switch is mainly for debugging the compiler and will likely be removed
18077 in a future version.
18079 @item -mno-vliw-branch
18080 @opindex mno-vliw-branch
18082 Do not run a pass to pack branches into VLIW instructions.
18084 This switch is mainly for debugging the compiler and will likely be removed
18085 in a future version.
18087 @item -mmulti-cond-exec
18088 @opindex mmulti-cond-exec
18090 Enable optimization of @code{&&} and @code{||} in conditional execution
18093 This switch is mainly for debugging the compiler and will likely be removed
18094 in a future version.
18096 @item -mno-multi-cond-exec
18097 @opindex mno-multi-cond-exec
18099 Disable optimization of @code{&&} and @code{||} in conditional execution.
18101 This switch is mainly for debugging the compiler and will likely be removed
18102 in a future version.
18104 @item -mnested-cond-exec
18105 @opindex mnested-cond-exec
18107 Enable nested conditional execution optimizations (default).
18109 This switch is mainly for debugging the compiler and will likely be removed
18110 in a future version.
18112 @item -mno-nested-cond-exec
18113 @opindex mno-nested-cond-exec
18115 Disable nested conditional execution optimizations.
18117 This switch is mainly for debugging the compiler and will likely be removed
18118 in a future version.
18120 @item -moptimize-membar
18121 @opindex moptimize-membar
18123 This switch removes redundant @code{membar} instructions from the
18124 compiler-generated code. It is enabled by default.
18126 @item -mno-optimize-membar
18127 @opindex mno-optimize-membar
18129 This switch disables the automatic removal of redundant @code{membar}
18130 instructions from the generated code.
18132 @item -mtomcat-stats
18133 @opindex mtomcat-stats
18135 Cause gas to print out tomcat statistics.
18137 @item -mcpu=@var{cpu}
18140 Select the processor type for which to generate code. Possible values are
18141 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
18142 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
18146 @node GNU/Linux Options
18147 @subsection GNU/Linux Options
18149 These @samp{-m} options are defined for GNU/Linux targets:
18154 Use the GNU C library. This is the default except
18155 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
18156 @samp{*-*-linux-*android*} targets.
18160 Use uClibc C library. This is the default on
18161 @samp{*-*-linux-*uclibc*} targets.
18165 Use the musl C library. This is the default on
18166 @samp{*-*-linux-*musl*} targets.
18170 Use Bionic C library. This is the default on
18171 @samp{*-*-linux-*android*} targets.
18175 Compile code compatible with Android platform. This is the default on
18176 @samp{*-*-linux-*android*} targets.
18178 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
18179 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
18180 this option makes the GCC driver pass Android-specific options to the linker.
18181 Finally, this option causes the preprocessor macro @code{__ANDROID__}
18184 @item -tno-android-cc
18185 @opindex tno-android-cc
18186 Disable compilation effects of @option{-mandroid}, i.e., do not enable
18187 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
18188 @option{-fno-rtti} by default.
18190 @item -tno-android-ld
18191 @opindex tno-android-ld
18192 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
18193 linking options to the linker.
18197 @node H8/300 Options
18198 @subsection H8/300 Options
18200 These @samp{-m} options are defined for the H8/300 implementations:
18205 Shorten some address references at link time, when possible; uses the
18206 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
18207 ld, Using ld}, for a fuller description.
18211 Generate code for the H8/300H@.
18215 Generate code for the H8S@.
18219 Generate code for the H8S and H8/300H in the normal mode. This switch
18220 must be used either with @option{-mh} or @option{-ms}.
18224 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
18228 Extended registers are stored on stack before execution of function
18229 with monitor attribute. Default option is @option{-mexr}.
18230 This option is valid only for H8S targets.
18234 Extended registers are not stored on stack before execution of function
18235 with monitor attribute. Default option is @option{-mno-exr}.
18236 This option is valid only for H8S targets.
18240 Make @code{int} data 32 bits by default.
18243 @opindex malign-300
18244 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
18245 The default for the H8/300H and H8S is to align longs and floats on
18247 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
18248 This option has no effect on the H8/300.
18252 @subsection HPPA Options
18253 @cindex HPPA Options
18255 These @samp{-m} options are defined for the HPPA family of computers:
18258 @item -march=@var{architecture-type}
18260 Generate code for the specified architecture. The choices for
18261 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
18262 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
18263 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
18264 architecture option for your machine. Code compiled for lower numbered
18265 architectures runs on higher numbered architectures, but not the
18268 @item -mpa-risc-1-0
18269 @itemx -mpa-risc-1-1
18270 @itemx -mpa-risc-2-0
18271 @opindex mpa-risc-1-0
18272 @opindex mpa-risc-1-1
18273 @opindex mpa-risc-2-0
18274 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
18276 @item -mcaller-copies
18277 @opindex mcaller-copies
18278 The caller copies function arguments passed by hidden reference. This
18279 option should be used with care as it is not compatible with the default
18280 32-bit runtime. However, only aggregates larger than eight bytes are
18281 passed by hidden reference and the option provides better compatibility
18284 @item -mjump-in-delay
18285 @opindex mjump-in-delay
18286 This option is ignored and provided for compatibility purposes only.
18288 @item -mdisable-fpregs
18289 @opindex mdisable-fpregs
18290 Prevent floating-point registers from being used in any manner. This is
18291 necessary for compiling kernels that perform lazy context switching of
18292 floating-point registers. If you use this option and attempt to perform
18293 floating-point operations, the compiler aborts.
18295 @item -mdisable-indexing
18296 @opindex mdisable-indexing
18297 Prevent the compiler from using indexing address modes. This avoids some
18298 rather obscure problems when compiling MIG generated code under MACH@.
18300 @item -mno-space-regs
18301 @opindex mno-space-regs
18302 Generate code that assumes the target has no space registers. This allows
18303 GCC to generate faster indirect calls and use unscaled index address modes.
18305 Such code is suitable for level 0 PA systems and kernels.
18307 @item -mfast-indirect-calls
18308 @opindex mfast-indirect-calls
18309 Generate code that assumes calls never cross space boundaries. This
18310 allows GCC to emit code that performs faster indirect calls.
18312 This option does not work in the presence of shared libraries or nested
18315 @item -mfixed-range=@var{register-range}
18316 @opindex mfixed-range
18317 Generate code treating the given register range as fixed registers.
18318 A fixed register is one that the register allocator cannot use. This is
18319 useful when compiling kernel code. A register range is specified as
18320 two registers separated by a dash. Multiple register ranges can be
18321 specified separated by a comma.
18323 @item -mlong-load-store
18324 @opindex mlong-load-store
18325 Generate 3-instruction load and store sequences as sometimes required by
18326 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
18329 @item -mportable-runtime
18330 @opindex mportable-runtime
18331 Use the portable calling conventions proposed by HP for ELF systems.
18335 Enable the use of assembler directives only GAS understands.
18337 @item -mschedule=@var{cpu-type}
18339 Schedule code according to the constraints for the machine type
18340 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
18341 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
18342 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
18343 proper scheduling option for your machine. The default scheduling is
18347 @opindex mlinker-opt
18348 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
18349 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
18350 linkers in which they give bogus error messages when linking some programs.
18353 @opindex msoft-float
18354 Generate output containing library calls for floating point.
18355 @strong{Warning:} the requisite libraries are not available for all HPPA
18356 targets. Normally the facilities of the machine's usual C compiler are
18357 used, but this cannot be done directly in cross-compilation. You must make
18358 your own arrangements to provide suitable library functions for
18361 @option{-msoft-float} changes the calling convention in the output file;
18362 therefore, it is only useful if you compile @emph{all} of a program with
18363 this option. In particular, you need to compile @file{libgcc.a}, the
18364 library that comes with GCC, with @option{-msoft-float} in order for
18369 Generate the predefine, @code{_SIO}, for server IO@. The default is
18370 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
18371 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
18372 options are available under HP-UX and HI-UX@.
18376 Use options specific to GNU @command{ld}.
18377 This passes @option{-shared} to @command{ld} when
18378 building a shared library. It is the default when GCC is configured,
18379 explicitly or implicitly, with the GNU linker. This option does not
18380 affect which @command{ld} is called; it only changes what parameters
18381 are passed to that @command{ld}.
18382 The @command{ld} that is called is determined by the
18383 @option{--with-ld} configure option, GCC's program search path, and
18384 finally by the user's @env{PATH}. The linker used by GCC can be printed
18385 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
18386 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18390 Use options specific to HP @command{ld}.
18391 This passes @option{-b} to @command{ld} when building
18392 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
18393 links. It is the default when GCC is configured, explicitly or
18394 implicitly, with the HP linker. This option does not affect
18395 which @command{ld} is called; it only changes what parameters are passed to that
18397 The @command{ld} that is called is determined by the @option{--with-ld}
18398 configure option, GCC's program search path, and finally by the user's
18399 @env{PATH}. The linker used by GCC can be printed using @samp{which
18400 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
18401 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18404 @opindex mno-long-calls
18405 Generate code that uses long call sequences. This ensures that a call
18406 is always able to reach linker generated stubs. The default is to generate
18407 long calls only when the distance from the call site to the beginning
18408 of the function or translation unit, as the case may be, exceeds a
18409 predefined limit set by the branch type being used. The limits for
18410 normal calls are 7,600,000 and 240,000 bytes, respectively for the
18411 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
18414 Distances are measured from the beginning of functions when using the
18415 @option{-ffunction-sections} option, or when using the @option{-mgas}
18416 and @option{-mno-portable-runtime} options together under HP-UX with
18419 It is normally not desirable to use this option as it degrades
18420 performance. However, it may be useful in large applications,
18421 particularly when partial linking is used to build the application.
18423 The types of long calls used depends on the capabilities of the
18424 assembler and linker, and the type of code being generated. The
18425 impact on systems that support long absolute calls, and long pic
18426 symbol-difference or pc-relative calls should be relatively small.
18427 However, an indirect call is used on 32-bit ELF systems in pic code
18428 and it is quite long.
18430 @item -munix=@var{unix-std}
18432 Generate compiler predefines and select a startfile for the specified
18433 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
18434 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
18435 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
18436 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
18437 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
18440 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
18441 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
18442 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
18443 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
18444 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
18445 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
18447 It is @emph{important} to note that this option changes the interfaces
18448 for various library routines. It also affects the operational behavior
18449 of the C library. Thus, @emph{extreme} care is needed in using this
18452 Library code that is intended to operate with more than one UNIX
18453 standard must test, set and restore the variable @code{__xpg4_extended_mask}
18454 as appropriate. Most GNU software doesn't provide this capability.
18458 Suppress the generation of link options to search libdld.sl when the
18459 @option{-static} option is specified on HP-UX 10 and later.
18463 The HP-UX implementation of setlocale in libc has a dependency on
18464 libdld.sl. There isn't an archive version of libdld.sl. Thus,
18465 when the @option{-static} option is specified, special link options
18466 are needed to resolve this dependency.
18468 On HP-UX 10 and later, the GCC driver adds the necessary options to
18469 link with libdld.sl when the @option{-static} option is specified.
18470 This causes the resulting binary to be dynamic. On the 64-bit port,
18471 the linkers generate dynamic binaries by default in any case. The
18472 @option{-nolibdld} option can be used to prevent the GCC driver from
18473 adding these link options.
18477 Add support for multithreading with the @dfn{dce thread} library
18478 under HP-UX@. This option sets flags for both the preprocessor and
18482 @node IA-64 Options
18483 @subsection IA-64 Options
18484 @cindex IA-64 Options
18486 These are the @samp{-m} options defined for the Intel IA-64 architecture.
18490 @opindex mbig-endian
18491 Generate code for a big-endian target. This is the default for HP-UX@.
18493 @item -mlittle-endian
18494 @opindex mlittle-endian
18495 Generate code for a little-endian target. This is the default for AIX5
18501 @opindex mno-gnu-as
18502 Generate (or don't) code for the GNU assembler. This is the default.
18503 @c Also, this is the default if the configure option @option{--with-gnu-as}
18509 @opindex mno-gnu-ld
18510 Generate (or don't) code for the GNU linker. This is the default.
18511 @c Also, this is the default if the configure option @option{--with-gnu-ld}
18516 Generate code that does not use a global pointer register. The result
18517 is not position independent code, and violates the IA-64 ABI@.
18519 @item -mvolatile-asm-stop
18520 @itemx -mno-volatile-asm-stop
18521 @opindex mvolatile-asm-stop
18522 @opindex mno-volatile-asm-stop
18523 Generate (or don't) a stop bit immediately before and after volatile asm
18526 @item -mregister-names
18527 @itemx -mno-register-names
18528 @opindex mregister-names
18529 @opindex mno-register-names
18530 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
18531 the stacked registers. This may make assembler output more readable.
18537 Disable (or enable) optimizations that use the small data section. This may
18538 be useful for working around optimizer bugs.
18540 @item -mconstant-gp
18541 @opindex mconstant-gp
18542 Generate code that uses a single constant global pointer value. This is
18543 useful when compiling kernel code.
18547 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
18548 This is useful when compiling firmware code.
18550 @item -minline-float-divide-min-latency
18551 @opindex minline-float-divide-min-latency
18552 Generate code for inline divides of floating-point values
18553 using the minimum latency algorithm.
18555 @item -minline-float-divide-max-throughput
18556 @opindex minline-float-divide-max-throughput
18557 Generate code for inline divides of floating-point values
18558 using the maximum throughput algorithm.
18560 @item -mno-inline-float-divide
18561 @opindex mno-inline-float-divide
18562 Do not generate inline code for divides of floating-point values.
18564 @item -minline-int-divide-min-latency
18565 @opindex minline-int-divide-min-latency
18566 Generate code for inline divides of integer values
18567 using the minimum latency algorithm.
18569 @item -minline-int-divide-max-throughput
18570 @opindex minline-int-divide-max-throughput
18571 Generate code for inline divides of integer values
18572 using the maximum throughput algorithm.
18574 @item -mno-inline-int-divide
18575 @opindex mno-inline-int-divide
18576 Do not generate inline code for divides of integer values.
18578 @item -minline-sqrt-min-latency
18579 @opindex minline-sqrt-min-latency
18580 Generate code for inline square roots
18581 using the minimum latency algorithm.
18583 @item -minline-sqrt-max-throughput
18584 @opindex minline-sqrt-max-throughput
18585 Generate code for inline square roots
18586 using the maximum throughput algorithm.
18588 @item -mno-inline-sqrt
18589 @opindex mno-inline-sqrt
18590 Do not generate inline code for @code{sqrt}.
18593 @itemx -mno-fused-madd
18594 @opindex mfused-madd
18595 @opindex mno-fused-madd
18596 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
18597 instructions. The default is to use these instructions.
18599 @item -mno-dwarf2-asm
18600 @itemx -mdwarf2-asm
18601 @opindex mno-dwarf2-asm
18602 @opindex mdwarf2-asm
18603 Don't (or do) generate assembler code for the DWARF line number debugging
18604 info. This may be useful when not using the GNU assembler.
18606 @item -mearly-stop-bits
18607 @itemx -mno-early-stop-bits
18608 @opindex mearly-stop-bits
18609 @opindex mno-early-stop-bits
18610 Allow stop bits to be placed earlier than immediately preceding the
18611 instruction that triggered the stop bit. This can improve instruction
18612 scheduling, but does not always do so.
18614 @item -mfixed-range=@var{register-range}
18615 @opindex mfixed-range
18616 Generate code treating the given register range as fixed registers.
18617 A fixed register is one that the register allocator cannot use. This is
18618 useful when compiling kernel code. A register range is specified as
18619 two registers separated by a dash. Multiple register ranges can be
18620 specified separated by a comma.
18622 @item -mtls-size=@var{tls-size}
18624 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
18627 @item -mtune=@var{cpu-type}
18629 Tune the instruction scheduling for a particular CPU, Valid values are
18630 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
18631 and @samp{mckinley}.
18637 Generate code for a 32-bit or 64-bit environment.
18638 The 32-bit environment sets int, long and pointer to 32 bits.
18639 The 64-bit environment sets int to 32 bits and long and pointer
18640 to 64 bits. These are HP-UX specific flags.
18642 @item -mno-sched-br-data-spec
18643 @itemx -msched-br-data-spec
18644 @opindex mno-sched-br-data-spec
18645 @opindex msched-br-data-spec
18646 (Dis/En)able data speculative scheduling before reload.
18647 This results in generation of @code{ld.a} instructions and
18648 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18649 The default setting is disabled.
18651 @item -msched-ar-data-spec
18652 @itemx -mno-sched-ar-data-spec
18653 @opindex msched-ar-data-spec
18654 @opindex mno-sched-ar-data-spec
18655 (En/Dis)able data speculative scheduling after reload.
18656 This results in generation of @code{ld.a} instructions and
18657 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18658 The default setting is enabled.
18660 @item -mno-sched-control-spec
18661 @itemx -msched-control-spec
18662 @opindex mno-sched-control-spec
18663 @opindex msched-control-spec
18664 (Dis/En)able control speculative scheduling. This feature is
18665 available only during region scheduling (i.e.@: before reload).
18666 This results in generation of the @code{ld.s} instructions and
18667 the corresponding check instructions @code{chk.s}.
18668 The default setting is disabled.
18670 @item -msched-br-in-data-spec
18671 @itemx -mno-sched-br-in-data-spec
18672 @opindex msched-br-in-data-spec
18673 @opindex mno-sched-br-in-data-spec
18674 (En/Dis)able speculative scheduling of the instructions that
18675 are dependent on the data speculative loads before reload.
18676 This is effective only with @option{-msched-br-data-spec} enabled.
18677 The default setting is enabled.
18679 @item -msched-ar-in-data-spec
18680 @itemx -mno-sched-ar-in-data-spec
18681 @opindex msched-ar-in-data-spec
18682 @opindex mno-sched-ar-in-data-spec
18683 (En/Dis)able speculative scheduling of the instructions that
18684 are dependent on the data speculative loads after reload.
18685 This is effective only with @option{-msched-ar-data-spec} enabled.
18686 The default setting is enabled.
18688 @item -msched-in-control-spec
18689 @itemx -mno-sched-in-control-spec
18690 @opindex msched-in-control-spec
18691 @opindex mno-sched-in-control-spec
18692 (En/Dis)able speculative scheduling of the instructions that
18693 are dependent on the control speculative loads.
18694 This is effective only with @option{-msched-control-spec} enabled.
18695 The default setting is enabled.
18697 @item -mno-sched-prefer-non-data-spec-insns
18698 @itemx -msched-prefer-non-data-spec-insns
18699 @opindex mno-sched-prefer-non-data-spec-insns
18700 @opindex msched-prefer-non-data-spec-insns
18701 If enabled, data-speculative instructions are chosen for schedule
18702 only if there are no other choices at the moment. This makes
18703 the use of the data speculation much more conservative.
18704 The default setting is disabled.
18706 @item -mno-sched-prefer-non-control-spec-insns
18707 @itemx -msched-prefer-non-control-spec-insns
18708 @opindex mno-sched-prefer-non-control-spec-insns
18709 @opindex msched-prefer-non-control-spec-insns
18710 If enabled, control-speculative instructions are chosen for schedule
18711 only if there are no other choices at the moment. This makes
18712 the use of the control speculation much more conservative.
18713 The default setting is disabled.
18715 @item -mno-sched-count-spec-in-critical-path
18716 @itemx -msched-count-spec-in-critical-path
18717 @opindex mno-sched-count-spec-in-critical-path
18718 @opindex msched-count-spec-in-critical-path
18719 If enabled, speculative dependencies are considered during
18720 computation of the instructions priorities. This makes the use of the
18721 speculation a bit more conservative.
18722 The default setting is disabled.
18724 @item -msched-spec-ldc
18725 @opindex msched-spec-ldc
18726 Use a simple data speculation check. This option is on by default.
18728 @item -msched-control-spec-ldc
18729 @opindex msched-spec-ldc
18730 Use a simple check for control speculation. This option is on by default.
18732 @item -msched-stop-bits-after-every-cycle
18733 @opindex msched-stop-bits-after-every-cycle
18734 Place a stop bit after every cycle when scheduling. This option is on
18737 @item -msched-fp-mem-deps-zero-cost
18738 @opindex msched-fp-mem-deps-zero-cost
18739 Assume that floating-point stores and loads are not likely to cause a conflict
18740 when placed into the same instruction group. This option is disabled by
18743 @item -msel-sched-dont-check-control-spec
18744 @opindex msel-sched-dont-check-control-spec
18745 Generate checks for control speculation in selective scheduling.
18746 This flag is disabled by default.
18748 @item -msched-max-memory-insns=@var{max-insns}
18749 @opindex msched-max-memory-insns
18750 Limit on the number of memory insns per instruction group, giving lower
18751 priority to subsequent memory insns attempting to schedule in the same
18752 instruction group. Frequently useful to prevent cache bank conflicts.
18753 The default value is 1.
18755 @item -msched-max-memory-insns-hard-limit
18756 @opindex msched-max-memory-insns-hard-limit
18757 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
18758 disallowing more than that number in an instruction group.
18759 Otherwise, the limit is ``soft'', meaning that non-memory operations
18760 are preferred when the limit is reached, but memory operations may still
18766 @subsection LM32 Options
18767 @cindex LM32 options
18769 These @option{-m} options are defined for the LatticeMico32 architecture:
18772 @item -mbarrel-shift-enabled
18773 @opindex mbarrel-shift-enabled
18774 Enable barrel-shift instructions.
18776 @item -mdivide-enabled
18777 @opindex mdivide-enabled
18778 Enable divide and modulus instructions.
18780 @item -mmultiply-enabled
18781 @opindex multiply-enabled
18782 Enable multiply instructions.
18784 @item -msign-extend-enabled
18785 @opindex msign-extend-enabled
18786 Enable sign extend instructions.
18788 @item -muser-enabled
18789 @opindex muser-enabled
18790 Enable user-defined instructions.
18795 @subsection M32C Options
18796 @cindex M32C options
18799 @item -mcpu=@var{name}
18801 Select the CPU for which code is generated. @var{name} may be one of
18802 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
18803 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
18804 the M32C/80 series.
18808 Specifies that the program will be run on the simulator. This causes
18809 an alternate runtime library to be linked in which supports, for
18810 example, file I/O@. You must not use this option when generating
18811 programs that will run on real hardware; you must provide your own
18812 runtime library for whatever I/O functions are needed.
18814 @item -memregs=@var{number}
18816 Specifies the number of memory-based pseudo-registers GCC uses
18817 during code generation. These pseudo-registers are used like real
18818 registers, so there is a tradeoff between GCC's ability to fit the
18819 code into available registers, and the performance penalty of using
18820 memory instead of registers. Note that all modules in a program must
18821 be compiled with the same value for this option. Because of that, you
18822 must not use this option with GCC's default runtime libraries.
18826 @node M32R/D Options
18827 @subsection M32R/D Options
18828 @cindex M32R/D options
18830 These @option{-m} options are defined for Renesas M32R/D architectures:
18835 Generate code for the M32R/2@.
18839 Generate code for the M32R/X@.
18843 Generate code for the M32R@. This is the default.
18845 @item -mmodel=small
18846 @opindex mmodel=small
18847 Assume all objects live in the lower 16MB of memory (so that their addresses
18848 can be loaded with the @code{ld24} instruction), and assume all subroutines
18849 are reachable with the @code{bl} instruction.
18850 This is the default.
18852 The addressability of a particular object can be set with the
18853 @code{model} attribute.
18855 @item -mmodel=medium
18856 @opindex mmodel=medium
18857 Assume objects may be anywhere in the 32-bit address space (the compiler
18858 generates @code{seth/add3} instructions to load their addresses), and
18859 assume all subroutines are reachable with the @code{bl} instruction.
18861 @item -mmodel=large
18862 @opindex mmodel=large
18863 Assume objects may be anywhere in the 32-bit address space (the compiler
18864 generates @code{seth/add3} instructions to load their addresses), and
18865 assume subroutines may not be reachable with the @code{bl} instruction
18866 (the compiler generates the much slower @code{seth/add3/jl}
18867 instruction sequence).
18870 @opindex msdata=none
18871 Disable use of the small data area. Variables are put into
18872 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
18873 @code{section} attribute has been specified).
18874 This is the default.
18876 The small data area consists of sections @code{.sdata} and @code{.sbss}.
18877 Objects may be explicitly put in the small data area with the
18878 @code{section} attribute using one of these sections.
18880 @item -msdata=sdata
18881 @opindex msdata=sdata
18882 Put small global and static data in the small data area, but do not
18883 generate special code to reference them.
18886 @opindex msdata=use
18887 Put small global and static data in the small data area, and generate
18888 special instructions to reference them.
18892 @cindex smaller data references
18893 Put global and static objects less than or equal to @var{num} bytes
18894 into the small data or BSS sections instead of the normal data or BSS
18895 sections. The default value of @var{num} is 8.
18896 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
18897 for this option to have any effect.
18899 All modules should be compiled with the same @option{-G @var{num}} value.
18900 Compiling with different values of @var{num} may or may not work; if it
18901 doesn't the linker gives an error message---incorrect code is not
18906 Makes the M32R-specific code in the compiler display some statistics
18907 that might help in debugging programs.
18909 @item -malign-loops
18910 @opindex malign-loops
18911 Align all loops to a 32-byte boundary.
18913 @item -mno-align-loops
18914 @opindex mno-align-loops
18915 Do not enforce a 32-byte alignment for loops. This is the default.
18917 @item -missue-rate=@var{number}
18918 @opindex missue-rate=@var{number}
18919 Issue @var{number} instructions per cycle. @var{number} can only be 1
18922 @item -mbranch-cost=@var{number}
18923 @opindex mbranch-cost=@var{number}
18924 @var{number} can only be 1 or 2. If it is 1 then branches are
18925 preferred over conditional code, if it is 2, then the opposite applies.
18927 @item -mflush-trap=@var{number}
18928 @opindex mflush-trap=@var{number}
18929 Specifies the trap number to use to flush the cache. The default is
18930 12. Valid numbers are between 0 and 15 inclusive.
18932 @item -mno-flush-trap
18933 @opindex mno-flush-trap
18934 Specifies that the cache cannot be flushed by using a trap.
18936 @item -mflush-func=@var{name}
18937 @opindex mflush-func=@var{name}
18938 Specifies the name of the operating system function to call to flush
18939 the cache. The default is @samp{_flush_cache}, but a function call
18940 is only used if a trap is not available.
18942 @item -mno-flush-func
18943 @opindex mno-flush-func
18944 Indicates that there is no OS function for flushing the cache.
18948 @node M680x0 Options
18949 @subsection M680x0 Options
18950 @cindex M680x0 options
18952 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
18953 The default settings depend on which architecture was selected when
18954 the compiler was configured; the defaults for the most common choices
18958 @item -march=@var{arch}
18960 Generate code for a specific M680x0 or ColdFire instruction set
18961 architecture. Permissible values of @var{arch} for M680x0
18962 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
18963 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
18964 architectures are selected according to Freescale's ISA classification
18965 and the permissible values are: @samp{isaa}, @samp{isaaplus},
18966 @samp{isab} and @samp{isac}.
18968 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
18969 code for a ColdFire target. The @var{arch} in this macro is one of the
18970 @option{-march} arguments given above.
18972 When used together, @option{-march} and @option{-mtune} select code
18973 that runs on a family of similar processors but that is optimized
18974 for a particular microarchitecture.
18976 @item -mcpu=@var{cpu}
18978 Generate code for a specific M680x0 or ColdFire processor.
18979 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
18980 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
18981 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
18982 below, which also classifies the CPUs into families:
18984 @multitable @columnfractions 0.20 0.80
18985 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
18986 @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}
18987 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
18988 @item @samp{5206e} @tab @samp{5206e}
18989 @item @samp{5208} @tab @samp{5207} @samp{5208}
18990 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
18991 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
18992 @item @samp{5216} @tab @samp{5214} @samp{5216}
18993 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
18994 @item @samp{5225} @tab @samp{5224} @samp{5225}
18995 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
18996 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
18997 @item @samp{5249} @tab @samp{5249}
18998 @item @samp{5250} @tab @samp{5250}
18999 @item @samp{5271} @tab @samp{5270} @samp{5271}
19000 @item @samp{5272} @tab @samp{5272}
19001 @item @samp{5275} @tab @samp{5274} @samp{5275}
19002 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
19003 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
19004 @item @samp{5307} @tab @samp{5307}
19005 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
19006 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
19007 @item @samp{5407} @tab @samp{5407}
19008 @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}
19011 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
19012 @var{arch} is compatible with @var{cpu}. Other combinations of
19013 @option{-mcpu} and @option{-march} are rejected.
19015 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
19016 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
19017 where the value of @var{family} is given by the table above.
19019 @item -mtune=@var{tune}
19021 Tune the code for a particular microarchitecture within the
19022 constraints set by @option{-march} and @option{-mcpu}.
19023 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
19024 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
19025 and @samp{cpu32}. The ColdFire microarchitectures
19026 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
19028 You can also use @option{-mtune=68020-40} for code that needs
19029 to run relatively well on 68020, 68030 and 68040 targets.
19030 @option{-mtune=68020-60} is similar but includes 68060 targets
19031 as well. These two options select the same tuning decisions as
19032 @option{-m68020-40} and @option{-m68020-60} respectively.
19034 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
19035 when tuning for 680x0 architecture @var{arch}. It also defines
19036 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
19037 option is used. If GCC is tuning for a range of architectures,
19038 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
19039 it defines the macros for every architecture in the range.
19041 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
19042 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
19043 of the arguments given above.
19049 Generate output for a 68000. This is the default
19050 when the compiler is configured for 68000-based systems.
19051 It is equivalent to @option{-march=68000}.
19053 Use this option for microcontrollers with a 68000 or EC000 core,
19054 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
19058 Generate output for a 68010. This is the default
19059 when the compiler is configured for 68010-based systems.
19060 It is equivalent to @option{-march=68010}.
19066 Generate output for a 68020. This is the default
19067 when the compiler is configured for 68020-based systems.
19068 It is equivalent to @option{-march=68020}.
19072 Generate output for a 68030. This is the default when the compiler is
19073 configured for 68030-based systems. It is equivalent to
19074 @option{-march=68030}.
19078 Generate output for a 68040. This is the default when the compiler is
19079 configured for 68040-based systems. It is equivalent to
19080 @option{-march=68040}.
19082 This option inhibits the use of 68881/68882 instructions that have to be
19083 emulated by software on the 68040. Use this option if your 68040 does not
19084 have code to emulate those instructions.
19088 Generate output for a 68060. This is the default when the compiler is
19089 configured for 68060-based systems. It is equivalent to
19090 @option{-march=68060}.
19092 This option inhibits the use of 68020 and 68881/68882 instructions that
19093 have to be emulated by software on the 68060. Use this option if your 68060
19094 does not have code to emulate those instructions.
19098 Generate output for a CPU32. This is the default
19099 when the compiler is configured for CPU32-based systems.
19100 It is equivalent to @option{-march=cpu32}.
19102 Use this option for microcontrollers with a
19103 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
19104 68336, 68340, 68341, 68349 and 68360.
19108 Generate output for a 520X ColdFire CPU@. This is the default
19109 when the compiler is configured for 520X-based systems.
19110 It is equivalent to @option{-mcpu=5206}, and is now deprecated
19111 in favor of that option.
19113 Use this option for microcontroller with a 5200 core, including
19114 the MCF5202, MCF5203, MCF5204 and MCF5206.
19118 Generate output for a 5206e ColdFire CPU@. The option is now
19119 deprecated in favor of the equivalent @option{-mcpu=5206e}.
19123 Generate output for a member of the ColdFire 528X family.
19124 The option is now deprecated in favor of the equivalent
19125 @option{-mcpu=528x}.
19129 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
19130 in favor of the equivalent @option{-mcpu=5307}.
19134 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
19135 in favor of the equivalent @option{-mcpu=5407}.
19139 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
19140 This includes use of hardware floating-point instructions.
19141 The option is equivalent to @option{-mcpu=547x}, and is now
19142 deprecated in favor of that option.
19146 Generate output for a 68040, without using any of the new instructions.
19147 This results in code that can run relatively efficiently on either a
19148 68020/68881 or a 68030 or a 68040. The generated code does use the
19149 68881 instructions that are emulated on the 68040.
19151 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
19155 Generate output for a 68060, without using any of the new instructions.
19156 This results in code that can run relatively efficiently on either a
19157 68020/68881 or a 68030 or a 68040. The generated code does use the
19158 68881 instructions that are emulated on the 68060.
19160 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
19164 @opindex mhard-float
19166 Generate floating-point instructions. This is the default for 68020
19167 and above, and for ColdFire devices that have an FPU@. It defines the
19168 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
19169 on ColdFire targets.
19172 @opindex msoft-float
19173 Do not generate floating-point instructions; use library calls instead.
19174 This is the default for 68000, 68010, and 68832 targets. It is also
19175 the default for ColdFire devices that have no FPU.
19181 Generate (do not generate) ColdFire hardware divide and remainder
19182 instructions. If @option{-march} is used without @option{-mcpu},
19183 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
19184 architectures. Otherwise, the default is taken from the target CPU
19185 (either the default CPU, or the one specified by @option{-mcpu}). For
19186 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
19187 @option{-mcpu=5206e}.
19189 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
19193 Consider type @code{int} to be 16 bits wide, like @code{short int}.
19194 Additionally, parameters passed on the stack are also aligned to a
19195 16-bit boundary even on targets whose API mandates promotion to 32-bit.
19199 Do not consider type @code{int} to be 16 bits wide. This is the default.
19202 @itemx -mno-bitfield
19203 @opindex mnobitfield
19204 @opindex mno-bitfield
19205 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
19206 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
19210 Do use the bit-field instructions. The @option{-m68020} option implies
19211 @option{-mbitfield}. This is the default if you use a configuration
19212 designed for a 68020.
19216 Use a different function-calling convention, in which functions
19217 that take a fixed number of arguments return with the @code{rtd}
19218 instruction, which pops their arguments while returning. This
19219 saves one instruction in the caller since there is no need to pop
19220 the arguments there.
19222 This calling convention is incompatible with the one normally
19223 used on Unix, so you cannot use it if you need to call libraries
19224 compiled with the Unix compiler.
19226 Also, you must provide function prototypes for all functions that
19227 take variable numbers of arguments (including @code{printf});
19228 otherwise incorrect code is generated for calls to those
19231 In addition, seriously incorrect code results if you call a
19232 function with too many arguments. (Normally, extra arguments are
19233 harmlessly ignored.)
19235 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
19236 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
19240 Do not use the calling conventions selected by @option{-mrtd}.
19241 This is the default.
19244 @itemx -mno-align-int
19245 @opindex malign-int
19246 @opindex mno-align-int
19247 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
19248 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
19249 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
19250 Aligning variables on 32-bit boundaries produces code that runs somewhat
19251 faster on processors with 32-bit busses at the expense of more memory.
19253 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
19254 aligns structures containing the above types differently than
19255 most published application binary interface specifications for the m68k.
19259 Use the pc-relative addressing mode of the 68000 directly, instead of
19260 using a global offset table. At present, this option implies @option{-fpic},
19261 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
19262 not presently supported with @option{-mpcrel}, though this could be supported for
19263 68020 and higher processors.
19265 @item -mno-strict-align
19266 @itemx -mstrict-align
19267 @opindex mno-strict-align
19268 @opindex mstrict-align
19269 Do not (do) assume that unaligned memory references are handled by
19273 Generate code that allows the data segment to be located in a different
19274 area of memory from the text segment. This allows for execute-in-place in
19275 an environment without virtual memory management. This option implies
19278 @item -mno-sep-data
19279 Generate code that assumes that the data segment follows the text segment.
19280 This is the default.
19282 @item -mid-shared-library
19283 Generate code that supports shared libraries via the library ID method.
19284 This allows for execute-in-place and shared libraries in an environment
19285 without virtual memory management. This option implies @option{-fPIC}.
19287 @item -mno-id-shared-library
19288 Generate code that doesn't assume ID-based shared libraries are being used.
19289 This is the default.
19291 @item -mshared-library-id=n
19292 Specifies the identification number of the ID-based shared library being
19293 compiled. Specifying a value of 0 generates more compact code; specifying
19294 other values forces the allocation of that number to the current
19295 library, but is no more space- or time-efficient than omitting this option.
19301 When generating position-independent code for ColdFire, generate code
19302 that works if the GOT has more than 8192 entries. This code is
19303 larger and slower than code generated without this option. On M680x0
19304 processors, this option is not needed; @option{-fPIC} suffices.
19306 GCC normally uses a single instruction to load values from the GOT@.
19307 While this is relatively efficient, it only works if the GOT
19308 is smaller than about 64k. Anything larger causes the linker
19309 to report an error such as:
19311 @cindex relocation truncated to fit (ColdFire)
19313 relocation truncated to fit: R_68K_GOT16O foobar
19316 If this happens, you should recompile your code with @option{-mxgot}.
19317 It should then work with very large GOTs. However, code generated with
19318 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
19319 the value of a global symbol.
19321 Note that some linkers, including newer versions of the GNU linker,
19322 can create multiple GOTs and sort GOT entries. If you have such a linker,
19323 you should only need to use @option{-mxgot} when compiling a single
19324 object file that accesses more than 8192 GOT entries. Very few do.
19326 These options have no effect unless GCC is generating
19327 position-independent code.
19329 @item -mlong-jump-table-offsets
19330 @opindex mlong-jump-table-offsets
19331 Use 32-bit offsets in @code{switch} tables. The default is to use
19336 @node MCore Options
19337 @subsection MCore Options
19338 @cindex MCore options
19340 These are the @samp{-m} options defined for the Motorola M*Core
19346 @itemx -mno-hardlit
19348 @opindex mno-hardlit
19349 Inline constants into the code stream if it can be done in two
19350 instructions or less.
19356 Use the divide instruction. (Enabled by default).
19358 @item -mrelax-immediate
19359 @itemx -mno-relax-immediate
19360 @opindex mrelax-immediate
19361 @opindex mno-relax-immediate
19362 Allow arbitrary-sized immediates in bit operations.
19364 @item -mwide-bitfields
19365 @itemx -mno-wide-bitfields
19366 @opindex mwide-bitfields
19367 @opindex mno-wide-bitfields
19368 Always treat bit-fields as @code{int}-sized.
19370 @item -m4byte-functions
19371 @itemx -mno-4byte-functions
19372 @opindex m4byte-functions
19373 @opindex mno-4byte-functions
19374 Force all functions to be aligned to a 4-byte boundary.
19376 @item -mcallgraph-data
19377 @itemx -mno-callgraph-data
19378 @opindex mcallgraph-data
19379 @opindex mno-callgraph-data
19380 Emit callgraph information.
19383 @itemx -mno-slow-bytes
19384 @opindex mslow-bytes
19385 @opindex mno-slow-bytes
19386 Prefer word access when reading byte quantities.
19388 @item -mlittle-endian
19389 @itemx -mbig-endian
19390 @opindex mlittle-endian
19391 @opindex mbig-endian
19392 Generate code for a little-endian target.
19398 Generate code for the 210 processor.
19402 Assume that runtime support has been provided and so omit the
19403 simulator library (@file{libsim.a)} from the linker command line.
19405 @item -mstack-increment=@var{size}
19406 @opindex mstack-increment
19407 Set the maximum amount for a single stack increment operation. Large
19408 values can increase the speed of programs that contain functions
19409 that need a large amount of stack space, but they can also trigger a
19410 segmentation fault if the stack is extended too much. The default
19416 @subsection MeP Options
19417 @cindex MeP options
19423 Enables the @code{abs} instruction, which is the absolute difference
19424 between two registers.
19428 Enables all the optional instructions---average, multiply, divide, bit
19429 operations, leading zero, absolute difference, min/max, clip, and
19435 Enables the @code{ave} instruction, which computes the average of two
19438 @item -mbased=@var{n}
19440 Variables of size @var{n} bytes or smaller are placed in the
19441 @code{.based} section by default. Based variables use the @code{$tp}
19442 register as a base register, and there is a 128-byte limit to the
19443 @code{.based} section.
19447 Enables the bit operation instructions---bit test (@code{btstm}), set
19448 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
19449 test-and-set (@code{tas}).
19451 @item -mc=@var{name}
19453 Selects which section constant data is placed in. @var{name} may
19454 be @samp{tiny}, @samp{near}, or @samp{far}.
19458 Enables the @code{clip} instruction. Note that @option{-mclip} is not
19459 useful unless you also provide @option{-mminmax}.
19461 @item -mconfig=@var{name}
19463 Selects one of the built-in core configurations. Each MeP chip has
19464 one or more modules in it; each module has a core CPU and a variety of
19465 coprocessors, optional instructions, and peripherals. The
19466 @code{MeP-Integrator} tool, not part of GCC, provides these
19467 configurations through this option; using this option is the same as
19468 using all the corresponding command-line options. The default
19469 configuration is @samp{default}.
19473 Enables the coprocessor instructions. By default, this is a 32-bit
19474 coprocessor. Note that the coprocessor is normally enabled via the
19475 @option{-mconfig=} option.
19479 Enables the 32-bit coprocessor's instructions.
19483 Enables the 64-bit coprocessor's instructions.
19487 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
19491 Causes constant variables to be placed in the @code{.near} section.
19495 Enables the @code{div} and @code{divu} instructions.
19499 Generate big-endian code.
19503 Generate little-endian code.
19505 @item -mio-volatile
19506 @opindex mio-volatile
19507 Tells the compiler that any variable marked with the @code{io}
19508 attribute is to be considered volatile.
19512 Causes variables to be assigned to the @code{.far} section by default.
19516 Enables the @code{leadz} (leading zero) instruction.
19520 Causes variables to be assigned to the @code{.near} section by default.
19524 Enables the @code{min} and @code{max} instructions.
19528 Enables the multiplication and multiply-accumulate instructions.
19532 Disables all the optional instructions enabled by @option{-mall-opts}.
19536 Enables the @code{repeat} and @code{erepeat} instructions, used for
19537 low-overhead looping.
19541 Causes all variables to default to the @code{.tiny} section. Note
19542 that there is a 65536-byte limit to this section. Accesses to these
19543 variables use the @code{%gp} base register.
19547 Enables the saturation instructions. Note that the compiler does not
19548 currently generate these itself, but this option is included for
19549 compatibility with other tools, like @code{as}.
19553 Link the SDRAM-based runtime instead of the default ROM-based runtime.
19557 Link the simulator run-time libraries.
19561 Link the simulator runtime libraries, excluding built-in support
19562 for reset and exception vectors and tables.
19566 Causes all functions to default to the @code{.far} section. Without
19567 this option, functions default to the @code{.near} section.
19569 @item -mtiny=@var{n}
19571 Variables that are @var{n} bytes or smaller are allocated to the
19572 @code{.tiny} section. These variables use the @code{$gp} base
19573 register. The default for this option is 4, but note that there's a
19574 65536-byte limit to the @code{.tiny} section.
19578 @node MicroBlaze Options
19579 @subsection MicroBlaze Options
19580 @cindex MicroBlaze Options
19585 @opindex msoft-float
19586 Use software emulation for floating point (default).
19589 @opindex mhard-float
19590 Use hardware floating-point instructions.
19594 Do not optimize block moves, use @code{memcpy}.
19596 @item -mno-clearbss
19597 @opindex mno-clearbss
19598 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
19600 @item -mcpu=@var{cpu-type}
19602 Use features of, and schedule code for, the given CPU.
19603 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
19604 where @var{X} is a major version, @var{YY} is the minor version, and
19605 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
19606 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
19608 @item -mxl-soft-mul
19609 @opindex mxl-soft-mul
19610 Use software multiply emulation (default).
19612 @item -mxl-soft-div
19613 @opindex mxl-soft-div
19614 Use software emulation for divides (default).
19616 @item -mxl-barrel-shift
19617 @opindex mxl-barrel-shift
19618 Use the hardware barrel shifter.
19620 @item -mxl-pattern-compare
19621 @opindex mxl-pattern-compare
19622 Use pattern compare instructions.
19624 @item -msmall-divides
19625 @opindex msmall-divides
19626 Use table lookup optimization for small signed integer divisions.
19628 @item -mxl-stack-check
19629 @opindex mxl-stack-check
19630 This option is deprecated. Use @option{-fstack-check} instead.
19633 @opindex mxl-gp-opt
19634 Use GP-relative @code{.sdata}/@code{.sbss} sections.
19636 @item -mxl-multiply-high
19637 @opindex mxl-multiply-high
19638 Use multiply high instructions for high part of 32x32 multiply.
19640 @item -mxl-float-convert
19641 @opindex mxl-float-convert
19642 Use hardware floating-point conversion instructions.
19644 @item -mxl-float-sqrt
19645 @opindex mxl-float-sqrt
19646 Use hardware floating-point square root instruction.
19649 @opindex mbig-endian
19650 Generate code for a big-endian target.
19652 @item -mlittle-endian
19653 @opindex mlittle-endian
19654 Generate code for a little-endian target.
19657 @opindex mxl-reorder
19658 Use reorder instructions (swap and byte reversed load/store).
19660 @item -mxl-mode-@var{app-model}
19661 Select application model @var{app-model}. Valid models are
19664 normal executable (default), uses startup code @file{crt0.o}.
19667 for use with Xilinx Microprocessor Debugger (XMD) based
19668 software intrusive debug agent called xmdstub. This uses startup file
19669 @file{crt1.o} and sets the start address of the program to 0x800.
19672 for applications that are loaded using a bootloader.
19673 This model uses startup file @file{crt2.o} which does not contain a processor
19674 reset vector handler. This is suitable for transferring control on a
19675 processor reset to the bootloader rather than the application.
19678 for applications that do not require any of the
19679 MicroBlaze vectors. This option may be useful for applications running
19680 within a monitoring application. This model uses @file{crt3.o} as a startup file.
19683 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
19684 @option{-mxl-mode-@var{app-model}}.
19689 @subsection MIPS Options
19690 @cindex MIPS options
19696 Generate big-endian code.
19700 Generate little-endian code. This is the default for @samp{mips*el-*-*}
19703 @item -march=@var{arch}
19705 Generate code that runs on @var{arch}, which can be the name of a
19706 generic MIPS ISA, or the name of a particular processor.
19708 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
19709 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
19710 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
19711 @samp{mips64r5} and @samp{mips64r6}.
19712 The processor names are:
19713 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
19714 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
19715 @samp{5kc}, @samp{5kf},
19717 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
19718 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
19719 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
19720 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
19721 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
19724 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
19726 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
19727 @samp{m5100}, @samp{m5101},
19728 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
19731 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
19732 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
19733 @samp{rm7000}, @samp{rm9000},
19734 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
19737 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
19738 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
19739 @samp{xlr} and @samp{xlp}.
19740 The special value @samp{from-abi} selects the
19741 most compatible architecture for the selected ABI (that is,
19742 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
19744 The native Linux/GNU toolchain also supports the value @samp{native},
19745 which selects the best architecture option for the host processor.
19746 @option{-march=native} has no effect if GCC does not recognize
19749 In processor names, a final @samp{000} can be abbreviated as @samp{k}
19750 (for example, @option{-march=r2k}). Prefixes are optional, and
19751 @samp{vr} may be written @samp{r}.
19753 Names of the form @samp{@var{n}f2_1} refer to processors with
19754 FPUs clocked at half the rate of the core, names of the form
19755 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
19756 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
19757 processors with FPUs clocked a ratio of 3:2 with respect to the core.
19758 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
19759 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
19760 accepted as synonyms for @samp{@var{n}f1_1}.
19762 GCC defines two macros based on the value of this option. The first
19763 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
19764 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
19765 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
19766 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
19767 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
19769 Note that the @code{_MIPS_ARCH} macro uses the processor names given
19770 above. In other words, it has the full prefix and does not
19771 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
19772 the macro names the resolved architecture (either @code{"mips1"} or
19773 @code{"mips3"}). It names the default architecture when no
19774 @option{-march} option is given.
19776 @item -mtune=@var{arch}
19778 Optimize for @var{arch}. Among other things, this option controls
19779 the way instructions are scheduled, and the perceived cost of arithmetic
19780 operations. The list of @var{arch} values is the same as for
19783 When this option is not used, GCC optimizes for the processor
19784 specified by @option{-march}. By using @option{-march} and
19785 @option{-mtune} together, it is possible to generate code that
19786 runs on a family of processors, but optimize the code for one
19787 particular member of that family.
19789 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
19790 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
19791 @option{-march} ones described above.
19795 Equivalent to @option{-march=mips1}.
19799 Equivalent to @option{-march=mips2}.
19803 Equivalent to @option{-march=mips3}.
19807 Equivalent to @option{-march=mips4}.
19811 Equivalent to @option{-march=mips32}.
19815 Equivalent to @option{-march=mips32r3}.
19819 Equivalent to @option{-march=mips32r5}.
19823 Equivalent to @option{-march=mips32r6}.
19827 Equivalent to @option{-march=mips64}.
19831 Equivalent to @option{-march=mips64r2}.
19835 Equivalent to @option{-march=mips64r3}.
19839 Equivalent to @option{-march=mips64r5}.
19843 Equivalent to @option{-march=mips64r6}.
19848 @opindex mno-mips16
19849 Generate (do not generate) MIPS16 code. If GCC is targeting a
19850 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
19852 MIPS16 code generation can also be controlled on a per-function basis
19853 by means of @code{mips16} and @code{nomips16} attributes.
19854 @xref{Function Attributes}, for more information.
19856 @item -mflip-mips16
19857 @opindex mflip-mips16
19858 Generate MIPS16 code on alternating functions. This option is provided
19859 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
19860 not intended for ordinary use in compiling user code.
19862 @item -minterlink-compressed
19863 @item -mno-interlink-compressed
19864 @opindex minterlink-compressed
19865 @opindex mno-interlink-compressed
19866 Require (do not require) that code using the standard (uncompressed) MIPS ISA
19867 be link-compatible with MIPS16 and microMIPS code, and vice versa.
19869 For example, code using the standard ISA encoding cannot jump directly
19870 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
19871 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
19872 knows that the target of the jump is not compressed.
19874 @item -minterlink-mips16
19875 @itemx -mno-interlink-mips16
19876 @opindex minterlink-mips16
19877 @opindex mno-interlink-mips16
19878 Aliases of @option{-minterlink-compressed} and
19879 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
19880 and are retained for backwards compatibility.
19892 Generate code for the given ABI@.
19894 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
19895 generates 64-bit code when you select a 64-bit architecture, but you
19896 can use @option{-mgp32} to get 32-bit code instead.
19898 For information about the O64 ABI, see
19899 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
19901 GCC supports a variant of the o32 ABI in which floating-point registers
19902 are 64 rather than 32 bits wide. You can select this combination with
19903 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
19904 and @code{mfhc1} instructions and is therefore only supported for
19905 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
19907 The register assignments for arguments and return values remain the
19908 same, but each scalar value is passed in a single 64-bit register
19909 rather than a pair of 32-bit registers. For example, scalar
19910 floating-point values are returned in @samp{$f0} only, not a
19911 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
19912 remains the same in that the even-numbered double-precision registers
19915 Two additional variants of the o32 ABI are supported to enable
19916 a transition from 32-bit to 64-bit registers. These are FPXX
19917 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
19918 The FPXX extension mandates that all code must execute correctly
19919 when run using 32-bit or 64-bit registers. The code can be interlinked
19920 with either FP32 or FP64, but not both.
19921 The FP64A extension is similar to the FP64 extension but forbids the
19922 use of odd-numbered single-precision registers. This can be used
19923 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
19924 processors and allows both FP32 and FP64A code to interlink and
19925 run in the same process without changing FPU modes.
19928 @itemx -mno-abicalls
19930 @opindex mno-abicalls
19931 Generate (do not generate) code that is suitable for SVR4-style
19932 dynamic objects. @option{-mabicalls} is the default for SVR4-based
19937 Generate (do not generate) code that is fully position-independent,
19938 and that can therefore be linked into shared libraries. This option
19939 only affects @option{-mabicalls}.
19941 All @option{-mabicalls} code has traditionally been position-independent,
19942 regardless of options like @option{-fPIC} and @option{-fpic}. However,
19943 as an extension, the GNU toolchain allows executables to use absolute
19944 accesses for locally-binding symbols. It can also use shorter GP
19945 initialization sequences and generate direct calls to locally-defined
19946 functions. This mode is selected by @option{-mno-shared}.
19948 @option{-mno-shared} depends on binutils 2.16 or higher and generates
19949 objects that can only be linked by the GNU linker. However, the option
19950 does not affect the ABI of the final executable; it only affects the ABI
19951 of relocatable objects. Using @option{-mno-shared} generally makes
19952 executables both smaller and quicker.
19954 @option{-mshared} is the default.
19960 Assume (do not assume) that the static and dynamic linkers
19961 support PLTs and copy relocations. This option only affects
19962 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
19963 has no effect without @option{-msym32}.
19965 You can make @option{-mplt} the default by configuring
19966 GCC with @option{--with-mips-plt}. The default is
19967 @option{-mno-plt} otherwise.
19973 Lift (do not lift) the usual restrictions on the size of the global
19976 GCC normally uses a single instruction to load values from the GOT@.
19977 While this is relatively efficient, it only works if the GOT
19978 is smaller than about 64k. Anything larger causes the linker
19979 to report an error such as:
19981 @cindex relocation truncated to fit (MIPS)
19983 relocation truncated to fit: R_MIPS_GOT16 foobar
19986 If this happens, you should recompile your code with @option{-mxgot}.
19987 This works with very large GOTs, although the code is also
19988 less efficient, since it takes three instructions to fetch the
19989 value of a global symbol.
19991 Note that some linkers can create multiple GOTs. If you have such a
19992 linker, you should only need to use @option{-mxgot} when a single object
19993 file accesses more than 64k's worth of GOT entries. Very few do.
19995 These options have no effect unless GCC is generating position
20000 Assume that general-purpose registers are 32 bits wide.
20004 Assume that general-purpose registers are 64 bits wide.
20008 Assume that floating-point registers are 32 bits wide.
20012 Assume that floating-point registers are 64 bits wide.
20016 Do not assume the width of floating-point registers.
20019 @opindex mhard-float
20020 Use floating-point coprocessor instructions.
20023 @opindex msoft-float
20024 Do not use floating-point coprocessor instructions. Implement
20025 floating-point calculations using library calls instead.
20029 Equivalent to @option{-msoft-float}, but additionally asserts that the
20030 program being compiled does not perform any floating-point operations.
20031 This option is presently supported only by some bare-metal MIPS
20032 configurations, where it may select a special set of libraries
20033 that lack all floating-point support (including, for example, the
20034 floating-point @code{printf} formats).
20035 If code compiled with @option{-mno-float} accidentally contains
20036 floating-point operations, it is likely to suffer a link-time
20037 or run-time failure.
20039 @item -msingle-float
20040 @opindex msingle-float
20041 Assume that the floating-point coprocessor only supports single-precision
20044 @item -mdouble-float
20045 @opindex mdouble-float
20046 Assume that the floating-point coprocessor supports double-precision
20047 operations. This is the default.
20050 @itemx -mno-odd-spreg
20051 @opindex modd-spreg
20052 @opindex mno-odd-spreg
20053 Enable the use of odd-numbered single-precision floating-point registers
20054 for the o32 ABI. This is the default for processors that are known to
20055 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
20059 @itemx -mabs=legacy
20061 @opindex mabs=legacy
20062 These options control the treatment of the special not-a-number (NaN)
20063 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
20064 @code{neg.@i{fmt}} machine instructions.
20066 By default or when @option{-mabs=legacy} is used the legacy
20067 treatment is selected. In this case these instructions are considered
20068 arithmetic and avoided where correct operation is required and the
20069 input operand might be a NaN. A longer sequence of instructions that
20070 manipulate the sign bit of floating-point datum manually is used
20071 instead unless the @option{-ffinite-math-only} option has also been
20074 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
20075 this case these instructions are considered non-arithmetic and therefore
20076 operating correctly in all cases, including in particular where the
20077 input operand is a NaN. These instructions are therefore always used
20078 for the respective operations.
20081 @itemx -mnan=legacy
20083 @opindex mnan=legacy
20084 These options control the encoding of the special not-a-number (NaN)
20085 IEEE 754 floating-point data.
20087 The @option{-mnan=legacy} option selects the legacy encoding. In this
20088 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
20089 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
20090 by the first bit of their trailing significand field being 1.
20092 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
20093 this case qNaNs are denoted by the first bit of their trailing
20094 significand field being 1, whereas sNaNs are denoted by the first bit of
20095 their trailing significand field being 0.
20097 The default is @option{-mnan=legacy} unless GCC has been configured with
20098 @option{--with-nan=2008}.
20104 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
20105 implement atomic memory built-in functions. When neither option is
20106 specified, GCC uses the instructions if the target architecture
20109 @option{-mllsc} is useful if the runtime environment can emulate the
20110 instructions and @option{-mno-llsc} can be useful when compiling for
20111 nonstandard ISAs. You can make either option the default by
20112 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
20113 respectively. @option{--with-llsc} is the default for some
20114 configurations; see the installation documentation for details.
20120 Use (do not use) revision 1 of the MIPS DSP ASE@.
20121 @xref{MIPS DSP Built-in Functions}. This option defines the
20122 preprocessor macro @code{__mips_dsp}. It also defines
20123 @code{__mips_dsp_rev} to 1.
20129 Use (do not use) revision 2 of the MIPS DSP ASE@.
20130 @xref{MIPS DSP Built-in Functions}. This option defines the
20131 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
20132 It also defines @code{__mips_dsp_rev} to 2.
20135 @itemx -mno-smartmips
20136 @opindex msmartmips
20137 @opindex mno-smartmips
20138 Use (do not use) the MIPS SmartMIPS ASE.
20140 @item -mpaired-single
20141 @itemx -mno-paired-single
20142 @opindex mpaired-single
20143 @opindex mno-paired-single
20144 Use (do not use) paired-single floating-point instructions.
20145 @xref{MIPS Paired-Single Support}. This option requires
20146 hardware floating-point support to be enabled.
20152 Use (do not use) MIPS Digital Media Extension instructions.
20153 This option can only be used when generating 64-bit code and requires
20154 hardware floating-point support to be enabled.
20159 @opindex mno-mips3d
20160 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
20161 The option @option{-mips3d} implies @option{-mpaired-single}.
20164 @itemx -mno-micromips
20165 @opindex mmicromips
20166 @opindex mno-mmicromips
20167 Generate (do not generate) microMIPS code.
20169 MicroMIPS code generation can also be controlled on a per-function basis
20170 by means of @code{micromips} and @code{nomicromips} attributes.
20171 @xref{Function Attributes}, for more information.
20177 Use (do not use) MT Multithreading instructions.
20183 Use (do not use) the MIPS MCU ASE instructions.
20189 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
20195 Use (do not use) the MIPS Virtualization (VZ) instructions.
20201 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
20205 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
20206 an explanation of the default and the way that the pointer size is
20211 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
20213 The default size of @code{int}s, @code{long}s and pointers depends on
20214 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
20215 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
20216 32-bit @code{long}s. Pointers are the same size as @code{long}s,
20217 or the same size as integer registers, whichever is smaller.
20223 Assume (do not assume) that all symbols have 32-bit values, regardless
20224 of the selected ABI@. This option is useful in combination with
20225 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
20226 to generate shorter and faster references to symbolic addresses.
20230 Put definitions of externally-visible data in a small data section
20231 if that data is no bigger than @var{num} bytes. GCC can then generate
20232 more efficient accesses to the data; see @option{-mgpopt} for details.
20234 The default @option{-G} option depends on the configuration.
20236 @item -mlocal-sdata
20237 @itemx -mno-local-sdata
20238 @opindex mlocal-sdata
20239 @opindex mno-local-sdata
20240 Extend (do not extend) the @option{-G} behavior to local data too,
20241 such as to static variables in C@. @option{-mlocal-sdata} is the
20242 default for all configurations.
20244 If the linker complains that an application is using too much small data,
20245 you might want to try rebuilding the less performance-critical parts with
20246 @option{-mno-local-sdata}. You might also want to build large
20247 libraries with @option{-mno-local-sdata}, so that the libraries leave
20248 more room for the main program.
20250 @item -mextern-sdata
20251 @itemx -mno-extern-sdata
20252 @opindex mextern-sdata
20253 @opindex mno-extern-sdata
20254 Assume (do not assume) that externally-defined data is in
20255 a small data section if the size of that data is within the @option{-G} limit.
20256 @option{-mextern-sdata} is the default for all configurations.
20258 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
20259 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
20260 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
20261 is placed in a small data section. If @var{Var} is defined by another
20262 module, you must either compile that module with a high-enough
20263 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
20264 definition. If @var{Var} is common, you must link the application
20265 with a high-enough @option{-G} setting.
20267 The easiest way of satisfying these restrictions is to compile
20268 and link every module with the same @option{-G} option. However,
20269 you may wish to build a library that supports several different
20270 small data limits. You can do this by compiling the library with
20271 the highest supported @option{-G} setting and additionally using
20272 @option{-mno-extern-sdata} to stop the library from making assumptions
20273 about externally-defined data.
20279 Use (do not use) GP-relative accesses for symbols that are known to be
20280 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
20281 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
20284 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
20285 might not hold the value of @code{_gp}. For example, if the code is
20286 part of a library that might be used in a boot monitor, programs that
20287 call boot monitor routines pass an unknown value in @code{$gp}.
20288 (In such situations, the boot monitor itself is usually compiled
20289 with @option{-G0}.)
20291 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
20292 @option{-mno-extern-sdata}.
20294 @item -membedded-data
20295 @itemx -mno-embedded-data
20296 @opindex membedded-data
20297 @opindex mno-embedded-data
20298 Allocate variables to the read-only data section first if possible, then
20299 next in the small data section if possible, otherwise in data. This gives
20300 slightly slower code than the default, but reduces the amount of RAM required
20301 when executing, and thus may be preferred for some embedded systems.
20303 @item -muninit-const-in-rodata
20304 @itemx -mno-uninit-const-in-rodata
20305 @opindex muninit-const-in-rodata
20306 @opindex mno-uninit-const-in-rodata
20307 Put uninitialized @code{const} variables in the read-only data section.
20308 This option is only meaningful in conjunction with @option{-membedded-data}.
20310 @item -mcode-readable=@var{setting}
20311 @opindex mcode-readable
20312 Specify whether GCC may generate code that reads from executable sections.
20313 There are three possible settings:
20316 @item -mcode-readable=yes
20317 Instructions may freely access executable sections. This is the
20320 @item -mcode-readable=pcrel
20321 MIPS16 PC-relative load instructions can access executable sections,
20322 but other instructions must not do so. This option is useful on 4KSc
20323 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
20324 It is also useful on processors that can be configured to have a dual
20325 instruction/data SRAM interface and that, like the M4K, automatically
20326 redirect PC-relative loads to the instruction RAM.
20328 @item -mcode-readable=no
20329 Instructions must not access executable sections. This option can be
20330 useful on targets that are configured to have a dual instruction/data
20331 SRAM interface but that (unlike the M4K) do not automatically redirect
20332 PC-relative loads to the instruction RAM.
20335 @item -msplit-addresses
20336 @itemx -mno-split-addresses
20337 @opindex msplit-addresses
20338 @opindex mno-split-addresses
20339 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
20340 relocation operators. This option has been superseded by
20341 @option{-mexplicit-relocs} but is retained for backwards compatibility.
20343 @item -mexplicit-relocs
20344 @itemx -mno-explicit-relocs
20345 @opindex mexplicit-relocs
20346 @opindex mno-explicit-relocs
20347 Use (do not use) assembler relocation operators when dealing with symbolic
20348 addresses. The alternative, selected by @option{-mno-explicit-relocs},
20349 is to use assembler macros instead.
20351 @option{-mexplicit-relocs} is the default if GCC was configured
20352 to use an assembler that supports relocation operators.
20354 @item -mcheck-zero-division
20355 @itemx -mno-check-zero-division
20356 @opindex mcheck-zero-division
20357 @opindex mno-check-zero-division
20358 Trap (do not trap) on integer division by zero.
20360 The default is @option{-mcheck-zero-division}.
20362 @item -mdivide-traps
20363 @itemx -mdivide-breaks
20364 @opindex mdivide-traps
20365 @opindex mdivide-breaks
20366 MIPS systems check for division by zero by generating either a
20367 conditional trap or a break instruction. Using traps results in
20368 smaller code, but is only supported on MIPS II and later. Also, some
20369 versions of the Linux kernel have a bug that prevents trap from
20370 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
20371 allow conditional traps on architectures that support them and
20372 @option{-mdivide-breaks} to force the use of breaks.
20374 The default is usually @option{-mdivide-traps}, but this can be
20375 overridden at configure time using @option{--with-divide=breaks}.
20376 Divide-by-zero checks can be completely disabled using
20377 @option{-mno-check-zero-division}.
20379 @item -mload-store-pairs
20380 @itemx -mno-load-store-pairs
20381 @opindex mload-store-pairs
20382 @opindex mno-load-store-pairs
20383 Enable (disable) an optimization that pairs consecutive load or store
20384 instructions to enable load/store bonding. This option is enabled by
20385 default but only takes effect when the selected architecture is known
20386 to support bonding.
20391 @opindex mno-memcpy
20392 Force (do not force) the use of @code{memcpy} for non-trivial block
20393 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
20394 most constant-sized copies.
20397 @itemx -mno-long-calls
20398 @opindex mlong-calls
20399 @opindex mno-long-calls
20400 Disable (do not disable) use of the @code{jal} instruction. Calling
20401 functions using @code{jal} is more efficient but requires the caller
20402 and callee to be in the same 256 megabyte segment.
20404 This option has no effect on abicalls code. The default is
20405 @option{-mno-long-calls}.
20411 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
20412 instructions, as provided by the R4650 ISA@.
20418 Enable (disable) use of the @code{madd} and @code{msub} integer
20419 instructions. The default is @option{-mimadd} on architectures
20420 that support @code{madd} and @code{msub} except for the 74k
20421 architecture where it was found to generate slower code.
20424 @itemx -mno-fused-madd
20425 @opindex mfused-madd
20426 @opindex mno-fused-madd
20427 Enable (disable) use of the floating-point multiply-accumulate
20428 instructions, when they are available. The default is
20429 @option{-mfused-madd}.
20431 On the R8000 CPU when multiply-accumulate instructions are used,
20432 the intermediate product is calculated to infinite precision
20433 and is not subject to the FCSR Flush to Zero bit. This may be
20434 undesirable in some circumstances. On other processors the result
20435 is numerically identical to the equivalent computation using
20436 separate multiply, add, subtract and negate instructions.
20440 Tell the MIPS assembler to not run its preprocessor over user
20441 assembler files (with a @samp{.s} suffix) when assembling them.
20446 @opindex mno-fix-24k
20447 Work around the 24K E48 (lost data on stores during refill) errata.
20448 The workarounds are implemented by the assembler rather than by GCC@.
20451 @itemx -mno-fix-r4000
20452 @opindex mfix-r4000
20453 @opindex mno-fix-r4000
20454 Work around certain R4000 CPU errata:
20457 A double-word or a variable shift may give an incorrect result if executed
20458 immediately after starting an integer division.
20460 A double-word or a variable shift may give an incorrect result if executed
20461 while an integer multiplication is in progress.
20463 An integer division may give an incorrect result if started in a delay slot
20464 of a taken branch or a jump.
20468 @itemx -mno-fix-r4400
20469 @opindex mfix-r4400
20470 @opindex mno-fix-r4400
20471 Work around certain R4400 CPU errata:
20474 A double-word or a variable shift may give an incorrect result if executed
20475 immediately after starting an integer division.
20479 @itemx -mno-fix-r10000
20480 @opindex mfix-r10000
20481 @opindex mno-fix-r10000
20482 Work around certain R10000 errata:
20485 @code{ll}/@code{sc} sequences may not behave atomically on revisions
20486 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
20489 This option can only be used if the target architecture supports
20490 branch-likely instructions. @option{-mfix-r10000} is the default when
20491 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
20495 @itemx -mno-fix-rm7000
20496 @opindex mfix-rm7000
20497 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
20498 workarounds are implemented by the assembler rather than by GCC@.
20501 @itemx -mno-fix-vr4120
20502 @opindex mfix-vr4120
20503 Work around certain VR4120 errata:
20506 @code{dmultu} does not always produce the correct result.
20508 @code{div} and @code{ddiv} do not always produce the correct result if one
20509 of the operands is negative.
20511 The workarounds for the division errata rely on special functions in
20512 @file{libgcc.a}. At present, these functions are only provided by
20513 the @code{mips64vr*-elf} configurations.
20515 Other VR4120 errata require a NOP to be inserted between certain pairs of
20516 instructions. These errata are handled by the assembler, not by GCC itself.
20519 @opindex mfix-vr4130
20520 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
20521 workarounds are implemented by the assembler rather than by GCC,
20522 although GCC avoids using @code{mflo} and @code{mfhi} if the
20523 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
20524 instructions are available instead.
20527 @itemx -mno-fix-sb1
20529 Work around certain SB-1 CPU core errata.
20530 (This flag currently works around the SB-1 revision 2
20531 ``F1'' and ``F2'' floating-point errata.)
20533 @item -mr10k-cache-barrier=@var{setting}
20534 @opindex mr10k-cache-barrier
20535 Specify whether GCC should insert cache barriers to avoid the
20536 side-effects of speculation on R10K processors.
20538 In common with many processors, the R10K tries to predict the outcome
20539 of a conditional branch and speculatively executes instructions from
20540 the ``taken'' branch. It later aborts these instructions if the
20541 predicted outcome is wrong. However, on the R10K, even aborted
20542 instructions can have side effects.
20544 This problem only affects kernel stores and, depending on the system,
20545 kernel loads. As an example, a speculatively-executed store may load
20546 the target memory into cache and mark the cache line as dirty, even if
20547 the store itself is later aborted. If a DMA operation writes to the
20548 same area of memory before the ``dirty'' line is flushed, the cached
20549 data overwrites the DMA-ed data. See the R10K processor manual
20550 for a full description, including other potential problems.
20552 One workaround is to insert cache barrier instructions before every memory
20553 access that might be speculatively executed and that might have side
20554 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
20555 controls GCC's implementation of this workaround. It assumes that
20556 aborted accesses to any byte in the following regions does not have
20561 the memory occupied by the current function's stack frame;
20564 the memory occupied by an incoming stack argument;
20567 the memory occupied by an object with a link-time-constant address.
20570 It is the kernel's responsibility to ensure that speculative
20571 accesses to these regions are indeed safe.
20573 If the input program contains a function declaration such as:
20579 then the implementation of @code{foo} must allow @code{j foo} and
20580 @code{jal foo} to be executed speculatively. GCC honors this
20581 restriction for functions it compiles itself. It expects non-GCC
20582 functions (such as hand-written assembly code) to do the same.
20584 The option has three forms:
20587 @item -mr10k-cache-barrier=load-store
20588 Insert a cache barrier before a load or store that might be
20589 speculatively executed and that might have side effects even
20592 @item -mr10k-cache-barrier=store
20593 Insert a cache barrier before a store that might be speculatively
20594 executed and that might have side effects even if aborted.
20596 @item -mr10k-cache-barrier=none
20597 Disable the insertion of cache barriers. This is the default setting.
20600 @item -mflush-func=@var{func}
20601 @itemx -mno-flush-func
20602 @opindex mflush-func
20603 Specifies the function to call to flush the I and D caches, or to not
20604 call any such function. If called, the function must take the same
20605 arguments as the common @code{_flush_func}, that is, the address of the
20606 memory range for which the cache is being flushed, the size of the
20607 memory range, and the number 3 (to flush both caches). The default
20608 depends on the target GCC was configured for, but commonly is either
20609 @code{_flush_func} or @code{__cpu_flush}.
20611 @item mbranch-cost=@var{num}
20612 @opindex mbranch-cost
20613 Set the cost of branches to roughly @var{num} ``simple'' instructions.
20614 This cost is only a heuristic and is not guaranteed to produce
20615 consistent results across releases. A zero cost redundantly selects
20616 the default, which is based on the @option{-mtune} setting.
20618 @item -mbranch-likely
20619 @itemx -mno-branch-likely
20620 @opindex mbranch-likely
20621 @opindex mno-branch-likely
20622 Enable or disable use of Branch Likely instructions, regardless of the
20623 default for the selected architecture. By default, Branch Likely
20624 instructions may be generated if they are supported by the selected
20625 architecture. An exception is for the MIPS32 and MIPS64 architectures
20626 and processors that implement those architectures; for those, Branch
20627 Likely instructions are not be generated by default because the MIPS32
20628 and MIPS64 architectures specifically deprecate their use.
20630 @item -mcompact-branches=never
20631 @itemx -mcompact-branches=optimal
20632 @itemx -mcompact-branches=always
20633 @opindex mcompact-branches=never
20634 @opindex mcompact-branches=optimal
20635 @opindex mcompact-branches=always
20636 These options control which form of branches will be generated. The
20637 default is @option{-mcompact-branches=optimal}.
20639 The @option{-mcompact-branches=never} option ensures that compact branch
20640 instructions will never be generated.
20642 The @option{-mcompact-branches=always} option ensures that a compact
20643 branch instruction will be generated if available. If a compact branch
20644 instruction is not available, a delay slot form of the branch will be
20647 This option is supported from MIPS Release 6 onwards.
20649 The @option{-mcompact-branches=optimal} option will cause a delay slot
20650 branch to be used if one is available in the current ISA and the delay
20651 slot is successfully filled. If the delay slot is not filled, a compact
20652 branch will be chosen if one is available.
20654 @item -mfp-exceptions
20655 @itemx -mno-fp-exceptions
20656 @opindex mfp-exceptions
20657 Specifies whether FP exceptions are enabled. This affects how
20658 FP instructions are scheduled for some processors.
20659 The default is that FP exceptions are
20662 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
20663 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
20666 @item -mvr4130-align
20667 @itemx -mno-vr4130-align
20668 @opindex mvr4130-align
20669 The VR4130 pipeline is two-way superscalar, but can only issue two
20670 instructions together if the first one is 8-byte aligned. When this
20671 option is enabled, GCC aligns pairs of instructions that it
20672 thinks should execute in parallel.
20674 This option only has an effect when optimizing for the VR4130.
20675 It normally makes code faster, but at the expense of making it bigger.
20676 It is enabled by default at optimization level @option{-O3}.
20681 Enable (disable) generation of @code{synci} instructions on
20682 architectures that support it. The @code{synci} instructions (if
20683 enabled) are generated when @code{__builtin___clear_cache} is
20686 This option defaults to @option{-mno-synci}, but the default can be
20687 overridden by configuring GCC with @option{--with-synci}.
20689 When compiling code for single processor systems, it is generally safe
20690 to use @code{synci}. However, on many multi-core (SMP) systems, it
20691 does not invalidate the instruction caches on all cores and may lead
20692 to undefined behavior.
20694 @item -mrelax-pic-calls
20695 @itemx -mno-relax-pic-calls
20696 @opindex mrelax-pic-calls
20697 Try to turn PIC calls that are normally dispatched via register
20698 @code{$25} into direct calls. This is only possible if the linker can
20699 resolve the destination at link time and if the destination is within
20700 range for a direct call.
20702 @option{-mrelax-pic-calls} is the default if GCC was configured to use
20703 an assembler and a linker that support the @code{.reloc} assembly
20704 directive and @option{-mexplicit-relocs} is in effect. With
20705 @option{-mno-explicit-relocs}, this optimization can be performed by the
20706 assembler and the linker alone without help from the compiler.
20708 @item -mmcount-ra-address
20709 @itemx -mno-mcount-ra-address
20710 @opindex mmcount-ra-address
20711 @opindex mno-mcount-ra-address
20712 Emit (do not emit) code that allows @code{_mcount} to modify the
20713 calling function's return address. When enabled, this option extends
20714 the usual @code{_mcount} interface with a new @var{ra-address}
20715 parameter, which has type @code{intptr_t *} and is passed in register
20716 @code{$12}. @code{_mcount} can then modify the return address by
20717 doing both of the following:
20720 Returning the new address in register @code{$31}.
20722 Storing the new address in @code{*@var{ra-address}},
20723 if @var{ra-address} is nonnull.
20726 The default is @option{-mno-mcount-ra-address}.
20728 @item -mframe-header-opt
20729 @itemx -mno-frame-header-opt
20730 @opindex mframe-header-opt
20731 Enable (disable) frame header optimization in the o32 ABI. When using the
20732 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
20733 function to write out register arguments. When enabled, this optimization
20734 will suppress the allocation of the frame header if it can be determined that
20737 This optimization is off by default at all optimization levels.
20740 @itemx -mno-lxc1-sxc1
20741 @opindex mlxc1-sxc1
20742 When applicable, enable (disable) the generation of @code{lwxc1},
20743 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
20748 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
20749 @code{madd.d} and related instructions. Enabled by default.
20754 @subsection MMIX Options
20755 @cindex MMIX Options
20757 These options are defined for the MMIX:
20761 @itemx -mno-libfuncs
20763 @opindex mno-libfuncs
20764 Specify that intrinsic library functions are being compiled, passing all
20765 values in registers, no matter the size.
20768 @itemx -mno-epsilon
20770 @opindex mno-epsilon
20771 Generate floating-point comparison instructions that compare with respect
20772 to the @code{rE} epsilon register.
20774 @item -mabi=mmixware
20776 @opindex mabi=mmixware
20778 Generate code that passes function parameters and return values that (in
20779 the called function) are seen as registers @code{$0} and up, as opposed to
20780 the GNU ABI which uses global registers @code{$231} and up.
20782 @item -mzero-extend
20783 @itemx -mno-zero-extend
20784 @opindex mzero-extend
20785 @opindex mno-zero-extend
20786 When reading data from memory in sizes shorter than 64 bits, use (do not
20787 use) zero-extending load instructions by default, rather than
20788 sign-extending ones.
20791 @itemx -mno-knuthdiv
20793 @opindex mno-knuthdiv
20794 Make the result of a division yielding a remainder have the same sign as
20795 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
20796 remainder follows the sign of the dividend. Both methods are
20797 arithmetically valid, the latter being almost exclusively used.
20799 @item -mtoplevel-symbols
20800 @itemx -mno-toplevel-symbols
20801 @opindex mtoplevel-symbols
20802 @opindex mno-toplevel-symbols
20803 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
20804 code can be used with the @code{PREFIX} assembly directive.
20808 Generate an executable in the ELF format, rather than the default
20809 @samp{mmo} format used by the @command{mmix} simulator.
20811 @item -mbranch-predict
20812 @itemx -mno-branch-predict
20813 @opindex mbranch-predict
20814 @opindex mno-branch-predict
20815 Use (do not use) the probable-branch instructions, when static branch
20816 prediction indicates a probable branch.
20818 @item -mbase-addresses
20819 @itemx -mno-base-addresses
20820 @opindex mbase-addresses
20821 @opindex mno-base-addresses
20822 Generate (do not generate) code that uses @emph{base addresses}. Using a
20823 base address automatically generates a request (handled by the assembler
20824 and the linker) for a constant to be set up in a global register. The
20825 register is used for one or more base address requests within the range 0
20826 to 255 from the value held in the register. The generally leads to short
20827 and fast code, but the number of different data items that can be
20828 addressed is limited. This means that a program that uses lots of static
20829 data may require @option{-mno-base-addresses}.
20831 @item -msingle-exit
20832 @itemx -mno-single-exit
20833 @opindex msingle-exit
20834 @opindex mno-single-exit
20835 Force (do not force) generated code to have a single exit point in each
20839 @node MN10300 Options
20840 @subsection MN10300 Options
20841 @cindex MN10300 options
20843 These @option{-m} options are defined for Matsushita MN10300 architectures:
20848 Generate code to avoid bugs in the multiply instructions for the MN10300
20849 processors. This is the default.
20851 @item -mno-mult-bug
20852 @opindex mno-mult-bug
20853 Do not generate code to avoid bugs in the multiply instructions for the
20854 MN10300 processors.
20858 Generate code using features specific to the AM33 processor.
20862 Do not generate code using features specific to the AM33 processor. This
20867 Generate code using features specific to the AM33/2.0 processor.
20871 Generate code using features specific to the AM34 processor.
20873 @item -mtune=@var{cpu-type}
20875 Use the timing characteristics of the indicated CPU type when
20876 scheduling instructions. This does not change the targeted processor
20877 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
20878 @samp{am33-2} or @samp{am34}.
20880 @item -mreturn-pointer-on-d0
20881 @opindex mreturn-pointer-on-d0
20882 When generating a function that returns a pointer, return the pointer
20883 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
20884 only in @code{a0}, and attempts to call such functions without a prototype
20885 result in errors. Note that this option is on by default; use
20886 @option{-mno-return-pointer-on-d0} to disable it.
20890 Do not link in the C run-time initialization object file.
20894 Indicate to the linker that it should perform a relaxation optimization pass
20895 to shorten branches, calls and absolute memory addresses. This option only
20896 has an effect when used on the command line for the final link step.
20898 This option makes symbolic debugging impossible.
20902 Allow the compiler to generate @emph{Long Instruction Word}
20903 instructions if the target is the @samp{AM33} or later. This is the
20904 default. This option defines the preprocessor macro @code{__LIW__}.
20908 Do not allow the compiler to generate @emph{Long Instruction Word}
20909 instructions. This option defines the preprocessor macro
20914 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
20915 instructions if the target is the @samp{AM33} or later. This is the
20916 default. This option defines the preprocessor macro @code{__SETLB__}.
20920 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
20921 instructions. This option defines the preprocessor macro
20922 @code{__NO_SETLB__}.
20926 @node Moxie Options
20927 @subsection Moxie Options
20928 @cindex Moxie Options
20934 Generate big-endian code. This is the default for @samp{moxie-*-*}
20939 Generate little-endian code.
20943 Generate mul.x and umul.x instructions. This is the default for
20944 @samp{moxiebox-*-*} configurations.
20948 Do not link in the C run-time initialization object file.
20952 @node MSP430 Options
20953 @subsection MSP430 Options
20954 @cindex MSP430 Options
20956 These options are defined for the MSP430:
20962 Force assembly output to always use hex constants. Normally such
20963 constants are signed decimals, but this option is available for
20964 testsuite and/or aesthetic purposes.
20968 Select the MCU to target. This is used to create a C preprocessor
20969 symbol based upon the MCU name, converted to upper case and pre- and
20970 post-fixed with @samp{__}. This in turn is used by the
20971 @file{msp430.h} header file to select an MCU-specific supplementary
20974 The option also sets the ISA to use. If the MCU name is one that is
20975 known to only support the 430 ISA then that is selected, otherwise the
20976 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
20977 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
20978 name selects the 430X ISA.
20980 In addition an MCU-specific linker script is added to the linker
20981 command line. The script's name is the name of the MCU with
20982 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
20983 command line defines the C preprocessor symbol @code{__XXX__} and
20984 cause the linker to search for a script called @file{xxx.ld}.
20986 This option is also passed on to the assembler.
20989 @itemx -mno-warn-mcu
20991 @opindex mno-warn-mcu
20992 This option enables or disables warnings about conflicts between the
20993 MCU name specified by the @option{-mmcu} option and the ISA set by the
20994 @option{-mcpu} option and/or the hardware multiply support set by the
20995 @option{-mhwmult} option. It also toggles warnings about unrecognized
20996 MCU names. This option is on by default.
21000 Specifies the ISA to use. Accepted values are @samp{msp430},
21001 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
21002 @option{-mmcu=} option should be used to select the ISA.
21006 Link to the simulator runtime libraries and linker script. Overrides
21007 any scripts that would be selected by the @option{-mmcu=} option.
21011 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
21015 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
21019 This option is passed to the assembler and linker, and allows the
21020 linker to perform certain optimizations that cannot be done until
21025 Describes the type of hardware multiply supported by the target.
21026 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
21027 for the original 16-bit-only multiply supported by early MCUs.
21028 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
21029 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
21030 A value of @samp{auto} can also be given. This tells GCC to deduce
21031 the hardware multiply support based upon the MCU name provided by the
21032 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
21033 the MCU name is not recognized then no hardware multiply support is
21034 assumed. @code{auto} is the default setting.
21036 Hardware multiplies are normally performed by calling a library
21037 routine. This saves space in the generated code. When compiling at
21038 @option{-O3} or higher however the hardware multiplier is invoked
21039 inline. This makes for bigger, but faster code.
21041 The hardware multiply routines disable interrupts whilst running and
21042 restore the previous interrupt state when they finish. This makes
21043 them safe to use inside interrupt handlers as well as in normal code.
21047 Enable the use of a minimum runtime environment - no static
21048 initializers or constructors. This is intended for memory-constrained
21049 devices. The compiler includes special symbols in some objects
21050 that tell the linker and runtime which code fragments are required.
21052 @item -mcode-region=
21053 @itemx -mdata-region=
21054 @opindex mcode-region
21055 @opindex mdata-region
21056 These options tell the compiler where to place functions and data that
21057 do not have one of the @code{lower}, @code{upper}, @code{either} or
21058 @code{section} attributes. Possible values are @code{lower},
21059 @code{upper}, @code{either} or @code{any}. The first three behave
21060 like the corresponding attribute. The fourth possible value -
21061 @code{any} - is the default. It leaves placement entirely up to the
21062 linker script and how it assigns the standard sections
21063 (@code{.text}, @code{.data}, etc) to the memory regions.
21065 @item -msilicon-errata=
21066 @opindex msilicon-errata
21067 This option passes on a request to assembler to enable the fixes for
21068 the named silicon errata.
21070 @item -msilicon-errata-warn=
21071 @opindex msilicon-errata-warn
21072 This option passes on a request to the assembler to enable warning
21073 messages when a silicon errata might need to be applied.
21077 @node NDS32 Options
21078 @subsection NDS32 Options
21079 @cindex NDS32 Options
21081 These options are defined for NDS32 implementations:
21086 @opindex mbig-endian
21087 Generate code in big-endian mode.
21089 @item -mlittle-endian
21090 @opindex mlittle-endian
21091 Generate code in little-endian mode.
21093 @item -mreduced-regs
21094 @opindex mreduced-regs
21095 Use reduced-set registers for register allocation.
21098 @opindex mfull-regs
21099 Use full-set registers for register allocation.
21103 Generate conditional move instructions.
21107 Do not generate conditional move instructions.
21111 Generate performance extension instructions.
21113 @item -mno-perf-ext
21114 @opindex mno-perf-ext
21115 Do not generate performance extension instructions.
21119 Generate v3 push25/pop25 instructions.
21122 @opindex mno-v3push
21123 Do not generate v3 push25/pop25 instructions.
21127 Generate 16-bit instructions.
21130 @opindex mno-16-bit
21131 Do not generate 16-bit instructions.
21133 @item -misr-vector-size=@var{num}
21134 @opindex misr-vector-size
21135 Specify the size of each interrupt vector, which must be 4 or 16.
21137 @item -mcache-block-size=@var{num}
21138 @opindex mcache-block-size
21139 Specify the size of each cache block,
21140 which must be a power of 2 between 4 and 512.
21142 @item -march=@var{arch}
21144 Specify the name of the target architecture.
21146 @item -mcmodel=@var{code-model}
21148 Set the code model to one of
21151 All the data and read-only data segments must be within 512KB addressing space.
21152 The text segment must be within 16MB addressing space.
21153 @item @samp{medium}
21154 The data segment must be within 512KB while the read-only data segment can be
21155 within 4GB addressing space. The text segment should be still within 16MB
21158 All the text and data segments can be within 4GB addressing space.
21162 @opindex mctor-dtor
21163 Enable constructor/destructor feature.
21167 Guide linker to relax instructions.
21171 @node Nios II Options
21172 @subsection Nios II Options
21173 @cindex Nios II options
21174 @cindex Altera Nios II options
21176 These are the options defined for the Altera Nios II processor.
21182 @cindex smaller data references
21183 Put global and static objects less than or equal to @var{num} bytes
21184 into the small data or BSS sections instead of the normal data or BSS
21185 sections. The default value of @var{num} is 8.
21187 @item -mgpopt=@var{option}
21192 Generate (do not generate) GP-relative accesses. The following
21193 @var{option} names are recognized:
21198 Do not generate GP-relative accesses.
21201 Generate GP-relative accesses for small data objects that are not
21202 external, weak, or uninitialized common symbols.
21203 Also use GP-relative addressing for objects that
21204 have been explicitly placed in a small data section via a @code{section}
21208 As for @samp{local}, but also generate GP-relative accesses for
21209 small data objects that are external, weak, or common. If you use this option,
21210 you must ensure that all parts of your program (including libraries) are
21211 compiled with the same @option{-G} setting.
21214 Generate GP-relative accesses for all data objects in the program. If you
21215 use this option, the entire data and BSS segments
21216 of your program must fit in 64K of memory and you must use an appropriate
21217 linker script to allocate them within the addressable range of the
21221 Generate GP-relative addresses for function pointers as well as data
21222 pointers. If you use this option, the entire text, data, and BSS segments
21223 of your program must fit in 64K of memory and you must use an appropriate
21224 linker script to allocate them within the addressable range of the
21229 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
21230 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
21232 The default is @option{-mgpopt} except when @option{-fpic} or
21233 @option{-fPIC} is specified to generate position-independent code.
21234 Note that the Nios II ABI does not permit GP-relative accesses from
21237 You may need to specify @option{-mno-gpopt} explicitly when building
21238 programs that include large amounts of small data, including large
21239 GOT data sections. In this case, the 16-bit offset for GP-relative
21240 addressing may not be large enough to allow access to the entire
21241 small data section.
21243 @item -mgprel-sec=@var{regexp}
21244 @opindex mgprel-sec
21245 This option specifies additional section names that can be accessed via
21246 GP-relative addressing. It is most useful in conjunction with
21247 @code{section} attributes on variable declarations
21248 (@pxref{Common Variable Attributes}) and a custom linker script.
21249 The @var{regexp} is a POSIX Extended Regular Expression.
21251 This option does not affect the behavior of the @option{-G} option, and
21252 and the specified sections are in addition to the standard @code{.sdata}
21253 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
21255 @item -mr0rel-sec=@var{regexp}
21256 @opindex mr0rel-sec
21257 This option specifies names of sections that can be accessed via a
21258 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
21259 of the 32-bit address space. It is most useful in conjunction with
21260 @code{section} attributes on variable declarations
21261 (@pxref{Common Variable Attributes}) and a custom linker script.
21262 The @var{regexp} is a POSIX Extended Regular Expression.
21264 In contrast to the use of GP-relative addressing for small data,
21265 zero-based addressing is never generated by default and there are no
21266 conventional section names used in standard linker scripts for sections
21267 in the low or high areas of memory.
21273 Generate little-endian (default) or big-endian (experimental) code,
21276 @item -march=@var{arch}
21278 This specifies the name of the target Nios II architecture. GCC uses this
21279 name to determine what kind of instructions it can emit when generating
21280 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
21282 The preprocessor macro @code{__nios2_arch__} is available to programs,
21283 with value 1 or 2, indicating the targeted ISA level.
21285 @item -mbypass-cache
21286 @itemx -mno-bypass-cache
21287 @opindex mno-bypass-cache
21288 @opindex mbypass-cache
21289 Force all load and store instructions to always bypass cache by
21290 using I/O variants of the instructions. The default is not to
21293 @item -mno-cache-volatile
21294 @itemx -mcache-volatile
21295 @opindex mcache-volatile
21296 @opindex mno-cache-volatile
21297 Volatile memory access bypass the cache using the I/O variants of
21298 the load and store instructions. The default is not to bypass the cache.
21300 @item -mno-fast-sw-div
21301 @itemx -mfast-sw-div
21302 @opindex mno-fast-sw-div
21303 @opindex mfast-sw-div
21304 Do not use table-based fast divide for small numbers. The default
21305 is to use the fast divide at @option{-O3} and above.
21309 @itemx -mno-hw-mulx
21313 @opindex mno-hw-mul
21315 @opindex mno-hw-mulx
21317 @opindex mno-hw-div
21319 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
21320 instructions by the compiler. The default is to emit @code{mul}
21321 and not emit @code{div} and @code{mulx}.
21327 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
21328 CDX (code density) instructions. Enabling these instructions also
21329 requires @option{-march=r2}. Since these instructions are optional
21330 extensions to the R2 architecture, the default is not to emit them.
21332 @item -mcustom-@var{insn}=@var{N}
21333 @itemx -mno-custom-@var{insn}
21334 @opindex mcustom-@var{insn}
21335 @opindex mno-custom-@var{insn}
21336 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
21337 custom instruction with encoding @var{N} when generating code that uses
21338 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
21339 instruction 253 for single-precision floating-point add operations instead
21340 of the default behavior of using a library call.
21342 The following values of @var{insn} are supported. Except as otherwise
21343 noted, floating-point operations are expected to be implemented with
21344 normal IEEE 754 semantics and correspond directly to the C operators or the
21345 equivalent GCC built-in functions (@pxref{Other Builtins}).
21347 Single-precision floating point:
21350 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
21351 Binary arithmetic operations.
21357 Unary absolute value.
21359 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
21360 Comparison operations.
21362 @item @samp{fmins}, @samp{fmaxs}
21363 Floating-point minimum and maximum. These instructions are only
21364 generated if @option{-ffinite-math-only} is specified.
21366 @item @samp{fsqrts}
21367 Unary square root operation.
21369 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
21370 Floating-point trigonometric and exponential functions. These instructions
21371 are only generated if @option{-funsafe-math-optimizations} is also specified.
21375 Double-precision floating point:
21378 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
21379 Binary arithmetic operations.
21385 Unary absolute value.
21387 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
21388 Comparison operations.
21390 @item @samp{fmind}, @samp{fmaxd}
21391 Double-precision minimum and maximum. These instructions are only
21392 generated if @option{-ffinite-math-only} is specified.
21394 @item @samp{fsqrtd}
21395 Unary square root operation.
21397 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
21398 Double-precision trigonometric and exponential functions. These instructions
21399 are only generated if @option{-funsafe-math-optimizations} is also specified.
21405 @item @samp{fextsd}
21406 Conversion from single precision to double precision.
21408 @item @samp{ftruncds}
21409 Conversion from double precision to single precision.
21411 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
21412 Conversion from floating point to signed or unsigned integer types, with
21413 truncation towards zero.
21416 Conversion from single-precision floating point to signed integer,
21417 rounding to the nearest integer and ties away from zero.
21418 This corresponds to the @code{__builtin_lroundf} function when
21419 @option{-fno-math-errno} is used.
21421 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
21422 Conversion from signed or unsigned integer types to floating-point types.
21426 In addition, all of the following transfer instructions for internal
21427 registers X and Y must be provided to use any of the double-precision
21428 floating-point instructions. Custom instructions taking two
21429 double-precision source operands expect the first operand in the
21430 64-bit register X. The other operand (or only operand of a unary
21431 operation) is given to the custom arithmetic instruction with the
21432 least significant half in source register @var{src1} and the most
21433 significant half in @var{src2}. A custom instruction that returns a
21434 double-precision result returns the most significant 32 bits in the
21435 destination register and the other half in 32-bit register Y.
21436 GCC automatically generates the necessary code sequences to write
21437 register X and/or read register Y when double-precision floating-point
21438 instructions are used.
21443 Write @var{src1} into the least significant half of X and @var{src2} into
21444 the most significant half of X.
21447 Write @var{src1} into Y.
21449 @item @samp{frdxhi}, @samp{frdxlo}
21450 Read the most or least (respectively) significant half of X and store it in
21454 Read the value of Y and store it into @var{dest}.
21457 Note that you can gain more local control over generation of Nios II custom
21458 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
21459 and @code{target("no-custom-@var{insn}")} function attributes
21460 (@pxref{Function Attributes})
21461 or pragmas (@pxref{Function Specific Option Pragmas}).
21463 @item -mcustom-fpu-cfg=@var{name}
21464 @opindex mcustom-fpu-cfg
21466 This option enables a predefined, named set of custom instruction encodings
21467 (see @option{-mcustom-@var{insn}} above).
21468 Currently, the following sets are defined:
21470 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
21471 @gccoptlist{-mcustom-fmuls=252 @gol
21472 -mcustom-fadds=253 @gol
21473 -mcustom-fsubs=254 @gol
21474 -fsingle-precision-constant}
21476 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
21477 @gccoptlist{-mcustom-fmuls=252 @gol
21478 -mcustom-fadds=253 @gol
21479 -mcustom-fsubs=254 @gol
21480 -mcustom-fdivs=255 @gol
21481 -fsingle-precision-constant}
21483 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
21484 @gccoptlist{-mcustom-floatus=243 @gol
21485 -mcustom-fixsi=244 @gol
21486 -mcustom-floatis=245 @gol
21487 -mcustom-fcmpgts=246 @gol
21488 -mcustom-fcmples=249 @gol
21489 -mcustom-fcmpeqs=250 @gol
21490 -mcustom-fcmpnes=251 @gol
21491 -mcustom-fmuls=252 @gol
21492 -mcustom-fadds=253 @gol
21493 -mcustom-fsubs=254 @gol
21494 -mcustom-fdivs=255 @gol
21495 -fsingle-precision-constant}
21497 Custom instruction assignments given by individual
21498 @option{-mcustom-@var{insn}=} options override those given by
21499 @option{-mcustom-fpu-cfg=}, regardless of the
21500 order of the options on the command line.
21502 Note that you can gain more local control over selection of a FPU
21503 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
21504 function attribute (@pxref{Function Attributes})
21505 or pragma (@pxref{Function Specific Option Pragmas}).
21509 These additional @samp{-m} options are available for the Altera Nios II
21510 ELF (bare-metal) target:
21516 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
21517 startup and termination code, and is typically used in conjunction with
21518 @option{-msys-crt0=} to specify the location of the alternate startup code
21519 provided by the HAL BSP.
21523 Link with a limited version of the C library, @option{-lsmallc}, rather than
21526 @item -msys-crt0=@var{startfile}
21528 @var{startfile} is the file name of the startfile (crt0) to use
21529 when linking. This option is only useful in conjunction with @option{-mhal}.
21531 @item -msys-lib=@var{systemlib}
21533 @var{systemlib} is the library name of the library that provides
21534 low-level system calls required by the C library,
21535 e.g. @code{read} and @code{write}.
21536 This option is typically used to link with a library provided by a HAL BSP.
21540 @node Nvidia PTX Options
21541 @subsection Nvidia PTX Options
21542 @cindex Nvidia PTX options
21543 @cindex nvptx options
21545 These options are defined for Nvidia PTX:
21553 Generate code for 32-bit or 64-bit ABI.
21556 @opindex mmainkernel
21557 Link in code for a __main kernel. This is for stand-alone instead of
21558 offloading execution.
21562 Apply partitioned execution optimizations. This is the default when any
21563 level of optimization is selected.
21566 @opindex msoft-stack
21567 Generate code that does not use @code{.local} memory
21568 directly for stack storage. Instead, a per-warp stack pointer is
21569 maintained explicitly. This enables variable-length stack allocation (with
21570 variable-length arrays or @code{alloca}), and when global memory is used for
21571 underlying storage, makes it possible to access automatic variables from other
21572 threads, or with atomic instructions. This code generation variant is used
21573 for OpenMP offloading, but the option is exposed on its own for the purpose
21574 of testing the compiler; to generate code suitable for linking into programs
21575 using OpenMP offloading, use option @option{-mgomp}.
21577 @item -muniform-simt
21578 @opindex muniform-simt
21579 Switch to code generation variant that allows to execute all threads in each
21580 warp, while maintaining memory state and side effects as if only one thread
21581 in each warp was active outside of OpenMP SIMD regions. All atomic operations
21582 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
21583 current lane index equals the master lane index), and the register being
21584 assigned is copied via a shuffle instruction from the master lane. Outside of
21585 SIMD regions lane 0 is the master; inside, each thread sees itself as the
21586 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
21587 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
21588 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
21589 with current lane index to compute the master lane index.
21593 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
21594 @option{-muniform-simt} options, and selects corresponding multilib variant.
21598 @node PDP-11 Options
21599 @subsection PDP-11 Options
21600 @cindex PDP-11 Options
21602 These options are defined for the PDP-11:
21607 Use hardware FPP floating point. This is the default. (FIS floating
21608 point on the PDP-11/40 is not supported.)
21611 @opindex msoft-float
21612 Do not use hardware floating point.
21616 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
21620 Return floating-point results in memory. This is the default.
21624 Generate code for a PDP-11/40.
21628 Generate code for a PDP-11/45. This is the default.
21632 Generate code for a PDP-11/10.
21634 @item -mbcopy-builtin
21635 @opindex mbcopy-builtin
21636 Use inline @code{movmemhi} patterns for copying memory. This is the
21641 Do not use inline @code{movmemhi} patterns for copying memory.
21647 Use 16-bit @code{int}. This is the default.
21653 Use 32-bit @code{int}.
21656 @itemx -mno-float32
21658 @opindex mno-float32
21659 Use 64-bit @code{float}. This is the default.
21662 @itemx -mno-float64
21664 @opindex mno-float64
21665 Use 32-bit @code{float}.
21669 Use @code{abshi2} pattern. This is the default.
21673 Do not use @code{abshi2} pattern.
21675 @item -mbranch-expensive
21676 @opindex mbranch-expensive
21677 Pretend that branches are expensive. This is for experimenting with
21678 code generation only.
21680 @item -mbranch-cheap
21681 @opindex mbranch-cheap
21682 Do not pretend that branches are expensive. This is the default.
21686 Use Unix assembler syntax. This is the default when configured for
21687 @samp{pdp11-*-bsd}.
21691 Use DEC assembler syntax. This is the default when configured for any
21692 PDP-11 target other than @samp{pdp11-*-bsd}.
21695 @node picoChip Options
21696 @subsection picoChip Options
21697 @cindex picoChip options
21699 These @samp{-m} options are defined for picoChip implementations:
21703 @item -mae=@var{ae_type}
21705 Set the instruction set, register set, and instruction scheduling
21706 parameters for array element type @var{ae_type}. Supported values
21707 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
21709 @option{-mae=ANY} selects a completely generic AE type. Code
21710 generated with this option runs on any of the other AE types. The
21711 code is not as efficient as it would be if compiled for a specific
21712 AE type, and some types of operation (e.g., multiplication) do not
21713 work properly on all types of AE.
21715 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
21716 for compiled code, and is the default.
21718 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
21719 option may suffer from poor performance of byte (char) manipulation,
21720 since the DSP AE does not provide hardware support for byte load/stores.
21722 @item -msymbol-as-address
21723 Enable the compiler to directly use a symbol name as an address in a
21724 load/store instruction, without first loading it into a
21725 register. Typically, the use of this option generates larger
21726 programs, which run faster than when the option isn't used. However, the
21727 results vary from program to program, so it is left as a user option,
21728 rather than being permanently enabled.
21730 @item -mno-inefficient-warnings
21731 Disables warnings about the generation of inefficient code. These
21732 warnings can be generated, for example, when compiling code that
21733 performs byte-level memory operations on the MAC AE type. The MAC AE has
21734 no hardware support for byte-level memory operations, so all byte
21735 load/stores must be synthesized from word load/store operations. This is
21736 inefficient and a warning is generated to indicate
21737 that you should rewrite the code to avoid byte operations, or to target
21738 an AE type that has the necessary hardware support. This option disables
21743 @node PowerPC Options
21744 @subsection PowerPC Options
21745 @cindex PowerPC options
21747 These are listed under @xref{RS/6000 and PowerPC Options}.
21749 @node RISC-V Options
21750 @subsection RISC-V Options
21751 @cindex RISC-V Options
21753 These command-line options are defined for RISC-V targets:
21756 @item -mbranch-cost=@var{n}
21757 @opindex mbranch-cost
21758 Set the cost of branches to roughly @var{n} instructions.
21763 Don't optimize block moves.
21768 When generating PIC code, allow the use of PLTs. Ignored for non-PIC.
21770 @item -mabi=@var{ABI-string}
21772 @item -mabi=@var{ABI-string}
21774 Specify integer and floating-point calling convention. @var{ABI-string}
21775 contains two parts: the size of integer types and the registers used for
21776 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
21777 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
21778 32-bit), and that floating-point values up to 64 bits wide are passed in F
21779 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
21780 allows the compiler to generate code that uses the F and D extensions but only
21781 allows floating-point values up to 32 bits long to be passed in registers; or
21782 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
21783 passed in registers.
21785 The default for this argument is system dependent, users who want a specific
21786 calling convention should specify one explicitly. The valid calling
21787 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
21788 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
21789 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
21790 invalid because the ABI requires 64-bit values be passed in F registers, but F
21791 registers are only 32 bits wide.
21796 Use hardware floating-point divide and square root instructions. This requires
21797 the F or D extensions for floating-point registers.
21802 Use hardware instructions for integer division. This requires the M extension.
21804 @item -march=@var{ISA-string}
21806 Generate code for given RISC-V ISA (e.g.@ @samp{rv64im}). ISA strings must be
21807 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, and @samp{rv32imaf}.
21809 @item -mtune=@var{processor-string}
21811 Optimize the output for the given processor, specified by microarchitecture
21814 @item -msmall-data-limit=@var{n}
21815 @opindex msmall-data-limit
21816 Put global and static data smaller than @var{n} bytes into a special section
21819 @item -msave-restore
21820 @itemx -mno-save-restore
21821 @opindex msave-restore
21822 Use smaller but slower prologue and epilogue code.
21824 @item -mstrict-align
21825 @itemx -mno-strict-align
21826 @opindex mstrict-align
21827 Do not generate unaligned memory accesses.
21829 @item -mcmodel=medlow
21830 @opindex mcmodel=medlow
21831 Generate code for the medium-low code model. The program and its statically
21832 defined symbols must lie within a single 2 GiB address range and must lie
21833 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
21834 statically or dynamically linked. This is the default code model.
21836 @item -mcmodel=medany
21837 @opindex mcmodel=medany
21838 Generate code for the medium-any code model. The program and its statically
21839 defined symbols must be within any single 2 GiB address range. Programs can be
21840 statically or dynamically linked.
21845 @subsection RL78 Options
21846 @cindex RL78 Options
21852 Links in additional target libraries to support operation within a
21861 Specifies the type of hardware multiplication and division support to
21862 be used. The simplest is @code{none}, which uses software for both
21863 multiplication and division. This is the default. The @code{g13}
21864 value is for the hardware multiply/divide peripheral found on the
21865 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
21866 the multiplication and division instructions supported by the RL78/G14
21867 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
21868 the value @code{mg10} is an alias for @code{none}.
21870 In addition a C preprocessor macro is defined, based upon the setting
21871 of this option. Possible values are: @code{__RL78_MUL_NONE__},
21872 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
21879 Specifies the RL78 core to target. The default is the G14 core, also
21880 known as an S3 core or just RL78. The G13 or S2 core does not have
21881 multiply or divide instructions, instead it uses a hardware peripheral
21882 for these operations. The G10 or S1 core does not have register
21883 banks, so it uses a different calling convention.
21885 If this option is set it also selects the type of hardware multiply
21886 support to use, unless this is overridden by an explicit
21887 @option{-mmul=none} option on the command line. Thus specifying
21888 @option{-mcpu=g13} enables the use of the G13 hardware multiply
21889 peripheral and specifying @option{-mcpu=g10} disables the use of
21890 hardware multiplications altogether.
21892 Note, although the RL78/G14 core is the default target, specifying
21893 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
21894 change the behavior of the toolchain since it also enables G14
21895 hardware multiply support. If these options are not specified on the
21896 command line then software multiplication routines will be used even
21897 though the code targets the RL78 core. This is for backwards
21898 compatibility with older toolchains which did not have hardware
21899 multiply and divide support.
21901 In addition a C preprocessor macro is defined, based upon the setting
21902 of this option. Possible values are: @code{__RL78_G10__},
21903 @code{__RL78_G13__} or @code{__RL78_G14__}.
21913 These are aliases for the corresponding @option{-mcpu=} option. They
21914 are provided for backwards compatibility.
21918 Allow the compiler to use all of the available registers. By default
21919 registers @code{r24..r31} are reserved for use in interrupt handlers.
21920 With this option enabled these registers can be used in ordinary
21923 @item -m64bit-doubles
21924 @itemx -m32bit-doubles
21925 @opindex m64bit-doubles
21926 @opindex m32bit-doubles
21927 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21928 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21929 @option{-m32bit-doubles}.
21931 @item -msave-mduc-in-interrupts
21932 @item -mno-save-mduc-in-interrupts
21933 @opindex msave-mduc-in-interrupts
21934 @opindex mno-save-mduc-in-interrupts
21935 Specifies that interrupt handler functions should preserve the
21936 MDUC registers. This is only necessary if normal code might use
21937 the MDUC registers, for example because it performs multiplication
21938 and division operations. The default is to ignore the MDUC registers
21939 as this makes the interrupt handlers faster. The target option -mg13
21940 needs to be passed for this to work as this feature is only available
21941 on the G13 target (S2 core). The MDUC registers will only be saved
21942 if the interrupt handler performs a multiplication or division
21943 operation or it calls another function.
21947 @node RS/6000 and PowerPC Options
21948 @subsection IBM RS/6000 and PowerPC Options
21949 @cindex RS/6000 and PowerPC Options
21950 @cindex IBM RS/6000 and PowerPC Options
21952 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
21954 @item -mpowerpc-gpopt
21955 @itemx -mno-powerpc-gpopt
21956 @itemx -mpowerpc-gfxopt
21957 @itemx -mno-powerpc-gfxopt
21960 @itemx -mno-powerpc64
21964 @itemx -mno-popcntb
21966 @itemx -mno-popcntd
21975 @itemx -mno-hard-dfp
21976 @opindex mpowerpc-gpopt
21977 @opindex mno-powerpc-gpopt
21978 @opindex mpowerpc-gfxopt
21979 @opindex mno-powerpc-gfxopt
21980 @opindex mpowerpc64
21981 @opindex mno-powerpc64
21985 @opindex mno-popcntb
21987 @opindex mno-popcntd
21993 @opindex mno-mfpgpr
21995 @opindex mno-hard-dfp
21996 You use these options to specify which instructions are available on the
21997 processor you are using. The default value of these options is
21998 determined when configuring GCC@. Specifying the
21999 @option{-mcpu=@var{cpu_type}} overrides the specification of these
22000 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
22001 rather than the options listed above.
22003 Specifying @option{-mpowerpc-gpopt} allows
22004 GCC to use the optional PowerPC architecture instructions in the
22005 General Purpose group, including floating-point square root. Specifying
22006 @option{-mpowerpc-gfxopt} allows GCC to
22007 use the optional PowerPC architecture instructions in the Graphics
22008 group, including floating-point select.
22010 The @option{-mmfcrf} option allows GCC to generate the move from
22011 condition register field instruction implemented on the POWER4
22012 processor and other processors that support the PowerPC V2.01
22014 The @option{-mpopcntb} option allows GCC to generate the popcount and
22015 double-precision FP reciprocal estimate instruction implemented on the
22016 POWER5 processor and other processors that support the PowerPC V2.02
22018 The @option{-mpopcntd} option allows GCC to generate the popcount
22019 instruction implemented on the POWER7 processor and other processors
22020 that support the PowerPC V2.06 architecture.
22021 The @option{-mfprnd} option allows GCC to generate the FP round to
22022 integer instructions implemented on the POWER5+ processor and other
22023 processors that support the PowerPC V2.03 architecture.
22024 The @option{-mcmpb} option allows GCC to generate the compare bytes
22025 instruction implemented on the POWER6 processor and other processors
22026 that support the PowerPC V2.05 architecture.
22027 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
22028 general-purpose register instructions implemented on the POWER6X
22029 processor and other processors that support the extended PowerPC V2.05
22031 The @option{-mhard-dfp} option allows GCC to generate the decimal
22032 floating-point instructions implemented on some POWER processors.
22034 The @option{-mpowerpc64} option allows GCC to generate the additional
22035 64-bit instructions that are found in the full PowerPC64 architecture
22036 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
22037 @option{-mno-powerpc64}.
22039 @item -mcpu=@var{cpu_type}
22041 Set architecture type, register usage, and
22042 instruction scheduling parameters for machine type @var{cpu_type}.
22043 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
22044 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
22045 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
22046 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
22047 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
22048 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
22049 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
22050 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
22051 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
22052 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
22053 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
22056 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
22057 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
22058 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
22059 architecture machine types, with an appropriate, generic processor
22060 model assumed for scheduling purposes.
22062 The other options specify a specific processor. Code generated under
22063 those options runs best on that processor, and may not run at all on
22066 The @option{-mcpu} options automatically enable or disable the
22069 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
22070 -mpopcntb -mpopcntd -mpowerpc64 @gol
22071 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
22072 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
22073 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
22074 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
22076 The particular options set for any particular CPU varies between
22077 compiler versions, depending on what setting seems to produce optimal
22078 code for that CPU; it doesn't necessarily reflect the actual hardware's
22079 capabilities. If you wish to set an individual option to a particular
22080 value, you may specify it after the @option{-mcpu} option, like
22081 @option{-mcpu=970 -mno-altivec}.
22083 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
22084 not enabled or disabled by the @option{-mcpu} option at present because
22085 AIX does not have full support for these options. You may still
22086 enable or disable them individually if you're sure it'll work in your
22089 @item -mtune=@var{cpu_type}
22091 Set the instruction scheduling parameters for machine type
22092 @var{cpu_type}, but do not set the architecture type or register usage,
22093 as @option{-mcpu=@var{cpu_type}} does. The same
22094 values for @var{cpu_type} are used for @option{-mtune} as for
22095 @option{-mcpu}. If both are specified, the code generated uses the
22096 architecture and registers set by @option{-mcpu}, but the
22097 scheduling parameters set by @option{-mtune}.
22099 @item -mcmodel=small
22100 @opindex mcmodel=small
22101 Generate PowerPC64 code for the small model: The TOC is limited to
22104 @item -mcmodel=medium
22105 @opindex mcmodel=medium
22106 Generate PowerPC64 code for the medium model: The TOC and other static
22107 data may be up to a total of 4G in size. This is the default for 64-bit
22110 @item -mcmodel=large
22111 @opindex mcmodel=large
22112 Generate PowerPC64 code for the large model: The TOC may be up to 4G
22113 in size. Other data and code is only limited by the 64-bit address
22117 @itemx -mno-altivec
22119 @opindex mno-altivec
22120 Generate code that uses (does not use) AltiVec instructions, and also
22121 enable the use of built-in functions that allow more direct access to
22122 the AltiVec instruction set. You may also need to set
22123 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
22126 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
22127 @option{-maltivec=be}, the element order for AltiVec intrinsics such
22128 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
22129 match array element order corresponding to the endianness of the
22130 target. That is, element zero identifies the leftmost element in a
22131 vector register when targeting a big-endian platform, and identifies
22132 the rightmost element in a vector register when targeting a
22133 little-endian platform.
22136 @opindex maltivec=be
22137 Generate AltiVec instructions using big-endian element order,
22138 regardless of whether the target is big- or little-endian. This is
22139 the default when targeting a big-endian platform.
22141 The element order is used to interpret element numbers in AltiVec
22142 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22143 @code{vec_insert}. By default, these match array element order
22144 corresponding to the endianness for the target.
22147 @opindex maltivec=le
22148 Generate AltiVec instructions using little-endian element order,
22149 regardless of whether the target is big- or little-endian. This is
22150 the default when targeting a little-endian platform. This option is
22151 currently ignored when targeting a big-endian platform.
22153 The element order is used to interpret element numbers in AltiVec
22154 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22155 @code{vec_insert}. By default, these match array element order
22156 corresponding to the endianness for the target.
22161 @opindex mno-vrsave
22162 Generate VRSAVE instructions when generating AltiVec code.
22165 @opindex msecure-plt
22166 Generate code that allows @command{ld} and @command{ld.so}
22167 to build executables and shared
22168 libraries with non-executable @code{.plt} and @code{.got} sections.
22170 32-bit SYSV ABI option.
22174 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
22176 requires @code{.plt} and @code{.got}
22177 sections that are both writable and executable.
22178 This is a PowerPC 32-bit SYSV ABI option.
22184 This switch enables or disables the generation of ISEL instructions.
22186 @item -misel=@var{yes/no}
22187 This switch has been deprecated. Use @option{-misel} and
22188 @option{-mno-isel} instead.
22194 This switch enables or disables the generation of SPE simd
22200 @opindex mno-paired
22201 This switch enables or disables the generation of PAIRED simd
22204 @item -mspe=@var{yes/no}
22205 This option has been deprecated. Use @option{-mspe} and
22206 @option{-mno-spe} instead.
22212 Generate code that uses (does not use) vector/scalar (VSX)
22213 instructions, and also enable the use of built-in functions that allow
22214 more direct access to the VSX instruction set.
22219 @opindex mno-crypto
22220 Enable the use (disable) of the built-in functions that allow direct
22221 access to the cryptographic instructions that were added in version
22222 2.07 of the PowerPC ISA.
22224 @item -mdirect-move
22225 @itemx -mno-direct-move
22226 @opindex mdirect-move
22227 @opindex mno-direct-move
22228 Generate code that uses (does not use) the instructions to move data
22229 between the general purpose registers and the vector/scalar (VSX)
22230 registers that were added in version 2.07 of the PowerPC ISA.
22236 Enable (disable) the use of the built-in functions that allow direct
22237 access to the Hardware Transactional Memory (HTM) instructions that
22238 were added in version 2.07 of the PowerPC ISA.
22240 @item -mpower8-fusion
22241 @itemx -mno-power8-fusion
22242 @opindex mpower8-fusion
22243 @opindex mno-power8-fusion
22244 Generate code that keeps (does not keeps) some integer operations
22245 adjacent so that the instructions can be fused together on power8 and
22248 @item -mpower8-vector
22249 @itemx -mno-power8-vector
22250 @opindex mpower8-vector
22251 @opindex mno-power8-vector
22252 Generate code that uses (does not use) the vector and scalar
22253 instructions that were added in version 2.07 of the PowerPC ISA. Also
22254 enable the use of built-in functions that allow more direct access to
22255 the vector instructions.
22257 @item -mquad-memory
22258 @itemx -mno-quad-memory
22259 @opindex mquad-memory
22260 @opindex mno-quad-memory
22261 Generate code that uses (does not use) the non-atomic quad word memory
22262 instructions. The @option{-mquad-memory} option requires use of
22265 @item -mquad-memory-atomic
22266 @itemx -mno-quad-memory-atomic
22267 @opindex mquad-memory-atomic
22268 @opindex mno-quad-memory-atomic
22269 Generate code that uses (does not use) the atomic quad word memory
22270 instructions. The @option{-mquad-memory-atomic} option requires use of
22274 @itemx -mno-float128
22276 @opindex mno-float128
22277 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
22278 and use either software emulation for IEEE 128-bit floating point or
22279 hardware instructions.
22281 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
22282 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
22283 use the IEEE 128-bit floating point support. The IEEE 128-bit
22284 floating point support only works on PowerPC Linux systems.
22286 The default for @option{-mfloat128} is enabled on PowerPC Linux
22287 systems using the VSX instruction set, and disabled on other systems.
22289 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
22290 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
22291 point support will also enable the generation of ISA 3.0 IEEE 128-bit
22292 floating point instructions. Otherwise, if you do not specify to
22293 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
22294 system, IEEE 128-bit floating point will be done with software
22297 @item -mfloat128-hardware
22298 @itemx -mno-float128-hardware
22299 @opindex mfloat128-hardware
22300 @opindex mno-float128-hardware
22301 Enable/disable using ISA 3.0 hardware instructions to support the
22302 @var{__float128} data type.
22304 The default for @option{-mfloat128-hardware} is enabled on PowerPC
22305 Linux systems using the ISA 3.0 instruction set, and disabled on other
22308 @item -mfloat-gprs=@var{yes/single/double/no}
22309 @itemx -mfloat-gprs
22310 @opindex mfloat-gprs
22311 This switch enables or disables the generation of floating-point
22312 operations on the general-purpose registers for architectures that
22315 The argument @samp{yes} or @samp{single} enables the use of
22316 single-precision floating-point operations.
22318 The argument @samp{double} enables the use of single and
22319 double-precision floating-point operations.
22321 The argument @samp{no} disables floating-point operations on the
22322 general-purpose registers.
22324 This option is currently only available on the MPC854x.
22330 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
22331 targets (including GNU/Linux). The 32-bit environment sets int, long
22332 and pointer to 32 bits and generates code that runs on any PowerPC
22333 variant. The 64-bit environment sets int to 32 bits and long and
22334 pointer to 64 bits, and generates code for PowerPC64, as for
22335 @option{-mpowerpc64}.
22338 @itemx -mno-fp-in-toc
22339 @itemx -mno-sum-in-toc
22340 @itemx -mminimal-toc
22342 @opindex mno-fp-in-toc
22343 @opindex mno-sum-in-toc
22344 @opindex mminimal-toc
22345 Modify generation of the TOC (Table Of Contents), which is created for
22346 every executable file. The @option{-mfull-toc} option is selected by
22347 default. In that case, GCC allocates at least one TOC entry for
22348 each unique non-automatic variable reference in your program. GCC
22349 also places floating-point constants in the TOC@. However, only
22350 16,384 entries are available in the TOC@.
22352 If you receive a linker error message that saying you have overflowed
22353 the available TOC space, you can reduce the amount of TOC space used
22354 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
22355 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
22356 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
22357 generate code to calculate the sum of an address and a constant at
22358 run time instead of putting that sum into the TOC@. You may specify one
22359 or both of these options. Each causes GCC to produce very slightly
22360 slower and larger code at the expense of conserving TOC space.
22362 If you still run out of space in the TOC even when you specify both of
22363 these options, specify @option{-mminimal-toc} instead. This option causes
22364 GCC to make only one TOC entry for every file. When you specify this
22365 option, GCC produces code that is slower and larger but which
22366 uses extremely little TOC space. You may wish to use this option
22367 only on files that contain less frequently-executed code.
22373 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
22374 @code{long} type, and the infrastructure needed to support them.
22375 Specifying @option{-maix64} implies @option{-mpowerpc64},
22376 while @option{-maix32} disables the 64-bit ABI and
22377 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
22380 @itemx -mno-xl-compat
22381 @opindex mxl-compat
22382 @opindex mno-xl-compat
22383 Produce code that conforms more closely to IBM XL compiler semantics
22384 when using AIX-compatible ABI@. Pass floating-point arguments to
22385 prototyped functions beyond the register save area (RSA) on the stack
22386 in addition to argument FPRs. Do not assume that most significant
22387 double in 128-bit long double value is properly rounded when comparing
22388 values and converting to double. Use XL symbol names for long double
22391 The AIX calling convention was extended but not initially documented to
22392 handle an obscure K&R C case of calling a function that takes the
22393 address of its arguments with fewer arguments than declared. IBM XL
22394 compilers access floating-point arguments that do not fit in the
22395 RSA from the stack when a subroutine is compiled without
22396 optimization. Because always storing floating-point arguments on the
22397 stack is inefficient and rarely needed, this option is not enabled by
22398 default and only is necessary when calling subroutines compiled by IBM
22399 XL compilers without optimization.
22403 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
22404 application written to use message passing with special startup code to
22405 enable the application to run. The system must have PE installed in the
22406 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
22407 must be overridden with the @option{-specs=} option to specify the
22408 appropriate directory location. The Parallel Environment does not
22409 support threads, so the @option{-mpe} option and the @option{-pthread}
22410 option are incompatible.
22412 @item -malign-natural
22413 @itemx -malign-power
22414 @opindex malign-natural
22415 @opindex malign-power
22416 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
22417 @option{-malign-natural} overrides the ABI-defined alignment of larger
22418 types, such as floating-point doubles, on their natural size-based boundary.
22419 The option @option{-malign-power} instructs GCC to follow the ABI-specified
22420 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
22422 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
22426 @itemx -mhard-float
22427 @opindex msoft-float
22428 @opindex mhard-float
22429 Generate code that does not use (uses) the floating-point register set.
22430 Software floating-point emulation is provided if you use the
22431 @option{-msoft-float} option, and pass the option to GCC when linking.
22433 @item -msingle-float
22434 @itemx -mdouble-float
22435 @opindex msingle-float
22436 @opindex mdouble-float
22437 Generate code for single- or double-precision floating-point operations.
22438 @option{-mdouble-float} implies @option{-msingle-float}.
22441 @opindex msimple-fpu
22442 Do not generate @code{sqrt} and @code{div} instructions for hardware
22443 floating-point unit.
22445 @item -mfpu=@var{name}
22447 Specify type of floating-point unit. Valid values for @var{name} are
22448 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
22449 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
22450 @samp{sp_full} (equivalent to @option{-msingle-float}),
22451 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
22454 @opindex mxilinx-fpu
22455 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
22458 @itemx -mno-multiple
22460 @opindex mno-multiple
22461 Generate code that uses (does not use) the load multiple word
22462 instructions and the store multiple word instructions. These
22463 instructions are generated by default on POWER systems, and not
22464 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
22465 PowerPC systems, since those instructions do not work when the
22466 processor is in little-endian mode. The exceptions are PPC740 and
22467 PPC750 which permit these instructions in little-endian mode.
22472 @opindex mno-string
22473 Generate code that uses (does not use) the load string instructions
22474 and the store string word instructions to save multiple registers and
22475 do small block moves. These instructions are generated by default on
22476 POWER systems, and not generated on PowerPC systems. Do not use
22477 @option{-mstring} on little-endian PowerPC systems, since those
22478 instructions do not work when the processor is in little-endian mode.
22479 The exceptions are PPC740 and PPC750 which permit these instructions
22480 in little-endian mode.
22485 @opindex mno-update
22486 Generate code that uses (does not use) the load or store instructions
22487 that update the base register to the address of the calculated memory
22488 location. These instructions are generated by default. If you use
22489 @option{-mno-update}, there is a small window between the time that the
22490 stack pointer is updated and the address of the previous frame is
22491 stored, which means code that walks the stack frame across interrupts or
22492 signals may get corrupted data.
22494 @item -mavoid-indexed-addresses
22495 @itemx -mno-avoid-indexed-addresses
22496 @opindex mavoid-indexed-addresses
22497 @opindex mno-avoid-indexed-addresses
22498 Generate code that tries to avoid (not avoid) the use of indexed load
22499 or store instructions. These instructions can incur a performance
22500 penalty on Power6 processors in certain situations, such as when
22501 stepping through large arrays that cross a 16M boundary. This option
22502 is enabled by default when targeting Power6 and disabled otherwise.
22505 @itemx -mno-fused-madd
22506 @opindex mfused-madd
22507 @opindex mno-fused-madd
22508 Generate code that uses (does not use) the floating-point multiply and
22509 accumulate instructions. These instructions are generated by default
22510 if hardware floating point is used. The machine-dependent
22511 @option{-mfused-madd} option is now mapped to the machine-independent
22512 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22513 mapped to @option{-ffp-contract=off}.
22519 Generate code that uses (does not use) the half-word multiply and
22520 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
22521 These instructions are generated by default when targeting those
22528 Generate code that uses (does not use) the string-search @samp{dlmzb}
22529 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
22530 generated by default when targeting those processors.
22532 @item -mno-bit-align
22534 @opindex mno-bit-align
22535 @opindex mbit-align
22536 On System V.4 and embedded PowerPC systems do not (do) force structures
22537 and unions that contain bit-fields to be aligned to the base type of the
22540 For example, by default a structure containing nothing but 8
22541 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
22542 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
22543 the structure is aligned to a 1-byte boundary and is 1 byte in
22546 @item -mno-strict-align
22547 @itemx -mstrict-align
22548 @opindex mno-strict-align
22549 @opindex mstrict-align
22550 On System V.4 and embedded PowerPC systems do not (do) assume that
22551 unaligned memory references are handled by the system.
22553 @item -mrelocatable
22554 @itemx -mno-relocatable
22555 @opindex mrelocatable
22556 @opindex mno-relocatable
22557 Generate code that allows (does not allow) a static executable to be
22558 relocated to a different address at run time. A simple embedded
22559 PowerPC system loader should relocate the entire contents of
22560 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
22561 a table of 32-bit addresses generated by this option. For this to
22562 work, all objects linked together must be compiled with
22563 @option{-mrelocatable} or @option{-mrelocatable-lib}.
22564 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
22566 @item -mrelocatable-lib
22567 @itemx -mno-relocatable-lib
22568 @opindex mrelocatable-lib
22569 @opindex mno-relocatable-lib
22570 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
22571 @code{.fixup} section to allow static executables to be relocated at
22572 run time, but @option{-mrelocatable-lib} does not use the smaller stack
22573 alignment of @option{-mrelocatable}. Objects compiled with
22574 @option{-mrelocatable-lib} may be linked with objects compiled with
22575 any combination of the @option{-mrelocatable} options.
22581 On System V.4 and embedded PowerPC systems do not (do) assume that
22582 register 2 contains a pointer to a global area pointing to the addresses
22583 used in the program.
22586 @itemx -mlittle-endian
22588 @opindex mlittle-endian
22589 On System V.4 and embedded PowerPC systems compile code for the
22590 processor in little-endian mode. The @option{-mlittle-endian} option is
22591 the same as @option{-mlittle}.
22594 @itemx -mbig-endian
22596 @opindex mbig-endian
22597 On System V.4 and embedded PowerPC systems compile code for the
22598 processor in big-endian mode. The @option{-mbig-endian} option is
22599 the same as @option{-mbig}.
22601 @item -mdynamic-no-pic
22602 @opindex mdynamic-no-pic
22603 On Darwin and Mac OS X systems, compile code so that it is not
22604 relocatable, but that its external references are relocatable. The
22605 resulting code is suitable for applications, but not shared
22608 @item -msingle-pic-base
22609 @opindex msingle-pic-base
22610 Treat the register used for PIC addressing as read-only, rather than
22611 loading it in the prologue for each function. The runtime system is
22612 responsible for initializing this register with an appropriate value
22613 before execution begins.
22615 @item -mprioritize-restricted-insns=@var{priority}
22616 @opindex mprioritize-restricted-insns
22617 This option controls the priority that is assigned to
22618 dispatch-slot restricted instructions during the second scheduling
22619 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
22620 or @samp{2} to assign no, highest, or second-highest (respectively)
22621 priority to dispatch-slot restricted
22624 @item -msched-costly-dep=@var{dependence_type}
22625 @opindex msched-costly-dep
22626 This option controls which dependences are considered costly
22627 by the target during instruction scheduling. The argument
22628 @var{dependence_type} takes one of the following values:
22632 No dependence is costly.
22635 All dependences are costly.
22637 @item @samp{true_store_to_load}
22638 A true dependence from store to load is costly.
22640 @item @samp{store_to_load}
22641 Any dependence from store to load is costly.
22644 Any dependence for which the latency is greater than or equal to
22645 @var{number} is costly.
22648 @item -minsert-sched-nops=@var{scheme}
22649 @opindex minsert-sched-nops
22650 This option controls which NOP insertion scheme is used during
22651 the second scheduling pass. The argument @var{scheme} takes one of the
22659 Pad with NOPs any dispatch group that has vacant issue slots,
22660 according to the scheduler's grouping.
22662 @item @samp{regroup_exact}
22663 Insert NOPs to force costly dependent insns into
22664 separate groups. Insert exactly as many NOPs as needed to force an insn
22665 to a new group, according to the estimated processor grouping.
22668 Insert NOPs to force costly dependent insns into
22669 separate groups. Insert @var{number} NOPs to force an insn to a new group.
22673 @opindex mcall-sysv
22674 On System V.4 and embedded PowerPC systems compile code using calling
22675 conventions that adhere to the March 1995 draft of the System V
22676 Application Binary Interface, PowerPC processor supplement. This is the
22677 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
22679 @item -mcall-sysv-eabi
22681 @opindex mcall-sysv-eabi
22682 @opindex mcall-eabi
22683 Specify both @option{-mcall-sysv} and @option{-meabi} options.
22685 @item -mcall-sysv-noeabi
22686 @opindex mcall-sysv-noeabi
22687 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
22689 @item -mcall-aixdesc
22691 On System V.4 and embedded PowerPC systems compile code for the AIX
22695 @opindex mcall-linux
22696 On System V.4 and embedded PowerPC systems compile code for the
22697 Linux-based GNU system.
22699 @item -mcall-freebsd
22700 @opindex mcall-freebsd
22701 On System V.4 and embedded PowerPC systems compile code for the
22702 FreeBSD operating system.
22704 @item -mcall-netbsd
22705 @opindex mcall-netbsd
22706 On System V.4 and embedded PowerPC systems compile code for the
22707 NetBSD operating system.
22709 @item -mcall-openbsd
22710 @opindex mcall-netbsd
22711 On System V.4 and embedded PowerPC systems compile code for the
22712 OpenBSD operating system.
22714 @item -maix-struct-return
22715 @opindex maix-struct-return
22716 Return all structures in memory (as specified by the AIX ABI)@.
22718 @item -msvr4-struct-return
22719 @opindex msvr4-struct-return
22720 Return structures smaller than 8 bytes in registers (as specified by the
22723 @item -mabi=@var{abi-type}
22725 Extend the current ABI with a particular extension, or remove such extension.
22726 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
22727 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
22728 @samp{elfv1}, @samp{elfv2}@.
22732 Extend the current ABI with SPE ABI extensions. This does not change
22733 the default ABI, instead it adds the SPE ABI extensions to the current
22737 @opindex mabi=no-spe
22738 Disable Book-E SPE ABI extensions for the current ABI@.
22740 @item -mabi=ibmlongdouble
22741 @opindex mabi=ibmlongdouble
22742 Change the current ABI to use IBM extended-precision long double.
22743 This is not likely to work if your system defaults to using IEEE
22744 extended-precision long double. If you change the long double type
22745 from IEEE extended-precision, the compiler will issue a warning unless
22746 you use the @option{-Wno-psabi} option.
22748 @item -mabi=ieeelongdouble
22749 @opindex mabi=ieeelongdouble
22750 Change the current ABI to use IEEE extended-precision long double.
22751 This is not likely to work if your system defaults to using IBM
22752 extended-precision long double. If you change the long double type
22753 from IBM extended-precision, the compiler will issue a warning unless
22754 you use the @option{-Wno-psabi} option.
22757 @opindex mabi=elfv1
22758 Change the current ABI to use the ELFv1 ABI.
22759 This is the default ABI for big-endian PowerPC 64-bit Linux.
22760 Overriding the default ABI requires special system support and is
22761 likely to fail in spectacular ways.
22764 @opindex mabi=elfv2
22765 Change the current ABI to use the ELFv2 ABI.
22766 This is the default ABI for little-endian PowerPC 64-bit Linux.
22767 Overriding the default ABI requires special system support and is
22768 likely to fail in spectacular ways.
22770 @item -mgnu-attribute
22771 @itemx -mno-gnu-attribute
22772 @opindex mgnu-attribute
22773 @opindex mno-gnu-attribute
22774 Emit .gnu_attribute assembly directives to set tag/value pairs in a
22775 .gnu.attributes section that specify ABI variations in function
22776 parameters or return values.
22779 @itemx -mno-prototype
22780 @opindex mprototype
22781 @opindex mno-prototype
22782 On System V.4 and embedded PowerPC systems assume that all calls to
22783 variable argument functions are properly prototyped. Otherwise, the
22784 compiler must insert an instruction before every non-prototyped call to
22785 set or clear bit 6 of the condition code register (@code{CR}) to
22786 indicate whether floating-point values are passed in the floating-point
22787 registers in case the function takes variable arguments. With
22788 @option{-mprototype}, only calls to prototyped variable argument functions
22789 set or clear the bit.
22793 On embedded PowerPC systems, assume that the startup module is called
22794 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
22795 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
22800 On embedded PowerPC systems, assume that the startup module is called
22801 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
22806 On embedded PowerPC systems, assume that the startup module is called
22807 @file{crt0.o} and the standard C libraries are @file{libads.a} and
22810 @item -myellowknife
22811 @opindex myellowknife
22812 On embedded PowerPC systems, assume that the startup module is called
22813 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
22818 On System V.4 and embedded PowerPC systems, specify that you are
22819 compiling for a VxWorks system.
22823 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
22824 header to indicate that @samp{eabi} extended relocations are used.
22830 On System V.4 and embedded PowerPC systems do (do not) adhere to the
22831 Embedded Applications Binary Interface (EABI), which is a set of
22832 modifications to the System V.4 specifications. Selecting @option{-meabi}
22833 means that the stack is aligned to an 8-byte boundary, a function
22834 @code{__eabi} is called from @code{main} to set up the EABI
22835 environment, and the @option{-msdata} option can use both @code{r2} and
22836 @code{r13} to point to two separate small data areas. Selecting
22837 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
22838 no EABI initialization function is called from @code{main}, and the
22839 @option{-msdata} option only uses @code{r13} to point to a single
22840 small data area. The @option{-meabi} option is on by default if you
22841 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
22844 @opindex msdata=eabi
22845 On System V.4 and embedded PowerPC systems, put small initialized
22846 @code{const} global and static data in the @code{.sdata2} section, which
22847 is pointed to by register @code{r2}. Put small initialized
22848 non-@code{const} global and static data in the @code{.sdata} section,
22849 which is pointed to by register @code{r13}. Put small uninitialized
22850 global and static data in the @code{.sbss} section, which is adjacent to
22851 the @code{.sdata} section. The @option{-msdata=eabi} option is
22852 incompatible with the @option{-mrelocatable} option. The
22853 @option{-msdata=eabi} option also sets the @option{-memb} option.
22856 @opindex msdata=sysv
22857 On System V.4 and embedded PowerPC systems, put small global and static
22858 data in the @code{.sdata} section, which is pointed to by register
22859 @code{r13}. Put small uninitialized global and static data in the
22860 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
22861 The @option{-msdata=sysv} option is incompatible with the
22862 @option{-mrelocatable} option.
22864 @item -msdata=default
22866 @opindex msdata=default
22868 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
22869 compile code the same as @option{-msdata=eabi}, otherwise compile code the
22870 same as @option{-msdata=sysv}.
22873 @opindex msdata=data
22874 On System V.4 and embedded PowerPC systems, put small global
22875 data in the @code{.sdata} section. Put small uninitialized global
22876 data in the @code{.sbss} section. Do not use register @code{r13}
22877 to address small data however. This is the default behavior unless
22878 other @option{-msdata} options are used.
22882 @opindex msdata=none
22884 On embedded PowerPC systems, put all initialized global and static data
22885 in the @code{.data} section, and all uninitialized data in the
22886 @code{.bss} section.
22888 @item -mblock-move-inline-limit=@var{num}
22889 @opindex mblock-move-inline-limit
22890 Inline all block moves (such as calls to @code{memcpy} or structure
22891 copies) less than or equal to @var{num} bytes. The minimum value for
22892 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
22893 targets. The default value is target-specific.
22897 @cindex smaller data references (PowerPC)
22898 @cindex .sdata/.sdata2 references (PowerPC)
22899 On embedded PowerPC systems, put global and static items less than or
22900 equal to @var{num} bytes into the small data or BSS sections instead of
22901 the normal data or BSS section. By default, @var{num} is 8. The
22902 @option{-G @var{num}} switch is also passed to the linker.
22903 All modules should be compiled with the same @option{-G @var{num}} value.
22906 @itemx -mno-regnames
22908 @opindex mno-regnames
22909 On System V.4 and embedded PowerPC systems do (do not) emit register
22910 names in the assembly language output using symbolic forms.
22913 @itemx -mno-longcall
22915 @opindex mno-longcall
22916 By default assume that all calls are far away so that a longer and more
22917 expensive calling sequence is required. This is required for calls
22918 farther than 32 megabytes (33,554,432 bytes) from the current location.
22919 A short call is generated if the compiler knows
22920 the call cannot be that far away. This setting can be overridden by
22921 the @code{shortcall} function attribute, or by @code{#pragma
22924 Some linkers are capable of detecting out-of-range calls and generating
22925 glue code on the fly. On these systems, long calls are unnecessary and
22926 generate slower code. As of this writing, the AIX linker can do this,
22927 as can the GNU linker for PowerPC/64. It is planned to add this feature
22928 to the GNU linker for 32-bit PowerPC systems as well.
22930 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
22931 callee, L42}, plus a @dfn{branch island} (glue code). The two target
22932 addresses represent the callee and the branch island. The
22933 Darwin/PPC linker prefers the first address and generates a @code{bl
22934 callee} if the PPC @code{bl} instruction reaches the callee directly;
22935 otherwise, the linker generates @code{bl L42} to call the branch
22936 island. The branch island is appended to the body of the
22937 calling function; it computes the full 32-bit address of the callee
22940 On Mach-O (Darwin) systems, this option directs the compiler emit to
22941 the glue for every direct call, and the Darwin linker decides whether
22942 to use or discard it.
22944 In the future, GCC may ignore all longcall specifications
22945 when the linker is known to generate glue.
22947 @item -mtls-markers
22948 @itemx -mno-tls-markers
22949 @opindex mtls-markers
22950 @opindex mno-tls-markers
22951 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22952 specifying the function argument. The relocation allows the linker to
22953 reliably associate function call with argument setup instructions for
22954 TLS optimization, which in turn allows GCC to better schedule the
22960 This option enables use of the reciprocal estimate and
22961 reciprocal square root estimate instructions with additional
22962 Newton-Raphson steps to increase precision instead of doing a divide or
22963 square root and divide for floating-point arguments. You should use
22964 the @option{-ffast-math} option when using @option{-mrecip} (or at
22965 least @option{-funsafe-math-optimizations},
22966 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22967 @option{-fno-trapping-math}). Note that while the throughput of the
22968 sequence is generally higher than the throughput of the non-reciprocal
22969 instruction, the precision of the sequence can be decreased by up to 2
22970 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
22973 @item -mrecip=@var{opt}
22974 @opindex mrecip=opt
22975 This option controls which reciprocal estimate instructions
22976 may be used. @var{opt} is a comma-separated list of options, which may
22977 be preceded by a @code{!} to invert the option:
22982 Enable all estimate instructions.
22985 Enable the default instructions, equivalent to @option{-mrecip}.
22988 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22991 Enable the reciprocal approximation instructions for both
22992 single and double precision.
22995 Enable the single-precision reciprocal approximation instructions.
22998 Enable the double-precision reciprocal approximation instructions.
23001 Enable the reciprocal square root approximation instructions for both
23002 single and double precision.
23005 Enable the single-precision reciprocal square root approximation instructions.
23008 Enable the double-precision reciprocal square root approximation instructions.
23012 So, for example, @option{-mrecip=all,!rsqrtd} enables
23013 all of the reciprocal estimate instructions, except for the
23014 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
23015 which handle the double-precision reciprocal square root calculations.
23017 @item -mrecip-precision
23018 @itemx -mno-recip-precision
23019 @opindex mrecip-precision
23020 Assume (do not assume) that the reciprocal estimate instructions
23021 provide higher-precision estimates than is mandated by the PowerPC
23022 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
23023 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
23024 The double-precision square root estimate instructions are not generated by
23025 default on low-precision machines, since they do not provide an
23026 estimate that converges after three steps.
23028 @item -mveclibabi=@var{type}
23029 @opindex mveclibabi
23030 Specifies the ABI type to use for vectorizing intrinsics using an
23031 external library. The only type supported at present is @samp{mass},
23032 which specifies to use IBM's Mathematical Acceleration Subsystem
23033 (MASS) libraries for vectorizing intrinsics using external libraries.
23034 GCC currently emits calls to @code{acosd2}, @code{acosf4},
23035 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
23036 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
23037 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
23038 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
23039 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
23040 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
23041 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
23042 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
23043 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
23044 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
23045 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
23046 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
23047 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
23048 for power7. Both @option{-ftree-vectorize} and
23049 @option{-funsafe-math-optimizations} must also be enabled. The MASS
23050 libraries must be specified at link time.
23055 Generate (do not generate) the @code{friz} instruction when the
23056 @option{-funsafe-math-optimizations} option is used to optimize
23057 rounding of floating-point values to 64-bit integer and back to floating
23058 point. The @code{friz} instruction does not return the same value if
23059 the floating-point number is too large to fit in an integer.
23061 @item -mpointers-to-nested-functions
23062 @itemx -mno-pointers-to-nested-functions
23063 @opindex mpointers-to-nested-functions
23064 Generate (do not generate) code to load up the static chain register
23065 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
23066 systems where a function pointer points to a 3-word descriptor giving
23067 the function address, TOC value to be loaded in register @code{r2}, and
23068 static chain value to be loaded in register @code{r11}. The
23069 @option{-mpointers-to-nested-functions} is on by default. You cannot
23070 call through pointers to nested functions or pointers
23071 to functions compiled in other languages that use the static chain if
23072 you use @option{-mno-pointers-to-nested-functions}.
23074 @item -msave-toc-indirect
23075 @itemx -mno-save-toc-indirect
23076 @opindex msave-toc-indirect
23077 Generate (do not generate) code to save the TOC value in the reserved
23078 stack location in the function prologue if the function calls through
23079 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
23080 saved in the prologue, it is saved just before the call through the
23081 pointer. The @option{-mno-save-toc-indirect} option is the default.
23083 @item -mcompat-align-parm
23084 @itemx -mno-compat-align-parm
23085 @opindex mcompat-align-parm
23086 Generate (do not generate) code to pass structure parameters with a
23087 maximum alignment of 64 bits, for compatibility with older versions
23090 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
23091 structure parameter on a 128-bit boundary when that structure contained
23092 a member requiring 128-bit alignment. This is corrected in more
23093 recent versions of GCC. This option may be used to generate code
23094 that is compatible with functions compiled with older versions of
23097 The @option{-mno-compat-align-parm} option is the default.
23099 @item -mstack-protector-guard=@var{guard}
23100 @itemx -mstack-protector-guard-reg=@var{reg}
23101 @itemx -mstack-protector-guard-offset=@var{offset}
23102 @itemx -mstack-protector-guard-symbol=@var{symbol}
23103 @opindex mstack-protector-guard
23104 @opindex mstack-protector-guard-reg
23105 @opindex mstack-protector-guard-offset
23106 @opindex mstack-protector-guard-symbol
23107 Generate stack protection code using canary at @var{guard}. Supported
23108 locations are @samp{global} for global canary or @samp{tls} for per-thread
23109 canary in the TLS block (the default with GNU libc version 2.4 or later).
23111 With the latter choice the options
23112 @option{-mstack-protector-guard-reg=@var{reg}} and
23113 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
23114 which register to use as base register for reading the canary, and from what
23115 offset from that base register. The default for those is as specified in the
23116 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
23117 the offset with a symbol reference to a canary in the TLS block.
23121 @subsection RX Options
23124 These command-line options are defined for RX targets:
23127 @item -m64bit-doubles
23128 @itemx -m32bit-doubles
23129 @opindex m64bit-doubles
23130 @opindex m32bit-doubles
23131 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23132 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23133 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
23134 works on 32-bit values, which is why the default is
23135 @option{-m32bit-doubles}.
23141 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
23142 floating-point hardware. The default is enabled for the RX600
23143 series and disabled for the RX200 series.
23145 Floating-point instructions are only generated for 32-bit floating-point
23146 values, however, so the FPU hardware is not used for doubles if the
23147 @option{-m64bit-doubles} option is used.
23149 @emph{Note} If the @option{-fpu} option is enabled then
23150 @option{-funsafe-math-optimizations} is also enabled automatically.
23151 This is because the RX FPU instructions are themselves unsafe.
23153 @item -mcpu=@var{name}
23155 Selects the type of RX CPU to be targeted. Currently three types are
23156 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
23157 the specific @samp{RX610} CPU. The default is @samp{RX600}.
23159 The only difference between @samp{RX600} and @samp{RX610} is that the
23160 @samp{RX610} does not support the @code{MVTIPL} instruction.
23162 The @samp{RX200} series does not have a hardware floating-point unit
23163 and so @option{-nofpu} is enabled by default when this type is
23166 @item -mbig-endian-data
23167 @itemx -mlittle-endian-data
23168 @opindex mbig-endian-data
23169 @opindex mlittle-endian-data
23170 Store data (but not code) in the big-endian format. The default is
23171 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
23174 @item -msmall-data-limit=@var{N}
23175 @opindex msmall-data-limit
23176 Specifies the maximum size in bytes of global and static variables
23177 which can be placed into the small data area. Using the small data
23178 area can lead to smaller and faster code, but the size of area is
23179 limited and it is up to the programmer to ensure that the area does
23180 not overflow. Also when the small data area is used one of the RX's
23181 registers (usually @code{r13}) is reserved for use pointing to this
23182 area, so it is no longer available for use by the compiler. This
23183 could result in slower and/or larger code if variables are pushed onto
23184 the stack instead of being held in this register.
23186 Note, common variables (variables that have not been initialized) and
23187 constants are not placed into the small data area as they are assigned
23188 to other sections in the output executable.
23190 The default value is zero, which disables this feature. Note, this
23191 feature is not enabled by default with higher optimization levels
23192 (@option{-O2} etc) because of the potentially detrimental effects of
23193 reserving a register. It is up to the programmer to experiment and
23194 discover whether this feature is of benefit to their program. See the
23195 description of the @option{-mpid} option for a description of how the
23196 actual register to hold the small data area pointer is chosen.
23202 Use the simulator runtime. The default is to use the libgloss
23203 board-specific runtime.
23205 @item -mas100-syntax
23206 @itemx -mno-as100-syntax
23207 @opindex mas100-syntax
23208 @opindex mno-as100-syntax
23209 When generating assembler output use a syntax that is compatible with
23210 Renesas's AS100 assembler. This syntax can also be handled by the GAS
23211 assembler, but it has some restrictions so it is not generated by default.
23213 @item -mmax-constant-size=@var{N}
23214 @opindex mmax-constant-size
23215 Specifies the maximum size, in bytes, of a constant that can be used as
23216 an operand in a RX instruction. Although the RX instruction set does
23217 allow constants of up to 4 bytes in length to be used in instructions,
23218 a longer value equates to a longer instruction. Thus in some
23219 circumstances it can be beneficial to restrict the size of constants
23220 that are used in instructions. Constants that are too big are instead
23221 placed into a constant pool and referenced via register indirection.
23223 The value @var{N} can be between 0 and 4. A value of 0 (the default)
23224 or 4 means that constants of any size are allowed.
23228 Enable linker relaxation. Linker relaxation is a process whereby the
23229 linker attempts to reduce the size of a program by finding shorter
23230 versions of various instructions. Disabled by default.
23232 @item -mint-register=@var{N}
23233 @opindex mint-register
23234 Specify the number of registers to reserve for fast interrupt handler
23235 functions. The value @var{N} can be between 0 and 4. A value of 1
23236 means that register @code{r13} is reserved for the exclusive use
23237 of fast interrupt handlers. A value of 2 reserves @code{r13} and
23238 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
23239 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
23240 A value of 0, the default, does not reserve any registers.
23242 @item -msave-acc-in-interrupts
23243 @opindex msave-acc-in-interrupts
23244 Specifies that interrupt handler functions should preserve the
23245 accumulator register. This is only necessary if normal code might use
23246 the accumulator register, for example because it performs 64-bit
23247 multiplications. The default is to ignore the accumulator as this
23248 makes the interrupt handlers faster.
23254 Enables the generation of position independent data. When enabled any
23255 access to constant data is done via an offset from a base address
23256 held in a register. This allows the location of constant data to be
23257 determined at run time without requiring the executable to be
23258 relocated, which is a benefit to embedded applications with tight
23259 memory constraints. Data that can be modified is not affected by this
23262 Note, using this feature reserves a register, usually @code{r13}, for
23263 the constant data base address. This can result in slower and/or
23264 larger code, especially in complicated functions.
23266 The actual register chosen to hold the constant data base address
23267 depends upon whether the @option{-msmall-data-limit} and/or the
23268 @option{-mint-register} command-line options are enabled. Starting
23269 with register @code{r13} and proceeding downwards, registers are
23270 allocated first to satisfy the requirements of @option{-mint-register},
23271 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
23272 is possible for the small data area register to be @code{r8} if both
23273 @option{-mint-register=4} and @option{-mpid} are specified on the
23276 By default this feature is not enabled. The default can be restored
23277 via the @option{-mno-pid} command-line option.
23279 @item -mno-warn-multiple-fast-interrupts
23280 @itemx -mwarn-multiple-fast-interrupts
23281 @opindex mno-warn-multiple-fast-interrupts
23282 @opindex mwarn-multiple-fast-interrupts
23283 Prevents GCC from issuing a warning message if it finds more than one
23284 fast interrupt handler when it is compiling a file. The default is to
23285 issue a warning for each extra fast interrupt handler found, as the RX
23286 only supports one such interrupt.
23288 @item -mallow-string-insns
23289 @itemx -mno-allow-string-insns
23290 @opindex mallow-string-insns
23291 @opindex mno-allow-string-insns
23292 Enables or disables the use of the string manipulation instructions
23293 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
23294 @code{SWHILE} and also the @code{RMPA} instruction. These
23295 instructions may prefetch data, which is not safe to do if accessing
23296 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
23297 for more information).
23299 The default is to allow these instructions, but it is not possible for
23300 GCC to reliably detect all circumstances where a string instruction
23301 might be used to access an I/O register, so their use cannot be
23302 disabled automatically. Instead it is reliant upon the programmer to
23303 use the @option{-mno-allow-string-insns} option if their program
23304 accesses I/O space.
23306 When the instructions are enabled GCC defines the C preprocessor
23307 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
23308 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
23314 Use only (or not only) @code{JSR} instructions to access functions.
23315 This option can be used when code size exceeds the range of @code{BSR}
23316 instructions. Note that @option{-mno-jsr} does not mean to not use
23317 @code{JSR} but instead means that any type of branch may be used.
23320 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
23321 has special significance to the RX port when used with the
23322 @code{interrupt} function attribute. This attribute indicates a
23323 function intended to process fast interrupts. GCC ensures
23324 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
23325 and/or @code{r13} and only provided that the normal use of the
23326 corresponding registers have been restricted via the
23327 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
23330 @node S/390 and zSeries Options
23331 @subsection S/390 and zSeries Options
23332 @cindex S/390 and zSeries Options
23334 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
23338 @itemx -msoft-float
23339 @opindex mhard-float
23340 @opindex msoft-float
23341 Use (do not use) the hardware floating-point instructions and registers
23342 for floating-point operations. When @option{-msoft-float} is specified,
23343 functions in @file{libgcc.a} are used to perform floating-point
23344 operations. When @option{-mhard-float} is specified, the compiler
23345 generates IEEE floating-point instructions. This is the default.
23348 @itemx -mno-hard-dfp
23350 @opindex mno-hard-dfp
23351 Use (do not use) the hardware decimal-floating-point instructions for
23352 decimal-floating-point operations. When @option{-mno-hard-dfp} is
23353 specified, functions in @file{libgcc.a} are used to perform
23354 decimal-floating-point operations. When @option{-mhard-dfp} is
23355 specified, the compiler generates decimal-floating-point hardware
23356 instructions. This is the default for @option{-march=z9-ec} or higher.
23358 @item -mlong-double-64
23359 @itemx -mlong-double-128
23360 @opindex mlong-double-64
23361 @opindex mlong-double-128
23362 These switches control the size of @code{long double} type. A size
23363 of 64 bits makes the @code{long double} type equivalent to the @code{double}
23364 type. This is the default.
23367 @itemx -mno-backchain
23368 @opindex mbackchain
23369 @opindex mno-backchain
23370 Store (do not store) the address of the caller's frame as backchain pointer
23371 into the callee's stack frame.
23372 A backchain may be needed to allow debugging using tools that do not understand
23373 DWARF call frame information.
23374 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
23375 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
23376 the backchain is placed into the topmost word of the 96/160 byte register
23379 In general, code compiled with @option{-mbackchain} is call-compatible with
23380 code compiled with @option{-mmo-backchain}; however, use of the backchain
23381 for debugging purposes usually requires that the whole binary is built with
23382 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
23383 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23384 to build a linux kernel use @option{-msoft-float}.
23386 The default is to not maintain the backchain.
23388 @item -mpacked-stack
23389 @itemx -mno-packed-stack
23390 @opindex mpacked-stack
23391 @opindex mno-packed-stack
23392 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
23393 specified, the compiler uses the all fields of the 96/160 byte register save
23394 area only for their default purpose; unused fields still take up stack space.
23395 When @option{-mpacked-stack} is specified, register save slots are densely
23396 packed at the top of the register save area; unused space is reused for other
23397 purposes, allowing for more efficient use of the available stack space.
23398 However, when @option{-mbackchain} is also in effect, the topmost word of
23399 the save area is always used to store the backchain, and the return address
23400 register is always saved two words below the backchain.
23402 As long as the stack frame backchain is not used, code generated with
23403 @option{-mpacked-stack} is call-compatible with code generated with
23404 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
23405 S/390 or zSeries generated code that uses the stack frame backchain at run
23406 time, not just for debugging purposes. Such code is not call-compatible
23407 with code compiled with @option{-mpacked-stack}. Also, note that the
23408 combination of @option{-mbackchain},
23409 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23410 to build a linux kernel use @option{-msoft-float}.
23412 The default is to not use the packed stack layout.
23415 @itemx -mno-small-exec
23416 @opindex msmall-exec
23417 @opindex mno-small-exec
23418 Generate (or do not generate) code using the @code{bras} instruction
23419 to do subroutine calls.
23420 This only works reliably if the total executable size does not
23421 exceed 64k. The default is to use the @code{basr} instruction instead,
23422 which does not have this limitation.
23428 When @option{-m31} is specified, generate code compliant to the
23429 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
23430 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
23431 particular to generate 64-bit instructions. For the @samp{s390}
23432 targets, the default is @option{-m31}, while the @samp{s390x}
23433 targets default to @option{-m64}.
23439 When @option{-mzarch} is specified, generate code using the
23440 instructions available on z/Architecture.
23441 When @option{-mesa} is specified, generate code using the
23442 instructions available on ESA/390. Note that @option{-mesa} is
23443 not possible with @option{-m64}.
23444 When generating code compliant to the GNU/Linux for S/390 ABI,
23445 the default is @option{-mesa}. When generating code compliant
23446 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
23452 The @option{-mhtm} option enables a set of builtins making use of
23453 instructions available with the transactional execution facility
23454 introduced with the IBM zEnterprise EC12 machine generation
23455 @ref{S/390 System z Built-in Functions}.
23456 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
23462 When @option{-mvx} is specified, generate code using the instructions
23463 available with the vector extension facility introduced with the IBM
23464 z13 machine generation.
23465 This option changes the ABI for some vector type values with regard to
23466 alignment and calling conventions. In case vector type values are
23467 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
23468 command will be added to mark the resulting binary with the ABI used.
23469 @option{-mvx} is enabled by default when using @option{-march=z13}.
23472 @itemx -mno-zvector
23474 @opindex mno-zvector
23475 The @option{-mzvector} option enables vector language extensions and
23476 builtins using instructions available with the vector extension
23477 facility introduced with the IBM z13 machine generation.
23478 This option adds support for @samp{vector} to be used as a keyword to
23479 define vector type variables and arguments. @samp{vector} is only
23480 available when GNU extensions are enabled. It will not be expanded
23481 when requesting strict standard compliance e.g. with @option{-std=c99}.
23482 In addition to the GCC low-level builtins @option{-mzvector} enables
23483 a set of builtins added for compatibility with AltiVec-style
23484 implementations like Power and Cell. In order to make use of these
23485 builtins the header file @file{vecintrin.h} needs to be included.
23486 @option{-mzvector} is disabled by default.
23492 Generate (or do not generate) code using the @code{mvcle} instruction
23493 to perform block moves. When @option{-mno-mvcle} is specified,
23494 use a @code{mvc} loop instead. This is the default unless optimizing for
23501 Print (or do not print) additional debug information when compiling.
23502 The default is to not print debug information.
23504 @item -march=@var{cpu-type}
23506 Generate code that runs on @var{cpu-type}, which is the name of a
23507 system representing a certain processor type. Possible values for
23508 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
23509 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
23510 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
23513 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
23514 @samp{g6} are deprecated and will be removed with future releases.
23516 Specifying @samp{native} as cpu type can be used to select the best
23517 architecture option for the host processor.
23518 @option{-march=native} has no effect if GCC does not recognize the
23521 @item -mtune=@var{cpu-type}
23523 Tune to @var{cpu-type} everything applicable about the generated code,
23524 except for the ABI and the set of available instructions.
23525 The list of @var{cpu-type} values is the same as for @option{-march}.
23526 The default is the value used for @option{-march}.
23529 @itemx -mno-tpf-trace
23530 @opindex mtpf-trace
23531 @opindex mno-tpf-trace
23532 Generate code that adds (does not add) in TPF OS specific branches to trace
23533 routines in the operating system. This option is off by default, even
23534 when compiling for the TPF OS@.
23537 @itemx -mno-fused-madd
23538 @opindex mfused-madd
23539 @opindex mno-fused-madd
23540 Generate code that uses (does not use) the floating-point multiply and
23541 accumulate instructions. These instructions are generated by default if
23542 hardware floating point is used.
23544 @item -mwarn-framesize=@var{framesize}
23545 @opindex mwarn-framesize
23546 Emit a warning if the current function exceeds the given frame size. Because
23547 this is a compile-time check it doesn't need to be a real problem when the program
23548 runs. It is intended to identify functions that most probably cause
23549 a stack overflow. It is useful to be used in an environment with limited stack
23550 size e.g.@: the linux kernel.
23552 @item -mwarn-dynamicstack
23553 @opindex mwarn-dynamicstack
23554 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
23555 arrays. This is generally a bad idea with a limited stack size.
23557 @item -mstack-guard=@var{stack-guard}
23558 @itemx -mstack-size=@var{stack-size}
23559 @opindex mstack-guard
23560 @opindex mstack-size
23561 If these options are provided the S/390 back end emits additional instructions in
23562 the function prologue that trigger a trap if the stack size is @var{stack-guard}
23563 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
23564 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
23565 the frame size of the compiled function is chosen.
23566 These options are intended to be used to help debugging stack overflow problems.
23567 The additionally emitted code causes only little overhead and hence can also be
23568 used in production-like systems without greater performance degradation. The given
23569 values have to be exact powers of 2 and @var{stack-size} has to be greater than
23570 @var{stack-guard} without exceeding 64k.
23571 In order to be efficient the extra code makes the assumption that the stack starts
23572 at an address aligned to the value given by @var{stack-size}.
23573 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
23575 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
23577 If the hotpatch option is enabled, a ``hot-patching'' function
23578 prologue is generated for all functions in the compilation unit.
23579 The funtion label is prepended with the given number of two-byte
23580 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
23581 the label, 2 * @var{post-halfwords} bytes are appended, using the
23582 largest NOP like instructions the architecture allows (maximum
23585 If both arguments are zero, hotpatching is disabled.
23587 This option can be overridden for individual functions with the
23588 @code{hotpatch} attribute.
23591 @node Score Options
23592 @subsection Score Options
23593 @cindex Score Options
23595 These options are defined for Score implementations:
23600 Compile code for big-endian mode. This is the default.
23604 Compile code for little-endian mode.
23608 Disable generation of @code{bcnz} instructions.
23612 Enable generation of unaligned load and store instructions.
23616 Enable the use of multiply-accumulate instructions. Disabled by default.
23620 Specify the SCORE5 as the target architecture.
23624 Specify the SCORE5U of the target architecture.
23628 Specify the SCORE7 as the target architecture. This is the default.
23632 Specify the SCORE7D as the target architecture.
23636 @subsection SH Options
23638 These @samp{-m} options are defined for the SH implementations:
23643 Generate code for the SH1.
23647 Generate code for the SH2.
23650 Generate code for the SH2e.
23654 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
23655 that the floating-point unit is not used.
23657 @item -m2a-single-only
23658 @opindex m2a-single-only
23659 Generate code for the SH2a-FPU, in such a way that no double-precision
23660 floating-point operations are used.
23663 @opindex m2a-single
23664 Generate code for the SH2a-FPU assuming the floating-point unit is in
23665 single-precision mode by default.
23669 Generate code for the SH2a-FPU assuming the floating-point unit is in
23670 double-precision mode by default.
23674 Generate code for the SH3.
23678 Generate code for the SH3e.
23682 Generate code for the SH4 without a floating-point unit.
23684 @item -m4-single-only
23685 @opindex m4-single-only
23686 Generate code for the SH4 with a floating-point unit that only
23687 supports single-precision arithmetic.
23691 Generate code for the SH4 assuming the floating-point unit is in
23692 single-precision mode by default.
23696 Generate code for the SH4.
23700 Generate code for SH4-100.
23702 @item -m4-100-nofpu
23703 @opindex m4-100-nofpu
23704 Generate code for SH4-100 in such a way that the
23705 floating-point unit is not used.
23707 @item -m4-100-single
23708 @opindex m4-100-single
23709 Generate code for SH4-100 assuming the floating-point unit is in
23710 single-precision mode by default.
23712 @item -m4-100-single-only
23713 @opindex m4-100-single-only
23714 Generate code for SH4-100 in such a way that no double-precision
23715 floating-point operations are used.
23719 Generate code for SH4-200.
23721 @item -m4-200-nofpu
23722 @opindex m4-200-nofpu
23723 Generate code for SH4-200 without in such a way that the
23724 floating-point unit is not used.
23726 @item -m4-200-single
23727 @opindex m4-200-single
23728 Generate code for SH4-200 assuming the floating-point unit is in
23729 single-precision mode by default.
23731 @item -m4-200-single-only
23732 @opindex m4-200-single-only
23733 Generate code for SH4-200 in such a way that no double-precision
23734 floating-point operations are used.
23738 Generate code for SH4-300.
23740 @item -m4-300-nofpu
23741 @opindex m4-300-nofpu
23742 Generate code for SH4-300 without in such a way that the
23743 floating-point unit is not used.
23745 @item -m4-300-single
23746 @opindex m4-300-single
23747 Generate code for SH4-300 in such a way that no double-precision
23748 floating-point operations are used.
23750 @item -m4-300-single-only
23751 @opindex m4-300-single-only
23752 Generate code for SH4-300 in such a way that no double-precision
23753 floating-point operations are used.
23757 Generate code for SH4-340 (no MMU, no FPU).
23761 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
23766 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
23767 floating-point unit is not used.
23769 @item -m4a-single-only
23770 @opindex m4a-single-only
23771 Generate code for the SH4a, in such a way that no double-precision
23772 floating-point operations are used.
23775 @opindex m4a-single
23776 Generate code for the SH4a assuming the floating-point unit is in
23777 single-precision mode by default.
23781 Generate code for the SH4a.
23785 Same as @option{-m4a-nofpu}, except that it implicitly passes
23786 @option{-dsp} to the assembler. GCC doesn't generate any DSP
23787 instructions at the moment.
23791 Compile code for the processor in big-endian mode.
23795 Compile code for the processor in little-endian mode.
23799 Align doubles at 64-bit boundaries. Note that this changes the calling
23800 conventions, and thus some functions from the standard C library do
23801 not work unless you recompile it first with @option{-mdalign}.
23805 Shorten some address references at link time, when possible; uses the
23806 linker option @option{-relax}.
23810 Use 32-bit offsets in @code{switch} tables. The default is to use
23815 Enable the use of bit manipulation instructions on SH2A.
23819 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
23820 alignment constraints.
23824 Comply with the calling conventions defined by Renesas.
23827 @opindex mno-renesas
23828 Comply with the calling conventions defined for GCC before the Renesas
23829 conventions were available. This option is the default for all
23830 targets of the SH toolchain.
23833 @opindex mnomacsave
23834 Mark the @code{MAC} register as call-clobbered, even if
23835 @option{-mrenesas} is given.
23841 Control the IEEE compliance of floating-point comparisons, which affects the
23842 handling of cases where the result of a comparison is unordered. By default
23843 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
23844 enabled @option{-mno-ieee} is implicitly set, which results in faster
23845 floating-point greater-equal and less-equal comparisons. The implicit settings
23846 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
23848 @item -minline-ic_invalidate
23849 @opindex minline-ic_invalidate
23850 Inline code to invalidate instruction cache entries after setting up
23851 nested function trampolines.
23852 This option has no effect if @option{-musermode} is in effect and the selected
23853 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
23855 If the selected code generation option does not allow the use of the @code{icbi}
23856 instruction, and @option{-musermode} is not in effect, the inlined code
23857 manipulates the instruction cache address array directly with an associative
23858 write. This not only requires privileged mode at run time, but it also
23859 fails if the cache line had been mapped via the TLB and has become unmapped.
23863 Dump instruction size and location in the assembly code.
23866 @opindex mpadstruct
23867 This option is deprecated. It pads structures to multiple of 4 bytes,
23868 which is incompatible with the SH ABI@.
23870 @item -matomic-model=@var{model}
23871 @opindex matomic-model=@var{model}
23872 Sets the model of atomic operations and additional parameters as a comma
23873 separated list. For details on the atomic built-in functions see
23874 @ref{__atomic Builtins}. The following models and parameters are supported:
23879 Disable compiler generated atomic sequences and emit library calls for atomic
23880 operations. This is the default if the target is not @code{sh*-*-linux*}.
23883 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
23884 built-in functions. The generated atomic sequences require additional support
23885 from the interrupt/exception handling code of the system and are only suitable
23886 for SH3* and SH4* single-core systems. This option is enabled by default when
23887 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
23888 this option also partially utilizes the hardware atomic instructions
23889 @code{movli.l} and @code{movco.l} to create more efficient code, unless
23890 @samp{strict} is specified.
23893 Generate software atomic sequences that use a variable in the thread control
23894 block. This is a variation of the gUSA sequences which can also be used on
23895 SH1* and SH2* targets. The generated atomic sequences require additional
23896 support from the interrupt/exception handling code of the system and are only
23897 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
23898 parameter has to be specified as well.
23901 Generate software atomic sequences that temporarily disable interrupts by
23902 setting @code{SR.IMASK = 1111}. This model works only when the program runs
23903 in privileged mode and is only suitable for single-core systems. Additional
23904 support from the interrupt/exception handling code of the system is not
23905 required. This model is enabled by default when the target is
23906 @code{sh*-*-linux*} and SH1* or SH2*.
23909 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
23910 instructions only. This is only available on SH4A and is suitable for
23911 multi-core systems. Since the hardware instructions support only 32 bit atomic
23912 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
23913 Code compiled with this option is also compatible with other software
23914 atomic model interrupt/exception handling systems if executed on an SH4A
23915 system. Additional support from the interrupt/exception handling code of the
23916 system is not required for this model.
23919 This parameter specifies the offset in bytes of the variable in the thread
23920 control block structure that should be used by the generated atomic sequences
23921 when the @samp{soft-tcb} model has been selected. For other models this
23922 parameter is ignored. The specified value must be an integer multiple of four
23923 and in the range 0-1020.
23926 This parameter prevents mixed usage of multiple atomic models, even if they
23927 are compatible, and makes the compiler generate atomic sequences of the
23928 specified model only.
23934 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
23935 Notice that depending on the particular hardware and software configuration
23936 this can degrade overall performance due to the operand cache line flushes
23937 that are implied by the @code{tas.b} instruction. On multi-core SH4A
23938 processors the @code{tas.b} instruction must be used with caution since it
23939 can result in data corruption for certain cache configurations.
23942 @opindex mprefergot
23943 When generating position-independent code, emit function calls using
23944 the Global Offset Table instead of the Procedure Linkage Table.
23947 @itemx -mno-usermode
23949 @opindex mno-usermode
23950 Don't allow (allow) the compiler generating privileged mode code. Specifying
23951 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
23952 inlined code would not work in user mode. @option{-musermode} is the default
23953 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
23954 @option{-musermode} has no effect, since there is no user mode.
23956 @item -multcost=@var{number}
23957 @opindex multcost=@var{number}
23958 Set the cost to assume for a multiply insn.
23960 @item -mdiv=@var{strategy}
23961 @opindex mdiv=@var{strategy}
23962 Set the division strategy to be used for integer division operations.
23963 @var{strategy} can be one of:
23968 Calls a library function that uses the single-step division instruction
23969 @code{div1} to perform the operation. Division by zero calculates an
23970 unspecified result and does not trap. This is the default except for SH4,
23971 SH2A and SHcompact.
23974 Calls a library function that performs the operation in double precision
23975 floating point. Division by zero causes a floating-point exception. This is
23976 the default for SHcompact with FPU. Specifying this for targets that do not
23977 have a double precision FPU defaults to @code{call-div1}.
23980 Calls a library function that uses a lookup table for small divisors and
23981 the @code{div1} instruction with case distinction for larger divisors. Division
23982 by zero calculates an unspecified result and does not trap. This is the default
23983 for SH4. Specifying this for targets that do not have dynamic shift
23984 instructions defaults to @code{call-div1}.
23988 When a division strategy has not been specified the default strategy is
23989 selected based on the current target. For SH2A the default strategy is to
23990 use the @code{divs} and @code{divu} instructions instead of library function
23993 @item -maccumulate-outgoing-args
23994 @opindex maccumulate-outgoing-args
23995 Reserve space once for outgoing arguments in the function prologue rather
23996 than around each call. Generally beneficial for performance and size. Also
23997 needed for unwinding to avoid changing the stack frame around conditional code.
23999 @item -mdivsi3_libfunc=@var{name}
24000 @opindex mdivsi3_libfunc=@var{name}
24001 Set the name of the library function used for 32-bit signed division to
24003 This only affects the name used in the @samp{call} division strategies, and
24004 the compiler still expects the same sets of input/output/clobbered registers as
24005 if this option were not present.
24007 @item -mfixed-range=@var{register-range}
24008 @opindex mfixed-range
24009 Generate code treating the given register range as fixed registers.
24010 A fixed register is one that the register allocator can not use. This is
24011 useful when compiling kernel code. A register range is specified as
24012 two registers separated by a dash. Multiple register ranges can be
24013 specified separated by a comma.
24015 @item -mbranch-cost=@var{num}
24016 @opindex mbranch-cost=@var{num}
24017 Assume @var{num} to be the cost for a branch instruction. Higher numbers
24018 make the compiler try to generate more branch-free code if possible.
24019 If not specified the value is selected depending on the processor type that
24020 is being compiled for.
24023 @itemx -mno-zdcbranch
24024 @opindex mzdcbranch
24025 @opindex mno-zdcbranch
24026 Assume (do not assume) that zero displacement conditional branch instructions
24027 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
24028 compiler prefers zero displacement branch code sequences. This is
24029 enabled by default when generating code for SH4 and SH4A. It can be explicitly
24030 disabled by specifying @option{-mno-zdcbranch}.
24032 @item -mcbranch-force-delay-slot
24033 @opindex mcbranch-force-delay-slot
24034 Force the usage of delay slots for conditional branches, which stuffs the delay
24035 slot with a @code{nop} if a suitable instruction cannot be found. By default
24036 this option is disabled. It can be enabled to work around hardware bugs as
24037 found in the original SH7055.
24040 @itemx -mno-fused-madd
24041 @opindex mfused-madd
24042 @opindex mno-fused-madd
24043 Generate code that uses (does not use) the floating-point multiply and
24044 accumulate instructions. These instructions are generated by default
24045 if hardware floating point is used. The machine-dependent
24046 @option{-mfused-madd} option is now mapped to the machine-independent
24047 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
24048 mapped to @option{-ffp-contract=off}.
24054 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
24055 and cosine approximations. The option @option{-mfsca} must be used in
24056 combination with @option{-funsafe-math-optimizations}. It is enabled by default
24057 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
24058 approximations even if @option{-funsafe-math-optimizations} is in effect.
24064 Allow or disallow the compiler to emit the @code{fsrra} instruction for
24065 reciprocal square root approximations. The option @option{-mfsrra} must be used
24066 in combination with @option{-funsafe-math-optimizations} and
24067 @option{-ffinite-math-only}. It is enabled by default when generating code for
24068 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
24069 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
24072 @item -mpretend-cmove
24073 @opindex mpretend-cmove
24074 Prefer zero-displacement conditional branches for conditional move instruction
24075 patterns. This can result in faster code on the SH4 processor.
24079 Generate code using the FDPIC ABI.
24083 @node Solaris 2 Options
24084 @subsection Solaris 2 Options
24085 @cindex Solaris 2 options
24087 These @samp{-m} options are supported on Solaris 2:
24090 @item -mclear-hwcap
24091 @opindex mclear-hwcap
24092 @option{-mclear-hwcap} tells the compiler to remove the hardware
24093 capabilities generated by the Solaris assembler. This is only necessary
24094 when object files use ISA extensions not supported by the current
24095 machine, but check at runtime whether or not to use them.
24097 @item -mimpure-text
24098 @opindex mimpure-text
24099 @option{-mimpure-text}, used in addition to @option{-shared}, tells
24100 the compiler to not pass @option{-z text} to the linker when linking a
24101 shared object. Using this option, you can link position-dependent
24102 code into a shared object.
24104 @option{-mimpure-text} suppresses the ``relocations remain against
24105 allocatable but non-writable sections'' linker error message.
24106 However, the necessary relocations trigger copy-on-write, and the
24107 shared object is not actually shared across processes. Instead of
24108 using @option{-mimpure-text}, you should compile all source code with
24109 @option{-fpic} or @option{-fPIC}.
24113 These switches are supported in addition to the above on Solaris 2:
24118 This is a synonym for @option{-pthread}.
24121 @node SPARC Options
24122 @subsection SPARC Options
24123 @cindex SPARC options
24125 These @samp{-m} options are supported on the SPARC:
24128 @item -mno-app-regs
24130 @opindex mno-app-regs
24132 Specify @option{-mapp-regs} to generate output using the global registers
24133 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
24134 global register 1, each global register 2 through 4 is then treated as an
24135 allocable register that is clobbered by function calls. This is the default.
24137 To be fully SVR4 ABI-compliant at the cost of some performance loss,
24138 specify @option{-mno-app-regs}. You should compile libraries and system
24139 software with this option.
24145 With @option{-mflat}, the compiler does not generate save/restore instructions
24146 and uses a ``flat'' or single register window model. This model is compatible
24147 with the regular register window model. The local registers and the input
24148 registers (0--5) are still treated as ``call-saved'' registers and are
24149 saved on the stack as needed.
24151 With @option{-mno-flat} (the default), the compiler generates save/restore
24152 instructions (except for leaf functions). This is the normal operating mode.
24155 @itemx -mhard-float
24157 @opindex mhard-float
24158 Generate output containing floating-point instructions. This is the
24162 @itemx -msoft-float
24164 @opindex msoft-float
24165 Generate output containing library calls for floating point.
24166 @strong{Warning:} the requisite libraries are not available for all SPARC
24167 targets. Normally the facilities of the machine's usual C compiler are
24168 used, but this cannot be done directly in cross-compilation. You must make
24169 your own arrangements to provide suitable library functions for
24170 cross-compilation. The embedded targets @samp{sparc-*-aout} and
24171 @samp{sparclite-*-*} do provide software floating-point support.
24173 @option{-msoft-float} changes the calling convention in the output file;
24174 therefore, it is only useful if you compile @emph{all} of a program with
24175 this option. In particular, you need to compile @file{libgcc.a}, the
24176 library that comes with GCC, with @option{-msoft-float} in order for
24179 @item -mhard-quad-float
24180 @opindex mhard-quad-float
24181 Generate output containing quad-word (long double) floating-point
24184 @item -msoft-quad-float
24185 @opindex msoft-quad-float
24186 Generate output containing library calls for quad-word (long double)
24187 floating-point instructions. The functions called are those specified
24188 in the SPARC ABI@. This is the default.
24190 As of this writing, there are no SPARC implementations that have hardware
24191 support for the quad-word floating-point instructions. They all invoke
24192 a trap handler for one of these instructions, and then the trap handler
24193 emulates the effect of the instruction. Because of the trap handler overhead,
24194 this is much slower than calling the ABI library routines. Thus the
24195 @option{-msoft-quad-float} option is the default.
24197 @item -mno-unaligned-doubles
24198 @itemx -munaligned-doubles
24199 @opindex mno-unaligned-doubles
24200 @opindex munaligned-doubles
24201 Assume that doubles have 8-byte alignment. This is the default.
24203 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
24204 alignment only if they are contained in another type, or if they have an
24205 absolute address. Otherwise, it assumes they have 4-byte alignment.
24206 Specifying this option avoids some rare compatibility problems with code
24207 generated by other compilers. It is not the default because it results
24208 in a performance loss, especially for floating-point code.
24211 @itemx -mno-user-mode
24212 @opindex muser-mode
24213 @opindex mno-user-mode
24214 Do not generate code that can only run in supervisor mode. This is relevant
24215 only for the @code{casa} instruction emitted for the LEON3 processor. This
24218 @item -mfaster-structs
24219 @itemx -mno-faster-structs
24220 @opindex mfaster-structs
24221 @opindex mno-faster-structs
24222 With @option{-mfaster-structs}, the compiler assumes that structures
24223 should have 8-byte alignment. This enables the use of pairs of
24224 @code{ldd} and @code{std} instructions for copies in structure
24225 assignment, in place of twice as many @code{ld} and @code{st} pairs.
24226 However, the use of this changed alignment directly violates the SPARC
24227 ABI@. Thus, it's intended only for use on targets where the developer
24228 acknowledges that their resulting code is not directly in line with
24229 the rules of the ABI@.
24231 @item -mstd-struct-return
24232 @itemx -mno-std-struct-return
24233 @opindex mstd-struct-return
24234 @opindex mno-std-struct-return
24235 With @option{-mstd-struct-return}, the compiler generates checking code
24236 in functions returning structures or unions to detect size mismatches
24237 between the two sides of function calls, as per the 32-bit ABI@.
24239 The default is @option{-mno-std-struct-return}. This option has no effect
24246 Enable Local Register Allocation. This is the default for SPARC since GCC 7
24247 so @option{-mno-lra} needs to be passed to get old Reload.
24249 @item -mcpu=@var{cpu_type}
24251 Set the instruction set, register set, and instruction scheduling parameters
24252 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24253 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
24254 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
24255 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
24256 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
24257 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
24259 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
24260 which selects the best architecture option for the host processor.
24261 @option{-mcpu=native} has no effect if GCC does not recognize
24264 Default instruction scheduling parameters are used for values that select
24265 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
24266 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
24268 Here is a list of each supported architecture and their supported
24276 supersparc, hypersparc, leon, leon3
24279 f930, f934, sparclite86x
24285 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
24289 By default (unless configured otherwise), GCC generates code for the V7
24290 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
24291 additionally optimizes it for the Cypress CY7C602 chip, as used in the
24292 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
24293 SPARCStation 1, 2, IPX etc.
24295 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
24296 architecture. The only difference from V7 code is that the compiler emits
24297 the integer multiply and integer divide instructions which exist in SPARC-V8
24298 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
24299 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
24302 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
24303 the SPARC architecture. This adds the integer multiply, integer divide step
24304 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
24305 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
24306 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
24307 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
24308 MB86934 chip, which is the more recent SPARClite with FPU@.
24310 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
24311 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
24312 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
24313 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
24314 optimizes it for the TEMIC SPARClet chip.
24316 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
24317 architecture. This adds 64-bit integer and floating-point move instructions,
24318 3 additional floating-point condition code registers and conditional move
24319 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
24320 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
24321 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
24322 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
24323 @option{-mcpu=niagara}, the compiler additionally optimizes it for
24324 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
24325 additionally optimizes it for Sun UltraSPARC T2 chips. With
24326 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
24327 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
24328 additionally optimizes it for Sun UltraSPARC T4 chips. With
24329 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
24330 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
24331 additionally optimizes it for Oracle M8 chips.
24333 @item -mtune=@var{cpu_type}
24335 Set the instruction scheduling parameters for machine type
24336 @var{cpu_type}, but do not set the instruction set or register set that the
24337 option @option{-mcpu=@var{cpu_type}} does.
24339 The same values for @option{-mcpu=@var{cpu_type}} can be used for
24340 @option{-mtune=@var{cpu_type}}, but the only useful values are those
24341 that select a particular CPU implementation. Those are
24342 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
24343 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
24344 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
24345 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
24346 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
24347 and GNU/Linux toolchains, @samp{native} can also be used.
24352 @opindex mno-v8plus
24353 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
24354 difference from the V8 ABI is that the global and out registers are
24355 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
24356 mode for all SPARC-V9 processors.
24362 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
24363 Visual Instruction Set extensions. The default is @option{-mno-vis}.
24369 With @option{-mvis2}, GCC generates code that takes advantage of
24370 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
24371 default is @option{-mvis2} when targeting a cpu that supports such
24372 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
24373 also sets @option{-mvis}.
24379 With @option{-mvis3}, GCC generates code that takes advantage of
24380 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
24381 default is @option{-mvis3} when targeting a cpu that supports such
24382 instructions, such as niagara-3 and later. Setting @option{-mvis3}
24383 also sets @option{-mvis2} and @option{-mvis}.
24389 With @option{-mvis4}, GCC generates code that takes advantage of
24390 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
24391 default is @option{-mvis4} when targeting a cpu that supports such
24392 instructions, such as niagara-7 and later. Setting @option{-mvis4}
24393 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
24399 With @option{-mvis4b}, GCC generates code that takes advantage of
24400 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
24401 the additional VIS instructions introduced in the Oracle SPARC
24402 Architecture 2017. The default is @option{-mvis4b} when targeting a
24403 cpu that supports such instructions, such as m8 and later. Setting
24404 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
24405 @option{-mvis2} and @option{-mvis}.
24410 @opindex mno-cbcond
24411 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
24412 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
24413 when targeting a CPU that supports such instructions, such as Niagara-4 and
24420 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
24421 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
24422 when targeting a CPU that supports such instructions, such as Niagara-3 and
24428 @opindex mno-fsmuld
24429 With @option{-mfsmuld}, GCC generates code that takes advantage of the
24430 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
24431 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
24432 or V9 with FPU except @option{-mcpu=leon}.
24438 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
24439 Population Count instruction. The default is @option{-mpopc}
24440 when targeting a CPU that supports such an instruction, such as Niagara-2 and
24447 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
24448 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
24449 when targeting a CPU that supports such an instruction, such as Niagara-7 and
24453 @opindex mfix-at697f
24454 Enable the documented workaround for the single erratum of the Atmel AT697F
24455 processor (which corresponds to erratum #13 of the AT697E processor).
24458 @opindex mfix-ut699
24459 Enable the documented workarounds for the floating-point errata and the data
24460 cache nullify errata of the UT699 processor.
24463 @opindex mfix-ut700
24464 Enable the documented workaround for the back-to-back store errata of
24465 the UT699E/UT700 processor.
24467 @item -mfix-gr712rc
24468 @opindex mfix-gr712rc
24469 Enable the documented workaround for the back-to-back store errata of
24470 the GR712RC processor.
24473 These @samp{-m} options are supported in addition to the above
24474 on SPARC-V9 processors in 64-bit environments:
24481 Generate code for a 32-bit or 64-bit environment.
24482 The 32-bit environment sets int, long and pointer to 32 bits.
24483 The 64-bit environment sets int to 32 bits and long and pointer
24486 @item -mcmodel=@var{which}
24488 Set the code model to one of
24492 The Medium/Low code model: 64-bit addresses, programs
24493 must be linked in the low 32 bits of memory. Programs can be statically
24494 or dynamically linked.
24497 The Medium/Middle code model: 64-bit addresses, programs
24498 must be linked in the low 44 bits of memory, the text and data segments must
24499 be less than 2GB in size and the data segment must be located within 2GB of
24503 The Medium/Anywhere code model: 64-bit addresses, programs
24504 may be linked anywhere in memory, the text and data segments must be less
24505 than 2GB in size and the data segment must be located within 2GB of the
24509 The Medium/Anywhere code model for embedded systems:
24510 64-bit addresses, the text and data segments must be less than 2GB in
24511 size, both starting anywhere in memory (determined at link time). The
24512 global register %g4 points to the base of the data segment. Programs
24513 are statically linked and PIC is not supported.
24516 @item -mmemory-model=@var{mem-model}
24517 @opindex mmemory-model
24518 Set the memory model in force on the processor to one of
24522 The default memory model for the processor and operating system.
24525 Relaxed Memory Order
24528 Partial Store Order
24534 Sequential Consistency
24537 These memory models are formally defined in Appendix D of the SPARC-V9
24538 architecture manual, as set in the processor's @code{PSTATE.MM} field.
24541 @itemx -mno-stack-bias
24542 @opindex mstack-bias
24543 @opindex mno-stack-bias
24544 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
24545 frame pointer if present, are offset by @minus{}2047 which must be added back
24546 when making stack frame references. This is the default in 64-bit mode.
24547 Otherwise, assume no such offset is present.
24551 @subsection SPU Options
24552 @cindex SPU options
24554 These @samp{-m} options are supported on the SPU:
24558 @itemx -merror-reloc
24559 @opindex mwarn-reloc
24560 @opindex merror-reloc
24562 The loader for SPU does not handle dynamic relocations. By default, GCC
24563 gives an error when it generates code that requires a dynamic
24564 relocation. @option{-mno-error-reloc} disables the error,
24565 @option{-mwarn-reloc} generates a warning instead.
24568 @itemx -munsafe-dma
24570 @opindex munsafe-dma
24572 Instructions that initiate or test completion of DMA must not be
24573 reordered with respect to loads and stores of the memory that is being
24575 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
24576 memory accesses, but that can lead to inefficient code in places where the
24577 memory is known to not change. Rather than mark the memory as volatile,
24578 you can use @option{-msafe-dma} to tell the compiler to treat
24579 the DMA instructions as potentially affecting all memory.
24581 @item -mbranch-hints
24582 @opindex mbranch-hints
24584 By default, GCC generates a branch hint instruction to avoid
24585 pipeline stalls for always-taken or probably-taken branches. A hint
24586 is not generated closer than 8 instructions away from its branch.
24587 There is little reason to disable them, except for debugging purposes,
24588 or to make an object a little bit smaller.
24592 @opindex msmall-mem
24593 @opindex mlarge-mem
24595 By default, GCC generates code assuming that addresses are never larger
24596 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
24597 a full 32-bit address.
24602 By default, GCC links against startup code that assumes the SPU-style
24603 main function interface (which has an unconventional parameter list).
24604 With @option{-mstdmain}, GCC links your program against startup
24605 code that assumes a C99-style interface to @code{main}, including a
24606 local copy of @code{argv} strings.
24608 @item -mfixed-range=@var{register-range}
24609 @opindex mfixed-range
24610 Generate code treating the given register range as fixed registers.
24611 A fixed register is one that the register allocator cannot use. This is
24612 useful when compiling kernel code. A register range is specified as
24613 two registers separated by a dash. Multiple register ranges can be
24614 specified separated by a comma.
24620 Compile code assuming that pointers to the PPU address space accessed
24621 via the @code{__ea} named address space qualifier are either 32 or 64
24622 bits wide. The default is 32 bits. As this is an ABI-changing option,
24623 all object code in an executable must be compiled with the same setting.
24625 @item -maddress-space-conversion
24626 @itemx -mno-address-space-conversion
24627 @opindex maddress-space-conversion
24628 @opindex mno-address-space-conversion
24629 Allow/disallow treating the @code{__ea} address space as superset
24630 of the generic address space. This enables explicit type casts
24631 between @code{__ea} and generic pointer as well as implicit
24632 conversions of generic pointers to @code{__ea} pointers. The
24633 default is to allow address space pointer conversions.
24635 @item -mcache-size=@var{cache-size}
24636 @opindex mcache-size
24637 This option controls the version of libgcc that the compiler links to an
24638 executable and selects a software-managed cache for accessing variables
24639 in the @code{__ea} address space with a particular cache size. Possible
24640 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
24641 and @samp{128}. The default cache size is 64KB.
24643 @item -matomic-updates
24644 @itemx -mno-atomic-updates
24645 @opindex matomic-updates
24646 @opindex mno-atomic-updates
24647 This option controls the version of libgcc that the compiler links to an
24648 executable and selects whether atomic updates to the software-managed
24649 cache of PPU-side variables are used. If you use atomic updates, changes
24650 to a PPU variable from SPU code using the @code{__ea} named address space
24651 qualifier do not interfere with changes to other PPU variables residing
24652 in the same cache line from PPU code. If you do not use atomic updates,
24653 such interference may occur; however, writing back cache lines is
24654 more efficient. The default behavior is to use atomic updates.
24657 @itemx -mdual-nops=@var{n}
24658 @opindex mdual-nops
24659 By default, GCC inserts NOPs to increase dual issue when it expects
24660 it to increase performance. @var{n} can be a value from 0 to 10. A
24661 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
24662 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
24664 @item -mhint-max-nops=@var{n}
24665 @opindex mhint-max-nops
24666 Maximum number of NOPs to insert for a branch hint. A branch hint must
24667 be at least 8 instructions away from the branch it is affecting. GCC
24668 inserts up to @var{n} NOPs to enforce this, otherwise it does not
24669 generate the branch hint.
24671 @item -mhint-max-distance=@var{n}
24672 @opindex mhint-max-distance
24673 The encoding of the branch hint instruction limits the hint to be within
24674 256 instructions of the branch it is affecting. By default, GCC makes
24675 sure it is within 125.
24678 @opindex msafe-hints
24679 Work around a hardware bug that causes the SPU to stall indefinitely.
24680 By default, GCC inserts the @code{hbrp} instruction to make sure
24681 this stall won't happen.
24685 @node System V Options
24686 @subsection Options for System V
24688 These additional options are available on System V Release 4 for
24689 compatibility with other compilers on those systems:
24694 Create a shared object.
24695 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
24699 Identify the versions of each tool used by the compiler, in a
24700 @code{.ident} assembler directive in the output.
24704 Refrain from adding @code{.ident} directives to the output file (this is
24707 @item -YP,@var{dirs}
24709 Search the directories @var{dirs}, and no others, for libraries
24710 specified with @option{-l}.
24712 @item -Ym,@var{dir}
24714 Look in the directory @var{dir} to find the M4 preprocessor.
24715 The assembler uses this option.
24716 @c This is supposed to go with a -Yd for predefined M4 macro files, but
24717 @c the generic assembler that comes with Solaris takes just -Ym.
24720 @node TILE-Gx Options
24721 @subsection TILE-Gx Options
24722 @cindex TILE-Gx options
24724 These @samp{-m} options are supported on the TILE-Gx:
24727 @item -mcmodel=small
24728 @opindex mcmodel=small
24729 Generate code for the small model. The distance for direct calls is
24730 limited to 500M in either direction. PC-relative addresses are 32
24731 bits. Absolute addresses support the full address range.
24733 @item -mcmodel=large
24734 @opindex mcmodel=large
24735 Generate code for the large model. There is no limitation on call
24736 distance, pc-relative addresses, or absolute addresses.
24738 @item -mcpu=@var{name}
24740 Selects the type of CPU to be targeted. Currently the only supported
24741 type is @samp{tilegx}.
24747 Generate code for a 32-bit or 64-bit environment. The 32-bit
24748 environment sets int, long, and pointer to 32 bits. The 64-bit
24749 environment sets int to 32 bits and long and pointer to 64 bits.
24752 @itemx -mlittle-endian
24753 @opindex mbig-endian
24754 @opindex mlittle-endian
24755 Generate code in big/little endian mode, respectively.
24758 @node TILEPro Options
24759 @subsection TILEPro Options
24760 @cindex TILEPro options
24762 These @samp{-m} options are supported on the TILEPro:
24765 @item -mcpu=@var{name}
24767 Selects the type of CPU to be targeted. Currently the only supported
24768 type is @samp{tilepro}.
24772 Generate code for a 32-bit environment, which sets int, long, and
24773 pointer to 32 bits. This is the only supported behavior so the flag
24774 is essentially ignored.
24778 @subsection V850 Options
24779 @cindex V850 Options
24781 These @samp{-m} options are defined for V850 implementations:
24785 @itemx -mno-long-calls
24786 @opindex mlong-calls
24787 @opindex mno-long-calls
24788 Treat all calls as being far away (near). If calls are assumed to be
24789 far away, the compiler always loads the function's address into a
24790 register, and calls indirect through the pointer.
24796 Do not optimize (do optimize) basic blocks that use the same index
24797 pointer 4 or more times to copy pointer into the @code{ep} register, and
24798 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
24799 option is on by default if you optimize.
24801 @item -mno-prolog-function
24802 @itemx -mprolog-function
24803 @opindex mno-prolog-function
24804 @opindex mprolog-function
24805 Do not use (do use) external functions to save and restore registers
24806 at the prologue and epilogue of a function. The external functions
24807 are slower, but use less code space if more than one function saves
24808 the same number of registers. The @option{-mprolog-function} option
24809 is on by default if you optimize.
24813 Try to make the code as small as possible. At present, this just turns
24814 on the @option{-mep} and @option{-mprolog-function} options.
24816 @item -mtda=@var{n}
24818 Put static or global variables whose size is @var{n} bytes or less into
24819 the tiny data area that register @code{ep} points to. The tiny data
24820 area can hold up to 256 bytes in total (128 bytes for byte references).
24822 @item -msda=@var{n}
24824 Put static or global variables whose size is @var{n} bytes or less into
24825 the small data area that register @code{gp} points to. The small data
24826 area can hold up to 64 kilobytes.
24828 @item -mzda=@var{n}
24830 Put static or global variables whose size is @var{n} bytes or less into
24831 the first 32 kilobytes of memory.
24835 Specify that the target processor is the V850.
24839 Specify that the target processor is the V850E3V5. The preprocessor
24840 constant @code{__v850e3v5__} is defined if this option is used.
24844 Specify that the target processor is the V850E3V5. This is an alias for
24845 the @option{-mv850e3v5} option.
24849 Specify that the target processor is the V850E2V3. The preprocessor
24850 constant @code{__v850e2v3__} is defined if this option is used.
24854 Specify that the target processor is the V850E2. The preprocessor
24855 constant @code{__v850e2__} is defined if this option is used.
24859 Specify that the target processor is the V850E1. The preprocessor
24860 constants @code{__v850e1__} and @code{__v850e__} are defined if
24861 this option is used.
24865 Specify that the target processor is the V850ES. This is an alias for
24866 the @option{-mv850e1} option.
24870 Specify that the target processor is the V850E@. The preprocessor
24871 constant @code{__v850e__} is defined if this option is used.
24873 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
24874 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
24875 are defined then a default target processor is chosen and the
24876 relevant @samp{__v850*__} preprocessor constant is defined.
24878 The preprocessor constants @code{__v850} and @code{__v851__} are always
24879 defined, regardless of which processor variant is the target.
24881 @item -mdisable-callt
24882 @itemx -mno-disable-callt
24883 @opindex mdisable-callt
24884 @opindex mno-disable-callt
24885 This option suppresses generation of the @code{CALLT} instruction for the
24886 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
24889 This option is enabled by default when the RH850 ABI is
24890 in use (see @option{-mrh850-abi}), and disabled by default when the
24891 GCC ABI is in use. If @code{CALLT} instructions are being generated
24892 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
24898 Pass on (or do not pass on) the @option{-mrelax} command-line option
24902 @itemx -mno-long-jumps
24903 @opindex mlong-jumps
24904 @opindex mno-long-jumps
24905 Disable (or re-enable) the generation of PC-relative jump instructions.
24908 @itemx -mhard-float
24909 @opindex msoft-float
24910 @opindex mhard-float
24911 Disable (or re-enable) the generation of hardware floating point
24912 instructions. This option is only significant when the target
24913 architecture is @samp{V850E2V3} or higher. If hardware floating point
24914 instructions are being generated then the C preprocessor symbol
24915 @code{__FPU_OK__} is defined, otherwise the symbol
24916 @code{__NO_FPU__} is defined.
24920 Enables the use of the e3v5 LOOP instruction. The use of this
24921 instruction is not enabled by default when the e3v5 architecture is
24922 selected because its use is still experimental.
24926 @opindex mrh850-abi
24928 Enables support for the RH850 version of the V850 ABI. This is the
24929 default. With this version of the ABI the following rules apply:
24933 Integer sized structures and unions are returned via a memory pointer
24934 rather than a register.
24937 Large structures and unions (more than 8 bytes in size) are passed by
24941 Functions are aligned to 16-bit boundaries.
24944 The @option{-m8byte-align} command-line option is supported.
24947 The @option{-mdisable-callt} command-line option is enabled by
24948 default. The @option{-mno-disable-callt} command-line option is not
24952 When this version of the ABI is enabled the C preprocessor symbol
24953 @code{__V850_RH850_ABI__} is defined.
24957 Enables support for the old GCC version of the V850 ABI. With this
24958 version of the ABI the following rules apply:
24962 Integer sized structures and unions are returned in register @code{r10}.
24965 Large structures and unions (more than 8 bytes in size) are passed by
24969 Functions are aligned to 32-bit boundaries, unless optimizing for
24973 The @option{-m8byte-align} command-line option is not supported.
24976 The @option{-mdisable-callt} command-line option is supported but not
24977 enabled by default.
24980 When this version of the ABI is enabled the C preprocessor symbol
24981 @code{__V850_GCC_ABI__} is defined.
24983 @item -m8byte-align
24984 @itemx -mno-8byte-align
24985 @opindex m8byte-align
24986 @opindex mno-8byte-align
24987 Enables support for @code{double} and @code{long long} types to be
24988 aligned on 8-byte boundaries. The default is to restrict the
24989 alignment of all objects to at most 4-bytes. When
24990 @option{-m8byte-align} is in effect the C preprocessor symbol
24991 @code{__V850_8BYTE_ALIGN__} is defined.
24994 @opindex mbig-switch
24995 Generate code suitable for big switch tables. Use this option only if
24996 the assembler/linker complain about out of range branches within a switch
25001 This option causes r2 and r5 to be used in the code generated by
25002 the compiler. This setting is the default.
25004 @item -mno-app-regs
25005 @opindex mno-app-regs
25006 This option causes r2 and r5 to be treated as fixed registers.
25011 @subsection VAX Options
25012 @cindex VAX options
25014 These @samp{-m} options are defined for the VAX:
25019 Do not output certain jump instructions (@code{aobleq} and so on)
25020 that the Unix assembler for the VAX cannot handle across long
25025 Do output those jump instructions, on the assumption that the
25026 GNU assembler is being used.
25030 Output code for G-format floating-point numbers instead of D-format.
25033 @node Visium Options
25034 @subsection Visium Options
25035 @cindex Visium options
25041 A program which performs file I/O and is destined to run on an MCM target
25042 should be linked with this option. It causes the libraries libc.a and
25043 libdebug.a to be linked. The program should be run on the target under
25044 the control of the GDB remote debugging stub.
25048 A program which performs file I/O and is destined to run on the simulator
25049 should be linked with option. This causes libraries libc.a and libsim.a to
25053 @itemx -mhard-float
25055 @opindex mhard-float
25056 Generate code containing floating-point instructions. This is the
25060 @itemx -msoft-float
25062 @opindex msoft-float
25063 Generate code containing library calls for floating-point.
25065 @option{-msoft-float} changes the calling convention in the output file;
25066 therefore, it is only useful if you compile @emph{all} of a program with
25067 this option. In particular, you need to compile @file{libgcc.a}, the
25068 library that comes with GCC, with @option{-msoft-float} in order for
25071 @item -mcpu=@var{cpu_type}
25073 Set the instruction set, register set, and instruction scheduling parameters
25074 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
25075 @samp{mcm}, @samp{gr5} and @samp{gr6}.
25077 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
25079 By default (unless configured otherwise), GCC generates code for the GR5
25080 variant of the Visium architecture.
25082 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
25083 architecture. The only difference from GR5 code is that the compiler will
25084 generate block move instructions.
25086 @item -mtune=@var{cpu_type}
25088 Set the instruction scheduling parameters for machine type @var{cpu_type},
25089 but do not set the instruction set or register set that the option
25090 @option{-mcpu=@var{cpu_type}} would.
25094 Generate code for the supervisor mode, where there are no restrictions on
25095 the access to general registers. This is the default.
25098 @opindex muser-mode
25099 Generate code for the user mode, where the access to some general registers
25100 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
25101 mode; on the GR6, only registers r29 to r31 are affected.
25105 @subsection VMS Options
25107 These @samp{-m} options are defined for the VMS implementations:
25110 @item -mvms-return-codes
25111 @opindex mvms-return-codes
25112 Return VMS condition codes from @code{main}. The default is to return POSIX-style
25113 condition (e.g.@ error) codes.
25115 @item -mdebug-main=@var{prefix}
25116 @opindex mdebug-main=@var{prefix}
25117 Flag the first routine whose name starts with @var{prefix} as the main
25118 routine for the debugger.
25122 Default to 64-bit memory allocation routines.
25124 @item -mpointer-size=@var{size}
25125 @opindex mpointer-size=@var{size}
25126 Set the default size of pointers. Possible options for @var{size} are
25127 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
25128 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
25129 The later option disables @code{pragma pointer_size}.
25132 @node VxWorks Options
25133 @subsection VxWorks Options
25134 @cindex VxWorks Options
25136 The options in this section are defined for all VxWorks targets.
25137 Options specific to the target hardware are listed with the other
25138 options for that target.
25143 GCC can generate code for both VxWorks kernels and real time processes
25144 (RTPs). This option switches from the former to the latter. It also
25145 defines the preprocessor macro @code{__RTP__}.
25148 @opindex non-static
25149 Link an RTP executable against shared libraries rather than static
25150 libraries. The options @option{-static} and @option{-shared} can
25151 also be used for RTPs (@pxref{Link Options}); @option{-static}
25158 These options are passed down to the linker. They are defined for
25159 compatibility with Diab.
25162 @opindex Xbind-lazy
25163 Enable lazy binding of function calls. This option is equivalent to
25164 @option{-Wl,-z,now} and is defined for compatibility with Diab.
25168 Disable lazy binding of function calls. This option is the default and
25169 is defined for compatibility with Diab.
25173 @subsection x86 Options
25174 @cindex x86 Options
25176 These @samp{-m} options are defined for the x86 family of computers.
25180 @item -march=@var{cpu-type}
25182 Generate instructions for the machine type @var{cpu-type}. In contrast to
25183 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
25184 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
25185 to generate code that may not run at all on processors other than the one
25186 indicated. Specifying @option{-march=@var{cpu-type}} implies
25187 @option{-mtune=@var{cpu-type}}.
25189 The choices for @var{cpu-type} are:
25193 This selects the CPU to generate code for at compilation time by determining
25194 the processor type of the compiling machine. Using @option{-march=native}
25195 enables all instruction subsets supported by the local machine (hence
25196 the result might not run on different machines). Using @option{-mtune=native}
25197 produces code optimized for the local machine under the constraints
25198 of the selected instruction set.
25201 Original Intel i386 CPU@.
25204 Intel i486 CPU@. (No scheduling is implemented for this chip.)
25208 Intel Pentium CPU with no MMX support.
25211 Intel Lakemont MCU, based on Intel Pentium CPU.
25214 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
25217 Intel Pentium Pro CPU@.
25220 When used with @option{-march}, the Pentium Pro
25221 instruction set is used, so the code runs on all i686 family chips.
25222 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
25225 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
25230 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
25234 Intel Pentium M; low-power version of Intel Pentium III CPU
25235 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
25239 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
25242 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
25246 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
25247 SSE2 and SSE3 instruction set support.
25250 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
25251 instruction set support.
25254 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25255 SSE4.1, SSE4.2 and POPCNT instruction set support.
25258 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25259 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
25262 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25263 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
25266 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25267 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
25268 instruction set support.
25271 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25272 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25273 BMI, BMI2 and F16C instruction set support.
25276 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25277 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25278 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
25281 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25282 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25283 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
25284 XSAVES instruction set support.
25287 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
25288 instruction set support.
25291 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25292 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
25295 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25296 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25297 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
25298 AVX512CD instruction set support.
25301 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25302 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25303 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25304 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
25306 @item skylake-avx512
25307 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25308 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25309 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
25310 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
25313 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
25314 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
25315 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
25316 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
25317 AVX512IFMA, SHA, CLWB and UMIP instruction set support.
25320 AMD K6 CPU with MMX instruction set support.
25324 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
25327 @itemx athlon-tbird
25328 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
25334 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
25335 instruction set support.
25341 Processors based on the AMD K8 core with x86-64 instruction set support,
25342 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
25343 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
25344 instruction set extensions.)
25347 @itemx opteron-sse3
25348 @itemx athlon64-sse3
25349 Improved versions of AMD K8 cores with SSE3 instruction set support.
25353 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
25354 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
25355 instruction set extensions.)
25358 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
25359 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
25360 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
25362 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25363 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
25364 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
25367 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25368 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
25369 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
25370 64-bit instruction set extensions.
25372 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25373 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
25374 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
25375 SSE4.2, ABM and 64-bit instruction set extensions.
25378 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
25379 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
25380 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
25381 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
25382 instruction set extensions.
25385 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
25386 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
25387 instruction set extensions.)
25390 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
25391 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
25392 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
25395 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
25399 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
25400 instruction set support.
25403 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
25404 (No scheduling is implemented for this chip.)
25407 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
25408 (No scheduling is implemented for this chip.)
25411 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25412 (No scheduling is implemented for this chip.)
25415 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
25416 (No scheduling is implemented for this chip.)
25419 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
25420 (No scheduling is implemented for this chip.)
25423 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25424 (No scheduling is implemented for this chip.)
25427 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
25428 (No scheduling is implemented for this chip.)
25431 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
25432 AVX and AVX2 instruction set support.
25433 (No scheduling is implemented for this chip.)
25436 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25437 instruction set support.
25438 (No scheduling is implemented for this chip.)
25441 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25442 instruction set support.
25443 (No scheduling is implemented for this chip.)
25446 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25447 instruction set support.
25448 (No scheduling is implemented for this chip.)
25451 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25452 instruction set support.
25453 (No scheduling is implemented for this chip.)
25456 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25457 instruction set support.
25458 (No scheduling is implemented for this chip.)
25461 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25462 instruction set support.
25463 (No scheduling is implemented for this chip.)
25466 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
25469 @item -mtune=@var{cpu-type}
25471 Tune to @var{cpu-type} everything applicable about the generated code, except
25472 for the ABI and the set of available instructions.
25473 While picking a specific @var{cpu-type} schedules things appropriately
25474 for that particular chip, the compiler does not generate any code that
25475 cannot run on the default machine type unless you use a
25476 @option{-march=@var{cpu-type}} option.
25477 For example, if GCC is configured for i686-pc-linux-gnu
25478 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
25479 but still runs on i686 machines.
25481 The choices for @var{cpu-type} are the same as for @option{-march}.
25482 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
25486 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
25487 If you know the CPU on which your code will run, then you should use
25488 the corresponding @option{-mtune} or @option{-march} option instead of
25489 @option{-mtune=generic}. But, if you do not know exactly what CPU users
25490 of your application will have, then you should use this option.
25492 As new processors are deployed in the marketplace, the behavior of this
25493 option will change. Therefore, if you upgrade to a newer version of
25494 GCC, code generation controlled by this option will change to reflect
25496 that are most common at the time that version of GCC is released.
25498 There is no @option{-march=generic} option because @option{-march}
25499 indicates the instruction set the compiler can use, and there is no
25500 generic instruction set applicable to all processors. In contrast,
25501 @option{-mtune} indicates the processor (or, in this case, collection of
25502 processors) for which the code is optimized.
25505 Produce code optimized for the most current Intel processors, which are
25506 Haswell and Silvermont for this version of GCC. If you know the CPU
25507 on which your code will run, then you should use the corresponding
25508 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
25509 But, if you want your application performs better on both Haswell and
25510 Silvermont, then you should use this option.
25512 As new Intel processors are deployed in the marketplace, the behavior of
25513 this option will change. Therefore, if you upgrade to a newer version of
25514 GCC, code generation controlled by this option will change to reflect
25515 the most current Intel processors at the time that version of GCC is
25518 There is no @option{-march=intel} option because @option{-march} indicates
25519 the instruction set the compiler can use, and there is no common
25520 instruction set applicable to all processors. In contrast,
25521 @option{-mtune} indicates the processor (or, in this case, collection of
25522 processors) for which the code is optimized.
25525 @item -mcpu=@var{cpu-type}
25527 A deprecated synonym for @option{-mtune}.
25529 @item -mfpmath=@var{unit}
25531 Generate floating-point arithmetic for selected unit @var{unit}. The choices
25532 for @var{unit} are:
25536 Use the standard 387 floating-point coprocessor present on the majority of chips and
25537 emulated otherwise. Code compiled with this option runs almost everywhere.
25538 The temporary results are computed in 80-bit precision instead of the precision
25539 specified by the type, resulting in slightly different results compared to most
25540 of other chips. See @option{-ffloat-store} for more detailed description.
25542 This is the default choice for non-Darwin x86-32 targets.
25545 Use scalar floating-point instructions present in the SSE instruction set.
25546 This instruction set is supported by Pentium III and newer chips,
25547 and in the AMD line
25548 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
25549 instruction set supports only single-precision arithmetic, thus the double and
25550 extended-precision arithmetic are still done using 387. A later version, present
25551 only in Pentium 4 and AMD x86-64 chips, supports double-precision
25554 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
25555 or @option{-msse2} switches to enable SSE extensions and make this option
25556 effective. For the x86-64 compiler, these extensions are enabled by default.
25558 The resulting code should be considerably faster in the majority of cases and avoid
25559 the numerical instability problems of 387 code, but may break some existing
25560 code that expects temporaries to be 80 bits.
25562 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
25563 and the default choice for x86-32 targets with the SSE2 instruction set
25564 when @option{-ffast-math} is enabled.
25569 Attempt to utilize both instruction sets at once. This effectively doubles the
25570 amount of available registers, and on chips with separate execution units for
25571 387 and SSE the execution resources too. Use this option with care, as it is
25572 still experimental, because the GCC register allocator does not model separate
25573 functional units well, resulting in unstable performance.
25576 @item -masm=@var{dialect}
25577 @opindex masm=@var{dialect}
25578 Output assembly instructions using selected @var{dialect}. Also affects
25579 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
25580 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
25581 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
25582 not support @samp{intel}.
25585 @itemx -mno-ieee-fp
25587 @opindex mno-ieee-fp
25588 Control whether or not the compiler uses IEEE floating-point
25589 comparisons. These correctly handle the case where the result of a
25590 comparison is unordered.
25595 @opindex mhard-float
25596 Generate output containing 80387 instructions for floating point.
25601 @opindex msoft-float
25602 Generate output containing library calls for floating point.
25604 @strong{Warning:} the requisite libraries are not part of GCC@.
25605 Normally the facilities of the machine's usual C compiler are used, but
25606 this cannot be done directly in cross-compilation. You must make your
25607 own arrangements to provide suitable library functions for
25610 On machines where a function returns floating-point results in the 80387
25611 register stack, some floating-point opcodes may be emitted even if
25612 @option{-msoft-float} is used.
25614 @item -mno-fp-ret-in-387
25615 @opindex mno-fp-ret-in-387
25616 Do not use the FPU registers for return values of functions.
25618 The usual calling convention has functions return values of types
25619 @code{float} and @code{double} in an FPU register, even if there
25620 is no FPU@. The idea is that the operating system should emulate
25623 The option @option{-mno-fp-ret-in-387} causes such values to be returned
25624 in ordinary CPU registers instead.
25626 @item -mno-fancy-math-387
25627 @opindex mno-fancy-math-387
25628 Some 387 emulators do not support the @code{sin}, @code{cos} and
25629 @code{sqrt} instructions for the 387. Specify this option to avoid
25630 generating those instructions. This option is the default on
25631 OpenBSD and NetBSD@. This option is overridden when @option{-march}
25632 indicates that the target CPU always has an FPU and so the
25633 instruction does not need emulation. These
25634 instructions are not generated unless you also use the
25635 @option{-funsafe-math-optimizations} switch.
25637 @item -malign-double
25638 @itemx -mno-align-double
25639 @opindex malign-double
25640 @opindex mno-align-double
25641 Control whether GCC aligns @code{double}, @code{long double}, and
25642 @code{long long} variables on a two-word boundary or a one-word
25643 boundary. Aligning @code{double} variables on a two-word boundary
25644 produces code that runs somewhat faster on a Pentium at the
25645 expense of more memory.
25647 On x86-64, @option{-malign-double} is enabled by default.
25649 @strong{Warning:} if you use the @option{-malign-double} switch,
25650 structures containing the above types are aligned differently than
25651 the published application binary interface specifications for the x86-32
25652 and are not binary compatible with structures in code compiled
25653 without that switch.
25655 @item -m96bit-long-double
25656 @itemx -m128bit-long-double
25657 @opindex m96bit-long-double
25658 @opindex m128bit-long-double
25659 These switches control the size of @code{long double} type. The x86-32
25660 application binary interface specifies the size to be 96 bits,
25661 so @option{-m96bit-long-double} is the default in 32-bit mode.
25663 Modern architectures (Pentium and newer) prefer @code{long double}
25664 to be aligned to an 8- or 16-byte boundary. In arrays or structures
25665 conforming to the ABI, this is not possible. So specifying
25666 @option{-m128bit-long-double} aligns @code{long double}
25667 to a 16-byte boundary by padding the @code{long double} with an additional
25670 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
25671 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
25673 Notice that neither of these options enable any extra precision over the x87
25674 standard of 80 bits for a @code{long double}.
25676 @strong{Warning:} if you override the default value for your target ABI, this
25677 changes the size of
25678 structures and arrays containing @code{long double} variables,
25679 as well as modifying the function calling convention for functions taking
25680 @code{long double}. Hence they are not binary-compatible
25681 with code compiled without that switch.
25683 @item -mlong-double-64
25684 @itemx -mlong-double-80
25685 @itemx -mlong-double-128
25686 @opindex mlong-double-64
25687 @opindex mlong-double-80
25688 @opindex mlong-double-128
25689 These switches control the size of @code{long double} type. A size
25690 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25691 type. This is the default for 32-bit Bionic C library. A size
25692 of 128 bits makes the @code{long double} type equivalent to the
25693 @code{__float128} type. This is the default for 64-bit Bionic C library.
25695 @strong{Warning:} if you override the default value for your target ABI, this
25696 changes the size of
25697 structures and arrays containing @code{long double} variables,
25698 as well as modifying the function calling convention for functions taking
25699 @code{long double}. Hence they are not binary-compatible
25700 with code compiled without that switch.
25702 @item -malign-data=@var{type}
25703 @opindex malign-data
25704 Control how GCC aligns variables. Supported values for @var{type} are
25705 @samp{compat} uses increased alignment value compatible uses GCC 4.8
25706 and earlier, @samp{abi} uses alignment value as specified by the
25707 psABI, and @samp{cacheline} uses increased alignment value to match
25708 the cache line size. @samp{compat} is the default.
25710 @item -mlarge-data-threshold=@var{threshold}
25711 @opindex mlarge-data-threshold
25712 When @option{-mcmodel=medium} is specified, data objects larger than
25713 @var{threshold} are placed in the large data section. This value must be the
25714 same across all objects linked into the binary, and defaults to 65535.
25718 Use a different function-calling convention, in which functions that
25719 take a fixed number of arguments return with the @code{ret @var{num}}
25720 instruction, which pops their arguments while returning. This saves one
25721 instruction in the caller since there is no need to pop the arguments
25724 You can specify that an individual function is called with this calling
25725 sequence with the function attribute @code{stdcall}. You can also
25726 override the @option{-mrtd} option by using the function attribute
25727 @code{cdecl}. @xref{Function Attributes}.
25729 @strong{Warning:} this calling convention is incompatible with the one
25730 normally used on Unix, so you cannot use it if you need to call
25731 libraries compiled with the Unix compiler.
25733 Also, you must provide function prototypes for all functions that
25734 take variable numbers of arguments (including @code{printf});
25735 otherwise incorrect code is generated for calls to those
25738 In addition, seriously incorrect code results if you call a
25739 function with too many arguments. (Normally, extra arguments are
25740 harmlessly ignored.)
25742 @item -mregparm=@var{num}
25744 Control how many registers are used to pass integer arguments. By
25745 default, no registers are used to pass arguments, and at most 3
25746 registers can be used. You can control this behavior for a specific
25747 function by using the function attribute @code{regparm}.
25748 @xref{Function Attributes}.
25750 @strong{Warning:} if you use this switch, and
25751 @var{num} is nonzero, then you must build all modules with the same
25752 value, including any libraries. This includes the system libraries and
25756 @opindex msseregparm
25757 Use SSE register passing conventions for float and double arguments
25758 and return values. You can control this behavior for a specific
25759 function by using the function attribute @code{sseregparm}.
25760 @xref{Function Attributes}.
25762 @strong{Warning:} if you use this switch then you must build all
25763 modules with the same value, including any libraries. This includes
25764 the system libraries and startup modules.
25766 @item -mvect8-ret-in-mem
25767 @opindex mvect8-ret-in-mem
25768 Return 8-byte vectors in memory instead of MMX registers. This is the
25769 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
25770 Studio compilers until version 12. Later compiler versions (starting
25771 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
25772 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
25773 you need to remain compatible with existing code produced by those
25774 previous compiler versions or older versions of GCC@.
25783 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
25784 is specified, the significands of results of floating-point operations are
25785 rounded to 24 bits (single precision); @option{-mpc64} rounds the
25786 significands of results of floating-point operations to 53 bits (double
25787 precision) and @option{-mpc80} rounds the significands of results of
25788 floating-point operations to 64 bits (extended double precision), which is
25789 the default. When this option is used, floating-point operations in higher
25790 precisions are not available to the programmer without setting the FPU
25791 control word explicitly.
25793 Setting the rounding of floating-point operations to less than the default
25794 80 bits can speed some programs by 2% or more. Note that some mathematical
25795 libraries assume that extended-precision (80-bit) floating-point operations
25796 are enabled by default; routines in such libraries could suffer significant
25797 loss of accuracy, typically through so-called ``catastrophic cancellation'',
25798 when this option is used to set the precision to less than extended precision.
25800 @item -mstackrealign
25801 @opindex mstackrealign
25802 Realign the stack at entry. On the x86, the @option{-mstackrealign}
25803 option generates an alternate prologue and epilogue that realigns the
25804 run-time stack if necessary. This supports mixing legacy codes that keep
25805 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
25806 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
25807 applicable to individual functions.
25809 @item -mpreferred-stack-boundary=@var{num}
25810 @opindex mpreferred-stack-boundary
25811 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
25812 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
25813 the default is 4 (16 bytes or 128 bits).
25815 @strong{Warning:} When generating code for the x86-64 architecture with
25816 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
25817 used to keep the stack boundary aligned to 8 byte boundary. Since
25818 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
25819 intended to be used in controlled environment where stack space is
25820 important limitation. This option leads to wrong code when functions
25821 compiled with 16 byte stack alignment (such as functions from a standard
25822 library) are called with misaligned stack. In this case, SSE
25823 instructions may lead to misaligned memory access traps. In addition,
25824 variable arguments are handled incorrectly for 16 byte aligned
25825 objects (including x87 long double and __int128), leading to wrong
25826 results. You must build all modules with
25827 @option{-mpreferred-stack-boundary=3}, including any libraries. This
25828 includes the system libraries and startup modules.
25830 @item -mincoming-stack-boundary=@var{num}
25831 @opindex mincoming-stack-boundary
25832 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
25833 boundary. If @option{-mincoming-stack-boundary} is not specified,
25834 the one specified by @option{-mpreferred-stack-boundary} is used.
25836 On Pentium and Pentium Pro, @code{double} and @code{long double} values
25837 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
25838 suffer significant run time performance penalties. On Pentium III, the
25839 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
25840 properly if it is not 16-byte aligned.
25842 To ensure proper alignment of this values on the stack, the stack boundary
25843 must be as aligned as that required by any value stored on the stack.
25844 Further, every function must be generated such that it keeps the stack
25845 aligned. Thus calling a function compiled with a higher preferred
25846 stack boundary from a function compiled with a lower preferred stack
25847 boundary most likely misaligns the stack. It is recommended that
25848 libraries that use callbacks always use the default setting.
25850 This extra alignment does consume extra stack space, and generally
25851 increases code size. Code that is sensitive to stack space usage, such
25852 as embedded systems and operating system kernels, may want to reduce the
25853 preferred alignment to @option{-mpreferred-stack-boundary=2}.
25910 @itemx -mavx512ifma
25911 @opindex mavx512ifma
25913 @itemx -mavx512vbmi
25914 @opindex mavx512vbmi
25926 @opindex mclfushopt
25943 @itemx -mprefetchwt1
25944 @opindex mprefetchwt1
26008 @itemx -mavx512vbmi2
26009 @opindex mavx512vbmi2
26013 These switches enable the use of instructions in the MMX, SSE,
26014 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
26015 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
26016 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, BMI, BMI2,
26017 FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, IBT, SHSTK, AVX512VBMI2,
26018 GFNI, 3DNow!@: or enhanced 3DNow!@: extended instruction sets. Each has a
26019 corresponding @option{-mno-} option to disable use of these instructions.
26021 These extensions are also available as built-in functions: see
26022 @ref{x86 Built-in Functions}, for details of the functions enabled and
26023 disabled by these switches.
26025 To generate SSE/SSE2 instructions automatically from floating-point
26026 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
26028 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
26029 generates new AVX instructions or AVX equivalence for all SSEx instructions
26032 These options enable GCC to use these extended instructions in
26033 generated code, even without @option{-mfpmath=sse}. Applications that
26034 perform run-time CPU detection must compile separate files for each
26035 supported architecture, using the appropriate flags. In particular,
26036 the file containing the CPU detection code should be compiled without
26039 The @option{-mcet} option turns on the @option{-mibt} and @option{-mshstk}
26040 options. The @option{-mibt} option enables indirect branch tracking support
26041 and the @option{-mshstk} option enables shadow stack support from
26042 Intel Control-flow Enforcement Technology (CET). The compiler also provides
26043 a number of built-in functions for fine-grained control in a CET-based
26044 application. See @xref{x86 Built-in Functions}, for more information.
26046 @item -mdump-tune-features
26047 @opindex mdump-tune-features
26048 This option instructs GCC to dump the names of the x86 performance
26049 tuning features and default settings. The names can be used in
26050 @option{-mtune-ctrl=@var{feature-list}}.
26052 @item -mtune-ctrl=@var{feature-list}
26053 @opindex mtune-ctrl=@var{feature-list}
26054 This option is used to do fine grain control of x86 code generation features.
26055 @var{feature-list} is a comma separated list of @var{feature} names. See also
26056 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
26057 on if it is not preceded with @samp{^}, otherwise, it is turned off.
26058 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
26059 developers. Using it may lead to code paths not covered by testing and can
26060 potentially result in compiler ICEs or runtime errors.
26063 @opindex mno-default
26064 This option instructs GCC to turn off all tunable features. See also
26065 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
26069 This option instructs GCC to emit a @code{cld} instruction in the prologue
26070 of functions that use string instructions. String instructions depend on
26071 the DF flag to select between autoincrement or autodecrement mode. While the
26072 ABI specifies the DF flag to be cleared on function entry, some operating
26073 systems violate this specification by not clearing the DF flag in their
26074 exception dispatchers. The exception handler can be invoked with the DF flag
26075 set, which leads to wrong direction mode when string instructions are used.
26076 This option can be enabled by default on 32-bit x86 targets by configuring
26077 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
26078 instructions can be suppressed with the @option{-mno-cld} compiler option
26082 @opindex mvzeroupper
26083 This option instructs GCC to emit a @code{vzeroupper} instruction
26084 before a transfer of control flow out of the function to minimize
26085 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
26088 @item -mprefer-avx128
26089 @opindex mprefer-avx128
26090 This option instructs GCC to use 128-bit AVX instructions instead of
26091 256-bit AVX instructions in the auto-vectorizer.
26093 @item -mprefer-vector-width=@var{opt}
26094 @opindex mprefer-vector-width
26095 This option instructs GCC to use @var{opt}-bit vector width in instructions
26096 instead of default on the selected platform.
26100 No extra limitations applied to GCC other than defined by the selected platform.
26103 Prefer 128-bit vector width for instructions.
26106 Prefer 256-bit vector width for instructions.
26109 Prefer 512-bit vector width for instructions.
26114 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
26115 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
26116 objects. This is useful for atomic updates of data structures exceeding one
26117 machine word in size. The compiler uses this instruction to implement
26118 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
26119 128-bit integers, a library call is always used.
26123 This option enables generation of @code{SAHF} instructions in 64-bit code.
26124 Early Intel Pentium 4 CPUs with Intel 64 support,
26125 prior to the introduction of Pentium 4 G1 step in December 2005,
26126 lacked the @code{LAHF} and @code{SAHF} instructions
26127 which are supported by AMD64.
26128 These are load and store instructions, respectively, for certain status flags.
26129 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
26130 @code{drem}, and @code{remainder} built-in functions;
26131 see @ref{Other Builtins} for details.
26135 This option enables use of the @code{movbe} instruction to implement
26136 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
26140 This option tells the compiler to use indirect branch tracking support
26141 (for indirect calls and jumps) from x86 Control-flow Enforcement
26142 Technology (CET). The option has effect only if the
26143 @option{-fcf-protection=full} or @option{-fcf-protection=branch} option
26144 is specified. The option @option{-mibt} is on by default when the
26145 @code{-mcet} option is specified.
26149 This option tells the compiler to use shadow stack support (return
26150 address tracking) from x86 Control-flow Enforcement Technology (CET).
26151 The option has effect only if the @option{-fcf-protection=full} or
26152 @option{-fcf-protection=return} option is specified. The option
26153 @option{-mshstk} is on by default when the @option{-mcet} option is
26158 This option enables built-in functions @code{__builtin_ia32_crc32qi},
26159 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
26160 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
26164 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
26165 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
26166 with an additional Newton-Raphson step
26167 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
26168 (and their vectorized
26169 variants) for single-precision floating-point arguments. These instructions
26170 are generated only when @option{-funsafe-math-optimizations} is enabled
26171 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
26172 Note that while the throughput of the sequence is higher than the throughput
26173 of the non-reciprocal instruction, the precision of the sequence can be
26174 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
26176 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
26177 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
26178 combination), and doesn't need @option{-mrecip}.
26180 Also note that GCC emits the above sequence with additional Newton-Raphson step
26181 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
26182 already with @option{-ffast-math} (or the above option combination), and
26183 doesn't need @option{-mrecip}.
26185 @item -mrecip=@var{opt}
26186 @opindex mrecip=opt
26187 This option controls which reciprocal estimate instructions
26188 may be used. @var{opt} is a comma-separated list of options, which may
26189 be preceded by a @samp{!} to invert the option:
26193 Enable all estimate instructions.
26196 Enable the default instructions, equivalent to @option{-mrecip}.
26199 Disable all estimate instructions, equivalent to @option{-mno-recip}.
26202 Enable the approximation for scalar division.
26205 Enable the approximation for vectorized division.
26208 Enable the approximation for scalar square root.
26211 Enable the approximation for vectorized square root.
26214 So, for example, @option{-mrecip=all,!sqrt} enables
26215 all of the reciprocal approximations, except for square root.
26217 @item -mveclibabi=@var{type}
26218 @opindex mveclibabi
26219 Specifies the ABI type to use for vectorizing intrinsics using an
26220 external library. Supported values for @var{type} are @samp{svml}
26221 for the Intel short
26222 vector math library and @samp{acml} for the AMD math core library.
26223 To use this option, both @option{-ftree-vectorize} and
26224 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
26225 ABI-compatible library must be specified at link time.
26227 GCC currently emits calls to @code{vmldExp2},
26228 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
26229 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
26230 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
26231 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
26232 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
26233 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
26234 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
26235 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
26236 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
26237 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
26238 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
26239 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
26240 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
26241 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
26242 when @option{-mveclibabi=acml} is used.
26244 @item -mabi=@var{name}
26246 Generate code for the specified calling convention. Permissible values
26247 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
26248 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
26249 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
26250 You can control this behavior for specific functions by
26251 using the function attributes @code{ms_abi} and @code{sysv_abi}.
26252 @xref{Function Attributes}.
26254 @item -mforce-indirect-call
26255 @opindex mforce-indirect-call
26256 Force all calls to functions to be indirect. This is useful
26257 when using Intel Processor Trace where it generates more precise timing
26258 information for function calls.
26260 @item -mcall-ms2sysv-xlogues
26261 @opindex mcall-ms2sysv-xlogues
26262 @opindex mno-call-ms2sysv-xlogues
26263 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
26264 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
26265 default, the code for saving and restoring these registers is emitted inline,
26266 resulting in fairly lengthy prologues and epilogues. Using
26267 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
26268 use stubs in the static portion of libgcc to perform these saves and restores,
26269 thus reducing function size at the cost of a few extra instructions.
26271 @item -mtls-dialect=@var{type}
26272 @opindex mtls-dialect
26273 Generate code to access thread-local storage using the @samp{gnu} or
26274 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
26275 @samp{gnu2} is more efficient, but it may add compile- and run-time
26276 requirements that cannot be satisfied on all systems.
26279 @itemx -mno-push-args
26280 @opindex mpush-args
26281 @opindex mno-push-args
26282 Use PUSH operations to store outgoing parameters. This method is shorter
26283 and usually equally fast as method using SUB/MOV operations and is enabled
26284 by default. In some cases disabling it may improve performance because of
26285 improved scheduling and reduced dependencies.
26287 @item -maccumulate-outgoing-args
26288 @opindex maccumulate-outgoing-args
26289 If enabled, the maximum amount of space required for outgoing arguments is
26290 computed in the function prologue. This is faster on most modern CPUs
26291 because of reduced dependencies, improved scheduling and reduced stack usage
26292 when the preferred stack boundary is not equal to 2. The drawback is a notable
26293 increase in code size. This switch implies @option{-mno-push-args}.
26297 Support thread-safe exception handling on MinGW. Programs that rely
26298 on thread-safe exception handling must compile and link all code with the
26299 @option{-mthreads} option. When compiling, @option{-mthreads} defines
26300 @option{-D_MT}; when linking, it links in a special thread helper library
26301 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
26303 @item -mms-bitfields
26304 @itemx -mno-ms-bitfields
26305 @opindex mms-bitfields
26306 @opindex mno-ms-bitfields
26308 Enable/disable bit-field layout compatible with the native Microsoft
26311 If @code{packed} is used on a structure, or if bit-fields are used,
26312 it may be that the Microsoft ABI lays out the structure differently
26313 than the way GCC normally does. Particularly when moving packed
26314 data between functions compiled with GCC and the native Microsoft compiler
26315 (either via function call or as data in a file), it may be necessary to access
26318 This option is enabled by default for Microsoft Windows
26319 targets. This behavior can also be controlled locally by use of variable
26320 or type attributes. For more information, see @ref{x86 Variable Attributes}
26321 and @ref{x86 Type Attributes}.
26323 The Microsoft structure layout algorithm is fairly simple with the exception
26324 of the bit-field packing.
26325 The padding and alignment of members of structures and whether a bit-field
26326 can straddle a storage-unit boundary are determine by these rules:
26329 @item Structure members are stored sequentially in the order in which they are
26330 declared: the first member has the lowest memory address and the last member
26333 @item Every data object has an alignment requirement. The alignment requirement
26334 for all data except structures, unions, and arrays is either the size of the
26335 object or the current packing size (specified with either the
26336 @code{aligned} attribute or the @code{pack} pragma),
26337 whichever is less. For structures, unions, and arrays,
26338 the alignment requirement is the largest alignment requirement of its members.
26339 Every object is allocated an offset so that:
26342 offset % alignment_requirement == 0
26345 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
26346 unit if the integral types are the same size and if the next bit-field fits
26347 into the current allocation unit without crossing the boundary imposed by the
26348 common alignment requirements of the bit-fields.
26351 MSVC interprets zero-length bit-fields in the following ways:
26354 @item If a zero-length bit-field is inserted between two bit-fields that
26355 are normally coalesced, the bit-fields are not coalesced.
26362 unsigned long bf_1 : 12;
26364 unsigned long bf_2 : 12;
26369 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
26370 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
26372 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
26373 alignment of the zero-length bit-field is greater than the member that follows it,
26374 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
26395 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
26396 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
26397 bit-field does not affect the alignment of @code{bar} or, as a result, the size
26400 Taking this into account, it is important to note the following:
26403 @item If a zero-length bit-field follows a normal bit-field, the type of the
26404 zero-length bit-field may affect the alignment of the structure as whole. For
26405 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
26406 normal bit-field, and is of type short.
26408 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
26409 still affect the alignment of the structure:
26420 Here, @code{t4} takes up 4 bytes.
26423 @item Zero-length bit-fields following non-bit-field members are ignored:
26435 Here, @code{t5} takes up 2 bytes.
26439 @item -mno-align-stringops
26440 @opindex mno-align-stringops
26441 Do not align the destination of inlined string operations. This switch reduces
26442 code size and improves performance in case the destination is already aligned,
26443 but GCC doesn't know about it.
26445 @item -minline-all-stringops
26446 @opindex minline-all-stringops
26447 By default GCC inlines string operations only when the destination is
26448 known to be aligned to least a 4-byte boundary.
26449 This enables more inlining and increases code
26450 size, but may improve performance of code that depends on fast
26451 @code{memcpy}, @code{strlen},
26452 and @code{memset} for short lengths.
26454 @item -minline-stringops-dynamically
26455 @opindex minline-stringops-dynamically
26456 For string operations of unknown size, use run-time checks with
26457 inline code for small blocks and a library call for large blocks.
26459 @item -mstringop-strategy=@var{alg}
26460 @opindex mstringop-strategy=@var{alg}
26461 Override the internal decision heuristic for the particular algorithm to use
26462 for inlining string operations. The allowed values for @var{alg} are:
26468 Expand using i386 @code{rep} prefix of the specified size.
26472 @itemx unrolled_loop
26473 Expand into an inline loop.
26476 Always use a library call.
26479 @item -mmemcpy-strategy=@var{strategy}
26480 @opindex mmemcpy-strategy=@var{strategy}
26481 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
26482 should be inlined and what inline algorithm to use when the expected size
26483 of the copy operation is known. @var{strategy}
26484 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
26485 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
26486 the max byte size with which inline algorithm @var{alg} is allowed. For the last
26487 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
26488 in the list must be specified in increasing order. The minimal byte size for
26489 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
26492 @item -mmemset-strategy=@var{strategy}
26493 @opindex mmemset-strategy=@var{strategy}
26494 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
26495 @code{__builtin_memset} expansion.
26497 @item -momit-leaf-frame-pointer
26498 @opindex momit-leaf-frame-pointer
26499 Don't keep the frame pointer in a register for leaf functions. This
26500 avoids the instructions to save, set up, and restore frame pointers and
26501 makes an extra register available in leaf functions. The option
26502 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
26503 which might make debugging harder.
26505 @item -mtls-direct-seg-refs
26506 @itemx -mno-tls-direct-seg-refs
26507 @opindex mtls-direct-seg-refs
26508 Controls whether TLS variables may be accessed with offsets from the
26509 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
26510 or whether the thread base pointer must be added. Whether or not this
26511 is valid depends on the operating system, and whether it maps the
26512 segment to cover the entire TLS area.
26514 For systems that use the GNU C Library, the default is on.
26517 @itemx -mno-sse2avx
26519 Specify that the assembler should encode SSE instructions with VEX
26520 prefix. The option @option{-mavx} turns this on by default.
26525 If profiling is active (@option{-pg}), put the profiling
26526 counter call before the prologue.
26527 Note: On x86 architectures the attribute @code{ms_hook_prologue}
26528 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
26530 @item -mrecord-mcount
26531 @itemx -mno-record-mcount
26532 @opindex mrecord-mcount
26533 If profiling is active (@option{-pg}), generate a __mcount_loc section
26534 that contains pointers to each profiling call. This is useful for
26535 automatically patching and out calls.
26538 @itemx -mno-nop-mcount
26539 @opindex mnop-mcount
26540 If profiling is active (@option{-pg}), generate the calls to
26541 the profiling functions as NOPs. This is useful when they
26542 should be patched in later dynamically. This is likely only
26543 useful together with @option{-mrecord-mcount}.
26545 @item -mskip-rax-setup
26546 @itemx -mno-skip-rax-setup
26547 @opindex mskip-rax-setup
26548 When generating code for the x86-64 architecture with SSE extensions
26549 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
26550 register when there are no variable arguments passed in vector registers.
26552 @strong{Warning:} Since RAX register is used to avoid unnecessarily
26553 saving vector registers on stack when passing variable arguments, the
26554 impacts of this option are callees may waste some stack space,
26555 misbehave or jump to a random location. GCC 4.4 or newer don't have
26556 those issues, regardless the RAX register value.
26559 @itemx -mno-8bit-idiv
26560 @opindex m8bit-idiv
26561 On some processors, like Intel Atom, 8-bit unsigned integer divide is
26562 much faster than 32-bit/64-bit integer divide. This option generates a
26563 run-time check. If both dividend and divisor are within range of 0
26564 to 255, 8-bit unsigned integer divide is used instead of
26565 32-bit/64-bit integer divide.
26567 @item -mavx256-split-unaligned-load
26568 @itemx -mavx256-split-unaligned-store
26569 @opindex mavx256-split-unaligned-load
26570 @opindex mavx256-split-unaligned-store
26571 Split 32-byte AVX unaligned load and store.
26573 @item -mstack-protector-guard=@var{guard}
26574 @itemx -mstack-protector-guard-reg=@var{reg}
26575 @itemx -mstack-protector-guard-offset=@var{offset}
26576 @opindex mstack-protector-guard
26577 @opindex mstack-protector-guard-reg
26578 @opindex mstack-protector-guard-offset
26579 Generate stack protection code using canary at @var{guard}. Supported
26580 locations are @samp{global} for global canary or @samp{tls} for per-thread
26581 canary in the TLS block (the default). This option has effect only when
26582 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
26584 With the latter choice the options
26585 @option{-mstack-protector-guard-reg=@var{reg}} and
26586 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
26587 which segment register (@code{%fs} or @code{%gs}) to use as base register
26588 for reading the canary, and from what offset from that base register.
26589 The default for those is as specified in the relevant ABI.
26591 @item -mmitigate-rop
26592 @opindex mmitigate-rop
26593 Try to avoid generating code sequences that contain unintended return
26594 opcodes, to mitigate against certain forms of attack. At the moment,
26595 this option is limited in what it can do and should not be relied
26596 on to provide serious protection.
26598 @item -mgeneral-regs-only
26599 @opindex mgeneral-regs-only
26600 Generate code that uses only the general-purpose registers. This
26601 prevents the compiler from using floating-point, vector, mask and bound
26606 These @samp{-m} switches are supported in addition to the above
26607 on x86-64 processors in 64-bit environments.
26620 Generate code for a 16-bit, 32-bit or 64-bit environment.
26621 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
26623 generates code that runs on any i386 system.
26625 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
26626 types to 64 bits, and generates code for the x86-64 architecture.
26627 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
26628 and @option{-mdynamic-no-pic} options.
26630 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
26632 generates code for the x86-64 architecture.
26634 The @option{-m16} option is the same as @option{-m32}, except for that
26635 it outputs the @code{.code16gcc} assembly directive at the beginning of
26636 the assembly output so that the binary can run in 16-bit mode.
26638 The @option{-miamcu} option generates code which conforms to Intel MCU
26639 psABI. It requires the @option{-m32} option to be turned on.
26641 @item -mno-red-zone
26642 @opindex mno-red-zone
26643 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
26644 by the x86-64 ABI; it is a 128-byte area beyond the location of the
26645 stack pointer that is not modified by signal or interrupt handlers
26646 and therefore can be used for temporary data without adjusting the stack
26647 pointer. The flag @option{-mno-red-zone} disables this red zone.
26649 @item -mcmodel=small
26650 @opindex mcmodel=small
26651 Generate code for the small code model: the program and its symbols must
26652 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
26653 Programs can be statically or dynamically linked. This is the default
26656 @item -mcmodel=kernel
26657 @opindex mcmodel=kernel
26658 Generate code for the kernel code model. The kernel runs in the
26659 negative 2 GB of the address space.
26660 This model has to be used for Linux kernel code.
26662 @item -mcmodel=medium
26663 @opindex mcmodel=medium
26664 Generate code for the medium model: the program is linked in the lower 2
26665 GB of the address space. Small symbols are also placed there. Symbols
26666 with sizes larger than @option{-mlarge-data-threshold} are put into
26667 large data or BSS sections and can be located above 2GB. Programs can
26668 be statically or dynamically linked.
26670 @item -mcmodel=large
26671 @opindex mcmodel=large
26672 Generate code for the large model. This model makes no assumptions
26673 about addresses and sizes of sections.
26675 @item -maddress-mode=long
26676 @opindex maddress-mode=long
26677 Generate code for long address mode. This is only supported for 64-bit
26678 and x32 environments. It is the default address mode for 64-bit
26681 @item -maddress-mode=short
26682 @opindex maddress-mode=short
26683 Generate code for short address mode. This is only supported for 32-bit
26684 and x32 environments. It is the default address mode for 32-bit and
26688 @node x86 Windows Options
26689 @subsection x86 Windows Options
26690 @cindex x86 Windows Options
26691 @cindex Windows Options for x86
26693 These additional options are available for Microsoft Windows targets:
26699 specifies that a console application is to be generated, by
26700 instructing the linker to set the PE header subsystem type
26701 required for console applications.
26702 This option is available for Cygwin and MinGW targets and is
26703 enabled by default on those targets.
26707 This option is available for Cygwin and MinGW targets. It
26708 specifies that a DLL---a dynamic link library---is to be
26709 generated, enabling the selection of the required runtime
26710 startup object and entry point.
26712 @item -mnop-fun-dllimport
26713 @opindex mnop-fun-dllimport
26714 This option is available for Cygwin and MinGW targets. It
26715 specifies that the @code{dllimport} attribute should be ignored.
26719 This option is available for MinGW targets. It specifies
26720 that MinGW-specific thread support is to be used.
26724 This option is available for MinGW-w64 targets. It causes
26725 the @code{UNICODE} preprocessor macro to be predefined, and
26726 chooses Unicode-capable runtime startup code.
26730 This option is available for Cygwin and MinGW targets. It
26731 specifies that the typical Microsoft Windows predefined macros are to
26732 be set in the pre-processor, but does not influence the choice
26733 of runtime library/startup code.
26737 This option is available for Cygwin and MinGW targets. It
26738 specifies that a GUI application is to be generated by
26739 instructing the linker to set the PE header subsystem type
26742 @item -fno-set-stack-executable
26743 @opindex fno-set-stack-executable
26744 This option is available for MinGW targets. It specifies that
26745 the executable flag for the stack used by nested functions isn't
26746 set. This is necessary for binaries running in kernel mode of
26747 Microsoft Windows, as there the User32 API, which is used to set executable
26748 privileges, isn't available.
26750 @item -fwritable-relocated-rdata
26751 @opindex fno-writable-relocated-rdata
26752 This option is available for MinGW and Cygwin targets. It specifies
26753 that relocated-data in read-only section is put into the @code{.data}
26754 section. This is a necessary for older runtimes not supporting
26755 modification of @code{.rdata} sections for pseudo-relocation.
26757 @item -mpe-aligned-commons
26758 @opindex mpe-aligned-commons
26759 This option is available for Cygwin and MinGW targets. It
26760 specifies that the GNU extension to the PE file format that
26761 permits the correct alignment of COMMON variables should be
26762 used when generating code. It is enabled by default if
26763 GCC detects that the target assembler found during configuration
26764 supports the feature.
26767 See also under @ref{x86 Options} for standard options.
26769 @node Xstormy16 Options
26770 @subsection Xstormy16 Options
26771 @cindex Xstormy16 Options
26773 These options are defined for Xstormy16:
26778 Choose startup files and linker script suitable for the simulator.
26781 @node Xtensa Options
26782 @subsection Xtensa Options
26783 @cindex Xtensa Options
26785 These options are supported for Xtensa targets:
26789 @itemx -mno-const16
26791 @opindex mno-const16
26792 Enable or disable use of @code{CONST16} instructions for loading
26793 constant values. The @code{CONST16} instruction is currently not a
26794 standard option from Tensilica. When enabled, @code{CONST16}
26795 instructions are always used in place of the standard @code{L32R}
26796 instructions. The use of @code{CONST16} is enabled by default only if
26797 the @code{L32R} instruction is not available.
26800 @itemx -mno-fused-madd
26801 @opindex mfused-madd
26802 @opindex mno-fused-madd
26803 Enable or disable use of fused multiply/add and multiply/subtract
26804 instructions in the floating-point option. This has no effect if the
26805 floating-point option is not also enabled. Disabling fused multiply/add
26806 and multiply/subtract instructions forces the compiler to use separate
26807 instructions for the multiply and add/subtract operations. This may be
26808 desirable in some cases where strict IEEE 754-compliant results are
26809 required: the fused multiply add/subtract instructions do not round the
26810 intermediate result, thereby producing results with @emph{more} bits of
26811 precision than specified by the IEEE standard. Disabling fused multiply
26812 add/subtract instructions also ensures that the program output is not
26813 sensitive to the compiler's ability to combine multiply and add/subtract
26816 @item -mserialize-volatile
26817 @itemx -mno-serialize-volatile
26818 @opindex mserialize-volatile
26819 @opindex mno-serialize-volatile
26820 When this option is enabled, GCC inserts @code{MEMW} instructions before
26821 @code{volatile} memory references to guarantee sequential consistency.
26822 The default is @option{-mserialize-volatile}. Use
26823 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
26825 @item -mforce-no-pic
26826 @opindex mforce-no-pic
26827 For targets, like GNU/Linux, where all user-mode Xtensa code must be
26828 position-independent code (PIC), this option disables PIC for compiling
26831 @item -mtext-section-literals
26832 @itemx -mno-text-section-literals
26833 @opindex mtext-section-literals
26834 @opindex mno-text-section-literals
26835 These options control the treatment of literal pools. The default is
26836 @option{-mno-text-section-literals}, which places literals in a separate
26837 section in the output file. This allows the literal pool to be placed
26838 in a data RAM/ROM, and it also allows the linker to combine literal
26839 pools from separate object files to remove redundant literals and
26840 improve code size. With @option{-mtext-section-literals}, the literals
26841 are interspersed in the text section in order to keep them as close as
26842 possible to their references. This may be necessary for large assembly
26843 files. Literals for each function are placed right before that function.
26845 @item -mauto-litpools
26846 @itemx -mno-auto-litpools
26847 @opindex mauto-litpools
26848 @opindex mno-auto-litpools
26849 These options control the treatment of literal pools. The default is
26850 @option{-mno-auto-litpools}, which places literals in a separate
26851 section in the output file unless @option{-mtext-section-literals} is
26852 used. With @option{-mauto-litpools} the literals are interspersed in
26853 the text section by the assembler. Compiler does not produce explicit
26854 @code{.literal} directives and loads literals into registers with
26855 @code{MOVI} instructions instead of @code{L32R} to let the assembler
26856 do relaxation and place literals as necessary. This option allows
26857 assembler to create several literal pools per function and assemble
26858 very big functions, which may not be possible with
26859 @option{-mtext-section-literals}.
26861 @item -mtarget-align
26862 @itemx -mno-target-align
26863 @opindex mtarget-align
26864 @opindex mno-target-align
26865 When this option is enabled, GCC instructs the assembler to
26866 automatically align instructions to reduce branch penalties at the
26867 expense of some code density. The assembler attempts to widen density
26868 instructions to align branch targets and the instructions following call
26869 instructions. If there are not enough preceding safe density
26870 instructions to align a target, no widening is performed. The
26871 default is @option{-mtarget-align}. These options do not affect the
26872 treatment of auto-aligned instructions like @code{LOOP}, which the
26873 assembler always aligns, either by widening density instructions or
26874 by inserting NOP instructions.
26877 @itemx -mno-longcalls
26878 @opindex mlongcalls
26879 @opindex mno-longcalls
26880 When this option is enabled, GCC instructs the assembler to translate
26881 direct calls to indirect calls unless it can determine that the target
26882 of a direct call is in the range allowed by the call instruction. This
26883 translation typically occurs for calls to functions in other source
26884 files. Specifically, the assembler translates a direct @code{CALL}
26885 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
26886 The default is @option{-mno-longcalls}. This option should be used in
26887 programs where the call target can potentially be out of range. This
26888 option is implemented in the assembler, not the compiler, so the
26889 assembly code generated by GCC still shows direct call
26890 instructions---look at the disassembled object code to see the actual
26891 instructions. Note that the assembler uses an indirect call for
26892 every cross-file call, not just those that really are out of range.
26895 @node zSeries Options
26896 @subsection zSeries Options
26897 @cindex zSeries options
26899 These are listed under @xref{S/390 and zSeries Options}.
26905 @section Specifying Subprocesses and the Switches to Pass to Them
26908 @command{gcc} is a driver program. It performs its job by invoking a
26909 sequence of other programs to do the work of compiling, assembling and
26910 linking. GCC interprets its command-line parameters and uses these to
26911 deduce which programs it should invoke, and which command-line options
26912 it ought to place on their command lines. This behavior is controlled
26913 by @dfn{spec strings}. In most cases there is one spec string for each
26914 program that GCC can invoke, but a few programs have multiple spec
26915 strings to control their behavior. The spec strings built into GCC can
26916 be overridden by using the @option{-specs=} command-line switch to specify
26919 @dfn{Spec files} are plain-text files that are used to construct spec
26920 strings. They consist of a sequence of directives separated by blank
26921 lines. The type of directive is determined by the first non-whitespace
26922 character on the line, which can be one of the following:
26925 @item %@var{command}
26926 Issues a @var{command} to the spec file processor. The commands that can
26930 @item %include <@var{file}>
26931 @cindex @code{%include}
26932 Search for @var{file} and insert its text at the current point in the
26935 @item %include_noerr <@var{file}>
26936 @cindex @code{%include_noerr}
26937 Just like @samp{%include}, but do not generate an error message if the include
26938 file cannot be found.
26940 @item %rename @var{old_name} @var{new_name}
26941 @cindex @code{%rename}
26942 Rename the spec string @var{old_name} to @var{new_name}.
26946 @item *[@var{spec_name}]:
26947 This tells the compiler to create, override or delete the named spec
26948 string. All lines after this directive up to the next directive or
26949 blank line are considered to be the text for the spec string. If this
26950 results in an empty string then the spec is deleted. (Or, if the
26951 spec did not exist, then nothing happens.) Otherwise, if the spec
26952 does not currently exist a new spec is created. If the spec does
26953 exist then its contents are overridden by the text of this
26954 directive, unless the first character of that text is the @samp{+}
26955 character, in which case the text is appended to the spec.
26957 @item [@var{suffix}]:
26958 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
26959 and up to the next directive or blank line are considered to make up the
26960 spec string for the indicated suffix. When the compiler encounters an
26961 input file with the named suffix, it processes the spec string in
26962 order to work out how to compile that file. For example:
26966 z-compile -input %i
26969 This says that any input file whose name ends in @samp{.ZZ} should be
26970 passed to the program @samp{z-compile}, which should be invoked with the
26971 command-line switch @option{-input} and with the result of performing the
26972 @samp{%i} substitution. (See below.)
26974 As an alternative to providing a spec string, the text following a
26975 suffix directive can be one of the following:
26978 @item @@@var{language}
26979 This says that the suffix is an alias for a known @var{language}. This is
26980 similar to using the @option{-x} command-line switch to GCC to specify a
26981 language explicitly. For example:
26988 Says that .ZZ files are, in fact, C++ source files.
26991 This causes an error messages saying:
26994 @var{name} compiler not installed on this system.
26998 GCC already has an extensive list of suffixes built into it.
26999 This directive adds an entry to the end of the list of suffixes, but
27000 since the list is searched from the end backwards, it is effectively
27001 possible to override earlier entries using this technique.
27005 GCC has the following spec strings built into it. Spec files can
27006 override these strings or create their own. Note that individual
27007 targets can also add their own spec strings to this list.
27010 asm Options to pass to the assembler
27011 asm_final Options to pass to the assembler post-processor
27012 cpp Options to pass to the C preprocessor
27013 cc1 Options to pass to the C compiler
27014 cc1plus Options to pass to the C++ compiler
27015 endfile Object files to include at the end of the link
27016 link Options to pass to the linker
27017 lib Libraries to include on the command line to the linker
27018 libgcc Decides which GCC support library to pass to the linker
27019 linker Sets the name of the linker
27020 predefines Defines to be passed to the C preprocessor
27021 signed_char Defines to pass to CPP to say whether @code{char} is signed
27023 startfile Object files to include at the start of the link
27026 Here is a small example of a spec file:
27029 %rename lib old_lib
27032 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
27035 This example renames the spec called @samp{lib} to @samp{old_lib} and
27036 then overrides the previous definition of @samp{lib} with a new one.
27037 The new definition adds in some extra command-line options before
27038 including the text of the old definition.
27040 @dfn{Spec strings} are a list of command-line options to be passed to their
27041 corresponding program. In addition, the spec strings can contain
27042 @samp{%}-prefixed sequences to substitute variable text or to
27043 conditionally insert text into the command line. Using these constructs
27044 it is possible to generate quite complex command lines.
27046 Here is a table of all defined @samp{%}-sequences for spec
27047 strings. Note that spaces are not generated automatically around the
27048 results of expanding these sequences. Therefore you can concatenate them
27049 together or combine them with constant text in a single argument.
27053 Substitute one @samp{%} into the program name or argument.
27056 Substitute the name of the input file being processed.
27059 Substitute the basename of the input file being processed.
27060 This is the substring up to (and not including) the last period
27061 and not including the directory.
27064 This is the same as @samp{%b}, but include the file suffix (text after
27068 Marks the argument containing or following the @samp{%d} as a
27069 temporary file name, so that that file is deleted if GCC exits
27070 successfully. Unlike @samp{%g}, this contributes no text to the
27073 @item %g@var{suffix}
27074 Substitute a file name that has suffix @var{suffix} and is chosen
27075 once per compilation, and mark the argument in the same way as
27076 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
27077 name is now chosen in a way that is hard to predict even when previously
27078 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
27079 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
27080 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
27081 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
27082 was simply substituted with a file name chosen once per compilation,
27083 without regard to any appended suffix (which was therefore treated
27084 just like ordinary text), making such attacks more likely to succeed.
27086 @item %u@var{suffix}
27087 Like @samp{%g}, but generates a new temporary file name
27088 each time it appears instead of once per compilation.
27090 @item %U@var{suffix}
27091 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
27092 new one if there is no such last file name. In the absence of any
27093 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
27094 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
27095 involves the generation of two distinct file names, one
27096 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
27097 simply substituted with a file name chosen for the previous @samp{%u},
27098 without regard to any appended suffix.
27100 @item %j@var{suffix}
27101 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
27102 writable, and if @option{-save-temps} is not used;
27103 otherwise, substitute the name
27104 of a temporary file, just like @samp{%u}. This temporary file is not
27105 meant for communication between processes, but rather as a junk
27106 disposal mechanism.
27108 @item %|@var{suffix}
27109 @itemx %m@var{suffix}
27110 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
27111 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
27112 all. These are the two most common ways to instruct a program that it
27113 should read from standard input or write to standard output. If you
27114 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
27115 construct: see for example @file{f/lang-specs.h}.
27117 @item %.@var{SUFFIX}
27118 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
27119 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
27120 terminated by the next space or %.
27123 Marks the argument containing or following the @samp{%w} as the
27124 designated output file of this compilation. This puts the argument
27125 into the sequence of arguments that @samp{%o} substitutes.
27128 Substitutes the names of all the output files, with spaces
27129 automatically placed around them. You should write spaces
27130 around the @samp{%o} as well or the results are undefined.
27131 @samp{%o} is for use in the specs for running the linker.
27132 Input files whose names have no recognized suffix are not compiled
27133 at all, but they are included among the output files, so they are
27137 Substitutes the suffix for object files. Note that this is
27138 handled specially when it immediately follows @samp{%g, %u, or %U},
27139 because of the need for those to form complete file names. The
27140 handling is such that @samp{%O} is treated exactly as if it had already
27141 been substituted, except that @samp{%g, %u, and %U} do not currently
27142 support additional @var{suffix} characters following @samp{%O} as they do
27143 following, for example, @samp{.o}.
27146 Substitutes the standard macro predefinitions for the
27147 current target machine. Use this when running @command{cpp}.
27150 Like @samp{%p}, but puts @samp{__} before and after the name of each
27151 predefined macro, except for macros that start with @samp{__} or with
27152 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
27156 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
27157 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
27158 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
27159 and @option{-imultilib} as necessary.
27162 Current argument is the name of a library or startup file of some sort.
27163 Search for that file in a standard list of directories and substitute
27164 the full name found. The current working directory is included in the
27165 list of directories scanned.
27168 Current argument is the name of a linker script. Search for that file
27169 in the current list of directories to scan for libraries. If the file
27170 is located insert a @option{--script} option into the command line
27171 followed by the full path name found. If the file is not found then
27172 generate an error message. Note: the current working directory is not
27176 Print @var{str} as an error message. @var{str} is terminated by a newline.
27177 Use this when inconsistent options are detected.
27179 @item %(@var{name})
27180 Substitute the contents of spec string @var{name} at this point.
27182 @item %x@{@var{option}@}
27183 Accumulate an option for @samp{%X}.
27186 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
27190 Output the accumulated assembler options specified by @option{-Wa}.
27193 Output the accumulated preprocessor options specified by @option{-Wp}.
27196 Process the @code{asm} spec. This is used to compute the
27197 switches to be passed to the assembler.
27200 Process the @code{asm_final} spec. This is a spec string for
27201 passing switches to an assembler post-processor, if such a program is
27205 Process the @code{link} spec. This is the spec for computing the
27206 command line passed to the linker. Typically it makes use of the
27207 @samp{%L %G %S %D and %E} sequences.
27210 Dump out a @option{-L} option for each directory that GCC believes might
27211 contain startup files. If the target supports multilibs then the
27212 current multilib directory is prepended to each of these paths.
27215 Process the @code{lib} spec. This is a spec string for deciding which
27216 libraries are included on the command line to the linker.
27219 Process the @code{libgcc} spec. This is a spec string for deciding
27220 which GCC support library is included on the command line to the linker.
27223 Process the @code{startfile} spec. This is a spec for deciding which
27224 object files are the first ones passed to the linker. Typically
27225 this might be a file named @file{crt0.o}.
27228 Process the @code{endfile} spec. This is a spec string that specifies
27229 the last object files that are passed to the linker.
27232 Process the @code{cpp} spec. This is used to construct the arguments
27233 to be passed to the C preprocessor.
27236 Process the @code{cc1} spec. This is used to construct the options to be
27237 passed to the actual C compiler (@command{cc1}).
27240 Process the @code{cc1plus} spec. This is used to construct the options to be
27241 passed to the actual C++ compiler (@command{cc1plus}).
27244 Substitute the variable part of a matched option. See below.
27245 Note that each comma in the substituted string is replaced by
27249 Remove all occurrences of @code{-S} from the command line. Note---this
27250 command is position dependent. @samp{%} commands in the spec string
27251 before this one see @code{-S}, @samp{%} commands in the spec string
27252 after this one do not.
27254 @item %:@var{function}(@var{args})
27255 Call the named function @var{function}, passing it @var{args}.
27256 @var{args} is first processed as a nested spec string, then split
27257 into an argument vector in the usual fashion. The function returns
27258 a string which is processed as if it had appeared literally as part
27259 of the current spec.
27261 The following built-in spec functions are provided:
27264 @item @code{getenv}
27265 The @code{getenv} spec function takes two arguments: an environment
27266 variable name and a string. If the environment variable is not
27267 defined, a fatal error is issued. Otherwise, the return value is the
27268 value of the environment variable concatenated with the string. For
27269 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
27272 %:getenv(TOPDIR /include)
27275 expands to @file{/path/to/top/include}.
27277 @item @code{if-exists}
27278 The @code{if-exists} spec function takes one argument, an absolute
27279 pathname to a file. If the file exists, @code{if-exists} returns the
27280 pathname. Here is a small example of its usage:
27284 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
27287 @item @code{if-exists-else}
27288 The @code{if-exists-else} spec function is similar to the @code{if-exists}
27289 spec function, except that it takes two arguments. The first argument is
27290 an absolute pathname to a file. If the file exists, @code{if-exists-else}
27291 returns the pathname. If it does not exist, it returns the second argument.
27292 This way, @code{if-exists-else} can be used to select one file or another,
27293 based on the existence of the first. Here is a small example of its usage:
27297 crt0%O%s %:if-exists(crti%O%s) \
27298 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
27301 @item @code{replace-outfile}
27302 The @code{replace-outfile} spec function takes two arguments. It looks for the
27303 first argument in the outfiles array and replaces it with the second argument. Here
27304 is a small example of its usage:
27307 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
27310 @item @code{remove-outfile}
27311 The @code{remove-outfile} spec function takes one argument. It looks for the
27312 first argument in the outfiles array and removes it. Here is a small example
27316 %:remove-outfile(-lm)
27319 @item @code{pass-through-libs}
27320 The @code{pass-through-libs} spec function takes any number of arguments. It
27321 finds any @option{-l} options and any non-options ending in @file{.a} (which it
27322 assumes are the names of linker input library archive files) and returns a
27323 result containing all the found arguments each prepended by
27324 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
27325 intended to be passed to the LTO linker plugin.
27328 %:pass-through-libs(%G %L %G)
27331 @item @code{print-asm-header}
27332 The @code{print-asm-header} function takes no arguments and simply
27333 prints a banner like:
27339 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
27342 It is used to separate compiler options from assembler options
27343 in the @option{--target-help} output.
27347 Substitutes the @code{-S} switch, if that switch is given to GCC@.
27348 If that switch is not specified, this substitutes nothing. Note that
27349 the leading dash is omitted when specifying this option, and it is
27350 automatically inserted if the substitution is performed. Thus the spec
27351 string @samp{%@{foo@}} matches the command-line option @option{-foo}
27352 and outputs the command-line option @option{-foo}.
27355 Like %@{@code{S}@} but mark last argument supplied within as a file to be
27356 deleted on failure.
27359 Substitutes all the switches specified to GCC whose names start
27360 with @code{-S}, but which also take an argument. This is used for
27361 switches like @option{-o}, @option{-D}, @option{-I}, etc.
27362 GCC considers @option{-o foo} as being
27363 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
27364 text, including the space. Thus two arguments are generated.
27367 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
27368 (the order of @code{S} and @code{T} in the spec is not significant).
27369 There can be any number of ampersand-separated variables; for each the
27370 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
27373 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
27376 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
27379 Substitutes @code{X} if one or more switches whose names start with
27380 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
27381 once, no matter how many such switches appeared. However, if @code{%*}
27382 appears somewhere in @code{X}, then @code{X} is substituted once
27383 for each matching switch, with the @code{%*} replaced by the part of
27384 that switch matching the @code{*}.
27386 If @code{%*} appears as the last part of a spec sequence then a space
27387 is added after the end of the last substitution. If there is more
27388 text in the sequence, however, then a space is not generated. This
27389 allows the @code{%*} substitution to be used as part of a larger
27390 string. For example, a spec string like this:
27393 %@{mcu=*:--script=%*/memory.ld@}
27397 when matching an option like @option{-mcu=newchip} produces:
27400 --script=newchip/memory.ld
27404 Substitutes @code{X}, if processing a file with suffix @code{S}.
27407 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
27410 Substitutes @code{X}, if processing a file for language @code{S}.
27413 Substitutes @code{X}, if not processing a file for language @code{S}.
27416 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
27417 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
27418 @code{*} sequences as well, although they have a stronger binding than
27419 the @samp{|}. If @code{%*} appears in @code{X}, all of the
27420 alternatives must be starred, and only the first matching alternative
27423 For example, a spec string like this:
27426 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
27430 outputs the following command-line options from the following input
27431 command-line options:
27436 -d fred.c -foo -baz -boggle
27437 -d jim.d -bar -baz -boggle
27440 @item %@{S:X; T:Y; :D@}
27442 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
27443 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
27444 be as many clauses as you need. This may be combined with @code{.},
27445 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
27450 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
27451 or similar construct can use a backslash to ignore the special meaning
27452 of the character following it, thus allowing literal matching of a
27453 character that is otherwise specially treated. For example,
27454 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
27455 @option{-std=iso9899:1999} option is given.
27457 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
27458 construct may contain other nested @samp{%} constructs or spaces, or
27459 even newlines. They are processed as usual, as described above.
27460 Trailing white space in @code{X} is ignored. White space may also
27461 appear anywhere on the left side of the colon in these constructs,
27462 except between @code{.} or @code{*} and the corresponding word.
27464 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
27465 handled specifically in these constructs. If another value of
27466 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
27467 @option{-W} switch is found later in the command line, the earlier
27468 switch value is ignored, except with @{@code{S}*@} where @code{S} is
27469 just one letter, which passes all matching options.
27471 The character @samp{|} at the beginning of the predicate text is used to
27472 indicate that a command should be piped to the following command, but
27473 only if @option{-pipe} is specified.
27475 It is built into GCC which switches take arguments and which do not.
27476 (You might think it would be useful to generalize this to allow each
27477 compiler's spec to say which switches take arguments. But this cannot
27478 be done in a consistent fashion. GCC cannot even decide which input
27479 files have been specified without knowing which switches take arguments,
27480 and it must know which input files to compile in order to tell which
27483 GCC also knows implicitly that arguments starting in @option{-l} are to be
27484 treated as compiler output files, and passed to the linker in their
27485 proper position among the other output files.
27487 @node Environment Variables
27488 @section Environment Variables Affecting GCC
27489 @cindex environment variables
27491 @c man begin ENVIRONMENT
27492 This section describes several environment variables that affect how GCC
27493 operates. Some of them work by specifying directories or prefixes to use
27494 when searching for various kinds of files. Some are used to specify other
27495 aspects of the compilation environment.
27497 Note that you can also specify places to search using options such as
27498 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
27499 take precedence over places specified using environment variables, which
27500 in turn take precedence over those specified by the configuration of GCC@.
27501 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
27502 GNU Compiler Collection (GCC) Internals}.
27507 @c @itemx LC_COLLATE
27509 @c @itemx LC_MONETARY
27510 @c @itemx LC_NUMERIC
27515 @c @findex LC_COLLATE
27516 @findex LC_MESSAGES
27517 @c @findex LC_MONETARY
27518 @c @findex LC_NUMERIC
27522 These environment variables control the way that GCC uses
27523 localization information which allows GCC to work with different
27524 national conventions. GCC inspects the locale categories
27525 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
27526 so. These locale categories can be set to any value supported by your
27527 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
27528 Kingdom encoded in UTF-8.
27530 The @env{LC_CTYPE} environment variable specifies character
27531 classification. GCC uses it to determine the character boundaries in
27532 a string; this is needed for some multibyte encodings that contain quote
27533 and escape characters that are otherwise interpreted as a string
27536 The @env{LC_MESSAGES} environment variable specifies the language to
27537 use in diagnostic messages.
27539 If the @env{LC_ALL} environment variable is set, it overrides the value
27540 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
27541 and @env{LC_MESSAGES} default to the value of the @env{LANG}
27542 environment variable. If none of these variables are set, GCC
27543 defaults to traditional C English behavior.
27547 If @env{TMPDIR} is set, it specifies the directory to use for temporary
27548 files. GCC uses temporary files to hold the output of one stage of
27549 compilation which is to be used as input to the next stage: for example,
27550 the output of the preprocessor, which is the input to the compiler
27553 @item GCC_COMPARE_DEBUG
27554 @findex GCC_COMPARE_DEBUG
27555 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
27556 @option{-fcompare-debug} to the compiler driver. See the documentation
27557 of this option for more details.
27559 @item GCC_EXEC_PREFIX
27560 @findex GCC_EXEC_PREFIX
27561 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
27562 names of the subprograms executed by the compiler. No slash is added
27563 when this prefix is combined with the name of a subprogram, but you can
27564 specify a prefix that ends with a slash if you wish.
27566 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
27567 an appropriate prefix to use based on the pathname it is invoked with.
27569 If GCC cannot find the subprogram using the specified prefix, it
27570 tries looking in the usual places for the subprogram.
27572 The default value of @env{GCC_EXEC_PREFIX} is
27573 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
27574 the installed compiler. In many cases @var{prefix} is the value
27575 of @code{prefix} when you ran the @file{configure} script.
27577 Other prefixes specified with @option{-B} take precedence over this prefix.
27579 This prefix is also used for finding files such as @file{crt0.o} that are
27582 In addition, the prefix is used in an unusual way in finding the
27583 directories to search for header files. For each of the standard
27584 directories whose name normally begins with @samp{/usr/local/lib/gcc}
27585 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
27586 replacing that beginning with the specified prefix to produce an
27587 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
27588 @file{foo/bar} just before it searches the standard directory
27589 @file{/usr/local/lib/bar}.
27590 If a standard directory begins with the configured
27591 @var{prefix} then the value of @var{prefix} is replaced by
27592 @env{GCC_EXEC_PREFIX} when looking for header files.
27594 @item COMPILER_PATH
27595 @findex COMPILER_PATH
27596 The value of @env{COMPILER_PATH} is a colon-separated list of
27597 directories, much like @env{PATH}. GCC tries the directories thus
27598 specified when searching for subprograms, if it cannot find the
27599 subprograms using @env{GCC_EXEC_PREFIX}.
27602 @findex LIBRARY_PATH
27603 The value of @env{LIBRARY_PATH} is a colon-separated list of
27604 directories, much like @env{PATH}. When configured as a native compiler,
27605 GCC tries the directories thus specified when searching for special
27606 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
27607 using GCC also uses these directories when searching for ordinary
27608 libraries for the @option{-l} option (but directories specified with
27609 @option{-L} come first).
27613 @cindex locale definition
27614 This variable is used to pass locale information to the compiler. One way in
27615 which this information is used is to determine the character set to be used
27616 when character literals, string literals and comments are parsed in C and C++.
27617 When the compiler is configured to allow multibyte characters,
27618 the following values for @env{LANG} are recognized:
27622 Recognize JIS characters.
27624 Recognize SJIS characters.
27626 Recognize EUCJP characters.
27629 If @env{LANG} is not defined, or if it has some other value, then the
27630 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
27631 recognize and translate multibyte characters.
27635 Some additional environment variables affect the behavior of the
27638 @include cppenv.texi
27642 @node Precompiled Headers
27643 @section Using Precompiled Headers
27644 @cindex precompiled headers
27645 @cindex speed of compilation
27647 Often large projects have many header files that are included in every
27648 source file. The time the compiler takes to process these header files
27649 over and over again can account for nearly all of the time required to
27650 build the project. To make builds faster, GCC allows you to
27651 @dfn{precompile} a header file.
27653 To create a precompiled header file, simply compile it as you would any
27654 other file, if necessary using the @option{-x} option to make the driver
27655 treat it as a C or C++ header file. You may want to use a
27656 tool like @command{make} to keep the precompiled header up-to-date when
27657 the headers it contains change.
27659 A precompiled header file is searched for when @code{#include} is
27660 seen in the compilation. As it searches for the included file
27661 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
27662 compiler looks for a precompiled header in each directory just before it
27663 looks for the include file in that directory. The name searched for is
27664 the name specified in the @code{#include} with @samp{.gch} appended. If
27665 the precompiled header file cannot be used, it is ignored.
27667 For instance, if you have @code{#include "all.h"}, and you have
27668 @file{all.h.gch} in the same directory as @file{all.h}, then the
27669 precompiled header file is used if possible, and the original
27670 header is used otherwise.
27672 Alternatively, you might decide to put the precompiled header file in a
27673 directory and use @option{-I} to ensure that directory is searched
27674 before (or instead of) the directory containing the original header.
27675 Then, if you want to check that the precompiled header file is always
27676 used, you can put a file of the same name as the original header in this
27677 directory containing an @code{#error} command.
27679 This also works with @option{-include}. So yet another way to use
27680 precompiled headers, good for projects not designed with precompiled
27681 header files in mind, is to simply take most of the header files used by
27682 a project, include them from another header file, precompile that header
27683 file, and @option{-include} the precompiled header. If the header files
27684 have guards against multiple inclusion, they are skipped because
27685 they've already been included (in the precompiled header).
27687 If you need to precompile the same header file for different
27688 languages, targets, or compiler options, you can instead make a
27689 @emph{directory} named like @file{all.h.gch}, and put each precompiled
27690 header in the directory, perhaps using @option{-o}. It doesn't matter
27691 what you call the files in the directory; every precompiled header in
27692 the directory is considered. The first precompiled header
27693 encountered in the directory that is valid for this compilation is
27694 used; they're searched in no particular order.
27696 There are many other possibilities, limited only by your imagination,
27697 good sense, and the constraints of your build system.
27699 A precompiled header file can be used only when these conditions apply:
27703 Only one precompiled header can be used in a particular compilation.
27706 A precompiled header cannot be used once the first C token is seen. You
27707 can have preprocessor directives before a precompiled header; you cannot
27708 include a precompiled header from inside another header.
27711 The precompiled header file must be produced for the same language as
27712 the current compilation. You cannot use a C precompiled header for a C++
27716 The precompiled header file must have been produced by the same compiler
27717 binary as the current compilation is using.
27720 Any macros defined before the precompiled header is included must
27721 either be defined in the same way as when the precompiled header was
27722 generated, or must not affect the precompiled header, which usually
27723 means that they don't appear in the precompiled header at all.
27725 The @option{-D} option is one way to define a macro before a
27726 precompiled header is included; using a @code{#define} can also do it.
27727 There are also some options that define macros implicitly, like
27728 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
27731 @item If debugging information is output when using the precompiled
27732 header, using @option{-g} or similar, the same kind of debugging information
27733 must have been output when building the precompiled header. However,
27734 a precompiled header built using @option{-g} can be used in a compilation
27735 when no debugging information is being output.
27737 @item The same @option{-m} options must generally be used when building
27738 and using the precompiled header. @xref{Submodel Options},
27739 for any cases where this rule is relaxed.
27741 @item Each of the following options must be the same when building and using
27742 the precompiled header:
27744 @gccoptlist{-fexceptions}
27747 Some other command-line options starting with @option{-f},
27748 @option{-p}, or @option{-O} must be defined in the same way as when
27749 the precompiled header was generated. At present, it's not clear
27750 which options are safe to change and which are not; the safest choice
27751 is to use exactly the same options when generating and using the
27752 precompiled header. The following are known to be safe:
27754 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
27755 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
27756 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
27761 For all of these except the last, the compiler automatically
27762 ignores the precompiled header if the conditions aren't met. If you
27763 find an option combination that doesn't work and doesn't cause the
27764 precompiled header to be ignored, please consider filing a bug report,
27767 If you do use differing options when generating and using the
27768 precompiled header, the actual behavior is a mixture of the
27769 behavior for the options. For instance, if you use @option{-g} to
27770 generate the precompiled header but not when using it, you may or may
27771 not get debugging information for routines in the precompiled header.