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
983 -mabi=@var{ABI-string} @gol
984 -mfdiv -mno-fdiv @gol
986 -march=@var{ISA-string} @gol
987 -mtune=@var{processor-string} @gol
988 -msmall-data-limit=@var{N-bytes} @gol
989 -msave-restore -mno-save-restore @gol
990 -mstrict-align -mno-strict-align @gol
991 -mcmodel=medlow -mcmodel=medany @gol
992 -mexplicit-relocs -mno-explicit-relocs @gol}
995 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
996 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
997 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
999 @emph{RS/6000 and PowerPC Options}
1000 @gccoptlist{-mcpu=@var{cpu-type} @gol
1001 -mtune=@var{cpu-type} @gol
1002 -mcmodel=@var{code-model} @gol
1004 -maltivec -mno-altivec @gol
1005 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1006 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1007 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1008 -mfprnd -mno-fprnd @gol
1009 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
1010 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1011 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1012 -malign-power -malign-natural @gol
1013 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1014 -msingle-float -mdouble-float -msimple-fpu @gol
1015 -mstring -mno-string -mupdate -mno-update @gol
1016 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1017 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1018 -mstrict-align -mno-strict-align -mrelocatable @gol
1019 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1020 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1021 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
1022 -mprioritize-restricted-insns=@var{priority} @gol
1023 -msched-costly-dep=@var{dependence_type} @gol
1024 -minsert-sched-nops=@var{scheme} @gol
1025 -mcall-sysv -mcall-netbsd @gol
1026 -maix-struct-return -msvr4-struct-return @gol
1027 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1028 -mblock-move-inline-limit=@var{num} @gol
1029 -misel -mno-isel @gol
1030 -misel=yes -misel=no @gol
1032 -mspe=yes -mspe=no @gol
1034 -mvrsave -mno-vrsave @gol
1035 -mmulhw -mno-mulhw @gol
1036 -mdlmzb -mno-dlmzb @gol
1037 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1038 -mprototype -mno-prototype @gol
1039 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1040 -msdata=@var{opt} -mvxworks -G @var{num} @gol
1041 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1042 -mno-recip-precision @gol
1043 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1044 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1045 -msave-toc-indirect -mno-save-toc-indirect @gol
1046 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1047 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1048 -mquad-memory -mno-quad-memory @gol
1049 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1050 -mcompat-align-parm -mno-compat-align-parm @gol
1051 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1052 -mgnu-attribute -mno-gnu-attribute @gol
1053 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1054 -mstack-protector-guard-offset=@var{offset}}
1057 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1059 -mbig-endian-data -mlittle-endian-data @gol
1062 -mas100-syntax -mno-as100-syntax@gol
1064 -mmax-constant-size=@gol
1067 -mallow-string-insns -mno-allow-string-insns@gol
1069 -mno-warn-multiple-fast-interrupts@gol
1070 -msave-acc-in-interrupts}
1072 @emph{S/390 and zSeries Options}
1073 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1074 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1075 -mlong-double-64 -mlong-double-128 @gol
1076 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1077 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1078 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1079 -mhtm -mvx -mzvector @gol
1080 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1081 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1082 -mhotpatch=@var{halfwords},@var{halfwords}}
1084 @emph{Score Options}
1085 @gccoptlist{-meb -mel @gol
1089 -mscore5 -mscore5u -mscore7 -mscore7d}
1092 @gccoptlist{-m1 -m2 -m2e @gol
1093 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1095 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1096 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1097 -mb -ml -mdalign -mrelax @gol
1098 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1099 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1100 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1101 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1102 -maccumulate-outgoing-args @gol
1103 -matomic-model=@var{atomic-model} @gol
1104 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1105 -mcbranch-force-delay-slot @gol
1106 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1107 -mpretend-cmove -mtas}
1109 @emph{Solaris 2 Options}
1110 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1113 @emph{SPARC Options}
1114 @gccoptlist{-mcpu=@var{cpu-type} @gol
1115 -mtune=@var{cpu-type} @gol
1116 -mcmodel=@var{code-model} @gol
1117 -mmemory-model=@var{mem-model} @gol
1118 -m32 -m64 -mapp-regs -mno-app-regs @gol
1119 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1120 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1121 -mhard-quad-float -msoft-quad-float @gol
1122 -mstack-bias -mno-stack-bias @gol
1123 -mstd-struct-return -mno-std-struct-return @gol
1124 -munaligned-doubles -mno-unaligned-doubles @gol
1125 -muser-mode -mno-user-mode @gol
1126 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1127 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1128 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1129 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1130 -mpopc -mno-popc -msubxc -mno-subxc @gol
1131 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1135 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1136 -msafe-dma -munsafe-dma @gol
1138 -msmall-mem -mlarge-mem -mstdmain @gol
1139 -mfixed-range=@var{register-range} @gol
1141 -maddress-space-conversion -mno-address-space-conversion @gol
1142 -mcache-size=@var{cache-size} @gol
1143 -matomic-updates -mno-atomic-updates}
1145 @emph{System V Options}
1146 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1148 @emph{TILE-Gx Options}
1149 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1150 -mcmodel=@var{code-model}}
1152 @emph{TILEPro Options}
1153 @gccoptlist{-mcpu=@var{cpu} -m32}
1156 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1157 -mprolog-function -mno-prolog-function -mspace @gol
1158 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1159 -mapp-regs -mno-app-regs @gol
1160 -mdisable-callt -mno-disable-callt @gol
1161 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1162 -mv850e -mv850 -mv850e3v5 @gol
1173 @gccoptlist{-mg -mgnu -munix}
1175 @emph{Visium Options}
1176 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1177 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1180 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1181 -mpointer-size=@var{size}}
1183 @emph{VxWorks Options}
1184 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1185 -Xbind-lazy -Xbind-now}
1188 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1189 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1190 -mfpmath=@var{unit} @gol
1191 -masm=@var{dialect} -mno-fancy-math-387 @gol
1192 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1193 -mno-wide-multiply -mrtd -malign-double @gol
1194 -mpreferred-stack-boundary=@var{num} @gol
1195 -mincoming-stack-boundary=@var{num} @gol
1196 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1197 -mrecip -mrecip=@var{opt} @gol
1198 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol
1199 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1200 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1201 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1202 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1203 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1204 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1205 -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx @gol
1206 -mmwaitx -mclzero -mpku -mthreads -mgfni @gol
1207 -mcet -mibt -mshstk -mforce-indirect-call -mavx512vbmi2 @gol
1208 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1209 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1210 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1211 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1212 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1213 -mregparm=@var{num} -msseregparm @gol
1214 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1215 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1216 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1217 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1218 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1219 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1220 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1221 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1222 -mstack-protector-guard-reg=@var{reg} @gol
1223 -mstack-protector-guard-offset=@var{offset} @gol
1224 -mstack-protector-guard-symbol=@var{symbol} -mmitigate-rop @gol
1225 -mgeneral-regs-only -mcall-ms2sysv-xlogues}
1227 @emph{x86 Windows Options}
1228 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1229 -mnop-fun-dllimport -mthread @gol
1230 -municode -mwin32 -mwindows -fno-set-stack-executable}
1232 @emph{Xstormy16 Options}
1235 @emph{Xtensa Options}
1236 @gccoptlist{-mconst16 -mno-const16 @gol
1237 -mfused-madd -mno-fused-madd @gol
1239 -mserialize-volatile -mno-serialize-volatile @gol
1240 -mtext-section-literals -mno-text-section-literals @gol
1241 -mauto-litpools -mno-auto-litpools @gol
1242 -mtarget-align -mno-target-align @gol
1243 -mlongcalls -mno-longcalls}
1245 @emph{zSeries Options}
1246 See S/390 and zSeries Options.
1250 @node Overall Options
1251 @section Options Controlling the Kind of Output
1253 Compilation can involve up to four stages: preprocessing, compilation
1254 proper, assembly and linking, always in that order. GCC is capable of
1255 preprocessing and compiling several files either into several
1256 assembler input files, or into one assembler input file; then each
1257 assembler input file produces an object file, and linking combines all
1258 the object files (those newly compiled, and those specified as input)
1259 into an executable file.
1261 @cindex file name suffix
1262 For any given input file, the file name suffix determines what kind of
1263 compilation is done:
1267 C source code that must be preprocessed.
1270 C source code that should not be preprocessed.
1273 C++ source code that should not be preprocessed.
1276 Objective-C source code. Note that you must link with the @file{libobjc}
1277 library to make an Objective-C program work.
1280 Objective-C source code that should not be preprocessed.
1284 Objective-C++ source code. Note that you must link with the @file{libobjc}
1285 library to make an Objective-C++ program work. Note that @samp{.M} refers
1286 to a literal capital M@.
1288 @item @var{file}.mii
1289 Objective-C++ source code that should not be preprocessed.
1292 C, C++, Objective-C or Objective-C++ header file to be turned into a
1293 precompiled header (default), or C, C++ header file to be turned into an
1294 Ada spec (via the @option{-fdump-ada-spec} switch).
1297 @itemx @var{file}.cp
1298 @itemx @var{file}.cxx
1299 @itemx @var{file}.cpp
1300 @itemx @var{file}.CPP
1301 @itemx @var{file}.c++
1303 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1304 the last two letters must both be literally @samp{x}. Likewise,
1305 @samp{.C} refers to a literal capital C@.
1309 Objective-C++ source code that must be preprocessed.
1311 @item @var{file}.mii
1312 Objective-C++ source code that should not be preprocessed.
1316 @itemx @var{file}.hp
1317 @itemx @var{file}.hxx
1318 @itemx @var{file}.hpp
1319 @itemx @var{file}.HPP
1320 @itemx @var{file}.h++
1321 @itemx @var{file}.tcc
1322 C++ header file to be turned into a precompiled header or Ada spec.
1325 @itemx @var{file}.for
1326 @itemx @var{file}.ftn
1327 Fixed form Fortran source code that should not be preprocessed.
1330 @itemx @var{file}.FOR
1331 @itemx @var{file}.fpp
1332 @itemx @var{file}.FPP
1333 @itemx @var{file}.FTN
1334 Fixed form Fortran source code that must be preprocessed (with the traditional
1337 @item @var{file}.f90
1338 @itemx @var{file}.f95
1339 @itemx @var{file}.f03
1340 @itemx @var{file}.f08
1341 Free form Fortran source code that should not be preprocessed.
1343 @item @var{file}.F90
1344 @itemx @var{file}.F95
1345 @itemx @var{file}.F03
1346 @itemx @var{file}.F08
1347 Free form Fortran source code that must be preprocessed (with the
1348 traditional preprocessor).
1353 @item @var{file}.brig
1354 BRIG files (binary representation of HSAIL).
1356 @item @var{file}.ads
1357 Ada source code file that contains a library unit declaration (a
1358 declaration of a package, subprogram, or generic, or a generic
1359 instantiation), or a library unit renaming declaration (a package,
1360 generic, or subprogram renaming declaration). Such files are also
1363 @item @var{file}.adb
1364 Ada source code file containing a library unit body (a subprogram or
1365 package body). Such files are also called @dfn{bodies}.
1367 @c GCC also knows about some suffixes for languages not yet included:
1378 @itemx @var{file}.sx
1379 Assembler code that must be preprocessed.
1382 An object file to be fed straight into linking.
1383 Any file name with no recognized suffix is treated this way.
1387 You can specify the input language explicitly with the @option{-x} option:
1390 @item -x @var{language}
1391 Specify explicitly the @var{language} for the following input files
1392 (rather than letting the compiler choose a default based on the file
1393 name suffix). This option applies to all following input files until
1394 the next @option{-x} option. Possible values for @var{language} are:
1396 c c-header cpp-output
1397 c++ c++-header c++-cpp-output
1398 objective-c objective-c-header objective-c-cpp-output
1399 objective-c++ objective-c++-header objective-c++-cpp-output
1400 assembler assembler-with-cpp
1402 f77 f77-cpp-input f95 f95-cpp-input
1408 Turn off any specification of a language, so that subsequent files are
1409 handled according to their file name suffixes (as they are if @option{-x}
1410 has not been used at all).
1413 If you only want some of the stages of compilation, you can use
1414 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1415 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1416 @command{gcc} is to stop. Note that some combinations (for example,
1417 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1422 Compile or assemble the source files, but do not link. The linking
1423 stage simply is not done. The ultimate output is in the form of an
1424 object file for each source file.
1426 By default, the object file name for a source file is made by replacing
1427 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1429 Unrecognized input files, not requiring compilation or assembly, are
1434 Stop after the stage of compilation proper; do not assemble. The output
1435 is in the form of an assembler code file for each non-assembler input
1438 By default, the assembler file name for a source file is made by
1439 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1441 Input files that don't require compilation are ignored.
1445 Stop after the preprocessing stage; do not run the compiler proper. The
1446 output is in the form of preprocessed source code, which is sent to the
1449 Input files that don't require preprocessing are ignored.
1451 @cindex output file option
1454 Place output in file @var{file}. This applies to whatever
1455 sort of output is being produced, whether it be an executable file,
1456 an object file, an assembler file or preprocessed C code.
1458 If @option{-o} is not specified, the default is to put an executable
1459 file in @file{a.out}, the object file for
1460 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1461 assembler file in @file{@var{source}.s}, a precompiled header file in
1462 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1467 Print (on standard error output) the commands executed to run the stages
1468 of compilation. Also print the version number of the compiler driver
1469 program and of the preprocessor and the compiler proper.
1473 Like @option{-v} except the commands are not executed and arguments
1474 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1475 This is useful for shell scripts to capture the driver-generated command lines.
1479 Print (on the standard output) a description of the command-line options
1480 understood by @command{gcc}. If the @option{-v} option is also specified
1481 then @option{--help} is also passed on to the various processes
1482 invoked by @command{gcc}, so that they can display the command-line options
1483 they accept. If the @option{-Wextra} option has also been specified
1484 (prior to the @option{--help} option), then command-line options that
1485 have no documentation associated with them are also displayed.
1488 @opindex target-help
1489 Print (on the standard output) a description of target-specific command-line
1490 options for each tool. For some targets extra target-specific
1491 information may also be printed.
1493 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1494 Print (on the standard output) a description of the command-line
1495 options understood by the compiler that fit into all specified classes
1496 and qualifiers. These are the supported classes:
1499 @item @samp{optimizers}
1500 Display all of the optimization options supported by the
1503 @item @samp{warnings}
1504 Display all of the options controlling warning messages
1505 produced by the compiler.
1508 Display target-specific options. Unlike the
1509 @option{--target-help} option however, target-specific options of the
1510 linker and assembler are not displayed. This is because those
1511 tools do not currently support the extended @option{--help=} syntax.
1514 Display the values recognized by the @option{--param}
1517 @item @var{language}
1518 Display the options supported for @var{language}, where
1519 @var{language} is the name of one of the languages supported in this
1523 Display the options that are common to all languages.
1526 These are the supported qualifiers:
1529 @item @samp{undocumented}
1530 Display only those options that are undocumented.
1533 Display options taking an argument that appears after an equal
1534 sign in the same continuous piece of text, such as:
1535 @samp{--help=target}.
1537 @item @samp{separate}
1538 Display options taking an argument that appears as a separate word
1539 following the original option, such as: @samp{-o output-file}.
1542 Thus for example to display all the undocumented target-specific
1543 switches supported by the compiler, use:
1546 --help=target,undocumented
1549 The sense of a qualifier can be inverted by prefixing it with the
1550 @samp{^} character, so for example to display all binary warning
1551 options (i.e., ones that are either on or off and that do not take an
1552 argument) that have a description, use:
1555 --help=warnings,^joined,^undocumented
1558 The argument to @option{--help=} should not consist solely of inverted
1561 Combining several classes is possible, although this usually
1562 restricts the output so much that there is nothing to display. One
1563 case where it does work, however, is when one of the classes is
1564 @var{target}. For example, to display all the target-specific
1565 optimization options, use:
1568 --help=target,optimizers
1571 The @option{--help=} option can be repeated on the command line. Each
1572 successive use displays its requested class of options, skipping
1573 those that have already been displayed.
1575 If the @option{-Q} option appears on the command line before the
1576 @option{--help=} option, then the descriptive text displayed by
1577 @option{--help=} is changed. Instead of describing the displayed
1578 options, an indication is given as to whether the option is enabled,
1579 disabled or set to a specific value (assuming that the compiler
1580 knows this at the point where the @option{--help=} option is used).
1582 Here is a truncated example from the ARM port of @command{gcc}:
1585 % gcc -Q -mabi=2 --help=target -c
1586 The following options are target specific:
1588 -mabort-on-noreturn [disabled]
1592 The output is sensitive to the effects of previous command-line
1593 options, so for example it is possible to find out which optimizations
1594 are enabled at @option{-O2} by using:
1597 -Q -O2 --help=optimizers
1600 Alternatively you can discover which binary optimizations are enabled
1601 by @option{-O3} by using:
1604 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1605 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1606 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1611 Display the version number and copyrights of the invoked GCC@.
1613 @item -pass-exit-codes
1614 @opindex pass-exit-codes
1615 Normally the @command{gcc} program exits with the code of 1 if any
1616 phase of the compiler returns a non-success return code. If you specify
1617 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1618 the numerically highest error produced by any phase returning an error
1619 indication. The C, C++, and Fortran front ends return 4 if an internal
1620 compiler error is encountered.
1624 Use pipes rather than temporary files for communication between the
1625 various stages of compilation. This fails to work on some systems where
1626 the assembler is unable to read from a pipe; but the GNU assembler has
1629 @item -specs=@var{file}
1631 Process @var{file} after the compiler reads in the standard @file{specs}
1632 file, in order to override the defaults which the @command{gcc} driver
1633 program uses when determining what switches to pass to @command{cc1},
1634 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1635 @option{-specs=@var{file}} can be specified on the command line, and they
1636 are processed in order, from left to right. @xref{Spec Files}, for
1637 information about the format of the @var{file}.
1641 Invoke all subcommands under a wrapper program. The name of the
1642 wrapper program and its parameters are passed as a comma separated
1646 gcc -c t.c -wrapper gdb,--args
1650 This invokes all subprograms of @command{gcc} under
1651 @samp{gdb --args}, thus the invocation of @command{cc1} is
1652 @samp{gdb --args cc1 @dots{}}.
1654 @item -fplugin=@var{name}.so
1656 Load the plugin code in file @var{name}.so, assumed to be a
1657 shared object to be dlopen'd by the compiler. The base name of
1658 the shared object file is used to identify the plugin for the
1659 purposes of argument parsing (See
1660 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1661 Each plugin should define the callback functions specified in the
1664 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1665 @opindex fplugin-arg
1666 Define an argument called @var{key} with a value of @var{value}
1667 for the plugin called @var{name}.
1669 @item -fdump-ada-spec@r{[}-slim@r{]}
1670 @opindex fdump-ada-spec
1671 For C and C++ source and include files, generate corresponding Ada specs.
1672 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1673 GNAT User's Guide}, which provides detailed documentation on this feature.
1675 @item -fada-spec-parent=@var{unit}
1676 @opindex fada-spec-parent
1677 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1678 Ada specs as child units of parent @var{unit}.
1680 @item -fdump-go-spec=@var{file}
1681 @opindex fdump-go-spec
1682 For input files in any language, generate corresponding Go
1683 declarations in @var{file}. This generates Go @code{const},
1684 @code{type}, @code{var}, and @code{func} declarations which may be a
1685 useful way to start writing a Go interface to code written in some
1688 @include @value{srcdir}/../libiberty/at-file.texi
1692 @section Compiling C++ Programs
1694 @cindex suffixes for C++ source
1695 @cindex C++ source file suffixes
1696 C++ source files conventionally use one of the suffixes @samp{.C},
1697 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1698 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1699 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1700 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1701 files with these names and compiles them as C++ programs even if you
1702 call the compiler the same way as for compiling C programs (usually
1703 with the name @command{gcc}).
1707 However, the use of @command{gcc} does not add the C++ library.
1708 @command{g++} is a program that calls GCC and automatically specifies linking
1709 against the C++ library. It treats @samp{.c},
1710 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1711 files unless @option{-x} is used. This program is also useful when
1712 precompiling a C header file with a @samp{.h} extension for use in C++
1713 compilations. On many systems, @command{g++} is also installed with
1714 the name @command{c++}.
1716 @cindex invoking @command{g++}
1717 When you compile C++ programs, you may specify many of the same
1718 command-line options that you use for compiling programs in any
1719 language; or command-line options meaningful for C and related
1720 languages; or options that are meaningful only for C++ programs.
1721 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1722 explanations of options for languages related to C@.
1723 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1724 explanations of options that are meaningful only for C++ programs.
1726 @node C Dialect Options
1727 @section Options Controlling C Dialect
1728 @cindex dialect options
1729 @cindex language dialect options
1730 @cindex options, dialect
1732 The following options control the dialect of C (or languages derived
1733 from C, such as C++, Objective-C and Objective-C++) that the compiler
1737 @cindex ANSI support
1741 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1742 equivalent to @option{-std=c++98}.
1744 This turns off certain features of GCC that are incompatible with ISO
1745 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1746 such as the @code{asm} and @code{typeof} keywords, and
1747 predefined macros such as @code{unix} and @code{vax} that identify the
1748 type of system you are using. It also enables the undesirable and
1749 rarely used ISO trigraph feature. For the C compiler,
1750 it disables recognition of C++ style @samp{//} comments as well as
1751 the @code{inline} keyword.
1753 The alternate keywords @code{__asm__}, @code{__extension__},
1754 @code{__inline__} and @code{__typeof__} continue to work despite
1755 @option{-ansi}. You would not want to use them in an ISO C program, of
1756 course, but it is useful to put them in header files that might be included
1757 in compilations done with @option{-ansi}. Alternate predefined macros
1758 such as @code{__unix__} and @code{__vax__} are also available, with or
1759 without @option{-ansi}.
1761 The @option{-ansi} option does not cause non-ISO programs to be
1762 rejected gratuitously. For that, @option{-Wpedantic} is required in
1763 addition to @option{-ansi}. @xref{Warning Options}.
1765 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1766 option is used. Some header files may notice this macro and refrain
1767 from declaring certain functions or defining certain macros that the
1768 ISO standard doesn't call for; this is to avoid interfering with any
1769 programs that might use these names for other things.
1771 Functions that are normally built in but do not have semantics
1772 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1773 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1774 built-in functions provided by GCC}, for details of the functions
1779 Determine the language standard. @xref{Standards,,Language Standards
1780 Supported by GCC}, for details of these standard versions. This option
1781 is currently only supported when compiling C or C++.
1783 The compiler can accept several base standards, such as @samp{c90} or
1784 @samp{c++98}, and GNU dialects of those standards, such as
1785 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1786 compiler accepts all programs following that standard plus those
1787 using GNU extensions that do not contradict it. For example,
1788 @option{-std=c90} turns off certain features of GCC that are
1789 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1790 keywords, but not other GNU extensions that do not have a meaning in
1791 ISO C90, such as omitting the middle term of a @code{?:}
1792 expression. On the other hand, when a GNU dialect of a standard is
1793 specified, all features supported by the compiler are enabled, even when
1794 those features change the meaning of the base standard. As a result, some
1795 strict-conforming programs may be rejected. The particular standard
1796 is used by @option{-Wpedantic} to identify which features are GNU
1797 extensions given that version of the standard. For example
1798 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1799 comments, while @option{-std=gnu99 -Wpedantic} does not.
1801 A value for this option must be provided; possible values are
1807 Support all ISO C90 programs (certain GNU extensions that conflict
1808 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1810 @item iso9899:199409
1811 ISO C90 as modified in amendment 1.
1817 ISO C99. This standard is substantially completely supported, modulo
1818 bugs and floating-point issues
1819 (mainly but not entirely relating to optional C99 features from
1820 Annexes F and G). See
1821 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1822 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1827 ISO C11, the 2011 revision of the ISO C standard. This standard is
1828 substantially completely supported, modulo bugs, floating-point issues
1829 (mainly but not entirely relating to optional C11 features from
1830 Annexes F and G) and the optional Annexes K (Bounds-checking
1831 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1837 ISO C17, the 2017 revision of the ISO C standard (expected to be
1838 published in 2018). This standard is
1839 same as C11 except for corrections of defects (all of which are also
1840 applied with @option{-std=c11}) and a new value of
1841 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1845 GNU dialect of ISO C90 (including some C99 features).
1849 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1853 GNU dialect of ISO C11.
1854 The name @samp{gnu1x} is deprecated.
1858 GNU dialect of ISO C17. This is the default for C code.
1862 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1863 additional defect reports. Same as @option{-ansi} for C++ code.
1867 GNU dialect of @option{-std=c++98}.
1871 The 2011 ISO C++ standard plus amendments.
1872 The name @samp{c++0x} is deprecated.
1876 GNU dialect of @option{-std=c++11}.
1877 The name @samp{gnu++0x} is deprecated.
1881 The 2014 ISO C++ standard plus amendments.
1882 The name @samp{c++1y} is deprecated.
1886 GNU dialect of @option{-std=c++14}.
1887 This is the default for C++ code.
1888 The name @samp{gnu++1y} is deprecated.
1892 The 2017 ISO C++ standard plus amendments.
1893 The name @samp{c++1z} is deprecated.
1897 GNU dialect of @option{-std=c++17}.
1898 The name @samp{gnu++1z} is deprecated.
1901 The next revision of the ISO C++ standard, tentatively planned for
1902 2020. Support is highly experimental, and will almost certainly
1903 change in incompatible ways in future releases.
1906 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
1907 and will almost certainly change in incompatible ways in future
1911 @item -fgnu89-inline
1912 @opindex fgnu89-inline
1913 The option @option{-fgnu89-inline} tells GCC to use the traditional
1914 GNU semantics for @code{inline} functions when in C99 mode.
1915 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1916 Using this option is roughly equivalent to adding the
1917 @code{gnu_inline} function attribute to all inline functions
1918 (@pxref{Function Attributes}).
1920 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1921 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1922 specifies the default behavior).
1923 This option is not supported in @option{-std=c90} or
1924 @option{-std=gnu90} mode.
1926 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1927 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1928 in effect for @code{inline} functions. @xref{Common Predefined
1929 Macros,,,cpp,The C Preprocessor}.
1931 @item -fpermitted-flt-eval-methods=@var{style}
1932 @opindex fpermitted-flt-eval-methods
1933 @opindex fpermitted-flt-eval-methods=c11
1934 @opindex fpermitted-flt-eval-methods=ts-18661-3
1935 ISO/IEC TS 18661-3 defines new permissible values for
1936 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
1937 a semantic type that is an interchange or extended format should be
1938 evaluated to the precision and range of that type. These new values are
1939 a superset of those permitted under C99/C11, which does not specify the
1940 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
1941 conforming to C11 may not have been written expecting the possibility of
1944 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
1945 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
1946 or the extended set of values specified in ISO/IEC TS 18661-3.
1948 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
1950 The default when in a standards compliant mode (@option{-std=c11} or similar)
1951 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
1952 dialect (@option{-std=gnu11} or similar) is
1953 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
1955 @item -aux-info @var{filename}
1957 Output to the given filename prototyped declarations for all functions
1958 declared and/or defined in a translation unit, including those in header
1959 files. This option is silently ignored in any language other than C@.
1961 Besides declarations, the file indicates, in comments, the origin of
1962 each declaration (source file and line), whether the declaration was
1963 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1964 @samp{O} for old, respectively, in the first character after the line
1965 number and the colon), and whether it came from a declaration or a
1966 definition (@samp{C} or @samp{F}, respectively, in the following
1967 character). In the case of function definitions, a K&R-style list of
1968 arguments followed by their declarations is also provided, inside
1969 comments, after the declaration.
1971 @item -fallow-parameterless-variadic-functions
1972 @opindex fallow-parameterless-variadic-functions
1973 Accept variadic functions without named parameters.
1975 Although it is possible to define such a function, this is not very
1976 useful as it is not possible to read the arguments. This is only
1977 supported for C as this construct is allowed by C++.
1981 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1982 keyword, so that code can use these words as identifiers. You can use
1983 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1984 instead. @option{-ansi} implies @option{-fno-asm}.
1986 In C++, this switch only affects the @code{typeof} keyword, since
1987 @code{asm} and @code{inline} are standard keywords. You may want to
1988 use the @option{-fno-gnu-keywords} flag instead, which has the same
1989 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1990 switch only affects the @code{asm} and @code{typeof} keywords, since
1991 @code{inline} is a standard keyword in ISO C99.
1994 @itemx -fno-builtin-@var{function}
1995 @opindex fno-builtin
1996 @cindex built-in functions
1997 Don't recognize built-in functions that do not begin with
1998 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1999 functions provided by GCC}, for details of the functions affected,
2000 including those which are not built-in functions when @option{-ansi} or
2001 @option{-std} options for strict ISO C conformance are used because they
2002 do not have an ISO standard meaning.
2004 GCC normally generates special code to handle certain built-in functions
2005 more efficiently; for instance, calls to @code{alloca} may become single
2006 instructions which adjust the stack directly, and calls to @code{memcpy}
2007 may become inline copy loops. The resulting code is often both smaller
2008 and faster, but since the function calls no longer appear as such, you
2009 cannot set a breakpoint on those calls, nor can you change the behavior
2010 of the functions by linking with a different library. In addition,
2011 when a function is recognized as a built-in function, GCC may use
2012 information about that function to warn about problems with calls to
2013 that function, or to generate more efficient code, even if the
2014 resulting code still contains calls to that function. For example,
2015 warnings are given with @option{-Wformat} for bad calls to
2016 @code{printf} when @code{printf} is built in and @code{strlen} is
2017 known not to modify global memory.
2019 With the @option{-fno-builtin-@var{function}} option
2020 only the built-in function @var{function} is
2021 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2022 function is named that is not built-in in this version of GCC, this
2023 option is ignored. There is no corresponding
2024 @option{-fbuiltin-@var{function}} option; if you wish to enable
2025 built-in functions selectively when using @option{-fno-builtin} or
2026 @option{-ffreestanding}, you may define macros such as:
2029 #define abs(n) __builtin_abs ((n))
2030 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2036 Enable parsing of function definitions marked with @code{__GIMPLE}.
2037 This is an experimental feature that allows unit testing of GIMPLE
2042 @cindex hosted environment
2044 Assert that compilation targets a hosted environment. This implies
2045 @option{-fbuiltin}. A hosted environment is one in which the
2046 entire standard library is available, and in which @code{main} has a return
2047 type of @code{int}. Examples are nearly everything except a kernel.
2048 This is equivalent to @option{-fno-freestanding}.
2050 @item -ffreestanding
2051 @opindex ffreestanding
2052 @cindex hosted environment
2054 Assert that compilation targets a freestanding environment. This
2055 implies @option{-fno-builtin}. A freestanding environment
2056 is one in which the standard library may not exist, and program startup may
2057 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2058 This is equivalent to @option{-fno-hosted}.
2060 @xref{Standards,,Language Standards Supported by GCC}, for details of
2061 freestanding and hosted environments.
2065 @cindex OpenACC accelerator programming
2066 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2067 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2068 compiler generates accelerated code according to the OpenACC Application
2069 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2070 implies @option{-pthread}, and thus is only supported on targets that
2071 have support for @option{-pthread}.
2073 @item -fopenacc-dim=@var{geom}
2074 @opindex fopenacc-dim
2075 @cindex OpenACC accelerator programming
2076 Specify default compute dimensions for parallel offload regions that do
2077 not explicitly specify. The @var{geom} value is a triple of
2078 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2079 can be omitted, to use a target-specific default value.
2083 @cindex OpenMP parallel
2084 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2085 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2086 compiler generates parallel code according to the OpenMP Application
2087 Program Interface v4.5 @w{@uref{http://www.openmp.org/}}. This option
2088 implies @option{-pthread}, and thus is only supported on targets that
2089 have support for @option{-pthread}. @option{-fopenmp} implies
2090 @option{-fopenmp-simd}.
2093 @opindex fopenmp-simd
2096 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2097 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2102 When the option @option{-fgnu-tm} is specified, the compiler
2103 generates code for the Linux variant of Intel's current Transactional
2104 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2105 an experimental feature whose interface may change in future versions
2106 of GCC, as the official specification changes. Please note that not
2107 all architectures are supported for this feature.
2109 For more information on GCC's support for transactional memory,
2110 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2111 Transactional Memory Library}.
2113 Note that the transactional memory feature is not supported with
2114 non-call exceptions (@option{-fnon-call-exceptions}).
2116 @item -fms-extensions
2117 @opindex fms-extensions
2118 Accept some non-standard constructs used in Microsoft header files.
2120 In C++ code, this allows member names in structures to be similar
2121 to previous types declarations.
2130 Some cases of unnamed fields in structures and unions are only
2131 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2132 fields within structs/unions}, for details.
2134 Note that this option is off for all targets but x86
2135 targets using ms-abi.
2137 @item -fplan9-extensions
2138 @opindex fplan9-extensions
2139 Accept some non-standard constructs used in Plan 9 code.
2141 This enables @option{-fms-extensions}, permits passing pointers to
2142 structures with anonymous fields to functions that expect pointers to
2143 elements of the type of the field, and permits referring to anonymous
2144 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2145 struct/union fields within structs/unions}, for details. This is only
2146 supported for C, not C++.
2148 @item -fcond-mismatch
2149 @opindex fcond-mismatch
2150 Allow conditional expressions with mismatched types in the second and
2151 third arguments. The value of such an expression is void. This option
2152 is not supported for C++.
2154 @item -flax-vector-conversions
2155 @opindex flax-vector-conversions
2156 Allow implicit conversions between vectors with differing numbers of
2157 elements and/or incompatible element types. This option should not be
2160 @item -funsigned-char
2161 @opindex funsigned-char
2162 Let the type @code{char} be unsigned, like @code{unsigned char}.
2164 Each kind of machine has a default for what @code{char} should
2165 be. It is either like @code{unsigned char} by default or like
2166 @code{signed char} by default.
2168 Ideally, a portable program should always use @code{signed char} or
2169 @code{unsigned char} when it depends on the signedness of an object.
2170 But many programs have been written to use plain @code{char} and
2171 expect it to be signed, or expect it to be unsigned, depending on the
2172 machines they were written for. This option, and its inverse, let you
2173 make such a program work with the opposite default.
2175 The type @code{char} is always a distinct type from each of
2176 @code{signed char} or @code{unsigned char}, even though its behavior
2177 is always just like one of those two.
2180 @opindex fsigned-char
2181 Let the type @code{char} be signed, like @code{signed char}.
2183 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2184 the negative form of @option{-funsigned-char}. Likewise, the option
2185 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2187 @item -fsigned-bitfields
2188 @itemx -funsigned-bitfields
2189 @itemx -fno-signed-bitfields
2190 @itemx -fno-unsigned-bitfields
2191 @opindex fsigned-bitfields
2192 @opindex funsigned-bitfields
2193 @opindex fno-signed-bitfields
2194 @opindex fno-unsigned-bitfields
2195 These options control whether a bit-field is signed or unsigned, when the
2196 declaration does not use either @code{signed} or @code{unsigned}. By
2197 default, such a bit-field is signed, because this is consistent: the
2198 basic integer types such as @code{int} are signed types.
2200 @item -fsso-struct=@var{endianness}
2201 @opindex fsso-struct
2202 Set the default scalar storage order of structures and unions to the
2203 specified endianness. The accepted values are @samp{big-endian},
2204 @samp{little-endian} and @samp{native} for the native endianness of
2205 the target (the default). This option is not supported for C++.
2207 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2208 code that is not binary compatible with code generated without it if the
2209 specified endianness is not the native endianness of the target.
2212 @node C++ Dialect Options
2213 @section Options Controlling C++ Dialect
2215 @cindex compiler options, C++
2216 @cindex C++ options, command-line
2217 @cindex options, C++
2218 This section describes the command-line options that are only meaningful
2219 for C++ programs. You can also use most of the GNU compiler options
2220 regardless of what language your program is in. For example, you
2221 might compile a file @file{firstClass.C} like this:
2224 g++ -g -fstrict-enums -O -c firstClass.C
2228 In this example, only @option{-fstrict-enums} is an option meant
2229 only for C++ programs; you can use the other options with any
2230 language supported by GCC@.
2232 Some options for compiling C programs, such as @option{-std}, are also
2233 relevant for C++ programs.
2234 @xref{C Dialect Options,,Options Controlling C Dialect}.
2236 Here is a list of options that are @emph{only} for compiling C++ programs:
2240 @item -fabi-version=@var{n}
2241 @opindex fabi-version
2242 Use version @var{n} of the C++ ABI@. The default is version 0.
2244 Version 0 refers to the version conforming most closely to
2245 the C++ ABI specification. Therefore, the ABI obtained using version 0
2246 will change in different versions of G++ as ABI bugs are fixed.
2248 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2250 Version 2 is the version of the C++ ABI that first appeared in G++
2251 3.4, and was the default through G++ 4.9.
2253 Version 3 corrects an error in mangling a constant address as a
2256 Version 4, which first appeared in G++ 4.5, implements a standard
2257 mangling for vector types.
2259 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2260 attribute const/volatile on function pointer types, decltype of a
2261 plain decl, and use of a function parameter in the declaration of
2264 Version 6, which first appeared in G++ 4.7, corrects the promotion
2265 behavior of C++11 scoped enums and the mangling of template argument
2266 packs, const/static_cast, prefix ++ and --, and a class scope function
2267 used as a template argument.
2269 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2270 builtin type and corrects the mangling of lambdas in default argument
2273 Version 8, which first appeared in G++ 4.9, corrects the substitution
2274 behavior of function types with function-cv-qualifiers.
2276 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2279 Version 10, which first appeared in G++ 6.1, adds mangling of
2280 attributes that affect type identity, such as ia32 calling convention
2281 attributes (e.g. @samp{stdcall}).
2283 Version 11, which first appeared in G++ 7, corrects the mangling of
2284 sizeof... expressions and operator names. For multiple entities with
2285 the same name within a function, that are declared in different scopes,
2286 the mangling now changes starting with the twelfth occurrence. It also
2287 implies @option{-fnew-inheriting-ctors}.
2289 See also @option{-Wabi}.
2291 @item -fabi-compat-version=@var{n}
2292 @opindex fabi-compat-version
2293 On targets that support strong aliases, G++
2294 works around mangling changes by creating an alias with the correct
2295 mangled name when defining a symbol with an incorrect mangled name.
2296 This switch specifies which ABI version to use for the alias.
2298 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2299 compatibility). If another ABI version is explicitly selected, this
2300 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2301 use @option{-fabi-compat-version=2}.
2303 If this option is not provided but @option{-Wabi=@var{n}} is, that
2304 version is used for compatibility aliases. If this option is provided
2305 along with @option{-Wabi} (without the version), the version from this
2306 option is used for the warning.
2308 @item -fno-access-control
2309 @opindex fno-access-control
2310 Turn off all access checking. This switch is mainly useful for working
2311 around bugs in the access control code.
2314 @opindex faligned-new
2315 Enable support for C++17 @code{new} of types that require more
2316 alignment than @code{void* ::operator new(std::size_t)} provides. A
2317 numeric argument such as @code{-faligned-new=32} can be used to
2318 specify how much alignment (in bytes) is provided by that function,
2319 but few users will need to override the default of
2320 @code{alignof(std::max_align_t)}.
2322 This flag is enabled by default for @option{-std=c++17}.
2326 Check that the pointer returned by @code{operator new} is non-null
2327 before attempting to modify the storage allocated. This check is
2328 normally unnecessary because the C++ standard specifies that
2329 @code{operator new} only returns @code{0} if it is declared
2330 @code{throw()}, in which case the compiler always checks the
2331 return value even without this option. In all other cases, when
2332 @code{operator new} has a non-empty exception specification, memory
2333 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2334 @samp{new (nothrow)}.
2338 Enable support for the C++ Extensions for Concepts Technical
2339 Specification, ISO 19217 (2015), which allows code like
2342 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2343 template <Addable T> T add (T a, T b) @{ return a + b; @}
2346 @item -fconstexpr-depth=@var{n}
2347 @opindex fconstexpr-depth
2348 Set the maximum nested evaluation depth for C++11 constexpr functions
2349 to @var{n}. A limit is needed to detect endless recursion during
2350 constant expression evaluation. The minimum specified by the standard
2353 @item -fconstexpr-loop-limit=@var{n}
2354 @opindex fconstexpr-loop-limit
2355 Set the maximum number of iterations for a loop in C++14 constexpr functions
2356 to @var{n}. A limit is needed to detect infinite loops during
2357 constant expression evaluation. The default is 262144 (1<<18).
2359 @item -fdeduce-init-list
2360 @opindex fdeduce-init-list
2361 Enable deduction of a template type parameter as
2362 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2365 template <class T> auto forward(T t) -> decltype (realfn (t))
2372 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2376 This deduction was implemented as a possible extension to the
2377 originally proposed semantics for the C++11 standard, but was not part
2378 of the final standard, so it is disabled by default. This option is
2379 deprecated, and may be removed in a future version of G++.
2381 @item -ffriend-injection
2382 @opindex ffriend-injection
2383 Inject friend functions into the enclosing namespace, so that they are
2384 visible outside the scope of the class in which they are declared.
2385 Friend functions were documented to work this way in the old Annotated
2386 C++ Reference Manual.
2387 However, in ISO C++ a friend function that is not declared
2388 in an enclosing scope can only be found using argument dependent
2389 lookup. GCC defaults to the standard behavior.
2391 This option is for compatibility, and may be removed in a future
2394 @item -fno-elide-constructors
2395 @opindex fno-elide-constructors
2396 The C++ standard allows an implementation to omit creating a temporary
2397 that is only used to initialize another object of the same type.
2398 Specifying this option disables that optimization, and forces G++ to
2399 call the copy constructor in all cases. This option also causes G++
2400 to call trivial member functions which otherwise would be expanded inline.
2402 In C++17, the compiler is required to omit these temporaries, but this
2403 option still affects trivial member functions.
2405 @item -fno-enforce-eh-specs
2406 @opindex fno-enforce-eh-specs
2407 Don't generate code to check for violation of exception specifications
2408 at run time. This option violates the C++ standard, but may be useful
2409 for reducing code size in production builds, much like defining
2410 @code{NDEBUG}. This does not give user code permission to throw
2411 exceptions in violation of the exception specifications; the compiler
2412 still optimizes based on the specifications, so throwing an
2413 unexpected exception results in undefined behavior at run time.
2415 @item -fextern-tls-init
2416 @itemx -fno-extern-tls-init
2417 @opindex fextern-tls-init
2418 @opindex fno-extern-tls-init
2419 The C++11 and OpenMP standards allow @code{thread_local} and
2420 @code{threadprivate} variables to have dynamic (runtime)
2421 initialization. To support this, any use of such a variable goes
2422 through a wrapper function that performs any necessary initialization.
2423 When the use and definition of the variable are in the same
2424 translation unit, this overhead can be optimized away, but when the
2425 use is in a different translation unit there is significant overhead
2426 even if the variable doesn't actually need dynamic initialization. If
2427 the programmer can be sure that no use of the variable in a
2428 non-defining TU needs to trigger dynamic initialization (either
2429 because the variable is statically initialized, or a use of the
2430 variable in the defining TU will be executed before any uses in
2431 another TU), they can avoid this overhead with the
2432 @option{-fno-extern-tls-init} option.
2434 On targets that support symbol aliases, the default is
2435 @option{-fextern-tls-init}. On targets that do not support symbol
2436 aliases, the default is @option{-fno-extern-tls-init}.
2439 @itemx -fno-for-scope
2441 @opindex fno-for-scope
2442 If @option{-ffor-scope} is specified, the scope of variables declared in
2443 a @i{for-init-statement} is limited to the @code{for} loop itself,
2444 as specified by the C++ standard.
2445 If @option{-fno-for-scope} is specified, the scope of variables declared in
2446 a @i{for-init-statement} extends to the end of the enclosing scope,
2447 as was the case in old versions of G++, and other (traditional)
2448 implementations of C++.
2450 If neither flag is given, the default is to follow the standard,
2451 but to allow and give a warning for old-style code that would
2452 otherwise be invalid, or have different behavior.
2454 @item -fno-gnu-keywords
2455 @opindex fno-gnu-keywords
2456 Do not recognize @code{typeof} as a keyword, so that code can use this
2457 word as an identifier. You can use the keyword @code{__typeof__} instead.
2458 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2459 @option{-std=c++98}, @option{-std=c++11}, etc.
2461 @item -fno-implicit-templates
2462 @opindex fno-implicit-templates
2463 Never emit code for non-inline templates that are instantiated
2464 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2465 @xref{Template Instantiation}, for more information.
2467 @item -fno-implicit-inline-templates
2468 @opindex fno-implicit-inline-templates
2469 Don't emit code for implicit instantiations of inline templates, either.
2470 The default is to handle inlines differently so that compiles with and
2471 without optimization need the same set of explicit instantiations.
2473 @item -fno-implement-inlines
2474 @opindex fno-implement-inlines
2475 To save space, do not emit out-of-line copies of inline functions
2476 controlled by @code{#pragma implementation}. This causes linker
2477 errors if these functions are not inlined everywhere they are called.
2479 @item -fms-extensions
2480 @opindex fms-extensions
2481 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2482 int and getting a pointer to member function via non-standard syntax.
2484 @item -fnew-inheriting-ctors
2485 @opindex fnew-inheriting-ctors
2486 Enable the P0136 adjustment to the semantics of C++11 constructor
2487 inheritance. This is part of C++17 but also considered to be a Defect
2488 Report against C++11 and C++14. This flag is enabled by default
2489 unless @option{-fabi-version=10} or lower is specified.
2491 @item -fnew-ttp-matching
2492 @opindex fnew-ttp-matching
2493 Enable the P0522 resolution to Core issue 150, template template
2494 parameters and default arguments: this allows a template with default
2495 template arguments as an argument for a template template parameter
2496 with fewer template parameters. This flag is enabled by default for
2497 @option{-std=c++17}.
2499 @item -fno-nonansi-builtins
2500 @opindex fno-nonansi-builtins
2501 Disable built-in declarations of functions that are not mandated by
2502 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2503 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2506 @opindex fnothrow-opt
2507 Treat a @code{throw()} exception specification as if it were a
2508 @code{noexcept} specification to reduce or eliminate the text size
2509 overhead relative to a function with no exception specification. If
2510 the function has local variables of types with non-trivial
2511 destructors, the exception specification actually makes the
2512 function smaller because the EH cleanups for those variables can be
2513 optimized away. The semantic effect is that an exception thrown out of
2514 a function with such an exception specification results in a call
2515 to @code{terminate} rather than @code{unexpected}.
2517 @item -fno-operator-names
2518 @opindex fno-operator-names
2519 Do not treat the operator name keywords @code{and}, @code{bitand},
2520 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2521 synonyms as keywords.
2523 @item -fno-optional-diags
2524 @opindex fno-optional-diags
2525 Disable diagnostics that the standard says a compiler does not need to
2526 issue. Currently, the only such diagnostic issued by G++ is the one for
2527 a name having multiple meanings within a class.
2530 @opindex fpermissive
2531 Downgrade some diagnostics about nonconformant code from errors to
2532 warnings. Thus, using @option{-fpermissive} allows some
2533 nonconforming code to compile.
2535 @item -fno-pretty-templates
2536 @opindex fno-pretty-templates
2537 When an error message refers to a specialization of a function
2538 template, the compiler normally prints the signature of the
2539 template followed by the template arguments and any typedefs or
2540 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2541 rather than @code{void f(int)}) so that it's clear which template is
2542 involved. When an error message refers to a specialization of a class
2543 template, the compiler omits any template arguments that match
2544 the default template arguments for that template. If either of these
2545 behaviors make it harder to understand the error message rather than
2546 easier, you can use @option{-fno-pretty-templates} to disable them.
2550 Enable automatic template instantiation at link time. This option also
2551 implies @option{-fno-implicit-templates}. @xref{Template
2552 Instantiation}, for more information.
2556 Disable generation of information about every class with virtual
2557 functions for use by the C++ run-time type identification features
2558 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2559 of the language, you can save some space by using this flag. Note that
2560 exception handling uses the same information, but G++ generates it as
2561 needed. The @code{dynamic_cast} operator can still be used for casts that
2562 do not require run-time type information, i.e.@: casts to @code{void *} or to
2563 unambiguous base classes.
2565 @item -fsized-deallocation
2566 @opindex fsized-deallocation
2567 Enable the built-in global declarations
2569 void operator delete (void *, std::size_t) noexcept;
2570 void operator delete[] (void *, std::size_t) noexcept;
2572 as introduced in C++14. This is useful for user-defined replacement
2573 deallocation functions that, for example, use the size of the object
2574 to make deallocation faster. Enabled by default under
2575 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2576 warns about places that might want to add a definition.
2578 @item -fstrict-enums
2579 @opindex fstrict-enums
2580 Allow the compiler to optimize using the assumption that a value of
2581 enumerated type can only be one of the values of the enumeration (as
2582 defined in the C++ standard; basically, a value that can be
2583 represented in the minimum number of bits needed to represent all the
2584 enumerators). This assumption may not be valid if the program uses a
2585 cast to convert an arbitrary integer value to the enumerated type.
2587 @item -fstrong-eval-order
2588 @opindex fstrong-eval-order
2589 Evaluate member access, array subscripting, and shift expressions in
2590 left-to-right order, and evaluate assignment in right-to-left order,
2591 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2592 @option{-fstrong-eval-order=some} enables just the ordering of member
2593 access and shift expressions, and is the default without
2594 @option{-std=c++17}.
2596 @item -ftemplate-backtrace-limit=@var{n}
2597 @opindex ftemplate-backtrace-limit
2598 Set the maximum number of template instantiation notes for a single
2599 warning or error to @var{n}. The default value is 10.
2601 @item -ftemplate-depth=@var{n}
2602 @opindex ftemplate-depth
2603 Set the maximum instantiation depth for template classes to @var{n}.
2604 A limit on the template instantiation depth is needed to detect
2605 endless recursions during template class instantiation. ANSI/ISO C++
2606 conforming programs must not rely on a maximum depth greater than 17
2607 (changed to 1024 in C++11). The default value is 900, as the compiler
2608 can run out of stack space before hitting 1024 in some situations.
2610 @item -fno-threadsafe-statics
2611 @opindex fno-threadsafe-statics
2612 Do not emit the extra code to use the routines specified in the C++
2613 ABI for thread-safe initialization of local statics. You can use this
2614 option to reduce code size slightly in code that doesn't need to be
2617 @item -fuse-cxa-atexit
2618 @opindex fuse-cxa-atexit
2619 Register destructors for objects with static storage duration with the
2620 @code{__cxa_atexit} function rather than the @code{atexit} function.
2621 This option is required for fully standards-compliant handling of static
2622 destructors, but only works if your C library supports
2623 @code{__cxa_atexit}.
2625 @item -fno-use-cxa-get-exception-ptr
2626 @opindex fno-use-cxa-get-exception-ptr
2627 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2628 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2629 if the runtime routine is not available.
2631 @item -fvisibility-inlines-hidden
2632 @opindex fvisibility-inlines-hidden
2633 This switch declares that the user does not attempt to compare
2634 pointers to inline functions or methods where the addresses of the two functions
2635 are taken in different shared objects.
2637 The effect of this is that GCC may, effectively, mark inline methods with
2638 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2639 appear in the export table of a DSO and do not require a PLT indirection
2640 when used within the DSO@. Enabling this option can have a dramatic effect
2641 on load and link times of a DSO as it massively reduces the size of the
2642 dynamic export table when the library makes heavy use of templates.
2644 The behavior of this switch is not quite the same as marking the
2645 methods as hidden directly, because it does not affect static variables
2646 local to the function or cause the compiler to deduce that
2647 the function is defined in only one shared object.
2649 You may mark a method as having a visibility explicitly to negate the
2650 effect of the switch for that method. For example, if you do want to
2651 compare pointers to a particular inline method, you might mark it as
2652 having default visibility. Marking the enclosing class with explicit
2653 visibility has no effect.
2655 Explicitly instantiated inline methods are unaffected by this option
2656 as their linkage might otherwise cross a shared library boundary.
2657 @xref{Template Instantiation}.
2659 @item -fvisibility-ms-compat
2660 @opindex fvisibility-ms-compat
2661 This flag attempts to use visibility settings to make GCC's C++
2662 linkage model compatible with that of Microsoft Visual Studio.
2664 The flag makes these changes to GCC's linkage model:
2668 It sets the default visibility to @code{hidden}, like
2669 @option{-fvisibility=hidden}.
2672 Types, but not their members, are not hidden by default.
2675 The One Definition Rule is relaxed for types without explicit
2676 visibility specifications that are defined in more than one
2677 shared object: those declarations are permitted if they are
2678 permitted when this option is not used.
2681 In new code it is better to use @option{-fvisibility=hidden} and
2682 export those classes that are intended to be externally visible.
2683 Unfortunately it is possible for code to rely, perhaps accidentally,
2684 on the Visual Studio behavior.
2686 Among the consequences of these changes are that static data members
2687 of the same type with the same name but defined in different shared
2688 objects are different, so changing one does not change the other;
2689 and that pointers to function members defined in different shared
2690 objects may not compare equal. When this flag is given, it is a
2691 violation of the ODR to define types with the same name differently.
2695 Do not use weak symbol support, even if it is provided by the linker.
2696 By default, G++ uses weak symbols if they are available. This
2697 option exists only for testing, and should not be used by end-users;
2698 it results in inferior code and has no benefits. This option may
2699 be removed in a future release of G++.
2703 Do not search for header files in the standard directories specific to
2704 C++, but do still search the other standard directories. (This option
2705 is used when building the C++ library.)
2708 In addition, these optimization, warning, and code generation options
2709 have meanings only for C++ programs:
2712 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2715 Warn when G++ it generates code that is probably not compatible with
2716 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2717 ABI with each major release, normally @option{-Wabi} will warn only if
2718 there is a check added later in a release series for an ABI issue
2719 discovered since the initial release. @option{-Wabi} will warn about
2720 more things if an older ABI version is selected (with
2721 @option{-fabi-version=@var{n}}).
2723 @option{-Wabi} can also be used with an explicit version number to
2724 warn about compatibility with a particular @option{-fabi-version}
2725 level, e.g. @option{-Wabi=2} to warn about changes relative to
2726 @option{-fabi-version=2}.
2728 If an explicit version number is provided and
2729 @option{-fabi-compat-version} is not specified, the version number
2730 from this option is used for compatibility aliases. If no explicit
2731 version number is provided with this option, but
2732 @option{-fabi-compat-version} is specified, that version number is
2733 used for ABI warnings.
2735 Although an effort has been made to warn about
2736 all such cases, there are probably some cases that are not warned about,
2737 even though G++ is generating incompatible code. There may also be
2738 cases where warnings are emitted even though the code that is generated
2741 You should rewrite your code to avoid these warnings if you are
2742 concerned about the fact that code generated by G++ may not be binary
2743 compatible with code generated by other compilers.
2745 Known incompatibilities in @option{-fabi-version=2} (which was the
2746 default from GCC 3.4 to 4.9) include:
2751 A template with a non-type template parameter of reference type was
2752 mangled incorrectly:
2755 template <int &> struct S @{@};
2759 This was fixed in @option{-fabi-version=3}.
2762 SIMD vector types declared using @code{__attribute ((vector_size))} were
2763 mangled in a non-standard way that does not allow for overloading of
2764 functions taking vectors of different sizes.
2766 The mangling was changed in @option{-fabi-version=4}.
2769 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2770 qualifiers, and @code{decltype} of a plain declaration was folded away.
2772 These mangling issues were fixed in @option{-fabi-version=5}.
2775 Scoped enumerators passed as arguments to a variadic function are
2776 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2777 On most targets this does not actually affect the parameter passing
2778 ABI, as there is no way to pass an argument smaller than @code{int}.
2780 Also, the ABI changed the mangling of template argument packs,
2781 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2782 a class scope function used as a template argument.
2784 These issues were corrected in @option{-fabi-version=6}.
2787 Lambdas in default argument scope were mangled incorrectly, and the
2788 ABI changed the mangling of @code{nullptr_t}.
2790 These issues were corrected in @option{-fabi-version=7}.
2793 When mangling a function type with function-cv-qualifiers, the
2794 un-qualified function type was incorrectly treated as a substitution
2797 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2800 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2801 unaligned accesses. Note that this did not affect the ABI of a
2802 function with a @code{nullptr_t} parameter, as parameters have a
2805 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2808 Target-specific attributes that affect the identity of a type, such as
2809 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2810 did not affect the mangled name, leading to name collisions when
2811 function pointers were used as template arguments.
2813 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2817 It also warns about psABI-related changes. The known psABI changes at this
2823 For SysV/x86-64, unions with @code{long double} members are
2824 passed in memory as specified in psABI. For example:
2834 @code{union U} is always passed in memory.
2838 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2841 Warn when a type with an ABI tag is used in a context that does not
2842 have that ABI tag. See @ref{C++ Attributes} for more information
2845 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2846 @opindex Wctor-dtor-privacy
2847 @opindex Wno-ctor-dtor-privacy
2848 Warn when a class seems unusable because all the constructors or
2849 destructors in that class are private, and it has neither friends nor
2850 public static member functions. Also warn if there are no non-private
2851 methods, and there's at least one private member function that isn't
2852 a constructor or destructor.
2854 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2855 @opindex Wdelete-non-virtual-dtor
2856 @opindex Wno-delete-non-virtual-dtor
2857 Warn when @code{delete} is used to destroy an instance of a class that
2858 has virtual functions and non-virtual destructor. It is unsafe to delete
2859 an instance of a derived class through a pointer to a base class if the
2860 base class does not have a virtual destructor. This warning is enabled
2863 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2864 @opindex Wliteral-suffix
2865 @opindex Wno-literal-suffix
2866 Warn when a string or character literal is followed by a ud-suffix which does
2867 not begin with an underscore. As a conforming extension, GCC treats such
2868 suffixes as separate preprocessing tokens in order to maintain backwards
2869 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2873 #define __STDC_FORMAT_MACROS
2874 #include <inttypes.h>
2879 printf("My int64: %" PRId64"\n", i64);
2883 In this case, @code{PRId64} is treated as a separate preprocessing token.
2885 Additionally, warn when a user-defined literal operator is declared with
2886 a literal suffix identifier that doesn't begin with an underscore. Literal
2887 suffix identifiers that don't begin with an underscore are reserved for
2888 future standardization.
2890 This warning is enabled by default.
2892 @item -Wlto-type-mismatch
2893 @opindex Wlto-type-mismatch
2894 @opindex Wno-lto-type-mismatch
2896 During the link-time optimization warn about type mismatches in
2897 global declarations from different compilation units.
2898 Requires @option{-flto} to be enabled. Enabled by default.
2900 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2902 @opindex Wno-narrowing
2903 For C++11 and later standards, narrowing conversions are diagnosed by default,
2904 as required by the standard. A narrowing conversion from a constant produces
2905 an error, and a narrowing conversion from a non-constant produces a warning,
2906 but @option{-Wno-narrowing} suppresses the diagnostic.
2907 Note that this does not affect the meaning of well-formed code;
2908 narrowing conversions are still considered ill-formed in SFINAE contexts.
2910 With @option{-Wnarrowing} in C++98, warn when a narrowing
2911 conversion prohibited by C++11 occurs within
2915 int i = @{ 2.2 @}; // error: narrowing from double to int
2918 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2920 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2922 @opindex Wno-noexcept
2923 Warn when a noexcept-expression evaluates to false because of a call
2924 to a function that does not have a non-throwing exception
2925 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2926 the compiler to never throw an exception.
2928 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
2929 @opindex Wnoexcept-type
2930 @opindex Wno-noexcept-type
2931 Warn if the C++17 feature making @code{noexcept} part of a function
2932 type changes the mangled name of a symbol relative to C++14. Enabled
2933 by @option{-Wabi} and @option{-Wc++17-compat}.
2936 template <class T> void f(T t) @{ t(); @};
2938 void h() @{ f(g); @} // in C++14 calls f<void(*)()>, in C++17 calls f<void(*)()noexcept>
2941 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
2942 @opindex Wclass-memaccess
2943 Warn when the destination of a call to a raw memory function such as
2944 @code{memset} or @code{memcpy} is an object of class type writing into which
2945 might bypass the class non-trivial or deleted constructor or copy assignment,
2946 violate const-correctness or encapsulation, or corrupt the virtual table.
2947 Modifying the representation of such objects may violate invariants maintained
2948 by member functions of the class. For example, the call to @code{memset}
2949 below is undefined becase it modifies a non-trivial class object and is,
2950 therefore, diagnosed. The safe way to either initialize or clear the storage
2951 of objects of such types is by using the appropriate constructor or assignment
2952 operator, if one is available.
2954 std::string str = "abc";
2955 memset (&str, 0, 3);
2957 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
2959 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2960 @opindex Wnon-virtual-dtor
2961 @opindex Wno-non-virtual-dtor
2962 Warn when a class has virtual functions and an accessible non-virtual
2963 destructor itself or in an accessible polymorphic base class, in which
2964 case it is possible but unsafe to delete an instance of a derived
2965 class through a pointer to the class itself or base class. This
2966 warning is automatically enabled if @option{-Weffc++} is specified.
2968 @item -Wregister @r{(C++ and Objective-C++ only)}
2970 @opindex Wno-register
2971 Warn on uses of the @code{register} storage class specifier, except
2972 when it is part of the GNU @ref{Explicit Register Variables} extension.
2973 The use of the @code{register} keyword as storage class specifier has
2974 been deprecated in C++11 and removed in C++17.
2975 Enabled by default with @option{-std=c++17}.
2977 @item -Wreorder @r{(C++ and Objective-C++ only)}
2979 @opindex Wno-reorder
2980 @cindex reordering, warning
2981 @cindex warning for reordering of member initializers
2982 Warn when the order of member initializers given in the code does not
2983 match the order in which they must be executed. For instance:
2989 A(): j (0), i (1) @{ @}
2994 The compiler rearranges the member initializers for @code{i}
2995 and @code{j} to match the declaration order of the members, emitting
2996 a warning to that effect. This warning is enabled by @option{-Wall}.
2998 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2999 @opindex fext-numeric-literals
3000 @opindex fno-ext-numeric-literals
3001 Accept imaginary, fixed-point, or machine-defined
3002 literal number suffixes as GNU extensions.
3003 When this option is turned off these suffixes are treated
3004 as C++11 user-defined literal numeric suffixes.
3005 This is on by default for all pre-C++11 dialects and all GNU dialects:
3006 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3007 @option{-std=gnu++14}.
3008 This option is off by default
3009 for ISO C++11 onwards (@option{-std=c++11}, ...).
3012 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3015 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3018 Warn about violations of the following style guidelines from Scott Meyers'
3019 @cite{Effective C++} series of books:
3023 Define a copy constructor and an assignment operator for classes
3024 with dynamically-allocated memory.
3027 Prefer initialization to assignment in constructors.
3030 Have @code{operator=} return a reference to @code{*this}.
3033 Don't try to return a reference when you must return an object.
3036 Distinguish between prefix and postfix forms of increment and
3037 decrement operators.
3040 Never overload @code{&&}, @code{||}, or @code{,}.
3044 This option also enables @option{-Wnon-virtual-dtor}, which is also
3045 one of the effective C++ recommendations. However, the check is
3046 extended to warn about the lack of virtual destructor in accessible
3047 non-polymorphic bases classes too.
3049 When selecting this option, be aware that the standard library
3050 headers do not obey all of these guidelines; use @samp{grep -v}
3051 to filter out those warnings.
3053 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3054 @opindex Wstrict-null-sentinel
3055 @opindex Wno-strict-null-sentinel
3056 Warn about the use of an uncasted @code{NULL} as sentinel. When
3057 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3058 to @code{__null}. Although it is a null pointer constant rather than a
3059 null pointer, it is guaranteed to be of the same size as a pointer.
3060 But this use is not portable across different compilers.
3062 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3063 @opindex Wno-non-template-friend
3064 @opindex Wnon-template-friend
3065 Disable warnings when non-template friend functions are declared
3066 within a template. In very old versions of GCC that predate implementation
3067 of the ISO standard, declarations such as
3068 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3069 could be interpreted as a particular specialization of a template
3070 function; the warning exists to diagnose compatibility problems,
3071 and is enabled by default.
3073 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3074 @opindex Wold-style-cast
3075 @opindex Wno-old-style-cast
3076 Warn if an old-style (C-style) cast to a non-void type is used within
3077 a C++ program. The new-style casts (@code{dynamic_cast},
3078 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3079 less vulnerable to unintended effects and much easier to search for.
3081 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3082 @opindex Woverloaded-virtual
3083 @opindex Wno-overloaded-virtual
3084 @cindex overloaded virtual function, warning
3085 @cindex warning for overloaded virtual function
3086 Warn when a function declaration hides virtual functions from a
3087 base class. For example, in:
3094 struct B: public A @{
3099 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3110 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3111 @opindex Wno-pmf-conversions
3112 @opindex Wpmf-conversions
3113 Disable the diagnostic for converting a bound pointer to member function
3116 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3117 @opindex Wsign-promo
3118 @opindex Wno-sign-promo
3119 Warn when overload resolution chooses a promotion from unsigned or
3120 enumerated type to a signed type, over a conversion to an unsigned type of
3121 the same size. Previous versions of G++ tried to preserve
3122 unsignedness, but the standard mandates the current behavior.
3124 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3126 Warn when a primary template declaration is encountered. Some coding
3127 rules disallow templates, and this may be used to enforce that rule.
3128 The warning is inactive inside a system header file, such as the STL, so
3129 one can still use the STL. One may also instantiate or specialize
3132 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3133 @opindex Wmultiple-inheritance
3134 Warn when a class is defined with multiple direct base classes. Some
3135 coding rules disallow multiple inheritance, and this may be used to
3136 enforce that rule. The warning is inactive inside a system header file,
3137 such as the STL, so one can still use the STL. One may also define
3138 classes that indirectly use multiple inheritance.
3140 @item -Wvirtual-inheritance
3141 @opindex Wvirtual-inheritance
3142 Warn when a class is defined with a virtual direct base class. Some
3143 coding rules disallow multiple inheritance, and this may be used to
3144 enforce that rule. The warning is inactive inside a system header file,
3145 such as the STL, so one can still use the STL. One may also define
3146 classes that indirectly use virtual inheritance.
3149 @opindex Wnamespaces
3150 Warn when a namespace definition is opened. Some coding rules disallow
3151 namespaces, and this may be used to enforce that rule. The warning is
3152 inactive inside a system header file, such as the STL, so one can still
3153 use the STL. One may also use using directives and qualified names.
3155 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3157 @opindex Wno-terminate
3158 Disable the warning about a throw-expression that will immediately
3159 result in a call to @code{terminate}.
3162 @node Objective-C and Objective-C++ Dialect Options
3163 @section Options Controlling Objective-C and Objective-C++ Dialects
3165 @cindex compiler options, Objective-C and Objective-C++
3166 @cindex Objective-C and Objective-C++ options, command-line
3167 @cindex options, Objective-C and Objective-C++
3168 (NOTE: This manual does not describe the Objective-C and Objective-C++
3169 languages themselves. @xref{Standards,,Language Standards
3170 Supported by GCC}, for references.)
3172 This section describes the command-line options that are only meaningful
3173 for Objective-C and Objective-C++ programs. You can also use most of
3174 the language-independent GNU compiler options.
3175 For example, you might compile a file @file{some_class.m} like this:
3178 gcc -g -fgnu-runtime -O -c some_class.m
3182 In this example, @option{-fgnu-runtime} is an option meant only for
3183 Objective-C and Objective-C++ programs; you can use the other options with
3184 any language supported by GCC@.
3186 Note that since Objective-C is an extension of the C language, Objective-C
3187 compilations may also use options specific to the C front-end (e.g.,
3188 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3189 C++-specific options (e.g., @option{-Wabi}).
3191 Here is a list of options that are @emph{only} for compiling Objective-C
3192 and Objective-C++ programs:
3195 @item -fconstant-string-class=@var{class-name}
3196 @opindex fconstant-string-class
3197 Use @var{class-name} as the name of the class to instantiate for each
3198 literal string specified with the syntax @code{@@"@dots{}"}. The default
3199 class name is @code{NXConstantString} if the GNU runtime is being used, and
3200 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3201 @option{-fconstant-cfstrings} option, if also present, overrides the
3202 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3203 to be laid out as constant CoreFoundation strings.
3206 @opindex fgnu-runtime
3207 Generate object code compatible with the standard GNU Objective-C
3208 runtime. This is the default for most types of systems.
3210 @item -fnext-runtime
3211 @opindex fnext-runtime
3212 Generate output compatible with the NeXT runtime. This is the default
3213 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3214 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3217 @item -fno-nil-receivers
3218 @opindex fno-nil-receivers
3219 Assume that all Objective-C message dispatches (@code{[receiver
3220 message:arg]}) in this translation unit ensure that the receiver is
3221 not @code{nil}. This allows for more efficient entry points in the
3222 runtime to be used. This option is only available in conjunction with
3223 the NeXT runtime and ABI version 0 or 1.
3225 @item -fobjc-abi-version=@var{n}
3226 @opindex fobjc-abi-version
3227 Use version @var{n} of the Objective-C ABI for the selected runtime.
3228 This option is currently supported only for the NeXT runtime. In that
3229 case, Version 0 is the traditional (32-bit) ABI without support for
3230 properties and other Objective-C 2.0 additions. Version 1 is the
3231 traditional (32-bit) ABI with support for properties and other
3232 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3233 nothing is specified, the default is Version 0 on 32-bit target
3234 machines, and Version 2 on 64-bit target machines.
3236 @item -fobjc-call-cxx-cdtors
3237 @opindex fobjc-call-cxx-cdtors
3238 For each Objective-C class, check if any of its instance variables is a
3239 C++ object with a non-trivial default constructor. If so, synthesize a
3240 special @code{- (id) .cxx_construct} instance method which runs
3241 non-trivial default constructors on any such instance variables, in order,
3242 and then return @code{self}. Similarly, check if any instance variable
3243 is a C++ object with a non-trivial destructor, and if so, synthesize a
3244 special @code{- (void) .cxx_destruct} method which runs
3245 all such default destructors, in reverse order.
3247 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3248 methods thusly generated only operate on instance variables
3249 declared in the current Objective-C class, and not those inherited
3250 from superclasses. It is the responsibility of the Objective-C
3251 runtime to invoke all such methods in an object's inheritance
3252 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3253 by the runtime immediately after a new object instance is allocated;
3254 the @code{- (void) .cxx_destruct} methods are invoked immediately
3255 before the runtime deallocates an object instance.
3257 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3258 support for invoking the @code{- (id) .cxx_construct} and
3259 @code{- (void) .cxx_destruct} methods.
3261 @item -fobjc-direct-dispatch
3262 @opindex fobjc-direct-dispatch
3263 Allow fast jumps to the message dispatcher. On Darwin this is
3264 accomplished via the comm page.
3266 @item -fobjc-exceptions
3267 @opindex fobjc-exceptions
3268 Enable syntactic support for structured exception handling in
3269 Objective-C, similar to what is offered by C++. This option
3270 is required to use the Objective-C keywords @code{@@try},
3271 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3272 @code{@@synchronized}. This option is available with both the GNU
3273 runtime and the NeXT runtime (but not available in conjunction with
3274 the NeXT runtime on Mac OS X 10.2 and earlier).
3278 Enable garbage collection (GC) in Objective-C and Objective-C++
3279 programs. This option is only available with the NeXT runtime; the
3280 GNU runtime has a different garbage collection implementation that
3281 does not require special compiler flags.
3283 @item -fobjc-nilcheck
3284 @opindex fobjc-nilcheck
3285 For the NeXT runtime with version 2 of the ABI, check for a nil
3286 receiver in method invocations before doing the actual method call.
3287 This is the default and can be disabled using
3288 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3289 checked for nil in this way no matter what this flag is set to.
3290 Currently this flag does nothing when the GNU runtime, or an older
3291 version of the NeXT runtime ABI, is used.
3293 @item -fobjc-std=objc1
3295 Conform to the language syntax of Objective-C 1.0, the language
3296 recognized by GCC 4.0. This only affects the Objective-C additions to
3297 the C/C++ language; it does not affect conformance to C/C++ standards,
3298 which is controlled by the separate C/C++ dialect option flags. When
3299 this option is used with the Objective-C or Objective-C++ compiler,
3300 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3301 This is useful if you need to make sure that your Objective-C code can
3302 be compiled with older versions of GCC@.
3304 @item -freplace-objc-classes
3305 @opindex freplace-objc-classes
3306 Emit a special marker instructing @command{ld(1)} not to statically link in
3307 the resulting object file, and allow @command{dyld(1)} to load it in at
3308 run time instead. This is used in conjunction with the Fix-and-Continue
3309 debugging mode, where the object file in question may be recompiled and
3310 dynamically reloaded in the course of program execution, without the need
3311 to restart the program itself. Currently, Fix-and-Continue functionality
3312 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3317 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3318 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3319 compile time) with static class references that get initialized at load time,
3320 which improves run-time performance. Specifying the @option{-fzero-link} flag
3321 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3322 to be retained. This is useful in Zero-Link debugging mode, since it allows
3323 for individual class implementations to be modified during program execution.
3324 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3325 regardless of command-line options.
3327 @item -fno-local-ivars
3328 @opindex fno-local-ivars
3329 @opindex flocal-ivars
3330 By default instance variables in Objective-C can be accessed as if
3331 they were local variables from within the methods of the class they're
3332 declared in. This can lead to shadowing between instance variables
3333 and other variables declared either locally inside a class method or
3334 globally with the same name. Specifying the @option{-fno-local-ivars}
3335 flag disables this behavior thus avoiding variable shadowing issues.
3337 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3338 @opindex fivar-visibility
3339 Set the default instance variable visibility to the specified option
3340 so that instance variables declared outside the scope of any access
3341 modifier directives default to the specified visibility.
3345 Dump interface declarations for all classes seen in the source file to a
3346 file named @file{@var{sourcename}.decl}.
3348 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3349 @opindex Wassign-intercept
3350 @opindex Wno-assign-intercept
3351 Warn whenever an Objective-C assignment is being intercepted by the
3354 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3355 @opindex Wno-protocol
3357 If a class is declared to implement a protocol, a warning is issued for
3358 every method in the protocol that is not implemented by the class. The
3359 default behavior is to issue a warning for every method not explicitly
3360 implemented in the class, even if a method implementation is inherited
3361 from the superclass. If you use the @option{-Wno-protocol} option, then
3362 methods inherited from the superclass are considered to be implemented,
3363 and no warning is issued for them.
3365 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3367 @opindex Wno-selector
3368 Warn if multiple methods of different types for the same selector are
3369 found during compilation. The check is performed on the list of methods
3370 in the final stage of compilation. Additionally, a check is performed
3371 for each selector appearing in a @code{@@selector(@dots{})}
3372 expression, and a corresponding method for that selector has been found
3373 during compilation. Because these checks scan the method table only at
3374 the end of compilation, these warnings are not produced if the final
3375 stage of compilation is not reached, for example because an error is
3376 found during compilation, or because the @option{-fsyntax-only} option is
3379 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3380 @opindex Wstrict-selector-match
3381 @opindex Wno-strict-selector-match
3382 Warn if multiple methods with differing argument and/or return types are
3383 found for a given selector when attempting to send a message using this
3384 selector to a receiver of type @code{id} or @code{Class}. When this flag
3385 is off (which is the default behavior), the compiler omits such warnings
3386 if any differences found are confined to types that share the same size
3389 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3390 @opindex Wundeclared-selector
3391 @opindex Wno-undeclared-selector
3392 Warn if a @code{@@selector(@dots{})} expression referring to an
3393 undeclared selector is found. A selector is considered undeclared if no
3394 method with that name has been declared before the
3395 @code{@@selector(@dots{})} expression, either explicitly in an
3396 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3397 an @code{@@implementation} section. This option always performs its
3398 checks as soon as a @code{@@selector(@dots{})} expression is found,
3399 while @option{-Wselector} only performs its checks in the final stage of
3400 compilation. This also enforces the coding style convention
3401 that methods and selectors must be declared before being used.
3403 @item -print-objc-runtime-info
3404 @opindex print-objc-runtime-info
3405 Generate C header describing the largest structure that is passed by
3410 @node Diagnostic Message Formatting Options
3411 @section Options to Control Diagnostic Messages Formatting
3412 @cindex options to control diagnostics formatting
3413 @cindex diagnostic messages
3414 @cindex message formatting
3416 Traditionally, diagnostic messages have been formatted irrespective of
3417 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3418 options described below
3419 to control the formatting algorithm for diagnostic messages,
3420 e.g.@: how many characters per line, how often source location
3421 information should be reported. Note that some language front ends may not
3422 honor these options.
3425 @item -fmessage-length=@var{n}
3426 @opindex fmessage-length
3427 Try to format error messages so that they fit on lines of about
3428 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3429 done; each error message appears on a single line. This is the
3430 default for all front ends.
3432 @item -fdiagnostics-show-location=once
3433 @opindex fdiagnostics-show-location
3434 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3435 reporter to emit source location information @emph{once}; that is, in
3436 case the message is too long to fit on a single physical line and has to
3437 be wrapped, the source location won't be emitted (as prefix) again,
3438 over and over, in subsequent continuation lines. This is the default
3441 @item -fdiagnostics-show-location=every-line
3442 Only meaningful in line-wrapping mode. Instructs the diagnostic
3443 messages reporter to emit the same source location information (as
3444 prefix) for physical lines that result from the process of breaking
3445 a message which is too long to fit on a single line.
3447 @item -fdiagnostics-color[=@var{WHEN}]
3448 @itemx -fno-diagnostics-color
3449 @opindex fdiagnostics-color
3450 @cindex highlight, color
3451 @vindex GCC_COLORS @r{environment variable}
3452 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3453 or @samp{auto}. The default depends on how the compiler has been configured,
3454 it can be any of the above @var{WHEN} options or also @samp{never}
3455 if @env{GCC_COLORS} environment variable isn't present in the environment,
3456 and @samp{auto} otherwise.
3457 @samp{auto} means to use color only when the standard error is a terminal.
3458 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3459 aliases for @option{-fdiagnostics-color=always} and
3460 @option{-fdiagnostics-color=never}, respectively.
3462 The colors are defined by the environment variable @env{GCC_COLORS}.
3463 Its value is a colon-separated list of capabilities and Select Graphic
3464 Rendition (SGR) substrings. SGR commands are interpreted by the
3465 terminal or terminal emulator. (See the section in the documentation
3466 of your text terminal for permitted values and their meanings as
3467 character attributes.) These substring values are integers in decimal
3468 representation and can be concatenated with semicolons.
3469 Common values to concatenate include
3471 @samp{4} for underline,
3473 @samp{7} for inverse,
3474 @samp{39} for default foreground color,
3475 @samp{30} to @samp{37} for foreground colors,
3476 @samp{90} to @samp{97} for 16-color mode foreground colors,
3477 @samp{38;5;0} to @samp{38;5;255}
3478 for 88-color and 256-color modes foreground colors,
3479 @samp{49} for default background color,
3480 @samp{40} to @samp{47} for background colors,
3481 @samp{100} to @samp{107} for 16-color mode background colors,
3482 and @samp{48;5;0} to @samp{48;5;255}
3483 for 88-color and 256-color modes background colors.
3485 The default @env{GCC_COLORS} is
3487 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3488 quote=01:fixit-insert=32:fixit-delete=31:\
3489 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3493 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3494 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3495 @samp{01} is bold, and @samp{31} is red.
3496 Setting @env{GCC_COLORS} to the empty string disables colors.
3497 Supported capabilities are as follows.
3501 @vindex error GCC_COLORS @r{capability}
3502 SGR substring for error: markers.
3505 @vindex warning GCC_COLORS @r{capability}
3506 SGR substring for warning: markers.
3509 @vindex note GCC_COLORS @r{capability}
3510 SGR substring for note: markers.
3513 @vindex range1 GCC_COLORS @r{capability}
3514 SGR substring for first additional range.
3517 @vindex range2 GCC_COLORS @r{capability}
3518 SGR substring for second additional range.
3521 @vindex locus GCC_COLORS @r{capability}
3522 SGR substring for location information, @samp{file:line} or
3523 @samp{file:line:column} etc.
3526 @vindex quote GCC_COLORS @r{capability}
3527 SGR substring for information printed within quotes.
3530 @vindex fixit-insert GCC_COLORS @r{capability}
3531 SGR substring for fix-it hints suggesting text to
3532 be inserted or replaced.
3535 @vindex fixit-delete GCC_COLORS @r{capability}
3536 SGR substring for fix-it hints suggesting text to
3539 @item diff-filename=
3540 @vindex diff-filename GCC_COLORS @r{capability}
3541 SGR substring for filename headers within generated patches.
3544 @vindex diff-hunk GCC_COLORS @r{capability}
3545 SGR substring for the starts of hunks within generated patches.
3548 @vindex diff-delete GCC_COLORS @r{capability}
3549 SGR substring for deleted lines within generated patches.
3552 @vindex diff-insert GCC_COLORS @r{capability}
3553 SGR substring for inserted lines within generated patches.
3556 @vindex type-diff GCC_COLORS @r{capability}
3557 SGR substring for highlighting mismatching types within template
3558 arguments in the C++ frontend.
3561 @item -fno-diagnostics-show-option
3562 @opindex fno-diagnostics-show-option
3563 @opindex fdiagnostics-show-option
3564 By default, each diagnostic emitted includes text indicating the
3565 command-line option that directly controls the diagnostic (if such an
3566 option is known to the diagnostic machinery). Specifying the
3567 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3569 @item -fno-diagnostics-show-caret
3570 @opindex fno-diagnostics-show-caret
3571 @opindex fdiagnostics-show-caret
3572 By default, each diagnostic emitted includes the original source line
3573 and a caret @samp{^} indicating the column. This option suppresses this
3574 information. The source line is truncated to @var{n} characters, if
3575 the @option{-fmessage-length=n} option is given. When the output is done
3576 to the terminal, the width is limited to the width given by the
3577 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3579 @item -fdiagnostics-parseable-fixits
3580 @opindex fdiagnostics-parseable-fixits
3581 Emit fix-it hints in a machine-parseable format, suitable for consumption
3582 by IDEs. For each fix-it, a line will be printed after the relevant
3583 diagnostic, starting with the string ``fix-it:''. For example:
3586 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3589 The location is expressed as a half-open range, expressed as a count of
3590 bytes, starting at byte 1 for the initial column. In the above example,
3591 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3595 00000000011111111112222222222
3596 12345678901234567890123456789
3597 gtk_widget_showall (dlg);
3602 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3603 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3604 (e.g. vertical tab as ``\013'').
3606 An empty replacement string indicates that the given range is to be removed.
3607 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3608 be inserted at the given position.
3610 @item -fdiagnostics-generate-patch
3611 @opindex fdiagnostics-generate-patch
3612 Print fix-it hints to stderr in unified diff format, after any diagnostics
3613 are printed. For example:
3620 void show_cb(GtkDialog *dlg)
3622 - gtk_widget_showall(dlg);
3623 + gtk_widget_show_all(dlg);
3628 The diff may or may not be colorized, following the same rules
3629 as for diagnostics (see @option{-fdiagnostics-color}).
3631 @item -fdiagnostics-show-template-tree
3632 @opindex fdiagnostics-show-template-tree
3634 In the C++ frontend, when printing diagnostics showing mismatching
3635 template types, such as:
3638 could not convert 'std::map<int, std::vector<double> >()'
3639 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3642 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3643 tree-like structure showing the common and differing parts of the types,
3653 The parts that differ are highlighted with color (``double'' and
3654 ``float'' in this case).
3656 @item -fno-elide-type
3657 @opindex fno-elide-type
3658 @opindex felide-type
3659 By default when the C++ frontend prints diagnostics showing mismatching
3660 template types, common parts of the types are printed as ``[...]'' to
3661 simplify the error message. For example:
3664 could not convert 'std::map<int, std::vector<double> >()'
3665 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3668 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
3669 This flag also affects the output of the
3670 @option{-fdiagnostics-show-template-tree} flag.
3672 @item -fno-show-column
3673 @opindex fno-show-column
3674 Do not print column numbers in diagnostics. This may be necessary if
3675 diagnostics are being scanned by a program that does not understand the
3676 column numbers, such as @command{dejagnu}.
3680 @node Warning Options
3681 @section Options to Request or Suppress Warnings
3682 @cindex options to control warnings
3683 @cindex warning messages
3684 @cindex messages, warning
3685 @cindex suppressing warnings
3687 Warnings are diagnostic messages that report constructions that
3688 are not inherently erroneous but that are risky or suggest there
3689 may have been an error.
3691 The following language-independent options do not enable specific
3692 warnings but control the kinds of diagnostics produced by GCC@.
3695 @cindex syntax checking
3697 @opindex fsyntax-only
3698 Check the code for syntax errors, but don't do anything beyond that.
3700 @item -fmax-errors=@var{n}
3701 @opindex fmax-errors
3702 Limits the maximum number of error messages to @var{n}, at which point
3703 GCC bails out rather than attempting to continue processing the source
3704 code. If @var{n} is 0 (the default), there is no limit on the number
3705 of error messages produced. If @option{-Wfatal-errors} is also
3706 specified, then @option{-Wfatal-errors} takes precedence over this
3711 Inhibit all warning messages.
3716 Make all warnings into errors.
3721 Make the specified warning into an error. The specifier for a warning
3722 is appended; for example @option{-Werror=switch} turns the warnings
3723 controlled by @option{-Wswitch} into errors. This switch takes a
3724 negative form, to be used to negate @option{-Werror} for specific
3725 warnings; for example @option{-Wno-error=switch} makes
3726 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3729 The warning message for each controllable warning includes the
3730 option that controls the warning. That option can then be used with
3731 @option{-Werror=} and @option{-Wno-error=} as described above.
3732 (Printing of the option in the warning message can be disabled using the
3733 @option{-fno-diagnostics-show-option} flag.)
3735 Note that specifying @option{-Werror=}@var{foo} automatically implies
3736 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3739 @item -Wfatal-errors
3740 @opindex Wfatal-errors
3741 @opindex Wno-fatal-errors
3742 This option causes the compiler to abort compilation on the first error
3743 occurred rather than trying to keep going and printing further error
3748 You can request many specific warnings with options beginning with
3749 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3750 implicit declarations. Each of these specific warning options also
3751 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3752 example, @option{-Wno-implicit}. This manual lists only one of the
3753 two forms, whichever is not the default. For further
3754 language-specific options also refer to @ref{C++ Dialect Options} and
3755 @ref{Objective-C and Objective-C++ Dialect Options}.
3757 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3758 options, such as @option{-Wunused}, which may turn on further options,
3759 such as @option{-Wunused-value}. The combined effect of positive and
3760 negative forms is that more specific options have priority over less
3761 specific ones, independently of their position in the command-line. For
3762 options of the same specificity, the last one takes effect. Options
3763 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3764 as if they appeared at the end of the command-line.
3766 When an unrecognized warning option is requested (e.g.,
3767 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3768 that the option is not recognized. However, if the @option{-Wno-} form
3769 is used, the behavior is slightly different: no diagnostic is
3770 produced for @option{-Wno-unknown-warning} unless other diagnostics
3771 are being produced. This allows the use of new @option{-Wno-} options
3772 with old compilers, but if something goes wrong, the compiler
3773 warns that an unrecognized option is present.
3780 Issue all the warnings demanded by strict ISO C and ISO C++;
3781 reject all programs that use forbidden extensions, and some other
3782 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3783 version of the ISO C standard specified by any @option{-std} option used.
3785 Valid ISO C and ISO C++ programs should compile properly with or without
3786 this option (though a rare few require @option{-ansi} or a
3787 @option{-std} option specifying the required version of ISO C)@. However,
3788 without this option, certain GNU extensions and traditional C and C++
3789 features are supported as well. With this option, they are rejected.
3791 @option{-Wpedantic} does not cause warning messages for use of the
3792 alternate keywords whose names begin and end with @samp{__}. Pedantic
3793 warnings are also disabled in the expression that follows
3794 @code{__extension__}. However, only system header files should use
3795 these escape routes; application programs should avoid them.
3796 @xref{Alternate Keywords}.
3798 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3799 C conformance. They soon find that it does not do quite what they want:
3800 it finds some non-ISO practices, but not all---only those for which
3801 ISO C @emph{requires} a diagnostic, and some others for which
3802 diagnostics have been added.
3804 A feature to report any failure to conform to ISO C might be useful in
3805 some instances, but would require considerable additional work and would
3806 be quite different from @option{-Wpedantic}. We don't have plans to
3807 support such a feature in the near future.
3809 Where the standard specified with @option{-std} represents a GNU
3810 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3811 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3812 extended dialect is based. Warnings from @option{-Wpedantic} are given
3813 where they are required by the base standard. (It does not make sense
3814 for such warnings to be given only for features not in the specified GNU
3815 C dialect, since by definition the GNU dialects of C include all
3816 features the compiler supports with the given option, and there would be
3817 nothing to warn about.)
3819 @item -pedantic-errors
3820 @opindex pedantic-errors
3821 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3822 requires a diagnostic, in some cases where there is undefined behavior
3823 at compile-time and in some other cases that do not prevent compilation
3824 of programs that are valid according to the standard. This is not
3825 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3826 by this option and not enabled by the latter and vice versa.
3831 This enables all the warnings about constructions that some users
3832 consider questionable, and that are easy to avoid (or modify to
3833 prevent the warning), even in conjunction with macros. This also
3834 enables some language-specific warnings described in @ref{C++ Dialect
3835 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3837 @option{-Wall} turns on the following warning flags:
3839 @gccoptlist{-Waddress @gol
3840 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3842 -Wbool-operation @gol
3843 -Wc++11-compat -Wc++14-compat @gol
3844 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
3845 -Wchar-subscripts @gol
3847 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3848 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3850 -Wint-in-bool-context @gol
3851 -Wimplicit @r{(C and Objective-C only)} @gol
3852 -Wimplicit-int @r{(C and Objective-C only)} @gol
3853 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3854 -Winit-self @r{(only for C++)} @gol
3855 -Wlogical-not-parentheses @gol
3856 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3857 -Wmaybe-uninitialized @gol
3858 -Wmemset-elt-size @gol
3859 -Wmemset-transposed-args @gol
3860 -Wmisleading-indentation @r{(only for C/C++)} @gol
3861 -Wmissing-braces @r{(only for C/ObjC)} @gol
3862 -Wmultistatement-macros @gol
3863 -Wnarrowing @r{(only for C++)} @gol
3865 -Wnonnull-compare @gol
3871 -Wsequence-point @gol
3872 -Wsign-compare @r{(only in C++)} @gol
3873 -Wsizeof-pointer-div @gol
3874 -Wsizeof-pointer-memaccess @gol
3875 -Wstrict-aliasing @gol
3876 -Wstrict-overflow=1 @gol
3878 -Wtautological-compare @gol
3880 -Wuninitialized @gol
3881 -Wunknown-pragmas @gol
3882 -Wunused-function @gol
3885 -Wunused-variable @gol
3886 -Wvolatile-register-var @gol
3889 Note that some warning flags are not implied by @option{-Wall}. Some of
3890 them warn about constructions that users generally do not consider
3891 questionable, but which occasionally you might wish to check for;
3892 others warn about constructions that are necessary or hard to avoid in
3893 some cases, and there is no simple way to modify the code to suppress
3894 the warning. Some of them are enabled by @option{-Wextra} but many of
3895 them must be enabled individually.
3901 This enables some extra warning flags that are not enabled by
3902 @option{-Wall}. (This option used to be called @option{-W}. The older
3903 name is still supported, but the newer name is more descriptive.)
3905 @gccoptlist{-Wclobbered @gol
3907 -Wignored-qualifiers @gol
3908 -Wimplicit-fallthrough=3 @gol
3909 -Wmissing-field-initializers @gol
3910 -Wmissing-parameter-type @r{(C only)} @gol
3911 -Wold-style-declaration @r{(C only)} @gol
3912 -Woverride-init @gol
3913 -Wsign-compare @r{(C only)} @gol
3915 -Wuninitialized @gol
3916 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3917 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3918 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3921 The option @option{-Wextra} also prints warning messages for the
3927 A pointer is compared against integer zero with @code{<}, @code{<=},
3928 @code{>}, or @code{>=}.
3931 (C++ only) An enumerator and a non-enumerator both appear in a
3932 conditional expression.
3935 (C++ only) Ambiguous virtual bases.
3938 (C++ only) Subscripting an array that has been declared @code{register}.
3941 (C++ only) Taking the address of a variable that has been declared
3945 (C++ only) A base class is not initialized in the copy constructor
3950 @item -Wchar-subscripts
3951 @opindex Wchar-subscripts
3952 @opindex Wno-char-subscripts
3953 Warn if an array subscript has type @code{char}. This is a common cause
3954 of error, as programmers often forget that this type is signed on some
3956 This warning is enabled by @option{-Wall}.
3960 Warn about an invalid memory access that is found by Pointer Bounds Checker
3961 (@option{-fcheck-pointer-bounds}).
3963 @item -Wno-coverage-mismatch
3964 @opindex Wno-coverage-mismatch
3965 Warn if feedback profiles do not match when using the
3966 @option{-fprofile-use} option.
3967 If a source file is changed between compiling with @option{-fprofile-gen} and
3968 with @option{-fprofile-use}, the files with the profile feedback can fail
3969 to match the source file and GCC cannot use the profile feedback
3970 information. By default, this warning is enabled and is treated as an
3971 error. @option{-Wno-coverage-mismatch} can be used to disable the
3972 warning or @option{-Wno-error=coverage-mismatch} can be used to
3973 disable the error. Disabling the error for this warning can result in
3974 poorly optimized code and is useful only in the
3975 case of very minor changes such as bug fixes to an existing code-base.
3976 Completely disabling the warning is not recommended.
3979 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3981 Suppress warning messages emitted by @code{#warning} directives.
3983 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3984 @opindex Wdouble-promotion
3985 @opindex Wno-double-promotion
3986 Give a warning when a value of type @code{float} is implicitly
3987 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3988 floating-point unit implement @code{float} in hardware, but emulate
3989 @code{double} in software. On such a machine, doing computations
3990 using @code{double} values is much more expensive because of the
3991 overhead required for software emulation.
3993 It is easy to accidentally do computations with @code{double} because
3994 floating-point literals are implicitly of type @code{double}. For
3998 float area(float radius)
4000 return 3.14159 * radius * radius;
4004 the compiler performs the entire computation with @code{double}
4005 because the floating-point literal is a @code{double}.
4007 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4008 @opindex Wduplicate-decl-specifier
4009 @opindex Wno-duplicate-decl-specifier
4010 Warn if a declaration has duplicate @code{const}, @code{volatile},
4011 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4015 @itemx -Wformat=@var{n}
4018 @opindex ffreestanding
4019 @opindex fno-builtin
4021 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4022 the arguments supplied have types appropriate to the format string
4023 specified, and that the conversions specified in the format string make
4024 sense. This includes standard functions, and others specified by format
4025 attributes (@pxref{Function Attributes}), in the @code{printf},
4026 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4027 not in the C standard) families (or other target-specific families).
4028 Which functions are checked without format attributes having been
4029 specified depends on the standard version selected, and such checks of
4030 functions without the attribute specified are disabled by
4031 @option{-ffreestanding} or @option{-fno-builtin}.
4033 The formats are checked against the format features supported by GNU
4034 libc version 2.2. These include all ISO C90 and C99 features, as well
4035 as features from the Single Unix Specification and some BSD and GNU
4036 extensions. Other library implementations may not support all these
4037 features; GCC does not support warning about features that go beyond a
4038 particular library's limitations. However, if @option{-Wpedantic} is used
4039 with @option{-Wformat}, warnings are given about format features not
4040 in the selected standard version (but not for @code{strfmon} formats,
4041 since those are not in any version of the C standard). @xref{C Dialect
4042 Options,,Options Controlling C Dialect}.
4049 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4050 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4051 @option{-Wformat} also checks for null format arguments for several
4052 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4053 aspects of this level of format checking can be disabled by the
4054 options: @option{-Wno-format-contains-nul},
4055 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4056 @option{-Wformat} is enabled by @option{-Wall}.
4058 @item -Wno-format-contains-nul
4059 @opindex Wno-format-contains-nul
4060 @opindex Wformat-contains-nul
4061 If @option{-Wformat} is specified, do not warn about format strings that
4064 @item -Wno-format-extra-args
4065 @opindex Wno-format-extra-args
4066 @opindex Wformat-extra-args
4067 If @option{-Wformat} is specified, do not warn about excess arguments to a
4068 @code{printf} or @code{scanf} format function. The C standard specifies
4069 that such arguments are ignored.
4071 Where the unused arguments lie between used arguments that are
4072 specified with @samp{$} operand number specifications, normally
4073 warnings are still given, since the implementation could not know what
4074 type to pass to @code{va_arg} to skip the unused arguments. However,
4075 in the case of @code{scanf} formats, this option suppresses the
4076 warning if the unused arguments are all pointers, since the Single
4077 Unix Specification says that such unused arguments are allowed.
4079 @item -Wformat-overflow
4080 @itemx -Wformat-overflow=@var{level}
4081 @opindex Wformat-overflow
4082 @opindex Wno-format-overflow
4083 Warn about calls to formatted input/output functions such as @code{sprintf}
4084 and @code{vsprintf} that might overflow the destination buffer. When the
4085 exact number of bytes written by a format directive cannot be determined
4086 at compile-time it is estimated based on heuristics that depend on the
4087 @var{level} argument and on optimization. While enabling optimization
4088 will in most cases improve the accuracy of the warning, it may also
4089 result in false positives.
4092 @item -Wformat-overflow
4093 @item -Wformat-overflow=1
4094 @opindex Wformat-overflow
4095 @opindex Wno-format-overflow
4096 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4097 employs a conservative approach that warns only about calls that most
4098 likely overflow the buffer. At this level, numeric arguments to format
4099 directives with unknown values are assumed to have the value of one, and
4100 strings of unknown length to be empty. Numeric arguments that are known
4101 to be bounded to a subrange of their type, or string arguments whose output
4102 is bounded either by their directive's precision or by a finite set of
4103 string literals, are assumed to take on the value within the range that
4104 results in the most bytes on output. For example, the call to @code{sprintf}
4105 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4106 the terminating NUL character (@code{'\0'}) appended by the function
4107 to the destination buffer will be written past its end. Increasing
4108 the size of the buffer by a single byte is sufficient to avoid the
4109 warning, though it may not be sufficient to avoid the overflow.
4112 void f (int a, int b)
4115 sprintf (buf, "a = %i, b = %i\n", a, b);
4119 @item -Wformat-overflow=2
4120 Level @var{2} warns also about calls that might overflow the destination
4121 buffer given an argument of sufficient length or magnitude. At level
4122 @var{2}, unknown numeric arguments are assumed to have the minimum
4123 representable value for signed types with a precision greater than 1, and
4124 the maximum representable value otherwise. Unknown string arguments whose
4125 length cannot be assumed to be bounded either by the directive's precision,
4126 or by a finite set of string literals they may evaluate to, or the character
4127 array they may point to, are assumed to be 1 character long.
4129 At level @var{2}, the call in the example above is again diagnosed, but
4130 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4131 @code{%i} directive will write some of its digits beyond the end of
4132 the destination buffer. To make the call safe regardless of the values
4133 of the two variables, the size of the destination buffer must be increased
4134 to at least 34 bytes. GCC includes the minimum size of the buffer in
4135 an informational note following the warning.
4137 An alternative to increasing the size of the destination buffer is to
4138 constrain the range of formatted values. The maximum length of string
4139 arguments can be bounded by specifying the precision in the format
4140 directive. When numeric arguments of format directives can be assumed
4141 to be bounded by less than the precision of their type, choosing
4142 an appropriate length modifier to the format specifier will reduce
4143 the required buffer size. For example, if @var{a} and @var{b} in the
4144 example above can be assumed to be within the precision of
4145 the @code{short int} type then using either the @code{%hi} format
4146 directive or casting the argument to @code{short} reduces the maximum
4147 required size of the buffer to 24 bytes.
4150 void f (int a, int b)
4153 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4158 @item -Wno-format-zero-length
4159 @opindex Wno-format-zero-length
4160 @opindex Wformat-zero-length
4161 If @option{-Wformat} is specified, do not warn about zero-length formats.
4162 The C standard specifies that zero-length formats are allowed.
4167 Enable @option{-Wformat} plus additional format checks. Currently
4168 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4171 @item -Wformat-nonliteral
4172 @opindex Wformat-nonliteral
4173 @opindex Wno-format-nonliteral
4174 If @option{-Wformat} is specified, also warn if the format string is not a
4175 string literal and so cannot be checked, unless the format function
4176 takes its format arguments as a @code{va_list}.
4178 @item -Wformat-security
4179 @opindex Wformat-security
4180 @opindex Wno-format-security
4181 If @option{-Wformat} is specified, also warn about uses of format
4182 functions that represent possible security problems. At present, this
4183 warns about calls to @code{printf} and @code{scanf} functions where the
4184 format string is not a string literal and there are no format arguments,
4185 as in @code{printf (foo);}. This may be a security hole if the format
4186 string came from untrusted input and contains @samp{%n}. (This is
4187 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4188 in future warnings may be added to @option{-Wformat-security} that are not
4189 included in @option{-Wformat-nonliteral}.)
4191 @item -Wformat-signedness
4192 @opindex Wformat-signedness
4193 @opindex Wno-format-signedness
4194 If @option{-Wformat} is specified, also warn if the format string
4195 requires an unsigned argument and the argument is signed and vice versa.
4197 @item -Wformat-truncation
4198 @itemx -Wformat-truncation=@var{level}
4199 @opindex Wformat-truncation
4200 @opindex Wno-format-truncation
4201 Warn about calls to formatted input/output functions such as @code{snprintf}
4202 and @code{vsnprintf} that might result in output truncation. When the exact
4203 number of bytes written by a format directive cannot be determined at
4204 compile-time it is estimated based on heuristics that depend on
4205 the @var{level} argument and on optimization. While enabling optimization
4206 will in most cases improve the accuracy of the warning, it may also result
4207 in false positives. Except as noted otherwise, the option uses the same
4208 logic @option{-Wformat-overflow}.
4211 @item -Wformat-truncation
4212 @item -Wformat-truncation=1
4213 @opindex Wformat-truncation
4214 @opindex Wno-format-overflow
4215 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4216 employs a conservative approach that warns only about calls to bounded
4217 functions whose return value is unused and that will most likely result
4218 in output truncation.
4220 @item -Wformat-truncation=2
4221 Level @var{2} warns also about calls to bounded functions whose return
4222 value is used and that might result in truncation given an argument of
4223 sufficient length or magnitude.
4227 @opindex Wformat-y2k
4228 @opindex Wno-format-y2k
4229 If @option{-Wformat} is specified, also warn about @code{strftime}
4230 formats that may yield only a two-digit year.
4235 @opindex Wno-nonnull
4236 Warn about passing a null pointer for arguments marked as
4237 requiring a non-null value by the @code{nonnull} function attribute.
4239 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4240 can be disabled with the @option{-Wno-nonnull} option.
4242 @item -Wnonnull-compare
4243 @opindex Wnonnull-compare
4244 @opindex Wno-nonnull-compare
4245 Warn when comparing an argument marked with the @code{nonnull}
4246 function attribute against null inside the function.
4248 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4249 can be disabled with the @option{-Wno-nonnull-compare} option.
4251 @item -Wnull-dereference
4252 @opindex Wnull-dereference
4253 @opindex Wno-null-dereference
4254 Warn if the compiler detects paths that trigger erroneous or
4255 undefined behavior due to dereferencing a null pointer. This option
4256 is only active when @option{-fdelete-null-pointer-checks} is active,
4257 which is enabled by optimizations in most targets. The precision of
4258 the warnings depends on the optimization options used.
4260 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4262 @opindex Wno-init-self
4263 Warn about uninitialized variables that are initialized with themselves.
4264 Note this option can only be used with the @option{-Wuninitialized} option.
4266 For example, GCC warns about @code{i} being uninitialized in the
4267 following snippet only when @option{-Winit-self} has been specified:
4278 This warning is enabled by @option{-Wall} in C++.
4280 @item -Wimplicit-int @r{(C and Objective-C only)}
4281 @opindex Wimplicit-int
4282 @opindex Wno-implicit-int
4283 Warn when a declaration does not specify a type.
4284 This warning is enabled by @option{-Wall}.
4286 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4287 @opindex Wimplicit-function-declaration
4288 @opindex Wno-implicit-function-declaration
4289 Give a warning whenever a function is used before being declared. In
4290 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4291 enabled by default and it is made into an error by
4292 @option{-pedantic-errors}. This warning is also enabled by
4295 @item -Wimplicit @r{(C and Objective-C only)}
4297 @opindex Wno-implicit
4298 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4299 This warning is enabled by @option{-Wall}.
4301 @item -Wimplicit-fallthrough
4302 @opindex Wimplicit-fallthrough
4303 @opindex Wno-implicit-fallthrough
4304 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4305 and @option{-Wno-implicit-fallthrough} is the same as
4306 @option{-Wimplicit-fallthrough=0}.
4308 @item -Wimplicit-fallthrough=@var{n}
4309 @opindex Wimplicit-fallthrough=
4310 Warn when a switch case falls through. For example:
4328 This warning does not warn when the last statement of a case cannot
4329 fall through, e.g. when there is a return statement or a call to function
4330 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4331 also takes into account control flow statements, such as ifs, and only
4332 warns when appropriate. E.g.@:
4342 @} else if (i < 1) @{
4352 Since there are occasions where a switch case fall through is desirable,
4353 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4354 to be used along with a null statement to suppress this warning that
4355 would normally occur:
4363 __attribute__ ((fallthrough));
4370 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4371 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4372 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4373 Instead of these attributes, it is also possible to add a fallthrough comment
4374 to silence the warning. The whole body of the C or C++ style comment should
4375 match the given regular expressions listed below. The option argument @var{n}
4376 specifies what kind of comments are accepted:
4380 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4382 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4383 expression, any comment is used as fallthrough comment.
4385 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4386 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4388 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4389 following regular expressions:
4393 @item @code{-fallthrough}
4395 @item @code{@@fallthrough@@}
4397 @item @code{lint -fallthrough[ \t]*}
4399 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4401 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4403 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4407 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4408 following regular expressions:
4412 @item @code{-fallthrough}
4414 @item @code{@@fallthrough@@}
4416 @item @code{lint -fallthrough[ \t]*}
4418 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4422 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4423 fallthrough comments, only attributes disable the warning.
4427 The comment needs to be followed after optional whitespace and other comments
4428 by @code{case} or @code{default} keywords or by a user label that precedes some
4429 @code{case} or @code{default} label.
4444 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4446 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4447 @opindex Wif-not-aligned
4448 @opindex Wno-if-not-aligned
4449 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4450 should be issued. This is is enabled by default.
4451 Use @option{-Wno-if-not-aligned} to disable it.
4453 @item -Wignored-qualifiers @r{(C and C++ only)}
4454 @opindex Wignored-qualifiers
4455 @opindex Wno-ignored-qualifiers
4456 Warn if the return type of a function has a type qualifier
4457 such as @code{const}. For ISO C such a type qualifier has no effect,
4458 since the value returned by a function is not an lvalue.
4459 For C++, the warning is only emitted for scalar types or @code{void}.
4460 ISO C prohibits qualified @code{void} return types on function
4461 definitions, so such return types always receive a warning
4462 even without this option.
4464 This warning is also enabled by @option{-Wextra}.
4466 @item -Wignored-attributes @r{(C and C++ only)}
4467 @opindex Wignored-attributes
4468 @opindex Wno-ignored-attributes
4469 Warn when an attribute is ignored. This is different from the
4470 @option{-Wattributes} option in that it warns whenever the compiler decides
4471 to drop an attribute, not that the attribute is either unknown, used in a
4472 wrong place, etc. This warning is enabled by default.
4477 Warn if the type of @code{main} is suspicious. @code{main} should be
4478 a function with external linkage, returning int, taking either zero
4479 arguments, two, or three arguments of appropriate types. This warning
4480 is enabled by default in C++ and is enabled by either @option{-Wall}
4481 or @option{-Wpedantic}.
4483 @item -Wmisleading-indentation @r{(C and C++ only)}
4484 @opindex Wmisleading-indentation
4485 @opindex Wno-misleading-indentation
4486 Warn when the indentation of the code does not reflect the block structure.
4487 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4488 @code{for} clauses with a guarded statement that does not use braces,
4489 followed by an unguarded statement with the same indentation.
4491 In the following example, the call to ``bar'' is misleadingly indented as
4492 if it were guarded by the ``if'' conditional.
4495 if (some_condition ())
4497 bar (); /* Gotcha: this is not guarded by the "if". */
4500 In the case of mixed tabs and spaces, the warning uses the
4501 @option{-ftabstop=} option to determine if the statements line up
4504 The warning is not issued for code involving multiline preprocessor logic
4505 such as the following example.
4510 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4516 The warning is not issued after a @code{#line} directive, since this
4517 typically indicates autogenerated code, and no assumptions can be made
4518 about the layout of the file that the directive references.
4520 This warning is enabled by @option{-Wall} in C and C++.
4522 @item -Wmissing-braces
4523 @opindex Wmissing-braces
4524 @opindex Wno-missing-braces
4525 Warn if an aggregate or union initializer is not fully bracketed. In
4526 the following example, the initializer for @code{a} is not fully
4527 bracketed, but that for @code{b} is fully bracketed. This warning is
4528 enabled by @option{-Wall} in C.
4531 int a[2][2] = @{ 0, 1, 2, 3 @};
4532 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4535 This warning is enabled by @option{-Wall}.
4537 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4538 @opindex Wmissing-include-dirs
4539 @opindex Wno-missing-include-dirs
4540 Warn if a user-supplied include directory does not exist.
4542 @item -Wmultistatement-macros
4543 @opindex Wmultistatement-macros
4544 @opindex Wno-multistatement-macros
4545 Warn about unsafe multiple statement macros that appear to be guarded
4546 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
4547 @code{while}, in which only the first statement is actually guarded after
4548 the macro is expanded.
4553 #define DOIT x++; y++
4558 will increment @code{y} unconditionally, not just when @code{c} holds.
4559 The can usually be fixed by wrapping the macro in a do-while loop:
4561 #define DOIT do @{ x++; y++; @} while (0)
4566 This warning is enabled by @option{-Wall} in C and C++.
4569 @opindex Wparentheses
4570 @opindex Wno-parentheses
4571 Warn if parentheses are omitted in certain contexts, such
4572 as when there is an assignment in a context where a truth value
4573 is expected, or when operators are nested whose precedence people
4574 often get confused about.
4576 Also warn if a comparison like @code{x<=y<=z} appears; this is
4577 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4578 interpretation from that of ordinary mathematical notation.
4580 Also warn for dangerous uses of the GNU extension to
4581 @code{?:} with omitted middle operand. When the condition
4582 in the @code{?}: operator is a boolean expression, the omitted value is
4583 always 1. Often programmers expect it to be a value computed
4584 inside the conditional expression instead.
4586 For C++ this also warns for some cases of unnecessary parentheses in
4587 declarations, which can indicate an attempt at a function call instead
4591 // Declares a local variable called mymutex.
4592 std::unique_lock<std::mutex> (mymutex);
4593 // User meant std::unique_lock<std::mutex> lock (mymutex);
4597 This warning is enabled by @option{-Wall}.
4599 @item -Wsequence-point
4600 @opindex Wsequence-point
4601 @opindex Wno-sequence-point
4602 Warn about code that may have undefined semantics because of violations
4603 of sequence point rules in the C and C++ standards.
4605 The C and C++ standards define the order in which expressions in a C/C++
4606 program are evaluated in terms of @dfn{sequence points}, which represent
4607 a partial ordering between the execution of parts of the program: those
4608 executed before the sequence point, and those executed after it. These
4609 occur after the evaluation of a full expression (one which is not part
4610 of a larger expression), after the evaluation of the first operand of a
4611 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4612 function is called (but after the evaluation of its arguments and the
4613 expression denoting the called function), and in certain other places.
4614 Other than as expressed by the sequence point rules, the order of
4615 evaluation of subexpressions of an expression is not specified. All
4616 these rules describe only a partial order rather than a total order,
4617 since, for example, if two functions are called within one expression
4618 with no sequence point between them, the order in which the functions
4619 are called is not specified. However, the standards committee have
4620 ruled that function calls do not overlap.
4622 It is not specified when between sequence points modifications to the
4623 values of objects take effect. Programs whose behavior depends on this
4624 have undefined behavior; the C and C++ standards specify that ``Between
4625 the previous and next sequence point an object shall have its stored
4626 value modified at most once by the evaluation of an expression.
4627 Furthermore, the prior value shall be read only to determine the value
4628 to be stored.''. If a program breaks these rules, the results on any
4629 particular implementation are entirely unpredictable.
4631 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4632 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4633 diagnosed by this option, and it may give an occasional false positive
4634 result, but in general it has been found fairly effective at detecting
4635 this sort of problem in programs.
4637 The C++17 standard will define the order of evaluation of operands in
4638 more cases: in particular it requires that the right-hand side of an
4639 assignment be evaluated before the left-hand side, so the above
4640 examples are no longer undefined. But this warning will still warn
4641 about them, to help people avoid writing code that is undefined in C
4642 and earlier revisions of C++.
4644 The standard is worded confusingly, therefore there is some debate
4645 over the precise meaning of the sequence point rules in subtle cases.
4646 Links to discussions of the problem, including proposed formal
4647 definitions, may be found on the GCC readings page, at
4648 @uref{http://gcc.gnu.org/@/readings.html}.
4650 This warning is enabled by @option{-Wall} for C and C++.
4652 @item -Wno-return-local-addr
4653 @opindex Wno-return-local-addr
4654 @opindex Wreturn-local-addr
4655 Do not warn about returning a pointer (or in C++, a reference) to a
4656 variable that goes out of scope after the function returns.
4659 @opindex Wreturn-type
4660 @opindex Wno-return-type
4661 Warn whenever a function is defined with a return type that defaults
4662 to @code{int}. Also warn about any @code{return} statement with no
4663 return value in a function whose return type is not @code{void}
4664 (falling off the end of the function body is considered returning
4667 For C only, warn about a @code{return} statement with an expression in a
4668 function whose return type is @code{void}, unless the expression type is
4669 also @code{void}. As a GNU extension, the latter case is accepted
4670 without a warning unless @option{-Wpedantic} is used.
4672 For C++, a function without return type always produces a diagnostic
4673 message, even when @option{-Wno-return-type} is specified. The only
4674 exceptions are @code{main} and functions defined in system headers.
4676 This warning is enabled by @option{-Wall}.
4678 @item -Wshift-count-negative
4679 @opindex Wshift-count-negative
4680 @opindex Wno-shift-count-negative
4681 Warn if shift count is negative. This warning is enabled by default.
4683 @item -Wshift-count-overflow
4684 @opindex Wshift-count-overflow
4685 @opindex Wno-shift-count-overflow
4686 Warn if shift count >= width of type. This warning is enabled by default.
4688 @item -Wshift-negative-value
4689 @opindex Wshift-negative-value
4690 @opindex Wno-shift-negative-value
4691 Warn if left shifting a negative value. This warning is enabled by
4692 @option{-Wextra} in C99 and C++11 modes (and newer).
4694 @item -Wshift-overflow
4695 @itemx -Wshift-overflow=@var{n}
4696 @opindex Wshift-overflow
4697 @opindex Wno-shift-overflow
4698 Warn about left shift overflows. This warning is enabled by
4699 default in C99 and C++11 modes (and newer).
4702 @item -Wshift-overflow=1
4703 This is the warning level of @option{-Wshift-overflow} and is enabled
4704 by default in C99 and C++11 modes (and newer). This warning level does
4705 not warn about left-shifting 1 into the sign bit. (However, in C, such
4706 an overflow is still rejected in contexts where an integer constant expression
4709 @item -Wshift-overflow=2
4710 This warning level also warns about left-shifting 1 into the sign bit,
4711 unless C++14 mode is active.
4717 Warn whenever a @code{switch} statement has an index of enumerated type
4718 and lacks a @code{case} for one or more of the named codes of that
4719 enumeration. (The presence of a @code{default} label prevents this
4720 warning.) @code{case} labels outside the enumeration range also
4721 provoke warnings when this option is used (even if there is a
4722 @code{default} label).
4723 This warning is enabled by @option{-Wall}.
4725 @item -Wswitch-default
4726 @opindex Wswitch-default
4727 @opindex Wno-switch-default
4728 Warn whenever a @code{switch} statement does not have a @code{default}
4732 @opindex Wswitch-enum
4733 @opindex Wno-switch-enum
4734 Warn whenever a @code{switch} statement has an index of enumerated type
4735 and lacks a @code{case} for one or more of the named codes of that
4736 enumeration. @code{case} labels outside the enumeration range also
4737 provoke warnings when this option is used. The only difference
4738 between @option{-Wswitch} and this option is that this option gives a
4739 warning about an omitted enumeration code even if there is a
4740 @code{default} label.
4743 @opindex Wswitch-bool
4744 @opindex Wno-switch-bool
4745 Warn whenever a @code{switch} statement has an index of boolean type
4746 and the case values are outside the range of a boolean type.
4747 It is possible to suppress this warning by casting the controlling
4748 expression to a type other than @code{bool}. For example:
4751 switch ((int) (a == 4))
4757 This warning is enabled by default for C and C++ programs.
4759 @item -Wswitch-unreachable
4760 @opindex Wswitch-unreachable
4761 @opindex Wno-switch-unreachable
4762 Warn whenever a @code{switch} statement contains statements between the
4763 controlling expression and the first case label, which will never be
4764 executed. For example:
4776 @option{-Wswitch-unreachable} does not warn if the statement between the
4777 controlling expression and the first case label is just a declaration:
4790 This warning is enabled by default for C and C++ programs.
4792 @item -Wsync-nand @r{(C and C++ only)}
4794 @opindex Wno-sync-nand
4795 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4796 built-in functions are used. These functions changed semantics in GCC 4.4.
4798 @item -Wunused-but-set-parameter
4799 @opindex Wunused-but-set-parameter
4800 @opindex Wno-unused-but-set-parameter
4801 Warn whenever a function parameter is assigned to, but otherwise unused
4802 (aside from its declaration).
4804 To suppress this warning use the @code{unused} attribute
4805 (@pxref{Variable Attributes}).
4807 This warning is also enabled by @option{-Wunused} together with
4810 @item -Wunused-but-set-variable
4811 @opindex Wunused-but-set-variable
4812 @opindex Wno-unused-but-set-variable
4813 Warn whenever a local variable is assigned to, but otherwise unused
4814 (aside from its declaration).
4815 This warning is enabled by @option{-Wall}.
4817 To suppress this warning use the @code{unused} attribute
4818 (@pxref{Variable Attributes}).
4820 This warning is also enabled by @option{-Wunused}, which is enabled
4823 @item -Wunused-function
4824 @opindex Wunused-function
4825 @opindex Wno-unused-function
4826 Warn whenever a static function is declared but not defined or a
4827 non-inline static function is unused.
4828 This warning is enabled by @option{-Wall}.
4830 @item -Wunused-label
4831 @opindex Wunused-label
4832 @opindex Wno-unused-label
4833 Warn whenever a label is declared but not used.
4834 This warning is enabled by @option{-Wall}.
4836 To suppress this warning use the @code{unused} attribute
4837 (@pxref{Variable Attributes}).
4839 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4840 @opindex Wunused-local-typedefs
4841 Warn when a typedef locally defined in a function is not used.
4842 This warning is enabled by @option{-Wall}.
4844 @item -Wunused-parameter
4845 @opindex Wunused-parameter
4846 @opindex Wno-unused-parameter
4847 Warn whenever a function parameter is unused aside from its declaration.
4849 To suppress this warning use the @code{unused} attribute
4850 (@pxref{Variable Attributes}).
4852 @item -Wno-unused-result
4853 @opindex Wunused-result
4854 @opindex Wno-unused-result
4855 Do not warn if a caller of a function marked with attribute
4856 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4857 its return value. The default is @option{-Wunused-result}.
4859 @item -Wunused-variable
4860 @opindex Wunused-variable
4861 @opindex Wno-unused-variable
4862 Warn whenever a local or static variable is unused aside from its
4863 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4864 but not for C++. This warning is enabled by @option{-Wall}.
4866 To suppress this warning use the @code{unused} attribute
4867 (@pxref{Variable Attributes}).
4869 @item -Wunused-const-variable
4870 @itemx -Wunused-const-variable=@var{n}
4871 @opindex Wunused-const-variable
4872 @opindex Wno-unused-const-variable
4873 Warn whenever a constant static variable is unused aside from its declaration.
4874 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4875 for C, but not for C++. In C this declares variable storage, but in C++ this
4876 is not an error since const variables take the place of @code{#define}s.
4878 To suppress this warning use the @code{unused} attribute
4879 (@pxref{Variable Attributes}).
4882 @item -Wunused-const-variable=1
4883 This is the warning level that is enabled by @option{-Wunused-variable} for
4884 C. It warns only about unused static const variables defined in the main
4885 compilation unit, but not about static const variables declared in any
4888 @item -Wunused-const-variable=2
4889 This warning level also warns for unused constant static variables in
4890 headers (excluding system headers). This is the warning level of
4891 @option{-Wunused-const-variable} and must be explicitly requested since
4892 in C++ this isn't an error and in C it might be harder to clean up all
4896 @item -Wunused-value
4897 @opindex Wunused-value
4898 @opindex Wno-unused-value
4899 Warn whenever a statement computes a result that is explicitly not
4900 used. To suppress this warning cast the unused expression to
4901 @code{void}. This includes an expression-statement or the left-hand
4902 side of a comma expression that contains no side effects. For example,
4903 an expression such as @code{x[i,j]} causes a warning, while
4904 @code{x[(void)i,j]} does not.
4906 This warning is enabled by @option{-Wall}.
4911 All the above @option{-Wunused} options combined.
4913 In order to get a warning about an unused function parameter, you must
4914 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4915 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4917 @item -Wuninitialized
4918 @opindex Wuninitialized
4919 @opindex Wno-uninitialized
4920 Warn if an automatic variable is used without first being initialized
4921 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4922 warn if a non-static reference or non-static @code{const} member
4923 appears in a class without constructors.
4925 If you want to warn about code that uses the uninitialized value of the
4926 variable in its own initializer, use the @option{-Winit-self} option.
4928 These warnings occur for individual uninitialized or clobbered
4929 elements of structure, union or array variables as well as for
4930 variables that are uninitialized or clobbered as a whole. They do
4931 not occur for variables or elements declared @code{volatile}. Because
4932 these warnings depend on optimization, the exact variables or elements
4933 for which there are warnings depends on the precise optimization
4934 options and version of GCC used.
4936 Note that there may be no warning about a variable that is used only
4937 to compute a value that itself is never used, because such
4938 computations may be deleted by data flow analysis before the warnings
4941 @item -Winvalid-memory-model
4942 @opindex Winvalid-memory-model
4943 @opindex Wno-invalid-memory-model
4944 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4945 and the C11 atomic generic functions with a memory consistency argument
4946 that is either invalid for the operation or outside the range of values
4947 of the @code{memory_order} enumeration. For example, since the
4948 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4949 defined for the relaxed, release, and sequentially consistent memory
4950 orders the following code is diagnosed:
4955 __atomic_store_n (i, 0, memory_order_consume);
4959 @option{-Winvalid-memory-model} is enabled by default.
4961 @item -Wmaybe-uninitialized
4962 @opindex Wmaybe-uninitialized
4963 @opindex Wno-maybe-uninitialized
4964 For an automatic (i.e.@ local) variable, if there exists a path from the
4965 function entry to a use of the variable that is initialized, but there exist
4966 some other paths for which the variable is not initialized, the compiler
4967 emits a warning if it cannot prove the uninitialized paths are not
4968 executed at run time.
4970 These warnings are only possible in optimizing compilation, because otherwise
4971 GCC does not keep track of the state of variables.
4973 These warnings are made optional because GCC may not be able to determine when
4974 the code is correct in spite of appearing to have an error. Here is one
4975 example of how this can happen:
4995 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4996 always initialized, but GCC doesn't know this. To suppress the
4997 warning, you need to provide a default case with assert(0) or
5000 @cindex @code{longjmp} warnings
5001 This option also warns when a non-volatile automatic variable might be
5002 changed by a call to @code{longjmp}.
5003 The compiler sees only the calls to @code{setjmp}. It cannot know
5004 where @code{longjmp} will be called; in fact, a signal handler could
5005 call it at any point in the code. As a result, you may get a warning
5006 even when there is in fact no problem because @code{longjmp} cannot
5007 in fact be called at the place that would cause a problem.
5009 Some spurious warnings can be avoided if you declare all the functions
5010 you use that never return as @code{noreturn}. @xref{Function
5013 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5015 @item -Wunknown-pragmas
5016 @opindex Wunknown-pragmas
5017 @opindex Wno-unknown-pragmas
5018 @cindex warning for unknown pragmas
5019 @cindex unknown pragmas, warning
5020 @cindex pragmas, warning of unknown
5021 Warn when a @code{#pragma} directive is encountered that is not understood by
5022 GCC@. If this command-line option is used, warnings are even issued
5023 for unknown pragmas in system header files. This is not the case if
5024 the warnings are only enabled by the @option{-Wall} command-line option.
5027 @opindex Wno-pragmas
5029 Do not warn about misuses of pragmas, such as incorrect parameters,
5030 invalid syntax, or conflicts between pragmas. See also
5031 @option{-Wunknown-pragmas}.
5033 @item -Wstrict-aliasing
5034 @opindex Wstrict-aliasing
5035 @opindex Wno-strict-aliasing
5036 This option is only active when @option{-fstrict-aliasing} is active.
5037 It warns about code that might break the strict aliasing rules that the
5038 compiler is using for optimization. The warning does not catch all
5039 cases, but does attempt to catch the more common pitfalls. It is
5040 included in @option{-Wall}.
5041 It is equivalent to @option{-Wstrict-aliasing=3}
5043 @item -Wstrict-aliasing=n
5044 @opindex Wstrict-aliasing=n
5045 This option is only active when @option{-fstrict-aliasing} is active.
5046 It warns about code that might break the strict aliasing rules that the
5047 compiler is using for optimization.
5048 Higher levels correspond to higher accuracy (fewer false positives).
5049 Higher levels also correspond to more effort, similar to the way @option{-O}
5051 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5053 Level 1: Most aggressive, quick, least accurate.
5054 Possibly useful when higher levels
5055 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5056 false negatives. However, it has many false positives.
5057 Warns for all pointer conversions between possibly incompatible types,
5058 even if never dereferenced. Runs in the front end only.
5060 Level 2: Aggressive, quick, not too precise.
5061 May still have many false positives (not as many as level 1 though),
5062 and few false negatives (but possibly more than level 1).
5063 Unlike level 1, it only warns when an address is taken. Warns about
5064 incomplete types. Runs in the front end only.
5066 Level 3 (default for @option{-Wstrict-aliasing}):
5067 Should have very few false positives and few false
5068 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5069 Takes care of the common pun+dereference pattern in the front end:
5070 @code{*(int*)&some_float}.
5071 If optimization is enabled, it also runs in the back end, where it deals
5072 with multiple statement cases using flow-sensitive points-to information.
5073 Only warns when the converted pointer is dereferenced.
5074 Does not warn about incomplete types.
5076 @item -Wstrict-overflow
5077 @itemx -Wstrict-overflow=@var{n}
5078 @opindex Wstrict-overflow
5079 @opindex Wno-strict-overflow
5080 This option is only active when signed overflow is undefined.
5081 It warns about cases where the compiler optimizes based on the
5082 assumption that signed overflow does not occur. Note that it does not
5083 warn about all cases where the code might overflow: it only warns
5084 about cases where the compiler implements some optimization. Thus
5085 this warning depends on the optimization level.
5087 An optimization that assumes that signed overflow does not occur is
5088 perfectly safe if the values of the variables involved are such that
5089 overflow never does, in fact, occur. Therefore this warning can
5090 easily give a false positive: a warning about code that is not
5091 actually a problem. To help focus on important issues, several
5092 warning levels are defined. No warnings are issued for the use of
5093 undefined signed overflow when estimating how many iterations a loop
5094 requires, in particular when determining whether a loop will be
5098 @item -Wstrict-overflow=1
5099 Warn about cases that are both questionable and easy to avoid. For
5100 example the compiler simplifies
5101 @code{x + 1 > x} to @code{1}. This level of
5102 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5103 are not, and must be explicitly requested.
5105 @item -Wstrict-overflow=2
5106 Also warn about other cases where a comparison is simplified to a
5107 constant. For example: @code{abs (x) >= 0}. This can only be
5108 simplified when signed integer overflow is undefined, because
5109 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5110 zero. @option{-Wstrict-overflow} (with no level) is the same as
5111 @option{-Wstrict-overflow=2}.
5113 @item -Wstrict-overflow=3
5114 Also warn about other cases where a comparison is simplified. For
5115 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5117 @item -Wstrict-overflow=4
5118 Also warn about other simplifications not covered by the above cases.
5119 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5121 @item -Wstrict-overflow=5
5122 Also warn about cases where the compiler reduces the magnitude of a
5123 constant involved in a comparison. For example: @code{x + 2 > y} is
5124 simplified to @code{x + 1 >= y}. This is reported only at the
5125 highest warning level because this simplification applies to many
5126 comparisons, so this warning level gives a very large number of
5130 @item -Wstringop-overflow
5131 @itemx -Wstringop-overflow=@var{type}
5132 @opindex Wstringop-overflow
5133 @opindex Wno-stringop-overflow
5134 Warn for calls to string manipulation functions such as @code{memcpy} and
5135 @code{strcpy} that are determined to overflow the destination buffer. The
5136 optional argument is one greater than the type of Object Size Checking to
5137 perform to determine the size of the destination. @xref{Object Size Checking}.
5138 The argument is meaningful only for functions that operate on character arrays
5139 but not for raw memory functions like @code{memcpy} which always make use
5140 of Object Size type-0. The option also warns for calls that specify a size
5141 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5142 The option produces the best results with optimization enabled but can detect
5143 a small subset of simple buffer overflows even without optimization in
5144 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5145 correspond to the standard functions. In any case, the option warns about
5146 just a subset of buffer overflows detected by the corresponding overflow
5147 checking built-ins. For example, the option will issue a warning for
5148 the @code{strcpy} call below because it copies at least 5 characters
5149 (the string @code{"blue"} including the terminating NUL) into the buffer
5153 enum Color @{ blue, purple, yellow @};
5154 const char* f (enum Color clr)
5156 static char buf [4];
5160 case blue: str = "blue"; break;
5161 case purple: str = "purple"; break;
5162 case yellow: str = "yellow"; break;
5165 return strcpy (buf, str); // warning here
5169 Option @option{-Wstringop-overflow=2} is enabled by default.
5172 @item -Wstringop-overflow
5173 @item -Wstringop-overflow=1
5174 @opindex Wstringop-overflow
5175 @opindex Wno-stringop-overflow
5176 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5177 to determine the sizes of destination objects. This is the default setting
5178 of the option. At this setting the option will not warn for writes past
5179 the end of subobjects of larger objects accessed by pointers unless the
5180 size of the largest surrounding object is known. When the destination may
5181 be one of several objects it is assumed to be the largest one of them. On
5182 Linux systems, when optimization is enabled at this setting the option warns
5183 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5186 @item -Wstringop-overflow=2
5187 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5188 to determine the sizes of destination objects. At this setting the option
5189 will warn about overflows when writing to members of the largest complete
5190 objects whose exact size is known. It will, however, not warn for excessive
5191 writes to the same members of unknown objects referenced by pointers since
5192 they may point to arrays containing unknown numbers of elements.
5194 @item -Wstringop-overflow=3
5195 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5196 to determine the sizes of destination objects. At this setting the option
5197 warns about overflowing the smallest object or data member. This is the
5198 most restrictive setting of the option that may result in warnings for safe
5201 @item -Wstringop-overflow=4
5202 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5203 to determine the sizes of destination objects. At this setting the option
5204 will warn about overflowing any data members, and when the destination is
5205 one of several objects it uses the size of the largest of them to decide
5206 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5207 setting of the option may result in warnings for benign code.
5210 @item -Wstringop-truncation
5211 @opindex Wstringop-truncation
5212 @opindex Wno-stringop-truncation
5213 Warn for calls to bounded string manipulation functions such as @code{strncat},
5214 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5215 or leave the destination unchanged.
5217 In the following example, the call to @code{strncat} specifies a bound that
5218 is less than the length of the source string. As a result, the copy of
5219 the source will be truncated and so the call is diagnosed. To avoid the
5220 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5223 void append (char *buf, size_t bufsize)
5225 strncat (buf, ".txt", 3);
5229 As another example, the following call to @code{strncpy} results in copying
5230 to @code{d} just the characters preceding the terminating NUL, without
5231 appending the NUL to the end. Assuming the result of @code{strncpy} is
5232 necessarily a NUL-terminated string is a common mistake, and so the call
5233 is diagnosed. To avoid the warning when the result is not expected to be
5234 NUL-terminated, call @code{memcpy} instead.
5237 void copy (char *d, const char *s)
5239 strncpy (d, s, strlen (s));
5243 In the following example, the call to @code{strncpy} specifies the size
5244 of the destination buffer as the bound. If the length of the source
5245 string is equal to or greater than this size the result of the copy will
5246 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5247 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5248 element of the buffer to @code{NUL}.
5251 void copy (const char *s)
5254 strncpy (buf, s, sizeof buf);
5259 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5260 @opindex Wsuggest-attribute=
5261 @opindex Wno-suggest-attribute=
5262 Warn for cases where adding an attribute may be beneficial. The
5263 attributes currently supported are listed below.
5266 @item -Wsuggest-attribute=pure
5267 @itemx -Wsuggest-attribute=const
5268 @itemx -Wsuggest-attribute=noreturn
5269 @itemx -Wsuggest-attribute=malloc
5270 @opindex Wsuggest-attribute=pure
5271 @opindex Wno-suggest-attribute=pure
5272 @opindex Wsuggest-attribute=const
5273 @opindex Wno-suggest-attribute=const
5274 @opindex Wsuggest-attribute=noreturn
5275 @opindex Wno-suggest-attribute=noreturn
5276 @opindex Wsuggest-attribute=malloc
5277 @opindex Wno-suggest-attribute=malloc
5279 Warn about functions that might be candidates for attributes
5280 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5281 only warns for functions visible in other compilation units or (in the case of
5282 @code{pure} and @code{const}) if it cannot prove that the function returns
5283 normally. A function returns normally if it doesn't contain an infinite loop or
5284 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5285 requires option @option{-fipa-pure-const}, which is enabled by default at
5286 @option{-O} and higher. Higher optimization levels improve the accuracy
5289 @item -Wsuggest-attribute=format
5290 @itemx -Wmissing-format-attribute
5291 @opindex Wsuggest-attribute=format
5292 @opindex Wmissing-format-attribute
5293 @opindex Wno-suggest-attribute=format
5294 @opindex Wno-missing-format-attribute
5298 Warn about function pointers that might be candidates for @code{format}
5299 attributes. Note these are only possible candidates, not absolute ones.
5300 GCC guesses that function pointers with @code{format} attributes that
5301 are used in assignment, initialization, parameter passing or return
5302 statements should have a corresponding @code{format} attribute in the
5303 resulting type. I.e.@: the left-hand side of the assignment or
5304 initialization, the type of the parameter variable, or the return type
5305 of the containing function respectively should also have a @code{format}
5306 attribute to avoid the warning.
5308 GCC also warns about function definitions that might be
5309 candidates for @code{format} attributes. Again, these are only
5310 possible candidates. GCC guesses that @code{format} attributes
5311 might be appropriate for any function that calls a function like
5312 @code{vprintf} or @code{vscanf}, but this might not always be the
5313 case, and some functions for which @code{format} attributes are
5314 appropriate may not be detected.
5316 @item -Wsuggest-attribute=cold
5317 @opindex Wsuggest-attribute=cold
5318 @opindex Wno-suggest-attribute=cold
5320 Warn about functions that might be candidates for @code{cold} attribute. This
5321 is based on static detection and generally will only warn about functions which
5322 always leads to a call to another @code{cold} function such as wrappers of
5323 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5326 @item -Wsuggest-final-types
5327 @opindex Wno-suggest-final-types
5328 @opindex Wsuggest-final-types
5329 Warn about types with virtual methods where code quality would be improved
5330 if the type were declared with the C++11 @code{final} specifier,
5332 declared in an anonymous namespace. This allows GCC to more aggressively
5333 devirtualize the polymorphic calls. This warning is more effective with link
5334 time optimization, where the information about the class hierarchy graph is
5337 @item -Wsuggest-final-methods
5338 @opindex Wno-suggest-final-methods
5339 @opindex Wsuggest-final-methods
5340 Warn about virtual methods where code quality would be improved if the method
5341 were declared with the C++11 @code{final} specifier,
5342 or, if possible, its type were
5343 declared in an anonymous namespace or with the @code{final} specifier.
5345 more effective with link-time optimization, where the information about the
5346 class hierarchy graph is more complete. It is recommended to first consider
5347 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5350 @item -Wsuggest-override
5351 Warn about overriding virtual functions that are not marked with the override
5355 @opindex Wno-alloc-zero
5356 @opindex Walloc-zero
5357 Warn about calls to allocation functions decorated with attribute
5358 @code{alloc_size} that specify zero bytes, including those to the built-in
5359 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5360 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5361 when called with a zero size differs among implementations (and in the case
5362 of @code{realloc} has been deprecated) relying on it may result in subtle
5363 portability bugs and should be avoided.
5365 @item -Walloc-size-larger-than=@var{n}
5366 Warn about calls to functions decorated with attribute @code{alloc_size}
5367 that attempt to allocate objects larger than the specified number of bytes,
5368 or where the result of the size computation in an integer type with infinite
5369 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5370 may end in one of the standard suffixes designating a multiple of bytes
5371 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5372 @code{MB} and @code{MiB} for megabyte and mebibyte, and so on.
5373 @xref{Function Attributes}.
5378 This option warns on all uses of @code{alloca} in the source.
5380 @item -Walloca-larger-than=@var{n}
5381 This option warns on calls to @code{alloca} that are not bounded by a
5382 controlling predicate limiting its argument of integer type to at most
5383 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5384 Arguments of non-integer types are considered unbounded even if they
5385 appear to be constrained to the expected range.
5387 For example, a bounded case of @code{alloca} could be:
5390 void func (size_t n)
5401 In the above example, passing @code{-Walloca-larger-than=1000} would not
5402 issue a warning because the call to @code{alloca} is known to be at most
5403 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5404 the compiler would emit a warning.
5406 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5407 controlling predicate constraining its integer argument. For example:
5412 void *p = alloca (n);
5417 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5418 a warning, but this time because of the lack of bounds checking.
5420 Note, that even seemingly correct code involving signed integers could
5424 void func (signed int n)
5434 In the above example, @var{n} could be negative, causing a larger than
5435 expected argument to be implicitly cast into the @code{alloca} call.
5437 This option also warns when @code{alloca} is used in a loop.
5439 This warning is not enabled by @option{-Wall}, and is only active when
5440 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5442 See also @option{-Wvla-larger-than=@var{n}}.
5444 @item -Warray-bounds
5445 @itemx -Warray-bounds=@var{n}
5446 @opindex Wno-array-bounds
5447 @opindex Warray-bounds
5448 This option is only active when @option{-ftree-vrp} is active
5449 (default for @option{-O2} and above). It warns about subscripts to arrays
5450 that are always out of bounds. This warning is enabled by @option{-Wall}.
5453 @item -Warray-bounds=1
5454 This is the warning level of @option{-Warray-bounds} and is enabled
5455 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5457 @item -Warray-bounds=2
5458 This warning level also warns about out of bounds access for
5459 arrays at the end of a struct and for arrays accessed through
5460 pointers. This warning level may give a larger number of
5461 false positives and is deactivated by default.
5464 @item -Wattribute-alias
5465 Warn about declarations using the @code{alias} and similar attributes whose
5466 target is incompatible with the type of the alias. @xref{Function Attributes,
5467 ,Declaring Attributes of Functions}.
5469 @item -Wbool-compare
5470 @opindex Wno-bool-compare
5471 @opindex Wbool-compare
5472 Warn about boolean expression compared with an integer value different from
5473 @code{true}/@code{false}. For instance, the following comparison is
5478 if ((n > 1) == 2) @{ @dots{} @}
5480 This warning is enabled by @option{-Wall}.
5482 @item -Wbool-operation
5483 @opindex Wno-bool-operation
5484 @opindex Wbool-operation
5485 Warn about suspicious operations on expressions of a boolean type. For
5486 instance, bitwise negation of a boolean is very likely a bug in the program.
5487 For C, this warning also warns about incrementing or decrementing a boolean,
5488 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5489 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
5491 This warning is enabled by @option{-Wall}.
5493 @item -Wduplicated-branches
5494 @opindex Wno-duplicated-branches
5495 @opindex Wduplicated-branches
5496 Warn when an if-else has identical branches. This warning detects cases like
5503 It doesn't warn when both branches contain just a null statement. This warning
5504 also warn for conditional operators:
5506 int i = x ? *p : *p;
5509 @item -Wduplicated-cond
5510 @opindex Wno-duplicated-cond
5511 @opindex Wduplicated-cond
5512 Warn about duplicated conditions in an if-else-if chain. For instance,
5513 warn for the following code:
5515 if (p->q != NULL) @{ @dots{} @}
5516 else if (p->q != NULL) @{ @dots{} @}
5519 @item -Wframe-address
5520 @opindex Wno-frame-address
5521 @opindex Wframe-address
5522 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5523 is called with an argument greater than 0. Such calls may return indeterminate
5524 values or crash the program. The warning is included in @option{-Wall}.
5526 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5527 @opindex Wno-discarded-qualifiers
5528 @opindex Wdiscarded-qualifiers
5529 Do not warn if type qualifiers on pointers are being discarded.
5530 Typically, the compiler warns if a @code{const char *} variable is
5531 passed to a function that takes a @code{char *} parameter. This option
5532 can be used to suppress such a warning.
5534 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5535 @opindex Wno-discarded-array-qualifiers
5536 @opindex Wdiscarded-array-qualifiers
5537 Do not warn if type qualifiers on arrays which are pointer targets
5538 are being discarded. Typically, the compiler warns if a
5539 @code{const int (*)[]} variable is passed to a function that
5540 takes a @code{int (*)[]} parameter. This option can be used to
5541 suppress such a warning.
5543 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5544 @opindex Wno-incompatible-pointer-types
5545 @opindex Wincompatible-pointer-types
5546 Do not warn when there is a conversion between pointers that have incompatible
5547 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5548 which warns for pointer argument passing or assignment with different
5551 @item -Wno-int-conversion @r{(C and Objective-C only)}
5552 @opindex Wno-int-conversion
5553 @opindex Wint-conversion
5554 Do not warn about incompatible integer to pointer and pointer to integer
5555 conversions. This warning is about implicit conversions; for explicit
5556 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5557 @option{-Wno-pointer-to-int-cast} may be used.
5559 @item -Wno-div-by-zero
5560 @opindex Wno-div-by-zero
5561 @opindex Wdiv-by-zero
5562 Do not warn about compile-time integer division by zero. Floating-point
5563 division by zero is not warned about, as it can be a legitimate way of
5564 obtaining infinities and NaNs.
5566 @item -Wsystem-headers
5567 @opindex Wsystem-headers
5568 @opindex Wno-system-headers
5569 @cindex warnings from system headers
5570 @cindex system headers, warnings from
5571 Print warning messages for constructs found in system header files.
5572 Warnings from system headers are normally suppressed, on the assumption
5573 that they usually do not indicate real problems and would only make the
5574 compiler output harder to read. Using this command-line option tells
5575 GCC to emit warnings from system headers as if they occurred in user
5576 code. However, note that using @option{-Wall} in conjunction with this
5577 option does @emph{not} warn about unknown pragmas in system
5578 headers---for that, @option{-Wunknown-pragmas} must also be used.
5580 @item -Wtautological-compare
5581 @opindex Wtautological-compare
5582 @opindex Wno-tautological-compare
5583 Warn if a self-comparison always evaluates to true or false. This
5584 warning detects various mistakes such as:
5588 if (i > i) @{ @dots{} @}
5591 This warning also warns about bitwise comparisons that always evaluate
5592 to true or false, for instance:
5594 if ((a & 16) == 10) @{ @dots{} @}
5596 will always be false.
5598 This warning is enabled by @option{-Wall}.
5601 @opindex Wtrampolines
5602 @opindex Wno-trampolines
5603 Warn about trampolines generated for pointers to nested functions.
5604 A trampoline is a small piece of data or code that is created at run
5605 time on the stack when the address of a nested function is taken, and is
5606 used to call the nested function indirectly. For some targets, it is
5607 made up of data only and thus requires no special treatment. But, for
5608 most targets, it is made up of code and thus requires the stack to be
5609 made executable in order for the program to work properly.
5612 @opindex Wfloat-equal
5613 @opindex Wno-float-equal
5614 Warn if floating-point values are used in equality comparisons.
5616 The idea behind this is that sometimes it is convenient (for the
5617 programmer) to consider floating-point values as approximations to
5618 infinitely precise real numbers. If you are doing this, then you need
5619 to compute (by analyzing the code, or in some other way) the maximum or
5620 likely maximum error that the computation introduces, and allow for it
5621 when performing comparisons (and when producing output, but that's a
5622 different problem). In particular, instead of testing for equality, you
5623 should check to see whether the two values have ranges that overlap; and
5624 this is done with the relational operators, so equality comparisons are
5627 @item -Wtraditional @r{(C and Objective-C only)}
5628 @opindex Wtraditional
5629 @opindex Wno-traditional
5630 Warn about certain constructs that behave differently in traditional and
5631 ISO C@. Also warn about ISO C constructs that have no traditional C
5632 equivalent, and/or problematic constructs that should be avoided.
5636 Macro parameters that appear within string literals in the macro body.
5637 In traditional C macro replacement takes place within string literals,
5638 but in ISO C it does not.
5641 In traditional C, some preprocessor directives did not exist.
5642 Traditional preprocessors only considered a line to be a directive
5643 if the @samp{#} appeared in column 1 on the line. Therefore
5644 @option{-Wtraditional} warns about directives that traditional C
5645 understands but ignores because the @samp{#} does not appear as the
5646 first character on the line. It also suggests you hide directives like
5647 @code{#pragma} not understood by traditional C by indenting them. Some
5648 traditional implementations do not recognize @code{#elif}, so this option
5649 suggests avoiding it altogether.
5652 A function-like macro that appears without arguments.
5655 The unary plus operator.
5658 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5659 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5660 constants.) Note, these suffixes appear in macros defined in the system
5661 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5662 Use of these macros in user code might normally lead to spurious
5663 warnings, however GCC's integrated preprocessor has enough context to
5664 avoid warning in these cases.
5667 A function declared external in one block and then used after the end of
5671 A @code{switch} statement has an operand of type @code{long}.
5674 A non-@code{static} function declaration follows a @code{static} one.
5675 This construct is not accepted by some traditional C compilers.
5678 The ISO type of an integer constant has a different width or
5679 signedness from its traditional type. This warning is only issued if
5680 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5681 typically represent bit patterns, are not warned about.
5684 Usage of ISO string concatenation is detected.
5687 Initialization of automatic aggregates.
5690 Identifier conflicts with labels. Traditional C lacks a separate
5691 namespace for labels.
5694 Initialization of unions. If the initializer is zero, the warning is
5695 omitted. This is done under the assumption that the zero initializer in
5696 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5697 initializer warnings and relies on default initialization to zero in the
5701 Conversions by prototypes between fixed/floating-point values and vice
5702 versa. The absence of these prototypes when compiling with traditional
5703 C causes serious problems. This is a subset of the possible
5704 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5707 Use of ISO C style function definitions. This warning intentionally is
5708 @emph{not} issued for prototype declarations or variadic functions
5709 because these ISO C features appear in your code when using
5710 libiberty's traditional C compatibility macros, @code{PARAMS} and
5711 @code{VPARAMS}. This warning is also bypassed for nested functions
5712 because that feature is already a GCC extension and thus not relevant to
5713 traditional C compatibility.
5716 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5717 @opindex Wtraditional-conversion
5718 @opindex Wno-traditional-conversion
5719 Warn if a prototype causes a type conversion that is different from what
5720 would happen to the same argument in the absence of a prototype. This
5721 includes conversions of fixed point to floating and vice versa, and
5722 conversions changing the width or signedness of a fixed-point argument
5723 except when the same as the default promotion.
5725 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5726 @opindex Wdeclaration-after-statement
5727 @opindex Wno-declaration-after-statement
5728 Warn when a declaration is found after a statement in a block. This
5729 construct, known from C++, was introduced with ISO C99 and is by default
5730 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5735 Warn whenever a local variable or type declaration shadows another
5736 variable, parameter, type, class member (in C++), or instance variable
5737 (in Objective-C) or whenever a built-in function is shadowed. Note
5738 that in C++, the compiler warns if a local variable shadows an
5739 explicit typedef, but not if it shadows a struct/class/enum.
5740 Same as @option{-Wshadow=global}.
5742 @item -Wno-shadow-ivar @r{(Objective-C only)}
5743 @opindex Wno-shadow-ivar
5744 @opindex Wshadow-ivar
5745 Do not warn whenever a local variable shadows an instance variable in an
5748 @item -Wshadow=global
5749 @opindex Wshadow=local
5750 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5752 @item -Wshadow=local
5753 @opindex Wshadow=local
5754 Warn when a local variable shadows another local variable or parameter.
5755 This warning is enabled by @option{-Wshadow=global}.
5757 @item -Wshadow=compatible-local
5758 @opindex Wshadow=compatible-local
5759 Warn when a local variable shadows another local variable or parameter
5760 whose type is compatible with that of the shadowing variable. In C++,
5761 type compatibility here means the type of the shadowing variable can be
5762 converted to that of the shadowed variable. The creation of this flag
5763 (in addition to @option{-Wshadow=local}) is based on the idea that when
5764 a local variable shadows another one of incompatible type, it is most
5765 likely intentional, not a bug or typo, as shown in the following example:
5769 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5771 for (int i = 0; i < N; ++i)
5780 Since the two variable @code{i} in the example above have incompatible types,
5781 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5782 Because their types are incompatible, if a programmer accidentally uses one
5783 in place of the other, type checking will catch that and emit an error or
5784 warning. So not warning (about shadowing) in this case will not lead to
5785 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5786 possibly reduce the number of warnings triggered by intentional shadowing.
5788 This warning is enabled by @option{-Wshadow=local}.
5790 @item -Wlarger-than=@var{len}
5791 @opindex Wlarger-than=@var{len}
5792 @opindex Wlarger-than-@var{len}
5793 Warn whenever an object of larger than @var{len} bytes is defined.
5795 @item -Wframe-larger-than=@var{len}
5796 @opindex Wframe-larger-than
5797 Warn if the size of a function frame is larger than @var{len} bytes.
5798 The computation done to determine the stack frame size is approximate
5799 and not conservative.
5800 The actual requirements may be somewhat greater than @var{len}
5801 even if you do not get a warning. In addition, any space allocated
5802 via @code{alloca}, variable-length arrays, or related constructs
5803 is not included by the compiler when determining
5804 whether or not to issue a warning.
5806 @item -Wno-free-nonheap-object
5807 @opindex Wno-free-nonheap-object
5808 @opindex Wfree-nonheap-object
5809 Do not warn when attempting to free an object that was not allocated
5812 @item -Wstack-usage=@var{len}
5813 @opindex Wstack-usage
5814 Warn if the stack usage of a function might be larger than @var{len} bytes.
5815 The computation done to determine the stack usage is conservative.
5816 Any space allocated via @code{alloca}, variable-length arrays, or related
5817 constructs is included by the compiler when determining whether or not to
5820 The message is in keeping with the output of @option{-fstack-usage}.
5824 If the stack usage is fully static but exceeds the specified amount, it's:
5827 warning: stack usage is 1120 bytes
5830 If the stack usage is (partly) dynamic but bounded, it's:
5833 warning: stack usage might be 1648 bytes
5836 If the stack usage is (partly) dynamic and not bounded, it's:
5839 warning: stack usage might be unbounded
5843 @item -Wunsafe-loop-optimizations
5844 @opindex Wunsafe-loop-optimizations
5845 @opindex Wno-unsafe-loop-optimizations
5846 Warn if the loop cannot be optimized because the compiler cannot
5847 assume anything on the bounds of the loop indices. With
5848 @option{-funsafe-loop-optimizations} warn if the compiler makes
5851 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5852 @opindex Wno-pedantic-ms-format
5853 @opindex Wpedantic-ms-format
5854 When used in combination with @option{-Wformat}
5855 and @option{-pedantic} without GNU extensions, this option
5856 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5857 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5858 which depend on the MS runtime.
5861 @opindex Waligned-new
5862 @opindex Wno-aligned-new
5863 Warn about a new-expression of a type that requires greater alignment
5864 than the @code{alignof(std::max_align_t)} but uses an allocation
5865 function without an explicit alignment parameter. This option is
5866 enabled by @option{-Wall}.
5868 Normally this only warns about global allocation functions, but
5869 @option{-Waligned-new=all} also warns about class member allocation
5872 @item -Wplacement-new
5873 @itemx -Wplacement-new=@var{n}
5874 @opindex Wplacement-new
5875 @opindex Wno-placement-new
5876 Warn about placement new expressions with undefined behavior, such as
5877 constructing an object in a buffer that is smaller than the type of
5878 the object. For example, the placement new expression below is diagnosed
5879 because it attempts to construct an array of 64 integers in a buffer only
5885 This warning is enabled by default.
5888 @item -Wplacement-new=1
5889 This is the default warning level of @option{-Wplacement-new}. At this
5890 level the warning is not issued for some strictly undefined constructs that
5891 GCC allows as extensions for compatibility with legacy code. For example,
5892 the following @code{new} expression is not diagnosed at this level even
5893 though it has undefined behavior according to the C++ standard because
5894 it writes past the end of the one-element array.
5896 struct S @{ int n, a[1]; @};
5897 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5898 new (s->a)int [32]();
5901 @item -Wplacement-new=2
5902 At this level, in addition to diagnosing all the same constructs as at level
5903 1, a diagnostic is also issued for placement new expressions that construct
5904 an object in the last member of structure whose type is an array of a single
5905 element and whose size is less than the size of the object being constructed.
5906 While the previous example would be diagnosed, the following construct makes
5907 use of the flexible member array extension to avoid the warning at level 2.
5909 struct S @{ int n, a[]; @};
5910 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5911 new (s->a)int [32]();
5916 @item -Wpointer-arith
5917 @opindex Wpointer-arith
5918 @opindex Wno-pointer-arith
5919 Warn about anything that depends on the ``size of'' a function type or
5920 of @code{void}. GNU C assigns these types a size of 1, for
5921 convenience in calculations with @code{void *} pointers and pointers
5922 to functions. In C++, warn also when an arithmetic operation involves
5923 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5925 @item -Wpointer-compare
5926 @opindex Wpointer-compare
5927 @opindex Wno-pointer-compare
5928 Warn if a pointer is compared with a zero character constant. This usually
5929 means that the pointer was meant to be dereferenced. For example:
5932 const char *p = foo ();
5937 Note that the code above is invalid in C++11.
5939 This warning is enabled by default.
5942 @opindex Wtype-limits
5943 @opindex Wno-type-limits
5944 Warn if a comparison is always true or always false due to the limited
5945 range of the data type, but do not warn for constant expressions. For
5946 example, warn if an unsigned variable is compared against zero with
5947 @code{<} or @code{>=}. This warning is also enabled by
5950 @include cppwarnopts.texi
5952 @item -Wbad-function-cast @r{(C and Objective-C only)}
5953 @opindex Wbad-function-cast
5954 @opindex Wno-bad-function-cast
5955 Warn when a function call is cast to a non-matching type.
5956 For example, warn if a call to a function returning an integer type
5957 is cast to a pointer type.
5959 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5960 @opindex Wc90-c99-compat
5961 @opindex Wno-c90-c99-compat
5962 Warn about features not present in ISO C90, but present in ISO C99.
5963 For instance, warn about use of variable length arrays, @code{long long}
5964 type, @code{bool} type, compound literals, designated initializers, and so
5965 on. This option is independent of the standards mode. Warnings are disabled
5966 in the expression that follows @code{__extension__}.
5968 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5969 @opindex Wc99-c11-compat
5970 @opindex Wno-c99-c11-compat
5971 Warn about features not present in ISO C99, but present in ISO C11.
5972 For instance, warn about use of anonymous structures and unions,
5973 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5974 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5975 and so on. This option is independent of the standards mode. Warnings are
5976 disabled in the expression that follows @code{__extension__}.
5978 @item -Wc++-compat @r{(C and Objective-C only)}
5979 @opindex Wc++-compat
5980 Warn about ISO C constructs that are outside of the common subset of
5981 ISO C and ISO C++, e.g.@: request for implicit conversion from
5982 @code{void *} to a pointer to non-@code{void} type.
5984 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5985 @opindex Wc++11-compat
5986 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5987 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5988 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5989 enabled by @option{-Wall}.
5991 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5992 @opindex Wc++14-compat
5993 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5994 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5996 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
5997 @opindex Wc++17-compat
5998 Warn about C++ constructs whose meaning differs between ISO C++ 2014
5999 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6003 @opindex Wno-cast-qual
6004 Warn whenever a pointer is cast so as to remove a type qualifier from
6005 the target type. For example, warn if a @code{const char *} is cast
6006 to an ordinary @code{char *}.
6008 Also warn when making a cast that introduces a type qualifier in an
6009 unsafe way. For example, casting @code{char **} to @code{const char **}
6010 is unsafe, as in this example:
6013 /* p is char ** value. */
6014 const char **q = (const char **) p;
6015 /* Assignment of readonly string to const char * is OK. */
6017 /* Now char** pointer points to read-only memory. */
6022 @opindex Wcast-align
6023 @opindex Wno-cast-align
6024 Warn whenever a pointer is cast such that the required alignment of the
6025 target is increased. For example, warn if a @code{char *} is cast to
6026 an @code{int *} on machines where integers can only be accessed at
6027 two- or four-byte boundaries.
6029 @item -Wcast-align=strict
6030 @opindex Wcast-align=strict
6031 Warn whenever a pointer is cast such that the required alignment of the
6032 target is increased. For example, warn if a @code{char *} is cast to
6033 an @code{int *} regardless of the target machine.
6035 @item -Wwrite-strings
6036 @opindex Wwrite-strings
6037 @opindex Wno-write-strings
6038 When compiling C, give string constants the type @code{const
6039 char[@var{length}]} so that copying the address of one into a
6040 non-@code{const} @code{char *} pointer produces a warning. These
6041 warnings help you find at compile time code that can try to write
6042 into a string constant, but only if you have been very careful about
6043 using @code{const} in declarations and prototypes. Otherwise, it is
6044 just a nuisance. This is why we did not make @option{-Wall} request
6047 When compiling C++, warn about the deprecated conversion from string
6048 literals to @code{char *}. This warning is enabled by default for C++
6052 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6053 @opindex Wcatch-value
6054 @opindex Wno-catch-value
6055 Warn about catch handlers that do not catch via reference.
6056 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6057 warn about polymorphic class types that are caught by value.
6058 With @option{-Wcatch-value=2} warn about all class types that are caught
6059 by value. With @option{-Wcatch-value=3} warn about all types that are
6060 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6064 @opindex Wno-clobbered
6065 Warn for variables that might be changed by @code{longjmp} or
6066 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6068 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6069 @opindex Wconditionally-supported
6070 @opindex Wno-conditionally-supported
6071 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6074 @opindex Wconversion
6075 @opindex Wno-conversion
6076 Warn for implicit conversions that may alter a value. This includes
6077 conversions between real and integer, like @code{abs (x)} when
6078 @code{x} is @code{double}; conversions between signed and unsigned,
6079 like @code{unsigned ui = -1}; and conversions to smaller types, like
6080 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6081 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6082 changed by the conversion like in @code{abs (2.0)}. Warnings about
6083 conversions between signed and unsigned integers can be disabled by
6084 using @option{-Wno-sign-conversion}.
6086 For C++, also warn for confusing overload resolution for user-defined
6087 conversions; and conversions that never use a type conversion
6088 operator: conversions to @code{void}, the same type, a base class or a
6089 reference to them. Warnings about conversions between signed and
6090 unsigned integers are disabled by default in C++ unless
6091 @option{-Wsign-conversion} is explicitly enabled.
6093 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6094 @opindex Wconversion-null
6095 @opindex Wno-conversion-null
6096 Do not warn for conversions between @code{NULL} and non-pointer
6097 types. @option{-Wconversion-null} is enabled by default.
6099 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6100 @opindex Wzero-as-null-pointer-constant
6101 @opindex Wno-zero-as-null-pointer-constant
6102 Warn when a literal @samp{0} is used as null pointer constant. This can
6103 be useful to facilitate the conversion to @code{nullptr} in C++11.
6105 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6106 @opindex Wsubobject-linkage
6107 @opindex Wno-subobject-linkage
6108 Warn if a class type has a base or a field whose type uses the anonymous
6109 namespace or depends on a type with no linkage. If a type A depends on
6110 a type B with no or internal linkage, defining it in multiple
6111 translation units would be an ODR violation because the meaning of B
6112 is different in each translation unit. If A only appears in a single
6113 translation unit, the best way to silence the warning is to give it
6114 internal linkage by putting it in an anonymous namespace as well. The
6115 compiler doesn't give this warning for types defined in the main .C
6116 file, as those are unlikely to have multiple definitions.
6117 @option{-Wsubobject-linkage} is enabled by default.
6119 @item -Wdangling-else
6120 @opindex Wdangling-else
6121 @opindex Wno-dangling-else
6122 Warn about constructions where there may be confusion to which
6123 @code{if} statement an @code{else} branch belongs. Here is an example of
6138 In C/C++, every @code{else} branch belongs to the innermost possible
6139 @code{if} statement, which in this example is @code{if (b)}. This is
6140 often not what the programmer expected, as illustrated in the above
6141 example by indentation the programmer chose. When there is the
6142 potential for this confusion, GCC issues a warning when this flag
6143 is specified. To eliminate the warning, add explicit braces around
6144 the innermost @code{if} statement so there is no way the @code{else}
6145 can belong to the enclosing @code{if}. The resulting code
6162 This warning is enabled by @option{-Wparentheses}.
6166 @opindex Wno-date-time
6167 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6168 are encountered as they might prevent bit-wise-identical reproducible
6171 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6172 @opindex Wdelete-incomplete
6173 @opindex Wno-delete-incomplete
6174 Warn when deleting a pointer to incomplete type, which may cause
6175 undefined behavior at runtime. This warning is enabled by default.
6177 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6178 @opindex Wuseless-cast
6179 @opindex Wno-useless-cast
6180 Warn when an expression is casted to its own type.
6183 @opindex Wempty-body
6184 @opindex Wno-empty-body
6185 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6186 while} statement. This warning is also enabled by @option{-Wextra}.
6188 @item -Wenum-compare
6189 @opindex Wenum-compare
6190 @opindex Wno-enum-compare
6191 Warn about a comparison between values of different enumerated types.
6192 In C++ enumerated type mismatches in conditional expressions are also
6193 diagnosed and the warning is enabled by default. In C this warning is
6194 enabled by @option{-Wall}.
6196 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6197 @opindex Wextra-semi
6198 @opindex Wno-extra-semi
6199 Warn about redundant semicolon after in-class function definition.
6201 @item -Wjump-misses-init @r{(C, Objective-C only)}
6202 @opindex Wjump-misses-init
6203 @opindex Wno-jump-misses-init
6204 Warn if a @code{goto} statement or a @code{switch} statement jumps
6205 forward across the initialization of a variable, or jumps backward to a
6206 label after the variable has been initialized. This only warns about
6207 variables that are initialized when they are declared. This warning is
6208 only supported for C and Objective-C; in C++ this sort of branch is an
6211 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6212 can be disabled with the @option{-Wno-jump-misses-init} option.
6214 @item -Wsign-compare
6215 @opindex Wsign-compare
6216 @opindex Wno-sign-compare
6217 @cindex warning for comparison of signed and unsigned values
6218 @cindex comparison of signed and unsigned values, warning
6219 @cindex signed and unsigned values, comparison warning
6220 Warn when a comparison between signed and unsigned values could produce
6221 an incorrect result when the signed value is converted to unsigned.
6222 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6223 also enabled by @option{-Wextra}.
6225 @item -Wsign-conversion
6226 @opindex Wsign-conversion
6227 @opindex Wno-sign-conversion
6228 Warn for implicit conversions that may change the sign of an integer
6229 value, like assigning a signed integer expression to an unsigned
6230 integer variable. An explicit cast silences the warning. In C, this
6231 option is enabled also by @option{-Wconversion}.
6233 @item -Wfloat-conversion
6234 @opindex Wfloat-conversion
6235 @opindex Wno-float-conversion
6236 Warn for implicit conversions that reduce the precision of a real value.
6237 This includes conversions from real to integer, and from higher precision
6238 real to lower precision real values. This option is also enabled by
6239 @option{-Wconversion}.
6241 @item -Wno-scalar-storage-order
6242 @opindex -Wno-scalar-storage-order
6243 @opindex -Wscalar-storage-order
6244 Do not warn on suspicious constructs involving reverse scalar storage order.
6246 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6247 @opindex Wsized-deallocation
6248 @opindex Wno-sized-deallocation
6249 Warn about a definition of an unsized deallocation function
6251 void operator delete (void *) noexcept;
6252 void operator delete[] (void *) noexcept;
6254 without a definition of the corresponding sized deallocation function
6256 void operator delete (void *, std::size_t) noexcept;
6257 void operator delete[] (void *, std::size_t) noexcept;
6259 or vice versa. Enabled by @option{-Wextra} along with
6260 @option{-fsized-deallocation}.
6262 @item -Wsizeof-pointer-div
6263 @opindex Wsizeof-pointer-div
6264 @opindex Wno-sizeof-pointer-div
6265 Warn for suspicious divisions of two sizeof expressions that divide
6266 the pointer size by the element size, which is the usual way to compute
6267 the array size but won't work out correctly with pointers. This warning
6268 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
6269 not an array, but a pointer. This warning is enabled by @option{-Wall}.
6271 @item -Wsizeof-pointer-memaccess
6272 @opindex Wsizeof-pointer-memaccess
6273 @opindex Wno-sizeof-pointer-memaccess
6274 Warn for suspicious length parameters to certain string and memory built-in
6275 functions if the argument uses @code{sizeof}. This warning triggers for
6276 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
6277 an array, but a pointer, and suggests a possible fix, or about
6278 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
6279 also warns about calls to bounded string copy functions like @code{strncat}
6280 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
6281 the source array. For example, in the following function the call to
6282 @code{strncat} specifies the size of the source string as the bound. That
6283 is almost certainly a mistake and so the call is diagnosed.
6285 void make_file (const char *name)
6287 char path[PATH_MAX];
6288 strncpy (path, name, sizeof path - 1);
6289 strncat (path, ".text", sizeof ".text");
6294 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
6296 @item -Wsizeof-array-argument
6297 @opindex Wsizeof-array-argument
6298 @opindex Wno-sizeof-array-argument
6299 Warn when the @code{sizeof} operator is applied to a parameter that is
6300 declared as an array in a function definition. This warning is enabled by
6301 default for C and C++ programs.
6303 @item -Wmemset-elt-size
6304 @opindex Wmemset-elt-size
6305 @opindex Wno-memset-elt-size
6306 Warn for suspicious calls to the @code{memset} built-in function, if the
6307 first argument references an array, and the third argument is a number
6308 equal to the number of elements, but not equal to the size of the array
6309 in memory. This indicates that the user has omitted a multiplication by
6310 the element size. This warning is enabled by @option{-Wall}.
6312 @item -Wmemset-transposed-args
6313 @opindex Wmemset-transposed-args
6314 @opindex Wno-memset-transposed-args
6315 Warn for suspicious calls to the @code{memset} built-in function, if the
6316 second argument is not zero and the third argument is zero. This warns e.g.@
6317 about @code{memset (buf, sizeof buf, 0)} where most probably
6318 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6319 is only emitted if the third argument is literal zero. If it is some
6320 expression that is folded to zero, a cast of zero to some type, etc.,
6321 it is far less likely that the user has mistakenly exchanged the arguments
6322 and no warning is emitted. This warning is enabled by @option{-Wall}.
6326 @opindex Wno-address
6327 Warn about suspicious uses of memory addresses. These include using
6328 the address of a function in a conditional expression, such as
6329 @code{void func(void); if (func)}, and comparisons against the memory
6330 address of a string literal, such as @code{if (x == "abc")}. Such
6331 uses typically indicate a programmer error: the address of a function
6332 always evaluates to true, so their use in a conditional usually
6333 indicate that the programmer forgot the parentheses in a function
6334 call; and comparisons against string literals result in unspecified
6335 behavior and are not portable in C, so they usually indicate that the
6336 programmer intended to use @code{strcmp}. This warning is enabled by
6340 @opindex Wlogical-op
6341 @opindex Wno-logical-op
6342 Warn about suspicious uses of logical operators in expressions.
6343 This includes using logical operators in contexts where a
6344 bit-wise operator is likely to be expected. Also warns when
6345 the operands of a logical operator are the same:
6348 if (a < 0 && a < 0) @{ @dots{} @}
6351 @item -Wlogical-not-parentheses
6352 @opindex Wlogical-not-parentheses
6353 @opindex Wno-logical-not-parentheses
6354 Warn about logical not used on the left hand side operand of a comparison.
6355 This option does not warn if the right operand is considered to be a boolean
6356 expression. Its purpose is to detect suspicious code like the following:
6360 if (!a > 1) @{ @dots{} @}
6363 It is possible to suppress the warning by wrapping the LHS into
6366 if ((!a) > 1) @{ @dots{} @}
6369 This warning is enabled by @option{-Wall}.
6371 @item -Waggregate-return
6372 @opindex Waggregate-return
6373 @opindex Wno-aggregate-return
6374 Warn if any functions that return structures or unions are defined or
6375 called. (In languages where you can return an array, this also elicits
6378 @item -Wno-aggressive-loop-optimizations
6379 @opindex Wno-aggressive-loop-optimizations
6380 @opindex Waggressive-loop-optimizations
6381 Warn if in a loop with constant number of iterations the compiler detects
6382 undefined behavior in some statement during one or more of the iterations.
6384 @item -Wno-attributes
6385 @opindex Wno-attributes
6386 @opindex Wattributes
6387 Do not warn if an unexpected @code{__attribute__} is used, such as
6388 unrecognized attributes, function attributes applied to variables,
6389 etc. This does not stop errors for incorrect use of supported
6392 @item -Wno-builtin-declaration-mismatch
6393 @opindex Wno-builtin-declaration-mismatch
6394 @opindex Wbuiltin-declaration-mismatch
6395 Warn if a built-in function is declared with the wrong signature or
6397 This warning is enabled by default.
6399 @item -Wno-builtin-macro-redefined
6400 @opindex Wno-builtin-macro-redefined
6401 @opindex Wbuiltin-macro-redefined
6402 Do not warn if certain built-in macros are redefined. This suppresses
6403 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6404 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6406 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6407 @opindex Wstrict-prototypes
6408 @opindex Wno-strict-prototypes
6409 Warn if a function is declared or defined without specifying the
6410 argument types. (An old-style function definition is permitted without
6411 a warning if preceded by a declaration that specifies the argument
6414 @item -Wold-style-declaration @r{(C and Objective-C only)}
6415 @opindex Wold-style-declaration
6416 @opindex Wno-old-style-declaration
6417 Warn for obsolescent usages, according to the C Standard, in a
6418 declaration. For example, warn if storage-class specifiers like
6419 @code{static} are not the first things in a declaration. This warning
6420 is also enabled by @option{-Wextra}.
6422 @item -Wold-style-definition @r{(C and Objective-C only)}
6423 @opindex Wold-style-definition
6424 @opindex Wno-old-style-definition
6425 Warn if an old-style function definition is used. A warning is given
6426 even if there is a previous prototype.
6428 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6429 @opindex Wmissing-parameter-type
6430 @opindex Wno-missing-parameter-type
6431 A function parameter is declared without a type specifier in K&R-style
6438 This warning is also enabled by @option{-Wextra}.
6440 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6441 @opindex Wmissing-prototypes
6442 @opindex Wno-missing-prototypes
6443 Warn if a global function is defined without a previous prototype
6444 declaration. This warning is issued even if the definition itself
6445 provides a prototype. Use this option to detect global functions
6446 that do not have a matching prototype declaration in a header file.
6447 This option is not valid for C++ because all function declarations
6448 provide prototypes and a non-matching declaration declares an
6449 overload rather than conflict with an earlier declaration.
6450 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6452 @item -Wmissing-declarations
6453 @opindex Wmissing-declarations
6454 @opindex Wno-missing-declarations
6455 Warn if a global function is defined without a previous declaration.
6456 Do so even if the definition itself provides a prototype.
6457 Use this option to detect global functions that are not declared in
6458 header files. In C, no warnings are issued for functions with previous
6459 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6460 missing prototypes. In C++, no warnings are issued for function templates,
6461 or for inline functions, or for functions in anonymous namespaces.
6463 @item -Wmissing-field-initializers
6464 @opindex Wmissing-field-initializers
6465 @opindex Wno-missing-field-initializers
6469 Warn if a structure's initializer has some fields missing. For
6470 example, the following code causes such a warning, because
6471 @code{x.h} is implicitly zero:
6474 struct s @{ int f, g, h; @};
6475 struct s x = @{ 3, 4 @};
6478 This option does not warn about designated initializers, so the following
6479 modification does not trigger a warning:
6482 struct s @{ int f, g, h; @};
6483 struct s x = @{ .f = 3, .g = 4 @};
6486 In C this option does not warn about the universal zero initializer
6490 struct s @{ int f, g, h; @};
6491 struct s x = @{ 0 @};
6494 Likewise, in C++ this option does not warn about the empty @{ @}
6495 initializer, for example:
6498 struct s @{ int f, g, h; @};
6502 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6503 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6505 @item -Wno-multichar
6506 @opindex Wno-multichar
6508 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6509 Usually they indicate a typo in the user's code, as they have
6510 implementation-defined values, and should not be used in portable code.
6512 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
6513 @opindex Wnormalized=
6514 @opindex Wnormalized
6515 @opindex Wno-normalized
6518 @cindex character set, input normalization
6519 In ISO C and ISO C++, two identifiers are different if they are
6520 different sequences of characters. However, sometimes when characters
6521 outside the basic ASCII character set are used, you can have two
6522 different character sequences that look the same. To avoid confusion,
6523 the ISO 10646 standard sets out some @dfn{normalization rules} which
6524 when applied ensure that two sequences that look the same are turned into
6525 the same sequence. GCC can warn you if you are using identifiers that
6526 have not been normalized; this option controls that warning.
6528 There are four levels of warning supported by GCC@. The default is
6529 @option{-Wnormalized=nfc}, which warns about any identifier that is
6530 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6531 recommended form for most uses. It is equivalent to
6532 @option{-Wnormalized}.
6534 Unfortunately, there are some characters allowed in identifiers by
6535 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6536 identifiers. That is, there's no way to use these symbols in portable
6537 ISO C or C++ and have all your identifiers in NFC@.
6538 @option{-Wnormalized=id} suppresses the warning for these characters.
6539 It is hoped that future versions of the standards involved will correct
6540 this, which is why this option is not the default.
6542 You can switch the warning off for all characters by writing
6543 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6544 only do this if you are using some other normalization scheme (like
6545 ``D''), because otherwise you can easily create bugs that are
6546 literally impossible to see.
6548 Some characters in ISO 10646 have distinct meanings but look identical
6549 in some fonts or display methodologies, especially once formatting has
6550 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6551 LETTER N'', displays just like a regular @code{n} that has been
6552 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6553 normalization scheme to convert all these into a standard form as
6554 well, and GCC warns if your code is not in NFKC if you use
6555 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6556 about every identifier that contains the letter O because it might be
6557 confused with the digit 0, and so is not the default, but may be
6558 useful as a local coding convention if the programming environment
6559 cannot be fixed to display these characters distinctly.
6561 @item -Wno-deprecated
6562 @opindex Wno-deprecated
6563 @opindex Wdeprecated
6564 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6566 @item -Wno-deprecated-declarations
6567 @opindex Wno-deprecated-declarations
6568 @opindex Wdeprecated-declarations
6569 Do not warn about uses of functions (@pxref{Function Attributes}),
6570 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6571 Attributes}) marked as deprecated by using the @code{deprecated}
6575 @opindex Wno-overflow
6577 Do not warn about compile-time overflow in constant expressions.
6582 Warn about One Definition Rule violations during link-time optimization.
6583 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6586 @opindex Wopenm-simd
6587 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6588 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6589 option can be used to relax the cost model.
6591 @item -Woverride-init @r{(C and Objective-C only)}
6592 @opindex Woverride-init
6593 @opindex Wno-override-init
6597 Warn if an initialized field without side effects is overridden when
6598 using designated initializers (@pxref{Designated Inits, , Designated
6601 This warning is included in @option{-Wextra}. To get other
6602 @option{-Wextra} warnings without this one, use @option{-Wextra
6603 -Wno-override-init}.
6605 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6606 @opindex Woverride-init-side-effects
6607 @opindex Wno-override-init-side-effects
6608 Warn if an initialized field with side effects is overridden when
6609 using designated initializers (@pxref{Designated Inits, , Designated
6610 Initializers}). This warning is enabled by default.
6615 Warn if a structure is given the packed attribute, but the packed
6616 attribute has no effect on the layout or size of the structure.
6617 Such structures may be mis-aligned for little benefit. For
6618 instance, in this code, the variable @code{f.x} in @code{struct bar}
6619 is misaligned even though @code{struct bar} does not itself
6620 have the packed attribute:
6627 @} __attribute__((packed));
6635 @item -Wpacked-bitfield-compat
6636 @opindex Wpacked-bitfield-compat
6637 @opindex Wno-packed-bitfield-compat
6638 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6639 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6640 the change can lead to differences in the structure layout. GCC
6641 informs you when the offset of such a field has changed in GCC 4.4.
6642 For example there is no longer a 4-bit padding between field @code{a}
6643 and @code{b} in this structure:
6650 @} __attribute__ ((packed));
6653 This warning is enabled by default. Use
6654 @option{-Wno-packed-bitfield-compat} to disable this warning.
6656 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
6657 @opindex Wpacked-not-aligned
6658 @opindex Wno-packed-not-aligned
6659 Warn if a structure field with explicitly specified alignment in a
6660 packed struct or union is misaligned. For example, a warning will
6661 be issued on @code{struct S}, like, @code{warning: alignment 1 of
6662 'struct S' is less than 8}, in this code:
6666 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
6667 struct __attribute__ ((packed)) S @{
6673 This warning is enabled by @option{-Wall}.
6678 Warn if padding is included in a structure, either to align an element
6679 of the structure or to align the whole structure. Sometimes when this
6680 happens it is possible to rearrange the fields of the structure to
6681 reduce the padding and so make the structure smaller.
6683 @item -Wredundant-decls
6684 @opindex Wredundant-decls
6685 @opindex Wno-redundant-decls
6686 Warn if anything is declared more than once in the same scope, even in
6687 cases where multiple declaration is valid and changes nothing.
6691 @opindex Wno-restrict
6692 Warn when an argument passed to a restrict-qualified parameter
6693 aliases with another argument.
6695 @item -Wnested-externs @r{(C and Objective-C only)}
6696 @opindex Wnested-externs
6697 @opindex Wno-nested-externs
6698 Warn if an @code{extern} declaration is encountered within a function.
6700 @item -Wno-inherited-variadic-ctor
6701 @opindex Winherited-variadic-ctor
6702 @opindex Wno-inherited-variadic-ctor
6703 Suppress warnings about use of C++11 inheriting constructors when the
6704 base class inherited from has a C variadic constructor; the warning is
6705 on by default because the ellipsis is not inherited.
6710 Warn if a function that is declared as inline cannot be inlined.
6711 Even with this option, the compiler does not warn about failures to
6712 inline functions declared in system headers.
6714 The compiler uses a variety of heuristics to determine whether or not
6715 to inline a function. For example, the compiler takes into account
6716 the size of the function being inlined and the amount of inlining
6717 that has already been done in the current function. Therefore,
6718 seemingly insignificant changes in the source program can cause the
6719 warnings produced by @option{-Winline} to appear or disappear.
6721 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6722 @opindex Wno-invalid-offsetof
6723 @opindex Winvalid-offsetof
6724 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6725 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6726 to a non-standard-layout type is undefined. In existing C++ implementations,
6727 however, @code{offsetof} typically gives meaningful results.
6728 This flag is for users who are aware that they are
6729 writing nonportable code and who have deliberately chosen to ignore the
6732 The restrictions on @code{offsetof} may be relaxed in a future version
6733 of the C++ standard.
6735 @item -Wint-in-bool-context
6736 @opindex Wint-in-bool-context
6737 @opindex Wno-int-in-bool-context
6738 Warn for suspicious use of integer values where boolean values are expected,
6739 such as conditional expressions (?:) using non-boolean integer constants in
6740 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6741 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6742 for all kinds of multiplications regardless of the data type.
6743 This warning is enabled by @option{-Wall}.
6745 @item -Wno-int-to-pointer-cast
6746 @opindex Wno-int-to-pointer-cast
6747 @opindex Wint-to-pointer-cast
6748 Suppress warnings from casts to pointer type of an integer of a
6749 different size. In C++, casting to a pointer type of smaller size is
6750 an error. @option{Wint-to-pointer-cast} is enabled by default.
6753 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6754 @opindex Wno-pointer-to-int-cast
6755 @opindex Wpointer-to-int-cast
6756 Suppress warnings from casts from a pointer to an integer type of a
6760 @opindex Winvalid-pch
6761 @opindex Wno-invalid-pch
6762 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6763 the search path but cannot be used.
6767 @opindex Wno-long-long
6768 Warn if @code{long long} type is used. This is enabled by either
6769 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6770 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6772 @item -Wvariadic-macros
6773 @opindex Wvariadic-macros
6774 @opindex Wno-variadic-macros
6775 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6776 alternate syntax is used in ISO C99 mode. This is enabled by either
6777 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6778 messages, use @option{-Wno-variadic-macros}.
6782 @opindex Wno-varargs
6783 Warn upon questionable usage of the macros used to handle variable
6784 arguments like @code{va_start}. This is default. To inhibit the
6785 warning messages, use @option{-Wno-varargs}.
6787 @item -Wvector-operation-performance
6788 @opindex Wvector-operation-performance
6789 @opindex Wno-vector-operation-performance
6790 Warn if vector operation is not implemented via SIMD capabilities of the
6791 architecture. Mainly useful for the performance tuning.
6792 Vector operation can be implemented @code{piecewise}, which means that the
6793 scalar operation is performed on every vector element;
6794 @code{in parallel}, which means that the vector operation is implemented
6795 using scalars of wider type, which normally is more performance efficient;
6796 and @code{as a single scalar}, which means that vector fits into a
6799 @item -Wno-virtual-move-assign
6800 @opindex Wvirtual-move-assign
6801 @opindex Wno-virtual-move-assign
6802 Suppress warnings about inheriting from a virtual base with a
6803 non-trivial C++11 move assignment operator. This is dangerous because
6804 if the virtual base is reachable along more than one path, it is
6805 moved multiple times, which can mean both objects end up in the
6806 moved-from state. If the move assignment operator is written to avoid
6807 moving from a moved-from object, this warning can be disabled.
6812 Warn if a variable-length array is used in the code.
6813 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6814 the variable-length array.
6816 @item -Wvla-larger-than=@var{n}
6817 If this option is used, the compiler will warn on uses of
6818 variable-length arrays where the size is either unbounded, or bounded
6819 by an argument that can be larger than @var{n} bytes. This is similar
6820 to how @option{-Walloca-larger-than=@var{n}} works, but with
6821 variable-length arrays.
6823 Note that GCC may optimize small variable-length arrays of a known
6824 value into plain arrays, so this warning may not get triggered for
6827 This warning is not enabled by @option{-Wall}, and is only active when
6828 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6830 See also @option{-Walloca-larger-than=@var{n}}.
6832 @item -Wvolatile-register-var
6833 @opindex Wvolatile-register-var
6834 @opindex Wno-volatile-register-var
6835 Warn if a register variable is declared volatile. The volatile
6836 modifier does not inhibit all optimizations that may eliminate reads
6837 and/or writes to register variables. This warning is enabled by
6840 @item -Wdisabled-optimization
6841 @opindex Wdisabled-optimization
6842 @opindex Wno-disabled-optimization
6843 Warn if a requested optimization pass is disabled. This warning does
6844 not generally indicate that there is anything wrong with your code; it
6845 merely indicates that GCC's optimizers are unable to handle the code
6846 effectively. Often, the problem is that your code is too big or too
6847 complex; GCC refuses to optimize programs when the optimization
6848 itself is likely to take inordinate amounts of time.
6850 @item -Wpointer-sign @r{(C and Objective-C only)}
6851 @opindex Wpointer-sign
6852 @opindex Wno-pointer-sign
6853 Warn for pointer argument passing or assignment with different signedness.
6854 This option is only supported for C and Objective-C@. It is implied by
6855 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6856 @option{-Wno-pointer-sign}.
6858 @item -Wstack-protector
6859 @opindex Wstack-protector
6860 @opindex Wno-stack-protector
6861 This option is only active when @option{-fstack-protector} is active. It
6862 warns about functions that are not protected against stack smashing.
6864 @item -Woverlength-strings
6865 @opindex Woverlength-strings
6866 @opindex Wno-overlength-strings
6867 Warn about string constants that are longer than the ``minimum
6868 maximum'' length specified in the C standard. Modern compilers
6869 generally allow string constants that are much longer than the
6870 standard's minimum limit, but very portable programs should avoid
6871 using longer strings.
6873 The limit applies @emph{after} string constant concatenation, and does
6874 not count the trailing NUL@. In C90, the limit was 509 characters; in
6875 C99, it was raised to 4095. C++98 does not specify a normative
6876 minimum maximum, so we do not diagnose overlength strings in C++@.
6878 This option is implied by @option{-Wpedantic}, and can be disabled with
6879 @option{-Wno-overlength-strings}.
6881 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6882 @opindex Wunsuffixed-float-constants
6884 Issue a warning for any floating constant that does not have
6885 a suffix. When used together with @option{-Wsystem-headers} it
6886 warns about such constants in system header files. This can be useful
6887 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6888 from the decimal floating-point extension to C99.
6890 @item -Wno-designated-init @r{(C and Objective-C only)}
6891 Suppress warnings when a positional initializer is used to initialize
6892 a structure that has been marked with the @code{designated_init}
6896 Issue a warning when HSAIL cannot be emitted for the compiled function or
6901 @node Debugging Options
6902 @section Options for Debugging Your Program
6903 @cindex options, debugging
6904 @cindex debugging information options
6906 To tell GCC to emit extra information for use by a debugger, in almost
6907 all cases you need only to add @option{-g} to your other options.
6909 GCC allows you to use @option{-g} with
6910 @option{-O}. The shortcuts taken by optimized code may occasionally
6911 be surprising: some variables you declared may not exist
6912 at all; flow of control may briefly move where you did not expect it;
6913 some statements may not be executed because they compute constant
6914 results or their values are already at hand; some statements may
6915 execute in different places because they have been moved out of loops.
6916 Nevertheless it is possible to debug optimized output. This makes
6917 it reasonable to use the optimizer for programs that might have bugs.
6919 If you are not using some other optimization option, consider
6920 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6921 With no @option{-O} option at all, some compiler passes that collect
6922 information useful for debugging do not run at all, so that
6923 @option{-Og} may result in a better debugging experience.
6928 Produce debugging information in the operating system's native format
6929 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6932 On most systems that use stabs format, @option{-g} enables use of extra
6933 debugging information that only GDB can use; this extra information
6934 makes debugging work better in GDB but probably makes other debuggers
6936 refuse to read the program. If you want to control for certain whether
6937 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6938 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6942 Produce debugging information for use by GDB@. This means to use the
6943 most expressive format available (DWARF, stabs, or the native format
6944 if neither of those are supported), including GDB extensions if at all
6948 @itemx -gdwarf-@var{version}
6950 Produce debugging information in DWARF format (if that is supported).
6951 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6952 for most targets is 4. DWARF Version 5 is only experimental.
6954 Note that with DWARF Version 2, some ports require and always
6955 use some non-conflicting DWARF 3 extensions in the unwind tables.
6957 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6958 for maximum benefit.
6960 GCC no longer supports DWARF Version 1, which is substantially
6961 different than Version 2 and later. For historical reasons, some
6962 other DWARF-related options such as
6963 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6964 in their names, but apply to all currently-supported versions of DWARF.
6968 Produce debugging information in stabs format (if that is supported),
6969 without GDB extensions. This is the format used by DBX on most BSD
6970 systems. On MIPS, Alpha and System V Release 4 systems this option
6971 produces stabs debugging output that is not understood by DBX@.
6972 On System V Release 4 systems this option requires the GNU assembler.
6976 Produce debugging information in stabs format (if that is supported),
6977 using GNU extensions understood only by the GNU debugger (GDB)@. The
6978 use of these extensions is likely to make other debuggers crash or
6979 refuse to read the program.
6983 Produce debugging information in XCOFF format (if that is supported).
6984 This is the format used by the DBX debugger on IBM RS/6000 systems.
6988 Produce debugging information in XCOFF format (if that is supported),
6989 using GNU extensions understood only by the GNU debugger (GDB)@. The
6990 use of these extensions is likely to make other debuggers crash or
6991 refuse to read the program, and may cause assemblers other than the GNU
6992 assembler (GAS) to fail with an error.
6996 Produce debugging information in Alpha/VMS debug format (if that is
6997 supported). This is the format used by DEBUG on Alpha/VMS systems.
7000 @itemx -ggdb@var{level}
7001 @itemx -gstabs@var{level}
7002 @itemx -gxcoff@var{level}
7003 @itemx -gvms@var{level}
7004 Request debugging information and also use @var{level} to specify how
7005 much information. The default level is 2.
7007 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7010 Level 1 produces minimal information, enough for making backtraces in
7011 parts of the program that you don't plan to debug. This includes
7012 descriptions of functions and external variables, and line number
7013 tables, but no information about local variables.
7015 Level 3 includes extra information, such as all the macro definitions
7016 present in the program. Some debuggers support macro expansion when
7017 you use @option{-g3}.
7019 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7020 confusion with @option{-gdwarf-@var{level}}.
7021 Instead use an additional @option{-g@var{level}} option to change the
7022 debug level for DWARF.
7024 @item -feliminate-unused-debug-symbols
7025 @opindex feliminate-unused-debug-symbols
7026 Produce debugging information in stabs format (if that is supported),
7027 for only symbols that are actually used.
7029 @item -femit-class-debug-always
7030 @opindex femit-class-debug-always
7031 Instead of emitting debugging information for a C++ class in only one
7032 object file, emit it in all object files using the class. This option
7033 should be used only with debuggers that are unable to handle the way GCC
7034 normally emits debugging information for classes because using this
7035 option increases the size of debugging information by as much as a
7038 @item -fno-merge-debug-strings
7039 @opindex fmerge-debug-strings
7040 @opindex fno-merge-debug-strings
7041 Direct the linker to not merge together strings in the debugging
7042 information that are identical in different object files. Merging is
7043 not supported by all assemblers or linkers. Merging decreases the size
7044 of the debug information in the output file at the cost of increasing
7045 link processing time. Merging is enabled by default.
7047 @item -fdebug-prefix-map=@var{old}=@var{new}
7048 @opindex fdebug-prefix-map
7049 When compiling files in directory @file{@var{old}}, record debugging
7050 information describing them as in @file{@var{new}} instead. This can be
7051 used to replace a build-time path with an install-time path in the debug info.
7052 It can also be used to change an absolute path to a relative path by using
7053 @file{.} for @var{new}. This can give more reproducible builds, which are
7054 location independent, but may require an extra command to tell GDB where to
7055 find the source files.
7057 @item -fvar-tracking
7058 @opindex fvar-tracking
7059 Run variable tracking pass. It computes where variables are stored at each
7060 position in code. Better debugging information is then generated
7061 (if the debugging information format supports this information).
7063 It is enabled by default when compiling with optimization (@option{-Os},
7064 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7065 the debug info format supports it.
7067 @item -fvar-tracking-assignments
7068 @opindex fvar-tracking-assignments
7069 @opindex fno-var-tracking-assignments
7070 Annotate assignments to user variables early in the compilation and
7071 attempt to carry the annotations over throughout the compilation all the
7072 way to the end, in an attempt to improve debug information while
7073 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7075 It can be enabled even if var-tracking is disabled, in which case
7076 annotations are created and maintained, but discarded at the end.
7077 By default, this flag is enabled together with @option{-fvar-tracking},
7078 except when selective scheduling is enabled.
7081 @opindex gsplit-dwarf
7082 Separate as much DWARF debugging information as possible into a
7083 separate output file with the extension @file{.dwo}. This option allows
7084 the build system to avoid linking files with debug information. To
7085 be useful, this option requires a debugger capable of reading @file{.dwo}
7090 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7092 @item -ggnu-pubnames
7093 @opindex ggnu-pubnames
7094 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7095 suitable for conversion into a GDB@ index. This option is only useful
7096 with a linker that can produce GDB@ index version 7.
7098 @item -fdebug-types-section
7099 @opindex fdebug-types-section
7100 @opindex fno-debug-types-section
7101 When using DWARF Version 4 or higher, type DIEs can be put into
7102 their own @code{.debug_types} section instead of making them part of the
7103 @code{.debug_info} section. It is more efficient to put them in a separate
7104 comdat sections since the linker can then remove duplicates.
7105 But not all DWARF consumers support @code{.debug_types} sections yet
7106 and on some objects @code{.debug_types} produces larger instead of smaller
7107 debugging information.
7109 @item -grecord-gcc-switches
7110 @item -gno-record-gcc-switches
7111 @opindex grecord-gcc-switches
7112 @opindex gno-record-gcc-switches
7113 This switch causes the command-line options used to invoke the
7114 compiler that may affect code generation to be appended to the
7115 DW_AT_producer attribute in DWARF debugging information. The options
7116 are concatenated with spaces separating them from each other and from
7117 the compiler version.
7118 It is enabled by default.
7119 See also @option{-frecord-gcc-switches} for another
7120 way of storing compiler options into the object file.
7122 @item -gstrict-dwarf
7123 @opindex gstrict-dwarf
7124 Disallow using extensions of later DWARF standard version than selected
7125 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7126 DWARF extensions from later standard versions is allowed.
7128 @item -gno-strict-dwarf
7129 @opindex gno-strict-dwarf
7130 Allow using extensions of later DWARF standard version than selected with
7131 @option{-gdwarf-@var{version}}.
7134 @item -gno-column-info
7135 @opindex gcolumn-info
7136 @opindex gno-column-info
7137 Emit location column information into DWARF debugging information, rather
7138 than just file and line.
7139 This option is enabled by default.
7141 @item -gz@r{[}=@var{type}@r{]}
7143 Produce compressed debug sections in DWARF format, if that is supported.
7144 If @var{type} is not given, the default type depends on the capabilities
7145 of the assembler and linker used. @var{type} may be one of
7146 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
7147 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
7148 compression in traditional GNU format). If the linker doesn't support
7149 writing compressed debug sections, the option is rejected. Otherwise,
7150 if the assembler does not support them, @option{-gz} is silently ignored
7151 when producing object files.
7153 @item -femit-struct-debug-baseonly
7154 @opindex femit-struct-debug-baseonly
7155 Emit debug information for struct-like types
7156 only when the base name of the compilation source file
7157 matches the base name of file in which the struct is defined.
7159 This option substantially reduces the size of debugging information,
7160 but at significant potential loss in type information to the debugger.
7161 See @option{-femit-struct-debug-reduced} for a less aggressive option.
7162 See @option{-femit-struct-debug-detailed} for more detailed control.
7164 This option works only with DWARF debug output.
7166 @item -femit-struct-debug-reduced
7167 @opindex femit-struct-debug-reduced
7168 Emit debug information for struct-like types
7169 only when the base name of the compilation source file
7170 matches the base name of file in which the type is defined,
7171 unless the struct is a template or defined in a system header.
7173 This option significantly reduces the size of debugging information,
7174 with some potential loss in type information to the debugger.
7175 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
7176 See @option{-femit-struct-debug-detailed} for more detailed control.
7178 This option works only with DWARF debug output.
7180 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
7181 @opindex femit-struct-debug-detailed
7182 Specify the struct-like types
7183 for which the compiler generates debug information.
7184 The intent is to reduce duplicate struct debug information
7185 between different object files within the same program.
7187 This option is a detailed version of
7188 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
7189 which serves for most needs.
7191 A specification has the syntax@*
7192 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
7194 The optional first word limits the specification to
7195 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
7196 A struct type is used directly when it is the type of a variable, member.
7197 Indirect uses arise through pointers to structs.
7198 That is, when use of an incomplete struct is valid, the use is indirect.
7200 @samp{struct one direct; struct two * indirect;}.
7202 The optional second word limits the specification to
7203 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
7204 Generic structs are a bit complicated to explain.
7205 For C++, these are non-explicit specializations of template classes,
7206 or non-template classes within the above.
7207 Other programming languages have generics,
7208 but @option{-femit-struct-debug-detailed} does not yet implement them.
7210 The third word specifies the source files for those
7211 structs for which the compiler should emit debug information.
7212 The values @samp{none} and @samp{any} have the normal meaning.
7213 The value @samp{base} means that
7214 the base of name of the file in which the type declaration appears
7215 must match the base of the name of the main compilation file.
7216 In practice, this means that when compiling @file{foo.c}, debug information
7217 is generated for types declared in that file and @file{foo.h},
7218 but not other header files.
7219 The value @samp{sys} means those types satisfying @samp{base}
7220 or declared in system or compiler headers.
7222 You may need to experiment to determine the best settings for your application.
7224 The default is @option{-femit-struct-debug-detailed=all}.
7226 This option works only with DWARF debug output.
7228 @item -fno-dwarf2-cfi-asm
7229 @opindex fdwarf2-cfi-asm
7230 @opindex fno-dwarf2-cfi-asm
7231 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
7232 instead of using GAS @code{.cfi_*} directives.
7234 @item -fno-eliminate-unused-debug-types
7235 @opindex feliminate-unused-debug-types
7236 @opindex fno-eliminate-unused-debug-types
7237 Normally, when producing DWARF output, GCC avoids producing debug symbol
7238 output for types that are nowhere used in the source file being compiled.
7239 Sometimes it is useful to have GCC emit debugging
7240 information for all types declared in a compilation
7241 unit, regardless of whether or not they are actually used
7242 in that compilation unit, for example
7243 if, in the debugger, you want to cast a value to a type that is
7244 not actually used in your program (but is declared). More often,
7245 however, this results in a significant amount of wasted space.
7248 @node Optimize Options
7249 @section Options That Control Optimization
7250 @cindex optimize options
7251 @cindex options, optimization
7253 These options control various sorts of optimizations.
7255 Without any optimization option, the compiler's goal is to reduce the
7256 cost of compilation and to make debugging produce the expected
7257 results. Statements are independent: if you stop the program with a
7258 breakpoint between statements, you can then assign a new value to any
7259 variable or change the program counter to any other statement in the
7260 function and get exactly the results you expect from the source
7263 Turning on optimization flags makes the compiler attempt to improve
7264 the performance and/or code size at the expense of compilation time
7265 and possibly the ability to debug the program.
7267 The compiler performs optimization based on the knowledge it has of the
7268 program. Compiling multiple files at once to a single output file mode allows
7269 the compiler to use information gained from all of the files when compiling
7272 Not all optimizations are controlled directly by a flag. Only
7273 optimizations that have a flag are listed in this section.
7275 Most optimizations are only enabled if an @option{-O} level is set on
7276 the command line. Otherwise they are disabled, even if individual
7277 optimization flags are specified.
7279 Depending on the target and how GCC was configured, a slightly different
7280 set of optimizations may be enabled at each @option{-O} level than
7281 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7282 to find out the exact set of optimizations that are enabled at each level.
7283 @xref{Overall Options}, for examples.
7290 Optimize. Optimizing compilation takes somewhat more time, and a lot
7291 more memory for a large function.
7293 With @option{-O}, the compiler tries to reduce code size and execution
7294 time, without performing any optimizations that take a great deal of
7297 @option{-O} turns on the following optimization flags:
7300 -fbranch-count-reg @gol
7301 -fcombine-stack-adjustments @gol
7303 -fcprop-registers @gol
7306 -fdelayed-branch @gol
7308 -fforward-propagate @gol
7309 -fguess-branch-probability @gol
7310 -fif-conversion2 @gol
7311 -fif-conversion @gol
7312 -finline-functions-called-once @gol
7313 -fipa-pure-const @gol
7315 -fipa-reference @gol
7316 -fmerge-constants @gol
7317 -fmove-loop-invariants @gol
7318 -fomit-frame-pointer @gol
7319 -freorder-blocks @gol
7321 -fshrink-wrap-separate @gol
7322 -fsplit-wide-types @gol
7328 -ftree-coalesce-vars @gol
7329 -ftree-copy-prop @gol
7331 -ftree-dominator-opts @gol
7333 -ftree-forwprop @gol
7345 Optimize even more. GCC performs nearly all supported optimizations
7346 that do not involve a space-speed tradeoff.
7347 As compared to @option{-O}, this option increases both compilation time
7348 and the performance of the generated code.
7350 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7351 also turns on the following optimization flags:
7352 @gccoptlist{-fthread-jumps @gol
7353 -falign-functions -falign-jumps @gol
7354 -falign-loops -falign-labels @gol
7357 -fcse-follow-jumps -fcse-skip-blocks @gol
7358 -fdelete-null-pointer-checks @gol
7359 -fdevirtualize -fdevirtualize-speculatively @gol
7360 -fexpensive-optimizations @gol
7361 -fgcse -fgcse-lm @gol
7362 -fhoist-adjacent-loads @gol
7363 -finline-small-functions @gol
7364 -findirect-inlining @gol
7370 -fisolate-erroneous-paths-dereference @gol
7372 -foptimize-sibling-calls @gol
7373 -foptimize-strlen @gol
7374 -fpartial-inlining @gol
7376 -freorder-blocks-algorithm=stc @gol
7377 -freorder-blocks-and-partition -freorder-functions @gol
7378 -frerun-cse-after-loop @gol
7379 -fsched-interblock -fsched-spec @gol
7380 -fschedule-insns -fschedule-insns2 @gol
7381 -fstore-merging @gol
7382 -fstrict-aliasing @gol
7383 -ftree-builtin-call-dce @gol
7384 -ftree-switch-conversion -ftree-tail-merge @gol
7385 -fcode-hoisting @gol
7390 Please note the warning under @option{-fgcse} about
7391 invoking @option{-O2} on programs that use computed gotos.
7395 Optimize yet more. @option{-O3} turns on all optimizations specified
7396 by @option{-O2} and also turns on the following optimization flags:
7397 @gccoptlist{-finline-functions @gol
7398 -funswitch-loops @gol
7399 -fpredictive-commoning @gol
7400 -fgcse-after-reload @gol
7401 -ftree-loop-vectorize @gol
7402 -ftree-loop-distribution @gol
7403 -ftree-loop-distribute-patterns @gol
7405 -ftree-slp-vectorize @gol
7406 -fvect-cost-model @gol
7407 -ftree-partial-pre @gol
7413 Reduce compilation time and make debugging produce the expected
7414 results. This is the default.
7418 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7419 do not typically increase code size. It also performs further
7420 optimizations designed to reduce code size.
7422 @option{-Os} disables the following optimization flags:
7423 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7424 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7425 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7429 Disregard strict standards compliance. @option{-Ofast} enables all
7430 @option{-O3} optimizations. It also enables optimizations that are not
7431 valid for all standard-compliant programs.
7432 It turns on @option{-ffast-math} and the Fortran-specific
7433 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
7434 specified, and @option{-fno-protect-parens}.
7438 Optimize debugging experience. @option{-Og} enables optimizations
7439 that do not interfere with debugging. It should be the optimization
7440 level of choice for the standard edit-compile-debug cycle, offering
7441 a reasonable level of optimization while maintaining fast compilation
7442 and a good debugging experience.
7445 If you use multiple @option{-O} options, with or without level numbers,
7446 the last such option is the one that is effective.
7448 Options of the form @option{-f@var{flag}} specify machine-independent
7449 flags. Most flags have both positive and negative forms; the negative
7450 form of @option{-ffoo} is @option{-fno-foo}. In the table
7451 below, only one of the forms is listed---the one you typically
7452 use. You can figure out the other form by either removing @samp{no-}
7455 The following options control specific optimizations. They are either
7456 activated by @option{-O} options or are related to ones that are. You
7457 can use the following flags in the rare cases when ``fine-tuning'' of
7458 optimizations to be performed is desired.
7461 @item -fno-defer-pop
7462 @opindex fno-defer-pop
7463 Always pop the arguments to each function call as soon as that function
7464 returns. For machines that must pop arguments after a function call,
7465 the compiler normally lets arguments accumulate on the stack for several
7466 function calls and pops them all at once.
7468 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7470 @item -fforward-propagate
7471 @opindex fforward-propagate
7472 Perform a forward propagation pass on RTL@. The pass tries to combine two
7473 instructions and checks if the result can be simplified. If loop unrolling
7474 is active, two passes are performed and the second is scheduled after
7477 This option is enabled by default at optimization levels @option{-O},
7478 @option{-O2}, @option{-O3}, @option{-Os}.
7480 @item -ffp-contract=@var{style}
7481 @opindex ffp-contract
7482 @option{-ffp-contract=off} disables floating-point expression contraction.
7483 @option{-ffp-contract=fast} enables floating-point expression contraction
7484 such as forming of fused multiply-add operations if the target has
7485 native support for them.
7486 @option{-ffp-contract=on} enables floating-point expression contraction
7487 if allowed by the language standard. This is currently not implemented
7488 and treated equal to @option{-ffp-contract=off}.
7490 The default is @option{-ffp-contract=fast}.
7492 @item -fomit-frame-pointer
7493 @opindex fomit-frame-pointer
7494 Omit the frame pointer in functions that don't need one. This avoids the
7495 instructions to save, set up and restore the frame pointer; on many targets
7496 it also makes an extra register available.
7498 On some targets this flag has no effect because the standard calling sequence
7499 always uses a frame pointer, so it cannot be omitted.
7501 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
7502 is used in all functions. Several targets always omit the frame pointer in
7505 Enabled by default at @option{-O} and higher.
7507 @item -foptimize-sibling-calls
7508 @opindex foptimize-sibling-calls
7509 Optimize sibling and tail recursive calls.
7511 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7513 @item -foptimize-strlen
7514 @opindex foptimize-strlen
7515 Optimize various standard C string functions (e.g. @code{strlen},
7516 @code{strchr} or @code{strcpy}) and
7517 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7519 Enabled at levels @option{-O2}, @option{-O3}.
7523 Do not expand any functions inline apart from those marked with
7524 the @code{always_inline} attribute. This is the default when not
7527 Single functions can be exempted from inlining by marking them
7528 with the @code{noinline} attribute.
7530 @item -finline-small-functions
7531 @opindex finline-small-functions
7532 Integrate functions into their callers when their body is smaller than expected
7533 function call code (so overall size of program gets smaller). The compiler
7534 heuristically decides which functions are simple enough to be worth integrating
7535 in this way. This inlining applies to all functions, even those not declared
7538 Enabled at level @option{-O2}.
7540 @item -findirect-inlining
7541 @opindex findirect-inlining
7542 Inline also indirect calls that are discovered to be known at compile
7543 time thanks to previous inlining. This option has any effect only
7544 when inlining itself is turned on by the @option{-finline-functions}
7545 or @option{-finline-small-functions} options.
7547 Enabled at level @option{-O2}.
7549 @item -finline-functions
7550 @opindex finline-functions
7551 Consider all functions for inlining, even if they are not declared inline.
7552 The compiler heuristically decides which functions are worth integrating
7555 If all calls to a given function are integrated, and the function is
7556 declared @code{static}, then the function is normally not output as
7557 assembler code in its own right.
7559 Enabled at level @option{-O3}.
7561 @item -finline-functions-called-once
7562 @opindex finline-functions-called-once
7563 Consider all @code{static} functions called once for inlining into their
7564 caller even if they are not marked @code{inline}. If a call to a given
7565 function is integrated, then the function is not output as assembler code
7568 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7570 @item -fearly-inlining
7571 @opindex fearly-inlining
7572 Inline functions marked by @code{always_inline} and functions whose body seems
7573 smaller than the function call overhead early before doing
7574 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7575 makes profiling significantly cheaper and usually inlining faster on programs
7576 having large chains of nested wrapper functions.
7582 Perform interprocedural scalar replacement of aggregates, removal of
7583 unused parameters and replacement of parameters passed by reference
7584 by parameters passed by value.
7586 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7588 @item -finline-limit=@var{n}
7589 @opindex finline-limit
7590 By default, GCC limits the size of functions that can be inlined. This flag
7591 allows coarse control of this limit. @var{n} is the size of functions that
7592 can be inlined in number of pseudo instructions.
7594 Inlining is actually controlled by a number of parameters, which may be
7595 specified individually by using @option{--param @var{name}=@var{value}}.
7596 The @option{-finline-limit=@var{n}} option sets some of these parameters
7600 @item max-inline-insns-single
7601 is set to @var{n}/2.
7602 @item max-inline-insns-auto
7603 is set to @var{n}/2.
7606 See below for a documentation of the individual
7607 parameters controlling inlining and for the defaults of these parameters.
7609 @emph{Note:} there may be no value to @option{-finline-limit} that results
7610 in default behavior.
7612 @emph{Note:} pseudo instruction represents, in this particular context, an
7613 abstract measurement of function's size. In no way does it represent a count
7614 of assembly instructions and as such its exact meaning might change from one
7615 release to an another.
7617 @item -fno-keep-inline-dllexport
7618 @opindex fno-keep-inline-dllexport
7619 This is a more fine-grained version of @option{-fkeep-inline-functions},
7620 which applies only to functions that are declared using the @code{dllexport}
7621 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
7624 @item -fkeep-inline-functions
7625 @opindex fkeep-inline-functions
7626 In C, emit @code{static} functions that are declared @code{inline}
7627 into the object file, even if the function has been inlined into all
7628 of its callers. This switch does not affect functions using the
7629 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7630 inline functions into the object file.
7632 @item -fkeep-static-functions
7633 @opindex fkeep-static-functions
7634 Emit @code{static} functions into the object file, even if the function
7637 @item -fkeep-static-consts
7638 @opindex fkeep-static-consts
7639 Emit variables declared @code{static const} when optimization isn't turned
7640 on, even if the variables aren't referenced.
7642 GCC enables this option by default. If you want to force the compiler to
7643 check if a variable is referenced, regardless of whether or not
7644 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7646 @item -fmerge-constants
7647 @opindex fmerge-constants
7648 Attempt to merge identical constants (string constants and floating-point
7649 constants) across compilation units.
7651 This option is the default for optimized compilation if the assembler and
7652 linker support it. Use @option{-fno-merge-constants} to inhibit this
7655 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7657 @item -fmerge-all-constants
7658 @opindex fmerge-all-constants
7659 Attempt to merge identical constants and identical variables.
7661 This option implies @option{-fmerge-constants}. In addition to
7662 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7663 arrays or initialized constant variables with integral or floating-point
7664 types. Languages like C or C++ require each variable, including multiple
7665 instances of the same variable in recursive calls, to have distinct locations,
7666 so using this option results in non-conforming
7669 @item -fmodulo-sched
7670 @opindex fmodulo-sched
7671 Perform swing modulo scheduling immediately before the first scheduling
7672 pass. This pass looks at innermost loops and reorders their
7673 instructions by overlapping different iterations.
7675 @item -fmodulo-sched-allow-regmoves
7676 @opindex fmodulo-sched-allow-regmoves
7677 Perform more aggressive SMS-based modulo scheduling with register moves
7678 allowed. By setting this flag certain anti-dependences edges are
7679 deleted, which triggers the generation of reg-moves based on the
7680 life-range analysis. This option is effective only with
7681 @option{-fmodulo-sched} enabled.
7683 @item -fno-branch-count-reg
7684 @opindex fno-branch-count-reg
7685 Avoid running a pass scanning for opportunities to use ``decrement and
7686 branch'' instructions on a count register instead of generating sequences
7687 of instructions that decrement a register, compare it against zero, and
7688 then branch based upon the result. This option is only meaningful on
7689 architectures that support such instructions, which include x86, PowerPC,
7690 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7691 doesn't remove the decrement and branch instructions from the generated
7692 instruction stream introduced by other optimization passes.
7694 Enabled by default at @option{-O1} and higher.
7696 The default is @option{-fbranch-count-reg}.
7698 @item -fno-function-cse
7699 @opindex fno-function-cse
7700 Do not put function addresses in registers; make each instruction that
7701 calls a constant function contain the function's address explicitly.
7703 This option results in less efficient code, but some strange hacks
7704 that alter the assembler output may be confused by the optimizations
7705 performed when this option is not used.
7707 The default is @option{-ffunction-cse}
7709 @item -fno-zero-initialized-in-bss
7710 @opindex fno-zero-initialized-in-bss
7711 If the target supports a BSS section, GCC by default puts variables that
7712 are initialized to zero into BSS@. This can save space in the resulting
7715 This option turns off this behavior because some programs explicitly
7716 rely on variables going to the data section---e.g., so that the
7717 resulting executable can find the beginning of that section and/or make
7718 assumptions based on that.
7720 The default is @option{-fzero-initialized-in-bss}.
7722 @item -fthread-jumps
7723 @opindex fthread-jumps
7724 Perform optimizations that check to see if a jump branches to a
7725 location where another comparison subsumed by the first is found. If
7726 so, the first branch is redirected to either the destination of the
7727 second branch or a point immediately following it, depending on whether
7728 the condition is known to be true or false.
7730 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7732 @item -fsplit-wide-types
7733 @opindex fsplit-wide-types
7734 When using a type that occupies multiple registers, such as @code{long
7735 long} on a 32-bit system, split the registers apart and allocate them
7736 independently. This normally generates better code for those types,
7737 but may make debugging more difficult.
7739 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7742 @item -fcse-follow-jumps
7743 @opindex fcse-follow-jumps
7744 In common subexpression elimination (CSE), scan through jump instructions
7745 when the target of the jump is not reached by any other path. For
7746 example, when CSE encounters an @code{if} statement with an
7747 @code{else} clause, CSE follows the jump when the condition
7750 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7752 @item -fcse-skip-blocks
7753 @opindex fcse-skip-blocks
7754 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7755 follow jumps that conditionally skip over blocks. When CSE
7756 encounters a simple @code{if} statement with no else clause,
7757 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7758 body of the @code{if}.
7760 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7762 @item -frerun-cse-after-loop
7763 @opindex frerun-cse-after-loop
7764 Re-run common subexpression elimination after loop optimizations are
7767 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7771 Perform a global common subexpression elimination pass.
7772 This pass also performs global constant and copy propagation.
7774 @emph{Note:} When compiling a program using computed gotos, a GCC
7775 extension, you may get better run-time performance if you disable
7776 the global common subexpression elimination pass by adding
7777 @option{-fno-gcse} to the command line.
7779 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7783 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7784 attempts to move loads that are only killed by stores into themselves. This
7785 allows a loop containing a load/store sequence to be changed to a load outside
7786 the loop, and a copy/store within the loop.
7788 Enabled by default when @option{-fgcse} is enabled.
7792 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7793 global common subexpression elimination. This pass attempts to move
7794 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7795 loops containing a load/store sequence can be changed to a load before
7796 the loop and a store after the loop.
7798 Not enabled at any optimization level.
7802 When @option{-fgcse-las} is enabled, the global common subexpression
7803 elimination pass eliminates redundant loads that come after stores to the
7804 same memory location (both partial and full redundancies).
7806 Not enabled at any optimization level.
7808 @item -fgcse-after-reload
7809 @opindex fgcse-after-reload
7810 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7811 pass is performed after reload. The purpose of this pass is to clean up
7814 @item -faggressive-loop-optimizations
7815 @opindex faggressive-loop-optimizations
7816 This option tells the loop optimizer to use language constraints to
7817 derive bounds for the number of iterations of a loop. This assumes that
7818 loop code does not invoke undefined behavior by for example causing signed
7819 integer overflows or out-of-bound array accesses. The bounds for the
7820 number of iterations of a loop are used to guide loop unrolling and peeling
7821 and loop exit test optimizations.
7822 This option is enabled by default.
7824 @item -funconstrained-commons
7825 @opindex funconstrained-commons
7826 This option tells the compiler that variables declared in common blocks
7827 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7828 prevents certain optimizations that depend on knowing the array bounds.
7830 @item -fcrossjumping
7831 @opindex fcrossjumping
7832 Perform cross-jumping transformation.
7833 This transformation unifies equivalent code and saves code size. The
7834 resulting code may or may not perform better than without cross-jumping.
7836 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7838 @item -fauto-inc-dec
7839 @opindex fauto-inc-dec
7840 Combine increments or decrements of addresses with memory accesses.
7841 This pass is always skipped on architectures that do not have
7842 instructions to support this. Enabled by default at @option{-O} and
7843 higher on architectures that support this.
7847 Perform dead code elimination (DCE) on RTL@.
7848 Enabled by default at @option{-O} and higher.
7852 Perform dead store elimination (DSE) on RTL@.
7853 Enabled by default at @option{-O} and higher.
7855 @item -fif-conversion
7856 @opindex fif-conversion
7857 Attempt to transform conditional jumps into branch-less equivalents. This
7858 includes use of conditional moves, min, max, set flags and abs instructions, and
7859 some tricks doable by standard arithmetics. The use of conditional execution
7860 on chips where it is available is controlled by @option{-fif-conversion2}.
7862 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7864 @item -fif-conversion2
7865 @opindex fif-conversion2
7866 Use conditional execution (where available) to transform conditional jumps into
7867 branch-less equivalents.
7869 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7871 @item -fdeclone-ctor-dtor
7872 @opindex fdeclone-ctor-dtor
7873 The C++ ABI requires multiple entry points for constructors and
7874 destructors: one for a base subobject, one for a complete object, and
7875 one for a virtual destructor that calls operator delete afterwards.
7876 For a hierarchy with virtual bases, the base and complete variants are
7877 clones, which means two copies of the function. With this option, the
7878 base and complete variants are changed to be thunks that call a common
7881 Enabled by @option{-Os}.
7883 @item -fdelete-null-pointer-checks
7884 @opindex fdelete-null-pointer-checks
7885 Assume that programs cannot safely dereference null pointers, and that
7886 no code or data element resides at address zero.
7887 This option enables simple constant
7888 folding optimizations at all optimization levels. In addition, other
7889 optimization passes in GCC use this flag to control global dataflow
7890 analyses that eliminate useless checks for null pointers; these assume
7891 that a memory access to address zero always results in a trap, so
7892 that if a pointer is checked after it has already been dereferenced,
7895 Note however that in some environments this assumption is not true.
7896 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7897 for programs that depend on that behavior.
7899 This option is enabled by default on most targets. On Nios II ELF, it
7900 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
7902 Passes that use the dataflow information
7903 are enabled independently at different optimization levels.
7905 @item -fdevirtualize
7906 @opindex fdevirtualize
7907 Attempt to convert calls to virtual functions to direct calls. This
7908 is done both within a procedure and interprocedurally as part of
7909 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7910 propagation (@option{-fipa-cp}).
7911 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7913 @item -fdevirtualize-speculatively
7914 @opindex fdevirtualize-speculatively
7915 Attempt to convert calls to virtual functions to speculative direct calls.
7916 Based on the analysis of the type inheritance graph, determine for a given call
7917 the set of likely targets. If the set is small, preferably of size 1, change
7918 the call into a conditional deciding between direct and indirect calls. The
7919 speculative calls enable more optimizations, such as inlining. When they seem
7920 useless after further optimization, they are converted back into original form.
7922 @item -fdevirtualize-at-ltrans
7923 @opindex fdevirtualize-at-ltrans
7924 Stream extra information needed for aggressive devirtualization when running
7925 the link-time optimizer in local transformation mode.
7926 This option enables more devirtualization but
7927 significantly increases the size of streamed data. For this reason it is
7928 disabled by default.
7930 @item -fexpensive-optimizations
7931 @opindex fexpensive-optimizations
7932 Perform a number of minor optimizations that are relatively expensive.
7934 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7938 Attempt to remove redundant extension instructions. This is especially
7939 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7940 registers after writing to their lower 32-bit half.
7942 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7943 @option{-O3}, @option{-Os}.
7945 @item -fno-lifetime-dse
7946 @opindex fno-lifetime-dse
7947 In C++ the value of an object is only affected by changes within its
7948 lifetime: when the constructor begins, the object has an indeterminate
7949 value, and any changes during the lifetime of the object are dead when
7950 the object is destroyed. Normally dead store elimination will take
7951 advantage of this; if your code relies on the value of the object
7952 storage persisting beyond the lifetime of the object, you can use this
7953 flag to disable this optimization. To preserve stores before the
7954 constructor starts (e.g. because your operator new clears the object
7955 storage) but still treat the object as dead after the destructor you,
7956 can use @option{-flifetime-dse=1}. The default behavior can be
7957 explicitly selected with @option{-flifetime-dse=2}.
7958 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7960 @item -flive-range-shrinkage
7961 @opindex flive-range-shrinkage
7962 Attempt to decrease register pressure through register live range
7963 shrinkage. This is helpful for fast processors with small or moderate
7966 @item -fira-algorithm=@var{algorithm}
7967 @opindex fira-algorithm
7968 Use the specified coloring algorithm for the integrated register
7969 allocator. The @var{algorithm} argument can be @samp{priority}, which
7970 specifies Chow's priority coloring, or @samp{CB}, which specifies
7971 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7972 for all architectures, but for those targets that do support it, it is
7973 the default because it generates better code.
7975 @item -fira-region=@var{region}
7976 @opindex fira-region
7977 Use specified regions for the integrated register allocator. The
7978 @var{region} argument should be one of the following:
7983 Use all loops as register allocation regions.
7984 This can give the best results for machines with a small and/or
7985 irregular register set.
7988 Use all loops except for loops with small register pressure
7989 as the regions. This value usually gives
7990 the best results in most cases and for most architectures,
7991 and is enabled by default when compiling with optimization for speed
7992 (@option{-O}, @option{-O2}, @dots{}).
7995 Use all functions as a single region.
7996 This typically results in the smallest code size, and is enabled by default for
7997 @option{-Os} or @option{-O0}.
8001 @item -fira-hoist-pressure
8002 @opindex fira-hoist-pressure
8003 Use IRA to evaluate register pressure in the code hoisting pass for
8004 decisions to hoist expressions. This option usually results in smaller
8005 code, but it can slow the compiler down.
8007 This option is enabled at level @option{-Os} for all targets.
8009 @item -fira-loop-pressure
8010 @opindex fira-loop-pressure
8011 Use IRA to evaluate register pressure in loops for decisions to move
8012 loop invariants. This option usually results in generation
8013 of faster and smaller code on machines with large register files (>= 32
8014 registers), but it can slow the compiler down.
8016 This option is enabled at level @option{-O3} for some targets.
8018 @item -fno-ira-share-save-slots
8019 @opindex fno-ira-share-save-slots
8020 Disable sharing of stack slots used for saving call-used hard
8021 registers living through a call. Each hard register gets a
8022 separate stack slot, and as a result function stack frames are
8025 @item -fno-ira-share-spill-slots
8026 @opindex fno-ira-share-spill-slots
8027 Disable sharing of stack slots allocated for pseudo-registers. Each
8028 pseudo-register that does not get a hard register gets a separate
8029 stack slot, and as a result function stack frames are larger.
8033 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8034 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8035 values if it is profitable.
8037 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8039 @item -fdelayed-branch
8040 @opindex fdelayed-branch
8041 If supported for the target machine, attempt to reorder instructions
8042 to exploit instruction slots available after delayed branch
8045 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8047 @item -fschedule-insns
8048 @opindex fschedule-insns
8049 If supported for the target machine, attempt to reorder instructions to
8050 eliminate execution stalls due to required data being unavailable. This
8051 helps machines that have slow floating point or memory load instructions
8052 by allowing other instructions to be issued until the result of the load
8053 or floating-point instruction is required.
8055 Enabled at levels @option{-O2}, @option{-O3}.
8057 @item -fschedule-insns2
8058 @opindex fschedule-insns2
8059 Similar to @option{-fschedule-insns}, but requests an additional pass of
8060 instruction scheduling after register allocation has been done. This is
8061 especially useful on machines with a relatively small number of
8062 registers and where memory load instructions take more than one cycle.
8064 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8066 @item -fno-sched-interblock
8067 @opindex fno-sched-interblock
8068 Don't schedule instructions across basic blocks. This is normally
8069 enabled by default when scheduling before register allocation, i.e.@:
8070 with @option{-fschedule-insns} or at @option{-O2} or higher.
8072 @item -fno-sched-spec
8073 @opindex fno-sched-spec
8074 Don't allow speculative motion of non-load instructions. This is normally
8075 enabled by default when scheduling before register allocation, i.e.@:
8076 with @option{-fschedule-insns} or at @option{-O2} or higher.
8078 @item -fsched-pressure
8079 @opindex fsched-pressure
8080 Enable register pressure sensitive insn scheduling before register
8081 allocation. This only makes sense when scheduling before register
8082 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8083 @option{-O2} or higher. Usage of this option can improve the
8084 generated code and decrease its size by preventing register pressure
8085 increase above the number of available hard registers and subsequent
8086 spills in register allocation.
8088 @item -fsched-spec-load
8089 @opindex fsched-spec-load
8090 Allow speculative motion of some load instructions. This only makes
8091 sense when scheduling before register allocation, i.e.@: with
8092 @option{-fschedule-insns} or at @option{-O2} or higher.
8094 @item -fsched-spec-load-dangerous
8095 @opindex fsched-spec-load-dangerous
8096 Allow speculative motion of more load instructions. This only makes
8097 sense when scheduling before register allocation, i.e.@: with
8098 @option{-fschedule-insns} or at @option{-O2} or higher.
8100 @item -fsched-stalled-insns
8101 @itemx -fsched-stalled-insns=@var{n}
8102 @opindex fsched-stalled-insns
8103 Define how many insns (if any) can be moved prematurely from the queue
8104 of stalled insns into the ready list during the second scheduling pass.
8105 @option{-fno-sched-stalled-insns} means that no insns are moved
8106 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8107 on how many queued insns can be moved prematurely.
8108 @option{-fsched-stalled-insns} without a value is equivalent to
8109 @option{-fsched-stalled-insns=1}.
8111 @item -fsched-stalled-insns-dep
8112 @itemx -fsched-stalled-insns-dep=@var{n}
8113 @opindex fsched-stalled-insns-dep
8114 Define how many insn groups (cycles) are examined for a dependency
8115 on a stalled insn that is a candidate for premature removal from the queue
8116 of stalled insns. This has an effect only during the second scheduling pass,
8117 and only if @option{-fsched-stalled-insns} is used.
8118 @option{-fno-sched-stalled-insns-dep} is equivalent to
8119 @option{-fsched-stalled-insns-dep=0}.
8120 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8121 @option{-fsched-stalled-insns-dep=1}.
8123 @item -fsched2-use-superblocks
8124 @opindex fsched2-use-superblocks
8125 When scheduling after register allocation, use superblock scheduling.
8126 This allows motion across basic block boundaries,
8127 resulting in faster schedules. This option is experimental, as not all machine
8128 descriptions used by GCC model the CPU closely enough to avoid unreliable
8129 results from the algorithm.
8131 This only makes sense when scheduling after register allocation, i.e.@: with
8132 @option{-fschedule-insns2} or at @option{-O2} or higher.
8134 @item -fsched-group-heuristic
8135 @opindex fsched-group-heuristic
8136 Enable the group heuristic in the scheduler. This heuristic favors
8137 the instruction that belongs to a schedule group. This is enabled
8138 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8139 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8141 @item -fsched-critical-path-heuristic
8142 @opindex fsched-critical-path-heuristic
8143 Enable the critical-path heuristic in the scheduler. This heuristic favors
8144 instructions on the critical path. This is enabled by default when
8145 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8146 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8148 @item -fsched-spec-insn-heuristic
8149 @opindex fsched-spec-insn-heuristic
8150 Enable the speculative instruction heuristic in the scheduler. This
8151 heuristic favors speculative instructions with greater dependency weakness.
8152 This is enabled by default when scheduling is enabled, i.e.@:
8153 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8154 or at @option{-O2} or higher.
8156 @item -fsched-rank-heuristic
8157 @opindex fsched-rank-heuristic
8158 Enable the rank heuristic in the scheduler. This heuristic favors
8159 the instruction belonging to a basic block with greater size or frequency.
8160 This is enabled by default when scheduling is enabled, i.e.@:
8161 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8162 at @option{-O2} or higher.
8164 @item -fsched-last-insn-heuristic
8165 @opindex fsched-last-insn-heuristic
8166 Enable the last-instruction heuristic in the scheduler. This heuristic
8167 favors the instruction that is less dependent on the last instruction
8168 scheduled. This is enabled by default when scheduling is enabled,
8169 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8170 at @option{-O2} or higher.
8172 @item -fsched-dep-count-heuristic
8173 @opindex fsched-dep-count-heuristic
8174 Enable the dependent-count heuristic in the scheduler. This heuristic
8175 favors the instruction that has more instructions depending on it.
8176 This is enabled by default when scheduling is enabled, i.e.@:
8177 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8178 at @option{-O2} or higher.
8180 @item -freschedule-modulo-scheduled-loops
8181 @opindex freschedule-modulo-scheduled-loops
8182 Modulo scheduling is performed before traditional scheduling. If a loop
8183 is modulo scheduled, later scheduling passes may change its schedule.
8184 Use this option to control that behavior.
8186 @item -fselective-scheduling
8187 @opindex fselective-scheduling
8188 Schedule instructions using selective scheduling algorithm. Selective
8189 scheduling runs instead of the first scheduler pass.
8191 @item -fselective-scheduling2
8192 @opindex fselective-scheduling2
8193 Schedule instructions using selective scheduling algorithm. Selective
8194 scheduling runs instead of the second scheduler pass.
8196 @item -fsel-sched-pipelining
8197 @opindex fsel-sched-pipelining
8198 Enable software pipelining of innermost loops during selective scheduling.
8199 This option has no effect unless one of @option{-fselective-scheduling} or
8200 @option{-fselective-scheduling2} is turned on.
8202 @item -fsel-sched-pipelining-outer-loops
8203 @opindex fsel-sched-pipelining-outer-loops
8204 When pipelining loops during selective scheduling, also pipeline outer loops.
8205 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8207 @item -fsemantic-interposition
8208 @opindex fsemantic-interposition
8209 Some object formats, like ELF, allow interposing of symbols by the
8211 This means that for symbols exported from the DSO, the compiler cannot perform
8212 interprocedural propagation, inlining and other optimizations in anticipation
8213 that the function or variable in question may change. While this feature is
8214 useful, for example, to rewrite memory allocation functions by a debugging
8215 implementation, it is expensive in the terms of code quality.
8216 With @option{-fno-semantic-interposition} the compiler assumes that
8217 if interposition happens for functions the overwriting function will have
8218 precisely the same semantics (and side effects).
8219 Similarly if interposition happens
8220 for variables, the constructor of the variable will be the same. The flag
8221 has no effect for functions explicitly declared inline
8222 (where it is never allowed for interposition to change semantics)
8223 and for symbols explicitly declared weak.
8226 @opindex fshrink-wrap
8227 Emit function prologues only before parts of the function that need it,
8228 rather than at the top of the function. This flag is enabled by default at
8229 @option{-O} and higher.
8231 @item -fshrink-wrap-separate
8232 @opindex fshrink-wrap-separate
8233 Shrink-wrap separate parts of the prologue and epilogue separately, so that
8234 those parts are only executed when needed.
8235 This option is on by default, but has no effect unless @option{-fshrink-wrap}
8236 is also turned on and the target supports this.
8238 @item -fcaller-saves
8239 @opindex fcaller-saves
8240 Enable allocation of values to registers that are clobbered by
8241 function calls, by emitting extra instructions to save and restore the
8242 registers around such calls. Such allocation is done only when it
8243 seems to result in better code.
8245 This option is always enabled by default on certain machines, usually
8246 those which have no call-preserved registers to use instead.
8248 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8250 @item -fcombine-stack-adjustments
8251 @opindex fcombine-stack-adjustments
8252 Tracks stack adjustments (pushes and pops) and stack memory references
8253 and then tries to find ways to combine them.
8255 Enabled by default at @option{-O1} and higher.
8259 Use caller save registers for allocation if those registers are not used by
8260 any called function. In that case it is not necessary to save and restore
8261 them around calls. This is only possible if called functions are part of
8262 same compilation unit as current function and they are compiled before it.
8264 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
8265 is disabled if generated code will be instrumented for profiling
8266 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
8267 exactly (this happens on targets that do not expose prologues
8268 and epilogues in RTL).
8270 @item -fconserve-stack
8271 @opindex fconserve-stack
8272 Attempt to minimize stack usage. The compiler attempts to use less
8273 stack space, even if that makes the program slower. This option
8274 implies setting the @option{large-stack-frame} parameter to 100
8275 and the @option{large-stack-frame-growth} parameter to 400.
8277 @item -ftree-reassoc
8278 @opindex ftree-reassoc
8279 Perform reassociation on trees. This flag is enabled by default
8280 at @option{-O} and higher.
8282 @item -fcode-hoisting
8283 @opindex fcode-hoisting
8284 Perform code hoisting. Code hoisting tries to move the
8285 evaluation of expressions executed on all paths to the function exit
8286 as early as possible. This is especially useful as a code size
8287 optimization, but it often helps for code speed as well.
8288 This flag is enabled by default at @option{-O2} and higher.
8292 Perform partial redundancy elimination (PRE) on trees. This flag is
8293 enabled by default at @option{-O2} and @option{-O3}.
8295 @item -ftree-partial-pre
8296 @opindex ftree-partial-pre
8297 Make partial redundancy elimination (PRE) more aggressive. This flag is
8298 enabled by default at @option{-O3}.
8300 @item -ftree-forwprop
8301 @opindex ftree-forwprop
8302 Perform forward propagation on trees. This flag is enabled by default
8303 at @option{-O} and higher.
8307 Perform full redundancy elimination (FRE) on trees. The difference
8308 between FRE and PRE is that FRE only considers expressions
8309 that are computed on all paths leading to the redundant computation.
8310 This analysis is faster than PRE, though it exposes fewer redundancies.
8311 This flag is enabled by default at @option{-O} and higher.
8313 @item -ftree-phiprop
8314 @opindex ftree-phiprop
8315 Perform hoisting of loads from conditional pointers on trees. This
8316 pass is enabled by default at @option{-O} and higher.
8318 @item -fhoist-adjacent-loads
8319 @opindex fhoist-adjacent-loads
8320 Speculatively hoist loads from both branches of an if-then-else if the
8321 loads are from adjacent locations in the same structure and the target
8322 architecture has a conditional move instruction. This flag is enabled
8323 by default at @option{-O2} and higher.
8325 @item -ftree-copy-prop
8326 @opindex ftree-copy-prop
8327 Perform copy propagation on trees. This pass eliminates unnecessary
8328 copy operations. This flag is enabled by default at @option{-O} and
8331 @item -fipa-pure-const
8332 @opindex fipa-pure-const
8333 Discover which functions are pure or constant.
8334 Enabled by default at @option{-O} and higher.
8336 @item -fipa-reference
8337 @opindex fipa-reference
8338 Discover which static variables do not escape the
8340 Enabled by default at @option{-O} and higher.
8344 Perform interprocedural pointer analysis and interprocedural modification
8345 and reference analysis. This option can cause excessive memory and
8346 compile-time usage on large compilation units. It is not enabled by
8347 default at any optimization level.
8350 @opindex fipa-profile
8351 Perform interprocedural profile propagation. The functions called only from
8352 cold functions are marked as cold. Also functions executed once (such as
8353 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8354 functions and loop less parts of functions executed once are then optimized for
8356 Enabled by default at @option{-O} and higher.
8360 Perform interprocedural constant propagation.
8361 This optimization analyzes the program to determine when values passed
8362 to functions are constants and then optimizes accordingly.
8363 This optimization can substantially increase performance
8364 if the application has constants passed to functions.
8365 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8367 @item -fipa-cp-clone
8368 @opindex fipa-cp-clone
8369 Perform function cloning to make interprocedural constant propagation stronger.
8370 When enabled, interprocedural constant propagation performs function cloning
8371 when externally visible function can be called with constant arguments.
8372 Because this optimization can create multiple copies of functions,
8373 it may significantly increase code size
8374 (see @option{--param ipcp-unit-growth=@var{value}}).
8375 This flag is enabled by default at @option{-O3}.
8378 @opindex -fipa-bit-cp
8379 When enabled, perform interprocedural bitwise constant
8380 propagation. This flag is enabled by default at @option{-O2}. It
8381 requires that @option{-fipa-cp} is enabled.
8385 When enabled, perform interprocedural propagation of value
8386 ranges. This flag is enabled by default at @option{-O2}. It requires
8387 that @option{-fipa-cp} is enabled.
8391 Perform Identical Code Folding for functions and read-only variables.
8392 The optimization reduces code size and may disturb unwind stacks by replacing
8393 a function by equivalent one with a different name. The optimization works
8394 more effectively with link-time optimization enabled.
8396 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8397 works on different levels and thus the optimizations are not same - there are
8398 equivalences that are found only by GCC and equivalences found only by Gold.
8400 This flag is enabled by default at @option{-O2} and @option{-Os}.
8402 @item -fisolate-erroneous-paths-dereference
8403 @opindex fisolate-erroneous-paths-dereference
8404 Detect paths that trigger erroneous or undefined behavior due to
8405 dereferencing a null pointer. Isolate those paths from the main control
8406 flow and turn the statement with erroneous or undefined behavior into a trap.
8407 This flag is enabled by default at @option{-O2} and higher and depends on
8408 @option{-fdelete-null-pointer-checks} also being enabled.
8410 @item -fisolate-erroneous-paths-attribute
8411 @opindex fisolate-erroneous-paths-attribute
8412 Detect paths that trigger erroneous or undefined behavior due a null value
8413 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8414 attribute. Isolate those paths from the main control flow and turn the
8415 statement with erroneous or undefined behavior into a trap. This is not
8416 currently enabled, but may be enabled by @option{-O2} in the future.
8420 Perform forward store motion on trees. This flag is
8421 enabled by default at @option{-O} and higher.
8423 @item -ftree-bit-ccp
8424 @opindex ftree-bit-ccp
8425 Perform sparse conditional bit constant propagation on trees and propagate
8426 pointer alignment information.
8427 This pass only operates on local scalar variables and is enabled by default
8428 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8432 Perform sparse conditional constant propagation (CCP) on trees. This
8433 pass only operates on local scalar variables and is enabled by default
8434 at @option{-O} and higher.
8436 @item -fssa-backprop
8437 @opindex fssa-backprop
8438 Propagate information about uses of a value up the definition chain
8439 in order to simplify the definitions. For example, this pass strips
8440 sign operations if the sign of a value never matters. The flag is
8441 enabled by default at @option{-O} and higher.
8444 @opindex fssa-phiopt
8445 Perform pattern matching on SSA PHI nodes to optimize conditional
8446 code. This pass is enabled by default at @option{-O} and higher.
8448 @item -ftree-switch-conversion
8449 @opindex ftree-switch-conversion
8450 Perform conversion of simple initializations in a switch to
8451 initializations from a scalar array. This flag is enabled by default
8452 at @option{-O2} and higher.
8454 @item -ftree-tail-merge
8455 @opindex ftree-tail-merge
8456 Look for identical code sequences. When found, replace one with a jump to the
8457 other. This optimization is known as tail merging or cross jumping. This flag
8458 is enabled by default at @option{-O2} and higher. The compilation time
8460 be limited using @option{max-tail-merge-comparisons} parameter and
8461 @option{max-tail-merge-iterations} parameter.
8465 Perform dead code elimination (DCE) on trees. This flag is enabled by
8466 default at @option{-O} and higher.
8468 @item -ftree-builtin-call-dce
8469 @opindex ftree-builtin-call-dce
8470 Perform conditional dead code elimination (DCE) for calls to built-in functions
8471 that may set @code{errno} but are otherwise side-effect free. This flag is
8472 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8475 @item -ftree-dominator-opts
8476 @opindex ftree-dominator-opts
8477 Perform a variety of simple scalar cleanups (constant/copy
8478 propagation, redundancy elimination, range propagation and expression
8479 simplification) based on a dominator tree traversal. This also
8480 performs jump threading (to reduce jumps to jumps). This flag is
8481 enabled by default at @option{-O} and higher.
8485 Perform dead store elimination (DSE) on trees. A dead store is a store into
8486 a memory location that is later overwritten by another store without
8487 any intervening loads. In this case the earlier store can be deleted. This
8488 flag is enabled by default at @option{-O} and higher.
8492 Perform loop header copying on trees. This is beneficial since it increases
8493 effectiveness of code motion optimizations. It also saves one jump. This flag
8494 is enabled by default at @option{-O} and higher. It is not enabled
8495 for @option{-Os}, since it usually increases code size.
8497 @item -ftree-loop-optimize
8498 @opindex ftree-loop-optimize
8499 Perform loop optimizations on trees. This flag is enabled by default
8500 at @option{-O} and higher.
8502 @item -ftree-loop-linear
8503 @itemx -floop-interchange
8504 @itemx -floop-strip-mine
8506 @itemx -floop-unroll-and-jam
8507 @opindex ftree-loop-linear
8508 @opindex floop-interchange
8509 @opindex floop-strip-mine
8510 @opindex floop-block
8511 @opindex floop-unroll-and-jam
8512 Perform loop nest optimizations. Same as
8513 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8514 to be configured with @option{--with-isl} to enable the Graphite loop
8515 transformation infrastructure.
8517 @item -fgraphite-identity
8518 @opindex fgraphite-identity
8519 Enable the identity transformation for graphite. For every SCoP we generate
8520 the polyhedral representation and transform it back to gimple. Using
8521 @option{-fgraphite-identity} we can check the costs or benefits of the
8522 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8523 are also performed by the code generator isl, like index splitting and
8524 dead code elimination in loops.
8526 @item -floop-nest-optimize
8527 @opindex floop-nest-optimize
8528 Enable the isl based loop nest optimizer. This is a generic loop nest
8529 optimizer based on the Pluto optimization algorithms. It calculates a loop
8530 structure optimized for data-locality and parallelism. This option
8533 @item -floop-parallelize-all
8534 @opindex floop-parallelize-all
8535 Use the Graphite data dependence analysis to identify loops that can
8536 be parallelized. Parallelize all the loops that can be analyzed to
8537 not contain loop carried dependences without checking that it is
8538 profitable to parallelize the loops.
8540 @item -ftree-coalesce-vars
8541 @opindex ftree-coalesce-vars
8542 While transforming the program out of the SSA representation, attempt to
8543 reduce copying by coalescing versions of different user-defined
8544 variables, instead of just compiler temporaries. This may severely
8545 limit the ability to debug an optimized program compiled with
8546 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8547 prevents SSA coalescing of user variables. This option is enabled by
8548 default if optimization is enabled, and it does very little otherwise.
8550 @item -ftree-loop-if-convert
8551 @opindex ftree-loop-if-convert
8552 Attempt to transform conditional jumps in the innermost loops to
8553 branch-less equivalents. The intent is to remove control-flow from
8554 the innermost loops in order to improve the ability of the
8555 vectorization pass to handle these loops. This is enabled by default
8556 if vectorization is enabled.
8558 @item -ftree-loop-distribution
8559 @opindex ftree-loop-distribution
8560 Perform loop distribution. This flag can improve cache performance on
8561 big loop bodies and allow further loop optimizations, like
8562 parallelization or vectorization, to take place. For example, the loop
8579 @item -ftree-loop-distribute-patterns
8580 @opindex ftree-loop-distribute-patterns
8581 Perform loop distribution of patterns that can be code generated with
8582 calls to a library. This flag is enabled by default at @option{-O3}.
8584 This pass distributes the initialization loops and generates a call to
8585 memset zero. For example, the loop
8601 and the initialization loop is transformed into a call to memset zero.
8603 @item -ftree-loop-im
8604 @opindex ftree-loop-im
8605 Perform loop invariant motion on trees. This pass moves only invariants that
8606 are hard to handle at RTL level (function calls, operations that expand to
8607 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8608 operands of conditions that are invariant out of the loop, so that we can use
8609 just trivial invariantness analysis in loop unswitching. The pass also includes
8612 @item -ftree-loop-ivcanon
8613 @opindex ftree-loop-ivcanon
8614 Create a canonical counter for number of iterations in loops for which
8615 determining number of iterations requires complicated analysis. Later
8616 optimizations then may determine the number easily. Useful especially
8617 in connection with unrolling.
8621 Perform induction variable optimizations (strength reduction, induction
8622 variable merging and induction variable elimination) on trees.
8624 @item -ftree-parallelize-loops=n
8625 @opindex ftree-parallelize-loops
8626 Parallelize loops, i.e., split their iteration space to run in n threads.
8627 This is only possible for loops whose iterations are independent
8628 and can be arbitrarily reordered. The optimization is only
8629 profitable on multiprocessor machines, for loops that are CPU-intensive,
8630 rather than constrained e.g.@: by memory bandwidth. This option
8631 implies @option{-pthread}, and thus is only supported on targets
8632 that have support for @option{-pthread}.
8636 Perform function-local points-to analysis on trees. This flag is
8637 enabled by default at @option{-O} and higher.
8641 Perform scalar replacement of aggregates. This pass replaces structure
8642 references with scalars to prevent committing structures to memory too
8643 early. This flag is enabled by default at @option{-O} and higher.
8645 @item -fstore-merging
8646 @opindex fstore-merging
8647 Perform merging of narrow stores to consecutive memory addresses. This pass
8648 merges contiguous stores of immediate values narrower than a word into fewer
8649 wider stores to reduce the number of instructions. This is enabled by default
8650 at @option{-O2} and higher as well as @option{-Os}.
8654 Perform temporary expression replacement during the SSA->normal phase. Single
8655 use/single def temporaries are replaced at their use location with their
8656 defining expression. This results in non-GIMPLE code, but gives the expanders
8657 much more complex trees to work on resulting in better RTL generation. This is
8658 enabled by default at @option{-O} and higher.
8662 Perform straight-line strength reduction on trees. This recognizes related
8663 expressions involving multiplications and replaces them by less expensive
8664 calculations when possible. This is enabled by default at @option{-O} and
8667 @item -ftree-vectorize
8668 @opindex ftree-vectorize
8669 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8670 and @option{-ftree-slp-vectorize} if not explicitly specified.
8672 @item -ftree-loop-vectorize
8673 @opindex ftree-loop-vectorize
8674 Perform loop vectorization on trees. This flag is enabled by default at
8675 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8677 @item -ftree-slp-vectorize
8678 @opindex ftree-slp-vectorize
8679 Perform basic block vectorization on trees. This flag is enabled by default at
8680 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8682 @item -fvect-cost-model=@var{model}
8683 @opindex fvect-cost-model
8684 Alter the cost model used for vectorization. The @var{model} argument
8685 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8686 With the @samp{unlimited} model the vectorized code-path is assumed
8687 to be profitable while with the @samp{dynamic} model a runtime check
8688 guards the vectorized code-path to enable it only for iteration
8689 counts that will likely execute faster than when executing the original
8690 scalar loop. The @samp{cheap} model disables vectorization of
8691 loops where doing so would be cost prohibitive for example due to
8692 required runtime checks for data dependence or alignment but otherwise
8693 is equal to the @samp{dynamic} model.
8694 The default cost model depends on other optimization flags and is
8695 either @samp{dynamic} or @samp{cheap}.
8697 @item -fsimd-cost-model=@var{model}
8698 @opindex fsimd-cost-model
8699 Alter the cost model used for vectorization of loops marked with the OpenMP
8700 or Cilk Plus simd directive. The @var{model} argument should be one of
8701 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8702 have the same meaning as described in @option{-fvect-cost-model} and by
8703 default a cost model defined with @option{-fvect-cost-model} is used.
8707 Perform Value Range Propagation on trees. This is similar to the
8708 constant propagation pass, but instead of values, ranges of values are
8709 propagated. This allows the optimizers to remove unnecessary range
8710 checks like array bound checks and null pointer checks. This is
8711 enabled by default at @option{-O2} and higher. Null pointer check
8712 elimination is only done if @option{-fdelete-null-pointer-checks} is
8716 @opindex fsplit-paths
8717 Split paths leading to loop backedges. This can improve dead code
8718 elimination and common subexpression elimination. This is enabled by
8719 default at @option{-O2} and above.
8721 @item -fsplit-ivs-in-unroller
8722 @opindex fsplit-ivs-in-unroller
8723 Enables expression of values of induction variables in later iterations
8724 of the unrolled loop using the value in the first iteration. This breaks
8725 long dependency chains, thus improving efficiency of the scheduling passes.
8727 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8728 same effect. However, that is not reliable in cases where the loop body
8729 is more complicated than a single basic block. It also does not work at all
8730 on some architectures due to restrictions in the CSE pass.
8732 This optimization is enabled by default.
8734 @item -fvariable-expansion-in-unroller
8735 @opindex fvariable-expansion-in-unroller
8736 With this option, the compiler creates multiple copies of some
8737 local variables when unrolling a loop, which can result in superior code.
8739 @item -fpartial-inlining
8740 @opindex fpartial-inlining
8741 Inline parts of functions. This option has any effect only
8742 when inlining itself is turned on by the @option{-finline-functions}
8743 or @option{-finline-small-functions} options.
8745 Enabled at level @option{-O2}.
8747 @item -fpredictive-commoning
8748 @opindex fpredictive-commoning
8749 Perform predictive commoning optimization, i.e., reusing computations
8750 (especially memory loads and stores) performed in previous
8751 iterations of loops.
8753 This option is enabled at level @option{-O3}.
8755 @item -fprefetch-loop-arrays
8756 @opindex fprefetch-loop-arrays
8757 If supported by the target machine, generate instructions to prefetch
8758 memory to improve the performance of loops that access large arrays.
8760 This option may generate better or worse code; results are highly
8761 dependent on the structure of loops within the source code.
8763 Disabled at level @option{-Os}.
8765 @item -fno-printf-return-value
8766 @opindex fno-printf-return-value
8767 Do not substitute constants for known return value of formatted output
8768 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
8769 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
8770 transformation allows GCC to optimize or even eliminate branches based
8771 on the known return value of these functions called with arguments that
8772 are either constant, or whose values are known to be in a range that
8773 makes determining the exact return value possible. For example, when
8774 @option{-fprintf-return-value} is in effect, both the branch and the
8775 body of the @code{if} statement (but not the call to @code{snprint})
8776 can be optimized away when @code{i} is a 32-bit or smaller integer
8777 because the return value is guaranteed to be at most 8.
8781 if (snprintf (buf, "%08x", i) >= sizeof buf)
8785 The @option{-fprintf-return-value} option relies on other optimizations
8786 and yields best results with @option{-O2}. It works in tandem with the
8787 @option{-Wformat-overflow} and @option{-Wformat-truncation} options.
8788 The @option{-fprintf-return-value} option is enabled by default.
8791 @itemx -fno-peephole2
8792 @opindex fno-peephole
8793 @opindex fno-peephole2
8794 Disable any machine-specific peephole optimizations. The difference
8795 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8796 are implemented in the compiler; some targets use one, some use the
8797 other, a few use both.
8799 @option{-fpeephole} is enabled by default.
8800 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8802 @item -fno-guess-branch-probability
8803 @opindex fno-guess-branch-probability
8804 Do not guess branch probabilities using heuristics.
8806 GCC uses heuristics to guess branch probabilities if they are
8807 not provided by profiling feedback (@option{-fprofile-arcs}). These
8808 heuristics are based on the control flow graph. If some branch probabilities
8809 are specified by @code{__builtin_expect}, then the heuristics are
8810 used to guess branch probabilities for the rest of the control flow graph,
8811 taking the @code{__builtin_expect} info into account. The interactions
8812 between the heuristics and @code{__builtin_expect} can be complex, and in
8813 some cases, it may be useful to disable the heuristics so that the effects
8814 of @code{__builtin_expect} are easier to understand.
8816 The default is @option{-fguess-branch-probability} at levels
8817 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8819 @item -freorder-blocks
8820 @opindex freorder-blocks
8821 Reorder basic blocks in the compiled function in order to reduce number of
8822 taken branches and improve code locality.
8824 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8826 @item -freorder-blocks-algorithm=@var{algorithm}
8827 @opindex freorder-blocks-algorithm
8828 Use the specified algorithm for basic block reordering. The
8829 @var{algorithm} argument can be @samp{simple}, which does not increase
8830 code size (except sometimes due to secondary effects like alignment),
8831 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8832 put all often executed code together, minimizing the number of branches
8833 executed by making extra copies of code.
8835 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8836 @samp{stc} at levels @option{-O2}, @option{-O3}.
8838 @item -freorder-blocks-and-partition
8839 @opindex freorder-blocks-and-partition
8840 In addition to reordering basic blocks in the compiled function, in order
8841 to reduce number of taken branches, partitions hot and cold basic blocks
8842 into separate sections of the assembly and @file{.o} files, to improve
8843 paging and cache locality performance.
8845 This optimization is automatically turned off in the presence of
8846 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
8847 section attribute and on any architecture that does not support named
8848 sections. When @option{-fsplit-stack} is used this option is not
8849 enabled by default (to avoid linker errors), but may be enabled
8850 explicitly (if using a working linker).
8852 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8854 @item -freorder-functions
8855 @opindex freorder-functions
8856 Reorder functions in the object file in order to
8857 improve code locality. This is implemented by using special
8858 subsections @code{.text.hot} for most frequently executed functions and
8859 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8860 the linker so object file format must support named sections and linker must
8861 place them in a reasonable way.
8863 Also profile feedback must be available to make this option effective. See
8864 @option{-fprofile-arcs} for details.
8866 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8868 @item -fstrict-aliasing
8869 @opindex fstrict-aliasing
8870 Allow the compiler to assume the strictest aliasing rules applicable to
8871 the language being compiled. For C (and C++), this activates
8872 optimizations based on the type of expressions. In particular, an
8873 object of one type is assumed never to reside at the same address as an
8874 object of a different type, unless the types are almost the same. For
8875 example, an @code{unsigned int} can alias an @code{int}, but not a
8876 @code{void*} or a @code{double}. A character type may alias any other
8879 @anchor{Type-punning}Pay special attention to code like this:
8892 The practice of reading from a different union member than the one most
8893 recently written to (called ``type-punning'') is common. Even with
8894 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8895 is accessed through the union type. So, the code above works as
8896 expected. @xref{Structures unions enumerations and bit-fields
8897 implementation}. However, this code might not:
8908 Similarly, access by taking the address, casting the resulting pointer
8909 and dereferencing the result has undefined behavior, even if the cast
8910 uses a union type, e.g.:
8914 return ((union a_union *) &d)->i;
8918 The @option{-fstrict-aliasing} option is enabled at levels
8919 @option{-O2}, @option{-O3}, @option{-Os}.
8921 @item -falign-functions
8922 @itemx -falign-functions=@var{n}
8923 @opindex falign-functions
8924 Align the start of functions to the next power-of-two greater than
8925 @var{n}, skipping up to @var{n} bytes. For instance,
8926 @option{-falign-functions=32} aligns functions to the next 32-byte
8927 boundary, but @option{-falign-functions=24} aligns to the next
8928 32-byte boundary only if this can be done by skipping 23 bytes or less.
8930 @option{-fno-align-functions} and @option{-falign-functions=1} are
8931 equivalent and mean that functions are not aligned.
8933 Some assemblers only support this flag when @var{n} is a power of two;
8934 in that case, it is rounded up.
8936 If @var{n} is not specified or is zero, use a machine-dependent default.
8938 Enabled at levels @option{-O2}, @option{-O3}.
8940 @item -flimit-function-alignment
8941 If this option is enabled, the compiler tries to avoid unnecessarily
8942 overaligning functions. It attempts to instruct the assembler to align
8943 by the amount specified by @option{-falign-functions}, but not to
8944 skip more bytes than the size of the function.
8946 @item -falign-labels
8947 @itemx -falign-labels=@var{n}
8948 @opindex falign-labels
8949 Align all branch targets to a power-of-two boundary, skipping up to
8950 @var{n} bytes like @option{-falign-functions}. This option can easily
8951 make code slower, because it must insert dummy operations for when the
8952 branch target is reached in the usual flow of the code.
8954 @option{-fno-align-labels} and @option{-falign-labels=1} are
8955 equivalent and mean that labels are not aligned.
8957 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8958 are greater than this value, then their values are used instead.
8960 If @var{n} is not specified or is zero, use a machine-dependent default
8961 which is very likely to be @samp{1}, meaning no alignment.
8963 Enabled at levels @option{-O2}, @option{-O3}.
8966 @itemx -falign-loops=@var{n}
8967 @opindex falign-loops
8968 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8969 like @option{-falign-functions}. If the loops are
8970 executed many times, this makes up for any execution of the dummy
8973 @option{-fno-align-loops} and @option{-falign-loops=1} are
8974 equivalent and mean that loops are not aligned.
8976 If @var{n} is not specified or is zero, use a machine-dependent default.
8978 Enabled at levels @option{-O2}, @option{-O3}.
8981 @itemx -falign-jumps=@var{n}
8982 @opindex falign-jumps
8983 Align branch targets to a power-of-two boundary, for branch targets
8984 where the targets can only be reached by jumping, skipping up to @var{n}
8985 bytes like @option{-falign-functions}. In this case, no dummy operations
8988 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8989 equivalent and mean that loops are not aligned.
8991 If @var{n} is not specified or is zero, use a machine-dependent default.
8993 Enabled at levels @option{-O2}, @option{-O3}.
8995 @item -funit-at-a-time
8996 @opindex funit-at-a-time
8997 This option is left for compatibility reasons. @option{-funit-at-a-time}
8998 has no effect, while @option{-fno-unit-at-a-time} implies
8999 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9003 @item -fno-toplevel-reorder
9004 @opindex fno-toplevel-reorder
9005 Do not reorder top-level functions, variables, and @code{asm}
9006 statements. Output them in the same order that they appear in the
9007 input file. When this option is used, unreferenced static variables
9008 are not removed. This option is intended to support existing code
9009 that relies on a particular ordering. For new code, it is better to
9010 use attributes when possible.
9012 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9013 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9018 Constructs webs as commonly used for register allocation purposes and assign
9019 each web individual pseudo register. This allows the register allocation pass
9020 to operate on pseudos directly, but also strengthens several other optimization
9021 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9022 however, make debugging impossible, since variables no longer stay in a
9025 Enabled by default with @option{-funroll-loops}.
9027 @item -fwhole-program
9028 @opindex fwhole-program
9029 Assume that the current compilation unit represents the whole program being
9030 compiled. All public functions and variables with the exception of @code{main}
9031 and those merged by attribute @code{externally_visible} become static functions
9032 and in effect are optimized more aggressively by interprocedural optimizers.
9034 This option should not be used in combination with @option{-flto}.
9035 Instead relying on a linker plugin should provide safer and more precise
9038 @item -flto[=@var{n}]
9040 This option runs the standard link-time optimizer. When invoked
9041 with source code, it generates GIMPLE (one of GCC's internal
9042 representations) and writes it to special ELF sections in the object
9043 file. When the object files are linked together, all the function
9044 bodies are read from these ELF sections and instantiated as if they
9045 had been part of the same translation unit.
9047 To use the link-time optimizer, @option{-flto} and optimization
9048 options should be specified at compile time and during the final link.
9049 It is recommended that you compile all the files participating in the
9050 same link with the same options and also specify those options at
9055 gcc -c -O2 -flto foo.c
9056 gcc -c -O2 -flto bar.c
9057 gcc -o myprog -flto -O2 foo.o bar.o
9060 The first two invocations to GCC save a bytecode representation
9061 of GIMPLE into special ELF sections inside @file{foo.o} and
9062 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9063 @file{foo.o} and @file{bar.o}, merges the two files into a single
9064 internal image, and compiles the result as usual. Since both
9065 @file{foo.o} and @file{bar.o} are merged into a single image, this
9066 causes all the interprocedural analyses and optimizations in GCC to
9067 work across the two files as if they were a single one. This means,
9068 for example, that the inliner is able to inline functions in
9069 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9071 Another (simpler) way to enable link-time optimization is:
9074 gcc -o myprog -flto -O2 foo.c bar.c
9077 The above generates bytecode for @file{foo.c} and @file{bar.c},
9078 merges them together into a single GIMPLE representation and optimizes
9079 them as usual to produce @file{myprog}.
9081 The only important thing to keep in mind is that to enable link-time
9082 optimizations you need to use the GCC driver to perform the link step.
9083 GCC then automatically performs link-time optimization if any of the
9084 objects involved were compiled with the @option{-flto} command-line option.
9086 should specify the optimization options to be used for link-time
9087 optimization though GCC tries to be clever at guessing an
9088 optimization level to use from the options used at compile time
9089 if you fail to specify one at link time. You can always override
9090 the automatic decision to do link-time optimization
9091 by passing @option{-fno-lto} to the link command.
9093 To make whole program optimization effective, it is necessary to make
9094 certain whole program assumptions. The compiler needs to know
9095 what functions and variables can be accessed by libraries and runtime
9096 outside of the link-time optimized unit. When supported by the linker,
9097 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9098 to the compiler about used and externally visible symbols. When
9099 the linker plugin is not available, @option{-fwhole-program} should be
9100 used to allow the compiler to make these assumptions, which leads
9101 to more aggressive optimization decisions.
9103 When @option{-fuse-linker-plugin} is not enabled, when a file is
9104 compiled with @option{-flto}, the generated object file is larger than
9105 a regular object file because it contains GIMPLE bytecodes and the usual
9106 final code (see @option{-ffat-lto-objects}. This means that
9107 object files with LTO information can be linked as normal object
9108 files; if @option{-fno-lto} is passed to the linker, no
9109 interprocedural optimizations are applied. Note that when
9110 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
9111 but you cannot perform a regular, non-LTO link on them.
9113 Additionally, the optimization flags used to compile individual files
9114 are not necessarily related to those used at link time. For instance,
9117 gcc -c -O0 -ffat-lto-objects -flto foo.c
9118 gcc -c -O0 -ffat-lto-objects -flto bar.c
9119 gcc -o myprog -O3 foo.o bar.o
9122 This produces individual object files with unoptimized assembler
9123 code, but the resulting binary @file{myprog} is optimized at
9124 @option{-O3}. If, instead, the final binary is generated with
9125 @option{-fno-lto}, then @file{myprog} is not optimized.
9127 When producing the final binary, GCC only
9128 applies link-time optimizations to those files that contain bytecode.
9129 Therefore, you can mix and match object files and libraries with
9130 GIMPLE bytecodes and final object code. GCC automatically selects
9131 which files to optimize in LTO mode and which files to link without
9134 There are some code generation flags preserved by GCC when
9135 generating bytecodes, as they need to be used during the final link
9136 stage. Generally options specified at link time override those
9137 specified at compile time.
9139 If you do not specify an optimization level option @option{-O} at
9140 link time, then GCC uses the highest optimization level
9141 used when compiling the object files.
9143 Currently, the following options and their settings are taken from
9144 the first object file that explicitly specifies them:
9145 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9146 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9147 and all the @option{-m} target flags.
9149 Certain ABI-changing flags are required to match in all compilation units,
9150 and trying to override this at link time with a conflicting value
9151 is ignored. This includes options such as @option{-freg-struct-return}
9152 and @option{-fpcc-struct-return}.
9154 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9155 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9156 are passed through to the link stage and merged conservatively for
9157 conflicting translation units. Specifically
9158 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9159 precedence; and for example @option{-ffp-contract=off} takes precedence
9160 over @option{-ffp-contract=fast}. You can override them at link time.
9162 If LTO encounters objects with C linkage declared with incompatible
9163 types in separate translation units to be linked together (undefined
9164 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9165 issued. The behavior is still undefined at run time. Similar
9166 diagnostics may be raised for other languages.
9168 Another feature of LTO is that it is possible to apply interprocedural
9169 optimizations on files written in different languages:
9174 gfortran -c -flto baz.f90
9175 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9178 Notice that the final link is done with @command{g++} to get the C++
9179 runtime libraries and @option{-lgfortran} is added to get the Fortran
9180 runtime libraries. In general, when mixing languages in LTO mode, you
9181 should use the same link command options as when mixing languages in a
9182 regular (non-LTO) compilation.
9184 If object files containing GIMPLE bytecode are stored in a library archive, say
9185 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9186 are using a linker with plugin support. To create static libraries suitable
9187 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9188 and @command{ranlib};
9189 to show the symbols of object files with GIMPLE bytecode, use
9190 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9191 and @command{nm} have been compiled with plugin support. At link time, use the the
9192 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9193 the LTO optimization process:
9196 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9199 With the linker plugin enabled, the linker extracts the needed
9200 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9201 to make them part of the aggregated GIMPLE image to be optimized.
9203 If you are not using a linker with plugin support and/or do not
9204 enable the linker plugin, then the objects inside @file{libfoo.a}
9205 are extracted and linked as usual, but they do not participate
9206 in the LTO optimization process. In order to make a static library suitable
9207 for both LTO optimization and usual linkage, compile its object files with
9208 @option{-flto} @option{-ffat-lto-objects}.
9210 Link-time optimizations do not require the presence of the whole program to
9211 operate. If the program does not require any symbols to be exported, it is
9212 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9213 the interprocedural optimizers to use more aggressive assumptions which may
9214 lead to improved optimization opportunities.
9215 Use of @option{-fwhole-program} is not needed when linker plugin is
9216 active (see @option{-fuse-linker-plugin}).
9218 The current implementation of LTO makes no
9219 attempt to generate bytecode that is portable between different
9220 types of hosts. The bytecode files are versioned and there is a
9221 strict version check, so bytecode files generated in one version of
9222 GCC do not work with an older or newer version of GCC.
9224 Link-time optimization does not work well with generation of debugging
9225 information. Combining @option{-flto} with
9226 @option{-g} is currently experimental and expected to produce unexpected
9229 If you specify the optional @var{n}, the optimization and code
9230 generation done at link time is executed in parallel using @var{n}
9231 parallel jobs by utilizing an installed @command{make} program. The
9232 environment variable @env{MAKE} may be used to override the program
9233 used. The default value for @var{n} is 1.
9235 You can also specify @option{-flto=jobserver} to use GNU make's
9236 job server mode to determine the number of parallel jobs. This
9237 is useful when the Makefile calling GCC is already executing in parallel.
9238 You must prepend a @samp{+} to the command recipe in the parent Makefile
9239 for this to work. This option likely only works if @env{MAKE} is
9242 @item -flto-partition=@var{alg}
9243 @opindex flto-partition
9244 Specify the partitioning algorithm used by the link-time optimizer.
9245 The value is either @samp{1to1} to specify a partitioning mirroring
9246 the original source files or @samp{balanced} to specify partitioning
9247 into equally sized chunks (whenever possible) or @samp{max} to create
9248 new partition for every symbol where possible. Specifying @samp{none}
9249 as an algorithm disables partitioning and streaming completely.
9250 The default value is @samp{balanced}. While @samp{1to1} can be used
9251 as an workaround for various code ordering issues, the @samp{max}
9252 partitioning is intended for internal testing only.
9253 The value @samp{one} specifies that exactly one partition should be
9254 used while the value @samp{none} bypasses partitioning and executes
9255 the link-time optimization step directly from the WPA phase.
9257 @item -flto-odr-type-merging
9258 @opindex flto-odr-type-merging
9259 Enable streaming of mangled types names of C++ types and their unification
9260 at link time. This increases size of LTO object files, but enables
9261 diagnostics about One Definition Rule violations.
9263 @item -flto-compression-level=@var{n}
9264 @opindex flto-compression-level
9265 This option specifies the level of compression used for intermediate
9266 language written to LTO object files, and is only meaningful in
9267 conjunction with LTO mode (@option{-flto}). Valid
9268 values are 0 (no compression) to 9 (maximum compression). Values
9269 outside this range are clamped to either 0 or 9. If the option is not
9270 given, a default balanced compression setting is used.
9272 @item -fuse-linker-plugin
9273 @opindex fuse-linker-plugin
9274 Enables the use of a linker plugin during link-time optimization. This
9275 option relies on plugin support in the linker, which is available in gold
9276 or in GNU ld 2.21 or newer.
9278 This option enables the extraction of object files with GIMPLE bytecode out
9279 of library archives. This improves the quality of optimization by exposing
9280 more code to the link-time optimizer. This information specifies what
9281 symbols can be accessed externally (by non-LTO object or during dynamic
9282 linking). Resulting code quality improvements on binaries (and shared
9283 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9284 See @option{-flto} for a description of the effect of this flag and how to
9287 This option is enabled by default when LTO support in GCC is enabled
9288 and GCC was configured for use with
9289 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9291 @item -ffat-lto-objects
9292 @opindex ffat-lto-objects
9293 Fat LTO objects are object files that contain both the intermediate language
9294 and the object code. This makes them usable for both LTO linking and normal
9295 linking. This option is effective only when compiling with @option{-flto}
9296 and is ignored at link time.
9298 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9299 requires the complete toolchain to be aware of LTO. It requires a linker with
9300 linker plugin support for basic functionality. Additionally,
9301 @command{nm}, @command{ar} and @command{ranlib}
9302 need to support linker plugins to allow a full-featured build environment
9303 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9304 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9305 to these tools. With non fat LTO makefiles need to be modified to use them.
9307 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9310 @item -fcompare-elim
9311 @opindex fcompare-elim
9312 After register allocation and post-register allocation instruction splitting,
9313 identify arithmetic instructions that compute processor flags similar to a
9314 comparison operation based on that arithmetic. If possible, eliminate the
9315 explicit comparison operation.
9317 This pass only applies to certain targets that cannot explicitly represent
9318 the comparison operation before register allocation is complete.
9320 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9322 @item -fcprop-registers
9323 @opindex fcprop-registers
9324 After register allocation and post-register allocation instruction splitting,
9325 perform a copy-propagation pass to try to reduce scheduling dependencies
9326 and occasionally eliminate the copy.
9328 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9330 @item -fprofile-correction
9331 @opindex fprofile-correction
9332 Profiles collected using an instrumented binary for multi-threaded programs may
9333 be inconsistent due to missed counter updates. When this option is specified,
9334 GCC uses heuristics to correct or smooth out such inconsistencies. By
9335 default, GCC emits an error message when an inconsistent profile is detected.
9338 @itemx -fprofile-use=@var{path}
9339 @opindex fprofile-use
9340 Enable profile feedback-directed optimizations,
9341 and the following optimizations
9342 which are generally profitable only with profile feedback available:
9343 @option{-fbranch-probabilities}, @option{-fvpt},
9344 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9345 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9347 Before you can use this option, you must first generate profiling information.
9348 @xref{Instrumentation Options}, for information about the
9349 @option{-fprofile-generate} option.
9351 By default, GCC emits an error message if the feedback profiles do not
9352 match the source code. This error can be turned into a warning by using
9353 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9356 If @var{path} is specified, GCC looks at the @var{path} to find
9357 the profile feedback data files. See @option{-fprofile-dir}.
9359 @item -fauto-profile
9360 @itemx -fauto-profile=@var{path}
9361 @opindex fauto-profile
9362 Enable sampling-based feedback-directed optimizations,
9363 and the following optimizations
9364 which are generally profitable only with profile feedback available:
9365 @option{-fbranch-probabilities}, @option{-fvpt},
9366 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9367 @option{-ftree-vectorize},
9368 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9369 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9370 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9372 @var{path} is the name of a file containing AutoFDO profile information.
9373 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9375 Producing an AutoFDO profile data file requires running your program
9376 with the @command{perf} utility on a supported GNU/Linux target system.
9377 For more information, see @uref{https://perf.wiki.kernel.org/}.
9381 perf record -e br_inst_retired:near_taken -b -o perf.data \
9385 Then use the @command{create_gcov} tool to convert the raw profile data
9386 to a format that can be used by GCC.@ You must also supply the
9387 unstripped binary for your program to this tool.
9388 See @uref{https://github.com/google/autofdo}.
9392 create_gcov --binary=your_program.unstripped --profile=perf.data \
9397 The following options control compiler behavior regarding floating-point
9398 arithmetic. These options trade off between speed and
9399 correctness. All must be specifically enabled.
9403 @opindex ffloat-store
9404 Do not store floating-point variables in registers, and inhibit other
9405 options that might change whether a floating-point value is taken from a
9408 @cindex floating-point precision
9409 This option prevents undesirable excess precision on machines such as
9410 the 68000 where the floating registers (of the 68881) keep more
9411 precision than a @code{double} is supposed to have. Similarly for the
9412 x86 architecture. For most programs, the excess precision does only
9413 good, but a few programs rely on the precise definition of IEEE floating
9414 point. Use @option{-ffloat-store} for such programs, after modifying
9415 them to store all pertinent intermediate computations into variables.
9417 @item -fexcess-precision=@var{style}
9418 @opindex fexcess-precision
9419 This option allows further control over excess precision on machines
9420 where floating-point operations occur in a format with more precision or
9421 range than the IEEE standard and interchange floating-point types. By
9422 default, @option{-fexcess-precision=fast} is in effect; this means that
9423 operations may be carried out in a wider precision than the types specified
9424 in the source if that would result in faster code, and it is unpredictable
9425 when rounding to the types specified in the source code takes place.
9426 When compiling C, if @option{-fexcess-precision=standard} is specified then
9427 excess precision follows the rules specified in ISO C99; in particular,
9428 both casts and assignments cause values to be rounded to their
9429 semantic types (whereas @option{-ffloat-store} only affects
9430 assignments). This option is enabled by default for C if a strict
9431 conformance option such as @option{-std=c99} is used.
9432 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
9433 regardless of whether a strict conformance option is used.
9436 @option{-fexcess-precision=standard} is not implemented for languages
9437 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
9438 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9439 semantics apply without excess precision, and in the latter, rounding
9444 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9445 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9446 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
9447 @option{-fexcess-precision=fast}.
9449 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9451 This option is not turned on by any @option{-O} option besides
9452 @option{-Ofast} since it can result in incorrect output for programs
9453 that depend on an exact implementation of IEEE or ISO rules/specifications
9454 for math functions. It may, however, yield faster code for programs
9455 that do not require the guarantees of these specifications.
9457 @item -fno-math-errno
9458 @opindex fno-math-errno
9459 Do not set @code{errno} after calling math functions that are executed
9460 with a single instruction, e.g., @code{sqrt}. A program that relies on
9461 IEEE exceptions for math error handling may want to use this flag
9462 for speed while maintaining IEEE arithmetic compatibility.
9464 This option is not turned on by any @option{-O} option since
9465 it can result in incorrect output for programs that depend on
9466 an exact implementation of IEEE or ISO rules/specifications for
9467 math functions. It may, however, yield faster code for programs
9468 that do not require the guarantees of these specifications.
9470 The default is @option{-fmath-errno}.
9472 On Darwin systems, the math library never sets @code{errno}. There is
9473 therefore no reason for the compiler to consider the possibility that
9474 it might, and @option{-fno-math-errno} is the default.
9476 @item -funsafe-math-optimizations
9477 @opindex funsafe-math-optimizations
9479 Allow optimizations for floating-point arithmetic that (a) assume
9480 that arguments and results are valid and (b) may violate IEEE or
9481 ANSI standards. When used at link time, it may include libraries
9482 or startup files that change the default FPU control word or other
9483 similar optimizations.
9485 This option is not turned on by any @option{-O} option since
9486 it can result in incorrect output for programs that depend on
9487 an exact implementation of IEEE or ISO rules/specifications for
9488 math functions. It may, however, yield faster code for programs
9489 that do not require the guarantees of these specifications.
9490 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9491 @option{-fassociative-math} and @option{-freciprocal-math}.
9493 The default is @option{-fno-unsafe-math-optimizations}.
9495 @item -fassociative-math
9496 @opindex fassociative-math
9498 Allow re-association of operands in series of floating-point operations.
9499 This violates the ISO C and C++ language standard by possibly changing
9500 computation result. NOTE: re-ordering may change the sign of zero as
9501 well as ignore NaNs and inhibit or create underflow or overflow (and
9502 thus cannot be used on code that relies on rounding behavior like
9503 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9504 and thus may not be used when ordered comparisons are required.
9505 This option requires that both @option{-fno-signed-zeros} and
9506 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9507 much sense with @option{-frounding-math}. For Fortran the option
9508 is automatically enabled when both @option{-fno-signed-zeros} and
9509 @option{-fno-trapping-math} are in effect.
9511 The default is @option{-fno-associative-math}.
9513 @item -freciprocal-math
9514 @opindex freciprocal-math
9516 Allow the reciprocal of a value to be used instead of dividing by
9517 the value if this enables optimizations. For example @code{x / y}
9518 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9519 is subject to common subexpression elimination. Note that this loses
9520 precision and increases the number of flops operating on the value.
9522 The default is @option{-fno-reciprocal-math}.
9524 @item -ffinite-math-only
9525 @opindex ffinite-math-only
9526 Allow optimizations for floating-point arithmetic that assume
9527 that arguments and results are not NaNs or +-Infs.
9529 This option is not turned on by any @option{-O} option since
9530 it can result in incorrect output for programs that depend on
9531 an exact implementation of IEEE or ISO rules/specifications for
9532 math functions. It may, however, yield faster code for programs
9533 that do not require the guarantees of these specifications.
9535 The default is @option{-fno-finite-math-only}.
9537 @item -fno-signed-zeros
9538 @opindex fno-signed-zeros
9539 Allow optimizations for floating-point arithmetic that ignore the
9540 signedness of zero. IEEE arithmetic specifies the behavior of
9541 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9542 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9543 This option implies that the sign of a zero result isn't significant.
9545 The default is @option{-fsigned-zeros}.
9547 @item -fno-trapping-math
9548 @opindex fno-trapping-math
9549 Compile code assuming that floating-point operations cannot generate
9550 user-visible traps. These traps include division by zero, overflow,
9551 underflow, inexact result and invalid operation. This option requires
9552 that @option{-fno-signaling-nans} be in effect. Setting this option may
9553 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9555 This option should never be turned on by any @option{-O} option since
9556 it can result in incorrect output for programs that depend on
9557 an exact implementation of IEEE or ISO rules/specifications for
9560 The default is @option{-ftrapping-math}.
9562 @item -frounding-math
9563 @opindex frounding-math
9564 Disable transformations and optimizations that assume default floating-point
9565 rounding behavior. This is round-to-zero for all floating point
9566 to integer conversions, and round-to-nearest for all other arithmetic
9567 truncations. This option should be specified for programs that change
9568 the FP rounding mode dynamically, or that may be executed with a
9569 non-default rounding mode. This option disables constant folding of
9570 floating-point expressions at compile time (which may be affected by
9571 rounding mode) and arithmetic transformations that are unsafe in the
9572 presence of sign-dependent rounding modes.
9574 The default is @option{-fno-rounding-math}.
9576 This option is experimental and does not currently guarantee to
9577 disable all GCC optimizations that are affected by rounding mode.
9578 Future versions of GCC may provide finer control of this setting
9579 using C99's @code{FENV_ACCESS} pragma. This command-line option
9580 will be used to specify the default state for @code{FENV_ACCESS}.
9582 @item -fsignaling-nans
9583 @opindex fsignaling-nans
9584 Compile code assuming that IEEE signaling NaNs may generate user-visible
9585 traps during floating-point operations. Setting this option disables
9586 optimizations that may change the number of exceptions visible with
9587 signaling NaNs. This option implies @option{-ftrapping-math}.
9589 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9592 The default is @option{-fno-signaling-nans}.
9594 This option is experimental and does not currently guarantee to
9595 disable all GCC optimizations that affect signaling NaN behavior.
9597 @item -fno-fp-int-builtin-inexact
9598 @opindex fno-fp-int-builtin-inexact
9599 Do not allow the built-in functions @code{ceil}, @code{floor},
9600 @code{round} and @code{trunc}, and their @code{float} and @code{long
9601 double} variants, to generate code that raises the ``inexact''
9602 floating-point exception for noninteger arguments. ISO C99 and C11
9603 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9604 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9607 The default is @option{-ffp-int-builtin-inexact}, allowing the
9608 exception to be raised. This option does nothing unless
9609 @option{-ftrapping-math} is in effect.
9611 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9612 generate a call to a library function then the ``inexact'' exception
9613 may be raised if the library implementation does not follow TS 18661.
9615 @item -fsingle-precision-constant
9616 @opindex fsingle-precision-constant
9617 Treat floating-point constants as single precision instead of
9618 implicitly converting them to double-precision constants.
9620 @item -fcx-limited-range
9621 @opindex fcx-limited-range
9622 When enabled, this option states that a range reduction step is not
9623 needed when performing complex division. Also, there is no checking
9624 whether the result of a complex multiplication or division is @code{NaN
9625 + I*NaN}, with an attempt to rescue the situation in that case. The
9626 default is @option{-fno-cx-limited-range}, but is enabled by
9627 @option{-ffast-math}.
9629 This option controls the default setting of the ISO C99
9630 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9633 @item -fcx-fortran-rules
9634 @opindex fcx-fortran-rules
9635 Complex multiplication and division follow Fortran rules. Range
9636 reduction is done as part of complex division, but there is no checking
9637 whether the result of a complex multiplication or division is @code{NaN
9638 + I*NaN}, with an attempt to rescue the situation in that case.
9640 The default is @option{-fno-cx-fortran-rules}.
9644 The following options control optimizations that may improve
9645 performance, but are not enabled by any @option{-O} options. This
9646 section includes experimental options that may produce broken code.
9649 @item -fbranch-probabilities
9650 @opindex fbranch-probabilities
9651 After running a program compiled with @option{-fprofile-arcs}
9652 (@pxref{Instrumentation Options}),
9653 you can compile it a second time using
9654 @option{-fbranch-probabilities}, to improve optimizations based on
9655 the number of times each branch was taken. When a program
9656 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9657 counts to a file called @file{@var{sourcename}.gcda} for each source
9658 file. The information in this data file is very dependent on the
9659 structure of the generated code, so you must use the same source code
9660 and the same optimization options for both compilations.
9662 With @option{-fbranch-probabilities}, GCC puts a
9663 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9664 These can be used to improve optimization. Currently, they are only
9665 used in one place: in @file{reorg.c}, instead of guessing which path a
9666 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9667 exactly determine which path is taken more often.
9669 @item -fprofile-values
9670 @opindex fprofile-values
9671 If combined with @option{-fprofile-arcs}, it adds code so that some
9672 data about values of expressions in the program is gathered.
9674 With @option{-fbranch-probabilities}, it reads back the data gathered
9675 from profiling values of expressions for usage in optimizations.
9677 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9679 @item -fprofile-reorder-functions
9680 @opindex fprofile-reorder-functions
9681 Function reordering based on profile instrumentation collects
9682 first time of execution of a function and orders these functions
9685 Enabled with @option{-fprofile-use}.
9689 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9690 to add code to gather information about values of expressions.
9692 With @option{-fbranch-probabilities}, it reads back the data gathered
9693 and actually performs the optimizations based on them.
9694 Currently the optimizations include specialization of division operations
9695 using the knowledge about the value of the denominator.
9697 @item -frename-registers
9698 @opindex frename-registers
9699 Attempt to avoid false dependencies in scheduled code by making use
9700 of registers left over after register allocation. This optimization
9701 most benefits processors with lots of registers. Depending on the
9702 debug information format adopted by the target, however, it can
9703 make debugging impossible, since variables no longer stay in
9704 a ``home register''.
9706 Enabled by default with @option{-funroll-loops}.
9708 @item -fschedule-fusion
9709 @opindex fschedule-fusion
9710 Performs a target dependent pass over the instruction stream to schedule
9711 instructions of same type together because target machine can execute them
9712 more efficiently if they are adjacent to each other in the instruction flow.
9714 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9718 Perform tail duplication to enlarge superblock size. This transformation
9719 simplifies the control flow of the function allowing other optimizations to do
9722 Enabled with @option{-fprofile-use}.
9724 @item -funroll-loops
9725 @opindex funroll-loops
9726 Unroll loops whose number of iterations can be determined at compile time or
9727 upon entry to the loop. @option{-funroll-loops} implies
9728 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9729 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9730 a small constant number of iterations). This option makes code larger, and may
9731 or may not make it run faster.
9733 Enabled with @option{-fprofile-use}.
9735 @item -funroll-all-loops
9736 @opindex funroll-all-loops
9737 Unroll all loops, even if their number of iterations is uncertain when
9738 the loop is entered. This usually makes programs run more slowly.
9739 @option{-funroll-all-loops} implies the same options as
9740 @option{-funroll-loops}.
9743 @opindex fpeel-loops
9744 Peels loops for which there is enough information that they do not
9745 roll much (from profile feedback or static analysis). It also turns on
9746 complete loop peeling (i.e.@: complete removal of loops with small constant
9747 number of iterations).
9749 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9751 @item -fmove-loop-invariants
9752 @opindex fmove-loop-invariants
9753 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9754 at level @option{-O1}
9757 @opindex fsplit-loops
9758 Split a loop into two if it contains a condition that's always true
9759 for one side of the iteration space and false for the other.
9761 @item -funswitch-loops
9762 @opindex funswitch-loops
9763 Move branches with loop invariant conditions out of the loop, with duplicates
9764 of the loop on both branches (modified according to result of the condition).
9766 @item -ffunction-sections
9767 @itemx -fdata-sections
9768 @opindex ffunction-sections
9769 @opindex fdata-sections
9770 Place each function or data item into its own section in the output
9771 file if the target supports arbitrary sections. The name of the
9772 function or the name of the data item determines the section's name
9775 Use these options on systems where the linker can perform optimizations to
9776 improve locality of reference in the instruction space. Most systems using the
9777 ELF object format have linkers with such optimizations. On AIX, the linker
9778 rearranges sections (CSECTs) based on the call graph. The performance impact
9781 Together with a linker garbage collection (linker @option{--gc-sections}
9782 option) these options may lead to smaller statically-linked executables (after
9785 On ELF/DWARF systems these options do not degenerate the quality of the debug
9786 information. There could be issues with other object files/debug info formats.
9788 Only use these options when there are significant benefits from doing so. When
9789 you specify these options, the assembler and linker create larger object and
9790 executable files and are also slower. These options affect code generation.
9791 They prevent optimizations by the compiler and assembler using relative
9792 locations inside a translation unit since the locations are unknown until
9793 link time. An example of such an optimization is relaxing calls to short call
9796 @item -fbranch-target-load-optimize
9797 @opindex fbranch-target-load-optimize
9798 Perform branch target register load optimization before prologue / epilogue
9800 The use of target registers can typically be exposed only during reload,
9801 thus hoisting loads out of loops and doing inter-block scheduling needs
9802 a separate optimization pass.
9804 @item -fbranch-target-load-optimize2
9805 @opindex fbranch-target-load-optimize2
9806 Perform branch target register load optimization after prologue / epilogue
9809 @item -fbtr-bb-exclusive
9810 @opindex fbtr-bb-exclusive
9811 When performing branch target register load optimization, don't reuse
9812 branch target registers within any basic block.
9815 @opindex fstdarg-opt
9816 Optimize the prologue of variadic argument functions with respect to usage of
9819 @item -fsection-anchors
9820 @opindex fsection-anchors
9821 Try to reduce the number of symbolic address calculations by using
9822 shared ``anchor'' symbols to address nearby objects. This transformation
9823 can help to reduce the number of GOT entries and GOT accesses on some
9826 For example, the implementation of the following function @code{foo}:
9830 int foo (void) @{ return a + b + c; @}
9834 usually calculates the addresses of all three variables, but if you
9835 compile it with @option{-fsection-anchors}, it accesses the variables
9836 from a common anchor point instead. The effect is similar to the
9837 following pseudocode (which isn't valid C):
9842 register int *xr = &x;
9843 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9847 Not all targets support this option.
9849 @item --param @var{name}=@var{value}
9851 In some places, GCC uses various constants to control the amount of
9852 optimization that is done. For example, GCC does not inline functions
9853 that contain more than a certain number of instructions. You can
9854 control some of these constants on the command line using the
9855 @option{--param} option.
9857 The names of specific parameters, and the meaning of the values, are
9858 tied to the internals of the compiler, and are subject to change
9859 without notice in future releases.
9861 In each case, the @var{value} is an integer. The allowable choices for
9865 @item predictable-branch-outcome
9866 When branch is predicted to be taken with probability lower than this threshold
9867 (in percent), then it is considered well predictable. The default is 10.
9869 @item max-rtl-if-conversion-insns
9870 RTL if-conversion tries to remove conditional branches around a block and
9871 replace them with conditionally executed instructions. This parameter
9872 gives the maximum number of instructions in a block which should be
9873 considered for if-conversion. The default is 10, though the compiler will
9874 also use other heuristics to decide whether if-conversion is likely to be
9877 @item max-rtl-if-conversion-predictable-cost
9878 @item max-rtl-if-conversion-unpredictable-cost
9879 RTL if-conversion will try to remove conditional branches around a block
9880 and replace them with conditionally executed instructions. These parameters
9881 give the maximum permissible cost for the sequence that would be generated
9882 by if-conversion depending on whether the branch is statically determined
9883 to be predictable or not. The units for this parameter are the same as
9884 those for the GCC internal seq_cost metric. The compiler will try to
9885 provide a reasonable default for this parameter using the BRANCH_COST
9888 @item max-crossjump-edges
9889 The maximum number of incoming edges to consider for cross-jumping.
9890 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9891 the number of edges incoming to each block. Increasing values mean
9892 more aggressive optimization, making the compilation time increase with
9893 probably small improvement in executable size.
9895 @item min-crossjump-insns
9896 The minimum number of instructions that must be matched at the end
9897 of two blocks before cross-jumping is performed on them. This
9898 value is ignored in the case where all instructions in the block being
9899 cross-jumped from are matched. The default value is 5.
9901 @item max-grow-copy-bb-insns
9902 The maximum code size expansion factor when copying basic blocks
9903 instead of jumping. The expansion is relative to a jump instruction.
9904 The default value is 8.
9906 @item max-goto-duplication-insns
9907 The maximum number of instructions to duplicate to a block that jumps
9908 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9909 passes, GCC factors computed gotos early in the compilation process,
9910 and unfactors them as late as possible. Only computed jumps at the
9911 end of a basic blocks with no more than max-goto-duplication-insns are
9912 unfactored. The default value is 8.
9914 @item max-delay-slot-insn-search
9915 The maximum number of instructions to consider when looking for an
9916 instruction to fill a delay slot. If more than this arbitrary number of
9917 instructions are searched, the time savings from filling the delay slot
9918 are minimal, so stop searching. Increasing values mean more
9919 aggressive optimization, making the compilation time increase with probably
9920 small improvement in execution time.
9922 @item max-delay-slot-live-search
9923 When trying to fill delay slots, the maximum number of instructions to
9924 consider when searching for a block with valid live register
9925 information. Increasing this arbitrarily chosen value means more
9926 aggressive optimization, increasing the compilation time. This parameter
9927 should be removed when the delay slot code is rewritten to maintain the
9930 @item max-gcse-memory
9931 The approximate maximum amount of memory that can be allocated in
9932 order to perform the global common subexpression elimination
9933 optimization. If more memory than specified is required, the
9934 optimization is not done.
9936 @item max-gcse-insertion-ratio
9937 If the ratio of expression insertions to deletions is larger than this value
9938 for any expression, then RTL PRE inserts or removes the expression and thus
9939 leaves partially redundant computations in the instruction stream. The default value is 20.
9941 @item max-pending-list-length
9942 The maximum number of pending dependencies scheduling allows
9943 before flushing the current state and starting over. Large functions
9944 with few branches or calls can create excessively large lists which
9945 needlessly consume memory and resources.
9947 @item max-modulo-backtrack-attempts
9948 The maximum number of backtrack attempts the scheduler should make
9949 when modulo scheduling a loop. Larger values can exponentially increase
9952 @item max-inline-insns-single
9953 Several parameters control the tree inliner used in GCC@.
9954 This number sets the maximum number of instructions (counted in GCC's
9955 internal representation) in a single function that the tree inliner
9956 considers for inlining. This only affects functions declared
9957 inline and methods implemented in a class declaration (C++).
9958 The default value is 400.
9960 @item max-inline-insns-auto
9961 When you use @option{-finline-functions} (included in @option{-O3}),
9962 a lot of functions that would otherwise not be considered for inlining
9963 by the compiler are investigated. To those functions, a different
9964 (more restrictive) limit compared to functions declared inline can
9966 The default value is 40.
9968 @item inline-min-speedup
9969 When estimated performance improvement of caller + callee runtime exceeds this
9970 threshold (in percent), the function can be inlined regardless of the limit on
9971 @option{--param max-inline-insns-single} and @option{--param
9972 max-inline-insns-auto}.
9974 @item large-function-insns
9975 The limit specifying really large functions. For functions larger than this
9976 limit after inlining, inlining is constrained by
9977 @option{--param large-function-growth}. This parameter is useful primarily
9978 to avoid extreme compilation time caused by non-linear algorithms used by the
9980 The default value is 2700.
9982 @item large-function-growth
9983 Specifies maximal growth of large function caused by inlining in percents.
9984 The default value is 100 which limits large function growth to 2.0 times
9987 @item large-unit-insns
9988 The limit specifying large translation unit. Growth caused by inlining of
9989 units larger than this limit is limited by @option{--param inline-unit-growth}.
9990 For small units this might be too tight.
9991 For example, consider a unit consisting of function A
9992 that is inline and B that just calls A three times. If B is small relative to
9993 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9994 large units consisting of small inlineable functions, however, the overall unit
9995 growth limit is needed to avoid exponential explosion of code size. Thus for
9996 smaller units, the size is increased to @option{--param large-unit-insns}
9997 before applying @option{--param inline-unit-growth}. The default is 10000.
9999 @item inline-unit-growth
10000 Specifies maximal overall growth of the compilation unit caused by inlining.
10001 The default value is 20 which limits unit growth to 1.2 times the original
10002 size. Cold functions (either marked cold via an attribute or by profile
10003 feedback) are not accounted into the unit size.
10005 @item ipcp-unit-growth
10006 Specifies maximal overall growth of the compilation unit caused by
10007 interprocedural constant propagation. The default value is 10 which limits
10008 unit growth to 1.1 times the original size.
10010 @item large-stack-frame
10011 The limit specifying large stack frames. While inlining the algorithm is trying
10012 to not grow past this limit too much. The default value is 256 bytes.
10014 @item large-stack-frame-growth
10015 Specifies maximal growth of large stack frames caused by inlining in percents.
10016 The default value is 1000 which limits large stack frame growth to 11 times
10019 @item max-inline-insns-recursive
10020 @itemx max-inline-insns-recursive-auto
10021 Specifies the maximum number of instructions an out-of-line copy of a
10022 self-recursive inline
10023 function can grow into by performing recursive inlining.
10025 @option{--param max-inline-insns-recursive} applies to functions
10027 For functions not declared inline, recursive inlining
10028 happens only when @option{-finline-functions} (included in @option{-O3}) is
10029 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10030 default value is 450.
10032 @item max-inline-recursive-depth
10033 @itemx max-inline-recursive-depth-auto
10034 Specifies the maximum recursion depth used for recursive inlining.
10036 @option{--param max-inline-recursive-depth} applies to functions
10037 declared inline. For functions not declared inline, recursive inlining
10038 happens only when @option{-finline-functions} (included in @option{-O3}) is
10039 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10040 default value is 8.
10042 @item min-inline-recursive-probability
10043 Recursive inlining is profitable only for function having deep recursion
10044 in average and can hurt for function having little recursion depth by
10045 increasing the prologue size or complexity of function body to other
10048 When profile feedback is available (see @option{-fprofile-generate}) the actual
10049 recursion depth can be guessed from the probability that function recurses
10050 via a given call expression. This parameter limits inlining only to call
10051 expressions whose probability exceeds the given threshold (in percents).
10052 The default value is 10.
10054 @item early-inlining-insns
10055 Specify growth that the early inliner can make. In effect it increases
10056 the amount of inlining for code having a large abstraction penalty.
10057 The default value is 14.
10059 @item max-early-inliner-iterations
10060 Limit of iterations of the early inliner. This basically bounds
10061 the number of nested indirect calls the early inliner can resolve.
10062 Deeper chains are still handled by late inlining.
10064 @item comdat-sharing-probability
10065 Probability (in percent) that C++ inline function with comdat visibility
10066 are shared across multiple compilation units. The default value is 20.
10068 @item profile-func-internal-id
10069 A parameter to control whether to use function internal id in profile
10070 database lookup. If the value is 0, the compiler uses an id that
10071 is based on function assembler name and filename, which makes old profile
10072 data more tolerant to source changes such as function reordering etc.
10073 The default value is 0.
10075 @item min-vect-loop-bound
10076 The minimum number of iterations under which loops are not vectorized
10077 when @option{-ftree-vectorize} is used. The number of iterations after
10078 vectorization needs to be greater than the value specified by this option
10079 to allow vectorization. The default value is 0.
10081 @item gcse-cost-distance-ratio
10082 Scaling factor in calculation of maximum distance an expression
10083 can be moved by GCSE optimizations. This is currently supported only in the
10084 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10085 is with simple expressions, i.e., the expressions that have cost
10086 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10087 hoisting of simple expressions. The default value is 10.
10089 @item gcse-unrestricted-cost
10090 Cost, roughly measured as the cost of a single typical machine
10091 instruction, at which GCSE optimizations do not constrain
10092 the distance an expression can travel. This is currently
10093 supported only in the code hoisting pass. The lesser the cost,
10094 the more aggressive code hoisting is. Specifying 0
10095 allows all expressions to travel unrestricted distances.
10096 The default value is 3.
10098 @item max-hoist-depth
10099 The depth of search in the dominator tree for expressions to hoist.
10100 This is used to avoid quadratic behavior in hoisting algorithm.
10101 The value of 0 does not limit on the search, but may slow down compilation
10102 of huge functions. The default value is 30.
10104 @item max-tail-merge-comparisons
10105 The maximum amount of similar bbs to compare a bb with. This is used to
10106 avoid quadratic behavior in tree tail merging. The default value is 10.
10108 @item max-tail-merge-iterations
10109 The maximum amount of iterations of the pass over the function. This is used to
10110 limit compilation time in tree tail merging. The default value is 2.
10112 @item store-merging-allow-unaligned
10113 Allow the store merging pass to introduce unaligned stores if it is legal to
10114 do so. The default value is 1.
10116 @item max-stores-to-merge
10117 The maximum number of stores to attempt to merge into wider stores in the store
10118 merging pass. The minimum value is 2 and the default is 64.
10120 @item max-unrolled-insns
10121 The maximum number of instructions that a loop may have to be unrolled.
10122 If a loop is unrolled, this parameter also determines how many times
10123 the loop code is unrolled.
10125 @item max-average-unrolled-insns
10126 The maximum number of instructions biased by probabilities of their execution
10127 that a loop may have to be unrolled. If a loop is unrolled,
10128 this parameter also determines how many times the loop code is unrolled.
10130 @item max-unroll-times
10131 The maximum number of unrollings of a single loop.
10133 @item max-peeled-insns
10134 The maximum number of instructions that a loop may have to be peeled.
10135 If a loop is peeled, this parameter also determines how many times
10136 the loop code is peeled.
10138 @item max-peel-times
10139 The maximum number of peelings of a single loop.
10141 @item max-peel-branches
10142 The maximum number of branches on the hot path through the peeled sequence.
10144 @item max-completely-peeled-insns
10145 The maximum number of insns of a completely peeled loop.
10147 @item max-completely-peel-times
10148 The maximum number of iterations of a loop to be suitable for complete peeling.
10150 @item max-completely-peel-loop-nest-depth
10151 The maximum depth of a loop nest suitable for complete peeling.
10153 @item max-unswitch-insns
10154 The maximum number of insns of an unswitched loop.
10156 @item max-unswitch-level
10157 The maximum number of branches unswitched in a single loop.
10159 @item max-loop-headers-insns
10160 The maximum number of insns in loop header duplicated by the copy loop headers
10163 @item lim-expensive
10164 The minimum cost of an expensive expression in the loop invariant motion.
10166 @item iv-consider-all-candidates-bound
10167 Bound on number of candidates for induction variables, below which
10168 all candidates are considered for each use in induction variable
10169 optimizations. If there are more candidates than this,
10170 only the most relevant ones are considered to avoid quadratic time complexity.
10172 @item iv-max-considered-uses
10173 The induction variable optimizations give up on loops that contain more
10174 induction variable uses.
10176 @item iv-always-prune-cand-set-bound
10177 If the number of candidates in the set is smaller than this value,
10178 always try to remove unnecessary ivs from the set
10179 when adding a new one.
10181 @item avg-loop-niter
10182 Average number of iterations of a loop.
10184 @item dse-max-object-size
10185 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
10186 Larger values may result in larger compilation times.
10188 @item scev-max-expr-size
10189 Bound on size of expressions used in the scalar evolutions analyzer.
10190 Large expressions slow the analyzer.
10192 @item scev-max-expr-complexity
10193 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10194 Complex expressions slow the analyzer.
10196 @item max-tree-if-conversion-phi-args
10197 Maximum number of arguments in a PHI supported by TREE if conversion
10198 unless the loop is marked with simd pragma.
10200 @item vect-max-version-for-alignment-checks
10201 The maximum number of run-time checks that can be performed when
10202 doing loop versioning for alignment in the vectorizer.
10204 @item vect-max-version-for-alias-checks
10205 The maximum number of run-time checks that can be performed when
10206 doing loop versioning for alias in the vectorizer.
10208 @item vect-max-peeling-for-alignment
10209 The maximum number of loop peels to enhance access alignment
10210 for vectorizer. Value -1 means no limit.
10212 @item max-iterations-to-track
10213 The maximum number of iterations of a loop the brute-force algorithm
10214 for analysis of the number of iterations of the loop tries to evaluate.
10216 @item hot-bb-count-ws-permille
10217 A basic block profile count is considered hot if it contributes to
10218 the given permillage (i.e. 0...1000) of the entire profiled execution.
10220 @item hot-bb-frequency-fraction
10221 Select fraction of the entry block frequency of executions of basic block in
10222 function given basic block needs to have to be considered hot.
10224 @item max-predicted-iterations
10225 The maximum number of loop iterations we predict statically. This is useful
10226 in cases where a function contains a single loop with known bound and
10227 another loop with unknown bound.
10228 The known number of iterations is predicted correctly, while
10229 the unknown number of iterations average to roughly 10. This means that the
10230 loop without bounds appears artificially cold relative to the other one.
10232 @item builtin-expect-probability
10233 Control the probability of the expression having the specified value. This
10234 parameter takes a percentage (i.e. 0 ... 100) as input.
10235 The default probability of 90 is obtained empirically.
10237 @item align-threshold
10239 Select fraction of the maximal frequency of executions of a basic block in
10240 a function to align the basic block.
10242 @item align-loop-iterations
10244 A loop expected to iterate at least the selected number of iterations is
10247 @item tracer-dynamic-coverage
10248 @itemx tracer-dynamic-coverage-feedback
10250 This value is used to limit superblock formation once the given percentage of
10251 executed instructions is covered. This limits unnecessary code size
10254 The @option{tracer-dynamic-coverage-feedback} parameter
10255 is used only when profile
10256 feedback is available. The real profiles (as opposed to statically estimated
10257 ones) are much less balanced allowing the threshold to be larger value.
10259 @item tracer-max-code-growth
10260 Stop tail duplication once code growth has reached given percentage. This is
10261 a rather artificial limit, as most of the duplicates are eliminated later in
10262 cross jumping, so it may be set to much higher values than is the desired code
10265 @item tracer-min-branch-ratio
10267 Stop reverse growth when the reverse probability of best edge is less than this
10268 threshold (in percent).
10270 @item tracer-min-branch-probability
10271 @itemx tracer-min-branch-probability-feedback
10273 Stop forward growth if the best edge has probability lower than this
10276 Similarly to @option{tracer-dynamic-coverage} two parameters are
10277 provided. @option{tracer-min-branch-probability-feedback} is used for
10278 compilation with profile feedback and @option{tracer-min-branch-probability}
10279 compilation without. The value for compilation with profile feedback
10280 needs to be more conservative (higher) in order to make tracer
10283 @item stack-clash-protection-guard-size
10284 Specify the size of the operating system provided stack guard as
10285 2 raised to @var{num} bytes. The default value is 12 (4096 bytes).
10286 Acceptable values are between 12 and 30. Higher values may reduce the
10287 number of explicit probes, but a value larger than the operating system
10288 provided guard will leave code vulnerable to stack clash style attacks.
10290 @item stack-clash-protection-probe-interval
10291 Stack clash protection involves probing stack space as it is allocated. This
10292 param controls the maximum distance between probes into the stack as 2 raised
10293 to @var{num} bytes. Acceptable values are between 10 and 16 and defaults to
10294 12. Higher values may reduce the number of explicit probes, but a value
10295 larger than the operating system provided guard will leave code vulnerable to
10296 stack clash style attacks.
10298 @item max-cse-path-length
10300 The maximum number of basic blocks on path that CSE considers.
10303 @item max-cse-insns
10304 The maximum number of instructions CSE processes before flushing.
10305 The default is 1000.
10307 @item ggc-min-expand
10309 GCC uses a garbage collector to manage its own memory allocation. This
10310 parameter specifies the minimum percentage by which the garbage
10311 collector's heap should be allowed to expand between collections.
10312 Tuning this may improve compilation speed; it has no effect on code
10315 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10316 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10317 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10318 GCC is not able to calculate RAM on a particular platform, the lower
10319 bound of 30% is used. Setting this parameter and
10320 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10321 every opportunity. This is extremely slow, but can be useful for
10324 @item ggc-min-heapsize
10326 Minimum size of the garbage collector's heap before it begins bothering
10327 to collect garbage. The first collection occurs after the heap expands
10328 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10329 tuning this may improve compilation speed, and has no effect on code
10332 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10333 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10334 with a lower bound of 4096 (four megabytes) and an upper bound of
10335 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10336 particular platform, the lower bound is used. Setting this parameter
10337 very large effectively disables garbage collection. Setting this
10338 parameter and @option{ggc-min-expand} to zero causes a full collection
10339 to occur at every opportunity.
10341 @item max-reload-search-insns
10342 The maximum number of instruction reload should look backward for equivalent
10343 register. Increasing values mean more aggressive optimization, making the
10344 compilation time increase with probably slightly better performance.
10345 The default value is 100.
10347 @item max-cselib-memory-locations
10348 The maximum number of memory locations cselib should take into account.
10349 Increasing values mean more aggressive optimization, making the compilation time
10350 increase with probably slightly better performance. The default value is 500.
10352 @item max-sched-ready-insns
10353 The maximum number of instructions ready to be issued the scheduler should
10354 consider at any given time during the first scheduling pass. Increasing
10355 values mean more thorough searches, making the compilation time increase
10356 with probably little benefit. The default value is 100.
10358 @item max-sched-region-blocks
10359 The maximum number of blocks in a region to be considered for
10360 interblock scheduling. The default value is 10.
10362 @item max-pipeline-region-blocks
10363 The maximum number of blocks in a region to be considered for
10364 pipelining in the selective scheduler. The default value is 15.
10366 @item max-sched-region-insns
10367 The maximum number of insns in a region to be considered for
10368 interblock scheduling. The default value is 100.
10370 @item max-pipeline-region-insns
10371 The maximum number of insns in a region to be considered for
10372 pipelining in the selective scheduler. The default value is 200.
10374 @item min-spec-prob
10375 The minimum probability (in percents) of reaching a source block
10376 for interblock speculative scheduling. The default value is 40.
10378 @item max-sched-extend-regions-iters
10379 The maximum number of iterations through CFG to extend regions.
10380 A value of 0 (the default) disables region extensions.
10382 @item max-sched-insn-conflict-delay
10383 The maximum conflict delay for an insn to be considered for speculative motion.
10384 The default value is 3.
10386 @item sched-spec-prob-cutoff
10387 The minimal probability of speculation success (in percents), so that
10388 speculative insns are scheduled.
10389 The default value is 40.
10391 @item sched-state-edge-prob-cutoff
10392 The minimum probability an edge must have for the scheduler to save its
10394 The default value is 10.
10396 @item sched-mem-true-dep-cost
10397 Minimal distance (in CPU cycles) between store and load targeting same
10398 memory locations. The default value is 1.
10400 @item selsched-max-lookahead
10401 The maximum size of the lookahead window of selective scheduling. It is a
10402 depth of search for available instructions.
10403 The default value is 50.
10405 @item selsched-max-sched-times
10406 The maximum number of times that an instruction is scheduled during
10407 selective scheduling. This is the limit on the number of iterations
10408 through which the instruction may be pipelined. The default value is 2.
10410 @item selsched-insns-to-rename
10411 The maximum number of best instructions in the ready list that are considered
10412 for renaming in the selective scheduler. The default value is 2.
10415 The minimum value of stage count that swing modulo scheduler
10416 generates. The default value is 2.
10418 @item max-last-value-rtl
10419 The maximum size measured as number of RTLs that can be recorded in an expression
10420 in combiner for a pseudo register as last known value of that register. The default
10423 @item max-combine-insns
10424 The maximum number of instructions the RTL combiner tries to combine.
10425 The default value is 2 at @option{-Og} and 4 otherwise.
10427 @item integer-share-limit
10428 Small integer constants can use a shared data structure, reducing the
10429 compiler's memory usage and increasing its speed. This sets the maximum
10430 value of a shared integer constant. The default value is 256.
10432 @item ssp-buffer-size
10433 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10434 protection when @option{-fstack-protection} is used.
10436 @item min-size-for-stack-sharing
10437 The minimum size of variables taking part in stack slot sharing when not
10438 optimizing. The default value is 32.
10440 @item max-jump-thread-duplication-stmts
10441 Maximum number of statements allowed in a block that needs to be
10442 duplicated when threading jumps.
10444 @item max-fields-for-field-sensitive
10445 Maximum number of fields in a structure treated in
10446 a field sensitive manner during pointer analysis. The default is zero
10447 for @option{-O0} and @option{-O1},
10448 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10450 @item prefetch-latency
10451 Estimate on average number of instructions that are executed before
10452 prefetch finishes. The distance prefetched ahead is proportional
10453 to this constant. Increasing this number may also lead to less
10454 streams being prefetched (see @option{simultaneous-prefetches}).
10456 @item simultaneous-prefetches
10457 Maximum number of prefetches that can run at the same time.
10459 @item l1-cache-line-size
10460 The size of cache line in L1 cache, in bytes.
10462 @item l1-cache-size
10463 The size of L1 cache, in kilobytes.
10465 @item l2-cache-size
10466 The size of L2 cache, in kilobytes.
10468 @item min-insn-to-prefetch-ratio
10469 The minimum ratio between the number of instructions and the
10470 number of prefetches to enable prefetching in a loop.
10472 @item prefetch-min-insn-to-mem-ratio
10473 The minimum ratio between the number of instructions and the
10474 number of memory references to enable prefetching in a loop.
10476 @item use-canonical-types
10477 Whether the compiler should use the ``canonical'' type system. By
10478 default, this should always be 1, which uses a more efficient internal
10479 mechanism for comparing types in C++ and Objective-C++. However, if
10480 bugs in the canonical type system are causing compilation failures,
10481 set this value to 0 to disable canonical types.
10483 @item switch-conversion-max-branch-ratio
10484 Switch initialization conversion refuses to create arrays that are
10485 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10486 branches in the switch.
10488 @item max-partial-antic-length
10489 Maximum length of the partial antic set computed during the tree
10490 partial redundancy elimination optimization (@option{-ftree-pre}) when
10491 optimizing at @option{-O3} and above. For some sorts of source code
10492 the enhanced partial redundancy elimination optimization can run away,
10493 consuming all of the memory available on the host machine. This
10494 parameter sets a limit on the length of the sets that are computed,
10495 which prevents the runaway behavior. Setting a value of 0 for
10496 this parameter allows an unlimited set length.
10498 @item sccvn-max-scc-size
10499 Maximum size of a strongly connected component (SCC) during SCCVN
10500 processing. If this limit is hit, SCCVN processing for the whole
10501 function is not done and optimizations depending on it are
10502 disabled. The default maximum SCC size is 10000.
10504 @item sccvn-max-alias-queries-per-access
10505 Maximum number of alias-oracle queries we perform when looking for
10506 redundancies for loads and stores. If this limit is hit the search
10507 is aborted and the load or store is not considered redundant. The
10508 number of queries is algorithmically limited to the number of
10509 stores on all paths from the load to the function entry.
10510 The default maximum number of queries is 1000.
10512 @item ira-max-loops-num
10513 IRA uses regional register allocation by default. If a function
10514 contains more loops than the number given by this parameter, only at most
10515 the given number of the most frequently-executed loops form regions
10516 for regional register allocation. The default value of the
10519 @item ira-max-conflict-table-size
10520 Although IRA uses a sophisticated algorithm to compress the conflict
10521 table, the table can still require excessive amounts of memory for
10522 huge functions. If the conflict table for a function could be more
10523 than the size in MB given by this parameter, the register allocator
10524 instead uses a faster, simpler, and lower-quality
10525 algorithm that does not require building a pseudo-register conflict table.
10526 The default value of the parameter is 2000.
10528 @item ira-loop-reserved-regs
10529 IRA can be used to evaluate more accurate register pressure in loops
10530 for decisions to move loop invariants (see @option{-O3}). The number
10531 of available registers reserved for some other purposes is given
10532 by this parameter. The default value of the parameter is 2, which is
10533 the minimal number of registers needed by typical instructions.
10534 This value is the best found from numerous experiments.
10536 @item lra-inheritance-ebb-probability-cutoff
10537 LRA tries to reuse values reloaded in registers in subsequent insns.
10538 This optimization is called inheritance. EBB is used as a region to
10539 do this optimization. The parameter defines a minimal fall-through
10540 edge probability in percentage used to add BB to inheritance EBB in
10541 LRA. The default value of the parameter is 40. The value was chosen
10542 from numerous runs of SPEC2000 on x86-64.
10544 @item loop-invariant-max-bbs-in-loop
10545 Loop invariant motion can be very expensive, both in compilation time and
10546 in amount of needed compile-time memory, with very large loops. Loops
10547 with more basic blocks than this parameter won't have loop invariant
10548 motion optimization performed on them. The default value of the
10549 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10551 @item loop-max-datarefs-for-datadeps
10552 Building data dependencies is expensive for very large loops. This
10553 parameter limits the number of data references in loops that are
10554 considered for data dependence analysis. These large loops are no
10555 handled by the optimizations using loop data dependencies.
10556 The default value is 1000.
10558 @item max-vartrack-size
10559 Sets a maximum number of hash table slots to use during variable
10560 tracking dataflow analysis of any function. If this limit is exceeded
10561 with variable tracking at assignments enabled, analysis for that
10562 function is retried without it, after removing all debug insns from
10563 the function. If the limit is exceeded even without debug insns, var
10564 tracking analysis is completely disabled for the function. Setting
10565 the parameter to zero makes it unlimited.
10567 @item max-vartrack-expr-depth
10568 Sets a maximum number of recursion levels when attempting to map
10569 variable names or debug temporaries to value expressions. This trades
10570 compilation time for more complete debug information. If this is set too
10571 low, value expressions that are available and could be represented in
10572 debug information may end up not being used; setting this higher may
10573 enable the compiler to find more complex debug expressions, but compile
10574 time and memory use may grow. The default is 12.
10576 @item min-nondebug-insn-uid
10577 Use uids starting at this parameter for nondebug insns. The range below
10578 the parameter is reserved exclusively for debug insns created by
10579 @option{-fvar-tracking-assignments}, but debug insns may get
10580 (non-overlapping) uids above it if the reserved range is exhausted.
10582 @item ipa-sra-ptr-growth-factor
10583 IPA-SRA replaces a pointer to an aggregate with one or more new
10584 parameters only when their cumulative size is less or equal to
10585 @option{ipa-sra-ptr-growth-factor} times the size of the original
10588 @item sra-max-scalarization-size-Ospeed
10589 @item sra-max-scalarization-size-Osize
10590 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10591 replace scalar parts of aggregates with uses of independent scalar
10592 variables. These parameters control the maximum size, in storage units,
10593 of aggregate which is considered for replacement when compiling for
10595 (@option{sra-max-scalarization-size-Ospeed}) or size
10596 (@option{sra-max-scalarization-size-Osize}) respectively.
10598 @item tm-max-aggregate-size
10599 When making copies of thread-local variables in a transaction, this
10600 parameter specifies the size in bytes after which variables are
10601 saved with the logging functions as opposed to save/restore code
10602 sequence pairs. This option only applies when using
10605 @item graphite-max-nb-scop-params
10606 To avoid exponential effects in the Graphite loop transforms, the
10607 number of parameters in a Static Control Part (SCoP) is bounded. The
10608 default value is 10 parameters, a value of zero can be used to lift
10609 the bound. A variable whose value is unknown at compilation time and
10610 defined outside a SCoP is a parameter of the SCoP.
10612 @item loop-block-tile-size
10613 Loop blocking or strip mining transforms, enabled with
10614 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10615 loop in the loop nest by a given number of iterations. The strip
10616 length can be changed using the @option{loop-block-tile-size}
10617 parameter. The default value is 51 iterations.
10619 @item loop-unroll-jam-size
10620 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10621 default value is 4.
10623 @item loop-unroll-jam-depth
10624 Specify the dimension to be unrolled (counting from the most inner loop)
10625 for the @option{-floop-unroll-and-jam}. The default value is 2.
10627 @item ipa-cp-value-list-size
10628 IPA-CP attempts to track all possible values and types passed to a function's
10629 parameter in order to propagate them and perform devirtualization.
10630 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10631 stores per one formal parameter of a function.
10633 @item ipa-cp-eval-threshold
10634 IPA-CP calculates its own score of cloning profitability heuristics
10635 and performs those cloning opportunities with scores that exceed
10636 @option{ipa-cp-eval-threshold}.
10638 @item ipa-cp-recursion-penalty
10639 Percentage penalty the recursive functions will receive when they
10640 are evaluated for cloning.
10642 @item ipa-cp-single-call-penalty
10643 Percentage penalty functions containing a single call to another
10644 function will receive when they are evaluated for cloning.
10647 @item ipa-max-agg-items
10648 IPA-CP is also capable to propagate a number of scalar values passed
10649 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10650 number of such values per one parameter.
10652 @item ipa-cp-loop-hint-bonus
10653 When IPA-CP determines that a cloning candidate would make the number
10654 of iterations of a loop known, it adds a bonus of
10655 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10658 @item ipa-cp-array-index-hint-bonus
10659 When IPA-CP determines that a cloning candidate would make the index of
10660 an array access known, it adds a bonus of
10661 @option{ipa-cp-array-index-hint-bonus} to the profitability
10662 score of the candidate.
10664 @item ipa-max-aa-steps
10665 During its analysis of function bodies, IPA-CP employs alias analysis
10666 in order to track values pointed to by function parameters. In order
10667 not spend too much time analyzing huge functions, it gives up and
10668 consider all memory clobbered after examining
10669 @option{ipa-max-aa-steps} statements modifying memory.
10671 @item lto-partitions
10672 Specify desired number of partitions produced during WHOPR compilation.
10673 The number of partitions should exceed the number of CPUs used for compilation.
10674 The default value is 32.
10676 @item lto-min-partition
10677 Size of minimal partition for WHOPR (in estimated instructions).
10678 This prevents expenses of splitting very small programs into too many
10681 @item lto-max-partition
10682 Size of max partition for WHOPR (in estimated instructions).
10683 to provide an upper bound for individual size of partition.
10684 Meant to be used only with balanced partitioning.
10686 @item cxx-max-namespaces-for-diagnostic-help
10687 The maximum number of namespaces to consult for suggestions when C++
10688 name lookup fails for an identifier. The default is 1000.
10690 @item sink-frequency-threshold
10691 The maximum relative execution frequency (in percents) of the target block
10692 relative to a statement's original block to allow statement sinking of a
10693 statement. Larger numbers result in more aggressive statement sinking.
10694 The default value is 75. A small positive adjustment is applied for
10695 statements with memory operands as those are even more profitable so sink.
10697 @item max-stores-to-sink
10698 The maximum number of conditional store pairs that can be sunk. Set to 0
10699 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10700 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10702 @item allow-store-data-races
10703 Allow optimizers to introduce new data races on stores.
10704 Set to 1 to allow, otherwise to 0. This option is enabled by default
10705 at optimization level @option{-Ofast}.
10707 @item case-values-threshold
10708 The smallest number of different values for which it is best to use a
10709 jump-table instead of a tree of conditional branches. If the value is
10710 0, use the default for the machine. The default is 0.
10712 @item tree-reassoc-width
10713 Set the maximum number of instructions executed in parallel in
10714 reassociated tree. This parameter overrides target dependent
10715 heuristics used by default if has non zero value.
10717 @item sched-pressure-algorithm
10718 Choose between the two available implementations of
10719 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10720 and is the more likely to prevent instructions from being reordered.
10721 Algorithm 2 was designed to be a compromise between the relatively
10722 conservative approach taken by algorithm 1 and the rather aggressive
10723 approach taken by the default scheduler. It relies more heavily on
10724 having a regular register file and accurate register pressure classes.
10725 See @file{haifa-sched.c} in the GCC sources for more details.
10727 The default choice depends on the target.
10729 @item max-slsr-cand-scan
10730 Set the maximum number of existing candidates that are considered when
10731 seeking a basis for a new straight-line strength reduction candidate.
10734 Enable buffer overflow detection for global objects. This kind
10735 of protection is enabled by default if you are using
10736 @option{-fsanitize=address} option.
10737 To disable global objects protection use @option{--param asan-globals=0}.
10740 Enable buffer overflow detection for stack objects. This kind of
10741 protection is enabled by default when using @option{-fsanitize=address}.
10742 To disable stack protection use @option{--param asan-stack=0} option.
10744 @item asan-instrument-reads
10745 Enable buffer overflow detection for memory reads. This kind of
10746 protection is enabled by default when using @option{-fsanitize=address}.
10747 To disable memory reads protection use
10748 @option{--param asan-instrument-reads=0}.
10750 @item asan-instrument-writes
10751 Enable buffer overflow detection for memory writes. This kind of
10752 protection is enabled by default when using @option{-fsanitize=address}.
10753 To disable memory writes protection use
10754 @option{--param asan-instrument-writes=0} option.
10756 @item asan-memintrin
10757 Enable detection for built-in functions. This kind of protection
10758 is enabled by default when using @option{-fsanitize=address}.
10759 To disable built-in functions protection use
10760 @option{--param asan-memintrin=0}.
10762 @item asan-use-after-return
10763 Enable detection of use-after-return. This kind of protection
10764 is enabled by default when using the @option{-fsanitize=address} option.
10765 To disable it use @option{--param asan-use-after-return=0}.
10767 Note: By default the check is disabled at run time. To enable it,
10768 add @code{detect_stack_use_after_return=1} to the environment variable
10769 @env{ASAN_OPTIONS}.
10771 @item asan-instrumentation-with-call-threshold
10772 If number of memory accesses in function being instrumented
10773 is greater or equal to this number, use callbacks instead of inline checks.
10774 E.g. to disable inline code use
10775 @option{--param asan-instrumentation-with-call-threshold=0}.
10777 @item use-after-scope-direct-emission-threshold
10778 If the size of a local variable in bytes is smaller or equal to this
10779 number, directly poison (or unpoison) shadow memory instead of using
10780 run-time callbacks. The default value is 256.
10782 @item chkp-max-ctor-size
10783 Static constructors generated by Pointer Bounds Checker may become very
10784 large and significantly increase compile time at optimization level
10785 @option{-O1} and higher. This parameter is a maximum number of statements
10786 in a single generated constructor. Default value is 5000.
10788 @item max-fsm-thread-path-insns
10789 Maximum number of instructions to copy when duplicating blocks on a
10790 finite state automaton jump thread path. The default is 100.
10792 @item max-fsm-thread-length
10793 Maximum number of basic blocks on a finite state automaton jump thread
10794 path. The default is 10.
10796 @item max-fsm-thread-paths
10797 Maximum number of new jump thread paths to create for a finite state
10798 automaton. The default is 50.
10800 @item parloops-chunk-size
10801 Chunk size of omp schedule for loops parallelized by parloops. The default
10804 @item parloops-schedule
10805 Schedule type of omp schedule for loops parallelized by parloops (static,
10806 dynamic, guided, auto, runtime). The default is static.
10808 @item parloops-min-per-thread
10809 The minimum number of iterations per thread of an innermost parallelized
10810 loop for which the parallelized variant is prefered over the single threaded
10811 one. The default is 100. Note that for a parallelized loop nest the
10812 minimum number of iterations of the outermost loop per thread is two.
10814 @item max-ssa-name-query-depth
10815 Maximum depth of recursion when querying properties of SSA names in things
10816 like fold routines. One level of recursion corresponds to following a
10819 @item hsa-gen-debug-stores
10820 Enable emission of special debug stores within HSA kernels which are
10821 then read and reported by libgomp plugin. Generation of these stores
10822 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10825 @item max-speculative-devirt-maydefs
10826 The maximum number of may-defs we analyze when looking for a must-def
10827 specifying the dynamic type of an object that invokes a virtual call
10828 we may be able to devirtualize speculatively.
10830 @item max-vrp-switch-assertions
10831 The maximum number of assertions to add along the default edge of a switch
10832 statement during VRP. The default is 10.
10836 @node Instrumentation Options
10837 @section Program Instrumentation Options
10838 @cindex instrumentation options
10839 @cindex program instrumentation options
10840 @cindex run-time error checking options
10841 @cindex profiling options
10842 @cindex options, program instrumentation
10843 @cindex options, run-time error checking
10844 @cindex options, profiling
10846 GCC supports a number of command-line options that control adding
10847 run-time instrumentation to the code it normally generates.
10848 For example, one purpose of instrumentation is collect profiling
10849 statistics for use in finding program hot spots, code coverage
10850 analysis, or profile-guided optimizations.
10851 Another class of program instrumentation is adding run-time checking
10852 to detect programming errors like invalid pointer
10853 dereferences or out-of-bounds array accesses, as well as deliberately
10854 hostile attacks such as stack smashing or C++ vtable hijacking.
10855 There is also a general hook which can be used to implement other
10856 forms of tracing or function-level instrumentation for debug or
10857 program analysis purposes.
10860 @cindex @command{prof}
10863 Generate extra code to write profile information suitable for the
10864 analysis program @command{prof}. You must use this option when compiling
10865 the source files you want data about, and you must also use it when
10868 @cindex @command{gprof}
10871 Generate extra code to write profile information suitable for the
10872 analysis program @command{gprof}. You must use this option when compiling
10873 the source files you want data about, and you must also use it when
10876 @item -fprofile-arcs
10877 @opindex fprofile-arcs
10878 Add code so that program flow @dfn{arcs} are instrumented. During
10879 execution the program records how many times each branch and call is
10880 executed and how many times it is taken or returns. On targets that support
10881 constructors with priority support, profiling properly handles constructors,
10882 destructors and C++ constructors (and destructors) of classes which are used
10883 as a type of a global variable.
10886 program exits it saves this data to a file called
10887 @file{@var{auxname}.gcda} for each source file. The data may be used for
10888 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10889 test coverage analysis (@option{-ftest-coverage}). Each object file's
10890 @var{auxname} is generated from the name of the output file, if
10891 explicitly specified and it is not the final executable, otherwise it is
10892 the basename of the source file. In both cases any suffix is removed
10893 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10894 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10895 @xref{Cross-profiling}.
10897 @cindex @command{gcov}
10901 This option is used to compile and link code instrumented for coverage
10902 analysis. The option is a synonym for @option{-fprofile-arcs}
10903 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10904 linking). See the documentation for those options for more details.
10909 Compile the source files with @option{-fprofile-arcs} plus optimization
10910 and code generation options. For test coverage analysis, use the
10911 additional @option{-ftest-coverage} option. You do not need to profile
10912 every source file in a program.
10915 Compile the source files additionally with @option{-fprofile-abs-path}
10916 to create absolute path names in the @file{.gcno} files. This allows
10917 @command{gcov} to find the correct sources in projects where compilations
10918 occur with different working directories.
10921 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10922 (the latter implies the former).
10925 Run the program on a representative workload to generate the arc profile
10926 information. This may be repeated any number of times. You can run
10927 concurrent instances of your program, and provided that the file system
10928 supports locking, the data files will be correctly updated. Unless
10929 a strict ISO C dialect option is in effect, @code{fork} calls are
10930 detected and correctly handled without double counting.
10933 For profile-directed optimizations, compile the source files again with
10934 the same optimization and code generation options plus
10935 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10936 Control Optimization}).
10939 For test coverage analysis, use @command{gcov} to produce human readable
10940 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10941 @command{gcov} documentation for further information.
10945 With @option{-fprofile-arcs}, for each function of your program GCC
10946 creates a program flow graph, then finds a spanning tree for the graph.
10947 Only arcs that are not on the spanning tree have to be instrumented: the
10948 compiler adds code to count the number of times that these arcs are
10949 executed. When an arc is the only exit or only entrance to a block, the
10950 instrumentation code can be added to the block; otherwise, a new basic
10951 block must be created to hold the instrumentation code.
10954 @item -ftest-coverage
10955 @opindex ftest-coverage
10956 Produce a notes file that the @command{gcov} code-coverage utility
10957 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10958 show program coverage. Each source file's note file is called
10959 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10960 above for a description of @var{auxname} and instructions on how to
10961 generate test coverage data. Coverage data matches the source files
10962 more closely if you do not optimize.
10964 @item -fprofile-abs-path
10965 @opindex fprofile-abs-path
10966 Automatically convert relative source file names to absolute path names
10967 in the @file{.gcno} files. This allows @command{gcov} to find the correct
10968 sources in projects where compilations occur with different working
10971 @item -fprofile-dir=@var{path}
10972 @opindex fprofile-dir
10974 Set the directory to search for the profile data files in to @var{path}.
10975 This option affects only the profile data generated by
10976 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10977 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10978 and its related options. Both absolute and relative paths can be used.
10979 By default, GCC uses the current directory as @var{path}, thus the
10980 profile data file appears in the same directory as the object file.
10982 @item -fprofile-generate
10983 @itemx -fprofile-generate=@var{path}
10984 @opindex fprofile-generate
10986 Enable options usually used for instrumenting application to produce
10987 profile useful for later recompilation with profile feedback based
10988 optimization. You must use @option{-fprofile-generate} both when
10989 compiling and when linking your program.
10991 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10993 If @var{path} is specified, GCC looks at the @var{path} to find
10994 the profile feedback data files. See @option{-fprofile-dir}.
10996 To optimize the program based on the collected profile information, use
10997 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10999 @item -fprofile-update=@var{method}
11000 @opindex fprofile-update
11002 Alter the update method for an application instrumented for profile
11003 feedback based optimization. The @var{method} argument should be one of
11004 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
11005 The first one is useful for single-threaded applications,
11006 while the second one prevents profile corruption by emitting thread-safe code.
11008 @strong{Warning:} When an application does not properly join all threads
11009 (or creates an detached thread), a profile file can be still corrupted.
11011 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
11012 when supported by a target, or to @samp{single} otherwise. The GCC driver
11013 automatically selects @samp{prefer-atomic} when @option{-pthread}
11014 is present in the command line.
11016 @item -fsanitize=address
11017 @opindex fsanitize=address
11018 Enable AddressSanitizer, a fast memory error detector.
11019 Memory access instructions are instrumented to detect
11020 out-of-bounds and use-after-free bugs.
11021 The option enables @option{-fsanitize-address-use-after-scope}.
11022 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
11023 more details. The run-time behavior can be influenced using the
11024 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
11025 the available options are shown at startup of the instrumented program. See
11026 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
11027 for a list of supported options.
11028 The option cannot be combined with @option{-fsanitize=thread}
11029 and/or @option{-fcheck-pointer-bounds}.
11031 @item -fsanitize=kernel-address
11032 @opindex fsanitize=kernel-address
11033 Enable AddressSanitizer for Linux kernel.
11034 See @uref{https://github.com/google/kasan/wiki} for more details.
11035 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11037 @item -fsanitize=pointer-compare
11038 @opindex fsanitize=pointer-compare
11039 Instrument comparison operation (<, <=, >, >=) with pointer operands.
11040 The option must be combined with either @option{-fsanitize=kernel-address} or
11041 @option{-fsanitize=address}
11042 The option cannot be combined with @option{-fsanitize=thread}
11043 and/or @option{-fcheck-pointer-bounds}.
11044 Note: By default the check is disabled at run time. To enable it,
11045 add @code{detect_invalid_pointer_pairs=1} to the environment variable
11046 @env{ASAN_OPTIONS}.
11048 @item -fsanitize=pointer-subtract
11049 @opindex fsanitize=pointer-subtract
11050 Instrument subtraction with pointer operands.
11051 The option must be combined with either @option{-fsanitize=kernel-address} or
11052 @option{-fsanitize=address}
11053 The option cannot be combined with @option{-fsanitize=thread}
11054 and/or @option{-fcheck-pointer-bounds}.
11055 Note: By default the check is disabled at run time. To enable it,
11056 add @code{detect_invalid_pointer_pairs=1} to the environment variable
11057 @env{ASAN_OPTIONS}.
11059 @item -fsanitize=thread
11060 @opindex fsanitize=thread
11061 Enable ThreadSanitizer, a fast data race detector.
11062 Memory access instructions are instrumented to detect
11063 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
11064 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
11065 environment variable; see
11066 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
11068 The option cannot be combined with @option{-fsanitize=address},
11069 @option{-fsanitize=leak} and/or @option{-fcheck-pointer-bounds}.
11071 Note that sanitized atomic builtins cannot throw exceptions when
11072 operating on invalid memory addresses with non-call exceptions
11073 (@option{-fnon-call-exceptions}).
11075 @item -fsanitize=leak
11076 @opindex fsanitize=leak
11077 Enable LeakSanitizer, a memory leak detector.
11078 This option only matters for linking of executables and
11079 the executable is linked against a library that overrides @code{malloc}
11080 and other allocator functions. See
11081 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
11082 details. The run-time behavior can be influenced using the
11083 @env{LSAN_OPTIONS} environment variable.
11084 The option cannot be combined with @option{-fsanitize=thread}.
11086 @item -fsanitize=undefined
11087 @opindex fsanitize=undefined
11088 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
11089 Various computations are instrumented to detect undefined behavior
11090 at runtime. Current suboptions are:
11094 @item -fsanitize=shift
11095 @opindex fsanitize=shift
11096 This option enables checking that the result of a shift operation is
11097 not undefined. Note that what exactly is considered undefined differs
11098 slightly between C and C++, as well as between ISO C90 and C99, etc.
11099 This option has two suboptions, @option{-fsanitize=shift-base} and
11100 @option{-fsanitize=shift-exponent}.
11102 @item -fsanitize=shift-exponent
11103 @opindex fsanitize=shift-exponent
11104 This option enables checking that the second argument of a shift operation
11105 is not negative and is smaller than the precision of the promoted first
11108 @item -fsanitize=shift-base
11109 @opindex fsanitize=shift-base
11110 If the second argument of a shift operation is within range, check that the
11111 result of a shift operation is not undefined. Note that what exactly is
11112 considered undefined differs slightly between C and C++, as well as between
11113 ISO C90 and C99, etc.
11115 @item -fsanitize=integer-divide-by-zero
11116 @opindex fsanitize=integer-divide-by-zero
11117 Detect integer division by zero as well as @code{INT_MIN / -1} division.
11119 @item -fsanitize=unreachable
11120 @opindex fsanitize=unreachable
11121 With this option, the compiler turns the @code{__builtin_unreachable}
11122 call into a diagnostics message call instead. When reaching the
11123 @code{__builtin_unreachable} call, the behavior is undefined.
11125 @item -fsanitize=vla-bound
11126 @opindex fsanitize=vla-bound
11127 This option instructs the compiler to check that the size of a variable
11128 length array is positive.
11130 @item -fsanitize=null
11131 @opindex fsanitize=null
11132 This option enables pointer checking. Particularly, the application
11133 built with this option turned on will issue an error message when it
11134 tries to dereference a NULL pointer, or if a reference (possibly an
11135 rvalue reference) is bound to a NULL pointer, or if a method is invoked
11136 on an object pointed by a NULL pointer.
11138 @item -fsanitize=return
11139 @opindex fsanitize=return
11140 This option enables return statement checking. Programs
11141 built with this option turned on will issue an error message
11142 when the end of a non-void function is reached without actually
11143 returning a value. This option works in C++ only.
11145 @item -fsanitize=signed-integer-overflow
11146 @opindex fsanitize=signed-integer-overflow
11147 This option enables signed integer overflow checking. We check that
11148 the result of @code{+}, @code{*}, and both unary and binary @code{-}
11149 does not overflow in the signed arithmetics. Note, integer promotion
11150 rules must be taken into account. That is, the following is not an
11153 signed char a = SCHAR_MAX;
11157 @item -fsanitize=bounds
11158 @opindex fsanitize=bounds
11159 This option enables instrumentation of array bounds. Various out of bounds
11160 accesses are detected. Flexible array members, flexible array member-like
11161 arrays, and initializers of variables with static storage are not instrumented.
11162 The option cannot be combined with @option{-fcheck-pointer-bounds}.
11164 @item -fsanitize=bounds-strict
11165 @opindex fsanitize=bounds-strict
11166 This option enables strict instrumentation of array bounds. Most out of bounds
11167 accesses are detected, including flexible array members and flexible array
11168 member-like arrays. Initializers of variables with static storage are not
11169 instrumented. The option cannot be combined
11170 with @option{-fcheck-pointer-bounds}.
11172 @item -fsanitize=alignment
11173 @opindex fsanitize=alignment
11175 This option enables checking of alignment of pointers when they are
11176 dereferenced, or when a reference is bound to insufficiently aligned target,
11177 or when a method or constructor is invoked on insufficiently aligned object.
11179 @item -fsanitize=object-size
11180 @opindex fsanitize=object-size
11181 This option enables instrumentation of memory references using the
11182 @code{__builtin_object_size} function. Various out of bounds pointer
11183 accesses are detected.
11185 @item -fsanitize=float-divide-by-zero
11186 @opindex fsanitize=float-divide-by-zero
11187 Detect floating-point division by zero. Unlike other similar options,
11188 @option{-fsanitize=float-divide-by-zero} is not enabled by
11189 @option{-fsanitize=undefined}, since floating-point division by zero can
11190 be a legitimate way of obtaining infinities and NaNs.
11192 @item -fsanitize=float-cast-overflow
11193 @opindex fsanitize=float-cast-overflow
11194 This option enables floating-point type to integer conversion checking.
11195 We check that the result of the conversion does not overflow.
11196 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
11197 not enabled by @option{-fsanitize=undefined}.
11198 This option does not work well with @code{FE_INVALID} exceptions enabled.
11200 @item -fsanitize=nonnull-attribute
11201 @opindex fsanitize=nonnull-attribute
11203 This option enables instrumentation of calls, checking whether null values
11204 are not passed to arguments marked as requiring a non-null value by the
11205 @code{nonnull} function attribute.
11207 @item -fsanitize=returns-nonnull-attribute
11208 @opindex fsanitize=returns-nonnull-attribute
11210 This option enables instrumentation of return statements in functions
11211 marked with @code{returns_nonnull} function attribute, to detect returning
11212 of null values from such functions.
11214 @item -fsanitize=bool
11215 @opindex fsanitize=bool
11217 This option enables instrumentation of loads from bool. If a value other
11218 than 0/1 is loaded, a run-time error is issued.
11220 @item -fsanitize=enum
11221 @opindex fsanitize=enum
11223 This option enables instrumentation of loads from an enum type. If
11224 a value outside the range of values for the enum type is loaded,
11225 a run-time error is issued.
11227 @item -fsanitize=vptr
11228 @opindex fsanitize=vptr
11230 This option enables instrumentation of C++ member function calls, member
11231 accesses and some conversions between pointers to base and derived classes,
11232 to verify the referenced object has the correct dynamic type.
11234 @item -fsanitize=pointer-overflow
11235 @opindex fsanitize=pointer-overflow
11237 This option enables instrumentation of pointer arithmetics. If the pointer
11238 arithmetics overflows, a run-time error is issued.
11240 @item -fsanitize=builtin
11241 @opindex fsanitize=builtin
11243 This option enables instrumentation of arguments to selected builtin
11244 functions. If an invalid value is passed to such arguments, a run-time
11245 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
11246 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
11251 While @option{-ftrapv} causes traps for signed overflows to be emitted,
11252 @option{-fsanitize=undefined} gives a diagnostic message.
11253 This currently works only for the C family of languages.
11255 @item -fno-sanitize=all
11256 @opindex fno-sanitize=all
11258 This option disables all previously enabled sanitizers.
11259 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
11262 @item -fasan-shadow-offset=@var{number}
11263 @opindex fasan-shadow-offset
11264 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
11265 It is useful for experimenting with different shadow memory layouts in
11266 Kernel AddressSanitizer.
11268 @item -fsanitize-sections=@var{s1},@var{s2},...
11269 @opindex fsanitize-sections
11270 Sanitize global variables in selected user-defined sections. @var{si} may
11273 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
11274 @opindex fsanitize-recover
11275 @opindex fno-sanitize-recover
11276 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
11277 mentioned in comma-separated list of @var{opts}. Enabling this option
11278 for a sanitizer component causes it to attempt to continue
11279 running the program as if no error happened. This means multiple
11280 runtime errors can be reported in a single program run, and the exit
11281 code of the program may indicate success even when errors
11282 have been reported. The @option{-fno-sanitize-recover=} option
11283 can be used to alter
11284 this behavior: only the first detected error is reported
11285 and program then exits with a non-zero exit code.
11287 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
11288 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
11289 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
11290 @option{-fsanitize=bounds-strict},
11291 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
11292 For these sanitizers error recovery is turned on by default,
11293 except @option{-fsanitize=address}, for which this feature is experimental.
11294 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
11295 accepted, the former enables recovery for all sanitizers that support it,
11296 the latter disables recovery for all sanitizers that support it.
11298 Even if a recovery mode is turned on the compiler side, it needs to be also
11299 enabled on the runtime library side, otherwise the failures are still fatal.
11300 The runtime library defaults to @code{halt_on_error=0} for
11301 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
11302 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
11303 setting the @code{halt_on_error} flag in the corresponding environment variable.
11305 Syntax without an explicit @var{opts} parameter is deprecated. It is
11306 equivalent to specifying an @var{opts} list of:
11309 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
11312 @item -fsanitize-address-use-after-scope
11313 @opindex fsanitize-address-use-after-scope
11314 Enable sanitization of local variables to detect use-after-scope bugs.
11315 The option sets @option{-fstack-reuse} to @samp{none}.
11317 @item -fsanitize-undefined-trap-on-error
11318 @opindex fsanitize-undefined-trap-on-error
11319 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
11320 report undefined behavior using @code{__builtin_trap} rather than
11321 a @code{libubsan} library routine. The advantage of this is that the
11322 @code{libubsan} library is not needed and is not linked in, so this
11323 is usable even in freestanding environments.
11325 @item -fsanitize-coverage=trace-pc
11326 @opindex fsanitize-coverage=trace-pc
11327 Enable coverage-guided fuzzing code instrumentation.
11328 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
11330 @item -fsanitize-coverage=trace-cmp
11331 @opindex fsanitize-coverage=trace-cmp
11332 Enable dataflow guided fuzzing code instrumentation.
11333 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
11334 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
11335 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
11336 variable or @code{__sanitizer_cov_trace_const_cmp1},
11337 @code{__sanitizer_cov_trace_const_cmp2},
11338 @code{__sanitizer_cov_trace_const_cmp4} or
11339 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
11340 operand constant, @code{__sanitizer_cov_trace_cmpf} or
11341 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
11342 @code{__sanitizer_cov_trace_switch} for switch statements.
11344 @item -fbounds-check
11345 @opindex fbounds-check
11346 For front ends that support it, generate additional code to check that
11347 indices used to access arrays are within the declared range. This is
11348 currently only supported by the Fortran front end, where this option
11351 @item -fcheck-pointer-bounds
11352 @opindex fcheck-pointer-bounds
11353 @opindex fno-check-pointer-bounds
11354 @cindex Pointer Bounds Checker options
11355 Enable Pointer Bounds Checker instrumentation. Each memory reference
11356 is instrumented with checks of the pointer used for memory access against
11357 bounds associated with that pointer.
11360 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
11361 and @option{-mmpx} are required to enable this feature.
11362 MPX-based instrumentation requires
11363 a runtime library to enable MPX in hardware and handle bounds
11364 violation signals. By default when @option{-fcheck-pointer-bounds}
11365 and @option{-mmpx} options are used to link a program, the GCC driver
11366 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
11367 Bounds checking on calls to dynamic libraries requires a linker
11368 with @option{-z bndplt} support; if GCC was configured with a linker
11369 without support for this option (including the Gold linker and older
11370 versions of ld), a warning is given if you link with @option{-mmpx}
11371 without also specifying @option{-static}, since the overall effectiveness
11372 of the bounds checking protection is reduced.
11373 See also @option{-static-libmpxwrappers}.
11375 MPX-based instrumentation
11376 may be used for debugging and also may be included in production code
11377 to increase program security. Depending on usage, you may
11378 have different requirements for the runtime library. The current version
11379 of the MPX runtime library is more oriented for use as a debugging
11380 tool. MPX runtime library usage implies @option{-lpthread}. See
11381 also @option{-static-libmpx}. The runtime library behavior can be
11382 influenced using various @env{CHKP_RT_*} environment variables. See
11383 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
11386 Generated instrumentation may be controlled by various
11387 @option{-fchkp-*} options and by the @code{bnd_variable_size}
11388 structure field attribute (@pxref{Type Attributes}) and
11389 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
11390 (@pxref{Function Attributes}). GCC also provides a number of built-in
11391 functions for controlling the Pointer Bounds Checker. @xref{Pointer
11392 Bounds Checker builtins}, for more information.
11394 @item -fchkp-check-incomplete-type
11395 @opindex fchkp-check-incomplete-type
11396 @opindex fno-chkp-check-incomplete-type
11397 Generate pointer bounds checks for variables with incomplete type.
11398 Enabled by default.
11400 @item -fchkp-narrow-bounds
11401 @opindex fchkp-narrow-bounds
11402 @opindex fno-chkp-narrow-bounds
11403 Controls bounds used by Pointer Bounds Checker for pointers to object
11404 fields. If narrowing is enabled then field bounds are used. Otherwise
11405 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
11406 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
11408 @item -fchkp-first-field-has-own-bounds
11409 @opindex fchkp-first-field-has-own-bounds
11410 @opindex fno-chkp-first-field-has-own-bounds
11411 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
11412 first field in the structure. By default a pointer to the first field has
11413 the same bounds as a pointer to the whole structure.
11415 @item -fchkp-flexible-struct-trailing-arrays
11416 @opindex fchkp-flexible-struct-trailing-arrays
11417 @opindex fno-chkp-flexible-struct-trailing-arrays
11418 Forces Pointer Bounds Checker to treat all trailing arrays in structures as
11419 possibly flexible. By default only array fields with zero length or that are
11420 marked with attribute bnd_variable_size are treated as flexible.
11422 @item -fchkp-narrow-to-innermost-array
11423 @opindex fchkp-narrow-to-innermost-array
11424 @opindex fno-chkp-narrow-to-innermost-array
11425 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
11426 case of nested static array access. By default this option is disabled and
11427 bounds of the outermost array are used.
11429 @item -fchkp-optimize
11430 @opindex fchkp-optimize
11431 @opindex fno-chkp-optimize
11432 Enables Pointer Bounds Checker optimizations. Enabled by default at
11433 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11435 @item -fchkp-use-fast-string-functions
11436 @opindex fchkp-use-fast-string-functions
11437 @opindex fno-chkp-use-fast-string-functions
11438 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11439 by Pointer Bounds Checker. Disabled by default.
11441 @item -fchkp-use-nochk-string-functions
11442 @opindex fchkp-use-nochk-string-functions
11443 @opindex fno-chkp-use-nochk-string-functions
11444 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11445 by Pointer Bounds Checker. Disabled by default.
11447 @item -fchkp-use-static-bounds
11448 @opindex fchkp-use-static-bounds
11449 @opindex fno-chkp-use-static-bounds
11450 Allow Pointer Bounds Checker to generate static bounds holding
11451 bounds of static variables. Enabled by default.
11453 @item -fchkp-use-static-const-bounds
11454 @opindex fchkp-use-static-const-bounds
11455 @opindex fno-chkp-use-static-const-bounds
11456 Use statically-initialized bounds for constant bounds instead of
11457 generating them each time they are required. By default enabled when
11458 @option{-fchkp-use-static-bounds} is enabled.
11460 @item -fchkp-treat-zero-dynamic-size-as-infinite
11461 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11462 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11463 With this option, objects with incomplete type whose
11464 dynamically-obtained size is zero are treated as having infinite size
11465 instead by Pointer Bounds
11466 Checker. This option may be helpful if a program is linked with a library
11467 missing size information for some symbols. Disabled by default.
11469 @item -fchkp-check-read
11470 @opindex fchkp-check-read
11471 @opindex fno-chkp-check-read
11472 Instructs Pointer Bounds Checker to generate checks for all read
11473 accesses to memory. Enabled by default.
11475 @item -fchkp-check-write
11476 @opindex fchkp-check-write
11477 @opindex fno-chkp-check-write
11478 Instructs Pointer Bounds Checker to generate checks for all write
11479 accesses to memory. Enabled by default.
11481 @item -fchkp-store-bounds
11482 @opindex fchkp-store-bounds
11483 @opindex fno-chkp-store-bounds
11484 Instructs Pointer Bounds Checker to generate bounds stores for
11485 pointer writes. Enabled by default.
11487 @item -fchkp-instrument-calls
11488 @opindex fchkp-instrument-calls
11489 @opindex fno-chkp-instrument-calls
11490 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11491 Enabled by default.
11493 @item -fchkp-instrument-marked-only
11494 @opindex fchkp-instrument-marked-only
11495 @opindex fno-chkp-instrument-marked-only
11496 Instructs Pointer Bounds Checker to instrument only functions
11497 marked with the @code{bnd_instrument} attribute
11498 (@pxref{Function Attributes}). Disabled by default.
11500 @item -fchkp-use-wrappers
11501 @opindex fchkp-use-wrappers
11502 @opindex fno-chkp-use-wrappers
11503 Allows Pointer Bounds Checker to replace calls to built-in functions
11504 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11505 is used to link a program, the GCC driver automatically links
11506 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11507 Enabled by default.
11509 @item -fcf-protection==@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
11510 @opindex fcf-protection
11511 Enable code instrumentation of control-flow transfers to increase
11512 program security by checking that target addresses of control-flow
11513 transfer instructions (such as indirect function call, function return,
11514 indirect jump) are valid. This prevents diverting the flow of control
11515 to an unexpected target. This is intended to protect against such
11516 threats as Return-oriented Programming (ROP), and similarly
11517 call/jmp-oriented programming (COP/JOP).
11519 The value @code{branch} tells the compiler to implement checking of
11520 validity of control-flow transfer at the point of indirect branch
11521 instructions, i.e. call/jmp instructions. The value @code{return}
11522 implements checking of validity at the point of returning from a
11523 function. The value @code{full} is an alias for specifying both
11524 @code{branch} and @code{return}. The value @code{none} turns off
11527 You can also use the @code{nocf_check} attribute to identify
11528 which functions and calls should be skipped from instrumentation
11529 (@pxref{Function Attributes}).
11531 Currently the x86 GNU/Linux target provides an implementation based
11532 on Intel Control-flow Enforcement Technology (CET). Instrumentation
11533 for x86 is controlled by target-specific options @option{-mcet},
11534 @option{-mibt} and @option{-mshstk} (@pxref{x86 Options}).
11536 @item -fstack-protector
11537 @opindex fstack-protector
11538 Emit extra code to check for buffer overflows, such as stack smashing
11539 attacks. This is done by adding a guard variable to functions with
11540 vulnerable objects. This includes functions that call @code{alloca}, and
11541 functions with buffers larger than 8 bytes. The guards are initialized
11542 when a function is entered and then checked when the function exits.
11543 If a guard check fails, an error message is printed and the program exits.
11545 @item -fstack-protector-all
11546 @opindex fstack-protector-all
11547 Like @option{-fstack-protector} except that all functions are protected.
11549 @item -fstack-protector-strong
11550 @opindex fstack-protector-strong
11551 Like @option{-fstack-protector} but includes additional functions to
11552 be protected --- those that have local array definitions, or have
11553 references to local frame addresses.
11555 @item -fstack-protector-explicit
11556 @opindex fstack-protector-explicit
11557 Like @option{-fstack-protector} but only protects those functions which
11558 have the @code{stack_protect} attribute.
11560 @item -fstack-check
11561 @opindex fstack-check
11562 Generate code to verify that you do not go beyond the boundary of the
11563 stack. You should specify this flag if you are running in an
11564 environment with multiple threads, but you only rarely need to specify it in
11565 a single-threaded environment since stack overflow is automatically
11566 detected on nearly all systems if there is only one stack.
11568 Note that this switch does not actually cause checking to be done; the
11569 operating system or the language runtime must do that. The switch causes
11570 generation of code to ensure that they see the stack being extended.
11572 You can additionally specify a string parameter: @samp{no} means no
11573 checking, @samp{generic} means force the use of old-style checking,
11574 @samp{specific} means use the best checking method and is equivalent
11575 to bare @option{-fstack-check}.
11577 Old-style checking is a generic mechanism that requires no specific
11578 target support in the compiler but comes with the following drawbacks:
11582 Modified allocation strategy for large objects: they are always
11583 allocated dynamically if their size exceeds a fixed threshold. Note this
11584 may change the semantics of some code.
11587 Fixed limit on the size of the static frame of functions: when it is
11588 topped by a particular function, stack checking is not reliable and
11589 a warning is issued by the compiler.
11592 Inefficiency: because of both the modified allocation strategy and the
11593 generic implementation, code performance is hampered.
11596 Note that old-style stack checking is also the fallback method for
11597 @samp{specific} if no target support has been added in the compiler.
11599 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
11600 and stack overflows. @samp{specific} is an excellent choice when compiling
11601 Ada code. It is not generally sufficient to protect against stack-clash
11602 attacks. To protect against those you want @samp{-fstack-clash-protection}.
11604 @item -fstack-clash-protection
11605 @opindex fstack-clash-protection
11606 Generate code to prevent stack clash style attacks. When this option is
11607 enabled, the compiler will only allocate one page of stack space at a time
11608 and each page is accessed immediately after allocation. Thus, it prevents
11609 allocations from jumping over any stack guard page provided by the
11612 Most targets do not fully support stack clash protection. However, on
11613 those targets @option{-fstack-clash-protection} will protect dynamic stack
11614 allocations. @option{-fstack-clash-protection} may also provide limited
11615 protection for static stack allocations if the target supports
11616 @option{-fstack-check=specific}.
11618 @item -fstack-limit-register=@var{reg}
11619 @itemx -fstack-limit-symbol=@var{sym}
11620 @itemx -fno-stack-limit
11621 @opindex fstack-limit-register
11622 @opindex fstack-limit-symbol
11623 @opindex fno-stack-limit
11624 Generate code to ensure that the stack does not grow beyond a certain value,
11625 either the value of a register or the address of a symbol. If a larger
11626 stack is required, a signal is raised at run time. For most targets,
11627 the signal is raised before the stack overruns the boundary, so
11628 it is possible to catch the signal without taking special precautions.
11630 For instance, if the stack starts at absolute address @samp{0x80000000}
11631 and grows downwards, you can use the flags
11632 @option{-fstack-limit-symbol=__stack_limit} and
11633 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11634 of 128KB@. Note that this may only work with the GNU linker.
11636 You can locally override stack limit checking by using the
11637 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11639 @item -fsplit-stack
11640 @opindex fsplit-stack
11641 Generate code to automatically split the stack before it overflows.
11642 The resulting program has a discontiguous stack which can only
11643 overflow if the program is unable to allocate any more memory. This
11644 is most useful when running threaded programs, as it is no longer
11645 necessary to calculate a good stack size to use for each thread. This
11646 is currently only implemented for the x86 targets running
11649 When code compiled with @option{-fsplit-stack} calls code compiled
11650 without @option{-fsplit-stack}, there may not be much stack space
11651 available for the latter code to run. If compiling all code,
11652 including library code, with @option{-fsplit-stack} is not an option,
11653 then the linker can fix up these calls so that the code compiled
11654 without @option{-fsplit-stack} always has a large stack. Support for
11655 this is implemented in the gold linker in GNU binutils release 2.21
11658 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11659 @opindex fvtable-verify
11660 This option is only available when compiling C++ code.
11661 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11662 feature that verifies at run time, for every virtual call, that
11663 the vtable pointer through which the call is made is valid for the type of
11664 the object, and has not been corrupted or overwritten. If an invalid vtable
11665 pointer is detected at run time, an error is reported and execution of the
11666 program is immediately halted.
11668 This option causes run-time data structures to be built at program startup,
11669 which are used for verifying the vtable pointers.
11670 The options @samp{std} and @samp{preinit}
11671 control the timing of when these data structures are built. In both cases the
11672 data structures are built before execution reaches @code{main}. Using
11673 @option{-fvtable-verify=std} causes the data structures to be built after
11674 shared libraries have been loaded and initialized.
11675 @option{-fvtable-verify=preinit} causes them to be built before shared
11676 libraries have been loaded and initialized.
11678 If this option appears multiple times in the command line with different
11679 values specified, @samp{none} takes highest priority over both @samp{std} and
11680 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11683 @opindex fvtv-debug
11684 When used in conjunction with @option{-fvtable-verify=std} or
11685 @option{-fvtable-verify=preinit}, causes debug versions of the
11686 runtime functions for the vtable verification feature to be called.
11687 This flag also causes the compiler to log information about which
11688 vtable pointers it finds for each class.
11689 This information is written to a file named @file{vtv_set_ptr_data.log}
11690 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11691 if that is defined or the current working directory otherwise.
11693 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11694 file, be sure to delete any existing one.
11697 @opindex fvtv-counts
11698 This is a debugging flag. When used in conjunction with
11699 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11700 causes the compiler to keep track of the total number of virtual calls
11701 it encounters and the number of verifications it inserts. It also
11702 counts the number of calls to certain run-time library functions
11703 that it inserts and logs this information for each compilation unit.
11704 The compiler writes this information to a file named
11705 @file{vtv_count_data.log} in the directory named by the environment
11706 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11707 directory otherwise. It also counts the size of the vtable pointer sets
11708 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11709 in the same directory.
11711 Note: This feature @emph{appends} data to the log files. To get fresh log
11712 files, be sure to delete any existing ones.
11714 @item -finstrument-functions
11715 @opindex finstrument-functions
11716 Generate instrumentation calls for entry and exit to functions. Just
11717 after function entry and just before function exit, the following
11718 profiling functions are called with the address of the current
11719 function and its call site. (On some platforms,
11720 @code{__builtin_return_address} does not work beyond the current
11721 function, so the call site information may not be available to the
11722 profiling functions otherwise.)
11725 void __cyg_profile_func_enter (void *this_fn,
11727 void __cyg_profile_func_exit (void *this_fn,
11731 The first argument is the address of the start of the current function,
11732 which may be looked up exactly in the symbol table.
11734 This instrumentation is also done for functions expanded inline in other
11735 functions. The profiling calls indicate where, conceptually, the
11736 inline function is entered and exited. This means that addressable
11737 versions of such functions must be available. If all your uses of a
11738 function are expanded inline, this may mean an additional expansion of
11739 code size. If you use @code{extern inline} in your C code, an
11740 addressable version of such functions must be provided. (This is
11741 normally the case anyway, but if you get lucky and the optimizer always
11742 expands the functions inline, you might have gotten away without
11743 providing static copies.)
11745 A function may be given the attribute @code{no_instrument_function}, in
11746 which case this instrumentation is not done. This can be used, for
11747 example, for the profiling functions listed above, high-priority
11748 interrupt routines, and any functions from which the profiling functions
11749 cannot safely be called (perhaps signal handlers, if the profiling
11750 routines generate output or allocate memory).
11752 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11753 @opindex finstrument-functions-exclude-file-list
11755 Set the list of functions that are excluded from instrumentation (see
11756 the description of @option{-finstrument-functions}). If the file that
11757 contains a function definition matches with one of @var{file}, then
11758 that function is not instrumented. The match is done on substrings:
11759 if the @var{file} parameter is a substring of the file name, it is
11760 considered to be a match.
11765 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11769 excludes any inline function defined in files whose pathnames
11770 contain @file{/bits/stl} or @file{include/sys}.
11772 If, for some reason, you want to include letter @samp{,} in one of
11773 @var{sym}, write @samp{\,}. For example,
11774 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11775 (note the single quote surrounding the option).
11777 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11778 @opindex finstrument-functions-exclude-function-list
11780 This is similar to @option{-finstrument-functions-exclude-file-list},
11781 but this option sets the list of function names to be excluded from
11782 instrumentation. The function name to be matched is its user-visible
11783 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11784 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11785 match is done on substrings: if the @var{sym} parameter is a substring
11786 of the function name, it is considered to be a match. For C99 and C++
11787 extended identifiers, the function name must be given in UTF-8, not
11788 using universal character names.
11790 @item -fpatchable-function-entry=@var{N}[,@var{M}]
11791 @opindex fpatchable-function-entry
11792 Generate @var{N} NOPs right at the beginning
11793 of each function, with the function entry point before the @var{M}th NOP.
11794 If @var{M} is omitted, it defaults to @code{0} so the
11795 function entry points to the address just at the first NOP.
11796 The NOP instructions reserve extra space which can be used to patch in
11797 any desired instrumentation at run time, provided that the code segment
11798 is writable. The amount of space is controllable indirectly via
11799 the number of NOPs; the NOP instruction used corresponds to the instruction
11800 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
11801 is target-specific and may also depend on the architecture variant and/or
11802 other compilation options.
11804 For run-time identification, the starting addresses of these areas,
11805 which correspond to their respective function entries minus @var{M},
11806 are additionally collected in the @code{__patchable_function_entries}
11807 section of the resulting binary.
11809 Note that the value of @code{__attribute__ ((patchable_function_entry
11810 (N,M)))} takes precedence over command-line option
11811 @option{-fpatchable-function-entry=N,M}. This can be used to increase
11812 the area size or to remove it completely on a single function.
11813 If @code{N=0}, no pad location is recorded.
11815 The NOP instructions are inserted at---and maybe before, depending on
11816 @var{M}---the function entry address, even before the prologue.
11821 @node Preprocessor Options
11822 @section Options Controlling the Preprocessor
11823 @cindex preprocessor options
11824 @cindex options, preprocessor
11826 These options control the C preprocessor, which is run on each C source
11827 file before actual compilation.
11829 If you use the @option{-E} option, nothing is done except preprocessing.
11830 Some of these options make sense only together with @option{-E} because
11831 they cause the preprocessor output to be unsuitable for actual
11834 In addition to the options listed here, there are a number of options
11835 to control search paths for include files documented in
11836 @ref{Directory Options}.
11837 Options to control preprocessor diagnostics are listed in
11838 @ref{Warning Options}.
11841 @include cppopts.texi
11843 @item -Wp,@var{option}
11845 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11846 and pass @var{option} directly through to the preprocessor. If
11847 @var{option} contains commas, it is split into multiple options at the
11848 commas. However, many options are modified, translated or interpreted
11849 by the compiler driver before being passed to the preprocessor, and
11850 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11851 interface is undocumented and subject to change, so whenever possible
11852 you should avoid using @option{-Wp} and let the driver handle the
11855 @item -Xpreprocessor @var{option}
11856 @opindex Xpreprocessor
11857 Pass @var{option} as an option to the preprocessor. You can use this to
11858 supply system-specific preprocessor options that GCC does not
11861 If you want to pass an option that takes an argument, you must use
11862 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11864 @item -no-integrated-cpp
11865 @opindex no-integrated-cpp
11866 Perform preprocessing as a separate pass before compilation.
11867 By default, GCC performs preprocessing as an integrated part of
11868 input tokenization and parsing.
11869 If this option is provided, the appropriate language front end
11870 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11871 and Objective-C, respectively) is instead invoked twice,
11872 once for preprocessing only and once for actual compilation
11873 of the preprocessed input.
11874 This option may be useful in conjunction with the @option{-B} or
11875 @option{-wrapper} options to specify an alternate preprocessor or
11876 perform additional processing of the program source between
11877 normal preprocessing and compilation.
11881 @node Assembler Options
11882 @section Passing Options to the Assembler
11884 @c prevent bad page break with this line
11885 You can pass options to the assembler.
11888 @item -Wa,@var{option}
11890 Pass @var{option} as an option to the assembler. If @var{option}
11891 contains commas, it is split into multiple options at the commas.
11893 @item -Xassembler @var{option}
11894 @opindex Xassembler
11895 Pass @var{option} as an option to the assembler. You can use this to
11896 supply system-specific assembler options that GCC does not
11899 If you want to pass an option that takes an argument, you must use
11900 @option{-Xassembler} twice, once for the option and once for the argument.
11905 @section Options for Linking
11906 @cindex link options
11907 @cindex options, linking
11909 These options come into play when the compiler links object files into
11910 an executable output file. They are meaningless if the compiler is
11911 not doing a link step.
11915 @item @var{object-file-name}
11916 A file name that does not end in a special recognized suffix is
11917 considered to name an object file or library. (Object files are
11918 distinguished from libraries by the linker according to the file
11919 contents.) If linking is done, these object files are used as input
11928 If any of these options is used, then the linker is not run, and
11929 object file names should not be used as arguments. @xref{Overall
11933 @opindex fuse-ld=bfd
11934 Use the @command{bfd} linker instead of the default linker.
11936 @item -fuse-ld=gold
11937 @opindex fuse-ld=gold
11938 Use the @command{gold} linker instead of the default linker.
11941 @item -l@var{library}
11942 @itemx -l @var{library}
11944 Search the library named @var{library} when linking. (The second
11945 alternative with the library as a separate argument is only for
11946 POSIX compliance and is not recommended.)
11948 It makes a difference where in the command you write this option; the
11949 linker searches and processes libraries and object files in the order they
11950 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11951 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11952 to functions in @samp{z}, those functions may not be loaded.
11954 The linker searches a standard list of directories for the library,
11955 which is actually a file named @file{lib@var{library}.a}. The linker
11956 then uses this file as if it had been specified precisely by name.
11958 The directories searched include several standard system directories
11959 plus any that you specify with @option{-L}.
11961 Normally the files found this way are library files---archive files
11962 whose members are object files. The linker handles an archive file by
11963 scanning through it for members which define symbols that have so far
11964 been referenced but not defined. But if the file that is found is an
11965 ordinary object file, it is linked in the usual fashion. The only
11966 difference between using an @option{-l} option and specifying a file name
11967 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11968 and searches several directories.
11972 You need this special case of the @option{-l} option in order to
11973 link an Objective-C or Objective-C++ program.
11975 @item -nostartfiles
11976 @opindex nostartfiles
11977 Do not use the standard system startup files when linking.
11978 The standard system libraries are used normally, unless @option{-nostdlib}
11979 or @option{-nodefaultlibs} is used.
11981 @item -nodefaultlibs
11982 @opindex nodefaultlibs
11983 Do not use the standard system libraries when linking.
11984 Only the libraries you specify are passed to the linker, and options
11985 specifying linkage of the system libraries, such as @option{-static-libgcc}
11986 or @option{-shared-libgcc}, are ignored.
11987 The standard startup files are used normally, unless @option{-nostartfiles}
11990 The compiler may generate calls to @code{memcmp},
11991 @code{memset}, @code{memcpy} and @code{memmove}.
11992 These entries are usually resolved by entries in
11993 libc. These entry points should be supplied through some other
11994 mechanism when this option is specified.
11998 Do not use the standard system startup files or libraries when linking.
11999 No startup files and only the libraries you specify are passed to
12000 the linker, and options specifying linkage of the system libraries, such as
12001 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
12003 The compiler may generate calls to @code{memcmp}, @code{memset},
12004 @code{memcpy} and @code{memmove}.
12005 These entries are usually resolved by entries in
12006 libc. These entry points should be supplied through some other
12007 mechanism when this option is specified.
12009 @cindex @option{-lgcc}, use with @option{-nostdlib}
12010 @cindex @option{-nostdlib} and unresolved references
12011 @cindex unresolved references and @option{-nostdlib}
12012 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
12013 @cindex @option{-nodefaultlibs} and unresolved references
12014 @cindex unresolved references and @option{-nodefaultlibs}
12015 One of the standard libraries bypassed by @option{-nostdlib} and
12016 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
12017 which GCC uses to overcome shortcomings of particular machines, or special
12018 needs for some languages.
12019 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
12020 Collection (GCC) Internals},
12021 for more discussion of @file{libgcc.a}.)
12022 In most cases, you need @file{libgcc.a} even when you want to avoid
12023 other standard libraries. In other words, when you specify @option{-nostdlib}
12024 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
12025 This ensures that you have no unresolved references to internal GCC
12026 library subroutines.
12027 (An example of such an internal subroutine is @code{__main}, used to ensure C++
12028 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
12029 GNU Compiler Collection (GCC) Internals}.)
12033 Produce a dynamically linked position independent executable on targets
12034 that support it. For predictable results, you must also specify the same
12035 set of options used for compilation (@option{-fpie}, @option{-fPIE},
12036 or model suboptions) when you specify this linker option.
12040 Don't produce a dynamically linked position independent executable.
12043 @opindex static-pie
12044 Produce a static position independent executable on targets that support
12045 it. A static position independent executable is similar to a static
12046 executable, but can be loaded at any address without a dynamic linker.
12047 For predictable results, you must also specify the same set of options
12048 used for compilation (@option{-fpie}, @option{-fPIE}, or model
12049 suboptions) when you specify this linker option.
12053 Link with the POSIX threads library. This option is supported on
12054 GNU/Linux targets, most other Unix derivatives, and also on
12055 x86 Cygwin and MinGW targets. On some targets this option also sets
12056 flags for the preprocessor, so it should be used consistently for both
12057 compilation and linking.
12061 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
12062 that support it. This instructs the linker to add all symbols, not
12063 only used ones, to the dynamic symbol table. This option is needed
12064 for some uses of @code{dlopen} or to allow obtaining backtraces
12065 from within a program.
12069 Remove all symbol table and relocation information from the executable.
12073 On systems that support dynamic linking, this overrides @option{-pie}
12074 and prevents linking with the shared libraries. On other systems, this
12075 option has no effect.
12079 Produce a shared object which can then be linked with other objects to
12080 form an executable. Not all systems support this option. For predictable
12081 results, you must also specify the same set of options used for compilation
12082 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
12083 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
12084 needs to build supplementary stub code for constructors to work. On
12085 multi-libbed systems, @samp{gcc -shared} must select the correct support
12086 libraries to link against. Failing to supply the correct flags may lead
12087 to subtle defects. Supplying them in cases where they are not necessary
12090 @item -shared-libgcc
12091 @itemx -static-libgcc
12092 @opindex shared-libgcc
12093 @opindex static-libgcc
12094 On systems that provide @file{libgcc} as a shared library, these options
12095 force the use of either the shared or static version, respectively.
12096 If no shared version of @file{libgcc} was built when the compiler was
12097 configured, these options have no effect.
12099 There are several situations in which an application should use the
12100 shared @file{libgcc} instead of the static version. The most common
12101 of these is when the application wishes to throw and catch exceptions
12102 across different shared libraries. In that case, each of the libraries
12103 as well as the application itself should use the shared @file{libgcc}.
12105 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
12106 whenever you build a shared library or a main executable, because C++
12107 programs typically use exceptions, so this is the right thing to do.
12109 If, instead, you use the GCC driver to create shared libraries, you may
12110 find that they are not always linked with the shared @file{libgcc}.
12111 If GCC finds, at its configuration time, that you have a non-GNU linker
12112 or a GNU linker that does not support option @option{--eh-frame-hdr},
12113 it links the shared version of @file{libgcc} into shared libraries
12114 by default. Otherwise, it takes advantage of the linker and optimizes
12115 away the linking with the shared version of @file{libgcc}, linking with
12116 the static version of libgcc by default. This allows exceptions to
12117 propagate through such shared libraries, without incurring relocation
12118 costs at library load time.
12120 However, if a library or main executable is supposed to throw or catch
12121 exceptions, you must link it using the G++ driver, as appropriate
12122 for the languages used in the program, or using the option
12123 @option{-shared-libgcc}, such that it is linked with the shared
12126 @item -static-libasan
12127 @opindex static-libasan
12128 When the @option{-fsanitize=address} option is used to link a program,
12129 the GCC driver automatically links against @option{libasan}. If
12130 @file{libasan} is available as a shared library, and the @option{-static}
12131 option is not used, then this links against the shared version of
12132 @file{libasan}. The @option{-static-libasan} option directs the GCC
12133 driver to link @file{libasan} statically, without necessarily linking
12134 other libraries statically.
12136 @item -static-libtsan
12137 @opindex static-libtsan
12138 When the @option{-fsanitize=thread} option is used to link a program,
12139 the GCC driver automatically links against @option{libtsan}. If
12140 @file{libtsan} is available as a shared library, and the @option{-static}
12141 option is not used, then this links against the shared version of
12142 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
12143 driver to link @file{libtsan} statically, without necessarily linking
12144 other libraries statically.
12146 @item -static-liblsan
12147 @opindex static-liblsan
12148 When the @option{-fsanitize=leak} option is used to link a program,
12149 the GCC driver automatically links against @option{liblsan}. If
12150 @file{liblsan} is available as a shared library, and the @option{-static}
12151 option is not used, then this links against the shared version of
12152 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
12153 driver to link @file{liblsan} statically, without necessarily linking
12154 other libraries statically.
12156 @item -static-libubsan
12157 @opindex static-libubsan
12158 When the @option{-fsanitize=undefined} option is used to link a program,
12159 the GCC driver automatically links against @option{libubsan}. If
12160 @file{libubsan} is available as a shared library, and the @option{-static}
12161 option is not used, then this links against the shared version of
12162 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
12163 driver to link @file{libubsan} statically, without necessarily linking
12164 other libraries statically.
12166 @item -static-libmpx
12167 @opindex static-libmpx
12168 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
12169 used to link a program, the GCC driver automatically links against
12170 @file{libmpx}. If @file{libmpx} is available as a shared library,
12171 and the @option{-static} option is not used, then this links against
12172 the shared version of @file{libmpx}. The @option{-static-libmpx}
12173 option directs the GCC driver to link @file{libmpx} statically,
12174 without necessarily linking other libraries statically.
12176 @item -static-libmpxwrappers
12177 @opindex static-libmpxwrappers
12178 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
12179 to link a program without also using @option{-fno-chkp-use-wrappers}, the
12180 GCC driver automatically links against @file{libmpxwrappers}. If
12181 @file{libmpxwrappers} is available as a shared library, and the
12182 @option{-static} option is not used, then this links against the shared
12183 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
12184 option directs the GCC driver to link @file{libmpxwrappers} statically,
12185 without necessarily linking other libraries statically.
12187 @item -static-libstdc++
12188 @opindex static-libstdc++
12189 When the @command{g++} program is used to link a C++ program, it
12190 normally automatically links against @option{libstdc++}. If
12191 @file{libstdc++} is available as a shared library, and the
12192 @option{-static} option is not used, then this links against the
12193 shared version of @file{libstdc++}. That is normally fine. However, it
12194 is sometimes useful to freeze the version of @file{libstdc++} used by
12195 the program without going all the way to a fully static link. The
12196 @option{-static-libstdc++} option directs the @command{g++} driver to
12197 link @file{libstdc++} statically, without necessarily linking other
12198 libraries statically.
12202 Bind references to global symbols when building a shared object. Warn
12203 about any unresolved references (unless overridden by the link editor
12204 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
12207 @item -T @var{script}
12209 @cindex linker script
12210 Use @var{script} as the linker script. This option is supported by most
12211 systems using the GNU linker. On some targets, such as bare-board
12212 targets without an operating system, the @option{-T} option may be required
12213 when linking to avoid references to undefined symbols.
12215 @item -Xlinker @var{option}
12217 Pass @var{option} as an option to the linker. You can use this to
12218 supply system-specific linker options that GCC does not recognize.
12220 If you want to pass an option that takes a separate argument, you must use
12221 @option{-Xlinker} twice, once for the option and once for the argument.
12222 For example, to pass @option{-assert definitions}, you must write
12223 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
12224 @option{-Xlinker "-assert definitions"}, because this passes the entire
12225 string as a single argument, which is not what the linker expects.
12227 When using the GNU linker, it is usually more convenient to pass
12228 arguments to linker options using the @option{@var{option}=@var{value}}
12229 syntax than as separate arguments. For example, you can specify
12230 @option{-Xlinker -Map=output.map} rather than
12231 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
12232 this syntax for command-line options.
12234 @item -Wl,@var{option}
12236 Pass @var{option} as an option to the linker. If @var{option} contains
12237 commas, it is split into multiple options at the commas. You can use this
12238 syntax to pass an argument to the option.
12239 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
12240 linker. When using the GNU linker, you can also get the same effect with
12241 @option{-Wl,-Map=output.map}.
12243 @item -u @var{symbol}
12245 Pretend the symbol @var{symbol} is undefined, to force linking of
12246 library modules to define it. You can use @option{-u} multiple times with
12247 different symbols to force loading of additional library modules.
12249 @item -z @var{keyword}
12251 @option{-z} is passed directly on to the linker along with the keyword
12252 @var{keyword}. See the section in the documentation of your linker for
12253 permitted values and their meanings.
12256 @node Directory Options
12257 @section Options for Directory Search
12258 @cindex directory options
12259 @cindex options, directory search
12260 @cindex search path
12262 These options specify directories to search for header files, for
12263 libraries and for parts of the compiler:
12266 @include cppdiropts.texi
12268 @item -iplugindir=@var{dir}
12269 @opindex iplugindir=
12270 Set the directory to search for plugins that are passed
12271 by @option{-fplugin=@var{name}} instead of
12272 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
12273 to be used by the user, but only passed by the driver.
12277 Add directory @var{dir} to the list of directories to be searched
12280 @item -B@var{prefix}
12282 This option specifies where to find the executables, libraries,
12283 include files, and data files of the compiler itself.
12285 The compiler driver program runs one or more of the subprograms
12286 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
12287 @var{prefix} as a prefix for each program it tries to run, both with and
12288 without @samp{@var{machine}/@var{version}/} for the corresponding target
12289 machine and compiler version.
12291 For each subprogram to be run, the compiler driver first tries the
12292 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
12293 is not specified, the driver tries two standard prefixes,
12294 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
12295 those results in a file name that is found, the unmodified program
12296 name is searched for using the directories specified in your
12297 @env{PATH} environment variable.
12299 The compiler checks to see if the path provided by @option{-B}
12300 refers to a directory, and if necessary it adds a directory
12301 separator character at the end of the path.
12303 @option{-B} prefixes that effectively specify directory names also apply
12304 to libraries in the linker, because the compiler translates these
12305 options into @option{-L} options for the linker. They also apply to
12306 include files in the preprocessor, because the compiler translates these
12307 options into @option{-isystem} options for the preprocessor. In this case,
12308 the compiler appends @samp{include} to the prefix.
12310 The runtime support file @file{libgcc.a} can also be searched for using
12311 the @option{-B} prefix, if needed. If it is not found there, the two
12312 standard prefixes above are tried, and that is all. The file is left
12313 out of the link if it is not found by those means.
12315 Another way to specify a prefix much like the @option{-B} prefix is to use
12316 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
12319 As a special kludge, if the path provided by @option{-B} is
12320 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
12321 9, then it is replaced by @file{[dir/]include}. This is to help
12322 with boot-strapping the compiler.
12324 @item -no-canonical-prefixes
12325 @opindex no-canonical-prefixes
12326 Do not expand any symbolic links, resolve references to @samp{/../}
12327 or @samp{/./}, or make the path absolute when generating a relative
12330 @item --sysroot=@var{dir}
12332 Use @var{dir} as the logical root directory for headers and libraries.
12333 For example, if the compiler normally searches for headers in
12334 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
12335 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
12337 If you use both this option and the @option{-isysroot} option, then
12338 the @option{--sysroot} option applies to libraries, but the
12339 @option{-isysroot} option applies to header files.
12341 The GNU linker (beginning with version 2.16) has the necessary support
12342 for this option. If your linker does not support this option, the
12343 header file aspect of @option{--sysroot} still works, but the
12344 library aspect does not.
12346 @item --no-sysroot-suffix
12347 @opindex no-sysroot-suffix
12348 For some targets, a suffix is added to the root directory specified
12349 with @option{--sysroot}, depending on the other options used, so that
12350 headers may for example be found in
12351 @file{@var{dir}/@var{suffix}/usr/include} instead of
12352 @file{@var{dir}/usr/include}. This option disables the addition of
12357 @node Code Gen Options
12358 @section Options for Code Generation Conventions
12359 @cindex code generation conventions
12360 @cindex options, code generation
12361 @cindex run-time options
12363 These machine-independent options control the interface conventions
12364 used in code generation.
12366 Most of them have both positive and negative forms; the negative form
12367 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
12368 one of the forms is listed---the one that is not the default. You
12369 can figure out the other form by either removing @samp{no-} or adding
12373 @item -fstack-reuse=@var{reuse-level}
12374 @opindex fstack_reuse
12375 This option controls stack space reuse for user declared local/auto variables
12376 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
12377 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
12378 local variables and temporaries, @samp{named_vars} enables the reuse only for
12379 user defined local variables with names, and @samp{none} disables stack reuse
12380 completely. The default value is @samp{all}. The option is needed when the
12381 program extends the lifetime of a scoped local variable or a compiler generated
12382 temporary beyond the end point defined by the language. When a lifetime of
12383 a variable ends, and if the variable lives in memory, the optimizing compiler
12384 has the freedom to reuse its stack space with other temporaries or scoped
12385 local variables whose live range does not overlap with it. Legacy code extending
12386 local lifetime is likely to break with the stack reuse optimization.
12405 if (*p == 10) // out of scope use of local1
12416 A(int k) : i(k), j(k) @{ @}
12423 void foo(const A& ar)
12430 foo(A(10)); // temp object's lifetime ends when foo returns
12436 ap->i+= 10; // ap references out of scope temp whose space
12437 // is reused with a. What is the value of ap->i?
12442 The lifetime of a compiler generated temporary is well defined by the C++
12443 standard. When a lifetime of a temporary ends, and if the temporary lives
12444 in memory, the optimizing compiler has the freedom to reuse its stack
12445 space with other temporaries or scoped local variables whose live range
12446 does not overlap with it. However some of the legacy code relies on
12447 the behavior of older compilers in which temporaries' stack space is
12448 not reused, the aggressive stack reuse can lead to runtime errors. This
12449 option is used to control the temporary stack reuse optimization.
12453 This option generates traps for signed overflow on addition, subtraction,
12454 multiplication operations.
12455 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12456 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12457 @option{-fwrapv} being effective. Note that only active options override, so
12458 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12459 results in @option{-ftrapv} being effective.
12463 This option instructs the compiler to assume that signed arithmetic
12464 overflow of addition, subtraction and multiplication wraps around
12465 using twos-complement representation. This flag enables some optimizations
12466 and disables others.
12467 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12468 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12469 @option{-fwrapv} being effective. Note that only active options override, so
12470 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12471 results in @option{-ftrapv} being effective.
12474 @opindex fexceptions
12475 Enable exception handling. Generates extra code needed to propagate
12476 exceptions. For some targets, this implies GCC generates frame
12477 unwind information for all functions, which can produce significant data
12478 size overhead, although it does not affect execution. If you do not
12479 specify this option, GCC enables it by default for languages like
12480 C++ that normally require exception handling, and disables it for
12481 languages like C that do not normally require it. However, you may need
12482 to enable this option when compiling C code that needs to interoperate
12483 properly with exception handlers written in C++. You may also wish to
12484 disable this option if you are compiling older C++ programs that don't
12485 use exception handling.
12487 @item -fnon-call-exceptions
12488 @opindex fnon-call-exceptions
12489 Generate code that allows trapping instructions to throw exceptions.
12490 Note that this requires platform-specific runtime support that does
12491 not exist everywhere. Moreover, it only allows @emph{trapping}
12492 instructions to throw exceptions, i.e.@: memory references or floating-point
12493 instructions. It does not allow exceptions to be thrown from
12494 arbitrary signal handlers such as @code{SIGALRM}.
12496 @item -fdelete-dead-exceptions
12497 @opindex fdelete-dead-exceptions
12498 Consider that instructions that may throw exceptions but don't otherwise
12499 contribute to the execution of the program can be optimized away.
12500 This option is enabled by default for the Ada front end, as permitted by
12501 the Ada language specification.
12502 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12504 @item -funwind-tables
12505 @opindex funwind-tables
12506 Similar to @option{-fexceptions}, except that it just generates any needed
12507 static data, but does not affect the generated code in any other way.
12508 You normally do not need to enable this option; instead, a language processor
12509 that needs this handling enables it on your behalf.
12511 @item -fasynchronous-unwind-tables
12512 @opindex fasynchronous-unwind-tables
12513 Generate unwind table in DWARF format, if supported by target machine. The
12514 table is exact at each instruction boundary, so it can be used for stack
12515 unwinding from asynchronous events (such as debugger or garbage collector).
12517 @item -fno-gnu-unique
12518 @opindex fno-gnu-unique
12519 On systems with recent GNU assembler and C library, the C++ compiler
12520 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12521 of template static data members and static local variables in inline
12522 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12523 is necessary to avoid problems with a library used by two different
12524 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12525 therefore disagreeing with the other one about the binding of the
12526 symbol. But this causes @code{dlclose} to be ignored for affected
12527 DSOs; if your program relies on reinitialization of a DSO via
12528 @code{dlclose} and @code{dlopen}, you can use
12529 @option{-fno-gnu-unique}.
12531 @item -fpcc-struct-return
12532 @opindex fpcc-struct-return
12533 Return ``short'' @code{struct} and @code{union} values in memory like
12534 longer ones, rather than in registers. This convention is less
12535 efficient, but it has the advantage of allowing intercallability between
12536 GCC-compiled files and files compiled with other compilers, particularly
12537 the Portable C Compiler (pcc).
12539 The precise convention for returning structures in memory depends
12540 on the target configuration macros.
12542 Short structures and unions are those whose size and alignment match
12543 that of some integer type.
12545 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12546 switch is not binary compatible with code compiled with the
12547 @option{-freg-struct-return} switch.
12548 Use it to conform to a non-default application binary interface.
12550 @item -freg-struct-return
12551 @opindex freg-struct-return
12552 Return @code{struct} and @code{union} values in registers when possible.
12553 This is more efficient for small structures than
12554 @option{-fpcc-struct-return}.
12556 If you specify neither @option{-fpcc-struct-return} nor
12557 @option{-freg-struct-return}, GCC defaults to whichever convention is
12558 standard for the target. If there is no standard convention, GCC
12559 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12560 the principal compiler. In those cases, we can choose the standard, and
12561 we chose the more efficient register return alternative.
12563 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12564 switch is not binary compatible with code compiled with the
12565 @option{-fpcc-struct-return} switch.
12566 Use it to conform to a non-default application binary interface.
12568 @item -fshort-enums
12569 @opindex fshort-enums
12570 Allocate to an @code{enum} type only as many bytes as it needs for the
12571 declared range of possible values. Specifically, the @code{enum} type
12572 is equivalent to the smallest integer type that has enough room.
12574 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12575 code that is not binary compatible with code generated without that switch.
12576 Use it to conform to a non-default application binary interface.
12578 @item -fshort-wchar
12579 @opindex fshort-wchar
12580 Override the underlying type for @code{wchar_t} to be @code{short
12581 unsigned int} instead of the default for the target. This option is
12582 useful for building programs to run under WINE@.
12584 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12585 code that is not binary compatible with code generated without that switch.
12586 Use it to conform to a non-default application binary interface.
12589 @opindex fno-common
12590 @cindex tentative definitions
12591 In C code, this option controls the placement of global variables
12592 defined without an initializer, known as @dfn{tentative definitions}
12593 in the C standard. Tentative definitions are distinct from declarations
12594 of a variable with the @code{extern} keyword, which do not allocate storage.
12596 Unix C compilers have traditionally allocated storage for
12597 uninitialized global variables in a common block. This allows the
12598 linker to resolve all tentative definitions of the same variable
12599 in different compilation units to the same object, or to a non-tentative
12601 This is the behavior specified by @option{-fcommon}, and is the default for
12602 GCC on most targets.
12603 On the other hand, this behavior is not required by ISO
12604 C, and on some targets may carry a speed or code size penalty on
12605 variable references.
12607 The @option{-fno-common} option specifies that the compiler should instead
12608 place uninitialized global variables in the data section of the object file.
12609 This inhibits the merging of tentative definitions by the linker so
12610 you get a multiple-definition error if the same
12611 variable is defined in more than one compilation unit.
12612 Compiling with @option{-fno-common} is useful on targets for which
12613 it provides better performance, or if you wish to verify that the
12614 program will work on other systems that always treat uninitialized
12615 variable definitions this way.
12619 Ignore the @code{#ident} directive.
12621 @item -finhibit-size-directive
12622 @opindex finhibit-size-directive
12623 Don't output a @code{.size} assembler directive, or anything else that
12624 would cause trouble if the function is split in the middle, and the
12625 two halves are placed at locations far apart in memory. This option is
12626 used when compiling @file{crtstuff.c}; you should not need to use it
12629 @item -fverbose-asm
12630 @opindex fverbose-asm
12631 Put extra commentary information in the generated assembly code to
12632 make it more readable. This option is generally only of use to those
12633 who actually need to read the generated assembly code (perhaps while
12634 debugging the compiler itself).
12636 @option{-fno-verbose-asm}, the default, causes the
12637 extra information to be omitted and is useful when comparing two assembler
12640 The added comments include:
12645 information on the compiler version and command-line options,
12648 the source code lines associated with the assembly instructions,
12649 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12652 hints on which high-level expressions correspond to
12653 the various assembly instruction operands.
12657 For example, given this C source file:
12665 for (i = 0; i < n; i++)
12672 compiling to (x86_64) assembly via @option{-S} and emitting the result
12673 direct to stdout via @option{-o} @option{-}
12676 gcc -S test.c -fverbose-asm -Os -o -
12679 gives output similar to this:
12683 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12690 .type test, @@function
12694 # test.c:4: int total = 0;
12695 xorl %eax, %eax # <retval>
12696 # test.c:6: for (i = 0; i < n; i++)
12697 xorl %edx, %edx # i
12699 # test.c:6: for (i = 0; i < n; i++)
12700 cmpl %edi, %edx # n, i
12702 # test.c:7: total += i * i;
12703 movl %edx, %ecx # i, tmp92
12704 imull %edx, %ecx # i, tmp92
12705 # test.c:6: for (i = 0; i < n; i++)
12707 # test.c:7: total += i * i;
12708 addl %ecx, %eax # tmp92, <retval>
12716 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12717 .section .note.GNU-stack,"",@@progbits
12720 The comments are intended for humans rather than machines and hence the
12721 precise format of the comments is subject to change.
12723 @item -frecord-gcc-switches
12724 @opindex frecord-gcc-switches
12725 This switch causes the command line used to invoke the
12726 compiler to be recorded into the object file that is being created.
12727 This switch is only implemented on some targets and the exact format
12728 of the recording is target and binary file format dependent, but it
12729 usually takes the form of a section containing ASCII text. This
12730 switch is related to the @option{-fverbose-asm} switch, but that
12731 switch only records information in the assembler output file as
12732 comments, so it never reaches the object file.
12733 See also @option{-grecord-gcc-switches} for another
12734 way of storing compiler options into the object file.
12738 @cindex global offset table
12740 Generate position-independent code (PIC) suitable for use in a shared
12741 library, if supported for the target machine. Such code accesses all
12742 constant addresses through a global offset table (GOT)@. The dynamic
12743 loader resolves the GOT entries when the program starts (the dynamic
12744 loader is not part of GCC; it is part of the operating system). If
12745 the GOT size for the linked executable exceeds a machine-specific
12746 maximum size, you get an error message from the linker indicating that
12747 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12748 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12749 on the m68k and RS/6000. The x86 has no such limit.)
12751 Position-independent code requires special support, and therefore works
12752 only on certain machines. For the x86, GCC supports PIC for System V
12753 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12754 position-independent.
12756 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12761 If supported for the target machine, emit position-independent code,
12762 suitable for dynamic linking and avoiding any limit on the size of the
12763 global offset table. This option makes a difference on AArch64, m68k,
12764 PowerPC and SPARC@.
12766 Position-independent code requires special support, and therefore works
12767 only on certain machines.
12769 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12776 These options are similar to @option{-fpic} and @option{-fPIC}, but
12777 generated position independent code can be only linked into executables.
12778 Usually these options are used when @option{-pie} GCC option is
12779 used during linking.
12781 @option{-fpie} and @option{-fPIE} both define the macros
12782 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12783 for @option{-fpie} and 2 for @option{-fPIE}.
12787 Do not use the PLT for external function calls in position-independent code.
12788 Instead, load the callee address at call sites from the GOT and branch to it.
12789 This leads to more efficient code by eliminating PLT stubs and exposing
12790 GOT loads to optimizations. On architectures such as 32-bit x86 where
12791 PLT stubs expect the GOT pointer in a specific register, this gives more
12792 register allocation freedom to the compiler.
12793 Lazy binding requires use of the PLT;
12794 with @option{-fno-plt} all external symbols are resolved at load time.
12796 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12797 through the PLT for specific external functions.
12799 In position-dependent code, a few targets also convert calls to
12800 functions that are marked to not use the PLT to use the GOT instead.
12802 @item -fno-jump-tables
12803 @opindex fno-jump-tables
12804 Do not use jump tables for switch statements even where it would be
12805 more efficient than other code generation strategies. This option is
12806 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12807 building code that forms part of a dynamic linker and cannot
12808 reference the address of a jump table. On some targets, jump tables
12809 do not require a GOT and this option is not needed.
12811 @item -ffixed-@var{reg}
12813 Treat the register named @var{reg} as a fixed register; generated code
12814 should never refer to it (except perhaps as a stack pointer, frame
12815 pointer or in some other fixed role).
12817 @var{reg} must be the name of a register. The register names accepted
12818 are machine-specific and are defined in the @code{REGISTER_NAMES}
12819 macro in the machine description macro file.
12821 This flag does not have a negative form, because it specifies a
12824 @item -fcall-used-@var{reg}
12825 @opindex fcall-used
12826 Treat the register named @var{reg} as an allocable register that is
12827 clobbered by function calls. It may be allocated for temporaries or
12828 variables that do not live across a call. Functions compiled this way
12829 do not save and restore the register @var{reg}.
12831 It is an error to use this flag with the frame pointer or stack pointer.
12832 Use of this flag for other registers that have fixed pervasive roles in
12833 the machine's execution model produces disastrous results.
12835 This flag does not have a negative form, because it specifies a
12838 @item -fcall-saved-@var{reg}
12839 @opindex fcall-saved
12840 Treat the register named @var{reg} as an allocable register saved by
12841 functions. It may be allocated even for temporaries or variables that
12842 live across a call. Functions compiled this way save and restore
12843 the register @var{reg} if they use it.
12845 It is an error to use this flag with the frame pointer or stack pointer.
12846 Use of this flag for other registers that have fixed pervasive roles in
12847 the machine's execution model produces disastrous results.
12849 A different sort of disaster results from the use of this flag for
12850 a register in which function values may be returned.
12852 This flag does not have a negative form, because it specifies a
12855 @item -fpack-struct[=@var{n}]
12856 @opindex fpack-struct
12857 Without a value specified, pack all structure members together without
12858 holes. When a value is specified (which must be a small power of two), pack
12859 structure members according to this value, representing the maximum
12860 alignment (that is, objects with default alignment requirements larger than
12861 this are output potentially unaligned at the next fitting location.
12863 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12864 code that is not binary compatible with code generated without that switch.
12865 Additionally, it makes the code suboptimal.
12866 Use it to conform to a non-default application binary interface.
12868 @item -fleading-underscore
12869 @opindex fleading-underscore
12870 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12871 change the way C symbols are represented in the object file. One use
12872 is to help link with legacy assembly code.
12874 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12875 generate code that is not binary compatible with code generated without that
12876 switch. Use it to conform to a non-default application binary interface.
12877 Not all targets provide complete support for this switch.
12879 @item -ftls-model=@var{model}
12880 @opindex ftls-model
12881 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12882 The @var{model} argument should be one of @samp{global-dynamic},
12883 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12884 Note that the choice is subject to optimization: the compiler may use
12885 a more efficient model for symbols not visible outside of the translation
12886 unit, or if @option{-fpic} is not given on the command line.
12888 The default without @option{-fpic} is @samp{initial-exec}; with
12889 @option{-fpic} the default is @samp{global-dynamic}.
12891 @item -ftrampolines
12892 @opindex ftrampolines
12893 For targets that normally need trampolines for nested functions, always
12894 generate them instead of using descriptors. Otherwise, for targets that
12895 do not need them, like for example HP-PA or IA-64, do nothing.
12897 A trampoline is a small piece of code that is created at run time on the
12898 stack when the address of a nested function is taken, and is used to call
12899 the nested function indirectly. Therefore, it requires the stack to be
12900 made executable in order for the program to work properly.
12902 @option{-fno-trampolines} is enabled by default on a language by language
12903 basis to let the compiler avoid generating them, if it computes that this
12904 is safe, and replace them with descriptors. Descriptors are made up of data
12905 only, but the generated code must be prepared to deal with them. As of this
12906 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12908 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12909 @option{-fno-trampolines} are not binary compatible if nested functions are
12910 present. This option must therefore be used on a program-wide basis and be
12911 manipulated with extreme care.
12913 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12914 @opindex fvisibility
12915 Set the default ELF image symbol visibility to the specified option---all
12916 symbols are marked with this unless overridden within the code.
12917 Using this feature can very substantially improve linking and
12918 load times of shared object libraries, produce more optimized
12919 code, provide near-perfect API export and prevent symbol clashes.
12920 It is @strong{strongly} recommended that you use this in any shared objects
12923 Despite the nomenclature, @samp{default} always means public; i.e.,
12924 available to be linked against from outside the shared object.
12925 @samp{protected} and @samp{internal} are pretty useless in real-world
12926 usage so the only other commonly used option is @samp{hidden}.
12927 The default if @option{-fvisibility} isn't specified is
12928 @samp{default}, i.e., make every symbol public.
12930 A good explanation of the benefits offered by ensuring ELF
12931 symbols have the correct visibility is given by ``How To Write
12932 Shared Libraries'' by Ulrich Drepper (which can be found at
12933 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
12934 solution made possible by this option to marking things hidden when
12935 the default is public is to make the default hidden and mark things
12936 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12937 and @code{__attribute__ ((visibility("default")))} instead of
12938 @code{__declspec(dllexport)} you get almost identical semantics with
12939 identical syntax. This is a great boon to those working with
12940 cross-platform projects.
12942 For those adding visibility support to existing code, you may find
12943 @code{#pragma GCC visibility} of use. This works by you enclosing
12944 the declarations you wish to set visibility for with (for example)
12945 @code{#pragma GCC visibility push(hidden)} and
12946 @code{#pragma GCC visibility pop}.
12947 Bear in mind that symbol visibility should be viewed @strong{as
12948 part of the API interface contract} and thus all new code should
12949 always specify visibility when it is not the default; i.e., declarations
12950 only for use within the local DSO should @strong{always} be marked explicitly
12951 as hidden as so to avoid PLT indirection overheads---making this
12952 abundantly clear also aids readability and self-documentation of the code.
12953 Note that due to ISO C++ specification requirements, @code{operator new} and
12954 @code{operator delete} must always be of default visibility.
12956 Be aware that headers from outside your project, in particular system
12957 headers and headers from any other library you use, may not be
12958 expecting to be compiled with visibility other than the default. You
12959 may need to explicitly say @code{#pragma GCC visibility push(default)}
12960 before including any such headers.
12962 @code{extern} declarations are not affected by @option{-fvisibility}, so
12963 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12964 no modifications. However, this means that calls to @code{extern}
12965 functions with no explicit visibility use the PLT, so it is more
12966 effective to use @code{__attribute ((visibility))} and/or
12967 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12968 declarations should be treated as hidden.
12970 Note that @option{-fvisibility} does affect C++ vague linkage
12971 entities. This means that, for instance, an exception class that is
12972 be thrown between DSOs must be explicitly marked with default
12973 visibility so that the @samp{type_info} nodes are unified between
12976 An overview of these techniques, their benefits and how to use them
12977 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12979 @item -fstrict-volatile-bitfields
12980 @opindex fstrict-volatile-bitfields
12981 This option should be used if accesses to volatile bit-fields (or other
12982 structure fields, although the compiler usually honors those types
12983 anyway) should use a single access of the width of the
12984 field's type, aligned to a natural alignment if possible. For
12985 example, targets with memory-mapped peripheral registers might require
12986 all such accesses to be 16 bits wide; with this flag you can
12987 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12988 is 16 bits on these targets) to force GCC to use 16-bit accesses
12989 instead of, perhaps, a more efficient 32-bit access.
12991 If this option is disabled, the compiler uses the most efficient
12992 instruction. In the previous example, that might be a 32-bit load
12993 instruction, even though that accesses bytes that do not contain
12994 any portion of the bit-field, or memory-mapped registers unrelated to
12995 the one being updated.
12997 In some cases, such as when the @code{packed} attribute is applied to a
12998 structure field, it may not be possible to access the field with a single
12999 read or write that is correctly aligned for the target machine. In this
13000 case GCC falls back to generating multiple accesses rather than code that
13001 will fault or truncate the result at run time.
13003 Note: Due to restrictions of the C/C++11 memory model, write accesses are
13004 not allowed to touch non bit-field members. It is therefore recommended
13005 to define all bits of the field's type as bit-field members.
13007 The default value of this option is determined by the application binary
13008 interface for the target processor.
13010 @item -fsync-libcalls
13011 @opindex fsync-libcalls
13012 This option controls whether any out-of-line instance of the @code{__sync}
13013 family of functions may be used to implement the C++11 @code{__atomic}
13014 family of functions.
13016 The default value of this option is enabled, thus the only useful form
13017 of the option is @option{-fno-sync-libcalls}. This option is used in
13018 the implementation of the @file{libatomic} runtime library.
13022 @node Developer Options
13023 @section GCC Developer Options
13024 @cindex developer options
13025 @cindex debugging GCC
13026 @cindex debug dump options
13027 @cindex dump options
13028 @cindex compilation statistics
13030 This section describes command-line options that are primarily of
13031 interest to GCC developers, including options to support compiler
13032 testing and investigation of compiler bugs and compile-time
13033 performance problems. This includes options that produce debug dumps
13034 at various points in the compilation; that print statistics such as
13035 memory use and execution time; and that print information about GCC's
13036 configuration, such as where it searches for libraries. You should
13037 rarely need to use any of these options for ordinary compilation and
13042 @item -d@var{letters}
13043 @itemx -fdump-rtl-@var{pass}
13044 @itemx -fdump-rtl-@var{pass}=@var{filename}
13046 @opindex fdump-rtl-@var{pass}
13047 Says to make debugging dumps during compilation at times specified by
13048 @var{letters}. This is used for debugging the RTL-based passes of the
13049 compiler. The file names for most of the dumps are made by appending
13050 a pass number and a word to the @var{dumpname}, and the files are
13051 created in the directory of the output file. In case of
13052 @option{=@var{filename}} option, the dump is output on the given file
13053 instead of the pass numbered dump files. Note that the pass number is
13054 assigned as passes are registered into the pass manager. Most passes
13055 are registered in the order that they will execute and for these passes
13056 the number corresponds to the pass execution order. However, passes
13057 registered by plugins, passes specific to compilation targets, or
13058 passes that are otherwise registered after all the other passes are
13059 numbered higher than a pass named "final", even if they are executed
13060 earlier. @var{dumpname} is generated from the name of the output
13061 file if explicitly specified and not an executable, otherwise it is
13062 the basename of the source file.
13064 Some @option{-d@var{letters}} switches have different meaning when
13065 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
13066 for information about preprocessor-specific dump options.
13068 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
13069 @option{-d} option @var{letters}. Here are the possible
13070 letters for use in @var{pass} and @var{letters}, and their meanings:
13074 @item -fdump-rtl-alignments
13075 @opindex fdump-rtl-alignments
13076 Dump after branch alignments have been computed.
13078 @item -fdump-rtl-asmcons
13079 @opindex fdump-rtl-asmcons
13080 Dump after fixing rtl statements that have unsatisfied in/out constraints.
13082 @item -fdump-rtl-auto_inc_dec
13083 @opindex fdump-rtl-auto_inc_dec
13084 Dump after auto-inc-dec discovery. This pass is only run on
13085 architectures that have auto inc or auto dec instructions.
13087 @item -fdump-rtl-barriers
13088 @opindex fdump-rtl-barriers
13089 Dump after cleaning up the barrier instructions.
13091 @item -fdump-rtl-bbpart
13092 @opindex fdump-rtl-bbpart
13093 Dump after partitioning hot and cold basic blocks.
13095 @item -fdump-rtl-bbro
13096 @opindex fdump-rtl-bbro
13097 Dump after block reordering.
13099 @item -fdump-rtl-btl1
13100 @itemx -fdump-rtl-btl2
13101 @opindex fdump-rtl-btl2
13102 @opindex fdump-rtl-btl2
13103 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
13104 after the two branch
13105 target load optimization passes.
13107 @item -fdump-rtl-bypass
13108 @opindex fdump-rtl-bypass
13109 Dump after jump bypassing and control flow optimizations.
13111 @item -fdump-rtl-combine
13112 @opindex fdump-rtl-combine
13113 Dump after the RTL instruction combination pass.
13115 @item -fdump-rtl-compgotos
13116 @opindex fdump-rtl-compgotos
13117 Dump after duplicating the computed gotos.
13119 @item -fdump-rtl-ce1
13120 @itemx -fdump-rtl-ce2
13121 @itemx -fdump-rtl-ce3
13122 @opindex fdump-rtl-ce1
13123 @opindex fdump-rtl-ce2
13124 @opindex fdump-rtl-ce3
13125 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
13126 @option{-fdump-rtl-ce3} enable dumping after the three
13127 if conversion passes.
13129 @item -fdump-rtl-cprop_hardreg
13130 @opindex fdump-rtl-cprop_hardreg
13131 Dump after hard register copy propagation.
13133 @item -fdump-rtl-csa
13134 @opindex fdump-rtl-csa
13135 Dump after combining stack adjustments.
13137 @item -fdump-rtl-cse1
13138 @itemx -fdump-rtl-cse2
13139 @opindex fdump-rtl-cse1
13140 @opindex fdump-rtl-cse2
13141 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
13142 the two common subexpression elimination passes.
13144 @item -fdump-rtl-dce
13145 @opindex fdump-rtl-dce
13146 Dump after the standalone dead code elimination passes.
13148 @item -fdump-rtl-dbr
13149 @opindex fdump-rtl-dbr
13150 Dump after delayed branch scheduling.
13152 @item -fdump-rtl-dce1
13153 @itemx -fdump-rtl-dce2
13154 @opindex fdump-rtl-dce1
13155 @opindex fdump-rtl-dce2
13156 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
13157 the two dead store elimination passes.
13159 @item -fdump-rtl-eh
13160 @opindex fdump-rtl-eh
13161 Dump after finalization of EH handling code.
13163 @item -fdump-rtl-eh_ranges
13164 @opindex fdump-rtl-eh_ranges
13165 Dump after conversion of EH handling range regions.
13167 @item -fdump-rtl-expand
13168 @opindex fdump-rtl-expand
13169 Dump after RTL generation.
13171 @item -fdump-rtl-fwprop1
13172 @itemx -fdump-rtl-fwprop2
13173 @opindex fdump-rtl-fwprop1
13174 @opindex fdump-rtl-fwprop2
13175 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
13176 dumping after the two forward propagation passes.
13178 @item -fdump-rtl-gcse1
13179 @itemx -fdump-rtl-gcse2
13180 @opindex fdump-rtl-gcse1
13181 @opindex fdump-rtl-gcse2
13182 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
13183 after global common subexpression elimination.
13185 @item -fdump-rtl-init-regs
13186 @opindex fdump-rtl-init-regs
13187 Dump after the initialization of the registers.
13189 @item -fdump-rtl-initvals
13190 @opindex fdump-rtl-initvals
13191 Dump after the computation of the initial value sets.
13193 @item -fdump-rtl-into_cfglayout
13194 @opindex fdump-rtl-into_cfglayout
13195 Dump after converting to cfglayout mode.
13197 @item -fdump-rtl-ira
13198 @opindex fdump-rtl-ira
13199 Dump after iterated register allocation.
13201 @item -fdump-rtl-jump
13202 @opindex fdump-rtl-jump
13203 Dump after the second jump optimization.
13205 @item -fdump-rtl-loop2
13206 @opindex fdump-rtl-loop2
13207 @option{-fdump-rtl-loop2} enables dumping after the rtl
13208 loop optimization passes.
13210 @item -fdump-rtl-mach
13211 @opindex fdump-rtl-mach
13212 Dump after performing the machine dependent reorganization pass, if that
13215 @item -fdump-rtl-mode_sw
13216 @opindex fdump-rtl-mode_sw
13217 Dump after removing redundant mode switches.
13219 @item -fdump-rtl-rnreg
13220 @opindex fdump-rtl-rnreg
13221 Dump after register renumbering.
13223 @item -fdump-rtl-outof_cfglayout
13224 @opindex fdump-rtl-outof_cfglayout
13225 Dump after converting from cfglayout mode.
13227 @item -fdump-rtl-peephole2
13228 @opindex fdump-rtl-peephole2
13229 Dump after the peephole pass.
13231 @item -fdump-rtl-postreload
13232 @opindex fdump-rtl-postreload
13233 Dump after post-reload optimizations.
13235 @item -fdump-rtl-pro_and_epilogue
13236 @opindex fdump-rtl-pro_and_epilogue
13237 Dump after generating the function prologues and epilogues.
13239 @item -fdump-rtl-sched1
13240 @itemx -fdump-rtl-sched2
13241 @opindex fdump-rtl-sched1
13242 @opindex fdump-rtl-sched2
13243 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
13244 after the basic block scheduling passes.
13246 @item -fdump-rtl-ree
13247 @opindex fdump-rtl-ree
13248 Dump after sign/zero extension elimination.
13250 @item -fdump-rtl-seqabstr
13251 @opindex fdump-rtl-seqabstr
13252 Dump after common sequence discovery.
13254 @item -fdump-rtl-shorten
13255 @opindex fdump-rtl-shorten
13256 Dump after shortening branches.
13258 @item -fdump-rtl-sibling
13259 @opindex fdump-rtl-sibling
13260 Dump after sibling call optimizations.
13262 @item -fdump-rtl-split1
13263 @itemx -fdump-rtl-split2
13264 @itemx -fdump-rtl-split3
13265 @itemx -fdump-rtl-split4
13266 @itemx -fdump-rtl-split5
13267 @opindex fdump-rtl-split1
13268 @opindex fdump-rtl-split2
13269 @opindex fdump-rtl-split3
13270 @opindex fdump-rtl-split4
13271 @opindex fdump-rtl-split5
13272 These options enable dumping after five rounds of
13273 instruction splitting.
13275 @item -fdump-rtl-sms
13276 @opindex fdump-rtl-sms
13277 Dump after modulo scheduling. This pass is only run on some
13280 @item -fdump-rtl-stack
13281 @opindex fdump-rtl-stack
13282 Dump after conversion from GCC's ``flat register file'' registers to the
13283 x87's stack-like registers. This pass is only run on x86 variants.
13285 @item -fdump-rtl-subreg1
13286 @itemx -fdump-rtl-subreg2
13287 @opindex fdump-rtl-subreg1
13288 @opindex fdump-rtl-subreg2
13289 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
13290 the two subreg expansion passes.
13292 @item -fdump-rtl-unshare
13293 @opindex fdump-rtl-unshare
13294 Dump after all rtl has been unshared.
13296 @item -fdump-rtl-vartrack
13297 @opindex fdump-rtl-vartrack
13298 Dump after variable tracking.
13300 @item -fdump-rtl-vregs
13301 @opindex fdump-rtl-vregs
13302 Dump after converting virtual registers to hard registers.
13304 @item -fdump-rtl-web
13305 @opindex fdump-rtl-web
13306 Dump after live range splitting.
13308 @item -fdump-rtl-regclass
13309 @itemx -fdump-rtl-subregs_of_mode_init
13310 @itemx -fdump-rtl-subregs_of_mode_finish
13311 @itemx -fdump-rtl-dfinit
13312 @itemx -fdump-rtl-dfinish
13313 @opindex fdump-rtl-regclass
13314 @opindex fdump-rtl-subregs_of_mode_init
13315 @opindex fdump-rtl-subregs_of_mode_finish
13316 @opindex fdump-rtl-dfinit
13317 @opindex fdump-rtl-dfinish
13318 These dumps are defined but always produce empty files.
13321 @itemx -fdump-rtl-all
13323 @opindex fdump-rtl-all
13324 Produce all the dumps listed above.
13328 Annotate the assembler output with miscellaneous debugging information.
13332 Dump all macro definitions, at the end of preprocessing, in addition to
13337 Produce a core dump whenever an error occurs.
13341 Annotate the assembler output with a comment indicating which
13342 pattern and alternative is used. The length and cost of each instruction are
13347 Dump the RTL in the assembler output as a comment before each instruction.
13348 Also turns on @option{-dp} annotation.
13352 Just generate RTL for a function instead of compiling it. Usually used
13353 with @option{-fdump-rtl-expand}.
13356 @item -fdump-noaddr
13357 @opindex fdump-noaddr
13358 When doing debugging dumps, suppress address output. This makes it more
13359 feasible to use diff on debugging dumps for compiler invocations with
13360 different compiler binaries and/or different
13361 text / bss / data / heap / stack / dso start locations.
13364 @opindex freport-bug
13365 Collect and dump debug information into a temporary file if an
13366 internal compiler error (ICE) occurs.
13368 @item -fdump-unnumbered
13369 @opindex fdump-unnumbered
13370 When doing debugging dumps, suppress instruction numbers and address output.
13371 This makes it more feasible to use diff on debugging dumps for compiler
13372 invocations with different options, in particular with and without
13375 @item -fdump-unnumbered-links
13376 @opindex fdump-unnumbered-links
13377 When doing debugging dumps (see @option{-d} option above), suppress
13378 instruction numbers for the links to the previous and next instructions
13381 @item -fdump-ipa-@var{switch}
13383 Control the dumping at various stages of inter-procedural analysis
13384 language tree to a file. The file name is generated by appending a
13385 switch specific suffix to the source file name, and the file is created
13386 in the same directory as the output file. The following dumps are
13391 Enables all inter-procedural analysis dumps.
13394 Dumps information about call-graph optimization, unused function removal,
13395 and inlining decisions.
13398 Dump after function inlining.
13402 @item -fdump-lang-all
13403 @itemx -fdump-lang-@var{switch}
13404 @itemx -fdump-lang-@var{switch}-@var{options}
13405 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
13406 @opindex fdump-lang-all
13407 @opindex fdump-lang
13408 Control the dumping of language-specific information. The @var{options}
13409 and @var{filename} portions behave as described in the
13410 @option{-fdump-tree} option. The following @var{switch} values are
13416 Enable all language-specific dumps.
13419 Dump class hierarchy information. Virtual table information is emitted
13420 unless '@option{slim}' is specified. This option is applicable to C++ only.
13423 Dump the raw internal tree data. This option is applicable to C++ only.
13427 @item -fdump-passes
13428 @opindex fdump-passes
13429 Print on @file{stderr} the list of optimization passes that are turned
13430 on and off by the current command-line options.
13432 @item -fdump-statistics-@var{option}
13433 @opindex fdump-statistics
13434 Enable and control dumping of pass statistics in a separate file. The
13435 file name is generated by appending a suffix ending in
13436 @samp{.statistics} to the source file name, and the file is created in
13437 the same directory as the output file. If the @samp{-@var{option}}
13438 form is used, @samp{-stats} causes counters to be summed over the
13439 whole compilation unit while @samp{-details} dumps every event as
13440 the passes generate them. The default with no option is to sum
13441 counters for each function compiled.
13443 @item -fdump-tree-all
13444 @itemx -fdump-tree-@var{switch}
13445 @itemx -fdump-tree-@var{switch}-@var{options}
13446 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
13447 @opindex fdump-tree-all
13448 @opindex fdump-tree
13449 Control the dumping at various stages of processing the intermediate
13450 language tree to a file. The file name is generated by appending a
13451 switch-specific suffix to the source file name, and the file is
13452 created in the same directory as the output file. In case of
13453 @option{=@var{filename}} option, the dump is output on the given file
13454 instead of the auto named dump files. If the @samp{-@var{options}}
13455 form is used, @var{options} is a list of @samp{-} separated options
13456 which control the details of the dump. Not all options are applicable
13457 to all dumps; those that are not meaningful are ignored. The
13458 following options are available
13462 Print the address of each node. Usually this is not meaningful as it
13463 changes according to the environment and source file. Its primary use
13464 is for tying up a dump file with a debug environment.
13466 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
13467 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
13468 use working backward from mangled names in the assembly file.
13470 When dumping front-end intermediate representations, inhibit dumping
13471 of members of a scope or body of a function merely because that scope
13472 has been reached. Only dump such items when they are directly reachable
13473 by some other path.
13475 When dumping pretty-printed trees, this option inhibits dumping the
13476 bodies of control structures.
13478 When dumping RTL, print the RTL in slim (condensed) form instead of
13479 the default LISP-like representation.
13481 Print a raw representation of the tree. By default, trees are
13482 pretty-printed into a C-like representation.
13484 Enable more detailed dumps (not honored by every dump option). Also
13485 include information from the optimization passes.
13487 Enable dumping various statistics about the pass (not honored by every dump
13490 Enable showing basic block boundaries (disabled in raw dumps).
13492 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13493 dump a representation of the control flow graph suitable for viewing with
13494 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13495 the file is pretty-printed as a subgraph, so that GraphViz can render them
13496 all in a single plot.
13498 This option currently only works for RTL dumps, and the RTL is always
13499 dumped in slim form.
13501 Enable showing virtual operands for every statement.
13503 Enable showing line numbers for statements.
13505 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13507 Enable showing the tree dump for each statement.
13509 Enable showing the EH region number holding each statement.
13511 Enable showing scalar evolution analysis details.
13513 Enable showing optimization information (only available in certain
13516 Enable showing missed optimization information (only available in certain
13519 Enable other detailed optimization information (only available in
13521 @item =@var{filename}
13522 Instead of an auto named dump file, output into the given file
13523 name. The file names @file{stdout} and @file{stderr} are treated
13524 specially and are considered already open standard streams. For
13528 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13529 -fdump-tree-pre=/dev/stderr file.c
13532 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13533 output on to @file{stderr}. If two conflicting dump filenames are
13534 given for the same pass, then the latter option overrides the earlier
13538 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13539 and @option{lineno}.
13542 Turn on all optimization options, i.e., @option{optimized},
13543 @option{missed}, and @option{note}.
13546 To determine what tree dumps are available or find the dump for a pass
13547 of interest follow the steps below.
13551 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
13552 look for a code that corresponds to the pass you are interested in.
13553 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
13554 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
13555 The number at the end distinguishes distinct invocations of the same pass.
13557 To enable the creation of the dump file, append the pass code to
13558 the @option{-fdump-} option prefix and invoke GCC with it. For example,
13559 to enable the dump from the Early Value Range Propagation pass, invoke
13560 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
13561 specify the name of the dump file. If you don't specify one, GCC
13562 creates as described below.
13564 Find the pass dump in a file whose name is composed of three components
13565 separated by a period: the name of the source file GCC was invoked to
13566 compile, a numeric suffix indicating the pass number followed by the
13567 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
13568 and finally the pass code. For example, the Early VRP pass dump might
13569 be in a file named @file{myfile.c.038t.evrp} in the current working
13570 directory. Note that the numeric codes are not stable and may change
13571 from one version of GCC to another.
13575 @itemx -fopt-info-@var{options}
13576 @itemx -fopt-info-@var{options}=@var{filename}
13578 Controls optimization dumps from various optimization passes. If the
13579 @samp{-@var{options}} form is used, @var{options} is a list of
13580 @samp{-} separated option keywords to select the dump details and
13583 The @var{options} can be divided into two groups: options describing the
13584 verbosity of the dump, and options describing which optimizations
13585 should be included. The options from both the groups can be freely
13586 mixed as they are non-overlapping. However, in case of any conflicts,
13587 the later options override the earlier options on the command
13590 The following options control the dump verbosity:
13594 Print information when an optimization is successfully applied. It is
13595 up to a pass to decide which information is relevant. For example, the
13596 vectorizer passes print the source location of loops which are
13597 successfully vectorized.
13599 Print information about missed optimizations. Individual passes
13600 control which information to include in the output.
13602 Print verbose information about optimizations, such as certain
13603 transformations, more detailed messages about decisions etc.
13605 Print detailed optimization information. This includes
13606 @samp{optimized}, @samp{missed}, and @samp{note}.
13609 One or more of the following option keywords can be used to describe a
13610 group of optimizations:
13614 Enable dumps from all interprocedural optimizations.
13616 Enable dumps from all loop optimizations.
13618 Enable dumps from all inlining optimizations.
13620 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
13622 Enable dumps from all vectorization optimizations.
13624 Enable dumps from all optimizations. This is a superset of
13625 the optimization groups listed above.
13628 If @var{options} is
13629 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13630 info about successful optimizations from all the passes.
13632 If the @var{filename} is provided, then the dumps from all the
13633 applicable optimizations are concatenated into the @var{filename}.
13634 Otherwise the dump is output onto @file{stderr}. Though multiple
13635 @option{-fopt-info} options are accepted, only one of them can include
13636 a @var{filename}. If other filenames are provided then all but the
13637 first such option are ignored.
13639 Note that the output @var{filename} is overwritten
13640 in case of multiple translation units. If a combined output from
13641 multiple translation units is desired, @file{stderr} should be used
13644 In the following example, the optimization info is output to
13653 gcc -O3 -fopt-info-missed=missed.all
13657 outputs missed optimization report from all the passes into
13658 @file{missed.all}, and this one:
13661 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13665 prints information about missed optimization opportunities from
13666 vectorization passes on @file{stderr}.
13667 Note that @option{-fopt-info-vec-missed} is equivalent to
13668 @option{-fopt-info-missed-vec}. The order of the optimization group
13669 names and message types listed after @option{-fopt-info} does not matter.
13671 As another example,
13673 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13677 outputs information about missed optimizations as well as
13678 optimized locations from all the inlining passes into
13684 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13688 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13689 in conflict since only one output file is allowed. In this case, only
13690 the first option takes effect and the subsequent options are
13691 ignored. Thus only @file{vec.miss} is produced which contains
13692 dumps from the vectorizer about missed opportunities.
13694 @item -fsched-verbose=@var{n}
13695 @opindex fsched-verbose
13696 On targets that use instruction scheduling, this option controls the
13697 amount of debugging output the scheduler prints to the dump files.
13699 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13700 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13701 For @var{n} greater than one, it also output basic block probabilities,
13702 detailed ready list information and unit/insn info. For @var{n} greater
13703 than two, it includes RTL at abort point, control-flow and regions info.
13704 And for @var{n} over four, @option{-fsched-verbose} also includes
13709 @item -fenable-@var{kind}-@var{pass}
13710 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13714 This is a set of options that are used to explicitly disable/enable
13715 optimization passes. These options are intended for use for debugging GCC.
13716 Compiler users should use regular options for enabling/disabling
13721 @item -fdisable-ipa-@var{pass}
13722 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13723 statically invoked in the compiler multiple times, the pass name should be
13724 appended with a sequential number starting from 1.
13726 @item -fdisable-rtl-@var{pass}
13727 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13728 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13729 statically invoked in the compiler multiple times, the pass name should be
13730 appended with a sequential number starting from 1. @var{range-list} is a
13731 comma-separated list of function ranges or assembler names. Each range is a number
13732 pair separated by a colon. The range is inclusive in both ends. If the range
13733 is trivial, the number pair can be simplified as a single number. If the
13734 function's call graph node's @var{uid} falls within one of the specified ranges,
13735 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13736 function header of a dump file, and the pass names can be dumped by using
13737 option @option{-fdump-passes}.
13739 @item -fdisable-tree-@var{pass}
13740 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13741 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13744 @item -fenable-ipa-@var{pass}
13745 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13746 statically invoked in the compiler multiple times, the pass name should be
13747 appended with a sequential number starting from 1.
13749 @item -fenable-rtl-@var{pass}
13750 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13751 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13752 description and examples.
13754 @item -fenable-tree-@var{pass}
13755 @itemx -fenable-tree-@var{pass}=@var{range-list}
13756 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13757 of option arguments.
13761 Here are some examples showing uses of these options.
13765 # disable ccp1 for all functions
13766 -fdisable-tree-ccp1
13767 # disable complete unroll for function whose cgraph node uid is 1
13768 -fenable-tree-cunroll=1
13769 # disable gcse2 for functions at the following ranges [1,1],
13770 # [300,400], and [400,1000]
13771 # disable gcse2 for functions foo and foo2
13772 -fdisable-rtl-gcse2=foo,foo2
13773 # disable early inlining
13774 -fdisable-tree-einline
13775 # disable ipa inlining
13776 -fdisable-ipa-inline
13777 # enable tree full unroll
13778 -fenable-tree-unroll
13783 @itemx -fchecking=@var{n}
13785 @opindex fno-checking
13786 Enable internal consistency checking. The default depends on
13787 the compiler configuration. @option{-fchecking=2} enables further
13788 internal consistency checking that might affect code generation.
13790 @item -frandom-seed=@var{string}
13791 @opindex frandom-seed
13792 This option provides a seed that GCC uses in place of
13793 random numbers in generating certain symbol names
13794 that have to be different in every compiled file. It is also used to
13795 place unique stamps in coverage data files and the object files that
13796 produce them. You can use the @option{-frandom-seed} option to produce
13797 reproducibly identical object files.
13799 The @var{string} can either be a number (decimal, octal or hex) or an
13800 arbitrary string (in which case it's converted to a number by
13803 The @var{string} should be different for every file you compile.
13806 @itemx -save-temps=cwd
13807 @opindex save-temps
13808 Store the usual ``temporary'' intermediate files permanently; place them
13809 in the current directory and name them based on the source file. Thus,
13810 compiling @file{foo.c} with @option{-c -save-temps} produces files
13811 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13812 preprocessed @file{foo.i} output file even though the compiler now
13813 normally uses an integrated preprocessor.
13815 When used in combination with the @option{-x} command-line option,
13816 @option{-save-temps} is sensible enough to avoid over writing an
13817 input source file with the same extension as an intermediate file.
13818 The corresponding intermediate file may be obtained by renaming the
13819 source file before using @option{-save-temps}.
13821 If you invoke GCC in parallel, compiling several different source
13822 files that share a common base name in different subdirectories or the
13823 same source file compiled for multiple output destinations, it is
13824 likely that the different parallel compilers will interfere with each
13825 other, and overwrite the temporary files. For instance:
13828 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13829 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13832 may result in @file{foo.i} and @file{foo.o} being written to
13833 simultaneously by both compilers.
13835 @item -save-temps=obj
13836 @opindex save-temps=obj
13837 Store the usual ``temporary'' intermediate files permanently. If the
13838 @option{-o} option is used, the temporary files are based on the
13839 object file. If the @option{-o} option is not used, the
13840 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13845 gcc -save-temps=obj -c foo.c
13846 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13847 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13851 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13852 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13853 @file{dir2/yfoobar.o}.
13855 @item -time@r{[}=@var{file}@r{]}
13857 Report the CPU time taken by each subprocess in the compilation
13858 sequence. For C source files, this is the compiler proper and assembler
13859 (plus the linker if linking is done).
13861 Without the specification of an output file, the output looks like this:
13868 The first number on each line is the ``user time'', that is time spent
13869 executing the program itself. The second number is ``system time'',
13870 time spent executing operating system routines on behalf of the program.
13871 Both numbers are in seconds.
13873 With the specification of an output file, the output is appended to the
13874 named file, and it looks like this:
13877 0.12 0.01 cc1 @var{options}
13878 0.00 0.01 as @var{options}
13881 The ``user time'' and the ``system time'' are moved before the program
13882 name, and the options passed to the program are displayed, so that one
13883 can later tell what file was being compiled, and with which options.
13885 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13886 @opindex fdump-final-insns
13887 Dump the final internal representation (RTL) to @var{file}. If the
13888 optional argument is omitted (or if @var{file} is @code{.}), the name
13889 of the dump file is determined by appending @code{.gkd} to the
13890 compilation output file name.
13892 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13893 @opindex fcompare-debug
13894 @opindex fno-compare-debug
13895 If no error occurs during compilation, run the compiler a second time,
13896 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13897 passed to the second compilation. Dump the final internal
13898 representation in both compilations, and print an error if they differ.
13900 If the equal sign is omitted, the default @option{-gtoggle} is used.
13902 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13903 and nonzero, implicitly enables @option{-fcompare-debug}. If
13904 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13905 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13908 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13909 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13910 of the final representation and the second compilation, preventing even
13911 @env{GCC_COMPARE_DEBUG} from taking effect.
13913 To verify full coverage during @option{-fcompare-debug} testing, set
13914 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13915 which GCC rejects as an invalid option in any actual compilation
13916 (rather than preprocessing, assembly or linking). To get just a
13917 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13918 not overridden} will do.
13920 @item -fcompare-debug-second
13921 @opindex fcompare-debug-second
13922 This option is implicitly passed to the compiler for the second
13923 compilation requested by @option{-fcompare-debug}, along with options to
13924 silence warnings, and omitting other options that would cause
13925 side-effect compiler outputs to files or to the standard output. Dump
13926 files and preserved temporary files are renamed so as to contain the
13927 @code{.gk} additional extension during the second compilation, to avoid
13928 overwriting those generated by the first.
13930 When this option is passed to the compiler driver, it causes the
13931 @emph{first} compilation to be skipped, which makes it useful for little
13932 other than debugging the compiler proper.
13936 Turn off generation of debug info, if leaving out this option
13937 generates it, or turn it on at level 2 otherwise. The position of this
13938 argument in the command line does not matter; it takes effect after all
13939 other options are processed, and it does so only once, no matter how
13940 many times it is given. This is mainly intended to be used with
13941 @option{-fcompare-debug}.
13943 @item -fvar-tracking-assignments-toggle
13944 @opindex fvar-tracking-assignments-toggle
13945 @opindex fno-var-tracking-assignments-toggle
13946 Toggle @option{-fvar-tracking-assignments}, in the same way that
13947 @option{-gtoggle} toggles @option{-g}.
13951 Makes the compiler print out each function name as it is compiled, and
13952 print some statistics about each pass when it finishes.
13954 @item -ftime-report
13955 @opindex ftime-report
13956 Makes the compiler print some statistics about the time consumed by each
13957 pass when it finishes.
13959 @item -ftime-report-details
13960 @opindex ftime-report-details
13961 Record the time consumed by infrastructure parts separately for each pass.
13963 @item -fira-verbose=@var{n}
13964 @opindex fira-verbose
13965 Control the verbosity of the dump file for the integrated register allocator.
13966 The default value is 5. If the value @var{n} is greater or equal to 10,
13967 the dump output is sent to stderr using the same format as @var{n} minus 10.
13970 @opindex flto-report
13971 Prints a report with internal details on the workings of the link-time
13972 optimizer. The contents of this report vary from version to version.
13973 It is meant to be useful to GCC developers when processing object
13974 files in LTO mode (via @option{-flto}).
13976 Disabled by default.
13978 @item -flto-report-wpa
13979 @opindex flto-report-wpa
13980 Like @option{-flto-report}, but only print for the WPA phase of Link
13984 @opindex fmem-report
13985 Makes the compiler print some statistics about permanent memory
13986 allocation when it finishes.
13988 @item -fmem-report-wpa
13989 @opindex fmem-report-wpa
13990 Makes the compiler print some statistics about permanent memory
13991 allocation for the WPA phase only.
13993 @item -fpre-ipa-mem-report
13994 @opindex fpre-ipa-mem-report
13995 @item -fpost-ipa-mem-report
13996 @opindex fpost-ipa-mem-report
13997 Makes the compiler print some statistics about permanent memory
13998 allocation before or after interprocedural optimization.
14000 @item -fprofile-report
14001 @opindex fprofile-report
14002 Makes the compiler print some statistics about consistency of the
14003 (estimated) profile and effect of individual passes.
14005 @item -fstack-usage
14006 @opindex fstack-usage
14007 Makes the compiler output stack usage information for the program, on a
14008 per-function basis. The filename for the dump is made by appending
14009 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
14010 the output file, if explicitly specified and it is not an executable,
14011 otherwise it is the basename of the source file. An entry is made up
14016 The name of the function.
14020 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
14023 The qualifier @code{static} means that the function manipulates the stack
14024 statically: a fixed number of bytes are allocated for the frame on function
14025 entry and released on function exit; no stack adjustments are otherwise made
14026 in the function. The second field is this fixed number of bytes.
14028 The qualifier @code{dynamic} means that the function manipulates the stack
14029 dynamically: in addition to the static allocation described above, stack
14030 adjustments are made in the body of the function, for example to push/pop
14031 arguments around function calls. If the qualifier @code{bounded} is also
14032 present, the amount of these adjustments is bounded at compile time and
14033 the second field is an upper bound of the total amount of stack used by
14034 the function. If it is not present, the amount of these adjustments is
14035 not bounded at compile time and the second field only represents the
14040 Emit statistics about front-end processing at the end of the compilation.
14041 This option is supported only by the C++ front end, and
14042 the information is generally only useful to the G++ development team.
14044 @item -fdbg-cnt-list
14045 @opindex fdbg-cnt-list
14046 Print the name and the counter upper bound for all debug counters.
14049 @item -fdbg-cnt=@var{counter-value-list}
14051 Set the internal debug counter upper bound. @var{counter-value-list}
14052 is a comma-separated list of @var{name}:@var{value} pairs
14053 which sets the upper bound of each debug counter @var{name} to @var{value}.
14054 All debug counters have the initial upper bound of @code{UINT_MAX};
14055 thus @code{dbg_cnt} returns true always unless the upper bound
14056 is set by this option.
14057 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
14058 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
14060 @item -print-file-name=@var{library}
14061 @opindex print-file-name
14062 Print the full absolute name of the library file @var{library} that
14063 would be used when linking---and don't do anything else. With this
14064 option, GCC does not compile or link anything; it just prints the
14067 @item -print-multi-directory
14068 @opindex print-multi-directory
14069 Print the directory name corresponding to the multilib selected by any
14070 other switches present in the command line. This directory is supposed
14071 to exist in @env{GCC_EXEC_PREFIX}.
14073 @item -print-multi-lib
14074 @opindex print-multi-lib
14075 Print the mapping from multilib directory names to compiler switches
14076 that enable them. The directory name is separated from the switches by
14077 @samp{;}, and each switch starts with an @samp{@@} instead of the
14078 @samp{-}, without spaces between multiple switches. This is supposed to
14079 ease shell processing.
14081 @item -print-multi-os-directory
14082 @opindex print-multi-os-directory
14083 Print the path to OS libraries for the selected
14084 multilib, relative to some @file{lib} subdirectory. If OS libraries are
14085 present in the @file{lib} subdirectory and no multilibs are used, this is
14086 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
14087 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
14088 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
14089 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
14091 @item -print-multiarch
14092 @opindex print-multiarch
14093 Print the path to OS libraries for the selected multiarch,
14094 relative to some @file{lib} subdirectory.
14096 @item -print-prog-name=@var{program}
14097 @opindex print-prog-name
14098 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
14100 @item -print-libgcc-file-name
14101 @opindex print-libgcc-file-name
14102 Same as @option{-print-file-name=libgcc.a}.
14104 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
14105 but you do want to link with @file{libgcc.a}. You can do:
14108 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
14111 @item -print-search-dirs
14112 @opindex print-search-dirs
14113 Print the name of the configured installation directory and a list of
14114 program and library directories @command{gcc} searches---and don't do anything else.
14116 This is useful when @command{gcc} prints the error message
14117 @samp{installation problem, cannot exec cpp0: No such file or directory}.
14118 To resolve this you either need to put @file{cpp0} and the other compiler
14119 components where @command{gcc} expects to find them, or you can set the environment
14120 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
14121 Don't forget the trailing @samp{/}.
14122 @xref{Environment Variables}.
14124 @item -print-sysroot
14125 @opindex print-sysroot
14126 Print the target sysroot directory that is used during
14127 compilation. This is the target sysroot specified either at configure
14128 time or using the @option{--sysroot} option, possibly with an extra
14129 suffix that depends on compilation options. If no target sysroot is
14130 specified, the option prints nothing.
14132 @item -print-sysroot-headers-suffix
14133 @opindex print-sysroot-headers-suffix
14134 Print the suffix added to the target sysroot when searching for
14135 headers, or give an error if the compiler is not configured with such
14136 a suffix---and don't do anything else.
14139 @opindex dumpmachine
14140 Print the compiler's target machine (for example,
14141 @samp{i686-pc-linux-gnu})---and don't do anything else.
14144 @opindex dumpversion
14145 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
14146 anything else. This is the compiler version used in filesystem paths,
14147 specs, can be depending on how the compiler has been configured just
14148 a single number (major version), two numbers separated by dot (major and
14149 minor version) or three numbers separated by dots (major, minor and patchlevel
14152 @item -dumpfullversion
14153 @opindex dumpfullversion
14154 Print the full compiler version, always 3 numbers separated by dots,
14155 major, minor and patchlevel version.
14159 Print the compiler's built-in specs---and don't do anything else. (This
14160 is used when GCC itself is being built.) @xref{Spec Files}.
14163 @node Submodel Options
14164 @section Machine-Dependent Options
14165 @cindex submodel options
14166 @cindex specifying hardware config
14167 @cindex hardware models and configurations, specifying
14168 @cindex target-dependent options
14169 @cindex machine-dependent options
14171 Each target machine supported by GCC can have its own options---for
14172 example, to allow you to compile for a particular processor variant or
14173 ABI, or to control optimizations specific to that machine. By
14174 convention, the names of machine-specific options start with
14177 Some configurations of the compiler also support additional target-specific
14178 options, usually for compatibility with other compilers on the same
14181 @c This list is ordered alphanumerically by subsection name.
14182 @c It should be the same order and spelling as these options are listed
14183 @c in Machine Dependent Options
14186 * AArch64 Options::
14187 * Adapteva Epiphany Options::
14191 * Blackfin Options::
14196 * DEC Alpha Options::
14200 * GNU/Linux Options::
14210 * MicroBlaze Options::
14213 * MN10300 Options::
14217 * Nios II Options::
14218 * Nvidia PTX Options::
14220 * picoChip Options::
14221 * PowerPC Options::
14224 * RS/6000 and PowerPC Options::
14226 * S/390 and zSeries Options::
14229 * Solaris 2 Options::
14232 * System V Options::
14233 * TILE-Gx Options::
14234 * TILEPro Options::
14239 * VxWorks Options::
14241 * x86 Windows Options::
14242 * Xstormy16 Options::
14244 * zSeries Options::
14247 @node AArch64 Options
14248 @subsection AArch64 Options
14249 @cindex AArch64 Options
14251 These options are defined for AArch64 implementations:
14255 @item -mabi=@var{name}
14257 Generate code for the specified data model. Permissible values
14258 are @samp{ilp32} for SysV-like data model where int, long int and pointers
14259 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
14260 but long int and pointers are 64 bits.
14262 The default depends on the specific target configuration. Note that
14263 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
14264 entire program with the same ABI, and link with a compatible set of libraries.
14267 @opindex mbig-endian
14268 Generate big-endian code. This is the default when GCC is configured for an
14269 @samp{aarch64_be-*-*} target.
14271 @item -mgeneral-regs-only
14272 @opindex mgeneral-regs-only
14273 Generate code which uses only the general-purpose registers. This will prevent
14274 the compiler from using floating-point and Advanced SIMD registers but will not
14275 impose any restrictions on the assembler.
14277 @item -mlittle-endian
14278 @opindex mlittle-endian
14279 Generate little-endian code. This is the default when GCC is configured for an
14280 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
14282 @item -mcmodel=tiny
14283 @opindex mcmodel=tiny
14284 Generate code for the tiny code model. The program and its statically defined
14285 symbols must be within 1MB of each other. Programs can be statically or
14286 dynamically linked.
14288 @item -mcmodel=small
14289 @opindex mcmodel=small
14290 Generate code for the small code model. The program and its statically defined
14291 symbols must be within 4GB of each other. Programs can be statically or
14292 dynamically linked. This is the default code model.
14294 @item -mcmodel=large
14295 @opindex mcmodel=large
14296 Generate code for the large code model. This makes no assumptions about
14297 addresses and sizes of sections. Programs can be statically linked only.
14299 @item -mstrict-align
14300 @opindex mstrict-align
14301 Avoid generating memory accesses that may not be aligned on a natural object
14302 boundary as described in the architecture specification.
14304 @item -momit-leaf-frame-pointer
14305 @itemx -mno-omit-leaf-frame-pointer
14306 @opindex momit-leaf-frame-pointer
14307 @opindex mno-omit-leaf-frame-pointer
14308 Omit or keep the frame pointer in leaf functions. The former behavior is the
14311 @item -mtls-dialect=desc
14312 @opindex mtls-dialect=desc
14313 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
14314 of TLS variables. This is the default.
14316 @item -mtls-dialect=traditional
14317 @opindex mtls-dialect=traditional
14318 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
14321 @item -mtls-size=@var{size}
14323 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
14324 This option requires binutils 2.26 or newer.
14326 @item -mfix-cortex-a53-835769
14327 @itemx -mno-fix-cortex-a53-835769
14328 @opindex mfix-cortex-a53-835769
14329 @opindex mno-fix-cortex-a53-835769
14330 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
14331 This involves inserting a NOP instruction between memory instructions and
14332 64-bit integer multiply-accumulate instructions.
14334 @item -mfix-cortex-a53-843419
14335 @itemx -mno-fix-cortex-a53-843419
14336 @opindex mfix-cortex-a53-843419
14337 @opindex mno-fix-cortex-a53-843419
14338 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
14339 This erratum workaround is made at link time and this will only pass the
14340 corresponding flag to the linker.
14342 @item -mlow-precision-recip-sqrt
14343 @item -mno-low-precision-recip-sqrt
14344 @opindex mlow-precision-recip-sqrt
14345 @opindex mno-low-precision-recip-sqrt
14346 Enable or disable the reciprocal square root 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 reciprocal square root results to about 16 bits for
14350 single precision and to 32 bits for double precision.
14352 @item -mlow-precision-sqrt
14353 @item -mno-low-precision-sqrt
14354 @opindex -mlow-precision-sqrt
14355 @opindex -mno-low-precision-sqrt
14356 Enable or disable the square root approximation.
14357 This option only has an effect if @option{-ffast-math} or
14358 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14359 precision of square root results to about 16 bits for
14360 single precision and to 32 bits for double precision.
14361 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
14363 @item -mlow-precision-div
14364 @item -mno-low-precision-div
14365 @opindex -mlow-precision-div
14366 @opindex -mno-low-precision-div
14367 Enable or disable the division approximation.
14368 This option only has an effect if @option{-ffast-math} or
14369 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
14370 precision of division results to about 16 bits for
14371 single precision and to 32 bits for double precision.
14373 @item -march=@var{name}
14375 Specify the name of the target architecture and, optionally, one or
14376 more feature modifiers. This option has the form
14377 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
14379 The permissible values for @var{arch} are @samp{armv8-a},
14380 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @var{native}.
14382 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
14383 support for the ARMv8.3-A architecture extensions.
14385 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
14386 support for the ARMv8.2-A architecture extensions.
14388 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
14389 support for the ARMv8.1-A architecture extension. In particular, it
14390 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
14392 The value @samp{native} is available on native AArch64 GNU/Linux and
14393 causes the compiler to pick the architecture of the host system. This
14394 option has no effect if the compiler is unable to recognize the
14395 architecture of the host system,
14397 The permissible values for @var{feature} are listed in the sub-section
14398 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14399 Feature Modifiers}. Where conflicting feature modifiers are
14400 specified, the right-most feature is used.
14402 GCC uses @var{name} to determine what kind of instructions it can emit
14403 when generating assembly code. If @option{-march} is specified
14404 without either of @option{-mtune} or @option{-mcpu} also being
14405 specified, the code is tuned to perform well across a range of target
14406 processors implementing the target architecture.
14408 @item -mtune=@var{name}
14410 Specify the name of the target processor for which GCC should tune the
14411 performance of the code. Permissible values for this option are:
14412 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
14413 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
14414 @samp{exynos-m1}, @samp{falkor}, @samp{qdf24xx}, @samp{saphira},
14415 @samp{xgene1}, @samp{vulcan}, @samp{thunderx},
14416 @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
14417 @samp{thunderxt83}, @samp{thunderx2t99}, @samp{cortex-a57.cortex-a53},
14418 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
14419 @samp{cortex-a73.cortex-a53}, @samp{cortex-a75.cortex-a55},
14422 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14423 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
14424 @samp{cortex-a75.cortex-a55} specify that GCC should tune for a
14427 Additionally on native AArch64 GNU/Linux systems the value
14428 @samp{native} tunes performance to the host system. This option has no effect
14429 if the compiler is unable to recognize the processor of the host system.
14431 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
14432 are specified, the code is tuned to perform well across a range
14433 of target processors.
14435 This option cannot be suffixed by feature modifiers.
14437 @item -mcpu=@var{name}
14439 Specify the name of the target processor, optionally suffixed by one
14440 or more feature modifiers. This option has the form
14441 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
14442 the permissible values for @var{cpu} are the same as those available
14443 for @option{-mtune}. The permissible values for @var{feature} are
14444 documented in the sub-section on
14445 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14446 Feature Modifiers}. Where conflicting feature modifiers are
14447 specified, the right-most feature is used.
14449 GCC uses @var{name} to determine what kind of instructions it can emit when
14450 generating assembly code (as if by @option{-march}) and to determine
14451 the target processor for which to tune for performance (as if
14452 by @option{-mtune}). Where this option is used in conjunction
14453 with @option{-march} or @option{-mtune}, those options take precedence
14454 over the appropriate part of this option.
14456 @item -moverride=@var{string}
14458 Override tuning decisions made by the back-end in response to a
14459 @option{-mtune=} switch. The syntax, semantics, and accepted values
14460 for @var{string} in this option are not guaranteed to be consistent
14463 This option is only intended to be useful when developing GCC.
14465 @item -mpc-relative-literal-loads
14466 @itemx -mno-pc-relative-literal-loads
14467 @opindex mpc-relative-literal-loads
14468 @opindex mno-pc-relative-literal-loads
14469 Enable or disable PC-relative literal loads. With this option literal pools are
14470 accessed using a single instruction and emitted after each function. This
14471 limits the maximum size of functions to 1MB. This is enabled by default for
14472 @option{-mcmodel=tiny}.
14474 @item -msign-return-address=@var{scope}
14475 @opindex msign-return-address
14476 Select the function scope on which return address signing will be applied.
14477 Permissible values are @samp{none}, which disables return address signing,
14478 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
14479 functions, and @samp{all}, which enables pointer signing for all functions. The
14480 default value is @samp{none}.
14484 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14485 @anchor{aarch64-feature-modifiers}
14486 @cindex @option{-march} feature modifiers
14487 @cindex @option{-mcpu} feature modifiers
14488 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14489 the following and their inverses @option{no@var{feature}}:
14493 Enable CRC extension. This is on by default for
14494 @option{-march=armv8.1-a}.
14496 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14499 Enable floating-point instructions. This is on by default for all possible
14500 values for options @option{-march} and @option{-mcpu}.
14502 Enable Advanced SIMD instructions. This also enables floating-point
14503 instructions. This is on by default for all possible values for options
14504 @option{-march} and @option{-mcpu}.
14506 Enable Large System Extension instructions. This is on by default for
14507 @option{-march=armv8.1-a}.
14509 Enable Round Double Multiply Accumulate instructions. This is on by default
14510 for @option{-march=armv8.1-a}.
14512 Enable FP16 extension. This also enables floating-point instructions.
14514 Enable the RcPc extension. This does not change code generation from GCC,
14515 but is passed on to the assembler, enabling inline asm statements to use
14516 instructions from the RcPc extension.
14518 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
14522 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
14523 Conversely, @option{nofp} implies @option{nosimd}, which implies
14526 @node Adapteva Epiphany Options
14527 @subsection Adapteva Epiphany Options
14529 These @samp{-m} options are defined for Adapteva Epiphany:
14532 @item -mhalf-reg-file
14533 @opindex mhalf-reg-file
14534 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14535 That allows code to run on hardware variants that lack these registers.
14537 @item -mprefer-short-insn-regs
14538 @opindex mprefer-short-insn-regs
14539 Preferentially allocate registers that allow short instruction generation.
14540 This can result in increased instruction count, so this may either reduce or
14541 increase overall code size.
14543 @item -mbranch-cost=@var{num}
14544 @opindex mbranch-cost
14545 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14546 This cost is only a heuristic and is not guaranteed to produce
14547 consistent results across releases.
14551 Enable the generation of conditional moves.
14553 @item -mnops=@var{num}
14555 Emit @var{num} NOPs before every other generated instruction.
14557 @item -mno-soft-cmpsf
14558 @opindex mno-soft-cmpsf
14559 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14560 and test the flags. This is faster than a software comparison, but can
14561 get incorrect results in the presence of NaNs, or when two different small
14562 numbers are compared such that their difference is calculated as zero.
14563 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14564 software comparisons.
14566 @item -mstack-offset=@var{num}
14567 @opindex mstack-offset
14568 Set the offset between the top of the stack and the stack pointer.
14569 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14570 can be used by leaf functions without stack allocation.
14571 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14572 Note also that this option changes the ABI; compiling a program with a
14573 different stack offset than the libraries have been compiled with
14574 generally does not work.
14575 This option can be useful if you want to evaluate if a different stack
14576 offset would give you better code, but to actually use a different stack
14577 offset to build working programs, it is recommended to configure the
14578 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14580 @item -mno-round-nearest
14581 @opindex mno-round-nearest
14582 Make the scheduler assume that the rounding mode has been set to
14583 truncating. The default is @option{-mround-nearest}.
14586 @opindex mlong-calls
14587 If not otherwise specified by an attribute, assume all calls might be beyond
14588 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14589 function address into a register before performing a (otherwise direct) call.
14590 This is the default.
14592 @item -mshort-calls
14593 @opindex short-calls
14594 If not otherwise specified by an attribute, assume all direct calls are
14595 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14596 for direct calls. The default is @option{-mlong-calls}.
14600 Assume addresses can be loaded as 16-bit unsigned values. This does not
14601 apply to function addresses for which @option{-mlong-calls} semantics
14604 @item -mfp-mode=@var{mode}
14606 Set the prevailing mode of the floating-point unit.
14607 This determines the floating-point mode that is provided and expected
14608 at function call and return time. Making this mode match the mode you
14609 predominantly need at function start can make your programs smaller and
14610 faster by avoiding unnecessary mode switches.
14612 @var{mode} can be set to one the following values:
14616 Any mode at function entry is valid, and retained or restored when
14617 the function returns, and when it calls other functions.
14618 This mode is useful for compiling libraries or other compilation units
14619 you might want to incorporate into different programs with different
14620 prevailing FPU modes, and the convenience of being able to use a single
14621 object file outweighs the size and speed overhead for any extra
14622 mode switching that might be needed, compared with what would be needed
14623 with a more specific choice of prevailing FPU mode.
14626 This is the mode used for floating-point calculations with
14627 truncating (i.e.@: round towards zero) rounding mode. That includes
14628 conversion from floating point to integer.
14630 @item round-nearest
14631 This is the mode used for floating-point calculations with
14632 round-to-nearest-or-even rounding mode.
14635 This is the mode used to perform integer calculations in the FPU, e.g.@:
14636 integer multiply, or integer multiply-and-accumulate.
14639 The default is @option{-mfp-mode=caller}
14641 @item -mnosplit-lohi
14642 @itemx -mno-postinc
14643 @itemx -mno-postmodify
14644 @opindex mnosplit-lohi
14645 @opindex mno-postinc
14646 @opindex mno-postmodify
14647 Code generation tweaks that disable, respectively, splitting of 32-bit
14648 loads, generation of post-increment addresses, and generation of
14649 post-modify addresses. The defaults are @option{msplit-lohi},
14650 @option{-mpost-inc}, and @option{-mpost-modify}.
14652 @item -mnovect-double
14653 @opindex mno-vect-double
14654 Change the preferred SIMD mode to SImode. The default is
14655 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14657 @item -max-vect-align=@var{num}
14658 @opindex max-vect-align
14659 The maximum alignment for SIMD vector mode types.
14660 @var{num} may be 4 or 8. The default is 8.
14661 Note that this is an ABI change, even though many library function
14662 interfaces are unaffected if they don't use SIMD vector modes
14663 in places that affect size and/or alignment of relevant types.
14665 @item -msplit-vecmove-early
14666 @opindex msplit-vecmove-early
14667 Split vector moves into single word moves before reload. In theory this
14668 can give better register allocation, but so far the reverse seems to be
14669 generally the case.
14671 @item -m1reg-@var{reg}
14673 Specify a register to hold the constant @minus{}1, which makes loading small negative
14674 constants and certain bitmasks faster.
14675 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14676 which specify use of that register as a fixed register,
14677 and @samp{none}, which means that no register is used for this
14678 purpose. The default is @option{-m1reg-none}.
14683 @subsection ARC Options
14684 @cindex ARC options
14686 The following options control the architecture variant for which code
14689 @c architecture variants
14692 @item -mbarrel-shifter
14693 @opindex mbarrel-shifter
14694 Generate instructions supported by barrel shifter. This is the default
14695 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14697 @item -mcpu=@var{cpu}
14699 Set architecture type, register usage, and instruction scheduling
14700 parameters for @var{cpu}. There are also shortcut alias options
14701 available for backward compatibility and convenience. Supported
14702 values for @var{cpu} are
14708 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14712 Compile for ARC601. Alias: @option{-mARC601}.
14717 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14718 This is the default when configured with @option{--with-cpu=arc700}@.
14721 Compile for ARC EM.
14724 Compile for ARC HS.
14727 Compile for ARC EM CPU with no hardware extensions.
14730 Compile for ARC EM4 CPU.
14733 Compile for ARC EM4 DMIPS CPU.
14736 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
14740 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
14741 double assist instructions.
14744 Compile for ARC HS CPU with no hardware extensions except the atomic
14748 Compile for ARC HS34 CPU.
14751 Compile for ARC HS38 CPU.
14754 Compile for ARC HS38 CPU with all hardware extensions on.
14757 Compile for ARC 600 CPU with @code{norm} instructions enabled.
14759 @item arc600_mul32x16
14760 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
14761 instructions enabled.
14764 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
14765 instructions enabled.
14768 Compile for ARC 601 CPU with @code{norm} instructions enabled.
14770 @item arc601_mul32x16
14771 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
14772 instructions enabled.
14775 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
14776 instructions enabled.
14779 Compile for ARC 700 on NPS400 chip.
14785 @itemx -mdpfp-compact
14786 @opindex mdpfp-compact
14787 Generate double-precision FPX instructions, tuned for the compact
14791 @opindex mdpfp-fast
14792 Generate double-precision FPX instructions, tuned for the fast
14795 @item -mno-dpfp-lrsr
14796 @opindex mno-dpfp-lrsr
14797 Disable @code{lr} and @code{sr} instructions from using FPX extension
14802 Generate extended arithmetic instructions. Currently only
14803 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14804 supported. This is always enabled for @option{-mcpu=ARC700}.
14808 Do not generate @code{mpy}-family instructions for ARC700. This option is
14813 Generate 32x16-bit multiply and multiply-accumulate instructions.
14817 Generate @code{mul64} and @code{mulu64} instructions.
14818 Only valid for @option{-mcpu=ARC600}.
14822 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
14827 @itemx -mspfp-compact
14828 @opindex mspfp-compact
14829 Generate single-precision FPX instructions, tuned for the compact
14833 @opindex mspfp-fast
14834 Generate single-precision FPX instructions, tuned for the fast
14839 Enable generation of ARC SIMD instructions via target-specific
14840 builtins. Only valid for @option{-mcpu=ARC700}.
14843 @opindex msoft-float
14844 This option ignored; it is provided for compatibility purposes only.
14845 Software floating-point code is emitted by default, and this default
14846 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
14847 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
14848 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
14852 Generate @code{swap} instructions.
14856 This enables use of the locked load/store conditional extension to implement
14857 atomic memory built-in functions. Not available for ARC 6xx or ARC
14862 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
14864 @item -mcode-density
14865 @opindex mcode-density
14866 Enable code density instructions for ARC EM.
14867 This option is on by default for ARC HS.
14871 Enable double load/store operations for ARC HS cores.
14873 @item -mtp-regno=@var{regno}
14875 Specify thread pointer register number.
14877 @item -mmpy-option=@var{multo}
14878 @opindex mmpy-option
14879 Compile ARCv2 code with a multiplier design option. You can specify
14880 the option using either a string or numeric value for @var{multo}.
14881 @samp{wlh1} is the default value. The recognized values are:
14886 No multiplier available.
14890 16x16 multiplier, fully pipelined.
14891 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
14895 32x32 multiplier, fully
14896 pipelined (1 stage). The following instructions are additionally
14897 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14901 32x32 multiplier, fully pipelined
14902 (2 stages). The following instructions are additionally enabled: @code{mpy},
14903 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14907 Two 16x16 multipliers, blocking,
14908 sequential. The following instructions are additionally enabled: @code{mpy},
14909 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14913 One 16x16 multiplier, blocking,
14914 sequential. The following instructions are additionally enabled: @code{mpy},
14915 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14919 One 32x4 multiplier, blocking,
14920 sequential. The following instructions are additionally enabled: @code{mpy},
14921 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14925 ARC HS SIMD support.
14929 ARC HS SIMD support.
14933 ARC HS SIMD support.
14937 This option is only available for ARCv2 cores@.
14939 @item -mfpu=@var{fpu}
14941 Enables support for specific floating-point hardware extensions for ARCv2
14942 cores. Supported values for @var{fpu} are:
14947 Enables support for single-precision floating-point hardware
14951 Enables support for double-precision floating-point hardware
14952 extensions. The single-precision floating-point extension is also
14953 enabled. Not available for ARC EM@.
14956 Enables support for double-precision floating-point hardware
14957 extensions using double-precision assist instructions. The single-precision
14958 floating-point extension is also enabled. This option is
14959 only available for ARC EM@.
14962 Enables support for double-precision floating-point hardware
14963 extensions using double-precision assist instructions.
14964 The single-precision floating-point, square-root, and divide
14965 extensions are also enabled. This option is
14966 only available for ARC EM@.
14969 Enables support for double-precision floating-point hardware
14970 extensions using double-precision assist instructions.
14971 The single-precision floating-point and fused multiply and add
14972 hardware extensions are also enabled. This option is
14973 only available for ARC EM@.
14976 Enables support for double-precision floating-point hardware
14977 extensions using double-precision assist instructions.
14978 All single-precision floating-point hardware extensions are also
14979 enabled. This option is only available for ARC EM@.
14982 Enables support for single-precision floating-point, square-root and divide
14983 hardware extensions@.
14986 Enables support for double-precision floating-point, square-root and divide
14987 hardware extensions. This option
14988 includes option @samp{fpus_div}. Not available for ARC EM@.
14991 Enables support for single-precision floating-point and
14992 fused multiply and add hardware extensions@.
14995 Enables support for double-precision floating-point and
14996 fused multiply and add hardware extensions. This option
14997 includes option @samp{fpus_fma}. Not available for ARC EM@.
15000 Enables support for all single-precision floating-point hardware
15004 Enables support for all single- and double-precision floating-point
15005 hardware extensions. Not available for ARC EM@.
15009 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
15010 @opindex mirq-ctrl-saved
15011 Specifies general-purposes registers that the processor automatically
15012 saves/restores on interrupt entry and exit. @var{register-range} is
15013 specified as two registers separated by a dash. The register range
15014 always starts with @code{r0}, the upper limit is @code{fp} register.
15015 @var{blink} and @var{lp_count} are optional. This option is only
15016 valid for ARC EM and ARC HS cores.
15018 @item -mrgf-banked-regs=@var{number}
15019 @opindex mrgf-banked-regs
15020 Specifies the number of registers replicated in second register bank
15021 on entry to fast interrupt. Fast interrupts are interrupts with the
15022 highest priority level P0. These interrupts save only PC and STATUS32
15023 registers to avoid memory transactions during interrupt entry and exit
15024 sequences. Use this option when you are using fast interrupts in an
15025 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
15027 @item -mlpc-width=@var{width}
15028 @opindex mlpc-width
15029 Specify the width of the @code{lp_count} register. Valid values for
15030 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
15031 fixed to 32 bits. If the width is less than 32, the compiler does not
15032 attempt to transform loops in your program to use the zero-delay loop
15033 mechanism unless it is known that the @code{lp_count} register can
15034 hold the required loop-counter value. Depending on the width
15035 specified, the compiler and run-time library might continue to use the
15036 loop mechanism for various needs. This option defines macro
15037 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
15041 The following options are passed through to the assembler, and also
15042 define preprocessor macro symbols.
15044 @c Flags used by the assembler, but for which we define preprocessor
15045 @c macro symbols as well.
15048 @opindex mdsp-packa
15049 Passed down to the assembler to enable the DSP Pack A extensions.
15050 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
15055 Passed down to the assembler to enable the dual Viterbi butterfly
15056 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
15057 option is deprecated.
15059 @c ARC700 4.10 extension instruction
15062 Passed down to the assembler to enable the locked load/store
15063 conditional extension. Also sets the preprocessor symbol
15068 Passed down to the assembler. Also sets the preprocessor symbol
15069 @code{__Xxmac_d16}. This option is deprecated.
15073 Passed down to the assembler. Also sets the preprocessor symbol
15074 @code{__Xxmac_24}. This option is deprecated.
15076 @c ARC700 4.10 extension instruction
15079 Passed down to the assembler to enable the 64-bit time-stamp counter
15080 extension instruction. Also sets the preprocessor symbol
15081 @code{__Xrtsc}. This option is deprecated.
15083 @c ARC700 4.10 extension instruction
15086 Passed down to the assembler to enable the swap byte ordering
15087 extension instruction. Also sets the preprocessor symbol
15091 @opindex mtelephony
15092 Passed down to the assembler to enable dual- and single-operand
15093 instructions for telephony. Also sets the preprocessor symbol
15094 @code{__Xtelephony}. This option is deprecated.
15098 Passed down to the assembler to enable the XY memory extension. Also
15099 sets the preprocessor symbol @code{__Xxy}.
15103 The following options control how the assembly code is annotated:
15105 @c Assembly annotation options
15109 Annotate assembler instructions with estimated addresses.
15111 @item -mannotate-align
15112 @opindex mannotate-align
15113 Explain what alignment considerations lead to the decision to make an
15114 instruction short or long.
15118 The following options are passed through to the linker:
15120 @c options passed through to the linker
15124 Passed through to the linker, to specify use of the @code{arclinux} emulation.
15125 This option is enabled by default in tool chains built for
15126 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
15127 when profiling is not requested.
15129 @item -marclinux_prof
15130 @opindex marclinux_prof
15131 Passed through to the linker, to specify use of the
15132 @code{arclinux_prof} emulation. This option is enabled by default in
15133 tool chains built for @w{@code{arc-linux-uclibc}} and
15134 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
15138 The following options control the semantics of generated code:
15140 @c semantically relevant code generation options
15143 @opindex mlong-calls
15144 Generate calls as register indirect calls, thus providing access
15145 to the full 32-bit address range.
15147 @item -mmedium-calls
15148 @opindex mmedium-calls
15149 Don't use less than 25-bit addressing range for calls, which is the
15150 offset available for an unconditional branch-and-link
15151 instruction. Conditional execution of function calls is suppressed, to
15152 allow use of the 25-bit range, rather than the 21-bit range with
15153 conditional branch-and-link. This is the default for tool chains built
15154 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
15158 Put definitions of externally-visible data in a small data section if
15159 that data is no bigger than @var{num} bytes. The default value of
15160 @var{num} is 4 for any ARC configuration, or 8 when we have double
15161 load/store operations.
15165 Do not generate sdata references. This is the default for tool chains
15166 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
15169 @item -mvolatile-cache
15170 @opindex mvolatile-cache
15171 Use ordinarily cached memory accesses for volatile references. This is the
15174 @item -mno-volatile-cache
15175 @opindex mno-volatile-cache
15176 Enable cache bypass for volatile references.
15180 The following options fine tune code generation:
15181 @c code generation tuning options
15184 @opindex malign-call
15185 Do alignment optimizations for call instructions.
15187 @item -mauto-modify-reg
15188 @opindex mauto-modify-reg
15189 Enable the use of pre/post modify with register displacement.
15191 @item -mbbit-peephole
15192 @opindex mbbit-peephole
15193 Enable bbit peephole2.
15197 This option disables a target-specific pass in @file{arc_reorg} to
15198 generate compare-and-branch (@code{br@var{cc}}) instructions.
15199 It has no effect on
15200 generation of these instructions driven by the combiner pass.
15202 @item -mcase-vector-pcrel
15203 @opindex mcase-vector-pcrel
15204 Use PC-relative switch case tables to enable case table shortening.
15205 This is the default for @option{-Os}.
15207 @item -mcompact-casesi
15208 @opindex mcompact-casesi
15209 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
15210 and only available for ARCv1 cores.
15212 @item -mno-cond-exec
15213 @opindex mno-cond-exec
15214 Disable the ARCompact-specific pass to generate conditional
15215 execution instructions.
15217 Due to delay slot scheduling and interactions between operand numbers,
15218 literal sizes, instruction lengths, and the support for conditional execution,
15219 the target-independent pass to generate conditional execution is often lacking,
15220 so the ARC port has kept a special pass around that tries to find more
15221 conditional execution generation opportunities after register allocation,
15222 branch shortening, and delay slot scheduling have been done. This pass
15223 generally, but not always, improves performance and code size, at the cost of
15224 extra compilation time, which is why there is an option to switch it off.
15225 If you have a problem with call instructions exceeding their allowable
15226 offset range because they are conditionalized, you should consider using
15227 @option{-mmedium-calls} instead.
15229 @item -mearly-cbranchsi
15230 @opindex mearly-cbranchsi
15231 Enable pre-reload use of the @code{cbranchsi} pattern.
15233 @item -mexpand-adddi
15234 @opindex mexpand-adddi
15235 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
15236 @code{add.f}, @code{adc} etc. This option is deprecated.
15238 @item -mindexed-loads
15239 @opindex mindexed-loads
15240 Enable the use of indexed loads. This can be problematic because some
15241 optimizers then assume that indexed stores exist, which is not
15245 Enable Local Register Allocation. This is still experimental for ARC,
15246 so by default the compiler uses standard reload
15247 (i.e. @option{-mno-lra}).
15249 @item -mlra-priority-none
15250 @opindex mlra-priority-none
15251 Don't indicate any priority for target registers.
15253 @item -mlra-priority-compact
15254 @opindex mlra-priority-compact
15255 Indicate target register priority for r0..r3 / r12..r15.
15257 @item -mlra-priority-noncompact
15258 @opindex mlra-priority-noncompact
15259 Reduce target register priority for r0..r3 / r12..r15.
15261 @item -mno-millicode
15262 @opindex mno-millicode
15263 When optimizing for size (using @option{-Os}), prologues and epilogues
15264 that have to save or restore a large number of registers are often
15265 shortened by using call to a special function in libgcc; this is
15266 referred to as a @emph{millicode} call. As these calls can pose
15267 performance issues, and/or cause linking issues when linking in a
15268 nonstandard way, this option is provided to turn off millicode call
15272 @opindex mmixed-code
15273 Tweak register allocation to help 16-bit instruction generation.
15274 This generally has the effect of decreasing the average instruction size
15275 while increasing the instruction count.
15279 Enable @samp{q} instruction alternatives.
15280 This is the default for @option{-Os}.
15284 Enable @samp{Rcq} constraint handling.
15285 Most short code generation depends on this.
15286 This is the default.
15290 Enable @samp{Rcw} constraint handling.
15291 Most ccfsm condexec mostly depends on this.
15292 This is the default.
15294 @item -msize-level=@var{level}
15295 @opindex msize-level
15296 Fine-tune size optimization with regards to instruction lengths and alignment.
15297 The recognized values for @var{level} are:
15300 No size optimization. This level is deprecated and treated like @samp{1}.
15303 Short instructions are used opportunistically.
15306 In addition, alignment of loops and of code after barriers are dropped.
15309 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
15313 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
15314 the behavior when this is not set is equivalent to level @samp{1}.
15316 @item -mtune=@var{cpu}
15318 Set instruction scheduling parameters for @var{cpu}, overriding any implied
15319 by @option{-mcpu=}.
15321 Supported values for @var{cpu} are
15325 Tune for ARC600 CPU.
15328 Tune for ARC601 CPU.
15331 Tune for ARC700 CPU with standard multiplier block.
15334 Tune for ARC700 CPU with XMAC block.
15337 Tune for ARC725D CPU.
15340 Tune for ARC750D CPU.
15344 @item -mmultcost=@var{num}
15346 Cost to assume for a multiply instruction, with @samp{4} being equal to a
15347 normal instruction.
15349 @item -munalign-prob-threshold=@var{probability}
15350 @opindex munalign-prob-threshold
15351 Set probability threshold for unaligning branches.
15352 When tuning for @samp{ARC700} and optimizing for speed, branches without
15353 filled delay slot are preferably emitted unaligned and long, unless
15354 profiling indicates that the probability for the branch to be taken
15355 is below @var{probability}. @xref{Cross-profiling}.
15356 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
15360 The following options are maintained for backward compatibility, but
15361 are now deprecated and will be removed in a future release:
15363 @c Deprecated options
15371 @opindex mbig-endian
15374 Compile code for big-endian targets. Use of these options is now
15375 deprecated. Big-endian code is supported by configuring GCC to build
15376 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
15377 for which big endian is the default.
15379 @item -mlittle-endian
15380 @opindex mlittle-endian
15383 Compile code for little-endian targets. Use of these options is now
15384 deprecated. Little-endian code is supported by configuring GCC to build
15385 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
15386 for which little endian is the default.
15388 @item -mbarrel_shifter
15389 @opindex mbarrel_shifter
15390 Replaced by @option{-mbarrel-shifter}.
15392 @item -mdpfp_compact
15393 @opindex mdpfp_compact
15394 Replaced by @option{-mdpfp-compact}.
15397 @opindex mdpfp_fast
15398 Replaced by @option{-mdpfp-fast}.
15401 @opindex mdsp_packa
15402 Replaced by @option{-mdsp-packa}.
15406 Replaced by @option{-mea}.
15410 Replaced by @option{-mmac-24}.
15414 Replaced by @option{-mmac-d16}.
15416 @item -mspfp_compact
15417 @opindex mspfp_compact
15418 Replaced by @option{-mspfp-compact}.
15421 @opindex mspfp_fast
15422 Replaced by @option{-mspfp-fast}.
15424 @item -mtune=@var{cpu}
15426 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
15427 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
15428 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
15430 @item -multcost=@var{num}
15432 Replaced by @option{-mmultcost}.
15437 @subsection ARM Options
15438 @cindex ARM options
15440 These @samp{-m} options are defined for the ARM port:
15443 @item -mabi=@var{name}
15445 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
15446 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
15449 @opindex mapcs-frame
15450 Generate a stack frame that is compliant with the ARM Procedure Call
15451 Standard for all functions, even if this is not strictly necessary for
15452 correct execution of the code. Specifying @option{-fomit-frame-pointer}
15453 with this option causes the stack frames not to be generated for
15454 leaf functions. The default is @option{-mno-apcs-frame}.
15455 This option is deprecated.
15459 This is a synonym for @option{-mapcs-frame} and is deprecated.
15462 @c not currently implemented
15463 @item -mapcs-stack-check
15464 @opindex mapcs-stack-check
15465 Generate code to check the amount of stack space available upon entry to
15466 every function (that actually uses some stack space). If there is
15467 insufficient space available then either the function
15468 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
15469 called, depending upon the amount of stack space required. The runtime
15470 system is required to provide these functions. The default is
15471 @option{-mno-apcs-stack-check}, since this produces smaller code.
15473 @c not currently implemented
15474 @item -mapcs-reentrant
15475 @opindex mapcs-reentrant
15476 Generate reentrant, position-independent code. The default is
15477 @option{-mno-apcs-reentrant}.
15480 @item -mthumb-interwork
15481 @opindex mthumb-interwork
15482 Generate code that supports calling between the ARM and Thumb
15483 instruction sets. Without this option, on pre-v5 architectures, the
15484 two instruction sets cannot be reliably used inside one program. The
15485 default is @option{-mno-thumb-interwork}, since slightly larger code
15486 is generated when @option{-mthumb-interwork} is specified. In AAPCS
15487 configurations this option is meaningless.
15489 @item -mno-sched-prolog
15490 @opindex mno-sched-prolog
15491 Prevent the reordering of instructions in the function prologue, or the
15492 merging of those instruction with the instructions in the function's
15493 body. This means that all functions start with a recognizable set
15494 of instructions (or in fact one of a choice from a small set of
15495 different function prologues), and this information can be used to
15496 locate the start of functions inside an executable piece of code. The
15497 default is @option{-msched-prolog}.
15499 @item -mfloat-abi=@var{name}
15500 @opindex mfloat-abi
15501 Specifies which floating-point ABI to use. Permissible values
15502 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15504 Specifying @samp{soft} causes GCC to generate output containing
15505 library calls for floating-point operations.
15506 @samp{softfp} allows the generation of code using hardware floating-point
15507 instructions, but still uses the soft-float calling conventions.
15508 @samp{hard} allows generation of floating-point instructions
15509 and uses FPU-specific calling conventions.
15511 The default depends on the specific target configuration. Note that
15512 the hard-float and soft-float ABIs are not link-compatible; you must
15513 compile your entire program with the same ABI, and link with a
15514 compatible set of libraries.
15516 @item -mlittle-endian
15517 @opindex mlittle-endian
15518 Generate code for a processor running in little-endian mode. This is
15519 the default for all standard configurations.
15522 @opindex mbig-endian
15523 Generate code for a processor running in big-endian mode; the default is
15524 to compile code for a little-endian processor.
15529 When linking a big-endian image select between BE8 and BE32 formats.
15530 The option has no effect for little-endian images and is ignored. The
15531 default is dependent on the selected target architecture. For ARMv6
15532 and later architectures the default is BE8, for older architectures
15533 the default is BE32. BE32 format has been deprecated by ARM.
15535 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
15537 This specifies the name of the target ARM architecture. GCC uses this
15538 name to determine what kind of instructions it can emit when generating
15539 assembly code. This option can be used in conjunction with or instead
15540 of the @option{-mcpu=} option.
15542 Permissible names are:
15544 @samp{armv5t}, @samp{armv5te},
15545 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
15546 @samp{armv6z}, @samp{armv6zk},
15547 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
15548 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
15551 @samp{armv6-m}, @samp{armv6s-m},
15552 @samp{armv7-m}, @samp{armv7e-m},
15553 @samp{armv8-m.base}, @samp{armv8-m.main},
15554 @samp{iwmmxt} and @samp{iwmmxt2}.
15556 Additionally, the following architectures, which lack support for the
15557 Thumb exection state, are recognized but support is deprecated:
15558 @samp{armv2}, @samp{armv2a}, @samp{armv3}, @samp{armv3m},
15559 @samp{armv4}, @samp{armv5} and @samp{armv5e}.
15561 Many of the architectures support extensions. These can be added by
15562 appending @samp{+@var{extension}} to the architecture name. Extension
15563 options are processed in order and capabilities accumulate. An extension
15564 will also enable any necessary base extensions
15565 upon which it depends. For example, the @samp{+crypto} extension
15566 will always enable the @samp{+simd} extension. The exception to the
15567 additive construction is for extensions that are prefixed with
15568 @samp{+no@dots{}}: these extensions disable the specified option and
15569 any other extensions that may depend on the presence of that
15572 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
15573 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
15574 entirely disabled by the @samp{+nofp} option that follows it.
15576 Most extension names are generically named, but have an effect that is
15577 dependent upon the architecture to which it is applied. For example,
15578 the @samp{+simd} option can be applied to both @samp{armv7-a} and
15579 @samp{armv8-a} architectures, but will enable the original ARMv7-A
15580 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
15581 variant for @samp{armv8-a}.
15583 The table below lists the supported extensions for each architecture.
15584 Architectures not mentioned do not support any extensions.
15598 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
15599 used as an alias for this extension.
15602 Disable the floating-point instructions.
15606 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
15609 The VFPv3 floating-point instructions, with 16 double-precision
15610 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15611 for this extension. Note that floating-point is not supported by the
15612 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
15613 ARMv7-R architectures.
15616 Disable the floating-point instructions.
15622 The VFPv3 floating-point instructions, with 16 double-precision
15623 registers. The extension @samp{+vfpv3-d16} can be used as an alias
15624 for this extension.
15627 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15628 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
15629 for this extension.
15632 The VFPv3 floating-point instructions, with 32 double-precision
15635 @item +vfpv3-d16-fp16
15636 The VFPv3 floating-point instructions, with 16 double-precision
15637 registers and the half-precision floating-point conversion operations.
15640 The VFPv3 floating-point instructions, with 32 double-precision
15641 registers and the half-precision floating-point conversion operations.
15644 The VFPv4 floating-point instructions, with 16 double-precision
15648 The VFPv4 floating-point instructions, with 32 double-precision
15652 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15653 the half-precision floating-point conversion operations.
15656 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
15659 Disable the Advanced SIMD instructions (does not disable floating point).
15662 Disable the floating-point and Advanced SIMD instructions.
15666 The extended version of the ARMv7-A architecture with support for
15670 The VFPv4 floating-point instructions, with 16 double-precision registers.
15671 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
15674 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
15675 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
15678 The VFPv3 floating-point instructions, with 16 double-precision
15682 The VFPv3 floating-point instructions, with 32 double-precision
15685 @item +vfpv3-d16-fp16
15686 The VFPv3 floating-point instructions, with 16 double-precision
15687 registers and the half-precision floating-point conversion operations.
15690 The VFPv3 floating-point instructions, with 32 double-precision
15691 registers and the half-precision floating-point conversion operations.
15694 The VFPv4 floating-point instructions, with 16 double-precision
15698 The VFPv4 floating-point instructions, with 32 double-precision
15702 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
15703 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
15706 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
15707 the half-precision floating-point conversion operations.
15710 Disable the Advanced SIMD instructions (does not disable floating point).
15713 Disable the floating-point and Advanced SIMD instructions.
15719 The Cyclic Redundancy Check (CRC) instructions.
15721 The ARMv8-A Advanced SIMD and floating-point instructions.
15723 The cryptographic instructions.
15725 Disable the cryptographic isntructions.
15727 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15733 The ARMv8.1-A Advanced SIMD and floating-point instructions.
15736 The cryptographic instructions. This also enables the Advanced SIMD and
15737 floating-point instructions.
15740 Disable the cryptographic isntructions.
15743 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15749 The half-precision floating-point data processing instructions.
15750 This also enables the Advanced SIMD and floating-point instructions.
15753 The ARMv8.1-A Advanced SIMD and floating-point instructions.
15756 The cryptographic instructions. This also enables the Advanced SIMD and
15757 floating-point instructions.
15760 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
15763 Disable the cryptographic extension.
15766 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15772 The single-precision VFPv3 floating-point instructions. The extension
15773 @samp{+vfpv3xd} can be used as an alias for this extension.
15776 The VFPv3 floating-point instructions with 16 double-precision registers.
15777 The extension +vfpv3-d16 can be used as an alias for this extension.
15780 Disable the floating-point extension.
15783 The ARM-state integer division instructions.
15786 Disable the ARM-state integer division extension.
15792 The single-precision VFPv4 floating-point instructions.
15795 The single-precision FPv5 floating-point instructions.
15798 The single- and double-precision FPv5 floating-point instructions.
15801 Disable the floating-point extensions.
15807 The DSP instructions.
15810 Disable the DSP extension.
15813 The single-precision floating-point instructions.
15816 The single- and double-precision floating-point instructions.
15819 Disable the floating-point extension.
15825 The Cyclic Redundancy Check (CRC) instructions.
15827 The single-precision FPv5 floating-point instructions.
15829 The ARMv8-A Advanced SIMD and floating-point instructions.
15831 The cryptographic instructions.
15833 Disable the cryptographic isntructions.
15835 Disable the floating-point, Advanced SIMD and cryptographic instructions.
15840 @option{-march=native} causes the compiler to auto-detect the architecture
15841 of the build computer. At present, this feature is only supported on
15842 GNU/Linux, and not all architectures are recognized. If the auto-detect
15843 is unsuccessful the option has no effect.
15845 @item -mtune=@var{name}
15847 This option specifies the name of the target ARM processor for
15848 which GCC should tune the performance of the code.
15849 For some ARM implementations better performance can be obtained by using
15851 Permissible names are: @samp{arm2}, @samp{arm250},
15852 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
15853 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
15854 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
15855 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
15857 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
15858 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
15859 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
15860 @samp{strongarm1110},
15861 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
15862 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
15863 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
15864 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
15865 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
15866 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
15867 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
15868 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
15869 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
15870 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15871 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15872 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7},
15873 @samp{cortex-r8}, @samp{cortex-r52},
15881 @samp{cortex-m0plus},
15882 @samp{cortex-m1.small-multiply},
15883 @samp{cortex-m0.small-multiply},
15884 @samp{cortex-m0plus.small-multiply},
15886 @samp{marvell-pj4},
15887 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
15888 @samp{fa526}, @samp{fa626},
15889 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
15892 Additionally, this option can specify that GCC should tune the performance
15893 of the code for a big.LITTLE system. Permissible names are:
15894 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
15895 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15896 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
15897 @samp{cortex-a75.cortex-a55}.
15899 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
15900 performance for a blend of processors within architecture @var{arch}.
15901 The aim is to generate code that run well on the current most popular
15902 processors, balancing between optimizations that benefit some CPUs in the
15903 range, and avoiding performance pitfalls of other CPUs. The effects of
15904 this option may change in future GCC versions as CPU models come and go.
15906 @option{-mtune} permits the same extension options as @option{-mcpu}, but
15907 the extension options do not affect the tuning of the generated code.
15909 @option{-mtune=native} causes the compiler to auto-detect the CPU
15910 of the build computer. At present, this feature is only supported on
15911 GNU/Linux, and not all architectures are recognized. If the auto-detect is
15912 unsuccessful the option has no effect.
15914 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
15916 This specifies the name of the target ARM processor. GCC uses this name
15917 to derive the name of the target ARM architecture (as if specified
15918 by @option{-march}) and the ARM processor type for which to tune for
15919 performance (as if specified by @option{-mtune}). Where this option
15920 is used in conjunction with @option{-march} or @option{-mtune},
15921 those options take precedence over the appropriate part of this option.
15923 Many of the supported CPUs implement optional architectural
15924 extensions. Where this is so the architectural extensions are
15925 normally enabled by default. If implementations that lack the
15926 extension exist, then the extension syntax can be used to disable
15927 those extensions that have been omitted. For floating-point and
15928 Advanced SIMD (Neon) instructions, the settings of the options
15929 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
15930 floating-point and Advanced SIMD instructions will only be used if
15931 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
15932 @option{-mfpu} other than @samp{auto} will override the available
15933 floating-point and SIMD extension instructions.
15935 For example, @samp{cortex-a9} can be found in three major
15936 configurations: integer only, with just a floating-point unit or with
15937 floating-point and Advanced SIMD. The default is to enable all the
15938 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
15939 be used to disable just the SIMD or both the SIMD and floating-point
15940 instructions respectively.
15942 Permissible names for this option are the same as those for
15945 The following extension options are common to the listed CPUs:
15949 Disable the DSP instructions on @samp{cortex-m33}.
15952 Disables the floating-point instructions on @samp{arm9e},
15953 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
15954 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
15955 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
15956 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
15957 Disables the floating-point and SIMD instructions on
15958 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
15959 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
15960 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
15961 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
15962 @samp{cortex-a53} and @samp{cortex-a55}.
15965 Disables the double-precision component of the floating-point instructions
15966 on @samp{cortex-r5}, @samp{cortex-r52} and @samp{cortex-m7}.
15969 Disables the SIMD (but not floating-point) instructions on
15970 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
15971 and @samp{cortex-a9}.
15974 Enables the cryptographic instructions on @samp{cortex-a32},
15975 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
15976 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
15977 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15978 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
15979 @samp{cortex-a75.cortex-a55}.
15982 Additionally the @samp{generic-armv7-a} pseudo target defaults to
15983 VFPv3 with 16 double-precision registers. It supports the following
15984 extension options: @samp{vfpv3-d16}, @samp{vfpv3},
15985 @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16}, @samp{vfpv4-d16},
15986 @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3}, @samp{neon-fp16},
15987 @samp{neon-vfpv4}. The meanings are the same as for the extensions to
15988 @option{-march=armv7-a}.
15990 @option{-mcpu=generic-@var{arch}} is also permissible, and is
15991 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
15992 See @option{-mtune} for more information.
15994 @option{-mcpu=native} causes the compiler to auto-detect the CPU
15995 of the build computer. At present, this feature is only supported on
15996 GNU/Linux, and not all architectures are recognized. If the auto-detect
15997 is unsuccessful the option has no effect.
15999 @item -mfpu=@var{name}
16001 This specifies what floating-point hardware (or hardware emulation) is
16002 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
16004 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
16005 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
16006 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
16007 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
16008 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
16009 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
16010 is an alias for @samp{vfpv2}.
16012 The setting @samp{auto} is the default and is special. It causes the
16013 compiler to select the floating-point and Advanced SIMD instructions
16014 based on the settings of @option{-mcpu} and @option{-march}.
16016 If the selected floating-point hardware includes the NEON extension
16017 (e.g. @option{-mfpu=neon}), note that floating-point
16018 operations are not generated by GCC's auto-vectorization pass unless
16019 @option{-funsafe-math-optimizations} is also specified. This is
16020 because NEON hardware does not fully implement the IEEE 754 standard for
16021 floating-point arithmetic (in particular denormal values are treated as
16022 zero), so the use of NEON instructions may lead to a loss of precision.
16024 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}).
16026 @item -mfp16-format=@var{name}
16027 @opindex mfp16-format
16028 Specify the format of the @code{__fp16} half-precision floating-point type.
16029 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
16030 the default is @samp{none}, in which case the @code{__fp16} type is not
16031 defined. @xref{Half-Precision}, for more information.
16033 @item -mstructure-size-boundary=@var{n}
16034 @opindex mstructure-size-boundary
16035 The sizes of all structures and unions are rounded up to a multiple
16036 of the number of bits set by this option. Permissible values are 8, 32
16037 and 64. The default value varies for different toolchains. For the COFF
16038 targeted toolchain the default value is 8. A value of 64 is only allowed
16039 if the underlying ABI supports it.
16041 Specifying a larger number can produce faster, more efficient code, but
16042 can also increase the size of the program. Different values are potentially
16043 incompatible. Code compiled with one value cannot necessarily expect to
16044 work with code or libraries compiled with another value, if they exchange
16045 information using structures or unions.
16047 This option is deprecated.
16049 @item -mabort-on-noreturn
16050 @opindex mabort-on-noreturn
16051 Generate a call to the function @code{abort} at the end of a
16052 @code{noreturn} function. It is executed if the function tries to
16056 @itemx -mno-long-calls
16057 @opindex mlong-calls
16058 @opindex mno-long-calls
16059 Tells the compiler to perform function calls by first loading the
16060 address of the function into a register and then performing a subroutine
16061 call on this register. This switch is needed if the target function
16062 lies outside of the 64-megabyte addressing range of the offset-based
16063 version of subroutine call instruction.
16065 Even if this switch is enabled, not all function calls are turned
16066 into long calls. The heuristic is that static functions, functions
16067 that have the @code{short_call} attribute, functions that are inside
16068 the scope of a @code{#pragma no_long_calls} directive, and functions whose
16069 definitions have already been compiled within the current compilation
16070 unit are not turned into long calls. The exceptions to this rule are
16071 that weak function definitions, functions with the @code{long_call}
16072 attribute or the @code{section} attribute, and functions that are within
16073 the scope of a @code{#pragma long_calls} directive are always
16074 turned into long calls.
16076 This feature is not enabled by default. Specifying
16077 @option{-mno-long-calls} restores the default behavior, as does
16078 placing the function calls within the scope of a @code{#pragma
16079 long_calls_off} directive. Note these switches have no effect on how
16080 the compiler generates code to handle function calls via function
16083 @item -msingle-pic-base
16084 @opindex msingle-pic-base
16085 Treat the register used for PIC addressing as read-only, rather than
16086 loading it in the prologue for each function. The runtime system is
16087 responsible for initializing this register with an appropriate value
16088 before execution begins.
16090 @item -mpic-register=@var{reg}
16091 @opindex mpic-register
16092 Specify the register to be used for PIC addressing.
16093 For standard PIC base case, the default is any suitable register
16094 determined by compiler. For single PIC base case, the default is
16095 @samp{R9} if target is EABI based or stack-checking is enabled,
16096 otherwise the default is @samp{R10}.
16098 @item -mpic-data-is-text-relative
16099 @opindex mpic-data-is-text-relative
16100 Assume that the displacement between the text and data segments is fixed
16101 at static link time. This permits using PC-relative addressing
16102 operations to access data known to be in the data segment. For
16103 non-VxWorks RTP targets, this option is enabled by default. When
16104 disabled on such targets, it will enable @option{-msingle-pic-base} by
16107 @item -mpoke-function-name
16108 @opindex mpoke-function-name
16109 Write the name of each function into the text section, directly
16110 preceding the function prologue. The generated code is similar to this:
16114 .ascii "arm_poke_function_name", 0
16117 .word 0xff000000 + (t1 - t0)
16118 arm_poke_function_name
16120 stmfd sp!, @{fp, ip, lr, pc@}
16124 When performing a stack backtrace, code can inspect the value of
16125 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
16126 location @code{pc - 12} and the top 8 bits are set, then we know that
16127 there is a function name embedded immediately preceding this location
16128 and has length @code{((pc[-3]) & 0xff000000)}.
16135 Select between generating code that executes in ARM and Thumb
16136 states. The default for most configurations is to generate code
16137 that executes in ARM state, but the default can be changed by
16138 configuring GCC with the @option{--with-mode=}@var{state}
16141 You can also override the ARM and Thumb mode for each function
16142 by using the @code{target("thumb")} and @code{target("arm")} function attributes
16143 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
16146 @opindex mtpcs-frame
16147 Generate a stack frame that is compliant with the Thumb Procedure Call
16148 Standard for all non-leaf functions. (A leaf function is one that does
16149 not call any other functions.) The default is @option{-mno-tpcs-frame}.
16151 @item -mtpcs-leaf-frame
16152 @opindex mtpcs-leaf-frame
16153 Generate a stack frame that is compliant with the Thumb Procedure Call
16154 Standard for all leaf functions. (A leaf function is one that does
16155 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
16157 @item -mcallee-super-interworking
16158 @opindex mcallee-super-interworking
16159 Gives all externally visible functions in the file being compiled an ARM
16160 instruction set header which switches to Thumb mode before executing the
16161 rest of the function. This allows these functions to be called from
16162 non-interworking code. This option is not valid in AAPCS configurations
16163 because interworking is enabled by default.
16165 @item -mcaller-super-interworking
16166 @opindex mcaller-super-interworking
16167 Allows calls via function pointers (including virtual functions) to
16168 execute correctly regardless of whether the target code has been
16169 compiled for interworking or not. There is a small overhead in the cost
16170 of executing a function pointer if this option is enabled. This option
16171 is not valid in AAPCS configurations because interworking is enabled
16174 @item -mtp=@var{name}
16176 Specify the access model for the thread local storage pointer. The valid
16177 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
16178 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
16179 (supported in the arm6k architecture), and @samp{auto}, which uses the
16180 best available method for the selected processor. The default setting is
16183 @item -mtls-dialect=@var{dialect}
16184 @opindex mtls-dialect
16185 Specify the dialect to use for accessing thread local storage. Two
16186 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
16187 @samp{gnu} dialect selects the original GNU scheme for supporting
16188 local and global dynamic TLS models. The @samp{gnu2} dialect
16189 selects the GNU descriptor scheme, which provides better performance
16190 for shared libraries. The GNU descriptor scheme is compatible with
16191 the original scheme, but does require new assembler, linker and
16192 library support. Initial and local exec TLS models are unaffected by
16193 this option and always use the original scheme.
16195 @item -mword-relocations
16196 @opindex mword-relocations
16197 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
16198 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
16199 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
16202 @item -mfix-cortex-m3-ldrd
16203 @opindex mfix-cortex-m3-ldrd
16204 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
16205 with overlapping destination and base registers are used. This option avoids
16206 generating these instructions. This option is enabled by default when
16207 @option{-mcpu=cortex-m3} is specified.
16209 @item -munaligned-access
16210 @itemx -mno-unaligned-access
16211 @opindex munaligned-access
16212 @opindex mno-unaligned-access
16213 Enables (or disables) reading and writing of 16- and 32- bit values
16214 from addresses that are not 16- or 32- bit aligned. By default
16215 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
16216 ARMv8-M Baseline architectures, and enabled for all other
16217 architectures. If unaligned access is not enabled then words in packed
16218 data structures are accessed a byte at a time.
16220 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
16221 generated object file to either true or false, depending upon the
16222 setting of this option. If unaligned access is enabled then the
16223 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
16226 @item -mneon-for-64bits
16227 @opindex mneon-for-64bits
16228 Enables using Neon to handle scalar 64-bits operations. This is
16229 disabled by default since the cost of moving data from core registers
16232 @item -mslow-flash-data
16233 @opindex mslow-flash-data
16234 Assume loading data from flash is slower than fetching instruction.
16235 Therefore literal load is minimized for better performance.
16236 This option is only supported when compiling for ARMv7 M-profile and
16239 @item -masm-syntax-unified
16240 @opindex masm-syntax-unified
16241 Assume inline assembler is using unified asm syntax. The default is
16242 currently off which implies divided syntax. This option has no impact
16243 on Thumb2. However, this may change in future releases of GCC.
16244 Divided syntax should be considered deprecated.
16246 @item -mrestrict-it
16247 @opindex mrestrict-it
16248 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
16249 IT blocks can only contain a single 16-bit instruction from a select
16250 set of instructions. This option is on by default for ARMv8-A Thumb mode.
16252 @item -mprint-tune-info
16253 @opindex mprint-tune-info
16254 Print CPU tuning information as comment in assembler file. This is
16255 an option used only for regression testing of the compiler and not
16256 intended for ordinary use in compiling code. This option is disabled
16260 @opindex mpure-code
16261 Do not allow constant data to be placed in code sections.
16262 Additionally, when compiling for ELF object format give all text sections the
16263 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
16264 is only available when generating non-pic code for M-profile targets with the
16269 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
16270 Development Tools Engineering Specification", which can be found on
16271 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
16275 @subsection AVR Options
16276 @cindex AVR Options
16278 These options are defined for AVR implementations:
16281 @item -mmcu=@var{mcu}
16283 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
16285 The default for this option is@tie{}@samp{avr2}.
16287 GCC supports the following AVR devices and ISAs:
16289 @include avr-mmcu.texi
16294 Assume that all data in static storage can be accessed by LDS / STS
16295 instructions. This option has only an effect on reduced Tiny devices like
16296 ATtiny40. See also the @code{absdata}
16297 @ref{AVR Variable Attributes,variable attribute}.
16299 @item -maccumulate-args
16300 @opindex maccumulate-args
16301 Accumulate outgoing function arguments and acquire/release the needed
16302 stack space for outgoing function arguments once in function
16303 prologue/epilogue. Without this option, outgoing arguments are pushed
16304 before calling a function and popped afterwards.
16306 Popping the arguments after the function call can be expensive on
16307 AVR so that accumulating the stack space might lead to smaller
16308 executables because arguments need not be removed from the
16309 stack after such a function call.
16311 This option can lead to reduced code size for functions that perform
16312 several calls to functions that get their arguments on the stack like
16313 calls to printf-like functions.
16315 @item -mbranch-cost=@var{cost}
16316 @opindex mbranch-cost
16317 Set the branch costs for conditional branch instructions to
16318 @var{cost}. Reasonable values for @var{cost} are small, non-negative
16319 integers. The default branch cost is 0.
16321 @item -mcall-prologues
16322 @opindex mcall-prologues
16323 Functions prologues/epilogues are expanded as calls to appropriate
16324 subroutines. Code size is smaller.
16326 @item -mgas-isr-prologues
16327 @opindex mgas-isr-prologues
16328 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
16329 instruction supported by GNU Binutils.
16330 If this option is on, the feature can still be disabled for individual
16331 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
16332 function attribute. This feature is activated per default
16333 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
16334 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
16338 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
16339 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
16340 and @code{long long} is 4 bytes. Please note that this option does not
16341 conform to the C standards, but it results in smaller code
16344 @item -mn-flash=@var{num}
16346 Assume that the flash memory has a size of
16347 @var{num} times 64@tie{}KiB.
16349 @item -mno-interrupts
16350 @opindex mno-interrupts
16351 Generated code is not compatible with hardware interrupts.
16352 Code size is smaller.
16356 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
16357 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
16358 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
16359 the assembler's command line and the @option{--relax} option to the
16360 linker's command line.
16362 Jump relaxing is performed by the linker because jump offsets are not
16363 known before code is located. Therefore, the assembler code generated by the
16364 compiler is the same, but the instructions in the executable may
16365 differ from instructions in the assembler code.
16367 Relaxing must be turned on if linker stubs are needed, see the
16368 section on @code{EIND} and linker stubs below.
16372 Assume that the device supports the Read-Modify-Write
16373 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
16375 @item -mshort-calls
16376 @opindex mshort-calls
16378 Assume that @code{RJMP} and @code{RCALL} can target the whole
16381 This option is used internally for multilib selection. It is
16382 not an optimization option, and you don't need to set it by hand.
16386 Treat the stack pointer register as an 8-bit register,
16387 i.e.@: assume the high byte of the stack pointer is zero.
16388 In general, you don't need to set this option by hand.
16390 This option is used internally by the compiler to select and
16391 build multilibs for architectures @code{avr2} and @code{avr25}.
16392 These architectures mix devices with and without @code{SPH}.
16393 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
16394 the compiler driver adds or removes this option from the compiler
16395 proper's command line, because the compiler then knows if the device
16396 or architecture has an 8-bit stack pointer and thus no @code{SPH}
16401 Use address register @code{X} in a way proposed by the hardware. This means
16402 that @code{X} is only used in indirect, post-increment or
16403 pre-decrement addressing.
16405 Without this option, the @code{X} register may be used in the same way
16406 as @code{Y} or @code{Z} which then is emulated by additional
16408 For example, loading a value with @code{X+const} addressing with a
16409 small non-negative @code{const < 64} to a register @var{Rn} is
16413 adiw r26, const ; X += const
16414 ld @var{Rn}, X ; @var{Rn} = *X
16415 sbiw r26, const ; X -= const
16419 @opindex mtiny-stack
16420 Only change the lower 8@tie{}bits of the stack pointer.
16422 @item -mfract-convert-truncate
16423 @opindex mfract-convert-truncate
16424 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
16427 @opindex nodevicelib
16428 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
16430 @item -Waddr-space-convert
16431 @opindex Waddr-space-convert
16432 Warn about conversions between address spaces in the case where the
16433 resulting address space is not contained in the incoming address space.
16435 @item -Wmisspelled-isr
16436 @opindex Wmisspelled-isr
16437 Warn if the ISR is misspelled, i.e. without __vector prefix.
16438 Enabled by default.
16441 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
16442 @cindex @code{EIND}
16443 Pointers in the implementation are 16@tie{}bits wide.
16444 The address of a function or label is represented as word address so
16445 that indirect jumps and calls can target any code address in the
16446 range of 64@tie{}Ki words.
16448 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
16449 bytes of program memory space, there is a special function register called
16450 @code{EIND} that serves as most significant part of the target address
16451 when @code{EICALL} or @code{EIJMP} instructions are used.
16453 Indirect jumps and calls on these devices are handled as follows by
16454 the compiler and are subject to some limitations:
16459 The compiler never sets @code{EIND}.
16462 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
16463 instructions or might read @code{EIND} directly in order to emulate an
16464 indirect call/jump by means of a @code{RET} instruction.
16467 The compiler assumes that @code{EIND} never changes during the startup
16468 code or during the application. In particular, @code{EIND} is not
16469 saved/restored in function or interrupt service routine
16473 For indirect calls to functions and computed goto, the linker
16474 generates @emph{stubs}. Stubs are jump pads sometimes also called
16475 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
16476 The stub contains a direct jump to the desired address.
16479 Linker relaxation must be turned on so that the linker generates
16480 the stubs correctly in all situations. See the compiler option
16481 @option{-mrelax} and the linker option @option{--relax}.
16482 There are corner cases where the linker is supposed to generate stubs
16483 but aborts without relaxation and without a helpful error message.
16486 The default linker script is arranged for code with @code{EIND = 0}.
16487 If code is supposed to work for a setup with @code{EIND != 0}, a custom
16488 linker script has to be used in order to place the sections whose
16489 name start with @code{.trampolines} into the segment where @code{EIND}
16493 The startup code from libgcc never sets @code{EIND}.
16494 Notice that startup code is a blend of code from libgcc and AVR-LibC.
16495 For the impact of AVR-LibC on @code{EIND}, see the
16496 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
16499 It is legitimate for user-specific startup code to set up @code{EIND}
16500 early, for example by means of initialization code located in
16501 section @code{.init3}. Such code runs prior to general startup code
16502 that initializes RAM and calls constructors, but after the bit
16503 of startup code from AVR-LibC that sets @code{EIND} to the segment
16504 where the vector table is located.
16506 #include <avr/io.h>
16509 __attribute__((section(".init3"),naked,used,no_instrument_function))
16510 init3_set_eind (void)
16512 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
16513 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
16518 The @code{__trampolines_start} symbol is defined in the linker script.
16521 Stubs are generated automatically by the linker if
16522 the following two conditions are met:
16525 @item The address of a label is taken by means of the @code{gs} modifier
16526 (short for @emph{generate stubs}) like so:
16528 LDI r24, lo8(gs(@var{func}))
16529 LDI r25, hi8(gs(@var{func}))
16531 @item The final location of that label is in a code segment
16532 @emph{outside} the segment where the stubs are located.
16536 The compiler emits such @code{gs} modifiers for code labels in the
16537 following situations:
16539 @item Taking address of a function or code label.
16540 @item Computed goto.
16541 @item If prologue-save function is used, see @option{-mcall-prologues}
16542 command-line option.
16543 @item Switch/case dispatch tables. If you do not want such dispatch
16544 tables you can specify the @option{-fno-jump-tables} command-line option.
16545 @item C and C++ constructors/destructors called during startup/shutdown.
16546 @item If the tools hit a @code{gs()} modifier explained above.
16550 Jumping to non-symbolic addresses like so is @emph{not} supported:
16555 /* Call function at word address 0x2 */
16556 return ((int(*)(void)) 0x2)();
16560 Instead, a stub has to be set up, i.e.@: the function has to be called
16561 through a symbol (@code{func_4} in the example):
16566 extern int func_4 (void);
16568 /* Call function at byte address 0x4 */
16573 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
16574 Alternatively, @code{func_4} can be defined in the linker script.
16577 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
16578 @cindex @code{RAMPD}
16579 @cindex @code{RAMPX}
16580 @cindex @code{RAMPY}
16581 @cindex @code{RAMPZ}
16582 Some AVR devices support memories larger than the 64@tie{}KiB range
16583 that can be accessed with 16-bit pointers. To access memory locations
16584 outside this 64@tie{}KiB range, the content of a @code{RAMP}
16585 register is used as high part of the address:
16586 The @code{X}, @code{Y}, @code{Z} address register is concatenated
16587 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
16588 register, respectively, to get a wide address. Similarly,
16589 @code{RAMPD} is used together with direct addressing.
16593 The startup code initializes the @code{RAMP} special function
16594 registers with zero.
16597 If a @ref{AVR Named Address Spaces,named address space} other than
16598 generic or @code{__flash} is used, then @code{RAMPZ} is set
16599 as needed before the operation.
16602 If the device supports RAM larger than 64@tie{}KiB and the compiler
16603 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
16604 is reset to zero after the operation.
16607 If the device comes with a specific @code{RAMP} register, the ISR
16608 prologue/epilogue saves/restores that SFR and initializes it with
16609 zero in case the ISR code might (implicitly) use it.
16612 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
16613 If you use inline assembler to read from locations outside the
16614 16-bit address range and change one of the @code{RAMP} registers,
16615 you must reset it to zero after the access.
16619 @subsubsection AVR Built-in Macros
16621 GCC defines several built-in macros so that the user code can test
16622 for the presence or absence of features. Almost any of the following
16623 built-in macros are deduced from device capabilities and thus
16624 triggered by the @option{-mmcu=} command-line option.
16626 For even more AVR-specific built-in macros see
16627 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
16632 Build-in macro that resolves to a decimal number that identifies the
16633 architecture and depends on the @option{-mmcu=@var{mcu}} option.
16634 Possible values are:
16636 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
16637 @code{4}, @code{5}, @code{51}, @code{6}
16639 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
16640 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
16645 @code{102}, @code{103}, @code{104},
16646 @code{105}, @code{106}, @code{107}
16648 for @var{mcu}=@code{avrtiny},
16649 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
16650 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
16651 If @var{mcu} specifies a device, this built-in macro is set
16652 accordingly. For example, with @option{-mmcu=atmega8} the macro is
16653 defined to @code{4}.
16655 @item __AVR_@var{Device}__
16656 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
16657 the device's name. For example, @option{-mmcu=atmega8} defines the
16658 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
16659 @code{__AVR_ATtiny261A__}, etc.
16661 The built-in macros' names follow
16662 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
16663 the device name as from the AVR user manual. The difference between
16664 @var{Device} in the built-in macro and @var{device} in
16665 @option{-mmcu=@var{device}} is that the latter is always lowercase.
16667 If @var{device} is not a device but only a core architecture like
16668 @samp{avr51}, this macro is not defined.
16670 @item __AVR_DEVICE_NAME__
16671 Setting @option{-mmcu=@var{device}} defines this built-in macro to
16672 the device's name. For example, with @option{-mmcu=atmega8} the macro
16673 is defined to @code{atmega8}.
16675 If @var{device} is not a device but only a core architecture like
16676 @samp{avr51}, this macro is not defined.
16678 @item __AVR_XMEGA__
16679 The device / architecture belongs to the XMEGA family of devices.
16681 @item __AVR_HAVE_ELPM__
16682 The device has the @code{ELPM} instruction.
16684 @item __AVR_HAVE_ELPMX__
16685 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
16686 R@var{n},Z+} instructions.
16688 @item __AVR_HAVE_MOVW__
16689 The device has the @code{MOVW} instruction to perform 16-bit
16690 register-register moves.
16692 @item __AVR_HAVE_LPMX__
16693 The device has the @code{LPM R@var{n},Z} and
16694 @code{LPM R@var{n},Z+} instructions.
16696 @item __AVR_HAVE_MUL__
16697 The device has a hardware multiplier.
16699 @item __AVR_HAVE_JMP_CALL__
16700 The device has the @code{JMP} and @code{CALL} instructions.
16701 This is the case for devices with more than 8@tie{}KiB of program
16704 @item __AVR_HAVE_EIJMP_EICALL__
16705 @itemx __AVR_3_BYTE_PC__
16706 The device has the @code{EIJMP} and @code{EICALL} instructions.
16707 This is the case for devices with more than 128@tie{}KiB of program memory.
16708 This also means that the program counter
16709 (PC) is 3@tie{}bytes wide.
16711 @item __AVR_2_BYTE_PC__
16712 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
16713 with up to 128@tie{}KiB of program memory.
16715 @item __AVR_HAVE_8BIT_SP__
16716 @itemx __AVR_HAVE_16BIT_SP__
16717 The stack pointer (SP) register is treated as 8-bit respectively
16718 16-bit register by the compiler.
16719 The definition of these macros is affected by @option{-mtiny-stack}.
16721 @item __AVR_HAVE_SPH__
16723 The device has the SPH (high part of stack pointer) special function
16724 register or has an 8-bit stack pointer, respectively.
16725 The definition of these macros is affected by @option{-mmcu=} and
16726 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
16729 @item __AVR_HAVE_RAMPD__
16730 @itemx __AVR_HAVE_RAMPX__
16731 @itemx __AVR_HAVE_RAMPY__
16732 @itemx __AVR_HAVE_RAMPZ__
16733 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
16734 @code{RAMPZ} special function register, respectively.
16736 @item __NO_INTERRUPTS__
16737 This macro reflects the @option{-mno-interrupts} command-line option.
16739 @item __AVR_ERRATA_SKIP__
16740 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
16741 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
16742 instructions because of a hardware erratum. Skip instructions are
16743 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
16744 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
16747 @item __AVR_ISA_RMW__
16748 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
16750 @item __AVR_SFR_OFFSET__=@var{offset}
16751 Instructions that can address I/O special function registers directly
16752 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
16753 address as if addressed by an instruction to access RAM like @code{LD}
16754 or @code{STS}. This offset depends on the device architecture and has
16755 to be subtracted from the RAM address in order to get the
16756 respective I/O@tie{}address.
16758 @item __AVR_SHORT_CALLS__
16759 The @option{-mshort-calls} command line option is set.
16761 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
16762 Some devices support reading from flash memory by means of @code{LD*}
16763 instructions. The flash memory is seen in the data address space
16764 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
16765 is not defined, this feature is not available. If defined,
16766 the address space is linear and there is no need to put
16767 @code{.rodata} into RAM. This is handled by the default linker
16768 description file, and is currently available for
16769 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
16770 there is no need to use address spaces like @code{__flash} or
16771 features like attribute @code{progmem} and @code{pgm_read_*}.
16773 @item __WITH_AVRLIBC__
16774 The compiler is configured to be used together with AVR-Libc.
16775 See the @option{--with-avrlibc} configure option.
16779 @node Blackfin Options
16780 @subsection Blackfin Options
16781 @cindex Blackfin Options
16784 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
16786 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
16787 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
16788 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
16789 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
16790 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
16791 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
16792 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
16793 @samp{bf561}, @samp{bf592}.
16795 The optional @var{sirevision} specifies the silicon revision of the target
16796 Blackfin processor. Any workarounds available for the targeted silicon revision
16797 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
16798 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
16799 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
16800 hexadecimal digits representing the major and minor numbers in the silicon
16801 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
16802 is not defined. If @var{sirevision} is @samp{any}, the
16803 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
16804 If this optional @var{sirevision} is not used, GCC assumes the latest known
16805 silicon revision of the targeted Blackfin processor.
16807 GCC defines a preprocessor macro for the specified @var{cpu}.
16808 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
16809 provided by libgloss to be linked in if @option{-msim} is not given.
16811 Without this option, @samp{bf532} is used as the processor by default.
16813 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
16814 only the preprocessor macro is defined.
16818 Specifies that the program will be run on the simulator. This causes
16819 the simulator BSP provided by libgloss to be linked in. This option
16820 has effect only for @samp{bfin-elf} toolchain.
16821 Certain other options, such as @option{-mid-shared-library} and
16822 @option{-mfdpic}, imply @option{-msim}.
16824 @item -momit-leaf-frame-pointer
16825 @opindex momit-leaf-frame-pointer
16826 Don't keep the frame pointer in a register for leaf functions. This
16827 avoids the instructions to save, set up and restore frame pointers and
16828 makes an extra register available in leaf functions.
16830 @item -mspecld-anomaly
16831 @opindex mspecld-anomaly
16832 When enabled, the compiler ensures that the generated code does not
16833 contain speculative loads after jump instructions. If this option is used,
16834 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
16836 @item -mno-specld-anomaly
16837 @opindex mno-specld-anomaly
16838 Don't generate extra code to prevent speculative loads from occurring.
16840 @item -mcsync-anomaly
16841 @opindex mcsync-anomaly
16842 When enabled, the compiler ensures that the generated code does not
16843 contain CSYNC or SSYNC instructions too soon after conditional branches.
16844 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
16846 @item -mno-csync-anomaly
16847 @opindex mno-csync-anomaly
16848 Don't generate extra code to prevent CSYNC or SSYNC instructions from
16849 occurring too soon after a conditional branch.
16853 When enabled, the compiler is free to take advantage of the knowledge that
16854 the entire program fits into the low 64k of memory.
16857 @opindex mno-low-64k
16858 Assume that the program is arbitrarily large. This is the default.
16860 @item -mstack-check-l1
16861 @opindex mstack-check-l1
16862 Do stack checking using information placed into L1 scratchpad memory by the
16865 @item -mid-shared-library
16866 @opindex mid-shared-library
16867 Generate code that supports shared libraries via the library ID method.
16868 This allows for execute in place and shared libraries in an environment
16869 without virtual memory management. This option implies @option{-fPIC}.
16870 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16872 @item -mno-id-shared-library
16873 @opindex mno-id-shared-library
16874 Generate code that doesn't assume ID-based shared libraries are being used.
16875 This is the default.
16877 @item -mleaf-id-shared-library
16878 @opindex mleaf-id-shared-library
16879 Generate code that supports shared libraries via the library ID method,
16880 but assumes that this library or executable won't link against any other
16881 ID shared libraries. That allows the compiler to use faster code for jumps
16884 @item -mno-leaf-id-shared-library
16885 @opindex mno-leaf-id-shared-library
16886 Do not assume that the code being compiled won't link against any ID shared
16887 libraries. Slower code is generated for jump and call insns.
16889 @item -mshared-library-id=n
16890 @opindex mshared-library-id
16891 Specifies the identification number of the ID-based shared library being
16892 compiled. Specifying a value of 0 generates more compact code; specifying
16893 other values forces the allocation of that number to the current
16894 library but is no more space- or time-efficient than omitting this option.
16898 Generate code that allows the data segment to be located in a different
16899 area of memory from the text segment. This allows for execute in place in
16900 an environment without virtual memory management by eliminating relocations
16901 against the text section.
16903 @item -mno-sep-data
16904 @opindex mno-sep-data
16905 Generate code that assumes that the data segment follows the text segment.
16906 This is the default.
16909 @itemx -mno-long-calls
16910 @opindex mlong-calls
16911 @opindex mno-long-calls
16912 Tells the compiler to perform function calls by first loading the
16913 address of the function into a register and then performing a subroutine
16914 call on this register. This switch is needed if the target function
16915 lies outside of the 24-bit addressing range of the offset-based
16916 version of subroutine call instruction.
16918 This feature is not enabled by default. Specifying
16919 @option{-mno-long-calls} restores the default behavior. Note these
16920 switches have no effect on how the compiler generates code to handle
16921 function calls via function pointers.
16925 Link with the fast floating-point library. This library relaxes some of
16926 the IEEE floating-point standard's rules for checking inputs against
16927 Not-a-Number (NAN), in the interest of performance.
16930 @opindex minline-plt
16931 Enable inlining of PLT entries in function calls to functions that are
16932 not known to bind locally. It has no effect without @option{-mfdpic}.
16935 @opindex mmulticore
16936 Build a standalone application for multicore Blackfin processors.
16937 This option causes proper start files and link scripts supporting
16938 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
16939 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
16941 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
16942 selects the one-application-per-core programming model. Without
16943 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
16944 programming model is used. In this model, the main function of Core B
16945 should be named as @code{coreb_main}.
16947 If this option is not used, the single-core application programming
16952 Build a standalone application for Core A of BF561 when using
16953 the one-application-per-core programming model. Proper start files
16954 and link scripts are used to support Core A, and the macro
16955 @code{__BFIN_COREA} is defined.
16956 This option can only be used in conjunction with @option{-mmulticore}.
16960 Build a standalone application for Core B of BF561 when using
16961 the one-application-per-core programming model. Proper start files
16962 and link scripts are used to support Core B, and the macro
16963 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
16964 should be used instead of @code{main}.
16965 This option can only be used in conjunction with @option{-mmulticore}.
16969 Build a standalone application for SDRAM. Proper start files and
16970 link scripts are used to put the application into SDRAM, and the macro
16971 @code{__BFIN_SDRAM} is defined.
16972 The loader should initialize SDRAM before loading the application.
16976 Assume that ICPLBs are enabled at run time. This has an effect on certain
16977 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
16978 are enabled; for standalone applications the default is off.
16982 @subsection C6X Options
16983 @cindex C6X Options
16986 @item -march=@var{name}
16988 This specifies the name of the target architecture. GCC uses this
16989 name to determine what kind of instructions it can emit when generating
16990 assembly code. Permissible names are: @samp{c62x},
16991 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
16994 @opindex mbig-endian
16995 Generate code for a big-endian target.
16997 @item -mlittle-endian
16998 @opindex mlittle-endian
16999 Generate code for a little-endian target. This is the default.
17003 Choose startup files and linker script suitable for the simulator.
17005 @item -msdata=default
17006 @opindex msdata=default
17007 Put small global and static data in the @code{.neardata} section,
17008 which is pointed to by register @code{B14}. Put small uninitialized
17009 global and static data in the @code{.bss} section, which is adjacent
17010 to the @code{.neardata} section. Put small read-only data into the
17011 @code{.rodata} section. The corresponding sections used for large
17012 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
17015 @opindex msdata=all
17016 Put all data, not just small objects, into the sections reserved for
17017 small data, and use addressing relative to the @code{B14} register to
17021 @opindex msdata=none
17022 Make no use of the sections reserved for small data, and use absolute
17023 addresses to access all data. Put all initialized global and static
17024 data in the @code{.fardata} section, and all uninitialized data in the
17025 @code{.far} section. Put all constant data into the @code{.const}
17030 @subsection CRIS Options
17031 @cindex CRIS Options
17033 These options are defined specifically for the CRIS ports.
17036 @item -march=@var{architecture-type}
17037 @itemx -mcpu=@var{architecture-type}
17040 Generate code for the specified architecture. The choices for
17041 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
17042 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
17043 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
17046 @item -mtune=@var{architecture-type}
17048 Tune to @var{architecture-type} everything applicable about the generated
17049 code, except for the ABI and the set of available instructions. The
17050 choices for @var{architecture-type} are the same as for
17051 @option{-march=@var{architecture-type}}.
17053 @item -mmax-stack-frame=@var{n}
17054 @opindex mmax-stack-frame
17055 Warn when the stack frame of a function exceeds @var{n} bytes.
17061 The options @option{-metrax4} and @option{-metrax100} are synonyms for
17062 @option{-march=v3} and @option{-march=v8} respectively.
17064 @item -mmul-bug-workaround
17065 @itemx -mno-mul-bug-workaround
17066 @opindex mmul-bug-workaround
17067 @opindex mno-mul-bug-workaround
17068 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
17069 models where it applies. This option is active by default.
17073 Enable CRIS-specific verbose debug-related information in the assembly
17074 code. This option also has the effect of turning off the @samp{#NO_APP}
17075 formatted-code indicator to the assembler at the beginning of the
17080 Do not use condition-code results from previous instruction; always emit
17081 compare and test instructions before use of condition codes.
17083 @item -mno-side-effects
17084 @opindex mno-side-effects
17085 Do not emit instructions with side effects in addressing modes other than
17088 @item -mstack-align
17089 @itemx -mno-stack-align
17090 @itemx -mdata-align
17091 @itemx -mno-data-align
17092 @itemx -mconst-align
17093 @itemx -mno-const-align
17094 @opindex mstack-align
17095 @opindex mno-stack-align
17096 @opindex mdata-align
17097 @opindex mno-data-align
17098 @opindex mconst-align
17099 @opindex mno-const-align
17100 These options (@samp{no-} options) arrange (eliminate arrangements) for the
17101 stack frame, individual data and constants to be aligned for the maximum
17102 single data access size for the chosen CPU model. The default is to
17103 arrange for 32-bit alignment. ABI details such as structure layout are
17104 not affected by these options.
17112 Similar to the stack- data- and const-align options above, these options
17113 arrange for stack frame, writable data and constants to all be 32-bit,
17114 16-bit or 8-bit aligned. The default is 32-bit alignment.
17116 @item -mno-prologue-epilogue
17117 @itemx -mprologue-epilogue
17118 @opindex mno-prologue-epilogue
17119 @opindex mprologue-epilogue
17120 With @option{-mno-prologue-epilogue}, the normal function prologue and
17121 epilogue which set up the stack frame are omitted and no return
17122 instructions or return sequences are generated in the code. Use this
17123 option only together with visual inspection of the compiled code: no
17124 warnings or errors are generated when call-saved registers must be saved,
17125 or storage for local variables needs to be allocated.
17129 @opindex mno-gotplt
17131 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
17132 instruction sequences that load addresses for functions from the PLT part
17133 of the GOT rather than (traditional on other architectures) calls to the
17134 PLT@. The default is @option{-mgotplt}.
17138 Legacy no-op option only recognized with the cris-axis-elf and
17139 cris-axis-linux-gnu targets.
17143 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
17147 This option, recognized for the cris-axis-elf, arranges
17148 to link with input-output functions from a simulator library. Code,
17149 initialized data and zero-initialized data are allocated consecutively.
17153 Like @option{-sim}, but pass linker options to locate initialized data at
17154 0x40000000 and zero-initialized data at 0x80000000.
17158 @subsection CR16 Options
17159 @cindex CR16 Options
17161 These options are defined specifically for the CR16 ports.
17167 Enable the use of multiply-accumulate instructions. Disabled by default.
17171 @opindex mcr16cplus
17173 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
17178 Links the library libsim.a which is in compatible with simulator. Applicable
17179 to ELF compiler only.
17183 Choose integer type as 32-bit wide.
17187 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
17189 @item -mdata-model=@var{model}
17190 @opindex mdata-model
17191 Choose a data model. The choices for @var{model} are @samp{near},
17192 @samp{far} or @samp{medium}. @samp{medium} is default.
17193 However, @samp{far} is not valid with @option{-mcr16c}, as the
17194 CR16C architecture does not support the far data model.
17197 @node Darwin Options
17198 @subsection Darwin Options
17199 @cindex Darwin options
17201 These options are defined for all architectures running the Darwin operating
17204 FSF GCC on Darwin does not create ``fat'' object files; it creates
17205 an object file for the single architecture that GCC was built to
17206 target. Apple's GCC on Darwin does create ``fat'' files if multiple
17207 @option{-arch} options are used; it does so by running the compiler or
17208 linker multiple times and joining the results together with
17211 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
17212 @samp{i686}) is determined by the flags that specify the ISA
17213 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
17214 @option{-force_cpusubtype_ALL} option can be used to override this.
17216 The Darwin tools vary in their behavior when presented with an ISA
17217 mismatch. The assembler, @file{as}, only permits instructions to
17218 be used that are valid for the subtype of the file it is generating,
17219 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
17220 The linker for shared libraries, @file{/usr/bin/libtool}, fails
17221 and prints an error if asked to create a shared library with a less
17222 restrictive subtype than its input files (for instance, trying to put
17223 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
17224 for executables, @command{ld}, quietly gives the executable the most
17225 restrictive subtype of any of its input files.
17230 Add the framework directory @var{dir} to the head of the list of
17231 directories to be searched for header files. These directories are
17232 interleaved with those specified by @option{-I} options and are
17233 scanned in a left-to-right order.
17235 A framework directory is a directory with frameworks in it. A
17236 framework is a directory with a @file{Headers} and/or
17237 @file{PrivateHeaders} directory contained directly in it that ends
17238 in @file{.framework}. The name of a framework is the name of this
17239 directory excluding the @file{.framework}. Headers associated with
17240 the framework are found in one of those two directories, with
17241 @file{Headers} being searched first. A subframework is a framework
17242 directory that is in a framework's @file{Frameworks} directory.
17243 Includes of subframework headers can only appear in a header of a
17244 framework that contains the subframework, or in a sibling subframework
17245 header. Two subframeworks are siblings if they occur in the same
17246 framework. A subframework should not have the same name as a
17247 framework; a warning is issued if this is violated. Currently a
17248 subframework cannot have subframeworks; in the future, the mechanism
17249 may be extended to support this. The standard frameworks can be found
17250 in @file{/System/Library/Frameworks} and
17251 @file{/Library/Frameworks}. An example include looks like
17252 @code{#include <Framework/header.h>}, where @file{Framework} denotes
17253 the name of the framework and @file{header.h} is found in the
17254 @file{PrivateHeaders} or @file{Headers} directory.
17256 @item -iframework@var{dir}
17257 @opindex iframework
17258 Like @option{-F} except the directory is a treated as a system
17259 directory. The main difference between this @option{-iframework} and
17260 @option{-F} is that with @option{-iframework} the compiler does not
17261 warn about constructs contained within header files found via
17262 @var{dir}. This option is valid only for the C family of languages.
17266 Emit debugging information for symbols that are used. For stabs
17267 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
17268 This is by default ON@.
17272 Emit debugging information for all symbols and types.
17274 @item -mmacosx-version-min=@var{version}
17275 The earliest version of MacOS X that this executable will run on
17276 is @var{version}. Typical values of @var{version} include @code{10.1},
17277 @code{10.2}, and @code{10.3.9}.
17279 If the compiler was built to use the system's headers by default,
17280 then the default for this option is the system version on which the
17281 compiler is running, otherwise the default is to make choices that
17282 are compatible with as many systems and code bases as possible.
17286 Enable kernel development mode. The @option{-mkernel} option sets
17287 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
17288 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
17289 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
17290 applicable. This mode also sets @option{-mno-altivec},
17291 @option{-msoft-float}, @option{-fno-builtin} and
17292 @option{-mlong-branch} for PowerPC targets.
17294 @item -mone-byte-bool
17295 @opindex mone-byte-bool
17296 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
17297 By default @code{sizeof(bool)} is @code{4} when compiling for
17298 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
17299 option has no effect on x86.
17301 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
17302 to generate code that is not binary compatible with code generated
17303 without that switch. Using this switch may require recompiling all
17304 other modules in a program, including system libraries. Use this
17305 switch to conform to a non-default data model.
17307 @item -mfix-and-continue
17308 @itemx -ffix-and-continue
17309 @itemx -findirect-data
17310 @opindex mfix-and-continue
17311 @opindex ffix-and-continue
17312 @opindex findirect-data
17313 Generate code suitable for fast turnaround development, such as to
17314 allow GDB to dynamically load @file{.o} files into already-running
17315 programs. @option{-findirect-data} and @option{-ffix-and-continue}
17316 are provided for backwards compatibility.
17320 Loads all members of static archive libraries.
17321 See man ld(1) for more information.
17323 @item -arch_errors_fatal
17324 @opindex arch_errors_fatal
17325 Cause the errors having to do with files that have the wrong architecture
17328 @item -bind_at_load
17329 @opindex bind_at_load
17330 Causes the output file to be marked such that the dynamic linker will
17331 bind all undefined references when the file is loaded or launched.
17335 Produce a Mach-o bundle format file.
17336 See man ld(1) for more information.
17338 @item -bundle_loader @var{executable}
17339 @opindex bundle_loader
17340 This option specifies the @var{executable} that will load the build
17341 output file being linked. See man ld(1) for more information.
17344 @opindex dynamiclib
17345 When passed this option, GCC produces a dynamic library instead of
17346 an executable when linking, using the Darwin @file{libtool} command.
17348 @item -force_cpusubtype_ALL
17349 @opindex force_cpusubtype_ALL
17350 This causes GCC's output file to have the @samp{ALL} subtype, instead of
17351 one controlled by the @option{-mcpu} or @option{-march} option.
17353 @item -allowable_client @var{client_name}
17354 @itemx -client_name
17355 @itemx -compatibility_version
17356 @itemx -current_version
17358 @itemx -dependency-file
17360 @itemx -dylinker_install_name
17362 @itemx -exported_symbols_list
17365 @itemx -flat_namespace
17366 @itemx -force_flat_namespace
17367 @itemx -headerpad_max_install_names
17370 @itemx -install_name
17371 @itemx -keep_private_externs
17372 @itemx -multi_module
17373 @itemx -multiply_defined
17374 @itemx -multiply_defined_unused
17377 @itemx -no_dead_strip_inits_and_terms
17378 @itemx -nofixprebinding
17379 @itemx -nomultidefs
17381 @itemx -noseglinkedit
17382 @itemx -pagezero_size
17384 @itemx -prebind_all_twolevel_modules
17385 @itemx -private_bundle
17387 @itemx -read_only_relocs
17389 @itemx -sectobjectsymbols
17393 @itemx -sectobjectsymbols
17396 @itemx -segs_read_only_addr
17398 @itemx -segs_read_write_addr
17399 @itemx -seg_addr_table
17400 @itemx -seg_addr_table_filename
17401 @itemx -seglinkedit
17403 @itemx -segs_read_only_addr
17404 @itemx -segs_read_write_addr
17405 @itemx -single_module
17407 @itemx -sub_library
17409 @itemx -sub_umbrella
17410 @itemx -twolevel_namespace
17413 @itemx -unexported_symbols_list
17414 @itemx -weak_reference_mismatches
17415 @itemx -whatsloaded
17416 @opindex allowable_client
17417 @opindex client_name
17418 @opindex compatibility_version
17419 @opindex current_version
17420 @opindex dead_strip
17421 @opindex dependency-file
17422 @opindex dylib_file
17423 @opindex dylinker_install_name
17425 @opindex exported_symbols_list
17427 @opindex flat_namespace
17428 @opindex force_flat_namespace
17429 @opindex headerpad_max_install_names
17430 @opindex image_base
17432 @opindex install_name
17433 @opindex keep_private_externs
17434 @opindex multi_module
17435 @opindex multiply_defined
17436 @opindex multiply_defined_unused
17437 @opindex noall_load
17438 @opindex no_dead_strip_inits_and_terms
17439 @opindex nofixprebinding
17440 @opindex nomultidefs
17442 @opindex noseglinkedit
17443 @opindex pagezero_size
17445 @opindex prebind_all_twolevel_modules
17446 @opindex private_bundle
17447 @opindex read_only_relocs
17449 @opindex sectobjectsymbols
17452 @opindex sectcreate
17453 @opindex sectobjectsymbols
17456 @opindex segs_read_only_addr
17457 @opindex segs_read_write_addr
17458 @opindex seg_addr_table
17459 @opindex seg_addr_table_filename
17460 @opindex seglinkedit
17462 @opindex segs_read_only_addr
17463 @opindex segs_read_write_addr
17464 @opindex single_module
17466 @opindex sub_library
17467 @opindex sub_umbrella
17468 @opindex twolevel_namespace
17471 @opindex unexported_symbols_list
17472 @opindex weak_reference_mismatches
17473 @opindex whatsloaded
17474 These options are passed to the Darwin linker. The Darwin linker man page
17475 describes them in detail.
17478 @node DEC Alpha Options
17479 @subsection DEC Alpha Options
17481 These @samp{-m} options are defined for the DEC Alpha implementations:
17484 @item -mno-soft-float
17485 @itemx -msoft-float
17486 @opindex mno-soft-float
17487 @opindex msoft-float
17488 Use (do not use) the hardware floating-point instructions for
17489 floating-point operations. When @option{-msoft-float} is specified,
17490 functions in @file{libgcc.a} are used to perform floating-point
17491 operations. Unless they are replaced by routines that emulate the
17492 floating-point operations, or compiled in such a way as to call such
17493 emulations routines, these routines issue floating-point
17494 operations. If you are compiling for an Alpha without floating-point
17495 operations, you must ensure that the library is built so as not to call
17498 Note that Alpha implementations without floating-point operations are
17499 required to have floating-point registers.
17502 @itemx -mno-fp-regs
17504 @opindex mno-fp-regs
17505 Generate code that uses (does not use) the floating-point register set.
17506 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
17507 register set is not used, floating-point operands are passed in integer
17508 registers as if they were integers and floating-point results are passed
17509 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
17510 so any function with a floating-point argument or return value called by code
17511 compiled with @option{-mno-fp-regs} must also be compiled with that
17514 A typical use of this option is building a kernel that does not use,
17515 and hence need not save and restore, any floating-point registers.
17519 The Alpha architecture implements floating-point hardware optimized for
17520 maximum performance. It is mostly compliant with the IEEE floating-point
17521 standard. However, for full compliance, software assistance is
17522 required. This option generates code fully IEEE-compliant code
17523 @emph{except} that the @var{inexact-flag} is not maintained (see below).
17524 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
17525 defined during compilation. The resulting code is less efficient but is
17526 able to correctly support denormalized numbers and exceptional IEEE
17527 values such as not-a-number and plus/minus infinity. Other Alpha
17528 compilers call this option @option{-ieee_with_no_inexact}.
17530 @item -mieee-with-inexact
17531 @opindex mieee-with-inexact
17532 This is like @option{-mieee} except the generated code also maintains
17533 the IEEE @var{inexact-flag}. Turning on this option causes the
17534 generated code to implement fully-compliant IEEE math. In addition to
17535 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
17536 macro. On some Alpha implementations the resulting code may execute
17537 significantly slower than the code generated by default. Since there is
17538 very little code that depends on the @var{inexact-flag}, you should
17539 normally not specify this option. Other Alpha compilers call this
17540 option @option{-ieee_with_inexact}.
17542 @item -mfp-trap-mode=@var{trap-mode}
17543 @opindex mfp-trap-mode
17544 This option controls what floating-point related traps are enabled.
17545 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
17546 The trap mode can be set to one of four values:
17550 This is the default (normal) setting. The only traps that are enabled
17551 are the ones that cannot be disabled in software (e.g., division by zero
17555 In addition to the traps enabled by @samp{n}, underflow traps are enabled
17559 Like @samp{u}, but the instructions are marked to be safe for software
17560 completion (see Alpha architecture manual for details).
17563 Like @samp{su}, but inexact traps are enabled as well.
17566 @item -mfp-rounding-mode=@var{rounding-mode}
17567 @opindex mfp-rounding-mode
17568 Selects the IEEE rounding mode. Other Alpha compilers call this option
17569 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
17574 Normal IEEE rounding mode. Floating-point numbers are rounded towards
17575 the nearest machine number or towards the even machine number in case
17579 Round towards minus infinity.
17582 Chopped rounding mode. Floating-point numbers are rounded towards zero.
17585 Dynamic rounding mode. A field in the floating-point control register
17586 (@var{fpcr}, see Alpha architecture reference manual) controls the
17587 rounding mode in effect. The C library initializes this register for
17588 rounding towards plus infinity. Thus, unless your program modifies the
17589 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
17592 @item -mtrap-precision=@var{trap-precision}
17593 @opindex mtrap-precision
17594 In the Alpha architecture, floating-point traps are imprecise. This
17595 means without software assistance it is impossible to recover from a
17596 floating trap and program execution normally needs to be terminated.
17597 GCC can generate code that can assist operating system trap handlers
17598 in determining the exact location that caused a floating-point trap.
17599 Depending on the requirements of an application, different levels of
17600 precisions can be selected:
17604 Program precision. This option is the default and means a trap handler
17605 can only identify which program caused a floating-point exception.
17608 Function precision. The trap handler can determine the function that
17609 caused a floating-point exception.
17612 Instruction precision. The trap handler can determine the exact
17613 instruction that caused a floating-point exception.
17616 Other Alpha compilers provide the equivalent options called
17617 @option{-scope_safe} and @option{-resumption_safe}.
17619 @item -mieee-conformant
17620 @opindex mieee-conformant
17621 This option marks the generated code as IEEE conformant. You must not
17622 use this option unless you also specify @option{-mtrap-precision=i} and either
17623 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
17624 is to emit the line @samp{.eflag 48} in the function prologue of the
17625 generated assembly file.
17627 @item -mbuild-constants
17628 @opindex mbuild-constants
17629 Normally GCC examines a 32- or 64-bit integer constant to
17630 see if it can construct it from smaller constants in two or three
17631 instructions. If it cannot, it outputs the constant as a literal and
17632 generates code to load it from the data segment at run time.
17634 Use this option to require GCC to construct @emph{all} integer constants
17635 using code, even if it takes more instructions (the maximum is six).
17637 You typically use this option to build a shared library dynamic
17638 loader. Itself a shared library, it must relocate itself in memory
17639 before it can find the variables and constants in its own data segment.
17657 Indicate whether GCC should generate code to use the optional BWX,
17658 CIX, FIX and MAX instruction sets. The default is to use the instruction
17659 sets supported by the CPU type specified via @option{-mcpu=} option or that
17660 of the CPU on which GCC was built if none is specified.
17663 @itemx -mfloat-ieee
17664 @opindex mfloat-vax
17665 @opindex mfloat-ieee
17666 Generate code that uses (does not use) VAX F and G floating-point
17667 arithmetic instead of IEEE single and double precision.
17669 @item -mexplicit-relocs
17670 @itemx -mno-explicit-relocs
17671 @opindex mexplicit-relocs
17672 @opindex mno-explicit-relocs
17673 Older Alpha assemblers provided no way to generate symbol relocations
17674 except via assembler macros. Use of these macros does not allow
17675 optimal instruction scheduling. GNU binutils as of version 2.12
17676 supports a new syntax that allows the compiler to explicitly mark
17677 which relocations should apply to which instructions. This option
17678 is mostly useful for debugging, as GCC detects the capabilities of
17679 the assembler when it is built and sets the default accordingly.
17682 @itemx -mlarge-data
17683 @opindex msmall-data
17684 @opindex mlarge-data
17685 When @option{-mexplicit-relocs} is in effect, static data is
17686 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
17687 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
17688 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
17689 16-bit relocations off of the @code{$gp} register. This limits the
17690 size of the small data area to 64KB, but allows the variables to be
17691 directly accessed via a single instruction.
17693 The default is @option{-mlarge-data}. With this option the data area
17694 is limited to just below 2GB@. Programs that require more than 2GB of
17695 data must use @code{malloc} or @code{mmap} to allocate the data in the
17696 heap instead of in the program's data segment.
17698 When generating code for shared libraries, @option{-fpic} implies
17699 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
17702 @itemx -mlarge-text
17703 @opindex msmall-text
17704 @opindex mlarge-text
17705 When @option{-msmall-text} is used, the compiler assumes that the
17706 code of the entire program (or shared library) fits in 4MB, and is
17707 thus reachable with a branch instruction. When @option{-msmall-data}
17708 is used, the compiler can assume that all local symbols share the
17709 same @code{$gp} value, and thus reduce the number of instructions
17710 required for a function call from 4 to 1.
17712 The default is @option{-mlarge-text}.
17714 @item -mcpu=@var{cpu_type}
17716 Set the instruction set and instruction scheduling parameters for
17717 machine type @var{cpu_type}. You can specify either the @samp{EV}
17718 style name or the corresponding chip number. GCC supports scheduling
17719 parameters for the EV4, EV5 and EV6 family of processors and
17720 chooses the default values for the instruction set from the processor
17721 you specify. If you do not specify a processor type, GCC defaults
17722 to the processor on which the compiler was built.
17724 Supported values for @var{cpu_type} are
17730 Schedules as an EV4 and has no instruction set extensions.
17734 Schedules as an EV5 and has no instruction set extensions.
17738 Schedules as an EV5 and supports the BWX extension.
17743 Schedules as an EV5 and supports the BWX and MAX extensions.
17747 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
17751 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
17754 Native toolchains also support the value @samp{native},
17755 which selects the best architecture option for the host processor.
17756 @option{-mcpu=native} has no effect if GCC does not recognize
17759 @item -mtune=@var{cpu_type}
17761 Set only the instruction scheduling parameters for machine type
17762 @var{cpu_type}. The instruction set is not changed.
17764 Native toolchains also support the value @samp{native},
17765 which selects the best architecture option for the host processor.
17766 @option{-mtune=native} has no effect if GCC does not recognize
17769 @item -mmemory-latency=@var{time}
17770 @opindex mmemory-latency
17771 Sets the latency the scheduler should assume for typical memory
17772 references as seen by the application. This number is highly
17773 dependent on the memory access patterns used by the application
17774 and the size of the external cache on the machine.
17776 Valid options for @var{time} are
17780 A decimal number representing clock cycles.
17786 The compiler contains estimates of the number of clock cycles for
17787 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
17788 (also called Dcache, Scache, and Bcache), as well as to main memory.
17789 Note that L3 is only valid for EV5.
17795 @subsection FR30 Options
17796 @cindex FR30 Options
17798 These options are defined specifically for the FR30 port.
17802 @item -msmall-model
17803 @opindex msmall-model
17804 Use the small address space model. This can produce smaller code, but
17805 it does assume that all symbolic values and addresses fit into a
17810 Assume that runtime support has been provided and so there is no need
17811 to include the simulator library (@file{libsim.a}) on the linker
17817 @subsection FT32 Options
17818 @cindex FT32 Options
17820 These options are defined specifically for the FT32 port.
17826 Specifies that the program will be run on the simulator. This causes
17827 an alternate runtime startup and library to be linked.
17828 You must not use this option when generating programs that will run on
17829 real hardware; you must provide your own runtime library for whatever
17830 I/O functions are needed.
17834 Enable Local Register Allocation. This is still experimental for FT32,
17835 so by default the compiler uses standard reload.
17839 Do not use div and mod instructions.
17843 Enable use of the extended instructions of the FT32B processor.
17847 Compress all code using the Ft32B code compression scheme.
17851 Do not generate code that reads program memory.
17856 @subsection FRV Options
17857 @cindex FRV Options
17863 Only use the first 32 general-purpose registers.
17868 Use all 64 general-purpose registers.
17873 Use only the first 32 floating-point registers.
17878 Use all 64 floating-point registers.
17881 @opindex mhard-float
17883 Use hardware instructions for floating-point operations.
17886 @opindex msoft-float
17888 Use library routines for floating-point operations.
17893 Dynamically allocate condition code registers.
17898 Do not try to dynamically allocate condition code registers, only
17899 use @code{icc0} and @code{fcc0}.
17904 Change ABI to use double word insns.
17909 Do not use double word instructions.
17914 Use floating-point double instructions.
17917 @opindex mno-double
17919 Do not use floating-point double instructions.
17924 Use media instructions.
17929 Do not use media instructions.
17934 Use multiply and add/subtract instructions.
17937 @opindex mno-muladd
17939 Do not use multiply and add/subtract instructions.
17944 Select the FDPIC ABI, which uses function descriptors to represent
17945 pointers to functions. Without any PIC/PIE-related options, it
17946 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
17947 assumes GOT entries and small data are within a 12-bit range from the
17948 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
17949 are computed with 32 bits.
17950 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17953 @opindex minline-plt
17955 Enable inlining of PLT entries in function calls to functions that are
17956 not known to bind locally. It has no effect without @option{-mfdpic}.
17957 It's enabled by default if optimizing for speed and compiling for
17958 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
17959 optimization option such as @option{-O3} or above is present in the
17965 Assume a large TLS segment when generating thread-local code.
17970 Do not assume a large TLS segment when generating thread-local code.
17975 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
17976 that is known to be in read-only sections. It's enabled by default,
17977 except for @option{-fpic} or @option{-fpie}: even though it may help
17978 make the global offset table smaller, it trades 1 instruction for 4.
17979 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
17980 one of which may be shared by multiple symbols, and it avoids the need
17981 for a GOT entry for the referenced symbol, so it's more likely to be a
17982 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
17984 @item -multilib-library-pic
17985 @opindex multilib-library-pic
17987 Link with the (library, not FD) pic libraries. It's implied by
17988 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
17989 @option{-fpic} without @option{-mfdpic}. You should never have to use
17993 @opindex mlinked-fp
17995 Follow the EABI requirement of always creating a frame pointer whenever
17996 a stack frame is allocated. This option is enabled by default and can
17997 be disabled with @option{-mno-linked-fp}.
18000 @opindex mlong-calls
18002 Use indirect addressing to call functions outside the current
18003 compilation unit. This allows the functions to be placed anywhere
18004 within the 32-bit address space.
18006 @item -malign-labels
18007 @opindex malign-labels
18009 Try to align labels to an 8-byte boundary by inserting NOPs into the
18010 previous packet. This option only has an effect when VLIW packing
18011 is enabled. It doesn't create new packets; it merely adds NOPs to
18014 @item -mlibrary-pic
18015 @opindex mlibrary-pic
18017 Generate position-independent EABI code.
18022 Use only the first four media accumulator registers.
18027 Use all eight media accumulator registers.
18032 Pack VLIW instructions.
18037 Do not pack VLIW instructions.
18040 @opindex mno-eflags
18042 Do not mark ABI switches in e_flags.
18045 @opindex mcond-move
18047 Enable the use of conditional-move instructions (default).
18049 This switch is mainly for debugging the compiler and will likely be removed
18050 in a future version.
18052 @item -mno-cond-move
18053 @opindex mno-cond-move
18055 Disable the use of conditional-move instructions.
18057 This switch is mainly for debugging the compiler and will likely be removed
18058 in a future version.
18063 Enable the use of conditional set instructions (default).
18065 This switch is mainly for debugging the compiler and will likely be removed
18066 in a future version.
18071 Disable the use of conditional set instructions.
18073 This switch is mainly for debugging the compiler and will likely be removed
18074 in a future version.
18077 @opindex mcond-exec
18079 Enable the use of conditional execution (default).
18081 This switch is mainly for debugging the compiler and will likely be removed
18082 in a future version.
18084 @item -mno-cond-exec
18085 @opindex mno-cond-exec
18087 Disable the use of conditional execution.
18089 This switch is mainly for debugging the compiler and will likely be removed
18090 in a future version.
18092 @item -mvliw-branch
18093 @opindex mvliw-branch
18095 Run a pass to pack branches into VLIW instructions (default).
18097 This switch is mainly for debugging the compiler and will likely be removed
18098 in a future version.
18100 @item -mno-vliw-branch
18101 @opindex mno-vliw-branch
18103 Do not run a pass to pack branches into VLIW instructions.
18105 This switch is mainly for debugging the compiler and will likely be removed
18106 in a future version.
18108 @item -mmulti-cond-exec
18109 @opindex mmulti-cond-exec
18111 Enable optimization of @code{&&} and @code{||} in conditional execution
18114 This switch is mainly for debugging the compiler and will likely be removed
18115 in a future version.
18117 @item -mno-multi-cond-exec
18118 @opindex mno-multi-cond-exec
18120 Disable optimization of @code{&&} and @code{||} in conditional execution.
18122 This switch is mainly for debugging the compiler and will likely be removed
18123 in a future version.
18125 @item -mnested-cond-exec
18126 @opindex mnested-cond-exec
18128 Enable nested conditional execution optimizations (default).
18130 This switch is mainly for debugging the compiler and will likely be removed
18131 in a future version.
18133 @item -mno-nested-cond-exec
18134 @opindex mno-nested-cond-exec
18136 Disable nested conditional execution optimizations.
18138 This switch is mainly for debugging the compiler and will likely be removed
18139 in a future version.
18141 @item -moptimize-membar
18142 @opindex moptimize-membar
18144 This switch removes redundant @code{membar} instructions from the
18145 compiler-generated code. It is enabled by default.
18147 @item -mno-optimize-membar
18148 @opindex mno-optimize-membar
18150 This switch disables the automatic removal of redundant @code{membar}
18151 instructions from the generated code.
18153 @item -mtomcat-stats
18154 @opindex mtomcat-stats
18156 Cause gas to print out tomcat statistics.
18158 @item -mcpu=@var{cpu}
18161 Select the processor type for which to generate code. Possible values are
18162 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
18163 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
18167 @node GNU/Linux Options
18168 @subsection GNU/Linux Options
18170 These @samp{-m} options are defined for GNU/Linux targets:
18175 Use the GNU C library. This is the default except
18176 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
18177 @samp{*-*-linux-*android*} targets.
18181 Use uClibc C library. This is the default on
18182 @samp{*-*-linux-*uclibc*} targets.
18186 Use the musl C library. This is the default on
18187 @samp{*-*-linux-*musl*} targets.
18191 Use Bionic C library. This is the default on
18192 @samp{*-*-linux-*android*} targets.
18196 Compile code compatible with Android platform. This is the default on
18197 @samp{*-*-linux-*android*} targets.
18199 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
18200 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
18201 this option makes the GCC driver pass Android-specific options to the linker.
18202 Finally, this option causes the preprocessor macro @code{__ANDROID__}
18205 @item -tno-android-cc
18206 @opindex tno-android-cc
18207 Disable compilation effects of @option{-mandroid}, i.e., do not enable
18208 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
18209 @option{-fno-rtti} by default.
18211 @item -tno-android-ld
18212 @opindex tno-android-ld
18213 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
18214 linking options to the linker.
18218 @node H8/300 Options
18219 @subsection H8/300 Options
18221 These @samp{-m} options are defined for the H8/300 implementations:
18226 Shorten some address references at link time, when possible; uses the
18227 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
18228 ld, Using ld}, for a fuller description.
18232 Generate code for the H8/300H@.
18236 Generate code for the H8S@.
18240 Generate code for the H8S and H8/300H in the normal mode. This switch
18241 must be used either with @option{-mh} or @option{-ms}.
18245 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
18249 Extended registers are stored on stack before execution of function
18250 with monitor attribute. Default option is @option{-mexr}.
18251 This option is valid only for H8S targets.
18255 Extended registers are not stored on stack before execution of function
18256 with monitor attribute. Default option is @option{-mno-exr}.
18257 This option is valid only for H8S targets.
18261 Make @code{int} data 32 bits by default.
18264 @opindex malign-300
18265 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
18266 The default for the H8/300H and H8S is to align longs and floats on
18268 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
18269 This option has no effect on the H8/300.
18273 @subsection HPPA Options
18274 @cindex HPPA Options
18276 These @samp{-m} options are defined for the HPPA family of computers:
18279 @item -march=@var{architecture-type}
18281 Generate code for the specified architecture. The choices for
18282 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
18283 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
18284 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
18285 architecture option for your machine. Code compiled for lower numbered
18286 architectures runs on higher numbered architectures, but not the
18289 @item -mpa-risc-1-0
18290 @itemx -mpa-risc-1-1
18291 @itemx -mpa-risc-2-0
18292 @opindex mpa-risc-1-0
18293 @opindex mpa-risc-1-1
18294 @opindex mpa-risc-2-0
18295 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
18297 @item -mcaller-copies
18298 @opindex mcaller-copies
18299 The caller copies function arguments passed by hidden reference. This
18300 option should be used with care as it is not compatible with the default
18301 32-bit runtime. However, only aggregates larger than eight bytes are
18302 passed by hidden reference and the option provides better compatibility
18305 @item -mjump-in-delay
18306 @opindex mjump-in-delay
18307 This option is ignored and provided for compatibility purposes only.
18309 @item -mdisable-fpregs
18310 @opindex mdisable-fpregs
18311 Prevent floating-point registers from being used in any manner. This is
18312 necessary for compiling kernels that perform lazy context switching of
18313 floating-point registers. If you use this option and attempt to perform
18314 floating-point operations, the compiler aborts.
18316 @item -mdisable-indexing
18317 @opindex mdisable-indexing
18318 Prevent the compiler from using indexing address modes. This avoids some
18319 rather obscure problems when compiling MIG generated code under MACH@.
18321 @item -mno-space-regs
18322 @opindex mno-space-regs
18323 Generate code that assumes the target has no space registers. This allows
18324 GCC to generate faster indirect calls and use unscaled index address modes.
18326 Such code is suitable for level 0 PA systems and kernels.
18328 @item -mfast-indirect-calls
18329 @opindex mfast-indirect-calls
18330 Generate code that assumes calls never cross space boundaries. This
18331 allows GCC to emit code that performs faster indirect calls.
18333 This option does not work in the presence of shared libraries or nested
18336 @item -mfixed-range=@var{register-range}
18337 @opindex mfixed-range
18338 Generate code treating the given register range as fixed registers.
18339 A fixed register is one that the register allocator cannot use. This is
18340 useful when compiling kernel code. A register range is specified as
18341 two registers separated by a dash. Multiple register ranges can be
18342 specified separated by a comma.
18344 @item -mlong-load-store
18345 @opindex mlong-load-store
18346 Generate 3-instruction load and store sequences as sometimes required by
18347 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
18350 @item -mportable-runtime
18351 @opindex mportable-runtime
18352 Use the portable calling conventions proposed by HP for ELF systems.
18356 Enable the use of assembler directives only GAS understands.
18358 @item -mschedule=@var{cpu-type}
18360 Schedule code according to the constraints for the machine type
18361 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
18362 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
18363 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
18364 proper scheduling option for your machine. The default scheduling is
18368 @opindex mlinker-opt
18369 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
18370 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
18371 linkers in which they give bogus error messages when linking some programs.
18374 @opindex msoft-float
18375 Generate output containing library calls for floating point.
18376 @strong{Warning:} the requisite libraries are not available for all HPPA
18377 targets. Normally the facilities of the machine's usual C compiler are
18378 used, but this cannot be done directly in cross-compilation. You must make
18379 your own arrangements to provide suitable library functions for
18382 @option{-msoft-float} changes the calling convention in the output file;
18383 therefore, it is only useful if you compile @emph{all} of a program with
18384 this option. In particular, you need to compile @file{libgcc.a}, the
18385 library that comes with GCC, with @option{-msoft-float} in order for
18390 Generate the predefine, @code{_SIO}, for server IO@. The default is
18391 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
18392 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
18393 options are available under HP-UX and HI-UX@.
18397 Use options specific to GNU @command{ld}.
18398 This passes @option{-shared} to @command{ld} when
18399 building a shared library. It is the default when GCC is configured,
18400 explicitly or implicitly, with the GNU linker. This option does not
18401 affect which @command{ld} is called; it only changes what parameters
18402 are passed to that @command{ld}.
18403 The @command{ld} that is called is determined by the
18404 @option{--with-ld} configure option, GCC's program search path, and
18405 finally by the user's @env{PATH}. The linker used by GCC can be printed
18406 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
18407 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18411 Use options specific to HP @command{ld}.
18412 This passes @option{-b} to @command{ld} when building
18413 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
18414 links. It is the default when GCC is configured, explicitly or
18415 implicitly, with the HP linker. This option does not affect
18416 which @command{ld} is called; it only changes what parameters are passed to that
18418 The @command{ld} that is called is determined by the @option{--with-ld}
18419 configure option, GCC's program search path, and finally by the user's
18420 @env{PATH}. The linker used by GCC can be printed using @samp{which
18421 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
18422 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
18425 @opindex mno-long-calls
18426 Generate code that uses long call sequences. This ensures that a call
18427 is always able to reach linker generated stubs. The default is to generate
18428 long calls only when the distance from the call site to the beginning
18429 of the function or translation unit, as the case may be, exceeds a
18430 predefined limit set by the branch type being used. The limits for
18431 normal calls are 7,600,000 and 240,000 bytes, respectively for the
18432 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
18435 Distances are measured from the beginning of functions when using the
18436 @option{-ffunction-sections} option, or when using the @option{-mgas}
18437 and @option{-mno-portable-runtime} options together under HP-UX with
18440 It is normally not desirable to use this option as it degrades
18441 performance. However, it may be useful in large applications,
18442 particularly when partial linking is used to build the application.
18444 The types of long calls used depends on the capabilities of the
18445 assembler and linker, and the type of code being generated. The
18446 impact on systems that support long absolute calls, and long pic
18447 symbol-difference or pc-relative calls should be relatively small.
18448 However, an indirect call is used on 32-bit ELF systems in pic code
18449 and it is quite long.
18451 @item -munix=@var{unix-std}
18453 Generate compiler predefines and select a startfile for the specified
18454 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
18455 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
18456 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
18457 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
18458 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
18461 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
18462 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
18463 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
18464 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
18465 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
18466 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
18468 It is @emph{important} to note that this option changes the interfaces
18469 for various library routines. It also affects the operational behavior
18470 of the C library. Thus, @emph{extreme} care is needed in using this
18473 Library code that is intended to operate with more than one UNIX
18474 standard must test, set and restore the variable @code{__xpg4_extended_mask}
18475 as appropriate. Most GNU software doesn't provide this capability.
18479 Suppress the generation of link options to search libdld.sl when the
18480 @option{-static} option is specified on HP-UX 10 and later.
18484 The HP-UX implementation of setlocale in libc has a dependency on
18485 libdld.sl. There isn't an archive version of libdld.sl. Thus,
18486 when the @option{-static} option is specified, special link options
18487 are needed to resolve this dependency.
18489 On HP-UX 10 and later, the GCC driver adds the necessary options to
18490 link with libdld.sl when the @option{-static} option is specified.
18491 This causes the resulting binary to be dynamic. On the 64-bit port,
18492 the linkers generate dynamic binaries by default in any case. The
18493 @option{-nolibdld} option can be used to prevent the GCC driver from
18494 adding these link options.
18498 Add support for multithreading with the @dfn{dce thread} library
18499 under HP-UX@. This option sets flags for both the preprocessor and
18503 @node IA-64 Options
18504 @subsection IA-64 Options
18505 @cindex IA-64 Options
18507 These are the @samp{-m} options defined for the Intel IA-64 architecture.
18511 @opindex mbig-endian
18512 Generate code for a big-endian target. This is the default for HP-UX@.
18514 @item -mlittle-endian
18515 @opindex mlittle-endian
18516 Generate code for a little-endian target. This is the default for AIX5
18522 @opindex mno-gnu-as
18523 Generate (or don't) code for the GNU assembler. This is the default.
18524 @c Also, this is the default if the configure option @option{--with-gnu-as}
18530 @opindex mno-gnu-ld
18531 Generate (or don't) code for the GNU linker. This is the default.
18532 @c Also, this is the default if the configure option @option{--with-gnu-ld}
18537 Generate code that does not use a global pointer register. The result
18538 is not position independent code, and violates the IA-64 ABI@.
18540 @item -mvolatile-asm-stop
18541 @itemx -mno-volatile-asm-stop
18542 @opindex mvolatile-asm-stop
18543 @opindex mno-volatile-asm-stop
18544 Generate (or don't) a stop bit immediately before and after volatile asm
18547 @item -mregister-names
18548 @itemx -mno-register-names
18549 @opindex mregister-names
18550 @opindex mno-register-names
18551 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
18552 the stacked registers. This may make assembler output more readable.
18558 Disable (or enable) optimizations that use the small data section. This may
18559 be useful for working around optimizer bugs.
18561 @item -mconstant-gp
18562 @opindex mconstant-gp
18563 Generate code that uses a single constant global pointer value. This is
18564 useful when compiling kernel code.
18568 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
18569 This is useful when compiling firmware code.
18571 @item -minline-float-divide-min-latency
18572 @opindex minline-float-divide-min-latency
18573 Generate code for inline divides of floating-point values
18574 using the minimum latency algorithm.
18576 @item -minline-float-divide-max-throughput
18577 @opindex minline-float-divide-max-throughput
18578 Generate code for inline divides of floating-point values
18579 using the maximum throughput algorithm.
18581 @item -mno-inline-float-divide
18582 @opindex mno-inline-float-divide
18583 Do not generate inline code for divides of floating-point values.
18585 @item -minline-int-divide-min-latency
18586 @opindex minline-int-divide-min-latency
18587 Generate code for inline divides of integer values
18588 using the minimum latency algorithm.
18590 @item -minline-int-divide-max-throughput
18591 @opindex minline-int-divide-max-throughput
18592 Generate code for inline divides of integer values
18593 using the maximum throughput algorithm.
18595 @item -mno-inline-int-divide
18596 @opindex mno-inline-int-divide
18597 Do not generate inline code for divides of integer values.
18599 @item -minline-sqrt-min-latency
18600 @opindex minline-sqrt-min-latency
18601 Generate code for inline square roots
18602 using the minimum latency algorithm.
18604 @item -minline-sqrt-max-throughput
18605 @opindex minline-sqrt-max-throughput
18606 Generate code for inline square roots
18607 using the maximum throughput algorithm.
18609 @item -mno-inline-sqrt
18610 @opindex mno-inline-sqrt
18611 Do not generate inline code for @code{sqrt}.
18614 @itemx -mno-fused-madd
18615 @opindex mfused-madd
18616 @opindex mno-fused-madd
18617 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
18618 instructions. The default is to use these instructions.
18620 @item -mno-dwarf2-asm
18621 @itemx -mdwarf2-asm
18622 @opindex mno-dwarf2-asm
18623 @opindex mdwarf2-asm
18624 Don't (or do) generate assembler code for the DWARF line number debugging
18625 info. This may be useful when not using the GNU assembler.
18627 @item -mearly-stop-bits
18628 @itemx -mno-early-stop-bits
18629 @opindex mearly-stop-bits
18630 @opindex mno-early-stop-bits
18631 Allow stop bits to be placed earlier than immediately preceding the
18632 instruction that triggered the stop bit. This can improve instruction
18633 scheduling, but does not always do so.
18635 @item -mfixed-range=@var{register-range}
18636 @opindex mfixed-range
18637 Generate code treating the given register range as fixed registers.
18638 A fixed register is one that the register allocator cannot use. This is
18639 useful when compiling kernel code. A register range is specified as
18640 two registers separated by a dash. Multiple register ranges can be
18641 specified separated by a comma.
18643 @item -mtls-size=@var{tls-size}
18645 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
18648 @item -mtune=@var{cpu-type}
18650 Tune the instruction scheduling for a particular CPU, Valid values are
18651 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
18652 and @samp{mckinley}.
18658 Generate code for a 32-bit or 64-bit environment.
18659 The 32-bit environment sets int, long and pointer to 32 bits.
18660 The 64-bit environment sets int to 32 bits and long and pointer
18661 to 64 bits. These are HP-UX specific flags.
18663 @item -mno-sched-br-data-spec
18664 @itemx -msched-br-data-spec
18665 @opindex mno-sched-br-data-spec
18666 @opindex msched-br-data-spec
18667 (Dis/En)able data speculative scheduling before reload.
18668 This results in generation of @code{ld.a} instructions and
18669 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18670 The default setting is disabled.
18672 @item -msched-ar-data-spec
18673 @itemx -mno-sched-ar-data-spec
18674 @opindex msched-ar-data-spec
18675 @opindex mno-sched-ar-data-spec
18676 (En/Dis)able data speculative scheduling after reload.
18677 This results in generation of @code{ld.a} instructions and
18678 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
18679 The default setting is enabled.
18681 @item -mno-sched-control-spec
18682 @itemx -msched-control-spec
18683 @opindex mno-sched-control-spec
18684 @opindex msched-control-spec
18685 (Dis/En)able control speculative scheduling. This feature is
18686 available only during region scheduling (i.e.@: before reload).
18687 This results in generation of the @code{ld.s} instructions and
18688 the corresponding check instructions @code{chk.s}.
18689 The default setting is disabled.
18691 @item -msched-br-in-data-spec
18692 @itemx -mno-sched-br-in-data-spec
18693 @opindex msched-br-in-data-spec
18694 @opindex mno-sched-br-in-data-spec
18695 (En/Dis)able speculative scheduling of the instructions that
18696 are dependent on the data speculative loads before reload.
18697 This is effective only with @option{-msched-br-data-spec} enabled.
18698 The default setting is enabled.
18700 @item -msched-ar-in-data-spec
18701 @itemx -mno-sched-ar-in-data-spec
18702 @opindex msched-ar-in-data-spec
18703 @opindex mno-sched-ar-in-data-spec
18704 (En/Dis)able speculative scheduling of the instructions that
18705 are dependent on the data speculative loads after reload.
18706 This is effective only with @option{-msched-ar-data-spec} enabled.
18707 The default setting is enabled.
18709 @item -msched-in-control-spec
18710 @itemx -mno-sched-in-control-spec
18711 @opindex msched-in-control-spec
18712 @opindex mno-sched-in-control-spec
18713 (En/Dis)able speculative scheduling of the instructions that
18714 are dependent on the control speculative loads.
18715 This is effective only with @option{-msched-control-spec} enabled.
18716 The default setting is enabled.
18718 @item -mno-sched-prefer-non-data-spec-insns
18719 @itemx -msched-prefer-non-data-spec-insns
18720 @opindex mno-sched-prefer-non-data-spec-insns
18721 @opindex msched-prefer-non-data-spec-insns
18722 If enabled, data-speculative instructions are chosen for schedule
18723 only if there are no other choices at the moment. This makes
18724 the use of the data speculation much more conservative.
18725 The default setting is disabled.
18727 @item -mno-sched-prefer-non-control-spec-insns
18728 @itemx -msched-prefer-non-control-spec-insns
18729 @opindex mno-sched-prefer-non-control-spec-insns
18730 @opindex msched-prefer-non-control-spec-insns
18731 If enabled, control-speculative instructions are chosen for schedule
18732 only if there are no other choices at the moment. This makes
18733 the use of the control speculation much more conservative.
18734 The default setting is disabled.
18736 @item -mno-sched-count-spec-in-critical-path
18737 @itemx -msched-count-spec-in-critical-path
18738 @opindex mno-sched-count-spec-in-critical-path
18739 @opindex msched-count-spec-in-critical-path
18740 If enabled, speculative dependencies are considered during
18741 computation of the instructions priorities. This makes the use of the
18742 speculation a bit more conservative.
18743 The default setting is disabled.
18745 @item -msched-spec-ldc
18746 @opindex msched-spec-ldc
18747 Use a simple data speculation check. This option is on by default.
18749 @item -msched-control-spec-ldc
18750 @opindex msched-spec-ldc
18751 Use a simple check for control speculation. This option is on by default.
18753 @item -msched-stop-bits-after-every-cycle
18754 @opindex msched-stop-bits-after-every-cycle
18755 Place a stop bit after every cycle when scheduling. This option is on
18758 @item -msched-fp-mem-deps-zero-cost
18759 @opindex msched-fp-mem-deps-zero-cost
18760 Assume that floating-point stores and loads are not likely to cause a conflict
18761 when placed into the same instruction group. This option is disabled by
18764 @item -msel-sched-dont-check-control-spec
18765 @opindex msel-sched-dont-check-control-spec
18766 Generate checks for control speculation in selective scheduling.
18767 This flag is disabled by default.
18769 @item -msched-max-memory-insns=@var{max-insns}
18770 @opindex msched-max-memory-insns
18771 Limit on the number of memory insns per instruction group, giving lower
18772 priority to subsequent memory insns attempting to schedule in the same
18773 instruction group. Frequently useful to prevent cache bank conflicts.
18774 The default value is 1.
18776 @item -msched-max-memory-insns-hard-limit
18777 @opindex msched-max-memory-insns-hard-limit
18778 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
18779 disallowing more than that number in an instruction group.
18780 Otherwise, the limit is ``soft'', meaning that non-memory operations
18781 are preferred when the limit is reached, but memory operations may still
18787 @subsection LM32 Options
18788 @cindex LM32 options
18790 These @option{-m} options are defined for the LatticeMico32 architecture:
18793 @item -mbarrel-shift-enabled
18794 @opindex mbarrel-shift-enabled
18795 Enable barrel-shift instructions.
18797 @item -mdivide-enabled
18798 @opindex mdivide-enabled
18799 Enable divide and modulus instructions.
18801 @item -mmultiply-enabled
18802 @opindex multiply-enabled
18803 Enable multiply instructions.
18805 @item -msign-extend-enabled
18806 @opindex msign-extend-enabled
18807 Enable sign extend instructions.
18809 @item -muser-enabled
18810 @opindex muser-enabled
18811 Enable user-defined instructions.
18816 @subsection M32C Options
18817 @cindex M32C options
18820 @item -mcpu=@var{name}
18822 Select the CPU for which code is generated. @var{name} may be one of
18823 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
18824 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
18825 the M32C/80 series.
18829 Specifies that the program will be run on the simulator. This causes
18830 an alternate runtime library to be linked in which supports, for
18831 example, file I/O@. You must not use this option when generating
18832 programs that will run on real hardware; you must provide your own
18833 runtime library for whatever I/O functions are needed.
18835 @item -memregs=@var{number}
18837 Specifies the number of memory-based pseudo-registers GCC uses
18838 during code generation. These pseudo-registers are used like real
18839 registers, so there is a tradeoff between GCC's ability to fit the
18840 code into available registers, and the performance penalty of using
18841 memory instead of registers. Note that all modules in a program must
18842 be compiled with the same value for this option. Because of that, you
18843 must not use this option with GCC's default runtime libraries.
18847 @node M32R/D Options
18848 @subsection M32R/D Options
18849 @cindex M32R/D options
18851 These @option{-m} options are defined for Renesas M32R/D architectures:
18856 Generate code for the M32R/2@.
18860 Generate code for the M32R/X@.
18864 Generate code for the M32R@. This is the default.
18866 @item -mmodel=small
18867 @opindex mmodel=small
18868 Assume all objects live in the lower 16MB of memory (so that their addresses
18869 can be loaded with the @code{ld24} instruction), and assume all subroutines
18870 are reachable with the @code{bl} instruction.
18871 This is the default.
18873 The addressability of a particular object can be set with the
18874 @code{model} attribute.
18876 @item -mmodel=medium
18877 @opindex mmodel=medium
18878 Assume objects may be anywhere in the 32-bit address space (the compiler
18879 generates @code{seth/add3} instructions to load their addresses), and
18880 assume all subroutines are reachable with the @code{bl} instruction.
18882 @item -mmodel=large
18883 @opindex mmodel=large
18884 Assume objects may be anywhere in the 32-bit address space (the compiler
18885 generates @code{seth/add3} instructions to load their addresses), and
18886 assume subroutines may not be reachable with the @code{bl} instruction
18887 (the compiler generates the much slower @code{seth/add3/jl}
18888 instruction sequence).
18891 @opindex msdata=none
18892 Disable use of the small data area. Variables are put into
18893 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
18894 @code{section} attribute has been specified).
18895 This is the default.
18897 The small data area consists of sections @code{.sdata} and @code{.sbss}.
18898 Objects may be explicitly put in the small data area with the
18899 @code{section} attribute using one of these sections.
18901 @item -msdata=sdata
18902 @opindex msdata=sdata
18903 Put small global and static data in the small data area, but do not
18904 generate special code to reference them.
18907 @opindex msdata=use
18908 Put small global and static data in the small data area, and generate
18909 special instructions to reference them.
18913 @cindex smaller data references
18914 Put global and static objects less than or equal to @var{num} bytes
18915 into the small data or BSS sections instead of the normal data or BSS
18916 sections. The default value of @var{num} is 8.
18917 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
18918 for this option to have any effect.
18920 All modules should be compiled with the same @option{-G @var{num}} value.
18921 Compiling with different values of @var{num} may or may not work; if it
18922 doesn't the linker gives an error message---incorrect code is not
18927 Makes the M32R-specific code in the compiler display some statistics
18928 that might help in debugging programs.
18930 @item -malign-loops
18931 @opindex malign-loops
18932 Align all loops to a 32-byte boundary.
18934 @item -mno-align-loops
18935 @opindex mno-align-loops
18936 Do not enforce a 32-byte alignment for loops. This is the default.
18938 @item -missue-rate=@var{number}
18939 @opindex missue-rate=@var{number}
18940 Issue @var{number} instructions per cycle. @var{number} can only be 1
18943 @item -mbranch-cost=@var{number}
18944 @opindex mbranch-cost=@var{number}
18945 @var{number} can only be 1 or 2. If it is 1 then branches are
18946 preferred over conditional code, if it is 2, then the opposite applies.
18948 @item -mflush-trap=@var{number}
18949 @opindex mflush-trap=@var{number}
18950 Specifies the trap number to use to flush the cache. The default is
18951 12. Valid numbers are between 0 and 15 inclusive.
18953 @item -mno-flush-trap
18954 @opindex mno-flush-trap
18955 Specifies that the cache cannot be flushed by using a trap.
18957 @item -mflush-func=@var{name}
18958 @opindex mflush-func=@var{name}
18959 Specifies the name of the operating system function to call to flush
18960 the cache. The default is @samp{_flush_cache}, but a function call
18961 is only used if a trap is not available.
18963 @item -mno-flush-func
18964 @opindex mno-flush-func
18965 Indicates that there is no OS function for flushing the cache.
18969 @node M680x0 Options
18970 @subsection M680x0 Options
18971 @cindex M680x0 options
18973 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
18974 The default settings depend on which architecture was selected when
18975 the compiler was configured; the defaults for the most common choices
18979 @item -march=@var{arch}
18981 Generate code for a specific M680x0 or ColdFire instruction set
18982 architecture. Permissible values of @var{arch} for M680x0
18983 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
18984 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
18985 architectures are selected according to Freescale's ISA classification
18986 and the permissible values are: @samp{isaa}, @samp{isaaplus},
18987 @samp{isab} and @samp{isac}.
18989 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
18990 code for a ColdFire target. The @var{arch} in this macro is one of the
18991 @option{-march} arguments given above.
18993 When used together, @option{-march} and @option{-mtune} select code
18994 that runs on a family of similar processors but that is optimized
18995 for a particular microarchitecture.
18997 @item -mcpu=@var{cpu}
18999 Generate code for a specific M680x0 or ColdFire processor.
19000 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
19001 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
19002 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
19003 below, which also classifies the CPUs into families:
19005 @multitable @columnfractions 0.20 0.80
19006 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
19007 @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}
19008 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
19009 @item @samp{5206e} @tab @samp{5206e}
19010 @item @samp{5208} @tab @samp{5207} @samp{5208}
19011 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
19012 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
19013 @item @samp{5216} @tab @samp{5214} @samp{5216}
19014 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
19015 @item @samp{5225} @tab @samp{5224} @samp{5225}
19016 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
19017 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
19018 @item @samp{5249} @tab @samp{5249}
19019 @item @samp{5250} @tab @samp{5250}
19020 @item @samp{5271} @tab @samp{5270} @samp{5271}
19021 @item @samp{5272} @tab @samp{5272}
19022 @item @samp{5275} @tab @samp{5274} @samp{5275}
19023 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
19024 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
19025 @item @samp{5307} @tab @samp{5307}
19026 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
19027 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
19028 @item @samp{5407} @tab @samp{5407}
19029 @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}
19032 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
19033 @var{arch} is compatible with @var{cpu}. Other combinations of
19034 @option{-mcpu} and @option{-march} are rejected.
19036 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
19037 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
19038 where the value of @var{family} is given by the table above.
19040 @item -mtune=@var{tune}
19042 Tune the code for a particular microarchitecture within the
19043 constraints set by @option{-march} and @option{-mcpu}.
19044 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
19045 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
19046 and @samp{cpu32}. The ColdFire microarchitectures
19047 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
19049 You can also use @option{-mtune=68020-40} for code that needs
19050 to run relatively well on 68020, 68030 and 68040 targets.
19051 @option{-mtune=68020-60} is similar but includes 68060 targets
19052 as well. These two options select the same tuning decisions as
19053 @option{-m68020-40} and @option{-m68020-60} respectively.
19055 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
19056 when tuning for 680x0 architecture @var{arch}. It also defines
19057 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
19058 option is used. If GCC is tuning for a range of architectures,
19059 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
19060 it defines the macros for every architecture in the range.
19062 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
19063 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
19064 of the arguments given above.
19070 Generate output for a 68000. This is the default
19071 when the compiler is configured for 68000-based systems.
19072 It is equivalent to @option{-march=68000}.
19074 Use this option for microcontrollers with a 68000 or EC000 core,
19075 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
19079 Generate output for a 68010. This is the default
19080 when the compiler is configured for 68010-based systems.
19081 It is equivalent to @option{-march=68010}.
19087 Generate output for a 68020. This is the default
19088 when the compiler is configured for 68020-based systems.
19089 It is equivalent to @option{-march=68020}.
19093 Generate output for a 68030. This is the default when the compiler is
19094 configured for 68030-based systems. It is equivalent to
19095 @option{-march=68030}.
19099 Generate output for a 68040. This is the default when the compiler is
19100 configured for 68040-based systems. It is equivalent to
19101 @option{-march=68040}.
19103 This option inhibits the use of 68881/68882 instructions that have to be
19104 emulated by software on the 68040. Use this option if your 68040 does not
19105 have code to emulate those instructions.
19109 Generate output for a 68060. This is the default when the compiler is
19110 configured for 68060-based systems. It is equivalent to
19111 @option{-march=68060}.
19113 This option inhibits the use of 68020 and 68881/68882 instructions that
19114 have to be emulated by software on the 68060. Use this option if your 68060
19115 does not have code to emulate those instructions.
19119 Generate output for a CPU32. This is the default
19120 when the compiler is configured for CPU32-based systems.
19121 It is equivalent to @option{-march=cpu32}.
19123 Use this option for microcontrollers with a
19124 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
19125 68336, 68340, 68341, 68349 and 68360.
19129 Generate output for a 520X ColdFire CPU@. This is the default
19130 when the compiler is configured for 520X-based systems.
19131 It is equivalent to @option{-mcpu=5206}, and is now deprecated
19132 in favor of that option.
19134 Use this option for microcontroller with a 5200 core, including
19135 the MCF5202, MCF5203, MCF5204 and MCF5206.
19139 Generate output for a 5206e ColdFire CPU@. The option is now
19140 deprecated in favor of the equivalent @option{-mcpu=5206e}.
19144 Generate output for a member of the ColdFire 528X family.
19145 The option is now deprecated in favor of the equivalent
19146 @option{-mcpu=528x}.
19150 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
19151 in favor of the equivalent @option{-mcpu=5307}.
19155 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
19156 in favor of the equivalent @option{-mcpu=5407}.
19160 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
19161 This includes use of hardware floating-point instructions.
19162 The option is equivalent to @option{-mcpu=547x}, and is now
19163 deprecated in favor of that option.
19167 Generate output for a 68040, without using any of the new instructions.
19168 This results in code that can run relatively efficiently on either a
19169 68020/68881 or a 68030 or a 68040. The generated code does use the
19170 68881 instructions that are emulated on the 68040.
19172 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
19176 Generate output for a 68060, without using any of the new instructions.
19177 This results in code that can run relatively efficiently on either a
19178 68020/68881 or a 68030 or a 68040. The generated code does use the
19179 68881 instructions that are emulated on the 68060.
19181 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
19185 @opindex mhard-float
19187 Generate floating-point instructions. This is the default for 68020
19188 and above, and for ColdFire devices that have an FPU@. It defines the
19189 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
19190 on ColdFire targets.
19193 @opindex msoft-float
19194 Do not generate floating-point instructions; use library calls instead.
19195 This is the default for 68000, 68010, and 68832 targets. It is also
19196 the default for ColdFire devices that have no FPU.
19202 Generate (do not generate) ColdFire hardware divide and remainder
19203 instructions. If @option{-march} is used without @option{-mcpu},
19204 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
19205 architectures. Otherwise, the default is taken from the target CPU
19206 (either the default CPU, or the one specified by @option{-mcpu}). For
19207 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
19208 @option{-mcpu=5206e}.
19210 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
19214 Consider type @code{int} to be 16 bits wide, like @code{short int}.
19215 Additionally, parameters passed on the stack are also aligned to a
19216 16-bit boundary even on targets whose API mandates promotion to 32-bit.
19220 Do not consider type @code{int} to be 16 bits wide. This is the default.
19223 @itemx -mno-bitfield
19224 @opindex mnobitfield
19225 @opindex mno-bitfield
19226 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
19227 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
19231 Do use the bit-field instructions. The @option{-m68020} option implies
19232 @option{-mbitfield}. This is the default if you use a configuration
19233 designed for a 68020.
19237 Use a different function-calling convention, in which functions
19238 that take a fixed number of arguments return with the @code{rtd}
19239 instruction, which pops their arguments while returning. This
19240 saves one instruction in the caller since there is no need to pop
19241 the arguments there.
19243 This calling convention is incompatible with the one normally
19244 used on Unix, so you cannot use it if you need to call libraries
19245 compiled with the Unix compiler.
19247 Also, you must provide function prototypes for all functions that
19248 take variable numbers of arguments (including @code{printf});
19249 otherwise incorrect code is generated for calls to those
19252 In addition, seriously incorrect code results if you call a
19253 function with too many arguments. (Normally, extra arguments are
19254 harmlessly ignored.)
19256 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
19257 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
19261 Do not use the calling conventions selected by @option{-mrtd}.
19262 This is the default.
19265 @itemx -mno-align-int
19266 @opindex malign-int
19267 @opindex mno-align-int
19268 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
19269 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
19270 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
19271 Aligning variables on 32-bit boundaries produces code that runs somewhat
19272 faster on processors with 32-bit busses at the expense of more memory.
19274 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
19275 aligns structures containing the above types differently than
19276 most published application binary interface specifications for the m68k.
19280 Use the pc-relative addressing mode of the 68000 directly, instead of
19281 using a global offset table. At present, this option implies @option{-fpic},
19282 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
19283 not presently supported with @option{-mpcrel}, though this could be supported for
19284 68020 and higher processors.
19286 @item -mno-strict-align
19287 @itemx -mstrict-align
19288 @opindex mno-strict-align
19289 @opindex mstrict-align
19290 Do not (do) assume that unaligned memory references are handled by
19294 Generate code that allows the data segment to be located in a different
19295 area of memory from the text segment. This allows for execute-in-place in
19296 an environment without virtual memory management. This option implies
19299 @item -mno-sep-data
19300 Generate code that assumes that the data segment follows the text segment.
19301 This is the default.
19303 @item -mid-shared-library
19304 Generate code that supports shared libraries via the library ID method.
19305 This allows for execute-in-place and shared libraries in an environment
19306 without virtual memory management. This option implies @option{-fPIC}.
19308 @item -mno-id-shared-library
19309 Generate code that doesn't assume ID-based shared libraries are being used.
19310 This is the default.
19312 @item -mshared-library-id=n
19313 Specifies the identification number of the ID-based shared library being
19314 compiled. Specifying a value of 0 generates more compact code; specifying
19315 other values forces the allocation of that number to the current
19316 library, but is no more space- or time-efficient than omitting this option.
19322 When generating position-independent code for ColdFire, generate code
19323 that works if the GOT has more than 8192 entries. This code is
19324 larger and slower than code generated without this option. On M680x0
19325 processors, this option is not needed; @option{-fPIC} suffices.
19327 GCC normally uses a single instruction to load values from the GOT@.
19328 While this is relatively efficient, it only works if the GOT
19329 is smaller than about 64k. Anything larger causes the linker
19330 to report an error such as:
19332 @cindex relocation truncated to fit (ColdFire)
19334 relocation truncated to fit: R_68K_GOT16O foobar
19337 If this happens, you should recompile your code with @option{-mxgot}.
19338 It should then work with very large GOTs. However, code generated with
19339 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
19340 the value of a global symbol.
19342 Note that some linkers, including newer versions of the GNU linker,
19343 can create multiple GOTs and sort GOT entries. If you have such a linker,
19344 you should only need to use @option{-mxgot} when compiling a single
19345 object file that accesses more than 8192 GOT entries. Very few do.
19347 These options have no effect unless GCC is generating
19348 position-independent code.
19350 @item -mlong-jump-table-offsets
19351 @opindex mlong-jump-table-offsets
19352 Use 32-bit offsets in @code{switch} tables. The default is to use
19357 @node MCore Options
19358 @subsection MCore Options
19359 @cindex MCore options
19361 These are the @samp{-m} options defined for the Motorola M*Core
19367 @itemx -mno-hardlit
19369 @opindex mno-hardlit
19370 Inline constants into the code stream if it can be done in two
19371 instructions or less.
19377 Use the divide instruction. (Enabled by default).
19379 @item -mrelax-immediate
19380 @itemx -mno-relax-immediate
19381 @opindex mrelax-immediate
19382 @opindex mno-relax-immediate
19383 Allow arbitrary-sized immediates in bit operations.
19385 @item -mwide-bitfields
19386 @itemx -mno-wide-bitfields
19387 @opindex mwide-bitfields
19388 @opindex mno-wide-bitfields
19389 Always treat bit-fields as @code{int}-sized.
19391 @item -m4byte-functions
19392 @itemx -mno-4byte-functions
19393 @opindex m4byte-functions
19394 @opindex mno-4byte-functions
19395 Force all functions to be aligned to a 4-byte boundary.
19397 @item -mcallgraph-data
19398 @itemx -mno-callgraph-data
19399 @opindex mcallgraph-data
19400 @opindex mno-callgraph-data
19401 Emit callgraph information.
19404 @itemx -mno-slow-bytes
19405 @opindex mslow-bytes
19406 @opindex mno-slow-bytes
19407 Prefer word access when reading byte quantities.
19409 @item -mlittle-endian
19410 @itemx -mbig-endian
19411 @opindex mlittle-endian
19412 @opindex mbig-endian
19413 Generate code for a little-endian target.
19419 Generate code for the 210 processor.
19423 Assume that runtime support has been provided and so omit the
19424 simulator library (@file{libsim.a)} from the linker command line.
19426 @item -mstack-increment=@var{size}
19427 @opindex mstack-increment
19428 Set the maximum amount for a single stack increment operation. Large
19429 values can increase the speed of programs that contain functions
19430 that need a large amount of stack space, but they can also trigger a
19431 segmentation fault if the stack is extended too much. The default
19437 @subsection MeP Options
19438 @cindex MeP options
19444 Enables the @code{abs} instruction, which is the absolute difference
19445 between two registers.
19449 Enables all the optional instructions---average, multiply, divide, bit
19450 operations, leading zero, absolute difference, min/max, clip, and
19456 Enables the @code{ave} instruction, which computes the average of two
19459 @item -mbased=@var{n}
19461 Variables of size @var{n} bytes or smaller are placed in the
19462 @code{.based} section by default. Based variables use the @code{$tp}
19463 register as a base register, and there is a 128-byte limit to the
19464 @code{.based} section.
19468 Enables the bit operation instructions---bit test (@code{btstm}), set
19469 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
19470 test-and-set (@code{tas}).
19472 @item -mc=@var{name}
19474 Selects which section constant data is placed in. @var{name} may
19475 be @samp{tiny}, @samp{near}, or @samp{far}.
19479 Enables the @code{clip} instruction. Note that @option{-mclip} is not
19480 useful unless you also provide @option{-mminmax}.
19482 @item -mconfig=@var{name}
19484 Selects one of the built-in core configurations. Each MeP chip has
19485 one or more modules in it; each module has a core CPU and a variety of
19486 coprocessors, optional instructions, and peripherals. The
19487 @code{MeP-Integrator} tool, not part of GCC, provides these
19488 configurations through this option; using this option is the same as
19489 using all the corresponding command-line options. The default
19490 configuration is @samp{default}.
19494 Enables the coprocessor instructions. By default, this is a 32-bit
19495 coprocessor. Note that the coprocessor is normally enabled via the
19496 @option{-mconfig=} option.
19500 Enables the 32-bit coprocessor's instructions.
19504 Enables the 64-bit coprocessor's instructions.
19508 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
19512 Causes constant variables to be placed in the @code{.near} section.
19516 Enables the @code{div} and @code{divu} instructions.
19520 Generate big-endian code.
19524 Generate little-endian code.
19526 @item -mio-volatile
19527 @opindex mio-volatile
19528 Tells the compiler that any variable marked with the @code{io}
19529 attribute is to be considered volatile.
19533 Causes variables to be assigned to the @code{.far} section by default.
19537 Enables the @code{leadz} (leading zero) instruction.
19541 Causes variables to be assigned to the @code{.near} section by default.
19545 Enables the @code{min} and @code{max} instructions.
19549 Enables the multiplication and multiply-accumulate instructions.
19553 Disables all the optional instructions enabled by @option{-mall-opts}.
19557 Enables the @code{repeat} and @code{erepeat} instructions, used for
19558 low-overhead looping.
19562 Causes all variables to default to the @code{.tiny} section. Note
19563 that there is a 65536-byte limit to this section. Accesses to these
19564 variables use the @code{%gp} base register.
19568 Enables the saturation instructions. Note that the compiler does not
19569 currently generate these itself, but this option is included for
19570 compatibility with other tools, like @code{as}.
19574 Link the SDRAM-based runtime instead of the default ROM-based runtime.
19578 Link the simulator run-time libraries.
19582 Link the simulator runtime libraries, excluding built-in support
19583 for reset and exception vectors and tables.
19587 Causes all functions to default to the @code{.far} section. Without
19588 this option, functions default to the @code{.near} section.
19590 @item -mtiny=@var{n}
19592 Variables that are @var{n} bytes or smaller are allocated to the
19593 @code{.tiny} section. These variables use the @code{$gp} base
19594 register. The default for this option is 4, but note that there's a
19595 65536-byte limit to the @code{.tiny} section.
19599 @node MicroBlaze Options
19600 @subsection MicroBlaze Options
19601 @cindex MicroBlaze Options
19606 @opindex msoft-float
19607 Use software emulation for floating point (default).
19610 @opindex mhard-float
19611 Use hardware floating-point instructions.
19615 Do not optimize block moves, use @code{memcpy}.
19617 @item -mno-clearbss
19618 @opindex mno-clearbss
19619 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
19621 @item -mcpu=@var{cpu-type}
19623 Use features of, and schedule code for, the given CPU.
19624 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
19625 where @var{X} is a major version, @var{YY} is the minor version, and
19626 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
19627 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
19629 @item -mxl-soft-mul
19630 @opindex mxl-soft-mul
19631 Use software multiply emulation (default).
19633 @item -mxl-soft-div
19634 @opindex mxl-soft-div
19635 Use software emulation for divides (default).
19637 @item -mxl-barrel-shift
19638 @opindex mxl-barrel-shift
19639 Use the hardware barrel shifter.
19641 @item -mxl-pattern-compare
19642 @opindex mxl-pattern-compare
19643 Use pattern compare instructions.
19645 @item -msmall-divides
19646 @opindex msmall-divides
19647 Use table lookup optimization for small signed integer divisions.
19649 @item -mxl-stack-check
19650 @opindex mxl-stack-check
19651 This option is deprecated. Use @option{-fstack-check} instead.
19654 @opindex mxl-gp-opt
19655 Use GP-relative @code{.sdata}/@code{.sbss} sections.
19657 @item -mxl-multiply-high
19658 @opindex mxl-multiply-high
19659 Use multiply high instructions for high part of 32x32 multiply.
19661 @item -mxl-float-convert
19662 @opindex mxl-float-convert
19663 Use hardware floating-point conversion instructions.
19665 @item -mxl-float-sqrt
19666 @opindex mxl-float-sqrt
19667 Use hardware floating-point square root instruction.
19670 @opindex mbig-endian
19671 Generate code for a big-endian target.
19673 @item -mlittle-endian
19674 @opindex mlittle-endian
19675 Generate code for a little-endian target.
19678 @opindex mxl-reorder
19679 Use reorder instructions (swap and byte reversed load/store).
19681 @item -mxl-mode-@var{app-model}
19682 Select application model @var{app-model}. Valid models are
19685 normal executable (default), uses startup code @file{crt0.o}.
19688 for use with Xilinx Microprocessor Debugger (XMD) based
19689 software intrusive debug agent called xmdstub. This uses startup file
19690 @file{crt1.o} and sets the start address of the program to 0x800.
19693 for applications that are loaded using a bootloader.
19694 This model uses startup file @file{crt2.o} which does not contain a processor
19695 reset vector handler. This is suitable for transferring control on a
19696 processor reset to the bootloader rather than the application.
19699 for applications that do not require any of the
19700 MicroBlaze vectors. This option may be useful for applications running
19701 within a monitoring application. This model uses @file{crt3.o} as a startup file.
19704 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
19705 @option{-mxl-mode-@var{app-model}}.
19710 @subsection MIPS Options
19711 @cindex MIPS options
19717 Generate big-endian code.
19721 Generate little-endian code. This is the default for @samp{mips*el-*-*}
19724 @item -march=@var{arch}
19726 Generate code that runs on @var{arch}, which can be the name of a
19727 generic MIPS ISA, or the name of a particular processor.
19729 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
19730 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
19731 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
19732 @samp{mips64r5} and @samp{mips64r6}.
19733 The processor names are:
19734 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
19735 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
19736 @samp{5kc}, @samp{5kf},
19738 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
19739 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
19740 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
19741 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
19742 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
19745 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
19747 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
19748 @samp{m5100}, @samp{m5101},
19749 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
19752 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
19753 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
19754 @samp{rm7000}, @samp{rm9000},
19755 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
19758 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
19759 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
19760 @samp{xlr} and @samp{xlp}.
19761 The special value @samp{from-abi} selects the
19762 most compatible architecture for the selected ABI (that is,
19763 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
19765 The native Linux/GNU toolchain also supports the value @samp{native},
19766 which selects the best architecture option for the host processor.
19767 @option{-march=native} has no effect if GCC does not recognize
19770 In processor names, a final @samp{000} can be abbreviated as @samp{k}
19771 (for example, @option{-march=r2k}). Prefixes are optional, and
19772 @samp{vr} may be written @samp{r}.
19774 Names of the form @samp{@var{n}f2_1} refer to processors with
19775 FPUs clocked at half the rate of the core, names of the form
19776 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
19777 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
19778 processors with FPUs clocked a ratio of 3:2 with respect to the core.
19779 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
19780 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
19781 accepted as synonyms for @samp{@var{n}f1_1}.
19783 GCC defines two macros based on the value of this option. The first
19784 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
19785 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
19786 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
19787 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
19788 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
19790 Note that the @code{_MIPS_ARCH} macro uses the processor names given
19791 above. In other words, it has the full prefix and does not
19792 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
19793 the macro names the resolved architecture (either @code{"mips1"} or
19794 @code{"mips3"}). It names the default architecture when no
19795 @option{-march} option is given.
19797 @item -mtune=@var{arch}
19799 Optimize for @var{arch}. Among other things, this option controls
19800 the way instructions are scheduled, and the perceived cost of arithmetic
19801 operations. The list of @var{arch} values is the same as for
19804 When this option is not used, GCC optimizes for the processor
19805 specified by @option{-march}. By using @option{-march} and
19806 @option{-mtune} together, it is possible to generate code that
19807 runs on a family of processors, but optimize the code for one
19808 particular member of that family.
19810 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
19811 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
19812 @option{-march} ones described above.
19816 Equivalent to @option{-march=mips1}.
19820 Equivalent to @option{-march=mips2}.
19824 Equivalent to @option{-march=mips3}.
19828 Equivalent to @option{-march=mips4}.
19832 Equivalent to @option{-march=mips32}.
19836 Equivalent to @option{-march=mips32r3}.
19840 Equivalent to @option{-march=mips32r5}.
19844 Equivalent to @option{-march=mips32r6}.
19848 Equivalent to @option{-march=mips64}.
19852 Equivalent to @option{-march=mips64r2}.
19856 Equivalent to @option{-march=mips64r3}.
19860 Equivalent to @option{-march=mips64r5}.
19864 Equivalent to @option{-march=mips64r6}.
19869 @opindex mno-mips16
19870 Generate (do not generate) MIPS16 code. If GCC is targeting a
19871 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
19873 MIPS16 code generation can also be controlled on a per-function basis
19874 by means of @code{mips16} and @code{nomips16} attributes.
19875 @xref{Function Attributes}, for more information.
19877 @item -mflip-mips16
19878 @opindex mflip-mips16
19879 Generate MIPS16 code on alternating functions. This option is provided
19880 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
19881 not intended for ordinary use in compiling user code.
19883 @item -minterlink-compressed
19884 @item -mno-interlink-compressed
19885 @opindex minterlink-compressed
19886 @opindex mno-interlink-compressed
19887 Require (do not require) that code using the standard (uncompressed) MIPS ISA
19888 be link-compatible with MIPS16 and microMIPS code, and vice versa.
19890 For example, code using the standard ISA encoding cannot jump directly
19891 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
19892 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
19893 knows that the target of the jump is not compressed.
19895 @item -minterlink-mips16
19896 @itemx -mno-interlink-mips16
19897 @opindex minterlink-mips16
19898 @opindex mno-interlink-mips16
19899 Aliases of @option{-minterlink-compressed} and
19900 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
19901 and are retained for backwards compatibility.
19913 Generate code for the given ABI@.
19915 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
19916 generates 64-bit code when you select a 64-bit architecture, but you
19917 can use @option{-mgp32} to get 32-bit code instead.
19919 For information about the O64 ABI, see
19920 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
19922 GCC supports a variant of the o32 ABI in which floating-point registers
19923 are 64 rather than 32 bits wide. You can select this combination with
19924 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
19925 and @code{mfhc1} instructions and is therefore only supported for
19926 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
19928 The register assignments for arguments and return values remain the
19929 same, but each scalar value is passed in a single 64-bit register
19930 rather than a pair of 32-bit registers. For example, scalar
19931 floating-point values are returned in @samp{$f0} only, not a
19932 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
19933 remains the same in that the even-numbered double-precision registers
19936 Two additional variants of the o32 ABI are supported to enable
19937 a transition from 32-bit to 64-bit registers. These are FPXX
19938 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
19939 The FPXX extension mandates that all code must execute correctly
19940 when run using 32-bit or 64-bit registers. The code can be interlinked
19941 with either FP32 or FP64, but not both.
19942 The FP64A extension is similar to the FP64 extension but forbids the
19943 use of odd-numbered single-precision registers. This can be used
19944 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
19945 processors and allows both FP32 and FP64A code to interlink and
19946 run in the same process without changing FPU modes.
19949 @itemx -mno-abicalls
19951 @opindex mno-abicalls
19952 Generate (do not generate) code that is suitable for SVR4-style
19953 dynamic objects. @option{-mabicalls} is the default for SVR4-based
19958 Generate (do not generate) code that is fully position-independent,
19959 and that can therefore be linked into shared libraries. This option
19960 only affects @option{-mabicalls}.
19962 All @option{-mabicalls} code has traditionally been position-independent,
19963 regardless of options like @option{-fPIC} and @option{-fpic}. However,
19964 as an extension, the GNU toolchain allows executables to use absolute
19965 accesses for locally-binding symbols. It can also use shorter GP
19966 initialization sequences and generate direct calls to locally-defined
19967 functions. This mode is selected by @option{-mno-shared}.
19969 @option{-mno-shared} depends on binutils 2.16 or higher and generates
19970 objects that can only be linked by the GNU linker. However, the option
19971 does not affect the ABI of the final executable; it only affects the ABI
19972 of relocatable objects. Using @option{-mno-shared} generally makes
19973 executables both smaller and quicker.
19975 @option{-mshared} is the default.
19981 Assume (do not assume) that the static and dynamic linkers
19982 support PLTs and copy relocations. This option only affects
19983 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
19984 has no effect without @option{-msym32}.
19986 You can make @option{-mplt} the default by configuring
19987 GCC with @option{--with-mips-plt}. The default is
19988 @option{-mno-plt} otherwise.
19994 Lift (do not lift) the usual restrictions on the size of the global
19997 GCC normally uses a single instruction to load values from the GOT@.
19998 While this is relatively efficient, it only works if the GOT
19999 is smaller than about 64k. Anything larger causes the linker
20000 to report an error such as:
20002 @cindex relocation truncated to fit (MIPS)
20004 relocation truncated to fit: R_MIPS_GOT16 foobar
20007 If this happens, you should recompile your code with @option{-mxgot}.
20008 This works with very large GOTs, although the code is also
20009 less efficient, since it takes three instructions to fetch the
20010 value of a global symbol.
20012 Note that some linkers can create multiple GOTs. If you have such a
20013 linker, you should only need to use @option{-mxgot} when a single object
20014 file accesses more than 64k's worth of GOT entries. Very few do.
20016 These options have no effect unless GCC is generating position
20021 Assume that general-purpose registers are 32 bits wide.
20025 Assume that general-purpose registers are 64 bits wide.
20029 Assume that floating-point registers are 32 bits wide.
20033 Assume that floating-point registers are 64 bits wide.
20037 Do not assume the width of floating-point registers.
20040 @opindex mhard-float
20041 Use floating-point coprocessor instructions.
20044 @opindex msoft-float
20045 Do not use floating-point coprocessor instructions. Implement
20046 floating-point calculations using library calls instead.
20050 Equivalent to @option{-msoft-float}, but additionally asserts that the
20051 program being compiled does not perform any floating-point operations.
20052 This option is presently supported only by some bare-metal MIPS
20053 configurations, where it may select a special set of libraries
20054 that lack all floating-point support (including, for example, the
20055 floating-point @code{printf} formats).
20056 If code compiled with @option{-mno-float} accidentally contains
20057 floating-point operations, it is likely to suffer a link-time
20058 or run-time failure.
20060 @item -msingle-float
20061 @opindex msingle-float
20062 Assume that the floating-point coprocessor only supports single-precision
20065 @item -mdouble-float
20066 @opindex mdouble-float
20067 Assume that the floating-point coprocessor supports double-precision
20068 operations. This is the default.
20071 @itemx -mno-odd-spreg
20072 @opindex modd-spreg
20073 @opindex mno-odd-spreg
20074 Enable the use of odd-numbered single-precision floating-point registers
20075 for the o32 ABI. This is the default for processors that are known to
20076 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
20080 @itemx -mabs=legacy
20082 @opindex mabs=legacy
20083 These options control the treatment of the special not-a-number (NaN)
20084 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
20085 @code{neg.@i{fmt}} machine instructions.
20087 By default or when @option{-mabs=legacy} is used the legacy
20088 treatment is selected. In this case these instructions are considered
20089 arithmetic and avoided where correct operation is required and the
20090 input operand might be a NaN. A longer sequence of instructions that
20091 manipulate the sign bit of floating-point datum manually is used
20092 instead unless the @option{-ffinite-math-only} option has also been
20095 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
20096 this case these instructions are considered non-arithmetic and therefore
20097 operating correctly in all cases, including in particular where the
20098 input operand is a NaN. These instructions are therefore always used
20099 for the respective operations.
20102 @itemx -mnan=legacy
20104 @opindex mnan=legacy
20105 These options control the encoding of the special not-a-number (NaN)
20106 IEEE 754 floating-point data.
20108 The @option{-mnan=legacy} option selects the legacy encoding. In this
20109 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
20110 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
20111 by the first bit of their trailing significand field being 1.
20113 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
20114 this case qNaNs are denoted by the first bit of their trailing
20115 significand field being 1, whereas sNaNs are denoted by the first bit of
20116 their trailing significand field being 0.
20118 The default is @option{-mnan=legacy} unless GCC has been configured with
20119 @option{--with-nan=2008}.
20125 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
20126 implement atomic memory built-in functions. When neither option is
20127 specified, GCC uses the instructions if the target architecture
20130 @option{-mllsc} is useful if the runtime environment can emulate the
20131 instructions and @option{-mno-llsc} can be useful when compiling for
20132 nonstandard ISAs. You can make either option the default by
20133 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
20134 respectively. @option{--with-llsc} is the default for some
20135 configurations; see the installation documentation for details.
20141 Use (do not use) revision 1 of the MIPS DSP ASE@.
20142 @xref{MIPS DSP Built-in Functions}. This option defines the
20143 preprocessor macro @code{__mips_dsp}. It also defines
20144 @code{__mips_dsp_rev} to 1.
20150 Use (do not use) revision 2 of the MIPS DSP ASE@.
20151 @xref{MIPS DSP Built-in Functions}. This option defines the
20152 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
20153 It also defines @code{__mips_dsp_rev} to 2.
20156 @itemx -mno-smartmips
20157 @opindex msmartmips
20158 @opindex mno-smartmips
20159 Use (do not use) the MIPS SmartMIPS ASE.
20161 @item -mpaired-single
20162 @itemx -mno-paired-single
20163 @opindex mpaired-single
20164 @opindex mno-paired-single
20165 Use (do not use) paired-single floating-point instructions.
20166 @xref{MIPS Paired-Single Support}. This option requires
20167 hardware floating-point support to be enabled.
20173 Use (do not use) MIPS Digital Media Extension instructions.
20174 This option can only be used when generating 64-bit code and requires
20175 hardware floating-point support to be enabled.
20180 @opindex mno-mips3d
20181 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
20182 The option @option{-mips3d} implies @option{-mpaired-single}.
20185 @itemx -mno-micromips
20186 @opindex mmicromips
20187 @opindex mno-mmicromips
20188 Generate (do not generate) microMIPS code.
20190 MicroMIPS code generation can also be controlled on a per-function basis
20191 by means of @code{micromips} and @code{nomicromips} attributes.
20192 @xref{Function Attributes}, for more information.
20198 Use (do not use) MT Multithreading instructions.
20204 Use (do not use) the MIPS MCU ASE instructions.
20210 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
20216 Use (do not use) the MIPS Virtualization (VZ) instructions.
20222 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
20226 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
20227 an explanation of the default and the way that the pointer size is
20232 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
20234 The default size of @code{int}s, @code{long}s and pointers depends on
20235 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
20236 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
20237 32-bit @code{long}s. Pointers are the same size as @code{long}s,
20238 or the same size as integer registers, whichever is smaller.
20244 Assume (do not assume) that all symbols have 32-bit values, regardless
20245 of the selected ABI@. This option is useful in combination with
20246 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
20247 to generate shorter and faster references to symbolic addresses.
20251 Put definitions of externally-visible data in a small data section
20252 if that data is no bigger than @var{num} bytes. GCC can then generate
20253 more efficient accesses to the data; see @option{-mgpopt} for details.
20255 The default @option{-G} option depends on the configuration.
20257 @item -mlocal-sdata
20258 @itemx -mno-local-sdata
20259 @opindex mlocal-sdata
20260 @opindex mno-local-sdata
20261 Extend (do not extend) the @option{-G} behavior to local data too,
20262 such as to static variables in C@. @option{-mlocal-sdata} is the
20263 default for all configurations.
20265 If the linker complains that an application is using too much small data,
20266 you might want to try rebuilding the less performance-critical parts with
20267 @option{-mno-local-sdata}. You might also want to build large
20268 libraries with @option{-mno-local-sdata}, so that the libraries leave
20269 more room for the main program.
20271 @item -mextern-sdata
20272 @itemx -mno-extern-sdata
20273 @opindex mextern-sdata
20274 @opindex mno-extern-sdata
20275 Assume (do not assume) that externally-defined data is in
20276 a small data section if the size of that data is within the @option{-G} limit.
20277 @option{-mextern-sdata} is the default for all configurations.
20279 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
20280 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
20281 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
20282 is placed in a small data section. If @var{Var} is defined by another
20283 module, you must either compile that module with a high-enough
20284 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
20285 definition. If @var{Var} is common, you must link the application
20286 with a high-enough @option{-G} setting.
20288 The easiest way of satisfying these restrictions is to compile
20289 and link every module with the same @option{-G} option. However,
20290 you may wish to build a library that supports several different
20291 small data limits. You can do this by compiling the library with
20292 the highest supported @option{-G} setting and additionally using
20293 @option{-mno-extern-sdata} to stop the library from making assumptions
20294 about externally-defined data.
20300 Use (do not use) GP-relative accesses for symbols that are known to be
20301 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
20302 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
20305 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
20306 might not hold the value of @code{_gp}. For example, if the code is
20307 part of a library that might be used in a boot monitor, programs that
20308 call boot monitor routines pass an unknown value in @code{$gp}.
20309 (In such situations, the boot monitor itself is usually compiled
20310 with @option{-G0}.)
20312 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
20313 @option{-mno-extern-sdata}.
20315 @item -membedded-data
20316 @itemx -mno-embedded-data
20317 @opindex membedded-data
20318 @opindex mno-embedded-data
20319 Allocate variables to the read-only data section first if possible, then
20320 next in the small data section if possible, otherwise in data. This gives
20321 slightly slower code than the default, but reduces the amount of RAM required
20322 when executing, and thus may be preferred for some embedded systems.
20324 @item -muninit-const-in-rodata
20325 @itemx -mno-uninit-const-in-rodata
20326 @opindex muninit-const-in-rodata
20327 @opindex mno-uninit-const-in-rodata
20328 Put uninitialized @code{const} variables in the read-only data section.
20329 This option is only meaningful in conjunction with @option{-membedded-data}.
20331 @item -mcode-readable=@var{setting}
20332 @opindex mcode-readable
20333 Specify whether GCC may generate code that reads from executable sections.
20334 There are three possible settings:
20337 @item -mcode-readable=yes
20338 Instructions may freely access executable sections. This is the
20341 @item -mcode-readable=pcrel
20342 MIPS16 PC-relative load instructions can access executable sections,
20343 but other instructions must not do so. This option is useful on 4KSc
20344 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
20345 It is also useful on processors that can be configured to have a dual
20346 instruction/data SRAM interface and that, like the M4K, automatically
20347 redirect PC-relative loads to the instruction RAM.
20349 @item -mcode-readable=no
20350 Instructions must not access executable sections. This option can be
20351 useful on targets that are configured to have a dual instruction/data
20352 SRAM interface but that (unlike the M4K) do not automatically redirect
20353 PC-relative loads to the instruction RAM.
20356 @item -msplit-addresses
20357 @itemx -mno-split-addresses
20358 @opindex msplit-addresses
20359 @opindex mno-split-addresses
20360 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
20361 relocation operators. This option has been superseded by
20362 @option{-mexplicit-relocs} but is retained for backwards compatibility.
20364 @item -mexplicit-relocs
20365 @itemx -mno-explicit-relocs
20366 @opindex mexplicit-relocs
20367 @opindex mno-explicit-relocs
20368 Use (do not use) assembler relocation operators when dealing with symbolic
20369 addresses. The alternative, selected by @option{-mno-explicit-relocs},
20370 is to use assembler macros instead.
20372 @option{-mexplicit-relocs} is the default if GCC was configured
20373 to use an assembler that supports relocation operators.
20375 @item -mcheck-zero-division
20376 @itemx -mno-check-zero-division
20377 @opindex mcheck-zero-division
20378 @opindex mno-check-zero-division
20379 Trap (do not trap) on integer division by zero.
20381 The default is @option{-mcheck-zero-division}.
20383 @item -mdivide-traps
20384 @itemx -mdivide-breaks
20385 @opindex mdivide-traps
20386 @opindex mdivide-breaks
20387 MIPS systems check for division by zero by generating either a
20388 conditional trap or a break instruction. Using traps results in
20389 smaller code, but is only supported on MIPS II and later. Also, some
20390 versions of the Linux kernel have a bug that prevents trap from
20391 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
20392 allow conditional traps on architectures that support them and
20393 @option{-mdivide-breaks} to force the use of breaks.
20395 The default is usually @option{-mdivide-traps}, but this can be
20396 overridden at configure time using @option{--with-divide=breaks}.
20397 Divide-by-zero checks can be completely disabled using
20398 @option{-mno-check-zero-division}.
20400 @item -mload-store-pairs
20401 @itemx -mno-load-store-pairs
20402 @opindex mload-store-pairs
20403 @opindex mno-load-store-pairs
20404 Enable (disable) an optimization that pairs consecutive load or store
20405 instructions to enable load/store bonding. This option is enabled by
20406 default but only takes effect when the selected architecture is known
20407 to support bonding.
20412 @opindex mno-memcpy
20413 Force (do not force) the use of @code{memcpy} for non-trivial block
20414 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
20415 most constant-sized copies.
20418 @itemx -mno-long-calls
20419 @opindex mlong-calls
20420 @opindex mno-long-calls
20421 Disable (do not disable) use of the @code{jal} instruction. Calling
20422 functions using @code{jal} is more efficient but requires the caller
20423 and callee to be in the same 256 megabyte segment.
20425 This option has no effect on abicalls code. The default is
20426 @option{-mno-long-calls}.
20432 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
20433 instructions, as provided by the R4650 ISA@.
20439 Enable (disable) use of the @code{madd} and @code{msub} integer
20440 instructions. The default is @option{-mimadd} on architectures
20441 that support @code{madd} and @code{msub} except for the 74k
20442 architecture where it was found to generate slower code.
20445 @itemx -mno-fused-madd
20446 @opindex mfused-madd
20447 @opindex mno-fused-madd
20448 Enable (disable) use of the floating-point multiply-accumulate
20449 instructions, when they are available. The default is
20450 @option{-mfused-madd}.
20452 On the R8000 CPU when multiply-accumulate instructions are used,
20453 the intermediate product is calculated to infinite precision
20454 and is not subject to the FCSR Flush to Zero bit. This may be
20455 undesirable in some circumstances. On other processors the result
20456 is numerically identical to the equivalent computation using
20457 separate multiply, add, subtract and negate instructions.
20461 Tell the MIPS assembler to not run its preprocessor over user
20462 assembler files (with a @samp{.s} suffix) when assembling them.
20467 @opindex mno-fix-24k
20468 Work around the 24K E48 (lost data on stores during refill) errata.
20469 The workarounds are implemented by the assembler rather than by GCC@.
20472 @itemx -mno-fix-r4000
20473 @opindex mfix-r4000
20474 @opindex mno-fix-r4000
20475 Work around certain R4000 CPU errata:
20478 A double-word or a variable shift may give an incorrect result if executed
20479 immediately after starting an integer division.
20481 A double-word or a variable shift may give an incorrect result if executed
20482 while an integer multiplication is in progress.
20484 An integer division may give an incorrect result if started in a delay slot
20485 of a taken branch or a jump.
20489 @itemx -mno-fix-r4400
20490 @opindex mfix-r4400
20491 @opindex mno-fix-r4400
20492 Work around certain R4400 CPU errata:
20495 A double-word or a variable shift may give an incorrect result if executed
20496 immediately after starting an integer division.
20500 @itemx -mno-fix-r10000
20501 @opindex mfix-r10000
20502 @opindex mno-fix-r10000
20503 Work around certain R10000 errata:
20506 @code{ll}/@code{sc} sequences may not behave atomically on revisions
20507 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
20510 This option can only be used if the target architecture supports
20511 branch-likely instructions. @option{-mfix-r10000} is the default when
20512 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
20516 @itemx -mno-fix-rm7000
20517 @opindex mfix-rm7000
20518 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
20519 workarounds are implemented by the assembler rather than by GCC@.
20522 @itemx -mno-fix-vr4120
20523 @opindex mfix-vr4120
20524 Work around certain VR4120 errata:
20527 @code{dmultu} does not always produce the correct result.
20529 @code{div} and @code{ddiv} do not always produce the correct result if one
20530 of the operands is negative.
20532 The workarounds for the division errata rely on special functions in
20533 @file{libgcc.a}. At present, these functions are only provided by
20534 the @code{mips64vr*-elf} configurations.
20536 Other VR4120 errata require a NOP to be inserted between certain pairs of
20537 instructions. These errata are handled by the assembler, not by GCC itself.
20540 @opindex mfix-vr4130
20541 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
20542 workarounds are implemented by the assembler rather than by GCC,
20543 although GCC avoids using @code{mflo} and @code{mfhi} if the
20544 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
20545 instructions are available instead.
20548 @itemx -mno-fix-sb1
20550 Work around certain SB-1 CPU core errata.
20551 (This flag currently works around the SB-1 revision 2
20552 ``F1'' and ``F2'' floating-point errata.)
20554 @item -mr10k-cache-barrier=@var{setting}
20555 @opindex mr10k-cache-barrier
20556 Specify whether GCC should insert cache barriers to avoid the
20557 side-effects of speculation on R10K processors.
20559 In common with many processors, the R10K tries to predict the outcome
20560 of a conditional branch and speculatively executes instructions from
20561 the ``taken'' branch. It later aborts these instructions if the
20562 predicted outcome is wrong. However, on the R10K, even aborted
20563 instructions can have side effects.
20565 This problem only affects kernel stores and, depending on the system,
20566 kernel loads. As an example, a speculatively-executed store may load
20567 the target memory into cache and mark the cache line as dirty, even if
20568 the store itself is later aborted. If a DMA operation writes to the
20569 same area of memory before the ``dirty'' line is flushed, the cached
20570 data overwrites the DMA-ed data. See the R10K processor manual
20571 for a full description, including other potential problems.
20573 One workaround is to insert cache barrier instructions before every memory
20574 access that might be speculatively executed and that might have side
20575 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
20576 controls GCC's implementation of this workaround. It assumes that
20577 aborted accesses to any byte in the following regions does not have
20582 the memory occupied by the current function's stack frame;
20585 the memory occupied by an incoming stack argument;
20588 the memory occupied by an object with a link-time-constant address.
20591 It is the kernel's responsibility to ensure that speculative
20592 accesses to these regions are indeed safe.
20594 If the input program contains a function declaration such as:
20600 then the implementation of @code{foo} must allow @code{j foo} and
20601 @code{jal foo} to be executed speculatively. GCC honors this
20602 restriction for functions it compiles itself. It expects non-GCC
20603 functions (such as hand-written assembly code) to do the same.
20605 The option has three forms:
20608 @item -mr10k-cache-barrier=load-store
20609 Insert a cache barrier before a load or store that might be
20610 speculatively executed and that might have side effects even
20613 @item -mr10k-cache-barrier=store
20614 Insert a cache barrier before a store that might be speculatively
20615 executed and that might have side effects even if aborted.
20617 @item -mr10k-cache-barrier=none
20618 Disable the insertion of cache barriers. This is the default setting.
20621 @item -mflush-func=@var{func}
20622 @itemx -mno-flush-func
20623 @opindex mflush-func
20624 Specifies the function to call to flush the I and D caches, or to not
20625 call any such function. If called, the function must take the same
20626 arguments as the common @code{_flush_func}, that is, the address of the
20627 memory range for which the cache is being flushed, the size of the
20628 memory range, and the number 3 (to flush both caches). The default
20629 depends on the target GCC was configured for, but commonly is either
20630 @code{_flush_func} or @code{__cpu_flush}.
20632 @item mbranch-cost=@var{num}
20633 @opindex mbranch-cost
20634 Set the cost of branches to roughly @var{num} ``simple'' instructions.
20635 This cost is only a heuristic and is not guaranteed to produce
20636 consistent results across releases. A zero cost redundantly selects
20637 the default, which is based on the @option{-mtune} setting.
20639 @item -mbranch-likely
20640 @itemx -mno-branch-likely
20641 @opindex mbranch-likely
20642 @opindex mno-branch-likely
20643 Enable or disable use of Branch Likely instructions, regardless of the
20644 default for the selected architecture. By default, Branch Likely
20645 instructions may be generated if they are supported by the selected
20646 architecture. An exception is for the MIPS32 and MIPS64 architectures
20647 and processors that implement those architectures; for those, Branch
20648 Likely instructions are not be generated by default because the MIPS32
20649 and MIPS64 architectures specifically deprecate their use.
20651 @item -mcompact-branches=never
20652 @itemx -mcompact-branches=optimal
20653 @itemx -mcompact-branches=always
20654 @opindex mcompact-branches=never
20655 @opindex mcompact-branches=optimal
20656 @opindex mcompact-branches=always
20657 These options control which form of branches will be generated. The
20658 default is @option{-mcompact-branches=optimal}.
20660 The @option{-mcompact-branches=never} option ensures that compact branch
20661 instructions will never be generated.
20663 The @option{-mcompact-branches=always} option ensures that a compact
20664 branch instruction will be generated if available. If a compact branch
20665 instruction is not available, a delay slot form of the branch will be
20668 This option is supported from MIPS Release 6 onwards.
20670 The @option{-mcompact-branches=optimal} option will cause a delay slot
20671 branch to be used if one is available in the current ISA and the delay
20672 slot is successfully filled. If the delay slot is not filled, a compact
20673 branch will be chosen if one is available.
20675 @item -mfp-exceptions
20676 @itemx -mno-fp-exceptions
20677 @opindex mfp-exceptions
20678 Specifies whether FP exceptions are enabled. This affects how
20679 FP instructions are scheduled for some processors.
20680 The default is that FP exceptions are
20683 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
20684 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
20687 @item -mvr4130-align
20688 @itemx -mno-vr4130-align
20689 @opindex mvr4130-align
20690 The VR4130 pipeline is two-way superscalar, but can only issue two
20691 instructions together if the first one is 8-byte aligned. When this
20692 option is enabled, GCC aligns pairs of instructions that it
20693 thinks should execute in parallel.
20695 This option only has an effect when optimizing for the VR4130.
20696 It normally makes code faster, but at the expense of making it bigger.
20697 It is enabled by default at optimization level @option{-O3}.
20702 Enable (disable) generation of @code{synci} instructions on
20703 architectures that support it. The @code{synci} instructions (if
20704 enabled) are generated when @code{__builtin___clear_cache} is
20707 This option defaults to @option{-mno-synci}, but the default can be
20708 overridden by configuring GCC with @option{--with-synci}.
20710 When compiling code for single processor systems, it is generally safe
20711 to use @code{synci}. However, on many multi-core (SMP) systems, it
20712 does not invalidate the instruction caches on all cores and may lead
20713 to undefined behavior.
20715 @item -mrelax-pic-calls
20716 @itemx -mno-relax-pic-calls
20717 @opindex mrelax-pic-calls
20718 Try to turn PIC calls that are normally dispatched via register
20719 @code{$25} into direct calls. This is only possible if the linker can
20720 resolve the destination at link time and if the destination is within
20721 range for a direct call.
20723 @option{-mrelax-pic-calls} is the default if GCC was configured to use
20724 an assembler and a linker that support the @code{.reloc} assembly
20725 directive and @option{-mexplicit-relocs} is in effect. With
20726 @option{-mno-explicit-relocs}, this optimization can be performed by the
20727 assembler and the linker alone without help from the compiler.
20729 @item -mmcount-ra-address
20730 @itemx -mno-mcount-ra-address
20731 @opindex mmcount-ra-address
20732 @opindex mno-mcount-ra-address
20733 Emit (do not emit) code that allows @code{_mcount} to modify the
20734 calling function's return address. When enabled, this option extends
20735 the usual @code{_mcount} interface with a new @var{ra-address}
20736 parameter, which has type @code{intptr_t *} and is passed in register
20737 @code{$12}. @code{_mcount} can then modify the return address by
20738 doing both of the following:
20741 Returning the new address in register @code{$31}.
20743 Storing the new address in @code{*@var{ra-address}},
20744 if @var{ra-address} is nonnull.
20747 The default is @option{-mno-mcount-ra-address}.
20749 @item -mframe-header-opt
20750 @itemx -mno-frame-header-opt
20751 @opindex mframe-header-opt
20752 Enable (disable) frame header optimization in the o32 ABI. When using the
20753 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
20754 function to write out register arguments. When enabled, this optimization
20755 will suppress the allocation of the frame header if it can be determined that
20758 This optimization is off by default at all optimization levels.
20761 @itemx -mno-lxc1-sxc1
20762 @opindex mlxc1-sxc1
20763 When applicable, enable (disable) the generation of @code{lwxc1},
20764 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
20769 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
20770 @code{madd.d} and related instructions. Enabled by default.
20775 @subsection MMIX Options
20776 @cindex MMIX Options
20778 These options are defined for the MMIX:
20782 @itemx -mno-libfuncs
20784 @opindex mno-libfuncs
20785 Specify that intrinsic library functions are being compiled, passing all
20786 values in registers, no matter the size.
20789 @itemx -mno-epsilon
20791 @opindex mno-epsilon
20792 Generate floating-point comparison instructions that compare with respect
20793 to the @code{rE} epsilon register.
20795 @item -mabi=mmixware
20797 @opindex mabi=mmixware
20799 Generate code that passes function parameters and return values that (in
20800 the called function) are seen as registers @code{$0} and up, as opposed to
20801 the GNU ABI which uses global registers @code{$231} and up.
20803 @item -mzero-extend
20804 @itemx -mno-zero-extend
20805 @opindex mzero-extend
20806 @opindex mno-zero-extend
20807 When reading data from memory in sizes shorter than 64 bits, use (do not
20808 use) zero-extending load instructions by default, rather than
20809 sign-extending ones.
20812 @itemx -mno-knuthdiv
20814 @opindex mno-knuthdiv
20815 Make the result of a division yielding a remainder have the same sign as
20816 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
20817 remainder follows the sign of the dividend. Both methods are
20818 arithmetically valid, the latter being almost exclusively used.
20820 @item -mtoplevel-symbols
20821 @itemx -mno-toplevel-symbols
20822 @opindex mtoplevel-symbols
20823 @opindex mno-toplevel-symbols
20824 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
20825 code can be used with the @code{PREFIX} assembly directive.
20829 Generate an executable in the ELF format, rather than the default
20830 @samp{mmo} format used by the @command{mmix} simulator.
20832 @item -mbranch-predict
20833 @itemx -mno-branch-predict
20834 @opindex mbranch-predict
20835 @opindex mno-branch-predict
20836 Use (do not use) the probable-branch instructions, when static branch
20837 prediction indicates a probable branch.
20839 @item -mbase-addresses
20840 @itemx -mno-base-addresses
20841 @opindex mbase-addresses
20842 @opindex mno-base-addresses
20843 Generate (do not generate) code that uses @emph{base addresses}. Using a
20844 base address automatically generates a request (handled by the assembler
20845 and the linker) for a constant to be set up in a global register. The
20846 register is used for one or more base address requests within the range 0
20847 to 255 from the value held in the register. The generally leads to short
20848 and fast code, but the number of different data items that can be
20849 addressed is limited. This means that a program that uses lots of static
20850 data may require @option{-mno-base-addresses}.
20852 @item -msingle-exit
20853 @itemx -mno-single-exit
20854 @opindex msingle-exit
20855 @opindex mno-single-exit
20856 Force (do not force) generated code to have a single exit point in each
20860 @node MN10300 Options
20861 @subsection MN10300 Options
20862 @cindex MN10300 options
20864 These @option{-m} options are defined for Matsushita MN10300 architectures:
20869 Generate code to avoid bugs in the multiply instructions for the MN10300
20870 processors. This is the default.
20872 @item -mno-mult-bug
20873 @opindex mno-mult-bug
20874 Do not generate code to avoid bugs in the multiply instructions for the
20875 MN10300 processors.
20879 Generate code using features specific to the AM33 processor.
20883 Do not generate code using features specific to the AM33 processor. This
20888 Generate code using features specific to the AM33/2.0 processor.
20892 Generate code using features specific to the AM34 processor.
20894 @item -mtune=@var{cpu-type}
20896 Use the timing characteristics of the indicated CPU type when
20897 scheduling instructions. This does not change the targeted processor
20898 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
20899 @samp{am33-2} or @samp{am34}.
20901 @item -mreturn-pointer-on-d0
20902 @opindex mreturn-pointer-on-d0
20903 When generating a function that returns a pointer, return the pointer
20904 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
20905 only in @code{a0}, and attempts to call such functions without a prototype
20906 result in errors. Note that this option is on by default; use
20907 @option{-mno-return-pointer-on-d0} to disable it.
20911 Do not link in the C run-time initialization object file.
20915 Indicate to the linker that it should perform a relaxation optimization pass
20916 to shorten branches, calls and absolute memory addresses. This option only
20917 has an effect when used on the command line for the final link step.
20919 This option makes symbolic debugging impossible.
20923 Allow the compiler to generate @emph{Long Instruction Word}
20924 instructions if the target is the @samp{AM33} or later. This is the
20925 default. This option defines the preprocessor macro @code{__LIW__}.
20929 Do not allow the compiler to generate @emph{Long Instruction Word}
20930 instructions. This option defines the preprocessor macro
20935 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
20936 instructions if the target is the @samp{AM33} or later. This is the
20937 default. This option defines the preprocessor macro @code{__SETLB__}.
20941 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
20942 instructions. This option defines the preprocessor macro
20943 @code{__NO_SETLB__}.
20947 @node Moxie Options
20948 @subsection Moxie Options
20949 @cindex Moxie Options
20955 Generate big-endian code. This is the default for @samp{moxie-*-*}
20960 Generate little-endian code.
20964 Generate mul.x and umul.x instructions. This is the default for
20965 @samp{moxiebox-*-*} configurations.
20969 Do not link in the C run-time initialization object file.
20973 @node MSP430 Options
20974 @subsection MSP430 Options
20975 @cindex MSP430 Options
20977 These options are defined for the MSP430:
20983 Force assembly output to always use hex constants. Normally such
20984 constants are signed decimals, but this option is available for
20985 testsuite and/or aesthetic purposes.
20989 Select the MCU to target. This is used to create a C preprocessor
20990 symbol based upon the MCU name, converted to upper case and pre- and
20991 post-fixed with @samp{__}. This in turn is used by the
20992 @file{msp430.h} header file to select an MCU-specific supplementary
20995 The option also sets the ISA to use. If the MCU name is one that is
20996 known to only support the 430 ISA then that is selected, otherwise the
20997 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
20998 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
20999 name selects the 430X ISA.
21001 In addition an MCU-specific linker script is added to the linker
21002 command line. The script's name is the name of the MCU with
21003 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
21004 command line defines the C preprocessor symbol @code{__XXX__} and
21005 cause the linker to search for a script called @file{xxx.ld}.
21007 This option is also passed on to the assembler.
21010 @itemx -mno-warn-mcu
21012 @opindex mno-warn-mcu
21013 This option enables or disables warnings about conflicts between the
21014 MCU name specified by the @option{-mmcu} option and the ISA set by the
21015 @option{-mcpu} option and/or the hardware multiply support set by the
21016 @option{-mhwmult} option. It also toggles warnings about unrecognized
21017 MCU names. This option is on by default.
21021 Specifies the ISA to use. Accepted values are @samp{msp430},
21022 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
21023 @option{-mmcu=} option should be used to select the ISA.
21027 Link to the simulator runtime libraries and linker script. Overrides
21028 any scripts that would be selected by the @option{-mmcu=} option.
21032 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
21036 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
21040 This option is passed to the assembler and linker, and allows the
21041 linker to perform certain optimizations that cannot be done until
21046 Describes the type of hardware multiply supported by the target.
21047 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
21048 for the original 16-bit-only multiply supported by early MCUs.
21049 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
21050 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
21051 A value of @samp{auto} can also be given. This tells GCC to deduce
21052 the hardware multiply support based upon the MCU name provided by the
21053 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
21054 the MCU name is not recognized then no hardware multiply support is
21055 assumed. @code{auto} is the default setting.
21057 Hardware multiplies are normally performed by calling a library
21058 routine. This saves space in the generated code. When compiling at
21059 @option{-O3} or higher however the hardware multiplier is invoked
21060 inline. This makes for bigger, but faster code.
21062 The hardware multiply routines disable interrupts whilst running and
21063 restore the previous interrupt state when they finish. This makes
21064 them safe to use inside interrupt handlers as well as in normal code.
21068 Enable the use of a minimum runtime environment - no static
21069 initializers or constructors. This is intended for memory-constrained
21070 devices. The compiler includes special symbols in some objects
21071 that tell the linker and runtime which code fragments are required.
21073 @item -mcode-region=
21074 @itemx -mdata-region=
21075 @opindex mcode-region
21076 @opindex mdata-region
21077 These options tell the compiler where to place functions and data that
21078 do not have one of the @code{lower}, @code{upper}, @code{either} or
21079 @code{section} attributes. Possible values are @code{lower},
21080 @code{upper}, @code{either} or @code{any}. The first three behave
21081 like the corresponding attribute. The fourth possible value -
21082 @code{any} - is the default. It leaves placement entirely up to the
21083 linker script and how it assigns the standard sections
21084 (@code{.text}, @code{.data}, etc) to the memory regions.
21086 @item -msilicon-errata=
21087 @opindex msilicon-errata
21088 This option passes on a request to assembler to enable the fixes for
21089 the named silicon errata.
21091 @item -msilicon-errata-warn=
21092 @opindex msilicon-errata-warn
21093 This option passes on a request to the assembler to enable warning
21094 messages when a silicon errata might need to be applied.
21098 @node NDS32 Options
21099 @subsection NDS32 Options
21100 @cindex NDS32 Options
21102 These options are defined for NDS32 implementations:
21107 @opindex mbig-endian
21108 Generate code in big-endian mode.
21110 @item -mlittle-endian
21111 @opindex mlittle-endian
21112 Generate code in little-endian mode.
21114 @item -mreduced-regs
21115 @opindex mreduced-regs
21116 Use reduced-set registers for register allocation.
21119 @opindex mfull-regs
21120 Use full-set registers for register allocation.
21124 Generate conditional move instructions.
21128 Do not generate conditional move instructions.
21132 Generate performance extension instructions.
21134 @item -mno-perf-ext
21135 @opindex mno-perf-ext
21136 Do not generate performance extension instructions.
21140 Generate v3 push25/pop25 instructions.
21143 @opindex mno-v3push
21144 Do not generate v3 push25/pop25 instructions.
21148 Generate 16-bit instructions.
21151 @opindex mno-16-bit
21152 Do not generate 16-bit instructions.
21154 @item -misr-vector-size=@var{num}
21155 @opindex misr-vector-size
21156 Specify the size of each interrupt vector, which must be 4 or 16.
21158 @item -mcache-block-size=@var{num}
21159 @opindex mcache-block-size
21160 Specify the size of each cache block,
21161 which must be a power of 2 between 4 and 512.
21163 @item -march=@var{arch}
21165 Specify the name of the target architecture.
21167 @item -mcmodel=@var{code-model}
21169 Set the code model to one of
21172 All the data and read-only data segments must be within 512KB addressing space.
21173 The text segment must be within 16MB addressing space.
21174 @item @samp{medium}
21175 The data segment must be within 512KB while the read-only data segment can be
21176 within 4GB addressing space. The text segment should be still within 16MB
21179 All the text and data segments can be within 4GB addressing space.
21183 @opindex mctor-dtor
21184 Enable constructor/destructor feature.
21188 Guide linker to relax instructions.
21192 @node Nios II Options
21193 @subsection Nios II Options
21194 @cindex Nios II options
21195 @cindex Altera Nios II options
21197 These are the options defined for the Altera Nios II processor.
21203 @cindex smaller data references
21204 Put global and static objects less than or equal to @var{num} bytes
21205 into the small data or BSS sections instead of the normal data or BSS
21206 sections. The default value of @var{num} is 8.
21208 @item -mgpopt=@var{option}
21213 Generate (do not generate) GP-relative accesses. The following
21214 @var{option} names are recognized:
21219 Do not generate GP-relative accesses.
21222 Generate GP-relative accesses for small data objects that are not
21223 external, weak, or uninitialized common symbols.
21224 Also use GP-relative addressing for objects that
21225 have been explicitly placed in a small data section via a @code{section}
21229 As for @samp{local}, but also generate GP-relative accesses for
21230 small data objects that are external, weak, or common. If you use this option,
21231 you must ensure that all parts of your program (including libraries) are
21232 compiled with the same @option{-G} setting.
21235 Generate GP-relative accesses for all data objects in the program. If you
21236 use this option, the entire data and BSS segments
21237 of your program must fit in 64K of memory and you must use an appropriate
21238 linker script to allocate them within the addressable range of the
21242 Generate GP-relative addresses for function pointers as well as data
21243 pointers. If you use this option, the entire text, data, and BSS segments
21244 of your program must fit in 64K of memory and you must use an appropriate
21245 linker script to allocate them within the addressable range of the
21250 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
21251 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
21253 The default is @option{-mgpopt} except when @option{-fpic} or
21254 @option{-fPIC} is specified to generate position-independent code.
21255 Note that the Nios II ABI does not permit GP-relative accesses from
21258 You may need to specify @option{-mno-gpopt} explicitly when building
21259 programs that include large amounts of small data, including large
21260 GOT data sections. In this case, the 16-bit offset for GP-relative
21261 addressing may not be large enough to allow access to the entire
21262 small data section.
21264 @item -mgprel-sec=@var{regexp}
21265 @opindex mgprel-sec
21266 This option specifies additional section names that can be accessed via
21267 GP-relative addressing. It is most useful in conjunction with
21268 @code{section} attributes on variable declarations
21269 (@pxref{Common Variable Attributes}) and a custom linker script.
21270 The @var{regexp} is a POSIX Extended Regular Expression.
21272 This option does not affect the behavior of the @option{-G} option, and
21273 and the specified sections are in addition to the standard @code{.sdata}
21274 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
21276 @item -mr0rel-sec=@var{regexp}
21277 @opindex mr0rel-sec
21278 This option specifies names of sections that can be accessed via a
21279 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
21280 of the 32-bit address space. It is most useful in conjunction with
21281 @code{section} attributes on variable declarations
21282 (@pxref{Common Variable Attributes}) and a custom linker script.
21283 The @var{regexp} is a POSIX Extended Regular Expression.
21285 In contrast to the use of GP-relative addressing for small data,
21286 zero-based addressing is never generated by default and there are no
21287 conventional section names used in standard linker scripts for sections
21288 in the low or high areas of memory.
21294 Generate little-endian (default) or big-endian (experimental) code,
21297 @item -march=@var{arch}
21299 This specifies the name of the target Nios II architecture. GCC uses this
21300 name to determine what kind of instructions it can emit when generating
21301 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
21303 The preprocessor macro @code{__nios2_arch__} is available to programs,
21304 with value 1 or 2, indicating the targeted ISA level.
21306 @item -mbypass-cache
21307 @itemx -mno-bypass-cache
21308 @opindex mno-bypass-cache
21309 @opindex mbypass-cache
21310 Force all load and store instructions to always bypass cache by
21311 using I/O variants of the instructions. The default is not to
21314 @item -mno-cache-volatile
21315 @itemx -mcache-volatile
21316 @opindex mcache-volatile
21317 @opindex mno-cache-volatile
21318 Volatile memory access bypass the cache using the I/O variants of
21319 the load and store instructions. The default is not to bypass the cache.
21321 @item -mno-fast-sw-div
21322 @itemx -mfast-sw-div
21323 @opindex mno-fast-sw-div
21324 @opindex mfast-sw-div
21325 Do not use table-based fast divide for small numbers. The default
21326 is to use the fast divide at @option{-O3} and above.
21330 @itemx -mno-hw-mulx
21334 @opindex mno-hw-mul
21336 @opindex mno-hw-mulx
21338 @opindex mno-hw-div
21340 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
21341 instructions by the compiler. The default is to emit @code{mul}
21342 and not emit @code{div} and @code{mulx}.
21348 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
21349 CDX (code density) instructions. Enabling these instructions also
21350 requires @option{-march=r2}. Since these instructions are optional
21351 extensions to the R2 architecture, the default is not to emit them.
21353 @item -mcustom-@var{insn}=@var{N}
21354 @itemx -mno-custom-@var{insn}
21355 @opindex mcustom-@var{insn}
21356 @opindex mno-custom-@var{insn}
21357 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
21358 custom instruction with encoding @var{N} when generating code that uses
21359 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
21360 instruction 253 for single-precision floating-point add operations instead
21361 of the default behavior of using a library call.
21363 The following values of @var{insn} are supported. Except as otherwise
21364 noted, floating-point operations are expected to be implemented with
21365 normal IEEE 754 semantics and correspond directly to the C operators or the
21366 equivalent GCC built-in functions (@pxref{Other Builtins}).
21368 Single-precision floating point:
21371 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
21372 Binary arithmetic operations.
21378 Unary absolute value.
21380 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
21381 Comparison operations.
21383 @item @samp{fmins}, @samp{fmaxs}
21384 Floating-point minimum and maximum. These instructions are only
21385 generated if @option{-ffinite-math-only} is specified.
21387 @item @samp{fsqrts}
21388 Unary square root operation.
21390 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
21391 Floating-point trigonometric and exponential functions. These instructions
21392 are only generated if @option{-funsafe-math-optimizations} is also specified.
21396 Double-precision floating point:
21399 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
21400 Binary arithmetic operations.
21406 Unary absolute value.
21408 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
21409 Comparison operations.
21411 @item @samp{fmind}, @samp{fmaxd}
21412 Double-precision minimum and maximum. These instructions are only
21413 generated if @option{-ffinite-math-only} is specified.
21415 @item @samp{fsqrtd}
21416 Unary square root operation.
21418 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
21419 Double-precision trigonometric and exponential functions. These instructions
21420 are only generated if @option{-funsafe-math-optimizations} is also specified.
21426 @item @samp{fextsd}
21427 Conversion from single precision to double precision.
21429 @item @samp{ftruncds}
21430 Conversion from double precision to single precision.
21432 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
21433 Conversion from floating point to signed or unsigned integer types, with
21434 truncation towards zero.
21437 Conversion from single-precision floating point to signed integer,
21438 rounding to the nearest integer and ties away from zero.
21439 This corresponds to the @code{__builtin_lroundf} function when
21440 @option{-fno-math-errno} is used.
21442 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
21443 Conversion from signed or unsigned integer types to floating-point types.
21447 In addition, all of the following transfer instructions for internal
21448 registers X and Y must be provided to use any of the double-precision
21449 floating-point instructions. Custom instructions taking two
21450 double-precision source operands expect the first operand in the
21451 64-bit register X. The other operand (or only operand of a unary
21452 operation) is given to the custom arithmetic instruction with the
21453 least significant half in source register @var{src1} and the most
21454 significant half in @var{src2}. A custom instruction that returns a
21455 double-precision result returns the most significant 32 bits in the
21456 destination register and the other half in 32-bit register Y.
21457 GCC automatically generates the necessary code sequences to write
21458 register X and/or read register Y when double-precision floating-point
21459 instructions are used.
21464 Write @var{src1} into the least significant half of X and @var{src2} into
21465 the most significant half of X.
21468 Write @var{src1} into Y.
21470 @item @samp{frdxhi}, @samp{frdxlo}
21471 Read the most or least (respectively) significant half of X and store it in
21475 Read the value of Y and store it into @var{dest}.
21478 Note that you can gain more local control over generation of Nios II custom
21479 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
21480 and @code{target("no-custom-@var{insn}")} function attributes
21481 (@pxref{Function Attributes})
21482 or pragmas (@pxref{Function Specific Option Pragmas}).
21484 @item -mcustom-fpu-cfg=@var{name}
21485 @opindex mcustom-fpu-cfg
21487 This option enables a predefined, named set of custom instruction encodings
21488 (see @option{-mcustom-@var{insn}} above).
21489 Currently, the following sets are defined:
21491 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
21492 @gccoptlist{-mcustom-fmuls=252 @gol
21493 -mcustom-fadds=253 @gol
21494 -mcustom-fsubs=254 @gol
21495 -fsingle-precision-constant}
21497 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
21498 @gccoptlist{-mcustom-fmuls=252 @gol
21499 -mcustom-fadds=253 @gol
21500 -mcustom-fsubs=254 @gol
21501 -mcustom-fdivs=255 @gol
21502 -fsingle-precision-constant}
21504 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
21505 @gccoptlist{-mcustom-floatus=243 @gol
21506 -mcustom-fixsi=244 @gol
21507 -mcustom-floatis=245 @gol
21508 -mcustom-fcmpgts=246 @gol
21509 -mcustom-fcmples=249 @gol
21510 -mcustom-fcmpeqs=250 @gol
21511 -mcustom-fcmpnes=251 @gol
21512 -mcustom-fmuls=252 @gol
21513 -mcustom-fadds=253 @gol
21514 -mcustom-fsubs=254 @gol
21515 -mcustom-fdivs=255 @gol
21516 -fsingle-precision-constant}
21518 Custom instruction assignments given by individual
21519 @option{-mcustom-@var{insn}=} options override those given by
21520 @option{-mcustom-fpu-cfg=}, regardless of the
21521 order of the options on the command line.
21523 Note that you can gain more local control over selection of a FPU
21524 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
21525 function attribute (@pxref{Function Attributes})
21526 or pragma (@pxref{Function Specific Option Pragmas}).
21530 These additional @samp{-m} options are available for the Altera Nios II
21531 ELF (bare-metal) target:
21537 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
21538 startup and termination code, and is typically used in conjunction with
21539 @option{-msys-crt0=} to specify the location of the alternate startup code
21540 provided by the HAL BSP.
21544 Link with a limited version of the C library, @option{-lsmallc}, rather than
21547 @item -msys-crt0=@var{startfile}
21549 @var{startfile} is the file name of the startfile (crt0) to use
21550 when linking. This option is only useful in conjunction with @option{-mhal}.
21552 @item -msys-lib=@var{systemlib}
21554 @var{systemlib} is the library name of the library that provides
21555 low-level system calls required by the C library,
21556 e.g. @code{read} and @code{write}.
21557 This option is typically used to link with a library provided by a HAL BSP.
21561 @node Nvidia PTX Options
21562 @subsection Nvidia PTX Options
21563 @cindex Nvidia PTX options
21564 @cindex nvptx options
21566 These options are defined for Nvidia PTX:
21574 Generate code for 32-bit or 64-bit ABI.
21577 @opindex mmainkernel
21578 Link in code for a __main kernel. This is for stand-alone instead of
21579 offloading execution.
21583 Apply partitioned execution optimizations. This is the default when any
21584 level of optimization is selected.
21587 @opindex msoft-stack
21588 Generate code that does not use @code{.local} memory
21589 directly for stack storage. Instead, a per-warp stack pointer is
21590 maintained explicitly. This enables variable-length stack allocation (with
21591 variable-length arrays or @code{alloca}), and when global memory is used for
21592 underlying storage, makes it possible to access automatic variables from other
21593 threads, or with atomic instructions. This code generation variant is used
21594 for OpenMP offloading, but the option is exposed on its own for the purpose
21595 of testing the compiler; to generate code suitable for linking into programs
21596 using OpenMP offloading, use option @option{-mgomp}.
21598 @item -muniform-simt
21599 @opindex muniform-simt
21600 Switch to code generation variant that allows to execute all threads in each
21601 warp, while maintaining memory state and side effects as if only one thread
21602 in each warp was active outside of OpenMP SIMD regions. All atomic operations
21603 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
21604 current lane index equals the master lane index), and the register being
21605 assigned is copied via a shuffle instruction from the master lane. Outside of
21606 SIMD regions lane 0 is the master; inside, each thread sees itself as the
21607 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
21608 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
21609 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
21610 with current lane index to compute the master lane index.
21614 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
21615 @option{-muniform-simt} options, and selects corresponding multilib variant.
21619 @node PDP-11 Options
21620 @subsection PDP-11 Options
21621 @cindex PDP-11 Options
21623 These options are defined for the PDP-11:
21628 Use hardware FPP floating point. This is the default. (FIS floating
21629 point on the PDP-11/40 is not supported.)
21632 @opindex msoft-float
21633 Do not use hardware floating point.
21637 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
21641 Return floating-point results in memory. This is the default.
21645 Generate code for a PDP-11/40.
21649 Generate code for a PDP-11/45. This is the default.
21653 Generate code for a PDP-11/10.
21655 @item -mbcopy-builtin
21656 @opindex mbcopy-builtin
21657 Use inline @code{movmemhi} patterns for copying memory. This is the
21662 Do not use inline @code{movmemhi} patterns for copying memory.
21668 Use 16-bit @code{int}. This is the default.
21674 Use 32-bit @code{int}.
21677 @itemx -mno-float32
21679 @opindex mno-float32
21680 Use 64-bit @code{float}. This is the default.
21683 @itemx -mno-float64
21685 @opindex mno-float64
21686 Use 32-bit @code{float}.
21690 Use @code{abshi2} pattern. This is the default.
21694 Do not use @code{abshi2} pattern.
21696 @item -mbranch-expensive
21697 @opindex mbranch-expensive
21698 Pretend that branches are expensive. This is for experimenting with
21699 code generation only.
21701 @item -mbranch-cheap
21702 @opindex mbranch-cheap
21703 Do not pretend that branches are expensive. This is the default.
21707 Use Unix assembler syntax. This is the default when configured for
21708 @samp{pdp11-*-bsd}.
21712 Use DEC assembler syntax. This is the default when configured for any
21713 PDP-11 target other than @samp{pdp11-*-bsd}.
21716 @node picoChip Options
21717 @subsection picoChip Options
21718 @cindex picoChip options
21720 These @samp{-m} options are defined for picoChip implementations:
21724 @item -mae=@var{ae_type}
21726 Set the instruction set, register set, and instruction scheduling
21727 parameters for array element type @var{ae_type}. Supported values
21728 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
21730 @option{-mae=ANY} selects a completely generic AE type. Code
21731 generated with this option runs on any of the other AE types. The
21732 code is not as efficient as it would be if compiled for a specific
21733 AE type, and some types of operation (e.g., multiplication) do not
21734 work properly on all types of AE.
21736 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
21737 for compiled code, and is the default.
21739 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
21740 option may suffer from poor performance of byte (char) manipulation,
21741 since the DSP AE does not provide hardware support for byte load/stores.
21743 @item -msymbol-as-address
21744 Enable the compiler to directly use a symbol name as an address in a
21745 load/store instruction, without first loading it into a
21746 register. Typically, the use of this option generates larger
21747 programs, which run faster than when the option isn't used. However, the
21748 results vary from program to program, so it is left as a user option,
21749 rather than being permanently enabled.
21751 @item -mno-inefficient-warnings
21752 Disables warnings about the generation of inefficient code. These
21753 warnings can be generated, for example, when compiling code that
21754 performs byte-level memory operations on the MAC AE type. The MAC AE has
21755 no hardware support for byte-level memory operations, so all byte
21756 load/stores must be synthesized from word load/store operations. This is
21757 inefficient and a warning is generated to indicate
21758 that you should rewrite the code to avoid byte operations, or to target
21759 an AE type that has the necessary hardware support. This option disables
21764 @node PowerPC Options
21765 @subsection PowerPC Options
21766 @cindex PowerPC options
21768 These are listed under @xref{RS/6000 and PowerPC Options}.
21770 @node RISC-V Options
21771 @subsection RISC-V Options
21772 @cindex RISC-V Options
21774 These command-line options are defined for RISC-V targets:
21777 @item -mbranch-cost=@var{n}
21778 @opindex mbranch-cost
21779 Set the cost of branches to roughly @var{n} instructions.
21784 When generating PIC code, do or don't allow the use of PLTs. Ignored for
21785 non-PIC. The default is @option{-mplt}.
21787 @item -mabi=@var{ABI-string}
21789 Specify integer and floating-point calling convention. @var{ABI-string}
21790 contains two parts: the size of integer types and the registers used for
21791 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
21792 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
21793 32-bit), and that floating-point values up to 64 bits wide are passed in F
21794 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
21795 allows the compiler to generate code that uses the F and D extensions but only
21796 allows floating-point values up to 32 bits long to be passed in registers; or
21797 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
21798 passed in registers.
21800 The default for this argument is system dependent, users who want a specific
21801 calling convention should specify one explicitly. The valid calling
21802 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
21803 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
21804 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
21805 invalid because the ABI requires 64-bit values be passed in F registers, but F
21806 registers are only 32 bits wide.
21811 Do or don't use hardware floating-point divide and square root instructions.
21812 This requires the F or D extensions for floating-point registers. The default
21813 is to use them if the specified architecture has these instructions.
21818 Do or don't use hardware instructions for integer division. This requires the
21819 M extension. The default is to use them if the specified architecture has
21820 these instructions.
21822 @item -march=@var{ISA-string}
21824 Generate code for given RISC-V ISA (e.g.@ @samp{rv64im}). ISA strings must be
21825 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, and @samp{rv32imaf}.
21827 @item -mtune=@var{processor-string}
21829 Optimize the output for the given processor, specified by microarchitecture
21832 @item -msmall-data-limit=@var{n}
21833 @opindex msmall-data-limit
21834 Put global and static data smaller than @var{n} bytes into a special section
21837 @item -msave-restore
21838 @itemx -mno-save-restore
21839 @opindex msave-restore
21840 Do or don't use smaller but slower prologue and epilogue code that uses
21841 library function calls. The default is to use fast inline prologues and
21844 @item -mstrict-align
21845 @itemx -mno-strict-align
21846 @opindex mstrict-align
21847 Do not or do generate unaligned memory accesses. The default is set depending
21848 on whether the processor we are optimizing for supports fast unaligned access
21851 @item -mcmodel=medlow
21852 @opindex mcmodel=medlow
21853 Generate code for the medium-low code model. The program and its statically
21854 defined symbols must lie within a single 2 GiB address range and must lie
21855 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
21856 statically or dynamically linked. This is the default code model.
21858 @item -mcmodel=medany
21859 @opindex mcmodel=medany
21860 Generate code for the medium-any code model. The program and its statically
21861 defined symbols must be within any single 2 GiB address range. Programs can be
21862 statically or dynamically linked.
21864 @item -mexplicit-relocs
21865 @itemx -mno-exlicit-relocs
21866 Use or do not use assembler relocation operators when dealing with symbolic
21867 addresses. The alternative is to use assembler macros instead, which may
21868 limit optimization.
21873 @subsection RL78 Options
21874 @cindex RL78 Options
21880 Links in additional target libraries to support operation within a
21889 Specifies the type of hardware multiplication and division support to
21890 be used. The simplest is @code{none}, which uses software for both
21891 multiplication and division. This is the default. The @code{g13}
21892 value is for the hardware multiply/divide peripheral found on the
21893 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
21894 the multiplication and division instructions supported by the RL78/G14
21895 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
21896 the value @code{mg10} is an alias for @code{none}.
21898 In addition a C preprocessor macro is defined, based upon the setting
21899 of this option. Possible values are: @code{__RL78_MUL_NONE__},
21900 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
21907 Specifies the RL78 core to target. The default is the G14 core, also
21908 known as an S3 core or just RL78. The G13 or S2 core does not have
21909 multiply or divide instructions, instead it uses a hardware peripheral
21910 for these operations. The G10 or S1 core does not have register
21911 banks, so it uses a different calling convention.
21913 If this option is set it also selects the type of hardware multiply
21914 support to use, unless this is overridden by an explicit
21915 @option{-mmul=none} option on the command line. Thus specifying
21916 @option{-mcpu=g13} enables the use of the G13 hardware multiply
21917 peripheral and specifying @option{-mcpu=g10} disables the use of
21918 hardware multiplications altogether.
21920 Note, although the RL78/G14 core is the default target, specifying
21921 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
21922 change the behavior of the toolchain since it also enables G14
21923 hardware multiply support. If these options are not specified on the
21924 command line then software multiplication routines will be used even
21925 though the code targets the RL78 core. This is for backwards
21926 compatibility with older toolchains which did not have hardware
21927 multiply and divide support.
21929 In addition a C preprocessor macro is defined, based upon the setting
21930 of this option. Possible values are: @code{__RL78_G10__},
21931 @code{__RL78_G13__} or @code{__RL78_G14__}.
21941 These are aliases for the corresponding @option{-mcpu=} option. They
21942 are provided for backwards compatibility.
21946 Allow the compiler to use all of the available registers. By default
21947 registers @code{r24..r31} are reserved for use in interrupt handlers.
21948 With this option enabled these registers can be used in ordinary
21951 @item -m64bit-doubles
21952 @itemx -m32bit-doubles
21953 @opindex m64bit-doubles
21954 @opindex m32bit-doubles
21955 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21956 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21957 @option{-m32bit-doubles}.
21959 @item -msave-mduc-in-interrupts
21960 @item -mno-save-mduc-in-interrupts
21961 @opindex msave-mduc-in-interrupts
21962 @opindex mno-save-mduc-in-interrupts
21963 Specifies that interrupt handler functions should preserve the
21964 MDUC registers. This is only necessary if normal code might use
21965 the MDUC registers, for example because it performs multiplication
21966 and division operations. The default is to ignore the MDUC registers
21967 as this makes the interrupt handlers faster. The target option -mg13
21968 needs to be passed for this to work as this feature is only available
21969 on the G13 target (S2 core). The MDUC registers will only be saved
21970 if the interrupt handler performs a multiplication or division
21971 operation or it calls another function.
21975 @node RS/6000 and PowerPC Options
21976 @subsection IBM RS/6000 and PowerPC Options
21977 @cindex RS/6000 and PowerPC Options
21978 @cindex IBM RS/6000 and PowerPC Options
21980 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
21982 @item -mpowerpc-gpopt
21983 @itemx -mno-powerpc-gpopt
21984 @itemx -mpowerpc-gfxopt
21985 @itemx -mno-powerpc-gfxopt
21988 @itemx -mno-powerpc64
21992 @itemx -mno-popcntb
21994 @itemx -mno-popcntd
22003 @itemx -mno-hard-dfp
22004 @opindex mpowerpc-gpopt
22005 @opindex mno-powerpc-gpopt
22006 @opindex mpowerpc-gfxopt
22007 @opindex mno-powerpc-gfxopt
22008 @opindex mpowerpc64
22009 @opindex mno-powerpc64
22013 @opindex mno-popcntb
22015 @opindex mno-popcntd
22021 @opindex mno-mfpgpr
22023 @opindex mno-hard-dfp
22024 You use these options to specify which instructions are available on the
22025 processor you are using. The default value of these options is
22026 determined when configuring GCC@. Specifying the
22027 @option{-mcpu=@var{cpu_type}} overrides the specification of these
22028 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
22029 rather than the options listed above.
22031 Specifying @option{-mpowerpc-gpopt} allows
22032 GCC to use the optional PowerPC architecture instructions in the
22033 General Purpose group, including floating-point square root. Specifying
22034 @option{-mpowerpc-gfxopt} allows GCC to
22035 use the optional PowerPC architecture instructions in the Graphics
22036 group, including floating-point select.
22038 The @option{-mmfcrf} option allows GCC to generate the move from
22039 condition register field instruction implemented on the POWER4
22040 processor and other processors that support the PowerPC V2.01
22042 The @option{-mpopcntb} option allows GCC to generate the popcount and
22043 double-precision FP reciprocal estimate instruction implemented on the
22044 POWER5 processor and other processors that support the PowerPC V2.02
22046 The @option{-mpopcntd} option allows GCC to generate the popcount
22047 instruction implemented on the POWER7 processor and other processors
22048 that support the PowerPC V2.06 architecture.
22049 The @option{-mfprnd} option allows GCC to generate the FP round to
22050 integer instructions implemented on the POWER5+ processor and other
22051 processors that support the PowerPC V2.03 architecture.
22052 The @option{-mcmpb} option allows GCC to generate the compare bytes
22053 instruction implemented on the POWER6 processor and other processors
22054 that support the PowerPC V2.05 architecture.
22055 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
22056 general-purpose register instructions implemented on the POWER6X
22057 processor and other processors that support the extended PowerPC V2.05
22059 The @option{-mhard-dfp} option allows GCC to generate the decimal
22060 floating-point instructions implemented on some POWER processors.
22062 The @option{-mpowerpc64} option allows GCC to generate the additional
22063 64-bit instructions that are found in the full PowerPC64 architecture
22064 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
22065 @option{-mno-powerpc64}.
22067 @item -mcpu=@var{cpu_type}
22069 Set architecture type, register usage, and
22070 instruction scheduling parameters for machine type @var{cpu_type}.
22071 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
22072 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
22073 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
22074 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
22075 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
22076 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
22077 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
22078 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
22079 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
22080 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
22081 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
22084 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
22085 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
22086 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
22087 architecture machine types, with an appropriate, generic processor
22088 model assumed for scheduling purposes.
22090 The other options specify a specific processor. Code generated under
22091 those options runs best on that processor, and may not run at all on
22094 The @option{-mcpu} options automatically enable or disable the
22097 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
22098 -mpopcntb -mpopcntd -mpowerpc64 @gol
22099 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
22100 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
22101 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
22102 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
22104 The particular options set for any particular CPU varies between
22105 compiler versions, depending on what setting seems to produce optimal
22106 code for that CPU; it doesn't necessarily reflect the actual hardware's
22107 capabilities. If you wish to set an individual option to a particular
22108 value, you may specify it after the @option{-mcpu} option, like
22109 @option{-mcpu=970 -mno-altivec}.
22111 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
22112 not enabled or disabled by the @option{-mcpu} option at present because
22113 AIX does not have full support for these options. You may still
22114 enable or disable them individually if you're sure it'll work in your
22117 @item -mtune=@var{cpu_type}
22119 Set the instruction scheduling parameters for machine type
22120 @var{cpu_type}, but do not set the architecture type or register usage,
22121 as @option{-mcpu=@var{cpu_type}} does. The same
22122 values for @var{cpu_type} are used for @option{-mtune} as for
22123 @option{-mcpu}. If both are specified, the code generated uses the
22124 architecture and registers set by @option{-mcpu}, but the
22125 scheduling parameters set by @option{-mtune}.
22127 @item -mcmodel=small
22128 @opindex mcmodel=small
22129 Generate PowerPC64 code for the small model: The TOC is limited to
22132 @item -mcmodel=medium
22133 @opindex mcmodel=medium
22134 Generate PowerPC64 code for the medium model: The TOC and other static
22135 data may be up to a total of 4G in size. This is the default for 64-bit
22138 @item -mcmodel=large
22139 @opindex mcmodel=large
22140 Generate PowerPC64 code for the large model: The TOC may be up to 4G
22141 in size. Other data and code is only limited by the 64-bit address
22145 @itemx -mno-altivec
22147 @opindex mno-altivec
22148 Generate code that uses (does not use) AltiVec instructions, and also
22149 enable the use of built-in functions that allow more direct access to
22150 the AltiVec instruction set. You may also need to set
22151 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
22154 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
22155 @option{-maltivec=be}, the element order for AltiVec intrinsics such
22156 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
22157 match array element order corresponding to the endianness of the
22158 target. That is, element zero identifies the leftmost element in a
22159 vector register when targeting a big-endian platform, and identifies
22160 the rightmost element in a vector register when targeting a
22161 little-endian platform.
22164 @opindex maltivec=be
22165 Generate AltiVec instructions using big-endian element order,
22166 regardless of whether the target is big- or little-endian. This is
22167 the default when targeting a big-endian platform.
22169 The element order is used to interpret element numbers in AltiVec
22170 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22171 @code{vec_insert}. By default, these match array element order
22172 corresponding to the endianness for the target.
22175 @opindex maltivec=le
22176 Generate AltiVec instructions using little-endian element order,
22177 regardless of whether the target is big- or little-endian. This is
22178 the default when targeting a little-endian platform. This option is
22179 currently ignored when targeting a big-endian platform.
22181 The element order is used to interpret element numbers in AltiVec
22182 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
22183 @code{vec_insert}. By default, these match array element order
22184 corresponding to the endianness for the target.
22189 @opindex mno-vrsave
22190 Generate VRSAVE instructions when generating AltiVec code.
22193 @opindex msecure-plt
22194 Generate code that allows @command{ld} and @command{ld.so}
22195 to build executables and shared
22196 libraries with non-executable @code{.plt} and @code{.got} sections.
22198 32-bit SYSV ABI option.
22202 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
22204 requires @code{.plt} and @code{.got}
22205 sections that are both writable and executable.
22206 This is a PowerPC 32-bit SYSV ABI option.
22212 This switch enables or disables the generation of ISEL instructions.
22214 @item -misel=@var{yes/no}
22215 This switch has been deprecated. Use @option{-misel} and
22216 @option{-mno-isel} instead.
22222 This switch enables or disables the generation of SPE simd
22228 @opindex mno-paired
22229 This switch enables or disables the generation of PAIRED simd
22232 @item -mspe=@var{yes/no}
22233 This option has been deprecated. Use @option{-mspe} and
22234 @option{-mno-spe} instead.
22240 Generate code that uses (does not use) vector/scalar (VSX)
22241 instructions, and also enable the use of built-in functions that allow
22242 more direct access to the VSX instruction set.
22247 @opindex mno-crypto
22248 Enable the use (disable) of the built-in functions that allow direct
22249 access to the cryptographic instructions that were added in version
22250 2.07 of the PowerPC ISA.
22252 @item -mdirect-move
22253 @itemx -mno-direct-move
22254 @opindex mdirect-move
22255 @opindex mno-direct-move
22256 Generate code that uses (does not use) the instructions to move data
22257 between the general purpose registers and the vector/scalar (VSX)
22258 registers that were added in version 2.07 of the PowerPC ISA.
22264 Enable (disable) the use of the built-in functions that allow direct
22265 access to the Hardware Transactional Memory (HTM) instructions that
22266 were added in version 2.07 of the PowerPC ISA.
22268 @item -mpower8-fusion
22269 @itemx -mno-power8-fusion
22270 @opindex mpower8-fusion
22271 @opindex mno-power8-fusion
22272 Generate code that keeps (does not keeps) some integer operations
22273 adjacent so that the instructions can be fused together on power8 and
22276 @item -mpower8-vector
22277 @itemx -mno-power8-vector
22278 @opindex mpower8-vector
22279 @opindex mno-power8-vector
22280 Generate code that uses (does not use) the vector and scalar
22281 instructions that were added in version 2.07 of the PowerPC ISA. Also
22282 enable the use of built-in functions that allow more direct access to
22283 the vector instructions.
22285 @item -mquad-memory
22286 @itemx -mno-quad-memory
22287 @opindex mquad-memory
22288 @opindex mno-quad-memory
22289 Generate code that uses (does not use) the non-atomic quad word memory
22290 instructions. The @option{-mquad-memory} option requires use of
22293 @item -mquad-memory-atomic
22294 @itemx -mno-quad-memory-atomic
22295 @opindex mquad-memory-atomic
22296 @opindex mno-quad-memory-atomic
22297 Generate code that uses (does not use) the atomic quad word memory
22298 instructions. The @option{-mquad-memory-atomic} option requires use of
22302 @itemx -mno-float128
22304 @opindex mno-float128
22305 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
22306 and use either software emulation for IEEE 128-bit floating point or
22307 hardware instructions.
22309 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
22310 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
22311 use the IEEE 128-bit floating point support. The IEEE 128-bit
22312 floating point support only works on PowerPC Linux systems.
22314 The default for @option{-mfloat128} is enabled on PowerPC Linux
22315 systems using the VSX instruction set, and disabled on other systems.
22317 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
22318 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
22319 point support will also enable the generation of ISA 3.0 IEEE 128-bit
22320 floating point instructions. Otherwise, if you do not specify to
22321 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
22322 system, IEEE 128-bit floating point will be done with software
22325 @item -mfloat128-hardware
22326 @itemx -mno-float128-hardware
22327 @opindex mfloat128-hardware
22328 @opindex mno-float128-hardware
22329 Enable/disable using ISA 3.0 hardware instructions to support the
22330 @var{__float128} data type.
22332 The default for @option{-mfloat128-hardware} is enabled on PowerPC
22333 Linux systems using the ISA 3.0 instruction set, and disabled on other
22336 @item -mfloat-gprs=@var{yes/single/double/no}
22337 @itemx -mfloat-gprs
22338 @opindex mfloat-gprs
22339 This switch enables or disables the generation of floating-point
22340 operations on the general-purpose registers for architectures that
22343 The argument @samp{yes} or @samp{single} enables the use of
22344 single-precision floating-point operations.
22346 The argument @samp{double} enables the use of single and
22347 double-precision floating-point operations.
22349 The argument @samp{no} disables floating-point operations on the
22350 general-purpose registers.
22352 This option is currently only available on the MPC854x.
22358 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
22359 targets (including GNU/Linux). The 32-bit environment sets int, long
22360 and pointer to 32 bits and generates code that runs on any PowerPC
22361 variant. The 64-bit environment sets int to 32 bits and long and
22362 pointer to 64 bits, and generates code for PowerPC64, as for
22363 @option{-mpowerpc64}.
22366 @itemx -mno-fp-in-toc
22367 @itemx -mno-sum-in-toc
22368 @itemx -mminimal-toc
22370 @opindex mno-fp-in-toc
22371 @opindex mno-sum-in-toc
22372 @opindex mminimal-toc
22373 Modify generation of the TOC (Table Of Contents), which is created for
22374 every executable file. The @option{-mfull-toc} option is selected by
22375 default. In that case, GCC allocates at least one TOC entry for
22376 each unique non-automatic variable reference in your program. GCC
22377 also places floating-point constants in the TOC@. However, only
22378 16,384 entries are available in the TOC@.
22380 If you receive a linker error message that saying you have overflowed
22381 the available TOC space, you can reduce the amount of TOC space used
22382 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
22383 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
22384 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
22385 generate code to calculate the sum of an address and a constant at
22386 run time instead of putting that sum into the TOC@. You may specify one
22387 or both of these options. Each causes GCC to produce very slightly
22388 slower and larger code at the expense of conserving TOC space.
22390 If you still run out of space in the TOC even when you specify both of
22391 these options, specify @option{-mminimal-toc} instead. This option causes
22392 GCC to make only one TOC entry for every file. When you specify this
22393 option, GCC produces code that is slower and larger but which
22394 uses extremely little TOC space. You may wish to use this option
22395 only on files that contain less frequently-executed code.
22401 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
22402 @code{long} type, and the infrastructure needed to support them.
22403 Specifying @option{-maix64} implies @option{-mpowerpc64},
22404 while @option{-maix32} disables the 64-bit ABI and
22405 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
22408 @itemx -mno-xl-compat
22409 @opindex mxl-compat
22410 @opindex mno-xl-compat
22411 Produce code that conforms more closely to IBM XL compiler semantics
22412 when using AIX-compatible ABI@. Pass floating-point arguments to
22413 prototyped functions beyond the register save area (RSA) on the stack
22414 in addition to argument FPRs. Do not assume that most significant
22415 double in 128-bit long double value is properly rounded when comparing
22416 values and converting to double. Use XL symbol names for long double
22419 The AIX calling convention was extended but not initially documented to
22420 handle an obscure K&R C case of calling a function that takes the
22421 address of its arguments with fewer arguments than declared. IBM XL
22422 compilers access floating-point arguments that do not fit in the
22423 RSA from the stack when a subroutine is compiled without
22424 optimization. Because always storing floating-point arguments on the
22425 stack is inefficient and rarely needed, this option is not enabled by
22426 default and only is necessary when calling subroutines compiled by IBM
22427 XL compilers without optimization.
22431 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
22432 application written to use message passing with special startup code to
22433 enable the application to run. The system must have PE installed in the
22434 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
22435 must be overridden with the @option{-specs=} option to specify the
22436 appropriate directory location. The Parallel Environment does not
22437 support threads, so the @option{-mpe} option and the @option{-pthread}
22438 option are incompatible.
22440 @item -malign-natural
22441 @itemx -malign-power
22442 @opindex malign-natural
22443 @opindex malign-power
22444 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
22445 @option{-malign-natural} overrides the ABI-defined alignment of larger
22446 types, such as floating-point doubles, on their natural size-based boundary.
22447 The option @option{-malign-power} instructs GCC to follow the ABI-specified
22448 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
22450 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
22454 @itemx -mhard-float
22455 @opindex msoft-float
22456 @opindex mhard-float
22457 Generate code that does not use (uses) the floating-point register set.
22458 Software floating-point emulation is provided if you use the
22459 @option{-msoft-float} option, and pass the option to GCC when linking.
22461 @item -msingle-float
22462 @itemx -mdouble-float
22463 @opindex msingle-float
22464 @opindex mdouble-float
22465 Generate code for single- or double-precision floating-point operations.
22466 @option{-mdouble-float} implies @option{-msingle-float}.
22469 @opindex msimple-fpu
22470 Do not generate @code{sqrt} and @code{div} instructions for hardware
22471 floating-point unit.
22473 @item -mfpu=@var{name}
22475 Specify type of floating-point unit. Valid values for @var{name} are
22476 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
22477 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
22478 @samp{sp_full} (equivalent to @option{-msingle-float}),
22479 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
22482 @opindex mxilinx-fpu
22483 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
22486 @itemx -mno-multiple
22488 @opindex mno-multiple
22489 Generate code that uses (does not use) the load multiple word
22490 instructions and the store multiple word instructions. These
22491 instructions are generated by default on POWER systems, and not
22492 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
22493 PowerPC systems, since those instructions do not work when the
22494 processor is in little-endian mode. The exceptions are PPC740 and
22495 PPC750 which permit these instructions in little-endian mode.
22500 @opindex mno-string
22501 Generate code that uses (does not use) the load string instructions
22502 and the store string word instructions to save multiple registers and
22503 do small block moves. These instructions are generated by default on
22504 POWER systems, and not generated on PowerPC systems. Do not use
22505 @option{-mstring} on little-endian PowerPC systems, since those
22506 instructions do not work when the processor is in little-endian mode.
22507 The exceptions are PPC740 and PPC750 which permit these instructions
22508 in little-endian mode.
22513 @opindex mno-update
22514 Generate code that uses (does not use) the load or store instructions
22515 that update the base register to the address of the calculated memory
22516 location. These instructions are generated by default. If you use
22517 @option{-mno-update}, there is a small window between the time that the
22518 stack pointer is updated and the address of the previous frame is
22519 stored, which means code that walks the stack frame across interrupts or
22520 signals may get corrupted data.
22522 @item -mavoid-indexed-addresses
22523 @itemx -mno-avoid-indexed-addresses
22524 @opindex mavoid-indexed-addresses
22525 @opindex mno-avoid-indexed-addresses
22526 Generate code that tries to avoid (not avoid) the use of indexed load
22527 or store instructions. These instructions can incur a performance
22528 penalty on Power6 processors in certain situations, such as when
22529 stepping through large arrays that cross a 16M boundary. This option
22530 is enabled by default when targeting Power6 and disabled otherwise.
22533 @itemx -mno-fused-madd
22534 @opindex mfused-madd
22535 @opindex mno-fused-madd
22536 Generate code that uses (does not use) the floating-point multiply and
22537 accumulate instructions. These instructions are generated by default
22538 if hardware floating point is used. The machine-dependent
22539 @option{-mfused-madd} option is now mapped to the machine-independent
22540 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22541 mapped to @option{-ffp-contract=off}.
22547 Generate code that uses (does not use) the half-word multiply and
22548 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
22549 These instructions are generated by default when targeting those
22556 Generate code that uses (does not use) the string-search @samp{dlmzb}
22557 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
22558 generated by default when targeting those processors.
22560 @item -mno-bit-align
22562 @opindex mno-bit-align
22563 @opindex mbit-align
22564 On System V.4 and embedded PowerPC systems do not (do) force structures
22565 and unions that contain bit-fields to be aligned to the base type of the
22568 For example, by default a structure containing nothing but 8
22569 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
22570 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
22571 the structure is aligned to a 1-byte boundary and is 1 byte in
22574 @item -mno-strict-align
22575 @itemx -mstrict-align
22576 @opindex mno-strict-align
22577 @opindex mstrict-align
22578 On System V.4 and embedded PowerPC systems do not (do) assume that
22579 unaligned memory references are handled by the system.
22581 @item -mrelocatable
22582 @itemx -mno-relocatable
22583 @opindex mrelocatable
22584 @opindex mno-relocatable
22585 Generate code that allows (does not allow) a static executable to be
22586 relocated to a different address at run time. A simple embedded
22587 PowerPC system loader should relocate the entire contents of
22588 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
22589 a table of 32-bit addresses generated by this option. For this to
22590 work, all objects linked together must be compiled with
22591 @option{-mrelocatable} or @option{-mrelocatable-lib}.
22592 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
22594 @item -mrelocatable-lib
22595 @itemx -mno-relocatable-lib
22596 @opindex mrelocatable-lib
22597 @opindex mno-relocatable-lib
22598 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
22599 @code{.fixup} section to allow static executables to be relocated at
22600 run time, but @option{-mrelocatable-lib} does not use the smaller stack
22601 alignment of @option{-mrelocatable}. Objects compiled with
22602 @option{-mrelocatable-lib} may be linked with objects compiled with
22603 any combination of the @option{-mrelocatable} options.
22609 On System V.4 and embedded PowerPC systems do not (do) assume that
22610 register 2 contains a pointer to a global area pointing to the addresses
22611 used in the program.
22614 @itemx -mlittle-endian
22616 @opindex mlittle-endian
22617 On System V.4 and embedded PowerPC systems compile code for the
22618 processor in little-endian mode. The @option{-mlittle-endian} option is
22619 the same as @option{-mlittle}.
22622 @itemx -mbig-endian
22624 @opindex mbig-endian
22625 On System V.4 and embedded PowerPC systems compile code for the
22626 processor in big-endian mode. The @option{-mbig-endian} option is
22627 the same as @option{-mbig}.
22629 @item -mdynamic-no-pic
22630 @opindex mdynamic-no-pic
22631 On Darwin and Mac OS X systems, compile code so that it is not
22632 relocatable, but that its external references are relocatable. The
22633 resulting code is suitable for applications, but not shared
22636 @item -msingle-pic-base
22637 @opindex msingle-pic-base
22638 Treat the register used for PIC addressing as read-only, rather than
22639 loading it in the prologue for each function. The runtime system is
22640 responsible for initializing this register with an appropriate value
22641 before execution begins.
22643 @item -mprioritize-restricted-insns=@var{priority}
22644 @opindex mprioritize-restricted-insns
22645 This option controls the priority that is assigned to
22646 dispatch-slot restricted instructions during the second scheduling
22647 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
22648 or @samp{2} to assign no, highest, or second-highest (respectively)
22649 priority to dispatch-slot restricted
22652 @item -msched-costly-dep=@var{dependence_type}
22653 @opindex msched-costly-dep
22654 This option controls which dependences are considered costly
22655 by the target during instruction scheduling. The argument
22656 @var{dependence_type} takes one of the following values:
22660 No dependence is costly.
22663 All dependences are costly.
22665 @item @samp{true_store_to_load}
22666 A true dependence from store to load is costly.
22668 @item @samp{store_to_load}
22669 Any dependence from store to load is costly.
22672 Any dependence for which the latency is greater than or equal to
22673 @var{number} is costly.
22676 @item -minsert-sched-nops=@var{scheme}
22677 @opindex minsert-sched-nops
22678 This option controls which NOP insertion scheme is used during
22679 the second scheduling pass. The argument @var{scheme} takes one of the
22687 Pad with NOPs any dispatch group that has vacant issue slots,
22688 according to the scheduler's grouping.
22690 @item @samp{regroup_exact}
22691 Insert NOPs to force costly dependent insns into
22692 separate groups. Insert exactly as many NOPs as needed to force an insn
22693 to a new group, according to the estimated processor grouping.
22696 Insert NOPs to force costly dependent insns into
22697 separate groups. Insert @var{number} NOPs to force an insn to a new group.
22701 @opindex mcall-sysv
22702 On System V.4 and embedded PowerPC systems compile code using calling
22703 conventions that adhere to the March 1995 draft of the System V
22704 Application Binary Interface, PowerPC processor supplement. This is the
22705 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
22707 @item -mcall-sysv-eabi
22709 @opindex mcall-sysv-eabi
22710 @opindex mcall-eabi
22711 Specify both @option{-mcall-sysv} and @option{-meabi} options.
22713 @item -mcall-sysv-noeabi
22714 @opindex mcall-sysv-noeabi
22715 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
22717 @item -mcall-aixdesc
22719 On System V.4 and embedded PowerPC systems compile code for the AIX
22723 @opindex mcall-linux
22724 On System V.4 and embedded PowerPC systems compile code for the
22725 Linux-based GNU system.
22727 @item -mcall-freebsd
22728 @opindex mcall-freebsd
22729 On System V.4 and embedded PowerPC systems compile code for the
22730 FreeBSD operating system.
22732 @item -mcall-netbsd
22733 @opindex mcall-netbsd
22734 On System V.4 and embedded PowerPC systems compile code for the
22735 NetBSD operating system.
22737 @item -mcall-openbsd
22738 @opindex mcall-netbsd
22739 On System V.4 and embedded PowerPC systems compile code for the
22740 OpenBSD operating system.
22742 @item -maix-struct-return
22743 @opindex maix-struct-return
22744 Return all structures in memory (as specified by the AIX ABI)@.
22746 @item -msvr4-struct-return
22747 @opindex msvr4-struct-return
22748 Return structures smaller than 8 bytes in registers (as specified by the
22751 @item -mabi=@var{abi-type}
22753 Extend the current ABI with a particular extension, or remove such extension.
22754 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
22755 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
22756 @samp{elfv1}, @samp{elfv2}@.
22760 Extend the current ABI with SPE ABI extensions. This does not change
22761 the default ABI, instead it adds the SPE ABI extensions to the current
22765 @opindex mabi=no-spe
22766 Disable Book-E SPE ABI extensions for the current ABI@.
22768 @item -mabi=ibmlongdouble
22769 @opindex mabi=ibmlongdouble
22770 Change the current ABI to use IBM extended-precision long double.
22771 This is not likely to work if your system defaults to using IEEE
22772 extended-precision long double. If you change the long double type
22773 from IEEE extended-precision, the compiler will issue a warning unless
22774 you use the @option{-Wno-psabi} option.
22776 @item -mabi=ieeelongdouble
22777 @opindex mabi=ieeelongdouble
22778 Change the current ABI to use IEEE extended-precision long double.
22779 This is not likely to work if your system defaults to using IBM
22780 extended-precision long double. If you change the long double type
22781 from IBM extended-precision, the compiler will issue a warning unless
22782 you use the @option{-Wno-psabi} option.
22785 @opindex mabi=elfv1
22786 Change the current ABI to use the ELFv1 ABI.
22787 This is the default ABI for big-endian PowerPC 64-bit Linux.
22788 Overriding the default ABI requires special system support and is
22789 likely to fail in spectacular ways.
22792 @opindex mabi=elfv2
22793 Change the current ABI to use the ELFv2 ABI.
22794 This is the default ABI for little-endian PowerPC 64-bit Linux.
22795 Overriding the default ABI requires special system support and is
22796 likely to fail in spectacular ways.
22798 @item -mgnu-attribute
22799 @itemx -mno-gnu-attribute
22800 @opindex mgnu-attribute
22801 @opindex mno-gnu-attribute
22802 Emit .gnu_attribute assembly directives to set tag/value pairs in a
22803 .gnu.attributes section that specify ABI variations in function
22804 parameters or return values.
22807 @itemx -mno-prototype
22808 @opindex mprototype
22809 @opindex mno-prototype
22810 On System V.4 and embedded PowerPC systems assume that all calls to
22811 variable argument functions are properly prototyped. Otherwise, the
22812 compiler must insert an instruction before every non-prototyped call to
22813 set or clear bit 6 of the condition code register (@code{CR}) to
22814 indicate whether floating-point values are passed in the floating-point
22815 registers in case the function takes variable arguments. With
22816 @option{-mprototype}, only calls to prototyped variable argument functions
22817 set or clear the bit.
22821 On embedded PowerPC systems, assume that the startup module is called
22822 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
22823 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
22828 On embedded PowerPC systems, assume that the startup module is called
22829 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
22834 On embedded PowerPC systems, assume that the startup module is called
22835 @file{crt0.o} and the standard C libraries are @file{libads.a} and
22838 @item -myellowknife
22839 @opindex myellowknife
22840 On embedded PowerPC systems, assume that the startup module is called
22841 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
22846 On System V.4 and embedded PowerPC systems, specify that you are
22847 compiling for a VxWorks system.
22851 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
22852 header to indicate that @samp{eabi} extended relocations are used.
22858 On System V.4 and embedded PowerPC systems do (do not) adhere to the
22859 Embedded Applications Binary Interface (EABI), which is a set of
22860 modifications to the System V.4 specifications. Selecting @option{-meabi}
22861 means that the stack is aligned to an 8-byte boundary, a function
22862 @code{__eabi} is called from @code{main} to set up the EABI
22863 environment, and the @option{-msdata} option can use both @code{r2} and
22864 @code{r13} to point to two separate small data areas. Selecting
22865 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
22866 no EABI initialization function is called from @code{main}, and the
22867 @option{-msdata} option only uses @code{r13} to point to a single
22868 small data area. The @option{-meabi} option is on by default if you
22869 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
22872 @opindex msdata=eabi
22873 On System V.4 and embedded PowerPC systems, put small initialized
22874 @code{const} global and static data in the @code{.sdata2} section, which
22875 is pointed to by register @code{r2}. Put small initialized
22876 non-@code{const} global and static data in the @code{.sdata} section,
22877 which is pointed to by register @code{r13}. Put small uninitialized
22878 global and static data in the @code{.sbss} section, which is adjacent to
22879 the @code{.sdata} section. The @option{-msdata=eabi} option is
22880 incompatible with the @option{-mrelocatable} option. The
22881 @option{-msdata=eabi} option also sets the @option{-memb} option.
22884 @opindex msdata=sysv
22885 On System V.4 and embedded PowerPC systems, put small global and static
22886 data in the @code{.sdata} section, which is pointed to by register
22887 @code{r13}. Put small uninitialized global and static data in the
22888 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
22889 The @option{-msdata=sysv} option is incompatible with the
22890 @option{-mrelocatable} option.
22892 @item -msdata=default
22894 @opindex msdata=default
22896 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
22897 compile code the same as @option{-msdata=eabi}, otherwise compile code the
22898 same as @option{-msdata=sysv}.
22901 @opindex msdata=data
22902 On System V.4 and embedded PowerPC systems, put small global
22903 data in the @code{.sdata} section. Put small uninitialized global
22904 data in the @code{.sbss} section. Do not use register @code{r13}
22905 to address small data however. This is the default behavior unless
22906 other @option{-msdata} options are used.
22910 @opindex msdata=none
22912 On embedded PowerPC systems, put all initialized global and static data
22913 in the @code{.data} section, and all uninitialized data in the
22914 @code{.bss} section.
22916 @item -mblock-move-inline-limit=@var{num}
22917 @opindex mblock-move-inline-limit
22918 Inline all block moves (such as calls to @code{memcpy} or structure
22919 copies) less than or equal to @var{num} bytes. The minimum value for
22920 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
22921 targets. The default value is target-specific.
22925 @cindex smaller data references (PowerPC)
22926 @cindex .sdata/.sdata2 references (PowerPC)
22927 On embedded PowerPC systems, put global and static items less than or
22928 equal to @var{num} bytes into the small data or BSS sections instead of
22929 the normal data or BSS section. By default, @var{num} is 8. The
22930 @option{-G @var{num}} switch is also passed to the linker.
22931 All modules should be compiled with the same @option{-G @var{num}} value.
22934 @itemx -mno-regnames
22936 @opindex mno-regnames
22937 On System V.4 and embedded PowerPC systems do (do not) emit register
22938 names in the assembly language output using symbolic forms.
22941 @itemx -mno-longcall
22943 @opindex mno-longcall
22944 By default assume that all calls are far away so that a longer and more
22945 expensive calling sequence is required. This is required for calls
22946 farther than 32 megabytes (33,554,432 bytes) from the current location.
22947 A short call is generated if the compiler knows
22948 the call cannot be that far away. This setting can be overridden by
22949 the @code{shortcall} function attribute, or by @code{#pragma
22952 Some linkers are capable of detecting out-of-range calls and generating
22953 glue code on the fly. On these systems, long calls are unnecessary and
22954 generate slower code. As of this writing, the AIX linker can do this,
22955 as can the GNU linker for PowerPC/64. It is planned to add this feature
22956 to the GNU linker for 32-bit PowerPC systems as well.
22958 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
22959 callee, L42}, plus a @dfn{branch island} (glue code). The two target
22960 addresses represent the callee and the branch island. The
22961 Darwin/PPC linker prefers the first address and generates a @code{bl
22962 callee} if the PPC @code{bl} instruction reaches the callee directly;
22963 otherwise, the linker generates @code{bl L42} to call the branch
22964 island. The branch island is appended to the body of the
22965 calling function; it computes the full 32-bit address of the callee
22968 On Mach-O (Darwin) systems, this option directs the compiler emit to
22969 the glue for every direct call, and the Darwin linker decides whether
22970 to use or discard it.
22972 In the future, GCC may ignore all longcall specifications
22973 when the linker is known to generate glue.
22975 @item -mtls-markers
22976 @itemx -mno-tls-markers
22977 @opindex mtls-markers
22978 @opindex mno-tls-markers
22979 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22980 specifying the function argument. The relocation allows the linker to
22981 reliably associate function call with argument setup instructions for
22982 TLS optimization, which in turn allows GCC to better schedule the
22988 This option enables use of the reciprocal estimate and
22989 reciprocal square root estimate instructions with additional
22990 Newton-Raphson steps to increase precision instead of doing a divide or
22991 square root and divide for floating-point arguments. You should use
22992 the @option{-ffast-math} option when using @option{-mrecip} (or at
22993 least @option{-funsafe-math-optimizations},
22994 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22995 @option{-fno-trapping-math}). Note that while the throughput of the
22996 sequence is generally higher than the throughput of the non-reciprocal
22997 instruction, the precision of the sequence can be decreased by up to 2
22998 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
23001 @item -mrecip=@var{opt}
23002 @opindex mrecip=opt
23003 This option controls which reciprocal estimate instructions
23004 may be used. @var{opt} is a comma-separated list of options, which may
23005 be preceded by a @code{!} to invert the option:
23010 Enable all estimate instructions.
23013 Enable the default instructions, equivalent to @option{-mrecip}.
23016 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23019 Enable the reciprocal approximation instructions for both
23020 single and double precision.
23023 Enable the single-precision reciprocal approximation instructions.
23026 Enable the double-precision reciprocal approximation instructions.
23029 Enable the reciprocal square root approximation instructions for both
23030 single and double precision.
23033 Enable the single-precision reciprocal square root approximation instructions.
23036 Enable the double-precision reciprocal square root approximation instructions.
23040 So, for example, @option{-mrecip=all,!rsqrtd} enables
23041 all of the reciprocal estimate instructions, except for the
23042 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
23043 which handle the double-precision reciprocal square root calculations.
23045 @item -mrecip-precision
23046 @itemx -mno-recip-precision
23047 @opindex mrecip-precision
23048 Assume (do not assume) that the reciprocal estimate instructions
23049 provide higher-precision estimates than is mandated by the PowerPC
23050 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
23051 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
23052 The double-precision square root estimate instructions are not generated by
23053 default on low-precision machines, since they do not provide an
23054 estimate that converges after three steps.
23056 @item -mveclibabi=@var{type}
23057 @opindex mveclibabi
23058 Specifies the ABI type to use for vectorizing intrinsics using an
23059 external library. The only type supported at present is @samp{mass},
23060 which specifies to use IBM's Mathematical Acceleration Subsystem
23061 (MASS) libraries for vectorizing intrinsics using external libraries.
23062 GCC currently emits calls to @code{acosd2}, @code{acosf4},
23063 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
23064 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
23065 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
23066 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
23067 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
23068 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
23069 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
23070 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
23071 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
23072 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
23073 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
23074 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
23075 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
23076 for power7. Both @option{-ftree-vectorize} and
23077 @option{-funsafe-math-optimizations} must also be enabled. The MASS
23078 libraries must be specified at link time.
23083 Generate (do not generate) the @code{friz} instruction when the
23084 @option{-funsafe-math-optimizations} option is used to optimize
23085 rounding of floating-point values to 64-bit integer and back to floating
23086 point. The @code{friz} instruction does not return the same value if
23087 the floating-point number is too large to fit in an integer.
23089 @item -mpointers-to-nested-functions
23090 @itemx -mno-pointers-to-nested-functions
23091 @opindex mpointers-to-nested-functions
23092 Generate (do not generate) code to load up the static chain register
23093 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
23094 systems where a function pointer points to a 3-word descriptor giving
23095 the function address, TOC value to be loaded in register @code{r2}, and
23096 static chain value to be loaded in register @code{r11}. The
23097 @option{-mpointers-to-nested-functions} is on by default. You cannot
23098 call through pointers to nested functions or pointers
23099 to functions compiled in other languages that use the static chain if
23100 you use @option{-mno-pointers-to-nested-functions}.
23102 @item -msave-toc-indirect
23103 @itemx -mno-save-toc-indirect
23104 @opindex msave-toc-indirect
23105 Generate (do not generate) code to save the TOC value in the reserved
23106 stack location in the function prologue if the function calls through
23107 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
23108 saved in the prologue, it is saved just before the call through the
23109 pointer. The @option{-mno-save-toc-indirect} option is the default.
23111 @item -mcompat-align-parm
23112 @itemx -mno-compat-align-parm
23113 @opindex mcompat-align-parm
23114 Generate (do not generate) code to pass structure parameters with a
23115 maximum alignment of 64 bits, for compatibility with older versions
23118 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
23119 structure parameter on a 128-bit boundary when that structure contained
23120 a member requiring 128-bit alignment. This is corrected in more
23121 recent versions of GCC. This option may be used to generate code
23122 that is compatible with functions compiled with older versions of
23125 The @option{-mno-compat-align-parm} option is the default.
23127 @item -mstack-protector-guard=@var{guard}
23128 @itemx -mstack-protector-guard-reg=@var{reg}
23129 @itemx -mstack-protector-guard-offset=@var{offset}
23130 @itemx -mstack-protector-guard-symbol=@var{symbol}
23131 @opindex mstack-protector-guard
23132 @opindex mstack-protector-guard-reg
23133 @opindex mstack-protector-guard-offset
23134 @opindex mstack-protector-guard-symbol
23135 Generate stack protection code using canary at @var{guard}. Supported
23136 locations are @samp{global} for global canary or @samp{tls} for per-thread
23137 canary in the TLS block (the default with GNU libc version 2.4 or later).
23139 With the latter choice the options
23140 @option{-mstack-protector-guard-reg=@var{reg}} and
23141 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
23142 which register to use as base register for reading the canary, and from what
23143 offset from that base register. The default for those is as specified in the
23144 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
23145 the offset with a symbol reference to a canary in the TLS block.
23149 @subsection RX Options
23152 These command-line options are defined for RX targets:
23155 @item -m64bit-doubles
23156 @itemx -m32bit-doubles
23157 @opindex m64bit-doubles
23158 @opindex m32bit-doubles
23159 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23160 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23161 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
23162 works on 32-bit values, which is why the default is
23163 @option{-m32bit-doubles}.
23169 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
23170 floating-point hardware. The default is enabled for the RX600
23171 series and disabled for the RX200 series.
23173 Floating-point instructions are only generated for 32-bit floating-point
23174 values, however, so the FPU hardware is not used for doubles if the
23175 @option{-m64bit-doubles} option is used.
23177 @emph{Note} If the @option{-fpu} option is enabled then
23178 @option{-funsafe-math-optimizations} is also enabled automatically.
23179 This is because the RX FPU instructions are themselves unsafe.
23181 @item -mcpu=@var{name}
23183 Selects the type of RX CPU to be targeted. Currently three types are
23184 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
23185 the specific @samp{RX610} CPU. The default is @samp{RX600}.
23187 The only difference between @samp{RX600} and @samp{RX610} is that the
23188 @samp{RX610} does not support the @code{MVTIPL} instruction.
23190 The @samp{RX200} series does not have a hardware floating-point unit
23191 and so @option{-nofpu} is enabled by default when this type is
23194 @item -mbig-endian-data
23195 @itemx -mlittle-endian-data
23196 @opindex mbig-endian-data
23197 @opindex mlittle-endian-data
23198 Store data (but not code) in the big-endian format. The default is
23199 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
23202 @item -msmall-data-limit=@var{N}
23203 @opindex msmall-data-limit
23204 Specifies the maximum size in bytes of global and static variables
23205 which can be placed into the small data area. Using the small data
23206 area can lead to smaller and faster code, but the size of area is
23207 limited and it is up to the programmer to ensure that the area does
23208 not overflow. Also when the small data area is used one of the RX's
23209 registers (usually @code{r13}) is reserved for use pointing to this
23210 area, so it is no longer available for use by the compiler. This
23211 could result in slower and/or larger code if variables are pushed onto
23212 the stack instead of being held in this register.
23214 Note, common variables (variables that have not been initialized) and
23215 constants are not placed into the small data area as they are assigned
23216 to other sections in the output executable.
23218 The default value is zero, which disables this feature. Note, this
23219 feature is not enabled by default with higher optimization levels
23220 (@option{-O2} etc) because of the potentially detrimental effects of
23221 reserving a register. It is up to the programmer to experiment and
23222 discover whether this feature is of benefit to their program. See the
23223 description of the @option{-mpid} option for a description of how the
23224 actual register to hold the small data area pointer is chosen.
23230 Use the simulator runtime. The default is to use the libgloss
23231 board-specific runtime.
23233 @item -mas100-syntax
23234 @itemx -mno-as100-syntax
23235 @opindex mas100-syntax
23236 @opindex mno-as100-syntax
23237 When generating assembler output use a syntax that is compatible with
23238 Renesas's AS100 assembler. This syntax can also be handled by the GAS
23239 assembler, but it has some restrictions so it is not generated by default.
23241 @item -mmax-constant-size=@var{N}
23242 @opindex mmax-constant-size
23243 Specifies the maximum size, in bytes, of a constant that can be used as
23244 an operand in a RX instruction. Although the RX instruction set does
23245 allow constants of up to 4 bytes in length to be used in instructions,
23246 a longer value equates to a longer instruction. Thus in some
23247 circumstances it can be beneficial to restrict the size of constants
23248 that are used in instructions. Constants that are too big are instead
23249 placed into a constant pool and referenced via register indirection.
23251 The value @var{N} can be between 0 and 4. A value of 0 (the default)
23252 or 4 means that constants of any size are allowed.
23256 Enable linker relaxation. Linker relaxation is a process whereby the
23257 linker attempts to reduce the size of a program by finding shorter
23258 versions of various instructions. Disabled by default.
23260 @item -mint-register=@var{N}
23261 @opindex mint-register
23262 Specify the number of registers to reserve for fast interrupt handler
23263 functions. The value @var{N} can be between 0 and 4. A value of 1
23264 means that register @code{r13} is reserved for the exclusive use
23265 of fast interrupt handlers. A value of 2 reserves @code{r13} and
23266 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
23267 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
23268 A value of 0, the default, does not reserve any registers.
23270 @item -msave-acc-in-interrupts
23271 @opindex msave-acc-in-interrupts
23272 Specifies that interrupt handler functions should preserve the
23273 accumulator register. This is only necessary if normal code might use
23274 the accumulator register, for example because it performs 64-bit
23275 multiplications. The default is to ignore the accumulator as this
23276 makes the interrupt handlers faster.
23282 Enables the generation of position independent data. When enabled any
23283 access to constant data is done via an offset from a base address
23284 held in a register. This allows the location of constant data to be
23285 determined at run time without requiring the executable to be
23286 relocated, which is a benefit to embedded applications with tight
23287 memory constraints. Data that can be modified is not affected by this
23290 Note, using this feature reserves a register, usually @code{r13}, for
23291 the constant data base address. This can result in slower and/or
23292 larger code, especially in complicated functions.
23294 The actual register chosen to hold the constant data base address
23295 depends upon whether the @option{-msmall-data-limit} and/or the
23296 @option{-mint-register} command-line options are enabled. Starting
23297 with register @code{r13} and proceeding downwards, registers are
23298 allocated first to satisfy the requirements of @option{-mint-register},
23299 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
23300 is possible for the small data area register to be @code{r8} if both
23301 @option{-mint-register=4} and @option{-mpid} are specified on the
23304 By default this feature is not enabled. The default can be restored
23305 via the @option{-mno-pid} command-line option.
23307 @item -mno-warn-multiple-fast-interrupts
23308 @itemx -mwarn-multiple-fast-interrupts
23309 @opindex mno-warn-multiple-fast-interrupts
23310 @opindex mwarn-multiple-fast-interrupts
23311 Prevents GCC from issuing a warning message if it finds more than one
23312 fast interrupt handler when it is compiling a file. The default is to
23313 issue a warning for each extra fast interrupt handler found, as the RX
23314 only supports one such interrupt.
23316 @item -mallow-string-insns
23317 @itemx -mno-allow-string-insns
23318 @opindex mallow-string-insns
23319 @opindex mno-allow-string-insns
23320 Enables or disables the use of the string manipulation instructions
23321 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
23322 @code{SWHILE} and also the @code{RMPA} instruction. These
23323 instructions may prefetch data, which is not safe to do if accessing
23324 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
23325 for more information).
23327 The default is to allow these instructions, but it is not possible for
23328 GCC to reliably detect all circumstances where a string instruction
23329 might be used to access an I/O register, so their use cannot be
23330 disabled automatically. Instead it is reliant upon the programmer to
23331 use the @option{-mno-allow-string-insns} option if their program
23332 accesses I/O space.
23334 When the instructions are enabled GCC defines the C preprocessor
23335 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
23336 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
23342 Use only (or not only) @code{JSR} instructions to access functions.
23343 This option can be used when code size exceeds the range of @code{BSR}
23344 instructions. Note that @option{-mno-jsr} does not mean to not use
23345 @code{JSR} but instead means that any type of branch may be used.
23348 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
23349 has special significance to the RX port when used with the
23350 @code{interrupt} function attribute. This attribute indicates a
23351 function intended to process fast interrupts. GCC ensures
23352 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
23353 and/or @code{r13} and only provided that the normal use of the
23354 corresponding registers have been restricted via the
23355 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
23358 @node S/390 and zSeries Options
23359 @subsection S/390 and zSeries Options
23360 @cindex S/390 and zSeries Options
23362 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
23366 @itemx -msoft-float
23367 @opindex mhard-float
23368 @opindex msoft-float
23369 Use (do not use) the hardware floating-point instructions and registers
23370 for floating-point operations. When @option{-msoft-float} is specified,
23371 functions in @file{libgcc.a} are used to perform floating-point
23372 operations. When @option{-mhard-float} is specified, the compiler
23373 generates IEEE floating-point instructions. This is the default.
23376 @itemx -mno-hard-dfp
23378 @opindex mno-hard-dfp
23379 Use (do not use) the hardware decimal-floating-point instructions for
23380 decimal-floating-point operations. When @option{-mno-hard-dfp} is
23381 specified, functions in @file{libgcc.a} are used to perform
23382 decimal-floating-point operations. When @option{-mhard-dfp} is
23383 specified, the compiler generates decimal-floating-point hardware
23384 instructions. This is the default for @option{-march=z9-ec} or higher.
23386 @item -mlong-double-64
23387 @itemx -mlong-double-128
23388 @opindex mlong-double-64
23389 @opindex mlong-double-128
23390 These switches control the size of @code{long double} type. A size
23391 of 64 bits makes the @code{long double} type equivalent to the @code{double}
23392 type. This is the default.
23395 @itemx -mno-backchain
23396 @opindex mbackchain
23397 @opindex mno-backchain
23398 Store (do not store) the address of the caller's frame as backchain pointer
23399 into the callee's stack frame.
23400 A backchain may be needed to allow debugging using tools that do not understand
23401 DWARF call frame information.
23402 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
23403 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
23404 the backchain is placed into the topmost word of the 96/160 byte register
23407 In general, code compiled with @option{-mbackchain} is call-compatible with
23408 code compiled with @option{-mmo-backchain}; however, use of the backchain
23409 for debugging purposes usually requires that the whole binary is built with
23410 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
23411 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23412 to build a linux kernel use @option{-msoft-float}.
23414 The default is to not maintain the backchain.
23416 @item -mpacked-stack
23417 @itemx -mno-packed-stack
23418 @opindex mpacked-stack
23419 @opindex mno-packed-stack
23420 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
23421 specified, the compiler uses the all fields of the 96/160 byte register save
23422 area only for their default purpose; unused fields still take up stack space.
23423 When @option{-mpacked-stack} is specified, register save slots are densely
23424 packed at the top of the register save area; unused space is reused for other
23425 purposes, allowing for more efficient use of the available stack space.
23426 However, when @option{-mbackchain} is also in effect, the topmost word of
23427 the save area is always used to store the backchain, and the return address
23428 register is always saved two words below the backchain.
23430 As long as the stack frame backchain is not used, code generated with
23431 @option{-mpacked-stack} is call-compatible with code generated with
23432 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
23433 S/390 or zSeries generated code that uses the stack frame backchain at run
23434 time, not just for debugging purposes. Such code is not call-compatible
23435 with code compiled with @option{-mpacked-stack}. Also, note that the
23436 combination of @option{-mbackchain},
23437 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
23438 to build a linux kernel use @option{-msoft-float}.
23440 The default is to not use the packed stack layout.
23443 @itemx -mno-small-exec
23444 @opindex msmall-exec
23445 @opindex mno-small-exec
23446 Generate (or do not generate) code using the @code{bras} instruction
23447 to do subroutine calls.
23448 This only works reliably if the total executable size does not
23449 exceed 64k. The default is to use the @code{basr} instruction instead,
23450 which does not have this limitation.
23456 When @option{-m31} is specified, generate code compliant to the
23457 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
23458 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
23459 particular to generate 64-bit instructions. For the @samp{s390}
23460 targets, the default is @option{-m31}, while the @samp{s390x}
23461 targets default to @option{-m64}.
23467 When @option{-mzarch} is specified, generate code using the
23468 instructions available on z/Architecture.
23469 When @option{-mesa} is specified, generate code using the
23470 instructions available on ESA/390. Note that @option{-mesa} is
23471 not possible with @option{-m64}.
23472 When generating code compliant to the GNU/Linux for S/390 ABI,
23473 the default is @option{-mesa}. When generating code compliant
23474 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
23480 The @option{-mhtm} option enables a set of builtins making use of
23481 instructions available with the transactional execution facility
23482 introduced with the IBM zEnterprise EC12 machine generation
23483 @ref{S/390 System z Built-in Functions}.
23484 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
23490 When @option{-mvx} is specified, generate code using the instructions
23491 available with the vector extension facility introduced with the IBM
23492 z13 machine generation.
23493 This option changes the ABI for some vector type values with regard to
23494 alignment and calling conventions. In case vector type values are
23495 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
23496 command will be added to mark the resulting binary with the ABI used.
23497 @option{-mvx} is enabled by default when using @option{-march=z13}.
23500 @itemx -mno-zvector
23502 @opindex mno-zvector
23503 The @option{-mzvector} option enables vector language extensions and
23504 builtins using instructions available with the vector extension
23505 facility introduced with the IBM z13 machine generation.
23506 This option adds support for @samp{vector} to be used as a keyword to
23507 define vector type variables and arguments. @samp{vector} is only
23508 available when GNU extensions are enabled. It will not be expanded
23509 when requesting strict standard compliance e.g. with @option{-std=c99}.
23510 In addition to the GCC low-level builtins @option{-mzvector} enables
23511 a set of builtins added for compatibility with AltiVec-style
23512 implementations like Power and Cell. In order to make use of these
23513 builtins the header file @file{vecintrin.h} needs to be included.
23514 @option{-mzvector} is disabled by default.
23520 Generate (or do not generate) code using the @code{mvcle} instruction
23521 to perform block moves. When @option{-mno-mvcle} is specified,
23522 use a @code{mvc} loop instead. This is the default unless optimizing for
23529 Print (or do not print) additional debug information when compiling.
23530 The default is to not print debug information.
23532 @item -march=@var{cpu-type}
23534 Generate code that runs on @var{cpu-type}, which is the name of a
23535 system representing a certain processor type. Possible values for
23536 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
23537 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
23538 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
23541 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
23542 @samp{g6} are deprecated and will be removed with future releases.
23544 Specifying @samp{native} as cpu type can be used to select the best
23545 architecture option for the host processor.
23546 @option{-march=native} has no effect if GCC does not recognize the
23549 @item -mtune=@var{cpu-type}
23551 Tune to @var{cpu-type} everything applicable about the generated code,
23552 except for the ABI and the set of available instructions.
23553 The list of @var{cpu-type} values is the same as for @option{-march}.
23554 The default is the value used for @option{-march}.
23557 @itemx -mno-tpf-trace
23558 @opindex mtpf-trace
23559 @opindex mno-tpf-trace
23560 Generate code that adds (does not add) in TPF OS specific branches to trace
23561 routines in the operating system. This option is off by default, even
23562 when compiling for the TPF OS@.
23565 @itemx -mno-fused-madd
23566 @opindex mfused-madd
23567 @opindex mno-fused-madd
23568 Generate code that uses (does not use) the floating-point multiply and
23569 accumulate instructions. These instructions are generated by default if
23570 hardware floating point is used.
23572 @item -mwarn-framesize=@var{framesize}
23573 @opindex mwarn-framesize
23574 Emit a warning if the current function exceeds the given frame size. Because
23575 this is a compile-time check it doesn't need to be a real problem when the program
23576 runs. It is intended to identify functions that most probably cause
23577 a stack overflow. It is useful to be used in an environment with limited stack
23578 size e.g.@: the linux kernel.
23580 @item -mwarn-dynamicstack
23581 @opindex mwarn-dynamicstack
23582 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
23583 arrays. This is generally a bad idea with a limited stack size.
23585 @item -mstack-guard=@var{stack-guard}
23586 @itemx -mstack-size=@var{stack-size}
23587 @opindex mstack-guard
23588 @opindex mstack-size
23589 If these options are provided the S/390 back end emits additional instructions in
23590 the function prologue that trigger a trap if the stack size is @var{stack-guard}
23591 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
23592 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
23593 the frame size of the compiled function is chosen.
23594 These options are intended to be used to help debugging stack overflow problems.
23595 The additionally emitted code causes only little overhead and hence can also be
23596 used in production-like systems without greater performance degradation. The given
23597 values have to be exact powers of 2 and @var{stack-size} has to be greater than
23598 @var{stack-guard} without exceeding 64k.
23599 In order to be efficient the extra code makes the assumption that the stack starts
23600 at an address aligned to the value given by @var{stack-size}.
23601 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
23603 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
23605 If the hotpatch option is enabled, a ``hot-patching'' function
23606 prologue is generated for all functions in the compilation unit.
23607 The funtion label is prepended with the given number of two-byte
23608 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
23609 the label, 2 * @var{post-halfwords} bytes are appended, using the
23610 largest NOP like instructions the architecture allows (maximum
23613 If both arguments are zero, hotpatching is disabled.
23615 This option can be overridden for individual functions with the
23616 @code{hotpatch} attribute.
23619 @node Score Options
23620 @subsection Score Options
23621 @cindex Score Options
23623 These options are defined for Score implementations:
23628 Compile code for big-endian mode. This is the default.
23632 Compile code for little-endian mode.
23636 Disable generation of @code{bcnz} instructions.
23640 Enable generation of unaligned load and store instructions.
23644 Enable the use of multiply-accumulate instructions. Disabled by default.
23648 Specify the SCORE5 as the target architecture.
23652 Specify the SCORE5U of the target architecture.
23656 Specify the SCORE7 as the target architecture. This is the default.
23660 Specify the SCORE7D as the target architecture.
23664 @subsection SH Options
23666 These @samp{-m} options are defined for the SH implementations:
23671 Generate code for the SH1.
23675 Generate code for the SH2.
23678 Generate code for the SH2e.
23682 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
23683 that the floating-point unit is not used.
23685 @item -m2a-single-only
23686 @opindex m2a-single-only
23687 Generate code for the SH2a-FPU, in such a way that no double-precision
23688 floating-point operations are used.
23691 @opindex m2a-single
23692 Generate code for the SH2a-FPU assuming the floating-point unit is in
23693 single-precision mode by default.
23697 Generate code for the SH2a-FPU assuming the floating-point unit is in
23698 double-precision mode by default.
23702 Generate code for the SH3.
23706 Generate code for the SH3e.
23710 Generate code for the SH4 without a floating-point unit.
23712 @item -m4-single-only
23713 @opindex m4-single-only
23714 Generate code for the SH4 with a floating-point unit that only
23715 supports single-precision arithmetic.
23719 Generate code for the SH4 assuming the floating-point unit is in
23720 single-precision mode by default.
23724 Generate code for the SH4.
23728 Generate code for SH4-100.
23730 @item -m4-100-nofpu
23731 @opindex m4-100-nofpu
23732 Generate code for SH4-100 in such a way that the
23733 floating-point unit is not used.
23735 @item -m4-100-single
23736 @opindex m4-100-single
23737 Generate code for SH4-100 assuming the floating-point unit is in
23738 single-precision mode by default.
23740 @item -m4-100-single-only
23741 @opindex m4-100-single-only
23742 Generate code for SH4-100 in such a way that no double-precision
23743 floating-point operations are used.
23747 Generate code for SH4-200.
23749 @item -m4-200-nofpu
23750 @opindex m4-200-nofpu
23751 Generate code for SH4-200 without in such a way that the
23752 floating-point unit is not used.
23754 @item -m4-200-single
23755 @opindex m4-200-single
23756 Generate code for SH4-200 assuming the floating-point unit is in
23757 single-precision mode by default.
23759 @item -m4-200-single-only
23760 @opindex m4-200-single-only
23761 Generate code for SH4-200 in such a way that no double-precision
23762 floating-point operations are used.
23766 Generate code for SH4-300.
23768 @item -m4-300-nofpu
23769 @opindex m4-300-nofpu
23770 Generate code for SH4-300 without in such a way that the
23771 floating-point unit is not used.
23773 @item -m4-300-single
23774 @opindex m4-300-single
23775 Generate code for SH4-300 in such a way that no double-precision
23776 floating-point operations are used.
23778 @item -m4-300-single-only
23779 @opindex m4-300-single-only
23780 Generate code for SH4-300 in such a way that no double-precision
23781 floating-point operations are used.
23785 Generate code for SH4-340 (no MMU, no FPU).
23789 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
23794 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
23795 floating-point unit is not used.
23797 @item -m4a-single-only
23798 @opindex m4a-single-only
23799 Generate code for the SH4a, in such a way that no double-precision
23800 floating-point operations are used.
23803 @opindex m4a-single
23804 Generate code for the SH4a assuming the floating-point unit is in
23805 single-precision mode by default.
23809 Generate code for the SH4a.
23813 Same as @option{-m4a-nofpu}, except that it implicitly passes
23814 @option{-dsp} to the assembler. GCC doesn't generate any DSP
23815 instructions at the moment.
23819 Compile code for the processor in big-endian mode.
23823 Compile code for the processor in little-endian mode.
23827 Align doubles at 64-bit boundaries. Note that this changes the calling
23828 conventions, and thus some functions from the standard C library do
23829 not work unless you recompile it first with @option{-mdalign}.
23833 Shorten some address references at link time, when possible; uses the
23834 linker option @option{-relax}.
23838 Use 32-bit offsets in @code{switch} tables. The default is to use
23843 Enable the use of bit manipulation instructions on SH2A.
23847 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
23848 alignment constraints.
23852 Comply with the calling conventions defined by Renesas.
23855 @opindex mno-renesas
23856 Comply with the calling conventions defined for GCC before the Renesas
23857 conventions were available. This option is the default for all
23858 targets of the SH toolchain.
23861 @opindex mnomacsave
23862 Mark the @code{MAC} register as call-clobbered, even if
23863 @option{-mrenesas} is given.
23869 Control the IEEE compliance of floating-point comparisons, which affects the
23870 handling of cases where the result of a comparison is unordered. By default
23871 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
23872 enabled @option{-mno-ieee} is implicitly set, which results in faster
23873 floating-point greater-equal and less-equal comparisons. The implicit settings
23874 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
23876 @item -minline-ic_invalidate
23877 @opindex minline-ic_invalidate
23878 Inline code to invalidate instruction cache entries after setting up
23879 nested function trampolines.
23880 This option has no effect if @option{-musermode} is in effect and the selected
23881 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
23883 If the selected code generation option does not allow the use of the @code{icbi}
23884 instruction, and @option{-musermode} is not in effect, the inlined code
23885 manipulates the instruction cache address array directly with an associative
23886 write. This not only requires privileged mode at run time, but it also
23887 fails if the cache line had been mapped via the TLB and has become unmapped.
23891 Dump instruction size and location in the assembly code.
23894 @opindex mpadstruct
23895 This option is deprecated. It pads structures to multiple of 4 bytes,
23896 which is incompatible with the SH ABI@.
23898 @item -matomic-model=@var{model}
23899 @opindex matomic-model=@var{model}
23900 Sets the model of atomic operations and additional parameters as a comma
23901 separated list. For details on the atomic built-in functions see
23902 @ref{__atomic Builtins}. The following models and parameters are supported:
23907 Disable compiler generated atomic sequences and emit library calls for atomic
23908 operations. This is the default if the target is not @code{sh*-*-linux*}.
23911 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
23912 built-in functions. The generated atomic sequences require additional support
23913 from the interrupt/exception handling code of the system and are only suitable
23914 for SH3* and SH4* single-core systems. This option is enabled by default when
23915 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
23916 this option also partially utilizes the hardware atomic instructions
23917 @code{movli.l} and @code{movco.l} to create more efficient code, unless
23918 @samp{strict} is specified.
23921 Generate software atomic sequences that use a variable in the thread control
23922 block. This is a variation of the gUSA sequences which can also be used on
23923 SH1* and SH2* targets. The generated atomic sequences require additional
23924 support from the interrupt/exception handling code of the system and are only
23925 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
23926 parameter has to be specified as well.
23929 Generate software atomic sequences that temporarily disable interrupts by
23930 setting @code{SR.IMASK = 1111}. This model works only when the program runs
23931 in privileged mode and is only suitable for single-core systems. Additional
23932 support from the interrupt/exception handling code of the system is not
23933 required. This model is enabled by default when the target is
23934 @code{sh*-*-linux*} and SH1* or SH2*.
23937 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
23938 instructions only. This is only available on SH4A and is suitable for
23939 multi-core systems. Since the hardware instructions support only 32 bit atomic
23940 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
23941 Code compiled with this option is also compatible with other software
23942 atomic model interrupt/exception handling systems if executed on an SH4A
23943 system. Additional support from the interrupt/exception handling code of the
23944 system is not required for this model.
23947 This parameter specifies the offset in bytes of the variable in the thread
23948 control block structure that should be used by the generated atomic sequences
23949 when the @samp{soft-tcb} model has been selected. For other models this
23950 parameter is ignored. The specified value must be an integer multiple of four
23951 and in the range 0-1020.
23954 This parameter prevents mixed usage of multiple atomic models, even if they
23955 are compatible, and makes the compiler generate atomic sequences of the
23956 specified model only.
23962 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
23963 Notice that depending on the particular hardware and software configuration
23964 this can degrade overall performance due to the operand cache line flushes
23965 that are implied by the @code{tas.b} instruction. On multi-core SH4A
23966 processors the @code{tas.b} instruction must be used with caution since it
23967 can result in data corruption for certain cache configurations.
23970 @opindex mprefergot
23971 When generating position-independent code, emit function calls using
23972 the Global Offset Table instead of the Procedure Linkage Table.
23975 @itemx -mno-usermode
23977 @opindex mno-usermode
23978 Don't allow (allow) the compiler generating privileged mode code. Specifying
23979 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
23980 inlined code would not work in user mode. @option{-musermode} is the default
23981 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
23982 @option{-musermode} has no effect, since there is no user mode.
23984 @item -multcost=@var{number}
23985 @opindex multcost=@var{number}
23986 Set the cost to assume for a multiply insn.
23988 @item -mdiv=@var{strategy}
23989 @opindex mdiv=@var{strategy}
23990 Set the division strategy to be used for integer division operations.
23991 @var{strategy} can be one of:
23996 Calls a library function that uses the single-step division instruction
23997 @code{div1} to perform the operation. Division by zero calculates an
23998 unspecified result and does not trap. This is the default except for SH4,
23999 SH2A and SHcompact.
24002 Calls a library function that performs the operation in double precision
24003 floating point. Division by zero causes a floating-point exception. This is
24004 the default for SHcompact with FPU. Specifying this for targets that do not
24005 have a double precision FPU defaults to @code{call-div1}.
24008 Calls a library function that uses a lookup table for small divisors and
24009 the @code{div1} instruction with case distinction for larger divisors. Division
24010 by zero calculates an unspecified result and does not trap. This is the default
24011 for SH4. Specifying this for targets that do not have dynamic shift
24012 instructions defaults to @code{call-div1}.
24016 When a division strategy has not been specified the default strategy is
24017 selected based on the current target. For SH2A the default strategy is to
24018 use the @code{divs} and @code{divu} instructions instead of library function
24021 @item -maccumulate-outgoing-args
24022 @opindex maccumulate-outgoing-args
24023 Reserve space once for outgoing arguments in the function prologue rather
24024 than around each call. Generally beneficial for performance and size. Also
24025 needed for unwinding to avoid changing the stack frame around conditional code.
24027 @item -mdivsi3_libfunc=@var{name}
24028 @opindex mdivsi3_libfunc=@var{name}
24029 Set the name of the library function used for 32-bit signed division to
24031 This only affects the name used in the @samp{call} division strategies, and
24032 the compiler still expects the same sets of input/output/clobbered registers as
24033 if this option were not present.
24035 @item -mfixed-range=@var{register-range}
24036 @opindex mfixed-range
24037 Generate code treating the given register range as fixed registers.
24038 A fixed register is one that the register allocator can not use. This is
24039 useful when compiling kernel code. A register range is specified as
24040 two registers separated by a dash. Multiple register ranges can be
24041 specified separated by a comma.
24043 @item -mbranch-cost=@var{num}
24044 @opindex mbranch-cost=@var{num}
24045 Assume @var{num} to be the cost for a branch instruction. Higher numbers
24046 make the compiler try to generate more branch-free code if possible.
24047 If not specified the value is selected depending on the processor type that
24048 is being compiled for.
24051 @itemx -mno-zdcbranch
24052 @opindex mzdcbranch
24053 @opindex mno-zdcbranch
24054 Assume (do not assume) that zero displacement conditional branch instructions
24055 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
24056 compiler prefers zero displacement branch code sequences. This is
24057 enabled by default when generating code for SH4 and SH4A. It can be explicitly
24058 disabled by specifying @option{-mno-zdcbranch}.
24060 @item -mcbranch-force-delay-slot
24061 @opindex mcbranch-force-delay-slot
24062 Force the usage of delay slots for conditional branches, which stuffs the delay
24063 slot with a @code{nop} if a suitable instruction cannot be found. By default
24064 this option is disabled. It can be enabled to work around hardware bugs as
24065 found in the original SH7055.
24068 @itemx -mno-fused-madd
24069 @opindex mfused-madd
24070 @opindex mno-fused-madd
24071 Generate code that uses (does not use) the floating-point multiply and
24072 accumulate instructions. These instructions are generated by default
24073 if hardware floating point is used. The machine-dependent
24074 @option{-mfused-madd} option is now mapped to the machine-independent
24075 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
24076 mapped to @option{-ffp-contract=off}.
24082 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
24083 and cosine approximations. The option @option{-mfsca} must be used in
24084 combination with @option{-funsafe-math-optimizations}. It is enabled by default
24085 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
24086 approximations even if @option{-funsafe-math-optimizations} is in effect.
24092 Allow or disallow the compiler to emit the @code{fsrra} instruction for
24093 reciprocal square root approximations. The option @option{-mfsrra} must be used
24094 in combination with @option{-funsafe-math-optimizations} and
24095 @option{-ffinite-math-only}. It is enabled by default when generating code for
24096 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
24097 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
24100 @item -mpretend-cmove
24101 @opindex mpretend-cmove
24102 Prefer zero-displacement conditional branches for conditional move instruction
24103 patterns. This can result in faster code on the SH4 processor.
24107 Generate code using the FDPIC ABI.
24111 @node Solaris 2 Options
24112 @subsection Solaris 2 Options
24113 @cindex Solaris 2 options
24115 These @samp{-m} options are supported on Solaris 2:
24118 @item -mclear-hwcap
24119 @opindex mclear-hwcap
24120 @option{-mclear-hwcap} tells the compiler to remove the hardware
24121 capabilities generated by the Solaris assembler. This is only necessary
24122 when object files use ISA extensions not supported by the current
24123 machine, but check at runtime whether or not to use them.
24125 @item -mimpure-text
24126 @opindex mimpure-text
24127 @option{-mimpure-text}, used in addition to @option{-shared}, tells
24128 the compiler to not pass @option{-z text} to the linker when linking a
24129 shared object. Using this option, you can link position-dependent
24130 code into a shared object.
24132 @option{-mimpure-text} suppresses the ``relocations remain against
24133 allocatable but non-writable sections'' linker error message.
24134 However, the necessary relocations trigger copy-on-write, and the
24135 shared object is not actually shared across processes. Instead of
24136 using @option{-mimpure-text}, you should compile all source code with
24137 @option{-fpic} or @option{-fPIC}.
24141 These switches are supported in addition to the above on Solaris 2:
24146 This is a synonym for @option{-pthread}.
24149 @node SPARC Options
24150 @subsection SPARC Options
24151 @cindex SPARC options
24153 These @samp{-m} options are supported on the SPARC:
24156 @item -mno-app-regs
24158 @opindex mno-app-regs
24160 Specify @option{-mapp-regs} to generate output using the global registers
24161 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
24162 global register 1, each global register 2 through 4 is then treated as an
24163 allocable register that is clobbered by function calls. This is the default.
24165 To be fully SVR4 ABI-compliant at the cost of some performance loss,
24166 specify @option{-mno-app-regs}. You should compile libraries and system
24167 software with this option.
24173 With @option{-mflat}, the compiler does not generate save/restore instructions
24174 and uses a ``flat'' or single register window model. This model is compatible
24175 with the regular register window model. The local registers and the input
24176 registers (0--5) are still treated as ``call-saved'' registers and are
24177 saved on the stack as needed.
24179 With @option{-mno-flat} (the default), the compiler generates save/restore
24180 instructions (except for leaf functions). This is the normal operating mode.
24183 @itemx -mhard-float
24185 @opindex mhard-float
24186 Generate output containing floating-point instructions. This is the
24190 @itemx -msoft-float
24192 @opindex msoft-float
24193 Generate output containing library calls for floating point.
24194 @strong{Warning:} the requisite libraries are not available for all SPARC
24195 targets. Normally the facilities of the machine's usual C compiler are
24196 used, but this cannot be done directly in cross-compilation. You must make
24197 your own arrangements to provide suitable library functions for
24198 cross-compilation. The embedded targets @samp{sparc-*-aout} and
24199 @samp{sparclite-*-*} do provide software floating-point support.
24201 @option{-msoft-float} changes the calling convention in the output file;
24202 therefore, it is only useful if you compile @emph{all} of a program with
24203 this option. In particular, you need to compile @file{libgcc.a}, the
24204 library that comes with GCC, with @option{-msoft-float} in order for
24207 @item -mhard-quad-float
24208 @opindex mhard-quad-float
24209 Generate output containing quad-word (long double) floating-point
24212 @item -msoft-quad-float
24213 @opindex msoft-quad-float
24214 Generate output containing library calls for quad-word (long double)
24215 floating-point instructions. The functions called are those specified
24216 in the SPARC ABI@. This is the default.
24218 As of this writing, there are no SPARC implementations that have hardware
24219 support for the quad-word floating-point instructions. They all invoke
24220 a trap handler for one of these instructions, and then the trap handler
24221 emulates the effect of the instruction. Because of the trap handler overhead,
24222 this is much slower than calling the ABI library routines. Thus the
24223 @option{-msoft-quad-float} option is the default.
24225 @item -mno-unaligned-doubles
24226 @itemx -munaligned-doubles
24227 @opindex mno-unaligned-doubles
24228 @opindex munaligned-doubles
24229 Assume that doubles have 8-byte alignment. This is the default.
24231 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
24232 alignment only if they are contained in another type, or if they have an
24233 absolute address. Otherwise, it assumes they have 4-byte alignment.
24234 Specifying this option avoids some rare compatibility problems with code
24235 generated by other compilers. It is not the default because it results
24236 in a performance loss, especially for floating-point code.
24239 @itemx -mno-user-mode
24240 @opindex muser-mode
24241 @opindex mno-user-mode
24242 Do not generate code that can only run in supervisor mode. This is relevant
24243 only for the @code{casa} instruction emitted for the LEON3 processor. This
24246 @item -mfaster-structs
24247 @itemx -mno-faster-structs
24248 @opindex mfaster-structs
24249 @opindex mno-faster-structs
24250 With @option{-mfaster-structs}, the compiler assumes that structures
24251 should have 8-byte alignment. This enables the use of pairs of
24252 @code{ldd} and @code{std} instructions for copies in structure
24253 assignment, in place of twice as many @code{ld} and @code{st} pairs.
24254 However, the use of this changed alignment directly violates the SPARC
24255 ABI@. Thus, it's intended only for use on targets where the developer
24256 acknowledges that their resulting code is not directly in line with
24257 the rules of the ABI@.
24259 @item -mstd-struct-return
24260 @itemx -mno-std-struct-return
24261 @opindex mstd-struct-return
24262 @opindex mno-std-struct-return
24263 With @option{-mstd-struct-return}, the compiler generates checking code
24264 in functions returning structures or unions to detect size mismatches
24265 between the two sides of function calls, as per the 32-bit ABI@.
24267 The default is @option{-mno-std-struct-return}. This option has no effect
24274 Enable Local Register Allocation. This is the default for SPARC since GCC 7
24275 so @option{-mno-lra} needs to be passed to get old Reload.
24277 @item -mcpu=@var{cpu_type}
24279 Set the instruction set, register set, and instruction scheduling parameters
24280 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24281 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
24282 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
24283 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
24284 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
24285 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
24287 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
24288 which selects the best architecture option for the host processor.
24289 @option{-mcpu=native} has no effect if GCC does not recognize
24292 Default instruction scheduling parameters are used for values that select
24293 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
24294 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
24296 Here is a list of each supported architecture and their supported
24304 supersparc, hypersparc, leon, leon3
24307 f930, f934, sparclite86x
24313 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
24317 By default (unless configured otherwise), GCC generates code for the V7
24318 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
24319 additionally optimizes it for the Cypress CY7C602 chip, as used in the
24320 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
24321 SPARCStation 1, 2, IPX etc.
24323 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
24324 architecture. The only difference from V7 code is that the compiler emits
24325 the integer multiply and integer divide instructions which exist in SPARC-V8
24326 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
24327 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
24330 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
24331 the SPARC architecture. This adds the integer multiply, integer divide step
24332 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
24333 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
24334 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
24335 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
24336 MB86934 chip, which is the more recent SPARClite with FPU@.
24338 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
24339 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
24340 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
24341 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
24342 optimizes it for the TEMIC SPARClet chip.
24344 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
24345 architecture. This adds 64-bit integer and floating-point move instructions,
24346 3 additional floating-point condition code registers and conditional move
24347 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
24348 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
24349 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
24350 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
24351 @option{-mcpu=niagara}, the compiler additionally optimizes it for
24352 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
24353 additionally optimizes it for Sun UltraSPARC T2 chips. With
24354 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
24355 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
24356 additionally optimizes it for Sun UltraSPARC T4 chips. With
24357 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
24358 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
24359 additionally optimizes it for Oracle M8 chips.
24361 @item -mtune=@var{cpu_type}
24363 Set the instruction scheduling parameters for machine type
24364 @var{cpu_type}, but do not set the instruction set or register set that the
24365 option @option{-mcpu=@var{cpu_type}} does.
24367 The same values for @option{-mcpu=@var{cpu_type}} can be used for
24368 @option{-mtune=@var{cpu_type}}, but the only useful values are those
24369 that select a particular CPU implementation. Those are
24370 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
24371 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
24372 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
24373 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
24374 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
24375 and GNU/Linux toolchains, @samp{native} can also be used.
24380 @opindex mno-v8plus
24381 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
24382 difference from the V8 ABI is that the global and out registers are
24383 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
24384 mode for all SPARC-V9 processors.
24390 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
24391 Visual Instruction Set extensions. The default is @option{-mno-vis}.
24397 With @option{-mvis2}, GCC generates code that takes advantage of
24398 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
24399 default is @option{-mvis2} when targeting a cpu that supports such
24400 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
24401 also sets @option{-mvis}.
24407 With @option{-mvis3}, GCC generates code that takes advantage of
24408 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
24409 default is @option{-mvis3} when targeting a cpu that supports such
24410 instructions, such as niagara-3 and later. Setting @option{-mvis3}
24411 also sets @option{-mvis2} and @option{-mvis}.
24417 With @option{-mvis4}, GCC generates code that takes advantage of
24418 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
24419 default is @option{-mvis4} when targeting a cpu that supports such
24420 instructions, such as niagara-7 and later. Setting @option{-mvis4}
24421 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
24427 With @option{-mvis4b}, GCC generates code that takes advantage of
24428 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
24429 the additional VIS instructions introduced in the Oracle SPARC
24430 Architecture 2017. The default is @option{-mvis4b} when targeting a
24431 cpu that supports such instructions, such as m8 and later. Setting
24432 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
24433 @option{-mvis2} and @option{-mvis}.
24438 @opindex mno-cbcond
24439 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
24440 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
24441 when targeting a CPU that supports such instructions, such as Niagara-4 and
24448 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
24449 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
24450 when targeting a CPU that supports such instructions, such as Niagara-3 and
24456 @opindex mno-fsmuld
24457 With @option{-mfsmuld}, GCC generates code that takes advantage of the
24458 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
24459 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
24460 or V9 with FPU except @option{-mcpu=leon}.
24466 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
24467 Population Count instruction. The default is @option{-mpopc}
24468 when targeting a CPU that supports such an instruction, such as Niagara-2 and
24475 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
24476 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
24477 when targeting a CPU that supports such an instruction, such as Niagara-7 and
24481 @opindex mfix-at697f
24482 Enable the documented workaround for the single erratum of the Atmel AT697F
24483 processor (which corresponds to erratum #13 of the AT697E processor).
24486 @opindex mfix-ut699
24487 Enable the documented workarounds for the floating-point errata and the data
24488 cache nullify errata of the UT699 processor.
24491 @opindex mfix-ut700
24492 Enable the documented workaround for the back-to-back store errata of
24493 the UT699E/UT700 processor.
24495 @item -mfix-gr712rc
24496 @opindex mfix-gr712rc
24497 Enable the documented workaround for the back-to-back store errata of
24498 the GR712RC processor.
24501 These @samp{-m} options are supported in addition to the above
24502 on SPARC-V9 processors in 64-bit environments:
24509 Generate code for a 32-bit or 64-bit environment.
24510 The 32-bit environment sets int, long and pointer to 32 bits.
24511 The 64-bit environment sets int to 32 bits and long and pointer
24514 @item -mcmodel=@var{which}
24516 Set the code model to one of
24520 The Medium/Low code model: 64-bit addresses, programs
24521 must be linked in the low 32 bits of memory. Programs can be statically
24522 or dynamically linked.
24525 The Medium/Middle code model: 64-bit addresses, programs
24526 must be linked in the low 44 bits of memory, the text and data segments must
24527 be less than 2GB in size and the data segment must be located within 2GB of
24531 The Medium/Anywhere code model: 64-bit addresses, programs
24532 may be linked anywhere in memory, the text and data segments must be less
24533 than 2GB in size and the data segment must be located within 2GB of the
24537 The Medium/Anywhere code model for embedded systems:
24538 64-bit addresses, the text and data segments must be less than 2GB in
24539 size, both starting anywhere in memory (determined at link time). The
24540 global register %g4 points to the base of the data segment. Programs
24541 are statically linked and PIC is not supported.
24544 @item -mmemory-model=@var{mem-model}
24545 @opindex mmemory-model
24546 Set the memory model in force on the processor to one of
24550 The default memory model for the processor and operating system.
24553 Relaxed Memory Order
24556 Partial Store Order
24562 Sequential Consistency
24565 These memory models are formally defined in Appendix D of the SPARC-V9
24566 architecture manual, as set in the processor's @code{PSTATE.MM} field.
24569 @itemx -mno-stack-bias
24570 @opindex mstack-bias
24571 @opindex mno-stack-bias
24572 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
24573 frame pointer if present, are offset by @minus{}2047 which must be added back
24574 when making stack frame references. This is the default in 64-bit mode.
24575 Otherwise, assume no such offset is present.
24579 @subsection SPU Options
24580 @cindex SPU options
24582 These @samp{-m} options are supported on the SPU:
24586 @itemx -merror-reloc
24587 @opindex mwarn-reloc
24588 @opindex merror-reloc
24590 The loader for SPU does not handle dynamic relocations. By default, GCC
24591 gives an error when it generates code that requires a dynamic
24592 relocation. @option{-mno-error-reloc} disables the error,
24593 @option{-mwarn-reloc} generates a warning instead.
24596 @itemx -munsafe-dma
24598 @opindex munsafe-dma
24600 Instructions that initiate or test completion of DMA must not be
24601 reordered with respect to loads and stores of the memory that is being
24603 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
24604 memory accesses, but that can lead to inefficient code in places where the
24605 memory is known to not change. Rather than mark the memory as volatile,
24606 you can use @option{-msafe-dma} to tell the compiler to treat
24607 the DMA instructions as potentially affecting all memory.
24609 @item -mbranch-hints
24610 @opindex mbranch-hints
24612 By default, GCC generates a branch hint instruction to avoid
24613 pipeline stalls for always-taken or probably-taken branches. A hint
24614 is not generated closer than 8 instructions away from its branch.
24615 There is little reason to disable them, except for debugging purposes,
24616 or to make an object a little bit smaller.
24620 @opindex msmall-mem
24621 @opindex mlarge-mem
24623 By default, GCC generates code assuming that addresses are never larger
24624 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
24625 a full 32-bit address.
24630 By default, GCC links against startup code that assumes the SPU-style
24631 main function interface (which has an unconventional parameter list).
24632 With @option{-mstdmain}, GCC links your program against startup
24633 code that assumes a C99-style interface to @code{main}, including a
24634 local copy of @code{argv} strings.
24636 @item -mfixed-range=@var{register-range}
24637 @opindex mfixed-range
24638 Generate code treating the given register range as fixed registers.
24639 A fixed register is one that the register allocator cannot use. This is
24640 useful when compiling kernel code. A register range is specified as
24641 two registers separated by a dash. Multiple register ranges can be
24642 specified separated by a comma.
24648 Compile code assuming that pointers to the PPU address space accessed
24649 via the @code{__ea} named address space qualifier are either 32 or 64
24650 bits wide. The default is 32 bits. As this is an ABI-changing option,
24651 all object code in an executable must be compiled with the same setting.
24653 @item -maddress-space-conversion
24654 @itemx -mno-address-space-conversion
24655 @opindex maddress-space-conversion
24656 @opindex mno-address-space-conversion
24657 Allow/disallow treating the @code{__ea} address space as superset
24658 of the generic address space. This enables explicit type casts
24659 between @code{__ea} and generic pointer as well as implicit
24660 conversions of generic pointers to @code{__ea} pointers. The
24661 default is to allow address space pointer conversions.
24663 @item -mcache-size=@var{cache-size}
24664 @opindex mcache-size
24665 This option controls the version of libgcc that the compiler links to an
24666 executable and selects a software-managed cache for accessing variables
24667 in the @code{__ea} address space with a particular cache size. Possible
24668 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
24669 and @samp{128}. The default cache size is 64KB.
24671 @item -matomic-updates
24672 @itemx -mno-atomic-updates
24673 @opindex matomic-updates
24674 @opindex mno-atomic-updates
24675 This option controls the version of libgcc that the compiler links to an
24676 executable and selects whether atomic updates to the software-managed
24677 cache of PPU-side variables are used. If you use atomic updates, changes
24678 to a PPU variable from SPU code using the @code{__ea} named address space
24679 qualifier do not interfere with changes to other PPU variables residing
24680 in the same cache line from PPU code. If you do not use atomic updates,
24681 such interference may occur; however, writing back cache lines is
24682 more efficient. The default behavior is to use atomic updates.
24685 @itemx -mdual-nops=@var{n}
24686 @opindex mdual-nops
24687 By default, GCC inserts NOPs to increase dual issue when it expects
24688 it to increase performance. @var{n} can be a value from 0 to 10. A
24689 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
24690 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
24692 @item -mhint-max-nops=@var{n}
24693 @opindex mhint-max-nops
24694 Maximum number of NOPs to insert for a branch hint. A branch hint must
24695 be at least 8 instructions away from the branch it is affecting. GCC
24696 inserts up to @var{n} NOPs to enforce this, otherwise it does not
24697 generate the branch hint.
24699 @item -mhint-max-distance=@var{n}
24700 @opindex mhint-max-distance
24701 The encoding of the branch hint instruction limits the hint to be within
24702 256 instructions of the branch it is affecting. By default, GCC makes
24703 sure it is within 125.
24706 @opindex msafe-hints
24707 Work around a hardware bug that causes the SPU to stall indefinitely.
24708 By default, GCC inserts the @code{hbrp} instruction to make sure
24709 this stall won't happen.
24713 @node System V Options
24714 @subsection Options for System V
24716 These additional options are available on System V Release 4 for
24717 compatibility with other compilers on those systems:
24722 Create a shared object.
24723 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
24727 Identify the versions of each tool used by the compiler, in a
24728 @code{.ident} assembler directive in the output.
24732 Refrain from adding @code{.ident} directives to the output file (this is
24735 @item -YP,@var{dirs}
24737 Search the directories @var{dirs}, and no others, for libraries
24738 specified with @option{-l}.
24740 @item -Ym,@var{dir}
24742 Look in the directory @var{dir} to find the M4 preprocessor.
24743 The assembler uses this option.
24744 @c This is supposed to go with a -Yd for predefined M4 macro files, but
24745 @c the generic assembler that comes with Solaris takes just -Ym.
24748 @node TILE-Gx Options
24749 @subsection TILE-Gx Options
24750 @cindex TILE-Gx options
24752 These @samp{-m} options are supported on the TILE-Gx:
24755 @item -mcmodel=small
24756 @opindex mcmodel=small
24757 Generate code for the small model. The distance for direct calls is
24758 limited to 500M in either direction. PC-relative addresses are 32
24759 bits. Absolute addresses support the full address range.
24761 @item -mcmodel=large
24762 @opindex mcmodel=large
24763 Generate code for the large model. There is no limitation on call
24764 distance, pc-relative addresses, or absolute addresses.
24766 @item -mcpu=@var{name}
24768 Selects the type of CPU to be targeted. Currently the only supported
24769 type is @samp{tilegx}.
24775 Generate code for a 32-bit or 64-bit environment. The 32-bit
24776 environment sets int, long, and pointer to 32 bits. The 64-bit
24777 environment sets int to 32 bits and long and pointer to 64 bits.
24780 @itemx -mlittle-endian
24781 @opindex mbig-endian
24782 @opindex mlittle-endian
24783 Generate code in big/little endian mode, respectively.
24786 @node TILEPro Options
24787 @subsection TILEPro Options
24788 @cindex TILEPro options
24790 These @samp{-m} options are supported on the TILEPro:
24793 @item -mcpu=@var{name}
24795 Selects the type of CPU to be targeted. Currently the only supported
24796 type is @samp{tilepro}.
24800 Generate code for a 32-bit environment, which sets int, long, and
24801 pointer to 32 bits. This is the only supported behavior so the flag
24802 is essentially ignored.
24806 @subsection V850 Options
24807 @cindex V850 Options
24809 These @samp{-m} options are defined for V850 implementations:
24813 @itemx -mno-long-calls
24814 @opindex mlong-calls
24815 @opindex mno-long-calls
24816 Treat all calls as being far away (near). If calls are assumed to be
24817 far away, the compiler always loads the function's address into a
24818 register, and calls indirect through the pointer.
24824 Do not optimize (do optimize) basic blocks that use the same index
24825 pointer 4 or more times to copy pointer into the @code{ep} register, and
24826 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
24827 option is on by default if you optimize.
24829 @item -mno-prolog-function
24830 @itemx -mprolog-function
24831 @opindex mno-prolog-function
24832 @opindex mprolog-function
24833 Do not use (do use) external functions to save and restore registers
24834 at the prologue and epilogue of a function. The external functions
24835 are slower, but use less code space if more than one function saves
24836 the same number of registers. The @option{-mprolog-function} option
24837 is on by default if you optimize.
24841 Try to make the code as small as possible. At present, this just turns
24842 on the @option{-mep} and @option{-mprolog-function} options.
24844 @item -mtda=@var{n}
24846 Put static or global variables whose size is @var{n} bytes or less into
24847 the tiny data area that register @code{ep} points to. The tiny data
24848 area can hold up to 256 bytes in total (128 bytes for byte references).
24850 @item -msda=@var{n}
24852 Put static or global variables whose size is @var{n} bytes or less into
24853 the small data area that register @code{gp} points to. The small data
24854 area can hold up to 64 kilobytes.
24856 @item -mzda=@var{n}
24858 Put static or global variables whose size is @var{n} bytes or less into
24859 the first 32 kilobytes of memory.
24863 Specify that the target processor is the V850.
24867 Specify that the target processor is the V850E3V5. The preprocessor
24868 constant @code{__v850e3v5__} is defined if this option is used.
24872 Specify that the target processor is the V850E3V5. This is an alias for
24873 the @option{-mv850e3v5} option.
24877 Specify that the target processor is the V850E2V3. The preprocessor
24878 constant @code{__v850e2v3__} is defined if this option is used.
24882 Specify that the target processor is the V850E2. The preprocessor
24883 constant @code{__v850e2__} is defined if this option is used.
24887 Specify that the target processor is the V850E1. The preprocessor
24888 constants @code{__v850e1__} and @code{__v850e__} are defined if
24889 this option is used.
24893 Specify that the target processor is the V850ES. This is an alias for
24894 the @option{-mv850e1} option.
24898 Specify that the target processor is the V850E@. The preprocessor
24899 constant @code{__v850e__} is defined if this option is used.
24901 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
24902 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
24903 are defined then a default target processor is chosen and the
24904 relevant @samp{__v850*__} preprocessor constant is defined.
24906 The preprocessor constants @code{__v850} and @code{__v851__} are always
24907 defined, regardless of which processor variant is the target.
24909 @item -mdisable-callt
24910 @itemx -mno-disable-callt
24911 @opindex mdisable-callt
24912 @opindex mno-disable-callt
24913 This option suppresses generation of the @code{CALLT} instruction for the
24914 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
24917 This option is enabled by default when the RH850 ABI is
24918 in use (see @option{-mrh850-abi}), and disabled by default when the
24919 GCC ABI is in use. If @code{CALLT} instructions are being generated
24920 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
24926 Pass on (or do not pass on) the @option{-mrelax} command-line option
24930 @itemx -mno-long-jumps
24931 @opindex mlong-jumps
24932 @opindex mno-long-jumps
24933 Disable (or re-enable) the generation of PC-relative jump instructions.
24936 @itemx -mhard-float
24937 @opindex msoft-float
24938 @opindex mhard-float
24939 Disable (or re-enable) the generation of hardware floating point
24940 instructions. This option is only significant when the target
24941 architecture is @samp{V850E2V3} or higher. If hardware floating point
24942 instructions are being generated then the C preprocessor symbol
24943 @code{__FPU_OK__} is defined, otherwise the symbol
24944 @code{__NO_FPU__} is defined.
24948 Enables the use of the e3v5 LOOP instruction. The use of this
24949 instruction is not enabled by default when the e3v5 architecture is
24950 selected because its use is still experimental.
24954 @opindex mrh850-abi
24956 Enables support for the RH850 version of the V850 ABI. This is the
24957 default. With this version of the ABI the following rules apply:
24961 Integer sized structures and unions are returned via a memory pointer
24962 rather than a register.
24965 Large structures and unions (more than 8 bytes in size) are passed by
24969 Functions are aligned to 16-bit boundaries.
24972 The @option{-m8byte-align} command-line option is supported.
24975 The @option{-mdisable-callt} command-line option is enabled by
24976 default. The @option{-mno-disable-callt} command-line option is not
24980 When this version of the ABI is enabled the C preprocessor symbol
24981 @code{__V850_RH850_ABI__} is defined.
24985 Enables support for the old GCC version of the V850 ABI. With this
24986 version of the ABI the following rules apply:
24990 Integer sized structures and unions are returned in register @code{r10}.
24993 Large structures and unions (more than 8 bytes in size) are passed by
24997 Functions are aligned to 32-bit boundaries, unless optimizing for
25001 The @option{-m8byte-align} command-line option is not supported.
25004 The @option{-mdisable-callt} command-line option is supported but not
25005 enabled by default.
25008 When this version of the ABI is enabled the C preprocessor symbol
25009 @code{__V850_GCC_ABI__} is defined.
25011 @item -m8byte-align
25012 @itemx -mno-8byte-align
25013 @opindex m8byte-align
25014 @opindex mno-8byte-align
25015 Enables support for @code{double} and @code{long long} types to be
25016 aligned on 8-byte boundaries. The default is to restrict the
25017 alignment of all objects to at most 4-bytes. When
25018 @option{-m8byte-align} is in effect the C preprocessor symbol
25019 @code{__V850_8BYTE_ALIGN__} is defined.
25022 @opindex mbig-switch
25023 Generate code suitable for big switch tables. Use this option only if
25024 the assembler/linker complain about out of range branches within a switch
25029 This option causes r2 and r5 to be used in the code generated by
25030 the compiler. This setting is the default.
25032 @item -mno-app-regs
25033 @opindex mno-app-regs
25034 This option causes r2 and r5 to be treated as fixed registers.
25039 @subsection VAX Options
25040 @cindex VAX options
25042 These @samp{-m} options are defined for the VAX:
25047 Do not output certain jump instructions (@code{aobleq} and so on)
25048 that the Unix assembler for the VAX cannot handle across long
25053 Do output those jump instructions, on the assumption that the
25054 GNU assembler is being used.
25058 Output code for G-format floating-point numbers instead of D-format.
25061 @node Visium Options
25062 @subsection Visium Options
25063 @cindex Visium options
25069 A program which performs file I/O and is destined to run on an MCM target
25070 should be linked with this option. It causes the libraries libc.a and
25071 libdebug.a to be linked. The program should be run on the target under
25072 the control of the GDB remote debugging stub.
25076 A program which performs file I/O and is destined to run on the simulator
25077 should be linked with option. This causes libraries libc.a and libsim.a to
25081 @itemx -mhard-float
25083 @opindex mhard-float
25084 Generate code containing floating-point instructions. This is the
25088 @itemx -msoft-float
25090 @opindex msoft-float
25091 Generate code containing library calls for floating-point.
25093 @option{-msoft-float} changes the calling convention in the output file;
25094 therefore, it is only useful if you compile @emph{all} of a program with
25095 this option. In particular, you need to compile @file{libgcc.a}, the
25096 library that comes with GCC, with @option{-msoft-float} in order for
25099 @item -mcpu=@var{cpu_type}
25101 Set the instruction set, register set, and instruction scheduling parameters
25102 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
25103 @samp{mcm}, @samp{gr5} and @samp{gr6}.
25105 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
25107 By default (unless configured otherwise), GCC generates code for the GR5
25108 variant of the Visium architecture.
25110 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
25111 architecture. The only difference from GR5 code is that the compiler will
25112 generate block move instructions.
25114 @item -mtune=@var{cpu_type}
25116 Set the instruction scheduling parameters for machine type @var{cpu_type},
25117 but do not set the instruction set or register set that the option
25118 @option{-mcpu=@var{cpu_type}} would.
25122 Generate code for the supervisor mode, where there are no restrictions on
25123 the access to general registers. This is the default.
25126 @opindex muser-mode
25127 Generate code for the user mode, where the access to some general registers
25128 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
25129 mode; on the GR6, only registers r29 to r31 are affected.
25133 @subsection VMS Options
25135 These @samp{-m} options are defined for the VMS implementations:
25138 @item -mvms-return-codes
25139 @opindex mvms-return-codes
25140 Return VMS condition codes from @code{main}. The default is to return POSIX-style
25141 condition (e.g.@ error) codes.
25143 @item -mdebug-main=@var{prefix}
25144 @opindex mdebug-main=@var{prefix}
25145 Flag the first routine whose name starts with @var{prefix} as the main
25146 routine for the debugger.
25150 Default to 64-bit memory allocation routines.
25152 @item -mpointer-size=@var{size}
25153 @opindex mpointer-size=@var{size}
25154 Set the default size of pointers. Possible options for @var{size} are
25155 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
25156 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
25157 The later option disables @code{pragma pointer_size}.
25160 @node VxWorks Options
25161 @subsection VxWorks Options
25162 @cindex VxWorks Options
25164 The options in this section are defined for all VxWorks targets.
25165 Options specific to the target hardware are listed with the other
25166 options for that target.
25171 GCC can generate code for both VxWorks kernels and real time processes
25172 (RTPs). This option switches from the former to the latter. It also
25173 defines the preprocessor macro @code{__RTP__}.
25176 @opindex non-static
25177 Link an RTP executable against shared libraries rather than static
25178 libraries. The options @option{-static} and @option{-shared} can
25179 also be used for RTPs (@pxref{Link Options}); @option{-static}
25186 These options are passed down to the linker. They are defined for
25187 compatibility with Diab.
25190 @opindex Xbind-lazy
25191 Enable lazy binding of function calls. This option is equivalent to
25192 @option{-Wl,-z,now} and is defined for compatibility with Diab.
25196 Disable lazy binding of function calls. This option is the default and
25197 is defined for compatibility with Diab.
25201 @subsection x86 Options
25202 @cindex x86 Options
25204 These @samp{-m} options are defined for the x86 family of computers.
25208 @item -march=@var{cpu-type}
25210 Generate instructions for the machine type @var{cpu-type}. In contrast to
25211 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
25212 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
25213 to generate code that may not run at all on processors other than the one
25214 indicated. Specifying @option{-march=@var{cpu-type}} implies
25215 @option{-mtune=@var{cpu-type}}.
25217 The choices for @var{cpu-type} are:
25221 This selects the CPU to generate code for at compilation time by determining
25222 the processor type of the compiling machine. Using @option{-march=native}
25223 enables all instruction subsets supported by the local machine (hence
25224 the result might not run on different machines). Using @option{-mtune=native}
25225 produces code optimized for the local machine under the constraints
25226 of the selected instruction set.
25229 Original Intel i386 CPU@.
25232 Intel i486 CPU@. (No scheduling is implemented for this chip.)
25236 Intel Pentium CPU with no MMX support.
25239 Intel Lakemont MCU, based on Intel Pentium CPU.
25242 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
25245 Intel Pentium Pro CPU@.
25248 When used with @option{-march}, the Pentium Pro
25249 instruction set is used, so the code runs on all i686 family chips.
25250 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
25253 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
25258 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
25262 Intel Pentium M; low-power version of Intel Pentium III CPU
25263 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
25267 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
25270 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
25274 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
25275 SSE2 and SSE3 instruction set support.
25278 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
25279 instruction set support.
25282 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25283 SSE4.1, SSE4.2 and POPCNT instruction set support.
25286 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25287 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
25290 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25291 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
25294 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
25295 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
25296 instruction set support.
25299 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25300 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25301 BMI, BMI2 and F16C instruction set support.
25304 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25305 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25306 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
25309 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25310 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25311 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
25312 XSAVES instruction set support.
25315 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
25316 instruction set support.
25319 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
25320 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
25323 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25324 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25325 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
25326 AVX512CD instruction set support.
25329 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25330 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25331 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25332 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
25334 @item skylake-avx512
25335 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
25336 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
25337 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
25338 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
25341 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
25342 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
25343 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
25344 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
25345 AVX512IFMA, SHA, CLWB and UMIP instruction set support.
25348 AMD K6 CPU with MMX instruction set support.
25352 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
25355 @itemx athlon-tbird
25356 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
25362 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
25363 instruction set support.
25369 Processors based on the AMD K8 core with x86-64 instruction set support,
25370 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
25371 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
25372 instruction set extensions.)
25375 @itemx opteron-sse3
25376 @itemx athlon64-sse3
25377 Improved versions of AMD K8 cores with SSE3 instruction set support.
25381 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
25382 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
25383 instruction set extensions.)
25386 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
25387 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
25388 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
25390 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25391 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
25392 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
25395 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25396 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
25397 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
25398 64-bit instruction set extensions.
25400 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
25401 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
25402 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
25403 SSE4.2, ABM and 64-bit instruction set extensions.
25406 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
25407 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
25408 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
25409 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
25410 instruction set extensions.
25413 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
25414 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
25415 instruction set extensions.)
25418 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
25419 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
25420 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
25423 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
25427 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
25428 instruction set support.
25431 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
25432 (No scheduling is implemented for this chip.)
25435 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
25436 (No scheduling is implemented for this chip.)
25439 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25440 (No scheduling is implemented for this chip.)
25443 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
25444 (No scheduling is implemented for this chip.)
25447 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
25448 (No scheduling is implemented for this chip.)
25451 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
25452 (No scheduling is implemented for this chip.)
25455 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
25456 (No scheduling is implemented for this chip.)
25459 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
25460 AVX and AVX2 instruction set support.
25461 (No scheduling is implemented for this chip.)
25464 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25465 instruction set support.
25466 (No scheduling is implemented for this chip.)
25469 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25470 instruction set support.
25471 (No scheduling is implemented for this chip.)
25474 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
25475 instruction set support.
25476 (No scheduling is implemented for this chip.)
25479 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25480 instruction set support.
25481 (No scheduling is implemented for this chip.)
25484 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25485 instruction set support.
25486 (No scheduling is implemented for this chip.)
25489 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
25490 instruction set support.
25491 (No scheduling is implemented for this chip.)
25494 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
25497 @item -mtune=@var{cpu-type}
25499 Tune to @var{cpu-type} everything applicable about the generated code, except
25500 for the ABI and the set of available instructions.
25501 While picking a specific @var{cpu-type} schedules things appropriately
25502 for that particular chip, the compiler does not generate any code that
25503 cannot run on the default machine type unless you use a
25504 @option{-march=@var{cpu-type}} option.
25505 For example, if GCC is configured for i686-pc-linux-gnu
25506 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
25507 but still runs on i686 machines.
25509 The choices for @var{cpu-type} are the same as for @option{-march}.
25510 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
25514 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
25515 If you know the CPU on which your code will run, then you should use
25516 the corresponding @option{-mtune} or @option{-march} option instead of
25517 @option{-mtune=generic}. But, if you do not know exactly what CPU users
25518 of your application will have, then you should use this option.
25520 As new processors are deployed in the marketplace, the behavior of this
25521 option will change. Therefore, if you upgrade to a newer version of
25522 GCC, code generation controlled by this option will change to reflect
25524 that are most common at the time that version of GCC is released.
25526 There is no @option{-march=generic} option because @option{-march}
25527 indicates the instruction set the compiler can use, and there is no
25528 generic instruction set applicable to all processors. In contrast,
25529 @option{-mtune} indicates the processor (or, in this case, collection of
25530 processors) for which the code is optimized.
25533 Produce code optimized for the most current Intel processors, which are
25534 Haswell and Silvermont for this version of GCC. If you know the CPU
25535 on which your code will run, then you should use the corresponding
25536 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
25537 But, if you want your application performs better on both Haswell and
25538 Silvermont, then you should use this option.
25540 As new Intel processors are deployed in the marketplace, the behavior of
25541 this option will change. Therefore, if you upgrade to a newer version of
25542 GCC, code generation controlled by this option will change to reflect
25543 the most current Intel processors at the time that version of GCC is
25546 There is no @option{-march=intel} option because @option{-march} indicates
25547 the instruction set the compiler can use, and there is no common
25548 instruction set applicable to all processors. In contrast,
25549 @option{-mtune} indicates the processor (or, in this case, collection of
25550 processors) for which the code is optimized.
25553 @item -mcpu=@var{cpu-type}
25555 A deprecated synonym for @option{-mtune}.
25557 @item -mfpmath=@var{unit}
25559 Generate floating-point arithmetic for selected unit @var{unit}. The choices
25560 for @var{unit} are:
25564 Use the standard 387 floating-point coprocessor present on the majority of chips and
25565 emulated otherwise. Code compiled with this option runs almost everywhere.
25566 The temporary results are computed in 80-bit precision instead of the precision
25567 specified by the type, resulting in slightly different results compared to most
25568 of other chips. See @option{-ffloat-store} for more detailed description.
25570 This is the default choice for non-Darwin x86-32 targets.
25573 Use scalar floating-point instructions present in the SSE instruction set.
25574 This instruction set is supported by Pentium III and newer chips,
25575 and in the AMD line
25576 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
25577 instruction set supports only single-precision arithmetic, thus the double and
25578 extended-precision arithmetic are still done using 387. A later version, present
25579 only in Pentium 4 and AMD x86-64 chips, supports double-precision
25582 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
25583 or @option{-msse2} switches to enable SSE extensions and make this option
25584 effective. For the x86-64 compiler, these extensions are enabled by default.
25586 The resulting code should be considerably faster in the majority of cases and avoid
25587 the numerical instability problems of 387 code, but may break some existing
25588 code that expects temporaries to be 80 bits.
25590 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
25591 and the default choice for x86-32 targets with the SSE2 instruction set
25592 when @option{-ffast-math} is enabled.
25597 Attempt to utilize both instruction sets at once. This effectively doubles the
25598 amount of available registers, and on chips with separate execution units for
25599 387 and SSE the execution resources too. Use this option with care, as it is
25600 still experimental, because the GCC register allocator does not model separate
25601 functional units well, resulting in unstable performance.
25604 @item -masm=@var{dialect}
25605 @opindex masm=@var{dialect}
25606 Output assembly instructions using selected @var{dialect}. Also affects
25607 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
25608 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
25609 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
25610 not support @samp{intel}.
25613 @itemx -mno-ieee-fp
25615 @opindex mno-ieee-fp
25616 Control whether or not the compiler uses IEEE floating-point
25617 comparisons. These correctly handle the case where the result of a
25618 comparison is unordered.
25623 @opindex mhard-float
25624 Generate output containing 80387 instructions for floating point.
25629 @opindex msoft-float
25630 Generate output containing library calls for floating point.
25632 @strong{Warning:} the requisite libraries are not part of GCC@.
25633 Normally the facilities of the machine's usual C compiler are used, but
25634 this cannot be done directly in cross-compilation. You must make your
25635 own arrangements to provide suitable library functions for
25638 On machines where a function returns floating-point results in the 80387
25639 register stack, some floating-point opcodes may be emitted even if
25640 @option{-msoft-float} is used.
25642 @item -mno-fp-ret-in-387
25643 @opindex mno-fp-ret-in-387
25644 Do not use the FPU registers for return values of functions.
25646 The usual calling convention has functions return values of types
25647 @code{float} and @code{double} in an FPU register, even if there
25648 is no FPU@. The idea is that the operating system should emulate
25651 The option @option{-mno-fp-ret-in-387} causes such values to be returned
25652 in ordinary CPU registers instead.
25654 @item -mno-fancy-math-387
25655 @opindex mno-fancy-math-387
25656 Some 387 emulators do not support the @code{sin}, @code{cos} and
25657 @code{sqrt} instructions for the 387. Specify this option to avoid
25658 generating those instructions. This option is the default on
25659 OpenBSD and NetBSD@. This option is overridden when @option{-march}
25660 indicates that the target CPU always has an FPU and so the
25661 instruction does not need emulation. These
25662 instructions are not generated unless you also use the
25663 @option{-funsafe-math-optimizations} switch.
25665 @item -malign-double
25666 @itemx -mno-align-double
25667 @opindex malign-double
25668 @opindex mno-align-double
25669 Control whether GCC aligns @code{double}, @code{long double}, and
25670 @code{long long} variables on a two-word boundary or a one-word
25671 boundary. Aligning @code{double} variables on a two-word boundary
25672 produces code that runs somewhat faster on a Pentium at the
25673 expense of more memory.
25675 On x86-64, @option{-malign-double} is enabled by default.
25677 @strong{Warning:} if you use the @option{-malign-double} switch,
25678 structures containing the above types are aligned differently than
25679 the published application binary interface specifications for the x86-32
25680 and are not binary compatible with structures in code compiled
25681 without that switch.
25683 @item -m96bit-long-double
25684 @itemx -m128bit-long-double
25685 @opindex m96bit-long-double
25686 @opindex m128bit-long-double
25687 These switches control the size of @code{long double} type. The x86-32
25688 application binary interface specifies the size to be 96 bits,
25689 so @option{-m96bit-long-double} is the default in 32-bit mode.
25691 Modern architectures (Pentium and newer) prefer @code{long double}
25692 to be aligned to an 8- or 16-byte boundary. In arrays or structures
25693 conforming to the ABI, this is not possible. So specifying
25694 @option{-m128bit-long-double} aligns @code{long double}
25695 to a 16-byte boundary by padding the @code{long double} with an additional
25698 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
25699 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
25701 Notice that neither of these options enable any extra precision over the x87
25702 standard of 80 bits for a @code{long double}.
25704 @strong{Warning:} if you override the default value for your target ABI, this
25705 changes the size of
25706 structures and arrays containing @code{long double} variables,
25707 as well as modifying the function calling convention for functions taking
25708 @code{long double}. Hence they are not binary-compatible
25709 with code compiled without that switch.
25711 @item -mlong-double-64
25712 @itemx -mlong-double-80
25713 @itemx -mlong-double-128
25714 @opindex mlong-double-64
25715 @opindex mlong-double-80
25716 @opindex mlong-double-128
25717 These switches control the size of @code{long double} type. A size
25718 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25719 type. This is the default for 32-bit Bionic C library. A size
25720 of 128 bits makes the @code{long double} type equivalent to the
25721 @code{__float128} type. This is the default for 64-bit Bionic C library.
25723 @strong{Warning:} if you override the default value for your target ABI, this
25724 changes the size of
25725 structures and arrays containing @code{long double} variables,
25726 as well as modifying the function calling convention for functions taking
25727 @code{long double}. Hence they are not binary-compatible
25728 with code compiled without that switch.
25730 @item -malign-data=@var{type}
25731 @opindex malign-data
25732 Control how GCC aligns variables. Supported values for @var{type} are
25733 @samp{compat} uses increased alignment value compatible uses GCC 4.8
25734 and earlier, @samp{abi} uses alignment value as specified by the
25735 psABI, and @samp{cacheline} uses increased alignment value to match
25736 the cache line size. @samp{compat} is the default.
25738 @item -mlarge-data-threshold=@var{threshold}
25739 @opindex mlarge-data-threshold
25740 When @option{-mcmodel=medium} is specified, data objects larger than
25741 @var{threshold} are placed in the large data section. This value must be the
25742 same across all objects linked into the binary, and defaults to 65535.
25746 Use a different function-calling convention, in which functions that
25747 take a fixed number of arguments return with the @code{ret @var{num}}
25748 instruction, which pops their arguments while returning. This saves one
25749 instruction in the caller since there is no need to pop the arguments
25752 You can specify that an individual function is called with this calling
25753 sequence with the function attribute @code{stdcall}. You can also
25754 override the @option{-mrtd} option by using the function attribute
25755 @code{cdecl}. @xref{Function Attributes}.
25757 @strong{Warning:} this calling convention is incompatible with the one
25758 normally used on Unix, so you cannot use it if you need to call
25759 libraries compiled with the Unix compiler.
25761 Also, you must provide function prototypes for all functions that
25762 take variable numbers of arguments (including @code{printf});
25763 otherwise incorrect code is generated for calls to those
25766 In addition, seriously incorrect code results if you call a
25767 function with too many arguments. (Normally, extra arguments are
25768 harmlessly ignored.)
25770 @item -mregparm=@var{num}
25772 Control how many registers are used to pass integer arguments. By
25773 default, no registers are used to pass arguments, and at most 3
25774 registers can be used. You can control this behavior for a specific
25775 function by using the function attribute @code{regparm}.
25776 @xref{Function Attributes}.
25778 @strong{Warning:} if you use this switch, and
25779 @var{num} is nonzero, then you must build all modules with the same
25780 value, including any libraries. This includes the system libraries and
25784 @opindex msseregparm
25785 Use SSE register passing conventions for float and double arguments
25786 and return values. You can control this behavior for a specific
25787 function by using the function attribute @code{sseregparm}.
25788 @xref{Function Attributes}.
25790 @strong{Warning:} if you use this switch then you must build all
25791 modules with the same value, including any libraries. This includes
25792 the system libraries and startup modules.
25794 @item -mvect8-ret-in-mem
25795 @opindex mvect8-ret-in-mem
25796 Return 8-byte vectors in memory instead of MMX registers. This is the
25797 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
25798 Studio compilers until version 12. Later compiler versions (starting
25799 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
25800 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
25801 you need to remain compatible with existing code produced by those
25802 previous compiler versions or older versions of GCC@.
25811 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
25812 is specified, the significands of results of floating-point operations are
25813 rounded to 24 bits (single precision); @option{-mpc64} rounds the
25814 significands of results of floating-point operations to 53 bits (double
25815 precision) and @option{-mpc80} rounds the significands of results of
25816 floating-point operations to 64 bits (extended double precision), which is
25817 the default. When this option is used, floating-point operations in higher
25818 precisions are not available to the programmer without setting the FPU
25819 control word explicitly.
25821 Setting the rounding of floating-point operations to less than the default
25822 80 bits can speed some programs by 2% or more. Note that some mathematical
25823 libraries assume that extended-precision (80-bit) floating-point operations
25824 are enabled by default; routines in such libraries could suffer significant
25825 loss of accuracy, typically through so-called ``catastrophic cancellation'',
25826 when this option is used to set the precision to less than extended precision.
25828 @item -mstackrealign
25829 @opindex mstackrealign
25830 Realign the stack at entry. On the x86, the @option{-mstackrealign}
25831 option generates an alternate prologue and epilogue that realigns the
25832 run-time stack if necessary. This supports mixing legacy codes that keep
25833 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
25834 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
25835 applicable to individual functions.
25837 @item -mpreferred-stack-boundary=@var{num}
25838 @opindex mpreferred-stack-boundary
25839 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
25840 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
25841 the default is 4 (16 bytes or 128 bits).
25843 @strong{Warning:} When generating code for the x86-64 architecture with
25844 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
25845 used to keep the stack boundary aligned to 8 byte boundary. Since
25846 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
25847 intended to be used in controlled environment where stack space is
25848 important limitation. This option leads to wrong code when functions
25849 compiled with 16 byte stack alignment (such as functions from a standard
25850 library) are called with misaligned stack. In this case, SSE
25851 instructions may lead to misaligned memory access traps. In addition,
25852 variable arguments are handled incorrectly for 16 byte aligned
25853 objects (including x87 long double and __int128), leading to wrong
25854 results. You must build all modules with
25855 @option{-mpreferred-stack-boundary=3}, including any libraries. This
25856 includes the system libraries and startup modules.
25858 @item -mincoming-stack-boundary=@var{num}
25859 @opindex mincoming-stack-boundary
25860 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
25861 boundary. If @option{-mincoming-stack-boundary} is not specified,
25862 the one specified by @option{-mpreferred-stack-boundary} is used.
25864 On Pentium and Pentium Pro, @code{double} and @code{long double} values
25865 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
25866 suffer significant run time performance penalties. On Pentium III, the
25867 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
25868 properly if it is not 16-byte aligned.
25870 To ensure proper alignment of this values on the stack, the stack boundary
25871 must be as aligned as that required by any value stored on the stack.
25872 Further, every function must be generated such that it keeps the stack
25873 aligned. Thus calling a function compiled with a higher preferred
25874 stack boundary from a function compiled with a lower preferred stack
25875 boundary most likely misaligns the stack. It is recommended that
25876 libraries that use callbacks always use the default setting.
25878 This extra alignment does consume extra stack space, and generally
25879 increases code size. Code that is sensitive to stack space usage, such
25880 as embedded systems and operating system kernels, may want to reduce the
25881 preferred alignment to @option{-mpreferred-stack-boundary=2}.
25938 @itemx -mavx512ifma
25939 @opindex mavx512ifma
25941 @itemx -mavx512vbmi
25942 @opindex mavx512vbmi
25954 @opindex mclfushopt
25971 @itemx -mprefetchwt1
25972 @opindex mprefetchwt1
26036 @itemx -mavx512vbmi2
26037 @opindex mavx512vbmi2
26041 These switches enable the use of instructions in the MMX, SSE,
26042 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
26043 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
26044 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, BMI, BMI2,
26045 FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, IBT, SHSTK, AVX512VBMI2,
26046 GFNI, 3DNow!@: or enhanced 3DNow!@: extended instruction sets. Each has a
26047 corresponding @option{-mno-} option to disable use of these instructions.
26049 These extensions are also available as built-in functions: see
26050 @ref{x86 Built-in Functions}, for details of the functions enabled and
26051 disabled by these switches.
26053 To generate SSE/SSE2 instructions automatically from floating-point
26054 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
26056 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
26057 generates new AVX instructions or AVX equivalence for all SSEx instructions
26060 These options enable GCC to use these extended instructions in
26061 generated code, even without @option{-mfpmath=sse}. Applications that
26062 perform run-time CPU detection must compile separate files for each
26063 supported architecture, using the appropriate flags. In particular,
26064 the file containing the CPU detection code should be compiled without
26067 The @option{-mcet} option turns on the @option{-mibt} and @option{-mshstk}
26068 options. The @option{-mibt} option enables indirect branch tracking support
26069 and the @option{-mshstk} option enables shadow stack support from
26070 Intel Control-flow Enforcement Technology (CET). The compiler also provides
26071 a number of built-in functions for fine-grained control in a CET-based
26072 application. See @xref{x86 Built-in Functions}, for more information.
26074 @item -mdump-tune-features
26075 @opindex mdump-tune-features
26076 This option instructs GCC to dump the names of the x86 performance
26077 tuning features and default settings. The names can be used in
26078 @option{-mtune-ctrl=@var{feature-list}}.
26080 @item -mtune-ctrl=@var{feature-list}
26081 @opindex mtune-ctrl=@var{feature-list}
26082 This option is used to do fine grain control of x86 code generation features.
26083 @var{feature-list} is a comma separated list of @var{feature} names. See also
26084 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
26085 on if it is not preceded with @samp{^}, otherwise, it is turned off.
26086 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
26087 developers. Using it may lead to code paths not covered by testing and can
26088 potentially result in compiler ICEs or runtime errors.
26091 @opindex mno-default
26092 This option instructs GCC to turn off all tunable features. See also
26093 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
26097 This option instructs GCC to emit a @code{cld} instruction in the prologue
26098 of functions that use string instructions. String instructions depend on
26099 the DF flag to select between autoincrement or autodecrement mode. While the
26100 ABI specifies the DF flag to be cleared on function entry, some operating
26101 systems violate this specification by not clearing the DF flag in their
26102 exception dispatchers. The exception handler can be invoked with the DF flag
26103 set, which leads to wrong direction mode when string instructions are used.
26104 This option can be enabled by default on 32-bit x86 targets by configuring
26105 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
26106 instructions can be suppressed with the @option{-mno-cld} compiler option
26110 @opindex mvzeroupper
26111 This option instructs GCC to emit a @code{vzeroupper} instruction
26112 before a transfer of control flow out of the function to minimize
26113 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
26116 @item -mprefer-avx128
26117 @opindex mprefer-avx128
26118 This option instructs GCC to use 128-bit AVX instructions instead of
26119 256-bit AVX instructions in the auto-vectorizer.
26121 @item -mprefer-vector-width=@var{opt}
26122 @opindex mprefer-vector-width
26123 This option instructs GCC to use @var{opt}-bit vector width in instructions
26124 instead of default on the selected platform.
26128 No extra limitations applied to GCC other than defined by the selected platform.
26131 Prefer 128-bit vector width for instructions.
26134 Prefer 256-bit vector width for instructions.
26137 Prefer 512-bit vector width for instructions.
26142 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
26143 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
26144 objects. This is useful for atomic updates of data structures exceeding one
26145 machine word in size. The compiler uses this instruction to implement
26146 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
26147 128-bit integers, a library call is always used.
26151 This option enables generation of @code{SAHF} instructions in 64-bit code.
26152 Early Intel Pentium 4 CPUs with Intel 64 support,
26153 prior to the introduction of Pentium 4 G1 step in December 2005,
26154 lacked the @code{LAHF} and @code{SAHF} instructions
26155 which are supported by AMD64.
26156 These are load and store instructions, respectively, for certain status flags.
26157 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
26158 @code{drem}, and @code{remainder} built-in functions;
26159 see @ref{Other Builtins} for details.
26163 This option enables use of the @code{movbe} instruction to implement
26164 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
26168 This option tells the compiler to use indirect branch tracking support
26169 (for indirect calls and jumps) from x86 Control-flow Enforcement
26170 Technology (CET). The option has effect only if the
26171 @option{-fcf-protection=full} or @option{-fcf-protection=branch} option
26172 is specified. The option @option{-mibt} is on by default when the
26173 @code{-mcet} option is specified.
26177 This option tells the compiler to use shadow stack support (return
26178 address tracking) from x86 Control-flow Enforcement Technology (CET).
26179 The option has effect only if the @option{-fcf-protection=full} or
26180 @option{-fcf-protection=return} option is specified. The option
26181 @option{-mshstk} is on by default when the @option{-mcet} option is
26186 This option enables built-in functions @code{__builtin_ia32_crc32qi},
26187 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
26188 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
26192 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
26193 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
26194 with an additional Newton-Raphson step
26195 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
26196 (and their vectorized
26197 variants) for single-precision floating-point arguments. These instructions
26198 are generated only when @option{-funsafe-math-optimizations} is enabled
26199 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
26200 Note that while the throughput of the sequence is higher than the throughput
26201 of the non-reciprocal instruction, the precision of the sequence can be
26202 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
26204 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
26205 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
26206 combination), and doesn't need @option{-mrecip}.
26208 Also note that GCC emits the above sequence with additional Newton-Raphson step
26209 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
26210 already with @option{-ffast-math} (or the above option combination), and
26211 doesn't need @option{-mrecip}.
26213 @item -mrecip=@var{opt}
26214 @opindex mrecip=opt
26215 This option controls which reciprocal estimate instructions
26216 may be used. @var{opt} is a comma-separated list of options, which may
26217 be preceded by a @samp{!} to invert the option:
26221 Enable all estimate instructions.
26224 Enable the default instructions, equivalent to @option{-mrecip}.
26227 Disable all estimate instructions, equivalent to @option{-mno-recip}.
26230 Enable the approximation for scalar division.
26233 Enable the approximation for vectorized division.
26236 Enable the approximation for scalar square root.
26239 Enable the approximation for vectorized square root.
26242 So, for example, @option{-mrecip=all,!sqrt} enables
26243 all of the reciprocal approximations, except for square root.
26245 @item -mveclibabi=@var{type}
26246 @opindex mveclibabi
26247 Specifies the ABI type to use for vectorizing intrinsics using an
26248 external library. Supported values for @var{type} are @samp{svml}
26249 for the Intel short
26250 vector math library and @samp{acml} for the AMD math core library.
26251 To use this option, both @option{-ftree-vectorize} and
26252 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
26253 ABI-compatible library must be specified at link time.
26255 GCC currently emits calls to @code{vmldExp2},
26256 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
26257 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
26258 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
26259 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
26260 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
26261 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
26262 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
26263 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
26264 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
26265 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
26266 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
26267 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
26268 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
26269 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
26270 when @option{-mveclibabi=acml} is used.
26272 @item -mabi=@var{name}
26274 Generate code for the specified calling convention. Permissible values
26275 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
26276 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
26277 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
26278 You can control this behavior for specific functions by
26279 using the function attributes @code{ms_abi} and @code{sysv_abi}.
26280 @xref{Function Attributes}.
26282 @item -mforce-indirect-call
26283 @opindex mforce-indirect-call
26284 Force all calls to functions to be indirect. This is useful
26285 when using Intel Processor Trace where it generates more precise timing
26286 information for function calls.
26288 @item -mcall-ms2sysv-xlogues
26289 @opindex mcall-ms2sysv-xlogues
26290 @opindex mno-call-ms2sysv-xlogues
26291 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
26292 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
26293 default, the code for saving and restoring these registers is emitted inline,
26294 resulting in fairly lengthy prologues and epilogues. Using
26295 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
26296 use stubs in the static portion of libgcc to perform these saves and restores,
26297 thus reducing function size at the cost of a few extra instructions.
26299 @item -mtls-dialect=@var{type}
26300 @opindex mtls-dialect
26301 Generate code to access thread-local storage using the @samp{gnu} or
26302 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
26303 @samp{gnu2} is more efficient, but it may add compile- and run-time
26304 requirements that cannot be satisfied on all systems.
26307 @itemx -mno-push-args
26308 @opindex mpush-args
26309 @opindex mno-push-args
26310 Use PUSH operations to store outgoing parameters. This method is shorter
26311 and usually equally fast as method using SUB/MOV operations and is enabled
26312 by default. In some cases disabling it may improve performance because of
26313 improved scheduling and reduced dependencies.
26315 @item -maccumulate-outgoing-args
26316 @opindex maccumulate-outgoing-args
26317 If enabled, the maximum amount of space required for outgoing arguments is
26318 computed in the function prologue. This is faster on most modern CPUs
26319 because of reduced dependencies, improved scheduling and reduced stack usage
26320 when the preferred stack boundary is not equal to 2. The drawback is a notable
26321 increase in code size. This switch implies @option{-mno-push-args}.
26325 Support thread-safe exception handling on MinGW. Programs that rely
26326 on thread-safe exception handling must compile and link all code with the
26327 @option{-mthreads} option. When compiling, @option{-mthreads} defines
26328 @option{-D_MT}; when linking, it links in a special thread helper library
26329 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
26331 @item -mms-bitfields
26332 @itemx -mno-ms-bitfields
26333 @opindex mms-bitfields
26334 @opindex mno-ms-bitfields
26336 Enable/disable bit-field layout compatible with the native Microsoft
26339 If @code{packed} is used on a structure, or if bit-fields are used,
26340 it may be that the Microsoft ABI lays out the structure differently
26341 than the way GCC normally does. Particularly when moving packed
26342 data between functions compiled with GCC and the native Microsoft compiler
26343 (either via function call or as data in a file), it may be necessary to access
26346 This option is enabled by default for Microsoft Windows
26347 targets. This behavior can also be controlled locally by use of variable
26348 or type attributes. For more information, see @ref{x86 Variable Attributes}
26349 and @ref{x86 Type Attributes}.
26351 The Microsoft structure layout algorithm is fairly simple with the exception
26352 of the bit-field packing.
26353 The padding and alignment of members of structures and whether a bit-field
26354 can straddle a storage-unit boundary are determine by these rules:
26357 @item Structure members are stored sequentially in the order in which they are
26358 declared: the first member has the lowest memory address and the last member
26361 @item Every data object has an alignment requirement. The alignment requirement
26362 for all data except structures, unions, and arrays is either the size of the
26363 object or the current packing size (specified with either the
26364 @code{aligned} attribute or the @code{pack} pragma),
26365 whichever is less. For structures, unions, and arrays,
26366 the alignment requirement is the largest alignment requirement of its members.
26367 Every object is allocated an offset so that:
26370 offset % alignment_requirement == 0
26373 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
26374 unit if the integral types are the same size and if the next bit-field fits
26375 into the current allocation unit without crossing the boundary imposed by the
26376 common alignment requirements of the bit-fields.
26379 MSVC interprets zero-length bit-fields in the following ways:
26382 @item If a zero-length bit-field is inserted between two bit-fields that
26383 are normally coalesced, the bit-fields are not coalesced.
26390 unsigned long bf_1 : 12;
26392 unsigned long bf_2 : 12;
26397 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
26398 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
26400 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
26401 alignment of the zero-length bit-field is greater than the member that follows it,
26402 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
26423 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
26424 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
26425 bit-field does not affect the alignment of @code{bar} or, as a result, the size
26428 Taking this into account, it is important to note the following:
26431 @item If a zero-length bit-field follows a normal bit-field, the type of the
26432 zero-length bit-field may affect the alignment of the structure as whole. For
26433 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
26434 normal bit-field, and is of type short.
26436 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
26437 still affect the alignment of the structure:
26448 Here, @code{t4} takes up 4 bytes.
26451 @item Zero-length bit-fields following non-bit-field members are ignored:
26463 Here, @code{t5} takes up 2 bytes.
26467 @item -mno-align-stringops
26468 @opindex mno-align-stringops
26469 Do not align the destination of inlined string operations. This switch reduces
26470 code size and improves performance in case the destination is already aligned,
26471 but GCC doesn't know about it.
26473 @item -minline-all-stringops
26474 @opindex minline-all-stringops
26475 By default GCC inlines string operations only when the destination is
26476 known to be aligned to least a 4-byte boundary.
26477 This enables more inlining and increases code
26478 size, but may improve performance of code that depends on fast
26479 @code{memcpy}, @code{strlen},
26480 and @code{memset} for short lengths.
26482 @item -minline-stringops-dynamically
26483 @opindex minline-stringops-dynamically
26484 For string operations of unknown size, use run-time checks with
26485 inline code for small blocks and a library call for large blocks.
26487 @item -mstringop-strategy=@var{alg}
26488 @opindex mstringop-strategy=@var{alg}
26489 Override the internal decision heuristic for the particular algorithm to use
26490 for inlining string operations. The allowed values for @var{alg} are:
26496 Expand using i386 @code{rep} prefix of the specified size.
26500 @itemx unrolled_loop
26501 Expand into an inline loop.
26504 Always use a library call.
26507 @item -mmemcpy-strategy=@var{strategy}
26508 @opindex mmemcpy-strategy=@var{strategy}
26509 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
26510 should be inlined and what inline algorithm to use when the expected size
26511 of the copy operation is known. @var{strategy}
26512 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
26513 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
26514 the max byte size with which inline algorithm @var{alg} is allowed. For the last
26515 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
26516 in the list must be specified in increasing order. The minimal byte size for
26517 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
26520 @item -mmemset-strategy=@var{strategy}
26521 @opindex mmemset-strategy=@var{strategy}
26522 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
26523 @code{__builtin_memset} expansion.
26525 @item -momit-leaf-frame-pointer
26526 @opindex momit-leaf-frame-pointer
26527 Don't keep the frame pointer in a register for leaf functions. This
26528 avoids the instructions to save, set up, and restore frame pointers and
26529 makes an extra register available in leaf functions. The option
26530 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
26531 which might make debugging harder.
26533 @item -mtls-direct-seg-refs
26534 @itemx -mno-tls-direct-seg-refs
26535 @opindex mtls-direct-seg-refs
26536 Controls whether TLS variables may be accessed with offsets from the
26537 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
26538 or whether the thread base pointer must be added. Whether or not this
26539 is valid depends on the operating system, and whether it maps the
26540 segment to cover the entire TLS area.
26542 For systems that use the GNU C Library, the default is on.
26545 @itemx -mno-sse2avx
26547 Specify that the assembler should encode SSE instructions with VEX
26548 prefix. The option @option{-mavx} turns this on by default.
26553 If profiling is active (@option{-pg}), put the profiling
26554 counter call before the prologue.
26555 Note: On x86 architectures the attribute @code{ms_hook_prologue}
26556 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
26558 @item -mrecord-mcount
26559 @itemx -mno-record-mcount
26560 @opindex mrecord-mcount
26561 If profiling is active (@option{-pg}), generate a __mcount_loc section
26562 that contains pointers to each profiling call. This is useful for
26563 automatically patching and out calls.
26566 @itemx -mno-nop-mcount
26567 @opindex mnop-mcount
26568 If profiling is active (@option{-pg}), generate the calls to
26569 the profiling functions as NOPs. This is useful when they
26570 should be patched in later dynamically. This is likely only
26571 useful together with @option{-mrecord-mcount}.
26573 @item -mskip-rax-setup
26574 @itemx -mno-skip-rax-setup
26575 @opindex mskip-rax-setup
26576 When generating code for the x86-64 architecture with SSE extensions
26577 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
26578 register when there are no variable arguments passed in vector registers.
26580 @strong{Warning:} Since RAX register is used to avoid unnecessarily
26581 saving vector registers on stack when passing variable arguments, the
26582 impacts of this option are callees may waste some stack space,
26583 misbehave or jump to a random location. GCC 4.4 or newer don't have
26584 those issues, regardless the RAX register value.
26587 @itemx -mno-8bit-idiv
26588 @opindex m8bit-idiv
26589 On some processors, like Intel Atom, 8-bit unsigned integer divide is
26590 much faster than 32-bit/64-bit integer divide. This option generates a
26591 run-time check. If both dividend and divisor are within range of 0
26592 to 255, 8-bit unsigned integer divide is used instead of
26593 32-bit/64-bit integer divide.
26595 @item -mavx256-split-unaligned-load
26596 @itemx -mavx256-split-unaligned-store
26597 @opindex mavx256-split-unaligned-load
26598 @opindex mavx256-split-unaligned-store
26599 Split 32-byte AVX unaligned load and store.
26601 @item -mstack-protector-guard=@var{guard}
26602 @itemx -mstack-protector-guard-reg=@var{reg}
26603 @itemx -mstack-protector-guard-offset=@var{offset}
26604 @opindex mstack-protector-guard
26605 @opindex mstack-protector-guard-reg
26606 @opindex mstack-protector-guard-offset
26607 Generate stack protection code using canary at @var{guard}. Supported
26608 locations are @samp{global} for global canary or @samp{tls} for per-thread
26609 canary in the TLS block (the default). This option has effect only when
26610 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
26612 With the latter choice the options
26613 @option{-mstack-protector-guard-reg=@var{reg}} and
26614 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
26615 which segment register (@code{%fs} or @code{%gs}) to use as base register
26616 for reading the canary, and from what offset from that base register.
26617 The default for those is as specified in the relevant ABI.
26619 @item -mmitigate-rop
26620 @opindex mmitigate-rop
26621 Try to avoid generating code sequences that contain unintended return
26622 opcodes, to mitigate against certain forms of attack. At the moment,
26623 this option is limited in what it can do and should not be relied
26624 on to provide serious protection.
26626 @item -mgeneral-regs-only
26627 @opindex mgeneral-regs-only
26628 Generate code that uses only the general-purpose registers. This
26629 prevents the compiler from using floating-point, vector, mask and bound
26634 These @samp{-m} switches are supported in addition to the above
26635 on x86-64 processors in 64-bit environments.
26648 Generate code for a 16-bit, 32-bit or 64-bit environment.
26649 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
26651 generates code that runs on any i386 system.
26653 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
26654 types to 64 bits, and generates code for the x86-64 architecture.
26655 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
26656 and @option{-mdynamic-no-pic} options.
26658 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
26660 generates code for the x86-64 architecture.
26662 The @option{-m16} option is the same as @option{-m32}, except for that
26663 it outputs the @code{.code16gcc} assembly directive at the beginning of
26664 the assembly output so that the binary can run in 16-bit mode.
26666 The @option{-miamcu} option generates code which conforms to Intel MCU
26667 psABI. It requires the @option{-m32} option to be turned on.
26669 @item -mno-red-zone
26670 @opindex mno-red-zone
26671 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
26672 by the x86-64 ABI; it is a 128-byte area beyond the location of the
26673 stack pointer that is not modified by signal or interrupt handlers
26674 and therefore can be used for temporary data without adjusting the stack
26675 pointer. The flag @option{-mno-red-zone} disables this red zone.
26677 @item -mcmodel=small
26678 @opindex mcmodel=small
26679 Generate code for the small code model: the program and its symbols must
26680 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
26681 Programs can be statically or dynamically linked. This is the default
26684 @item -mcmodel=kernel
26685 @opindex mcmodel=kernel
26686 Generate code for the kernel code model. The kernel runs in the
26687 negative 2 GB of the address space.
26688 This model has to be used for Linux kernel code.
26690 @item -mcmodel=medium
26691 @opindex mcmodel=medium
26692 Generate code for the medium model: the program is linked in the lower 2
26693 GB of the address space. Small symbols are also placed there. Symbols
26694 with sizes larger than @option{-mlarge-data-threshold} are put into
26695 large data or BSS sections and can be located above 2GB. Programs can
26696 be statically or dynamically linked.
26698 @item -mcmodel=large
26699 @opindex mcmodel=large
26700 Generate code for the large model. This model makes no assumptions
26701 about addresses and sizes of sections.
26703 @item -maddress-mode=long
26704 @opindex maddress-mode=long
26705 Generate code for long address mode. This is only supported for 64-bit
26706 and x32 environments. It is the default address mode for 64-bit
26709 @item -maddress-mode=short
26710 @opindex maddress-mode=short
26711 Generate code for short address mode. This is only supported for 32-bit
26712 and x32 environments. It is the default address mode for 32-bit and
26716 @node x86 Windows Options
26717 @subsection x86 Windows Options
26718 @cindex x86 Windows Options
26719 @cindex Windows Options for x86
26721 These additional options are available for Microsoft Windows targets:
26727 specifies that a console application is to be generated, by
26728 instructing the linker to set the PE header subsystem type
26729 required for console applications.
26730 This option is available for Cygwin and MinGW targets and is
26731 enabled by default on those targets.
26735 This option is available for Cygwin and MinGW targets. It
26736 specifies that a DLL---a dynamic link library---is to be
26737 generated, enabling the selection of the required runtime
26738 startup object and entry point.
26740 @item -mnop-fun-dllimport
26741 @opindex mnop-fun-dllimport
26742 This option is available for Cygwin and MinGW targets. It
26743 specifies that the @code{dllimport} attribute should be ignored.
26747 This option is available for MinGW targets. It specifies
26748 that MinGW-specific thread support is to be used.
26752 This option is available for MinGW-w64 targets. It causes
26753 the @code{UNICODE} preprocessor macro to be predefined, and
26754 chooses Unicode-capable runtime startup code.
26758 This option is available for Cygwin and MinGW targets. It
26759 specifies that the typical Microsoft Windows predefined macros are to
26760 be set in the pre-processor, but does not influence the choice
26761 of runtime library/startup code.
26765 This option is available for Cygwin and MinGW targets. It
26766 specifies that a GUI application is to be generated by
26767 instructing the linker to set the PE header subsystem type
26770 @item -fno-set-stack-executable
26771 @opindex fno-set-stack-executable
26772 This option is available for MinGW targets. It specifies that
26773 the executable flag for the stack used by nested functions isn't
26774 set. This is necessary for binaries running in kernel mode of
26775 Microsoft Windows, as there the User32 API, which is used to set executable
26776 privileges, isn't available.
26778 @item -fwritable-relocated-rdata
26779 @opindex fno-writable-relocated-rdata
26780 This option is available for MinGW and Cygwin targets. It specifies
26781 that relocated-data in read-only section is put into the @code{.data}
26782 section. This is a necessary for older runtimes not supporting
26783 modification of @code{.rdata} sections for pseudo-relocation.
26785 @item -mpe-aligned-commons
26786 @opindex mpe-aligned-commons
26787 This option is available for Cygwin and MinGW targets. It
26788 specifies that the GNU extension to the PE file format that
26789 permits the correct alignment of COMMON variables should be
26790 used when generating code. It is enabled by default if
26791 GCC detects that the target assembler found during configuration
26792 supports the feature.
26795 See also under @ref{x86 Options} for standard options.
26797 @node Xstormy16 Options
26798 @subsection Xstormy16 Options
26799 @cindex Xstormy16 Options
26801 These options are defined for Xstormy16:
26806 Choose startup files and linker script suitable for the simulator.
26809 @node Xtensa Options
26810 @subsection Xtensa Options
26811 @cindex Xtensa Options
26813 These options are supported for Xtensa targets:
26817 @itemx -mno-const16
26819 @opindex mno-const16
26820 Enable or disable use of @code{CONST16} instructions for loading
26821 constant values. The @code{CONST16} instruction is currently not a
26822 standard option from Tensilica. When enabled, @code{CONST16}
26823 instructions are always used in place of the standard @code{L32R}
26824 instructions. The use of @code{CONST16} is enabled by default only if
26825 the @code{L32R} instruction is not available.
26828 @itemx -mno-fused-madd
26829 @opindex mfused-madd
26830 @opindex mno-fused-madd
26831 Enable or disable use of fused multiply/add and multiply/subtract
26832 instructions in the floating-point option. This has no effect if the
26833 floating-point option is not also enabled. Disabling fused multiply/add
26834 and multiply/subtract instructions forces the compiler to use separate
26835 instructions for the multiply and add/subtract operations. This may be
26836 desirable in some cases where strict IEEE 754-compliant results are
26837 required: the fused multiply add/subtract instructions do not round the
26838 intermediate result, thereby producing results with @emph{more} bits of
26839 precision than specified by the IEEE standard. Disabling fused multiply
26840 add/subtract instructions also ensures that the program output is not
26841 sensitive to the compiler's ability to combine multiply and add/subtract
26844 @item -mserialize-volatile
26845 @itemx -mno-serialize-volatile
26846 @opindex mserialize-volatile
26847 @opindex mno-serialize-volatile
26848 When this option is enabled, GCC inserts @code{MEMW} instructions before
26849 @code{volatile} memory references to guarantee sequential consistency.
26850 The default is @option{-mserialize-volatile}. Use
26851 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
26853 @item -mforce-no-pic
26854 @opindex mforce-no-pic
26855 For targets, like GNU/Linux, where all user-mode Xtensa code must be
26856 position-independent code (PIC), this option disables PIC for compiling
26859 @item -mtext-section-literals
26860 @itemx -mno-text-section-literals
26861 @opindex mtext-section-literals
26862 @opindex mno-text-section-literals
26863 These options control the treatment of literal pools. The default is
26864 @option{-mno-text-section-literals}, which places literals in a separate
26865 section in the output file. This allows the literal pool to be placed
26866 in a data RAM/ROM, and it also allows the linker to combine literal
26867 pools from separate object files to remove redundant literals and
26868 improve code size. With @option{-mtext-section-literals}, the literals
26869 are interspersed in the text section in order to keep them as close as
26870 possible to their references. This may be necessary for large assembly
26871 files. Literals for each function are placed right before that function.
26873 @item -mauto-litpools
26874 @itemx -mno-auto-litpools
26875 @opindex mauto-litpools
26876 @opindex mno-auto-litpools
26877 These options control the treatment of literal pools. The default is
26878 @option{-mno-auto-litpools}, which places literals in a separate
26879 section in the output file unless @option{-mtext-section-literals} is
26880 used. With @option{-mauto-litpools} the literals are interspersed in
26881 the text section by the assembler. Compiler does not produce explicit
26882 @code{.literal} directives and loads literals into registers with
26883 @code{MOVI} instructions instead of @code{L32R} to let the assembler
26884 do relaxation and place literals as necessary. This option allows
26885 assembler to create several literal pools per function and assemble
26886 very big functions, which may not be possible with
26887 @option{-mtext-section-literals}.
26889 @item -mtarget-align
26890 @itemx -mno-target-align
26891 @opindex mtarget-align
26892 @opindex mno-target-align
26893 When this option is enabled, GCC instructs the assembler to
26894 automatically align instructions to reduce branch penalties at the
26895 expense of some code density. The assembler attempts to widen density
26896 instructions to align branch targets and the instructions following call
26897 instructions. If there are not enough preceding safe density
26898 instructions to align a target, no widening is performed. The
26899 default is @option{-mtarget-align}. These options do not affect the
26900 treatment of auto-aligned instructions like @code{LOOP}, which the
26901 assembler always aligns, either by widening density instructions or
26902 by inserting NOP instructions.
26905 @itemx -mno-longcalls
26906 @opindex mlongcalls
26907 @opindex mno-longcalls
26908 When this option is enabled, GCC instructs the assembler to translate
26909 direct calls to indirect calls unless it can determine that the target
26910 of a direct call is in the range allowed by the call instruction. This
26911 translation typically occurs for calls to functions in other source
26912 files. Specifically, the assembler translates a direct @code{CALL}
26913 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
26914 The default is @option{-mno-longcalls}. This option should be used in
26915 programs where the call target can potentially be out of range. This
26916 option is implemented in the assembler, not the compiler, so the
26917 assembly code generated by GCC still shows direct call
26918 instructions---look at the disassembled object code to see the actual
26919 instructions. Note that the assembler uses an indirect call for
26920 every cross-file call, not just those that really are out of range.
26923 @node zSeries Options
26924 @subsection zSeries Options
26925 @cindex zSeries options
26927 These are listed under @xref{S/390 and zSeries Options}.
26933 @section Specifying Subprocesses and the Switches to Pass to Them
26936 @command{gcc} is a driver program. It performs its job by invoking a
26937 sequence of other programs to do the work of compiling, assembling and
26938 linking. GCC interprets its command-line parameters and uses these to
26939 deduce which programs it should invoke, and which command-line options
26940 it ought to place on their command lines. This behavior is controlled
26941 by @dfn{spec strings}. In most cases there is one spec string for each
26942 program that GCC can invoke, but a few programs have multiple spec
26943 strings to control their behavior. The spec strings built into GCC can
26944 be overridden by using the @option{-specs=} command-line switch to specify
26947 @dfn{Spec files} are plain-text files that are used to construct spec
26948 strings. They consist of a sequence of directives separated by blank
26949 lines. The type of directive is determined by the first non-whitespace
26950 character on the line, which can be one of the following:
26953 @item %@var{command}
26954 Issues a @var{command} to the spec file processor. The commands that can
26958 @item %include <@var{file}>
26959 @cindex @code{%include}
26960 Search for @var{file} and insert its text at the current point in the
26963 @item %include_noerr <@var{file}>
26964 @cindex @code{%include_noerr}
26965 Just like @samp{%include}, but do not generate an error message if the include
26966 file cannot be found.
26968 @item %rename @var{old_name} @var{new_name}
26969 @cindex @code{%rename}
26970 Rename the spec string @var{old_name} to @var{new_name}.
26974 @item *[@var{spec_name}]:
26975 This tells the compiler to create, override or delete the named spec
26976 string. All lines after this directive up to the next directive or
26977 blank line are considered to be the text for the spec string. If this
26978 results in an empty string then the spec is deleted. (Or, if the
26979 spec did not exist, then nothing happens.) Otherwise, if the spec
26980 does not currently exist a new spec is created. If the spec does
26981 exist then its contents are overridden by the text of this
26982 directive, unless the first character of that text is the @samp{+}
26983 character, in which case the text is appended to the spec.
26985 @item [@var{suffix}]:
26986 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
26987 and up to the next directive or blank line are considered to make up the
26988 spec string for the indicated suffix. When the compiler encounters an
26989 input file with the named suffix, it processes the spec string in
26990 order to work out how to compile that file. For example:
26994 z-compile -input %i
26997 This says that any input file whose name ends in @samp{.ZZ} should be
26998 passed to the program @samp{z-compile}, which should be invoked with the
26999 command-line switch @option{-input} and with the result of performing the
27000 @samp{%i} substitution. (See below.)
27002 As an alternative to providing a spec string, the text following a
27003 suffix directive can be one of the following:
27006 @item @@@var{language}
27007 This says that the suffix is an alias for a known @var{language}. This is
27008 similar to using the @option{-x} command-line switch to GCC to specify a
27009 language explicitly. For example:
27016 Says that .ZZ files are, in fact, C++ source files.
27019 This causes an error messages saying:
27022 @var{name} compiler not installed on this system.
27026 GCC already has an extensive list of suffixes built into it.
27027 This directive adds an entry to the end of the list of suffixes, but
27028 since the list is searched from the end backwards, it is effectively
27029 possible to override earlier entries using this technique.
27033 GCC has the following spec strings built into it. Spec files can
27034 override these strings or create their own. Note that individual
27035 targets can also add their own spec strings to this list.
27038 asm Options to pass to the assembler
27039 asm_final Options to pass to the assembler post-processor
27040 cpp Options to pass to the C preprocessor
27041 cc1 Options to pass to the C compiler
27042 cc1plus Options to pass to the C++ compiler
27043 endfile Object files to include at the end of the link
27044 link Options to pass to the linker
27045 lib Libraries to include on the command line to the linker
27046 libgcc Decides which GCC support library to pass to the linker
27047 linker Sets the name of the linker
27048 predefines Defines to be passed to the C preprocessor
27049 signed_char Defines to pass to CPP to say whether @code{char} is signed
27051 startfile Object files to include at the start of the link
27054 Here is a small example of a spec file:
27057 %rename lib old_lib
27060 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
27063 This example renames the spec called @samp{lib} to @samp{old_lib} and
27064 then overrides the previous definition of @samp{lib} with a new one.
27065 The new definition adds in some extra command-line options before
27066 including the text of the old definition.
27068 @dfn{Spec strings} are a list of command-line options to be passed to their
27069 corresponding program. In addition, the spec strings can contain
27070 @samp{%}-prefixed sequences to substitute variable text or to
27071 conditionally insert text into the command line. Using these constructs
27072 it is possible to generate quite complex command lines.
27074 Here is a table of all defined @samp{%}-sequences for spec
27075 strings. Note that spaces are not generated automatically around the
27076 results of expanding these sequences. Therefore you can concatenate them
27077 together or combine them with constant text in a single argument.
27081 Substitute one @samp{%} into the program name or argument.
27084 Substitute the name of the input file being processed.
27087 Substitute the basename of the input file being processed.
27088 This is the substring up to (and not including) the last period
27089 and not including the directory.
27092 This is the same as @samp{%b}, but include the file suffix (text after
27096 Marks the argument containing or following the @samp{%d} as a
27097 temporary file name, so that that file is deleted if GCC exits
27098 successfully. Unlike @samp{%g}, this contributes no text to the
27101 @item %g@var{suffix}
27102 Substitute a file name that has suffix @var{suffix} and is chosen
27103 once per compilation, and mark the argument in the same way as
27104 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
27105 name is now chosen in a way that is hard to predict even when previously
27106 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
27107 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
27108 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
27109 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
27110 was simply substituted with a file name chosen once per compilation,
27111 without regard to any appended suffix (which was therefore treated
27112 just like ordinary text), making such attacks more likely to succeed.
27114 @item %u@var{suffix}
27115 Like @samp{%g}, but generates a new temporary file name
27116 each time it appears instead of once per compilation.
27118 @item %U@var{suffix}
27119 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
27120 new one if there is no such last file name. In the absence of any
27121 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
27122 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
27123 involves the generation of two distinct file names, one
27124 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
27125 simply substituted with a file name chosen for the previous @samp{%u},
27126 without regard to any appended suffix.
27128 @item %j@var{suffix}
27129 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
27130 writable, and if @option{-save-temps} is not used;
27131 otherwise, substitute the name
27132 of a temporary file, just like @samp{%u}. This temporary file is not
27133 meant for communication between processes, but rather as a junk
27134 disposal mechanism.
27136 @item %|@var{suffix}
27137 @itemx %m@var{suffix}
27138 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
27139 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
27140 all. These are the two most common ways to instruct a program that it
27141 should read from standard input or write to standard output. If you
27142 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
27143 construct: see for example @file{f/lang-specs.h}.
27145 @item %.@var{SUFFIX}
27146 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
27147 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
27148 terminated by the next space or %.
27151 Marks the argument containing or following the @samp{%w} as the
27152 designated output file of this compilation. This puts the argument
27153 into the sequence of arguments that @samp{%o} substitutes.
27156 Substitutes the names of all the output files, with spaces
27157 automatically placed around them. You should write spaces
27158 around the @samp{%o} as well or the results are undefined.
27159 @samp{%o} is for use in the specs for running the linker.
27160 Input files whose names have no recognized suffix are not compiled
27161 at all, but they are included among the output files, so they are
27165 Substitutes the suffix for object files. Note that this is
27166 handled specially when it immediately follows @samp{%g, %u, or %U},
27167 because of the need for those to form complete file names. The
27168 handling is such that @samp{%O} is treated exactly as if it had already
27169 been substituted, except that @samp{%g, %u, and %U} do not currently
27170 support additional @var{suffix} characters following @samp{%O} as they do
27171 following, for example, @samp{.o}.
27174 Substitutes the standard macro predefinitions for the
27175 current target machine. Use this when running @command{cpp}.
27178 Like @samp{%p}, but puts @samp{__} before and after the name of each
27179 predefined macro, except for macros that start with @samp{__} or with
27180 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
27184 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
27185 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
27186 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
27187 and @option{-imultilib} as necessary.
27190 Current argument is the name of a library or startup file of some sort.
27191 Search for that file in a standard list of directories and substitute
27192 the full name found. The current working directory is included in the
27193 list of directories scanned.
27196 Current argument is the name of a linker script. Search for that file
27197 in the current list of directories to scan for libraries. If the file
27198 is located insert a @option{--script} option into the command line
27199 followed by the full path name found. If the file is not found then
27200 generate an error message. Note: the current working directory is not
27204 Print @var{str} as an error message. @var{str} is terminated by a newline.
27205 Use this when inconsistent options are detected.
27207 @item %(@var{name})
27208 Substitute the contents of spec string @var{name} at this point.
27210 @item %x@{@var{option}@}
27211 Accumulate an option for @samp{%X}.
27214 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
27218 Output the accumulated assembler options specified by @option{-Wa}.
27221 Output the accumulated preprocessor options specified by @option{-Wp}.
27224 Process the @code{asm} spec. This is used to compute the
27225 switches to be passed to the assembler.
27228 Process the @code{asm_final} spec. This is a spec string for
27229 passing switches to an assembler post-processor, if such a program is
27233 Process the @code{link} spec. This is the spec for computing the
27234 command line passed to the linker. Typically it makes use of the
27235 @samp{%L %G %S %D and %E} sequences.
27238 Dump out a @option{-L} option for each directory that GCC believes might
27239 contain startup files. If the target supports multilibs then the
27240 current multilib directory is prepended to each of these paths.
27243 Process the @code{lib} spec. This is a spec string for deciding which
27244 libraries are included on the command line to the linker.
27247 Process the @code{libgcc} spec. This is a spec string for deciding
27248 which GCC support library is included on the command line to the linker.
27251 Process the @code{startfile} spec. This is a spec for deciding which
27252 object files are the first ones passed to the linker. Typically
27253 this might be a file named @file{crt0.o}.
27256 Process the @code{endfile} spec. This is a spec string that specifies
27257 the last object files that are passed to the linker.
27260 Process the @code{cpp} spec. This is used to construct the arguments
27261 to be passed to the C preprocessor.
27264 Process the @code{cc1} spec. This is used to construct the options to be
27265 passed to the actual C compiler (@command{cc1}).
27268 Process the @code{cc1plus} spec. This is used to construct the options to be
27269 passed to the actual C++ compiler (@command{cc1plus}).
27272 Substitute the variable part of a matched option. See below.
27273 Note that each comma in the substituted string is replaced by
27277 Remove all occurrences of @code{-S} from the command line. Note---this
27278 command is position dependent. @samp{%} commands in the spec string
27279 before this one see @code{-S}, @samp{%} commands in the spec string
27280 after this one do not.
27282 @item %:@var{function}(@var{args})
27283 Call the named function @var{function}, passing it @var{args}.
27284 @var{args} is first processed as a nested spec string, then split
27285 into an argument vector in the usual fashion. The function returns
27286 a string which is processed as if it had appeared literally as part
27287 of the current spec.
27289 The following built-in spec functions are provided:
27292 @item @code{getenv}
27293 The @code{getenv} spec function takes two arguments: an environment
27294 variable name and a string. If the environment variable is not
27295 defined, a fatal error is issued. Otherwise, the return value is the
27296 value of the environment variable concatenated with the string. For
27297 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
27300 %:getenv(TOPDIR /include)
27303 expands to @file{/path/to/top/include}.
27305 @item @code{if-exists}
27306 The @code{if-exists} spec function takes one argument, an absolute
27307 pathname to a file. If the file exists, @code{if-exists} returns the
27308 pathname. Here is a small example of its usage:
27312 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
27315 @item @code{if-exists-else}
27316 The @code{if-exists-else} spec function is similar to the @code{if-exists}
27317 spec function, except that it takes two arguments. The first argument is
27318 an absolute pathname to a file. If the file exists, @code{if-exists-else}
27319 returns the pathname. If it does not exist, it returns the second argument.
27320 This way, @code{if-exists-else} can be used to select one file or another,
27321 based on the existence of the first. Here is a small example of its usage:
27325 crt0%O%s %:if-exists(crti%O%s) \
27326 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
27329 @item @code{replace-outfile}
27330 The @code{replace-outfile} spec function takes two arguments. It looks for the
27331 first argument in the outfiles array and replaces it with the second argument. Here
27332 is a small example of its usage:
27335 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
27338 @item @code{remove-outfile}
27339 The @code{remove-outfile} spec function takes one argument. It looks for the
27340 first argument in the outfiles array and removes it. Here is a small example
27344 %:remove-outfile(-lm)
27347 @item @code{pass-through-libs}
27348 The @code{pass-through-libs} spec function takes any number of arguments. It
27349 finds any @option{-l} options and any non-options ending in @file{.a} (which it
27350 assumes are the names of linker input library archive files) and returns a
27351 result containing all the found arguments each prepended by
27352 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
27353 intended to be passed to the LTO linker plugin.
27356 %:pass-through-libs(%G %L %G)
27359 @item @code{print-asm-header}
27360 The @code{print-asm-header} function takes no arguments and simply
27361 prints a banner like:
27367 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
27370 It is used to separate compiler options from assembler options
27371 in the @option{--target-help} output.
27375 Substitutes the @code{-S} switch, if that switch is given to GCC@.
27376 If that switch is not specified, this substitutes nothing. Note that
27377 the leading dash is omitted when specifying this option, and it is
27378 automatically inserted if the substitution is performed. Thus the spec
27379 string @samp{%@{foo@}} matches the command-line option @option{-foo}
27380 and outputs the command-line option @option{-foo}.
27383 Like %@{@code{S}@} but mark last argument supplied within as a file to be
27384 deleted on failure.
27387 Substitutes all the switches specified to GCC whose names start
27388 with @code{-S}, but which also take an argument. This is used for
27389 switches like @option{-o}, @option{-D}, @option{-I}, etc.
27390 GCC considers @option{-o foo} as being
27391 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
27392 text, including the space. Thus two arguments are generated.
27395 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
27396 (the order of @code{S} and @code{T} in the spec is not significant).
27397 There can be any number of ampersand-separated variables; for each the
27398 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
27401 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
27404 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
27407 Substitutes @code{X} if one or more switches whose names start with
27408 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
27409 once, no matter how many such switches appeared. However, if @code{%*}
27410 appears somewhere in @code{X}, then @code{X} is substituted once
27411 for each matching switch, with the @code{%*} replaced by the part of
27412 that switch matching the @code{*}.
27414 If @code{%*} appears as the last part of a spec sequence then a space
27415 is added after the end of the last substitution. If there is more
27416 text in the sequence, however, then a space is not generated. This
27417 allows the @code{%*} substitution to be used as part of a larger
27418 string. For example, a spec string like this:
27421 %@{mcu=*:--script=%*/memory.ld@}
27425 when matching an option like @option{-mcu=newchip} produces:
27428 --script=newchip/memory.ld
27432 Substitutes @code{X}, if processing a file with suffix @code{S}.
27435 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
27438 Substitutes @code{X}, if processing a file for language @code{S}.
27441 Substitutes @code{X}, if not processing a file for language @code{S}.
27444 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
27445 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
27446 @code{*} sequences as well, although they have a stronger binding than
27447 the @samp{|}. If @code{%*} appears in @code{X}, all of the
27448 alternatives must be starred, and only the first matching alternative
27451 For example, a spec string like this:
27454 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
27458 outputs the following command-line options from the following input
27459 command-line options:
27464 -d fred.c -foo -baz -boggle
27465 -d jim.d -bar -baz -boggle
27468 @item %@{S:X; T:Y; :D@}
27470 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
27471 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
27472 be as many clauses as you need. This may be combined with @code{.},
27473 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
27478 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
27479 or similar construct can use a backslash to ignore the special meaning
27480 of the character following it, thus allowing literal matching of a
27481 character that is otherwise specially treated. For example,
27482 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
27483 @option{-std=iso9899:1999} option is given.
27485 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
27486 construct may contain other nested @samp{%} constructs or spaces, or
27487 even newlines. They are processed as usual, as described above.
27488 Trailing white space in @code{X} is ignored. White space may also
27489 appear anywhere on the left side of the colon in these constructs,
27490 except between @code{.} or @code{*} and the corresponding word.
27492 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
27493 handled specifically in these constructs. If another value of
27494 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
27495 @option{-W} switch is found later in the command line, the earlier
27496 switch value is ignored, except with @{@code{S}*@} where @code{S} is
27497 just one letter, which passes all matching options.
27499 The character @samp{|} at the beginning of the predicate text is used to
27500 indicate that a command should be piped to the following command, but
27501 only if @option{-pipe} is specified.
27503 It is built into GCC which switches take arguments and which do not.
27504 (You might think it would be useful to generalize this to allow each
27505 compiler's spec to say which switches take arguments. But this cannot
27506 be done in a consistent fashion. GCC cannot even decide which input
27507 files have been specified without knowing which switches take arguments,
27508 and it must know which input files to compile in order to tell which
27511 GCC also knows implicitly that arguments starting in @option{-l} are to be
27512 treated as compiler output files, and passed to the linker in their
27513 proper position among the other output files.
27515 @node Environment Variables
27516 @section Environment Variables Affecting GCC
27517 @cindex environment variables
27519 @c man begin ENVIRONMENT
27520 This section describes several environment variables that affect how GCC
27521 operates. Some of them work by specifying directories or prefixes to use
27522 when searching for various kinds of files. Some are used to specify other
27523 aspects of the compilation environment.
27525 Note that you can also specify places to search using options such as
27526 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
27527 take precedence over places specified using environment variables, which
27528 in turn take precedence over those specified by the configuration of GCC@.
27529 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
27530 GNU Compiler Collection (GCC) Internals}.
27535 @c @itemx LC_COLLATE
27537 @c @itemx LC_MONETARY
27538 @c @itemx LC_NUMERIC
27543 @c @findex LC_COLLATE
27544 @findex LC_MESSAGES
27545 @c @findex LC_MONETARY
27546 @c @findex LC_NUMERIC
27550 These environment variables control the way that GCC uses
27551 localization information which allows GCC to work with different
27552 national conventions. GCC inspects the locale categories
27553 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
27554 so. These locale categories can be set to any value supported by your
27555 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
27556 Kingdom encoded in UTF-8.
27558 The @env{LC_CTYPE} environment variable specifies character
27559 classification. GCC uses it to determine the character boundaries in
27560 a string; this is needed for some multibyte encodings that contain quote
27561 and escape characters that are otherwise interpreted as a string
27564 The @env{LC_MESSAGES} environment variable specifies the language to
27565 use in diagnostic messages.
27567 If the @env{LC_ALL} environment variable is set, it overrides the value
27568 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
27569 and @env{LC_MESSAGES} default to the value of the @env{LANG}
27570 environment variable. If none of these variables are set, GCC
27571 defaults to traditional C English behavior.
27575 If @env{TMPDIR} is set, it specifies the directory to use for temporary
27576 files. GCC uses temporary files to hold the output of one stage of
27577 compilation which is to be used as input to the next stage: for example,
27578 the output of the preprocessor, which is the input to the compiler
27581 @item GCC_COMPARE_DEBUG
27582 @findex GCC_COMPARE_DEBUG
27583 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
27584 @option{-fcompare-debug} to the compiler driver. See the documentation
27585 of this option for more details.
27587 @item GCC_EXEC_PREFIX
27588 @findex GCC_EXEC_PREFIX
27589 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
27590 names of the subprograms executed by the compiler. No slash is added
27591 when this prefix is combined with the name of a subprogram, but you can
27592 specify a prefix that ends with a slash if you wish.
27594 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
27595 an appropriate prefix to use based on the pathname it is invoked with.
27597 If GCC cannot find the subprogram using the specified prefix, it
27598 tries looking in the usual places for the subprogram.
27600 The default value of @env{GCC_EXEC_PREFIX} is
27601 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
27602 the installed compiler. In many cases @var{prefix} is the value
27603 of @code{prefix} when you ran the @file{configure} script.
27605 Other prefixes specified with @option{-B} take precedence over this prefix.
27607 This prefix is also used for finding files such as @file{crt0.o} that are
27610 In addition, the prefix is used in an unusual way in finding the
27611 directories to search for header files. For each of the standard
27612 directories whose name normally begins with @samp{/usr/local/lib/gcc}
27613 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
27614 replacing that beginning with the specified prefix to produce an
27615 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
27616 @file{foo/bar} just before it searches the standard directory
27617 @file{/usr/local/lib/bar}.
27618 If a standard directory begins with the configured
27619 @var{prefix} then the value of @var{prefix} is replaced by
27620 @env{GCC_EXEC_PREFIX} when looking for header files.
27622 @item COMPILER_PATH
27623 @findex COMPILER_PATH
27624 The value of @env{COMPILER_PATH} is a colon-separated list of
27625 directories, much like @env{PATH}. GCC tries the directories thus
27626 specified when searching for subprograms, if it cannot find the
27627 subprograms using @env{GCC_EXEC_PREFIX}.
27630 @findex LIBRARY_PATH
27631 The value of @env{LIBRARY_PATH} is a colon-separated list of
27632 directories, much like @env{PATH}. When configured as a native compiler,
27633 GCC tries the directories thus specified when searching for special
27634 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
27635 using GCC also uses these directories when searching for ordinary
27636 libraries for the @option{-l} option (but directories specified with
27637 @option{-L} come first).
27641 @cindex locale definition
27642 This variable is used to pass locale information to the compiler. One way in
27643 which this information is used is to determine the character set to be used
27644 when character literals, string literals and comments are parsed in C and C++.
27645 When the compiler is configured to allow multibyte characters,
27646 the following values for @env{LANG} are recognized:
27650 Recognize JIS characters.
27652 Recognize SJIS characters.
27654 Recognize EUCJP characters.
27657 If @env{LANG} is not defined, or if it has some other value, then the
27658 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
27659 recognize and translate multibyte characters.
27663 Some additional environment variables affect the behavior of the
27666 @include cppenv.texi
27670 @node Precompiled Headers
27671 @section Using Precompiled Headers
27672 @cindex precompiled headers
27673 @cindex speed of compilation
27675 Often large projects have many header files that are included in every
27676 source file. The time the compiler takes to process these header files
27677 over and over again can account for nearly all of the time required to
27678 build the project. To make builds faster, GCC allows you to
27679 @dfn{precompile} a header file.
27681 To create a precompiled header file, simply compile it as you would any
27682 other file, if necessary using the @option{-x} option to make the driver
27683 treat it as a C or C++ header file. You may want to use a
27684 tool like @command{make} to keep the precompiled header up-to-date when
27685 the headers it contains change.
27687 A precompiled header file is searched for when @code{#include} is
27688 seen in the compilation. As it searches for the included file
27689 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
27690 compiler looks for a precompiled header in each directory just before it
27691 looks for the include file in that directory. The name searched for is
27692 the name specified in the @code{#include} with @samp{.gch} appended. If
27693 the precompiled header file cannot be used, it is ignored.
27695 For instance, if you have @code{#include "all.h"}, and you have
27696 @file{all.h.gch} in the same directory as @file{all.h}, then the
27697 precompiled header file is used if possible, and the original
27698 header is used otherwise.
27700 Alternatively, you might decide to put the precompiled header file in a
27701 directory and use @option{-I} to ensure that directory is searched
27702 before (or instead of) the directory containing the original header.
27703 Then, if you want to check that the precompiled header file is always
27704 used, you can put a file of the same name as the original header in this
27705 directory containing an @code{#error} command.
27707 This also works with @option{-include}. So yet another way to use
27708 precompiled headers, good for projects not designed with precompiled
27709 header files in mind, is to simply take most of the header files used by
27710 a project, include them from another header file, precompile that header
27711 file, and @option{-include} the precompiled header. If the header files
27712 have guards against multiple inclusion, they are skipped because
27713 they've already been included (in the precompiled header).
27715 If you need to precompile the same header file for different
27716 languages, targets, or compiler options, you can instead make a
27717 @emph{directory} named like @file{all.h.gch}, and put each precompiled
27718 header in the directory, perhaps using @option{-o}. It doesn't matter
27719 what you call the files in the directory; every precompiled header in
27720 the directory is considered. The first precompiled header
27721 encountered in the directory that is valid for this compilation is
27722 used; they're searched in no particular order.
27724 There are many other possibilities, limited only by your imagination,
27725 good sense, and the constraints of your build system.
27727 A precompiled header file can be used only when these conditions apply:
27731 Only one precompiled header can be used in a particular compilation.
27734 A precompiled header cannot be used once the first C token is seen. You
27735 can have preprocessor directives before a precompiled header; you cannot
27736 include a precompiled header from inside another header.
27739 The precompiled header file must be produced for the same language as
27740 the current compilation. You cannot use a C precompiled header for a C++
27744 The precompiled header file must have been produced by the same compiler
27745 binary as the current compilation is using.
27748 Any macros defined before the precompiled header is included must
27749 either be defined in the same way as when the precompiled header was
27750 generated, or must not affect the precompiled header, which usually
27751 means that they don't appear in the precompiled header at all.
27753 The @option{-D} option is one way to define a macro before a
27754 precompiled header is included; using a @code{#define} can also do it.
27755 There are also some options that define macros implicitly, like
27756 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
27759 @item If debugging information is output when using the precompiled
27760 header, using @option{-g} or similar, the same kind of debugging information
27761 must have been output when building the precompiled header. However,
27762 a precompiled header built using @option{-g} can be used in a compilation
27763 when no debugging information is being output.
27765 @item The same @option{-m} options must generally be used when building
27766 and using the precompiled header. @xref{Submodel Options},
27767 for any cases where this rule is relaxed.
27769 @item Each of the following options must be the same when building and using
27770 the precompiled header:
27772 @gccoptlist{-fexceptions}
27775 Some other command-line options starting with @option{-f},
27776 @option{-p}, or @option{-O} must be defined in the same way as when
27777 the precompiled header was generated. At present, it's not clear
27778 which options are safe to change and which are not; the safest choice
27779 is to use exactly the same options when generating and using the
27780 precompiled header. The following are known to be safe:
27782 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
27783 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
27784 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
27789 For all of these except the last, the compiler automatically
27790 ignores the precompiled header if the conditions aren't met. If you
27791 find an option combination that doesn't work and doesn't cause the
27792 precompiled header to be ignored, please consider filing a bug report,
27795 If you do use differing options when generating and using the
27796 precompiled header, the actual behavior is a mixture of the
27797 behavior for the options. For instance, if you use @option{-g} to
27798 generate the precompiled header but not when using it, you may or may
27799 not get debugging information for routines in the precompiled header.