1 @c Copyright (C) 1988-2017 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1988-2017 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
126 @xref{Option Index}, for an index to GCC's options.
129 * Option Summary:: Brief list of all options, without explanations.
130 * Overall Options:: Controlling the kind of output:
131 an executable, object files, assembler files,
132 or preprocessed source.
133 * Invoking G++:: Compiling C++ programs.
134 * C Dialect Options:: Controlling the variant of C language compiled.
135 * C++ Dialect Options:: Variations on C++.
136 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
138 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
140 * Warning Options:: How picky should the compiler be?
141 * Debugging Options:: Producing debuggable code.
142 * Optimize Options:: How much optimization?
143 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
144 * Preprocessor Options:: Controlling header files and macro definitions.
145 Also, getting dependency information for Make.
146 * Assembler Options:: Passing options to the assembler.
147 * Link Options:: Specifying libraries and so on.
148 * Directory Options:: Where to find header files and libraries.
149 Where to find the compiler executable files.
150 * Code Gen Options:: Specifying conventions for function calls, data layout
152 * Developer Options:: Printing GCC configuration info, statistics, and
154 * Submodel Options:: Target-specific options, such as compiling for a
155 specific processor variant.
156 * Spec Files:: How to pass switches to sub-processes.
157 * Environment Variables:: Env vars that affect GCC.
158 * Precompiled Headers:: Compiling a header once, and using it many times.
164 @section Option Summary
166 Here is a summary of all the options, grouped by type. Explanations are
167 in the following sections.
170 @item Overall Options
171 @xref{Overall Options,,Options Controlling the Kind of Output}.
172 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
173 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
174 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
175 @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
176 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
178 @item C Language Options
179 @xref{C Dialect Options,,Options Controlling C Dialect}.
180 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
181 -fpermitted-flt-eval-methods=@var{standard} @gol
182 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
183 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
184 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
185 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
186 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
187 -fsigned-bitfields -fsigned-char @gol
188 -funsigned-bitfields -funsigned-char}
190 @item C++ Language Options
191 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
192 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
193 -faligned-new=@var{n} -fargs-in-order=@var{n} -fcheck-new @gol
194 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
195 -ffriend-injection @gol
196 -fno-elide-constructors @gol
197 -fno-enforce-eh-specs @gol
198 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
199 -fno-implicit-templates @gol
200 -fno-implicit-inline-templates @gol
201 -fno-implement-inlines -fms-extensions @gol
202 -fnew-inheriting-ctors @gol
203 -fnew-ttp-matching @gol
204 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
205 -fno-optional-diags -fpermissive @gol
206 -fno-pretty-templates @gol
207 -frepo -fno-rtti -fsized-deallocation @gol
208 -ftemplate-backtrace-limit=@var{n} @gol
209 -ftemplate-depth=@var{n} @gol
210 -fno-threadsafe-statics -fuse-cxa-atexit @gol
211 -fno-weak -nostdinc++ @gol
212 -fvisibility-inlines-hidden @gol
213 -fvisibility-ms-compat @gol
214 -fext-numeric-literals @gol
215 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
216 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
217 -Wnamespaces -Wnarrowing @gol
218 -Wnoexcept -Wnoexcept-type -Wnon-virtual-dtor -Wreorder -Wregister @gol
219 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
220 -Wno-non-template-friend -Wold-style-cast @gol
221 -Woverloaded-virtual -Wno-pmf-conversions @gol
222 -Wsign-promo -Wvirtual-inheritance}
224 @item Objective-C and Objective-C++ Language Options
225 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
226 Objective-C and Objective-C++ Dialects}.
227 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
228 -fgnu-runtime -fnext-runtime @gol
229 -fno-nil-receivers @gol
230 -fobjc-abi-version=@var{n} @gol
231 -fobjc-call-cxx-cdtors @gol
232 -fobjc-direct-dispatch @gol
233 -fobjc-exceptions @gol
236 -fobjc-std=objc1 @gol
237 -fno-local-ivars @gol
238 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
239 -freplace-objc-classes @gol
242 -Wassign-intercept @gol
243 -Wno-protocol -Wselector @gol
244 -Wstrict-selector-match @gol
245 -Wundeclared-selector}
247 @item Diagnostic Message Formatting Options
248 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
249 @gccoptlist{-fmessage-length=@var{n} @gol
250 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
251 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
252 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
253 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
256 @item Warning Options
257 @xref{Warning Options,,Options to Request or Suppress Warnings}.
258 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
259 -pedantic-errors @gol
260 -w -Wextra -Wall -Waddress -Waggregate-return @gol
261 -Walloc-zero -Walloc-size-larger-than=@var{n}
262 -Walloca -Walloca-larger-than=@var{n} @gol
263 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
264 -Wno-attributes -Wbool-compare -Wbool-operation @gol
265 -Wno-builtin-declaration-mismatch @gol
266 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
267 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
268 -Wchar-subscripts -Wchkp -Wclobbered -Wcomment @gol
269 -Wconditionally-supported @gol
270 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
271 -Wdelete-incomplete @gol
272 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
273 -Wdisabled-optimization @gol
274 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
275 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
276 -Wempty-body -Wenum-compare -Wno-endif-labels -Wexpansion-to-defined @gol
277 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
278 -Wfloat-equal -Wformat -Wformat=2 @gol
279 -Wno-format-contains-nul -Wno-format-extra-args @gol
280 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
281 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
282 -Wformat-y2k -Wframe-address @gol
283 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
284 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
285 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
286 -Wimplicit-function-declaration -Wimplicit-int @gol
287 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
288 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
289 -Winvalid-pch -Wlarger-than=@var{len} @gol
290 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
291 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
292 -Wmisleading-indentation -Wmissing-braces @gol
293 -Wmissing-field-initializers -Wmissing-include-dirs @gol
294 -Wno-multichar -Wnonnull -Wnonnull-compare @gol
295 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
296 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
297 -Woverride-init-side-effects -Woverlength-strings @gol
298 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
299 -Wparentheses -Wno-pedantic-ms-format @gol
300 -Wplacement-new -Wplacement-new=@var{n} @gol
301 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
302 -Wno-pragmas -Wredundant-decls -Wrestrict -Wno-return-local-addr @gol
303 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
304 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
305 -Wshift-overflow -Wshift-overflow=@var{n} @gol
306 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
307 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
308 -Wno-scalar-storage-order @gol
309 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
310 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
311 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
312 -Wstringop-overflow=@var{n} @gol
313 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
314 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
315 -Wmissing-format-attribute -Wsubobject-linkage @gol
316 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
317 -Wswitch-unreachable -Wsync-nand @gol
318 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
319 -Wtype-limits -Wundef @gol
320 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
321 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
322 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
323 -Wunused-parameter -Wno-unused-result @gol
324 -Wunused-value -Wunused-variable @gol
325 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
326 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
327 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
328 -Wvla -Wvla-larger-than=@var{n} -Wvolatile-register-var -Wwrite-strings @gol
329 -Wzero-as-null-pointer-constant -Whsa}
331 @item C and Objective-C-only Warning Options
332 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
333 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
334 -Wold-style-declaration -Wold-style-definition @gol
335 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
336 -Wdeclaration-after-statement -Wpointer-sign}
338 @item Debugging Options
339 @xref{Debugging Options,,Options for Debugging Your Program}.
340 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
341 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
342 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
343 -gcolumn-info -gno-column-info @gol
344 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
345 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
346 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
347 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
348 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
349 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
350 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
351 -fvar-tracking -fvar-tracking-assignments}
353 @item Optimization Options
354 @xref{Optimize Options,,Options that Control Optimization}.
355 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
356 -falign-jumps[=@var{n}] @gol
357 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
358 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
359 -fauto-inc-dec -fbranch-probabilities @gol
360 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
361 -fbtr-bb-exclusive -fcaller-saves @gol
362 -fcombine-stack-adjustments -fconserve-stack @gol
363 -fcompare-elim -fcprop-registers -fcrossjumping @gol
364 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
365 -fcx-limited-range @gol
366 -fdata-sections -fdce -fdelayed-branch @gol
367 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
368 -fdevirtualize-at-ltrans -fdse @gol
369 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
370 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
371 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
372 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
373 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
374 -fif-conversion2 -findirect-inlining @gol
375 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
376 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
377 -fipa-bit-cp -fipa-vrp @gol
378 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
379 -fira-algorithm=@var{algorithm} @gol
380 -fira-region=@var{region} -fira-hoist-pressure @gol
381 -fira-loop-pressure -fno-ira-share-save-slots @gol
382 -fno-ira-share-spill-slots @gol
383 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
384 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
385 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
386 -floop-block -floop-interchange -floop-strip-mine @gol
387 -floop-unroll-and-jam -floop-nest-optimize @gol
388 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
389 -flto-partition=@var{alg} -fmerge-all-constants @gol
390 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
391 -fmove-loop-invariants -fno-branch-count-reg @gol
392 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
393 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
394 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
395 -fno-sched-spec -fno-signed-zeros @gol
396 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
397 -fomit-frame-pointer -foptimize-sibling-calls @gol
398 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
399 -fprefetch-loop-arrays @gol
400 -fprofile-correction @gol
401 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
402 -fprofile-reorder-functions @gol
403 -freciprocal-math -free -frename-registers -freorder-blocks @gol
404 -freorder-blocks-algorithm=@var{algorithm} @gol
405 -freorder-blocks-and-partition -freorder-functions @gol
406 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
407 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
408 -fsched-spec-load -fsched-spec-load-dangerous @gol
409 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
410 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
411 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
412 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
413 -fschedule-fusion @gol
414 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
415 -fselective-scheduling -fselective-scheduling2 @gol
416 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
417 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
418 -fsignaling-nans @gol
419 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
421 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
422 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
423 -fthread-jumps -ftracer -ftree-bit-ccp @gol
424 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
425 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
426 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
427 -ftree-loop-if-convert -ftree-loop-im @gol
428 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
429 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
430 -ftree-loop-vectorize @gol
431 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
432 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
433 -ftree-switch-conversion -ftree-tail-merge @gol
434 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
435 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
436 -funsafe-math-optimizations -funswitch-loops @gol
437 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
438 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
439 --param @var{name}=@var{value}
440 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
442 @item Program Instrumentation Options
443 @xref{Instrumentation Options,,Program Instrumentation Options}.
444 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
445 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
446 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
447 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
448 -fsanitize-undefined-trap-on-error -fbounds-check @gol
449 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
450 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
451 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
452 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
453 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
454 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
455 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
456 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
457 -fchkp-use-wrappers -fchkp-flexible-struct-trailing-arrays@gol
458 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
459 -fstack-protector-explicit -fstack-check @gol
460 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
461 -fno-stack-limit -fsplit-stack @gol
462 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
463 -fvtv-counts -fvtv-debug @gol
464 -finstrument-functions @gol
465 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
466 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
468 @item Preprocessor Options
469 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
470 @gccoptlist{-A@var{question}=@var{answer} @gol
471 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
472 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
473 -dD -dI -dM -dN -dU @gol
474 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
475 -fexec-charset=@var{charset} -fextended-identifiers @gol
476 -finput-charset=@var{charset} -fno-canonical-system-headers @gol
477 -fpch-deps -fpch-preprocess -fpreprocessed @gol
478 -ftabstop=@var{width} -ftrack-macro-expansion @gol
479 -fwide-exec-charset=@var{charset} -fworking-directory @gol
480 -H -imacros @var{file} -include @var{file} @gol
481 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
482 -no-integrated-cpp -P -pthread -remap @gol
483 -traditional -traditional-cpp -trigraphs @gol
484 -U@var{macro} -undef @gol
485 -Wp,@var{option} -Xpreprocessor @var{option}}
487 @item Assembler Options
488 @xref{Assembler Options,,Passing Options to the Assembler}.
489 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
492 @xref{Link Options,,Options for Linking}.
493 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
494 -nostartfiles -nodefaultlibs -nostdlib -pie -pthread -rdynamic @gol
495 -s -static -static-libgcc -static-libstdc++ @gol
496 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
497 -static-libmpx -static-libmpxwrappers @gol
498 -shared -shared-libgcc -symbolic @gol
499 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
500 -u @var{symbol} -z @var{keyword}}
502 @item Directory Options
503 @xref{Directory Options,,Options for Directory Search}.
504 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
505 -idirafter @var{dir} @gol
506 -imacros @var{file} -imultilib @var{dir} @gol
507 -iplugindir=@var{dir} -iprefix @var{file} @gol
508 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
509 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
510 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
511 -nostdinc -nostdinc++ --sysroot=@var{dir}}
513 @item Code Generation Options
514 @xref{Code Gen Options,,Options for Code Generation Conventions}.
515 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
516 -ffixed-@var{reg} -fexceptions @gol
517 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
518 -fasynchronous-unwind-tables @gol
520 -finhibit-size-directive -fno-common -fno-ident @gol
521 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
522 -fno-jump-tables @gol
523 -frecord-gcc-switches @gol
524 -freg-struct-return -fshort-enums -fshort-wchar @gol
525 -fverbose-asm -fpack-struct[=@var{n}] @gol
526 -fleading-underscore -ftls-model=@var{model} @gol
527 -fstack-reuse=@var{reuse_level} @gol
528 -ftrampolines -ftrapv -fwrapv @gol
529 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
530 -fstrict-volatile-bitfields -fsync-libcalls}
532 @item Developer Options
533 @xref{Developer Options,,GCC Developer Options}.
534 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
535 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
536 -fdbg-cnt=@var{counter-value-list} @gol
537 -fdisable-ipa-@var{pass_name} @gol
538 -fdisable-rtl-@var{pass_name} @gol
539 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
540 -fdisable-tree-@var{pass_name} @gol
541 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
542 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
543 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
544 -fdump-final-insns@r{[}=@var{file}@r{]}
545 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
547 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
548 -fdump-statistics @gol
549 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
551 -fdump-tree-@var{switch} @gol
552 -fdump-tree-@var{switch}-@var{options} @gol
553 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
554 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
555 -fenable-@var{kind}-@var{pass} @gol
556 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
557 -fira-verbose=@var{n} @gol
558 -flto-report -flto-report-wpa -fmem-report-wpa @gol
559 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
560 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
561 -fprofile-report @gol
562 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
563 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
564 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
565 -fvar-tracking-assignments-toggle -gtoggle @gol
566 -print-file-name=@var{library} -print-libgcc-file-name @gol
567 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
568 -print-prog-name=@var{program} -print-search-dirs -Q @gol
569 -print-sysroot -print-sysroot-headers-suffix @gol
570 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
572 @item Machine-Dependent Options
573 @xref{Submodel Options,,Machine-Dependent Options}.
574 @c This list is ordered alphanumerically by subsection name.
575 @c Try and put the significant identifier (CPU or system) first,
576 @c so users have a clue at guessing where the ones they want will be.
578 @emph{AArch64 Options}
579 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
580 -mgeneral-regs-only @gol
581 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
583 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
584 -mtls-dialect=desc -mtls-dialect=traditional @gol
585 -mtls-size=@var{size} @gol
586 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
587 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
588 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
589 -mlow-precision-sqrt -mno-low-precision-sqrt@gol
590 -mlow-precision-div -mno-low-precision-div @gol
591 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
593 @emph{Adapteva Epiphany Options}
594 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
595 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
596 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
597 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
598 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
599 -msplit-vecmove-early -m1reg-@var{reg}}
602 @gccoptlist{-mbarrel-shifter @gol
603 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
604 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
605 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
606 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
607 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
608 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
609 -mlong-calls -mmedium-calls -msdata @gol
610 -mvolatile-cache -mtp-regno=@var{regno} @gol
611 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
612 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
613 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
614 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
615 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
616 -mtune=@var{cpu} -mmultcost=@var{num} @gol
617 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
618 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
621 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
622 -mabi=@var{name} @gol
623 -mapcs-stack-check -mno-apcs-stack-check @gol
624 -mapcs-reentrant -mno-apcs-reentrant @gol
625 -msched-prolog -mno-sched-prolog @gol
626 -mlittle-endian -mbig-endian @gol
627 -mfloat-abi=@var{name} @gol
628 -mfp16-format=@var{name}
629 -mthumb-interwork -mno-thumb-interwork @gol
630 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
631 -mtune=@var{name} -mprint-tune-info @gol
632 -mstructure-size-boundary=@var{n} @gol
633 -mabort-on-noreturn @gol
634 -mlong-calls -mno-long-calls @gol
635 -msingle-pic-base -mno-single-pic-base @gol
636 -mpic-register=@var{reg} @gol
637 -mnop-fun-dllimport @gol
638 -mpoke-function-name @gol
640 -mtpcs-frame -mtpcs-leaf-frame @gol
641 -mcaller-super-interworking -mcallee-super-interworking @gol
642 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
643 -mword-relocations @gol
644 -mfix-cortex-m3-ldrd @gol
645 -munaligned-access @gol
646 -mneon-for-64bits @gol
647 -mslow-flash-data @gol
648 -masm-syntax-unified @gol
654 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
655 -mbranch-cost=@var{cost} @gol
656 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
657 -mrelax -mrmw -mstrict-X -mtiny-stack -mfract-convert-truncate @gol
659 -Waddr-space-convert -Wmisspelled-isr}
661 @emph{Blackfin Options}
662 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
663 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
664 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
665 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
666 -mno-id-shared-library -mshared-library-id=@var{n} @gol
667 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
668 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
669 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
673 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
674 -msim -msdata=@var{sdata-type}}
677 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
678 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
679 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
680 -mstack-align -mdata-align -mconst-align @gol
681 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
682 -melf -maout -melinux -mlinux -sim -sim2 @gol
683 -mmul-bug-workaround -mno-mul-bug-workaround}
686 @gccoptlist{-mmac @gol
687 -mcr16cplus -mcr16c @gol
688 -msim -mint32 -mbit-ops
689 -mdata-model=@var{model}}
691 @emph{Darwin Options}
692 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
693 -arch_only -bind_at_load -bundle -bundle_loader @gol
694 -client_name -compatibility_version -current_version @gol
696 -dependency-file -dylib_file -dylinker_install_name @gol
697 -dynamic -dynamiclib -exported_symbols_list @gol
698 -filelist -flat_namespace -force_cpusubtype_ALL @gol
699 -force_flat_namespace -headerpad_max_install_names @gol
701 -image_base -init -install_name -keep_private_externs @gol
702 -multi_module -multiply_defined -multiply_defined_unused @gol
703 -noall_load -no_dead_strip_inits_and_terms @gol
704 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
705 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
706 -private_bundle -read_only_relocs -sectalign @gol
707 -sectobjectsymbols -whyload -seg1addr @gol
708 -sectcreate -sectobjectsymbols -sectorder @gol
709 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
710 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
711 -segprot -segs_read_only_addr -segs_read_write_addr @gol
712 -single_module -static -sub_library -sub_umbrella @gol
713 -twolevel_namespace -umbrella -undefined @gol
714 -unexported_symbols_list -weak_reference_mismatches @gol
715 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
716 -mkernel -mone-byte-bool}
718 @emph{DEC Alpha Options}
719 @gccoptlist{-mno-fp-regs -msoft-float @gol
720 -mieee -mieee-with-inexact -mieee-conformant @gol
721 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
722 -mtrap-precision=@var{mode} -mbuild-constants @gol
723 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
724 -mbwx -mmax -mfix -mcix @gol
725 -mfloat-vax -mfloat-ieee @gol
726 -mexplicit-relocs -msmall-data -mlarge-data @gol
727 -msmall-text -mlarge-text @gol
728 -mmemory-latency=@var{time}}
731 @gccoptlist{-msmall-model -mno-lsim}
734 @gccoptlist{-msim -mlra -mnodiv}
737 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
738 -mhard-float -msoft-float @gol
739 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
740 -mdouble -mno-double @gol
741 -mmedia -mno-media -mmuladd -mno-muladd @gol
742 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
743 -mlinked-fp -mlong-calls -malign-labels @gol
744 -mlibrary-pic -macc-4 -macc-8 @gol
745 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
746 -moptimize-membar -mno-optimize-membar @gol
747 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
748 -mvliw-branch -mno-vliw-branch @gol
749 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
750 -mno-nested-cond-exec -mtomcat-stats @gol
754 @emph{GNU/Linux Options}
755 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
756 -tno-android-cc -tno-android-ld}
758 @emph{H8/300 Options}
759 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
762 @gccoptlist{-march=@var{architecture-type} @gol
763 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
764 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
765 -mfixed-range=@var{register-range} @gol
766 -mjump-in-delay -mlinker-opt -mlong-calls @gol
767 -mlong-load-store -mno-disable-fpregs @gol
768 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
769 -mno-jump-in-delay -mno-long-load-store @gol
770 -mno-portable-runtime -mno-soft-float @gol
771 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
772 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
773 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
774 -munix=@var{unix-std} -nolibdld -static -threads}
777 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
778 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
779 -mconstant-gp -mauto-pic -mfused-madd @gol
780 -minline-float-divide-min-latency @gol
781 -minline-float-divide-max-throughput @gol
782 -mno-inline-float-divide @gol
783 -minline-int-divide-min-latency @gol
784 -minline-int-divide-max-throughput @gol
785 -mno-inline-int-divide @gol
786 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
787 -mno-inline-sqrt @gol
788 -mdwarf2-asm -mearly-stop-bits @gol
789 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
790 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
791 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
792 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
793 -msched-spec-ldc -msched-spec-control-ldc @gol
794 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
795 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
796 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
797 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
800 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
801 -msign-extend-enabled -muser-enabled}
803 @emph{M32R/D Options}
804 @gccoptlist{-m32r2 -m32rx -m32r @gol
806 -malign-loops -mno-align-loops @gol
807 -missue-rate=@var{number} @gol
808 -mbranch-cost=@var{number} @gol
809 -mmodel=@var{code-size-model-type} @gol
810 -msdata=@var{sdata-type} @gol
811 -mno-flush-func -mflush-func=@var{name} @gol
812 -mno-flush-trap -mflush-trap=@var{number} @gol
816 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
818 @emph{M680x0 Options}
819 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
820 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
821 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
822 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
823 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
824 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
825 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
826 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
827 -mxgot -mno-xgot -mlong-jump-table-offsets}
830 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
831 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
832 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
833 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
834 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
837 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
838 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
839 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
840 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
843 @emph{MicroBlaze Options}
844 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
845 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
846 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
847 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
848 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
851 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
852 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
853 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
854 -mips16 -mno-mips16 -mflip-mips16 @gol
855 -minterlink-compressed -mno-interlink-compressed @gol
856 -minterlink-mips16 -mno-interlink-mips16 @gol
857 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
858 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
859 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
860 -mno-float -msingle-float -mdouble-float @gol
861 -modd-spreg -mno-odd-spreg @gol
862 -mabs=@var{mode} -mnan=@var{encoding} @gol
863 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
866 -mvirt -mno-virt @gol
868 -mmicromips -mno-micromips @gol
870 -mfpu=@var{fpu-type} @gol
871 -msmartmips -mno-smartmips @gol
872 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
873 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
874 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
875 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
876 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
877 -membedded-data -mno-embedded-data @gol
878 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
879 -mcode-readable=@var{setting} @gol
880 -msplit-addresses -mno-split-addresses @gol
881 -mexplicit-relocs -mno-explicit-relocs @gol
882 -mcheck-zero-division -mno-check-zero-division @gol
883 -mdivide-traps -mdivide-breaks @gol
884 -mload-store-pairs -mno-load-store-pairs @gol
885 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
886 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
887 -mfix-24k -mno-fix-24k @gol
888 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
889 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
890 -mfix-vr4120 -mno-fix-vr4120 @gol
891 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
892 -mflush-func=@var{func} -mno-flush-func @gol
893 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
894 -mcompact-branches=@var{policy} @gol
895 -mfp-exceptions -mno-fp-exceptions @gol
896 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
897 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
898 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
899 -mframe-header-opt -mno-frame-header-opt}
902 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
903 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
904 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
905 -mno-base-addresses -msingle-exit -mno-single-exit}
907 @emph{MN10300 Options}
908 @gccoptlist{-mmult-bug -mno-mult-bug @gol
909 -mno-am33 -mam33 -mam33-2 -mam34 @gol
910 -mtune=@var{cpu-type} @gol
911 -mreturn-pointer-on-d0 @gol
912 -mno-crt0 -mrelax -mliw -msetlb}
915 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
917 @emph{MSP430 Options}
918 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
920 -mcode-region= -mdata-region= @gol
921 -msilicon-errata= -msilicon-errata-warn= @gol
925 @gccoptlist{-mbig-endian -mlittle-endian @gol
926 -mreduced-regs -mfull-regs @gol
927 -mcmov -mno-cmov @gol
928 -mperf-ext -mno-perf-ext @gol
929 -mv3push -mno-v3push @gol
930 -m16bit -mno-16bit @gol
931 -misr-vector-size=@var{num} @gol
932 -mcache-block-size=@var{num} @gol
933 -march=@var{arch} @gol
934 -mcmodel=@var{code-model} @gol
937 @emph{Nios II Options}
938 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
940 -mno-bypass-cache -mbypass-cache @gol
941 -mno-cache-volatile -mcache-volatile @gol
942 -mno-fast-sw-div -mfast-sw-div @gol
943 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
944 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
945 -mcustom-fpu-cfg=@var{name} @gol
946 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
947 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
949 @emph{Nvidia PTX Options}
950 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
952 @emph{PDP-11 Options}
953 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
954 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
955 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
956 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
957 -mbranch-expensive -mbranch-cheap @gol
958 -munix-asm -mdec-asm}
960 @emph{picoChip Options}
961 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
962 -msymbol-as-address -mno-inefficient-warnings}
964 @emph{PowerPC Options}
965 See RS/6000 and PowerPC Options.
967 @emph{RISC-V Options}
968 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
969 -mmemcpy -mno-memcpy @gol
971 -mabi=@var{ABI-string} @gol
972 -mfdiv -mno-fdiv @gol
974 -march=@var{ISA-string} @gol
975 -mtune=@var{processor-string} @gol
976 -msmall-data-limit=@var{N-bytes} @gol
977 -msave-restore -mno-save-restore @gol
978 -mstrict-align -mno-strict-align @gol
979 -mcmodel=@var{code-model} @gol
980 -mexplicit-relocs -mno-explicit-relocs @gol}
983 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
984 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
985 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
987 @emph{RS/6000 and PowerPC Options}
988 @gccoptlist{-mcpu=@var{cpu-type} @gol
989 -mtune=@var{cpu-type} @gol
990 -mcmodel=@var{code-model} @gol
992 -maltivec -mno-altivec @gol
993 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
994 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
995 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
996 -mfprnd -mno-fprnd @gol
997 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
998 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
999 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1000 -malign-power -malign-natural @gol
1001 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1002 -msingle-float -mdouble-float -msimple-fpu @gol
1003 -mstring -mno-string -mupdate -mno-update @gol
1004 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1005 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1006 -mstrict-align -mno-strict-align -mrelocatable @gol
1007 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1008 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1009 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
1010 -mprioritize-restricted-insns=@var{priority} @gol
1011 -msched-costly-dep=@var{dependence_type} @gol
1012 -minsert-sched-nops=@var{scheme} @gol
1013 -mcall-sysv -mcall-netbsd @gol
1014 -maix-struct-return -msvr4-struct-return @gol
1015 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1016 -mblock-move-inline-limit=@var{num} @gol
1017 -misel -mno-isel @gol
1018 -misel=yes -misel=no @gol
1020 -mspe=yes -mspe=no @gol
1022 -mgen-cell-microcode -mwarn-cell-microcode @gol
1023 -mvrsave -mno-vrsave @gol
1024 -mmulhw -mno-mulhw @gol
1025 -mdlmzb -mno-dlmzb @gol
1026 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1027 -mprototype -mno-prototype @gol
1028 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1029 -msdata=@var{opt} -mvxworks -G @var{num} @gol
1030 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1031 -mno-recip-precision @gol
1032 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1033 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1034 -msave-toc-indirect -mno-save-toc-indirect @gol
1035 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1036 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1037 -mquad-memory -mno-quad-memory @gol
1038 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1039 -mcompat-align-parm -mno-compat-align-parm @gol
1040 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1041 -mupper-regs-di -mno-upper-regs-di @gol
1042 -mupper-regs -mno-upper-regs @gol
1043 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1044 -mgnu-attribute -mno-gnu-attribute @gol
1045 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1046 -mstack-protector-guard-offset=@var{offset} @gol
1050 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1052 -mbig-endian-data -mlittle-endian-data @gol
1055 -mas100-syntax -mno-as100-syntax@gol
1057 -mmax-constant-size=@gol
1060 -mallow-string-insns -mno-allow-string-insns@gol
1062 -mno-warn-multiple-fast-interrupts@gol
1063 -msave-acc-in-interrupts}
1065 @emph{S/390 and zSeries Options}
1066 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1067 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1068 -mlong-double-64 -mlong-double-128 @gol
1069 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1070 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1071 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1072 -mhtm -mvx -mzvector @gol
1073 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1074 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1075 -mhotpatch=@var{halfwords},@var{halfwords}}
1077 @emph{Score Options}
1078 @gccoptlist{-meb -mel @gol
1082 -mscore5 -mscore5u -mscore7 -mscore7d}
1085 @gccoptlist{-m1 -m2 -m2e @gol
1086 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1088 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1089 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1090 -mb -ml -mdalign -mrelax @gol
1091 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1092 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1093 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1094 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1095 -maccumulate-outgoing-args @gol
1096 -matomic-model=@var{atomic-model} @gol
1097 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1098 -mcbranch-force-delay-slot @gol
1099 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1100 -mpretend-cmove -mtas}
1102 @emph{Solaris 2 Options}
1103 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1106 @emph{SPARC Options}
1107 @gccoptlist{-mcpu=@var{cpu-type} @gol
1108 -mtune=@var{cpu-type} @gol
1109 -mcmodel=@var{code-model} @gol
1110 -mmemory-model=@var{mem-model} @gol
1111 -m32 -m64 -mapp-regs -mno-app-regs @gol
1112 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1113 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1114 -mhard-quad-float -msoft-quad-float @gol
1115 -mstack-bias -mno-stack-bias @gol
1116 -mstd-struct-return -mno-std-struct-return @gol
1117 -munaligned-doubles -mno-unaligned-doubles @gol
1118 -muser-mode -mno-user-mode @gol
1119 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1120 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1121 -mcbcond -mno-cbcond -mfmaf -mno-fmaf @gol
1122 -mpopc -mno-popc -msubxc -mno-subxc@gol
1123 -mfix-at697f -mfix-ut699 @gol
1127 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1128 -msafe-dma -munsafe-dma @gol
1130 -msmall-mem -mlarge-mem -mstdmain @gol
1131 -mfixed-range=@var{register-range} @gol
1133 -maddress-space-conversion -mno-address-space-conversion @gol
1134 -mcache-size=@var{cache-size} @gol
1135 -matomic-updates -mno-atomic-updates}
1137 @emph{System V Options}
1138 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1140 @emph{TILE-Gx Options}
1141 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1142 -mcmodel=@var{code-model}}
1144 @emph{TILEPro Options}
1145 @gccoptlist{-mcpu=@var{cpu} -m32}
1148 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1149 -mprolog-function -mno-prolog-function -mspace @gol
1150 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1151 -mapp-regs -mno-app-regs @gol
1152 -mdisable-callt -mno-disable-callt @gol
1153 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1154 -mv850e -mv850 -mv850e3v5 @gol
1165 @gccoptlist{-mg -mgnu -munix}
1167 @emph{Visium Options}
1168 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1169 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1172 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1173 -mpointer-size=@var{size}}
1175 @emph{VxWorks Options}
1176 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1177 -Xbind-lazy -Xbind-now}
1180 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1181 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1182 -mfpmath=@var{unit} @gol
1183 -masm=@var{dialect} -mno-fancy-math-387 @gol
1184 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1185 -mno-wide-multiply -mrtd -malign-double @gol
1186 -mpreferred-stack-boundary=@var{num} @gol
1187 -mincoming-stack-boundary=@var{num} @gol
1188 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1189 -mrecip -mrecip=@var{opt} @gol
1190 -mvzeroupper -mprefer-avx128 @gol
1191 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1192 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1193 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1194 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1195 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1196 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1197 -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx @gol
1198 -mmwaitx -mclzero -mpku -mthreads @gol
1199 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1200 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1201 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1202 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1203 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1204 -mregparm=@var{num} -msseregparm @gol
1205 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1206 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1207 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1208 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1209 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1210 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1211 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1212 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1213 -mmitigate-rop -mgeneral-regs-only}
1215 @emph{x86 Windows Options}
1216 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1217 -mnop-fun-dllimport -mthread @gol
1218 -municode -mwin32 -mwindows -fno-set-stack-executable}
1220 @emph{Xstormy16 Options}
1223 @emph{Xtensa Options}
1224 @gccoptlist{-mconst16 -mno-const16 @gol
1225 -mfused-madd -mno-fused-madd @gol
1227 -mserialize-volatile -mno-serialize-volatile @gol
1228 -mtext-section-literals -mno-text-section-literals @gol
1229 -mauto-litpools -mno-auto-litpools @gol
1230 -mtarget-align -mno-target-align @gol
1231 -mlongcalls -mno-longcalls}
1233 @emph{zSeries Options}
1234 See S/390 and zSeries Options.
1238 @node Overall Options
1239 @section Options Controlling the Kind of Output
1241 Compilation can involve up to four stages: preprocessing, compilation
1242 proper, assembly and linking, always in that order. GCC is capable of
1243 preprocessing and compiling several files either into several
1244 assembler input files, or into one assembler input file; then each
1245 assembler input file produces an object file, and linking combines all
1246 the object files (those newly compiled, and those specified as input)
1247 into an executable file.
1249 @cindex file name suffix
1250 For any given input file, the file name suffix determines what kind of
1251 compilation is done:
1255 C source code that must be preprocessed.
1258 C source code that should not be preprocessed.
1261 C++ source code that should not be preprocessed.
1264 Objective-C source code. Note that you must link with the @file{libobjc}
1265 library to make an Objective-C program work.
1268 Objective-C source code that should not be preprocessed.
1272 Objective-C++ source code. Note that you must link with the @file{libobjc}
1273 library to make an Objective-C++ program work. Note that @samp{.M} refers
1274 to a literal capital M@.
1276 @item @var{file}.mii
1277 Objective-C++ source code that should not be preprocessed.
1280 C, C++, Objective-C or Objective-C++ header file to be turned into a
1281 precompiled header (default), or C, C++ header file to be turned into an
1282 Ada spec (via the @option{-fdump-ada-spec} switch).
1285 @itemx @var{file}.cp
1286 @itemx @var{file}.cxx
1287 @itemx @var{file}.cpp
1288 @itemx @var{file}.CPP
1289 @itemx @var{file}.c++
1291 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1292 the last two letters must both be literally @samp{x}. Likewise,
1293 @samp{.C} refers to a literal capital C@.
1297 Objective-C++ source code that must be preprocessed.
1299 @item @var{file}.mii
1300 Objective-C++ source code that should not be preprocessed.
1304 @itemx @var{file}.hp
1305 @itemx @var{file}.hxx
1306 @itemx @var{file}.hpp
1307 @itemx @var{file}.HPP
1308 @itemx @var{file}.h++
1309 @itemx @var{file}.tcc
1310 C++ header file to be turned into a precompiled header or Ada spec.
1313 @itemx @var{file}.for
1314 @itemx @var{file}.ftn
1315 Fixed form Fortran source code that should not be preprocessed.
1318 @itemx @var{file}.FOR
1319 @itemx @var{file}.fpp
1320 @itemx @var{file}.FPP
1321 @itemx @var{file}.FTN
1322 Fixed form Fortran source code that must be preprocessed (with the traditional
1325 @item @var{file}.f90
1326 @itemx @var{file}.f95
1327 @itemx @var{file}.f03
1328 @itemx @var{file}.f08
1329 Free form Fortran source code that should not be preprocessed.
1331 @item @var{file}.F90
1332 @itemx @var{file}.F95
1333 @itemx @var{file}.F03
1334 @itemx @var{file}.F08
1335 Free form Fortran source code that must be preprocessed (with the
1336 traditional preprocessor).
1341 @item @var{file}.brig
1342 BRIG files (binary representation of HSAIL).
1344 @item @var{file}.ads
1345 Ada source code file that contains a library unit declaration (a
1346 declaration of a package, subprogram, or generic, or a generic
1347 instantiation), or a library unit renaming declaration (a package,
1348 generic, or subprogram renaming declaration). Such files are also
1351 @item @var{file}.adb
1352 Ada source code file containing a library unit body (a subprogram or
1353 package body). Such files are also called @dfn{bodies}.
1355 @c GCC also knows about some suffixes for languages not yet included:
1366 @itemx @var{file}.sx
1367 Assembler code that must be preprocessed.
1370 An object file to be fed straight into linking.
1371 Any file name with no recognized suffix is treated this way.
1375 You can specify the input language explicitly with the @option{-x} option:
1378 @item -x @var{language}
1379 Specify explicitly the @var{language} for the following input files
1380 (rather than letting the compiler choose a default based on the file
1381 name suffix). This option applies to all following input files until
1382 the next @option{-x} option. Possible values for @var{language} are:
1384 c c-header cpp-output
1385 c++ c++-header c++-cpp-output
1386 objective-c objective-c-header objective-c-cpp-output
1387 objective-c++ objective-c++-header objective-c++-cpp-output
1388 assembler assembler-with-cpp
1390 f77 f77-cpp-input f95 f95-cpp-input
1396 Turn off any specification of a language, so that subsequent files are
1397 handled according to their file name suffixes (as they are if @option{-x}
1398 has not been used at all).
1401 If you only want some of the stages of compilation, you can use
1402 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1403 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1404 @command{gcc} is to stop. Note that some combinations (for example,
1405 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1410 Compile or assemble the source files, but do not link. The linking
1411 stage simply is not done. The ultimate output is in the form of an
1412 object file for each source file.
1414 By default, the object file name for a source file is made by replacing
1415 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1417 Unrecognized input files, not requiring compilation or assembly, are
1422 Stop after the stage of compilation proper; do not assemble. The output
1423 is in the form of an assembler code file for each non-assembler input
1426 By default, the assembler file name for a source file is made by
1427 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1429 Input files that don't require compilation are ignored.
1433 Stop after the preprocessing stage; do not run the compiler proper. The
1434 output is in the form of preprocessed source code, which is sent to the
1437 Input files that don't require preprocessing are ignored.
1439 @cindex output file option
1442 Place output in file @var{file}. This applies to whatever
1443 sort of output is being produced, whether it be an executable file,
1444 an object file, an assembler file or preprocessed C code.
1446 If @option{-o} is not specified, the default is to put an executable
1447 file in @file{a.out}, the object file for
1448 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1449 assembler file in @file{@var{source}.s}, a precompiled header file in
1450 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1455 Print (on standard error output) the commands executed to run the stages
1456 of compilation. Also print the version number of the compiler driver
1457 program and of the preprocessor and the compiler proper.
1461 Like @option{-v} except the commands are not executed and arguments
1462 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1463 This is useful for shell scripts to capture the driver-generated command lines.
1467 Print (on the standard output) a description of the command-line options
1468 understood by @command{gcc}. If the @option{-v} option is also specified
1469 then @option{--help} is also passed on to the various processes
1470 invoked by @command{gcc}, so that they can display the command-line options
1471 they accept. If the @option{-Wextra} option has also been specified
1472 (prior to the @option{--help} option), then command-line options that
1473 have no documentation associated with them are also displayed.
1476 @opindex target-help
1477 Print (on the standard output) a description of target-specific command-line
1478 options for each tool. For some targets extra target-specific
1479 information may also be printed.
1481 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1482 Print (on the standard output) a description of the command-line
1483 options understood by the compiler that fit into all specified classes
1484 and qualifiers. These are the supported classes:
1487 @item @samp{optimizers}
1488 Display all of the optimization options supported by the
1491 @item @samp{warnings}
1492 Display all of the options controlling warning messages
1493 produced by the compiler.
1496 Display target-specific options. Unlike the
1497 @option{--target-help} option however, target-specific options of the
1498 linker and assembler are not displayed. This is because those
1499 tools do not currently support the extended @option{--help=} syntax.
1502 Display the values recognized by the @option{--param}
1505 @item @var{language}
1506 Display the options supported for @var{language}, where
1507 @var{language} is the name of one of the languages supported in this
1511 Display the options that are common to all languages.
1514 These are the supported qualifiers:
1517 @item @samp{undocumented}
1518 Display only those options that are undocumented.
1521 Display options taking an argument that appears after an equal
1522 sign in the same continuous piece of text, such as:
1523 @samp{--help=target}.
1525 @item @samp{separate}
1526 Display options taking an argument that appears as a separate word
1527 following the original option, such as: @samp{-o output-file}.
1530 Thus for example to display all the undocumented target-specific
1531 switches supported by the compiler, use:
1534 --help=target,undocumented
1537 The sense of a qualifier can be inverted by prefixing it with the
1538 @samp{^} character, so for example to display all binary warning
1539 options (i.e., ones that are either on or off and that do not take an
1540 argument) that have a description, use:
1543 --help=warnings,^joined,^undocumented
1546 The argument to @option{--help=} should not consist solely of inverted
1549 Combining several classes is possible, although this usually
1550 restricts the output so much that there is nothing to display. One
1551 case where it does work, however, is when one of the classes is
1552 @var{target}. For example, to display all the target-specific
1553 optimization options, use:
1556 --help=target,optimizers
1559 The @option{--help=} option can be repeated on the command line. Each
1560 successive use displays its requested class of options, skipping
1561 those that have already been displayed.
1563 If the @option{-Q} option appears on the command line before the
1564 @option{--help=} option, then the descriptive text displayed by
1565 @option{--help=} is changed. Instead of describing the displayed
1566 options, an indication is given as to whether the option is enabled,
1567 disabled or set to a specific value (assuming that the compiler
1568 knows this at the point where the @option{--help=} option is used).
1570 Here is a truncated example from the ARM port of @command{gcc}:
1573 % gcc -Q -mabi=2 --help=target -c
1574 The following options are target specific:
1576 -mabort-on-noreturn [disabled]
1580 The output is sensitive to the effects of previous command-line
1581 options, so for example it is possible to find out which optimizations
1582 are enabled at @option{-O2} by using:
1585 -Q -O2 --help=optimizers
1588 Alternatively you can discover which binary optimizations are enabled
1589 by @option{-O3} by using:
1592 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1593 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1594 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1599 Display the version number and copyrights of the invoked GCC@.
1601 @item -pass-exit-codes
1602 @opindex pass-exit-codes
1603 Normally the @command{gcc} program exits with the code of 1 if any
1604 phase of the compiler returns a non-success return code. If you specify
1605 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1606 the numerically highest error produced by any phase returning an error
1607 indication. The C, C++, and Fortran front ends return 4 if an internal
1608 compiler error is encountered.
1612 Use pipes rather than temporary files for communication between the
1613 various stages of compilation. This fails to work on some systems where
1614 the assembler is unable to read from a pipe; but the GNU assembler has
1617 @item -specs=@var{file}
1619 Process @var{file} after the compiler reads in the standard @file{specs}
1620 file, in order to override the defaults which the @command{gcc} driver
1621 program uses when determining what switches to pass to @command{cc1},
1622 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1623 @option{-specs=@var{file}} can be specified on the command line, and they
1624 are processed in order, from left to right. @xref{Spec Files}, for
1625 information about the format of the @var{file}.
1629 Invoke all subcommands under a wrapper program. The name of the
1630 wrapper program and its parameters are passed as a comma separated
1634 gcc -c t.c -wrapper gdb,--args
1638 This invokes all subprograms of @command{gcc} under
1639 @samp{gdb --args}, thus the invocation of @command{cc1} is
1640 @samp{gdb --args cc1 @dots{}}.
1642 @item -fplugin=@var{name}.so
1644 Load the plugin code in file @var{name}.so, assumed to be a
1645 shared object to be dlopen'd by the compiler. The base name of
1646 the shared object file is used to identify the plugin for the
1647 purposes of argument parsing (See
1648 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1649 Each plugin should define the callback functions specified in the
1652 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1653 @opindex fplugin-arg
1654 Define an argument called @var{key} with a value of @var{value}
1655 for the plugin called @var{name}.
1657 @item -fdump-ada-spec@r{[}-slim@r{]}
1658 @opindex fdump-ada-spec
1659 For C and C++ source and include files, generate corresponding Ada specs.
1660 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1661 GNAT User's Guide}, which provides detailed documentation on this feature.
1663 @item -fada-spec-parent=@var{unit}
1664 @opindex fada-spec-parent
1665 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1666 Ada specs as child units of parent @var{unit}.
1668 @item -fdump-go-spec=@var{file}
1669 @opindex fdump-go-spec
1670 For input files in any language, generate corresponding Go
1671 declarations in @var{file}. This generates Go @code{const},
1672 @code{type}, @code{var}, and @code{func} declarations which may be a
1673 useful way to start writing a Go interface to code written in some
1676 @include @value{srcdir}/../libiberty/at-file.texi
1680 @section Compiling C++ Programs
1682 @cindex suffixes for C++ source
1683 @cindex C++ source file suffixes
1684 C++ source files conventionally use one of the suffixes @samp{.C},
1685 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1686 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1687 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1688 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1689 files with these names and compiles them as C++ programs even if you
1690 call the compiler the same way as for compiling C programs (usually
1691 with the name @command{gcc}).
1695 However, the use of @command{gcc} does not add the C++ library.
1696 @command{g++} is a program that calls GCC and automatically specifies linking
1697 against the C++ library. It treats @samp{.c},
1698 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1699 files unless @option{-x} is used. This program is also useful when
1700 precompiling a C header file with a @samp{.h} extension for use in C++
1701 compilations. On many systems, @command{g++} is also installed with
1702 the name @command{c++}.
1704 @cindex invoking @command{g++}
1705 When you compile C++ programs, you may specify many of the same
1706 command-line options that you use for compiling programs in any
1707 language; or command-line options meaningful for C and related
1708 languages; or options that are meaningful only for C++ programs.
1709 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1710 explanations of options for languages related to C@.
1711 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1712 explanations of options that are meaningful only for C++ programs.
1714 @node C Dialect Options
1715 @section Options Controlling C Dialect
1716 @cindex dialect options
1717 @cindex language dialect options
1718 @cindex options, dialect
1720 The following options control the dialect of C (or languages derived
1721 from C, such as C++, Objective-C and Objective-C++) that the compiler
1725 @cindex ANSI support
1729 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1730 equivalent to @option{-std=c++98}.
1732 This turns off certain features of GCC that are incompatible with ISO
1733 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1734 such as the @code{asm} and @code{typeof} keywords, and
1735 predefined macros such as @code{unix} and @code{vax} that identify the
1736 type of system you are using. It also enables the undesirable and
1737 rarely used ISO trigraph feature. For the C compiler,
1738 it disables recognition of C++ style @samp{//} comments as well as
1739 the @code{inline} keyword.
1741 The alternate keywords @code{__asm__}, @code{__extension__},
1742 @code{__inline__} and @code{__typeof__} continue to work despite
1743 @option{-ansi}. You would not want to use them in an ISO C program, of
1744 course, but it is useful to put them in header files that might be included
1745 in compilations done with @option{-ansi}. Alternate predefined macros
1746 such as @code{__unix__} and @code{__vax__} are also available, with or
1747 without @option{-ansi}.
1749 The @option{-ansi} option does not cause non-ISO programs to be
1750 rejected gratuitously. For that, @option{-Wpedantic} is required in
1751 addition to @option{-ansi}. @xref{Warning Options}.
1753 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1754 option is used. Some header files may notice this macro and refrain
1755 from declaring certain functions or defining certain macros that the
1756 ISO standard doesn't call for; this is to avoid interfering with any
1757 programs that might use these names for other things.
1759 Functions that are normally built in but do not have semantics
1760 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1761 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1762 built-in functions provided by GCC}, for details of the functions
1767 Determine the language standard. @xref{Standards,,Language Standards
1768 Supported by GCC}, for details of these standard versions. This option
1769 is currently only supported when compiling C or C++.
1771 The compiler can accept several base standards, such as @samp{c90} or
1772 @samp{c++98}, and GNU dialects of those standards, such as
1773 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1774 compiler accepts all programs following that standard plus those
1775 using GNU extensions that do not contradict it. For example,
1776 @option{-std=c90} turns off certain features of GCC that are
1777 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1778 keywords, but not other GNU extensions that do not have a meaning in
1779 ISO C90, such as omitting the middle term of a @code{?:}
1780 expression. On the other hand, when a GNU dialect of a standard is
1781 specified, all features supported by the compiler are enabled, even when
1782 those features change the meaning of the base standard. As a result, some
1783 strict-conforming programs may be rejected. The particular standard
1784 is used by @option{-Wpedantic} to identify which features are GNU
1785 extensions given that version of the standard. For example
1786 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1787 comments, while @option{-std=gnu99 -Wpedantic} does not.
1789 A value for this option must be provided; possible values are
1795 Support all ISO C90 programs (certain GNU extensions that conflict
1796 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1798 @item iso9899:199409
1799 ISO C90 as modified in amendment 1.
1805 ISO C99. This standard is substantially completely supported, modulo
1806 bugs and floating-point issues
1807 (mainly but not entirely relating to optional C99 features from
1808 Annexes F and G). See
1809 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1810 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1815 ISO C11, the 2011 revision of the ISO C standard. This standard is
1816 substantially completely supported, modulo bugs, floating-point issues
1817 (mainly but not entirely relating to optional C11 features from
1818 Annexes F and G) and the optional Annexes K (Bounds-checking
1819 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1823 GNU dialect of ISO C90 (including some C99 features).
1827 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1831 GNU dialect of ISO C11. This is the default for C code.
1832 The name @samp{gnu1x} is deprecated.
1836 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1837 additional defect reports. Same as @option{-ansi} for C++ code.
1841 GNU dialect of @option{-std=c++98}.
1845 The 2011 ISO C++ standard plus amendments.
1846 The name @samp{c++0x} is deprecated.
1850 GNU dialect of @option{-std=c++11}.
1851 The name @samp{gnu++0x} is deprecated.
1855 The 2014 ISO C++ standard plus amendments.
1856 The name @samp{c++1y} is deprecated.
1860 GNU dialect of @option{-std=c++14}.
1861 This is the default for C++ code.
1862 The name @samp{gnu++1y} is deprecated.
1865 The next revision of the ISO C++ standard, tentatively planned for
1866 2017. Support is highly experimental, and will almost certainly
1867 change in incompatible ways in future releases.
1870 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1871 and will almost certainly change in incompatible ways in future
1875 @item -fgnu89-inline
1876 @opindex fgnu89-inline
1877 The option @option{-fgnu89-inline} tells GCC to use the traditional
1878 GNU semantics for @code{inline} functions when in C99 mode.
1879 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1880 Using this option is roughly equivalent to adding the
1881 @code{gnu_inline} function attribute to all inline functions
1882 (@pxref{Function Attributes}).
1884 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1885 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1886 specifies the default behavior).
1887 This option is not supported in @option{-std=c90} or
1888 @option{-std=gnu90} mode.
1890 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1891 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1892 in effect for @code{inline} functions. @xref{Common Predefined
1893 Macros,,,cpp,The C Preprocessor}.
1895 @item -fpermitted-flt-eval-methods=@var{style}
1896 @opindex fpermitted-flt-eval-methods
1897 @opindex fpermitted-flt-eval-methods=c11
1898 @opindex fpermitted-flt-eval-methods=ts-18661-3
1899 ISO/IEC TS 18661-3 defines new permissible values for
1900 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
1901 a semantic type that is an interchange or extended format should be
1902 evaluated to the precision and range of that type. These new values are
1903 a superset of those permitted under C99/C11, which does not specify the
1904 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
1905 conforming to C11 may not have been written expecting the possibility of
1908 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
1909 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
1910 or the extended set of values specified in ISO/IEC TS 18661-3.
1912 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
1914 The default when in a standards compliant mode (@option{-std=c11} or similar)
1915 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
1916 dialect (@option{-std=gnu11} or similar) is
1917 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
1919 @item -aux-info @var{filename}
1921 Output to the given filename prototyped declarations for all functions
1922 declared and/or defined in a translation unit, including those in header
1923 files. This option is silently ignored in any language other than C@.
1925 Besides declarations, the file indicates, in comments, the origin of
1926 each declaration (source file and line), whether the declaration was
1927 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1928 @samp{O} for old, respectively, in the first character after the line
1929 number and the colon), and whether it came from a declaration or a
1930 definition (@samp{C} or @samp{F}, respectively, in the following
1931 character). In the case of function definitions, a K&R-style list of
1932 arguments followed by their declarations is also provided, inside
1933 comments, after the declaration.
1935 @item -fallow-parameterless-variadic-functions
1936 @opindex fallow-parameterless-variadic-functions
1937 Accept variadic functions without named parameters.
1939 Although it is possible to define such a function, this is not very
1940 useful as it is not possible to read the arguments. This is only
1941 supported for C as this construct is allowed by C++.
1945 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1946 keyword, so that code can use these words as identifiers. You can use
1947 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1948 instead. @option{-ansi} implies @option{-fno-asm}.
1950 In C++, this switch only affects the @code{typeof} keyword, since
1951 @code{asm} and @code{inline} are standard keywords. You may want to
1952 use the @option{-fno-gnu-keywords} flag instead, which has the same
1953 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1954 switch only affects the @code{asm} and @code{typeof} keywords, since
1955 @code{inline} is a standard keyword in ISO C99.
1958 @itemx -fno-builtin-@var{function}
1959 @opindex fno-builtin
1960 @cindex built-in functions
1961 Don't recognize built-in functions that do not begin with
1962 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1963 functions provided by GCC}, for details of the functions affected,
1964 including those which are not built-in functions when @option{-ansi} or
1965 @option{-std} options for strict ISO C conformance are used because they
1966 do not have an ISO standard meaning.
1968 GCC normally generates special code to handle certain built-in functions
1969 more efficiently; for instance, calls to @code{alloca} may become single
1970 instructions which adjust the stack directly, and calls to @code{memcpy}
1971 may become inline copy loops. The resulting code is often both smaller
1972 and faster, but since the function calls no longer appear as such, you
1973 cannot set a breakpoint on those calls, nor can you change the behavior
1974 of the functions by linking with a different library. In addition,
1975 when a function is recognized as a built-in function, GCC may use
1976 information about that function to warn about problems with calls to
1977 that function, or to generate more efficient code, even if the
1978 resulting code still contains calls to that function. For example,
1979 warnings are given with @option{-Wformat} for bad calls to
1980 @code{printf} when @code{printf} is built in and @code{strlen} is
1981 known not to modify global memory.
1983 With the @option{-fno-builtin-@var{function}} option
1984 only the built-in function @var{function} is
1985 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1986 function is named that is not built-in in this version of GCC, this
1987 option is ignored. There is no corresponding
1988 @option{-fbuiltin-@var{function}} option; if you wish to enable
1989 built-in functions selectively when using @option{-fno-builtin} or
1990 @option{-ffreestanding}, you may define macros such as:
1993 #define abs(n) __builtin_abs ((n))
1994 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2000 Enable parsing of function definitions marked with @code{__GIMPLE}.
2001 This is an experimental feature that allows unit testing of GIMPLE
2006 @cindex hosted environment
2008 Assert that compilation targets a hosted environment. This implies
2009 @option{-fbuiltin}. A hosted environment is one in which the
2010 entire standard library is available, and in which @code{main} has a return
2011 type of @code{int}. Examples are nearly everything except a kernel.
2012 This is equivalent to @option{-fno-freestanding}.
2014 @item -ffreestanding
2015 @opindex ffreestanding
2016 @cindex hosted environment
2018 Assert that compilation targets a freestanding environment. This
2019 implies @option{-fno-builtin}. A freestanding environment
2020 is one in which the standard library may not exist, and program startup may
2021 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2022 This is equivalent to @option{-fno-hosted}.
2024 @xref{Standards,,Language Standards Supported by GCC}, for details of
2025 freestanding and hosted environments.
2029 @cindex OpenACC accelerator programming
2030 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2031 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2032 compiler generates accelerated code according to the OpenACC Application
2033 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
2034 implies @option{-pthread}, and thus is only supported on targets that
2035 have support for @option{-pthread}.
2037 @item -fopenacc-dim=@var{geom}
2038 @opindex fopenacc-dim
2039 @cindex OpenACC accelerator programming
2040 Specify default compute dimensions for parallel offload regions that do
2041 not explicitly specify. The @var{geom} value is a triple of
2042 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2043 can be omitted, to use a target-specific default value.
2047 @cindex OpenMP parallel
2048 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2049 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2050 compiler generates parallel code according to the OpenMP Application
2051 Program Interface v4.5 @w{@uref{http://www.openmp.org/}}. This option
2052 implies @option{-pthread}, and thus is only supported on targets that
2053 have support for @option{-pthread}. @option{-fopenmp} implies
2054 @option{-fopenmp-simd}.
2057 @opindex fopenmp-simd
2060 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2061 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2066 @cindex Enable Cilk Plus
2067 Enable the usage of Cilk Plus language extension features for C/C++.
2068 When the option @option{-fcilkplus} is specified, enable the usage of
2069 the Cilk Plus Language extension features for C/C++. The present
2070 implementation follows ABI version 1.2. This is an experimental
2071 feature that is only partially complete, and whose interface may
2072 change in future versions of GCC as the official specification
2073 changes. Currently, all features but @code{_Cilk_for} have been
2078 When the option @option{-fgnu-tm} is specified, the compiler
2079 generates code for the Linux variant of Intel's current Transactional
2080 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2081 an experimental feature whose interface may change in future versions
2082 of GCC, as the official specification changes. Please note that not
2083 all architectures are supported for this feature.
2085 For more information on GCC's support for transactional memory,
2086 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2087 Transactional Memory Library}.
2089 Note that the transactional memory feature is not supported with
2090 non-call exceptions (@option{-fnon-call-exceptions}).
2092 @item -fms-extensions
2093 @opindex fms-extensions
2094 Accept some non-standard constructs used in Microsoft header files.
2096 In C++ code, this allows member names in structures to be similar
2097 to previous types declarations.
2106 Some cases of unnamed fields in structures and unions are only
2107 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2108 fields within structs/unions}, for details.
2110 Note that this option is off for all targets but x86
2111 targets using ms-abi.
2113 @item -fplan9-extensions
2114 @opindex fplan9-extensions
2115 Accept some non-standard constructs used in Plan 9 code.
2117 This enables @option{-fms-extensions}, permits passing pointers to
2118 structures with anonymous fields to functions that expect pointers to
2119 elements of the type of the field, and permits referring to anonymous
2120 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2121 struct/union fields within structs/unions}, for details. This is only
2122 supported for C, not C++.
2124 @item -fcond-mismatch
2125 @opindex fcond-mismatch
2126 Allow conditional expressions with mismatched types in the second and
2127 third arguments. The value of such an expression is void. This option
2128 is not supported for C++.
2130 @item -flax-vector-conversions
2131 @opindex flax-vector-conversions
2132 Allow implicit conversions between vectors with differing numbers of
2133 elements and/or incompatible element types. This option should not be
2136 @item -funsigned-char
2137 @opindex funsigned-char
2138 Let the type @code{char} be unsigned, like @code{unsigned char}.
2140 Each kind of machine has a default for what @code{char} should
2141 be. It is either like @code{unsigned char} by default or like
2142 @code{signed char} by default.
2144 Ideally, a portable program should always use @code{signed char} or
2145 @code{unsigned char} when it depends on the signedness of an object.
2146 But many programs have been written to use plain @code{char} and
2147 expect it to be signed, or expect it to be unsigned, depending on the
2148 machines they were written for. This option, and its inverse, let you
2149 make such a program work with the opposite default.
2151 The type @code{char} is always a distinct type from each of
2152 @code{signed char} or @code{unsigned char}, even though its behavior
2153 is always just like one of those two.
2156 @opindex fsigned-char
2157 Let the type @code{char} be signed, like @code{signed char}.
2159 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2160 the negative form of @option{-funsigned-char}. Likewise, the option
2161 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2163 @item -fsigned-bitfields
2164 @itemx -funsigned-bitfields
2165 @itemx -fno-signed-bitfields
2166 @itemx -fno-unsigned-bitfields
2167 @opindex fsigned-bitfields
2168 @opindex funsigned-bitfields
2169 @opindex fno-signed-bitfields
2170 @opindex fno-unsigned-bitfields
2171 These options control whether a bit-field is signed or unsigned, when the
2172 declaration does not use either @code{signed} or @code{unsigned}. By
2173 default, such a bit-field is signed, because this is consistent: the
2174 basic integer types such as @code{int} are signed types.
2176 @item -fsso-struct=@var{endianness}
2177 @opindex fsso-struct
2178 Set the default scalar storage order of structures and unions to the
2179 specified endianness. The accepted values are @samp{big-endian},
2180 @samp{little-endian} and @samp{native} for the native endianness of
2181 the target (the default). This option is not supported for C++.
2183 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2184 code that is not binary compatible with code generated without it if the
2185 specified endianness is not the native endianness of the target.
2188 @node C++ Dialect Options
2189 @section Options Controlling C++ Dialect
2191 @cindex compiler options, C++
2192 @cindex C++ options, command-line
2193 @cindex options, C++
2194 This section describes the command-line options that are only meaningful
2195 for C++ programs. You can also use most of the GNU compiler options
2196 regardless of what language your program is in. For example, you
2197 might compile a file @file{firstClass.C} like this:
2200 g++ -g -fstrict-enums -O -c firstClass.C
2204 In this example, only @option{-fstrict-enums} is an option meant
2205 only for C++ programs; you can use the other options with any
2206 language supported by GCC@.
2208 Some options for compiling C programs, such as @option{-std}, are also
2209 relevant for C++ programs.
2210 @xref{C Dialect Options,,Options Controlling C Dialect}.
2212 Here is a list of options that are @emph{only} for compiling C++ programs:
2216 @item -fabi-version=@var{n}
2217 @opindex fabi-version
2218 Use version @var{n} of the C++ ABI@. The default is version 0.
2220 Version 0 refers to the version conforming most closely to
2221 the C++ ABI specification. Therefore, the ABI obtained using version 0
2222 will change in different versions of G++ as ABI bugs are fixed.
2224 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2226 Version 2 is the version of the C++ ABI that first appeared in G++
2227 3.4, and was the default through G++ 4.9.
2229 Version 3 corrects an error in mangling a constant address as a
2232 Version 4, which first appeared in G++ 4.5, implements a standard
2233 mangling for vector types.
2235 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2236 attribute const/volatile on function pointer types, decltype of a
2237 plain decl, and use of a function parameter in the declaration of
2240 Version 6, which first appeared in G++ 4.7, corrects the promotion
2241 behavior of C++11 scoped enums and the mangling of template argument
2242 packs, const/static_cast, prefix ++ and --, and a class scope function
2243 used as a template argument.
2245 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2246 builtin type and corrects the mangling of lambdas in default argument
2249 Version 8, which first appeared in G++ 4.9, corrects the substitution
2250 behavior of function types with function-cv-qualifiers.
2252 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2255 Version 10, which first appeared in G++ 6.1, adds mangling of
2256 attributes that affect type identity, such as ia32 calling convention
2257 attributes (e.g. @samp{stdcall}).
2259 Version 11, which first appeared in G++ 7, corrects the mangling of
2260 sizeof... expressions and operator names. For multiple entities with
2261 the same name within a function, that are declared in different scopes,
2262 the mangling now changes starting with the twelfth occurrence. It also
2263 implies @option{-fnew-inheriting-ctors}.
2265 See also @option{-Wabi}.
2267 @item -fabi-compat-version=@var{n}
2268 @opindex fabi-compat-version
2269 On targets that support strong aliases, G++
2270 works around mangling changes by creating an alias with the correct
2271 mangled name when defining a symbol with an incorrect mangled name.
2272 This switch specifies which ABI version to use for the alias.
2274 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2275 compatibility). If another ABI version is explicitly selected, this
2276 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2277 use @option{-fabi-compat-version=2}.
2279 If this option is not provided but @option{-Wabi=@var{n}} is, that
2280 version is used for compatibility aliases. If this option is provided
2281 along with @option{-Wabi} (without the version), the version from this
2282 option is used for the warning.
2284 @item -fno-access-control
2285 @opindex fno-access-control
2286 Turn off all access checking. This switch is mainly useful for working
2287 around bugs in the access control code.
2290 @opindex faligned-new
2291 Enable support for C++17 @code{new} of types that require more
2292 alignment than @code{void* ::operator new(std::size_t)} provides. A
2293 numeric argument such as @code{-faligned-new=32} can be used to
2294 specify how much alignment (in bytes) is provided by that function,
2295 but few users will need to override the default of
2296 @code{alignof(std::max_align_t)}.
2298 This flag is enabled by default for @option{-std=c++1z}.
2302 Check that the pointer returned by @code{operator new} is non-null
2303 before attempting to modify the storage allocated. This check is
2304 normally unnecessary because the C++ standard specifies that
2305 @code{operator new} only returns @code{0} if it is declared
2306 @code{throw()}, in which case the compiler always checks the
2307 return value even without this option. In all other cases, when
2308 @code{operator new} has a non-empty exception specification, memory
2309 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2310 @samp{new (nothrow)}.
2314 Enable support for the C++ Extensions for Concepts Technical
2315 Specification, ISO 19217 (2015), which allows code like
2318 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2319 template <Addable T> T add (T a, T b) @{ return a + b; @}
2322 @item -fconstexpr-depth=@var{n}
2323 @opindex fconstexpr-depth
2324 Set the maximum nested evaluation depth for C++11 constexpr functions
2325 to @var{n}. A limit is needed to detect endless recursion during
2326 constant expression evaluation. The minimum specified by the standard
2329 @item -fconstexpr-loop-limit=@var{n}
2330 @opindex fconstexpr-loop-limit
2331 Set the maximum number of iterations for a loop in C++14 constexpr functions
2332 to @var{n}. A limit is needed to detect infinite loops during
2333 constant expression evaluation. The default is 262144 (1<<18).
2335 @item -fdeduce-init-list
2336 @opindex fdeduce-init-list
2337 Enable deduction of a template type parameter as
2338 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2341 template <class T> auto forward(T t) -> decltype (realfn (t))
2348 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2352 This deduction was implemented as a possible extension to the
2353 originally proposed semantics for the C++11 standard, but was not part
2354 of the final standard, so it is disabled by default. This option is
2355 deprecated, and may be removed in a future version of G++.
2357 @item -ffriend-injection
2358 @opindex ffriend-injection
2359 Inject friend functions into the enclosing namespace, so that they are
2360 visible outside the scope of the class in which they are declared.
2361 Friend functions were documented to work this way in the old Annotated
2362 C++ Reference Manual.
2363 However, in ISO C++ a friend function that is not declared
2364 in an enclosing scope can only be found using argument dependent
2365 lookup. GCC defaults to the standard behavior.
2367 This option is for compatibility, and may be removed in a future
2370 @item -fno-elide-constructors
2371 @opindex fno-elide-constructors
2372 The C++ standard allows an implementation to omit creating a temporary
2373 that is only used to initialize another object of the same type.
2374 Specifying this option disables that optimization, and forces G++ to
2375 call the copy constructor in all cases. This option also causes G++
2376 to call trivial member functions which otherwise would be expanded inline.
2378 In C++17, the compiler is required to omit these temporaries, but this
2379 option still affects trivial member functions.
2381 @item -fno-enforce-eh-specs
2382 @opindex fno-enforce-eh-specs
2383 Don't generate code to check for violation of exception specifications
2384 at run time. This option violates the C++ standard, but may be useful
2385 for reducing code size in production builds, much like defining
2386 @code{NDEBUG}. This does not give user code permission to throw
2387 exceptions in violation of the exception specifications; the compiler
2388 still optimizes based on the specifications, so throwing an
2389 unexpected exception results in undefined behavior at run time.
2391 @item -fextern-tls-init
2392 @itemx -fno-extern-tls-init
2393 @opindex fextern-tls-init
2394 @opindex fno-extern-tls-init
2395 The C++11 and OpenMP standards allow @code{thread_local} and
2396 @code{threadprivate} variables to have dynamic (runtime)
2397 initialization. To support this, any use of such a variable goes
2398 through a wrapper function that performs any necessary initialization.
2399 When the use and definition of the variable are in the same
2400 translation unit, this overhead can be optimized away, but when the
2401 use is in a different translation unit there is significant overhead
2402 even if the variable doesn't actually need dynamic initialization. If
2403 the programmer can be sure that no use of the variable in a
2404 non-defining TU needs to trigger dynamic initialization (either
2405 because the variable is statically initialized, or a use of the
2406 variable in the defining TU will be executed before any uses in
2407 another TU), they can avoid this overhead with the
2408 @option{-fno-extern-tls-init} option.
2410 On targets that support symbol aliases, the default is
2411 @option{-fextern-tls-init}. On targets that do not support symbol
2412 aliases, the default is @option{-fno-extern-tls-init}.
2415 @itemx -fno-for-scope
2417 @opindex fno-for-scope
2418 If @option{-ffor-scope} is specified, the scope of variables declared in
2419 a @i{for-init-statement} is limited to the @code{for} loop itself,
2420 as specified by the C++ standard.
2421 If @option{-fno-for-scope} is specified, the scope of variables declared in
2422 a @i{for-init-statement} extends to the end of the enclosing scope,
2423 as was the case in old versions of G++, and other (traditional)
2424 implementations of C++.
2426 If neither flag is given, the default is to follow the standard,
2427 but to allow and give a warning for old-style code that would
2428 otherwise be invalid, or have different behavior.
2430 @item -fno-gnu-keywords
2431 @opindex fno-gnu-keywords
2432 Do not recognize @code{typeof} as a keyword, so that code can use this
2433 word as an identifier. You can use the keyword @code{__typeof__} instead.
2434 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2435 @option{-std=c++98}, @option{-std=c++11}, etc.
2437 @item -fno-implicit-templates
2438 @opindex fno-implicit-templates
2439 Never emit code for non-inline templates that are instantiated
2440 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2441 @xref{Template Instantiation}, for more information.
2443 @item -fno-implicit-inline-templates
2444 @opindex fno-implicit-inline-templates
2445 Don't emit code for implicit instantiations of inline templates, either.
2446 The default is to handle inlines differently so that compiles with and
2447 without optimization need the same set of explicit instantiations.
2449 @item -fno-implement-inlines
2450 @opindex fno-implement-inlines
2451 To save space, do not emit out-of-line copies of inline functions
2452 controlled by @code{#pragma implementation}. This causes linker
2453 errors if these functions are not inlined everywhere they are called.
2455 @item -fms-extensions
2456 @opindex fms-extensions
2457 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2458 int and getting a pointer to member function via non-standard syntax.
2460 @item -fnew-inheriting-ctors
2461 @opindex fnew-inheriting-ctors
2462 Enable the P0136 adjustment to the semantics of C++11 constructor
2463 inheritance. This is part of C++17 but also considered to be a Defect
2464 Report against C++11 and C++14. This flag is enabled by default
2465 unless @option{-fabi-version=10} or lower is specified.
2467 @item -fnew-ttp-matching
2468 @opindex fnew-ttp-matching
2469 Enable the P0522 resolution to Core issue 150, template template
2470 parameters and default arguments: this allows a template with default
2471 template arguments as an argument for a template template parameter
2472 with fewer template parameters. This flag is enabled by default for
2473 @option{-std=c++1z}.
2475 @item -fno-nonansi-builtins
2476 @opindex fno-nonansi-builtins
2477 Disable built-in declarations of functions that are not mandated by
2478 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2479 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2482 @opindex fnothrow-opt
2483 Treat a @code{throw()} exception specification as if it were a
2484 @code{noexcept} specification to reduce or eliminate the text size
2485 overhead relative to a function with no exception specification. If
2486 the function has local variables of types with non-trivial
2487 destructors, the exception specification actually makes the
2488 function smaller because the EH cleanups for those variables can be
2489 optimized away. The semantic effect is that an exception thrown out of
2490 a function with such an exception specification results in a call
2491 to @code{terminate} rather than @code{unexpected}.
2493 @item -fno-operator-names
2494 @opindex fno-operator-names
2495 Do not treat the operator name keywords @code{and}, @code{bitand},
2496 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2497 synonyms as keywords.
2499 @item -fno-optional-diags
2500 @opindex fno-optional-diags
2501 Disable diagnostics that the standard says a compiler does not need to
2502 issue. Currently, the only such diagnostic issued by G++ is the one for
2503 a name having multiple meanings within a class.
2506 @opindex fpermissive
2507 Downgrade some diagnostics about nonconformant code from errors to
2508 warnings. Thus, using @option{-fpermissive} allows some
2509 nonconforming code to compile.
2511 @item -fno-pretty-templates
2512 @opindex fno-pretty-templates
2513 When an error message refers to a specialization of a function
2514 template, the compiler normally prints the signature of the
2515 template followed by the template arguments and any typedefs or
2516 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2517 rather than @code{void f(int)}) so that it's clear which template is
2518 involved. When an error message refers to a specialization of a class
2519 template, the compiler omits any template arguments that match
2520 the default template arguments for that template. If either of these
2521 behaviors make it harder to understand the error message rather than
2522 easier, you can use @option{-fno-pretty-templates} to disable them.
2526 Enable automatic template instantiation at link time. This option also
2527 implies @option{-fno-implicit-templates}. @xref{Template
2528 Instantiation}, for more information.
2532 Disable generation of information about every class with virtual
2533 functions for use by the C++ run-time type identification features
2534 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2535 of the language, you can save some space by using this flag. Note that
2536 exception handling uses the same information, but G++ generates it as
2537 needed. The @code{dynamic_cast} operator can still be used for casts that
2538 do not require run-time type information, i.e.@: casts to @code{void *} or to
2539 unambiguous base classes.
2541 @item -fsized-deallocation
2542 @opindex fsized-deallocation
2543 Enable the built-in global declarations
2545 void operator delete (void *, std::size_t) noexcept;
2546 void operator delete[] (void *, std::size_t) noexcept;
2548 as introduced in C++14. This is useful for user-defined replacement
2549 deallocation functions that, for example, use the size of the object
2550 to make deallocation faster. Enabled by default under
2551 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2552 warns about places that might want to add a definition.
2554 @item -fstrict-enums
2555 @opindex fstrict-enums
2556 Allow the compiler to optimize using the assumption that a value of
2557 enumerated type can only be one of the values of the enumeration (as
2558 defined in the C++ standard; basically, a value that can be
2559 represented in the minimum number of bits needed to represent all the
2560 enumerators). This assumption may not be valid if the program uses a
2561 cast to convert an arbitrary integer value to the enumerated type.
2563 @item -fstrong-eval-order
2564 @opindex fstrong-eval-order
2565 Evaluate member access, array subscripting, and shift expressions in
2566 left-to-right order, and evaluate assignment in right-to-left order,
2567 as adopted for C++17. Enabled by default with @option{-std=c++1z}.
2568 @option{-fstrong-eval-order=some} enables just the ordering of member
2569 access and shift expressions, and is the default without
2570 @option{-std=c++1z}.
2572 @item -ftemplate-backtrace-limit=@var{n}
2573 @opindex ftemplate-backtrace-limit
2574 Set the maximum number of template instantiation notes for a single
2575 warning or error to @var{n}. The default value is 10.
2577 @item -ftemplate-depth=@var{n}
2578 @opindex ftemplate-depth
2579 Set the maximum instantiation depth for template classes to @var{n}.
2580 A limit on the template instantiation depth is needed to detect
2581 endless recursions during template class instantiation. ANSI/ISO C++
2582 conforming programs must not rely on a maximum depth greater than 17
2583 (changed to 1024 in C++11). The default value is 900, as the compiler
2584 can run out of stack space before hitting 1024 in some situations.
2586 @item -fno-threadsafe-statics
2587 @opindex fno-threadsafe-statics
2588 Do not emit the extra code to use the routines specified in the C++
2589 ABI for thread-safe initialization of local statics. You can use this
2590 option to reduce code size slightly in code that doesn't need to be
2593 @item -fuse-cxa-atexit
2594 @opindex fuse-cxa-atexit
2595 Register destructors for objects with static storage duration with the
2596 @code{__cxa_atexit} function rather than the @code{atexit} function.
2597 This option is required for fully standards-compliant handling of static
2598 destructors, but only works if your C library supports
2599 @code{__cxa_atexit}.
2601 @item -fno-use-cxa-get-exception-ptr
2602 @opindex fno-use-cxa-get-exception-ptr
2603 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2604 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2605 if the runtime routine is not available.
2607 @item -fvisibility-inlines-hidden
2608 @opindex fvisibility-inlines-hidden
2609 This switch declares that the user does not attempt to compare
2610 pointers to inline functions or methods where the addresses of the two functions
2611 are taken in different shared objects.
2613 The effect of this is that GCC may, effectively, mark inline methods with
2614 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2615 appear in the export table of a DSO and do not require a PLT indirection
2616 when used within the DSO@. Enabling this option can have a dramatic effect
2617 on load and link times of a DSO as it massively reduces the size of the
2618 dynamic export table when the library makes heavy use of templates.
2620 The behavior of this switch is not quite the same as marking the
2621 methods as hidden directly, because it does not affect static variables
2622 local to the function or cause the compiler to deduce that
2623 the function is defined in only one shared object.
2625 You may mark a method as having a visibility explicitly to negate the
2626 effect of the switch for that method. For example, if you do want to
2627 compare pointers to a particular inline method, you might mark it as
2628 having default visibility. Marking the enclosing class with explicit
2629 visibility has no effect.
2631 Explicitly instantiated inline methods are unaffected by this option
2632 as their linkage might otherwise cross a shared library boundary.
2633 @xref{Template Instantiation}.
2635 @item -fvisibility-ms-compat
2636 @opindex fvisibility-ms-compat
2637 This flag attempts to use visibility settings to make GCC's C++
2638 linkage model compatible with that of Microsoft Visual Studio.
2640 The flag makes these changes to GCC's linkage model:
2644 It sets the default visibility to @code{hidden}, like
2645 @option{-fvisibility=hidden}.
2648 Types, but not their members, are not hidden by default.
2651 The One Definition Rule is relaxed for types without explicit
2652 visibility specifications that are defined in more than one
2653 shared object: those declarations are permitted if they are
2654 permitted when this option is not used.
2657 In new code it is better to use @option{-fvisibility=hidden} and
2658 export those classes that are intended to be externally visible.
2659 Unfortunately it is possible for code to rely, perhaps accidentally,
2660 on the Visual Studio behavior.
2662 Among the consequences of these changes are that static data members
2663 of the same type with the same name but defined in different shared
2664 objects are different, so changing one does not change the other;
2665 and that pointers to function members defined in different shared
2666 objects may not compare equal. When this flag is given, it is a
2667 violation of the ODR to define types with the same name differently.
2671 Do not use weak symbol support, even if it is provided by the linker.
2672 By default, G++ uses weak symbols if they are available. This
2673 option exists only for testing, and should not be used by end-users;
2674 it results in inferior code and has no benefits. This option may
2675 be removed in a future release of G++.
2679 Do not search for header files in the standard directories specific to
2680 C++, but do still search the other standard directories. (This option
2681 is used when building the C++ library.)
2684 In addition, these optimization, warning, and code generation options
2685 have meanings only for C++ programs:
2688 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2691 Warn when G++ it generates code that is probably not compatible with
2692 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2693 ABI with each major release, normally @option{-Wabi} will warn only if
2694 there is a check added later in a release series for an ABI issue
2695 discovered since the initial release. @option{-Wabi} will warn about
2696 more things if an older ABI version is selected (with
2697 @option{-fabi-version=@var{n}}).
2699 @option{-Wabi} can also be used with an explicit version number to
2700 warn about compatibility with a particular @option{-fabi-version}
2701 level, e.g. @option{-Wabi=2} to warn about changes relative to
2702 @option{-fabi-version=2}.
2704 If an explicit version number is provided and
2705 @option{-fabi-compat-version} is not specified, the version number
2706 from this option is used for compatibility aliases. If no explicit
2707 version number is provided with this option, but
2708 @option{-fabi-compat-version} is specified, that version number is
2709 used for ABI warnings.
2711 Although an effort has been made to warn about
2712 all such cases, there are probably some cases that are not warned about,
2713 even though G++ is generating incompatible code. There may also be
2714 cases where warnings are emitted even though the code that is generated
2717 You should rewrite your code to avoid these warnings if you are
2718 concerned about the fact that code generated by G++ may not be binary
2719 compatible with code generated by other compilers.
2721 Known incompatibilities in @option{-fabi-version=2} (which was the
2722 default from GCC 3.4 to 4.9) include:
2727 A template with a non-type template parameter of reference type was
2728 mangled incorrectly:
2731 template <int &> struct S @{@};
2735 This was fixed in @option{-fabi-version=3}.
2738 SIMD vector types declared using @code{__attribute ((vector_size))} were
2739 mangled in a non-standard way that does not allow for overloading of
2740 functions taking vectors of different sizes.
2742 The mangling was changed in @option{-fabi-version=4}.
2745 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2746 qualifiers, and @code{decltype} of a plain declaration was folded away.
2748 These mangling issues were fixed in @option{-fabi-version=5}.
2751 Scoped enumerators passed as arguments to a variadic function are
2752 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2753 On most targets this does not actually affect the parameter passing
2754 ABI, as there is no way to pass an argument smaller than @code{int}.
2756 Also, the ABI changed the mangling of template argument packs,
2757 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2758 a class scope function used as a template argument.
2760 These issues were corrected in @option{-fabi-version=6}.
2763 Lambdas in default argument scope were mangled incorrectly, and the
2764 ABI changed the mangling of @code{nullptr_t}.
2766 These issues were corrected in @option{-fabi-version=7}.
2769 When mangling a function type with function-cv-qualifiers, the
2770 un-qualified function type was incorrectly treated as a substitution
2773 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2776 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2777 unaligned accesses. Note that this did not affect the ABI of a
2778 function with a @code{nullptr_t} parameter, as parameters have a
2781 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2784 Target-specific attributes that affect the identity of a type, such as
2785 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2786 did not affect the mangled name, leading to name collisions when
2787 function pointers were used as template arguments.
2789 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2793 It also warns about psABI-related changes. The known psABI changes at this
2799 For SysV/x86-64, unions with @code{long double} members are
2800 passed in memory as specified in psABI. For example:
2810 @code{union U} is always passed in memory.
2814 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2817 Warn when a type with an ABI tag is used in a context that does not
2818 have that ABI tag. See @ref{C++ Attributes} for more information
2821 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2822 @opindex Wctor-dtor-privacy
2823 @opindex Wno-ctor-dtor-privacy
2824 Warn when a class seems unusable because all the constructors or
2825 destructors in that class are private, and it has neither friends nor
2826 public static member functions. Also warn if there are no non-private
2827 methods, and there's at least one private member function that isn't
2828 a constructor or destructor.
2830 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2831 @opindex Wdelete-non-virtual-dtor
2832 @opindex Wno-delete-non-virtual-dtor
2833 Warn when @code{delete} is used to destroy an instance of a class that
2834 has virtual functions and non-virtual destructor. It is unsafe to delete
2835 an instance of a derived class through a pointer to a base class if the
2836 base class does not have a virtual destructor. This warning is enabled
2839 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2840 @opindex Wliteral-suffix
2841 @opindex Wno-literal-suffix
2842 Warn when a string or character literal is followed by a ud-suffix which does
2843 not begin with an underscore. As a conforming extension, GCC treats such
2844 suffixes as separate preprocessing tokens in order to maintain backwards
2845 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2849 #define __STDC_FORMAT_MACROS
2850 #include <inttypes.h>
2855 printf("My int64: %" PRId64"\n", i64);
2859 In this case, @code{PRId64} is treated as a separate preprocessing token.
2861 Additionally, warn when a user-defined literal operator is declared with
2862 a literal suffix identifier that doesn't begin with an underscore. Literal
2863 suffix identifiers that don't begin with an underscore are reserved for
2864 future standardization.
2866 This warning is enabled by default.
2868 @item -Wlto-type-mismatch
2869 @opindex Wlto-type-mismatch
2870 @opindex Wno-lto-type-mismatch
2872 During the link-time optimization warn about type mismatches in
2873 global declarations from different compilation units.
2874 Requires @option{-flto} to be enabled. Enabled by default.
2876 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2878 @opindex Wno-narrowing
2879 For C++11 and later standards, narrowing conversions are diagnosed by default,
2880 as required by the standard. A narrowing conversion from a constant produces
2881 an error, and a narrowing conversion from a non-constant produces a warning,
2882 but @option{-Wno-narrowing} suppresses the diagnostic.
2883 Note that this does not affect the meaning of well-formed code;
2884 narrowing conversions are still considered ill-formed in SFINAE contexts.
2886 With @option{-Wnarrowing} in C++98, warn when a narrowing
2887 conversion prohibited by C++11 occurs within
2891 int i = @{ 2.2 @}; // error: narrowing from double to int
2894 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2896 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2898 @opindex Wno-noexcept
2899 Warn when a noexcept-expression evaluates to false because of a call
2900 to a function that does not have a non-throwing exception
2901 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2902 the compiler to never throw an exception.
2904 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2905 @opindex Wnoexcept-type
2906 @opindex Wno-noexcept-type
2907 Warn if the C++1z feature making @code{noexcept} part of a function
2908 type changes the mangled name of a symbol relative to C++14. Enabled
2909 by @option{-Wabi} and @option{-Wc++1z-compat}.
2912 template <class T> void f(T t) @{ t(); @};
2914 void h() @{ f(g); @} // in C++14 calls f<void(*)()>, in C++1z calls f<void(*)()noexcept>
2918 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2919 @opindex Wnon-virtual-dtor
2920 @opindex Wno-non-virtual-dtor
2921 Warn when a class has virtual functions and an accessible non-virtual
2922 destructor itself or in an accessible polymorphic base class, in which
2923 case it is possible but unsafe to delete an instance of a derived
2924 class through a pointer to the class itself or base class. This
2925 warning is automatically enabled if @option{-Weffc++} is specified.
2927 @item -Wregister @r{(C++ and Objective-C++ only)}
2929 @opindex Wno-register
2930 Warn on uses of the @code{register} storage class specifier, except
2931 when it is part of the GNU @ref{Explicit Register Variables} extension.
2932 The use of the @code{register} keyword as storage class specifier has
2933 been deprecated in C++11 and removed in C++17.
2934 Enabled by default with @option{-std=c++1z}.
2936 @item -Wreorder @r{(C++ and Objective-C++ only)}
2938 @opindex Wno-reorder
2939 @cindex reordering, warning
2940 @cindex warning for reordering of member initializers
2941 Warn when the order of member initializers given in the code does not
2942 match the order in which they must be executed. For instance:
2948 A(): j (0), i (1) @{ @}
2953 The compiler rearranges the member initializers for @code{i}
2954 and @code{j} to match the declaration order of the members, emitting
2955 a warning to that effect. This warning is enabled by @option{-Wall}.
2957 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2958 @opindex fext-numeric-literals
2959 @opindex fno-ext-numeric-literals
2960 Accept imaginary, fixed-point, or machine-defined
2961 literal number suffixes as GNU extensions.
2962 When this option is turned off these suffixes are treated
2963 as C++11 user-defined literal numeric suffixes.
2964 This is on by default for all pre-C++11 dialects and all GNU dialects:
2965 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2966 @option{-std=gnu++14}.
2967 This option is off by default
2968 for ISO C++11 onwards (@option{-std=c++11}, ...).
2971 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2974 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2977 Warn about violations of the following style guidelines from Scott Meyers'
2978 @cite{Effective C++} series of books:
2982 Define a copy constructor and an assignment operator for classes
2983 with dynamically-allocated memory.
2986 Prefer initialization to assignment in constructors.
2989 Have @code{operator=} return a reference to @code{*this}.
2992 Don't try to return a reference when you must return an object.
2995 Distinguish between prefix and postfix forms of increment and
2996 decrement operators.
2999 Never overload @code{&&}, @code{||}, or @code{,}.
3003 This option also enables @option{-Wnon-virtual-dtor}, which is also
3004 one of the effective C++ recommendations. However, the check is
3005 extended to warn about the lack of virtual destructor in accessible
3006 non-polymorphic bases classes too.
3008 When selecting this option, be aware that the standard library
3009 headers do not obey all of these guidelines; use @samp{grep -v}
3010 to filter out those warnings.
3012 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3013 @opindex Wstrict-null-sentinel
3014 @opindex Wno-strict-null-sentinel
3015 Warn about the use of an uncasted @code{NULL} as sentinel. When
3016 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3017 to @code{__null}. Although it is a null pointer constant rather than a
3018 null pointer, it is guaranteed to be of the same size as a pointer.
3019 But this use is not portable across different compilers.
3021 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3022 @opindex Wno-non-template-friend
3023 @opindex Wnon-template-friend
3024 Disable warnings when non-template friend functions are declared
3025 within a template. In very old versions of GCC that predate implementation
3026 of the ISO standard, declarations such as
3027 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3028 could be interpreted as a particular specialization of a template
3029 function; the warning exists to diagnose compatibility problems,
3030 and is enabled by default.
3032 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3033 @opindex Wold-style-cast
3034 @opindex Wno-old-style-cast
3035 Warn if an old-style (C-style) cast to a non-void type is used within
3036 a C++ program. The new-style casts (@code{dynamic_cast},
3037 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3038 less vulnerable to unintended effects and much easier to search for.
3040 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3041 @opindex Woverloaded-virtual
3042 @opindex Wno-overloaded-virtual
3043 @cindex overloaded virtual function, warning
3044 @cindex warning for overloaded virtual function
3045 Warn when a function declaration hides virtual functions from a
3046 base class. For example, in:
3053 struct B: public A @{
3058 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3069 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3070 @opindex Wno-pmf-conversions
3071 @opindex Wpmf-conversions
3072 Disable the diagnostic for converting a bound pointer to member function
3075 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3076 @opindex Wsign-promo
3077 @opindex Wno-sign-promo
3078 Warn when overload resolution chooses a promotion from unsigned or
3079 enumerated type to a signed type, over a conversion to an unsigned type of
3080 the same size. Previous versions of G++ tried to preserve
3081 unsignedness, but the standard mandates the current behavior.
3083 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3085 Warn when a primary template declaration is encountered. Some coding
3086 rules disallow templates, and this may be used to enforce that rule.
3087 The warning is inactive inside a system header file, such as the STL, so
3088 one can still use the STL. One may also instantiate or specialize
3091 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3092 @opindex Wmultiple-inheritance
3093 Warn when a class is defined with multiple direct base classes. Some
3094 coding rules disallow multiple inheritance, and this may be used to
3095 enforce that rule. The warning is inactive inside a system header file,
3096 such as the STL, so one can still use the STL. One may also define
3097 classes that indirectly use multiple inheritance.
3099 @item -Wvirtual-inheritance
3100 @opindex Wvirtual-inheritance
3101 Warn when a class is defined with a virtual direct base class. Some
3102 coding rules disallow multiple inheritance, and this may be used to
3103 enforce that rule. The warning is inactive inside a system header file,
3104 such as the STL, so one can still use the STL. One may also define
3105 classes that indirectly use virtual inheritance.
3108 @opindex Wnamespaces
3109 Warn when a namespace definition is opened. Some coding rules disallow
3110 namespaces, and this may be used to enforce that rule. The warning is
3111 inactive inside a system header file, such as the STL, so one can still
3112 use the STL. One may also use using directives and qualified names.
3114 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3116 @opindex Wno-terminate
3117 Disable the warning about a throw-expression that will immediately
3118 result in a call to @code{terminate}.
3121 @node Objective-C and Objective-C++ Dialect Options
3122 @section Options Controlling Objective-C and Objective-C++ Dialects
3124 @cindex compiler options, Objective-C and Objective-C++
3125 @cindex Objective-C and Objective-C++ options, command-line
3126 @cindex options, Objective-C and Objective-C++
3127 (NOTE: This manual does not describe the Objective-C and Objective-C++
3128 languages themselves. @xref{Standards,,Language Standards
3129 Supported by GCC}, for references.)
3131 This section describes the command-line options that are only meaningful
3132 for Objective-C and Objective-C++ programs. You can also use most of
3133 the language-independent GNU compiler options.
3134 For example, you might compile a file @file{some_class.m} like this:
3137 gcc -g -fgnu-runtime -O -c some_class.m
3141 In this example, @option{-fgnu-runtime} is an option meant only for
3142 Objective-C and Objective-C++ programs; you can use the other options with
3143 any language supported by GCC@.
3145 Note that since Objective-C is an extension of the C language, Objective-C
3146 compilations may also use options specific to the C front-end (e.g.,
3147 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3148 C++-specific options (e.g., @option{-Wabi}).
3150 Here is a list of options that are @emph{only} for compiling Objective-C
3151 and Objective-C++ programs:
3154 @item -fconstant-string-class=@var{class-name}
3155 @opindex fconstant-string-class
3156 Use @var{class-name} as the name of the class to instantiate for each
3157 literal string specified with the syntax @code{@@"@dots{}"}. The default
3158 class name is @code{NXConstantString} if the GNU runtime is being used, and
3159 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3160 @option{-fconstant-cfstrings} option, if also present, overrides the
3161 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3162 to be laid out as constant CoreFoundation strings.
3165 @opindex fgnu-runtime
3166 Generate object code compatible with the standard GNU Objective-C
3167 runtime. This is the default for most types of systems.
3169 @item -fnext-runtime
3170 @opindex fnext-runtime
3171 Generate output compatible with the NeXT runtime. This is the default
3172 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3173 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3176 @item -fno-nil-receivers
3177 @opindex fno-nil-receivers
3178 Assume that all Objective-C message dispatches (@code{[receiver
3179 message:arg]}) in this translation unit ensure that the receiver is
3180 not @code{nil}. This allows for more efficient entry points in the
3181 runtime to be used. This option is only available in conjunction with
3182 the NeXT runtime and ABI version 0 or 1.
3184 @item -fobjc-abi-version=@var{n}
3185 @opindex fobjc-abi-version
3186 Use version @var{n} of the Objective-C ABI for the selected runtime.
3187 This option is currently supported only for the NeXT runtime. In that
3188 case, Version 0 is the traditional (32-bit) ABI without support for
3189 properties and other Objective-C 2.0 additions. Version 1 is the
3190 traditional (32-bit) ABI with support for properties and other
3191 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3192 nothing is specified, the default is Version 0 on 32-bit target
3193 machines, and Version 2 on 64-bit target machines.
3195 @item -fobjc-call-cxx-cdtors
3196 @opindex fobjc-call-cxx-cdtors
3197 For each Objective-C class, check if any of its instance variables is a
3198 C++ object with a non-trivial default constructor. If so, synthesize a
3199 special @code{- (id) .cxx_construct} instance method which runs
3200 non-trivial default constructors on any such instance variables, in order,
3201 and then return @code{self}. Similarly, check if any instance variable
3202 is a C++ object with a non-trivial destructor, and if so, synthesize a
3203 special @code{- (void) .cxx_destruct} method which runs
3204 all such default destructors, in reverse order.
3206 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3207 methods thusly generated only operate on instance variables
3208 declared in the current Objective-C class, and not those inherited
3209 from superclasses. It is the responsibility of the Objective-C
3210 runtime to invoke all such methods in an object's inheritance
3211 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3212 by the runtime immediately after a new object instance is allocated;
3213 the @code{- (void) .cxx_destruct} methods are invoked immediately
3214 before the runtime deallocates an object instance.
3216 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3217 support for invoking the @code{- (id) .cxx_construct} and
3218 @code{- (void) .cxx_destruct} methods.
3220 @item -fobjc-direct-dispatch
3221 @opindex fobjc-direct-dispatch
3222 Allow fast jumps to the message dispatcher. On Darwin this is
3223 accomplished via the comm page.
3225 @item -fobjc-exceptions
3226 @opindex fobjc-exceptions
3227 Enable syntactic support for structured exception handling in
3228 Objective-C, similar to what is offered by C++. This option
3229 is required to use the Objective-C keywords @code{@@try},
3230 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3231 @code{@@synchronized}. This option is available with both the GNU
3232 runtime and the NeXT runtime (but not available in conjunction with
3233 the NeXT runtime on Mac OS X 10.2 and earlier).
3237 Enable garbage collection (GC) in Objective-C and Objective-C++
3238 programs. This option is only available with the NeXT runtime; the
3239 GNU runtime has a different garbage collection implementation that
3240 does not require special compiler flags.
3242 @item -fobjc-nilcheck
3243 @opindex fobjc-nilcheck
3244 For the NeXT runtime with version 2 of the ABI, check for a nil
3245 receiver in method invocations before doing the actual method call.
3246 This is the default and can be disabled using
3247 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3248 checked for nil in this way no matter what this flag is set to.
3249 Currently this flag does nothing when the GNU runtime, or an older
3250 version of the NeXT runtime ABI, is used.
3252 @item -fobjc-std=objc1
3254 Conform to the language syntax of Objective-C 1.0, the language
3255 recognized by GCC 4.0. This only affects the Objective-C additions to
3256 the C/C++ language; it does not affect conformance to C/C++ standards,
3257 which is controlled by the separate C/C++ dialect option flags. When
3258 this option is used with the Objective-C or Objective-C++ compiler,
3259 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3260 This is useful if you need to make sure that your Objective-C code can
3261 be compiled with older versions of GCC@.
3263 @item -freplace-objc-classes
3264 @opindex freplace-objc-classes
3265 Emit a special marker instructing @command{ld(1)} not to statically link in
3266 the resulting object file, and allow @command{dyld(1)} to load it in at
3267 run time instead. This is used in conjunction with the Fix-and-Continue
3268 debugging mode, where the object file in question may be recompiled and
3269 dynamically reloaded in the course of program execution, without the need
3270 to restart the program itself. Currently, Fix-and-Continue functionality
3271 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3276 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3277 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3278 compile time) with static class references that get initialized at load time,
3279 which improves run-time performance. Specifying the @option{-fzero-link} flag
3280 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3281 to be retained. This is useful in Zero-Link debugging mode, since it allows
3282 for individual class implementations to be modified during program execution.
3283 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3284 regardless of command-line options.
3286 @item -fno-local-ivars
3287 @opindex fno-local-ivars
3288 @opindex flocal-ivars
3289 By default instance variables in Objective-C can be accessed as if
3290 they were local variables from within the methods of the class they're
3291 declared in. This can lead to shadowing between instance variables
3292 and other variables declared either locally inside a class method or
3293 globally with the same name. Specifying the @option{-fno-local-ivars}
3294 flag disables this behavior thus avoiding variable shadowing issues.
3296 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3297 @opindex fivar-visibility
3298 Set the default instance variable visibility to the specified option
3299 so that instance variables declared outside the scope of any access
3300 modifier directives default to the specified visibility.
3304 Dump interface declarations for all classes seen in the source file to a
3305 file named @file{@var{sourcename}.decl}.
3307 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3308 @opindex Wassign-intercept
3309 @opindex Wno-assign-intercept
3310 Warn whenever an Objective-C assignment is being intercepted by the
3313 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3314 @opindex Wno-protocol
3316 If a class is declared to implement a protocol, a warning is issued for
3317 every method in the protocol that is not implemented by the class. The
3318 default behavior is to issue a warning for every method not explicitly
3319 implemented in the class, even if a method implementation is inherited
3320 from the superclass. If you use the @option{-Wno-protocol} option, then
3321 methods inherited from the superclass are considered to be implemented,
3322 and no warning is issued for them.
3324 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3326 @opindex Wno-selector
3327 Warn if multiple methods of different types for the same selector are
3328 found during compilation. The check is performed on the list of methods
3329 in the final stage of compilation. Additionally, a check is performed
3330 for each selector appearing in a @code{@@selector(@dots{})}
3331 expression, and a corresponding method for that selector has been found
3332 during compilation. Because these checks scan the method table only at
3333 the end of compilation, these warnings are not produced if the final
3334 stage of compilation is not reached, for example because an error is
3335 found during compilation, or because the @option{-fsyntax-only} option is
3338 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3339 @opindex Wstrict-selector-match
3340 @opindex Wno-strict-selector-match
3341 Warn if multiple methods with differing argument and/or return types are
3342 found for a given selector when attempting to send a message using this
3343 selector to a receiver of type @code{id} or @code{Class}. When this flag
3344 is off (which is the default behavior), the compiler omits such warnings
3345 if any differences found are confined to types that share the same size
3348 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3349 @opindex Wundeclared-selector
3350 @opindex Wno-undeclared-selector
3351 Warn if a @code{@@selector(@dots{})} expression referring to an
3352 undeclared selector is found. A selector is considered undeclared if no
3353 method with that name has been declared before the
3354 @code{@@selector(@dots{})} expression, either explicitly in an
3355 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3356 an @code{@@implementation} section. This option always performs its
3357 checks as soon as a @code{@@selector(@dots{})} expression is found,
3358 while @option{-Wselector} only performs its checks in the final stage of
3359 compilation. This also enforces the coding style convention
3360 that methods and selectors must be declared before being used.
3362 @item -print-objc-runtime-info
3363 @opindex print-objc-runtime-info
3364 Generate C header describing the largest structure that is passed by
3369 @node Diagnostic Message Formatting Options
3370 @section Options to Control Diagnostic Messages Formatting
3371 @cindex options to control diagnostics formatting
3372 @cindex diagnostic messages
3373 @cindex message formatting
3375 Traditionally, diagnostic messages have been formatted irrespective of
3376 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3377 options described below
3378 to control the formatting algorithm for diagnostic messages,
3379 e.g.@: how many characters per line, how often source location
3380 information should be reported. Note that some language front ends may not
3381 honor these options.
3384 @item -fmessage-length=@var{n}
3385 @opindex fmessage-length
3386 Try to format error messages so that they fit on lines of about
3387 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3388 done; each error message appears on a single line. This is the
3389 default for all front ends.
3391 @item -fdiagnostics-show-location=once
3392 @opindex fdiagnostics-show-location
3393 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3394 reporter to emit source location information @emph{once}; that is, in
3395 case the message is too long to fit on a single physical line and has to
3396 be wrapped, the source location won't be emitted (as prefix) again,
3397 over and over, in subsequent continuation lines. This is the default
3400 @item -fdiagnostics-show-location=every-line
3401 Only meaningful in line-wrapping mode. Instructs the diagnostic
3402 messages reporter to emit the same source location information (as
3403 prefix) for physical lines that result from the process of breaking
3404 a message which is too long to fit on a single line.
3406 @item -fdiagnostics-color[=@var{WHEN}]
3407 @itemx -fno-diagnostics-color
3408 @opindex fdiagnostics-color
3409 @cindex highlight, color
3410 @vindex GCC_COLORS @r{environment variable}
3411 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3412 or @samp{auto}. The default depends on how the compiler has been configured,
3413 it can be any of the above @var{WHEN} options or also @samp{never}
3414 if @env{GCC_COLORS} environment variable isn't present in the environment,
3415 and @samp{auto} otherwise.
3416 @samp{auto} means to use color only when the standard error is a terminal.
3417 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3418 aliases for @option{-fdiagnostics-color=always} and
3419 @option{-fdiagnostics-color=never}, respectively.
3421 The colors are defined by the environment variable @env{GCC_COLORS}.
3422 Its value is a colon-separated list of capabilities and Select Graphic
3423 Rendition (SGR) substrings. SGR commands are interpreted by the
3424 terminal or terminal emulator. (See the section in the documentation
3425 of your text terminal for permitted values and their meanings as
3426 character attributes.) These substring values are integers in decimal
3427 representation and can be concatenated with semicolons.
3428 Common values to concatenate include
3430 @samp{4} for underline,
3432 @samp{7} for inverse,
3433 @samp{39} for default foreground color,
3434 @samp{30} to @samp{37} for foreground colors,
3435 @samp{90} to @samp{97} for 16-color mode foreground colors,
3436 @samp{38;5;0} to @samp{38;5;255}
3437 for 88-color and 256-color modes foreground colors,
3438 @samp{49} for default background color,
3439 @samp{40} to @samp{47} for background colors,
3440 @samp{100} to @samp{107} for 16-color mode background colors,
3441 and @samp{48;5;0} to @samp{48;5;255}
3442 for 88-color and 256-color modes background colors.
3444 The default @env{GCC_COLORS} is
3446 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3447 quote=01:fixit-insert=32:fixit-delete=31:\
3448 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32
3451 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3452 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3453 @samp{01} is bold, and @samp{31} is red.
3454 Setting @env{GCC_COLORS} to the empty string disables colors.
3455 Supported capabilities are as follows.
3459 @vindex error GCC_COLORS @r{capability}
3460 SGR substring for error: markers.
3463 @vindex warning GCC_COLORS @r{capability}
3464 SGR substring for warning: markers.
3467 @vindex note GCC_COLORS @r{capability}
3468 SGR substring for note: markers.
3471 @vindex range1 GCC_COLORS @r{capability}
3472 SGR substring for first additional range.
3475 @vindex range2 GCC_COLORS @r{capability}
3476 SGR substring for second additional range.
3479 @vindex locus GCC_COLORS @r{capability}
3480 SGR substring for location information, @samp{file:line} or
3481 @samp{file:line:column} etc.
3484 @vindex quote GCC_COLORS @r{capability}
3485 SGR substring for information printed within quotes.
3488 @vindex fixit-insert GCC_COLORS @r{capability}
3489 SGR substring for fix-it hints suggesting text to
3490 be inserted or replaced.
3493 @vindex fixit-delete GCC_COLORS @r{capability}
3494 SGR substring for fix-it hints suggesting text to
3497 @item diff-filename=
3498 @vindex diff-filename GCC_COLORS @r{capability}
3499 SGR substring for filename headers within generated patches.
3502 @vindex diff-hunk GCC_COLORS @r{capability}
3503 SGR substring for the starts of hunks within generated patches.
3506 @vindex diff-delete GCC_COLORS @r{capability}
3507 SGR substring for deleted lines within generated patches.
3510 @vindex diff-insert GCC_COLORS @r{capability}
3511 SGR substring for inserted lines within generated patches.
3514 @item -fno-diagnostics-show-option
3515 @opindex fno-diagnostics-show-option
3516 @opindex fdiagnostics-show-option
3517 By default, each diagnostic emitted includes text indicating the
3518 command-line option that directly controls the diagnostic (if such an
3519 option is known to the diagnostic machinery). Specifying the
3520 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3522 @item -fno-diagnostics-show-caret
3523 @opindex fno-diagnostics-show-caret
3524 @opindex fdiagnostics-show-caret
3525 By default, each diagnostic emitted includes the original source line
3526 and a caret @samp{^} indicating the column. This option suppresses this
3527 information. The source line is truncated to @var{n} characters, if
3528 the @option{-fmessage-length=n} option is given. When the output is done
3529 to the terminal, the width is limited to the width given by the
3530 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3532 @item -fdiagnostics-parseable-fixits
3533 @opindex fdiagnostics-parseable-fixits
3534 Emit fix-it hints in a machine-parseable format, suitable for consumption
3535 by IDEs. For each fix-it, a line will be printed after the relevant
3536 diagnostic, starting with the string ``fix-it:''. For example:
3539 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3542 The location is expressed as a half-open range, expressed as a count of
3543 bytes, starting at byte 1 for the initial column. In the above example,
3544 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3548 00000000011111111112222222222
3549 12345678901234567890123456789
3550 gtk_widget_showall (dlg);
3555 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3556 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3557 (e.g. vertical tab as ``\013'').
3559 An empty replacement string indicates that the given range is to be removed.
3560 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3561 be inserted at the given position.
3563 @item -fdiagnostics-generate-patch
3564 @opindex fdiagnostics-generate-patch
3565 Print fix-it hints to stderr in unified diff format, after any diagnostics
3566 are printed. For example:
3573 void show_cb(GtkDialog *dlg)
3575 - gtk_widget_showall(dlg);
3576 + gtk_widget_show_all(dlg);
3581 The diff may or may not be colorized, following the same rules
3582 as for diagnostics (see @option{-fdiagnostics-color}).
3584 @item -fno-show-column
3585 @opindex fno-show-column
3586 Do not print column numbers in diagnostics. This may be necessary if
3587 diagnostics are being scanned by a program that does not understand the
3588 column numbers, such as @command{dejagnu}.
3592 @node Warning Options
3593 @section Options to Request or Suppress Warnings
3594 @cindex options to control warnings
3595 @cindex warning messages
3596 @cindex messages, warning
3597 @cindex suppressing warnings
3599 Warnings are diagnostic messages that report constructions that
3600 are not inherently erroneous but that are risky or suggest there
3601 may have been an error.
3603 The following language-independent options do not enable specific
3604 warnings but control the kinds of diagnostics produced by GCC@.
3607 @cindex syntax checking
3609 @opindex fsyntax-only
3610 Check the code for syntax errors, but don't do anything beyond that.
3612 @item -fmax-errors=@var{n}
3613 @opindex fmax-errors
3614 Limits the maximum number of error messages to @var{n}, at which point
3615 GCC bails out rather than attempting to continue processing the source
3616 code. If @var{n} is 0 (the default), there is no limit on the number
3617 of error messages produced. If @option{-Wfatal-errors} is also
3618 specified, then @option{-Wfatal-errors} takes precedence over this
3623 Inhibit all warning messages.
3628 Make all warnings into errors.
3633 Make the specified warning into an error. The specifier for a warning
3634 is appended; for example @option{-Werror=switch} turns the warnings
3635 controlled by @option{-Wswitch} into errors. This switch takes a
3636 negative form, to be used to negate @option{-Werror} for specific
3637 warnings; for example @option{-Wno-error=switch} makes
3638 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3641 The warning message for each controllable warning includes the
3642 option that controls the warning. That option can then be used with
3643 @option{-Werror=} and @option{-Wno-error=} as described above.
3644 (Printing of the option in the warning message can be disabled using the
3645 @option{-fno-diagnostics-show-option} flag.)
3647 Note that specifying @option{-Werror=}@var{foo} automatically implies
3648 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3651 @item -Wfatal-errors
3652 @opindex Wfatal-errors
3653 @opindex Wno-fatal-errors
3654 This option causes the compiler to abort compilation on the first error
3655 occurred rather than trying to keep going and printing further error
3660 You can request many specific warnings with options beginning with
3661 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3662 implicit declarations. Each of these specific warning options also
3663 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3664 example, @option{-Wno-implicit}. This manual lists only one of the
3665 two forms, whichever is not the default. For further
3666 language-specific options also refer to @ref{C++ Dialect Options} and
3667 @ref{Objective-C and Objective-C++ Dialect Options}.
3669 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3670 options, such as @option{-Wunused}, which may turn on further options,
3671 such as @option{-Wunused-value}. The combined effect of positive and
3672 negative forms is that more specific options have priority over less
3673 specific ones, independently of their position in the command-line. For
3674 options of the same specificity, the last one takes effect. Options
3675 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3676 as if they appeared at the end of the command-line.
3678 When an unrecognized warning option is requested (e.g.,
3679 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3680 that the option is not recognized. However, if the @option{-Wno-} form
3681 is used, the behavior is slightly different: no diagnostic is
3682 produced for @option{-Wno-unknown-warning} unless other diagnostics
3683 are being produced. This allows the use of new @option{-Wno-} options
3684 with old compilers, but if something goes wrong, the compiler
3685 warns that an unrecognized option is present.
3692 Issue all the warnings demanded by strict ISO C and ISO C++;
3693 reject all programs that use forbidden extensions, and some other
3694 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3695 version of the ISO C standard specified by any @option{-std} option used.
3697 Valid ISO C and ISO C++ programs should compile properly with or without
3698 this option (though a rare few require @option{-ansi} or a
3699 @option{-std} option specifying the required version of ISO C)@. However,
3700 without this option, certain GNU extensions and traditional C and C++
3701 features are supported as well. With this option, they are rejected.
3703 @option{-Wpedantic} does not cause warning messages for use of the
3704 alternate keywords whose names begin and end with @samp{__}. Pedantic
3705 warnings are also disabled in the expression that follows
3706 @code{__extension__}. However, only system header files should use
3707 these escape routes; application programs should avoid them.
3708 @xref{Alternate Keywords}.
3710 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3711 C conformance. They soon find that it does not do quite what they want:
3712 it finds some non-ISO practices, but not all---only those for which
3713 ISO C @emph{requires} a diagnostic, and some others for which
3714 diagnostics have been added.
3716 A feature to report any failure to conform to ISO C might be useful in
3717 some instances, but would require considerable additional work and would
3718 be quite different from @option{-Wpedantic}. We don't have plans to
3719 support such a feature in the near future.
3721 Where the standard specified with @option{-std} represents a GNU
3722 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3723 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3724 extended dialect is based. Warnings from @option{-Wpedantic} are given
3725 where they are required by the base standard. (It does not make sense
3726 for such warnings to be given only for features not in the specified GNU
3727 C dialect, since by definition the GNU dialects of C include all
3728 features the compiler supports with the given option, and there would be
3729 nothing to warn about.)
3731 @item -pedantic-errors
3732 @opindex pedantic-errors
3733 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3734 requires a diagnostic, in some cases where there is undefined behavior
3735 at compile-time and in some other cases that do not prevent compilation
3736 of programs that are valid according to the standard. This is not
3737 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3738 by this option and not enabled by the latter and vice versa.
3743 This enables all the warnings about constructions that some users
3744 consider questionable, and that are easy to avoid (or modify to
3745 prevent the warning), even in conjunction with macros. This also
3746 enables some language-specific warnings described in @ref{C++ Dialect
3747 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3749 @option{-Wall} turns on the following warning flags:
3751 @gccoptlist{-Waddress @gol
3752 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3754 -Wbool-operation @gol
3755 -Wc++11-compat -Wc++14-compat @gol
3756 -Wchar-subscripts @gol
3758 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3759 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3761 -Wint-in-bool-context @gol
3762 -Wimplicit @r{(C and Objective-C only)} @gol
3763 -Wimplicit-int @r{(C and Objective-C only)} @gol
3764 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3765 -Winit-self @r{(only for C++)} @gol
3766 -Wlogical-not-parentheses @gol
3767 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3768 -Wmaybe-uninitialized @gol
3769 -Wmemset-elt-size @gol
3770 -Wmemset-transposed-args @gol
3771 -Wmisleading-indentation @r{(only for C/C++)} @gol
3772 -Wmissing-braces @r{(only for C/ObjC)} @gol
3773 -Wnarrowing @r{(only for C++)} @gol
3775 -Wnonnull-compare @gol
3781 -Wsequence-point @gol
3782 -Wsign-compare @r{(only in C++)} @gol
3783 -Wsizeof-pointer-memaccess @gol
3784 -Wstrict-aliasing @gol
3785 -Wstrict-overflow=1 @gol
3787 -Wtautological-compare @gol
3789 -Wuninitialized @gol
3790 -Wunknown-pragmas @gol
3791 -Wunused-function @gol
3794 -Wunused-variable @gol
3795 -Wvolatile-register-var @gol
3798 Note that some warning flags are not implied by @option{-Wall}. Some of
3799 them warn about constructions that users generally do not consider
3800 questionable, but which occasionally you might wish to check for;
3801 others warn about constructions that are necessary or hard to avoid in
3802 some cases, and there is no simple way to modify the code to suppress
3803 the warning. Some of them are enabled by @option{-Wextra} but many of
3804 them must be enabled individually.
3810 This enables some extra warning flags that are not enabled by
3811 @option{-Wall}. (This option used to be called @option{-W}. The older
3812 name is still supported, but the newer name is more descriptive.)
3814 @gccoptlist{-Wclobbered @gol
3816 -Wignored-qualifiers @gol
3817 -Wimplicit-fallthrough=3 @gol
3818 -Wmissing-field-initializers @gol
3819 -Wmissing-parameter-type @r{(C only)} @gol
3820 -Wold-style-declaration @r{(C only)} @gol
3821 -Woverride-init @gol
3822 -Wsign-compare @r{(C only)} @gol
3824 -Wuninitialized @gol
3825 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3826 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3827 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3830 The option @option{-Wextra} also prints warning messages for the
3836 A pointer is compared against integer zero with @code{<}, @code{<=},
3837 @code{>}, or @code{>=}.
3840 (C++ only) An enumerator and a non-enumerator both appear in a
3841 conditional expression.
3844 (C++ only) Ambiguous virtual bases.
3847 (C++ only) Subscripting an array that has been declared @code{register}.
3850 (C++ only) Taking the address of a variable that has been declared
3854 (C++ only) A base class is not initialized in the copy constructor
3859 @item -Wchar-subscripts
3860 @opindex Wchar-subscripts
3861 @opindex Wno-char-subscripts
3862 Warn if an array subscript has type @code{char}. This is a common cause
3863 of error, as programmers often forget that this type is signed on some
3865 This warning is enabled by @option{-Wall}.
3869 Warn about an invalid memory access that is found by Pointer Bounds Checker
3870 (@option{-fcheck-pointer-bounds}).
3872 @item -Wno-coverage-mismatch
3873 @opindex Wno-coverage-mismatch
3874 Warn if feedback profiles do not match when using the
3875 @option{-fprofile-use} option.
3876 If a source file is changed between compiling with @option{-fprofile-gen} and
3877 with @option{-fprofile-use}, the files with the profile feedback can fail
3878 to match the source file and GCC cannot use the profile feedback
3879 information. By default, this warning is enabled and is treated as an
3880 error. @option{-Wno-coverage-mismatch} can be used to disable the
3881 warning or @option{-Wno-error=coverage-mismatch} can be used to
3882 disable the error. Disabling the error for this warning can result in
3883 poorly optimized code and is useful only in the
3884 case of very minor changes such as bug fixes to an existing code-base.
3885 Completely disabling the warning is not recommended.
3888 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3890 Suppress warning messages emitted by @code{#warning} directives.
3892 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3893 @opindex Wdouble-promotion
3894 @opindex Wno-double-promotion
3895 Give a warning when a value of type @code{float} is implicitly
3896 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3897 floating-point unit implement @code{float} in hardware, but emulate
3898 @code{double} in software. On such a machine, doing computations
3899 using @code{double} values is much more expensive because of the
3900 overhead required for software emulation.
3902 It is easy to accidentally do computations with @code{double} because
3903 floating-point literals are implicitly of type @code{double}. For
3907 float area(float radius)
3909 return 3.14159 * radius * radius;
3913 the compiler performs the entire computation with @code{double}
3914 because the floating-point literal is a @code{double}.
3916 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
3917 @opindex Wduplicate-decl-specifier
3918 @opindex Wno-duplicate-decl-specifier
3919 Warn if a declaration has duplicate @code{const}, @code{volatile},
3920 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
3924 @itemx -Wformat=@var{n}
3927 @opindex ffreestanding
3928 @opindex fno-builtin
3930 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3931 the arguments supplied have types appropriate to the format string
3932 specified, and that the conversions specified in the format string make
3933 sense. This includes standard functions, and others specified by format
3934 attributes (@pxref{Function Attributes}), in the @code{printf},
3935 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3936 not in the C standard) families (or other target-specific families).
3937 Which functions are checked without format attributes having been
3938 specified depends on the standard version selected, and such checks of
3939 functions without the attribute specified are disabled by
3940 @option{-ffreestanding} or @option{-fno-builtin}.
3942 The formats are checked against the format features supported by GNU
3943 libc version 2.2. These include all ISO C90 and C99 features, as well
3944 as features from the Single Unix Specification and some BSD and GNU
3945 extensions. Other library implementations may not support all these
3946 features; GCC does not support warning about features that go beyond a
3947 particular library's limitations. However, if @option{-Wpedantic} is used
3948 with @option{-Wformat}, warnings are given about format features not
3949 in the selected standard version (but not for @code{strfmon} formats,
3950 since those are not in any version of the C standard). @xref{C Dialect
3951 Options,,Options Controlling C Dialect}.
3958 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3959 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3960 @option{-Wformat} also checks for null format arguments for several
3961 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3962 aspects of this level of format checking can be disabled by the
3963 options: @option{-Wno-format-contains-nul},
3964 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3965 @option{-Wformat} is enabled by @option{-Wall}.
3967 @item -Wno-format-contains-nul
3968 @opindex Wno-format-contains-nul
3969 @opindex Wformat-contains-nul
3970 If @option{-Wformat} is specified, do not warn about format strings that
3973 @item -Wno-format-extra-args
3974 @opindex Wno-format-extra-args
3975 @opindex Wformat-extra-args
3976 If @option{-Wformat} is specified, do not warn about excess arguments to a
3977 @code{printf} or @code{scanf} format function. The C standard specifies
3978 that such arguments are ignored.
3980 Where the unused arguments lie between used arguments that are
3981 specified with @samp{$} operand number specifications, normally
3982 warnings are still given, since the implementation could not know what
3983 type to pass to @code{va_arg} to skip the unused arguments. However,
3984 in the case of @code{scanf} formats, this option suppresses the
3985 warning if the unused arguments are all pointers, since the Single
3986 Unix Specification says that such unused arguments are allowed.
3988 @item -Wformat-overflow
3989 @itemx -Wformat-overflow=@var{level}
3990 @opindex Wformat-overflow
3991 @opindex Wno-format-overflow
3992 Warn about calls to formatted input/output functions such as @code{sprintf}
3993 and @code{vsprintf} that might overflow the destination buffer. When the
3994 exact number of bytes written by a format directive cannot be determined
3995 at compile-time it is estimated based on heuristics that depend on the
3996 @var{level} argument and on optimization. While enabling optimization
3997 will in most cases improve the accuracy of the warning, it may also
3998 result in false positives.
4001 @item -Wformat-overflow
4002 @item -Wformat-overflow=1
4003 @opindex Wformat-overflow
4004 @opindex Wno-format-overflow
4005 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4006 employs a conservative approach that warns only about calls that most
4007 likely overflow the buffer. At this level, numeric arguments to format
4008 directives with unknown values are assumed to have the value of one, and
4009 strings of unknown length to be empty. Numeric arguments that are known
4010 to be bounded to a subrange of their type, or string arguments whose output
4011 is bounded either by their directive's precision or by a finite set of
4012 string literals, are assumed to take on the value within the range that
4013 results in the most bytes on output. For example, the call to @code{sprintf}
4014 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4015 the terminating NUL character (@code{'\0'}) appended by the function
4016 to the destination buffer will be written past its end. Increasing
4017 the size of the buffer by a single byte is sufficient to avoid the
4018 warning, though it may not be sufficient to avoid the overflow.
4021 void f (int a, int b)
4024 sprintf (buf, "a = %i, b = %i\n", a, b);
4028 @item -Wformat-overflow=2
4029 Level @var{2} warns also about calls that might overflow the destination
4030 buffer given an argument of sufficient length or magnitude. At level
4031 @var{2}, unknown numeric arguments are assumed to have the minimum
4032 representable value for signed types with a precision greater than 1, and
4033 the maximum representable value otherwise. Unknown string arguments whose
4034 length cannot be assumed to be bounded either by the directive's precision,
4035 or by a finite set of string literals they may evaluate to, or the character
4036 array they may point to, are assumed to be 1 character long.
4038 At level @var{2}, the call in the example above is again diagnosed, but
4039 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4040 @code{%i} directive will write some of its digits beyond the end of
4041 the destination buffer. To make the call safe regardless of the values
4042 of the two variables, the size of the destination buffer must be increased
4043 to at least 34 bytes. GCC includes the minimum size of the buffer in
4044 an informational note following the warning.
4046 An alternative to increasing the size of the destination buffer is to
4047 constrain the range of formatted values. The maximum length of string
4048 arguments can be bounded by specifying the precision in the format
4049 directive. When numeric arguments of format directives can be assumed
4050 to be bounded by less than the precision of their type, choosing
4051 an appropriate length modifier to the format specifier will reduce
4052 the required buffer size. For example, if @var{a} and @var{b} in the
4053 example above can be assumed to be within the precision of
4054 the @code{short int} type then using either the @code{%hi} format
4055 directive or casting the argument to @code{short} reduces the maximum
4056 required size of the buffer to 24 bytes.
4059 void f (int a, int b)
4062 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4067 @item -Wno-format-zero-length
4068 @opindex Wno-format-zero-length
4069 @opindex Wformat-zero-length
4070 If @option{-Wformat} is specified, do not warn about zero-length formats.
4071 The C standard specifies that zero-length formats are allowed.
4076 Enable @option{-Wformat} plus additional format checks. Currently
4077 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4080 @item -Wformat-nonliteral
4081 @opindex Wformat-nonliteral
4082 @opindex Wno-format-nonliteral
4083 If @option{-Wformat} is specified, also warn if the format string is not a
4084 string literal and so cannot be checked, unless the format function
4085 takes its format arguments as a @code{va_list}.
4087 @item -Wformat-security
4088 @opindex Wformat-security
4089 @opindex Wno-format-security
4090 If @option{-Wformat} is specified, also warn about uses of format
4091 functions that represent possible security problems. At present, this
4092 warns about calls to @code{printf} and @code{scanf} functions where the
4093 format string is not a string literal and there are no format arguments,
4094 as in @code{printf (foo);}. This may be a security hole if the format
4095 string came from untrusted input and contains @samp{%n}. (This is
4096 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4097 in future warnings may be added to @option{-Wformat-security} that are not
4098 included in @option{-Wformat-nonliteral}.)
4100 @item -Wformat-signedness
4101 @opindex Wformat-signedness
4102 @opindex Wno-format-signedness
4103 If @option{-Wformat} is specified, also warn if the format string
4104 requires an unsigned argument and the argument is signed and vice versa.
4106 @item -Wformat-truncation
4107 @itemx -Wformat-truncation=@var{level}
4108 @opindex Wformat-truncation
4109 @opindex Wno-format-truncation
4110 Warn about calls to formatted input/output functions such as @code{snprintf}
4111 and @code{vsnprintf} that might result in output truncation. When the exact
4112 number of bytes written by a format directive cannot be determined at
4113 compile-time it is estimated based on heuristics that depend on
4114 the @var{level} argument and on optimization. While enabling optimization
4115 will in most cases improve the accuracy of the warning, it may also result
4116 in false positives. Except as noted otherwise, the option uses the same
4117 logic @option{-Wformat-overflow}.
4120 @item -Wformat-truncation
4121 @item -Wformat-truncation=1
4122 @opindex Wformat-truncation
4123 @opindex Wno-format-overflow
4124 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4125 employs a conservative approach that warns only about calls to bounded
4126 functions whose return value is unused and that will most likely result
4127 in output truncation.
4129 @item -Wformat-truncation=2
4130 Level @var{2} warns also about calls to bounded functions whose return
4131 value is used and that might result in truncation given an argument of
4132 sufficient length or magnitude.
4136 @opindex Wformat-y2k
4137 @opindex Wno-format-y2k
4138 If @option{-Wformat} is specified, also warn about @code{strftime}
4139 formats that may yield only a two-digit year.
4144 @opindex Wno-nonnull
4145 Warn about passing a null pointer for arguments marked as
4146 requiring a non-null value by the @code{nonnull} function attribute.
4148 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4149 can be disabled with the @option{-Wno-nonnull} option.
4151 @item -Wnonnull-compare
4152 @opindex Wnonnull-compare
4153 @opindex Wno-nonnull-compare
4154 Warn when comparing an argument marked with the @code{nonnull}
4155 function attribute against null inside the function.
4157 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4158 can be disabled with the @option{-Wno-nonnull-compare} option.
4160 @item -Wnull-dereference
4161 @opindex Wnull-dereference
4162 @opindex Wno-null-dereference
4163 Warn if the compiler detects paths that trigger erroneous or
4164 undefined behavior due to dereferencing a null pointer. This option
4165 is only active when @option{-fdelete-null-pointer-checks} is active,
4166 which is enabled by optimizations in most targets. The precision of
4167 the warnings depends on the optimization options used.
4169 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4171 @opindex Wno-init-self
4172 Warn about uninitialized variables that are initialized with themselves.
4173 Note this option can only be used with the @option{-Wuninitialized} option.
4175 For example, GCC warns about @code{i} being uninitialized in the
4176 following snippet only when @option{-Winit-self} has been specified:
4187 This warning is enabled by @option{-Wall} in C++.
4189 @item -Wimplicit-int @r{(C and Objective-C only)}
4190 @opindex Wimplicit-int
4191 @opindex Wno-implicit-int
4192 Warn when a declaration does not specify a type.
4193 This warning is enabled by @option{-Wall}.
4195 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4196 @opindex Wimplicit-function-declaration
4197 @opindex Wno-implicit-function-declaration
4198 Give a warning whenever a function is used before being declared. In
4199 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4200 enabled by default and it is made into an error by
4201 @option{-pedantic-errors}. This warning is also enabled by
4204 @item -Wimplicit @r{(C and Objective-C only)}
4206 @opindex Wno-implicit
4207 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4208 This warning is enabled by @option{-Wall}.
4210 @item -Wimplicit-fallthrough
4211 @opindex Wimplicit-fallthrough
4212 @opindex Wno-implicit-fallthrough
4213 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4214 and @option{-Wno-implicit-fallthrough} is the same as
4215 @option{-Wimplicit-fallthrough=0}.
4217 @item -Wimplicit-fallthrough=@var{n}
4218 @opindex Wimplicit-fallthrough=
4219 Warn when a switch case falls through. For example:
4237 This warning does not warn when the last statement of a case cannot
4238 fall through, e.g. when there is a return statement or a call to function
4239 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4240 also takes into account control flow statements, such as ifs, and only
4241 warns when appropriate. E.g.@:
4251 @} else if (i < 1) @{
4261 Since there are occasions where a switch case fall through is desirable,
4262 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4263 to be used along with a null statement to suppress this warning that
4264 would normally occur:
4272 __attribute__ ((fallthrough));
4279 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4280 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4281 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4282 Instead of these attributes, it is also possible to add a fallthrough comment
4283 to silence the warning. The whole body of the C or C++ style comment should
4284 match the given regular expressions listed below. The option argument @var{n}
4285 specifies what kind of comments are accepted:
4289 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4291 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4292 expression, any comment is used as fallthrough comment.
4294 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4295 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4297 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4298 following regular expressions:
4302 @item @code{-fallthrough}
4304 @item @code{@@fallthrough@@}
4306 @item @code{lint -fallthrough[ \t]*}
4308 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4310 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4312 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4316 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4317 following regular expressions:
4321 @item @code{-fallthrough}
4323 @item @code{@@fallthrough@@}
4325 @item @code{lint -fallthrough[ \t]*}
4327 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4331 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4332 fallthrough comments, only attributes disable the warning.
4336 The comment needs to be followed after optional whitespace and other comments
4337 by @code{case} or @code{default} keywords or by a user label that precedes some
4338 @code{case} or @code{default} label.
4353 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4355 @item -Wignored-qualifiers @r{(C and C++ only)}
4356 @opindex Wignored-qualifiers
4357 @opindex Wno-ignored-qualifiers
4358 Warn if the return type of a function has a type qualifier
4359 such as @code{const}. For ISO C such a type qualifier has no effect,
4360 since the value returned by a function is not an lvalue.
4361 For C++, the warning is only emitted for scalar types or @code{void}.
4362 ISO C prohibits qualified @code{void} return types on function
4363 definitions, so such return types always receive a warning
4364 even without this option.
4366 This warning is also enabled by @option{-Wextra}.
4368 @item -Wignored-attributes @r{(C and C++ only)}
4369 @opindex Wignored-attributes
4370 @opindex Wno-ignored-attributes
4371 Warn when an attribute is ignored. This is different from the
4372 @option{-Wattributes} option in that it warns whenever the compiler decides
4373 to drop an attribute, not that the attribute is either unknown, used in a
4374 wrong place, etc. This warning is enabled by default.
4379 Warn if the type of @code{main} is suspicious. @code{main} should be
4380 a function with external linkage, returning int, taking either zero
4381 arguments, two, or three arguments of appropriate types. This warning
4382 is enabled by default in C++ and is enabled by either @option{-Wall}
4383 or @option{-Wpedantic}.
4385 @item -Wmisleading-indentation @r{(C and C++ only)}
4386 @opindex Wmisleading-indentation
4387 @opindex Wno-misleading-indentation
4388 Warn when the indentation of the code does not reflect the block structure.
4389 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4390 @code{for} clauses with a guarded statement that does not use braces,
4391 followed by an unguarded statement with the same indentation.
4393 In the following example, the call to ``bar'' is misleadingly indented as
4394 if it were guarded by the ``if'' conditional.
4397 if (some_condition ())
4399 bar (); /* Gotcha: this is not guarded by the "if". */
4402 In the case of mixed tabs and spaces, the warning uses the
4403 @option{-ftabstop=} option to determine if the statements line up
4406 The warning is not issued for code involving multiline preprocessor logic
4407 such as the following example.
4412 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4418 The warning is not issued after a @code{#line} directive, since this
4419 typically indicates autogenerated code, and no assumptions can be made
4420 about the layout of the file that the directive references.
4422 This warning is enabled by @option{-Wall} in C and C++.
4424 @item -Wmissing-braces
4425 @opindex Wmissing-braces
4426 @opindex Wno-missing-braces
4427 Warn if an aggregate or union initializer is not fully bracketed. In
4428 the following example, the initializer for @code{a} is not fully
4429 bracketed, but that for @code{b} is fully bracketed. This warning is
4430 enabled by @option{-Wall} in C.
4433 int a[2][2] = @{ 0, 1, 2, 3 @};
4434 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4437 This warning is enabled by @option{-Wall}.
4439 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4440 @opindex Wmissing-include-dirs
4441 @opindex Wno-missing-include-dirs
4442 Warn if a user-supplied include directory does not exist.
4445 @opindex Wparentheses
4446 @opindex Wno-parentheses
4447 Warn if parentheses are omitted in certain contexts, such
4448 as when there is an assignment in a context where a truth value
4449 is expected, or when operators are nested whose precedence people
4450 often get confused about.
4452 Also warn if a comparison like @code{x<=y<=z} appears; this is
4453 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4454 interpretation from that of ordinary mathematical notation.
4456 Also warn for dangerous uses of the GNU extension to
4457 @code{?:} with omitted middle operand. When the condition
4458 in the @code{?}: operator is a boolean expression, the omitted value is
4459 always 1. Often programmers expect it to be a value computed
4460 inside the conditional expression instead.
4462 This warning is enabled by @option{-Wall}.
4464 @item -Wsequence-point
4465 @opindex Wsequence-point
4466 @opindex Wno-sequence-point
4467 Warn about code that may have undefined semantics because of violations
4468 of sequence point rules in the C and C++ standards.
4470 The C and C++ standards define the order in which expressions in a C/C++
4471 program are evaluated in terms of @dfn{sequence points}, which represent
4472 a partial ordering between the execution of parts of the program: those
4473 executed before the sequence point, and those executed after it. These
4474 occur after the evaluation of a full expression (one which is not part
4475 of a larger expression), after the evaluation of the first operand of a
4476 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4477 function is called (but after the evaluation of its arguments and the
4478 expression denoting the called function), and in certain other places.
4479 Other than as expressed by the sequence point rules, the order of
4480 evaluation of subexpressions of an expression is not specified. All
4481 these rules describe only a partial order rather than a total order,
4482 since, for example, if two functions are called within one expression
4483 with no sequence point between them, the order in which the functions
4484 are called is not specified. However, the standards committee have
4485 ruled that function calls do not overlap.
4487 It is not specified when between sequence points modifications to the
4488 values of objects take effect. Programs whose behavior depends on this
4489 have undefined behavior; the C and C++ standards specify that ``Between
4490 the previous and next sequence point an object shall have its stored
4491 value modified at most once by the evaluation of an expression.
4492 Furthermore, the prior value shall be read only to determine the value
4493 to be stored.''. If a program breaks these rules, the results on any
4494 particular implementation are entirely unpredictable.
4496 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4497 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4498 diagnosed by this option, and it may give an occasional false positive
4499 result, but in general it has been found fairly effective at detecting
4500 this sort of problem in programs.
4502 The C++17 standard will define the order of evaluation of operands in
4503 more cases: in particular it requires that the right-hand side of an
4504 assignment be evaluated before the left-hand side, so the above
4505 examples are no longer undefined. But this warning will still warn
4506 about them, to help people avoid writing code that is undefined in C
4507 and earlier revisions of C++.
4509 The standard is worded confusingly, therefore there is some debate
4510 over the precise meaning of the sequence point rules in subtle cases.
4511 Links to discussions of the problem, including proposed formal
4512 definitions, may be found on the GCC readings page, at
4513 @uref{http://gcc.gnu.org/@/readings.html}.
4515 This warning is enabled by @option{-Wall} for C and C++.
4517 @item -Wno-return-local-addr
4518 @opindex Wno-return-local-addr
4519 @opindex Wreturn-local-addr
4520 Do not warn about returning a pointer (or in C++, a reference) to a
4521 variable that goes out of scope after the function returns.
4524 @opindex Wreturn-type
4525 @opindex Wno-return-type
4526 Warn whenever a function is defined with a return type that defaults
4527 to @code{int}. Also warn about any @code{return} statement with no
4528 return value in a function whose return type is not @code{void}
4529 (falling off the end of the function body is considered returning
4532 For C only, warn about a @code{return} statement with an expression in a
4533 function whose return type is @code{void}, unless the expression type is
4534 also @code{void}. As a GNU extension, the latter case is accepted
4535 without a warning unless @option{-Wpedantic} is used.
4537 For C++, a function without return type always produces a diagnostic
4538 message, even when @option{-Wno-return-type} is specified. The only
4539 exceptions are @code{main} and functions defined in system headers.
4541 This warning is enabled by @option{-Wall}.
4543 @item -Wshift-count-negative
4544 @opindex Wshift-count-negative
4545 @opindex Wno-shift-count-negative
4546 Warn if shift count is negative. This warning is enabled by default.
4548 @item -Wshift-count-overflow
4549 @opindex Wshift-count-overflow
4550 @opindex Wno-shift-count-overflow
4551 Warn if shift count >= width of type. This warning is enabled by default.
4553 @item -Wshift-negative-value
4554 @opindex Wshift-negative-value
4555 @opindex Wno-shift-negative-value
4556 Warn if left shifting a negative value. This warning is enabled by
4557 @option{-Wextra} in C99 and C++11 modes (and newer).
4559 @item -Wshift-overflow
4560 @itemx -Wshift-overflow=@var{n}
4561 @opindex Wshift-overflow
4562 @opindex Wno-shift-overflow
4563 Warn about left shift overflows. This warning is enabled by
4564 default in C99 and C++11 modes (and newer).
4567 @item -Wshift-overflow=1
4568 This is the warning level of @option{-Wshift-overflow} and is enabled
4569 by default in C99 and C++11 modes (and newer). This warning level does
4570 not warn about left-shifting 1 into the sign bit. (However, in C, such
4571 an overflow is still rejected in contexts where an integer constant expression
4574 @item -Wshift-overflow=2
4575 This warning level also warns about left-shifting 1 into the sign bit,
4576 unless C++14 mode is active.
4582 Warn whenever a @code{switch} statement has an index of enumerated type
4583 and lacks a @code{case} for one or more of the named codes of that
4584 enumeration. (The presence of a @code{default} label prevents this
4585 warning.) @code{case} labels outside the enumeration range also
4586 provoke warnings when this option is used (even if there is a
4587 @code{default} label).
4588 This warning is enabled by @option{-Wall}.
4590 @item -Wswitch-default
4591 @opindex Wswitch-default
4592 @opindex Wno-switch-default
4593 Warn whenever a @code{switch} statement does not have a @code{default}
4597 @opindex Wswitch-enum
4598 @opindex Wno-switch-enum
4599 Warn whenever a @code{switch} statement has an index of enumerated type
4600 and lacks a @code{case} for one or more of the named codes of that
4601 enumeration. @code{case} labels outside the enumeration range also
4602 provoke warnings when this option is used. The only difference
4603 between @option{-Wswitch} and this option is that this option gives a
4604 warning about an omitted enumeration code even if there is a
4605 @code{default} label.
4608 @opindex Wswitch-bool
4609 @opindex Wno-switch-bool
4610 Warn whenever a @code{switch} statement has an index of boolean type
4611 and the case values are outside the range of a boolean type.
4612 It is possible to suppress this warning by casting the controlling
4613 expression to a type other than @code{bool}. For example:
4616 switch ((int) (a == 4))
4622 This warning is enabled by default for C and C++ programs.
4624 @item -Wswitch-unreachable
4625 @opindex Wswitch-unreachable
4626 @opindex Wno-switch-unreachable
4627 Warn whenever a @code{switch} statement contains statements between the
4628 controlling expression and the first case label, which will never be
4629 executed. For example:
4641 @option{-Wswitch-unreachable} does not warn if the statement between the
4642 controlling expression and the first case label is just a declaration:
4655 This warning is enabled by default for C and C++ programs.
4657 @item -Wsync-nand @r{(C and C++ only)}
4659 @opindex Wno-sync-nand
4660 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4661 built-in functions are used. These functions changed semantics in GCC 4.4.
4663 @item -Wunused-but-set-parameter
4664 @opindex Wunused-but-set-parameter
4665 @opindex Wno-unused-but-set-parameter
4666 Warn whenever a function parameter is assigned to, but otherwise unused
4667 (aside from its declaration).
4669 To suppress this warning use the @code{unused} attribute
4670 (@pxref{Variable Attributes}).
4672 This warning is also enabled by @option{-Wunused} together with
4675 @item -Wunused-but-set-variable
4676 @opindex Wunused-but-set-variable
4677 @opindex Wno-unused-but-set-variable
4678 Warn whenever a local variable is assigned to, but otherwise unused
4679 (aside from its declaration).
4680 This warning is enabled by @option{-Wall}.
4682 To suppress this warning use the @code{unused} attribute
4683 (@pxref{Variable Attributes}).
4685 This warning is also enabled by @option{-Wunused}, which is enabled
4688 @item -Wunused-function
4689 @opindex Wunused-function
4690 @opindex Wno-unused-function
4691 Warn whenever a static function is declared but not defined or a
4692 non-inline static function is unused.
4693 This warning is enabled by @option{-Wall}.
4695 @item -Wunused-label
4696 @opindex Wunused-label
4697 @opindex Wno-unused-label
4698 Warn whenever a label is declared but not used.
4699 This warning is enabled by @option{-Wall}.
4701 To suppress this warning use the @code{unused} attribute
4702 (@pxref{Variable Attributes}).
4704 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4705 @opindex Wunused-local-typedefs
4706 Warn when a typedef locally defined in a function is not used.
4707 This warning is enabled by @option{-Wall}.
4709 @item -Wunused-parameter
4710 @opindex Wunused-parameter
4711 @opindex Wno-unused-parameter
4712 Warn whenever a function parameter is unused aside from its declaration.
4714 To suppress this warning use the @code{unused} attribute
4715 (@pxref{Variable Attributes}).
4717 @item -Wno-unused-result
4718 @opindex Wunused-result
4719 @opindex Wno-unused-result
4720 Do not warn if a caller of a function marked with attribute
4721 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4722 its return value. The default is @option{-Wunused-result}.
4724 @item -Wunused-variable
4725 @opindex Wunused-variable
4726 @opindex Wno-unused-variable
4727 Warn whenever a local or static variable is unused aside from its
4728 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4729 but not for C++. This warning is enabled by @option{-Wall}.
4731 To suppress this warning use the @code{unused} attribute
4732 (@pxref{Variable Attributes}).
4734 @item -Wunused-const-variable
4735 @itemx -Wunused-const-variable=@var{n}
4736 @opindex Wunused-const-variable
4737 @opindex Wno-unused-const-variable
4738 Warn whenever a constant static variable is unused aside from its declaration.
4739 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4740 for C, but not for C++. In C this declares variable storage, but in C++ this
4741 is not an error since const variables take the place of @code{#define}s.
4743 To suppress this warning use the @code{unused} attribute
4744 (@pxref{Variable Attributes}).
4747 @item -Wunused-const-variable=1
4748 This is the warning level that is enabled by @option{-Wunused-variable} for
4749 C. It warns only about unused static const variables defined in the main
4750 compilation unit, but not about static const variables declared in any
4753 @item -Wunused-const-variable=2
4754 This warning level also warns for unused constant static variables in
4755 headers (excluding system headers). This is the warning level of
4756 @option{-Wunused-const-variable} and must be explicitly requested since
4757 in C++ this isn't an error and in C it might be harder to clean up all
4761 @item -Wunused-value
4762 @opindex Wunused-value
4763 @opindex Wno-unused-value
4764 Warn whenever a statement computes a result that is explicitly not
4765 used. To suppress this warning cast the unused expression to
4766 @code{void}. This includes an expression-statement or the left-hand
4767 side of a comma expression that contains no side effects. For example,
4768 an expression such as @code{x[i,j]} causes a warning, while
4769 @code{x[(void)i,j]} does not.
4771 This warning is enabled by @option{-Wall}.
4776 All the above @option{-Wunused} options combined.
4778 In order to get a warning about an unused function parameter, you must
4779 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4780 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4782 @item -Wuninitialized
4783 @opindex Wuninitialized
4784 @opindex Wno-uninitialized
4785 Warn if an automatic variable is used without first being initialized
4786 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4787 warn if a non-static reference or non-static @code{const} member
4788 appears in a class without constructors.
4790 If you want to warn about code that uses the uninitialized value of the
4791 variable in its own initializer, use the @option{-Winit-self} option.
4793 These warnings occur for individual uninitialized or clobbered
4794 elements of structure, union or array variables as well as for
4795 variables that are uninitialized or clobbered as a whole. They do
4796 not occur for variables or elements declared @code{volatile}. Because
4797 these warnings depend on optimization, the exact variables or elements
4798 for which there are warnings depends on the precise optimization
4799 options and version of GCC used.
4801 Note that there may be no warning about a variable that is used only
4802 to compute a value that itself is never used, because such
4803 computations may be deleted by data flow analysis before the warnings
4806 @item -Winvalid-memory-model
4807 @opindex Winvalid-memory-model
4808 @opindex Wno-invalid-memory-model
4809 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4810 and the C11 atomic generic functions with a memory consistency argument
4811 that is either invalid for the operation or outside the range of values
4812 of the @code{memory_order} enumeration. For example, since the
4813 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4814 defined for the relaxed, release, and sequentially consistent memory
4815 orders the following code is diagnosed:
4820 __atomic_store_n (i, 0, memory_order_consume);
4824 @option{-Winvalid-memory-model} is enabled by default.
4826 @item -Wmaybe-uninitialized
4827 @opindex Wmaybe-uninitialized
4828 @opindex Wno-maybe-uninitialized
4829 For an automatic variable, if there exists a path from the function
4830 entry to a use of the variable that is initialized, but there exist
4831 some other paths for which the variable is not initialized, the compiler
4832 emits a warning if it cannot prove the uninitialized paths are not
4833 executed at run time. These warnings are made optional because GCC is
4834 not smart enough to see all the reasons why the code might be correct
4835 in spite of appearing to have an error. Here is one example of how
4856 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4857 always initialized, but GCC doesn't know this. To suppress the
4858 warning, you need to provide a default case with assert(0) or
4861 @cindex @code{longjmp} warnings
4862 This option also warns when a non-volatile automatic variable might be
4863 changed by a call to @code{longjmp}. These warnings as well are possible
4864 only in optimizing compilation.
4866 The compiler sees only the calls to @code{setjmp}. It cannot know
4867 where @code{longjmp} will be called; in fact, a signal handler could
4868 call it at any point in the code. As a result, you may get a warning
4869 even when there is in fact no problem because @code{longjmp} cannot
4870 in fact be called at the place that would cause a problem.
4872 Some spurious warnings can be avoided if you declare all the functions
4873 you use that never return as @code{noreturn}. @xref{Function
4876 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4878 @item -Wunknown-pragmas
4879 @opindex Wunknown-pragmas
4880 @opindex Wno-unknown-pragmas
4881 @cindex warning for unknown pragmas
4882 @cindex unknown pragmas, warning
4883 @cindex pragmas, warning of unknown
4884 Warn when a @code{#pragma} directive is encountered that is not understood by
4885 GCC@. If this command-line option is used, warnings are even issued
4886 for unknown pragmas in system header files. This is not the case if
4887 the warnings are only enabled by the @option{-Wall} command-line option.
4890 @opindex Wno-pragmas
4892 Do not warn about misuses of pragmas, such as incorrect parameters,
4893 invalid syntax, or conflicts between pragmas. See also
4894 @option{-Wunknown-pragmas}.
4896 @item -Wstrict-aliasing
4897 @opindex Wstrict-aliasing
4898 @opindex Wno-strict-aliasing
4899 This option is only active when @option{-fstrict-aliasing} is active.
4900 It warns about code that might break the strict aliasing rules that the
4901 compiler is using for optimization. The warning does not catch all
4902 cases, but does attempt to catch the more common pitfalls. It is
4903 included in @option{-Wall}.
4904 It is equivalent to @option{-Wstrict-aliasing=3}
4906 @item -Wstrict-aliasing=n
4907 @opindex Wstrict-aliasing=n
4908 This option is only active when @option{-fstrict-aliasing} is active.
4909 It warns about code that might break the strict aliasing rules that the
4910 compiler is using for optimization.
4911 Higher levels correspond to higher accuracy (fewer false positives).
4912 Higher levels also correspond to more effort, similar to the way @option{-O}
4914 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4916 Level 1: Most aggressive, quick, least accurate.
4917 Possibly useful when higher levels
4918 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4919 false negatives. However, it has many false positives.
4920 Warns for all pointer conversions between possibly incompatible types,
4921 even if never dereferenced. Runs in the front end only.
4923 Level 2: Aggressive, quick, not too precise.
4924 May still have many false positives (not as many as level 1 though),
4925 and few false negatives (but possibly more than level 1).
4926 Unlike level 1, it only warns when an address is taken. Warns about
4927 incomplete types. Runs in the front end only.
4929 Level 3 (default for @option{-Wstrict-aliasing}):
4930 Should have very few false positives and few false
4931 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4932 Takes care of the common pun+dereference pattern in the front end:
4933 @code{*(int*)&some_float}.
4934 If optimization is enabled, it also runs in the back end, where it deals
4935 with multiple statement cases using flow-sensitive points-to information.
4936 Only warns when the converted pointer is dereferenced.
4937 Does not warn about incomplete types.
4939 @item -Wstrict-overflow
4940 @itemx -Wstrict-overflow=@var{n}
4941 @opindex Wstrict-overflow
4942 @opindex Wno-strict-overflow
4943 This option is only active when signed overflow is undefined.
4944 It warns about cases where the compiler optimizes based on the
4945 assumption that signed overflow does not occur. Note that it does not
4946 warn about all cases where the code might overflow: it only warns
4947 about cases where the compiler implements some optimization. Thus
4948 this warning depends on the optimization level.
4950 An optimization that assumes that signed overflow does not occur is
4951 perfectly safe if the values of the variables involved are such that
4952 overflow never does, in fact, occur. Therefore this warning can
4953 easily give a false positive: a warning about code that is not
4954 actually a problem. To help focus on important issues, several
4955 warning levels are defined. No warnings are issued for the use of
4956 undefined signed overflow when estimating how many iterations a loop
4957 requires, in particular when determining whether a loop will be
4961 @item -Wstrict-overflow=1
4962 Warn about cases that are both questionable and easy to avoid. For
4963 example the compiler simplifies
4964 @code{x + 1 > x} to @code{1}. This level of
4965 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4966 are not, and must be explicitly requested.
4968 @item -Wstrict-overflow=2
4969 Also warn about other cases where a comparison is simplified to a
4970 constant. For example: @code{abs (x) >= 0}. This can only be
4971 simplified when signed integer overflow is undefined, because
4972 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4973 zero. @option{-Wstrict-overflow} (with no level) is the same as
4974 @option{-Wstrict-overflow=2}.
4976 @item -Wstrict-overflow=3
4977 Also warn about other cases where a comparison is simplified. For
4978 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4980 @item -Wstrict-overflow=4
4981 Also warn about other simplifications not covered by the above cases.
4982 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4984 @item -Wstrict-overflow=5
4985 Also warn about cases where the compiler reduces the magnitude of a
4986 constant involved in a comparison. For example: @code{x + 2 > y} is
4987 simplified to @code{x + 1 >= y}. This is reported only at the
4988 highest warning level because this simplification applies to many
4989 comparisons, so this warning level gives a very large number of
4993 @item -Wstringop-overflow
4994 @itemx -Wstringop-overflow=@var{type}
4995 @opindex Wstringop-overflow
4996 @opindex Wno-stringop-overflow
4997 Warn for calls to string manipulation functions such as @code{memcpy} and
4998 @code{strcpy} that are determined to overflow the destination buffer. The
4999 optional argument is one greater than the type of Object Size Checking to
5000 perform to determine the size of the destination. @xref{Object Size Checking}.
5001 The argument is meaningful only for functions that operate on character arrays
5002 but not for raw memory functions like @code{memcpy} which always make use
5003 of Object Size type-0. The option also warns for calls that specify a size
5004 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5005 The option produces the best results with optimization enabled but can detect
5006 a small subset of simple buffer overflows even without optimization in
5007 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5008 correspond to the standard functions. In any case, the option warns about
5009 just a subset of buffer overflows detected by the corresponding overflow
5010 checking built-ins. For example, the option will issue a warning for
5011 the @code{strcpy} call below because it copies at least 5 characters
5012 (the string @code{"blue"} including the terminating NUL) into the buffer
5016 enum Color @{ blue, purple, yellow @};
5017 const char* f (enum Color clr)
5019 static char buf [4];
5023 case blue: str = "blue"; break;
5024 case purple: str = "purple"; break;
5025 case yellow: str = "yellow"; break;
5028 return strcpy (buf, str); // warning here
5032 Option @option{-Wstringop-overflow=2} is enabled by default.
5035 @item -Wstringop-overflow
5036 @item -Wstringop-overflow=1
5037 @opindex Wstringop-overflow
5038 @opindex Wno-stringop-overflow
5039 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5040 to determine the sizes of destination objects. This is the default setting
5041 of the option. At this setting the option will not warn for writes past
5042 the end of subobjects of larger objects accessed by pointers unless the
5043 size of the largest surrounding object is known. When the destination may
5044 be one of several objects it is assumed to be the largest one of them. On
5045 Linux systems, when optimization is enabled at this setting the option warns
5046 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5049 @item -Wstringop-overflow=2
5050 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5051 to determine the sizes of destination objects. At this setting the option
5052 will warn about overflows when writing to members of the largest complete
5053 objects whose exact size is known. It will, however, not warn for excessive
5054 writes to the same members of unknown objects referenced by pointers since
5055 they may point to arrays containing unknown numbers of elements.
5057 @item -Wstringop-overflow=3
5058 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5059 to determine the sizes of destination objects. At this setting the option
5060 warns about overflowing the smallest object or data member. This is the
5061 most restrictive setting of the option that may result in warnings for safe
5064 @item -Wstringop-overflow=4
5065 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5066 to determine the sizes of destination objects. At this setting the option
5067 will warn about overflowing any data members, and when the destination is
5068 one of several objects it uses the size of the largest of them to decide
5069 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5070 setting of the option may result in warnings for benign code.
5073 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
5074 @opindex Wsuggest-attribute=
5075 @opindex Wno-suggest-attribute=
5076 Warn for cases where adding an attribute may be beneficial. The
5077 attributes currently supported are listed below.
5080 @item -Wsuggest-attribute=pure
5081 @itemx -Wsuggest-attribute=const
5082 @itemx -Wsuggest-attribute=noreturn
5083 @opindex Wsuggest-attribute=pure
5084 @opindex Wno-suggest-attribute=pure
5085 @opindex Wsuggest-attribute=const
5086 @opindex Wno-suggest-attribute=const
5087 @opindex Wsuggest-attribute=noreturn
5088 @opindex Wno-suggest-attribute=noreturn
5090 Warn about functions that might be candidates for attributes
5091 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
5092 functions visible in other compilation units or (in the case of @code{pure} and
5093 @code{const}) if it cannot prove that the function returns normally. A function
5094 returns normally if it doesn't contain an infinite loop or return abnormally
5095 by throwing, calling @code{abort} or trapping. This analysis requires option
5096 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
5097 higher. Higher optimization levels improve the accuracy of the analysis.
5099 @item -Wsuggest-attribute=format
5100 @itemx -Wmissing-format-attribute
5101 @opindex Wsuggest-attribute=format
5102 @opindex Wmissing-format-attribute
5103 @opindex Wno-suggest-attribute=format
5104 @opindex Wno-missing-format-attribute
5108 Warn about function pointers that might be candidates for @code{format}
5109 attributes. Note these are only possible candidates, not absolute ones.
5110 GCC guesses that function pointers with @code{format} attributes that
5111 are used in assignment, initialization, parameter passing or return
5112 statements should have a corresponding @code{format} attribute in the
5113 resulting type. I.e.@: the left-hand side of the assignment or
5114 initialization, the type of the parameter variable, or the return type
5115 of the containing function respectively should also have a @code{format}
5116 attribute to avoid the warning.
5118 GCC also warns about function definitions that might be
5119 candidates for @code{format} attributes. Again, these are only
5120 possible candidates. GCC guesses that @code{format} attributes
5121 might be appropriate for any function that calls a function like
5122 @code{vprintf} or @code{vscanf}, but this might not always be the
5123 case, and some functions for which @code{format} attributes are
5124 appropriate may not be detected.
5127 @item -Wsuggest-final-types
5128 @opindex Wno-suggest-final-types
5129 @opindex Wsuggest-final-types
5130 Warn about types with virtual methods where code quality would be improved
5131 if the type were declared with the C++11 @code{final} specifier,
5133 declared in an anonymous namespace. This allows GCC to more aggressively
5134 devirtualize the polymorphic calls. This warning is more effective with link
5135 time optimization, where the information about the class hierarchy graph is
5138 @item -Wsuggest-final-methods
5139 @opindex Wno-suggest-final-methods
5140 @opindex Wsuggest-final-methods
5141 Warn about virtual methods where code quality would be improved if the method
5142 were declared with the C++11 @code{final} specifier,
5143 or, if possible, its type were
5144 declared in an anonymous namespace or with the @code{final} specifier.
5146 more effective with link-time optimization, where the information about the
5147 class hierarchy graph is more complete. It is recommended to first consider
5148 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5151 @item -Wsuggest-override
5152 Warn about overriding virtual functions that are not marked with the override
5156 @opindex Wno-alloc-zero
5157 @opindex Walloc-zero
5158 Warn about calls to allocation functions decorated with attribute
5159 @code{alloc_size} that specify zero bytes, including those to the built-in
5160 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5161 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5162 when called with a zero size differs among implementations (and in the case
5163 of @code{realloc} has been deprecated) relying on it may result in subtle
5164 portability bugs and should be avoided.
5166 @item -Walloc-size-larger-than=@var{n}
5167 Warn about calls to functions decorated with attribute @code{alloc_size}
5168 that attempt to allocate objects larger than the specified number of bytes,
5169 or where the result of the size computation in an integer type with infinite
5170 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5171 may end in one of the standard suffixes designating a multiple of bytes
5172 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5173 @code{MB} and @code{MiB} for megabyte and mebibyte, and so on.
5174 @xref{Function Attributes}.
5179 This option warns on all uses of @code{alloca} in the source.
5181 @item -Walloca-larger-than=@var{n}
5182 This option warns on calls to @code{alloca} that are not bounded by a
5183 controlling predicate limiting its argument of integer type to at most
5184 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5185 Arguments of non-integer types are considered unbounded even if they
5186 appear to be constrained to the expected range.
5188 For example, a bounded case of @code{alloca} could be:
5191 void func (size_t n)
5202 In the above example, passing @code{-Walloca-larger-than=1000} would not
5203 issue a warning because the call to @code{alloca} is known to be at most
5204 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5205 the compiler would emit a warning.
5207 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5208 controlling predicate constraining its integer argument. For example:
5213 void *p = alloca (n);
5218 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5219 a warning, but this time because of the lack of bounds checking.
5221 Note, that even seemingly correct code involving signed integers could
5225 void func (signed int n)
5235 In the above example, @var{n} could be negative, causing a larger than
5236 expected argument to be implicitly cast into the @code{alloca} call.
5238 This option also warns when @code{alloca} is used in a loop.
5240 This warning is not enabled by @option{-Wall}, and is only active when
5241 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5243 See also @option{-Wvla-larger-than=@var{n}}.
5245 @item -Warray-bounds
5246 @itemx -Warray-bounds=@var{n}
5247 @opindex Wno-array-bounds
5248 @opindex Warray-bounds
5249 This option is only active when @option{-ftree-vrp} is active
5250 (default for @option{-O2} and above). It warns about subscripts to arrays
5251 that are always out of bounds. This warning is enabled by @option{-Wall}.
5254 @item -Warray-bounds=1
5255 This is the warning level of @option{-Warray-bounds} and is enabled
5256 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5258 @item -Warray-bounds=2
5259 This warning level also warns about out of bounds access for
5260 arrays at the end of a struct and for arrays accessed through
5261 pointers. This warning level may give a larger number of
5262 false positives and is deactivated by default.
5265 @item -Wbool-compare
5266 @opindex Wno-bool-compare
5267 @opindex Wbool-compare
5268 Warn about boolean expression compared with an integer value different from
5269 @code{true}/@code{false}. For instance, the following comparison is
5274 if ((n > 1) == 2) @{ @dots{} @}
5276 This warning is enabled by @option{-Wall}.
5278 @item -Wbool-operation
5279 @opindex Wno-bool-operation
5280 @opindex Wbool-operation
5281 Warn about suspicious operations on expressions of a boolean type. For
5282 instance, bitwise negation of a boolean is very likely a bug in the program.
5283 For C, this warning also warns about incrementing or decrementing a boolean,
5284 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5285 Incrementing a boolean is invalid in C++1z, and deprecated otherwise.)
5287 This warning is enabled by @option{-Wall}.
5289 @item -Wduplicated-branches
5290 @opindex Wno-duplicated-branches
5291 @opindex Wduplicated-branches
5292 Warn when an if-else has identical branches. This warning detects cases like
5299 It doesn't warn when both branches contain just a null statement. This warning
5300 also warn for conditional operators:
5302 int i = x ? *p : *p;
5305 @item -Wduplicated-cond
5306 @opindex Wno-duplicated-cond
5307 @opindex Wduplicated-cond
5308 Warn about duplicated conditions in an if-else-if chain. For instance,
5309 warn for the following code:
5311 if (p->q != NULL) @{ @dots{} @}
5312 else if (p->q != NULL) @{ @dots{} @}
5315 @item -Wframe-address
5316 @opindex Wno-frame-address
5317 @opindex Wframe-address
5318 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5319 is called with an argument greater than 0. Such calls may return indeterminate
5320 values or crash the program. The warning is included in @option{-Wall}.
5322 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5323 @opindex Wno-discarded-qualifiers
5324 @opindex Wdiscarded-qualifiers
5325 Do not warn if type qualifiers on pointers are being discarded.
5326 Typically, the compiler warns if a @code{const char *} variable is
5327 passed to a function that takes a @code{char *} parameter. This option
5328 can be used to suppress such a warning.
5330 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5331 @opindex Wno-discarded-array-qualifiers
5332 @opindex Wdiscarded-array-qualifiers
5333 Do not warn if type qualifiers on arrays which are pointer targets
5334 are being discarded. Typically, the compiler warns if a
5335 @code{const int (*)[]} variable is passed to a function that
5336 takes a @code{int (*)[]} parameter. This option can be used to
5337 suppress such a warning.
5339 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5340 @opindex Wno-incompatible-pointer-types
5341 @opindex Wincompatible-pointer-types
5342 Do not warn when there is a conversion between pointers that have incompatible
5343 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5344 which warns for pointer argument passing or assignment with different
5347 @item -Wno-int-conversion @r{(C and Objective-C only)}
5348 @opindex Wno-int-conversion
5349 @opindex Wint-conversion
5350 Do not warn about incompatible integer to pointer and pointer to integer
5351 conversions. This warning is about implicit conversions; for explicit
5352 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5353 @option{-Wno-pointer-to-int-cast} may be used.
5355 @item -Wno-div-by-zero
5356 @opindex Wno-div-by-zero
5357 @opindex Wdiv-by-zero
5358 Do not warn about compile-time integer division by zero. Floating-point
5359 division by zero is not warned about, as it can be a legitimate way of
5360 obtaining infinities and NaNs.
5362 @item -Wsystem-headers
5363 @opindex Wsystem-headers
5364 @opindex Wno-system-headers
5365 @cindex warnings from system headers
5366 @cindex system headers, warnings from
5367 Print warning messages for constructs found in system header files.
5368 Warnings from system headers are normally suppressed, on the assumption
5369 that they usually do not indicate real problems and would only make the
5370 compiler output harder to read. Using this command-line option tells
5371 GCC to emit warnings from system headers as if they occurred in user
5372 code. However, note that using @option{-Wall} in conjunction with this
5373 option does @emph{not} warn about unknown pragmas in system
5374 headers---for that, @option{-Wunknown-pragmas} must also be used.
5376 @item -Wtautological-compare
5377 @opindex Wtautological-compare
5378 @opindex Wno-tautological-compare
5379 Warn if a self-comparison always evaluates to true or false. This
5380 warning detects various mistakes such as:
5384 if (i > i) @{ @dots{} @}
5386 This warning is enabled by @option{-Wall}.
5389 @opindex Wtrampolines
5390 @opindex Wno-trampolines
5391 Warn about trampolines generated for pointers to nested functions.
5392 A trampoline is a small piece of data or code that is created at run
5393 time on the stack when the address of a nested function is taken, and is
5394 used to call the nested function indirectly. For some targets, it is
5395 made up of data only and thus requires no special treatment. But, for
5396 most targets, it is made up of code and thus requires the stack to be
5397 made executable in order for the program to work properly.
5400 @opindex Wfloat-equal
5401 @opindex Wno-float-equal
5402 Warn if floating-point values are used in equality comparisons.
5404 The idea behind this is that sometimes it is convenient (for the
5405 programmer) to consider floating-point values as approximations to
5406 infinitely precise real numbers. If you are doing this, then you need
5407 to compute (by analyzing the code, or in some other way) the maximum or
5408 likely maximum error that the computation introduces, and allow for it
5409 when performing comparisons (and when producing output, but that's a
5410 different problem). In particular, instead of testing for equality, you
5411 should check to see whether the two values have ranges that overlap; and
5412 this is done with the relational operators, so equality comparisons are
5415 @item -Wtraditional @r{(C and Objective-C only)}
5416 @opindex Wtraditional
5417 @opindex Wno-traditional
5418 Warn about certain constructs that behave differently in traditional and
5419 ISO C@. Also warn about ISO C constructs that have no traditional C
5420 equivalent, and/or problematic constructs that should be avoided.
5424 Macro parameters that appear within string literals in the macro body.
5425 In traditional C macro replacement takes place within string literals,
5426 but in ISO C it does not.
5429 In traditional C, some preprocessor directives did not exist.
5430 Traditional preprocessors only considered a line to be a directive
5431 if the @samp{#} appeared in column 1 on the line. Therefore
5432 @option{-Wtraditional} warns about directives that traditional C
5433 understands but ignores because the @samp{#} does not appear as the
5434 first character on the line. It also suggests you hide directives like
5435 @code{#pragma} not understood by traditional C by indenting them. Some
5436 traditional implementations do not recognize @code{#elif}, so this option
5437 suggests avoiding it altogether.
5440 A function-like macro that appears without arguments.
5443 The unary plus operator.
5446 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5447 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5448 constants.) Note, these suffixes appear in macros defined in the system
5449 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5450 Use of these macros in user code might normally lead to spurious
5451 warnings, however GCC's integrated preprocessor has enough context to
5452 avoid warning in these cases.
5455 A function declared external in one block and then used after the end of
5459 A @code{switch} statement has an operand of type @code{long}.
5462 A non-@code{static} function declaration follows a @code{static} one.
5463 This construct is not accepted by some traditional C compilers.
5466 The ISO type of an integer constant has a different width or
5467 signedness from its traditional type. This warning is only issued if
5468 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5469 typically represent bit patterns, are not warned about.
5472 Usage of ISO string concatenation is detected.
5475 Initialization of automatic aggregates.
5478 Identifier conflicts with labels. Traditional C lacks a separate
5479 namespace for labels.
5482 Initialization of unions. If the initializer is zero, the warning is
5483 omitted. This is done under the assumption that the zero initializer in
5484 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5485 initializer warnings and relies on default initialization to zero in the
5489 Conversions by prototypes between fixed/floating-point values and vice
5490 versa. The absence of these prototypes when compiling with traditional
5491 C causes serious problems. This is a subset of the possible
5492 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5495 Use of ISO C style function definitions. This warning intentionally is
5496 @emph{not} issued for prototype declarations or variadic functions
5497 because these ISO C features appear in your code when using
5498 libiberty's traditional C compatibility macros, @code{PARAMS} and
5499 @code{VPARAMS}. This warning is also bypassed for nested functions
5500 because that feature is already a GCC extension and thus not relevant to
5501 traditional C compatibility.
5504 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5505 @opindex Wtraditional-conversion
5506 @opindex Wno-traditional-conversion
5507 Warn if a prototype causes a type conversion that is different from what
5508 would happen to the same argument in the absence of a prototype. This
5509 includes conversions of fixed point to floating and vice versa, and
5510 conversions changing the width or signedness of a fixed-point argument
5511 except when the same as the default promotion.
5513 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5514 @opindex Wdeclaration-after-statement
5515 @opindex Wno-declaration-after-statement
5516 Warn when a declaration is found after a statement in a block. This
5517 construct, known from C++, was introduced with ISO C99 and is by default
5518 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5523 Warn whenever a local variable or type declaration shadows another
5524 variable, parameter, type, class member (in C++), or instance variable
5525 (in Objective-C) or whenever a built-in function is shadowed. Note
5526 that in C++, the compiler warns if a local variable shadows an
5527 explicit typedef, but not if it shadows a struct/class/enum.
5528 Same as @option{-Wshadow=global}.
5530 @item -Wno-shadow-ivar @r{(Objective-C only)}
5531 @opindex Wno-shadow-ivar
5532 @opindex Wshadow-ivar
5533 Do not warn whenever a local variable shadows an instance variable in an
5536 @item -Wshadow=global
5537 @opindex Wshadow=local
5538 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5540 @item -Wshadow=local
5541 @opindex Wshadow=local
5542 Warn when a local variable shadows another local variable or parameter.
5543 This warning is enabled by @option{-Wshadow=global}.
5545 @item -Wshadow=compatible-local
5546 @opindex Wshadow=compatible-local
5547 Warn when a local variable shadows another local variable or parameter
5548 whose type is compatible with that of the shadowing variable. In C++,
5549 type compatibility here means the type of the shadowing variable can be
5550 converted to that of the shadowed variable. The creation of this flag
5551 (in addition to @option{-Wshadow=local}) is based on the idea that when
5552 a local variable shadows another one of incompatible type, it is most
5553 likely intentional, not a bug or typo, as shown in the following example:
5557 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5559 for (int i = 0; i < N; ++i)
5568 Since the two variable @code{i} in the example above have incompatible types,
5569 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5570 Because their types are incompatible, if a programmer accidentally uses one
5571 in place of the other, type checking will catch that and emit an error or
5572 warning. So not warning (about shadowing) in this case will not lead to
5573 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5574 possibly reduce the number of warnings triggered by intentional shadowing.
5576 This warning is enabled by @option{-Wshadow=local}.
5578 @item -Wlarger-than=@var{len}
5579 @opindex Wlarger-than=@var{len}
5580 @opindex Wlarger-than-@var{len}
5581 Warn whenever an object of larger than @var{len} bytes is defined.
5583 @item -Wframe-larger-than=@var{len}
5584 @opindex Wframe-larger-than
5585 Warn if the size of a function frame is larger than @var{len} bytes.
5586 The computation done to determine the stack frame size is approximate
5587 and not conservative.
5588 The actual requirements may be somewhat greater than @var{len}
5589 even if you do not get a warning. In addition, any space allocated
5590 via @code{alloca}, variable-length arrays, or related constructs
5591 is not included by the compiler when determining
5592 whether or not to issue a warning.
5594 @item -Wno-free-nonheap-object
5595 @opindex Wno-free-nonheap-object
5596 @opindex Wfree-nonheap-object
5597 Do not warn when attempting to free an object that was not allocated
5600 @item -Wstack-usage=@var{len}
5601 @opindex Wstack-usage
5602 Warn if the stack usage of a function might be larger than @var{len} bytes.
5603 The computation done to determine the stack usage is conservative.
5604 Any space allocated via @code{alloca}, variable-length arrays, or related
5605 constructs is included by the compiler when determining whether or not to
5608 The message is in keeping with the output of @option{-fstack-usage}.
5612 If the stack usage is fully static but exceeds the specified amount, it's:
5615 warning: stack usage is 1120 bytes
5618 If the stack usage is (partly) dynamic but bounded, it's:
5621 warning: stack usage might be 1648 bytes
5624 If the stack usage is (partly) dynamic and not bounded, it's:
5627 warning: stack usage might be unbounded
5631 @item -Wunsafe-loop-optimizations
5632 @opindex Wunsafe-loop-optimizations
5633 @opindex Wno-unsafe-loop-optimizations
5634 Warn if the loop cannot be optimized because the compiler cannot
5635 assume anything on the bounds of the loop indices. With
5636 @option{-funsafe-loop-optimizations} warn if the compiler makes
5639 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5640 @opindex Wno-pedantic-ms-format
5641 @opindex Wpedantic-ms-format
5642 When used in combination with @option{-Wformat}
5643 and @option{-pedantic} without GNU extensions, this option
5644 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5645 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5646 which depend on the MS runtime.
5649 @opindex Waligned-new
5650 @opindex Wno-aligned-new
5651 Warn about a new-expression of a type that requires greater alignment
5652 than the @code{alignof(std::max_align_t)} but uses an allocation
5653 function without an explicit alignment parameter. This option is
5654 enabled by @option{-Wall}.
5656 Normally this only warns about global allocation functions, but
5657 @option{-Waligned-new=all} also warns about class member allocation
5660 @item -Wplacement-new
5661 @itemx -Wplacement-new=@var{n}
5662 @opindex Wplacement-new
5663 @opindex Wno-placement-new
5664 Warn about placement new expressions with undefined behavior, such as
5665 constructing an object in a buffer that is smaller than the type of
5666 the object. For example, the placement new expression below is diagnosed
5667 because it attempts to construct an array of 64 integers in a buffer only
5673 This warning is enabled by default.
5676 @item -Wplacement-new=1
5677 This is the default warning level of @option{-Wplacement-new}. At this
5678 level the warning is not issued for some strictly undefined constructs that
5679 GCC allows as extensions for compatibility with legacy code. For example,
5680 the following @code{new} expression is not diagnosed at this level even
5681 though it has undefined behavior according to the C++ standard because
5682 it writes past the end of the one-element array.
5684 struct S @{ int n, a[1]; @};
5685 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5686 new (s->a)int [32]();
5689 @item -Wplacement-new=2
5690 At this level, in addition to diagnosing all the same constructs as at level
5691 1, a diagnostic is also issued for placement new expressions that construct
5692 an object in the last member of structure whose type is an array of a single
5693 element and whose size is less than the size of the object being constructed.
5694 While the previous example would be diagnosed, the following construct makes
5695 use of the flexible member array extension to avoid the warning at level 2.
5697 struct S @{ int n, a[]; @};
5698 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5699 new (s->a)int [32]();
5704 @item -Wpointer-arith
5705 @opindex Wpointer-arith
5706 @opindex Wno-pointer-arith
5707 Warn about anything that depends on the ``size of'' a function type or
5708 of @code{void}. GNU C assigns these types a size of 1, for
5709 convenience in calculations with @code{void *} pointers and pointers
5710 to functions. In C++, warn also when an arithmetic operation involves
5711 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5713 @item -Wpointer-compare
5714 @opindex Wpointer-compare
5715 @opindex Wno-pointer-compare
5716 Warn if a pointer is compared with a zero character constant. This usually
5717 means that the pointer was meant to be dereferenced. For example:
5720 const char *p = foo ();
5725 Note that the code above is invalid in C++11.
5727 This warning is enabled by default.
5730 @opindex Wtype-limits
5731 @opindex Wno-type-limits
5732 Warn if a comparison is always true or always false due to the limited
5733 range of the data type, but do not warn for constant expressions. For
5734 example, warn if an unsigned variable is compared against zero with
5735 @code{<} or @code{>=}. This warning is also enabled by
5738 @include cppwarnopts.texi
5740 @item -Wbad-function-cast @r{(C and Objective-C only)}
5741 @opindex Wbad-function-cast
5742 @opindex Wno-bad-function-cast
5743 Warn when a function call is cast to a non-matching type.
5744 For example, warn if a call to a function returning an integer type
5745 is cast to a pointer type.
5747 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5748 @opindex Wc90-c99-compat
5749 @opindex Wno-c90-c99-compat
5750 Warn about features not present in ISO C90, but present in ISO C99.
5751 For instance, warn about use of variable length arrays, @code{long long}
5752 type, @code{bool} type, compound literals, designated initializers, and so
5753 on. This option is independent of the standards mode. Warnings are disabled
5754 in the expression that follows @code{__extension__}.
5756 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5757 @opindex Wc99-c11-compat
5758 @opindex Wno-c99-c11-compat
5759 Warn about features not present in ISO C99, but present in ISO C11.
5760 For instance, warn about use of anonymous structures and unions,
5761 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5762 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5763 and so on. This option is independent of the standards mode. Warnings are
5764 disabled in the expression that follows @code{__extension__}.
5766 @item -Wc++-compat @r{(C and Objective-C only)}
5767 @opindex Wc++-compat
5768 Warn about ISO C constructs that are outside of the common subset of
5769 ISO C and ISO C++, e.g.@: request for implicit conversion from
5770 @code{void *} to a pointer to non-@code{void} type.
5772 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5773 @opindex Wc++11-compat
5774 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5775 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5776 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5777 enabled by @option{-Wall}.
5779 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5780 @opindex Wc++14-compat
5781 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5782 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5784 @item -Wc++1z-compat @r{(C++ and Objective-C++ only)}
5785 @opindex Wc++1z-compat
5786 Warn about C++ constructs whose meaning differs between ISO C++ 2014
5787 and the forthoming ISO C++ 2017(?). This warning is enabled by @option{-Wall}.
5791 @opindex Wno-cast-qual
5792 Warn whenever a pointer is cast so as to remove a type qualifier from
5793 the target type. For example, warn if a @code{const char *} is cast
5794 to an ordinary @code{char *}.
5796 Also warn when making a cast that introduces a type qualifier in an
5797 unsafe way. For example, casting @code{char **} to @code{const char **}
5798 is unsafe, as in this example:
5801 /* p is char ** value. */
5802 const char **q = (const char **) p;
5803 /* Assignment of readonly string to const char * is OK. */
5805 /* Now char** pointer points to read-only memory. */
5810 @opindex Wcast-align
5811 @opindex Wno-cast-align
5812 Warn whenever a pointer is cast such that the required alignment of the
5813 target is increased. For example, warn if a @code{char *} is cast to
5814 an @code{int *} on machines where integers can only be accessed at
5815 two- or four-byte boundaries.
5817 @item -Wwrite-strings
5818 @opindex Wwrite-strings
5819 @opindex Wno-write-strings
5820 When compiling C, give string constants the type @code{const
5821 char[@var{length}]} so that copying the address of one into a
5822 non-@code{const} @code{char *} pointer produces a warning. These
5823 warnings help you find at compile time code that can try to write
5824 into a string constant, but only if you have been very careful about
5825 using @code{const} in declarations and prototypes. Otherwise, it is
5826 just a nuisance. This is why we did not make @option{-Wall} request
5829 When compiling C++, warn about the deprecated conversion from string
5830 literals to @code{char *}. This warning is enabled by default for C++
5835 @opindex Wno-clobbered
5836 Warn for variables that might be changed by @code{longjmp} or
5837 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5839 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5840 @opindex Wconditionally-supported
5841 @opindex Wno-conditionally-supported
5842 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5845 @opindex Wconversion
5846 @opindex Wno-conversion
5847 Warn for implicit conversions that may alter a value. This includes
5848 conversions between real and integer, like @code{abs (x)} when
5849 @code{x} is @code{double}; conversions between signed and unsigned,
5850 like @code{unsigned ui = -1}; and conversions to smaller types, like
5851 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5852 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5853 changed by the conversion like in @code{abs (2.0)}. Warnings about
5854 conversions between signed and unsigned integers can be disabled by
5855 using @option{-Wno-sign-conversion}.
5857 For C++, also warn for confusing overload resolution for user-defined
5858 conversions; and conversions that never use a type conversion
5859 operator: conversions to @code{void}, the same type, a base class or a
5860 reference to them. Warnings about conversions between signed and
5861 unsigned integers are disabled by default in C++ unless
5862 @option{-Wsign-conversion} is explicitly enabled.
5864 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5865 @opindex Wconversion-null
5866 @opindex Wno-conversion-null
5867 Do not warn for conversions between @code{NULL} and non-pointer
5868 types. @option{-Wconversion-null} is enabled by default.
5870 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5871 @opindex Wzero-as-null-pointer-constant
5872 @opindex Wno-zero-as-null-pointer-constant
5873 Warn when a literal @samp{0} is used as null pointer constant. This can
5874 be useful to facilitate the conversion to @code{nullptr} in C++11.
5876 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5877 @opindex Wsubobject-linkage
5878 @opindex Wno-subobject-linkage
5879 Warn if a class type has a base or a field whose type uses the anonymous
5880 namespace or depends on a type with no linkage. If a type A depends on
5881 a type B with no or internal linkage, defining it in multiple
5882 translation units would be an ODR violation because the meaning of B
5883 is different in each translation unit. If A only appears in a single
5884 translation unit, the best way to silence the warning is to give it
5885 internal linkage by putting it in an anonymous namespace as well. The
5886 compiler doesn't give this warning for types defined in the main .C
5887 file, as those are unlikely to have multiple definitions.
5888 @option{-Wsubobject-linkage} is enabled by default.
5890 @item -Wdangling-else
5891 @opindex Wdangling-else
5892 @opindex Wno-dangling-else
5893 Warn about constructions where there may be confusion to which
5894 @code{if} statement an @code{else} branch belongs. Here is an example of
5909 In C/C++, every @code{else} branch belongs to the innermost possible
5910 @code{if} statement, which in this example is @code{if (b)}. This is
5911 often not what the programmer expected, as illustrated in the above
5912 example by indentation the programmer chose. When there is the
5913 potential for this confusion, GCC issues a warning when this flag
5914 is specified. To eliminate the warning, add explicit braces around
5915 the innermost @code{if} statement so there is no way the @code{else}
5916 can belong to the enclosing @code{if}. The resulting code
5933 This warning is enabled by @option{-Wparentheses}.
5937 @opindex Wno-date-time
5938 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5939 are encountered as they might prevent bit-wise-identical reproducible
5942 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5943 @opindex Wdelete-incomplete
5944 @opindex Wno-delete-incomplete
5945 Warn when deleting a pointer to incomplete type, which may cause
5946 undefined behavior at runtime. This warning is enabled by default.
5948 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5949 @opindex Wuseless-cast
5950 @opindex Wno-useless-cast
5951 Warn when an expression is casted to its own type.
5954 @opindex Wempty-body
5955 @opindex Wno-empty-body
5956 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5957 while} statement. This warning is also enabled by @option{-Wextra}.
5959 @item -Wenum-compare
5960 @opindex Wenum-compare
5961 @opindex Wno-enum-compare
5962 Warn about a comparison between values of different enumerated types.
5963 In C++ enumerated type mismatches in conditional expressions are also
5964 diagnosed and the warning is enabled by default. In C this warning is
5965 enabled by @option{-Wall}.
5967 @item -Wextra-semi @r{(C++, Objective-C++ only)}
5968 @opindex Wextra-semi
5969 @opindex Wno-extra-semi
5970 Warn about redundant semicolon after in-class function definition.
5972 @item -Wjump-misses-init @r{(C, Objective-C only)}
5973 @opindex Wjump-misses-init
5974 @opindex Wno-jump-misses-init
5975 Warn if a @code{goto} statement or a @code{switch} statement jumps
5976 forward across the initialization of a variable, or jumps backward to a
5977 label after the variable has been initialized. This only warns about
5978 variables that are initialized when they are declared. This warning is
5979 only supported for C and Objective-C; in C++ this sort of branch is an
5982 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5983 can be disabled with the @option{-Wno-jump-misses-init} option.
5985 @item -Wsign-compare
5986 @opindex Wsign-compare
5987 @opindex Wno-sign-compare
5988 @cindex warning for comparison of signed and unsigned values
5989 @cindex comparison of signed and unsigned values, warning
5990 @cindex signed and unsigned values, comparison warning
5991 Warn when a comparison between signed and unsigned values could produce
5992 an incorrect result when the signed value is converted to unsigned.
5993 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5994 also enabled by @option{-Wextra}.
5996 @item -Wsign-conversion
5997 @opindex Wsign-conversion
5998 @opindex Wno-sign-conversion
5999 Warn for implicit conversions that may change the sign of an integer
6000 value, like assigning a signed integer expression to an unsigned
6001 integer variable. An explicit cast silences the warning. In C, this
6002 option is enabled also by @option{-Wconversion}.
6004 @item -Wfloat-conversion
6005 @opindex Wfloat-conversion
6006 @opindex Wno-float-conversion
6007 Warn for implicit conversions that reduce the precision of a real value.
6008 This includes conversions from real to integer, and from higher precision
6009 real to lower precision real values. This option is also enabled by
6010 @option{-Wconversion}.
6012 @item -Wno-scalar-storage-order
6013 @opindex -Wno-scalar-storage-order
6014 @opindex -Wscalar-storage-order
6015 Do not warn on suspicious constructs involving reverse scalar storage order.
6017 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6018 @opindex Wsized-deallocation
6019 @opindex Wno-sized-deallocation
6020 Warn about a definition of an unsized deallocation function
6022 void operator delete (void *) noexcept;
6023 void operator delete[] (void *) noexcept;
6025 without a definition of the corresponding sized deallocation function
6027 void operator delete (void *, std::size_t) noexcept;
6028 void operator delete[] (void *, std::size_t) noexcept;
6030 or vice versa. Enabled by @option{-Wextra} along with
6031 @option{-fsized-deallocation}.
6033 @item -Wsizeof-pointer-memaccess
6034 @opindex Wsizeof-pointer-memaccess
6035 @opindex Wno-sizeof-pointer-memaccess
6036 Warn for suspicious length parameters to certain string and memory built-in
6037 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
6038 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
6039 but a pointer, and suggests a possible fix, or about
6040 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
6043 @item -Wsizeof-array-argument
6044 @opindex Wsizeof-array-argument
6045 @opindex Wno-sizeof-array-argument
6046 Warn when the @code{sizeof} operator is applied to a parameter that is
6047 declared as an array in a function definition. This warning is enabled by
6048 default for C and C++ programs.
6050 @item -Wmemset-elt-size
6051 @opindex Wmemset-elt-size
6052 @opindex Wno-memset-elt-size
6053 Warn for suspicious calls to the @code{memset} built-in function, if the
6054 first argument references an array, and the third argument is a number
6055 equal to the number of elements, but not equal to the size of the array
6056 in memory. This indicates that the user has omitted a multiplication by
6057 the element size. This warning is enabled by @option{-Wall}.
6059 @item -Wmemset-transposed-args
6060 @opindex Wmemset-transposed-args
6061 @opindex Wno-memset-transposed-args
6062 Warn for suspicious calls to the @code{memset} built-in function, if the
6063 second argument is not zero and the third argument is zero. This warns e.g.@
6064 about @code{memset (buf, sizeof buf, 0)} where most probably
6065 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6066 is only emitted if the third argument is literal zero. If it is some
6067 expression that is folded to zero, a cast of zero to some type, etc.,
6068 it is far less likely that the user has mistakenly exchanged the arguments
6069 and no warning is emitted. This warning is enabled by @option{-Wall}.
6073 @opindex Wno-address
6074 Warn about suspicious uses of memory addresses. These include using
6075 the address of a function in a conditional expression, such as
6076 @code{void func(void); if (func)}, and comparisons against the memory
6077 address of a string literal, such as @code{if (x == "abc")}. Such
6078 uses typically indicate a programmer error: the address of a function
6079 always evaluates to true, so their use in a conditional usually
6080 indicate that the programmer forgot the parentheses in a function
6081 call; and comparisons against string literals result in unspecified
6082 behavior and are not portable in C, so they usually indicate that the
6083 programmer intended to use @code{strcmp}. This warning is enabled by
6087 @opindex Wlogical-op
6088 @opindex Wno-logical-op
6089 Warn about suspicious uses of logical operators in expressions.
6090 This includes using logical operators in contexts where a
6091 bit-wise operator is likely to be expected. Also warns when
6092 the operands of a logical operator are the same:
6095 if (a < 0 && a < 0) @{ @dots{} @}
6098 @item -Wlogical-not-parentheses
6099 @opindex Wlogical-not-parentheses
6100 @opindex Wno-logical-not-parentheses
6101 Warn about logical not used on the left hand side operand of a comparison.
6102 This option does not warn if the right operand is considered to be a boolean
6103 expression. Its purpose is to detect suspicious code like the following:
6107 if (!a > 1) @{ @dots{} @}
6110 It is possible to suppress the warning by wrapping the LHS into
6113 if ((!a) > 1) @{ @dots{} @}
6116 This warning is enabled by @option{-Wall}.
6118 @item -Waggregate-return
6119 @opindex Waggregate-return
6120 @opindex Wno-aggregate-return
6121 Warn if any functions that return structures or unions are defined or
6122 called. (In languages where you can return an array, this also elicits
6125 @item -Wno-aggressive-loop-optimizations
6126 @opindex Wno-aggressive-loop-optimizations
6127 @opindex Waggressive-loop-optimizations
6128 Warn if in a loop with constant number of iterations the compiler detects
6129 undefined behavior in some statement during one or more of the iterations.
6131 @item -Wno-attributes
6132 @opindex Wno-attributes
6133 @opindex Wattributes
6134 Do not warn if an unexpected @code{__attribute__} is used, such as
6135 unrecognized attributes, function attributes applied to variables,
6136 etc. This does not stop errors for incorrect use of supported
6139 @item -Wno-builtin-declaration-mismatch
6140 @opindex Wno-builtin-declaration-mismatch
6141 @opindex Wbuiltin-declaration-mismatch
6142 Warn if a built-in function is declared with the wrong signature.
6143 This warning is enabled by default.
6145 @item -Wno-builtin-macro-redefined
6146 @opindex Wno-builtin-macro-redefined
6147 @opindex Wbuiltin-macro-redefined
6148 Do not warn if certain built-in macros are redefined. This suppresses
6149 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6150 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6152 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6153 @opindex Wstrict-prototypes
6154 @opindex Wno-strict-prototypes
6155 Warn if a function is declared or defined without specifying the
6156 argument types. (An old-style function definition is permitted without
6157 a warning if preceded by a declaration that specifies the argument
6160 @item -Wold-style-declaration @r{(C and Objective-C only)}
6161 @opindex Wold-style-declaration
6162 @opindex Wno-old-style-declaration
6163 Warn for obsolescent usages, according to the C Standard, in a
6164 declaration. For example, warn if storage-class specifiers like
6165 @code{static} are not the first things in a declaration. This warning
6166 is also enabled by @option{-Wextra}.
6168 @item -Wold-style-definition @r{(C and Objective-C only)}
6169 @opindex Wold-style-definition
6170 @opindex Wno-old-style-definition
6171 Warn if an old-style function definition is used. A warning is given
6172 even if there is a previous prototype.
6174 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6175 @opindex Wmissing-parameter-type
6176 @opindex Wno-missing-parameter-type
6177 A function parameter is declared without a type specifier in K&R-style
6184 This warning is also enabled by @option{-Wextra}.
6186 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6187 @opindex Wmissing-prototypes
6188 @opindex Wno-missing-prototypes
6189 Warn if a global function is defined without a previous prototype
6190 declaration. This warning is issued even if the definition itself
6191 provides a prototype. Use this option to detect global functions
6192 that do not have a matching prototype declaration in a header file.
6193 This option is not valid for C++ because all function declarations
6194 provide prototypes and a non-matching declaration declares an
6195 overload rather than conflict with an earlier declaration.
6196 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6198 @item -Wmissing-declarations
6199 @opindex Wmissing-declarations
6200 @opindex Wno-missing-declarations
6201 Warn if a global function is defined without a previous declaration.
6202 Do so even if the definition itself provides a prototype.
6203 Use this option to detect global functions that are not declared in
6204 header files. In C, no warnings are issued for functions with previous
6205 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6206 missing prototypes. In C++, no warnings are issued for function templates,
6207 or for inline functions, or for functions in anonymous namespaces.
6209 @item -Wmissing-field-initializers
6210 @opindex Wmissing-field-initializers
6211 @opindex Wno-missing-field-initializers
6215 Warn if a structure's initializer has some fields missing. For
6216 example, the following code causes such a warning, because
6217 @code{x.h} is implicitly zero:
6220 struct s @{ int f, g, h; @};
6221 struct s x = @{ 3, 4 @};
6224 This option does not warn about designated initializers, so the following
6225 modification does not trigger a warning:
6228 struct s @{ int f, g, h; @};
6229 struct s x = @{ .f = 3, .g = 4 @};
6232 In C this option does not warn about the universal zero initializer
6236 struct s @{ int f, g, h; @};
6237 struct s x = @{ 0 @};
6240 Likewise, in C++ this option does not warn about the empty @{ @}
6241 initializer, for example:
6244 struct s @{ int f, g, h; @};
6248 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6249 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6251 @item -Wno-multichar
6252 @opindex Wno-multichar
6254 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6255 Usually they indicate a typo in the user's code, as they have
6256 implementation-defined values, and should not be used in portable code.
6258 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
6259 @opindex Wnormalized=
6260 @opindex Wnormalized
6261 @opindex Wno-normalized
6264 @cindex character set, input normalization
6265 In ISO C and ISO C++, two identifiers are different if they are
6266 different sequences of characters. However, sometimes when characters
6267 outside the basic ASCII character set are used, you can have two
6268 different character sequences that look the same. To avoid confusion,
6269 the ISO 10646 standard sets out some @dfn{normalization rules} which
6270 when applied ensure that two sequences that look the same are turned into
6271 the same sequence. GCC can warn you if you are using identifiers that
6272 have not been normalized; this option controls that warning.
6274 There are four levels of warning supported by GCC@. The default is
6275 @option{-Wnormalized=nfc}, which warns about any identifier that is
6276 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6277 recommended form for most uses. It is equivalent to
6278 @option{-Wnormalized}.
6280 Unfortunately, there are some characters allowed in identifiers by
6281 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6282 identifiers. That is, there's no way to use these symbols in portable
6283 ISO C or C++ and have all your identifiers in NFC@.
6284 @option{-Wnormalized=id} suppresses the warning for these characters.
6285 It is hoped that future versions of the standards involved will correct
6286 this, which is why this option is not the default.
6288 You can switch the warning off for all characters by writing
6289 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6290 only do this if you are using some other normalization scheme (like
6291 ``D''), because otherwise you can easily create bugs that are
6292 literally impossible to see.
6294 Some characters in ISO 10646 have distinct meanings but look identical
6295 in some fonts or display methodologies, especially once formatting has
6296 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6297 LETTER N'', displays just like a regular @code{n} that has been
6298 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6299 normalization scheme to convert all these into a standard form as
6300 well, and GCC warns if your code is not in NFKC if you use
6301 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6302 about every identifier that contains the letter O because it might be
6303 confused with the digit 0, and so is not the default, but may be
6304 useful as a local coding convention if the programming environment
6305 cannot be fixed to display these characters distinctly.
6307 @item -Wno-deprecated
6308 @opindex Wno-deprecated
6309 @opindex Wdeprecated
6310 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6312 @item -Wno-deprecated-declarations
6313 @opindex Wno-deprecated-declarations
6314 @opindex Wdeprecated-declarations
6315 Do not warn about uses of functions (@pxref{Function Attributes}),
6316 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6317 Attributes}) marked as deprecated by using the @code{deprecated}
6321 @opindex Wno-overflow
6323 Do not warn about compile-time overflow in constant expressions.
6328 Warn about One Definition Rule violations during link-time optimization.
6329 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6332 @opindex Wopenm-simd
6333 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6334 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6335 option can be used to relax the cost model.
6337 @item -Woverride-init @r{(C and Objective-C only)}
6338 @opindex Woverride-init
6339 @opindex Wno-override-init
6343 Warn if an initialized field without side effects is overridden when
6344 using designated initializers (@pxref{Designated Inits, , Designated
6347 This warning is included in @option{-Wextra}. To get other
6348 @option{-Wextra} warnings without this one, use @option{-Wextra
6349 -Wno-override-init}.
6351 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6352 @opindex Woverride-init-side-effects
6353 @opindex Wno-override-init-side-effects
6354 Warn if an initialized field with side effects is overridden when
6355 using designated initializers (@pxref{Designated Inits, , Designated
6356 Initializers}). This warning is enabled by default.
6361 Warn if a structure is given the packed attribute, but the packed
6362 attribute has no effect on the layout or size of the structure.
6363 Such structures may be mis-aligned for little benefit. For
6364 instance, in this code, the variable @code{f.x} in @code{struct bar}
6365 is misaligned even though @code{struct bar} does not itself
6366 have the packed attribute:
6373 @} __attribute__((packed));
6381 @item -Wpacked-bitfield-compat
6382 @opindex Wpacked-bitfield-compat
6383 @opindex Wno-packed-bitfield-compat
6384 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6385 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6386 the change can lead to differences in the structure layout. GCC
6387 informs you when the offset of such a field has changed in GCC 4.4.
6388 For example there is no longer a 4-bit padding between field @code{a}
6389 and @code{b} in this structure:
6396 @} __attribute__ ((packed));
6399 This warning is enabled by default. Use
6400 @option{-Wno-packed-bitfield-compat} to disable this warning.
6405 Warn if padding is included in a structure, either to align an element
6406 of the structure or to align the whole structure. Sometimes when this
6407 happens it is possible to rearrange the fields of the structure to
6408 reduce the padding and so make the structure smaller.
6410 @item -Wredundant-decls
6411 @opindex Wredundant-decls
6412 @opindex Wno-redundant-decls
6413 Warn if anything is declared more than once in the same scope, even in
6414 cases where multiple declaration is valid and changes nothing.
6418 @opindex Wno-restrict
6419 Warn when an argument passed to a restrict-qualified parameter
6420 aliases with another argument.
6422 @item -Wnested-externs @r{(C and Objective-C only)}
6423 @opindex Wnested-externs
6424 @opindex Wno-nested-externs
6425 Warn if an @code{extern} declaration is encountered within a function.
6427 @item -Wno-inherited-variadic-ctor
6428 @opindex Winherited-variadic-ctor
6429 @opindex Wno-inherited-variadic-ctor
6430 Suppress warnings about use of C++11 inheriting constructors when the
6431 base class inherited from has a C variadic constructor; the warning is
6432 on by default because the ellipsis is not inherited.
6437 Warn if a function that is declared as inline cannot be inlined.
6438 Even with this option, the compiler does not warn about failures to
6439 inline functions declared in system headers.
6441 The compiler uses a variety of heuristics to determine whether or not
6442 to inline a function. For example, the compiler takes into account
6443 the size of the function being inlined and the amount of inlining
6444 that has already been done in the current function. Therefore,
6445 seemingly insignificant changes in the source program can cause the
6446 warnings produced by @option{-Winline} to appear or disappear.
6448 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6449 @opindex Wno-invalid-offsetof
6450 @opindex Winvalid-offsetof
6451 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6452 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6453 to a non-standard-layout type is undefined. In existing C++ implementations,
6454 however, @code{offsetof} typically gives meaningful results.
6455 This flag is for users who are aware that they are
6456 writing nonportable code and who have deliberately chosen to ignore the
6459 The restrictions on @code{offsetof} may be relaxed in a future version
6460 of the C++ standard.
6462 @item -Wint-in-bool-context
6463 @opindex Wint-in-bool-context
6464 @opindex Wno-int-in-bool-context
6465 Warn for suspicious use of integer values where boolean values are expected,
6466 such as conditional expressions (?:) using non-boolean integer constants in
6467 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6468 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6469 for all kinds of multiplications regardless of the data type.
6470 This warning is enabled by @option{-Wall}.
6472 @item -Wno-int-to-pointer-cast
6473 @opindex Wno-int-to-pointer-cast
6474 @opindex Wint-to-pointer-cast
6475 Suppress warnings from casts to pointer type of an integer of a
6476 different size. In C++, casting to a pointer type of smaller size is
6477 an error. @option{Wint-to-pointer-cast} is enabled by default.
6480 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6481 @opindex Wno-pointer-to-int-cast
6482 @opindex Wpointer-to-int-cast
6483 Suppress warnings from casts from a pointer to an integer type of a
6487 @opindex Winvalid-pch
6488 @opindex Wno-invalid-pch
6489 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6490 the search path but cannot be used.
6494 @opindex Wno-long-long
6495 Warn if @code{long long} type is used. This is enabled by either
6496 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6497 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6499 @item -Wvariadic-macros
6500 @opindex Wvariadic-macros
6501 @opindex Wno-variadic-macros
6502 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6503 alternate syntax is used in ISO C99 mode. This is enabled by either
6504 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6505 messages, use @option{-Wno-variadic-macros}.
6509 @opindex Wno-varargs
6510 Warn upon questionable usage of the macros used to handle variable
6511 arguments like @code{va_start}. This is default. To inhibit the
6512 warning messages, use @option{-Wno-varargs}.
6514 @item -Wvector-operation-performance
6515 @opindex Wvector-operation-performance
6516 @opindex Wno-vector-operation-performance
6517 Warn if vector operation is not implemented via SIMD capabilities of the
6518 architecture. Mainly useful for the performance tuning.
6519 Vector operation can be implemented @code{piecewise}, which means that the
6520 scalar operation is performed on every vector element;
6521 @code{in parallel}, which means that the vector operation is implemented
6522 using scalars of wider type, which normally is more performance efficient;
6523 and @code{as a single scalar}, which means that vector fits into a
6526 @item -Wno-virtual-move-assign
6527 @opindex Wvirtual-move-assign
6528 @opindex Wno-virtual-move-assign
6529 Suppress warnings about inheriting from a virtual base with a
6530 non-trivial C++11 move assignment operator. This is dangerous because
6531 if the virtual base is reachable along more than one path, it is
6532 moved multiple times, which can mean both objects end up in the
6533 moved-from state. If the move assignment operator is written to avoid
6534 moving from a moved-from object, this warning can be disabled.
6539 Warn if a variable-length array is used in the code.
6540 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6541 the variable-length array.
6543 @item -Wvla-larger-than=@var{n}
6544 If this option is used, the compiler will warn on uses of
6545 variable-length arrays where the size is either unbounded, or bounded
6546 by an argument that can be larger than @var{n} bytes. This is similar
6547 to how @option{-Walloca-larger-than=@var{n}} works, but with
6548 variable-length arrays.
6550 Note that GCC may optimize small variable-length arrays of a known
6551 value into plain arrays, so this warning may not get triggered for
6554 This warning is not enabled by @option{-Wall}, and is only active when
6555 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6557 See also @option{-Walloca-larger-than=@var{n}}.
6559 @item -Wvolatile-register-var
6560 @opindex Wvolatile-register-var
6561 @opindex Wno-volatile-register-var
6562 Warn if a register variable is declared volatile. The volatile
6563 modifier does not inhibit all optimizations that may eliminate reads
6564 and/or writes to register variables. This warning is enabled by
6567 @item -Wdisabled-optimization
6568 @opindex Wdisabled-optimization
6569 @opindex Wno-disabled-optimization
6570 Warn if a requested optimization pass is disabled. This warning does
6571 not generally indicate that there is anything wrong with your code; it
6572 merely indicates that GCC's optimizers are unable to handle the code
6573 effectively. Often, the problem is that your code is too big or too
6574 complex; GCC refuses to optimize programs when the optimization
6575 itself is likely to take inordinate amounts of time.
6577 @item -Wpointer-sign @r{(C and Objective-C only)}
6578 @opindex Wpointer-sign
6579 @opindex Wno-pointer-sign
6580 Warn for pointer argument passing or assignment with different signedness.
6581 This option is only supported for C and Objective-C@. It is implied by
6582 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6583 @option{-Wno-pointer-sign}.
6585 @item -Wstack-protector
6586 @opindex Wstack-protector
6587 @opindex Wno-stack-protector
6588 This option is only active when @option{-fstack-protector} is active. It
6589 warns about functions that are not protected against stack smashing.
6591 @item -Woverlength-strings
6592 @opindex Woverlength-strings
6593 @opindex Wno-overlength-strings
6594 Warn about string constants that are longer than the ``minimum
6595 maximum'' length specified in the C standard. Modern compilers
6596 generally allow string constants that are much longer than the
6597 standard's minimum limit, but very portable programs should avoid
6598 using longer strings.
6600 The limit applies @emph{after} string constant concatenation, and does
6601 not count the trailing NUL@. In C90, the limit was 509 characters; in
6602 C99, it was raised to 4095. C++98 does not specify a normative
6603 minimum maximum, so we do not diagnose overlength strings in C++@.
6605 This option is implied by @option{-Wpedantic}, and can be disabled with
6606 @option{-Wno-overlength-strings}.
6608 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6609 @opindex Wunsuffixed-float-constants
6611 Issue a warning for any floating constant that does not have
6612 a suffix. When used together with @option{-Wsystem-headers} it
6613 warns about such constants in system header files. This can be useful
6614 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6615 from the decimal floating-point extension to C99.
6617 @item -Wno-designated-init @r{(C and Objective-C only)}
6618 Suppress warnings when a positional initializer is used to initialize
6619 a structure that has been marked with the @code{designated_init}
6623 Issue a warning when HSAIL cannot be emitted for the compiled function or
6628 @node Debugging Options
6629 @section Options for Debugging Your Program
6630 @cindex options, debugging
6631 @cindex debugging information options
6633 To tell GCC to emit extra information for use by a debugger, in almost
6634 all cases you need only to add @option{-g} to your other options.
6636 GCC allows you to use @option{-g} with
6637 @option{-O}. The shortcuts taken by optimized code may occasionally
6638 be surprising: some variables you declared may not exist
6639 at all; flow of control may briefly move where you did not expect it;
6640 some statements may not be executed because they compute constant
6641 results or their values are already at hand; some statements may
6642 execute in different places because they have been moved out of loops.
6643 Nevertheless it is possible to debug optimized output. This makes
6644 it reasonable to use the optimizer for programs that might have bugs.
6646 If you are not using some other optimization option, consider
6647 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6648 With no @option{-O} option at all, some compiler passes that collect
6649 information useful for debugging do not run at all, so that
6650 @option{-Og} may result in a better debugging experience.
6655 Produce debugging information in the operating system's native format
6656 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6659 On most systems that use stabs format, @option{-g} enables use of extra
6660 debugging information that only GDB can use; this extra information
6661 makes debugging work better in GDB but probably makes other debuggers
6663 refuse to read the program. If you want to control for certain whether
6664 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6665 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6669 Produce debugging information for use by GDB@. This means to use the
6670 most expressive format available (DWARF, stabs, or the native format
6671 if neither of those are supported), including GDB extensions if at all
6675 @itemx -gdwarf-@var{version}
6677 Produce debugging information in DWARF format (if that is supported).
6678 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6679 for most targets is 4. DWARF Version 5 is only experimental.
6681 Note that with DWARF Version 2, some ports require and always
6682 use some non-conflicting DWARF 3 extensions in the unwind tables.
6684 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6685 for maximum benefit.
6687 GCC no longer supports DWARF Version 1, which is substantially
6688 different than Version 2 and later. For historical reasons, some
6689 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
6690 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6691 in their names, but apply to all currently-supported versions of DWARF.
6695 Produce debugging information in stabs format (if that is supported),
6696 without GDB extensions. This is the format used by DBX on most BSD
6697 systems. On MIPS, Alpha and System V Release 4 systems this option
6698 produces stabs debugging output that is not understood by DBX or SDB@.
6699 On System V Release 4 systems this option requires the GNU assembler.
6703 Produce debugging information in stabs format (if that is supported),
6704 using GNU extensions understood only by the GNU debugger (GDB)@. The
6705 use of these extensions is likely to make other debuggers crash or
6706 refuse to read the program.
6710 Produce debugging information in COFF format (if that is supported).
6711 This is the format used by SDB on most System V systems prior to
6716 Produce debugging information in XCOFF format (if that is supported).
6717 This is the format used by the DBX debugger on IBM RS/6000 systems.
6721 Produce debugging information in XCOFF format (if that is supported),
6722 using GNU extensions understood only by the GNU debugger (GDB)@. The
6723 use of these extensions is likely to make other debuggers crash or
6724 refuse to read the program, and may cause assemblers other than the GNU
6725 assembler (GAS) to fail with an error.
6729 Produce debugging information in Alpha/VMS debug format (if that is
6730 supported). This is the format used by DEBUG on Alpha/VMS systems.
6733 @itemx -ggdb@var{level}
6734 @itemx -gstabs@var{level}
6735 @itemx -gcoff@var{level}
6736 @itemx -gxcoff@var{level}
6737 @itemx -gvms@var{level}
6738 Request debugging information and also use @var{level} to specify how
6739 much information. The default level is 2.
6741 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6744 Level 1 produces minimal information, enough for making backtraces in
6745 parts of the program that you don't plan to debug. This includes
6746 descriptions of functions and external variables, and line number
6747 tables, but no information about local variables.
6749 Level 3 includes extra information, such as all the macro definitions
6750 present in the program. Some debuggers support macro expansion when
6751 you use @option{-g3}.
6753 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6754 confusion with @option{-gdwarf-@var{level}}.
6755 Instead use an additional @option{-g@var{level}} option to change the
6756 debug level for DWARF.
6758 @item -feliminate-unused-debug-symbols
6759 @opindex feliminate-unused-debug-symbols
6760 Produce debugging information in stabs format (if that is supported),
6761 for only symbols that are actually used.
6763 @item -femit-class-debug-always
6764 @opindex femit-class-debug-always
6765 Instead of emitting debugging information for a C++ class in only one
6766 object file, emit it in all object files using the class. This option
6767 should be used only with debuggers that are unable to handle the way GCC
6768 normally emits debugging information for classes because using this
6769 option increases the size of debugging information by as much as a
6772 @item -fno-merge-debug-strings
6773 @opindex fmerge-debug-strings
6774 @opindex fno-merge-debug-strings
6775 Direct the linker to not merge together strings in the debugging
6776 information that are identical in different object files. Merging is
6777 not supported by all assemblers or linkers. Merging decreases the size
6778 of the debug information in the output file at the cost of increasing
6779 link processing time. Merging is enabled by default.
6781 @item -fdebug-prefix-map=@var{old}=@var{new}
6782 @opindex fdebug-prefix-map
6783 When compiling files in directory @file{@var{old}}, record debugging
6784 information describing them as in @file{@var{new}} instead.
6786 @item -fvar-tracking
6787 @opindex fvar-tracking
6788 Run variable tracking pass. It computes where variables are stored at each
6789 position in code. Better debugging information is then generated
6790 (if the debugging information format supports this information).
6792 It is enabled by default when compiling with optimization (@option{-Os},
6793 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6794 the debug info format supports it.
6796 @item -fvar-tracking-assignments
6797 @opindex fvar-tracking-assignments
6798 @opindex fno-var-tracking-assignments
6799 Annotate assignments to user variables early in the compilation and
6800 attempt to carry the annotations over throughout the compilation all the
6801 way to the end, in an attempt to improve debug information while
6802 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6804 It can be enabled even if var-tracking is disabled, in which case
6805 annotations are created and maintained, but discarded at the end.
6806 By default, this flag is enabled together with @option{-fvar-tracking},
6807 except when selective scheduling is enabled.
6810 @opindex gsplit-dwarf
6811 Separate as much DWARF debugging information as possible into a
6812 separate output file with the extension @file{.dwo}. This option allows
6813 the build system to avoid linking files with debug information. To
6814 be useful, this option requires a debugger capable of reading @file{.dwo}
6819 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
6821 @item -ggnu-pubnames
6822 @opindex ggnu-pubnames
6823 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
6824 suitable for conversion into a GDB@ index. This option is only useful
6825 with a linker that can produce GDB@ index version 7.
6827 @item -fdebug-types-section
6828 @opindex fdebug-types-section
6829 @opindex fno-debug-types-section
6830 When using DWARF Version 4 or higher, type DIEs can be put into
6831 their own @code{.debug_types} section instead of making them part of the
6832 @code{.debug_info} section. It is more efficient to put them in a separate
6833 comdat sections since the linker can then remove duplicates.
6834 But not all DWARF consumers support @code{.debug_types} sections yet
6835 and on some objects @code{.debug_types} produces larger instead of smaller
6836 debugging information.
6838 @item -grecord-gcc-switches
6839 @item -gno-record-gcc-switches
6840 @opindex grecord-gcc-switches
6841 @opindex gno-record-gcc-switches
6842 This switch causes the command-line options used to invoke the
6843 compiler that may affect code generation to be appended to the
6844 DW_AT_producer attribute in DWARF debugging information. The options
6845 are concatenated with spaces separating them from each other and from
6846 the compiler version.
6847 It is enabled by default.
6848 See also @option{-frecord-gcc-switches} for another
6849 way of storing compiler options into the object file.
6851 @item -gstrict-dwarf
6852 @opindex gstrict-dwarf
6853 Disallow using extensions of later DWARF standard version than selected
6854 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6855 DWARF extensions from later standard versions is allowed.
6857 @item -gno-strict-dwarf
6858 @opindex gno-strict-dwarf
6859 Allow using extensions of later DWARF standard version than selected with
6860 @option{-gdwarf-@var{version}}.
6863 @item -gno-column-info
6864 @opindex gcolumn-info
6865 @opindex gno-column-info
6866 Emit location column information into DWARF debugging information, rather
6867 than just file and line.
6868 This option is disabled by default.
6870 @item -gz@r{[}=@var{type}@r{]}
6872 Produce compressed debug sections in DWARF format, if that is supported.
6873 If @var{type} is not given, the default type depends on the capabilities
6874 of the assembler and linker used. @var{type} may be one of
6875 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6876 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6877 compression in traditional GNU format). If the linker doesn't support
6878 writing compressed debug sections, the option is rejected. Otherwise,
6879 if the assembler does not support them, @option{-gz} is silently ignored
6880 when producing object files.
6882 @item -feliminate-dwarf2-dups
6883 @opindex feliminate-dwarf2-dups
6884 Compress DWARF debugging information by eliminating duplicated
6885 information about each symbol. This option only makes sense when
6886 generating DWARF debugging information.
6888 @item -femit-struct-debug-baseonly
6889 @opindex femit-struct-debug-baseonly
6890 Emit debug information for struct-like types
6891 only when the base name of the compilation source file
6892 matches the base name of file in which the struct is defined.
6894 This option substantially reduces the size of debugging information,
6895 but at significant potential loss in type information to the debugger.
6896 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6897 See @option{-femit-struct-debug-detailed} for more detailed control.
6899 This option works only with DWARF debug output.
6901 @item -femit-struct-debug-reduced
6902 @opindex femit-struct-debug-reduced
6903 Emit debug information for struct-like types
6904 only when the base name of the compilation source file
6905 matches the base name of file in which the type is defined,
6906 unless the struct is a template or defined in a system header.
6908 This option significantly reduces the size of debugging information,
6909 with some potential loss in type information to the debugger.
6910 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6911 See @option{-femit-struct-debug-detailed} for more detailed control.
6913 This option works only with DWARF debug output.
6915 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6916 @opindex femit-struct-debug-detailed
6917 Specify the struct-like types
6918 for which the compiler generates debug information.
6919 The intent is to reduce duplicate struct debug information
6920 between different object files within the same program.
6922 This option is a detailed version of
6923 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6924 which serves for most needs.
6926 A specification has the syntax@*
6927 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6929 The optional first word limits the specification to
6930 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6931 A struct type is used directly when it is the type of a variable, member.
6932 Indirect uses arise through pointers to structs.
6933 That is, when use of an incomplete struct is valid, the use is indirect.
6935 @samp{struct one direct; struct two * indirect;}.
6937 The optional second word limits the specification to
6938 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6939 Generic structs are a bit complicated to explain.
6940 For C++, these are non-explicit specializations of template classes,
6941 or non-template classes within the above.
6942 Other programming languages have generics,
6943 but @option{-femit-struct-debug-detailed} does not yet implement them.
6945 The third word specifies the source files for those
6946 structs for which the compiler should emit debug information.
6947 The values @samp{none} and @samp{any} have the normal meaning.
6948 The value @samp{base} means that
6949 the base of name of the file in which the type declaration appears
6950 must match the base of the name of the main compilation file.
6951 In practice, this means that when compiling @file{foo.c}, debug information
6952 is generated for types declared in that file and @file{foo.h},
6953 but not other header files.
6954 The value @samp{sys} means those types satisfying @samp{base}
6955 or declared in system or compiler headers.
6957 You may need to experiment to determine the best settings for your application.
6959 The default is @option{-femit-struct-debug-detailed=all}.
6961 This option works only with DWARF debug output.
6963 @item -fno-dwarf2-cfi-asm
6964 @opindex fdwarf2-cfi-asm
6965 @opindex fno-dwarf2-cfi-asm
6966 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6967 instead of using GAS @code{.cfi_*} directives.
6969 @item -fno-eliminate-unused-debug-types
6970 @opindex feliminate-unused-debug-types
6971 @opindex fno-eliminate-unused-debug-types
6972 Normally, when producing DWARF output, GCC avoids producing debug symbol
6973 output for types that are nowhere used in the source file being compiled.
6974 Sometimes it is useful to have GCC emit debugging
6975 information for all types declared in a compilation
6976 unit, regardless of whether or not they are actually used
6977 in that compilation unit, for example
6978 if, in the debugger, you want to cast a value to a type that is
6979 not actually used in your program (but is declared). More often,
6980 however, this results in a significant amount of wasted space.
6983 @node Optimize Options
6984 @section Options That Control Optimization
6985 @cindex optimize options
6986 @cindex options, optimization
6988 These options control various sorts of optimizations.
6990 Without any optimization option, the compiler's goal is to reduce the
6991 cost of compilation and to make debugging produce the expected
6992 results. Statements are independent: if you stop the program with a
6993 breakpoint between statements, you can then assign a new value to any
6994 variable or change the program counter to any other statement in the
6995 function and get exactly the results you expect from the source
6998 Turning on optimization flags makes the compiler attempt to improve
6999 the performance and/or code size at the expense of compilation time
7000 and possibly the ability to debug the program.
7002 The compiler performs optimization based on the knowledge it has of the
7003 program. Compiling multiple files at once to a single output file mode allows
7004 the compiler to use information gained from all of the files when compiling
7007 Not all optimizations are controlled directly by a flag. Only
7008 optimizations that have a flag are listed in this section.
7010 Most optimizations are only enabled if an @option{-O} level is set on
7011 the command line. Otherwise they are disabled, even if individual
7012 optimization flags are specified.
7014 Depending on the target and how GCC was configured, a slightly different
7015 set of optimizations may be enabled at each @option{-O} level than
7016 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7017 to find out the exact set of optimizations that are enabled at each level.
7018 @xref{Overall Options}, for examples.
7025 Optimize. Optimizing compilation takes somewhat more time, and a lot
7026 more memory for a large function.
7028 With @option{-O}, the compiler tries to reduce code size and execution
7029 time, without performing any optimizations that take a great deal of
7032 @option{-O} turns on the following optimization flags:
7035 -fbranch-count-reg @gol
7036 -fcombine-stack-adjustments @gol
7038 -fcprop-registers @gol
7041 -fdelayed-branch @gol
7043 -fforward-propagate @gol
7044 -fguess-branch-probability @gol
7045 -fif-conversion2 @gol
7046 -fif-conversion @gol
7047 -finline-functions-called-once @gol
7048 -fipa-pure-const @gol
7050 -fipa-reference @gol
7051 -fmerge-constants @gol
7052 -fmove-loop-invariants @gol
7053 -freorder-blocks @gol
7055 -fshrink-wrap-separate @gol
7056 -fsplit-wide-types @gol
7062 -ftree-coalesce-vars @gol
7063 -ftree-copy-prop @gol
7065 -ftree-dominator-opts @gol
7067 -ftree-forwprop @gol
7077 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7078 where doing so does not interfere with debugging.
7082 Optimize even more. GCC performs nearly all supported optimizations
7083 that do not involve a space-speed tradeoff.
7084 As compared to @option{-O}, this option increases both compilation time
7085 and the performance of the generated code.
7087 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7088 also turns on the following optimization flags:
7089 @gccoptlist{-fthread-jumps @gol
7090 -falign-functions -falign-jumps @gol
7091 -falign-loops -falign-labels @gol
7094 -fcse-follow-jumps -fcse-skip-blocks @gol
7095 -fdelete-null-pointer-checks @gol
7096 -fdevirtualize -fdevirtualize-speculatively @gol
7097 -fexpensive-optimizations @gol
7098 -fgcse -fgcse-lm @gol
7099 -fhoist-adjacent-loads @gol
7100 -finline-small-functions @gol
7101 -findirect-inlining @gol
7107 -fisolate-erroneous-paths-dereference @gol
7109 -foptimize-sibling-calls @gol
7110 -foptimize-strlen @gol
7111 -fpartial-inlining @gol
7113 -freorder-blocks-algorithm=stc @gol
7114 -freorder-blocks-and-partition -freorder-functions @gol
7115 -frerun-cse-after-loop @gol
7116 -fsched-interblock -fsched-spec @gol
7117 -fschedule-insns -fschedule-insns2 @gol
7118 -fstore-merging @gol
7119 -fstrict-aliasing @gol
7120 -ftree-builtin-call-dce @gol
7121 -ftree-switch-conversion -ftree-tail-merge @gol
7122 -fcode-hoisting @gol
7127 Please note the warning under @option{-fgcse} about
7128 invoking @option{-O2} on programs that use computed gotos.
7132 Optimize yet more. @option{-O3} turns on all optimizations specified
7133 by @option{-O2} and also turns on the @option{-finline-functions},
7134 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7135 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7136 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
7137 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7138 @option{-ftree-partial-pre}, @option{-fpeel-loops}
7139 and @option{-fipa-cp-clone} options.
7143 Reduce compilation time and make debugging produce the expected
7144 results. This is the default.
7148 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7149 do not typically increase code size. It also performs further
7150 optimizations designed to reduce code size.
7152 @option{-Os} disables the following optimization flags:
7153 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7154 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7155 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7159 Disregard strict standards compliance. @option{-Ofast} enables all
7160 @option{-O3} optimizations. It also enables optimizations that are not
7161 valid for all standard-compliant programs.
7162 It turns on @option{-ffast-math} and the Fortran-specific
7163 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7167 Optimize debugging experience. @option{-Og} enables optimizations
7168 that do not interfere with debugging. It should be the optimization
7169 level of choice for the standard edit-compile-debug cycle, offering
7170 a reasonable level of optimization while maintaining fast compilation
7171 and a good debugging experience.
7174 If you use multiple @option{-O} options, with or without level numbers,
7175 the last such option is the one that is effective.
7177 Options of the form @option{-f@var{flag}} specify machine-independent
7178 flags. Most flags have both positive and negative forms; the negative
7179 form of @option{-ffoo} is @option{-fno-foo}. In the table
7180 below, only one of the forms is listed---the one you typically
7181 use. You can figure out the other form by either removing @samp{no-}
7184 The following options control specific optimizations. They are either
7185 activated by @option{-O} options or are related to ones that are. You
7186 can use the following flags in the rare cases when ``fine-tuning'' of
7187 optimizations to be performed is desired.
7190 @item -fno-defer-pop
7191 @opindex fno-defer-pop
7192 Always pop the arguments to each function call as soon as that function
7193 returns. For machines that must pop arguments after a function call,
7194 the compiler normally lets arguments accumulate on the stack for several
7195 function calls and pops them all at once.
7197 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7199 @item -fforward-propagate
7200 @opindex fforward-propagate
7201 Perform a forward propagation pass on RTL@. The pass tries to combine two
7202 instructions and checks if the result can be simplified. If loop unrolling
7203 is active, two passes are performed and the second is scheduled after
7206 This option is enabled by default at optimization levels @option{-O},
7207 @option{-O2}, @option{-O3}, @option{-Os}.
7209 @item -ffp-contract=@var{style}
7210 @opindex ffp-contract
7211 @option{-ffp-contract=off} disables floating-point expression contraction.
7212 @option{-ffp-contract=fast} enables floating-point expression contraction
7213 such as forming of fused multiply-add operations if the target has
7214 native support for them.
7215 @option{-ffp-contract=on} enables floating-point expression contraction
7216 if allowed by the language standard. This is currently not implemented
7217 and treated equal to @option{-ffp-contract=off}.
7219 The default is @option{-ffp-contract=fast}.
7221 @item -fomit-frame-pointer
7222 @opindex fomit-frame-pointer
7223 Don't keep the frame pointer in a register for functions that
7224 don't need one. This avoids the instructions to save, set up and
7225 restore frame pointers; it also makes an extra register available
7226 in many functions. @strong{It also makes debugging impossible on
7229 On some machines, such as the VAX, this flag has no effect, because
7230 the standard calling sequence automatically handles the frame pointer
7231 and nothing is saved by pretending it doesn't exist. The
7232 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7233 whether a target machine supports this flag. @xref{Registers,,Register
7234 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7236 The default setting (when not optimizing for
7237 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7238 @option{-fomit-frame-pointer}. You can configure GCC with the
7239 @option{--enable-frame-pointer} configure option to change the default.
7241 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7243 @item -foptimize-sibling-calls
7244 @opindex foptimize-sibling-calls
7245 Optimize sibling and tail recursive calls.
7247 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7249 @item -foptimize-strlen
7250 @opindex foptimize-strlen
7251 Optimize various standard C string functions (e.g. @code{strlen},
7252 @code{strchr} or @code{strcpy}) and
7253 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7255 Enabled at levels @option{-O2}, @option{-O3}.
7259 Do not expand any functions inline apart from those marked with
7260 the @code{always_inline} attribute. This is the default when not
7263 Single functions can be exempted from inlining by marking them
7264 with the @code{noinline} attribute.
7266 @item -finline-small-functions
7267 @opindex finline-small-functions
7268 Integrate functions into their callers when their body is smaller than expected
7269 function call code (so overall size of program gets smaller). The compiler
7270 heuristically decides which functions are simple enough to be worth integrating
7271 in this way. This inlining applies to all functions, even those not declared
7274 Enabled at level @option{-O2}.
7276 @item -findirect-inlining
7277 @opindex findirect-inlining
7278 Inline also indirect calls that are discovered to be known at compile
7279 time thanks to previous inlining. This option has any effect only
7280 when inlining itself is turned on by the @option{-finline-functions}
7281 or @option{-finline-small-functions} options.
7283 Enabled at level @option{-O2}.
7285 @item -finline-functions
7286 @opindex finline-functions
7287 Consider all functions for inlining, even if they are not declared inline.
7288 The compiler heuristically decides which functions are worth integrating
7291 If all calls to a given function are integrated, and the function is
7292 declared @code{static}, then the function is normally not output as
7293 assembler code in its own right.
7295 Enabled at level @option{-O3}.
7297 @item -finline-functions-called-once
7298 @opindex finline-functions-called-once
7299 Consider all @code{static} functions called once for inlining into their
7300 caller even if they are not marked @code{inline}. If a call to a given
7301 function is integrated, then the function is not output as assembler code
7304 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7306 @item -fearly-inlining
7307 @opindex fearly-inlining
7308 Inline functions marked by @code{always_inline} and functions whose body seems
7309 smaller than the function call overhead early before doing
7310 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7311 makes profiling significantly cheaper and usually inlining faster on programs
7312 having large chains of nested wrapper functions.
7318 Perform interprocedural scalar replacement of aggregates, removal of
7319 unused parameters and replacement of parameters passed by reference
7320 by parameters passed by value.
7322 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7324 @item -finline-limit=@var{n}
7325 @opindex finline-limit
7326 By default, GCC limits the size of functions that can be inlined. This flag
7327 allows coarse control of this limit. @var{n} is the size of functions that
7328 can be inlined in number of pseudo instructions.
7330 Inlining is actually controlled by a number of parameters, which may be
7331 specified individually by using @option{--param @var{name}=@var{value}}.
7332 The @option{-finline-limit=@var{n}} option sets some of these parameters
7336 @item max-inline-insns-single
7337 is set to @var{n}/2.
7338 @item max-inline-insns-auto
7339 is set to @var{n}/2.
7342 See below for a documentation of the individual
7343 parameters controlling inlining and for the defaults of these parameters.
7345 @emph{Note:} there may be no value to @option{-finline-limit} that results
7346 in default behavior.
7348 @emph{Note:} pseudo instruction represents, in this particular context, an
7349 abstract measurement of function's size. In no way does it represent a count
7350 of assembly instructions and as such its exact meaning might change from one
7351 release to an another.
7353 @item -fno-keep-inline-dllexport
7354 @opindex fno-keep-inline-dllexport
7355 This is a more fine-grained version of @option{-fkeep-inline-functions},
7356 which applies only to functions that are declared using the @code{dllexport}
7357 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
7360 @item -fkeep-inline-functions
7361 @opindex fkeep-inline-functions
7362 In C, emit @code{static} functions that are declared @code{inline}
7363 into the object file, even if the function has been inlined into all
7364 of its callers. This switch does not affect functions using the
7365 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7366 inline functions into the object file.
7368 @item -fkeep-static-functions
7369 @opindex fkeep-static-functions
7370 Emit @code{static} functions into the object file, even if the function
7373 @item -fkeep-static-consts
7374 @opindex fkeep-static-consts
7375 Emit variables declared @code{static const} when optimization isn't turned
7376 on, even if the variables aren't referenced.
7378 GCC enables this option by default. If you want to force the compiler to
7379 check if a variable is referenced, regardless of whether or not
7380 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7382 @item -fmerge-constants
7383 @opindex fmerge-constants
7384 Attempt to merge identical constants (string constants and floating-point
7385 constants) across compilation units.
7387 This option is the default for optimized compilation if the assembler and
7388 linker support it. Use @option{-fno-merge-constants} to inhibit this
7391 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7393 @item -fmerge-all-constants
7394 @opindex fmerge-all-constants
7395 Attempt to merge identical constants and identical variables.
7397 This option implies @option{-fmerge-constants}. In addition to
7398 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7399 arrays or initialized constant variables with integral or floating-point
7400 types. Languages like C or C++ require each variable, including multiple
7401 instances of the same variable in recursive calls, to have distinct locations,
7402 so using this option results in non-conforming
7405 @item -fmodulo-sched
7406 @opindex fmodulo-sched
7407 Perform swing modulo scheduling immediately before the first scheduling
7408 pass. This pass looks at innermost loops and reorders their
7409 instructions by overlapping different iterations.
7411 @item -fmodulo-sched-allow-regmoves
7412 @opindex fmodulo-sched-allow-regmoves
7413 Perform more aggressive SMS-based modulo scheduling with register moves
7414 allowed. By setting this flag certain anti-dependences edges are
7415 deleted, which triggers the generation of reg-moves based on the
7416 life-range analysis. This option is effective only with
7417 @option{-fmodulo-sched} enabled.
7419 @item -fno-branch-count-reg
7420 @opindex fno-branch-count-reg
7421 Avoid running a pass scanning for opportunities to use ``decrement and
7422 branch'' instructions on a count register instead of generating sequences
7423 of instructions that decrement a register, compare it against zero, and
7424 then branch based upon the result. This option is only meaningful on
7425 architectures that support such instructions, which include x86, PowerPC,
7426 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7427 doesn't remove the decrement and branch instructions from the generated
7428 instruction stream introduced by other optimization passes.
7430 Enabled by default at @option{-O1} and higher.
7432 The default is @option{-fbranch-count-reg}.
7434 @item -fno-function-cse
7435 @opindex fno-function-cse
7436 Do not put function addresses in registers; make each instruction that
7437 calls a constant function contain the function's address explicitly.
7439 This option results in less efficient code, but some strange hacks
7440 that alter the assembler output may be confused by the optimizations
7441 performed when this option is not used.
7443 The default is @option{-ffunction-cse}
7445 @item -fno-zero-initialized-in-bss
7446 @opindex fno-zero-initialized-in-bss
7447 If the target supports a BSS section, GCC by default puts variables that
7448 are initialized to zero into BSS@. This can save space in the resulting
7451 This option turns off this behavior because some programs explicitly
7452 rely on variables going to the data section---e.g., so that the
7453 resulting executable can find the beginning of that section and/or make
7454 assumptions based on that.
7456 The default is @option{-fzero-initialized-in-bss}.
7458 @item -fthread-jumps
7459 @opindex fthread-jumps
7460 Perform optimizations that check to see if a jump branches to a
7461 location where another comparison subsumed by the first is found. If
7462 so, the first branch is redirected to either the destination of the
7463 second branch or a point immediately following it, depending on whether
7464 the condition is known to be true or false.
7466 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7468 @item -fsplit-wide-types
7469 @opindex fsplit-wide-types
7470 When using a type that occupies multiple registers, such as @code{long
7471 long} on a 32-bit system, split the registers apart and allocate them
7472 independently. This normally generates better code for those types,
7473 but may make debugging more difficult.
7475 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7478 @item -fcse-follow-jumps
7479 @opindex fcse-follow-jumps
7480 In common subexpression elimination (CSE), scan through jump instructions
7481 when the target of the jump is not reached by any other path. For
7482 example, when CSE encounters an @code{if} statement with an
7483 @code{else} clause, CSE follows the jump when the condition
7486 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7488 @item -fcse-skip-blocks
7489 @opindex fcse-skip-blocks
7490 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7491 follow jumps that conditionally skip over blocks. When CSE
7492 encounters a simple @code{if} statement with no else clause,
7493 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7494 body of the @code{if}.
7496 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7498 @item -frerun-cse-after-loop
7499 @opindex frerun-cse-after-loop
7500 Re-run common subexpression elimination after loop optimizations are
7503 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7507 Perform a global common subexpression elimination pass.
7508 This pass also performs global constant and copy propagation.
7510 @emph{Note:} When compiling a program using computed gotos, a GCC
7511 extension, you may get better run-time performance if you disable
7512 the global common subexpression elimination pass by adding
7513 @option{-fno-gcse} to the command line.
7515 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7519 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7520 attempts to move loads that are only killed by stores into themselves. This
7521 allows a loop containing a load/store sequence to be changed to a load outside
7522 the loop, and a copy/store within the loop.
7524 Enabled by default when @option{-fgcse} is enabled.
7528 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7529 global common subexpression elimination. This pass attempts to move
7530 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7531 loops containing a load/store sequence can be changed to a load before
7532 the loop and a store after the loop.
7534 Not enabled at any optimization level.
7538 When @option{-fgcse-las} is enabled, the global common subexpression
7539 elimination pass eliminates redundant loads that come after stores to the
7540 same memory location (both partial and full redundancies).
7542 Not enabled at any optimization level.
7544 @item -fgcse-after-reload
7545 @opindex fgcse-after-reload
7546 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7547 pass is performed after reload. The purpose of this pass is to clean up
7550 @item -faggressive-loop-optimizations
7551 @opindex faggressive-loop-optimizations
7552 This option tells the loop optimizer to use language constraints to
7553 derive bounds for the number of iterations of a loop. This assumes that
7554 loop code does not invoke undefined behavior by for example causing signed
7555 integer overflows or out-of-bound array accesses. The bounds for the
7556 number of iterations of a loop are used to guide loop unrolling and peeling
7557 and loop exit test optimizations.
7558 This option is enabled by default.
7560 @item -funconstrained-commons
7561 @opindex funconstrained-commons
7562 This option tells the compiler that variables declared in common blocks
7563 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7564 prevents certain optimizations that depend on knowing the array bounds.
7566 @item -fcrossjumping
7567 @opindex fcrossjumping
7568 Perform cross-jumping transformation.
7569 This transformation unifies equivalent code and saves code size. The
7570 resulting code may or may not perform better than without cross-jumping.
7572 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7574 @item -fauto-inc-dec
7575 @opindex fauto-inc-dec
7576 Combine increments or decrements of addresses with memory accesses.
7577 This pass is always skipped on architectures that do not have
7578 instructions to support this. Enabled by default at @option{-O} and
7579 higher on architectures that support this.
7583 Perform dead code elimination (DCE) on RTL@.
7584 Enabled by default at @option{-O} and higher.
7588 Perform dead store elimination (DSE) on RTL@.
7589 Enabled by default at @option{-O} and higher.
7591 @item -fif-conversion
7592 @opindex fif-conversion
7593 Attempt to transform conditional jumps into branch-less equivalents. This
7594 includes use of conditional moves, min, max, set flags and abs instructions, and
7595 some tricks doable by standard arithmetics. The use of conditional execution
7596 on chips where it is available is controlled by @option{-fif-conversion2}.
7598 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7600 @item -fif-conversion2
7601 @opindex fif-conversion2
7602 Use conditional execution (where available) to transform conditional jumps into
7603 branch-less equivalents.
7605 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7607 @item -fdeclone-ctor-dtor
7608 @opindex fdeclone-ctor-dtor
7609 The C++ ABI requires multiple entry points for constructors and
7610 destructors: one for a base subobject, one for a complete object, and
7611 one for a virtual destructor that calls operator delete afterwards.
7612 For a hierarchy with virtual bases, the base and complete variants are
7613 clones, which means two copies of the function. With this option, the
7614 base and complete variants are changed to be thunks that call a common
7617 Enabled by @option{-Os}.
7619 @item -fdelete-null-pointer-checks
7620 @opindex fdelete-null-pointer-checks
7621 Assume that programs cannot safely dereference null pointers, and that
7622 no code or data element resides at address zero.
7623 This option enables simple constant
7624 folding optimizations at all optimization levels. In addition, other
7625 optimization passes in GCC use this flag to control global dataflow
7626 analyses that eliminate useless checks for null pointers; these assume
7627 that a memory access to address zero always results in a trap, so
7628 that if a pointer is checked after it has already been dereferenced,
7631 Note however that in some environments this assumption is not true.
7632 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7633 for programs that depend on that behavior.
7635 This option is enabled by default on most targets. On Nios II ELF, it
7636 defaults to off. On AVR and CR16, this option is completely disabled.
7638 Passes that use the dataflow information
7639 are enabled independently at different optimization levels.
7641 @item -fdevirtualize
7642 @opindex fdevirtualize
7643 Attempt to convert calls to virtual functions to direct calls. This
7644 is done both within a procedure and interprocedurally as part of
7645 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7646 propagation (@option{-fipa-cp}).
7647 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7649 @item -fdevirtualize-speculatively
7650 @opindex fdevirtualize-speculatively
7651 Attempt to convert calls to virtual functions to speculative direct calls.
7652 Based on the analysis of the type inheritance graph, determine for a given call
7653 the set of likely targets. If the set is small, preferably of size 1, change
7654 the call into a conditional deciding between direct and indirect calls. The
7655 speculative calls enable more optimizations, such as inlining. When they seem
7656 useless after further optimization, they are converted back into original form.
7658 @item -fdevirtualize-at-ltrans
7659 @opindex fdevirtualize-at-ltrans
7660 Stream extra information needed for aggressive devirtualization when running
7661 the link-time optimizer in local transformation mode.
7662 This option enables more devirtualization but
7663 significantly increases the size of streamed data. For this reason it is
7664 disabled by default.
7666 @item -fexpensive-optimizations
7667 @opindex fexpensive-optimizations
7668 Perform a number of minor optimizations that are relatively expensive.
7670 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7674 Attempt to remove redundant extension instructions. This is especially
7675 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7676 registers after writing to their lower 32-bit half.
7678 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7679 @option{-O3}, @option{-Os}.
7681 @item -fno-lifetime-dse
7682 @opindex fno-lifetime-dse
7683 In C++ the value of an object is only affected by changes within its
7684 lifetime: when the constructor begins, the object has an indeterminate
7685 value, and any changes during the lifetime of the object are dead when
7686 the object is destroyed. Normally dead store elimination will take
7687 advantage of this; if your code relies on the value of the object
7688 storage persisting beyond the lifetime of the object, you can use this
7689 flag to disable this optimization. To preserve stores before the
7690 constructor starts (e.g. because your operator new clears the object
7691 storage) but still treat the object as dead after the destructor you,
7692 can use @option{-flifetime-dse=1}. The default behavior can be
7693 explicitly selected with @option{-flifetime-dse=2}.
7694 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7696 @item -flive-range-shrinkage
7697 @opindex flive-range-shrinkage
7698 Attempt to decrease register pressure through register live range
7699 shrinkage. This is helpful for fast processors with small or moderate
7702 @item -fira-algorithm=@var{algorithm}
7703 @opindex fira-algorithm
7704 Use the specified coloring algorithm for the integrated register
7705 allocator. The @var{algorithm} argument can be @samp{priority}, which
7706 specifies Chow's priority coloring, or @samp{CB}, which specifies
7707 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7708 for all architectures, but for those targets that do support it, it is
7709 the default because it generates better code.
7711 @item -fira-region=@var{region}
7712 @opindex fira-region
7713 Use specified regions for the integrated register allocator. The
7714 @var{region} argument should be one of the following:
7719 Use all loops as register allocation regions.
7720 This can give the best results for machines with a small and/or
7721 irregular register set.
7724 Use all loops except for loops with small register pressure
7725 as the regions. This value usually gives
7726 the best results in most cases and for most architectures,
7727 and is enabled by default when compiling with optimization for speed
7728 (@option{-O}, @option{-O2}, @dots{}).
7731 Use all functions as a single region.
7732 This typically results in the smallest code size, and is enabled by default for
7733 @option{-Os} or @option{-O0}.
7737 @item -fira-hoist-pressure
7738 @opindex fira-hoist-pressure
7739 Use IRA to evaluate register pressure in the code hoisting pass for
7740 decisions to hoist expressions. This option usually results in smaller
7741 code, but it can slow the compiler down.
7743 This option is enabled at level @option{-Os} for all targets.
7745 @item -fira-loop-pressure
7746 @opindex fira-loop-pressure
7747 Use IRA to evaluate register pressure in loops for decisions to move
7748 loop invariants. This option usually results in generation
7749 of faster and smaller code on machines with large register files (>= 32
7750 registers), but it can slow the compiler down.
7752 This option is enabled at level @option{-O3} for some targets.
7754 @item -fno-ira-share-save-slots
7755 @opindex fno-ira-share-save-slots
7756 Disable sharing of stack slots used for saving call-used hard
7757 registers living through a call. Each hard register gets a
7758 separate stack slot, and as a result function stack frames are
7761 @item -fno-ira-share-spill-slots
7762 @opindex fno-ira-share-spill-slots
7763 Disable sharing of stack slots allocated for pseudo-registers. Each
7764 pseudo-register that does not get a hard register gets a separate
7765 stack slot, and as a result function stack frames are larger.
7769 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7770 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7771 values if it is profitable.
7773 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7775 @item -fdelayed-branch
7776 @opindex fdelayed-branch
7777 If supported for the target machine, attempt to reorder instructions
7778 to exploit instruction slots available after delayed branch
7781 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7783 @item -fschedule-insns
7784 @opindex fschedule-insns
7785 If supported for the target machine, attempt to reorder instructions to
7786 eliminate execution stalls due to required data being unavailable. This
7787 helps machines that have slow floating point or memory load instructions
7788 by allowing other instructions to be issued until the result of the load
7789 or floating-point instruction is required.
7791 Enabled at levels @option{-O2}, @option{-O3}.
7793 @item -fschedule-insns2
7794 @opindex fschedule-insns2
7795 Similar to @option{-fschedule-insns}, but requests an additional pass of
7796 instruction scheduling after register allocation has been done. This is
7797 especially useful on machines with a relatively small number of
7798 registers and where memory load instructions take more than one cycle.
7800 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7802 @item -fno-sched-interblock
7803 @opindex fno-sched-interblock
7804 Don't schedule instructions across basic blocks. This is normally
7805 enabled by default when scheduling before register allocation, i.e.@:
7806 with @option{-fschedule-insns} or at @option{-O2} or higher.
7808 @item -fno-sched-spec
7809 @opindex fno-sched-spec
7810 Don't allow speculative motion of non-load instructions. This is normally
7811 enabled by default when scheduling before register allocation, i.e.@:
7812 with @option{-fschedule-insns} or at @option{-O2} or higher.
7814 @item -fsched-pressure
7815 @opindex fsched-pressure
7816 Enable register pressure sensitive insn scheduling before register
7817 allocation. This only makes sense when scheduling before register
7818 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7819 @option{-O2} or higher. Usage of this option can improve the
7820 generated code and decrease its size by preventing register pressure
7821 increase above the number of available hard registers and subsequent
7822 spills in register allocation.
7824 @item -fsched-spec-load
7825 @opindex fsched-spec-load
7826 Allow speculative motion of some load instructions. This only makes
7827 sense when scheduling before register allocation, i.e.@: with
7828 @option{-fschedule-insns} or at @option{-O2} or higher.
7830 @item -fsched-spec-load-dangerous
7831 @opindex fsched-spec-load-dangerous
7832 Allow speculative motion of more load instructions. This only makes
7833 sense when scheduling before register allocation, i.e.@: with
7834 @option{-fschedule-insns} or at @option{-O2} or higher.
7836 @item -fsched-stalled-insns
7837 @itemx -fsched-stalled-insns=@var{n}
7838 @opindex fsched-stalled-insns
7839 Define how many insns (if any) can be moved prematurely from the queue
7840 of stalled insns into the ready list during the second scheduling pass.
7841 @option{-fno-sched-stalled-insns} means that no insns are moved
7842 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7843 on how many queued insns can be moved prematurely.
7844 @option{-fsched-stalled-insns} without a value is equivalent to
7845 @option{-fsched-stalled-insns=1}.
7847 @item -fsched-stalled-insns-dep
7848 @itemx -fsched-stalled-insns-dep=@var{n}
7849 @opindex fsched-stalled-insns-dep
7850 Define how many insn groups (cycles) are examined for a dependency
7851 on a stalled insn that is a candidate for premature removal from the queue
7852 of stalled insns. This has an effect only during the second scheduling pass,
7853 and only if @option{-fsched-stalled-insns} is used.
7854 @option{-fno-sched-stalled-insns-dep} is equivalent to
7855 @option{-fsched-stalled-insns-dep=0}.
7856 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7857 @option{-fsched-stalled-insns-dep=1}.
7859 @item -fsched2-use-superblocks
7860 @opindex fsched2-use-superblocks
7861 When scheduling after register allocation, use superblock scheduling.
7862 This allows motion across basic block boundaries,
7863 resulting in faster schedules. This option is experimental, as not all machine
7864 descriptions used by GCC model the CPU closely enough to avoid unreliable
7865 results from the algorithm.
7867 This only makes sense when scheduling after register allocation, i.e.@: with
7868 @option{-fschedule-insns2} or at @option{-O2} or higher.
7870 @item -fsched-group-heuristic
7871 @opindex fsched-group-heuristic
7872 Enable the group heuristic in the scheduler. This heuristic favors
7873 the instruction that belongs to a schedule group. This is enabled
7874 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7875 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7877 @item -fsched-critical-path-heuristic
7878 @opindex fsched-critical-path-heuristic
7879 Enable the critical-path heuristic in the scheduler. This heuristic favors
7880 instructions on the critical path. This is enabled by default when
7881 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7882 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7884 @item -fsched-spec-insn-heuristic
7885 @opindex fsched-spec-insn-heuristic
7886 Enable the speculative instruction heuristic in the scheduler. This
7887 heuristic favors speculative instructions with greater dependency weakness.
7888 This is enabled by default when scheduling is enabled, i.e.@:
7889 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7890 or at @option{-O2} or higher.
7892 @item -fsched-rank-heuristic
7893 @opindex fsched-rank-heuristic
7894 Enable the rank heuristic in the scheduler. This heuristic favors
7895 the instruction belonging to a basic block with greater size or frequency.
7896 This is enabled by default when scheduling is enabled, i.e.@:
7897 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7898 at @option{-O2} or higher.
7900 @item -fsched-last-insn-heuristic
7901 @opindex fsched-last-insn-heuristic
7902 Enable the last-instruction heuristic in the scheduler. This heuristic
7903 favors the instruction that is less dependent on the last instruction
7904 scheduled. This is enabled by default when scheduling is enabled,
7905 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7906 at @option{-O2} or higher.
7908 @item -fsched-dep-count-heuristic
7909 @opindex fsched-dep-count-heuristic
7910 Enable the dependent-count heuristic in the scheduler. This heuristic
7911 favors the instruction that has more instructions depending on it.
7912 This is enabled by default when scheduling is enabled, i.e.@:
7913 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7914 at @option{-O2} or higher.
7916 @item -freschedule-modulo-scheduled-loops
7917 @opindex freschedule-modulo-scheduled-loops
7918 Modulo scheduling is performed before traditional scheduling. If a loop
7919 is modulo scheduled, later scheduling passes may change its schedule.
7920 Use this option to control that behavior.
7922 @item -fselective-scheduling
7923 @opindex fselective-scheduling
7924 Schedule instructions using selective scheduling algorithm. Selective
7925 scheduling runs instead of the first scheduler pass.
7927 @item -fselective-scheduling2
7928 @opindex fselective-scheduling2
7929 Schedule instructions using selective scheduling algorithm. Selective
7930 scheduling runs instead of the second scheduler pass.
7932 @item -fsel-sched-pipelining
7933 @opindex fsel-sched-pipelining
7934 Enable software pipelining of innermost loops during selective scheduling.
7935 This option has no effect unless one of @option{-fselective-scheduling} or
7936 @option{-fselective-scheduling2} is turned on.
7938 @item -fsel-sched-pipelining-outer-loops
7939 @opindex fsel-sched-pipelining-outer-loops
7940 When pipelining loops during selective scheduling, also pipeline outer loops.
7941 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7943 @item -fsemantic-interposition
7944 @opindex fsemantic-interposition
7945 Some object formats, like ELF, allow interposing of symbols by the
7947 This means that for symbols exported from the DSO, the compiler cannot perform
7948 interprocedural propagation, inlining and other optimizations in anticipation
7949 that the function or variable in question may change. While this feature is
7950 useful, for example, to rewrite memory allocation functions by a debugging
7951 implementation, it is expensive in the terms of code quality.
7952 With @option{-fno-semantic-interposition} the compiler assumes that
7953 if interposition happens for functions the overwriting function will have
7954 precisely the same semantics (and side effects).
7955 Similarly if interposition happens
7956 for variables, the constructor of the variable will be the same. The flag
7957 has no effect for functions explicitly declared inline
7958 (where it is never allowed for interposition to change semantics)
7959 and for symbols explicitly declared weak.
7962 @opindex fshrink-wrap
7963 Emit function prologues only before parts of the function that need it,
7964 rather than at the top of the function. This flag is enabled by default at
7965 @option{-O} and higher.
7967 @item -fshrink-wrap-separate
7968 @opindex fshrink-wrap-separate
7969 Shrink-wrap separate parts of the prologue and epilogue separately, so that
7970 those parts are only executed when needed.
7971 This option is on by default, but has no effect unless @option{-fshrink-wrap}
7972 is also turned on and the target supports this.
7974 @item -fcaller-saves
7975 @opindex fcaller-saves
7976 Enable allocation of values to registers that are clobbered by
7977 function calls, by emitting extra instructions to save and restore the
7978 registers around such calls. Such allocation is done only when it
7979 seems to result in better code.
7981 This option is always enabled by default on certain machines, usually
7982 those which have no call-preserved registers to use instead.
7984 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7986 @item -fcombine-stack-adjustments
7987 @opindex fcombine-stack-adjustments
7988 Tracks stack adjustments (pushes and pops) and stack memory references
7989 and then tries to find ways to combine them.
7991 Enabled by default at @option{-O1} and higher.
7995 Use caller save registers for allocation if those registers are not used by
7996 any called function. In that case it is not necessary to save and restore
7997 them around calls. This is only possible if called functions are part of
7998 same compilation unit as current function and they are compiled before it.
8000 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
8001 is disabled if generated code will be instrumented for profiling
8002 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
8003 exactly (this happens on targets that do not expose prologues
8004 and epilogues in RTL).
8006 @item -fconserve-stack
8007 @opindex fconserve-stack
8008 Attempt to minimize stack usage. The compiler attempts to use less
8009 stack space, even if that makes the program slower. This option
8010 implies setting the @option{large-stack-frame} parameter to 100
8011 and the @option{large-stack-frame-growth} parameter to 400.
8013 @item -ftree-reassoc
8014 @opindex ftree-reassoc
8015 Perform reassociation on trees. This flag is enabled by default
8016 at @option{-O} and higher.
8018 @item -fcode-hoisting
8019 @opindex fcode-hoisting
8020 Perform code hoisting. Code hoisting tries to move the
8021 evaluation of expressions executed on all paths to the function exit
8022 as early as possible. This is especially useful as a code size
8023 optimization, but it often helps for code speed as well.
8024 This flag is enabled by default at @option{-O2} and higher.
8028 Perform partial redundancy elimination (PRE) on trees. This flag is
8029 enabled by default at @option{-O2} and @option{-O3}.
8031 @item -ftree-partial-pre
8032 @opindex ftree-partial-pre
8033 Make partial redundancy elimination (PRE) more aggressive. This flag is
8034 enabled by default at @option{-O3}.
8036 @item -ftree-forwprop
8037 @opindex ftree-forwprop
8038 Perform forward propagation on trees. This flag is enabled by default
8039 at @option{-O} and higher.
8043 Perform full redundancy elimination (FRE) on trees. The difference
8044 between FRE and PRE is that FRE only considers expressions
8045 that are computed on all paths leading to the redundant computation.
8046 This analysis is faster than PRE, though it exposes fewer redundancies.
8047 This flag is enabled by default at @option{-O} and higher.
8049 @item -ftree-phiprop
8050 @opindex ftree-phiprop
8051 Perform hoisting of loads from conditional pointers on trees. This
8052 pass is enabled by default at @option{-O} and higher.
8054 @item -fhoist-adjacent-loads
8055 @opindex fhoist-adjacent-loads
8056 Speculatively hoist loads from both branches of an if-then-else if the
8057 loads are from adjacent locations in the same structure and the target
8058 architecture has a conditional move instruction. This flag is enabled
8059 by default at @option{-O2} and higher.
8061 @item -ftree-copy-prop
8062 @opindex ftree-copy-prop
8063 Perform copy propagation on trees. This pass eliminates unnecessary
8064 copy operations. This flag is enabled by default at @option{-O} and
8067 @item -fipa-pure-const
8068 @opindex fipa-pure-const
8069 Discover which functions are pure or constant.
8070 Enabled by default at @option{-O} and higher.
8072 @item -fipa-reference
8073 @opindex fipa-reference
8074 Discover which static variables do not escape the
8076 Enabled by default at @option{-O} and higher.
8080 Perform interprocedural pointer analysis and interprocedural modification
8081 and reference analysis. This option can cause excessive memory and
8082 compile-time usage on large compilation units. It is not enabled by
8083 default at any optimization level.
8086 @opindex fipa-profile
8087 Perform interprocedural profile propagation. The functions called only from
8088 cold functions are marked as cold. Also functions executed once (such as
8089 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8090 functions and loop less parts of functions executed once are then optimized for
8092 Enabled by default at @option{-O} and higher.
8096 Perform interprocedural constant propagation.
8097 This optimization analyzes the program to determine when values passed
8098 to functions are constants and then optimizes accordingly.
8099 This optimization can substantially increase performance
8100 if the application has constants passed to functions.
8101 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8103 @item -fipa-cp-clone
8104 @opindex fipa-cp-clone
8105 Perform function cloning to make interprocedural constant propagation stronger.
8106 When enabled, interprocedural constant propagation performs function cloning
8107 when externally visible function can be called with constant arguments.
8108 Because this optimization can create multiple copies of functions,
8109 it may significantly increase code size
8110 (see @option{--param ipcp-unit-growth=@var{value}}).
8111 This flag is enabled by default at @option{-O3}.
8114 @opindex -fipa-bit-cp
8115 When enabled, perform interprocedural bitwise constant
8116 propagation. This flag is enabled by default at @option{-O2}. It
8117 requires that @option{-fipa-cp} is enabled.
8121 When enabled, perform interprocedural propagation of value
8122 ranges. This flag is enabled by default at @option{-O2}. It requires
8123 that @option{-fipa-cp} is enabled.
8127 Perform Identical Code Folding for functions and read-only variables.
8128 The optimization reduces code size and may disturb unwind stacks by replacing
8129 a function by equivalent one with a different name. The optimization works
8130 more effectively with link-time optimization enabled.
8132 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8133 works on different levels and thus the optimizations are not same - there are
8134 equivalences that are found only by GCC and equivalences found only by Gold.
8136 This flag is enabled by default at @option{-O2} and @option{-Os}.
8138 @item -fisolate-erroneous-paths-dereference
8139 @opindex fisolate-erroneous-paths-dereference
8140 Detect paths that trigger erroneous or undefined behavior due to
8141 dereferencing a null pointer. Isolate those paths from the main control
8142 flow and turn the statement with erroneous or undefined behavior into a trap.
8143 This flag is enabled by default at @option{-O2} and higher and depends on
8144 @option{-fdelete-null-pointer-checks} also being enabled.
8146 @item -fisolate-erroneous-paths-attribute
8147 @opindex fisolate-erroneous-paths-attribute
8148 Detect paths that trigger erroneous or undefined behavior due a null value
8149 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8150 attribute. Isolate those paths from the main control flow and turn the
8151 statement with erroneous or undefined behavior into a trap. This is not
8152 currently enabled, but may be enabled by @option{-O2} in the future.
8156 Perform forward store motion on trees. This flag is
8157 enabled by default at @option{-O} and higher.
8159 @item -ftree-bit-ccp
8160 @opindex ftree-bit-ccp
8161 Perform sparse conditional bit constant propagation on trees and propagate
8162 pointer alignment information.
8163 This pass only operates on local scalar variables and is enabled by default
8164 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8168 Perform sparse conditional constant propagation (CCP) on trees. This
8169 pass only operates on local scalar variables and is enabled by default
8170 at @option{-O} and higher.
8172 @item -fssa-backprop
8173 @opindex fssa-backprop
8174 Propagate information about uses of a value up the definition chain
8175 in order to simplify the definitions. For example, this pass strips
8176 sign operations if the sign of a value never matters. The flag is
8177 enabled by default at @option{-O} and higher.
8180 @opindex fssa-phiopt
8181 Perform pattern matching on SSA PHI nodes to optimize conditional
8182 code. This pass is enabled by default at @option{-O} and higher.
8184 @item -ftree-switch-conversion
8185 @opindex ftree-switch-conversion
8186 Perform conversion of simple initializations in a switch to
8187 initializations from a scalar array. This flag is enabled by default
8188 at @option{-O2} and higher.
8190 @item -ftree-tail-merge
8191 @opindex ftree-tail-merge
8192 Look for identical code sequences. When found, replace one with a jump to the
8193 other. This optimization is known as tail merging or cross jumping. This flag
8194 is enabled by default at @option{-O2} and higher. The compilation time
8196 be limited using @option{max-tail-merge-comparisons} parameter and
8197 @option{max-tail-merge-iterations} parameter.
8201 Perform dead code elimination (DCE) on trees. This flag is enabled by
8202 default at @option{-O} and higher.
8204 @item -ftree-builtin-call-dce
8205 @opindex ftree-builtin-call-dce
8206 Perform conditional dead code elimination (DCE) for calls to built-in functions
8207 that may set @code{errno} but are otherwise side-effect free. This flag is
8208 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8211 @item -ftree-dominator-opts
8212 @opindex ftree-dominator-opts
8213 Perform a variety of simple scalar cleanups (constant/copy
8214 propagation, redundancy elimination, range propagation and expression
8215 simplification) based on a dominator tree traversal. This also
8216 performs jump threading (to reduce jumps to jumps). This flag is
8217 enabled by default at @option{-O} and higher.
8221 Perform dead store elimination (DSE) on trees. A dead store is a store into
8222 a memory location that is later overwritten by another store without
8223 any intervening loads. In this case the earlier store can be deleted. This
8224 flag is enabled by default at @option{-O} and higher.
8228 Perform loop header copying on trees. This is beneficial since it increases
8229 effectiveness of code motion optimizations. It also saves one jump. This flag
8230 is enabled by default at @option{-O} and higher. It is not enabled
8231 for @option{-Os}, since it usually increases code size.
8233 @item -ftree-loop-optimize
8234 @opindex ftree-loop-optimize
8235 Perform loop optimizations on trees. This flag is enabled by default
8236 at @option{-O} and higher.
8238 @item -ftree-loop-linear
8239 @itemx -floop-interchange
8240 @itemx -floop-strip-mine
8242 @itemx -floop-unroll-and-jam
8243 @opindex ftree-loop-linear
8244 @opindex floop-interchange
8245 @opindex floop-strip-mine
8246 @opindex floop-block
8247 @opindex floop-unroll-and-jam
8248 Perform loop nest optimizations. Same as
8249 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8250 to be configured with @option{--with-isl} to enable the Graphite loop
8251 transformation infrastructure.
8253 @item -fgraphite-identity
8254 @opindex fgraphite-identity
8255 Enable the identity transformation for graphite. For every SCoP we generate
8256 the polyhedral representation and transform it back to gimple. Using
8257 @option{-fgraphite-identity} we can check the costs or benefits of the
8258 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8259 are also performed by the code generator isl, like index splitting and
8260 dead code elimination in loops.
8262 @item -floop-nest-optimize
8263 @opindex floop-nest-optimize
8264 Enable the isl based loop nest optimizer. This is a generic loop nest
8265 optimizer based on the Pluto optimization algorithms. It calculates a loop
8266 structure optimized for data-locality and parallelism. This option
8269 @item -floop-parallelize-all
8270 @opindex floop-parallelize-all
8271 Use the Graphite data dependence analysis to identify loops that can
8272 be parallelized. Parallelize all the loops that can be analyzed to
8273 not contain loop carried dependences without checking that it is
8274 profitable to parallelize the loops.
8276 @item -ftree-coalesce-vars
8277 @opindex ftree-coalesce-vars
8278 While transforming the program out of the SSA representation, attempt to
8279 reduce copying by coalescing versions of different user-defined
8280 variables, instead of just compiler temporaries. This may severely
8281 limit the ability to debug an optimized program compiled with
8282 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8283 prevents SSA coalescing of user variables. This option is enabled by
8284 default if optimization is enabled, and it does very little otherwise.
8286 @item -ftree-loop-if-convert
8287 @opindex ftree-loop-if-convert
8288 Attempt to transform conditional jumps in the innermost loops to
8289 branch-less equivalents. The intent is to remove control-flow from
8290 the innermost loops in order to improve the ability of the
8291 vectorization pass to handle these loops. This is enabled by default
8292 if vectorization is enabled.
8294 @item -ftree-loop-distribution
8295 @opindex ftree-loop-distribution
8296 Perform loop distribution. This flag can improve cache performance on
8297 big loop bodies and allow further loop optimizations, like
8298 parallelization or vectorization, to take place. For example, the loop
8315 @item -ftree-loop-distribute-patterns
8316 @opindex ftree-loop-distribute-patterns
8317 Perform loop distribution of patterns that can be code generated with
8318 calls to a library. This flag is enabled by default at @option{-O3}.
8320 This pass distributes the initialization loops and generates a call to
8321 memset zero. For example, the loop
8337 and the initialization loop is transformed into a call to memset zero.
8339 @item -ftree-loop-im
8340 @opindex ftree-loop-im
8341 Perform loop invariant motion on trees. This pass moves only invariants that
8342 are hard to handle at RTL level (function calls, operations that expand to
8343 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8344 operands of conditions that are invariant out of the loop, so that we can use
8345 just trivial invariantness analysis in loop unswitching. The pass also includes
8348 @item -ftree-loop-ivcanon
8349 @opindex ftree-loop-ivcanon
8350 Create a canonical counter for number of iterations in loops for which
8351 determining number of iterations requires complicated analysis. Later
8352 optimizations then may determine the number easily. Useful especially
8353 in connection with unrolling.
8357 Perform induction variable optimizations (strength reduction, induction
8358 variable merging and induction variable elimination) on trees.
8360 @item -ftree-parallelize-loops=n
8361 @opindex ftree-parallelize-loops
8362 Parallelize loops, i.e., split their iteration space to run in n threads.
8363 This is only possible for loops whose iterations are independent
8364 and can be arbitrarily reordered. The optimization is only
8365 profitable on multiprocessor machines, for loops that are CPU-intensive,
8366 rather than constrained e.g.@: by memory bandwidth. This option
8367 implies @option{-pthread}, and thus is only supported on targets
8368 that have support for @option{-pthread}.
8372 Perform function-local points-to analysis on trees. This flag is
8373 enabled by default at @option{-O} and higher.
8377 Perform scalar replacement of aggregates. This pass replaces structure
8378 references with scalars to prevent committing structures to memory too
8379 early. This flag is enabled by default at @option{-O} and higher.
8381 @item -fstore-merging
8382 @opindex fstore-merging
8383 Perform merging of narrow stores to consecutive memory addresses. This pass
8384 merges contiguous stores of immediate values narrower than a word into fewer
8385 wider stores to reduce the number of instructions. This is enabled by default
8386 at @option{-O2} and higher as well as @option{-Os}.
8390 Perform temporary expression replacement during the SSA->normal phase. Single
8391 use/single def temporaries are replaced at their use location with their
8392 defining expression. This results in non-GIMPLE code, but gives the expanders
8393 much more complex trees to work on resulting in better RTL generation. This is
8394 enabled by default at @option{-O} and higher.
8398 Perform straight-line strength reduction on trees. This recognizes related
8399 expressions involving multiplications and replaces them by less expensive
8400 calculations when possible. This is enabled by default at @option{-O} and
8403 @item -ftree-vectorize
8404 @opindex ftree-vectorize
8405 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8406 and @option{-ftree-slp-vectorize} if not explicitly specified.
8408 @item -ftree-loop-vectorize
8409 @opindex ftree-loop-vectorize
8410 Perform loop vectorization on trees. This flag is enabled by default at
8411 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8413 @item -ftree-slp-vectorize
8414 @opindex ftree-slp-vectorize
8415 Perform basic block vectorization on trees. This flag is enabled by default at
8416 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8418 @item -fvect-cost-model=@var{model}
8419 @opindex fvect-cost-model
8420 Alter the cost model used for vectorization. The @var{model} argument
8421 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8422 With the @samp{unlimited} model the vectorized code-path is assumed
8423 to be profitable while with the @samp{dynamic} model a runtime check
8424 guards the vectorized code-path to enable it only for iteration
8425 counts that will likely execute faster than when executing the original
8426 scalar loop. The @samp{cheap} model disables vectorization of
8427 loops where doing so would be cost prohibitive for example due to
8428 required runtime checks for data dependence or alignment but otherwise
8429 is equal to the @samp{dynamic} model.
8430 The default cost model depends on other optimization flags and is
8431 either @samp{dynamic} or @samp{cheap}.
8433 @item -fsimd-cost-model=@var{model}
8434 @opindex fsimd-cost-model
8435 Alter the cost model used for vectorization of loops marked with the OpenMP
8436 or Cilk Plus simd directive. The @var{model} argument should be one of
8437 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8438 have the same meaning as described in @option{-fvect-cost-model} and by
8439 default a cost model defined with @option{-fvect-cost-model} is used.
8443 Perform Value Range Propagation on trees. This is similar to the
8444 constant propagation pass, but instead of values, ranges of values are
8445 propagated. This allows the optimizers to remove unnecessary range
8446 checks like array bound checks and null pointer checks. This is
8447 enabled by default at @option{-O2} and higher. Null pointer check
8448 elimination is only done if @option{-fdelete-null-pointer-checks} is
8452 @opindex fsplit-paths
8453 Split paths leading to loop backedges. This can improve dead code
8454 elimination and common subexpression elimination. This is enabled by
8455 default at @option{-O2} and above.
8457 @item -fsplit-ivs-in-unroller
8458 @opindex fsplit-ivs-in-unroller
8459 Enables expression of values of induction variables in later iterations
8460 of the unrolled loop using the value in the first iteration. This breaks
8461 long dependency chains, thus improving efficiency of the scheduling passes.
8463 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8464 same effect. However, that is not reliable in cases where the loop body
8465 is more complicated than a single basic block. It also does not work at all
8466 on some architectures due to restrictions in the CSE pass.
8468 This optimization is enabled by default.
8470 @item -fvariable-expansion-in-unroller
8471 @opindex fvariable-expansion-in-unroller
8472 With this option, the compiler creates multiple copies of some
8473 local variables when unrolling a loop, which can result in superior code.
8475 @item -fpartial-inlining
8476 @opindex fpartial-inlining
8477 Inline parts of functions. This option has any effect only
8478 when inlining itself is turned on by the @option{-finline-functions}
8479 or @option{-finline-small-functions} options.
8481 Enabled at level @option{-O2}.
8483 @item -fpredictive-commoning
8484 @opindex fpredictive-commoning
8485 Perform predictive commoning optimization, i.e., reusing computations
8486 (especially memory loads and stores) performed in previous
8487 iterations of loops.
8489 This option is enabled at level @option{-O3}.
8491 @item -fprefetch-loop-arrays
8492 @opindex fprefetch-loop-arrays
8493 If supported by the target machine, generate instructions to prefetch
8494 memory to improve the performance of loops that access large arrays.
8496 This option may generate better or worse code; results are highly
8497 dependent on the structure of loops within the source code.
8499 Disabled at level @option{-Os}.
8501 @item -fno-printf-return-value
8502 @opindex fno-printf-return-value
8503 Do not substitute constants for known return value of formatted output
8504 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
8505 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
8506 transformation allows GCC to optimize or even eliminate branches based
8507 on the known return value of these functions called with arguments that
8508 are either constant, or whose values are known to be in a range that
8509 makes determining the exact return value possible. For example, when
8510 @option{-fprintf-return-value} is in effect, both the branch and the
8511 body of the @code{if} statement (but not the call to @code{snprint})
8512 can be optimized away when @code{i} is a 32-bit or smaller integer
8513 because the return value is guaranteed to be at most 8.
8517 if (snprintf (buf, "%08x", i) >= sizeof buf)
8521 The @option{-fprintf-return-value} option relies on other optimizations
8522 and yields best results with @option{-O2}. It works in tandem with the
8523 @option{-Wformat-overflow} and @option{-Wformat-truncation} options.
8524 The @option{-fprintf-return-value} option is enabled by default.
8527 @itemx -fno-peephole2
8528 @opindex fno-peephole
8529 @opindex fno-peephole2
8530 Disable any machine-specific peephole optimizations. The difference
8531 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8532 are implemented in the compiler; some targets use one, some use the
8533 other, a few use both.
8535 @option{-fpeephole} is enabled by default.
8536 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8538 @item -fno-guess-branch-probability
8539 @opindex fno-guess-branch-probability
8540 Do not guess branch probabilities using heuristics.
8542 GCC uses heuristics to guess branch probabilities if they are
8543 not provided by profiling feedback (@option{-fprofile-arcs}). These
8544 heuristics are based on the control flow graph. If some branch probabilities
8545 are specified by @code{__builtin_expect}, then the heuristics are
8546 used to guess branch probabilities for the rest of the control flow graph,
8547 taking the @code{__builtin_expect} info into account. The interactions
8548 between the heuristics and @code{__builtin_expect} can be complex, and in
8549 some cases, it may be useful to disable the heuristics so that the effects
8550 of @code{__builtin_expect} are easier to understand.
8552 The default is @option{-fguess-branch-probability} at levels
8553 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8555 @item -freorder-blocks
8556 @opindex freorder-blocks
8557 Reorder basic blocks in the compiled function in order to reduce number of
8558 taken branches and improve code locality.
8560 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8562 @item -freorder-blocks-algorithm=@var{algorithm}
8563 @opindex freorder-blocks-algorithm
8564 Use the specified algorithm for basic block reordering. The
8565 @var{algorithm} argument can be @samp{simple}, which does not increase
8566 code size (except sometimes due to secondary effects like alignment),
8567 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8568 put all often executed code together, minimizing the number of branches
8569 executed by making extra copies of code.
8571 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8572 @samp{stc} at levels @option{-O2}, @option{-O3}.
8574 @item -freorder-blocks-and-partition
8575 @opindex freorder-blocks-and-partition
8576 In addition to reordering basic blocks in the compiled function, in order
8577 to reduce number of taken branches, partitions hot and cold basic blocks
8578 into separate sections of the assembly and @file{.o} files, to improve
8579 paging and cache locality performance.
8581 This optimization is automatically turned off in the presence of
8582 exception handling, for linkonce sections, for functions with a user-defined
8583 section attribute and on any architecture that does not support named
8586 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8588 @item -freorder-functions
8589 @opindex freorder-functions
8590 Reorder functions in the object file in order to
8591 improve code locality. This is implemented by using special
8592 subsections @code{.text.hot} for most frequently executed functions and
8593 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8594 the linker so object file format must support named sections and linker must
8595 place them in a reasonable way.
8597 Also profile feedback must be available to make this option effective. See
8598 @option{-fprofile-arcs} for details.
8600 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8602 @item -fstrict-aliasing
8603 @opindex fstrict-aliasing
8604 Allow the compiler to assume the strictest aliasing rules applicable to
8605 the language being compiled. For C (and C++), this activates
8606 optimizations based on the type of expressions. In particular, an
8607 object of one type is assumed never to reside at the same address as an
8608 object of a different type, unless the types are almost the same. For
8609 example, an @code{unsigned int} can alias an @code{int}, but not a
8610 @code{void*} or a @code{double}. A character type may alias any other
8613 @anchor{Type-punning}Pay special attention to code like this:
8626 The practice of reading from a different union member than the one most
8627 recently written to (called ``type-punning'') is common. Even with
8628 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8629 is accessed through the union type. So, the code above works as
8630 expected. @xref{Structures unions enumerations and bit-fields
8631 implementation}. However, this code might not:
8642 Similarly, access by taking the address, casting the resulting pointer
8643 and dereferencing the result has undefined behavior, even if the cast
8644 uses a union type, e.g.:
8648 return ((union a_union *) &d)->i;
8652 The @option{-fstrict-aliasing} option is enabled at levels
8653 @option{-O2}, @option{-O3}, @option{-Os}.
8655 @item -falign-functions
8656 @itemx -falign-functions=@var{n}
8657 @opindex falign-functions
8658 Align the start of functions to the next power-of-two greater than
8659 @var{n}, skipping up to @var{n} bytes. For instance,
8660 @option{-falign-functions=32} aligns functions to the next 32-byte
8661 boundary, but @option{-falign-functions=24} aligns to the next
8662 32-byte boundary only if this can be done by skipping 23 bytes or less.
8664 @option{-fno-align-functions} and @option{-falign-functions=1} are
8665 equivalent and mean that functions are not aligned.
8667 Some assemblers only support this flag when @var{n} is a power of two;
8668 in that case, it is rounded up.
8670 If @var{n} is not specified or is zero, use a machine-dependent default.
8672 Enabled at levels @option{-O2}, @option{-O3}.
8674 @item -flimit-function-alignment
8675 If this option is enabled, the compiler tries to avoid unnecessarily
8676 overaligning functions. It attempts to instruct the assembler to align
8677 by the amount specified by @option{-falign-functions}, but not to
8678 skip more bytes than the size of the function.
8680 @item -falign-labels
8681 @itemx -falign-labels=@var{n}
8682 @opindex falign-labels
8683 Align all branch targets to a power-of-two boundary, skipping up to
8684 @var{n} bytes like @option{-falign-functions}. This option can easily
8685 make code slower, because it must insert dummy operations for when the
8686 branch target is reached in the usual flow of the code.
8688 @option{-fno-align-labels} and @option{-falign-labels=1} are
8689 equivalent and mean that labels are not aligned.
8691 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8692 are greater than this value, then their values are used instead.
8694 If @var{n} is not specified or is zero, use a machine-dependent default
8695 which is very likely to be @samp{1}, meaning no alignment.
8697 Enabled at levels @option{-O2}, @option{-O3}.
8700 @itemx -falign-loops=@var{n}
8701 @opindex falign-loops
8702 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8703 like @option{-falign-functions}. If the loops are
8704 executed many times, this makes up for any execution of the dummy
8707 @option{-fno-align-loops} and @option{-falign-loops=1} are
8708 equivalent and mean that loops are not aligned.
8710 If @var{n} is not specified or is zero, use a machine-dependent default.
8712 Enabled at levels @option{-O2}, @option{-O3}.
8715 @itemx -falign-jumps=@var{n}
8716 @opindex falign-jumps
8717 Align branch targets to a power-of-two boundary, for branch targets
8718 where the targets can only be reached by jumping, skipping up to @var{n}
8719 bytes like @option{-falign-functions}. In this case, no dummy operations
8722 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8723 equivalent and mean that loops are not aligned.
8725 If @var{n} is not specified or is zero, use a machine-dependent default.
8727 Enabled at levels @option{-O2}, @option{-O3}.
8729 @item -funit-at-a-time
8730 @opindex funit-at-a-time
8731 This option is left for compatibility reasons. @option{-funit-at-a-time}
8732 has no effect, while @option{-fno-unit-at-a-time} implies
8733 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8737 @item -fno-toplevel-reorder
8738 @opindex fno-toplevel-reorder
8739 Do not reorder top-level functions, variables, and @code{asm}
8740 statements. Output them in the same order that they appear in the
8741 input file. When this option is used, unreferenced static variables
8742 are not removed. This option is intended to support existing code
8743 that relies on a particular ordering. For new code, it is better to
8744 use attributes when possible.
8746 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8747 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8752 Constructs webs as commonly used for register allocation purposes and assign
8753 each web individual pseudo register. This allows the register allocation pass
8754 to operate on pseudos directly, but also strengthens several other optimization
8755 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8756 however, make debugging impossible, since variables no longer stay in a
8759 Enabled by default with @option{-funroll-loops}.
8761 @item -fwhole-program
8762 @opindex fwhole-program
8763 Assume that the current compilation unit represents the whole program being
8764 compiled. All public functions and variables with the exception of @code{main}
8765 and those merged by attribute @code{externally_visible} become static functions
8766 and in effect are optimized more aggressively by interprocedural optimizers.
8768 This option should not be used in combination with @option{-flto}.
8769 Instead relying on a linker plugin should provide safer and more precise
8772 @item -flto[=@var{n}]
8774 This option runs the standard link-time optimizer. When invoked
8775 with source code, it generates GIMPLE (one of GCC's internal
8776 representations) and writes it to special ELF sections in the object
8777 file. When the object files are linked together, all the function
8778 bodies are read from these ELF sections and instantiated as if they
8779 had been part of the same translation unit.
8781 To use the link-time optimizer, @option{-flto} and optimization
8782 options should be specified at compile time and during the final link.
8783 It is recommended that you compile all the files participating in the
8784 same link with the same options and also specify those options at
8789 gcc -c -O2 -flto foo.c
8790 gcc -c -O2 -flto bar.c
8791 gcc -o myprog -flto -O2 foo.o bar.o
8794 The first two invocations to GCC save a bytecode representation
8795 of GIMPLE into special ELF sections inside @file{foo.o} and
8796 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8797 @file{foo.o} and @file{bar.o}, merges the two files into a single
8798 internal image, and compiles the result as usual. Since both
8799 @file{foo.o} and @file{bar.o} are merged into a single image, this
8800 causes all the interprocedural analyses and optimizations in GCC to
8801 work across the two files as if they were a single one. This means,
8802 for example, that the inliner is able to inline functions in
8803 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8805 Another (simpler) way to enable link-time optimization is:
8808 gcc -o myprog -flto -O2 foo.c bar.c
8811 The above generates bytecode for @file{foo.c} and @file{bar.c},
8812 merges them together into a single GIMPLE representation and optimizes
8813 them as usual to produce @file{myprog}.
8815 The only important thing to keep in mind is that to enable link-time
8816 optimizations you need to use the GCC driver to perform the link step.
8817 GCC then automatically performs link-time optimization if any of the
8818 objects involved were compiled with the @option{-flto} command-line option.
8820 should specify the optimization options to be used for link-time
8821 optimization though GCC tries to be clever at guessing an
8822 optimization level to use from the options used at compile time
8823 if you fail to specify one at link time. You can always override
8824 the automatic decision to do link-time optimization
8825 by passing @option{-fno-lto} to the link command.
8827 To make whole program optimization effective, it is necessary to make
8828 certain whole program assumptions. The compiler needs to know
8829 what functions and variables can be accessed by libraries and runtime
8830 outside of the link-time optimized unit. When supported by the linker,
8831 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8832 to the compiler about used and externally visible symbols. When
8833 the linker plugin is not available, @option{-fwhole-program} should be
8834 used to allow the compiler to make these assumptions, which leads
8835 to more aggressive optimization decisions.
8837 When @option{-fuse-linker-plugin} is not enabled, when a file is
8838 compiled with @option{-flto}, the generated object file is larger than
8839 a regular object file because it contains GIMPLE bytecodes and the usual
8840 final code (see @option{-ffat-lto-objects}. This means that
8841 object files with LTO information can be linked as normal object
8842 files; if @option{-fno-lto} is passed to the linker, no
8843 interprocedural optimizations are applied. Note that when
8844 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
8845 but you cannot perform a regular, non-LTO link on them.
8847 Additionally, the optimization flags used to compile individual files
8848 are not necessarily related to those used at link time. For instance,
8851 gcc -c -O0 -ffat-lto-objects -flto foo.c
8852 gcc -c -O0 -ffat-lto-objects -flto bar.c
8853 gcc -o myprog -O3 foo.o bar.o
8856 This produces individual object files with unoptimized assembler
8857 code, but the resulting binary @file{myprog} is optimized at
8858 @option{-O3}. If, instead, the final binary is generated with
8859 @option{-fno-lto}, then @file{myprog} is not optimized.
8861 When producing the final binary, GCC only
8862 applies link-time optimizations to those files that contain bytecode.
8863 Therefore, you can mix and match object files and libraries with
8864 GIMPLE bytecodes and final object code. GCC automatically selects
8865 which files to optimize in LTO mode and which files to link without
8868 There are some code generation flags preserved by GCC when
8869 generating bytecodes, as they need to be used during the final link
8870 stage. Generally options specified at link time override those
8871 specified at compile time.
8873 If you do not specify an optimization level option @option{-O} at
8874 link time, then GCC uses the highest optimization level
8875 used when compiling the object files.
8877 Currently, the following options and their settings are taken from
8878 the first object file that explicitly specifies them:
8879 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8880 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8881 and all the @option{-m} target flags.
8883 Certain ABI-changing flags are required to match in all compilation units,
8884 and trying to override this at link time with a conflicting value
8885 is ignored. This includes options such as @option{-freg-struct-return}
8886 and @option{-fpcc-struct-return}.
8888 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8889 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8890 are passed through to the link stage and merged conservatively for
8891 conflicting translation units. Specifically
8892 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8893 precedence; and for example @option{-ffp-contract=off} takes precedence
8894 over @option{-ffp-contract=fast}. You can override them at link time.
8896 If LTO encounters objects with C linkage declared with incompatible
8897 types in separate translation units to be linked together (undefined
8898 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8899 issued. The behavior is still undefined at run time. Similar
8900 diagnostics may be raised for other languages.
8902 Another feature of LTO is that it is possible to apply interprocedural
8903 optimizations on files written in different languages:
8908 gfortran -c -flto baz.f90
8909 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8912 Notice that the final link is done with @command{g++} to get the C++
8913 runtime libraries and @option{-lgfortran} is added to get the Fortran
8914 runtime libraries. In general, when mixing languages in LTO mode, you
8915 should use the same link command options as when mixing languages in a
8916 regular (non-LTO) compilation.
8918 If object files containing GIMPLE bytecode are stored in a library archive, say
8919 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8920 are using a linker with plugin support. To create static libraries suitable
8921 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8922 and @command{ranlib};
8923 to show the symbols of object files with GIMPLE bytecode, use
8924 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8925 and @command{nm} have been compiled with plugin support. At link time, use the the
8926 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8927 the LTO optimization process:
8930 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8933 With the linker plugin enabled, the linker extracts the needed
8934 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8935 to make them part of the aggregated GIMPLE image to be optimized.
8937 If you are not using a linker with plugin support and/or do not
8938 enable the linker plugin, then the objects inside @file{libfoo.a}
8939 are extracted and linked as usual, but they do not participate
8940 in the LTO optimization process. In order to make a static library suitable
8941 for both LTO optimization and usual linkage, compile its object files with
8942 @option{-flto} @option{-ffat-lto-objects}.
8944 Link-time optimizations do not require the presence of the whole program to
8945 operate. If the program does not require any symbols to be exported, it is
8946 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8947 the interprocedural optimizers to use more aggressive assumptions which may
8948 lead to improved optimization opportunities.
8949 Use of @option{-fwhole-program} is not needed when linker plugin is
8950 active (see @option{-fuse-linker-plugin}).
8952 The current implementation of LTO makes no
8953 attempt to generate bytecode that is portable between different
8954 types of hosts. The bytecode files are versioned and there is a
8955 strict version check, so bytecode files generated in one version of
8956 GCC do not work with an older or newer version of GCC.
8958 Link-time optimization does not work well with generation of debugging
8959 information. Combining @option{-flto} with
8960 @option{-g} is currently experimental and expected to produce unexpected
8963 If you specify the optional @var{n}, the optimization and code
8964 generation done at link time is executed in parallel using @var{n}
8965 parallel jobs by utilizing an installed @command{make} program. The
8966 environment variable @env{MAKE} may be used to override the program
8967 used. The default value for @var{n} is 1.
8969 You can also specify @option{-flto=jobserver} to use GNU make's
8970 job server mode to determine the number of parallel jobs. This
8971 is useful when the Makefile calling GCC is already executing in parallel.
8972 You must prepend a @samp{+} to the command recipe in the parent Makefile
8973 for this to work. This option likely only works if @env{MAKE} is
8976 @item -flto-partition=@var{alg}
8977 @opindex flto-partition
8978 Specify the partitioning algorithm used by the link-time optimizer.
8979 The value is either @samp{1to1} to specify a partitioning mirroring
8980 the original source files or @samp{balanced} to specify partitioning
8981 into equally sized chunks (whenever possible) or @samp{max} to create
8982 new partition for every symbol where possible. Specifying @samp{none}
8983 as an algorithm disables partitioning and streaming completely.
8984 The default value is @samp{balanced}. While @samp{1to1} can be used
8985 as an workaround for various code ordering issues, the @samp{max}
8986 partitioning is intended for internal testing only.
8987 The value @samp{one} specifies that exactly one partition should be
8988 used while the value @samp{none} bypasses partitioning and executes
8989 the link-time optimization step directly from the WPA phase.
8991 @item -flto-odr-type-merging
8992 @opindex flto-odr-type-merging
8993 Enable streaming of mangled types names of C++ types and their unification
8994 at link time. This increases size of LTO object files, but enables
8995 diagnostics about One Definition Rule violations.
8997 @item -flto-compression-level=@var{n}
8998 @opindex flto-compression-level
8999 This option specifies the level of compression used for intermediate
9000 language written to LTO object files, and is only meaningful in
9001 conjunction with LTO mode (@option{-flto}). Valid
9002 values are 0 (no compression) to 9 (maximum compression). Values
9003 outside this range are clamped to either 0 or 9. If the option is not
9004 given, a default balanced compression setting is used.
9006 @item -fuse-linker-plugin
9007 @opindex fuse-linker-plugin
9008 Enables the use of a linker plugin during link-time optimization. This
9009 option relies on plugin support in the linker, which is available in gold
9010 or in GNU ld 2.21 or newer.
9012 This option enables the extraction of object files with GIMPLE bytecode out
9013 of library archives. This improves the quality of optimization by exposing
9014 more code to the link-time optimizer. This information specifies what
9015 symbols can be accessed externally (by non-LTO object or during dynamic
9016 linking). Resulting code quality improvements on binaries (and shared
9017 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9018 See @option{-flto} for a description of the effect of this flag and how to
9021 This option is enabled by default when LTO support in GCC is enabled
9022 and GCC was configured for use with
9023 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9025 @item -ffat-lto-objects
9026 @opindex ffat-lto-objects
9027 Fat LTO objects are object files that contain both the intermediate language
9028 and the object code. This makes them usable for both LTO linking and normal
9029 linking. This option is effective only when compiling with @option{-flto}
9030 and is ignored at link time.
9032 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9033 requires the complete toolchain to be aware of LTO. It requires a linker with
9034 linker plugin support for basic functionality. Additionally,
9035 @command{nm}, @command{ar} and @command{ranlib}
9036 need to support linker plugins to allow a full-featured build environment
9037 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9038 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9039 to these tools. With non fat LTO makefiles need to be modified to use them.
9041 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9044 @item -fcompare-elim
9045 @opindex fcompare-elim
9046 After register allocation and post-register allocation instruction splitting,
9047 identify arithmetic instructions that compute processor flags similar to a
9048 comparison operation based on that arithmetic. If possible, eliminate the
9049 explicit comparison operation.
9051 This pass only applies to certain targets that cannot explicitly represent
9052 the comparison operation before register allocation is complete.
9054 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9056 @item -fcprop-registers
9057 @opindex fcprop-registers
9058 After register allocation and post-register allocation instruction splitting,
9059 perform a copy-propagation pass to try to reduce scheduling dependencies
9060 and occasionally eliminate the copy.
9062 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9064 @item -fprofile-correction
9065 @opindex fprofile-correction
9066 Profiles collected using an instrumented binary for multi-threaded programs may
9067 be inconsistent due to missed counter updates. When this option is specified,
9068 GCC uses heuristics to correct or smooth out such inconsistencies. By
9069 default, GCC emits an error message when an inconsistent profile is detected.
9072 @itemx -fprofile-use=@var{path}
9073 @opindex fprofile-use
9074 Enable profile feedback-directed optimizations,
9075 and the following optimizations
9076 which are generally profitable only with profile feedback available:
9077 @option{-fbranch-probabilities}, @option{-fvpt},
9078 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9079 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9081 Before you can use this option, you must first generate profiling information.
9082 @xref{Instrumentation Options}, for information about the
9083 @option{-fprofile-generate} option.
9085 By default, GCC emits an error message if the feedback profiles do not
9086 match the source code. This error can be turned into a warning by using
9087 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9090 If @var{path} is specified, GCC looks at the @var{path} to find
9091 the profile feedback data files. See @option{-fprofile-dir}.
9093 @item -fauto-profile
9094 @itemx -fauto-profile=@var{path}
9095 @opindex fauto-profile
9096 Enable sampling-based feedback-directed optimizations,
9097 and the following optimizations
9098 which are generally profitable only with profile feedback available:
9099 @option{-fbranch-probabilities}, @option{-fvpt},
9100 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9101 @option{-ftree-vectorize},
9102 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9103 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9104 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9106 @var{path} is the name of a file containing AutoFDO profile information.
9107 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9109 Producing an AutoFDO profile data file requires running your program
9110 with the @command{perf} utility on a supported GNU/Linux target system.
9111 For more information, see @uref{https://perf.wiki.kernel.org/}.
9115 perf record -e br_inst_retired:near_taken -b -o perf.data \
9119 Then use the @command{create_gcov} tool to convert the raw profile data
9120 to a format that can be used by GCC.@ You must also supply the
9121 unstripped binary for your program to this tool.
9122 See @uref{https://github.com/google/autofdo}.
9126 create_gcov --binary=your_program.unstripped --profile=perf.data \
9131 The following options control compiler behavior regarding floating-point
9132 arithmetic. These options trade off between speed and
9133 correctness. All must be specifically enabled.
9137 @opindex ffloat-store
9138 Do not store floating-point variables in registers, and inhibit other
9139 options that might change whether a floating-point value is taken from a
9142 @cindex floating-point precision
9143 This option prevents undesirable excess precision on machines such as
9144 the 68000 where the floating registers (of the 68881) keep more
9145 precision than a @code{double} is supposed to have. Similarly for the
9146 x86 architecture. For most programs, the excess precision does only
9147 good, but a few programs rely on the precise definition of IEEE floating
9148 point. Use @option{-ffloat-store} for such programs, after modifying
9149 them to store all pertinent intermediate computations into variables.
9151 @item -fexcess-precision=@var{style}
9152 @opindex fexcess-precision
9153 This option allows further control over excess precision on machines
9154 where floating-point operations occur in a format with more precision or
9155 range than the IEEE standard and interchange floating-point types. By
9156 default, @option{-fexcess-precision=fast} is in effect; this means that
9157 operations may be carried out in a wider precision than the types specified
9158 in the source if that would result in faster code, and it is unpredictable
9159 when rounding to the types specified in the source code takes place.
9160 When compiling C, if @option{-fexcess-precision=standard} is specified then
9161 excess precision follows the rules specified in ISO C99; in particular,
9162 both casts and assignments cause values to be rounded to their
9163 semantic types (whereas @option{-ffloat-store} only affects
9164 assignments). This option is enabled by default for C if a strict
9165 conformance option such as @option{-std=c99} is used.
9166 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
9167 regardless of whether a strict conformance option is used.
9170 @option{-fexcess-precision=standard} is not implemented for languages
9171 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
9172 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9173 semantics apply without excess precision, and in the latter, rounding
9178 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9179 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9180 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
9181 @option{-fexcess-precision=fast}.
9183 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9185 This option is not turned on by any @option{-O} option besides
9186 @option{-Ofast} since it can result in incorrect output for programs
9187 that depend on an exact implementation of IEEE or ISO rules/specifications
9188 for math functions. It may, however, yield faster code for programs
9189 that do not require the guarantees of these specifications.
9191 @item -fno-math-errno
9192 @opindex fno-math-errno
9193 Do not set @code{errno} after calling math functions that are executed
9194 with a single instruction, e.g., @code{sqrt}. A program that relies on
9195 IEEE exceptions for math error handling may want to use this flag
9196 for speed while maintaining IEEE arithmetic compatibility.
9198 This option is not turned on by any @option{-O} option since
9199 it can result in incorrect output for programs that depend on
9200 an exact implementation of IEEE or ISO rules/specifications for
9201 math functions. It may, however, yield faster code for programs
9202 that do not require the guarantees of these specifications.
9204 The default is @option{-fmath-errno}.
9206 On Darwin systems, the math library never sets @code{errno}. There is
9207 therefore no reason for the compiler to consider the possibility that
9208 it might, and @option{-fno-math-errno} is the default.
9210 @item -funsafe-math-optimizations
9211 @opindex funsafe-math-optimizations
9213 Allow optimizations for floating-point arithmetic that (a) assume
9214 that arguments and results are valid and (b) may violate IEEE or
9215 ANSI standards. When used at link time, it may include libraries
9216 or startup files that change the default FPU control word or other
9217 similar optimizations.
9219 This option is not turned on by any @option{-O} option since
9220 it can result in incorrect output for programs that depend on
9221 an exact implementation of IEEE or ISO rules/specifications for
9222 math functions. It may, however, yield faster code for programs
9223 that do not require the guarantees of these specifications.
9224 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9225 @option{-fassociative-math} and @option{-freciprocal-math}.
9227 The default is @option{-fno-unsafe-math-optimizations}.
9229 @item -fassociative-math
9230 @opindex fassociative-math
9232 Allow re-association of operands in series of floating-point operations.
9233 This violates the ISO C and C++ language standard by possibly changing
9234 computation result. NOTE: re-ordering may change the sign of zero as
9235 well as ignore NaNs and inhibit or create underflow or overflow (and
9236 thus cannot be used on code that relies on rounding behavior like
9237 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9238 and thus may not be used when ordered comparisons are required.
9239 This option requires that both @option{-fno-signed-zeros} and
9240 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9241 much sense with @option{-frounding-math}. For Fortran the option
9242 is automatically enabled when both @option{-fno-signed-zeros} and
9243 @option{-fno-trapping-math} are in effect.
9245 The default is @option{-fno-associative-math}.
9247 @item -freciprocal-math
9248 @opindex freciprocal-math
9250 Allow the reciprocal of a value to be used instead of dividing by
9251 the value if this enables optimizations. For example @code{x / y}
9252 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9253 is subject to common subexpression elimination. Note that this loses
9254 precision and increases the number of flops operating on the value.
9256 The default is @option{-fno-reciprocal-math}.
9258 @item -ffinite-math-only
9259 @opindex ffinite-math-only
9260 Allow optimizations for floating-point arithmetic that assume
9261 that arguments and results are not NaNs or +-Infs.
9263 This option is not turned on by any @option{-O} option since
9264 it can result in incorrect output for programs that depend on
9265 an exact implementation of IEEE or ISO rules/specifications for
9266 math functions. It may, however, yield faster code for programs
9267 that do not require the guarantees of these specifications.
9269 The default is @option{-fno-finite-math-only}.
9271 @item -fno-signed-zeros
9272 @opindex fno-signed-zeros
9273 Allow optimizations for floating-point arithmetic that ignore the
9274 signedness of zero. IEEE arithmetic specifies the behavior of
9275 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9276 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9277 This option implies that the sign of a zero result isn't significant.
9279 The default is @option{-fsigned-zeros}.
9281 @item -fno-trapping-math
9282 @opindex fno-trapping-math
9283 Compile code assuming that floating-point operations cannot generate
9284 user-visible traps. These traps include division by zero, overflow,
9285 underflow, inexact result and invalid operation. This option requires
9286 that @option{-fno-signaling-nans} be in effect. Setting this option may
9287 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9289 This option should never be turned on by any @option{-O} option since
9290 it can result in incorrect output for programs that depend on
9291 an exact implementation of IEEE or ISO rules/specifications for
9294 The default is @option{-ftrapping-math}.
9296 @item -frounding-math
9297 @opindex frounding-math
9298 Disable transformations and optimizations that assume default floating-point
9299 rounding behavior. This is round-to-zero for all floating point
9300 to integer conversions, and round-to-nearest for all other arithmetic
9301 truncations. This option should be specified for programs that change
9302 the FP rounding mode dynamically, or that may be executed with a
9303 non-default rounding mode. This option disables constant folding of
9304 floating-point expressions at compile time (which may be affected by
9305 rounding mode) and arithmetic transformations that are unsafe in the
9306 presence of sign-dependent rounding modes.
9308 The default is @option{-fno-rounding-math}.
9310 This option is experimental and does not currently guarantee to
9311 disable all GCC optimizations that are affected by rounding mode.
9312 Future versions of GCC may provide finer control of this setting
9313 using C99's @code{FENV_ACCESS} pragma. This command-line option
9314 will be used to specify the default state for @code{FENV_ACCESS}.
9316 @item -fsignaling-nans
9317 @opindex fsignaling-nans
9318 Compile code assuming that IEEE signaling NaNs may generate user-visible
9319 traps during floating-point operations. Setting this option disables
9320 optimizations that may change the number of exceptions visible with
9321 signaling NaNs. This option implies @option{-ftrapping-math}.
9323 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9326 The default is @option{-fno-signaling-nans}.
9328 This option is experimental and does not currently guarantee to
9329 disable all GCC optimizations that affect signaling NaN behavior.
9331 @item -fno-fp-int-builtin-inexact
9332 @opindex fno-fp-int-builtin-inexact
9333 Do not allow the built-in functions @code{ceil}, @code{floor},
9334 @code{round} and @code{trunc}, and their @code{float} and @code{long
9335 double} variants, to generate code that raises the ``inexact''
9336 floating-point exception for noninteger arguments. ISO C99 and C11
9337 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9338 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9341 The default is @option{-ffp-int-builtin-inexact}, allowing the
9342 exception to be raised. This option does nothing unless
9343 @option{-ftrapping-math} is in effect.
9345 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9346 generate a call to a library function then the ``inexact'' exception
9347 may be raised if the library implementation does not follow TS 18661.
9349 @item -fsingle-precision-constant
9350 @opindex fsingle-precision-constant
9351 Treat floating-point constants as single precision instead of
9352 implicitly converting them to double-precision constants.
9354 @item -fcx-limited-range
9355 @opindex fcx-limited-range
9356 When enabled, this option states that a range reduction step is not
9357 needed when performing complex division. Also, there is no checking
9358 whether the result of a complex multiplication or division is @code{NaN
9359 + I*NaN}, with an attempt to rescue the situation in that case. The
9360 default is @option{-fno-cx-limited-range}, but is enabled by
9361 @option{-ffast-math}.
9363 This option controls the default setting of the ISO C99
9364 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9367 @item -fcx-fortran-rules
9368 @opindex fcx-fortran-rules
9369 Complex multiplication and division follow Fortran rules. Range
9370 reduction is done as part of complex division, but there is no checking
9371 whether the result of a complex multiplication or division is @code{NaN
9372 + I*NaN}, with an attempt to rescue the situation in that case.
9374 The default is @option{-fno-cx-fortran-rules}.
9378 The following options control optimizations that may improve
9379 performance, but are not enabled by any @option{-O} options. This
9380 section includes experimental options that may produce broken code.
9383 @item -fbranch-probabilities
9384 @opindex fbranch-probabilities
9385 After running a program compiled with @option{-fprofile-arcs}
9386 (@pxref{Instrumentation Options}),
9387 you can compile it a second time using
9388 @option{-fbranch-probabilities}, to improve optimizations based on
9389 the number of times each branch was taken. When a program
9390 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9391 counts to a file called @file{@var{sourcename}.gcda} for each source
9392 file. The information in this data file is very dependent on the
9393 structure of the generated code, so you must use the same source code
9394 and the same optimization options for both compilations.
9396 With @option{-fbranch-probabilities}, GCC puts a
9397 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9398 These can be used to improve optimization. Currently, they are only
9399 used in one place: in @file{reorg.c}, instead of guessing which path a
9400 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9401 exactly determine which path is taken more often.
9403 @item -fprofile-values
9404 @opindex fprofile-values
9405 If combined with @option{-fprofile-arcs}, it adds code so that some
9406 data about values of expressions in the program is gathered.
9408 With @option{-fbranch-probabilities}, it reads back the data gathered
9409 from profiling values of expressions for usage in optimizations.
9411 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9413 @item -fprofile-reorder-functions
9414 @opindex fprofile-reorder-functions
9415 Function reordering based on profile instrumentation collects
9416 first time of execution of a function and orders these functions
9419 Enabled with @option{-fprofile-use}.
9423 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9424 to add code to gather information about values of expressions.
9426 With @option{-fbranch-probabilities}, it reads back the data gathered
9427 and actually performs the optimizations based on them.
9428 Currently the optimizations include specialization of division operations
9429 using the knowledge about the value of the denominator.
9431 @item -frename-registers
9432 @opindex frename-registers
9433 Attempt to avoid false dependencies in scheduled code by making use
9434 of registers left over after register allocation. This optimization
9435 most benefits processors with lots of registers. Depending on the
9436 debug information format adopted by the target, however, it can
9437 make debugging impossible, since variables no longer stay in
9438 a ``home register''.
9440 Enabled by default with @option{-funroll-loops}.
9442 @item -fschedule-fusion
9443 @opindex fschedule-fusion
9444 Performs a target dependent pass over the instruction stream to schedule
9445 instructions of same type together because target machine can execute them
9446 more efficiently if they are adjacent to each other in the instruction flow.
9448 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9452 Perform tail duplication to enlarge superblock size. This transformation
9453 simplifies the control flow of the function allowing other optimizations to do
9456 Enabled with @option{-fprofile-use}.
9458 @item -funroll-loops
9459 @opindex funroll-loops
9460 Unroll loops whose number of iterations can be determined at compile time or
9461 upon entry to the loop. @option{-funroll-loops} implies
9462 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9463 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9464 a small constant number of iterations). This option makes code larger, and may
9465 or may not make it run faster.
9467 Enabled with @option{-fprofile-use}.
9469 @item -funroll-all-loops
9470 @opindex funroll-all-loops
9471 Unroll all loops, even if their number of iterations is uncertain when
9472 the loop is entered. This usually makes programs run more slowly.
9473 @option{-funroll-all-loops} implies the same options as
9474 @option{-funroll-loops}.
9477 @opindex fpeel-loops
9478 Peels loops for which there is enough information that they do not
9479 roll much (from profile feedback or static analysis). It also turns on
9480 complete loop peeling (i.e.@: complete removal of loops with small constant
9481 number of iterations).
9483 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9485 @item -fmove-loop-invariants
9486 @opindex fmove-loop-invariants
9487 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9488 at level @option{-O1}
9491 @opindex fsplit-loops
9492 Split a loop into two if it contains a condition that's always true
9493 for one side of the iteration space and false for the other.
9495 @item -funswitch-loops
9496 @opindex funswitch-loops
9497 Move branches with loop invariant conditions out of the loop, with duplicates
9498 of the loop on both branches (modified according to result of the condition).
9500 @item -ffunction-sections
9501 @itemx -fdata-sections
9502 @opindex ffunction-sections
9503 @opindex fdata-sections
9504 Place each function or data item into its own section in the output
9505 file if the target supports arbitrary sections. The name of the
9506 function or the name of the data item determines the section's name
9509 Use these options on systems where the linker can perform optimizations
9510 to improve locality of reference in the instruction space. Most systems
9511 using the ELF object format and SPARC processors running Solaris 2 have
9512 linkers with such optimizations. AIX may have these optimizations in
9515 Only use these options when there are significant benefits from doing
9516 so. When you specify these options, the assembler and linker
9517 create larger object and executable files and are also slower.
9518 You cannot use @command{gprof} on all systems if you
9519 specify this option, and you may have problems with debugging if
9520 you specify both this option and @option{-g}.
9522 @item -fbranch-target-load-optimize
9523 @opindex fbranch-target-load-optimize
9524 Perform branch target register load optimization before prologue / epilogue
9526 The use of target registers can typically be exposed only during reload,
9527 thus hoisting loads out of loops and doing inter-block scheduling needs
9528 a separate optimization pass.
9530 @item -fbranch-target-load-optimize2
9531 @opindex fbranch-target-load-optimize2
9532 Perform branch target register load optimization after prologue / epilogue
9535 @item -fbtr-bb-exclusive
9536 @opindex fbtr-bb-exclusive
9537 When performing branch target register load optimization, don't reuse
9538 branch target registers within any basic block.
9541 @opindex fstdarg-opt
9542 Optimize the prologue of variadic argument functions with respect to usage of
9545 @item -fsection-anchors
9546 @opindex fsection-anchors
9547 Try to reduce the number of symbolic address calculations by using
9548 shared ``anchor'' symbols to address nearby objects. This transformation
9549 can help to reduce the number of GOT entries and GOT accesses on some
9552 For example, the implementation of the following function @code{foo}:
9556 int foo (void) @{ return a + b + c; @}
9560 usually calculates the addresses of all three variables, but if you
9561 compile it with @option{-fsection-anchors}, it accesses the variables
9562 from a common anchor point instead. The effect is similar to the
9563 following pseudocode (which isn't valid C):
9568 register int *xr = &x;
9569 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9573 Not all targets support this option.
9575 @item --param @var{name}=@var{value}
9577 In some places, GCC uses various constants to control the amount of
9578 optimization that is done. For example, GCC does not inline functions
9579 that contain more than a certain number of instructions. You can
9580 control some of these constants on the command line using the
9581 @option{--param} option.
9583 The names of specific parameters, and the meaning of the values, are
9584 tied to the internals of the compiler, and are subject to change
9585 without notice in future releases.
9587 In each case, the @var{value} is an integer. The allowable choices for
9591 @item predictable-branch-outcome
9592 When branch is predicted to be taken with probability lower than this threshold
9593 (in percent), then it is considered well predictable. The default is 10.
9595 @item max-rtl-if-conversion-insns
9596 RTL if-conversion tries to remove conditional branches around a block and
9597 replace them with conditionally executed instructions. This parameter
9598 gives the maximum number of instructions in a block which should be
9599 considered for if-conversion. The default is 10, though the compiler will
9600 also use other heuristics to decide whether if-conversion is likely to be
9603 @item max-rtl-if-conversion-predictable-cost
9604 @item max-rtl-if-conversion-unpredictable-cost
9605 RTL if-conversion will try to remove conditional branches around a block
9606 and replace them with conditionally executed instructions. These parameters
9607 give the maximum permissible cost for the sequence that would be generated
9608 by if-conversion depending on whether the branch is statically determined
9609 to be predictable or not. The units for this parameter are the same as
9610 those for the GCC internal seq_cost metric. The compiler will try to
9611 provide a reasonable default for this parameter using the BRANCH_COST
9614 @item max-crossjump-edges
9615 The maximum number of incoming edges to consider for cross-jumping.
9616 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9617 the number of edges incoming to each block. Increasing values mean
9618 more aggressive optimization, making the compilation time increase with
9619 probably small improvement in executable size.
9621 @item min-crossjump-insns
9622 The minimum number of instructions that must be matched at the end
9623 of two blocks before cross-jumping is performed on them. This
9624 value is ignored in the case where all instructions in the block being
9625 cross-jumped from are matched. The default value is 5.
9627 @item max-grow-copy-bb-insns
9628 The maximum code size expansion factor when copying basic blocks
9629 instead of jumping. The expansion is relative to a jump instruction.
9630 The default value is 8.
9632 @item max-goto-duplication-insns
9633 The maximum number of instructions to duplicate to a block that jumps
9634 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9635 passes, GCC factors computed gotos early in the compilation process,
9636 and unfactors them as late as possible. Only computed jumps at the
9637 end of a basic blocks with no more than max-goto-duplication-insns are
9638 unfactored. The default value is 8.
9640 @item max-delay-slot-insn-search
9641 The maximum number of instructions to consider when looking for an
9642 instruction to fill a delay slot. If more than this arbitrary number of
9643 instructions are searched, the time savings from filling the delay slot
9644 are minimal, so stop searching. Increasing values mean more
9645 aggressive optimization, making the compilation time increase with probably
9646 small improvement in execution time.
9648 @item max-delay-slot-live-search
9649 When trying to fill delay slots, the maximum number of instructions to
9650 consider when searching for a block with valid live register
9651 information. Increasing this arbitrarily chosen value means more
9652 aggressive optimization, increasing the compilation time. This parameter
9653 should be removed when the delay slot code is rewritten to maintain the
9656 @item max-gcse-memory
9657 The approximate maximum amount of memory that can be allocated in
9658 order to perform the global common subexpression elimination
9659 optimization. If more memory than specified is required, the
9660 optimization is not done.
9662 @item max-gcse-insertion-ratio
9663 If the ratio of expression insertions to deletions is larger than this value
9664 for any expression, then RTL PRE inserts or removes the expression and thus
9665 leaves partially redundant computations in the instruction stream. The default value is 20.
9667 @item max-pending-list-length
9668 The maximum number of pending dependencies scheduling allows
9669 before flushing the current state and starting over. Large functions
9670 with few branches or calls can create excessively large lists which
9671 needlessly consume memory and resources.
9673 @item max-modulo-backtrack-attempts
9674 The maximum number of backtrack attempts the scheduler should make
9675 when modulo scheduling a loop. Larger values can exponentially increase
9678 @item max-inline-insns-single
9679 Several parameters control the tree inliner used in GCC@.
9680 This number sets the maximum number of instructions (counted in GCC's
9681 internal representation) in a single function that the tree inliner
9682 considers for inlining. This only affects functions declared
9683 inline and methods implemented in a class declaration (C++).
9684 The default value is 400.
9686 @item max-inline-insns-auto
9687 When you use @option{-finline-functions} (included in @option{-O3}),
9688 a lot of functions that would otherwise not be considered for inlining
9689 by the compiler are investigated. To those functions, a different
9690 (more restrictive) limit compared to functions declared inline can
9692 The default value is 40.
9694 @item inline-min-speedup
9695 When estimated performance improvement of caller + callee runtime exceeds this
9696 threshold (in percent), the function can be inlined regardless of the limit on
9697 @option{--param max-inline-insns-single} and @option{--param
9698 max-inline-insns-auto}.
9700 @item large-function-insns
9701 The limit specifying really large functions. For functions larger than this
9702 limit after inlining, inlining is constrained by
9703 @option{--param large-function-growth}. This parameter is useful primarily
9704 to avoid extreme compilation time caused by non-linear algorithms used by the
9706 The default value is 2700.
9708 @item large-function-growth
9709 Specifies maximal growth of large function caused by inlining in percents.
9710 The default value is 100 which limits large function growth to 2.0 times
9713 @item large-unit-insns
9714 The limit specifying large translation unit. Growth caused by inlining of
9715 units larger than this limit is limited by @option{--param inline-unit-growth}.
9716 For small units this might be too tight.
9717 For example, consider a unit consisting of function A
9718 that is inline and B that just calls A three times. If B is small relative to
9719 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9720 large units consisting of small inlineable functions, however, the overall unit
9721 growth limit is needed to avoid exponential explosion of code size. Thus for
9722 smaller units, the size is increased to @option{--param large-unit-insns}
9723 before applying @option{--param inline-unit-growth}. The default is 10000.
9725 @item inline-unit-growth
9726 Specifies maximal overall growth of the compilation unit caused by inlining.
9727 The default value is 20 which limits unit growth to 1.2 times the original
9728 size. Cold functions (either marked cold via an attribute or by profile
9729 feedback) are not accounted into the unit size.
9731 @item ipcp-unit-growth
9732 Specifies maximal overall growth of the compilation unit caused by
9733 interprocedural constant propagation. The default value is 10 which limits
9734 unit growth to 1.1 times the original size.
9736 @item large-stack-frame
9737 The limit specifying large stack frames. While inlining the algorithm is trying
9738 to not grow past this limit too much. The default value is 256 bytes.
9740 @item large-stack-frame-growth
9741 Specifies maximal growth of large stack frames caused by inlining in percents.
9742 The default value is 1000 which limits large stack frame growth to 11 times
9745 @item max-inline-insns-recursive
9746 @itemx max-inline-insns-recursive-auto
9747 Specifies the maximum number of instructions an out-of-line copy of a
9748 self-recursive inline
9749 function can grow into by performing recursive inlining.
9751 @option{--param max-inline-insns-recursive} applies to functions
9753 For functions not declared inline, recursive inlining
9754 happens only when @option{-finline-functions} (included in @option{-O3}) is
9755 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
9756 default value is 450.
9758 @item max-inline-recursive-depth
9759 @itemx max-inline-recursive-depth-auto
9760 Specifies the maximum recursion depth used for recursive inlining.
9762 @option{--param max-inline-recursive-depth} applies to functions
9763 declared inline. For functions not declared inline, recursive inlining
9764 happens only when @option{-finline-functions} (included in @option{-O3}) is
9765 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9768 @item min-inline-recursive-probability
9769 Recursive inlining is profitable only for function having deep recursion
9770 in average and can hurt for function having little recursion depth by
9771 increasing the prologue size or complexity of function body to other
9774 When profile feedback is available (see @option{-fprofile-generate}) the actual
9775 recursion depth can be guessed from the probability that function recurses
9776 via a given call expression. This parameter limits inlining only to call
9777 expressions whose probability exceeds the given threshold (in percents).
9778 The default value is 10.
9780 @item early-inlining-insns
9781 Specify growth that the early inliner can make. In effect it increases
9782 the amount of inlining for code having a large abstraction penalty.
9783 The default value is 14.
9785 @item max-early-inliner-iterations
9786 Limit of iterations of the early inliner. This basically bounds
9787 the number of nested indirect calls the early inliner can resolve.
9788 Deeper chains are still handled by late inlining.
9790 @item comdat-sharing-probability
9791 Probability (in percent) that C++ inline function with comdat visibility
9792 are shared across multiple compilation units. The default value is 20.
9794 @item profile-func-internal-id
9795 A parameter to control whether to use function internal id in profile
9796 database lookup. If the value is 0, the compiler uses an id that
9797 is based on function assembler name and filename, which makes old profile
9798 data more tolerant to source changes such as function reordering etc.
9799 The default value is 0.
9801 @item min-vect-loop-bound
9802 The minimum number of iterations under which loops are not vectorized
9803 when @option{-ftree-vectorize} is used. The number of iterations after
9804 vectorization needs to be greater than the value specified by this option
9805 to allow vectorization. The default value is 0.
9807 @item gcse-cost-distance-ratio
9808 Scaling factor in calculation of maximum distance an expression
9809 can be moved by GCSE optimizations. This is currently supported only in the
9810 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9811 is with simple expressions, i.e., the expressions that have cost
9812 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9813 hoisting of simple expressions. The default value is 10.
9815 @item gcse-unrestricted-cost
9816 Cost, roughly measured as the cost of a single typical machine
9817 instruction, at which GCSE optimizations do not constrain
9818 the distance an expression can travel. This is currently
9819 supported only in the code hoisting pass. The lesser the cost,
9820 the more aggressive code hoisting is. Specifying 0
9821 allows all expressions to travel unrestricted distances.
9822 The default value is 3.
9824 @item max-hoist-depth
9825 The depth of search in the dominator tree for expressions to hoist.
9826 This is used to avoid quadratic behavior in hoisting algorithm.
9827 The value of 0 does not limit on the search, but may slow down compilation
9828 of huge functions. The default value is 30.
9830 @item max-tail-merge-comparisons
9831 The maximum amount of similar bbs to compare a bb with. This is used to
9832 avoid quadratic behavior in tree tail merging. The default value is 10.
9834 @item max-tail-merge-iterations
9835 The maximum amount of iterations of the pass over the function. This is used to
9836 limit compilation time in tree tail merging. The default value is 2.
9838 @item store-merging-allow-unaligned
9839 Allow the store merging pass to introduce unaligned stores if it is legal to
9840 do so. The default value is 1.
9842 @item max-stores-to-merge
9843 The maximum number of stores to attempt to merge into wider stores in the store
9844 merging pass. The minimum value is 2 and the default is 64.
9846 @item max-unrolled-insns
9847 The maximum number of instructions that a loop may have to be unrolled.
9848 If a loop is unrolled, this parameter also determines how many times
9849 the loop code is unrolled.
9851 @item max-average-unrolled-insns
9852 The maximum number of instructions biased by probabilities of their execution
9853 that a loop may have to be unrolled. If a loop is unrolled,
9854 this parameter also determines how many times the loop code is unrolled.
9856 @item max-unroll-times
9857 The maximum number of unrollings of a single loop.
9859 @item max-peeled-insns
9860 The maximum number of instructions that a loop may have to be peeled.
9861 If a loop is peeled, this parameter also determines how many times
9862 the loop code is peeled.
9864 @item max-peel-times
9865 The maximum number of peelings of a single loop.
9867 @item max-peel-branches
9868 The maximum number of branches on the hot path through the peeled sequence.
9870 @item max-completely-peeled-insns
9871 The maximum number of insns of a completely peeled loop.
9873 @item max-completely-peel-times
9874 The maximum number of iterations of a loop to be suitable for complete peeling.
9876 @item max-completely-peel-loop-nest-depth
9877 The maximum depth of a loop nest suitable for complete peeling.
9879 @item max-unswitch-insns
9880 The maximum number of insns of an unswitched loop.
9882 @item max-unswitch-level
9883 The maximum number of branches unswitched in a single loop.
9885 @item max-loop-headers-insns
9886 The maximum number of insns in loop header duplicated by the copy loop headers
9890 The minimum cost of an expensive expression in the loop invariant motion.
9892 @item iv-consider-all-candidates-bound
9893 Bound on number of candidates for induction variables, below which
9894 all candidates are considered for each use in induction variable
9895 optimizations. If there are more candidates than this,
9896 only the most relevant ones are considered to avoid quadratic time complexity.
9898 @item iv-max-considered-uses
9899 The induction variable optimizations give up on loops that contain more
9900 induction variable uses.
9902 @item iv-always-prune-cand-set-bound
9903 If the number of candidates in the set is smaller than this value,
9904 always try to remove unnecessary ivs from the set
9905 when adding a new one.
9907 @item avg-loop-niter
9908 Average number of iterations of a loop.
9910 @item dse-max-object-size
9911 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
9912 Larger values may result in larger compilation times.
9914 @item scev-max-expr-size
9915 Bound on size of expressions used in the scalar evolutions analyzer.
9916 Large expressions slow the analyzer.
9918 @item scev-max-expr-complexity
9919 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9920 Complex expressions slow the analyzer.
9922 @item max-tree-if-conversion-phi-args
9923 Maximum number of arguments in a PHI supported by TREE if conversion
9924 unless the loop is marked with simd pragma.
9926 @item vect-max-version-for-alignment-checks
9927 The maximum number of run-time checks that can be performed when
9928 doing loop versioning for alignment in the vectorizer.
9930 @item vect-max-version-for-alias-checks
9931 The maximum number of run-time checks that can be performed when
9932 doing loop versioning for alias in the vectorizer.
9934 @item vect-max-peeling-for-alignment
9935 The maximum number of loop peels to enhance access alignment
9936 for vectorizer. Value -1 means no limit.
9938 @item max-iterations-to-track
9939 The maximum number of iterations of a loop the brute-force algorithm
9940 for analysis of the number of iterations of the loop tries to evaluate.
9942 @item hot-bb-count-ws-permille
9943 A basic block profile count is considered hot if it contributes to
9944 the given permillage (i.e. 0...1000) of the entire profiled execution.
9946 @item hot-bb-frequency-fraction
9947 Select fraction of the entry block frequency of executions of basic block in
9948 function given basic block needs to have to be considered hot.
9950 @item max-predicted-iterations
9951 The maximum number of loop iterations we predict statically. This is useful
9952 in cases where a function contains a single loop with known bound and
9953 another loop with unknown bound.
9954 The known number of iterations is predicted correctly, while
9955 the unknown number of iterations average to roughly 10. This means that the
9956 loop without bounds appears artificially cold relative to the other one.
9958 @item builtin-expect-probability
9959 Control the probability of the expression having the specified value. This
9960 parameter takes a percentage (i.e. 0 ... 100) as input.
9961 The default probability of 90 is obtained empirically.
9963 @item align-threshold
9965 Select fraction of the maximal frequency of executions of a basic block in
9966 a function to align the basic block.
9968 @item align-loop-iterations
9970 A loop expected to iterate at least the selected number of iterations is
9973 @item tracer-dynamic-coverage
9974 @itemx tracer-dynamic-coverage-feedback
9976 This value is used to limit superblock formation once the given percentage of
9977 executed instructions is covered. This limits unnecessary code size
9980 The @option{tracer-dynamic-coverage-feedback} parameter
9981 is used only when profile
9982 feedback is available. The real profiles (as opposed to statically estimated
9983 ones) are much less balanced allowing the threshold to be larger value.
9985 @item tracer-max-code-growth
9986 Stop tail duplication once code growth has reached given percentage. This is
9987 a rather artificial limit, as most of the duplicates are eliminated later in
9988 cross jumping, so it may be set to much higher values than is the desired code
9991 @item tracer-min-branch-ratio
9993 Stop reverse growth when the reverse probability of best edge is less than this
9994 threshold (in percent).
9996 @item tracer-min-branch-probability
9997 @itemx tracer-min-branch-probability-feedback
9999 Stop forward growth if the best edge has probability lower than this
10002 Similarly to @option{tracer-dynamic-coverage} two parameters are
10003 provided. @option{tracer-min-branch-probability-feedback} is used for
10004 compilation with profile feedback and @option{tracer-min-branch-probability}
10005 compilation without. The value for compilation with profile feedback
10006 needs to be more conservative (higher) in order to make tracer
10009 @item max-cse-path-length
10011 The maximum number of basic blocks on path that CSE considers.
10014 @item max-cse-insns
10015 The maximum number of instructions CSE processes before flushing.
10016 The default is 1000.
10018 @item ggc-min-expand
10020 GCC uses a garbage collector to manage its own memory allocation. This
10021 parameter specifies the minimum percentage by which the garbage
10022 collector's heap should be allowed to expand between collections.
10023 Tuning this may improve compilation speed; it has no effect on code
10026 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10027 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10028 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10029 GCC is not able to calculate RAM on a particular platform, the lower
10030 bound of 30% is used. Setting this parameter and
10031 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10032 every opportunity. This is extremely slow, but can be useful for
10035 @item ggc-min-heapsize
10037 Minimum size of the garbage collector's heap before it begins bothering
10038 to collect garbage. The first collection occurs after the heap expands
10039 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10040 tuning this may improve compilation speed, and has no effect on code
10043 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10044 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10045 with a lower bound of 4096 (four megabytes) and an upper bound of
10046 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10047 particular platform, the lower bound is used. Setting this parameter
10048 very large effectively disables garbage collection. Setting this
10049 parameter and @option{ggc-min-expand} to zero causes a full collection
10050 to occur at every opportunity.
10052 @item max-reload-search-insns
10053 The maximum number of instruction reload should look backward for equivalent
10054 register. Increasing values mean more aggressive optimization, making the
10055 compilation time increase with probably slightly better performance.
10056 The default value is 100.
10058 @item max-cselib-memory-locations
10059 The maximum number of memory locations cselib should take into account.
10060 Increasing values mean more aggressive optimization, making the compilation time
10061 increase with probably slightly better performance. The default value is 500.
10063 @item max-sched-ready-insns
10064 The maximum number of instructions ready to be issued the scheduler should
10065 consider at any given time during the first scheduling pass. Increasing
10066 values mean more thorough searches, making the compilation time increase
10067 with probably little benefit. The default value is 100.
10069 @item max-sched-region-blocks
10070 The maximum number of blocks in a region to be considered for
10071 interblock scheduling. The default value is 10.
10073 @item max-pipeline-region-blocks
10074 The maximum number of blocks in a region to be considered for
10075 pipelining in the selective scheduler. The default value is 15.
10077 @item max-sched-region-insns
10078 The maximum number of insns in a region to be considered for
10079 interblock scheduling. The default value is 100.
10081 @item max-pipeline-region-insns
10082 The maximum number of insns in a region to be considered for
10083 pipelining in the selective scheduler. The default value is 200.
10085 @item min-spec-prob
10086 The minimum probability (in percents) of reaching a source block
10087 for interblock speculative scheduling. The default value is 40.
10089 @item max-sched-extend-regions-iters
10090 The maximum number of iterations through CFG to extend regions.
10091 A value of 0 (the default) disables region extensions.
10093 @item max-sched-insn-conflict-delay
10094 The maximum conflict delay for an insn to be considered for speculative motion.
10095 The default value is 3.
10097 @item sched-spec-prob-cutoff
10098 The minimal probability of speculation success (in percents), so that
10099 speculative insns are scheduled.
10100 The default value is 40.
10102 @item sched-state-edge-prob-cutoff
10103 The minimum probability an edge must have for the scheduler to save its
10105 The default value is 10.
10107 @item sched-mem-true-dep-cost
10108 Minimal distance (in CPU cycles) between store and load targeting same
10109 memory locations. The default value is 1.
10111 @item selsched-max-lookahead
10112 The maximum size of the lookahead window of selective scheduling. It is a
10113 depth of search for available instructions.
10114 The default value is 50.
10116 @item selsched-max-sched-times
10117 The maximum number of times that an instruction is scheduled during
10118 selective scheduling. This is the limit on the number of iterations
10119 through which the instruction may be pipelined. The default value is 2.
10121 @item selsched-insns-to-rename
10122 The maximum number of best instructions in the ready list that are considered
10123 for renaming in the selective scheduler. The default value is 2.
10126 The minimum value of stage count that swing modulo scheduler
10127 generates. The default value is 2.
10129 @item max-last-value-rtl
10130 The maximum size measured as number of RTLs that can be recorded in an expression
10131 in combiner for a pseudo register as last known value of that register. The default
10134 @item max-combine-insns
10135 The maximum number of instructions the RTL combiner tries to combine.
10136 The default value is 2 at @option{-Og} and 4 otherwise.
10138 @item integer-share-limit
10139 Small integer constants can use a shared data structure, reducing the
10140 compiler's memory usage and increasing its speed. This sets the maximum
10141 value of a shared integer constant. The default value is 256.
10143 @item ssp-buffer-size
10144 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10145 protection when @option{-fstack-protection} is used.
10147 @item min-size-for-stack-sharing
10148 The minimum size of variables taking part in stack slot sharing when not
10149 optimizing. The default value is 32.
10151 @item max-jump-thread-duplication-stmts
10152 Maximum number of statements allowed in a block that needs to be
10153 duplicated when threading jumps.
10155 @item max-fields-for-field-sensitive
10156 Maximum number of fields in a structure treated in
10157 a field sensitive manner during pointer analysis. The default is zero
10158 for @option{-O0} and @option{-O1},
10159 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10161 @item prefetch-latency
10162 Estimate on average number of instructions that are executed before
10163 prefetch finishes. The distance prefetched ahead is proportional
10164 to this constant. Increasing this number may also lead to less
10165 streams being prefetched (see @option{simultaneous-prefetches}).
10167 @item simultaneous-prefetches
10168 Maximum number of prefetches that can run at the same time.
10170 @item l1-cache-line-size
10171 The size of cache line in L1 cache, in bytes.
10173 @item l1-cache-size
10174 The size of L1 cache, in kilobytes.
10176 @item l2-cache-size
10177 The size of L2 cache, in kilobytes.
10179 @item min-insn-to-prefetch-ratio
10180 The minimum ratio between the number of instructions and the
10181 number of prefetches to enable prefetching in a loop.
10183 @item prefetch-min-insn-to-mem-ratio
10184 The minimum ratio between the number of instructions and the
10185 number of memory references to enable prefetching in a loop.
10187 @item use-canonical-types
10188 Whether the compiler should use the ``canonical'' type system. By
10189 default, this should always be 1, which uses a more efficient internal
10190 mechanism for comparing types in C++ and Objective-C++. However, if
10191 bugs in the canonical type system are causing compilation failures,
10192 set this value to 0 to disable canonical types.
10194 @item switch-conversion-max-branch-ratio
10195 Switch initialization conversion refuses to create arrays that are
10196 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10197 branches in the switch.
10199 @item max-partial-antic-length
10200 Maximum length of the partial antic set computed during the tree
10201 partial redundancy elimination optimization (@option{-ftree-pre}) when
10202 optimizing at @option{-O3} and above. For some sorts of source code
10203 the enhanced partial redundancy elimination optimization can run away,
10204 consuming all of the memory available on the host machine. This
10205 parameter sets a limit on the length of the sets that are computed,
10206 which prevents the runaway behavior. Setting a value of 0 for
10207 this parameter allows an unlimited set length.
10209 @item sccvn-max-scc-size
10210 Maximum size of a strongly connected component (SCC) during SCCVN
10211 processing. If this limit is hit, SCCVN processing for the whole
10212 function is not done and optimizations depending on it are
10213 disabled. The default maximum SCC size is 10000.
10215 @item sccvn-max-alias-queries-per-access
10216 Maximum number of alias-oracle queries we perform when looking for
10217 redundancies for loads and stores. If this limit is hit the search
10218 is aborted and the load or store is not considered redundant. The
10219 number of queries is algorithmically limited to the number of
10220 stores on all paths from the load to the function entry.
10221 The default maximum number of queries is 1000.
10223 @item ira-max-loops-num
10224 IRA uses regional register allocation by default. If a function
10225 contains more loops than the number given by this parameter, only at most
10226 the given number of the most frequently-executed loops form regions
10227 for regional register allocation. The default value of the
10230 @item ira-max-conflict-table-size
10231 Although IRA uses a sophisticated algorithm to compress the conflict
10232 table, the table can still require excessive amounts of memory for
10233 huge functions. If the conflict table for a function could be more
10234 than the size in MB given by this parameter, the register allocator
10235 instead uses a faster, simpler, and lower-quality
10236 algorithm that does not require building a pseudo-register conflict table.
10237 The default value of the parameter is 2000.
10239 @item ira-loop-reserved-regs
10240 IRA can be used to evaluate more accurate register pressure in loops
10241 for decisions to move loop invariants (see @option{-O3}). The number
10242 of available registers reserved for some other purposes is given
10243 by this parameter. The default value of the parameter is 2, which is
10244 the minimal number of registers needed by typical instructions.
10245 This value is the best found from numerous experiments.
10247 @item lra-inheritance-ebb-probability-cutoff
10248 LRA tries to reuse values reloaded in registers in subsequent insns.
10249 This optimization is called inheritance. EBB is used as a region to
10250 do this optimization. The parameter defines a minimal fall-through
10251 edge probability in percentage used to add BB to inheritance EBB in
10252 LRA. The default value of the parameter is 40. The value was chosen
10253 from numerous runs of SPEC2000 on x86-64.
10255 @item loop-invariant-max-bbs-in-loop
10256 Loop invariant motion can be very expensive, both in compilation time and
10257 in amount of needed compile-time memory, with very large loops. Loops
10258 with more basic blocks than this parameter won't have loop invariant
10259 motion optimization performed on them. The default value of the
10260 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10262 @item loop-max-datarefs-for-datadeps
10263 Building data dependencies is expensive for very large loops. This
10264 parameter limits the number of data references in loops that are
10265 considered for data dependence analysis. These large loops are no
10266 handled by the optimizations using loop data dependencies.
10267 The default value is 1000.
10269 @item max-vartrack-size
10270 Sets a maximum number of hash table slots to use during variable
10271 tracking dataflow analysis of any function. If this limit is exceeded
10272 with variable tracking at assignments enabled, analysis for that
10273 function is retried without it, after removing all debug insns from
10274 the function. If the limit is exceeded even without debug insns, var
10275 tracking analysis is completely disabled for the function. Setting
10276 the parameter to zero makes it unlimited.
10278 @item max-vartrack-expr-depth
10279 Sets a maximum number of recursion levels when attempting to map
10280 variable names or debug temporaries to value expressions. This trades
10281 compilation time for more complete debug information. If this is set too
10282 low, value expressions that are available and could be represented in
10283 debug information may end up not being used; setting this higher may
10284 enable the compiler to find more complex debug expressions, but compile
10285 time and memory use may grow. The default is 12.
10287 @item min-nondebug-insn-uid
10288 Use uids starting at this parameter for nondebug insns. The range below
10289 the parameter is reserved exclusively for debug insns created by
10290 @option{-fvar-tracking-assignments}, but debug insns may get
10291 (non-overlapping) uids above it if the reserved range is exhausted.
10293 @item ipa-sra-ptr-growth-factor
10294 IPA-SRA replaces a pointer to an aggregate with one or more new
10295 parameters only when their cumulative size is less or equal to
10296 @option{ipa-sra-ptr-growth-factor} times the size of the original
10299 @item sra-max-scalarization-size-Ospeed
10300 @item sra-max-scalarization-size-Osize
10301 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10302 replace scalar parts of aggregates with uses of independent scalar
10303 variables. These parameters control the maximum size, in storage units,
10304 of aggregate which is considered for replacement when compiling for
10306 (@option{sra-max-scalarization-size-Ospeed}) or size
10307 (@option{sra-max-scalarization-size-Osize}) respectively.
10309 @item tm-max-aggregate-size
10310 When making copies of thread-local variables in a transaction, this
10311 parameter specifies the size in bytes after which variables are
10312 saved with the logging functions as opposed to save/restore code
10313 sequence pairs. This option only applies when using
10316 @item graphite-max-nb-scop-params
10317 To avoid exponential effects in the Graphite loop transforms, the
10318 number of parameters in a Static Control Part (SCoP) is bounded. The
10319 default value is 10 parameters. A variable whose value is unknown at
10320 compilation time and defined outside a SCoP is a parameter of the SCoP.
10322 @item graphite-max-bbs-per-function
10323 To avoid exponential effects in the detection of SCoPs, the size of
10324 the functions analyzed by Graphite is bounded. The default value is
10327 @item loop-block-tile-size
10328 Loop blocking or strip mining transforms, enabled with
10329 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10330 loop in the loop nest by a given number of iterations. The strip
10331 length can be changed using the @option{loop-block-tile-size}
10332 parameter. The default value is 51 iterations.
10334 @item loop-unroll-jam-size
10335 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10336 default value is 4.
10338 @item loop-unroll-jam-depth
10339 Specify the dimension to be unrolled (counting from the most inner loop)
10340 for the @option{-floop-unroll-and-jam}. The default value is 2.
10342 @item ipa-cp-value-list-size
10343 IPA-CP attempts to track all possible values and types passed to a function's
10344 parameter in order to propagate them and perform devirtualization.
10345 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10346 stores per one formal parameter of a function.
10348 @item ipa-cp-eval-threshold
10349 IPA-CP calculates its own score of cloning profitability heuristics
10350 and performs those cloning opportunities with scores that exceed
10351 @option{ipa-cp-eval-threshold}.
10353 @item ipa-cp-recursion-penalty
10354 Percentage penalty the recursive functions will receive when they
10355 are evaluated for cloning.
10357 @item ipa-cp-single-call-penalty
10358 Percentage penalty functions containing a single call to another
10359 function will receive when they are evaluated for cloning.
10362 @item ipa-max-agg-items
10363 IPA-CP is also capable to propagate a number of scalar values passed
10364 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10365 number of such values per one parameter.
10367 @item ipa-cp-loop-hint-bonus
10368 When IPA-CP determines that a cloning candidate would make the number
10369 of iterations of a loop known, it adds a bonus of
10370 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10373 @item ipa-cp-array-index-hint-bonus
10374 When IPA-CP determines that a cloning candidate would make the index of
10375 an array access known, it adds a bonus of
10376 @option{ipa-cp-array-index-hint-bonus} to the profitability
10377 score of the candidate.
10379 @item ipa-max-aa-steps
10380 During its analysis of function bodies, IPA-CP employs alias analysis
10381 in order to track values pointed to by function parameters. In order
10382 not spend too much time analyzing huge functions, it gives up and
10383 consider all memory clobbered after examining
10384 @option{ipa-max-aa-steps} statements modifying memory.
10386 @item lto-partitions
10387 Specify desired number of partitions produced during WHOPR compilation.
10388 The number of partitions should exceed the number of CPUs used for compilation.
10389 The default value is 32.
10391 @item lto-min-partition
10392 Size of minimal partition for WHOPR (in estimated instructions).
10393 This prevents expenses of splitting very small programs into too many
10396 @item lto-max-partition
10397 Size of max partition for WHOPR (in estimated instructions).
10398 to provide an upper bound for individual size of partition.
10399 Meant to be used only with balanced partitioning.
10401 @item cxx-max-namespaces-for-diagnostic-help
10402 The maximum number of namespaces to consult for suggestions when C++
10403 name lookup fails for an identifier. The default is 1000.
10405 @item sink-frequency-threshold
10406 The maximum relative execution frequency (in percents) of the target block
10407 relative to a statement's original block to allow statement sinking of a
10408 statement. Larger numbers result in more aggressive statement sinking.
10409 The default value is 75. A small positive adjustment is applied for
10410 statements with memory operands as those are even more profitable so sink.
10412 @item max-stores-to-sink
10413 The maximum number of conditional store pairs that can be sunk. Set to 0
10414 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10415 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10417 @item allow-store-data-races
10418 Allow optimizers to introduce new data races on stores.
10419 Set to 1 to allow, otherwise to 0. This option is enabled by default
10420 at optimization level @option{-Ofast}.
10422 @item case-values-threshold
10423 The smallest number of different values for which it is best to use a
10424 jump-table instead of a tree of conditional branches. If the value is
10425 0, use the default for the machine. The default is 0.
10427 @item tree-reassoc-width
10428 Set the maximum number of instructions executed in parallel in
10429 reassociated tree. This parameter overrides target dependent
10430 heuristics used by default if has non zero value.
10432 @item sched-pressure-algorithm
10433 Choose between the two available implementations of
10434 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10435 and is the more likely to prevent instructions from being reordered.
10436 Algorithm 2 was designed to be a compromise between the relatively
10437 conservative approach taken by algorithm 1 and the rather aggressive
10438 approach taken by the default scheduler. It relies more heavily on
10439 having a regular register file and accurate register pressure classes.
10440 See @file{haifa-sched.c} in the GCC sources for more details.
10442 The default choice depends on the target.
10444 @item max-slsr-cand-scan
10445 Set the maximum number of existing candidates that are considered when
10446 seeking a basis for a new straight-line strength reduction candidate.
10449 Enable buffer overflow detection for global objects. This kind
10450 of protection is enabled by default if you are using
10451 @option{-fsanitize=address} option.
10452 To disable global objects protection use @option{--param asan-globals=0}.
10455 Enable buffer overflow detection for stack objects. This kind of
10456 protection is enabled by default when using @option{-fsanitize=address}.
10457 To disable stack protection use @option{--param asan-stack=0} option.
10459 @item asan-instrument-reads
10460 Enable buffer overflow detection for memory reads. This kind of
10461 protection is enabled by default when using @option{-fsanitize=address}.
10462 To disable memory reads protection use
10463 @option{--param asan-instrument-reads=0}.
10465 @item asan-instrument-writes
10466 Enable buffer overflow detection for memory writes. This kind of
10467 protection is enabled by default when using @option{-fsanitize=address}.
10468 To disable memory writes protection use
10469 @option{--param asan-instrument-writes=0} option.
10471 @item asan-memintrin
10472 Enable detection for built-in functions. This kind of protection
10473 is enabled by default when using @option{-fsanitize=address}.
10474 To disable built-in functions protection use
10475 @option{--param asan-memintrin=0}.
10477 @item asan-use-after-return
10478 Enable detection of use-after-return. This kind of protection
10479 is enabled by default when using the @option{-fsanitize=address} option.
10480 To disable it use @option{--param asan-use-after-return=0}.
10482 Note: By default the check is disabled at run time. To enable it,
10483 add @code{detect_stack_use_after_return=1} to the environment variable
10484 @env{ASAN_OPTIONS}.
10486 @item asan-instrumentation-with-call-threshold
10487 If number of memory accesses in function being instrumented
10488 is greater or equal to this number, use callbacks instead of inline checks.
10489 E.g. to disable inline code use
10490 @option{--param asan-instrumentation-with-call-threshold=0}.
10492 @item use-after-scope-direct-emission-threshold
10493 If the size of a local variable in bytes is smaller or equal to this
10494 number, directly poison (or unpoison) shadow memory instead of using
10495 run-time callbacks. The default value is 256.
10497 @item chkp-max-ctor-size
10498 Static constructors generated by Pointer Bounds Checker may become very
10499 large and significantly increase compile time at optimization level
10500 @option{-O1} and higher. This parameter is a maximum number of statements
10501 in a single generated constructor. Default value is 5000.
10503 @item max-fsm-thread-path-insns
10504 Maximum number of instructions to copy when duplicating blocks on a
10505 finite state automaton jump thread path. The default is 100.
10507 @item max-fsm-thread-length
10508 Maximum number of basic blocks on a finite state automaton jump thread
10509 path. The default is 10.
10511 @item max-fsm-thread-paths
10512 Maximum number of new jump thread paths to create for a finite state
10513 automaton. The default is 50.
10515 @item parloops-chunk-size
10516 Chunk size of omp schedule for loops parallelized by parloops. The default
10519 @item parloops-schedule
10520 Schedule type of omp schedule for loops parallelized by parloops (static,
10521 dynamic, guided, auto, runtime). The default is static.
10523 @item max-ssa-name-query-depth
10524 Maximum depth of recursion when querying properties of SSA names in things
10525 like fold routines. One level of recursion corresponds to following a
10528 @item hsa-gen-debug-stores
10529 Enable emission of special debug stores within HSA kernels which are
10530 then read and reported by libgomp plugin. Generation of these stores
10531 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10534 @item max-speculative-devirt-maydefs
10535 The maximum number of may-defs we analyze when looking for a must-def
10536 specifying the dynamic type of an object that invokes a virtual call
10537 we may be able to devirtualize speculatively.
10539 @item max-vrp-switch-assertions
10540 The maximum number of assertions to add along the default edge of a switch
10541 statement during VRP. The default is 10.
10545 @node Instrumentation Options
10546 @section Program Instrumentation Options
10547 @cindex instrumentation options
10548 @cindex program instrumentation options
10549 @cindex run-time error checking options
10550 @cindex profiling options
10551 @cindex options, program instrumentation
10552 @cindex options, run-time error checking
10553 @cindex options, profiling
10555 GCC supports a number of command-line options that control adding
10556 run-time instrumentation to the code it normally generates.
10557 For example, one purpose of instrumentation is collect profiling
10558 statistics for use in finding program hot spots, code coverage
10559 analysis, or profile-guided optimizations.
10560 Another class of program instrumentation is adding run-time checking
10561 to detect programming errors like invalid pointer
10562 dereferences or out-of-bounds array accesses, as well as deliberately
10563 hostile attacks such as stack smashing or C++ vtable hijacking.
10564 There is also a general hook which can be used to implement other
10565 forms of tracing or function-level instrumentation for debug or
10566 program analysis purposes.
10569 @cindex @command{prof}
10572 Generate extra code to write profile information suitable for the
10573 analysis program @command{prof}. You must use this option when compiling
10574 the source files you want data about, and you must also use it when
10577 @cindex @command{gprof}
10580 Generate extra code to write profile information suitable for the
10581 analysis program @command{gprof}. You must use this option when compiling
10582 the source files you want data about, and you must also use it when
10585 @item -fprofile-arcs
10586 @opindex fprofile-arcs
10587 Add code so that program flow @dfn{arcs} are instrumented. During
10588 execution the program records how many times each branch and call is
10589 executed and how many times it is taken or returns. On targets that support
10590 constructors with priority support, profiling properly handles constructors,
10591 destructors and C++ constructors (and destructors) of classes which are used
10592 as a type of a global variable.
10595 program exits it saves this data to a file called
10596 @file{@var{auxname}.gcda} for each source file. The data may be used for
10597 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10598 test coverage analysis (@option{-ftest-coverage}). Each object file's
10599 @var{auxname} is generated from the name of the output file, if
10600 explicitly specified and it is not the final executable, otherwise it is
10601 the basename of the source file. In both cases any suffix is removed
10602 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10603 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10604 @xref{Cross-profiling}.
10606 @cindex @command{gcov}
10610 This option is used to compile and link code instrumented for coverage
10611 analysis. The option is a synonym for @option{-fprofile-arcs}
10612 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10613 linking). See the documentation for those options for more details.
10618 Compile the source files with @option{-fprofile-arcs} plus optimization
10619 and code generation options. For test coverage analysis, use the
10620 additional @option{-ftest-coverage} option. You do not need to profile
10621 every source file in a program.
10624 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10625 (the latter implies the former).
10628 Run the program on a representative workload to generate the arc profile
10629 information. This may be repeated any number of times. You can run
10630 concurrent instances of your program, and provided that the file system
10631 supports locking, the data files will be correctly updated. Also
10632 @code{fork} calls are detected and correctly handled (double counting
10636 For profile-directed optimizations, compile the source files again with
10637 the same optimization and code generation options plus
10638 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10639 Control Optimization}).
10642 For test coverage analysis, use @command{gcov} to produce human readable
10643 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10644 @command{gcov} documentation for further information.
10648 With @option{-fprofile-arcs}, for each function of your program GCC
10649 creates a program flow graph, then finds a spanning tree for the graph.
10650 Only arcs that are not on the spanning tree have to be instrumented: the
10651 compiler adds code to count the number of times that these arcs are
10652 executed. When an arc is the only exit or only entrance to a block, the
10653 instrumentation code can be added to the block; otherwise, a new basic
10654 block must be created to hold the instrumentation code.
10657 @item -ftest-coverage
10658 @opindex ftest-coverage
10659 Produce a notes file that the @command{gcov} code-coverage utility
10660 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10661 show program coverage. Each source file's note file is called
10662 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10663 above for a description of @var{auxname} and instructions on how to
10664 generate test coverage data. Coverage data matches the source files
10665 more closely if you do not optimize.
10667 @item -fprofile-dir=@var{path}
10668 @opindex fprofile-dir
10670 Set the directory to search for the profile data files in to @var{path}.
10671 This option affects only the profile data generated by
10672 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10673 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10674 and its related options. Both absolute and relative paths can be used.
10675 By default, GCC uses the current directory as @var{path}, thus the
10676 profile data file appears in the same directory as the object file.
10678 @item -fprofile-generate
10679 @itemx -fprofile-generate=@var{path}
10680 @opindex fprofile-generate
10682 Enable options usually used for instrumenting application to produce
10683 profile useful for later recompilation with profile feedback based
10684 optimization. You must use @option{-fprofile-generate} both when
10685 compiling and when linking your program.
10687 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10689 If @var{path} is specified, GCC looks at the @var{path} to find
10690 the profile feedback data files. See @option{-fprofile-dir}.
10692 To optimize the program based on the collected profile information, use
10693 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10695 @item -fprofile-update=@var{method}
10696 @opindex fprofile-update
10698 Alter the update method for an application instrumented for profile
10699 feedback based optimization. The @var{method} argument should be one of
10700 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
10701 The first one is useful for single-threaded applications,
10702 while the second one prevents profile corruption by emitting thread-safe code.
10704 @strong{Warning:} When an application does not properly join all threads
10705 (or creates an detached thread), a profile file can be still corrupted.
10707 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
10708 when supported by a target, or to @samp{single} otherwise. The GCC driver
10709 automatically selects @samp{prefer-atomic} when @option{-pthread}
10710 is present in the command line.
10712 @item -fsanitize=address
10713 @opindex fsanitize=address
10714 Enable AddressSanitizer, a fast memory error detector.
10715 Memory access instructions are instrumented to detect
10716 out-of-bounds and use-after-free bugs.
10717 The option enables @option{-fsanitize-address-use-after-scope}.
10718 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
10719 more details. The run-time behavior can be influenced using the
10720 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
10721 the available options are shown at startup of the instrumented program. See
10722 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
10723 for a list of supported options.
10724 The option cannot be combined with @option{-fsanitize=thread}
10725 and/or @option{-fcheck-pointer-bounds}.
10727 @item -fsanitize=kernel-address
10728 @opindex fsanitize=kernel-address
10729 Enable AddressSanitizer for Linux kernel.
10730 See @uref{https://github.com/google/kasan/wiki} for more details.
10731 The option cannot be combined with @option{-fcheck-pointer-bounds}.
10733 @item -fsanitize=thread
10734 @opindex fsanitize=thread
10735 Enable ThreadSanitizer, a fast data race detector.
10736 Memory access instructions are instrumented to detect
10737 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
10738 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
10739 environment variable; see
10740 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
10742 The option cannot be combined with @option{-fsanitize=address},
10743 @option{-fsanitize=leak} and/or @option{-fcheck-pointer-bounds}.
10745 Note that sanitized atomic builtins cannot throw exceptions when
10746 operating on invalid memory addresses with non-call exceptions
10747 (@option{-fnon-call-exceptions}).
10749 @item -fsanitize=leak
10750 @opindex fsanitize=leak
10751 Enable LeakSanitizer, a memory leak detector.
10752 This option only matters for linking of executables and
10753 the executable is linked against a library that overrides @code{malloc}
10754 and other allocator functions. See
10755 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
10756 details. The run-time behavior can be influenced using the
10757 @env{LSAN_OPTIONS} environment variable.
10758 The option cannot be combined with @option{-fsanitize=thread}.
10760 @item -fsanitize=undefined
10761 @opindex fsanitize=undefined
10762 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
10763 Various computations are instrumented to detect undefined behavior
10764 at runtime. Current suboptions are:
10768 @item -fsanitize=shift
10769 @opindex fsanitize=shift
10770 This option enables checking that the result of a shift operation is
10771 not undefined. Note that what exactly is considered undefined differs
10772 slightly between C and C++, as well as between ISO C90 and C99, etc.
10773 This option has two suboptions, @option{-fsanitize=shift-base} and
10774 @option{-fsanitize=shift-exponent}.
10776 @item -fsanitize=shift-exponent
10777 @opindex fsanitize=shift-exponent
10778 This option enables checking that the second argument of a shift operation
10779 is not negative and is smaller than the precision of the promoted first
10782 @item -fsanitize=shift-base
10783 @opindex fsanitize=shift-base
10784 If the second argument of a shift operation is within range, check that the
10785 result of a shift operation is not undefined. Note that what exactly is
10786 considered undefined differs slightly between C and C++, as well as between
10787 ISO C90 and C99, etc.
10789 @item -fsanitize=integer-divide-by-zero
10790 @opindex fsanitize=integer-divide-by-zero
10791 Detect integer division by zero as well as @code{INT_MIN / -1} division.
10793 @item -fsanitize=unreachable
10794 @opindex fsanitize=unreachable
10795 With this option, the compiler turns the @code{__builtin_unreachable}
10796 call into a diagnostics message call instead. When reaching the
10797 @code{__builtin_unreachable} call, the behavior is undefined.
10799 @item -fsanitize=vla-bound
10800 @opindex fsanitize=vla-bound
10801 This option instructs the compiler to check that the size of a variable
10802 length array is positive.
10804 @item -fsanitize=null
10805 @opindex fsanitize=null
10806 This option enables pointer checking. Particularly, the application
10807 built with this option turned on will issue an error message when it
10808 tries to dereference a NULL pointer, or if a reference (possibly an
10809 rvalue reference) is bound to a NULL pointer, or if a method is invoked
10810 on an object pointed by a NULL pointer.
10812 @item -fsanitize=return
10813 @opindex fsanitize=return
10814 This option enables return statement checking. Programs
10815 built with this option turned on will issue an error message
10816 when the end of a non-void function is reached without actually
10817 returning a value. This option works in C++ only.
10819 @item -fsanitize=signed-integer-overflow
10820 @opindex fsanitize=signed-integer-overflow
10821 This option enables signed integer overflow checking. We check that
10822 the result of @code{+}, @code{*}, and both unary and binary @code{-}
10823 does not overflow in the signed arithmetics. Note, integer promotion
10824 rules must be taken into account. That is, the following is not an
10827 signed char a = SCHAR_MAX;
10831 @item -fsanitize=bounds
10832 @opindex fsanitize=bounds
10833 This option enables instrumentation of array bounds. Various out of bounds
10834 accesses are detected. Flexible array members, flexible array member-like
10835 arrays, and initializers of variables with static storage are not instrumented.
10836 The option cannot be combined with @option{-fcheck-pointer-bounds}.
10838 @item -fsanitize=bounds-strict
10839 @opindex fsanitize=bounds-strict
10840 This option enables strict instrumentation of array bounds. Most out of bounds
10841 accesses are detected, including flexible array members and flexible array
10842 member-like arrays. Initializers of variables with static storage are not
10843 instrumented. The option cannot be combined
10844 with @option{-fcheck-pointer-bounds}.
10846 @item -fsanitize=alignment
10847 @opindex fsanitize=alignment
10849 This option enables checking of alignment of pointers when they are
10850 dereferenced, or when a reference is bound to insufficiently aligned target,
10851 or when a method or constructor is invoked on insufficiently aligned object.
10853 @item -fsanitize=object-size
10854 @opindex fsanitize=object-size
10855 This option enables instrumentation of memory references using the
10856 @code{__builtin_object_size} function. Various out of bounds pointer
10857 accesses are detected.
10859 @item -fsanitize=float-divide-by-zero
10860 @opindex fsanitize=float-divide-by-zero
10861 Detect floating-point division by zero. Unlike other similar options,
10862 @option{-fsanitize=float-divide-by-zero} is not enabled by
10863 @option{-fsanitize=undefined}, since floating-point division by zero can
10864 be a legitimate way of obtaining infinities and NaNs.
10866 @item -fsanitize=float-cast-overflow
10867 @opindex fsanitize=float-cast-overflow
10868 This option enables floating-point type to integer conversion checking.
10869 We check that the result of the conversion does not overflow.
10870 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
10871 not enabled by @option{-fsanitize=undefined}.
10872 This option does not work well with @code{FE_INVALID} exceptions enabled.
10874 @item -fsanitize=nonnull-attribute
10875 @opindex fsanitize=nonnull-attribute
10877 This option enables instrumentation of calls, checking whether null values
10878 are not passed to arguments marked as requiring a non-null value by the
10879 @code{nonnull} function attribute.
10881 @item -fsanitize=returns-nonnull-attribute
10882 @opindex fsanitize=returns-nonnull-attribute
10884 This option enables instrumentation of return statements in functions
10885 marked with @code{returns_nonnull} function attribute, to detect returning
10886 of null values from such functions.
10888 @item -fsanitize=bool
10889 @opindex fsanitize=bool
10891 This option enables instrumentation of loads from bool. If a value other
10892 than 0/1 is loaded, a run-time error is issued.
10894 @item -fsanitize=enum
10895 @opindex fsanitize=enum
10897 This option enables instrumentation of loads from an enum type. If
10898 a value outside the range of values for the enum type is loaded,
10899 a run-time error is issued.
10901 @item -fsanitize=vptr
10902 @opindex fsanitize=vptr
10904 This option enables instrumentation of C++ member function calls, member
10905 accesses and some conversions between pointers to base and derived classes,
10906 to verify the referenced object has the correct dynamic type.
10910 While @option{-ftrapv} causes traps for signed overflows to be emitted,
10911 @option{-fsanitize=undefined} gives a diagnostic message.
10912 This currently works only for the C family of languages.
10914 @item -fno-sanitize=all
10915 @opindex fno-sanitize=all
10917 This option disables all previously enabled sanitizers.
10918 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
10921 @item -fasan-shadow-offset=@var{number}
10922 @opindex fasan-shadow-offset
10923 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
10924 It is useful for experimenting with different shadow memory layouts in
10925 Kernel AddressSanitizer.
10927 @item -fsanitize-sections=@var{s1},@var{s2},...
10928 @opindex fsanitize-sections
10929 Sanitize global variables in selected user-defined sections. @var{si} may
10932 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
10933 @opindex fsanitize-recover
10934 @opindex fno-sanitize-recover
10935 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
10936 mentioned in comma-separated list of @var{opts}. Enabling this option
10937 for a sanitizer component causes it to attempt to continue
10938 running the program as if no error happened. This means multiple
10939 runtime errors can be reported in a single program run, and the exit
10940 code of the program may indicate success even when errors
10941 have been reported. The @option{-fno-sanitize-recover=} option
10942 can be used to alter
10943 this behavior: only the first detected error is reported
10944 and program then exits with a non-zero exit code.
10946 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
10947 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
10948 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
10949 @option{-fsanitize=bounds-strict},
10950 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
10951 For these sanitizers error recovery is turned on by default,
10952 except @option{-fsanitize=address}, for which this feature is experimental.
10953 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
10954 accepted, the former enables recovery for all sanitizers that support it,
10955 the latter disables recovery for all sanitizers that support it.
10957 Even if a recovery mode is turned on the compiler side, it needs to be also
10958 enabled on the runtime library side, otherwise the failures are still fatal.
10959 The runtime library defaults to @code{halt_on_error=0} for
10960 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
10961 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
10962 setting the @code{halt_on_error} flag in the corresponding environment variable.
10964 Syntax without an explicit @var{opts} parameter is deprecated. It is
10965 equivalent to specifying an @var{opts} list of:
10968 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10971 @item -fsanitize-address-use-after-scope
10972 @opindex fsanitize-address-use-after-scope
10973 Enable sanitization of local variables to detect use-after-scope bugs.
10974 The option sets @option{-fstack-reuse} to @samp{none}.
10976 @item -fsanitize-undefined-trap-on-error
10977 @opindex fsanitize-undefined-trap-on-error
10978 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
10979 report undefined behavior using @code{__builtin_trap} rather than
10980 a @code{libubsan} library routine. The advantage of this is that the
10981 @code{libubsan} library is not needed and is not linked in, so this
10982 is usable even in freestanding environments.
10984 @item -fsanitize-coverage=trace-pc
10985 @opindex fsanitize-coverage=trace-pc
10986 Enable coverage-guided fuzzing code instrumentation.
10987 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
10989 @item -fbounds-check
10990 @opindex fbounds-check
10991 For front ends that support it, generate additional code to check that
10992 indices used to access arrays are within the declared range. This is
10993 currently only supported by the Fortran front end, where this option
10996 @item -fcheck-pointer-bounds
10997 @opindex fcheck-pointer-bounds
10998 @opindex fno-check-pointer-bounds
10999 @cindex Pointer Bounds Checker options
11000 Enable Pointer Bounds Checker instrumentation. Each memory reference
11001 is instrumented with checks of the pointer used for memory access against
11002 bounds associated with that pointer.
11005 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
11006 and @option{-mmpx} are required to enable this feature.
11007 MPX-based instrumentation requires
11008 a runtime library to enable MPX in hardware and handle bounds
11009 violation signals. By default when @option{-fcheck-pointer-bounds}
11010 and @option{-mmpx} options are used to link a program, the GCC driver
11011 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
11012 Bounds checking on calls to dynamic libraries requires a linker
11013 with @option{-z bndplt} support; if GCC was configured with a linker
11014 without support for this option (including the Gold linker and older
11015 versions of ld), a warning is given if you link with @option{-mmpx}
11016 without also specifying @option{-static}, since the overall effectiveness
11017 of the bounds checking protection is reduced.
11018 See also @option{-static-libmpxwrappers}.
11020 MPX-based instrumentation
11021 may be used for debugging and also may be included in production code
11022 to increase program security. Depending on usage, you may
11023 have different requirements for the runtime library. The current version
11024 of the MPX runtime library is more oriented for use as a debugging
11025 tool. MPX runtime library usage implies @option{-lpthread}. See
11026 also @option{-static-libmpx}. The runtime library behavior can be
11027 influenced using various @env{CHKP_RT_*} environment variables. See
11028 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
11031 Generated instrumentation may be controlled by various
11032 @option{-fchkp-*} options and by the @code{bnd_variable_size}
11033 structure field attribute (@pxref{Type Attributes}) and
11034 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
11035 (@pxref{Function Attributes}). GCC also provides a number of built-in
11036 functions for controlling the Pointer Bounds Checker. @xref{Pointer
11037 Bounds Checker builtins}, for more information.
11039 @item -fchkp-check-incomplete-type
11040 @opindex fchkp-check-incomplete-type
11041 @opindex fno-chkp-check-incomplete-type
11042 Generate pointer bounds checks for variables with incomplete type.
11043 Enabled by default.
11045 @item -fchkp-narrow-bounds
11046 @opindex fchkp-narrow-bounds
11047 @opindex fno-chkp-narrow-bounds
11048 Controls bounds used by Pointer Bounds Checker for pointers to object
11049 fields. If narrowing is enabled then field bounds are used. Otherwise
11050 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
11051 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
11053 @item -fchkp-first-field-has-own-bounds
11054 @opindex fchkp-first-field-has-own-bounds
11055 @opindex fno-chkp-first-field-has-own-bounds
11056 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
11057 first field in the structure. By default a pointer to the first field has
11058 the same bounds as a pointer to the whole structure.
11060 @item -fchkp-flexible-struct-trailing-arrays
11061 @opindex fchkp-flexible-struct-trailing-arrays
11062 @opindex fno-chkp-flexible-struct-trailing-arrays
11063 Forces Pointer Bounds Checker to treat all trailing arrays in structures as
11064 possibly flexible. By default only array fields with zero length or that are
11065 marked with attribute bnd_variable_size are treated as flexible.
11067 @item -fchkp-narrow-to-innermost-array
11068 @opindex fchkp-narrow-to-innermost-array
11069 @opindex fno-chkp-narrow-to-innermost-array
11070 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
11071 case of nested static array access. By default this option is disabled and
11072 bounds of the outermost array are used.
11074 @item -fchkp-optimize
11075 @opindex fchkp-optimize
11076 @opindex fno-chkp-optimize
11077 Enables Pointer Bounds Checker optimizations. Enabled by default at
11078 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11080 @item -fchkp-use-fast-string-functions
11081 @opindex fchkp-use-fast-string-functions
11082 @opindex fno-chkp-use-fast-string-functions
11083 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11084 by Pointer Bounds Checker. Disabled by default.
11086 @item -fchkp-use-nochk-string-functions
11087 @opindex fchkp-use-nochk-string-functions
11088 @opindex fno-chkp-use-nochk-string-functions
11089 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11090 by Pointer Bounds Checker. Disabled by default.
11092 @item -fchkp-use-static-bounds
11093 @opindex fchkp-use-static-bounds
11094 @opindex fno-chkp-use-static-bounds
11095 Allow Pointer Bounds Checker to generate static bounds holding
11096 bounds of static variables. Enabled by default.
11098 @item -fchkp-use-static-const-bounds
11099 @opindex fchkp-use-static-const-bounds
11100 @opindex fno-chkp-use-static-const-bounds
11101 Use statically-initialized bounds for constant bounds instead of
11102 generating them each time they are required. By default enabled when
11103 @option{-fchkp-use-static-bounds} is enabled.
11105 @item -fchkp-treat-zero-dynamic-size-as-infinite
11106 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11107 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11108 With this option, objects with incomplete type whose
11109 dynamically-obtained size is zero are treated as having infinite size
11110 instead by Pointer Bounds
11111 Checker. This option may be helpful if a program is linked with a library
11112 missing size information for some symbols. Disabled by default.
11114 @item -fchkp-check-read
11115 @opindex fchkp-check-read
11116 @opindex fno-chkp-check-read
11117 Instructs Pointer Bounds Checker to generate checks for all read
11118 accesses to memory. Enabled by default.
11120 @item -fchkp-check-write
11121 @opindex fchkp-check-write
11122 @opindex fno-chkp-check-write
11123 Instructs Pointer Bounds Checker to generate checks for all write
11124 accesses to memory. Enabled by default.
11126 @item -fchkp-store-bounds
11127 @opindex fchkp-store-bounds
11128 @opindex fno-chkp-store-bounds
11129 Instructs Pointer Bounds Checker to generate bounds stores for
11130 pointer writes. Enabled by default.
11132 @item -fchkp-instrument-calls
11133 @opindex fchkp-instrument-calls
11134 @opindex fno-chkp-instrument-calls
11135 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11136 Enabled by default.
11138 @item -fchkp-instrument-marked-only
11139 @opindex fchkp-instrument-marked-only
11140 @opindex fno-chkp-instrument-marked-only
11141 Instructs Pointer Bounds Checker to instrument only functions
11142 marked with the @code{bnd_instrument} attribute
11143 (@pxref{Function Attributes}). Disabled by default.
11145 @item -fchkp-use-wrappers
11146 @opindex fchkp-use-wrappers
11147 @opindex fno-chkp-use-wrappers
11148 Allows Pointer Bounds Checker to replace calls to built-in functions
11149 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11150 is used to link a program, the GCC driver automatically links
11151 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11152 Enabled by default.
11154 @item -fstack-protector
11155 @opindex fstack-protector
11156 Emit extra code to check for buffer overflows, such as stack smashing
11157 attacks. This is done by adding a guard variable to functions with
11158 vulnerable objects. This includes functions that call @code{alloca}, and
11159 functions with buffers larger than 8 bytes. The guards are initialized
11160 when a function is entered and then checked when the function exits.
11161 If a guard check fails, an error message is printed and the program exits.
11163 @item -fstack-protector-all
11164 @opindex fstack-protector-all
11165 Like @option{-fstack-protector} except that all functions are protected.
11167 @item -fstack-protector-strong
11168 @opindex fstack-protector-strong
11169 Like @option{-fstack-protector} but includes additional functions to
11170 be protected --- those that have local array definitions, or have
11171 references to local frame addresses.
11173 @item -fstack-protector-explicit
11174 @opindex fstack-protector-explicit
11175 Like @option{-fstack-protector} but only protects those functions which
11176 have the @code{stack_protect} attribute.
11178 @item -fstack-check
11179 @opindex fstack-check
11180 Generate code to verify that you do not go beyond the boundary of the
11181 stack. You should specify this flag if you are running in an
11182 environment with multiple threads, but you only rarely need to specify it in
11183 a single-threaded environment since stack overflow is automatically
11184 detected on nearly all systems if there is only one stack.
11186 Note that this switch does not actually cause checking to be done; the
11187 operating system or the language runtime must do that. The switch causes
11188 generation of code to ensure that they see the stack being extended.
11190 You can additionally specify a string parameter: @samp{no} means no
11191 checking, @samp{generic} means force the use of old-style checking,
11192 @samp{specific} means use the best checking method and is equivalent
11193 to bare @option{-fstack-check}.
11195 Old-style checking is a generic mechanism that requires no specific
11196 target support in the compiler but comes with the following drawbacks:
11200 Modified allocation strategy for large objects: they are always
11201 allocated dynamically if their size exceeds a fixed threshold.
11204 Fixed limit on the size of the static frame of functions: when it is
11205 topped by a particular function, stack checking is not reliable and
11206 a warning is issued by the compiler.
11209 Inefficiency: because of both the modified allocation strategy and the
11210 generic implementation, code performance is hampered.
11213 Note that old-style stack checking is also the fallback method for
11214 @samp{specific} if no target support has been added in the compiler.
11216 @item -fstack-limit-register=@var{reg}
11217 @itemx -fstack-limit-symbol=@var{sym}
11218 @itemx -fno-stack-limit
11219 @opindex fstack-limit-register
11220 @opindex fstack-limit-symbol
11221 @opindex fno-stack-limit
11222 Generate code to ensure that the stack does not grow beyond a certain value,
11223 either the value of a register or the address of a symbol. If a larger
11224 stack is required, a signal is raised at run time. For most targets,
11225 the signal is raised before the stack overruns the boundary, so
11226 it is possible to catch the signal without taking special precautions.
11228 For instance, if the stack starts at absolute address @samp{0x80000000}
11229 and grows downwards, you can use the flags
11230 @option{-fstack-limit-symbol=__stack_limit} and
11231 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11232 of 128KB@. Note that this may only work with the GNU linker.
11234 You can locally override stack limit checking by using the
11235 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11237 @item -fsplit-stack
11238 @opindex fsplit-stack
11239 Generate code to automatically split the stack before it overflows.
11240 The resulting program has a discontiguous stack which can only
11241 overflow if the program is unable to allocate any more memory. This
11242 is most useful when running threaded programs, as it is no longer
11243 necessary to calculate a good stack size to use for each thread. This
11244 is currently only implemented for the x86 targets running
11247 When code compiled with @option{-fsplit-stack} calls code compiled
11248 without @option{-fsplit-stack}, there may not be much stack space
11249 available for the latter code to run. If compiling all code,
11250 including library code, with @option{-fsplit-stack} is not an option,
11251 then the linker can fix up these calls so that the code compiled
11252 without @option{-fsplit-stack} always has a large stack. Support for
11253 this is implemented in the gold linker in GNU binutils release 2.21
11256 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11257 @opindex fvtable-verify
11258 This option is only available when compiling C++ code.
11259 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11260 feature that verifies at run time, for every virtual call, that
11261 the vtable pointer through which the call is made is valid for the type of
11262 the object, and has not been corrupted or overwritten. If an invalid vtable
11263 pointer is detected at run time, an error is reported and execution of the
11264 program is immediately halted.
11266 This option causes run-time data structures to be built at program startup,
11267 which are used for verifying the vtable pointers.
11268 The options @samp{std} and @samp{preinit}
11269 control the timing of when these data structures are built. In both cases the
11270 data structures are built before execution reaches @code{main}. Using
11271 @option{-fvtable-verify=std} causes the data structures to be built after
11272 shared libraries have been loaded and initialized.
11273 @option{-fvtable-verify=preinit} causes them to be built before shared
11274 libraries have been loaded and initialized.
11276 If this option appears multiple times in the command line with different
11277 values specified, @samp{none} takes highest priority over both @samp{std} and
11278 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11281 @opindex fvtv-debug
11282 When used in conjunction with @option{-fvtable-verify=std} or
11283 @option{-fvtable-verify=preinit}, causes debug versions of the
11284 runtime functions for the vtable verification feature to be called.
11285 This flag also causes the compiler to log information about which
11286 vtable pointers it finds for each class.
11287 This information is written to a file named @file{vtv_set_ptr_data.log}
11288 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11289 if that is defined or the current working directory otherwise.
11291 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11292 file, be sure to delete any existing one.
11295 @opindex fvtv-counts
11296 This is a debugging flag. When used in conjunction with
11297 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11298 causes the compiler to keep track of the total number of virtual calls
11299 it encounters and the number of verifications it inserts. It also
11300 counts the number of calls to certain run-time library functions
11301 that it inserts and logs this information for each compilation unit.
11302 The compiler writes this information to a file named
11303 @file{vtv_count_data.log} in the directory named by the environment
11304 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11305 directory otherwise. It also counts the size of the vtable pointer sets
11306 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11307 in the same directory.
11309 Note: This feature @emph{appends} data to the log files. To get fresh log
11310 files, be sure to delete any existing ones.
11312 @item -finstrument-functions
11313 @opindex finstrument-functions
11314 Generate instrumentation calls for entry and exit to functions. Just
11315 after function entry and just before function exit, the following
11316 profiling functions are called with the address of the current
11317 function and its call site. (On some platforms,
11318 @code{__builtin_return_address} does not work beyond the current
11319 function, so the call site information may not be available to the
11320 profiling functions otherwise.)
11323 void __cyg_profile_func_enter (void *this_fn,
11325 void __cyg_profile_func_exit (void *this_fn,
11329 The first argument is the address of the start of the current function,
11330 which may be looked up exactly in the symbol table.
11332 This instrumentation is also done for functions expanded inline in other
11333 functions. The profiling calls indicate where, conceptually, the
11334 inline function is entered and exited. This means that addressable
11335 versions of such functions must be available. If all your uses of a
11336 function are expanded inline, this may mean an additional expansion of
11337 code size. If you use @code{extern inline} in your C code, an
11338 addressable version of such functions must be provided. (This is
11339 normally the case anyway, but if you get lucky and the optimizer always
11340 expands the functions inline, you might have gotten away without
11341 providing static copies.)
11343 A function may be given the attribute @code{no_instrument_function}, in
11344 which case this instrumentation is not done. This can be used, for
11345 example, for the profiling functions listed above, high-priority
11346 interrupt routines, and any functions from which the profiling functions
11347 cannot safely be called (perhaps signal handlers, if the profiling
11348 routines generate output or allocate memory).
11350 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11351 @opindex finstrument-functions-exclude-file-list
11353 Set the list of functions that are excluded from instrumentation (see
11354 the description of @option{-finstrument-functions}). If the file that
11355 contains a function definition matches with one of @var{file}, then
11356 that function is not instrumented. The match is done on substrings:
11357 if the @var{file} parameter is a substring of the file name, it is
11358 considered to be a match.
11363 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11367 excludes any inline function defined in files whose pathnames
11368 contain @file{/bits/stl} or @file{include/sys}.
11370 If, for some reason, you want to include letter @samp{,} in one of
11371 @var{sym}, write @samp{\,}. For example,
11372 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11373 (note the single quote surrounding the option).
11375 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11376 @opindex finstrument-functions-exclude-function-list
11378 This is similar to @option{-finstrument-functions-exclude-file-list},
11379 but this option sets the list of function names to be excluded from
11380 instrumentation. The function name to be matched is its user-visible
11381 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11382 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11383 match is done on substrings: if the @var{sym} parameter is a substring
11384 of the function name, it is considered to be a match. For C99 and C++
11385 extended identifiers, the function name must be given in UTF-8, not
11386 using universal character names.
11391 @node Preprocessor Options
11392 @section Options Controlling the Preprocessor
11393 @cindex preprocessor options
11394 @cindex options, preprocessor
11396 These options control the C preprocessor, which is run on each C source
11397 file before actual compilation.
11399 If you use the @option{-E} option, nothing is done except preprocessing.
11400 Some of these options make sense only together with @option{-E} because
11401 they cause the preprocessor output to be unsuitable for actual
11404 In addition to the options listed here, there are a number of options
11405 to control search paths for include files documented in
11406 @ref{Directory Options}.
11407 Options to control preprocessor diagnostics are listed in
11408 @ref{Warning Options}.
11411 @include cppopts.texi
11413 @item -Wp,@var{option}
11415 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11416 and pass @var{option} directly through to the preprocessor. If
11417 @var{option} contains commas, it is split into multiple options at the
11418 commas. However, many options are modified, translated or interpreted
11419 by the compiler driver before being passed to the preprocessor, and
11420 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11421 interface is undocumented and subject to change, so whenever possible
11422 you should avoid using @option{-Wp} and let the driver handle the
11425 @item -Xpreprocessor @var{option}
11426 @opindex Xpreprocessor
11427 Pass @var{option} as an option to the preprocessor. You can use this to
11428 supply system-specific preprocessor options that GCC does not
11431 If you want to pass an option that takes an argument, you must use
11432 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11434 @item -no-integrated-cpp
11435 @opindex no-integrated-cpp
11436 Perform preprocessing as a separate pass before compilation.
11437 By default, GCC performs preprocessing as an integrated part of
11438 input tokenization and parsing.
11439 If this option is provided, the appropriate language front end
11440 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11441 and Objective-C, respectively) is instead invoked twice,
11442 once for preprocessing only and once for actual compilation
11443 of the preprocessed input.
11444 This option may be useful in conjunction with the @option{-B} or
11445 @option{-wrapper} options to specify an alternate preprocessor or
11446 perform additional processing of the program source between
11447 normal preprocessing and compilation.
11451 @node Assembler Options
11452 @section Passing Options to the Assembler
11454 @c prevent bad page break with this line
11455 You can pass options to the assembler.
11458 @item -Wa,@var{option}
11460 Pass @var{option} as an option to the assembler. If @var{option}
11461 contains commas, it is split into multiple options at the commas.
11463 @item -Xassembler @var{option}
11464 @opindex Xassembler
11465 Pass @var{option} as an option to the assembler. You can use this to
11466 supply system-specific assembler options that GCC does not
11469 If you want to pass an option that takes an argument, you must use
11470 @option{-Xassembler} twice, once for the option and once for the argument.
11475 @section Options for Linking
11476 @cindex link options
11477 @cindex options, linking
11479 These options come into play when the compiler links object files into
11480 an executable output file. They are meaningless if the compiler is
11481 not doing a link step.
11485 @item @var{object-file-name}
11486 A file name that does not end in a special recognized suffix is
11487 considered to name an object file or library. (Object files are
11488 distinguished from libraries by the linker according to the file
11489 contents.) If linking is done, these object files are used as input
11498 If any of these options is used, then the linker is not run, and
11499 object file names should not be used as arguments. @xref{Overall
11503 @opindex fuse-ld=bfd
11504 Use the @command{bfd} linker instead of the default linker.
11506 @item -fuse-ld=gold
11507 @opindex fuse-ld=gold
11508 Use the @command{gold} linker instead of the default linker.
11511 @item -l@var{library}
11512 @itemx -l @var{library}
11514 Search the library named @var{library} when linking. (The second
11515 alternative with the library as a separate argument is only for
11516 POSIX compliance and is not recommended.)
11518 It makes a difference where in the command you write this option; the
11519 linker searches and processes libraries and object files in the order they
11520 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11521 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11522 to functions in @samp{z}, those functions may not be loaded.
11524 The linker searches a standard list of directories for the library,
11525 which is actually a file named @file{lib@var{library}.a}. The linker
11526 then uses this file as if it had been specified precisely by name.
11528 The directories searched include several standard system directories
11529 plus any that you specify with @option{-L}.
11531 Normally the files found this way are library files---archive files
11532 whose members are object files. The linker handles an archive file by
11533 scanning through it for members which define symbols that have so far
11534 been referenced but not defined. But if the file that is found is an
11535 ordinary object file, it is linked in the usual fashion. The only
11536 difference between using an @option{-l} option and specifying a file name
11537 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11538 and searches several directories.
11542 You need this special case of the @option{-l} option in order to
11543 link an Objective-C or Objective-C++ program.
11545 @item -nostartfiles
11546 @opindex nostartfiles
11547 Do not use the standard system startup files when linking.
11548 The standard system libraries are used normally, unless @option{-nostdlib}
11549 or @option{-nodefaultlibs} is used.
11551 @item -nodefaultlibs
11552 @opindex nodefaultlibs
11553 Do not use the standard system libraries when linking.
11554 Only the libraries you specify are passed to the linker, and options
11555 specifying linkage of the system libraries, such as @option{-static-libgcc}
11556 or @option{-shared-libgcc}, are ignored.
11557 The standard startup files are used normally, unless @option{-nostartfiles}
11560 The compiler may generate calls to @code{memcmp},
11561 @code{memset}, @code{memcpy} and @code{memmove}.
11562 These entries are usually resolved by entries in
11563 libc. These entry points should be supplied through some other
11564 mechanism when this option is specified.
11568 Do not use the standard system startup files or libraries when linking.
11569 No startup files and only the libraries you specify are passed to
11570 the linker, and options specifying linkage of the system libraries, such as
11571 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11573 The compiler may generate calls to @code{memcmp}, @code{memset},
11574 @code{memcpy} and @code{memmove}.
11575 These entries are usually resolved by entries in
11576 libc. These entry points should be supplied through some other
11577 mechanism when this option is specified.
11579 @cindex @option{-lgcc}, use with @option{-nostdlib}
11580 @cindex @option{-nostdlib} and unresolved references
11581 @cindex unresolved references and @option{-nostdlib}
11582 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11583 @cindex @option{-nodefaultlibs} and unresolved references
11584 @cindex unresolved references and @option{-nodefaultlibs}
11585 One of the standard libraries bypassed by @option{-nostdlib} and
11586 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11587 which GCC uses to overcome shortcomings of particular machines, or special
11588 needs for some languages.
11589 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11590 Collection (GCC) Internals},
11591 for more discussion of @file{libgcc.a}.)
11592 In most cases, you need @file{libgcc.a} even when you want to avoid
11593 other standard libraries. In other words, when you specify @option{-nostdlib}
11594 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11595 This ensures that you have no unresolved references to internal GCC
11596 library subroutines.
11597 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11598 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11599 GNU Compiler Collection (GCC) Internals}.)
11603 Produce a position independent executable on targets that support it.
11604 For predictable results, you must also specify the same set of options
11605 used for compilation (@option{-fpie}, @option{-fPIE},
11606 or model suboptions) when you specify this linker option.
11610 Don't produce a position independent executable.
11614 Link with the POSIX threads library. This option is supported on
11615 GNU/Linux targets, most other Unix derivatives, and also on
11616 x86 Cygwin and MinGW targets. On some targets this option also sets
11617 flags for the preprocessor, so it should be used consistently for both
11618 compilation and linking.
11622 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11623 that support it. This instructs the linker to add all symbols, not
11624 only used ones, to the dynamic symbol table. This option is needed
11625 for some uses of @code{dlopen} or to allow obtaining backtraces
11626 from within a program.
11630 Remove all symbol table and relocation information from the executable.
11634 On systems that support dynamic linking, this prevents linking with the shared
11635 libraries. On other systems, this option has no effect.
11639 Produce a shared object which can then be linked with other objects to
11640 form an executable. Not all systems support this option. For predictable
11641 results, you must also specify the same set of options used for compilation
11642 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11643 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11644 needs to build supplementary stub code for constructors to work. On
11645 multi-libbed systems, @samp{gcc -shared} must select the correct support
11646 libraries to link against. Failing to supply the correct flags may lead
11647 to subtle defects. Supplying them in cases where they are not necessary
11650 @item -shared-libgcc
11651 @itemx -static-libgcc
11652 @opindex shared-libgcc
11653 @opindex static-libgcc
11654 On systems that provide @file{libgcc} as a shared library, these options
11655 force the use of either the shared or static version, respectively.
11656 If no shared version of @file{libgcc} was built when the compiler was
11657 configured, these options have no effect.
11659 There are several situations in which an application should use the
11660 shared @file{libgcc} instead of the static version. The most common
11661 of these is when the application wishes to throw and catch exceptions
11662 across different shared libraries. In that case, each of the libraries
11663 as well as the application itself should use the shared @file{libgcc}.
11665 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
11666 whenever you build a shared library or a main executable, because C++
11667 programs typically use exceptions, so this is the right thing to do.
11669 If, instead, you use the GCC driver to create shared libraries, you may
11670 find that they are not always linked with the shared @file{libgcc}.
11671 If GCC finds, at its configuration time, that you have a non-GNU linker
11672 or a GNU linker that does not support option @option{--eh-frame-hdr},
11673 it links the shared version of @file{libgcc} into shared libraries
11674 by default. Otherwise, it takes advantage of the linker and optimizes
11675 away the linking with the shared version of @file{libgcc}, linking with
11676 the static version of libgcc by default. This allows exceptions to
11677 propagate through such shared libraries, without incurring relocation
11678 costs at library load time.
11680 However, if a library or main executable is supposed to throw or catch
11681 exceptions, you must link it using the G++ driver, as appropriate
11682 for the languages used in the program, or using the option
11683 @option{-shared-libgcc}, such that it is linked with the shared
11686 @item -static-libasan
11687 @opindex static-libasan
11688 When the @option{-fsanitize=address} option is used to link a program,
11689 the GCC driver automatically links against @option{libasan}. If
11690 @file{libasan} is available as a shared library, and the @option{-static}
11691 option is not used, then this links against the shared version of
11692 @file{libasan}. The @option{-static-libasan} option directs the GCC
11693 driver to link @file{libasan} statically, without necessarily linking
11694 other libraries statically.
11696 @item -static-libtsan
11697 @opindex static-libtsan
11698 When the @option{-fsanitize=thread} option is used to link a program,
11699 the GCC driver automatically links against @option{libtsan}. If
11700 @file{libtsan} is available as a shared library, and the @option{-static}
11701 option is not used, then this links against the shared version of
11702 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11703 driver to link @file{libtsan} statically, without necessarily linking
11704 other libraries statically.
11706 @item -static-liblsan
11707 @opindex static-liblsan
11708 When the @option{-fsanitize=leak} option is used to link a program,
11709 the GCC driver automatically links against @option{liblsan}. If
11710 @file{liblsan} is available as a shared library, and the @option{-static}
11711 option is not used, then this links against the shared version of
11712 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11713 driver to link @file{liblsan} statically, without necessarily linking
11714 other libraries statically.
11716 @item -static-libubsan
11717 @opindex static-libubsan
11718 When the @option{-fsanitize=undefined} option is used to link a program,
11719 the GCC driver automatically links against @option{libubsan}. If
11720 @file{libubsan} is available as a shared library, and the @option{-static}
11721 option is not used, then this links against the shared version of
11722 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11723 driver to link @file{libubsan} statically, without necessarily linking
11724 other libraries statically.
11726 @item -static-libmpx
11727 @opindex static-libmpx
11728 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11729 used to link a program, the GCC driver automatically links against
11730 @file{libmpx}. If @file{libmpx} is available as a shared library,
11731 and the @option{-static} option is not used, then this links against
11732 the shared version of @file{libmpx}. The @option{-static-libmpx}
11733 option directs the GCC driver to link @file{libmpx} statically,
11734 without necessarily linking other libraries statically.
11736 @item -static-libmpxwrappers
11737 @opindex static-libmpxwrappers
11738 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11739 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11740 GCC driver automatically links against @file{libmpxwrappers}. If
11741 @file{libmpxwrappers} is available as a shared library, and the
11742 @option{-static} option is not used, then this links against the shared
11743 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11744 option directs the GCC driver to link @file{libmpxwrappers} statically,
11745 without necessarily linking other libraries statically.
11747 @item -static-libstdc++
11748 @opindex static-libstdc++
11749 When the @command{g++} program is used to link a C++ program, it
11750 normally automatically links against @option{libstdc++}. If
11751 @file{libstdc++} is available as a shared library, and the
11752 @option{-static} option is not used, then this links against the
11753 shared version of @file{libstdc++}. That is normally fine. However, it
11754 is sometimes useful to freeze the version of @file{libstdc++} used by
11755 the program without going all the way to a fully static link. The
11756 @option{-static-libstdc++} option directs the @command{g++} driver to
11757 link @file{libstdc++} statically, without necessarily linking other
11758 libraries statically.
11762 Bind references to global symbols when building a shared object. Warn
11763 about any unresolved references (unless overridden by the link editor
11764 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11767 @item -T @var{script}
11769 @cindex linker script
11770 Use @var{script} as the linker script. This option is supported by most
11771 systems using the GNU linker. On some targets, such as bare-board
11772 targets without an operating system, the @option{-T} option may be required
11773 when linking to avoid references to undefined symbols.
11775 @item -Xlinker @var{option}
11777 Pass @var{option} as an option to the linker. You can use this to
11778 supply system-specific linker options that GCC does not recognize.
11780 If you want to pass an option that takes a separate argument, you must use
11781 @option{-Xlinker} twice, once for the option and once for the argument.
11782 For example, to pass @option{-assert definitions}, you must write
11783 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11784 @option{-Xlinker "-assert definitions"}, because this passes the entire
11785 string as a single argument, which is not what the linker expects.
11787 When using the GNU linker, it is usually more convenient to pass
11788 arguments to linker options using the @option{@var{option}=@var{value}}
11789 syntax than as separate arguments. For example, you can specify
11790 @option{-Xlinker -Map=output.map} rather than
11791 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11792 this syntax for command-line options.
11794 @item -Wl,@var{option}
11796 Pass @var{option} as an option to the linker. If @var{option} contains
11797 commas, it is split into multiple options at the commas. You can use this
11798 syntax to pass an argument to the option.
11799 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11800 linker. When using the GNU linker, you can also get the same effect with
11801 @option{-Wl,-Map=output.map}.
11803 @item -u @var{symbol}
11805 Pretend the symbol @var{symbol} is undefined, to force linking of
11806 library modules to define it. You can use @option{-u} multiple times with
11807 different symbols to force loading of additional library modules.
11809 @item -z @var{keyword}
11811 @option{-z} is passed directly on to the linker along with the keyword
11812 @var{keyword}. See the section in the documentation of your linker for
11813 permitted values and their meanings.
11816 @node Directory Options
11817 @section Options for Directory Search
11818 @cindex directory options
11819 @cindex options, directory search
11820 @cindex search path
11822 These options specify directories to search for header files, for
11823 libraries and for parts of the compiler:
11826 @include cppdiropts.texi
11828 @item -iplugindir=@var{dir}
11829 @opindex iplugindir=
11830 Set the directory to search for plugins that are passed
11831 by @option{-fplugin=@var{name}} instead of
11832 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11833 to be used by the user, but only passed by the driver.
11837 Add directory @var{dir} to the list of directories to be searched
11840 @item -B@var{prefix}
11842 This option specifies where to find the executables, libraries,
11843 include files, and data files of the compiler itself.
11845 The compiler driver program runs one or more of the subprograms
11846 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11847 @var{prefix} as a prefix for each program it tries to run, both with and
11848 without @samp{@var{machine}/@var{version}/} for the corresponding target
11849 machine and compiler version.
11851 For each subprogram to be run, the compiler driver first tries the
11852 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11853 is not specified, the driver tries two standard prefixes,
11854 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11855 those results in a file name that is found, the unmodified program
11856 name is searched for using the directories specified in your
11857 @env{PATH} environment variable.
11859 The compiler checks to see if the path provided by @option{-B}
11860 refers to a directory, and if necessary it adds a directory
11861 separator character at the end of the path.
11863 @option{-B} prefixes that effectively specify directory names also apply
11864 to libraries in the linker, because the compiler translates these
11865 options into @option{-L} options for the linker. They also apply to
11866 include files in the preprocessor, because the compiler translates these
11867 options into @option{-isystem} options for the preprocessor. In this case,
11868 the compiler appends @samp{include} to the prefix.
11870 The runtime support file @file{libgcc.a} can also be searched for using
11871 the @option{-B} prefix, if needed. If it is not found there, the two
11872 standard prefixes above are tried, and that is all. The file is left
11873 out of the link if it is not found by those means.
11875 Another way to specify a prefix much like the @option{-B} prefix is to use
11876 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11879 As a special kludge, if the path provided by @option{-B} is
11880 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11881 9, then it is replaced by @file{[dir/]include}. This is to help
11882 with boot-strapping the compiler.
11884 @item -no-canonical-prefixes
11885 @opindex no-canonical-prefixes
11886 Do not expand any symbolic links, resolve references to @samp{/../}
11887 or @samp{/./}, or make the path absolute when generating a relative
11890 @item --sysroot=@var{dir}
11892 Use @var{dir} as the logical root directory for headers and libraries.
11893 For example, if the compiler normally searches for headers in
11894 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11895 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11897 If you use both this option and the @option{-isysroot} option, then
11898 the @option{--sysroot} option applies to libraries, but the
11899 @option{-isysroot} option applies to header files.
11901 The GNU linker (beginning with version 2.16) has the necessary support
11902 for this option. If your linker does not support this option, the
11903 header file aspect of @option{--sysroot} still works, but the
11904 library aspect does not.
11906 @item --no-sysroot-suffix
11907 @opindex no-sysroot-suffix
11908 For some targets, a suffix is added to the root directory specified
11909 with @option{--sysroot}, depending on the other options used, so that
11910 headers may for example be found in
11911 @file{@var{dir}/@var{suffix}/usr/include} instead of
11912 @file{@var{dir}/usr/include}. This option disables the addition of
11917 @node Code Gen Options
11918 @section Options for Code Generation Conventions
11919 @cindex code generation conventions
11920 @cindex options, code generation
11921 @cindex run-time options
11923 These machine-independent options control the interface conventions
11924 used in code generation.
11926 Most of them have both positive and negative forms; the negative form
11927 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
11928 one of the forms is listed---the one that is not the default. You
11929 can figure out the other form by either removing @samp{no-} or adding
11933 @item -fstack-reuse=@var{reuse-level}
11934 @opindex fstack_reuse
11935 This option controls stack space reuse for user declared local/auto variables
11936 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
11937 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
11938 local variables and temporaries, @samp{named_vars} enables the reuse only for
11939 user defined local variables with names, and @samp{none} disables stack reuse
11940 completely. The default value is @samp{all}. The option is needed when the
11941 program extends the lifetime of a scoped local variable or a compiler generated
11942 temporary beyond the end point defined by the language. When a lifetime of
11943 a variable ends, and if the variable lives in memory, the optimizing compiler
11944 has the freedom to reuse its stack space with other temporaries or scoped
11945 local variables whose live range does not overlap with it. Legacy code extending
11946 local lifetime is likely to break with the stack reuse optimization.
11965 if (*p == 10) // out of scope use of local1
11976 A(int k) : i(k), j(k) @{ @}
11983 void foo(const A& ar)
11990 foo(A(10)); // temp object's lifetime ends when foo returns
11996 ap->i+= 10; // ap references out of scope temp whose space
11997 // is reused with a. What is the value of ap->i?
12002 The lifetime of a compiler generated temporary is well defined by the C++
12003 standard. When a lifetime of a temporary ends, and if the temporary lives
12004 in memory, the optimizing compiler has the freedom to reuse its stack
12005 space with other temporaries or scoped local variables whose live range
12006 does not overlap with it. However some of the legacy code relies on
12007 the behavior of older compilers in which temporaries' stack space is
12008 not reused, the aggressive stack reuse can lead to runtime errors. This
12009 option is used to control the temporary stack reuse optimization.
12013 This option generates traps for signed overflow on addition, subtraction,
12014 multiplication operations.
12015 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12016 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12017 @option{-fwrapv} being effective. Note that only active options override, so
12018 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12019 results in @option{-ftrapv} being effective.
12023 This option instructs the compiler to assume that signed arithmetic
12024 overflow of addition, subtraction and multiplication wraps around
12025 using twos-complement representation. This flag enables some optimizations
12026 and disables others.
12027 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12028 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12029 @option{-fwrapv} being effective. Note that only active options override, so
12030 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12031 results in @option{-ftrapv} being effective.
12034 @opindex fexceptions
12035 Enable exception handling. Generates extra code needed to propagate
12036 exceptions. For some targets, this implies GCC generates frame
12037 unwind information for all functions, which can produce significant data
12038 size overhead, although it does not affect execution. If you do not
12039 specify this option, GCC enables it by default for languages like
12040 C++ that normally require exception handling, and disables it for
12041 languages like C that do not normally require it. However, you may need
12042 to enable this option when compiling C code that needs to interoperate
12043 properly with exception handlers written in C++. You may also wish to
12044 disable this option if you are compiling older C++ programs that don't
12045 use exception handling.
12047 @item -fnon-call-exceptions
12048 @opindex fnon-call-exceptions
12049 Generate code that allows trapping instructions to throw exceptions.
12050 Note that this requires platform-specific runtime support that does
12051 not exist everywhere. Moreover, it only allows @emph{trapping}
12052 instructions to throw exceptions, i.e.@: memory references or floating-point
12053 instructions. It does not allow exceptions to be thrown from
12054 arbitrary signal handlers such as @code{SIGALRM}.
12056 @item -fdelete-dead-exceptions
12057 @opindex fdelete-dead-exceptions
12058 Consider that instructions that may throw exceptions but don't otherwise
12059 contribute to the execution of the program can be optimized away.
12060 This option is enabled by default for the Ada front end, as permitted by
12061 the Ada language specification.
12062 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12064 @item -funwind-tables
12065 @opindex funwind-tables
12066 Similar to @option{-fexceptions}, except that it just generates any needed
12067 static data, but does not affect the generated code in any other way.
12068 You normally do not need to enable this option; instead, a language processor
12069 that needs this handling enables it on your behalf.
12071 @item -fasynchronous-unwind-tables
12072 @opindex fasynchronous-unwind-tables
12073 Generate unwind table in DWARF format, if supported by target machine. The
12074 table is exact at each instruction boundary, so it can be used for stack
12075 unwinding from asynchronous events (such as debugger or garbage collector).
12077 @item -fno-gnu-unique
12078 @opindex fno-gnu-unique
12079 On systems with recent GNU assembler and C library, the C++ compiler
12080 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12081 of template static data members and static local variables in inline
12082 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12083 is necessary to avoid problems with a library used by two different
12084 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12085 therefore disagreeing with the other one about the binding of the
12086 symbol. But this causes @code{dlclose} to be ignored for affected
12087 DSOs; if your program relies on reinitialization of a DSO via
12088 @code{dlclose} and @code{dlopen}, you can use
12089 @option{-fno-gnu-unique}.
12091 @item -fpcc-struct-return
12092 @opindex fpcc-struct-return
12093 Return ``short'' @code{struct} and @code{union} values in memory like
12094 longer ones, rather than in registers. This convention is less
12095 efficient, but it has the advantage of allowing intercallability between
12096 GCC-compiled files and files compiled with other compilers, particularly
12097 the Portable C Compiler (pcc).
12099 The precise convention for returning structures in memory depends
12100 on the target configuration macros.
12102 Short structures and unions are those whose size and alignment match
12103 that of some integer type.
12105 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12106 switch is not binary compatible with code compiled with the
12107 @option{-freg-struct-return} switch.
12108 Use it to conform to a non-default application binary interface.
12110 @item -freg-struct-return
12111 @opindex freg-struct-return
12112 Return @code{struct} and @code{union} values in registers when possible.
12113 This is more efficient for small structures than
12114 @option{-fpcc-struct-return}.
12116 If you specify neither @option{-fpcc-struct-return} nor
12117 @option{-freg-struct-return}, GCC defaults to whichever convention is
12118 standard for the target. If there is no standard convention, GCC
12119 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12120 the principal compiler. In those cases, we can choose the standard, and
12121 we chose the more efficient register return alternative.
12123 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12124 switch is not binary compatible with code compiled with the
12125 @option{-fpcc-struct-return} switch.
12126 Use it to conform to a non-default application binary interface.
12128 @item -fshort-enums
12129 @opindex fshort-enums
12130 Allocate to an @code{enum} type only as many bytes as it needs for the
12131 declared range of possible values. Specifically, the @code{enum} type
12132 is equivalent to the smallest integer type that has enough room.
12134 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12135 code that is not binary compatible with code generated without that switch.
12136 Use it to conform to a non-default application binary interface.
12138 @item -fshort-wchar
12139 @opindex fshort-wchar
12140 Override the underlying type for @code{wchar_t} to be @code{short
12141 unsigned int} instead of the default for the target. This option is
12142 useful for building programs to run under WINE@.
12144 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12145 code that is not binary compatible with code generated without that switch.
12146 Use it to conform to a non-default application binary interface.
12149 @opindex fno-common
12150 @cindex tentative definitions
12151 In C code, this option controls the placement of global variables
12152 defined without an initializer, known as @dfn{tentative definitions}
12153 in the C standard. Tentative definitions are distinct from declarations
12154 of a variable with the @code{extern} keyword, which do not allocate storage.
12156 Unix C compilers have traditionally allocated storage for
12157 uninitialized global variables in a common block. This allows the
12158 linker to resolve all tentative definitions of the same variable
12159 in different compilation units to the same object, or to a non-tentative
12161 This is the behavior specified by @option{-fcommon}, and is the default for
12162 GCC on most targets.
12163 On the other hand, this behavior is not required by ISO
12164 C, and on some targets may carry a speed or code size penalty on
12165 variable references.
12167 The @option{-fno-common} option specifies that the compiler should instead
12168 place uninitialized global variables in the data section of the object file.
12169 This inhibits the merging of tentative definitions by the linker so
12170 you get a multiple-definition error if the same
12171 variable is defined in more than one compilation unit.
12172 Compiling with @option{-fno-common} is useful on targets for which
12173 it provides better performance, or if you wish to verify that the
12174 program will work on other systems that always treat uninitialized
12175 variable definitions this way.
12179 Ignore the @code{#ident} directive.
12181 @item -finhibit-size-directive
12182 @opindex finhibit-size-directive
12183 Don't output a @code{.size} assembler directive, or anything else that
12184 would cause trouble if the function is split in the middle, and the
12185 two halves are placed at locations far apart in memory. This option is
12186 used when compiling @file{crtstuff.c}; you should not need to use it
12189 @item -fverbose-asm
12190 @opindex fverbose-asm
12191 Put extra commentary information in the generated assembly code to
12192 make it more readable. This option is generally only of use to those
12193 who actually need to read the generated assembly code (perhaps while
12194 debugging the compiler itself).
12196 @option{-fno-verbose-asm}, the default, causes the
12197 extra information to be omitted and is useful when comparing two assembler
12200 The added comments include:
12205 information on the compiler version and command-line options,
12208 the source code lines associated with the assembly instructions,
12209 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12212 hints on which high-level expressions correspond to
12213 the various assembly instruction operands.
12217 For example, given this C source file:
12225 for (i = 0; i < n; i++)
12232 compiling to (x86_64) assembly via @option{-S} and emitting the result
12233 direct to stdout via @option{-o} @option{-}
12236 gcc -S test.c -fverbose-asm -Os -o -
12239 gives output similar to this:
12243 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12250 .type test, @@function
12254 # test.c:4: int total = 0;
12255 xorl %eax, %eax # <retval>
12256 # test.c:6: for (i = 0; i < n; i++)
12257 xorl %edx, %edx # i
12259 # test.c:6: for (i = 0; i < n; i++)
12260 cmpl %edi, %edx # n, i
12262 # test.c:7: total += i * i;
12263 movl %edx, %ecx # i, tmp92
12264 imull %edx, %ecx # i, tmp92
12265 # test.c:6: for (i = 0; i < n; i++)
12267 # test.c:7: total += i * i;
12268 addl %ecx, %eax # tmp92, <retval>
12276 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12277 .section .note.GNU-stack,"",@@progbits
12280 The comments are intended for humans rather than machines and hence the
12281 precise format of the comments is subject to change.
12283 @item -frecord-gcc-switches
12284 @opindex frecord-gcc-switches
12285 This switch causes the command line used to invoke the
12286 compiler to be recorded into the object file that is being created.
12287 This switch is only implemented on some targets and the exact format
12288 of the recording is target and binary file format dependent, but it
12289 usually takes the form of a section containing ASCII text. This
12290 switch is related to the @option{-fverbose-asm} switch, but that
12291 switch only records information in the assembler output file as
12292 comments, so it never reaches the object file.
12293 See also @option{-grecord-gcc-switches} for another
12294 way of storing compiler options into the object file.
12298 @cindex global offset table
12300 Generate position-independent code (PIC) suitable for use in a shared
12301 library, if supported for the target machine. Such code accesses all
12302 constant addresses through a global offset table (GOT)@. The dynamic
12303 loader resolves the GOT entries when the program starts (the dynamic
12304 loader is not part of GCC; it is part of the operating system). If
12305 the GOT size for the linked executable exceeds a machine-specific
12306 maximum size, you get an error message from the linker indicating that
12307 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12308 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12309 on the m68k and RS/6000. The x86 has no such limit.)
12311 Position-independent code requires special support, and therefore works
12312 only on certain machines. For the x86, GCC supports PIC for System V
12313 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12314 position-independent.
12316 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12321 If supported for the target machine, emit position-independent code,
12322 suitable for dynamic linking and avoiding any limit on the size of the
12323 global offset table. This option makes a difference on AArch64, m68k,
12324 PowerPC and SPARC@.
12326 Position-independent code requires special support, and therefore works
12327 only on certain machines.
12329 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12336 These options are similar to @option{-fpic} and @option{-fPIC}, but
12337 generated position independent code can be only linked into executables.
12338 Usually these options are used when @option{-pie} GCC option is
12339 used during linking.
12341 @option{-fpie} and @option{-fPIE} both define the macros
12342 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12343 for @option{-fpie} and 2 for @option{-fPIE}.
12347 Do not use the PLT for external function calls in position-independent code.
12348 Instead, load the callee address at call sites from the GOT and branch to it.
12349 This leads to more efficient code by eliminating PLT stubs and exposing
12350 GOT loads to optimizations. On architectures such as 32-bit x86 where
12351 PLT stubs expect the GOT pointer in a specific register, this gives more
12352 register allocation freedom to the compiler.
12353 Lazy binding requires use of the PLT;
12354 with @option{-fno-plt} all external symbols are resolved at load time.
12356 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12357 through the PLT for specific external functions.
12359 In position-dependent code, a few targets also convert calls to
12360 functions that are marked to not use the PLT to use the GOT instead.
12362 @item -fno-jump-tables
12363 @opindex fno-jump-tables
12364 Do not use jump tables for switch statements even where it would be
12365 more efficient than other code generation strategies. This option is
12366 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12367 building code that forms part of a dynamic linker and cannot
12368 reference the address of a jump table. On some targets, jump tables
12369 do not require a GOT and this option is not needed.
12371 @item -ffixed-@var{reg}
12373 Treat the register named @var{reg} as a fixed register; generated code
12374 should never refer to it (except perhaps as a stack pointer, frame
12375 pointer or in some other fixed role).
12377 @var{reg} must be the name of a register. The register names accepted
12378 are machine-specific and are defined in the @code{REGISTER_NAMES}
12379 macro in the machine description macro file.
12381 This flag does not have a negative form, because it specifies a
12384 @item -fcall-used-@var{reg}
12385 @opindex fcall-used
12386 Treat the register named @var{reg} as an allocable register that is
12387 clobbered by function calls. It may be allocated for temporaries or
12388 variables that do not live across a call. Functions compiled this way
12389 do not save and restore the register @var{reg}.
12391 It is an error to use this flag with the frame pointer or stack pointer.
12392 Use of this flag for other registers that have fixed pervasive roles in
12393 the machine's execution model produces disastrous results.
12395 This flag does not have a negative form, because it specifies a
12398 @item -fcall-saved-@var{reg}
12399 @opindex fcall-saved
12400 Treat the register named @var{reg} as an allocable register saved by
12401 functions. It may be allocated even for temporaries or variables that
12402 live across a call. Functions compiled this way save and restore
12403 the register @var{reg} if they use it.
12405 It is an error to use this flag with the frame pointer or stack pointer.
12406 Use of this flag for other registers that have fixed pervasive roles in
12407 the machine's execution model produces disastrous results.
12409 A different sort of disaster results from the use of this flag for
12410 a register in which function values may be returned.
12412 This flag does not have a negative form, because it specifies a
12415 @item -fpack-struct[=@var{n}]
12416 @opindex fpack-struct
12417 Without a value specified, pack all structure members together without
12418 holes. When a value is specified (which must be a small power of two), pack
12419 structure members according to this value, representing the maximum
12420 alignment (that is, objects with default alignment requirements larger than
12421 this are output potentially unaligned at the next fitting location.
12423 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12424 code that is not binary compatible with code generated without that switch.
12425 Additionally, it makes the code suboptimal.
12426 Use it to conform to a non-default application binary interface.
12428 @item -fleading-underscore
12429 @opindex fleading-underscore
12430 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12431 change the way C symbols are represented in the object file. One use
12432 is to help link with legacy assembly code.
12434 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12435 generate code that is not binary compatible with code generated without that
12436 switch. Use it to conform to a non-default application binary interface.
12437 Not all targets provide complete support for this switch.
12439 @item -ftls-model=@var{model}
12440 @opindex ftls-model
12441 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12442 The @var{model} argument should be one of @samp{global-dynamic},
12443 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12444 Note that the choice is subject to optimization: the compiler may use
12445 a more efficient model for symbols not visible outside of the translation
12446 unit, or if @option{-fpic} is not given on the command line.
12448 The default without @option{-fpic} is @samp{initial-exec}; with
12449 @option{-fpic} the default is @samp{global-dynamic}.
12451 @item -ftrampolines
12452 @opindex ftrampolines
12453 For targets that normally need trampolines for nested functions, always
12454 generate them instead of using descriptors. Otherwise, for targets that
12455 do not need them, like for example HP-PA or IA-64, do nothing.
12457 A trampoline is a small piece of code that is created at run time on the
12458 stack when the address of a nested function is taken, and is used to call
12459 the nested function indirectly. Therefore, it requires the stack to be
12460 made executable in order for the program to work properly.
12462 @option{-fno-trampolines} is enabled by default on a language by language
12463 basis to let the compiler avoid generating them, if it computes that this
12464 is safe, and replace them with descriptors. Descriptors are made up of data
12465 only, but the generated code must be prepared to deal with them. As of this
12466 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12468 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12469 @option{-fno-trampolines} are not binary compatible if nested functions are
12470 present. This option must therefore be used on a program-wide basis and be
12471 manipulated with extreme care.
12473 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12474 @opindex fvisibility
12475 Set the default ELF image symbol visibility to the specified option---all
12476 symbols are marked with this unless overridden within the code.
12477 Using this feature can very substantially improve linking and
12478 load times of shared object libraries, produce more optimized
12479 code, provide near-perfect API export and prevent symbol clashes.
12480 It is @strong{strongly} recommended that you use this in any shared objects
12483 Despite the nomenclature, @samp{default} always means public; i.e.,
12484 available to be linked against from outside the shared object.
12485 @samp{protected} and @samp{internal} are pretty useless in real-world
12486 usage so the only other commonly used option is @samp{hidden}.
12487 The default if @option{-fvisibility} isn't specified is
12488 @samp{default}, i.e., make every symbol public.
12490 A good explanation of the benefits offered by ensuring ELF
12491 symbols have the correct visibility is given by ``How To Write
12492 Shared Libraries'' by Ulrich Drepper (which can be found at
12493 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
12494 solution made possible by this option to marking things hidden when
12495 the default is public is to make the default hidden and mark things
12496 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12497 and @code{__attribute__ ((visibility("default")))} instead of
12498 @code{__declspec(dllexport)} you get almost identical semantics with
12499 identical syntax. This is a great boon to those working with
12500 cross-platform projects.
12502 For those adding visibility support to existing code, you may find
12503 @code{#pragma GCC visibility} of use. This works by you enclosing
12504 the declarations you wish to set visibility for with (for example)
12505 @code{#pragma GCC visibility push(hidden)} and
12506 @code{#pragma GCC visibility pop}.
12507 Bear in mind that symbol visibility should be viewed @strong{as
12508 part of the API interface contract} and thus all new code should
12509 always specify visibility when it is not the default; i.e., declarations
12510 only for use within the local DSO should @strong{always} be marked explicitly
12511 as hidden as so to avoid PLT indirection overheads---making this
12512 abundantly clear also aids readability and self-documentation of the code.
12513 Note that due to ISO C++ specification requirements, @code{operator new} and
12514 @code{operator delete} must always be of default visibility.
12516 Be aware that headers from outside your project, in particular system
12517 headers and headers from any other library you use, may not be
12518 expecting to be compiled with visibility other than the default. You
12519 may need to explicitly say @code{#pragma GCC visibility push(default)}
12520 before including any such headers.
12522 @code{extern} declarations are not affected by @option{-fvisibility}, so
12523 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12524 no modifications. However, this means that calls to @code{extern}
12525 functions with no explicit visibility use the PLT, so it is more
12526 effective to use @code{__attribute ((visibility))} and/or
12527 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12528 declarations should be treated as hidden.
12530 Note that @option{-fvisibility} does affect C++ vague linkage
12531 entities. This means that, for instance, an exception class that is
12532 be thrown between DSOs must be explicitly marked with default
12533 visibility so that the @samp{type_info} nodes are unified between
12536 An overview of these techniques, their benefits and how to use them
12537 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12539 @item -fstrict-volatile-bitfields
12540 @opindex fstrict-volatile-bitfields
12541 This option should be used if accesses to volatile bit-fields (or other
12542 structure fields, although the compiler usually honors those types
12543 anyway) should use a single access of the width of the
12544 field's type, aligned to a natural alignment if possible. For
12545 example, targets with memory-mapped peripheral registers might require
12546 all such accesses to be 16 bits wide; with this flag you can
12547 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12548 is 16 bits on these targets) to force GCC to use 16-bit accesses
12549 instead of, perhaps, a more efficient 32-bit access.
12551 If this option is disabled, the compiler uses the most efficient
12552 instruction. In the previous example, that might be a 32-bit load
12553 instruction, even though that accesses bytes that do not contain
12554 any portion of the bit-field, or memory-mapped registers unrelated to
12555 the one being updated.
12557 In some cases, such as when the @code{packed} attribute is applied to a
12558 structure field, it may not be possible to access the field with a single
12559 read or write that is correctly aligned for the target machine. In this
12560 case GCC falls back to generating multiple accesses rather than code that
12561 will fault or truncate the result at run time.
12563 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12564 not allowed to touch non bit-field members. It is therefore recommended
12565 to define all bits of the field's type as bit-field members.
12567 The default value of this option is determined by the application binary
12568 interface for the target processor.
12570 @item -fsync-libcalls
12571 @opindex fsync-libcalls
12572 This option controls whether any out-of-line instance of the @code{__sync}
12573 family of functions may be used to implement the C++11 @code{__atomic}
12574 family of functions.
12576 The default value of this option is enabled, thus the only useful form
12577 of the option is @option{-fno-sync-libcalls}. This option is used in
12578 the implementation of the @file{libatomic} runtime library.
12582 @node Developer Options
12583 @section GCC Developer Options
12584 @cindex developer options
12585 @cindex debugging GCC
12586 @cindex debug dump options
12587 @cindex dump options
12588 @cindex compilation statistics
12590 This section describes command-line options that are primarily of
12591 interest to GCC developers, including options to support compiler
12592 testing and investigation of compiler bugs and compile-time
12593 performance problems. This includes options that produce debug dumps
12594 at various points in the compilation; that print statistics such as
12595 memory use and execution time; and that print information about GCC's
12596 configuration, such as where it searches for libraries. You should
12597 rarely need to use any of these options for ordinary compilation and
12602 @item -d@var{letters}
12603 @itemx -fdump-rtl-@var{pass}
12604 @itemx -fdump-rtl-@var{pass}=@var{filename}
12606 @opindex fdump-rtl-@var{pass}
12607 Says to make debugging dumps during compilation at times specified by
12608 @var{letters}. This is used for debugging the RTL-based passes of the
12609 compiler. The file names for most of the dumps are made by appending
12610 a pass number and a word to the @var{dumpname}, and the files are
12611 created in the directory of the output file. In case of
12612 @option{=@var{filename}} option, the dump is output on the given file
12613 instead of the pass numbered dump files. Note that the pass number is
12614 assigned as passes are registered into the pass manager. Most passes
12615 are registered in the order that they will execute and for these passes
12616 the number corresponds to the pass execution order. However, passes
12617 registered by plugins, passes specific to compilation targets, or
12618 passes that are otherwise registered after all the other passes are
12619 numbered higher than a pass named "final", even if they are executed
12620 earlier. @var{dumpname} is generated from the name of the output
12621 file if explicitly specified and not an executable, otherwise it is
12622 the basename of the source file.
12624 Some @option{-d@var{letters}} switches have different meaning when
12625 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
12626 for information about preprocessor-specific dump options.
12628 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
12629 @option{-d} option @var{letters}. Here are the possible
12630 letters for use in @var{pass} and @var{letters}, and their meanings:
12634 @item -fdump-rtl-alignments
12635 @opindex fdump-rtl-alignments
12636 Dump after branch alignments have been computed.
12638 @item -fdump-rtl-asmcons
12639 @opindex fdump-rtl-asmcons
12640 Dump after fixing rtl statements that have unsatisfied in/out constraints.
12642 @item -fdump-rtl-auto_inc_dec
12643 @opindex fdump-rtl-auto_inc_dec
12644 Dump after auto-inc-dec discovery. This pass is only run on
12645 architectures that have auto inc or auto dec instructions.
12647 @item -fdump-rtl-barriers
12648 @opindex fdump-rtl-barriers
12649 Dump after cleaning up the barrier instructions.
12651 @item -fdump-rtl-bbpart
12652 @opindex fdump-rtl-bbpart
12653 Dump after partitioning hot and cold basic blocks.
12655 @item -fdump-rtl-bbro
12656 @opindex fdump-rtl-bbro
12657 Dump after block reordering.
12659 @item -fdump-rtl-btl1
12660 @itemx -fdump-rtl-btl2
12661 @opindex fdump-rtl-btl2
12662 @opindex fdump-rtl-btl2
12663 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
12664 after the two branch
12665 target load optimization passes.
12667 @item -fdump-rtl-bypass
12668 @opindex fdump-rtl-bypass
12669 Dump after jump bypassing and control flow optimizations.
12671 @item -fdump-rtl-combine
12672 @opindex fdump-rtl-combine
12673 Dump after the RTL instruction combination pass.
12675 @item -fdump-rtl-compgotos
12676 @opindex fdump-rtl-compgotos
12677 Dump after duplicating the computed gotos.
12679 @item -fdump-rtl-ce1
12680 @itemx -fdump-rtl-ce2
12681 @itemx -fdump-rtl-ce3
12682 @opindex fdump-rtl-ce1
12683 @opindex fdump-rtl-ce2
12684 @opindex fdump-rtl-ce3
12685 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
12686 @option{-fdump-rtl-ce3} enable dumping after the three
12687 if conversion passes.
12689 @item -fdump-rtl-cprop_hardreg
12690 @opindex fdump-rtl-cprop_hardreg
12691 Dump after hard register copy propagation.
12693 @item -fdump-rtl-csa
12694 @opindex fdump-rtl-csa
12695 Dump after combining stack adjustments.
12697 @item -fdump-rtl-cse1
12698 @itemx -fdump-rtl-cse2
12699 @opindex fdump-rtl-cse1
12700 @opindex fdump-rtl-cse2
12701 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
12702 the two common subexpression elimination passes.
12704 @item -fdump-rtl-dce
12705 @opindex fdump-rtl-dce
12706 Dump after the standalone dead code elimination passes.
12708 @item -fdump-rtl-dbr
12709 @opindex fdump-rtl-dbr
12710 Dump after delayed branch scheduling.
12712 @item -fdump-rtl-dce1
12713 @itemx -fdump-rtl-dce2
12714 @opindex fdump-rtl-dce1
12715 @opindex fdump-rtl-dce2
12716 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
12717 the two dead store elimination passes.
12719 @item -fdump-rtl-eh
12720 @opindex fdump-rtl-eh
12721 Dump after finalization of EH handling code.
12723 @item -fdump-rtl-eh_ranges
12724 @opindex fdump-rtl-eh_ranges
12725 Dump after conversion of EH handling range regions.
12727 @item -fdump-rtl-expand
12728 @opindex fdump-rtl-expand
12729 Dump after RTL generation.
12731 @item -fdump-rtl-fwprop1
12732 @itemx -fdump-rtl-fwprop2
12733 @opindex fdump-rtl-fwprop1
12734 @opindex fdump-rtl-fwprop2
12735 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
12736 dumping after the two forward propagation passes.
12738 @item -fdump-rtl-gcse1
12739 @itemx -fdump-rtl-gcse2
12740 @opindex fdump-rtl-gcse1
12741 @opindex fdump-rtl-gcse2
12742 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
12743 after global common subexpression elimination.
12745 @item -fdump-rtl-init-regs
12746 @opindex fdump-rtl-init-regs
12747 Dump after the initialization of the registers.
12749 @item -fdump-rtl-initvals
12750 @opindex fdump-rtl-initvals
12751 Dump after the computation of the initial value sets.
12753 @item -fdump-rtl-into_cfglayout
12754 @opindex fdump-rtl-into_cfglayout
12755 Dump after converting to cfglayout mode.
12757 @item -fdump-rtl-ira
12758 @opindex fdump-rtl-ira
12759 Dump after iterated register allocation.
12761 @item -fdump-rtl-jump
12762 @opindex fdump-rtl-jump
12763 Dump after the second jump optimization.
12765 @item -fdump-rtl-loop2
12766 @opindex fdump-rtl-loop2
12767 @option{-fdump-rtl-loop2} enables dumping after the rtl
12768 loop optimization passes.
12770 @item -fdump-rtl-mach
12771 @opindex fdump-rtl-mach
12772 Dump after performing the machine dependent reorganization pass, if that
12775 @item -fdump-rtl-mode_sw
12776 @opindex fdump-rtl-mode_sw
12777 Dump after removing redundant mode switches.
12779 @item -fdump-rtl-rnreg
12780 @opindex fdump-rtl-rnreg
12781 Dump after register renumbering.
12783 @item -fdump-rtl-outof_cfglayout
12784 @opindex fdump-rtl-outof_cfglayout
12785 Dump after converting from cfglayout mode.
12787 @item -fdump-rtl-peephole2
12788 @opindex fdump-rtl-peephole2
12789 Dump after the peephole pass.
12791 @item -fdump-rtl-postreload
12792 @opindex fdump-rtl-postreload
12793 Dump after post-reload optimizations.
12795 @item -fdump-rtl-pro_and_epilogue
12796 @opindex fdump-rtl-pro_and_epilogue
12797 Dump after generating the function prologues and epilogues.
12799 @item -fdump-rtl-sched1
12800 @itemx -fdump-rtl-sched2
12801 @opindex fdump-rtl-sched1
12802 @opindex fdump-rtl-sched2
12803 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
12804 after the basic block scheduling passes.
12806 @item -fdump-rtl-ree
12807 @opindex fdump-rtl-ree
12808 Dump after sign/zero extension elimination.
12810 @item -fdump-rtl-seqabstr
12811 @opindex fdump-rtl-seqabstr
12812 Dump after common sequence discovery.
12814 @item -fdump-rtl-shorten
12815 @opindex fdump-rtl-shorten
12816 Dump after shortening branches.
12818 @item -fdump-rtl-sibling
12819 @opindex fdump-rtl-sibling
12820 Dump after sibling call optimizations.
12822 @item -fdump-rtl-split1
12823 @itemx -fdump-rtl-split2
12824 @itemx -fdump-rtl-split3
12825 @itemx -fdump-rtl-split4
12826 @itemx -fdump-rtl-split5
12827 @opindex fdump-rtl-split1
12828 @opindex fdump-rtl-split2
12829 @opindex fdump-rtl-split3
12830 @opindex fdump-rtl-split4
12831 @opindex fdump-rtl-split5
12832 These options enable dumping after five rounds of
12833 instruction splitting.
12835 @item -fdump-rtl-sms
12836 @opindex fdump-rtl-sms
12837 Dump after modulo scheduling. This pass is only run on some
12840 @item -fdump-rtl-stack
12841 @opindex fdump-rtl-stack
12842 Dump after conversion from GCC's ``flat register file'' registers to the
12843 x87's stack-like registers. This pass is only run on x86 variants.
12845 @item -fdump-rtl-subreg1
12846 @itemx -fdump-rtl-subreg2
12847 @opindex fdump-rtl-subreg1
12848 @opindex fdump-rtl-subreg2
12849 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
12850 the two subreg expansion passes.
12852 @item -fdump-rtl-unshare
12853 @opindex fdump-rtl-unshare
12854 Dump after all rtl has been unshared.
12856 @item -fdump-rtl-vartrack
12857 @opindex fdump-rtl-vartrack
12858 Dump after variable tracking.
12860 @item -fdump-rtl-vregs
12861 @opindex fdump-rtl-vregs
12862 Dump after converting virtual registers to hard registers.
12864 @item -fdump-rtl-web
12865 @opindex fdump-rtl-web
12866 Dump after live range splitting.
12868 @item -fdump-rtl-regclass
12869 @itemx -fdump-rtl-subregs_of_mode_init
12870 @itemx -fdump-rtl-subregs_of_mode_finish
12871 @itemx -fdump-rtl-dfinit
12872 @itemx -fdump-rtl-dfinish
12873 @opindex fdump-rtl-regclass
12874 @opindex fdump-rtl-subregs_of_mode_init
12875 @opindex fdump-rtl-subregs_of_mode_finish
12876 @opindex fdump-rtl-dfinit
12877 @opindex fdump-rtl-dfinish
12878 These dumps are defined but always produce empty files.
12881 @itemx -fdump-rtl-all
12883 @opindex fdump-rtl-all
12884 Produce all the dumps listed above.
12888 Annotate the assembler output with miscellaneous debugging information.
12892 Dump all macro definitions, at the end of preprocessing, in addition to
12897 Produce a core dump whenever an error occurs.
12901 Annotate the assembler output with a comment indicating which
12902 pattern and alternative is used. The length of each instruction is
12907 Dump the RTL in the assembler output as a comment before each instruction.
12908 Also turns on @option{-dp} annotation.
12912 Just generate RTL for a function instead of compiling it. Usually used
12913 with @option{-fdump-rtl-expand}.
12916 @item -fdump-noaddr
12917 @opindex fdump-noaddr
12918 When doing debugging dumps, suppress address output. This makes it more
12919 feasible to use diff on debugging dumps for compiler invocations with
12920 different compiler binaries and/or different
12921 text / bss / data / heap / stack / dso start locations.
12924 @opindex freport-bug
12925 Collect and dump debug information into a temporary file if an
12926 internal compiler error (ICE) occurs.
12928 @item -fdump-unnumbered
12929 @opindex fdump-unnumbered
12930 When doing debugging dumps, suppress instruction numbers and address output.
12931 This makes it more feasible to use diff on debugging dumps for compiler
12932 invocations with different options, in particular with and without
12935 @item -fdump-unnumbered-links
12936 @opindex fdump-unnumbered-links
12937 When doing debugging dumps (see @option{-d} option above), suppress
12938 instruction numbers for the links to the previous and next instructions
12941 @item -fdump-class-hierarchy @r{(C++ only)}
12942 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
12943 @opindex fdump-class-hierarchy
12944 Dump a representation of each class's hierarchy and virtual function
12945 table layout to a file. The file name is made by appending
12946 @file{.class} to the source file name, and the file is created in the
12947 same directory as the output file. If the @samp{-@var{options}} form
12948 is used, @var{options} controls the details of the dump as described
12949 for the @option{-fdump-tree} options.
12951 @item -fdump-ipa-@var{switch}
12953 Control the dumping at various stages of inter-procedural analysis
12954 language tree to a file. The file name is generated by appending a
12955 switch specific suffix to the source file name, and the file is created
12956 in the same directory as the output file. The following dumps are
12961 Enables all inter-procedural analysis dumps.
12964 Dumps information about call-graph optimization, unused function removal,
12965 and inlining decisions.
12968 Dump after function inlining.
12972 @item -fdump-passes
12973 @opindex fdump-passes
12974 Print on @file{stderr} the list of optimization passes that are turned
12975 on and off by the current command-line options.
12977 @item -fdump-statistics-@var{option}
12978 @opindex fdump-statistics
12979 Enable and control dumping of pass statistics in a separate file. The
12980 file name is generated by appending a suffix ending in
12981 @samp{.statistics} to the source file name, and the file is created in
12982 the same directory as the output file. If the @samp{-@var{option}}
12983 form is used, @samp{-stats} causes counters to be summed over the
12984 whole compilation unit while @samp{-details} dumps every event as
12985 the passes generate them. The default with no option is to sum
12986 counters for each function compiled.
12988 @item -fdump-translation-unit @r{(C++ only)}
12989 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
12990 @opindex fdump-translation-unit
12991 Dump a representation of the tree structure for the entire translation
12992 unit to a file. The file name is made by appending @file{.tu} to the
12993 source file name, and the file is created in the same directory as the
12994 output file. If the @samp{-@var{options}} form is used, @var{options}
12995 controls the details of the dump as described for the
12996 @option{-fdump-tree} options.
12998 @item -fdump-tree-all
12999 @itemx -fdump-tree-@var{switch}
13000 @itemx -fdump-tree-@var{switch}-@var{options}
13001 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
13002 @opindex fdump-tree-all
13003 @opindex fdump-tree
13004 Control the dumping at various stages of processing the intermediate
13005 language tree to a file. The file name is generated by appending a
13006 switch-specific suffix to the source file name, and the file is
13007 created in the same directory as the output file. In case of
13008 @option{=@var{filename}} option, the dump is output on the given file
13009 instead of the auto named dump files. If the @samp{-@var{options}}
13010 form is used, @var{options} is a list of @samp{-} separated options
13011 which control the details of the dump. Not all options are applicable
13012 to all dumps; those that are not meaningful are ignored. The
13013 following options are available
13017 Print the address of each node. Usually this is not meaningful as it
13018 changes according to the environment and source file. Its primary use
13019 is for tying up a dump file with a debug environment.
13021 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
13022 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
13023 use working backward from mangled names in the assembly file.
13025 When dumping front-end intermediate representations, inhibit dumping
13026 of members of a scope or body of a function merely because that scope
13027 has been reached. Only dump such items when they are directly reachable
13028 by some other path.
13030 When dumping pretty-printed trees, this option inhibits dumping the
13031 bodies of control structures.
13033 When dumping RTL, print the RTL in slim (condensed) form instead of
13034 the default LISP-like representation.
13036 Print a raw representation of the tree. By default, trees are
13037 pretty-printed into a C-like representation.
13039 Enable more detailed dumps (not honored by every dump option). Also
13040 include information from the optimization passes.
13042 Enable dumping various statistics about the pass (not honored by every dump
13045 Enable showing basic block boundaries (disabled in raw dumps).
13047 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13048 dump a representation of the control flow graph suitable for viewing with
13049 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13050 the file is pretty-printed as a subgraph, so that GraphViz can render them
13051 all in a single plot.
13053 This option currently only works for RTL dumps, and the RTL is always
13054 dumped in slim form.
13056 Enable showing virtual operands for every statement.
13058 Enable showing line numbers for statements.
13060 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13062 Enable showing the tree dump for each statement.
13064 Enable showing the EH region number holding each statement.
13066 Enable showing scalar evolution analysis details.
13068 Enable showing optimization information (only available in certain
13071 Enable showing missed optimization information (only available in certain
13074 Enable other detailed optimization information (only available in
13076 @item =@var{filename}
13077 Instead of an auto named dump file, output into the given file
13078 name. The file names @file{stdout} and @file{stderr} are treated
13079 specially and are considered already open standard streams. For
13083 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13084 -fdump-tree-pre=/dev/stderr file.c
13087 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13088 output on to @file{stderr}. If two conflicting dump filenames are
13089 given for the same pass, then the latter option overrides the earlier
13093 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13094 and @option{lineno}.
13097 Turn on all optimization options, i.e., @option{optimized},
13098 @option{missed}, and @option{note}.
13101 To determine what tree dumps are available or find the dump for a pass
13102 of interest follow the steps below.
13106 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
13107 look for a code that corresponds to the pass you are interested in.
13108 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
13109 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
13110 The number at the end distinguishes distinct invocations of the same pass.
13112 To enable the creation of the dump file, append the pass code to
13113 the @option{-fdump-} option prefix and invoke GCC with it. For example,
13114 to enable the dump from the Early Value Range Propagation pass, invoke
13115 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
13116 specify the name of the dump file. If you don't specify one, GCC
13117 creates as described below.
13119 Find the pass dump in a file whose name is composed of three components
13120 separated by a period: the name of the source file GCC was invoked to
13121 compile, a numeric suffix indicating the pass number followed by the
13122 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
13123 and finally the pass code. For example, the Early VRP pass dump might
13124 be in a file named @file{myfile.c.038t.evrp} in the current working
13125 directory. Note that the numeric codes are not stable and may change
13126 from one version of GCC to another.
13130 @itemx -fopt-info-@var{options}
13131 @itemx -fopt-info-@var{options}=@var{filename}
13133 Controls optimization dumps from various optimization passes. If the
13134 @samp{-@var{options}} form is used, @var{options} is a list of
13135 @samp{-} separated option keywords to select the dump details and
13138 The @var{options} can be divided into two groups: options describing the
13139 verbosity of the dump, and options describing which optimizations
13140 should be included. The options from both the groups can be freely
13141 mixed as they are non-overlapping. However, in case of any conflicts,
13142 the later options override the earlier options on the command
13145 The following options control the dump verbosity:
13149 Print information when an optimization is successfully applied. It is
13150 up to a pass to decide which information is relevant. For example, the
13151 vectorizer passes print the source location of loops which are
13152 successfully vectorized.
13154 Print information about missed optimizations. Individual passes
13155 control which information to include in the output.
13157 Print verbose information about optimizations, such as certain
13158 transformations, more detailed messages about decisions etc.
13160 Print detailed optimization information. This includes
13161 @samp{optimized}, @samp{missed}, and @samp{note}.
13164 One or more of the following option keywords can be used to describe a
13165 group of optimizations:
13169 Enable dumps from all interprocedural optimizations.
13171 Enable dumps from all loop optimizations.
13173 Enable dumps from all inlining optimizations.
13175 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
13177 Enable dumps from all vectorization optimizations.
13179 Enable dumps from all optimizations. This is a superset of
13180 the optimization groups listed above.
13183 If @var{options} is
13184 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13185 info about successful optimizations from all the passes.
13187 If the @var{filename} is provided, then the dumps from all the
13188 applicable optimizations are concatenated into the @var{filename}.
13189 Otherwise the dump is output onto @file{stderr}. Though multiple
13190 @option{-fopt-info} options are accepted, only one of them can include
13191 a @var{filename}. If other filenames are provided then all but the
13192 first such option are ignored.
13194 Note that the output @var{filename} is overwritten
13195 in case of multiple translation units. If a combined output from
13196 multiple translation units is desired, @file{stderr} should be used
13199 In the following example, the optimization info is output to
13208 gcc -O3 -fopt-info-missed=missed.all
13212 outputs missed optimization report from all the passes into
13213 @file{missed.all}, and this one:
13216 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13220 prints information about missed optimization opportunities from
13221 vectorization passes on @file{stderr}.
13222 Note that @option{-fopt-info-vec-missed} is equivalent to
13223 @option{-fopt-info-missed-vec}. The order of the optimization group
13224 names and message types listed after @option{-fopt-info} does not matter.
13226 As another example,
13228 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13232 outputs information about missed optimizations as well as
13233 optimized locations from all the inlining passes into
13239 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13243 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13244 in conflict since only one output file is allowed. In this case, only
13245 the first option takes effect and the subsequent options are
13246 ignored. Thus only @file{vec.miss} is produced which contains
13247 dumps from the vectorizer about missed opportunities.
13249 @item -fsched-verbose=@var{n}
13250 @opindex fsched-verbose
13251 On targets that use instruction scheduling, this option controls the
13252 amount of debugging output the scheduler prints to the dump files.
13254 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13255 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13256 For @var{n} greater than one, it also output basic block probabilities,
13257 detailed ready list information and unit/insn info. For @var{n} greater
13258 than two, it includes RTL at abort point, control-flow and regions info.
13259 And for @var{n} over four, @option{-fsched-verbose} also includes
13264 @item -fenable-@var{kind}-@var{pass}
13265 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13269 This is a set of options that are used to explicitly disable/enable
13270 optimization passes. These options are intended for use for debugging GCC.
13271 Compiler users should use regular options for enabling/disabling
13276 @item -fdisable-ipa-@var{pass}
13277 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13278 statically invoked in the compiler multiple times, the pass name should be
13279 appended with a sequential number starting from 1.
13281 @item -fdisable-rtl-@var{pass}
13282 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13283 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13284 statically invoked in the compiler multiple times, the pass name should be
13285 appended with a sequential number starting from 1. @var{range-list} is a
13286 comma-separated list of function ranges or assembler names. Each range is a number
13287 pair separated by a colon. The range is inclusive in both ends. If the range
13288 is trivial, the number pair can be simplified as a single number. If the
13289 function's call graph node's @var{uid} falls within one of the specified ranges,
13290 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13291 function header of a dump file, and the pass names can be dumped by using
13292 option @option{-fdump-passes}.
13294 @item -fdisable-tree-@var{pass}
13295 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13296 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13299 @item -fenable-ipa-@var{pass}
13300 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13301 statically invoked in the compiler multiple times, the pass name should be
13302 appended with a sequential number starting from 1.
13304 @item -fenable-rtl-@var{pass}
13305 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13306 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13307 description and examples.
13309 @item -fenable-tree-@var{pass}
13310 @itemx -fenable-tree-@var{pass}=@var{range-list}
13311 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13312 of option arguments.
13316 Here are some examples showing uses of these options.
13320 # disable ccp1 for all functions
13321 -fdisable-tree-ccp1
13322 # disable complete unroll for function whose cgraph node uid is 1
13323 -fenable-tree-cunroll=1
13324 # disable gcse2 for functions at the following ranges [1,1],
13325 # [300,400], and [400,1000]
13326 # disable gcse2 for functions foo and foo2
13327 -fdisable-rtl-gcse2=foo,foo2
13328 # disable early inlining
13329 -fdisable-tree-einline
13330 # disable ipa inlining
13331 -fdisable-ipa-inline
13332 # enable tree full unroll
13333 -fenable-tree-unroll
13338 @itemx -fchecking=@var{n}
13340 @opindex fno-checking
13341 Enable internal consistency checking. The default depends on
13342 the compiler configuration. @option{-fchecking=2} enables further
13343 internal consistency checking that might affect code generation.
13345 @item -frandom-seed=@var{string}
13346 @opindex frandom-seed
13347 This option provides a seed that GCC uses in place of
13348 random numbers in generating certain symbol names
13349 that have to be different in every compiled file. It is also used to
13350 place unique stamps in coverage data files and the object files that
13351 produce them. You can use the @option{-frandom-seed} option to produce
13352 reproducibly identical object files.
13354 The @var{string} can either be a number (decimal, octal or hex) or an
13355 arbitrary string (in which case it's converted to a number by
13358 The @var{string} should be different for every file you compile.
13361 @itemx -save-temps=cwd
13362 @opindex save-temps
13363 Store the usual ``temporary'' intermediate files permanently; place them
13364 in the current directory and name them based on the source file. Thus,
13365 compiling @file{foo.c} with @option{-c -save-temps} produces files
13366 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13367 preprocessed @file{foo.i} output file even though the compiler now
13368 normally uses an integrated preprocessor.
13370 When used in combination with the @option{-x} command-line option,
13371 @option{-save-temps} is sensible enough to avoid over writing an
13372 input source file with the same extension as an intermediate file.
13373 The corresponding intermediate file may be obtained by renaming the
13374 source file before using @option{-save-temps}.
13376 If you invoke GCC in parallel, compiling several different source
13377 files that share a common base name in different subdirectories or the
13378 same source file compiled for multiple output destinations, it is
13379 likely that the different parallel compilers will interfere with each
13380 other, and overwrite the temporary files. For instance:
13383 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13384 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13387 may result in @file{foo.i} and @file{foo.o} being written to
13388 simultaneously by both compilers.
13390 @item -save-temps=obj
13391 @opindex save-temps=obj
13392 Store the usual ``temporary'' intermediate files permanently. If the
13393 @option{-o} option is used, the temporary files are based on the
13394 object file. If the @option{-o} option is not used, the
13395 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13400 gcc -save-temps=obj -c foo.c
13401 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13402 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13406 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13407 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13408 @file{dir2/yfoobar.o}.
13410 @item -time@r{[}=@var{file}@r{]}
13412 Report the CPU time taken by each subprocess in the compilation
13413 sequence. For C source files, this is the compiler proper and assembler
13414 (plus the linker if linking is done).
13416 Without the specification of an output file, the output looks like this:
13423 The first number on each line is the ``user time'', that is time spent
13424 executing the program itself. The second number is ``system time'',
13425 time spent executing operating system routines on behalf of the program.
13426 Both numbers are in seconds.
13428 With the specification of an output file, the output is appended to the
13429 named file, and it looks like this:
13432 0.12 0.01 cc1 @var{options}
13433 0.00 0.01 as @var{options}
13436 The ``user time'' and the ``system time'' are moved before the program
13437 name, and the options passed to the program are displayed, so that one
13438 can later tell what file was being compiled, and with which options.
13440 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13441 @opindex fdump-final-insns
13442 Dump the final internal representation (RTL) to @var{file}. If the
13443 optional argument is omitted (or if @var{file} is @code{.}), the name
13444 of the dump file is determined by appending @code{.gkd} to the
13445 compilation output file name.
13447 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13448 @opindex fcompare-debug
13449 @opindex fno-compare-debug
13450 If no error occurs during compilation, run the compiler a second time,
13451 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13452 passed to the second compilation. Dump the final internal
13453 representation in both compilations, and print an error if they differ.
13455 If the equal sign is omitted, the default @option{-gtoggle} is used.
13457 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13458 and nonzero, implicitly enables @option{-fcompare-debug}. If
13459 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13460 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13463 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13464 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13465 of the final representation and the second compilation, preventing even
13466 @env{GCC_COMPARE_DEBUG} from taking effect.
13468 To verify full coverage during @option{-fcompare-debug} testing, set
13469 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13470 which GCC rejects as an invalid option in any actual compilation
13471 (rather than preprocessing, assembly or linking). To get just a
13472 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13473 not overridden} will do.
13475 @item -fcompare-debug-second
13476 @opindex fcompare-debug-second
13477 This option is implicitly passed to the compiler for the second
13478 compilation requested by @option{-fcompare-debug}, along with options to
13479 silence warnings, and omitting other options that would cause
13480 side-effect compiler outputs to files or to the standard output. Dump
13481 files and preserved temporary files are renamed so as to contain the
13482 @code{.gk} additional extension during the second compilation, to avoid
13483 overwriting those generated by the first.
13485 When this option is passed to the compiler driver, it causes the
13486 @emph{first} compilation to be skipped, which makes it useful for little
13487 other than debugging the compiler proper.
13491 Turn off generation of debug info, if leaving out this option
13492 generates it, or turn it on at level 2 otherwise. The position of this
13493 argument in the command line does not matter; it takes effect after all
13494 other options are processed, and it does so only once, no matter how
13495 many times it is given. This is mainly intended to be used with
13496 @option{-fcompare-debug}.
13498 @item -fvar-tracking-assignments-toggle
13499 @opindex fvar-tracking-assignments-toggle
13500 @opindex fno-var-tracking-assignments-toggle
13501 Toggle @option{-fvar-tracking-assignments}, in the same way that
13502 @option{-gtoggle} toggles @option{-g}.
13506 Makes the compiler print out each function name as it is compiled, and
13507 print some statistics about each pass when it finishes.
13509 @item -ftime-report
13510 @opindex ftime-report
13511 Makes the compiler print some statistics about the time consumed by each
13512 pass when it finishes.
13514 @item -ftime-report-details
13515 @opindex ftime-report-details
13516 Record the time consumed by infrastructure parts separately for each pass.
13518 @item -fira-verbose=@var{n}
13519 @opindex fira-verbose
13520 Control the verbosity of the dump file for the integrated register allocator.
13521 The default value is 5. If the value @var{n} is greater or equal to 10,
13522 the dump output is sent to stderr using the same format as @var{n} minus 10.
13525 @opindex flto-report
13526 Prints a report with internal details on the workings of the link-time
13527 optimizer. The contents of this report vary from version to version.
13528 It is meant to be useful to GCC developers when processing object
13529 files in LTO mode (via @option{-flto}).
13531 Disabled by default.
13533 @item -flto-report-wpa
13534 @opindex flto-report-wpa
13535 Like @option{-flto-report}, but only print for the WPA phase of Link
13539 @opindex fmem-report
13540 Makes the compiler print some statistics about permanent memory
13541 allocation when it finishes.
13543 @item -fmem-report-wpa
13544 @opindex fmem-report-wpa
13545 Makes the compiler print some statistics about permanent memory
13546 allocation for the WPA phase only.
13548 @item -fpre-ipa-mem-report
13549 @opindex fpre-ipa-mem-report
13550 @item -fpost-ipa-mem-report
13551 @opindex fpost-ipa-mem-report
13552 Makes the compiler print some statistics about permanent memory
13553 allocation before or after interprocedural optimization.
13555 @item -fprofile-report
13556 @opindex fprofile-report
13557 Makes the compiler print some statistics about consistency of the
13558 (estimated) profile and effect of individual passes.
13560 @item -fstack-usage
13561 @opindex fstack-usage
13562 Makes the compiler output stack usage information for the program, on a
13563 per-function basis. The filename for the dump is made by appending
13564 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13565 the output file, if explicitly specified and it is not an executable,
13566 otherwise it is the basename of the source file. An entry is made up
13571 The name of the function.
13575 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
13578 The qualifier @code{static} means that the function manipulates the stack
13579 statically: a fixed number of bytes are allocated for the frame on function
13580 entry and released on function exit; no stack adjustments are otherwise made
13581 in the function. The second field is this fixed number of bytes.
13583 The qualifier @code{dynamic} means that the function manipulates the stack
13584 dynamically: in addition to the static allocation described above, stack
13585 adjustments are made in the body of the function, for example to push/pop
13586 arguments around function calls. If the qualifier @code{bounded} is also
13587 present, the amount of these adjustments is bounded at compile time and
13588 the second field is an upper bound of the total amount of stack used by
13589 the function. If it is not present, the amount of these adjustments is
13590 not bounded at compile time and the second field only represents the
13595 Emit statistics about front-end processing at the end of the compilation.
13596 This option is supported only by the C++ front end, and
13597 the information is generally only useful to the G++ development team.
13599 @item -fdbg-cnt-list
13600 @opindex fdbg-cnt-list
13601 Print the name and the counter upper bound for all debug counters.
13604 @item -fdbg-cnt=@var{counter-value-list}
13606 Set the internal debug counter upper bound. @var{counter-value-list}
13607 is a comma-separated list of @var{name}:@var{value} pairs
13608 which sets the upper bound of each debug counter @var{name} to @var{value}.
13609 All debug counters have the initial upper bound of @code{UINT_MAX};
13610 thus @code{dbg_cnt} returns true always unless the upper bound
13611 is set by this option.
13612 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
13613 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
13615 @item -print-file-name=@var{library}
13616 @opindex print-file-name
13617 Print the full absolute name of the library file @var{library} that
13618 would be used when linking---and don't do anything else. With this
13619 option, GCC does not compile or link anything; it just prints the
13622 @item -print-multi-directory
13623 @opindex print-multi-directory
13624 Print the directory name corresponding to the multilib selected by any
13625 other switches present in the command line. This directory is supposed
13626 to exist in @env{GCC_EXEC_PREFIX}.
13628 @item -print-multi-lib
13629 @opindex print-multi-lib
13630 Print the mapping from multilib directory names to compiler switches
13631 that enable them. The directory name is separated from the switches by
13632 @samp{;}, and each switch starts with an @samp{@@} instead of the
13633 @samp{-}, without spaces between multiple switches. This is supposed to
13634 ease shell processing.
13636 @item -print-multi-os-directory
13637 @opindex print-multi-os-directory
13638 Print the path to OS libraries for the selected
13639 multilib, relative to some @file{lib} subdirectory. If OS libraries are
13640 present in the @file{lib} subdirectory and no multilibs are used, this is
13641 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
13642 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
13643 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
13644 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
13646 @item -print-multiarch
13647 @opindex print-multiarch
13648 Print the path to OS libraries for the selected multiarch,
13649 relative to some @file{lib} subdirectory.
13651 @item -print-prog-name=@var{program}
13652 @opindex print-prog-name
13653 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
13655 @item -print-libgcc-file-name
13656 @opindex print-libgcc-file-name
13657 Same as @option{-print-file-name=libgcc.a}.
13659 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
13660 but you do want to link with @file{libgcc.a}. You can do:
13663 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
13666 @item -print-search-dirs
13667 @opindex print-search-dirs
13668 Print the name of the configured installation directory and a list of
13669 program and library directories @command{gcc} searches---and don't do anything else.
13671 This is useful when @command{gcc} prints the error message
13672 @samp{installation problem, cannot exec cpp0: No such file or directory}.
13673 To resolve this you either need to put @file{cpp0} and the other compiler
13674 components where @command{gcc} expects to find them, or you can set the environment
13675 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
13676 Don't forget the trailing @samp{/}.
13677 @xref{Environment Variables}.
13679 @item -print-sysroot
13680 @opindex print-sysroot
13681 Print the target sysroot directory that is used during
13682 compilation. This is the target sysroot specified either at configure
13683 time or using the @option{--sysroot} option, possibly with an extra
13684 suffix that depends on compilation options. If no target sysroot is
13685 specified, the option prints nothing.
13687 @item -print-sysroot-headers-suffix
13688 @opindex print-sysroot-headers-suffix
13689 Print the suffix added to the target sysroot when searching for
13690 headers, or give an error if the compiler is not configured with such
13691 a suffix---and don't do anything else.
13694 @opindex dumpmachine
13695 Print the compiler's target machine (for example,
13696 @samp{i686-pc-linux-gnu})---and don't do anything else.
13699 @opindex dumpversion
13700 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
13701 anything else. This is the compiler version used in filesystem paths,
13702 specs, can be depending on how the compiler has been configured just
13703 a single number (major version), two numbers separated by dot (major and
13704 minor version) or three numbers separated by dots (major, minor and patchlevel
13707 @item -dumpfullversion
13708 @opindex dumpfullversion
13709 Print the full compiler version, always 3 numbers separated by dots,
13710 major, minor and patchlevel version.
13714 Print the compiler's built-in specs---and don't do anything else. (This
13715 is used when GCC itself is being built.) @xref{Spec Files}.
13718 @node Submodel Options
13719 @section Machine-Dependent Options
13720 @cindex submodel options
13721 @cindex specifying hardware config
13722 @cindex hardware models and configurations, specifying
13723 @cindex target-dependent options
13724 @cindex machine-dependent options
13726 Each target machine supported by GCC can have its own options---for
13727 example, to allow you to compile for a particular processor variant or
13728 ABI, or to control optimizations specific to that machine. By
13729 convention, the names of machine-specific options start with
13732 Some configurations of the compiler also support additional target-specific
13733 options, usually for compatibility with other compilers on the same
13736 @c This list is ordered alphanumerically by subsection name.
13737 @c It should be the same order and spelling as these options are listed
13738 @c in Machine Dependent Options
13741 * AArch64 Options::
13742 * Adapteva Epiphany Options::
13746 * Blackfin Options::
13751 * DEC Alpha Options::
13755 * GNU/Linux Options::
13765 * MicroBlaze Options::
13768 * MN10300 Options::
13772 * Nios II Options::
13773 * Nvidia PTX Options::
13775 * picoChip Options::
13776 * PowerPC Options::
13779 * RS/6000 and PowerPC Options::
13781 * S/390 and zSeries Options::
13784 * Solaris 2 Options::
13787 * System V Options::
13788 * TILE-Gx Options::
13789 * TILEPro Options::
13794 * VxWorks Options::
13796 * x86 Windows Options::
13797 * Xstormy16 Options::
13799 * zSeries Options::
13802 @node AArch64 Options
13803 @subsection AArch64 Options
13804 @cindex AArch64 Options
13806 These options are defined for AArch64 implementations:
13810 @item -mabi=@var{name}
13812 Generate code for the specified data model. Permissible values
13813 are @samp{ilp32} for SysV-like data model where int, long int and pointers
13814 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
13815 but long int and pointers are 64 bits.
13817 The default depends on the specific target configuration. Note that
13818 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
13819 entire program with the same ABI, and link with a compatible set of libraries.
13822 @opindex mbig-endian
13823 Generate big-endian code. This is the default when GCC is configured for an
13824 @samp{aarch64_be-*-*} target.
13826 @item -mgeneral-regs-only
13827 @opindex mgeneral-regs-only
13828 Generate code which uses only the general-purpose registers. This will prevent
13829 the compiler from using floating-point and Advanced SIMD registers but will not
13830 impose any restrictions on the assembler.
13832 @item -mlittle-endian
13833 @opindex mlittle-endian
13834 Generate little-endian code. This is the default when GCC is configured for an
13835 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
13837 @item -mcmodel=tiny
13838 @opindex mcmodel=tiny
13839 Generate code for the tiny code model. The program and its statically defined
13840 symbols must be within 1MB of each other. Programs can be statically or
13841 dynamically linked.
13843 @item -mcmodel=small
13844 @opindex mcmodel=small
13845 Generate code for the small code model. The program and its statically defined
13846 symbols must be within 4GB of each other. Programs can be statically or
13847 dynamically linked. This is the default code model.
13849 @item -mcmodel=large
13850 @opindex mcmodel=large
13851 Generate code for the large code model. This makes no assumptions about
13852 addresses and sizes of sections. Programs can be statically linked only.
13854 @item -mstrict-align
13855 @opindex mstrict-align
13856 Avoid generating memory accesses that may not be aligned on a natural object
13857 boundary as described in the architecture specification.
13859 @item -momit-leaf-frame-pointer
13860 @itemx -mno-omit-leaf-frame-pointer
13861 @opindex momit-leaf-frame-pointer
13862 @opindex mno-omit-leaf-frame-pointer
13863 Omit or keep the frame pointer in leaf functions. The former behavior is the
13866 @item -mtls-dialect=desc
13867 @opindex mtls-dialect=desc
13868 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
13869 of TLS variables. This is the default.
13871 @item -mtls-dialect=traditional
13872 @opindex mtls-dialect=traditional
13873 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
13876 @item -mtls-size=@var{size}
13878 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
13879 This option requires binutils 2.26 or newer.
13881 @item -mfix-cortex-a53-835769
13882 @itemx -mno-fix-cortex-a53-835769
13883 @opindex mfix-cortex-a53-835769
13884 @opindex mno-fix-cortex-a53-835769
13885 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
13886 This involves inserting a NOP instruction between memory instructions and
13887 64-bit integer multiply-accumulate instructions.
13889 @item -mfix-cortex-a53-843419
13890 @itemx -mno-fix-cortex-a53-843419
13891 @opindex mfix-cortex-a53-843419
13892 @opindex mno-fix-cortex-a53-843419
13893 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
13894 This erratum workaround is made at link time and this will only pass the
13895 corresponding flag to the linker.
13897 @item -mlow-precision-recip-sqrt
13898 @item -mno-low-precision-recip-sqrt
13899 @opindex mlow-precision-recip-sqrt
13900 @opindex mno-low-precision-recip-sqrt
13901 Enable or disable the reciprocal square root approximation.
13902 This option only has an effect if @option{-ffast-math} or
13903 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13904 precision of reciprocal square root results to about 16 bits for
13905 single precision and to 32 bits for double precision.
13907 @item -mlow-precision-sqrt
13908 @item -mno-low-precision-sqrt
13909 @opindex -mlow-precision-sqrt
13910 @opindex -mno-low-precision-sqrt
13911 Enable or disable the square root approximation.
13912 This option only has an effect if @option{-ffast-math} or
13913 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13914 precision of square root results to about 16 bits for
13915 single precision and to 32 bits for double precision.
13916 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
13918 @item -mlow-precision-div
13919 @item -mno-low-precision-div
13920 @opindex -mlow-precision-div
13921 @opindex -mno-low-precision-div
13922 Enable or disable the division approximation.
13923 This option only has an effect if @option{-ffast-math} or
13924 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13925 precision of division results to about 16 bits for
13926 single precision and to 32 bits for double precision.
13928 @item -march=@var{name}
13930 Specify the name of the target architecture and, optionally, one or
13931 more feature modifiers. This option has the form
13932 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
13934 The permissible values for @var{arch} are @samp{armv8-a},
13935 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @var{native}.
13937 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
13938 support for the ARMv8.3-A architecture extensions.
13940 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
13941 support for the ARMv8.2-A architecture extensions.
13943 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
13944 support for the ARMv8.1-A architecture extension. In particular, it
13945 enables the @samp{+crc} and @samp{+lse} features.
13947 The value @samp{native} is available on native AArch64 GNU/Linux and
13948 causes the compiler to pick the architecture of the host system. This
13949 option has no effect if the compiler is unable to recognize the
13950 architecture of the host system,
13952 The permissible values for @var{feature} are listed in the sub-section
13953 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13954 Feature Modifiers}. Where conflicting feature modifiers are
13955 specified, the right-most feature is used.
13957 GCC uses @var{name} to determine what kind of instructions it can emit
13958 when generating assembly code. If @option{-march} is specified
13959 without either of @option{-mtune} or @option{-mcpu} also being
13960 specified, the code is tuned to perform well across a range of target
13961 processors implementing the target architecture.
13963 @item -mtune=@var{name}
13965 Specify the name of the target processor for which GCC should tune the
13966 performance of the code. Permissible values for this option are:
13967 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
13968 @samp{cortex-a72}, @samp{cortex-a73}, @samp{exynos-m1}, @samp{falkor},
13969 @samp{qdf24xx}, @samp{xgene1}, @samp{vulcan}, @samp{thunderx},
13970 @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
13971 @samp{thunderxt83}, @samp{thunderx2t99}, @samp{cortex-a57.cortex-a53},
13972 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
13973 @samp{cortex-a73.cortex-a53}, @samp{native}.
13975 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
13976 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}
13977 specify that GCC should tune for a big.LITTLE system.
13979 Additionally on native AArch64 GNU/Linux systems the value
13980 @samp{native} tunes performance to the host system. This option has no effect
13981 if the compiler is unable to recognize the processor of the host system.
13983 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
13984 are specified, the code is tuned to perform well across a range
13985 of target processors.
13987 This option cannot be suffixed by feature modifiers.
13989 @item -mcpu=@var{name}
13991 Specify the name of the target processor, optionally suffixed by one
13992 or more feature modifiers. This option has the form
13993 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
13994 the permissible values for @var{cpu} are the same as those available
13995 for @option{-mtune}. The permissible values for @var{feature} are
13996 documented in the sub-section on
13997 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13998 Feature Modifiers}. Where conflicting feature modifiers are
13999 specified, the right-most feature is used.
14001 GCC uses @var{name} to determine what kind of instructions it can emit when
14002 generating assembly code (as if by @option{-march}) and to determine
14003 the target processor for which to tune for performance (as if
14004 by @option{-mtune}). Where this option is used in conjunction
14005 with @option{-march} or @option{-mtune}, those options take precedence
14006 over the appropriate part of this option.
14008 @item -moverride=@var{string}
14010 Override tuning decisions made by the back-end in response to a
14011 @option{-mtune=} switch. The syntax, semantics, and accepted values
14012 for @var{string} in this option are not guaranteed to be consistent
14015 This option is only intended to be useful when developing GCC.
14017 @item -mpc-relative-literal-loads
14018 @opindex mpc-relative-literal-loads
14019 Enable PC-relative literal loads. With this option literal pools are
14020 accessed using a single instruction and emitted after each function. This
14021 limits the maximum size of functions to 1MB. This is enabled by default for
14022 @option{-mcmodel=tiny}.
14024 @item -msign-return-address=@var{scope}
14025 @opindex msign-return-address
14026 Select the function scope on which return address signing will be applied.
14027 Permissible values are @samp{none}, which disables return address signing,
14028 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
14029 functions, and @samp{all}, which enables pointer signing for all functions. The
14030 default value is @samp{none}.
14034 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14035 @anchor{aarch64-feature-modifiers}
14036 @cindex @option{-march} feature modifiers
14037 @cindex @option{-mcpu} feature modifiers
14038 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14039 the following and their inverses @option{no@var{feature}}:
14043 Enable CRC extension. This is on by default for
14044 @option{-march=armv8.1-a}.
14046 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14049 Enable floating-point instructions. This is on by default for all possible
14050 values for options @option{-march} and @option{-mcpu}.
14052 Enable Advanced SIMD instructions. This also enables floating-point
14053 instructions. This is on by default for all possible values for options
14054 @option{-march} and @option{-mcpu}.
14056 Enable Large System Extension instructions. This is on by default for
14057 @option{-march=armv8.1-a}.
14059 Enable FP16 extension. This also enables floating-point instructions.
14063 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
14064 Conversely, @option{nofp} implies @option{nosimd}, which implies
14067 @node Adapteva Epiphany Options
14068 @subsection Adapteva Epiphany Options
14070 These @samp{-m} options are defined for Adapteva Epiphany:
14073 @item -mhalf-reg-file
14074 @opindex mhalf-reg-file
14075 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14076 That allows code to run on hardware variants that lack these registers.
14078 @item -mprefer-short-insn-regs
14079 @opindex mprefer-short-insn-regs
14080 Preferentially allocate registers that allow short instruction generation.
14081 This can result in increased instruction count, so this may either reduce or
14082 increase overall code size.
14084 @item -mbranch-cost=@var{num}
14085 @opindex mbranch-cost
14086 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14087 This cost is only a heuristic and is not guaranteed to produce
14088 consistent results across releases.
14092 Enable the generation of conditional moves.
14094 @item -mnops=@var{num}
14096 Emit @var{num} NOPs before every other generated instruction.
14098 @item -mno-soft-cmpsf
14099 @opindex mno-soft-cmpsf
14100 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14101 and test the flags. This is faster than a software comparison, but can
14102 get incorrect results in the presence of NaNs, or when two different small
14103 numbers are compared such that their difference is calculated as zero.
14104 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14105 software comparisons.
14107 @item -mstack-offset=@var{num}
14108 @opindex mstack-offset
14109 Set the offset between the top of the stack and the stack pointer.
14110 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14111 can be used by leaf functions without stack allocation.
14112 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14113 Note also that this option changes the ABI; compiling a program with a
14114 different stack offset than the libraries have been compiled with
14115 generally does not work.
14116 This option can be useful if you want to evaluate if a different stack
14117 offset would give you better code, but to actually use a different stack
14118 offset to build working programs, it is recommended to configure the
14119 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14121 @item -mno-round-nearest
14122 @opindex mno-round-nearest
14123 Make the scheduler assume that the rounding mode has been set to
14124 truncating. The default is @option{-mround-nearest}.
14127 @opindex mlong-calls
14128 If not otherwise specified by an attribute, assume all calls might be beyond
14129 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14130 function address into a register before performing a (otherwise direct) call.
14131 This is the default.
14133 @item -mshort-calls
14134 @opindex short-calls
14135 If not otherwise specified by an attribute, assume all direct calls are
14136 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14137 for direct calls. The default is @option{-mlong-calls}.
14141 Assume addresses can be loaded as 16-bit unsigned values. This does not
14142 apply to function addresses for which @option{-mlong-calls} semantics
14145 @item -mfp-mode=@var{mode}
14147 Set the prevailing mode of the floating-point unit.
14148 This determines the floating-point mode that is provided and expected
14149 at function call and return time. Making this mode match the mode you
14150 predominantly need at function start can make your programs smaller and
14151 faster by avoiding unnecessary mode switches.
14153 @var{mode} can be set to one the following values:
14157 Any mode at function entry is valid, and retained or restored when
14158 the function returns, and when it calls other functions.
14159 This mode is useful for compiling libraries or other compilation units
14160 you might want to incorporate into different programs with different
14161 prevailing FPU modes, and the convenience of being able to use a single
14162 object file outweighs the size and speed overhead for any extra
14163 mode switching that might be needed, compared with what would be needed
14164 with a more specific choice of prevailing FPU mode.
14167 This is the mode used for floating-point calculations with
14168 truncating (i.e.@: round towards zero) rounding mode. That includes
14169 conversion from floating point to integer.
14171 @item round-nearest
14172 This is the mode used for floating-point calculations with
14173 round-to-nearest-or-even rounding mode.
14176 This is the mode used to perform integer calculations in the FPU, e.g.@:
14177 integer multiply, or integer multiply-and-accumulate.
14180 The default is @option{-mfp-mode=caller}
14182 @item -mnosplit-lohi
14183 @itemx -mno-postinc
14184 @itemx -mno-postmodify
14185 @opindex mnosplit-lohi
14186 @opindex mno-postinc
14187 @opindex mno-postmodify
14188 Code generation tweaks that disable, respectively, splitting of 32-bit
14189 loads, generation of post-increment addresses, and generation of
14190 post-modify addresses. The defaults are @option{msplit-lohi},
14191 @option{-mpost-inc}, and @option{-mpost-modify}.
14193 @item -mnovect-double
14194 @opindex mno-vect-double
14195 Change the preferred SIMD mode to SImode. The default is
14196 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14198 @item -max-vect-align=@var{num}
14199 @opindex max-vect-align
14200 The maximum alignment for SIMD vector mode types.
14201 @var{num} may be 4 or 8. The default is 8.
14202 Note that this is an ABI change, even though many library function
14203 interfaces are unaffected if they don't use SIMD vector modes
14204 in places that affect size and/or alignment of relevant types.
14206 @item -msplit-vecmove-early
14207 @opindex msplit-vecmove-early
14208 Split vector moves into single word moves before reload. In theory this
14209 can give better register allocation, but so far the reverse seems to be
14210 generally the case.
14212 @item -m1reg-@var{reg}
14214 Specify a register to hold the constant @minus{}1, which makes loading small negative
14215 constants and certain bitmasks faster.
14216 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14217 which specify use of that register as a fixed register,
14218 and @samp{none}, which means that no register is used for this
14219 purpose. The default is @option{-m1reg-none}.
14224 @subsection ARC Options
14225 @cindex ARC options
14227 The following options control the architecture variant for which code
14230 @c architecture variants
14233 @item -mbarrel-shifter
14234 @opindex mbarrel-shifter
14235 Generate instructions supported by barrel shifter. This is the default
14236 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14238 @item -mcpu=@var{cpu}
14240 Set architecture type, register usage, and instruction scheduling
14241 parameters for @var{cpu}. There are also shortcut alias options
14242 available for backward compatibility and convenience. Supported
14243 values for @var{cpu} are
14249 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14253 Compile for ARC601. Alias: @option{-mARC601}.
14258 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14259 This is the default when configured with @option{--with-cpu=arc700}@.
14262 Compile for ARC EM.
14265 Compile for ARC HS.
14268 Compile for ARC EM CPU with no hardware extensions.
14271 Compile for ARC EM4 CPU.
14274 Compile for ARC EM4 DMIPS CPU.
14277 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
14281 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
14282 double assist instructions.
14285 Compile for ARC HS CPU with no hardware extensions except the atomic
14289 Compile for ARC HS34 CPU.
14292 Compile for ARC HS38 CPU.
14295 Compile for ARC HS38 CPU with all hardware extensions on.
14298 Compile for ARC 600 CPU with @code{norm} instructions enabled.
14300 @item arc600_mul32x16
14301 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
14302 instructions enabled.
14305 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
14306 instructions enabled.
14309 Compile for ARC 601 CPU with @code{norm} instructions enabled.
14311 @item arc601_mul32x16
14312 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
14313 instructions enabled.
14316 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
14317 instructions enabled.
14320 Compile for ARC 700 on NPS400 chip.
14326 @itemx -mdpfp-compact
14327 @opindex mdpfp-compact
14328 Generate double-precision FPX instructions, tuned for the compact
14332 @opindex mdpfp-fast
14333 Generate double-precision FPX instructions, tuned for the fast
14336 @item -mno-dpfp-lrsr
14337 @opindex mno-dpfp-lrsr
14338 Disable @code{lr} and @code{sr} instructions from using FPX extension
14343 Generate extended arithmetic instructions. Currently only
14344 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14345 supported. This is always enabled for @option{-mcpu=ARC700}.
14349 Do not generate @code{mpy}-family instructions for ARC700. This option is
14354 Generate 32x16-bit multiply and multiply-accumulate instructions.
14358 Generate @code{mul64} and @code{mulu64} instructions.
14359 Only valid for @option{-mcpu=ARC600}.
14363 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
14368 @itemx -mspfp-compact
14369 @opindex mspfp-compact
14370 Generate single-precision FPX instructions, tuned for the compact
14374 @opindex mspfp-fast
14375 Generate single-precision FPX instructions, tuned for the fast
14380 Enable generation of ARC SIMD instructions via target-specific
14381 builtins. Only valid for @option{-mcpu=ARC700}.
14384 @opindex msoft-float
14385 This option ignored; it is provided for compatibility purposes only.
14386 Software floating-point code is emitted by default, and this default
14387 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
14388 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
14389 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
14393 Generate @code{swap} instructions.
14397 This enables use of the locked load/store conditional extension to implement
14398 atomic memory built-in functions. Not available for ARC 6xx or ARC
14403 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
14405 @item -mcode-density
14406 @opindex mcode-density
14407 Enable code density instructions for ARC EM.
14408 This option is on by default for ARC HS.
14412 Enable double load/store operations for ARC HS cores.
14414 @item -mtp-regno=@var{regno}
14416 Specify thread pointer register number.
14418 @item -mmpy-option=@var{multo}
14419 @opindex mmpy-option
14420 Compile ARCv2 code with a multiplier design option. You can specify
14421 the option using either a string or numeric value for @var{multo}.
14422 @samp{wlh1} is the default value. The recognized values are:
14427 No multiplier available.
14431 16x16 multiplier, fully pipelined.
14432 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
14436 32x32 multiplier, fully
14437 pipelined (1 stage). The following instructions are additionally
14438 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14442 32x32 multiplier, fully pipelined
14443 (2 stages). The following instructions are additionally enabled: @code{mpy},
14444 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14448 Two 16x16 multipliers, blocking,
14449 sequential. The following instructions are additionally enabled: @code{mpy},
14450 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14454 One 16x16 multiplier, blocking,
14455 sequential. The following instructions are additionally enabled: @code{mpy},
14456 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14460 One 32x4 multiplier, blocking,
14461 sequential. The following instructions are additionally enabled: @code{mpy},
14462 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
14466 ARC HS SIMD support.
14470 ARC HS SIMD support.
14474 ARC HS SIMD support.
14478 This option is only available for ARCv2 cores@.
14480 @item -mfpu=@var{fpu}
14482 Enables support for specific floating-point hardware extensions for ARCv2
14483 cores. Supported values for @var{fpu} are:
14488 Enables support for single-precision floating-point hardware
14492 Enables support for double-precision floating-point hardware
14493 extensions. The single-precision floating-point extension is also
14494 enabled. Not available for ARC EM@.
14497 Enables support for double-precision floating-point hardware
14498 extensions using double-precision assist instructions. The single-precision
14499 floating-point extension is also enabled. This option is
14500 only available for ARC EM@.
14503 Enables support for double-precision floating-point hardware
14504 extensions using double-precision assist instructions.
14505 The single-precision floating-point, square-root, and divide
14506 extensions are also enabled. This option is
14507 only available for ARC EM@.
14510 Enables support for double-precision floating-point hardware
14511 extensions using double-precision assist instructions.
14512 The single-precision floating-point and fused multiply and add
14513 hardware extensions are also enabled. This option is
14514 only available for ARC EM@.
14517 Enables support for double-precision floating-point hardware
14518 extensions using double-precision assist instructions.
14519 All single-precision floating-point hardware extensions are also
14520 enabled. This option is only available for ARC EM@.
14523 Enables support for single-precision floating-point, square-root and divide
14524 hardware extensions@.
14527 Enables support for double-precision floating-point, square-root and divide
14528 hardware extensions. This option
14529 includes option @samp{fpus_div}. Not available for ARC EM@.
14532 Enables support for single-precision floating-point and
14533 fused multiply and add hardware extensions@.
14536 Enables support for double-precision floating-point and
14537 fused multiply and add hardware extensions. This option
14538 includes option @samp{fpus_fma}. Not available for ARC EM@.
14541 Enables support for all single-precision floating-point hardware
14545 Enables support for all single- and double-precision floating-point
14546 hardware extensions. Not available for ARC EM@.
14552 The following options are passed through to the assembler, and also
14553 define preprocessor macro symbols.
14555 @c Flags used by the assembler, but for which we define preprocessor
14556 @c macro symbols as well.
14559 @opindex mdsp-packa
14560 Passed down to the assembler to enable the DSP Pack A extensions.
14561 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
14566 Passed down to the assembler to enable the dual Viterbi butterfly
14567 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
14568 option is deprecated.
14570 @c ARC700 4.10 extension instruction
14573 Passed down to the assembler to enable the locked load/store
14574 conditional extension. Also sets the preprocessor symbol
14579 Passed down to the assembler. Also sets the preprocessor symbol
14580 @code{__Xxmac_d16}. This option is deprecated.
14584 Passed down to the assembler. Also sets the preprocessor symbol
14585 @code{__Xxmac_24}. This option is deprecated.
14587 @c ARC700 4.10 extension instruction
14590 Passed down to the assembler to enable the 64-bit time-stamp counter
14591 extension instruction. Also sets the preprocessor symbol
14592 @code{__Xrtsc}. This option is deprecated.
14594 @c ARC700 4.10 extension instruction
14597 Passed down to the assembler to enable the swap byte ordering
14598 extension instruction. Also sets the preprocessor symbol
14602 @opindex mtelephony
14603 Passed down to the assembler to enable dual- and single-operand
14604 instructions for telephony. Also sets the preprocessor symbol
14605 @code{__Xtelephony}. This option is deprecated.
14609 Passed down to the assembler to enable the XY memory extension. Also
14610 sets the preprocessor symbol @code{__Xxy}.
14614 The following options control how the assembly code is annotated:
14616 @c Assembly annotation options
14620 Annotate assembler instructions with estimated addresses.
14622 @item -mannotate-align
14623 @opindex mannotate-align
14624 Explain what alignment considerations lead to the decision to make an
14625 instruction short or long.
14629 The following options are passed through to the linker:
14631 @c options passed through to the linker
14635 Passed through to the linker, to specify use of the @code{arclinux} emulation.
14636 This option is enabled by default in tool chains built for
14637 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
14638 when profiling is not requested.
14640 @item -marclinux_prof
14641 @opindex marclinux_prof
14642 Passed through to the linker, to specify use of the
14643 @code{arclinux_prof} emulation. This option is enabled by default in
14644 tool chains built for @w{@code{arc-linux-uclibc}} and
14645 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
14649 The following options control the semantics of generated code:
14651 @c semantically relevant code generation options
14654 @opindex mlong-calls
14655 Generate calls as register indirect calls, thus providing access
14656 to the full 32-bit address range.
14658 @item -mmedium-calls
14659 @opindex mmedium-calls
14660 Don't use less than 25-bit addressing range for calls, which is the
14661 offset available for an unconditional branch-and-link
14662 instruction. Conditional execution of function calls is suppressed, to
14663 allow use of the 25-bit range, rather than the 21-bit range with
14664 conditional branch-and-link. This is the default for tool chains built
14665 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
14669 Do not generate sdata references. This is the default for tool chains
14670 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
14673 @item -mvolatile-cache
14674 @opindex mvolatile-cache
14675 Use ordinarily cached memory accesses for volatile references. This is the
14678 @item -mno-volatile-cache
14679 @opindex mno-volatile-cache
14680 Enable cache bypass for volatile references.
14684 The following options fine tune code generation:
14685 @c code generation tuning options
14688 @opindex malign-call
14689 Do alignment optimizations for call instructions.
14691 @item -mauto-modify-reg
14692 @opindex mauto-modify-reg
14693 Enable the use of pre/post modify with register displacement.
14695 @item -mbbit-peephole
14696 @opindex mbbit-peephole
14697 Enable bbit peephole2.
14701 This option disables a target-specific pass in @file{arc_reorg} to
14702 generate compare-and-branch (@code{br@var{cc}}) instructions.
14703 It has no effect on
14704 generation of these instructions driven by the combiner pass.
14706 @item -mcase-vector-pcrel
14707 @opindex mcase-vector-pcrel
14708 Use PC-relative switch case tables to enable case table shortening.
14709 This is the default for @option{-Os}.
14711 @item -mcompact-casesi
14712 @opindex mcompact-casesi
14713 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
14714 and only available for ARCv1 cores.
14716 @item -mno-cond-exec
14717 @opindex mno-cond-exec
14718 Disable the ARCompact-specific pass to generate conditional
14719 execution instructions.
14721 Due to delay slot scheduling and interactions between operand numbers,
14722 literal sizes, instruction lengths, and the support for conditional execution,
14723 the target-independent pass to generate conditional execution is often lacking,
14724 so the ARC port has kept a special pass around that tries to find more
14725 conditional execution generation opportunities after register allocation,
14726 branch shortening, and delay slot scheduling have been done. This pass
14727 generally, but not always, improves performance and code size, at the cost of
14728 extra compilation time, which is why there is an option to switch it off.
14729 If you have a problem with call instructions exceeding their allowable
14730 offset range because they are conditionalized, you should consider using
14731 @option{-mmedium-calls} instead.
14733 @item -mearly-cbranchsi
14734 @opindex mearly-cbranchsi
14735 Enable pre-reload use of the @code{cbranchsi} pattern.
14737 @item -mexpand-adddi
14738 @opindex mexpand-adddi
14739 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
14740 @code{add.f}, @code{adc} etc.
14742 @item -mindexed-loads
14743 @opindex mindexed-loads
14744 Enable the use of indexed loads. This can be problematic because some
14745 optimizers then assume that indexed stores exist, which is not
14749 Enable Local Register Allocation. This is still experimental for ARC,
14750 so by default the compiler uses standard reload
14751 (i.e. @option{-mno-lra}).
14753 @item -mlra-priority-none
14754 @opindex mlra-priority-none
14755 Don't indicate any priority for target registers.
14757 @item -mlra-priority-compact
14758 @opindex mlra-priority-compact
14759 Indicate target register priority for r0..r3 / r12..r15.
14761 @item -mlra-priority-noncompact
14762 @opindex mlra-priority-noncompact
14763 Reduce target register priority for r0..r3 / r12..r15.
14765 @item -mno-millicode
14766 @opindex mno-millicode
14767 When optimizing for size (using @option{-Os}), prologues and epilogues
14768 that have to save or restore a large number of registers are often
14769 shortened by using call to a special function in libgcc; this is
14770 referred to as a @emph{millicode} call. As these calls can pose
14771 performance issues, and/or cause linking issues when linking in a
14772 nonstandard way, this option is provided to turn off millicode call
14776 @opindex mmixed-code
14777 Tweak register allocation to help 16-bit instruction generation.
14778 This generally has the effect of decreasing the average instruction size
14779 while increasing the instruction count.
14783 Enable @samp{q} instruction alternatives.
14784 This is the default for @option{-Os}.
14788 Enable @samp{Rcq} constraint handling.
14789 Most short code generation depends on this.
14790 This is the default.
14794 Enable @samp{Rcw} constraint handling.
14795 Most ccfsm condexec mostly depends on this.
14796 This is the default.
14798 @item -msize-level=@var{level}
14799 @opindex msize-level
14800 Fine-tune size optimization with regards to instruction lengths and alignment.
14801 The recognized values for @var{level} are:
14804 No size optimization. This level is deprecated and treated like @samp{1}.
14807 Short instructions are used opportunistically.
14810 In addition, alignment of loops and of code after barriers are dropped.
14813 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
14817 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
14818 the behavior when this is not set is equivalent to level @samp{1}.
14820 @item -mtune=@var{cpu}
14822 Set instruction scheduling parameters for @var{cpu}, overriding any implied
14823 by @option{-mcpu=}.
14825 Supported values for @var{cpu} are
14829 Tune for ARC600 CPU.
14832 Tune for ARC601 CPU.
14835 Tune for ARC700 CPU with standard multiplier block.
14838 Tune for ARC700 CPU with XMAC block.
14841 Tune for ARC725D CPU.
14844 Tune for ARC750D CPU.
14848 @item -mmultcost=@var{num}
14850 Cost to assume for a multiply instruction, with @samp{4} being equal to a
14851 normal instruction.
14853 @item -munalign-prob-threshold=@var{probability}
14854 @opindex munalign-prob-threshold
14855 Set probability threshold for unaligning branches.
14856 When tuning for @samp{ARC700} and optimizing for speed, branches without
14857 filled delay slot are preferably emitted unaligned and long, unless
14858 profiling indicates that the probability for the branch to be taken
14859 is below @var{probability}. @xref{Cross-profiling}.
14860 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
14864 The following options are maintained for backward compatibility, but
14865 are now deprecated and will be removed in a future release:
14867 @c Deprecated options
14875 @opindex mbig-endian
14878 Compile code for big-endian targets. Use of these options is now
14879 deprecated. Big-endian code is supported by configuring GCC to build
14880 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
14881 for which big endian is the default.
14883 @item -mlittle-endian
14884 @opindex mlittle-endian
14887 Compile code for little-endian targets. Use of these options is now
14888 deprecated. Little-endian code is supported by configuring GCC to build
14889 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
14890 for which little endian is the default.
14892 @item -mbarrel_shifter
14893 @opindex mbarrel_shifter
14894 Replaced by @option{-mbarrel-shifter}.
14896 @item -mdpfp_compact
14897 @opindex mdpfp_compact
14898 Replaced by @option{-mdpfp-compact}.
14901 @opindex mdpfp_fast
14902 Replaced by @option{-mdpfp-fast}.
14905 @opindex mdsp_packa
14906 Replaced by @option{-mdsp-packa}.
14910 Replaced by @option{-mea}.
14914 Replaced by @option{-mmac-24}.
14918 Replaced by @option{-mmac-d16}.
14920 @item -mspfp_compact
14921 @opindex mspfp_compact
14922 Replaced by @option{-mspfp-compact}.
14925 @opindex mspfp_fast
14926 Replaced by @option{-mspfp-fast}.
14928 @item -mtune=@var{cpu}
14930 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
14931 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
14932 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
14934 @item -multcost=@var{num}
14936 Replaced by @option{-mmultcost}.
14941 @subsection ARM Options
14942 @cindex ARM options
14944 These @samp{-m} options are defined for the ARM port:
14947 @item -mabi=@var{name}
14949 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
14950 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
14953 @opindex mapcs-frame
14954 Generate a stack frame that is compliant with the ARM Procedure Call
14955 Standard for all functions, even if this is not strictly necessary for
14956 correct execution of the code. Specifying @option{-fomit-frame-pointer}
14957 with this option causes the stack frames not to be generated for
14958 leaf functions. The default is @option{-mno-apcs-frame}.
14959 This option is deprecated.
14963 This is a synonym for @option{-mapcs-frame} and is deprecated.
14966 @c not currently implemented
14967 @item -mapcs-stack-check
14968 @opindex mapcs-stack-check
14969 Generate code to check the amount of stack space available upon entry to
14970 every function (that actually uses some stack space). If there is
14971 insufficient space available then either the function
14972 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
14973 called, depending upon the amount of stack space required. The runtime
14974 system is required to provide these functions. The default is
14975 @option{-mno-apcs-stack-check}, since this produces smaller code.
14977 @c not currently implemented
14978 @item -mapcs-reentrant
14979 @opindex mapcs-reentrant
14980 Generate reentrant, position-independent code. The default is
14981 @option{-mno-apcs-reentrant}.
14984 @item -mthumb-interwork
14985 @opindex mthumb-interwork
14986 Generate code that supports calling between the ARM and Thumb
14987 instruction sets. Without this option, on pre-v5 architectures, the
14988 two instruction sets cannot be reliably used inside one program. The
14989 default is @option{-mno-thumb-interwork}, since slightly larger code
14990 is generated when @option{-mthumb-interwork} is specified. In AAPCS
14991 configurations this option is meaningless.
14993 @item -mno-sched-prolog
14994 @opindex mno-sched-prolog
14995 Prevent the reordering of instructions in the function prologue, or the
14996 merging of those instruction with the instructions in the function's
14997 body. This means that all functions start with a recognizable set
14998 of instructions (or in fact one of a choice from a small set of
14999 different function prologues), and this information can be used to
15000 locate the start of functions inside an executable piece of code. The
15001 default is @option{-msched-prolog}.
15003 @item -mfloat-abi=@var{name}
15004 @opindex mfloat-abi
15005 Specifies which floating-point ABI to use. Permissible values
15006 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15008 Specifying @samp{soft} causes GCC to generate output containing
15009 library calls for floating-point operations.
15010 @samp{softfp} allows the generation of code using hardware floating-point
15011 instructions, but still uses the soft-float calling conventions.
15012 @samp{hard} allows generation of floating-point instructions
15013 and uses FPU-specific calling conventions.
15015 The default depends on the specific target configuration. Note that
15016 the hard-float and soft-float ABIs are not link-compatible; you must
15017 compile your entire program with the same ABI, and link with a
15018 compatible set of libraries.
15020 @item -mlittle-endian
15021 @opindex mlittle-endian
15022 Generate code for a processor running in little-endian mode. This is
15023 the default for all standard configurations.
15026 @opindex mbig-endian
15027 Generate code for a processor running in big-endian mode; the default is
15028 to compile code for a little-endian processor.
15030 @item -march=@var{name}
15032 This specifies the name of the target ARM architecture. GCC uses this
15033 name to determine what kind of instructions it can emit when generating
15034 assembly code. This option can be used in conjunction with or instead
15035 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
15036 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
15037 @samp{armv5}, @samp{armv5e}, @samp{armv5t}, @samp{armv5te},
15038 @samp{armv6}, @samp{armv6-m}, @samp{armv6j}, @samp{armv6k},
15039 @samp{armv6kz}, @samp{armv6s-m},
15040 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk},
15041 @samp{armv7}, @samp{armv7-a}, @samp{armv7-m}, @samp{armv7-r}, @samp{armv7e-m},
15042 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
15043 @samp{armv8.1-a+crc}, @samp{armv8-m.base}, @samp{armv8-m.main},
15044 @samp{armv8-m.main+dsp}, @samp{iwmmxt}, @samp{iwmmxt2}.
15046 Architecture revisions older than @samp{armv4t} are deprecated.
15048 @option{-march=armv6s-m} is the @samp{armv6-m} architecture with support for
15049 the (now mandatory) SVC instruction.
15051 @option{-march=armv6zk} is an alias for @samp{armv6kz}, existing for backwards
15054 @option{-march=armv7ve} is the @samp{armv7-a} architecture with virtualization
15057 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
15058 architecture together with the optional CRC32 extensions.
15060 @option{-march=armv8.1-a} enables compiler support for the ARMv8.1-A
15061 architecture. This also enables the features provided by
15062 @option{-march=armv8-a+crc}.
15064 @option{-march=armv8.2-a} enables compiler support for the ARMv8.2-A
15065 architecture. This also enables the features provided by
15066 @option{-march=armv8.1-a}.
15068 @option{-march=armv8.2-a+fp16} enables compiler support for the
15069 ARMv8.2-A architecture with the optional FP16 instructions extension.
15070 This also enables the features provided by @option{-march=armv8.1-a}
15071 and implies @option{-mfp16-format=ieee}.
15073 @option{-march=native} causes the compiler to auto-detect the architecture
15074 of the build computer. At present, this feature is only supported on
15075 GNU/Linux, and not all architectures are recognized. If the auto-detect
15076 is unsuccessful the option has no effect.
15078 @item -mtune=@var{name}
15080 This option specifies the name of the target ARM processor for
15081 which GCC should tune the performance of the code.
15082 For some ARM implementations better performance can be obtained by using
15084 Permissible names are: @samp{arm2}, @samp{arm250},
15085 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
15086 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
15087 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
15088 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
15090 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
15091 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
15092 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
15093 @samp{strongarm1110},
15094 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
15095 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
15096 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
15097 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
15098 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
15099 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
15100 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
15101 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
15102 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
15103 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
15104 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-r4},
15105 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
15113 @samp{cortex-m0plus},
15114 @samp{cortex-m1.small-multiply},
15115 @samp{cortex-m0.small-multiply},
15116 @samp{cortex-m0plus.small-multiply},
15120 @samp{marvell-pj4},
15121 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
15122 @samp{fa526}, @samp{fa626},
15123 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
15126 Additionally, this option can specify that GCC should tune the performance
15127 of the code for a big.LITTLE system. Permissible names are:
15128 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
15129 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15130 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}.
15132 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
15133 performance for a blend of processors within architecture @var{arch}.
15134 The aim is to generate code that run well on the current most popular
15135 processors, balancing between optimizations that benefit some CPUs in the
15136 range, and avoiding performance pitfalls of other CPUs. The effects of
15137 this option may change in future GCC versions as CPU models come and go.
15139 @option{-mtune=native} causes the compiler to auto-detect the CPU
15140 of the build computer. At present, this feature is only supported on
15141 GNU/Linux, and not all architectures are recognized. If the auto-detect is
15142 unsuccessful the option has no effect.
15144 @item -mcpu=@var{name}
15146 This specifies the name of the target ARM processor. GCC uses this name
15147 to derive the name of the target ARM architecture (as if specified
15148 by @option{-march}) and the ARM processor type for which to tune for
15149 performance (as if specified by @option{-mtune}). Where this option
15150 is used in conjunction with @option{-march} or @option{-mtune},
15151 those options take precedence over the appropriate part of this option.
15153 Permissible names for this option are the same as those for
15156 @option{-mcpu=generic-@var{arch}} is also permissible, and is
15157 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
15158 See @option{-mtune} for more information.
15160 @option{-mcpu=native} causes the compiler to auto-detect the CPU
15161 of the build computer. At present, this feature is only supported on
15162 GNU/Linux, and not all architectures are recognized. If the auto-detect
15163 is unsuccessful the option has no effect.
15165 @item -mfpu=@var{name}
15167 This specifies what floating-point hardware (or hardware emulation) is
15168 available on the target. Permissible names are: @samp{vfpv2}, @samp{vfpv3},
15169 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
15170 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
15171 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
15172 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
15173 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
15174 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
15175 is an alias for @samp{vfpv2}.
15177 If @option{-msoft-float} is specified this specifies the format of
15178 floating-point values.
15180 If the selected floating-point hardware includes the NEON extension
15181 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
15182 operations are not generated by GCC's auto-vectorization pass unless
15183 @option{-funsafe-math-optimizations} is also specified. This is
15184 because NEON hardware does not fully implement the IEEE 754 standard for
15185 floating-point arithmetic (in particular denormal values are treated as
15186 zero), so the use of NEON instructions may lead to a loss of precision.
15188 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}).
15190 @item -mfp16-format=@var{name}
15191 @opindex mfp16-format
15192 Specify the format of the @code{__fp16} half-precision floating-point type.
15193 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
15194 the default is @samp{none}, in which case the @code{__fp16} type is not
15195 defined. @xref{Half-Precision}, for more information.
15197 @item -mstructure-size-boundary=@var{n}
15198 @opindex mstructure-size-boundary
15199 The sizes of all structures and unions are rounded up to a multiple
15200 of the number of bits set by this option. Permissible values are 8, 32
15201 and 64. The default value varies for different toolchains. For the COFF
15202 targeted toolchain the default value is 8. A value of 64 is only allowed
15203 if the underlying ABI supports it.
15205 Specifying a larger number can produce faster, more efficient code, but
15206 can also increase the size of the program. Different values are potentially
15207 incompatible. Code compiled with one value cannot necessarily expect to
15208 work with code or libraries compiled with another value, if they exchange
15209 information using structures or unions.
15211 @item -mabort-on-noreturn
15212 @opindex mabort-on-noreturn
15213 Generate a call to the function @code{abort} at the end of a
15214 @code{noreturn} function. It is executed if the function tries to
15218 @itemx -mno-long-calls
15219 @opindex mlong-calls
15220 @opindex mno-long-calls
15221 Tells the compiler to perform function calls by first loading the
15222 address of the function into a register and then performing a subroutine
15223 call on this register. This switch is needed if the target function
15224 lies outside of the 64-megabyte addressing range of the offset-based
15225 version of subroutine call instruction.
15227 Even if this switch is enabled, not all function calls are turned
15228 into long calls. The heuristic is that static functions, functions
15229 that have the @code{short_call} attribute, functions that are inside
15230 the scope of a @code{#pragma no_long_calls} directive, and functions whose
15231 definitions have already been compiled within the current compilation
15232 unit are not turned into long calls. The exceptions to this rule are
15233 that weak function definitions, functions with the @code{long_call}
15234 attribute or the @code{section} attribute, and functions that are within
15235 the scope of a @code{#pragma long_calls} directive are always
15236 turned into long calls.
15238 This feature is not enabled by default. Specifying
15239 @option{-mno-long-calls} restores the default behavior, as does
15240 placing the function calls within the scope of a @code{#pragma
15241 long_calls_off} directive. Note these switches have no effect on how
15242 the compiler generates code to handle function calls via function
15245 @item -msingle-pic-base
15246 @opindex msingle-pic-base
15247 Treat the register used for PIC addressing as read-only, rather than
15248 loading it in the prologue for each function. The runtime system is
15249 responsible for initializing this register with an appropriate value
15250 before execution begins.
15252 @item -mpic-register=@var{reg}
15253 @opindex mpic-register
15254 Specify the register to be used for PIC addressing.
15255 For standard PIC base case, the default is any suitable register
15256 determined by compiler. For single PIC base case, the default is
15257 @samp{R9} if target is EABI based or stack-checking is enabled,
15258 otherwise the default is @samp{R10}.
15260 @item -mpic-data-is-text-relative
15261 @opindex mpic-data-is-text-relative
15262 Assume that the displacement between the text and data segments is fixed
15263 at static link time. This permits using PC-relative addressing
15264 operations to access data known to be in the data segment. For
15265 non-VxWorks RTP targets, this option is enabled by default. When
15266 disabled on such targets, it will enable @option{-msingle-pic-base} by
15269 @item -mpoke-function-name
15270 @opindex mpoke-function-name
15271 Write the name of each function into the text section, directly
15272 preceding the function prologue. The generated code is similar to this:
15276 .ascii "arm_poke_function_name", 0
15279 .word 0xff000000 + (t1 - t0)
15280 arm_poke_function_name
15282 stmfd sp!, @{fp, ip, lr, pc@}
15286 When performing a stack backtrace, code can inspect the value of
15287 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
15288 location @code{pc - 12} and the top 8 bits are set, then we know that
15289 there is a function name embedded immediately preceding this location
15290 and has length @code{((pc[-3]) & 0xff000000)}.
15297 Select between generating code that executes in ARM and Thumb
15298 states. The default for most configurations is to generate code
15299 that executes in ARM state, but the default can be changed by
15300 configuring GCC with the @option{--with-mode=}@var{state}
15303 You can also override the ARM and Thumb mode for each function
15304 by using the @code{target("thumb")} and @code{target("arm")} function attributes
15305 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
15308 @opindex mtpcs-frame
15309 Generate a stack frame that is compliant with the Thumb Procedure Call
15310 Standard for all non-leaf functions. (A leaf function is one that does
15311 not call any other functions.) The default is @option{-mno-tpcs-frame}.
15313 @item -mtpcs-leaf-frame
15314 @opindex mtpcs-leaf-frame
15315 Generate a stack frame that is compliant with the Thumb Procedure Call
15316 Standard for all leaf functions. (A leaf function is one that does
15317 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
15319 @item -mcallee-super-interworking
15320 @opindex mcallee-super-interworking
15321 Gives all externally visible functions in the file being compiled an ARM
15322 instruction set header which switches to Thumb mode before executing the
15323 rest of the function. This allows these functions to be called from
15324 non-interworking code. This option is not valid in AAPCS configurations
15325 because interworking is enabled by default.
15327 @item -mcaller-super-interworking
15328 @opindex mcaller-super-interworking
15329 Allows calls via function pointers (including virtual functions) to
15330 execute correctly regardless of whether the target code has been
15331 compiled for interworking or not. There is a small overhead in the cost
15332 of executing a function pointer if this option is enabled. This option
15333 is not valid in AAPCS configurations because interworking is enabled
15336 @item -mtp=@var{name}
15338 Specify the access model for the thread local storage pointer. The valid
15339 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
15340 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
15341 (supported in the arm6k architecture), and @samp{auto}, which uses the
15342 best available method for the selected processor. The default setting is
15345 @item -mtls-dialect=@var{dialect}
15346 @opindex mtls-dialect
15347 Specify the dialect to use for accessing thread local storage. Two
15348 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
15349 @samp{gnu} dialect selects the original GNU scheme for supporting
15350 local and global dynamic TLS models. The @samp{gnu2} dialect
15351 selects the GNU descriptor scheme, which provides better performance
15352 for shared libraries. The GNU descriptor scheme is compatible with
15353 the original scheme, but does require new assembler, linker and
15354 library support. Initial and local exec TLS models are unaffected by
15355 this option and always use the original scheme.
15357 @item -mword-relocations
15358 @opindex mword-relocations
15359 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
15360 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
15361 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
15364 @item -mfix-cortex-m3-ldrd
15365 @opindex mfix-cortex-m3-ldrd
15366 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
15367 with overlapping destination and base registers are used. This option avoids
15368 generating these instructions. This option is enabled by default when
15369 @option{-mcpu=cortex-m3} is specified.
15371 @item -munaligned-access
15372 @itemx -mno-unaligned-access
15373 @opindex munaligned-access
15374 @opindex mno-unaligned-access
15375 Enables (or disables) reading and writing of 16- and 32- bit values
15376 from addresses that are not 16- or 32- bit aligned. By default
15377 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
15378 ARMv8-M Baseline architectures, and enabled for all other
15379 architectures. If unaligned access is not enabled then words in packed
15380 data structures are accessed a byte at a time.
15382 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
15383 generated object file to either true or false, depending upon the
15384 setting of this option. If unaligned access is enabled then the
15385 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
15388 @item -mneon-for-64bits
15389 @opindex mneon-for-64bits
15390 Enables using Neon to handle scalar 64-bits operations. This is
15391 disabled by default since the cost of moving data from core registers
15394 @item -mslow-flash-data
15395 @opindex mslow-flash-data
15396 Assume loading data from flash is slower than fetching instruction.
15397 Therefore literal load is minimized for better performance.
15398 This option is only supported when compiling for ARMv7 M-profile and
15401 @item -masm-syntax-unified
15402 @opindex masm-syntax-unified
15403 Assume inline assembler is using unified asm syntax. The default is
15404 currently off which implies divided syntax. This option has no impact
15405 on Thumb2. However, this may change in future releases of GCC.
15406 Divided syntax should be considered deprecated.
15408 @item -mrestrict-it
15409 @opindex mrestrict-it
15410 Restricts generation of IT blocks to conform to the rules of ARMv8.
15411 IT blocks can only contain a single 16-bit instruction from a select
15412 set of instructions. This option is on by default for ARMv8 Thumb mode.
15414 @item -mprint-tune-info
15415 @opindex mprint-tune-info
15416 Print CPU tuning information as comment in assembler file. This is
15417 an option used only for regression testing of the compiler and not
15418 intended for ordinary use in compiling code. This option is disabled
15422 @opindex mpure-code
15423 Do not allow constant data to be placed in code sections.
15424 Additionally, when compiling for ELF object format give all text sections the
15425 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
15426 is only available when generating non-pic code for M-profile targets with the
15431 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
15432 Development Tools Engineering Specification", which can be found on
15433 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
15437 @subsection AVR Options
15438 @cindex AVR Options
15440 These options are defined for AVR implementations:
15443 @item -mmcu=@var{mcu}
15445 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
15447 The default for this option is@tie{}@samp{avr2}.
15449 GCC supports the following AVR devices and ISAs:
15451 @include avr-mmcu.texi
15456 Assume that all data in static storage can be accessed by LDS / STS
15457 instructions. This option has only an effect on reduced Tiny devices like
15458 ATtiny40. See also the @code{absdata}
15459 @ref{AVR Variable Attributes,variable attribute}.
15461 @item -maccumulate-args
15462 @opindex maccumulate-args
15463 Accumulate outgoing function arguments and acquire/release the needed
15464 stack space for outgoing function arguments once in function
15465 prologue/epilogue. Without this option, outgoing arguments are pushed
15466 before calling a function and popped afterwards.
15468 Popping the arguments after the function call can be expensive on
15469 AVR so that accumulating the stack space might lead to smaller
15470 executables because arguments need not be removed from the
15471 stack after such a function call.
15473 This option can lead to reduced code size for functions that perform
15474 several calls to functions that get their arguments on the stack like
15475 calls to printf-like functions.
15477 @item -mbranch-cost=@var{cost}
15478 @opindex mbranch-cost
15479 Set the branch costs for conditional branch instructions to
15480 @var{cost}. Reasonable values for @var{cost} are small, non-negative
15481 integers. The default branch cost is 0.
15483 @item -mcall-prologues
15484 @opindex mcall-prologues
15485 Functions prologues/epilogues are expanded as calls to appropriate
15486 subroutines. Code size is smaller.
15490 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
15491 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
15492 and @code{long long} is 4 bytes. Please note that this option does not
15493 conform to the C standards, but it results in smaller code
15496 @item -mn-flash=@var{num}
15498 Assume that the flash memory has a size of
15499 @var{num} times 64@tie{}KiB.
15501 @item -mno-interrupts
15502 @opindex mno-interrupts
15503 Generated code is not compatible with hardware interrupts.
15504 Code size is smaller.
15508 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
15509 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
15510 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
15511 the assembler's command line and the @option{--relax} option to the
15512 linker's command line.
15514 Jump relaxing is performed by the linker because jump offsets are not
15515 known before code is located. Therefore, the assembler code generated by the
15516 compiler is the same, but the instructions in the executable may
15517 differ from instructions in the assembler code.
15519 Relaxing must be turned on if linker stubs are needed, see the
15520 section on @code{EIND} and linker stubs below.
15524 Assume that the device supports the Read-Modify-Write
15525 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
15529 Treat the stack pointer register as an 8-bit register,
15530 i.e.@: assume the high byte of the stack pointer is zero.
15531 In general, you don't need to set this option by hand.
15533 This option is used internally by the compiler to select and
15534 build multilibs for architectures @code{avr2} and @code{avr25}.
15535 These architectures mix devices with and without @code{SPH}.
15536 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
15537 the compiler driver adds or removes this option from the compiler
15538 proper's command line, because the compiler then knows if the device
15539 or architecture has an 8-bit stack pointer and thus no @code{SPH}
15544 Use address register @code{X} in a way proposed by the hardware. This means
15545 that @code{X} is only used in indirect, post-increment or
15546 pre-decrement addressing.
15548 Without this option, the @code{X} register may be used in the same way
15549 as @code{Y} or @code{Z} which then is emulated by additional
15551 For example, loading a value with @code{X+const} addressing with a
15552 small non-negative @code{const < 64} to a register @var{Rn} is
15556 adiw r26, const ; X += const
15557 ld @var{Rn}, X ; @var{Rn} = *X
15558 sbiw r26, const ; X -= const
15562 @opindex mtiny-stack
15563 Only change the lower 8@tie{}bits of the stack pointer.
15565 @item -mfract-convert-truncate
15566 @opindex mfract-convert-truncate
15567 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
15570 @opindex nodevicelib
15571 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
15573 @item -Waddr-space-convert
15574 @opindex Waddr-space-convert
15575 Warn about conversions between address spaces in the case where the
15576 resulting address space is not contained in the incoming address space.
15578 @item -Wmisspelled-isr
15579 @opindex Wmisspelled-isr
15580 Warn if the ISR is misspelled, i.e. without __vector prefix.
15581 Enabled by default.
15584 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
15585 @cindex @code{EIND}
15586 Pointers in the implementation are 16@tie{}bits wide.
15587 The address of a function or label is represented as word address so
15588 that indirect jumps and calls can target any code address in the
15589 range of 64@tie{}Ki words.
15591 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
15592 bytes of program memory space, there is a special function register called
15593 @code{EIND} that serves as most significant part of the target address
15594 when @code{EICALL} or @code{EIJMP} instructions are used.
15596 Indirect jumps and calls on these devices are handled as follows by
15597 the compiler and are subject to some limitations:
15602 The compiler never sets @code{EIND}.
15605 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
15606 instructions or might read @code{EIND} directly in order to emulate an
15607 indirect call/jump by means of a @code{RET} instruction.
15610 The compiler assumes that @code{EIND} never changes during the startup
15611 code or during the application. In particular, @code{EIND} is not
15612 saved/restored in function or interrupt service routine
15616 For indirect calls to functions and computed goto, the linker
15617 generates @emph{stubs}. Stubs are jump pads sometimes also called
15618 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
15619 The stub contains a direct jump to the desired address.
15622 Linker relaxation must be turned on so that the linker generates
15623 the stubs correctly in all situations. See the compiler option
15624 @option{-mrelax} and the linker option @option{--relax}.
15625 There are corner cases where the linker is supposed to generate stubs
15626 but aborts without relaxation and without a helpful error message.
15629 The default linker script is arranged for code with @code{EIND = 0}.
15630 If code is supposed to work for a setup with @code{EIND != 0}, a custom
15631 linker script has to be used in order to place the sections whose
15632 name start with @code{.trampolines} into the segment where @code{EIND}
15636 The startup code from libgcc never sets @code{EIND}.
15637 Notice that startup code is a blend of code from libgcc and AVR-LibC.
15638 For the impact of AVR-LibC on @code{EIND}, see the
15639 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
15642 It is legitimate for user-specific startup code to set up @code{EIND}
15643 early, for example by means of initialization code located in
15644 section @code{.init3}. Such code runs prior to general startup code
15645 that initializes RAM and calls constructors, but after the bit
15646 of startup code from AVR-LibC that sets @code{EIND} to the segment
15647 where the vector table is located.
15649 #include <avr/io.h>
15652 __attribute__((section(".init3"),naked,used,no_instrument_function))
15653 init3_set_eind (void)
15655 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
15656 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
15661 The @code{__trampolines_start} symbol is defined in the linker script.
15664 Stubs are generated automatically by the linker if
15665 the following two conditions are met:
15668 @item The address of a label is taken by means of the @code{gs} modifier
15669 (short for @emph{generate stubs}) like so:
15671 LDI r24, lo8(gs(@var{func}))
15672 LDI r25, hi8(gs(@var{func}))
15674 @item The final location of that label is in a code segment
15675 @emph{outside} the segment where the stubs are located.
15679 The compiler emits such @code{gs} modifiers for code labels in the
15680 following situations:
15682 @item Taking address of a function or code label.
15683 @item Computed goto.
15684 @item If prologue-save function is used, see @option{-mcall-prologues}
15685 command-line option.
15686 @item Switch/case dispatch tables. If you do not want such dispatch
15687 tables you can specify the @option{-fno-jump-tables} command-line option.
15688 @item C and C++ constructors/destructors called during startup/shutdown.
15689 @item If the tools hit a @code{gs()} modifier explained above.
15693 Jumping to non-symbolic addresses like so is @emph{not} supported:
15698 /* Call function at word address 0x2 */
15699 return ((int(*)(void)) 0x2)();
15703 Instead, a stub has to be set up, i.e.@: the function has to be called
15704 through a symbol (@code{func_4} in the example):
15709 extern int func_4 (void);
15711 /* Call function at byte address 0x4 */
15716 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
15717 Alternatively, @code{func_4} can be defined in the linker script.
15720 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
15721 @cindex @code{RAMPD}
15722 @cindex @code{RAMPX}
15723 @cindex @code{RAMPY}
15724 @cindex @code{RAMPZ}
15725 Some AVR devices support memories larger than the 64@tie{}KiB range
15726 that can be accessed with 16-bit pointers. To access memory locations
15727 outside this 64@tie{}KiB range, the content of a @code{RAMP}
15728 register is used as high part of the address:
15729 The @code{X}, @code{Y}, @code{Z} address register is concatenated
15730 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
15731 register, respectively, to get a wide address. Similarly,
15732 @code{RAMPD} is used together with direct addressing.
15736 The startup code initializes the @code{RAMP} special function
15737 registers with zero.
15740 If a @ref{AVR Named Address Spaces,named address space} other than
15741 generic or @code{__flash} is used, then @code{RAMPZ} is set
15742 as needed before the operation.
15745 If the device supports RAM larger than 64@tie{}KiB and the compiler
15746 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
15747 is reset to zero after the operation.
15750 If the device comes with a specific @code{RAMP} register, the ISR
15751 prologue/epilogue saves/restores that SFR and initializes it with
15752 zero in case the ISR code might (implicitly) use it.
15755 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
15756 If you use inline assembler to read from locations outside the
15757 16-bit address range and change one of the @code{RAMP} registers,
15758 you must reset it to zero after the access.
15762 @subsubsection AVR Built-in Macros
15764 GCC defines several built-in macros so that the user code can test
15765 for the presence or absence of features. Almost any of the following
15766 built-in macros are deduced from device capabilities and thus
15767 triggered by the @option{-mmcu=} command-line option.
15769 For even more AVR-specific built-in macros see
15770 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
15775 Build-in macro that resolves to a decimal number that identifies the
15776 architecture and depends on the @option{-mmcu=@var{mcu}} option.
15777 Possible values are:
15779 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
15780 @code{4}, @code{5}, @code{51}, @code{6}
15782 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
15783 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
15787 @code{100}, @code{102}, @code{104},
15788 @code{105}, @code{106}, @code{107}
15790 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
15791 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
15792 If @var{mcu} specifies a device, this built-in macro is set
15793 accordingly. For example, with @option{-mmcu=atmega8} the macro is
15794 defined to @code{4}.
15796 @item __AVR_@var{Device}__
15797 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
15798 the device's name. For example, @option{-mmcu=atmega8} defines the
15799 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
15800 @code{__AVR_ATtiny261A__}, etc.
15802 The built-in macros' names follow
15803 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
15804 the device name as from the AVR user manual. The difference between
15805 @var{Device} in the built-in macro and @var{device} in
15806 @option{-mmcu=@var{device}} is that the latter is always lowercase.
15808 If @var{device} is not a device but only a core architecture like
15809 @samp{avr51}, this macro is not defined.
15811 @item __AVR_DEVICE_NAME__
15812 Setting @option{-mmcu=@var{device}} defines this built-in macro to
15813 the device's name. For example, with @option{-mmcu=atmega8} the macro
15814 is defined to @code{atmega8}.
15816 If @var{device} is not a device but only a core architecture like
15817 @samp{avr51}, this macro is not defined.
15819 @item __AVR_XMEGA__
15820 The device / architecture belongs to the XMEGA family of devices.
15822 @item __AVR_HAVE_ELPM__
15823 The device has the @code{ELPM} instruction.
15825 @item __AVR_HAVE_ELPMX__
15826 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
15827 R@var{n},Z+} instructions.
15829 @item __AVR_HAVE_MOVW__
15830 The device has the @code{MOVW} instruction to perform 16-bit
15831 register-register moves.
15833 @item __AVR_HAVE_LPMX__
15834 The device has the @code{LPM R@var{n},Z} and
15835 @code{LPM R@var{n},Z+} instructions.
15837 @item __AVR_HAVE_MUL__
15838 The device has a hardware multiplier.
15840 @item __AVR_HAVE_JMP_CALL__
15841 The device has the @code{JMP} and @code{CALL} instructions.
15842 This is the case for devices with at least 16@tie{}KiB of program
15845 @item __AVR_HAVE_EIJMP_EICALL__
15846 @itemx __AVR_3_BYTE_PC__
15847 The device has the @code{EIJMP} and @code{EICALL} instructions.
15848 This is the case for devices with more than 128@tie{}KiB of program memory.
15849 This also means that the program counter
15850 (PC) is 3@tie{}bytes wide.
15852 @item __AVR_2_BYTE_PC__
15853 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
15854 with up to 128@tie{}KiB of program memory.
15856 @item __AVR_HAVE_8BIT_SP__
15857 @itemx __AVR_HAVE_16BIT_SP__
15858 The stack pointer (SP) register is treated as 8-bit respectively
15859 16-bit register by the compiler.
15860 The definition of these macros is affected by @option{-mtiny-stack}.
15862 @item __AVR_HAVE_SPH__
15864 The device has the SPH (high part of stack pointer) special function
15865 register or has an 8-bit stack pointer, respectively.
15866 The definition of these macros is affected by @option{-mmcu=} and
15867 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
15870 @item __AVR_HAVE_RAMPD__
15871 @itemx __AVR_HAVE_RAMPX__
15872 @itemx __AVR_HAVE_RAMPY__
15873 @itemx __AVR_HAVE_RAMPZ__
15874 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
15875 @code{RAMPZ} special function register, respectively.
15877 @item __NO_INTERRUPTS__
15878 This macro reflects the @option{-mno-interrupts} command-line option.
15880 @item __AVR_ERRATA_SKIP__
15881 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
15882 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
15883 instructions because of a hardware erratum. Skip instructions are
15884 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
15885 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
15888 @item __AVR_ISA_RMW__
15889 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
15891 @item __AVR_SFR_OFFSET__=@var{offset}
15892 Instructions that can address I/O special function registers directly
15893 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
15894 address as if addressed by an instruction to access RAM like @code{LD}
15895 or @code{STS}. This offset depends on the device architecture and has
15896 to be subtracted from the RAM address in order to get the
15897 respective I/O@tie{}address.
15899 @item __WITH_AVRLIBC__
15900 The compiler is configured to be used together with AVR-Libc.
15901 See the @option{--with-avrlibc} configure option.
15905 @node Blackfin Options
15906 @subsection Blackfin Options
15907 @cindex Blackfin Options
15910 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
15912 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
15913 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
15914 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
15915 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
15916 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
15917 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
15918 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
15919 @samp{bf561}, @samp{bf592}.
15921 The optional @var{sirevision} specifies the silicon revision of the target
15922 Blackfin processor. Any workarounds available for the targeted silicon revision
15923 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
15924 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
15925 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
15926 hexadecimal digits representing the major and minor numbers in the silicon
15927 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
15928 is not defined. If @var{sirevision} is @samp{any}, the
15929 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
15930 If this optional @var{sirevision} is not used, GCC assumes the latest known
15931 silicon revision of the targeted Blackfin processor.
15933 GCC defines a preprocessor macro for the specified @var{cpu}.
15934 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
15935 provided by libgloss to be linked in if @option{-msim} is not given.
15937 Without this option, @samp{bf532} is used as the processor by default.
15939 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
15940 only the preprocessor macro is defined.
15944 Specifies that the program will be run on the simulator. This causes
15945 the simulator BSP provided by libgloss to be linked in. This option
15946 has effect only for @samp{bfin-elf} toolchain.
15947 Certain other options, such as @option{-mid-shared-library} and
15948 @option{-mfdpic}, imply @option{-msim}.
15950 @item -momit-leaf-frame-pointer
15951 @opindex momit-leaf-frame-pointer
15952 Don't keep the frame pointer in a register for leaf functions. This
15953 avoids the instructions to save, set up and restore frame pointers and
15954 makes an extra register available in leaf functions. The option
15955 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
15956 which might make debugging harder.
15958 @item -mspecld-anomaly
15959 @opindex mspecld-anomaly
15960 When enabled, the compiler ensures that the generated code does not
15961 contain speculative loads after jump instructions. If this option is used,
15962 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
15964 @item -mno-specld-anomaly
15965 @opindex mno-specld-anomaly
15966 Don't generate extra code to prevent speculative loads from occurring.
15968 @item -mcsync-anomaly
15969 @opindex mcsync-anomaly
15970 When enabled, the compiler ensures that the generated code does not
15971 contain CSYNC or SSYNC instructions too soon after conditional branches.
15972 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
15974 @item -mno-csync-anomaly
15975 @opindex mno-csync-anomaly
15976 Don't generate extra code to prevent CSYNC or SSYNC instructions from
15977 occurring too soon after a conditional branch.
15981 When enabled, the compiler is free to take advantage of the knowledge that
15982 the entire program fits into the low 64k of memory.
15985 @opindex mno-low-64k
15986 Assume that the program is arbitrarily large. This is the default.
15988 @item -mstack-check-l1
15989 @opindex mstack-check-l1
15990 Do stack checking using information placed into L1 scratchpad memory by the
15993 @item -mid-shared-library
15994 @opindex mid-shared-library
15995 Generate code that supports shared libraries via the library ID method.
15996 This allows for execute in place and shared libraries in an environment
15997 without virtual memory management. This option implies @option{-fPIC}.
15998 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16000 @item -mno-id-shared-library
16001 @opindex mno-id-shared-library
16002 Generate code that doesn't assume ID-based shared libraries are being used.
16003 This is the default.
16005 @item -mleaf-id-shared-library
16006 @opindex mleaf-id-shared-library
16007 Generate code that supports shared libraries via the library ID method,
16008 but assumes that this library or executable won't link against any other
16009 ID shared libraries. That allows the compiler to use faster code for jumps
16012 @item -mno-leaf-id-shared-library
16013 @opindex mno-leaf-id-shared-library
16014 Do not assume that the code being compiled won't link against any ID shared
16015 libraries. Slower code is generated for jump and call insns.
16017 @item -mshared-library-id=n
16018 @opindex mshared-library-id
16019 Specifies the identification number of the ID-based shared library being
16020 compiled. Specifying a value of 0 generates more compact code; specifying
16021 other values forces the allocation of that number to the current
16022 library but is no more space- or time-efficient than omitting this option.
16026 Generate code that allows the data segment to be located in a different
16027 area of memory from the text segment. This allows for execute in place in
16028 an environment without virtual memory management by eliminating relocations
16029 against the text section.
16031 @item -mno-sep-data
16032 @opindex mno-sep-data
16033 Generate code that assumes that the data segment follows the text segment.
16034 This is the default.
16037 @itemx -mno-long-calls
16038 @opindex mlong-calls
16039 @opindex mno-long-calls
16040 Tells the compiler to perform function calls by first loading the
16041 address of the function into a register and then performing a subroutine
16042 call on this register. This switch is needed if the target function
16043 lies outside of the 24-bit addressing range of the offset-based
16044 version of subroutine call instruction.
16046 This feature is not enabled by default. Specifying
16047 @option{-mno-long-calls} restores the default behavior. Note these
16048 switches have no effect on how the compiler generates code to handle
16049 function calls via function pointers.
16053 Link with the fast floating-point library. This library relaxes some of
16054 the IEEE floating-point standard's rules for checking inputs against
16055 Not-a-Number (NAN), in the interest of performance.
16058 @opindex minline-plt
16059 Enable inlining of PLT entries in function calls to functions that are
16060 not known to bind locally. It has no effect without @option{-mfdpic}.
16063 @opindex mmulticore
16064 Build a standalone application for multicore Blackfin processors.
16065 This option causes proper start files and link scripts supporting
16066 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
16067 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
16069 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
16070 selects the one-application-per-core programming model. Without
16071 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
16072 programming model is used. In this model, the main function of Core B
16073 should be named as @code{coreb_main}.
16075 If this option is not used, the single-core application programming
16080 Build a standalone application for Core A of BF561 when using
16081 the one-application-per-core programming model. Proper start files
16082 and link scripts are used to support Core A, and the macro
16083 @code{__BFIN_COREA} is defined.
16084 This option can only be used in conjunction with @option{-mmulticore}.
16088 Build a standalone application for Core B of BF561 when using
16089 the one-application-per-core programming model. Proper start files
16090 and link scripts are used to support Core B, and the macro
16091 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
16092 should be used instead of @code{main}.
16093 This option can only be used in conjunction with @option{-mmulticore}.
16097 Build a standalone application for SDRAM. Proper start files and
16098 link scripts are used to put the application into SDRAM, and the macro
16099 @code{__BFIN_SDRAM} is defined.
16100 The loader should initialize SDRAM before loading the application.
16104 Assume that ICPLBs are enabled at run time. This has an effect on certain
16105 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
16106 are enabled; for standalone applications the default is off.
16110 @subsection C6X Options
16111 @cindex C6X Options
16114 @item -march=@var{name}
16116 This specifies the name of the target architecture. GCC uses this
16117 name to determine what kind of instructions it can emit when generating
16118 assembly code. Permissible names are: @samp{c62x},
16119 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
16122 @opindex mbig-endian
16123 Generate code for a big-endian target.
16125 @item -mlittle-endian
16126 @opindex mlittle-endian
16127 Generate code for a little-endian target. This is the default.
16131 Choose startup files and linker script suitable for the simulator.
16133 @item -msdata=default
16134 @opindex msdata=default
16135 Put small global and static data in the @code{.neardata} section,
16136 which is pointed to by register @code{B14}. Put small uninitialized
16137 global and static data in the @code{.bss} section, which is adjacent
16138 to the @code{.neardata} section. Put small read-only data into the
16139 @code{.rodata} section. The corresponding sections used for large
16140 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
16143 @opindex msdata=all
16144 Put all data, not just small objects, into the sections reserved for
16145 small data, and use addressing relative to the @code{B14} register to
16149 @opindex msdata=none
16150 Make no use of the sections reserved for small data, and use absolute
16151 addresses to access all data. Put all initialized global and static
16152 data in the @code{.fardata} section, and all uninitialized data in the
16153 @code{.far} section. Put all constant data into the @code{.const}
16158 @subsection CRIS Options
16159 @cindex CRIS Options
16161 These options are defined specifically for the CRIS ports.
16164 @item -march=@var{architecture-type}
16165 @itemx -mcpu=@var{architecture-type}
16168 Generate code for the specified architecture. The choices for
16169 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
16170 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
16171 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
16174 @item -mtune=@var{architecture-type}
16176 Tune to @var{architecture-type} everything applicable about the generated
16177 code, except for the ABI and the set of available instructions. The
16178 choices for @var{architecture-type} are the same as for
16179 @option{-march=@var{architecture-type}}.
16181 @item -mmax-stack-frame=@var{n}
16182 @opindex mmax-stack-frame
16183 Warn when the stack frame of a function exceeds @var{n} bytes.
16189 The options @option{-metrax4} and @option{-metrax100} are synonyms for
16190 @option{-march=v3} and @option{-march=v8} respectively.
16192 @item -mmul-bug-workaround
16193 @itemx -mno-mul-bug-workaround
16194 @opindex mmul-bug-workaround
16195 @opindex mno-mul-bug-workaround
16196 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
16197 models where it applies. This option is active by default.
16201 Enable CRIS-specific verbose debug-related information in the assembly
16202 code. This option also has the effect of turning off the @samp{#NO_APP}
16203 formatted-code indicator to the assembler at the beginning of the
16208 Do not use condition-code results from previous instruction; always emit
16209 compare and test instructions before use of condition codes.
16211 @item -mno-side-effects
16212 @opindex mno-side-effects
16213 Do not emit instructions with side effects in addressing modes other than
16216 @item -mstack-align
16217 @itemx -mno-stack-align
16218 @itemx -mdata-align
16219 @itemx -mno-data-align
16220 @itemx -mconst-align
16221 @itemx -mno-const-align
16222 @opindex mstack-align
16223 @opindex mno-stack-align
16224 @opindex mdata-align
16225 @opindex mno-data-align
16226 @opindex mconst-align
16227 @opindex mno-const-align
16228 These options (@samp{no-} options) arrange (eliminate arrangements) for the
16229 stack frame, individual data and constants to be aligned for the maximum
16230 single data access size for the chosen CPU model. The default is to
16231 arrange for 32-bit alignment. ABI details such as structure layout are
16232 not affected by these options.
16240 Similar to the stack- data- and const-align options above, these options
16241 arrange for stack frame, writable data and constants to all be 32-bit,
16242 16-bit or 8-bit aligned. The default is 32-bit alignment.
16244 @item -mno-prologue-epilogue
16245 @itemx -mprologue-epilogue
16246 @opindex mno-prologue-epilogue
16247 @opindex mprologue-epilogue
16248 With @option{-mno-prologue-epilogue}, the normal function prologue and
16249 epilogue which set up the stack frame are omitted and no return
16250 instructions or return sequences are generated in the code. Use this
16251 option only together with visual inspection of the compiled code: no
16252 warnings or errors are generated when call-saved registers must be saved,
16253 or storage for local variables needs to be allocated.
16257 @opindex mno-gotplt
16259 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
16260 instruction sequences that load addresses for functions from the PLT part
16261 of the GOT rather than (traditional on other architectures) calls to the
16262 PLT@. The default is @option{-mgotplt}.
16266 Legacy no-op option only recognized with the cris-axis-elf and
16267 cris-axis-linux-gnu targets.
16271 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
16275 This option, recognized for the cris-axis-elf, arranges
16276 to link with input-output functions from a simulator library. Code,
16277 initialized data and zero-initialized data are allocated consecutively.
16281 Like @option{-sim}, but pass linker options to locate initialized data at
16282 0x40000000 and zero-initialized data at 0x80000000.
16286 @subsection CR16 Options
16287 @cindex CR16 Options
16289 These options are defined specifically for the CR16 ports.
16295 Enable the use of multiply-accumulate instructions. Disabled by default.
16299 @opindex mcr16cplus
16301 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
16306 Links the library libsim.a which is in compatible with simulator. Applicable
16307 to ELF compiler only.
16311 Choose integer type as 32-bit wide.
16315 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
16317 @item -mdata-model=@var{model}
16318 @opindex mdata-model
16319 Choose a data model. The choices for @var{model} are @samp{near},
16320 @samp{far} or @samp{medium}. @samp{medium} is default.
16321 However, @samp{far} is not valid with @option{-mcr16c}, as the
16322 CR16C architecture does not support the far data model.
16325 @node Darwin Options
16326 @subsection Darwin Options
16327 @cindex Darwin options
16329 These options are defined for all architectures running the Darwin operating
16332 FSF GCC on Darwin does not create ``fat'' object files; it creates
16333 an object file for the single architecture that GCC was built to
16334 target. Apple's GCC on Darwin does create ``fat'' files if multiple
16335 @option{-arch} options are used; it does so by running the compiler or
16336 linker multiple times and joining the results together with
16339 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
16340 @samp{i686}) is determined by the flags that specify the ISA
16341 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
16342 @option{-force_cpusubtype_ALL} option can be used to override this.
16344 The Darwin tools vary in their behavior when presented with an ISA
16345 mismatch. The assembler, @file{as}, only permits instructions to
16346 be used that are valid for the subtype of the file it is generating,
16347 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
16348 The linker for shared libraries, @file{/usr/bin/libtool}, fails
16349 and prints an error if asked to create a shared library with a less
16350 restrictive subtype than its input files (for instance, trying to put
16351 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
16352 for executables, @command{ld}, quietly gives the executable the most
16353 restrictive subtype of any of its input files.
16358 Add the framework directory @var{dir} to the head of the list of
16359 directories to be searched for header files. These directories are
16360 interleaved with those specified by @option{-I} options and are
16361 scanned in a left-to-right order.
16363 A framework directory is a directory with frameworks in it. A
16364 framework is a directory with a @file{Headers} and/or
16365 @file{PrivateHeaders} directory contained directly in it that ends
16366 in @file{.framework}. The name of a framework is the name of this
16367 directory excluding the @file{.framework}. Headers associated with
16368 the framework are found in one of those two directories, with
16369 @file{Headers} being searched first. A subframework is a framework
16370 directory that is in a framework's @file{Frameworks} directory.
16371 Includes of subframework headers can only appear in a header of a
16372 framework that contains the subframework, or in a sibling subframework
16373 header. Two subframeworks are siblings if they occur in the same
16374 framework. A subframework should not have the same name as a
16375 framework; a warning is issued if this is violated. Currently a
16376 subframework cannot have subframeworks; in the future, the mechanism
16377 may be extended to support this. The standard frameworks can be found
16378 in @file{/System/Library/Frameworks} and
16379 @file{/Library/Frameworks}. An example include looks like
16380 @code{#include <Framework/header.h>}, where @file{Framework} denotes
16381 the name of the framework and @file{header.h} is found in the
16382 @file{PrivateHeaders} or @file{Headers} directory.
16384 @item -iframework@var{dir}
16385 @opindex iframework
16386 Like @option{-F} except the directory is a treated as a system
16387 directory. The main difference between this @option{-iframework} and
16388 @option{-F} is that with @option{-iframework} the compiler does not
16389 warn about constructs contained within header files found via
16390 @var{dir}. This option is valid only for the C family of languages.
16394 Emit debugging information for symbols that are used. For stabs
16395 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
16396 This is by default ON@.
16400 Emit debugging information for all symbols and types.
16402 @item -mmacosx-version-min=@var{version}
16403 The earliest version of MacOS X that this executable will run on
16404 is @var{version}. Typical values of @var{version} include @code{10.1},
16405 @code{10.2}, and @code{10.3.9}.
16407 If the compiler was built to use the system's headers by default,
16408 then the default for this option is the system version on which the
16409 compiler is running, otherwise the default is to make choices that
16410 are compatible with as many systems and code bases as possible.
16414 Enable kernel development mode. The @option{-mkernel} option sets
16415 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
16416 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
16417 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
16418 applicable. This mode also sets @option{-mno-altivec},
16419 @option{-msoft-float}, @option{-fno-builtin} and
16420 @option{-mlong-branch} for PowerPC targets.
16422 @item -mone-byte-bool
16423 @opindex mone-byte-bool
16424 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
16425 By default @code{sizeof(bool)} is @code{4} when compiling for
16426 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
16427 option has no effect on x86.
16429 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
16430 to generate code that is not binary compatible with code generated
16431 without that switch. Using this switch may require recompiling all
16432 other modules in a program, including system libraries. Use this
16433 switch to conform to a non-default data model.
16435 @item -mfix-and-continue
16436 @itemx -ffix-and-continue
16437 @itemx -findirect-data
16438 @opindex mfix-and-continue
16439 @opindex ffix-and-continue
16440 @opindex findirect-data
16441 Generate code suitable for fast turnaround development, such as to
16442 allow GDB to dynamically load @file{.o} files into already-running
16443 programs. @option{-findirect-data} and @option{-ffix-and-continue}
16444 are provided for backwards compatibility.
16448 Loads all members of static archive libraries.
16449 See man ld(1) for more information.
16451 @item -arch_errors_fatal
16452 @opindex arch_errors_fatal
16453 Cause the errors having to do with files that have the wrong architecture
16456 @item -bind_at_load
16457 @opindex bind_at_load
16458 Causes the output file to be marked such that the dynamic linker will
16459 bind all undefined references when the file is loaded or launched.
16463 Produce a Mach-o bundle format file.
16464 See man ld(1) for more information.
16466 @item -bundle_loader @var{executable}
16467 @opindex bundle_loader
16468 This option specifies the @var{executable} that will load the build
16469 output file being linked. See man ld(1) for more information.
16472 @opindex dynamiclib
16473 When passed this option, GCC produces a dynamic library instead of
16474 an executable when linking, using the Darwin @file{libtool} command.
16476 @item -force_cpusubtype_ALL
16477 @opindex force_cpusubtype_ALL
16478 This causes GCC's output file to have the @samp{ALL} subtype, instead of
16479 one controlled by the @option{-mcpu} or @option{-march} option.
16481 @item -allowable_client @var{client_name}
16482 @itemx -client_name
16483 @itemx -compatibility_version
16484 @itemx -current_version
16486 @itemx -dependency-file
16488 @itemx -dylinker_install_name
16490 @itemx -exported_symbols_list
16493 @itemx -flat_namespace
16494 @itemx -force_flat_namespace
16495 @itemx -headerpad_max_install_names
16498 @itemx -install_name
16499 @itemx -keep_private_externs
16500 @itemx -multi_module
16501 @itemx -multiply_defined
16502 @itemx -multiply_defined_unused
16505 @itemx -no_dead_strip_inits_and_terms
16506 @itemx -nofixprebinding
16507 @itemx -nomultidefs
16509 @itemx -noseglinkedit
16510 @itemx -pagezero_size
16512 @itemx -prebind_all_twolevel_modules
16513 @itemx -private_bundle
16515 @itemx -read_only_relocs
16517 @itemx -sectobjectsymbols
16521 @itemx -sectobjectsymbols
16524 @itemx -segs_read_only_addr
16526 @itemx -segs_read_write_addr
16527 @itemx -seg_addr_table
16528 @itemx -seg_addr_table_filename
16529 @itemx -seglinkedit
16531 @itemx -segs_read_only_addr
16532 @itemx -segs_read_write_addr
16533 @itemx -single_module
16535 @itemx -sub_library
16537 @itemx -sub_umbrella
16538 @itemx -twolevel_namespace
16541 @itemx -unexported_symbols_list
16542 @itemx -weak_reference_mismatches
16543 @itemx -whatsloaded
16544 @opindex allowable_client
16545 @opindex client_name
16546 @opindex compatibility_version
16547 @opindex current_version
16548 @opindex dead_strip
16549 @opindex dependency-file
16550 @opindex dylib_file
16551 @opindex dylinker_install_name
16553 @opindex exported_symbols_list
16555 @opindex flat_namespace
16556 @opindex force_flat_namespace
16557 @opindex headerpad_max_install_names
16558 @opindex image_base
16560 @opindex install_name
16561 @opindex keep_private_externs
16562 @opindex multi_module
16563 @opindex multiply_defined
16564 @opindex multiply_defined_unused
16565 @opindex noall_load
16566 @opindex no_dead_strip_inits_and_terms
16567 @opindex nofixprebinding
16568 @opindex nomultidefs
16570 @opindex noseglinkedit
16571 @opindex pagezero_size
16573 @opindex prebind_all_twolevel_modules
16574 @opindex private_bundle
16575 @opindex read_only_relocs
16577 @opindex sectobjectsymbols
16580 @opindex sectcreate
16581 @opindex sectobjectsymbols
16584 @opindex segs_read_only_addr
16585 @opindex segs_read_write_addr
16586 @opindex seg_addr_table
16587 @opindex seg_addr_table_filename
16588 @opindex seglinkedit
16590 @opindex segs_read_only_addr
16591 @opindex segs_read_write_addr
16592 @opindex single_module
16594 @opindex sub_library
16595 @opindex sub_umbrella
16596 @opindex twolevel_namespace
16599 @opindex unexported_symbols_list
16600 @opindex weak_reference_mismatches
16601 @opindex whatsloaded
16602 These options are passed to the Darwin linker. The Darwin linker man page
16603 describes them in detail.
16606 @node DEC Alpha Options
16607 @subsection DEC Alpha Options
16609 These @samp{-m} options are defined for the DEC Alpha implementations:
16612 @item -mno-soft-float
16613 @itemx -msoft-float
16614 @opindex mno-soft-float
16615 @opindex msoft-float
16616 Use (do not use) the hardware floating-point instructions for
16617 floating-point operations. When @option{-msoft-float} is specified,
16618 functions in @file{libgcc.a} are used to perform floating-point
16619 operations. Unless they are replaced by routines that emulate the
16620 floating-point operations, or compiled in such a way as to call such
16621 emulations routines, these routines issue floating-point
16622 operations. If you are compiling for an Alpha without floating-point
16623 operations, you must ensure that the library is built so as not to call
16626 Note that Alpha implementations without floating-point operations are
16627 required to have floating-point registers.
16630 @itemx -mno-fp-regs
16632 @opindex mno-fp-regs
16633 Generate code that uses (does not use) the floating-point register set.
16634 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
16635 register set is not used, floating-point operands are passed in integer
16636 registers as if they were integers and floating-point results are passed
16637 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
16638 so any function with a floating-point argument or return value called by code
16639 compiled with @option{-mno-fp-regs} must also be compiled with that
16642 A typical use of this option is building a kernel that does not use,
16643 and hence need not save and restore, any floating-point registers.
16647 The Alpha architecture implements floating-point hardware optimized for
16648 maximum performance. It is mostly compliant with the IEEE floating-point
16649 standard. However, for full compliance, software assistance is
16650 required. This option generates code fully IEEE-compliant code
16651 @emph{except} that the @var{inexact-flag} is not maintained (see below).
16652 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
16653 defined during compilation. The resulting code is less efficient but is
16654 able to correctly support denormalized numbers and exceptional IEEE
16655 values such as not-a-number and plus/minus infinity. Other Alpha
16656 compilers call this option @option{-ieee_with_no_inexact}.
16658 @item -mieee-with-inexact
16659 @opindex mieee-with-inexact
16660 This is like @option{-mieee} except the generated code also maintains
16661 the IEEE @var{inexact-flag}. Turning on this option causes the
16662 generated code to implement fully-compliant IEEE math. In addition to
16663 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
16664 macro. On some Alpha implementations the resulting code may execute
16665 significantly slower than the code generated by default. Since there is
16666 very little code that depends on the @var{inexact-flag}, you should
16667 normally not specify this option. Other Alpha compilers call this
16668 option @option{-ieee_with_inexact}.
16670 @item -mfp-trap-mode=@var{trap-mode}
16671 @opindex mfp-trap-mode
16672 This option controls what floating-point related traps are enabled.
16673 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
16674 The trap mode can be set to one of four values:
16678 This is the default (normal) setting. The only traps that are enabled
16679 are the ones that cannot be disabled in software (e.g., division by zero
16683 In addition to the traps enabled by @samp{n}, underflow traps are enabled
16687 Like @samp{u}, but the instructions are marked to be safe for software
16688 completion (see Alpha architecture manual for details).
16691 Like @samp{su}, but inexact traps are enabled as well.
16694 @item -mfp-rounding-mode=@var{rounding-mode}
16695 @opindex mfp-rounding-mode
16696 Selects the IEEE rounding mode. Other Alpha compilers call this option
16697 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
16702 Normal IEEE rounding mode. Floating-point numbers are rounded towards
16703 the nearest machine number or towards the even machine number in case
16707 Round towards minus infinity.
16710 Chopped rounding mode. Floating-point numbers are rounded towards zero.
16713 Dynamic rounding mode. A field in the floating-point control register
16714 (@var{fpcr}, see Alpha architecture reference manual) controls the
16715 rounding mode in effect. The C library initializes this register for
16716 rounding towards plus infinity. Thus, unless your program modifies the
16717 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
16720 @item -mtrap-precision=@var{trap-precision}
16721 @opindex mtrap-precision
16722 In the Alpha architecture, floating-point traps are imprecise. This
16723 means without software assistance it is impossible to recover from a
16724 floating trap and program execution normally needs to be terminated.
16725 GCC can generate code that can assist operating system trap handlers
16726 in determining the exact location that caused a floating-point trap.
16727 Depending on the requirements of an application, different levels of
16728 precisions can be selected:
16732 Program precision. This option is the default and means a trap handler
16733 can only identify which program caused a floating-point exception.
16736 Function precision. The trap handler can determine the function that
16737 caused a floating-point exception.
16740 Instruction precision. The trap handler can determine the exact
16741 instruction that caused a floating-point exception.
16744 Other Alpha compilers provide the equivalent options called
16745 @option{-scope_safe} and @option{-resumption_safe}.
16747 @item -mieee-conformant
16748 @opindex mieee-conformant
16749 This option marks the generated code as IEEE conformant. You must not
16750 use this option unless you also specify @option{-mtrap-precision=i} and either
16751 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
16752 is to emit the line @samp{.eflag 48} in the function prologue of the
16753 generated assembly file.
16755 @item -mbuild-constants
16756 @opindex mbuild-constants
16757 Normally GCC examines a 32- or 64-bit integer constant to
16758 see if it can construct it from smaller constants in two or three
16759 instructions. If it cannot, it outputs the constant as a literal and
16760 generates code to load it from the data segment at run time.
16762 Use this option to require GCC to construct @emph{all} integer constants
16763 using code, even if it takes more instructions (the maximum is six).
16765 You typically use this option to build a shared library dynamic
16766 loader. Itself a shared library, it must relocate itself in memory
16767 before it can find the variables and constants in its own data segment.
16785 Indicate whether GCC should generate code to use the optional BWX,
16786 CIX, FIX and MAX instruction sets. The default is to use the instruction
16787 sets supported by the CPU type specified via @option{-mcpu=} option or that
16788 of the CPU on which GCC was built if none is specified.
16791 @itemx -mfloat-ieee
16792 @opindex mfloat-vax
16793 @opindex mfloat-ieee
16794 Generate code that uses (does not use) VAX F and G floating-point
16795 arithmetic instead of IEEE single and double precision.
16797 @item -mexplicit-relocs
16798 @itemx -mno-explicit-relocs
16799 @opindex mexplicit-relocs
16800 @opindex mno-explicit-relocs
16801 Older Alpha assemblers provided no way to generate symbol relocations
16802 except via assembler macros. Use of these macros does not allow
16803 optimal instruction scheduling. GNU binutils as of version 2.12
16804 supports a new syntax that allows the compiler to explicitly mark
16805 which relocations should apply to which instructions. This option
16806 is mostly useful for debugging, as GCC detects the capabilities of
16807 the assembler when it is built and sets the default accordingly.
16810 @itemx -mlarge-data
16811 @opindex msmall-data
16812 @opindex mlarge-data
16813 When @option{-mexplicit-relocs} is in effect, static data is
16814 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
16815 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
16816 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
16817 16-bit relocations off of the @code{$gp} register. This limits the
16818 size of the small data area to 64KB, but allows the variables to be
16819 directly accessed via a single instruction.
16821 The default is @option{-mlarge-data}. With this option the data area
16822 is limited to just below 2GB@. Programs that require more than 2GB of
16823 data must use @code{malloc} or @code{mmap} to allocate the data in the
16824 heap instead of in the program's data segment.
16826 When generating code for shared libraries, @option{-fpic} implies
16827 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
16830 @itemx -mlarge-text
16831 @opindex msmall-text
16832 @opindex mlarge-text
16833 When @option{-msmall-text} is used, the compiler assumes that the
16834 code of the entire program (or shared library) fits in 4MB, and is
16835 thus reachable with a branch instruction. When @option{-msmall-data}
16836 is used, the compiler can assume that all local symbols share the
16837 same @code{$gp} value, and thus reduce the number of instructions
16838 required for a function call from 4 to 1.
16840 The default is @option{-mlarge-text}.
16842 @item -mcpu=@var{cpu_type}
16844 Set the instruction set and instruction scheduling parameters for
16845 machine type @var{cpu_type}. You can specify either the @samp{EV}
16846 style name or the corresponding chip number. GCC supports scheduling
16847 parameters for the EV4, EV5 and EV6 family of processors and
16848 chooses the default values for the instruction set from the processor
16849 you specify. If you do not specify a processor type, GCC defaults
16850 to the processor on which the compiler was built.
16852 Supported values for @var{cpu_type} are
16858 Schedules as an EV4 and has no instruction set extensions.
16862 Schedules as an EV5 and has no instruction set extensions.
16866 Schedules as an EV5 and supports the BWX extension.
16871 Schedules as an EV5 and supports the BWX and MAX extensions.
16875 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
16879 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
16882 Native toolchains also support the value @samp{native},
16883 which selects the best architecture option for the host processor.
16884 @option{-mcpu=native} has no effect if GCC does not recognize
16887 @item -mtune=@var{cpu_type}
16889 Set only the instruction scheduling parameters for machine type
16890 @var{cpu_type}. The instruction set is not changed.
16892 Native toolchains also support the value @samp{native},
16893 which selects the best architecture option for the host processor.
16894 @option{-mtune=native} has no effect if GCC does not recognize
16897 @item -mmemory-latency=@var{time}
16898 @opindex mmemory-latency
16899 Sets the latency the scheduler should assume for typical memory
16900 references as seen by the application. This number is highly
16901 dependent on the memory access patterns used by the application
16902 and the size of the external cache on the machine.
16904 Valid options for @var{time} are
16908 A decimal number representing clock cycles.
16914 The compiler contains estimates of the number of clock cycles for
16915 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
16916 (also called Dcache, Scache, and Bcache), as well as to main memory.
16917 Note that L3 is only valid for EV5.
16923 @subsection FR30 Options
16924 @cindex FR30 Options
16926 These options are defined specifically for the FR30 port.
16930 @item -msmall-model
16931 @opindex msmall-model
16932 Use the small address space model. This can produce smaller code, but
16933 it does assume that all symbolic values and addresses fit into a
16938 Assume that runtime support has been provided and so there is no need
16939 to include the simulator library (@file{libsim.a}) on the linker
16945 @subsection FT32 Options
16946 @cindex FT32 Options
16948 These options are defined specifically for the FT32 port.
16954 Specifies that the program will be run on the simulator. This causes
16955 an alternate runtime startup and library to be linked.
16956 You must not use this option when generating programs that will run on
16957 real hardware; you must provide your own runtime library for whatever
16958 I/O functions are needed.
16962 Enable Local Register Allocation. This is still experimental for FT32,
16963 so by default the compiler uses standard reload.
16967 Do not use div and mod instructions.
16972 @subsection FRV Options
16973 @cindex FRV Options
16979 Only use the first 32 general-purpose registers.
16984 Use all 64 general-purpose registers.
16989 Use only the first 32 floating-point registers.
16994 Use all 64 floating-point registers.
16997 @opindex mhard-float
16999 Use hardware instructions for floating-point operations.
17002 @opindex msoft-float
17004 Use library routines for floating-point operations.
17009 Dynamically allocate condition code registers.
17014 Do not try to dynamically allocate condition code registers, only
17015 use @code{icc0} and @code{fcc0}.
17020 Change ABI to use double word insns.
17025 Do not use double word instructions.
17030 Use floating-point double instructions.
17033 @opindex mno-double
17035 Do not use floating-point double instructions.
17040 Use media instructions.
17045 Do not use media instructions.
17050 Use multiply and add/subtract instructions.
17053 @opindex mno-muladd
17055 Do not use multiply and add/subtract instructions.
17060 Select the FDPIC ABI, which uses function descriptors to represent
17061 pointers to functions. Without any PIC/PIE-related options, it
17062 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
17063 assumes GOT entries and small data are within a 12-bit range from the
17064 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
17065 are computed with 32 bits.
17066 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17069 @opindex minline-plt
17071 Enable inlining of PLT entries in function calls to functions that are
17072 not known to bind locally. It has no effect without @option{-mfdpic}.
17073 It's enabled by default if optimizing for speed and compiling for
17074 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
17075 optimization option such as @option{-O3} or above is present in the
17081 Assume a large TLS segment when generating thread-local code.
17086 Do not assume a large TLS segment when generating thread-local code.
17091 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
17092 that is known to be in read-only sections. It's enabled by default,
17093 except for @option{-fpic} or @option{-fpie}: even though it may help
17094 make the global offset table smaller, it trades 1 instruction for 4.
17095 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
17096 one of which may be shared by multiple symbols, and it avoids the need
17097 for a GOT entry for the referenced symbol, so it's more likely to be a
17098 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
17100 @item -multilib-library-pic
17101 @opindex multilib-library-pic
17103 Link with the (library, not FD) pic libraries. It's implied by
17104 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
17105 @option{-fpic} without @option{-mfdpic}. You should never have to use
17109 @opindex mlinked-fp
17111 Follow the EABI requirement of always creating a frame pointer whenever
17112 a stack frame is allocated. This option is enabled by default and can
17113 be disabled with @option{-mno-linked-fp}.
17116 @opindex mlong-calls
17118 Use indirect addressing to call functions outside the current
17119 compilation unit. This allows the functions to be placed anywhere
17120 within the 32-bit address space.
17122 @item -malign-labels
17123 @opindex malign-labels
17125 Try to align labels to an 8-byte boundary by inserting NOPs into the
17126 previous packet. This option only has an effect when VLIW packing
17127 is enabled. It doesn't create new packets; it merely adds NOPs to
17130 @item -mlibrary-pic
17131 @opindex mlibrary-pic
17133 Generate position-independent EABI code.
17138 Use only the first four media accumulator registers.
17143 Use all eight media accumulator registers.
17148 Pack VLIW instructions.
17153 Do not pack VLIW instructions.
17156 @opindex mno-eflags
17158 Do not mark ABI switches in e_flags.
17161 @opindex mcond-move
17163 Enable the use of conditional-move instructions (default).
17165 This switch is mainly for debugging the compiler and will likely be removed
17166 in a future version.
17168 @item -mno-cond-move
17169 @opindex mno-cond-move
17171 Disable the use of conditional-move instructions.
17173 This switch is mainly for debugging the compiler and will likely be removed
17174 in a future version.
17179 Enable the use of conditional set instructions (default).
17181 This switch is mainly for debugging the compiler and will likely be removed
17182 in a future version.
17187 Disable the use of conditional set instructions.
17189 This switch is mainly for debugging the compiler and will likely be removed
17190 in a future version.
17193 @opindex mcond-exec
17195 Enable the use of conditional execution (default).
17197 This switch is mainly for debugging the compiler and will likely be removed
17198 in a future version.
17200 @item -mno-cond-exec
17201 @opindex mno-cond-exec
17203 Disable the use of conditional execution.
17205 This switch is mainly for debugging the compiler and will likely be removed
17206 in a future version.
17208 @item -mvliw-branch
17209 @opindex mvliw-branch
17211 Run a pass to pack branches into VLIW instructions (default).
17213 This switch is mainly for debugging the compiler and will likely be removed
17214 in a future version.
17216 @item -mno-vliw-branch
17217 @opindex mno-vliw-branch
17219 Do not run a pass to pack branches into VLIW instructions.
17221 This switch is mainly for debugging the compiler and will likely be removed
17222 in a future version.
17224 @item -mmulti-cond-exec
17225 @opindex mmulti-cond-exec
17227 Enable optimization of @code{&&} and @code{||} in conditional execution
17230 This switch is mainly for debugging the compiler and will likely be removed
17231 in a future version.
17233 @item -mno-multi-cond-exec
17234 @opindex mno-multi-cond-exec
17236 Disable optimization of @code{&&} and @code{||} in conditional execution.
17238 This switch is mainly for debugging the compiler and will likely be removed
17239 in a future version.
17241 @item -mnested-cond-exec
17242 @opindex mnested-cond-exec
17244 Enable nested conditional execution optimizations (default).
17246 This switch is mainly for debugging the compiler and will likely be removed
17247 in a future version.
17249 @item -mno-nested-cond-exec
17250 @opindex mno-nested-cond-exec
17252 Disable nested conditional execution optimizations.
17254 This switch is mainly for debugging the compiler and will likely be removed
17255 in a future version.
17257 @item -moptimize-membar
17258 @opindex moptimize-membar
17260 This switch removes redundant @code{membar} instructions from the
17261 compiler-generated code. It is enabled by default.
17263 @item -mno-optimize-membar
17264 @opindex mno-optimize-membar
17266 This switch disables the automatic removal of redundant @code{membar}
17267 instructions from the generated code.
17269 @item -mtomcat-stats
17270 @opindex mtomcat-stats
17272 Cause gas to print out tomcat statistics.
17274 @item -mcpu=@var{cpu}
17277 Select the processor type for which to generate code. Possible values are
17278 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
17279 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
17283 @node GNU/Linux Options
17284 @subsection GNU/Linux Options
17286 These @samp{-m} options are defined for GNU/Linux targets:
17291 Use the GNU C library. This is the default except
17292 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
17293 @samp{*-*-linux-*android*} targets.
17297 Use uClibc C library. This is the default on
17298 @samp{*-*-linux-*uclibc*} targets.
17302 Use the musl C library. This is the default on
17303 @samp{*-*-linux-*musl*} targets.
17307 Use Bionic C library. This is the default on
17308 @samp{*-*-linux-*android*} targets.
17312 Compile code compatible with Android platform. This is the default on
17313 @samp{*-*-linux-*android*} targets.
17315 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
17316 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
17317 this option makes the GCC driver pass Android-specific options to the linker.
17318 Finally, this option causes the preprocessor macro @code{__ANDROID__}
17321 @item -tno-android-cc
17322 @opindex tno-android-cc
17323 Disable compilation effects of @option{-mandroid}, i.e., do not enable
17324 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
17325 @option{-fno-rtti} by default.
17327 @item -tno-android-ld
17328 @opindex tno-android-ld
17329 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
17330 linking options to the linker.
17334 @node H8/300 Options
17335 @subsection H8/300 Options
17337 These @samp{-m} options are defined for the H8/300 implementations:
17342 Shorten some address references at link time, when possible; uses the
17343 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
17344 ld, Using ld}, for a fuller description.
17348 Generate code for the H8/300H@.
17352 Generate code for the H8S@.
17356 Generate code for the H8S and H8/300H in the normal mode. This switch
17357 must be used either with @option{-mh} or @option{-ms}.
17361 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
17365 Extended registers are stored on stack before execution of function
17366 with monitor attribute. Default option is @option{-mexr}.
17367 This option is valid only for H8S targets.
17371 Extended registers are not stored on stack before execution of function
17372 with monitor attribute. Default option is @option{-mno-exr}.
17373 This option is valid only for H8S targets.
17377 Make @code{int} data 32 bits by default.
17380 @opindex malign-300
17381 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
17382 The default for the H8/300H and H8S is to align longs and floats on
17384 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
17385 This option has no effect on the H8/300.
17389 @subsection HPPA Options
17390 @cindex HPPA Options
17392 These @samp{-m} options are defined for the HPPA family of computers:
17395 @item -march=@var{architecture-type}
17397 Generate code for the specified architecture. The choices for
17398 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
17399 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
17400 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
17401 architecture option for your machine. Code compiled for lower numbered
17402 architectures runs on higher numbered architectures, but not the
17405 @item -mpa-risc-1-0
17406 @itemx -mpa-risc-1-1
17407 @itemx -mpa-risc-2-0
17408 @opindex mpa-risc-1-0
17409 @opindex mpa-risc-1-1
17410 @opindex mpa-risc-2-0
17411 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
17413 @item -mcaller-copies
17414 @opindex mcaller-copies
17415 The caller copies function arguments passed by hidden reference. This
17416 option should be used with care as it is not compatible with the default
17417 32-bit runtime. However, only aggregates larger than eight bytes are
17418 passed by hidden reference and the option provides better compatibility
17421 @item -mjump-in-delay
17422 @opindex mjump-in-delay
17423 This option is ignored and provided for compatibility purposes only.
17425 @item -mdisable-fpregs
17426 @opindex mdisable-fpregs
17427 Prevent floating-point registers from being used in any manner. This is
17428 necessary for compiling kernels that perform lazy context switching of
17429 floating-point registers. If you use this option and attempt to perform
17430 floating-point operations, the compiler aborts.
17432 @item -mdisable-indexing
17433 @opindex mdisable-indexing
17434 Prevent the compiler from using indexing address modes. This avoids some
17435 rather obscure problems when compiling MIG generated code under MACH@.
17437 @item -mno-space-regs
17438 @opindex mno-space-regs
17439 Generate code that assumes the target has no space registers. This allows
17440 GCC to generate faster indirect calls and use unscaled index address modes.
17442 Such code is suitable for level 0 PA systems and kernels.
17444 @item -mfast-indirect-calls
17445 @opindex mfast-indirect-calls
17446 Generate code that assumes calls never cross space boundaries. This
17447 allows GCC to emit code that performs faster indirect calls.
17449 This option does not work in the presence of shared libraries or nested
17452 @item -mfixed-range=@var{register-range}
17453 @opindex mfixed-range
17454 Generate code treating the given register range as fixed registers.
17455 A fixed register is one that the register allocator cannot use. This is
17456 useful when compiling kernel code. A register range is specified as
17457 two registers separated by a dash. Multiple register ranges can be
17458 specified separated by a comma.
17460 @item -mlong-load-store
17461 @opindex mlong-load-store
17462 Generate 3-instruction load and store sequences as sometimes required by
17463 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
17466 @item -mportable-runtime
17467 @opindex mportable-runtime
17468 Use the portable calling conventions proposed by HP for ELF systems.
17472 Enable the use of assembler directives only GAS understands.
17474 @item -mschedule=@var{cpu-type}
17476 Schedule code according to the constraints for the machine type
17477 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
17478 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
17479 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
17480 proper scheduling option for your machine. The default scheduling is
17484 @opindex mlinker-opt
17485 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
17486 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
17487 linkers in which they give bogus error messages when linking some programs.
17490 @opindex msoft-float
17491 Generate output containing library calls for floating point.
17492 @strong{Warning:} the requisite libraries are not available for all HPPA
17493 targets. Normally the facilities of the machine's usual C compiler are
17494 used, but this cannot be done directly in cross-compilation. You must make
17495 your own arrangements to provide suitable library functions for
17498 @option{-msoft-float} changes the calling convention in the output file;
17499 therefore, it is only useful if you compile @emph{all} of a program with
17500 this option. In particular, you need to compile @file{libgcc.a}, the
17501 library that comes with GCC, with @option{-msoft-float} in order for
17506 Generate the predefine, @code{_SIO}, for server IO@. The default is
17507 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
17508 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
17509 options are available under HP-UX and HI-UX@.
17513 Use options specific to GNU @command{ld}.
17514 This passes @option{-shared} to @command{ld} when
17515 building a shared library. It is the default when GCC is configured,
17516 explicitly or implicitly, with the GNU linker. This option does not
17517 affect which @command{ld} is called; it only changes what parameters
17518 are passed to that @command{ld}.
17519 The @command{ld} that is called is determined by the
17520 @option{--with-ld} configure option, GCC's program search path, and
17521 finally by the user's @env{PATH}. The linker used by GCC can be printed
17522 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
17523 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17527 Use options specific to HP @command{ld}.
17528 This passes @option{-b} to @command{ld} when building
17529 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
17530 links. It is the default when GCC is configured, explicitly or
17531 implicitly, with the HP linker. This option does not affect
17532 which @command{ld} is called; it only changes what parameters are passed to that
17534 The @command{ld} that is called is determined by the @option{--with-ld}
17535 configure option, GCC's program search path, and finally by the user's
17536 @env{PATH}. The linker used by GCC can be printed using @samp{which
17537 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
17538 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17541 @opindex mno-long-calls
17542 Generate code that uses long call sequences. This ensures that a call
17543 is always able to reach linker generated stubs. The default is to generate
17544 long calls only when the distance from the call site to the beginning
17545 of the function or translation unit, as the case may be, exceeds a
17546 predefined limit set by the branch type being used. The limits for
17547 normal calls are 7,600,000 and 240,000 bytes, respectively for the
17548 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
17551 Distances are measured from the beginning of functions when using the
17552 @option{-ffunction-sections} option, or when using the @option{-mgas}
17553 and @option{-mno-portable-runtime} options together under HP-UX with
17556 It is normally not desirable to use this option as it degrades
17557 performance. However, it may be useful in large applications,
17558 particularly when partial linking is used to build the application.
17560 The types of long calls used depends on the capabilities of the
17561 assembler and linker, and the type of code being generated. The
17562 impact on systems that support long absolute calls, and long pic
17563 symbol-difference or pc-relative calls should be relatively small.
17564 However, an indirect call is used on 32-bit ELF systems in pic code
17565 and it is quite long.
17567 @item -munix=@var{unix-std}
17569 Generate compiler predefines and select a startfile for the specified
17570 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
17571 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
17572 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
17573 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
17574 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
17577 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
17578 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
17579 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
17580 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
17581 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
17582 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
17584 It is @emph{important} to note that this option changes the interfaces
17585 for various library routines. It also affects the operational behavior
17586 of the C library. Thus, @emph{extreme} care is needed in using this
17589 Library code that is intended to operate with more than one UNIX
17590 standard must test, set and restore the variable @code{__xpg4_extended_mask}
17591 as appropriate. Most GNU software doesn't provide this capability.
17595 Suppress the generation of link options to search libdld.sl when the
17596 @option{-static} option is specified on HP-UX 10 and later.
17600 The HP-UX implementation of setlocale in libc has a dependency on
17601 libdld.sl. There isn't an archive version of libdld.sl. Thus,
17602 when the @option{-static} option is specified, special link options
17603 are needed to resolve this dependency.
17605 On HP-UX 10 and later, the GCC driver adds the necessary options to
17606 link with libdld.sl when the @option{-static} option is specified.
17607 This causes the resulting binary to be dynamic. On the 64-bit port,
17608 the linkers generate dynamic binaries by default in any case. The
17609 @option{-nolibdld} option can be used to prevent the GCC driver from
17610 adding these link options.
17614 Add support for multithreading with the @dfn{dce thread} library
17615 under HP-UX@. This option sets flags for both the preprocessor and
17619 @node IA-64 Options
17620 @subsection IA-64 Options
17621 @cindex IA-64 Options
17623 These are the @samp{-m} options defined for the Intel IA-64 architecture.
17627 @opindex mbig-endian
17628 Generate code for a big-endian target. This is the default for HP-UX@.
17630 @item -mlittle-endian
17631 @opindex mlittle-endian
17632 Generate code for a little-endian target. This is the default for AIX5
17638 @opindex mno-gnu-as
17639 Generate (or don't) code for the GNU assembler. This is the default.
17640 @c Also, this is the default if the configure option @option{--with-gnu-as}
17646 @opindex mno-gnu-ld
17647 Generate (or don't) code for the GNU linker. This is the default.
17648 @c Also, this is the default if the configure option @option{--with-gnu-ld}
17653 Generate code that does not use a global pointer register. The result
17654 is not position independent code, and violates the IA-64 ABI@.
17656 @item -mvolatile-asm-stop
17657 @itemx -mno-volatile-asm-stop
17658 @opindex mvolatile-asm-stop
17659 @opindex mno-volatile-asm-stop
17660 Generate (or don't) a stop bit immediately before and after volatile asm
17663 @item -mregister-names
17664 @itemx -mno-register-names
17665 @opindex mregister-names
17666 @opindex mno-register-names
17667 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
17668 the stacked registers. This may make assembler output more readable.
17674 Disable (or enable) optimizations that use the small data section. This may
17675 be useful for working around optimizer bugs.
17677 @item -mconstant-gp
17678 @opindex mconstant-gp
17679 Generate code that uses a single constant global pointer value. This is
17680 useful when compiling kernel code.
17684 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
17685 This is useful when compiling firmware code.
17687 @item -minline-float-divide-min-latency
17688 @opindex minline-float-divide-min-latency
17689 Generate code for inline divides of floating-point values
17690 using the minimum latency algorithm.
17692 @item -minline-float-divide-max-throughput
17693 @opindex minline-float-divide-max-throughput
17694 Generate code for inline divides of floating-point values
17695 using the maximum throughput algorithm.
17697 @item -mno-inline-float-divide
17698 @opindex mno-inline-float-divide
17699 Do not generate inline code for divides of floating-point values.
17701 @item -minline-int-divide-min-latency
17702 @opindex minline-int-divide-min-latency
17703 Generate code for inline divides of integer values
17704 using the minimum latency algorithm.
17706 @item -minline-int-divide-max-throughput
17707 @opindex minline-int-divide-max-throughput
17708 Generate code for inline divides of integer values
17709 using the maximum throughput algorithm.
17711 @item -mno-inline-int-divide
17712 @opindex mno-inline-int-divide
17713 Do not generate inline code for divides of integer values.
17715 @item -minline-sqrt-min-latency
17716 @opindex minline-sqrt-min-latency
17717 Generate code for inline square roots
17718 using the minimum latency algorithm.
17720 @item -minline-sqrt-max-throughput
17721 @opindex minline-sqrt-max-throughput
17722 Generate code for inline square roots
17723 using the maximum throughput algorithm.
17725 @item -mno-inline-sqrt
17726 @opindex mno-inline-sqrt
17727 Do not generate inline code for @code{sqrt}.
17730 @itemx -mno-fused-madd
17731 @opindex mfused-madd
17732 @opindex mno-fused-madd
17733 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
17734 instructions. The default is to use these instructions.
17736 @item -mno-dwarf2-asm
17737 @itemx -mdwarf2-asm
17738 @opindex mno-dwarf2-asm
17739 @opindex mdwarf2-asm
17740 Don't (or do) generate assembler code for the DWARF line number debugging
17741 info. This may be useful when not using the GNU assembler.
17743 @item -mearly-stop-bits
17744 @itemx -mno-early-stop-bits
17745 @opindex mearly-stop-bits
17746 @opindex mno-early-stop-bits
17747 Allow stop bits to be placed earlier than immediately preceding the
17748 instruction that triggered the stop bit. This can improve instruction
17749 scheduling, but does not always do so.
17751 @item -mfixed-range=@var{register-range}
17752 @opindex mfixed-range
17753 Generate code treating the given register range as fixed registers.
17754 A fixed register is one that the register allocator cannot use. This is
17755 useful when compiling kernel code. A register range is specified as
17756 two registers separated by a dash. Multiple register ranges can be
17757 specified separated by a comma.
17759 @item -mtls-size=@var{tls-size}
17761 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
17764 @item -mtune=@var{cpu-type}
17766 Tune the instruction scheduling for a particular CPU, Valid values are
17767 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
17768 and @samp{mckinley}.
17774 Generate code for a 32-bit or 64-bit environment.
17775 The 32-bit environment sets int, long and pointer to 32 bits.
17776 The 64-bit environment sets int to 32 bits and long and pointer
17777 to 64 bits. These are HP-UX specific flags.
17779 @item -mno-sched-br-data-spec
17780 @itemx -msched-br-data-spec
17781 @opindex mno-sched-br-data-spec
17782 @opindex msched-br-data-spec
17783 (Dis/En)able data speculative scheduling before reload.
17784 This results in generation of @code{ld.a} instructions and
17785 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17786 The default setting is disabled.
17788 @item -msched-ar-data-spec
17789 @itemx -mno-sched-ar-data-spec
17790 @opindex msched-ar-data-spec
17791 @opindex mno-sched-ar-data-spec
17792 (En/Dis)able data speculative scheduling after reload.
17793 This results in generation of @code{ld.a} instructions and
17794 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17795 The default setting is enabled.
17797 @item -mno-sched-control-spec
17798 @itemx -msched-control-spec
17799 @opindex mno-sched-control-spec
17800 @opindex msched-control-spec
17801 (Dis/En)able control speculative scheduling. This feature is
17802 available only during region scheduling (i.e.@: before reload).
17803 This results in generation of the @code{ld.s} instructions and
17804 the corresponding check instructions @code{chk.s}.
17805 The default setting is disabled.
17807 @item -msched-br-in-data-spec
17808 @itemx -mno-sched-br-in-data-spec
17809 @opindex msched-br-in-data-spec
17810 @opindex mno-sched-br-in-data-spec
17811 (En/Dis)able speculative scheduling of the instructions that
17812 are dependent on the data speculative loads before reload.
17813 This is effective only with @option{-msched-br-data-spec} enabled.
17814 The default setting is enabled.
17816 @item -msched-ar-in-data-spec
17817 @itemx -mno-sched-ar-in-data-spec
17818 @opindex msched-ar-in-data-spec
17819 @opindex mno-sched-ar-in-data-spec
17820 (En/Dis)able speculative scheduling of the instructions that
17821 are dependent on the data speculative loads after reload.
17822 This is effective only with @option{-msched-ar-data-spec} enabled.
17823 The default setting is enabled.
17825 @item -msched-in-control-spec
17826 @itemx -mno-sched-in-control-spec
17827 @opindex msched-in-control-spec
17828 @opindex mno-sched-in-control-spec
17829 (En/Dis)able speculative scheduling of the instructions that
17830 are dependent on the control speculative loads.
17831 This is effective only with @option{-msched-control-spec} enabled.
17832 The default setting is enabled.
17834 @item -mno-sched-prefer-non-data-spec-insns
17835 @itemx -msched-prefer-non-data-spec-insns
17836 @opindex mno-sched-prefer-non-data-spec-insns
17837 @opindex msched-prefer-non-data-spec-insns
17838 If enabled, data-speculative instructions are chosen for schedule
17839 only if there are no other choices at the moment. This makes
17840 the use of the data speculation much more conservative.
17841 The default setting is disabled.
17843 @item -mno-sched-prefer-non-control-spec-insns
17844 @itemx -msched-prefer-non-control-spec-insns
17845 @opindex mno-sched-prefer-non-control-spec-insns
17846 @opindex msched-prefer-non-control-spec-insns
17847 If enabled, control-speculative instructions are chosen for schedule
17848 only if there are no other choices at the moment. This makes
17849 the use of the control speculation much more conservative.
17850 The default setting is disabled.
17852 @item -mno-sched-count-spec-in-critical-path
17853 @itemx -msched-count-spec-in-critical-path
17854 @opindex mno-sched-count-spec-in-critical-path
17855 @opindex msched-count-spec-in-critical-path
17856 If enabled, speculative dependencies are considered during
17857 computation of the instructions priorities. This makes the use of the
17858 speculation a bit more conservative.
17859 The default setting is disabled.
17861 @item -msched-spec-ldc
17862 @opindex msched-spec-ldc
17863 Use a simple data speculation check. This option is on by default.
17865 @item -msched-control-spec-ldc
17866 @opindex msched-spec-ldc
17867 Use a simple check for control speculation. This option is on by default.
17869 @item -msched-stop-bits-after-every-cycle
17870 @opindex msched-stop-bits-after-every-cycle
17871 Place a stop bit after every cycle when scheduling. This option is on
17874 @item -msched-fp-mem-deps-zero-cost
17875 @opindex msched-fp-mem-deps-zero-cost
17876 Assume that floating-point stores and loads are not likely to cause a conflict
17877 when placed into the same instruction group. This option is disabled by
17880 @item -msel-sched-dont-check-control-spec
17881 @opindex msel-sched-dont-check-control-spec
17882 Generate checks for control speculation in selective scheduling.
17883 This flag is disabled by default.
17885 @item -msched-max-memory-insns=@var{max-insns}
17886 @opindex msched-max-memory-insns
17887 Limit on the number of memory insns per instruction group, giving lower
17888 priority to subsequent memory insns attempting to schedule in the same
17889 instruction group. Frequently useful to prevent cache bank conflicts.
17890 The default value is 1.
17892 @item -msched-max-memory-insns-hard-limit
17893 @opindex msched-max-memory-insns-hard-limit
17894 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
17895 disallowing more than that number in an instruction group.
17896 Otherwise, the limit is ``soft'', meaning that non-memory operations
17897 are preferred when the limit is reached, but memory operations may still
17903 @subsection LM32 Options
17904 @cindex LM32 options
17906 These @option{-m} options are defined for the LatticeMico32 architecture:
17909 @item -mbarrel-shift-enabled
17910 @opindex mbarrel-shift-enabled
17911 Enable barrel-shift instructions.
17913 @item -mdivide-enabled
17914 @opindex mdivide-enabled
17915 Enable divide and modulus instructions.
17917 @item -mmultiply-enabled
17918 @opindex multiply-enabled
17919 Enable multiply instructions.
17921 @item -msign-extend-enabled
17922 @opindex msign-extend-enabled
17923 Enable sign extend instructions.
17925 @item -muser-enabled
17926 @opindex muser-enabled
17927 Enable user-defined instructions.
17932 @subsection M32C Options
17933 @cindex M32C options
17936 @item -mcpu=@var{name}
17938 Select the CPU for which code is generated. @var{name} may be one of
17939 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
17940 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
17941 the M32C/80 series.
17945 Specifies that the program will be run on the simulator. This causes
17946 an alternate runtime library to be linked in which supports, for
17947 example, file I/O@. You must not use this option when generating
17948 programs that will run on real hardware; you must provide your own
17949 runtime library for whatever I/O functions are needed.
17951 @item -memregs=@var{number}
17953 Specifies the number of memory-based pseudo-registers GCC uses
17954 during code generation. These pseudo-registers are used like real
17955 registers, so there is a tradeoff between GCC's ability to fit the
17956 code into available registers, and the performance penalty of using
17957 memory instead of registers. Note that all modules in a program must
17958 be compiled with the same value for this option. Because of that, you
17959 must not use this option with GCC's default runtime libraries.
17963 @node M32R/D Options
17964 @subsection M32R/D Options
17965 @cindex M32R/D options
17967 These @option{-m} options are defined for Renesas M32R/D architectures:
17972 Generate code for the M32R/2@.
17976 Generate code for the M32R/X@.
17980 Generate code for the M32R@. This is the default.
17982 @item -mmodel=small
17983 @opindex mmodel=small
17984 Assume all objects live in the lower 16MB of memory (so that their addresses
17985 can be loaded with the @code{ld24} instruction), and assume all subroutines
17986 are reachable with the @code{bl} instruction.
17987 This is the default.
17989 The addressability of a particular object can be set with the
17990 @code{model} attribute.
17992 @item -mmodel=medium
17993 @opindex mmodel=medium
17994 Assume objects may be anywhere in the 32-bit address space (the compiler
17995 generates @code{seth/add3} instructions to load their addresses), and
17996 assume all subroutines are reachable with the @code{bl} instruction.
17998 @item -mmodel=large
17999 @opindex mmodel=large
18000 Assume objects may be anywhere in the 32-bit address space (the compiler
18001 generates @code{seth/add3} instructions to load their addresses), and
18002 assume subroutines may not be reachable with the @code{bl} instruction
18003 (the compiler generates the much slower @code{seth/add3/jl}
18004 instruction sequence).
18007 @opindex msdata=none
18008 Disable use of the small data area. Variables are put into
18009 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
18010 @code{section} attribute has been specified).
18011 This is the default.
18013 The small data area consists of sections @code{.sdata} and @code{.sbss}.
18014 Objects may be explicitly put in the small data area with the
18015 @code{section} attribute using one of these sections.
18017 @item -msdata=sdata
18018 @opindex msdata=sdata
18019 Put small global and static data in the small data area, but do not
18020 generate special code to reference them.
18023 @opindex msdata=use
18024 Put small global and static data in the small data area, and generate
18025 special instructions to reference them.
18029 @cindex smaller data references
18030 Put global and static objects less than or equal to @var{num} bytes
18031 into the small data or BSS sections instead of the normal data or BSS
18032 sections. The default value of @var{num} is 8.
18033 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
18034 for this option to have any effect.
18036 All modules should be compiled with the same @option{-G @var{num}} value.
18037 Compiling with different values of @var{num} may or may not work; if it
18038 doesn't the linker gives an error message---incorrect code is not
18043 Makes the M32R-specific code in the compiler display some statistics
18044 that might help in debugging programs.
18046 @item -malign-loops
18047 @opindex malign-loops
18048 Align all loops to a 32-byte boundary.
18050 @item -mno-align-loops
18051 @opindex mno-align-loops
18052 Do not enforce a 32-byte alignment for loops. This is the default.
18054 @item -missue-rate=@var{number}
18055 @opindex missue-rate=@var{number}
18056 Issue @var{number} instructions per cycle. @var{number} can only be 1
18059 @item -mbranch-cost=@var{number}
18060 @opindex mbranch-cost=@var{number}
18061 @var{number} can only be 1 or 2. If it is 1 then branches are
18062 preferred over conditional code, if it is 2, then the opposite applies.
18064 @item -mflush-trap=@var{number}
18065 @opindex mflush-trap=@var{number}
18066 Specifies the trap number to use to flush the cache. The default is
18067 12. Valid numbers are between 0 and 15 inclusive.
18069 @item -mno-flush-trap
18070 @opindex mno-flush-trap
18071 Specifies that the cache cannot be flushed by using a trap.
18073 @item -mflush-func=@var{name}
18074 @opindex mflush-func=@var{name}
18075 Specifies the name of the operating system function to call to flush
18076 the cache. The default is @samp{_flush_cache}, but a function call
18077 is only used if a trap is not available.
18079 @item -mno-flush-func
18080 @opindex mno-flush-func
18081 Indicates that there is no OS function for flushing the cache.
18085 @node M680x0 Options
18086 @subsection M680x0 Options
18087 @cindex M680x0 options
18089 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
18090 The default settings depend on which architecture was selected when
18091 the compiler was configured; the defaults for the most common choices
18095 @item -march=@var{arch}
18097 Generate code for a specific M680x0 or ColdFire instruction set
18098 architecture. Permissible values of @var{arch} for M680x0
18099 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
18100 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
18101 architectures are selected according to Freescale's ISA classification
18102 and the permissible values are: @samp{isaa}, @samp{isaaplus},
18103 @samp{isab} and @samp{isac}.
18105 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
18106 code for a ColdFire target. The @var{arch} in this macro is one of the
18107 @option{-march} arguments given above.
18109 When used together, @option{-march} and @option{-mtune} select code
18110 that runs on a family of similar processors but that is optimized
18111 for a particular microarchitecture.
18113 @item -mcpu=@var{cpu}
18115 Generate code for a specific M680x0 or ColdFire processor.
18116 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
18117 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
18118 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
18119 below, which also classifies the CPUs into families:
18121 @multitable @columnfractions 0.20 0.80
18122 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
18123 @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}
18124 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
18125 @item @samp{5206e} @tab @samp{5206e}
18126 @item @samp{5208} @tab @samp{5207} @samp{5208}
18127 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
18128 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
18129 @item @samp{5216} @tab @samp{5214} @samp{5216}
18130 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
18131 @item @samp{5225} @tab @samp{5224} @samp{5225}
18132 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
18133 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
18134 @item @samp{5249} @tab @samp{5249}
18135 @item @samp{5250} @tab @samp{5250}
18136 @item @samp{5271} @tab @samp{5270} @samp{5271}
18137 @item @samp{5272} @tab @samp{5272}
18138 @item @samp{5275} @tab @samp{5274} @samp{5275}
18139 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
18140 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
18141 @item @samp{5307} @tab @samp{5307}
18142 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
18143 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
18144 @item @samp{5407} @tab @samp{5407}
18145 @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}
18148 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
18149 @var{arch} is compatible with @var{cpu}. Other combinations of
18150 @option{-mcpu} and @option{-march} are rejected.
18152 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
18153 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
18154 where the value of @var{family} is given by the table above.
18156 @item -mtune=@var{tune}
18158 Tune the code for a particular microarchitecture within the
18159 constraints set by @option{-march} and @option{-mcpu}.
18160 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
18161 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
18162 and @samp{cpu32}. The ColdFire microarchitectures
18163 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
18165 You can also use @option{-mtune=68020-40} for code that needs
18166 to run relatively well on 68020, 68030 and 68040 targets.
18167 @option{-mtune=68020-60} is similar but includes 68060 targets
18168 as well. These two options select the same tuning decisions as
18169 @option{-m68020-40} and @option{-m68020-60} respectively.
18171 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
18172 when tuning for 680x0 architecture @var{arch}. It also defines
18173 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
18174 option is used. If GCC is tuning for a range of architectures,
18175 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
18176 it defines the macros for every architecture in the range.
18178 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
18179 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
18180 of the arguments given above.
18186 Generate output for a 68000. This is the default
18187 when the compiler is configured for 68000-based systems.
18188 It is equivalent to @option{-march=68000}.
18190 Use this option for microcontrollers with a 68000 or EC000 core,
18191 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
18195 Generate output for a 68010. This is the default
18196 when the compiler is configured for 68010-based systems.
18197 It is equivalent to @option{-march=68010}.
18203 Generate output for a 68020. This is the default
18204 when the compiler is configured for 68020-based systems.
18205 It is equivalent to @option{-march=68020}.
18209 Generate output for a 68030. This is the default when the compiler is
18210 configured for 68030-based systems. It is equivalent to
18211 @option{-march=68030}.
18215 Generate output for a 68040. This is the default when the compiler is
18216 configured for 68040-based systems. It is equivalent to
18217 @option{-march=68040}.
18219 This option inhibits the use of 68881/68882 instructions that have to be
18220 emulated by software on the 68040. Use this option if your 68040 does not
18221 have code to emulate those instructions.
18225 Generate output for a 68060. This is the default when the compiler is
18226 configured for 68060-based systems. It is equivalent to
18227 @option{-march=68060}.
18229 This option inhibits the use of 68020 and 68881/68882 instructions that
18230 have to be emulated by software on the 68060. Use this option if your 68060
18231 does not have code to emulate those instructions.
18235 Generate output for a CPU32. This is the default
18236 when the compiler is configured for CPU32-based systems.
18237 It is equivalent to @option{-march=cpu32}.
18239 Use this option for microcontrollers with a
18240 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
18241 68336, 68340, 68341, 68349 and 68360.
18245 Generate output for a 520X ColdFire CPU@. This is the default
18246 when the compiler is configured for 520X-based systems.
18247 It is equivalent to @option{-mcpu=5206}, and is now deprecated
18248 in favor of that option.
18250 Use this option for microcontroller with a 5200 core, including
18251 the MCF5202, MCF5203, MCF5204 and MCF5206.
18255 Generate output for a 5206e ColdFire CPU@. The option is now
18256 deprecated in favor of the equivalent @option{-mcpu=5206e}.
18260 Generate output for a member of the ColdFire 528X family.
18261 The option is now deprecated in favor of the equivalent
18262 @option{-mcpu=528x}.
18266 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
18267 in favor of the equivalent @option{-mcpu=5307}.
18271 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
18272 in favor of the equivalent @option{-mcpu=5407}.
18276 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
18277 This includes use of hardware floating-point instructions.
18278 The option is equivalent to @option{-mcpu=547x}, and is now
18279 deprecated in favor of that option.
18283 Generate output for a 68040, without using any of the new instructions.
18284 This results in code that can run relatively efficiently on either a
18285 68020/68881 or a 68030 or a 68040. The generated code does use the
18286 68881 instructions that are emulated on the 68040.
18288 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
18292 Generate output for a 68060, without using any of the new instructions.
18293 This results in code that can run relatively efficiently on either a
18294 68020/68881 or a 68030 or a 68040. The generated code does use the
18295 68881 instructions that are emulated on the 68060.
18297 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
18301 @opindex mhard-float
18303 Generate floating-point instructions. This is the default for 68020
18304 and above, and for ColdFire devices that have an FPU@. It defines the
18305 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
18306 on ColdFire targets.
18309 @opindex msoft-float
18310 Do not generate floating-point instructions; use library calls instead.
18311 This is the default for 68000, 68010, and 68832 targets. It is also
18312 the default for ColdFire devices that have no FPU.
18318 Generate (do not generate) ColdFire hardware divide and remainder
18319 instructions. If @option{-march} is used without @option{-mcpu},
18320 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
18321 architectures. Otherwise, the default is taken from the target CPU
18322 (either the default CPU, or the one specified by @option{-mcpu}). For
18323 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
18324 @option{-mcpu=5206e}.
18326 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
18330 Consider type @code{int} to be 16 bits wide, like @code{short int}.
18331 Additionally, parameters passed on the stack are also aligned to a
18332 16-bit boundary even on targets whose API mandates promotion to 32-bit.
18336 Do not consider type @code{int} to be 16 bits wide. This is the default.
18339 @itemx -mno-bitfield
18340 @opindex mnobitfield
18341 @opindex mno-bitfield
18342 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
18343 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
18347 Do use the bit-field instructions. The @option{-m68020} option implies
18348 @option{-mbitfield}. This is the default if you use a configuration
18349 designed for a 68020.
18353 Use a different function-calling convention, in which functions
18354 that take a fixed number of arguments return with the @code{rtd}
18355 instruction, which pops their arguments while returning. This
18356 saves one instruction in the caller since there is no need to pop
18357 the arguments there.
18359 This calling convention is incompatible with the one normally
18360 used on Unix, so you cannot use it if you need to call libraries
18361 compiled with the Unix compiler.
18363 Also, you must provide function prototypes for all functions that
18364 take variable numbers of arguments (including @code{printf});
18365 otherwise incorrect code is generated for calls to those
18368 In addition, seriously incorrect code results if you call a
18369 function with too many arguments. (Normally, extra arguments are
18370 harmlessly ignored.)
18372 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
18373 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
18377 Do not use the calling conventions selected by @option{-mrtd}.
18378 This is the default.
18381 @itemx -mno-align-int
18382 @opindex malign-int
18383 @opindex mno-align-int
18384 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
18385 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
18386 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
18387 Aligning variables on 32-bit boundaries produces code that runs somewhat
18388 faster on processors with 32-bit busses at the expense of more memory.
18390 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
18391 aligns structures containing the above types differently than
18392 most published application binary interface specifications for the m68k.
18396 Use the pc-relative addressing mode of the 68000 directly, instead of
18397 using a global offset table. At present, this option implies @option{-fpic},
18398 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
18399 not presently supported with @option{-mpcrel}, though this could be supported for
18400 68020 and higher processors.
18402 @item -mno-strict-align
18403 @itemx -mstrict-align
18404 @opindex mno-strict-align
18405 @opindex mstrict-align
18406 Do not (do) assume that unaligned memory references are handled by
18410 Generate code that allows the data segment to be located in a different
18411 area of memory from the text segment. This allows for execute-in-place in
18412 an environment without virtual memory management. This option implies
18415 @item -mno-sep-data
18416 Generate code that assumes that the data segment follows the text segment.
18417 This is the default.
18419 @item -mid-shared-library
18420 Generate code that supports shared libraries via the library ID method.
18421 This allows for execute-in-place and shared libraries in an environment
18422 without virtual memory management. This option implies @option{-fPIC}.
18424 @item -mno-id-shared-library
18425 Generate code that doesn't assume ID-based shared libraries are being used.
18426 This is the default.
18428 @item -mshared-library-id=n
18429 Specifies the identification number of the ID-based shared library being
18430 compiled. Specifying a value of 0 generates more compact code; specifying
18431 other values forces the allocation of that number to the current
18432 library, but is no more space- or time-efficient than omitting this option.
18438 When generating position-independent code for ColdFire, generate code
18439 that works if the GOT has more than 8192 entries. This code is
18440 larger and slower than code generated without this option. On M680x0
18441 processors, this option is not needed; @option{-fPIC} suffices.
18443 GCC normally uses a single instruction to load values from the GOT@.
18444 While this is relatively efficient, it only works if the GOT
18445 is smaller than about 64k. Anything larger causes the linker
18446 to report an error such as:
18448 @cindex relocation truncated to fit (ColdFire)
18450 relocation truncated to fit: R_68K_GOT16O foobar
18453 If this happens, you should recompile your code with @option{-mxgot}.
18454 It should then work with very large GOTs. However, code generated with
18455 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
18456 the value of a global symbol.
18458 Note that some linkers, including newer versions of the GNU linker,
18459 can create multiple GOTs and sort GOT entries. If you have such a linker,
18460 you should only need to use @option{-mxgot} when compiling a single
18461 object file that accesses more than 8192 GOT entries. Very few do.
18463 These options have no effect unless GCC is generating
18464 position-independent code.
18466 @item -mlong-jump-table-offsets
18467 @opindex mlong-jump-table-offsets
18468 Use 32-bit offsets in @code{switch} tables. The default is to use
18473 @node MCore Options
18474 @subsection MCore Options
18475 @cindex MCore options
18477 These are the @samp{-m} options defined for the Motorola M*Core
18483 @itemx -mno-hardlit
18485 @opindex mno-hardlit
18486 Inline constants into the code stream if it can be done in two
18487 instructions or less.
18493 Use the divide instruction. (Enabled by default).
18495 @item -mrelax-immediate
18496 @itemx -mno-relax-immediate
18497 @opindex mrelax-immediate
18498 @opindex mno-relax-immediate
18499 Allow arbitrary-sized immediates in bit operations.
18501 @item -mwide-bitfields
18502 @itemx -mno-wide-bitfields
18503 @opindex mwide-bitfields
18504 @opindex mno-wide-bitfields
18505 Always treat bit-fields as @code{int}-sized.
18507 @item -m4byte-functions
18508 @itemx -mno-4byte-functions
18509 @opindex m4byte-functions
18510 @opindex mno-4byte-functions
18511 Force all functions to be aligned to a 4-byte boundary.
18513 @item -mcallgraph-data
18514 @itemx -mno-callgraph-data
18515 @opindex mcallgraph-data
18516 @opindex mno-callgraph-data
18517 Emit callgraph information.
18520 @itemx -mno-slow-bytes
18521 @opindex mslow-bytes
18522 @opindex mno-slow-bytes
18523 Prefer word access when reading byte quantities.
18525 @item -mlittle-endian
18526 @itemx -mbig-endian
18527 @opindex mlittle-endian
18528 @opindex mbig-endian
18529 Generate code for a little-endian target.
18535 Generate code for the 210 processor.
18539 Assume that runtime support has been provided and so omit the
18540 simulator library (@file{libsim.a)} from the linker command line.
18542 @item -mstack-increment=@var{size}
18543 @opindex mstack-increment
18544 Set the maximum amount for a single stack increment operation. Large
18545 values can increase the speed of programs that contain functions
18546 that need a large amount of stack space, but they can also trigger a
18547 segmentation fault if the stack is extended too much. The default
18553 @subsection MeP Options
18554 @cindex MeP options
18560 Enables the @code{abs} instruction, which is the absolute difference
18561 between two registers.
18565 Enables all the optional instructions---average, multiply, divide, bit
18566 operations, leading zero, absolute difference, min/max, clip, and
18572 Enables the @code{ave} instruction, which computes the average of two
18575 @item -mbased=@var{n}
18577 Variables of size @var{n} bytes or smaller are placed in the
18578 @code{.based} section by default. Based variables use the @code{$tp}
18579 register as a base register, and there is a 128-byte limit to the
18580 @code{.based} section.
18584 Enables the bit operation instructions---bit test (@code{btstm}), set
18585 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
18586 test-and-set (@code{tas}).
18588 @item -mc=@var{name}
18590 Selects which section constant data is placed in. @var{name} may
18591 be @samp{tiny}, @samp{near}, or @samp{far}.
18595 Enables the @code{clip} instruction. Note that @option{-mclip} is not
18596 useful unless you also provide @option{-mminmax}.
18598 @item -mconfig=@var{name}
18600 Selects one of the built-in core configurations. Each MeP chip has
18601 one or more modules in it; each module has a core CPU and a variety of
18602 coprocessors, optional instructions, and peripherals. The
18603 @code{MeP-Integrator} tool, not part of GCC, provides these
18604 configurations through this option; using this option is the same as
18605 using all the corresponding command-line options. The default
18606 configuration is @samp{default}.
18610 Enables the coprocessor instructions. By default, this is a 32-bit
18611 coprocessor. Note that the coprocessor is normally enabled via the
18612 @option{-mconfig=} option.
18616 Enables the 32-bit coprocessor's instructions.
18620 Enables the 64-bit coprocessor's instructions.
18624 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
18628 Causes constant variables to be placed in the @code{.near} section.
18632 Enables the @code{div} and @code{divu} instructions.
18636 Generate big-endian code.
18640 Generate little-endian code.
18642 @item -mio-volatile
18643 @opindex mio-volatile
18644 Tells the compiler that any variable marked with the @code{io}
18645 attribute is to be considered volatile.
18649 Causes variables to be assigned to the @code{.far} section by default.
18653 Enables the @code{leadz} (leading zero) instruction.
18657 Causes variables to be assigned to the @code{.near} section by default.
18661 Enables the @code{min} and @code{max} instructions.
18665 Enables the multiplication and multiply-accumulate instructions.
18669 Disables all the optional instructions enabled by @option{-mall-opts}.
18673 Enables the @code{repeat} and @code{erepeat} instructions, used for
18674 low-overhead looping.
18678 Causes all variables to default to the @code{.tiny} section. Note
18679 that there is a 65536-byte limit to this section. Accesses to these
18680 variables use the @code{%gp} base register.
18684 Enables the saturation instructions. Note that the compiler does not
18685 currently generate these itself, but this option is included for
18686 compatibility with other tools, like @code{as}.
18690 Link the SDRAM-based runtime instead of the default ROM-based runtime.
18694 Link the simulator run-time libraries.
18698 Link the simulator runtime libraries, excluding built-in support
18699 for reset and exception vectors and tables.
18703 Causes all functions to default to the @code{.far} section. Without
18704 this option, functions default to the @code{.near} section.
18706 @item -mtiny=@var{n}
18708 Variables that are @var{n} bytes or smaller are allocated to the
18709 @code{.tiny} section. These variables use the @code{$gp} base
18710 register. The default for this option is 4, but note that there's a
18711 65536-byte limit to the @code{.tiny} section.
18715 @node MicroBlaze Options
18716 @subsection MicroBlaze Options
18717 @cindex MicroBlaze Options
18722 @opindex msoft-float
18723 Use software emulation for floating point (default).
18726 @opindex mhard-float
18727 Use hardware floating-point instructions.
18731 Do not optimize block moves, use @code{memcpy}.
18733 @item -mno-clearbss
18734 @opindex mno-clearbss
18735 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
18737 @item -mcpu=@var{cpu-type}
18739 Use features of, and schedule code for, the given CPU.
18740 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
18741 where @var{X} is a major version, @var{YY} is the minor version, and
18742 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
18743 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
18745 @item -mxl-soft-mul
18746 @opindex mxl-soft-mul
18747 Use software multiply emulation (default).
18749 @item -mxl-soft-div
18750 @opindex mxl-soft-div
18751 Use software emulation for divides (default).
18753 @item -mxl-barrel-shift
18754 @opindex mxl-barrel-shift
18755 Use the hardware barrel shifter.
18757 @item -mxl-pattern-compare
18758 @opindex mxl-pattern-compare
18759 Use pattern compare instructions.
18761 @item -msmall-divides
18762 @opindex msmall-divides
18763 Use table lookup optimization for small signed integer divisions.
18765 @item -mxl-stack-check
18766 @opindex mxl-stack-check
18767 This option is deprecated. Use @option{-fstack-check} instead.
18770 @opindex mxl-gp-opt
18771 Use GP-relative @code{.sdata}/@code{.sbss} sections.
18773 @item -mxl-multiply-high
18774 @opindex mxl-multiply-high
18775 Use multiply high instructions for high part of 32x32 multiply.
18777 @item -mxl-float-convert
18778 @opindex mxl-float-convert
18779 Use hardware floating-point conversion instructions.
18781 @item -mxl-float-sqrt
18782 @opindex mxl-float-sqrt
18783 Use hardware floating-point square root instruction.
18786 @opindex mbig-endian
18787 Generate code for a big-endian target.
18789 @item -mlittle-endian
18790 @opindex mlittle-endian
18791 Generate code for a little-endian target.
18794 @opindex mxl-reorder
18795 Use reorder instructions (swap and byte reversed load/store).
18797 @item -mxl-mode-@var{app-model}
18798 Select application model @var{app-model}. Valid models are
18801 normal executable (default), uses startup code @file{crt0.o}.
18804 for use with Xilinx Microprocessor Debugger (XMD) based
18805 software intrusive debug agent called xmdstub. This uses startup file
18806 @file{crt1.o} and sets the start address of the program to 0x800.
18809 for applications that are loaded using a bootloader.
18810 This model uses startup file @file{crt2.o} which does not contain a processor
18811 reset vector handler. This is suitable for transferring control on a
18812 processor reset to the bootloader rather than the application.
18815 for applications that do not require any of the
18816 MicroBlaze vectors. This option may be useful for applications running
18817 within a monitoring application. This model uses @file{crt3.o} as a startup file.
18820 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
18821 @option{-mxl-mode-@var{app-model}}.
18826 @subsection MIPS Options
18827 @cindex MIPS options
18833 Generate big-endian code.
18837 Generate little-endian code. This is the default for @samp{mips*el-*-*}
18840 @item -march=@var{arch}
18842 Generate code that runs on @var{arch}, which can be the name of a
18843 generic MIPS ISA, or the name of a particular processor.
18845 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
18846 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
18847 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
18848 @samp{mips64r5} and @samp{mips64r6}.
18849 The processor names are:
18850 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
18851 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
18852 @samp{5kc}, @samp{5kf},
18854 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
18855 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
18856 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
18857 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
18858 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
18861 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
18863 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
18864 @samp{m5100}, @samp{m5101},
18865 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
18868 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
18869 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
18870 @samp{rm7000}, @samp{rm9000},
18871 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
18874 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
18875 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
18876 @samp{xlr} and @samp{xlp}.
18877 The special value @samp{from-abi} selects the
18878 most compatible architecture for the selected ABI (that is,
18879 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
18881 The native Linux/GNU toolchain also supports the value @samp{native},
18882 which selects the best architecture option for the host processor.
18883 @option{-march=native} has no effect if GCC does not recognize
18886 In processor names, a final @samp{000} can be abbreviated as @samp{k}
18887 (for example, @option{-march=r2k}). Prefixes are optional, and
18888 @samp{vr} may be written @samp{r}.
18890 Names of the form @samp{@var{n}f2_1} refer to processors with
18891 FPUs clocked at half the rate of the core, names of the form
18892 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
18893 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
18894 processors with FPUs clocked a ratio of 3:2 with respect to the core.
18895 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
18896 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
18897 accepted as synonyms for @samp{@var{n}f1_1}.
18899 GCC defines two macros based on the value of this option. The first
18900 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
18901 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
18902 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
18903 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
18904 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
18906 Note that the @code{_MIPS_ARCH} macro uses the processor names given
18907 above. In other words, it has the full prefix and does not
18908 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
18909 the macro names the resolved architecture (either @code{"mips1"} or
18910 @code{"mips3"}). It names the default architecture when no
18911 @option{-march} option is given.
18913 @item -mtune=@var{arch}
18915 Optimize for @var{arch}. Among other things, this option controls
18916 the way instructions are scheduled, and the perceived cost of arithmetic
18917 operations. The list of @var{arch} values is the same as for
18920 When this option is not used, GCC optimizes for the processor
18921 specified by @option{-march}. By using @option{-march} and
18922 @option{-mtune} together, it is possible to generate code that
18923 runs on a family of processors, but optimize the code for one
18924 particular member of that family.
18926 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
18927 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
18928 @option{-march} ones described above.
18932 Equivalent to @option{-march=mips1}.
18936 Equivalent to @option{-march=mips2}.
18940 Equivalent to @option{-march=mips3}.
18944 Equivalent to @option{-march=mips4}.
18948 Equivalent to @option{-march=mips32}.
18952 Equivalent to @option{-march=mips32r3}.
18956 Equivalent to @option{-march=mips32r5}.
18960 Equivalent to @option{-march=mips32r6}.
18964 Equivalent to @option{-march=mips64}.
18968 Equivalent to @option{-march=mips64r2}.
18972 Equivalent to @option{-march=mips64r3}.
18976 Equivalent to @option{-march=mips64r5}.
18980 Equivalent to @option{-march=mips64r6}.
18985 @opindex mno-mips16
18986 Generate (do not generate) MIPS16 code. If GCC is targeting a
18987 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
18989 MIPS16 code generation can also be controlled on a per-function basis
18990 by means of @code{mips16} and @code{nomips16} attributes.
18991 @xref{Function Attributes}, for more information.
18993 @item -mflip-mips16
18994 @opindex mflip-mips16
18995 Generate MIPS16 code on alternating functions. This option is provided
18996 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
18997 not intended for ordinary use in compiling user code.
18999 @item -minterlink-compressed
19000 @item -mno-interlink-compressed
19001 @opindex minterlink-compressed
19002 @opindex mno-interlink-compressed
19003 Require (do not require) that code using the standard (uncompressed) MIPS ISA
19004 be link-compatible with MIPS16 and microMIPS code, and vice versa.
19006 For example, code using the standard ISA encoding cannot jump directly
19007 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
19008 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
19009 knows that the target of the jump is not compressed.
19011 @item -minterlink-mips16
19012 @itemx -mno-interlink-mips16
19013 @opindex minterlink-mips16
19014 @opindex mno-interlink-mips16
19015 Aliases of @option{-minterlink-compressed} and
19016 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
19017 and are retained for backwards compatibility.
19029 Generate code for the given ABI@.
19031 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
19032 generates 64-bit code when you select a 64-bit architecture, but you
19033 can use @option{-mgp32} to get 32-bit code instead.
19035 For information about the O64 ABI, see
19036 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
19038 GCC supports a variant of the o32 ABI in which floating-point registers
19039 are 64 rather than 32 bits wide. You can select this combination with
19040 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
19041 and @code{mfhc1} instructions and is therefore only supported for
19042 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
19044 The register assignments for arguments and return values remain the
19045 same, but each scalar value is passed in a single 64-bit register
19046 rather than a pair of 32-bit registers. For example, scalar
19047 floating-point values are returned in @samp{$f0} only, not a
19048 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
19049 remains the same in that the even-numbered double-precision registers
19052 Two additional variants of the o32 ABI are supported to enable
19053 a transition from 32-bit to 64-bit registers. These are FPXX
19054 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
19055 The FPXX extension mandates that all code must execute correctly
19056 when run using 32-bit or 64-bit registers. The code can be interlinked
19057 with either FP32 or FP64, but not both.
19058 The FP64A extension is similar to the FP64 extension but forbids the
19059 use of odd-numbered single-precision registers. This can be used
19060 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
19061 processors and allows both FP32 and FP64A code to interlink and
19062 run in the same process without changing FPU modes.
19065 @itemx -mno-abicalls
19067 @opindex mno-abicalls
19068 Generate (do not generate) code that is suitable for SVR4-style
19069 dynamic objects. @option{-mabicalls} is the default for SVR4-based
19074 Generate (do not generate) code that is fully position-independent,
19075 and that can therefore be linked into shared libraries. This option
19076 only affects @option{-mabicalls}.
19078 All @option{-mabicalls} code has traditionally been position-independent,
19079 regardless of options like @option{-fPIC} and @option{-fpic}. However,
19080 as an extension, the GNU toolchain allows executables to use absolute
19081 accesses for locally-binding symbols. It can also use shorter GP
19082 initialization sequences and generate direct calls to locally-defined
19083 functions. This mode is selected by @option{-mno-shared}.
19085 @option{-mno-shared} depends on binutils 2.16 or higher and generates
19086 objects that can only be linked by the GNU linker. However, the option
19087 does not affect the ABI of the final executable; it only affects the ABI
19088 of relocatable objects. Using @option{-mno-shared} generally makes
19089 executables both smaller and quicker.
19091 @option{-mshared} is the default.
19097 Assume (do not assume) that the static and dynamic linkers
19098 support PLTs and copy relocations. This option only affects
19099 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
19100 has no effect without @option{-msym32}.
19102 You can make @option{-mplt} the default by configuring
19103 GCC with @option{--with-mips-plt}. The default is
19104 @option{-mno-plt} otherwise.
19110 Lift (do not lift) the usual restrictions on the size of the global
19113 GCC normally uses a single instruction to load values from the GOT@.
19114 While this is relatively efficient, it only works if the GOT
19115 is smaller than about 64k. Anything larger causes the linker
19116 to report an error such as:
19118 @cindex relocation truncated to fit (MIPS)
19120 relocation truncated to fit: R_MIPS_GOT16 foobar
19123 If this happens, you should recompile your code with @option{-mxgot}.
19124 This works with very large GOTs, although the code is also
19125 less efficient, since it takes three instructions to fetch the
19126 value of a global symbol.
19128 Note that some linkers can create multiple GOTs. If you have such a
19129 linker, you should only need to use @option{-mxgot} when a single object
19130 file accesses more than 64k's worth of GOT entries. Very few do.
19132 These options have no effect unless GCC is generating position
19137 Assume that general-purpose registers are 32 bits wide.
19141 Assume that general-purpose registers are 64 bits wide.
19145 Assume that floating-point registers are 32 bits wide.
19149 Assume that floating-point registers are 64 bits wide.
19153 Do not assume the width of floating-point registers.
19156 @opindex mhard-float
19157 Use floating-point coprocessor instructions.
19160 @opindex msoft-float
19161 Do not use floating-point coprocessor instructions. Implement
19162 floating-point calculations using library calls instead.
19166 Equivalent to @option{-msoft-float}, but additionally asserts that the
19167 program being compiled does not perform any floating-point operations.
19168 This option is presently supported only by some bare-metal MIPS
19169 configurations, where it may select a special set of libraries
19170 that lack all floating-point support (including, for example, the
19171 floating-point @code{printf} formats).
19172 If code compiled with @option{-mno-float} accidentally contains
19173 floating-point operations, it is likely to suffer a link-time
19174 or run-time failure.
19176 @item -msingle-float
19177 @opindex msingle-float
19178 Assume that the floating-point coprocessor only supports single-precision
19181 @item -mdouble-float
19182 @opindex mdouble-float
19183 Assume that the floating-point coprocessor supports double-precision
19184 operations. This is the default.
19187 @itemx -mno-odd-spreg
19188 @opindex modd-spreg
19189 @opindex mno-odd-spreg
19190 Enable the use of odd-numbered single-precision floating-point registers
19191 for the o32 ABI. This is the default for processors that are known to
19192 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
19196 @itemx -mabs=legacy
19198 @opindex mabs=legacy
19199 These options control the treatment of the special not-a-number (NaN)
19200 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
19201 @code{neg.@i{fmt}} machine instructions.
19203 By default or when @option{-mabs=legacy} is used the legacy
19204 treatment is selected. In this case these instructions are considered
19205 arithmetic and avoided where correct operation is required and the
19206 input operand might be a NaN. A longer sequence of instructions that
19207 manipulate the sign bit of floating-point datum manually is used
19208 instead unless the @option{-ffinite-math-only} option has also been
19211 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
19212 this case these instructions are considered non-arithmetic and therefore
19213 operating correctly in all cases, including in particular where the
19214 input operand is a NaN. These instructions are therefore always used
19215 for the respective operations.
19218 @itemx -mnan=legacy
19220 @opindex mnan=legacy
19221 These options control the encoding of the special not-a-number (NaN)
19222 IEEE 754 floating-point data.
19224 The @option{-mnan=legacy} option selects the legacy encoding. In this
19225 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
19226 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
19227 by the first bit of their trailing significand field being 1.
19229 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
19230 this case qNaNs are denoted by the first bit of their trailing
19231 significand field being 1, whereas sNaNs are denoted by the first bit of
19232 their trailing significand field being 0.
19234 The default is @option{-mnan=legacy} unless GCC has been configured with
19235 @option{--with-nan=2008}.
19241 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
19242 implement atomic memory built-in functions. When neither option is
19243 specified, GCC uses the instructions if the target architecture
19246 @option{-mllsc} is useful if the runtime environment can emulate the
19247 instructions and @option{-mno-llsc} can be useful when compiling for
19248 nonstandard ISAs. You can make either option the default by
19249 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
19250 respectively. @option{--with-llsc} is the default for some
19251 configurations; see the installation documentation for details.
19257 Use (do not use) revision 1 of the MIPS DSP ASE@.
19258 @xref{MIPS DSP Built-in Functions}. This option defines the
19259 preprocessor macro @code{__mips_dsp}. It also defines
19260 @code{__mips_dsp_rev} to 1.
19266 Use (do not use) revision 2 of the MIPS DSP ASE@.
19267 @xref{MIPS DSP Built-in Functions}. This option defines the
19268 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
19269 It also defines @code{__mips_dsp_rev} to 2.
19272 @itemx -mno-smartmips
19273 @opindex msmartmips
19274 @opindex mno-smartmips
19275 Use (do not use) the MIPS SmartMIPS ASE.
19277 @item -mpaired-single
19278 @itemx -mno-paired-single
19279 @opindex mpaired-single
19280 @opindex mno-paired-single
19281 Use (do not use) paired-single floating-point instructions.
19282 @xref{MIPS Paired-Single Support}. This option requires
19283 hardware floating-point support to be enabled.
19289 Use (do not use) MIPS Digital Media Extension instructions.
19290 This option can only be used when generating 64-bit code and requires
19291 hardware floating-point support to be enabled.
19296 @opindex mno-mips3d
19297 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
19298 The option @option{-mips3d} implies @option{-mpaired-single}.
19301 @itemx -mno-micromips
19302 @opindex mmicromips
19303 @opindex mno-mmicromips
19304 Generate (do not generate) microMIPS code.
19306 MicroMIPS code generation can also be controlled on a per-function basis
19307 by means of @code{micromips} and @code{nomicromips} attributes.
19308 @xref{Function Attributes}, for more information.
19314 Use (do not use) MT Multithreading instructions.
19320 Use (do not use) the MIPS MCU ASE instructions.
19326 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
19332 Use (do not use) the MIPS Virtualization (VZ) instructions.
19338 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
19342 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
19343 an explanation of the default and the way that the pointer size is
19348 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
19350 The default size of @code{int}s, @code{long}s and pointers depends on
19351 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
19352 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
19353 32-bit @code{long}s. Pointers are the same size as @code{long}s,
19354 or the same size as integer registers, whichever is smaller.
19360 Assume (do not assume) that all symbols have 32-bit values, regardless
19361 of the selected ABI@. This option is useful in combination with
19362 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
19363 to generate shorter and faster references to symbolic addresses.
19367 Put definitions of externally-visible data in a small data section
19368 if that data is no bigger than @var{num} bytes. GCC can then generate
19369 more efficient accesses to the data; see @option{-mgpopt} for details.
19371 The default @option{-G} option depends on the configuration.
19373 @item -mlocal-sdata
19374 @itemx -mno-local-sdata
19375 @opindex mlocal-sdata
19376 @opindex mno-local-sdata
19377 Extend (do not extend) the @option{-G} behavior to local data too,
19378 such as to static variables in C@. @option{-mlocal-sdata} is the
19379 default for all configurations.
19381 If the linker complains that an application is using too much small data,
19382 you might want to try rebuilding the less performance-critical parts with
19383 @option{-mno-local-sdata}. You might also want to build large
19384 libraries with @option{-mno-local-sdata}, so that the libraries leave
19385 more room for the main program.
19387 @item -mextern-sdata
19388 @itemx -mno-extern-sdata
19389 @opindex mextern-sdata
19390 @opindex mno-extern-sdata
19391 Assume (do not assume) that externally-defined data is in
19392 a small data section if the size of that data is within the @option{-G} limit.
19393 @option{-mextern-sdata} is the default for all configurations.
19395 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
19396 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
19397 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
19398 is placed in a small data section. If @var{Var} is defined by another
19399 module, you must either compile that module with a high-enough
19400 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
19401 definition. If @var{Var} is common, you must link the application
19402 with a high-enough @option{-G} setting.
19404 The easiest way of satisfying these restrictions is to compile
19405 and link every module with the same @option{-G} option. However,
19406 you may wish to build a library that supports several different
19407 small data limits. You can do this by compiling the library with
19408 the highest supported @option{-G} setting and additionally using
19409 @option{-mno-extern-sdata} to stop the library from making assumptions
19410 about externally-defined data.
19416 Use (do not use) GP-relative accesses for symbols that are known to be
19417 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
19418 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
19421 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
19422 might not hold the value of @code{_gp}. For example, if the code is
19423 part of a library that might be used in a boot monitor, programs that
19424 call boot monitor routines pass an unknown value in @code{$gp}.
19425 (In such situations, the boot monitor itself is usually compiled
19426 with @option{-G0}.)
19428 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
19429 @option{-mno-extern-sdata}.
19431 @item -membedded-data
19432 @itemx -mno-embedded-data
19433 @opindex membedded-data
19434 @opindex mno-embedded-data
19435 Allocate variables to the read-only data section first if possible, then
19436 next in the small data section if possible, otherwise in data. This gives
19437 slightly slower code than the default, but reduces the amount of RAM required
19438 when executing, and thus may be preferred for some embedded systems.
19440 @item -muninit-const-in-rodata
19441 @itemx -mno-uninit-const-in-rodata
19442 @opindex muninit-const-in-rodata
19443 @opindex mno-uninit-const-in-rodata
19444 Put uninitialized @code{const} variables in the read-only data section.
19445 This option is only meaningful in conjunction with @option{-membedded-data}.
19447 @item -mcode-readable=@var{setting}
19448 @opindex mcode-readable
19449 Specify whether GCC may generate code that reads from executable sections.
19450 There are three possible settings:
19453 @item -mcode-readable=yes
19454 Instructions may freely access executable sections. This is the
19457 @item -mcode-readable=pcrel
19458 MIPS16 PC-relative load instructions can access executable sections,
19459 but other instructions must not do so. This option is useful on 4KSc
19460 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
19461 It is also useful on processors that can be configured to have a dual
19462 instruction/data SRAM interface and that, like the M4K, automatically
19463 redirect PC-relative loads to the instruction RAM.
19465 @item -mcode-readable=no
19466 Instructions must not access executable sections. This option can be
19467 useful on targets that are configured to have a dual instruction/data
19468 SRAM interface but that (unlike the M4K) do not automatically redirect
19469 PC-relative loads to the instruction RAM.
19472 @item -msplit-addresses
19473 @itemx -mno-split-addresses
19474 @opindex msplit-addresses
19475 @opindex mno-split-addresses
19476 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
19477 relocation operators. This option has been superseded by
19478 @option{-mexplicit-relocs} but is retained for backwards compatibility.
19480 @item -mexplicit-relocs
19481 @itemx -mno-explicit-relocs
19482 @opindex mexplicit-relocs
19483 @opindex mno-explicit-relocs
19484 Use (do not use) assembler relocation operators when dealing with symbolic
19485 addresses. The alternative, selected by @option{-mno-explicit-relocs},
19486 is to use assembler macros instead.
19488 @option{-mexplicit-relocs} is the default if GCC was configured
19489 to use an assembler that supports relocation operators.
19491 @item -mcheck-zero-division
19492 @itemx -mno-check-zero-division
19493 @opindex mcheck-zero-division
19494 @opindex mno-check-zero-division
19495 Trap (do not trap) on integer division by zero.
19497 The default is @option{-mcheck-zero-division}.
19499 @item -mdivide-traps
19500 @itemx -mdivide-breaks
19501 @opindex mdivide-traps
19502 @opindex mdivide-breaks
19503 MIPS systems check for division by zero by generating either a
19504 conditional trap or a break instruction. Using traps results in
19505 smaller code, but is only supported on MIPS II and later. Also, some
19506 versions of the Linux kernel have a bug that prevents trap from
19507 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
19508 allow conditional traps on architectures that support them and
19509 @option{-mdivide-breaks} to force the use of breaks.
19511 The default is usually @option{-mdivide-traps}, but this can be
19512 overridden at configure time using @option{--with-divide=breaks}.
19513 Divide-by-zero checks can be completely disabled using
19514 @option{-mno-check-zero-division}.
19516 @item -mload-store-pairs
19517 @itemx -mno-load-store-pairs
19518 @opindex mload-store-pairs
19519 @opindex mno-load-store-pairs
19520 Enable (disable) an optimization that pairs consecutive load or store
19521 instructions to enable load/store bonding. This option is enabled by
19522 default but only takes effect when the selected architecture is known
19523 to support bonding.
19528 @opindex mno-memcpy
19529 Force (do not force) the use of @code{memcpy} for non-trivial block
19530 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
19531 most constant-sized copies.
19534 @itemx -mno-long-calls
19535 @opindex mlong-calls
19536 @opindex mno-long-calls
19537 Disable (do not disable) use of the @code{jal} instruction. Calling
19538 functions using @code{jal} is more efficient but requires the caller
19539 and callee to be in the same 256 megabyte segment.
19541 This option has no effect on abicalls code. The default is
19542 @option{-mno-long-calls}.
19548 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
19549 instructions, as provided by the R4650 ISA@.
19555 Enable (disable) use of the @code{madd} and @code{msub} integer
19556 instructions. The default is @option{-mimadd} on architectures
19557 that support @code{madd} and @code{msub} except for the 74k
19558 architecture where it was found to generate slower code.
19561 @itemx -mno-fused-madd
19562 @opindex mfused-madd
19563 @opindex mno-fused-madd
19564 Enable (disable) use of the floating-point multiply-accumulate
19565 instructions, when they are available. The default is
19566 @option{-mfused-madd}.
19568 On the R8000 CPU when multiply-accumulate instructions are used,
19569 the intermediate product is calculated to infinite precision
19570 and is not subject to the FCSR Flush to Zero bit. This may be
19571 undesirable in some circumstances. On other processors the result
19572 is numerically identical to the equivalent computation using
19573 separate multiply, add, subtract and negate instructions.
19577 Tell the MIPS assembler to not run its preprocessor over user
19578 assembler files (with a @samp{.s} suffix) when assembling them.
19583 @opindex mno-fix-24k
19584 Work around the 24K E48 (lost data on stores during refill) errata.
19585 The workarounds are implemented by the assembler rather than by GCC@.
19588 @itemx -mno-fix-r4000
19589 @opindex mfix-r4000
19590 @opindex mno-fix-r4000
19591 Work around certain R4000 CPU errata:
19594 A double-word or a variable shift may give an incorrect result if executed
19595 immediately after starting an integer division.
19597 A double-word or a variable shift may give an incorrect result if executed
19598 while an integer multiplication is in progress.
19600 An integer division may give an incorrect result if started in a delay slot
19601 of a taken branch or a jump.
19605 @itemx -mno-fix-r4400
19606 @opindex mfix-r4400
19607 @opindex mno-fix-r4400
19608 Work around certain R4400 CPU errata:
19611 A double-word or a variable shift may give an incorrect result if executed
19612 immediately after starting an integer division.
19616 @itemx -mno-fix-r10000
19617 @opindex mfix-r10000
19618 @opindex mno-fix-r10000
19619 Work around certain R10000 errata:
19622 @code{ll}/@code{sc} sequences may not behave atomically on revisions
19623 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
19626 This option can only be used if the target architecture supports
19627 branch-likely instructions. @option{-mfix-r10000} is the default when
19628 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
19632 @itemx -mno-fix-rm7000
19633 @opindex mfix-rm7000
19634 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
19635 workarounds are implemented by the assembler rather than by GCC@.
19638 @itemx -mno-fix-vr4120
19639 @opindex mfix-vr4120
19640 Work around certain VR4120 errata:
19643 @code{dmultu} does not always produce the correct result.
19645 @code{div} and @code{ddiv} do not always produce the correct result if one
19646 of the operands is negative.
19648 The workarounds for the division errata rely on special functions in
19649 @file{libgcc.a}. At present, these functions are only provided by
19650 the @code{mips64vr*-elf} configurations.
19652 Other VR4120 errata require a NOP to be inserted between certain pairs of
19653 instructions. These errata are handled by the assembler, not by GCC itself.
19656 @opindex mfix-vr4130
19657 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
19658 workarounds are implemented by the assembler rather than by GCC,
19659 although GCC avoids using @code{mflo} and @code{mfhi} if the
19660 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
19661 instructions are available instead.
19664 @itemx -mno-fix-sb1
19666 Work around certain SB-1 CPU core errata.
19667 (This flag currently works around the SB-1 revision 2
19668 ``F1'' and ``F2'' floating-point errata.)
19670 @item -mr10k-cache-barrier=@var{setting}
19671 @opindex mr10k-cache-barrier
19672 Specify whether GCC should insert cache barriers to avoid the
19673 side-effects of speculation on R10K processors.
19675 In common with many processors, the R10K tries to predict the outcome
19676 of a conditional branch and speculatively executes instructions from
19677 the ``taken'' branch. It later aborts these instructions if the
19678 predicted outcome is wrong. However, on the R10K, even aborted
19679 instructions can have side effects.
19681 This problem only affects kernel stores and, depending on the system,
19682 kernel loads. As an example, a speculatively-executed store may load
19683 the target memory into cache and mark the cache line as dirty, even if
19684 the store itself is later aborted. If a DMA operation writes to the
19685 same area of memory before the ``dirty'' line is flushed, the cached
19686 data overwrites the DMA-ed data. See the R10K processor manual
19687 for a full description, including other potential problems.
19689 One workaround is to insert cache barrier instructions before every memory
19690 access that might be speculatively executed and that might have side
19691 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
19692 controls GCC's implementation of this workaround. It assumes that
19693 aborted accesses to any byte in the following regions does not have
19698 the memory occupied by the current function's stack frame;
19701 the memory occupied by an incoming stack argument;
19704 the memory occupied by an object with a link-time-constant address.
19707 It is the kernel's responsibility to ensure that speculative
19708 accesses to these regions are indeed safe.
19710 If the input program contains a function declaration such as:
19716 then the implementation of @code{foo} must allow @code{j foo} and
19717 @code{jal foo} to be executed speculatively. GCC honors this
19718 restriction for functions it compiles itself. It expects non-GCC
19719 functions (such as hand-written assembly code) to do the same.
19721 The option has three forms:
19724 @item -mr10k-cache-barrier=load-store
19725 Insert a cache barrier before a load or store that might be
19726 speculatively executed and that might have side effects even
19729 @item -mr10k-cache-barrier=store
19730 Insert a cache barrier before a store that might be speculatively
19731 executed and that might have side effects even if aborted.
19733 @item -mr10k-cache-barrier=none
19734 Disable the insertion of cache barriers. This is the default setting.
19737 @item -mflush-func=@var{func}
19738 @itemx -mno-flush-func
19739 @opindex mflush-func
19740 Specifies the function to call to flush the I and D caches, or to not
19741 call any such function. If called, the function must take the same
19742 arguments as the common @code{_flush_func}, that is, the address of the
19743 memory range for which the cache is being flushed, the size of the
19744 memory range, and the number 3 (to flush both caches). The default
19745 depends on the target GCC was configured for, but commonly is either
19746 @code{_flush_func} or @code{__cpu_flush}.
19748 @item mbranch-cost=@var{num}
19749 @opindex mbranch-cost
19750 Set the cost of branches to roughly @var{num} ``simple'' instructions.
19751 This cost is only a heuristic and is not guaranteed to produce
19752 consistent results across releases. A zero cost redundantly selects
19753 the default, which is based on the @option{-mtune} setting.
19755 @item -mbranch-likely
19756 @itemx -mno-branch-likely
19757 @opindex mbranch-likely
19758 @opindex mno-branch-likely
19759 Enable or disable use of Branch Likely instructions, regardless of the
19760 default for the selected architecture. By default, Branch Likely
19761 instructions may be generated if they are supported by the selected
19762 architecture. An exception is for the MIPS32 and MIPS64 architectures
19763 and processors that implement those architectures; for those, Branch
19764 Likely instructions are not be generated by default because the MIPS32
19765 and MIPS64 architectures specifically deprecate their use.
19767 @item -mcompact-branches=never
19768 @itemx -mcompact-branches=optimal
19769 @itemx -mcompact-branches=always
19770 @opindex mcompact-branches=never
19771 @opindex mcompact-branches=optimal
19772 @opindex mcompact-branches=always
19773 These options control which form of branches will be generated. The
19774 default is @option{-mcompact-branches=optimal}.
19776 The @option{-mcompact-branches=never} option ensures that compact branch
19777 instructions will never be generated.
19779 The @option{-mcompact-branches=always} option ensures that a compact
19780 branch instruction will be generated if available. If a compact branch
19781 instruction is not available, a delay slot form of the branch will be
19784 This option is supported from MIPS Release 6 onwards.
19786 The @option{-mcompact-branches=optimal} option will cause a delay slot
19787 branch to be used if one is available in the current ISA and the delay
19788 slot is successfully filled. If the delay slot is not filled, a compact
19789 branch will be chosen if one is available.
19791 @item -mfp-exceptions
19792 @itemx -mno-fp-exceptions
19793 @opindex mfp-exceptions
19794 Specifies whether FP exceptions are enabled. This affects how
19795 FP instructions are scheduled for some processors.
19796 The default is that FP exceptions are
19799 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
19800 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
19803 @item -mvr4130-align
19804 @itemx -mno-vr4130-align
19805 @opindex mvr4130-align
19806 The VR4130 pipeline is two-way superscalar, but can only issue two
19807 instructions together if the first one is 8-byte aligned. When this
19808 option is enabled, GCC aligns pairs of instructions that it
19809 thinks should execute in parallel.
19811 This option only has an effect when optimizing for the VR4130.
19812 It normally makes code faster, but at the expense of making it bigger.
19813 It is enabled by default at optimization level @option{-O3}.
19818 Enable (disable) generation of @code{synci} instructions on
19819 architectures that support it. The @code{synci} instructions (if
19820 enabled) are generated when @code{__builtin___clear_cache} is
19823 This option defaults to @option{-mno-synci}, but the default can be
19824 overridden by configuring GCC with @option{--with-synci}.
19826 When compiling code for single processor systems, it is generally safe
19827 to use @code{synci}. However, on many multi-core (SMP) systems, it
19828 does not invalidate the instruction caches on all cores and may lead
19829 to undefined behavior.
19831 @item -mrelax-pic-calls
19832 @itemx -mno-relax-pic-calls
19833 @opindex mrelax-pic-calls
19834 Try to turn PIC calls that are normally dispatched via register
19835 @code{$25} into direct calls. This is only possible if the linker can
19836 resolve the destination at link time and if the destination is within
19837 range for a direct call.
19839 @option{-mrelax-pic-calls} is the default if GCC was configured to use
19840 an assembler and a linker that support the @code{.reloc} assembly
19841 directive and @option{-mexplicit-relocs} is in effect. With
19842 @option{-mno-explicit-relocs}, this optimization can be performed by the
19843 assembler and the linker alone without help from the compiler.
19845 @item -mmcount-ra-address
19846 @itemx -mno-mcount-ra-address
19847 @opindex mmcount-ra-address
19848 @opindex mno-mcount-ra-address
19849 Emit (do not emit) code that allows @code{_mcount} to modify the
19850 calling function's return address. When enabled, this option extends
19851 the usual @code{_mcount} interface with a new @var{ra-address}
19852 parameter, which has type @code{intptr_t *} and is passed in register
19853 @code{$12}. @code{_mcount} can then modify the return address by
19854 doing both of the following:
19857 Returning the new address in register @code{$31}.
19859 Storing the new address in @code{*@var{ra-address}},
19860 if @var{ra-address} is nonnull.
19863 The default is @option{-mno-mcount-ra-address}.
19865 @item -mframe-header-opt
19866 @itemx -mno-frame-header-opt
19867 @opindex mframe-header-opt
19868 Enable (disable) frame header optimization in the o32 ABI. When using the
19869 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
19870 function to write out register arguments. When enabled, this optimization
19871 will suppress the allocation of the frame header if it can be determined that
19874 This optimization is off by default at all optimization levels.
19877 @itemx -mno-lxc1-sxc1
19878 @opindex mlxc1-sxc1
19879 When applicable, enable (disable) the generation of @code{lwxc1},
19880 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
19885 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
19886 @code{madd.d} and related instructions. Enabled by default.
19891 @subsection MMIX Options
19892 @cindex MMIX Options
19894 These options are defined for the MMIX:
19898 @itemx -mno-libfuncs
19900 @opindex mno-libfuncs
19901 Specify that intrinsic library functions are being compiled, passing all
19902 values in registers, no matter the size.
19905 @itemx -mno-epsilon
19907 @opindex mno-epsilon
19908 Generate floating-point comparison instructions that compare with respect
19909 to the @code{rE} epsilon register.
19911 @item -mabi=mmixware
19913 @opindex mabi=mmixware
19915 Generate code that passes function parameters and return values that (in
19916 the called function) are seen as registers @code{$0} and up, as opposed to
19917 the GNU ABI which uses global registers @code{$231} and up.
19919 @item -mzero-extend
19920 @itemx -mno-zero-extend
19921 @opindex mzero-extend
19922 @opindex mno-zero-extend
19923 When reading data from memory in sizes shorter than 64 bits, use (do not
19924 use) zero-extending load instructions by default, rather than
19925 sign-extending ones.
19928 @itemx -mno-knuthdiv
19930 @opindex mno-knuthdiv
19931 Make the result of a division yielding a remainder have the same sign as
19932 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
19933 remainder follows the sign of the dividend. Both methods are
19934 arithmetically valid, the latter being almost exclusively used.
19936 @item -mtoplevel-symbols
19937 @itemx -mno-toplevel-symbols
19938 @opindex mtoplevel-symbols
19939 @opindex mno-toplevel-symbols
19940 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
19941 code can be used with the @code{PREFIX} assembly directive.
19945 Generate an executable in the ELF format, rather than the default
19946 @samp{mmo} format used by the @command{mmix} simulator.
19948 @item -mbranch-predict
19949 @itemx -mno-branch-predict
19950 @opindex mbranch-predict
19951 @opindex mno-branch-predict
19952 Use (do not use) the probable-branch instructions, when static branch
19953 prediction indicates a probable branch.
19955 @item -mbase-addresses
19956 @itemx -mno-base-addresses
19957 @opindex mbase-addresses
19958 @opindex mno-base-addresses
19959 Generate (do not generate) code that uses @emph{base addresses}. Using a
19960 base address automatically generates a request (handled by the assembler
19961 and the linker) for a constant to be set up in a global register. The
19962 register is used for one or more base address requests within the range 0
19963 to 255 from the value held in the register. The generally leads to short
19964 and fast code, but the number of different data items that can be
19965 addressed is limited. This means that a program that uses lots of static
19966 data may require @option{-mno-base-addresses}.
19968 @item -msingle-exit
19969 @itemx -mno-single-exit
19970 @opindex msingle-exit
19971 @opindex mno-single-exit
19972 Force (do not force) generated code to have a single exit point in each
19976 @node MN10300 Options
19977 @subsection MN10300 Options
19978 @cindex MN10300 options
19980 These @option{-m} options are defined for Matsushita MN10300 architectures:
19985 Generate code to avoid bugs in the multiply instructions for the MN10300
19986 processors. This is the default.
19988 @item -mno-mult-bug
19989 @opindex mno-mult-bug
19990 Do not generate code to avoid bugs in the multiply instructions for the
19991 MN10300 processors.
19995 Generate code using features specific to the AM33 processor.
19999 Do not generate code using features specific to the AM33 processor. This
20004 Generate code using features specific to the AM33/2.0 processor.
20008 Generate code using features specific to the AM34 processor.
20010 @item -mtune=@var{cpu-type}
20012 Use the timing characteristics of the indicated CPU type when
20013 scheduling instructions. This does not change the targeted processor
20014 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
20015 @samp{am33-2} or @samp{am34}.
20017 @item -mreturn-pointer-on-d0
20018 @opindex mreturn-pointer-on-d0
20019 When generating a function that returns a pointer, return the pointer
20020 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
20021 only in @code{a0}, and attempts to call such functions without a prototype
20022 result in errors. Note that this option is on by default; use
20023 @option{-mno-return-pointer-on-d0} to disable it.
20027 Do not link in the C run-time initialization object file.
20031 Indicate to the linker that it should perform a relaxation optimization pass
20032 to shorten branches, calls and absolute memory addresses. This option only
20033 has an effect when used on the command line for the final link step.
20035 This option makes symbolic debugging impossible.
20039 Allow the compiler to generate @emph{Long Instruction Word}
20040 instructions if the target is the @samp{AM33} or later. This is the
20041 default. This option defines the preprocessor macro @code{__LIW__}.
20045 Do not allow the compiler to generate @emph{Long Instruction Word}
20046 instructions. This option defines the preprocessor macro
20051 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
20052 instructions if the target is the @samp{AM33} or later. This is the
20053 default. This option defines the preprocessor macro @code{__SETLB__}.
20057 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
20058 instructions. This option defines the preprocessor macro
20059 @code{__NO_SETLB__}.
20063 @node Moxie Options
20064 @subsection Moxie Options
20065 @cindex Moxie Options
20071 Generate big-endian code. This is the default for @samp{moxie-*-*}
20076 Generate little-endian code.
20080 Generate mul.x and umul.x instructions. This is the default for
20081 @samp{moxiebox-*-*} configurations.
20085 Do not link in the C run-time initialization object file.
20089 @node MSP430 Options
20090 @subsection MSP430 Options
20091 @cindex MSP430 Options
20093 These options are defined for the MSP430:
20099 Force assembly output to always use hex constants. Normally such
20100 constants are signed decimals, but this option is available for
20101 testsuite and/or aesthetic purposes.
20105 Select the MCU to target. This is used to create a C preprocessor
20106 symbol based upon the MCU name, converted to upper case and pre- and
20107 post-fixed with @samp{__}. This in turn is used by the
20108 @file{msp430.h} header file to select an MCU-specific supplementary
20111 The option also sets the ISA to use. If the MCU name is one that is
20112 known to only support the 430 ISA then that is selected, otherwise the
20113 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
20114 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
20115 name selects the 430X ISA.
20117 In addition an MCU-specific linker script is added to the linker
20118 command line. The script's name is the name of the MCU with
20119 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
20120 command line defines the C preprocessor symbol @code{__XXX__} and
20121 cause the linker to search for a script called @file{xxx.ld}.
20123 This option is also passed on to the assembler.
20126 @itemx -mno-warn-mcu
20128 @opindex mno-warn-mcu
20129 This option enables or disables warnings about conflicts between the
20130 MCU name specified by the @option{-mmcu} option and the ISA set by the
20131 @option{-mcpu} option and/or the hardware multiply support set by the
20132 @option{-mhwmult} option. It also toggles warnings about unrecognized
20133 MCU names. This option is on by default.
20137 Specifies the ISA to use. Accepted values are @samp{msp430},
20138 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
20139 @option{-mmcu=} option should be used to select the ISA.
20143 Link to the simulator runtime libraries and linker script. Overrides
20144 any scripts that would be selected by the @option{-mmcu=} option.
20148 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
20152 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
20156 This option is passed to the assembler and linker, and allows the
20157 linker to perform certain optimizations that cannot be done until
20162 Describes the type of hardware multiply supported by the target.
20163 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
20164 for the original 16-bit-only multiply supported by early MCUs.
20165 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
20166 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
20167 A value of @samp{auto} can also be given. This tells GCC to deduce
20168 the hardware multiply support based upon the MCU name provided by the
20169 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
20170 the MCU name is not recognized then no hardware multiply support is
20171 assumed. @code{auto} is the default setting.
20173 Hardware multiplies are normally performed by calling a library
20174 routine. This saves space in the generated code. When compiling at
20175 @option{-O3} or higher however the hardware multiplier is invoked
20176 inline. This makes for bigger, but faster code.
20178 The hardware multiply routines disable interrupts whilst running and
20179 restore the previous interrupt state when they finish. This makes
20180 them safe to use inside interrupt handlers as well as in normal code.
20184 Enable the use of a minimum runtime environment - no static
20185 initializers or constructors. This is intended for memory-constrained
20186 devices. The compiler includes special symbols in some objects
20187 that tell the linker and runtime which code fragments are required.
20189 @item -mcode-region=
20190 @itemx -mdata-region=
20191 @opindex mcode-region
20192 @opindex mdata-region
20193 These options tell the compiler where to place functions and data that
20194 do not have one of the @code{lower}, @code{upper}, @code{either} or
20195 @code{section} attributes. Possible values are @code{lower},
20196 @code{upper}, @code{either} or @code{any}. The first three behave
20197 like the corresponding attribute. The fourth possible value -
20198 @code{any} - is the default. It leaves placement entirely up to the
20199 linker script and how it assigns the standard sections
20200 (@code{.text}, @code{.data}, etc) to the memory regions.
20202 @item -msilicon-errata=
20203 @opindex msilicon-errata
20204 This option passes on a request to assembler to enable the fixes for
20205 the named silicon errata.
20207 @item -msilicon-errata-warn=
20208 @opindex msilicon-errata-warn
20209 This option passes on a request to the assembler to enable warning
20210 messages when a silicon errata might need to be applied.
20214 @node NDS32 Options
20215 @subsection NDS32 Options
20216 @cindex NDS32 Options
20218 These options are defined for NDS32 implementations:
20223 @opindex mbig-endian
20224 Generate code in big-endian mode.
20226 @item -mlittle-endian
20227 @opindex mlittle-endian
20228 Generate code in little-endian mode.
20230 @item -mreduced-regs
20231 @opindex mreduced-regs
20232 Use reduced-set registers for register allocation.
20235 @opindex mfull-regs
20236 Use full-set registers for register allocation.
20240 Generate conditional move instructions.
20244 Do not generate conditional move instructions.
20248 Generate performance extension instructions.
20250 @item -mno-perf-ext
20251 @opindex mno-perf-ext
20252 Do not generate performance extension instructions.
20256 Generate v3 push25/pop25 instructions.
20259 @opindex mno-v3push
20260 Do not generate v3 push25/pop25 instructions.
20264 Generate 16-bit instructions.
20267 @opindex mno-16-bit
20268 Do not generate 16-bit instructions.
20270 @item -misr-vector-size=@var{num}
20271 @opindex misr-vector-size
20272 Specify the size of each interrupt vector, which must be 4 or 16.
20274 @item -mcache-block-size=@var{num}
20275 @opindex mcache-block-size
20276 Specify the size of each cache block,
20277 which must be a power of 2 between 4 and 512.
20279 @item -march=@var{arch}
20281 Specify the name of the target architecture.
20283 @item -mcmodel=@var{code-model}
20285 Set the code model to one of
20288 All the data and read-only data segments must be within 512KB addressing space.
20289 The text segment must be within 16MB addressing space.
20290 @item @samp{medium}
20291 The data segment must be within 512KB while the read-only data segment can be
20292 within 4GB addressing space. The text segment should be still within 16MB
20295 All the text and data segments can be within 4GB addressing space.
20299 @opindex mctor-dtor
20300 Enable constructor/destructor feature.
20304 Guide linker to relax instructions.
20308 @node Nios II Options
20309 @subsection Nios II Options
20310 @cindex Nios II options
20311 @cindex Altera Nios II options
20313 These are the options defined for the Altera Nios II processor.
20319 @cindex smaller data references
20320 Put global and static objects less than or equal to @var{num} bytes
20321 into the small data or BSS sections instead of the normal data or BSS
20322 sections. The default value of @var{num} is 8.
20324 @item -mgpopt=@var{option}
20329 Generate (do not generate) GP-relative accesses. The following
20330 @var{option} names are recognized:
20335 Do not generate GP-relative accesses.
20338 Generate GP-relative accesses for small data objects that are not
20339 external, weak, or uninitialized common symbols.
20340 Also use GP-relative addressing for objects that
20341 have been explicitly placed in a small data section via a @code{section}
20345 As for @samp{local}, but also generate GP-relative accesses for
20346 small data objects that are external, weak, or common. If you use this option,
20347 you must ensure that all parts of your program (including libraries) are
20348 compiled with the same @option{-G} setting.
20351 Generate GP-relative accesses for all data objects in the program. If you
20352 use this option, the entire data and BSS segments
20353 of your program must fit in 64K of memory and you must use an appropriate
20354 linker script to allocate them within the addressable range of the
20358 Generate GP-relative addresses for function pointers as well as data
20359 pointers. If you use this option, the entire text, data, and BSS segments
20360 of your program must fit in 64K of memory and you must use an appropriate
20361 linker script to allocate them within the addressable range of the
20366 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
20367 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
20369 The default is @option{-mgpopt} except when @option{-fpic} or
20370 @option{-fPIC} is specified to generate position-independent code.
20371 Note that the Nios II ABI does not permit GP-relative accesses from
20374 You may need to specify @option{-mno-gpopt} explicitly when building
20375 programs that include large amounts of small data, including large
20376 GOT data sections. In this case, the 16-bit offset for GP-relative
20377 addressing may not be large enough to allow access to the entire
20378 small data section.
20384 Generate little-endian (default) or big-endian (experimental) code,
20387 @item -march=@var{arch}
20389 This specifies the name of the target Nios II architecture. GCC uses this
20390 name to determine what kind of instructions it can emit when generating
20391 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
20393 The preprocessor macro @code{__nios2_arch__} is available to programs,
20394 with value 1 or 2, indicating the targeted ISA level.
20396 @item -mbypass-cache
20397 @itemx -mno-bypass-cache
20398 @opindex mno-bypass-cache
20399 @opindex mbypass-cache
20400 Force all load and store instructions to always bypass cache by
20401 using I/O variants of the instructions. The default is not to
20404 @item -mno-cache-volatile
20405 @itemx -mcache-volatile
20406 @opindex mcache-volatile
20407 @opindex mno-cache-volatile
20408 Volatile memory access bypass the cache using the I/O variants of
20409 the load and store instructions. The default is not to bypass the cache.
20411 @item -mno-fast-sw-div
20412 @itemx -mfast-sw-div
20413 @opindex mno-fast-sw-div
20414 @opindex mfast-sw-div
20415 Do not use table-based fast divide for small numbers. The default
20416 is to use the fast divide at @option{-O3} and above.
20420 @itemx -mno-hw-mulx
20424 @opindex mno-hw-mul
20426 @opindex mno-hw-mulx
20428 @opindex mno-hw-div
20430 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
20431 instructions by the compiler. The default is to emit @code{mul}
20432 and not emit @code{div} and @code{mulx}.
20438 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
20439 CDX (code density) instructions. Enabling these instructions also
20440 requires @option{-march=r2}. Since these instructions are optional
20441 extensions to the R2 architecture, the default is not to emit them.
20443 @item -mcustom-@var{insn}=@var{N}
20444 @itemx -mno-custom-@var{insn}
20445 @opindex mcustom-@var{insn}
20446 @opindex mno-custom-@var{insn}
20447 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
20448 custom instruction with encoding @var{N} when generating code that uses
20449 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
20450 instruction 253 for single-precision floating-point add operations instead
20451 of the default behavior of using a library call.
20453 The following values of @var{insn} are supported. Except as otherwise
20454 noted, floating-point operations are expected to be implemented with
20455 normal IEEE 754 semantics and correspond directly to the C operators or the
20456 equivalent GCC built-in functions (@pxref{Other Builtins}).
20458 Single-precision floating point:
20461 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
20462 Binary arithmetic operations.
20468 Unary absolute value.
20470 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
20471 Comparison operations.
20473 @item @samp{fmins}, @samp{fmaxs}
20474 Floating-point minimum and maximum. These instructions are only
20475 generated if @option{-ffinite-math-only} is specified.
20477 @item @samp{fsqrts}
20478 Unary square root operation.
20480 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
20481 Floating-point trigonometric and exponential functions. These instructions
20482 are only generated if @option{-funsafe-math-optimizations} is also specified.
20486 Double-precision floating point:
20489 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
20490 Binary arithmetic operations.
20496 Unary absolute value.
20498 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
20499 Comparison operations.
20501 @item @samp{fmind}, @samp{fmaxd}
20502 Double-precision minimum and maximum. These instructions are only
20503 generated if @option{-ffinite-math-only} is specified.
20505 @item @samp{fsqrtd}
20506 Unary square root operation.
20508 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
20509 Double-precision trigonometric and exponential functions. These instructions
20510 are only generated if @option{-funsafe-math-optimizations} is also specified.
20516 @item @samp{fextsd}
20517 Conversion from single precision to double precision.
20519 @item @samp{ftruncds}
20520 Conversion from double precision to single precision.
20522 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
20523 Conversion from floating point to signed or unsigned integer types, with
20524 truncation towards zero.
20527 Conversion from single-precision floating point to signed integer,
20528 rounding to the nearest integer and ties away from zero.
20529 This corresponds to the @code{__builtin_lroundf} function when
20530 @option{-fno-math-errno} is used.
20532 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
20533 Conversion from signed or unsigned integer types to floating-point types.
20537 In addition, all of the following transfer instructions for internal
20538 registers X and Y must be provided to use any of the double-precision
20539 floating-point instructions. Custom instructions taking two
20540 double-precision source operands expect the first operand in the
20541 64-bit register X. The other operand (or only operand of a unary
20542 operation) is given to the custom arithmetic instruction with the
20543 least significant half in source register @var{src1} and the most
20544 significant half in @var{src2}. A custom instruction that returns a
20545 double-precision result returns the most significant 32 bits in the
20546 destination register and the other half in 32-bit register Y.
20547 GCC automatically generates the necessary code sequences to write
20548 register X and/or read register Y when double-precision floating-point
20549 instructions are used.
20554 Write @var{src1} into the least significant half of X and @var{src2} into
20555 the most significant half of X.
20558 Write @var{src1} into Y.
20560 @item @samp{frdxhi}, @samp{frdxlo}
20561 Read the most or least (respectively) significant half of X and store it in
20565 Read the value of Y and store it into @var{dest}.
20568 Note that you can gain more local control over generation of Nios II custom
20569 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
20570 and @code{target("no-custom-@var{insn}")} function attributes
20571 (@pxref{Function Attributes})
20572 or pragmas (@pxref{Function Specific Option Pragmas}).
20574 @item -mcustom-fpu-cfg=@var{name}
20575 @opindex mcustom-fpu-cfg
20577 This option enables a predefined, named set of custom instruction encodings
20578 (see @option{-mcustom-@var{insn}} above).
20579 Currently, the following sets are defined:
20581 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
20582 @gccoptlist{-mcustom-fmuls=252 @gol
20583 -mcustom-fadds=253 @gol
20584 -mcustom-fsubs=254 @gol
20585 -fsingle-precision-constant}
20587 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
20588 @gccoptlist{-mcustom-fmuls=252 @gol
20589 -mcustom-fadds=253 @gol
20590 -mcustom-fsubs=254 @gol
20591 -mcustom-fdivs=255 @gol
20592 -fsingle-precision-constant}
20594 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
20595 @gccoptlist{-mcustom-floatus=243 @gol
20596 -mcustom-fixsi=244 @gol
20597 -mcustom-floatis=245 @gol
20598 -mcustom-fcmpgts=246 @gol
20599 -mcustom-fcmples=249 @gol
20600 -mcustom-fcmpeqs=250 @gol
20601 -mcustom-fcmpnes=251 @gol
20602 -mcustom-fmuls=252 @gol
20603 -mcustom-fadds=253 @gol
20604 -mcustom-fsubs=254 @gol
20605 -mcustom-fdivs=255 @gol
20606 -fsingle-precision-constant}
20608 Custom instruction assignments given by individual
20609 @option{-mcustom-@var{insn}=} options override those given by
20610 @option{-mcustom-fpu-cfg=}, regardless of the
20611 order of the options on the command line.
20613 Note that you can gain more local control over selection of a FPU
20614 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
20615 function attribute (@pxref{Function Attributes})
20616 or pragma (@pxref{Function Specific Option Pragmas}).
20620 These additional @samp{-m} options are available for the Altera Nios II
20621 ELF (bare-metal) target:
20627 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
20628 startup and termination code, and is typically used in conjunction with
20629 @option{-msys-crt0=} to specify the location of the alternate startup code
20630 provided by the HAL BSP.
20634 Link with a limited version of the C library, @option{-lsmallc}, rather than
20637 @item -msys-crt0=@var{startfile}
20639 @var{startfile} is the file name of the startfile (crt0) to use
20640 when linking. This option is only useful in conjunction with @option{-mhal}.
20642 @item -msys-lib=@var{systemlib}
20644 @var{systemlib} is the library name of the library that provides
20645 low-level system calls required by the C library,
20646 e.g. @code{read} and @code{write}.
20647 This option is typically used to link with a library provided by a HAL BSP.
20651 @node Nvidia PTX Options
20652 @subsection Nvidia PTX Options
20653 @cindex Nvidia PTX options
20654 @cindex nvptx options
20656 These options are defined for Nvidia PTX:
20664 Generate code for 32-bit or 64-bit ABI.
20667 @opindex mmainkernel
20668 Link in code for a __main kernel. This is for stand-alone instead of
20669 offloading execution.
20673 Apply partitioned execution optimizations. This is the default when any
20674 level of optimization is selected.
20677 @opindex msoft-stack
20678 Generate code that does not use @code{.local} memory
20679 directly for stack storage. Instead, a per-warp stack pointer is
20680 maintained explicitly. This enables variable-length stack allocation (with
20681 variable-length arrays or @code{alloca}), and when global memory is used for
20682 underlying storage, makes it possible to access automatic variables from other
20683 threads, or with atomic instructions. This code generation variant is used
20684 for OpenMP offloading, but the option is exposed on its own for the purpose
20685 of testing the compiler; to generate code suitable for linking into programs
20686 using OpenMP offloading, use option @option{-mgomp}.
20688 @item -muniform-simt
20689 @opindex muniform-simt
20690 Switch to code generation variant that allows to execute all threads in each
20691 warp, while maintaining memory state and side effects as if only one thread
20692 in each warp was active outside of OpenMP SIMD regions. All atomic operations
20693 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
20694 current lane index equals the master lane index), and the register being
20695 assigned is copied via a shuffle instruction from the master lane. Outside of
20696 SIMD regions lane 0 is the master; inside, each thread sees itself as the
20697 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
20698 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
20699 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
20700 with current lane index to compute the master lane index.
20704 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
20705 @option{-muniform-simt} options, and selects corresponding multilib variant.
20709 @node PDP-11 Options
20710 @subsection PDP-11 Options
20711 @cindex PDP-11 Options
20713 These options are defined for the PDP-11:
20718 Use hardware FPP floating point. This is the default. (FIS floating
20719 point on the PDP-11/40 is not supported.)
20722 @opindex msoft-float
20723 Do not use hardware floating point.
20727 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
20731 Return floating-point results in memory. This is the default.
20735 Generate code for a PDP-11/40.
20739 Generate code for a PDP-11/45. This is the default.
20743 Generate code for a PDP-11/10.
20745 @item -mbcopy-builtin
20746 @opindex mbcopy-builtin
20747 Use inline @code{movmemhi} patterns for copying memory. This is the
20752 Do not use inline @code{movmemhi} patterns for copying memory.
20758 Use 16-bit @code{int}. This is the default.
20764 Use 32-bit @code{int}.
20767 @itemx -mno-float32
20769 @opindex mno-float32
20770 Use 64-bit @code{float}. This is the default.
20773 @itemx -mno-float64
20775 @opindex mno-float64
20776 Use 32-bit @code{float}.
20780 Use @code{abshi2} pattern. This is the default.
20784 Do not use @code{abshi2} pattern.
20786 @item -mbranch-expensive
20787 @opindex mbranch-expensive
20788 Pretend that branches are expensive. This is for experimenting with
20789 code generation only.
20791 @item -mbranch-cheap
20792 @opindex mbranch-cheap
20793 Do not pretend that branches are expensive. This is the default.
20797 Use Unix assembler syntax. This is the default when configured for
20798 @samp{pdp11-*-bsd}.
20802 Use DEC assembler syntax. This is the default when configured for any
20803 PDP-11 target other than @samp{pdp11-*-bsd}.
20806 @node picoChip Options
20807 @subsection picoChip Options
20808 @cindex picoChip options
20810 These @samp{-m} options are defined for picoChip implementations:
20814 @item -mae=@var{ae_type}
20816 Set the instruction set, register set, and instruction scheduling
20817 parameters for array element type @var{ae_type}. Supported values
20818 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
20820 @option{-mae=ANY} selects a completely generic AE type. Code
20821 generated with this option runs on any of the other AE types. The
20822 code is not as efficient as it would be if compiled for a specific
20823 AE type, and some types of operation (e.g., multiplication) do not
20824 work properly on all types of AE.
20826 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
20827 for compiled code, and is the default.
20829 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
20830 option may suffer from poor performance of byte (char) manipulation,
20831 since the DSP AE does not provide hardware support for byte load/stores.
20833 @item -msymbol-as-address
20834 Enable the compiler to directly use a symbol name as an address in a
20835 load/store instruction, without first loading it into a
20836 register. Typically, the use of this option generates larger
20837 programs, which run faster than when the option isn't used. However, the
20838 results vary from program to program, so it is left as a user option,
20839 rather than being permanently enabled.
20841 @item -mno-inefficient-warnings
20842 Disables warnings about the generation of inefficient code. These
20843 warnings can be generated, for example, when compiling code that
20844 performs byte-level memory operations on the MAC AE type. The MAC AE has
20845 no hardware support for byte-level memory operations, so all byte
20846 load/stores must be synthesized from word load/store operations. This is
20847 inefficient and a warning is generated to indicate
20848 that you should rewrite the code to avoid byte operations, or to target
20849 an AE type that has the necessary hardware support. This option disables
20854 @node PowerPC Options
20855 @subsection PowerPC Options
20856 @cindex PowerPC options
20858 These are listed under @xref{RS/6000 and PowerPC Options}.
20860 @node RISC-V Options
20861 @subsection RISC-V Options
20862 @cindex RISC-V Options
20864 These command-line options are defined for RISC-V targets:
20867 @item -mbranch-cost=@var{n}
20868 @opindex mbranch-cost
20869 Set the cost of branches to roughly @var{n} instructions.
20874 Don't optimize block moves.
20879 When generating PIC code, allow the use of PLTs. Ignored for non-PIC.
20881 @item -mabi=@var{ABI-string}
20883 Specify integer and floating-point calling convention. This defaults to the
20884 natural calling convention: e.g.@ LP64 for RV64I, ILP32 for RV32I, LP64D for
20890 Use hardware floating-point divide and square root instructions. This requires
20891 the F or D extensions for floating-point registers.
20896 Use hardware instructions for integer division. This requires the M extension.
20898 @item -march=@var{ISA-string}
20900 Generate code for given RISC-V ISA (e.g.@ @samp{rv64im}). ISA strings must be
20901 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, and @samp{rv32imaf}.
20903 @item -mtune=@var{processor-string}
20905 Optimize the output for the given processor, specified by microarchitecture
20908 @item -msmall-data-limit=@var{n}
20909 @opindex msmall-data-limit
20910 Put global and static data smaller than @var{n} bytes into a special section
20913 @item -msave-restore
20914 @itemx -mno-save-restore
20915 @opindex msave-restore
20916 Use smaller but slower prologue and epilogue code.
20918 @item -mstrict-align
20919 @itemx -mno-strict-align
20920 @opindex mstrict-align
20921 Do not generate unaligned memory accesses.
20923 @item -mcmodel=@var{code-model}
20925 Specify the code model.
20930 @subsection RL78 Options
20931 @cindex RL78 Options
20937 Links in additional target libraries to support operation within a
20946 Specifies the type of hardware multiplication and division support to
20947 be used. The simplest is @code{none}, which uses software for both
20948 multiplication and division. This is the default. The @code{g13}
20949 value is for the hardware multiply/divide peripheral found on the
20950 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
20951 the multiplication and division instructions supported by the RL78/G14
20952 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
20953 the value @code{mg10} is an alias for @code{none}.
20955 In addition a C preprocessor macro is defined, based upon the setting
20956 of this option. Possible values are: @code{__RL78_MUL_NONE__},
20957 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
20964 Specifies the RL78 core to target. The default is the G14 core, also
20965 known as an S3 core or just RL78. The G13 or S2 core does not have
20966 multiply or divide instructions, instead it uses a hardware peripheral
20967 for these operations. The G10 or S1 core does not have register
20968 banks, so it uses a different calling convention.
20970 If this option is set it also selects the type of hardware multiply
20971 support to use, unless this is overridden by an explicit
20972 @option{-mmul=none} option on the command line. Thus specifying
20973 @option{-mcpu=g13} enables the use of the G13 hardware multiply
20974 peripheral and specifying @option{-mcpu=g10} disables the use of
20975 hardware multiplications altogether.
20977 Note, although the RL78/G14 core is the default target, specifying
20978 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
20979 change the behavior of the toolchain since it also enables G14
20980 hardware multiply support. If these options are not specified on the
20981 command line then software multiplication routines will be used even
20982 though the code targets the RL78 core. This is for backwards
20983 compatibility with older toolchains which did not have hardware
20984 multiply and divide support.
20986 In addition a C preprocessor macro is defined, based upon the setting
20987 of this option. Possible values are: @code{__RL78_G10__},
20988 @code{__RL78_G13__} or @code{__RL78_G14__}.
20998 These are aliases for the corresponding @option{-mcpu=} option. They
20999 are provided for backwards compatibility.
21003 Allow the compiler to use all of the available registers. By default
21004 registers @code{r24..r31} are reserved for use in interrupt handlers.
21005 With this option enabled these registers can be used in ordinary
21008 @item -m64bit-doubles
21009 @itemx -m32bit-doubles
21010 @opindex m64bit-doubles
21011 @opindex m32bit-doubles
21012 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21013 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21014 @option{-m32bit-doubles}.
21016 @item -msave-mduc-in-interrupts
21017 @item -mno-save-mduc-in-interrupts
21018 @opindex msave-mduc-in-interrupts
21019 @opindex mno-save-mduc-in-interrupts
21020 Specifies that interrupt handler functions should preserve the
21021 MDUC registers. This is only necessary if normal code might use
21022 the MDUC registers, for example because it performs multiplication
21023 and division operations. The default is to ignore the MDUC registers
21024 as this makes the interrupt handlers faster. The target option -mg13
21025 needs to be passed for this to work as this feature is only available
21026 on the G13 target (S2 core). The MDUC registers will only be saved
21027 if the interrupt handler performs a multiplication or division
21028 operation or it calls another function.
21032 @node RS/6000 and PowerPC Options
21033 @subsection IBM RS/6000 and PowerPC Options
21034 @cindex RS/6000 and PowerPC Options
21035 @cindex IBM RS/6000 and PowerPC Options
21037 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
21039 @item -mpowerpc-gpopt
21040 @itemx -mno-powerpc-gpopt
21041 @itemx -mpowerpc-gfxopt
21042 @itemx -mno-powerpc-gfxopt
21045 @itemx -mno-powerpc64
21049 @itemx -mno-popcntb
21051 @itemx -mno-popcntd
21060 @itemx -mno-hard-dfp
21061 @opindex mpowerpc-gpopt
21062 @opindex mno-powerpc-gpopt
21063 @opindex mpowerpc-gfxopt
21064 @opindex mno-powerpc-gfxopt
21065 @opindex mpowerpc64
21066 @opindex mno-powerpc64
21070 @opindex mno-popcntb
21072 @opindex mno-popcntd
21078 @opindex mno-mfpgpr
21080 @opindex mno-hard-dfp
21081 You use these options to specify which instructions are available on the
21082 processor you are using. The default value of these options is
21083 determined when configuring GCC@. Specifying the
21084 @option{-mcpu=@var{cpu_type}} overrides the specification of these
21085 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
21086 rather than the options listed above.
21088 Specifying @option{-mpowerpc-gpopt} allows
21089 GCC to use the optional PowerPC architecture instructions in the
21090 General Purpose group, including floating-point square root. Specifying
21091 @option{-mpowerpc-gfxopt} allows GCC to
21092 use the optional PowerPC architecture instructions in the Graphics
21093 group, including floating-point select.
21095 The @option{-mmfcrf} option allows GCC to generate the move from
21096 condition register field instruction implemented on the POWER4
21097 processor and other processors that support the PowerPC V2.01
21099 The @option{-mpopcntb} option allows GCC to generate the popcount and
21100 double-precision FP reciprocal estimate instruction implemented on the
21101 POWER5 processor and other processors that support the PowerPC V2.02
21103 The @option{-mpopcntd} option allows GCC to generate the popcount
21104 instruction implemented on the POWER7 processor and other processors
21105 that support the PowerPC V2.06 architecture.
21106 The @option{-mfprnd} option allows GCC to generate the FP round to
21107 integer instructions implemented on the POWER5+ processor and other
21108 processors that support the PowerPC V2.03 architecture.
21109 The @option{-mcmpb} option allows GCC to generate the compare bytes
21110 instruction implemented on the POWER6 processor and other processors
21111 that support the PowerPC V2.05 architecture.
21112 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
21113 general-purpose register instructions implemented on the POWER6X
21114 processor and other processors that support the extended PowerPC V2.05
21116 The @option{-mhard-dfp} option allows GCC to generate the decimal
21117 floating-point instructions implemented on some POWER processors.
21119 The @option{-mpowerpc64} option allows GCC to generate the additional
21120 64-bit instructions that are found in the full PowerPC64 architecture
21121 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
21122 @option{-mno-powerpc64}.
21124 @item -mcpu=@var{cpu_type}
21126 Set architecture type, register usage, and
21127 instruction scheduling parameters for machine type @var{cpu_type}.
21128 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
21129 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
21130 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
21131 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
21132 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
21133 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
21134 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
21135 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
21136 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
21137 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
21138 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
21141 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
21142 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
21143 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
21144 architecture machine types, with an appropriate, generic processor
21145 model assumed for scheduling purposes.
21147 The other options specify a specific processor. Code generated under
21148 those options runs best on that processor, and may not run at all on
21151 The @option{-mcpu} options automatically enable or disable the
21154 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
21155 -mpopcntb -mpopcntd -mpowerpc64 @gol
21156 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
21157 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
21158 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
21159 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
21161 The particular options set for any particular CPU varies between
21162 compiler versions, depending on what setting seems to produce optimal
21163 code for that CPU; it doesn't necessarily reflect the actual hardware's
21164 capabilities. If you wish to set an individual option to a particular
21165 value, you may specify it after the @option{-mcpu} option, like
21166 @option{-mcpu=970 -mno-altivec}.
21168 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
21169 not enabled or disabled by the @option{-mcpu} option at present because
21170 AIX does not have full support for these options. You may still
21171 enable or disable them individually if you're sure it'll work in your
21174 @item -mtune=@var{cpu_type}
21176 Set the instruction scheduling parameters for machine type
21177 @var{cpu_type}, but do not set the architecture type or register usage,
21178 as @option{-mcpu=@var{cpu_type}} does. The same
21179 values for @var{cpu_type} are used for @option{-mtune} as for
21180 @option{-mcpu}. If both are specified, the code generated uses the
21181 architecture and registers set by @option{-mcpu}, but the
21182 scheduling parameters set by @option{-mtune}.
21184 @item -mcmodel=small
21185 @opindex mcmodel=small
21186 Generate PowerPC64 code for the small model: The TOC is limited to
21189 @item -mcmodel=medium
21190 @opindex mcmodel=medium
21191 Generate PowerPC64 code for the medium model: The TOC and other static
21192 data may be up to a total of 4G in size. This is the default for 64-bit
21195 @item -mcmodel=large
21196 @opindex mcmodel=large
21197 Generate PowerPC64 code for the large model: The TOC may be up to 4G
21198 in size. Other data and code is only limited by the 64-bit address
21202 @itemx -mno-altivec
21204 @opindex mno-altivec
21205 Generate code that uses (does not use) AltiVec instructions, and also
21206 enable the use of built-in functions that allow more direct access to
21207 the AltiVec instruction set. You may also need to set
21208 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
21211 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
21212 @option{-maltivec=be}, the element order for AltiVec intrinsics such
21213 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
21214 match array element order corresponding to the endianness of the
21215 target. That is, element zero identifies the leftmost element in a
21216 vector register when targeting a big-endian platform, and identifies
21217 the rightmost element in a vector register when targeting a
21218 little-endian platform.
21221 @opindex maltivec=be
21222 Generate AltiVec instructions using big-endian element order,
21223 regardless of whether the target is big- or little-endian. This is
21224 the default when targeting a big-endian platform.
21226 The element order is used to interpret element numbers in AltiVec
21227 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
21228 @code{vec_insert}. By default, these match array element order
21229 corresponding to the endianness for the target.
21232 @opindex maltivec=le
21233 Generate AltiVec instructions using little-endian element order,
21234 regardless of whether the target is big- or little-endian. This is
21235 the default when targeting a little-endian platform. This option is
21236 currently ignored when targeting a big-endian platform.
21238 The element order is used to interpret element numbers in AltiVec
21239 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
21240 @code{vec_insert}. By default, these match array element order
21241 corresponding to the endianness for the target.
21246 @opindex mno-vrsave
21247 Generate VRSAVE instructions when generating AltiVec code.
21249 @item -mgen-cell-microcode
21250 @opindex mgen-cell-microcode
21251 Generate Cell microcode instructions.
21253 @item -mwarn-cell-microcode
21254 @opindex mwarn-cell-microcode
21255 Warn when a Cell microcode instruction is emitted. An example
21256 of a Cell microcode instruction is a variable shift.
21259 @opindex msecure-plt
21260 Generate code that allows @command{ld} and @command{ld.so}
21261 to build executables and shared
21262 libraries with non-executable @code{.plt} and @code{.got} sections.
21264 32-bit SYSV ABI option.
21268 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
21270 requires @code{.plt} and @code{.got}
21271 sections that are both writable and executable.
21272 This is a PowerPC 32-bit SYSV ABI option.
21278 This switch enables or disables the generation of ISEL instructions.
21280 @item -misel=@var{yes/no}
21281 This switch has been deprecated. Use @option{-misel} and
21282 @option{-mno-isel} instead.
21286 Enable Local Register Allocation. By default the port uses LRA.
21287 (i.e. @option{-mno-lra}).
21293 This switch enables or disables the generation of SPE simd
21299 @opindex mno-paired
21300 This switch enables or disables the generation of PAIRED simd
21303 @item -mspe=@var{yes/no}
21304 This option has been deprecated. Use @option{-mspe} and
21305 @option{-mno-spe} instead.
21311 Generate code that uses (does not use) vector/scalar (VSX)
21312 instructions, and also enable the use of built-in functions that allow
21313 more direct access to the VSX instruction set.
21318 @opindex mno-crypto
21319 Enable the use (disable) of the built-in functions that allow direct
21320 access to the cryptographic instructions that were added in version
21321 2.07 of the PowerPC ISA.
21323 @item -mdirect-move
21324 @itemx -mno-direct-move
21325 @opindex mdirect-move
21326 @opindex mno-direct-move
21327 Generate code that uses (does not use) the instructions to move data
21328 between the general purpose registers and the vector/scalar (VSX)
21329 registers that were added in version 2.07 of the PowerPC ISA.
21335 Enable (disable) the use of the built-in functions that allow direct
21336 access to the Hardware Transactional Memory (HTM) instructions that
21337 were added in version 2.07 of the PowerPC ISA.
21339 @item -mpower8-fusion
21340 @itemx -mno-power8-fusion
21341 @opindex mpower8-fusion
21342 @opindex mno-power8-fusion
21343 Generate code that keeps (does not keeps) some integer operations
21344 adjacent so that the instructions can be fused together on power8 and
21347 @item -mpower8-vector
21348 @itemx -mno-power8-vector
21349 @opindex mpower8-vector
21350 @opindex mno-power8-vector
21351 Generate code that uses (does not use) the vector and scalar
21352 instructions that were added in version 2.07 of the PowerPC ISA. Also
21353 enable the use of built-in functions that allow more direct access to
21354 the vector instructions.
21356 @item -mquad-memory
21357 @itemx -mno-quad-memory
21358 @opindex mquad-memory
21359 @opindex mno-quad-memory
21360 Generate code that uses (does not use) the non-atomic quad word memory
21361 instructions. The @option{-mquad-memory} option requires use of
21364 @item -mquad-memory-atomic
21365 @itemx -mno-quad-memory-atomic
21366 @opindex mquad-memory-atomic
21367 @opindex mno-quad-memory-atomic
21368 Generate code that uses (does not use) the atomic quad word memory
21369 instructions. The @option{-mquad-memory-atomic} option requires use of
21372 @item -mupper-regs-di
21373 @itemx -mno-upper-regs-di
21374 @opindex mupper-regs-di
21375 @opindex mno-upper-regs-di
21376 Generate code that uses (does not use) the scalar instructions that
21377 target all 64 registers in the vector/scalar floating point register
21378 set that were added in version 2.06 of the PowerPC ISA when processing
21379 integers. @option{-mupper-regs-di} is turned on by default if you use
21380 any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21381 @option{-mcpu=power9}, or @option{-mvsx} options.
21383 @item -mupper-regs-df
21384 @itemx -mno-upper-regs-df
21385 @opindex mupper-regs-df
21386 @opindex mno-upper-regs-df
21387 Generate code that uses (does not use) the scalar double precision
21388 instructions that target all 64 registers in the vector/scalar
21389 floating point register set that were added in version 2.06 of the
21390 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
21391 use any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21392 @option{-mcpu=power9}, or @option{-mvsx} options.
21394 @item -mupper-regs-sf
21395 @itemx -mno-upper-regs-sf
21396 @opindex mupper-regs-sf
21397 @opindex mno-upper-regs-sf
21398 Generate code that uses (does not use) the scalar single precision
21399 instructions that target all 64 registers in the vector/scalar
21400 floating point register set that were added in version 2.07 of the
21401 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
21402 use either of the @option{-mcpu=power8}, @option{-mpower8-vector}, or
21403 @option{-mcpu=power9} options.
21406 @itemx -mno-upper-regs
21407 @opindex mupper-regs
21408 @opindex mno-upper-regs
21409 Generate code that uses (does not use) the scalar
21410 instructions that target all 64 registers in the vector/scalar
21411 floating point register set, depending on the model of the machine.
21413 If the @option{-mno-upper-regs} option is used, it turns off both
21414 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
21417 @itemx -mno-float128
21419 @opindex mno-float128
21420 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
21421 and use either software emulation for IEEE 128-bit floating point or
21422 hardware instructions.
21424 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
21425 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
21426 option. The @option{-mfloat128} option only works on PowerPC 64-bit
21429 If you use the ISA 3.0 instruction set (@option{-mcpu=power9}), the
21430 @option{-mfloat128} option will also enable the generation of ISA 3.0
21431 IEEE 128-bit floating point instructions. Otherwise, IEEE 128-bit
21432 floating point will be done with software emulation.
21434 @item -mfloat128-hardware
21435 @itemx -mno-float128-hardware
21436 @opindex mfloat128-hardware
21437 @opindex mno-float128-hardware
21438 Enable/disable using ISA 3.0 hardware instructions to support the
21439 @var{__float128} data type.
21441 If you use @option{-mfloat128-hardware}, it will enable the option
21442 @option{-mfloat128} as well.
21444 If you select ISA 3.0 instructions with @option{-mcpu=power9}, but do
21445 not use either @option{-mfloat128} or @option{-mfloat128-hardware},
21446 the IEEE 128-bit floating point support will not be enabled.
21448 @item -mfloat-gprs=@var{yes/single/double/no}
21449 @itemx -mfloat-gprs
21450 @opindex mfloat-gprs
21451 This switch enables or disables the generation of floating-point
21452 operations on the general-purpose registers for architectures that
21455 The argument @samp{yes} or @samp{single} enables the use of
21456 single-precision floating-point operations.
21458 The argument @samp{double} enables the use of single and
21459 double-precision floating-point operations.
21461 The argument @samp{no} disables floating-point operations on the
21462 general-purpose registers.
21464 This option is currently only available on the MPC854x.
21470 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
21471 targets (including GNU/Linux). The 32-bit environment sets int, long
21472 and pointer to 32 bits and generates code that runs on any PowerPC
21473 variant. The 64-bit environment sets int to 32 bits and long and
21474 pointer to 64 bits, and generates code for PowerPC64, as for
21475 @option{-mpowerpc64}.
21478 @itemx -mno-fp-in-toc
21479 @itemx -mno-sum-in-toc
21480 @itemx -mminimal-toc
21482 @opindex mno-fp-in-toc
21483 @opindex mno-sum-in-toc
21484 @opindex mminimal-toc
21485 Modify generation of the TOC (Table Of Contents), which is created for
21486 every executable file. The @option{-mfull-toc} option is selected by
21487 default. In that case, GCC allocates at least one TOC entry for
21488 each unique non-automatic variable reference in your program. GCC
21489 also places floating-point constants in the TOC@. However, only
21490 16,384 entries are available in the TOC@.
21492 If you receive a linker error message that saying you have overflowed
21493 the available TOC space, you can reduce the amount of TOC space used
21494 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
21495 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
21496 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
21497 generate code to calculate the sum of an address and a constant at
21498 run time instead of putting that sum into the TOC@. You may specify one
21499 or both of these options. Each causes GCC to produce very slightly
21500 slower and larger code at the expense of conserving TOC space.
21502 If you still run out of space in the TOC even when you specify both of
21503 these options, specify @option{-mminimal-toc} instead. This option causes
21504 GCC to make only one TOC entry for every file. When you specify this
21505 option, GCC produces code that is slower and larger but which
21506 uses extremely little TOC space. You may wish to use this option
21507 only on files that contain less frequently-executed code.
21513 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
21514 @code{long} type, and the infrastructure needed to support them.
21515 Specifying @option{-maix64} implies @option{-mpowerpc64},
21516 while @option{-maix32} disables the 64-bit ABI and
21517 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
21520 @itemx -mno-xl-compat
21521 @opindex mxl-compat
21522 @opindex mno-xl-compat
21523 Produce code that conforms more closely to IBM XL compiler semantics
21524 when using AIX-compatible ABI@. Pass floating-point arguments to
21525 prototyped functions beyond the register save area (RSA) on the stack
21526 in addition to argument FPRs. Do not assume that most significant
21527 double in 128-bit long double value is properly rounded when comparing
21528 values and converting to double. Use XL symbol names for long double
21531 The AIX calling convention was extended but not initially documented to
21532 handle an obscure K&R C case of calling a function that takes the
21533 address of its arguments with fewer arguments than declared. IBM XL
21534 compilers access floating-point arguments that do not fit in the
21535 RSA from the stack when a subroutine is compiled without
21536 optimization. Because always storing floating-point arguments on the
21537 stack is inefficient and rarely needed, this option is not enabled by
21538 default and only is necessary when calling subroutines compiled by IBM
21539 XL compilers without optimization.
21543 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
21544 application written to use message passing with special startup code to
21545 enable the application to run. The system must have PE installed in the
21546 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
21547 must be overridden with the @option{-specs=} option to specify the
21548 appropriate directory location. The Parallel Environment does not
21549 support threads, so the @option{-mpe} option and the @option{-pthread}
21550 option are incompatible.
21552 @item -malign-natural
21553 @itemx -malign-power
21554 @opindex malign-natural
21555 @opindex malign-power
21556 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
21557 @option{-malign-natural} overrides the ABI-defined alignment of larger
21558 types, such as floating-point doubles, on their natural size-based boundary.
21559 The option @option{-malign-power} instructs GCC to follow the ABI-specified
21560 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
21562 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
21566 @itemx -mhard-float
21567 @opindex msoft-float
21568 @opindex mhard-float
21569 Generate code that does not use (uses) the floating-point register set.
21570 Software floating-point emulation is provided if you use the
21571 @option{-msoft-float} option, and pass the option to GCC when linking.
21573 @item -msingle-float
21574 @itemx -mdouble-float
21575 @opindex msingle-float
21576 @opindex mdouble-float
21577 Generate code for single- or double-precision floating-point operations.
21578 @option{-mdouble-float} implies @option{-msingle-float}.
21581 @opindex msimple-fpu
21582 Do not generate @code{sqrt} and @code{div} instructions for hardware
21583 floating-point unit.
21585 @item -mfpu=@var{name}
21587 Specify type of floating-point unit. Valid values for @var{name} are
21588 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
21589 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
21590 @samp{sp_full} (equivalent to @option{-msingle-float}),
21591 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
21594 @opindex mxilinx-fpu
21595 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
21598 @itemx -mno-multiple
21600 @opindex mno-multiple
21601 Generate code that uses (does not use) the load multiple word
21602 instructions and the store multiple word instructions. These
21603 instructions are generated by default on POWER systems, and not
21604 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
21605 PowerPC systems, since those instructions do not work when the
21606 processor is in little-endian mode. The exceptions are PPC740 and
21607 PPC750 which permit these instructions in little-endian mode.
21612 @opindex mno-string
21613 Generate code that uses (does not use) the load string instructions
21614 and the store string word instructions to save multiple registers and
21615 do small block moves. These instructions are generated by default on
21616 POWER systems, and not generated on PowerPC systems. Do not use
21617 @option{-mstring} on little-endian PowerPC systems, since those
21618 instructions do not work when the processor is in little-endian mode.
21619 The exceptions are PPC740 and PPC750 which permit these instructions
21620 in little-endian mode.
21625 @opindex mno-update
21626 Generate code that uses (does not use) the load or store instructions
21627 that update the base register to the address of the calculated memory
21628 location. These instructions are generated by default. If you use
21629 @option{-mno-update}, there is a small window between the time that the
21630 stack pointer is updated and the address of the previous frame is
21631 stored, which means code that walks the stack frame across interrupts or
21632 signals may get corrupted data.
21634 @item -mavoid-indexed-addresses
21635 @itemx -mno-avoid-indexed-addresses
21636 @opindex mavoid-indexed-addresses
21637 @opindex mno-avoid-indexed-addresses
21638 Generate code that tries to avoid (not avoid) the use of indexed load
21639 or store instructions. These instructions can incur a performance
21640 penalty on Power6 processors in certain situations, such as when
21641 stepping through large arrays that cross a 16M boundary. This option
21642 is enabled by default when targeting Power6 and disabled otherwise.
21645 @itemx -mno-fused-madd
21646 @opindex mfused-madd
21647 @opindex mno-fused-madd
21648 Generate code that uses (does not use) the floating-point multiply and
21649 accumulate instructions. These instructions are generated by default
21650 if hardware floating point is used. The machine-dependent
21651 @option{-mfused-madd} option is now mapped to the machine-independent
21652 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21653 mapped to @option{-ffp-contract=off}.
21659 Generate code that uses (does not use) the half-word multiply and
21660 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
21661 These instructions are generated by default when targeting those
21668 Generate code that uses (does not use) the string-search @samp{dlmzb}
21669 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
21670 generated by default when targeting those processors.
21672 @item -mno-bit-align
21674 @opindex mno-bit-align
21675 @opindex mbit-align
21676 On System V.4 and embedded PowerPC systems do not (do) force structures
21677 and unions that contain bit-fields to be aligned to the base type of the
21680 For example, by default a structure containing nothing but 8
21681 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
21682 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
21683 the structure is aligned to a 1-byte boundary and is 1 byte in
21686 @item -mno-strict-align
21687 @itemx -mstrict-align
21688 @opindex mno-strict-align
21689 @opindex mstrict-align
21690 On System V.4 and embedded PowerPC systems do not (do) assume that
21691 unaligned memory references are handled by the system.
21693 @item -mrelocatable
21694 @itemx -mno-relocatable
21695 @opindex mrelocatable
21696 @opindex mno-relocatable
21697 Generate code that allows (does not allow) a static executable to be
21698 relocated to a different address at run time. A simple embedded
21699 PowerPC system loader should relocate the entire contents of
21700 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
21701 a table of 32-bit addresses generated by this option. For this to
21702 work, all objects linked together must be compiled with
21703 @option{-mrelocatable} or @option{-mrelocatable-lib}.
21704 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
21706 @item -mrelocatable-lib
21707 @itemx -mno-relocatable-lib
21708 @opindex mrelocatable-lib
21709 @opindex mno-relocatable-lib
21710 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
21711 @code{.fixup} section to allow static executables to be relocated at
21712 run time, but @option{-mrelocatable-lib} does not use the smaller stack
21713 alignment of @option{-mrelocatable}. Objects compiled with
21714 @option{-mrelocatable-lib} may be linked with objects compiled with
21715 any combination of the @option{-mrelocatable} options.
21721 On System V.4 and embedded PowerPC systems do not (do) assume that
21722 register 2 contains a pointer to a global area pointing to the addresses
21723 used in the program.
21726 @itemx -mlittle-endian
21728 @opindex mlittle-endian
21729 On System V.4 and embedded PowerPC systems compile code for the
21730 processor in little-endian mode. The @option{-mlittle-endian} option is
21731 the same as @option{-mlittle}.
21734 @itemx -mbig-endian
21736 @opindex mbig-endian
21737 On System V.4 and embedded PowerPC systems compile code for the
21738 processor in big-endian mode. The @option{-mbig-endian} option is
21739 the same as @option{-mbig}.
21741 @item -mdynamic-no-pic
21742 @opindex mdynamic-no-pic
21743 On Darwin and Mac OS X systems, compile code so that it is not
21744 relocatable, but that its external references are relocatable. The
21745 resulting code is suitable for applications, but not shared
21748 @item -msingle-pic-base
21749 @opindex msingle-pic-base
21750 Treat the register used for PIC addressing as read-only, rather than
21751 loading it in the prologue for each function. The runtime system is
21752 responsible for initializing this register with an appropriate value
21753 before execution begins.
21755 @item -mprioritize-restricted-insns=@var{priority}
21756 @opindex mprioritize-restricted-insns
21757 This option controls the priority that is assigned to
21758 dispatch-slot restricted instructions during the second scheduling
21759 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
21760 or @samp{2} to assign no, highest, or second-highest (respectively)
21761 priority to dispatch-slot restricted
21764 @item -msched-costly-dep=@var{dependence_type}
21765 @opindex msched-costly-dep
21766 This option controls which dependences are considered costly
21767 by the target during instruction scheduling. The argument
21768 @var{dependence_type} takes one of the following values:
21772 No dependence is costly.
21775 All dependences are costly.
21777 @item @samp{true_store_to_load}
21778 A true dependence from store to load is costly.
21780 @item @samp{store_to_load}
21781 Any dependence from store to load is costly.
21784 Any dependence for which the latency is greater than or equal to
21785 @var{number} is costly.
21788 @item -minsert-sched-nops=@var{scheme}
21789 @opindex minsert-sched-nops
21790 This option controls which NOP insertion scheme is used during
21791 the second scheduling pass. The argument @var{scheme} takes one of the
21799 Pad with NOPs any dispatch group that has vacant issue slots,
21800 according to the scheduler's grouping.
21802 @item @samp{regroup_exact}
21803 Insert NOPs to force costly dependent insns into
21804 separate groups. Insert exactly as many NOPs as needed to force an insn
21805 to a new group, according to the estimated processor grouping.
21808 Insert NOPs to force costly dependent insns into
21809 separate groups. Insert @var{number} NOPs to force an insn to a new group.
21813 @opindex mcall-sysv
21814 On System V.4 and embedded PowerPC systems compile code using calling
21815 conventions that adhere to the March 1995 draft of the System V
21816 Application Binary Interface, PowerPC processor supplement. This is the
21817 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
21819 @item -mcall-sysv-eabi
21821 @opindex mcall-sysv-eabi
21822 @opindex mcall-eabi
21823 Specify both @option{-mcall-sysv} and @option{-meabi} options.
21825 @item -mcall-sysv-noeabi
21826 @opindex mcall-sysv-noeabi
21827 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
21829 @item -mcall-aixdesc
21831 On System V.4 and embedded PowerPC systems compile code for the AIX
21835 @opindex mcall-linux
21836 On System V.4 and embedded PowerPC systems compile code for the
21837 Linux-based GNU system.
21839 @item -mcall-freebsd
21840 @opindex mcall-freebsd
21841 On System V.4 and embedded PowerPC systems compile code for the
21842 FreeBSD operating system.
21844 @item -mcall-netbsd
21845 @opindex mcall-netbsd
21846 On System V.4 and embedded PowerPC systems compile code for the
21847 NetBSD operating system.
21849 @item -mcall-openbsd
21850 @opindex mcall-netbsd
21851 On System V.4 and embedded PowerPC systems compile code for the
21852 OpenBSD operating system.
21854 @item -maix-struct-return
21855 @opindex maix-struct-return
21856 Return all structures in memory (as specified by the AIX ABI)@.
21858 @item -msvr4-struct-return
21859 @opindex msvr4-struct-return
21860 Return structures smaller than 8 bytes in registers (as specified by the
21863 @item -mabi=@var{abi-type}
21865 Extend the current ABI with a particular extension, or remove such extension.
21866 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
21867 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
21868 @samp{elfv1}, @samp{elfv2}@.
21872 Extend the current ABI with SPE ABI extensions. This does not change
21873 the default ABI, instead it adds the SPE ABI extensions to the current
21877 @opindex mabi=no-spe
21878 Disable Book-E SPE ABI extensions for the current ABI@.
21880 @item -mabi=ibmlongdouble
21881 @opindex mabi=ibmlongdouble
21882 Change the current ABI to use IBM extended-precision long double.
21883 This is a PowerPC 32-bit SYSV ABI option.
21885 @item -mabi=ieeelongdouble
21886 @opindex mabi=ieeelongdouble
21887 Change the current ABI to use IEEE extended-precision long double.
21888 This is a PowerPC 32-bit Linux ABI option.
21891 @opindex mabi=elfv1
21892 Change the current ABI to use the ELFv1 ABI.
21893 This is the default ABI for big-endian PowerPC 64-bit Linux.
21894 Overriding the default ABI requires special system support and is
21895 likely to fail in spectacular ways.
21898 @opindex mabi=elfv2
21899 Change the current ABI to use the ELFv2 ABI.
21900 This is the default ABI for little-endian PowerPC 64-bit Linux.
21901 Overriding the default ABI requires special system support and is
21902 likely to fail in spectacular ways.
21904 @item -mgnu-attribute
21905 @itemx -mno-gnu-attribute
21906 @opindex mgnu-attribute
21907 @opindex mno-gnu-attribute
21908 Emit .gnu_attribute assembly directives to set tag/value pairs in a
21909 .gnu.attributes section that specify ABI variations in function
21910 parameters or return values.
21913 @itemx -mno-prototype
21914 @opindex mprototype
21915 @opindex mno-prototype
21916 On System V.4 and embedded PowerPC systems assume that all calls to
21917 variable argument functions are properly prototyped. Otherwise, the
21918 compiler must insert an instruction before every non-prototyped call to
21919 set or clear bit 6 of the condition code register (@code{CR}) to
21920 indicate whether floating-point values are passed in the floating-point
21921 registers in case the function takes variable arguments. With
21922 @option{-mprototype}, only calls to prototyped variable argument functions
21923 set or clear the bit.
21927 On embedded PowerPC systems, assume that the startup module is called
21928 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
21929 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
21934 On embedded PowerPC systems, assume that the startup module is called
21935 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
21940 On embedded PowerPC systems, assume that the startup module is called
21941 @file{crt0.o} and the standard C libraries are @file{libads.a} and
21944 @item -myellowknife
21945 @opindex myellowknife
21946 On embedded PowerPC systems, assume that the startup module is called
21947 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
21952 On System V.4 and embedded PowerPC systems, specify that you are
21953 compiling for a VxWorks system.
21957 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
21958 header to indicate that @samp{eabi} extended relocations are used.
21964 On System V.4 and embedded PowerPC systems do (do not) adhere to the
21965 Embedded Applications Binary Interface (EABI), which is a set of
21966 modifications to the System V.4 specifications. Selecting @option{-meabi}
21967 means that the stack is aligned to an 8-byte boundary, a function
21968 @code{__eabi} is called from @code{main} to set up the EABI
21969 environment, and the @option{-msdata} option can use both @code{r2} and
21970 @code{r13} to point to two separate small data areas. Selecting
21971 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
21972 no EABI initialization function is called from @code{main}, and the
21973 @option{-msdata} option only uses @code{r13} to point to a single
21974 small data area. The @option{-meabi} option is on by default if you
21975 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
21978 @opindex msdata=eabi
21979 On System V.4 and embedded PowerPC systems, put small initialized
21980 @code{const} global and static data in the @code{.sdata2} section, which
21981 is pointed to by register @code{r2}. Put small initialized
21982 non-@code{const} global and static data in the @code{.sdata} section,
21983 which is pointed to by register @code{r13}. Put small uninitialized
21984 global and static data in the @code{.sbss} section, which is adjacent to
21985 the @code{.sdata} section. The @option{-msdata=eabi} option is
21986 incompatible with the @option{-mrelocatable} option. The
21987 @option{-msdata=eabi} option also sets the @option{-memb} option.
21990 @opindex msdata=sysv
21991 On System V.4 and embedded PowerPC systems, put small global and static
21992 data in the @code{.sdata} section, which is pointed to by register
21993 @code{r13}. Put small uninitialized global and static data in the
21994 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
21995 The @option{-msdata=sysv} option is incompatible with the
21996 @option{-mrelocatable} option.
21998 @item -msdata=default
22000 @opindex msdata=default
22002 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
22003 compile code the same as @option{-msdata=eabi}, otherwise compile code the
22004 same as @option{-msdata=sysv}.
22007 @opindex msdata=data
22008 On System V.4 and embedded PowerPC systems, put small global
22009 data in the @code{.sdata} section. Put small uninitialized global
22010 data in the @code{.sbss} section. Do not use register @code{r13}
22011 to address small data however. This is the default behavior unless
22012 other @option{-msdata} options are used.
22016 @opindex msdata=none
22018 On embedded PowerPC systems, put all initialized global and static data
22019 in the @code{.data} section, and all uninitialized data in the
22020 @code{.bss} section.
22022 @item -mblock-move-inline-limit=@var{num}
22023 @opindex mblock-move-inline-limit
22024 Inline all block moves (such as calls to @code{memcpy} or structure
22025 copies) less than or equal to @var{num} bytes. The minimum value for
22026 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
22027 targets. The default value is target-specific.
22031 @cindex smaller data references (PowerPC)
22032 @cindex .sdata/.sdata2 references (PowerPC)
22033 On embedded PowerPC systems, put global and static items less than or
22034 equal to @var{num} bytes into the small data or BSS sections instead of
22035 the normal data or BSS section. By default, @var{num} is 8. The
22036 @option{-G @var{num}} switch is also passed to the linker.
22037 All modules should be compiled with the same @option{-G @var{num}} value.
22040 @itemx -mno-regnames
22042 @opindex mno-regnames
22043 On System V.4 and embedded PowerPC systems do (do not) emit register
22044 names in the assembly language output using symbolic forms.
22047 @itemx -mno-longcall
22049 @opindex mno-longcall
22050 By default assume that all calls are far away so that a longer and more
22051 expensive calling sequence is required. This is required for calls
22052 farther than 32 megabytes (33,554,432 bytes) from the current location.
22053 A short call is generated if the compiler knows
22054 the call cannot be that far away. This setting can be overridden by
22055 the @code{shortcall} function attribute, or by @code{#pragma
22058 Some linkers are capable of detecting out-of-range calls and generating
22059 glue code on the fly. On these systems, long calls are unnecessary and
22060 generate slower code. As of this writing, the AIX linker can do this,
22061 as can the GNU linker for PowerPC/64. It is planned to add this feature
22062 to the GNU linker for 32-bit PowerPC systems as well.
22064 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
22065 callee, L42}, plus a @dfn{branch island} (glue code). The two target
22066 addresses represent the callee and the branch island. The
22067 Darwin/PPC linker prefers the first address and generates a @code{bl
22068 callee} if the PPC @code{bl} instruction reaches the callee directly;
22069 otherwise, the linker generates @code{bl L42} to call the branch
22070 island. The branch island is appended to the body of the
22071 calling function; it computes the full 32-bit address of the callee
22074 On Mach-O (Darwin) systems, this option directs the compiler emit to
22075 the glue for every direct call, and the Darwin linker decides whether
22076 to use or discard it.
22078 In the future, GCC may ignore all longcall specifications
22079 when the linker is known to generate glue.
22081 @item -mtls-markers
22082 @itemx -mno-tls-markers
22083 @opindex mtls-markers
22084 @opindex mno-tls-markers
22085 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22086 specifying the function argument. The relocation allows the linker to
22087 reliably associate function call with argument setup instructions for
22088 TLS optimization, which in turn allows GCC to better schedule the
22094 This option enables use of the reciprocal estimate and
22095 reciprocal square root estimate instructions with additional
22096 Newton-Raphson steps to increase precision instead of doing a divide or
22097 square root and divide for floating-point arguments. You should use
22098 the @option{-ffast-math} option when using @option{-mrecip} (or at
22099 least @option{-funsafe-math-optimizations},
22100 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22101 @option{-fno-trapping-math}). Note that while the throughput of the
22102 sequence is generally higher than the throughput of the non-reciprocal
22103 instruction, the precision of the sequence can be decreased by up to 2
22104 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
22107 @item -mrecip=@var{opt}
22108 @opindex mrecip=opt
22109 This option controls which reciprocal estimate instructions
22110 may be used. @var{opt} is a comma-separated list of options, which may
22111 be preceded by a @code{!} to invert the option:
22116 Enable all estimate instructions.
22119 Enable the default instructions, equivalent to @option{-mrecip}.
22122 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22125 Enable the reciprocal approximation instructions for both
22126 single and double precision.
22129 Enable the single-precision reciprocal approximation instructions.
22132 Enable the double-precision reciprocal approximation instructions.
22135 Enable the reciprocal square root approximation instructions for both
22136 single and double precision.
22139 Enable the single-precision reciprocal square root approximation instructions.
22142 Enable the double-precision reciprocal square root approximation instructions.
22146 So, for example, @option{-mrecip=all,!rsqrtd} enables
22147 all of the reciprocal estimate instructions, except for the
22148 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
22149 which handle the double-precision reciprocal square root calculations.
22151 @item -mrecip-precision
22152 @itemx -mno-recip-precision
22153 @opindex mrecip-precision
22154 Assume (do not assume) that the reciprocal estimate instructions
22155 provide higher-precision estimates than is mandated by the PowerPC
22156 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
22157 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
22158 The double-precision square root estimate instructions are not generated by
22159 default on low-precision machines, since they do not provide an
22160 estimate that converges after three steps.
22162 @item -mveclibabi=@var{type}
22163 @opindex mveclibabi
22164 Specifies the ABI type to use for vectorizing intrinsics using an
22165 external library. The only type supported at present is @samp{mass},
22166 which specifies to use IBM's Mathematical Acceleration Subsystem
22167 (MASS) libraries for vectorizing intrinsics using external libraries.
22168 GCC currently emits calls to @code{acosd2}, @code{acosf4},
22169 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
22170 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
22171 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
22172 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
22173 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
22174 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
22175 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
22176 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
22177 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
22178 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
22179 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
22180 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
22181 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
22182 for power7. Both @option{-ftree-vectorize} and
22183 @option{-funsafe-math-optimizations} must also be enabled. The MASS
22184 libraries must be specified at link time.
22189 Generate (do not generate) the @code{friz} instruction when the
22190 @option{-funsafe-math-optimizations} option is used to optimize
22191 rounding of floating-point values to 64-bit integer and back to floating
22192 point. The @code{friz} instruction does not return the same value if
22193 the floating-point number is too large to fit in an integer.
22195 @item -mpointers-to-nested-functions
22196 @itemx -mno-pointers-to-nested-functions
22197 @opindex mpointers-to-nested-functions
22198 Generate (do not generate) code to load up the static chain register
22199 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
22200 systems where a function pointer points to a 3-word descriptor giving
22201 the function address, TOC value to be loaded in register @code{r2}, and
22202 static chain value to be loaded in register @code{r11}. The
22203 @option{-mpointers-to-nested-functions} is on by default. You cannot
22204 call through pointers to nested functions or pointers
22205 to functions compiled in other languages that use the static chain if
22206 you use @option{-mno-pointers-to-nested-functions}.
22208 @item -msave-toc-indirect
22209 @itemx -mno-save-toc-indirect
22210 @opindex msave-toc-indirect
22211 Generate (do not generate) code to save the TOC value in the reserved
22212 stack location in the function prologue if the function calls through
22213 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
22214 saved in the prologue, it is saved just before the call through the
22215 pointer. The @option{-mno-save-toc-indirect} option is the default.
22217 @item -mcompat-align-parm
22218 @itemx -mno-compat-align-parm
22219 @opindex mcompat-align-parm
22220 Generate (do not generate) code to pass structure parameters with a
22221 maximum alignment of 64 bits, for compatibility with older versions
22224 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
22225 structure parameter on a 128-bit boundary when that structure contained
22226 a member requiring 128-bit alignment. This is corrected in more
22227 recent versions of GCC. This option may be used to generate code
22228 that is compatible with functions compiled with older versions of
22231 The @option{-mno-compat-align-parm} option is the default.
22233 @item -mstack-protector-guard=@var{guard}
22234 @itemx -mstack-protector-guard-reg=@var{reg}
22235 @itemx -mstack-protector-guard-offset=@var{offset}
22236 @opindex mstack-protector-guard
22237 @opindex mstack-protector-guard-reg
22238 @opindex mstack-protector-guard-offset
22239 Generate stack protection code using canary at @var{guard}. Supported
22240 locations are @samp{global} for global canary or @samp{tls} for per-thread
22241 canary in the TLS block (the default with GNU libc version 2.4 or later).
22243 With the latter choice the options
22244 @option{-mstack-protector-guard-reg=@var{reg}} and
22245 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
22246 which register to use as base register for reading the canary, and from what
22247 offset from that base register. The default for those is as specified in the
22252 @subsection RX Options
22255 These command-line options are defined for RX targets:
22258 @item -m64bit-doubles
22259 @itemx -m32bit-doubles
22260 @opindex m64bit-doubles
22261 @opindex m32bit-doubles
22262 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
22263 or 32 bits (@option{-m32bit-doubles}) in size. The default is
22264 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
22265 works on 32-bit values, which is why the default is
22266 @option{-m32bit-doubles}.
22272 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
22273 floating-point hardware. The default is enabled for the RX600
22274 series and disabled for the RX200 series.
22276 Floating-point instructions are only generated for 32-bit floating-point
22277 values, however, so the FPU hardware is not used for doubles if the
22278 @option{-m64bit-doubles} option is used.
22280 @emph{Note} If the @option{-fpu} option is enabled then
22281 @option{-funsafe-math-optimizations} is also enabled automatically.
22282 This is because the RX FPU instructions are themselves unsafe.
22284 @item -mcpu=@var{name}
22286 Selects the type of RX CPU to be targeted. Currently three types are
22287 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
22288 the specific @samp{RX610} CPU. The default is @samp{RX600}.
22290 The only difference between @samp{RX600} and @samp{RX610} is that the
22291 @samp{RX610} does not support the @code{MVTIPL} instruction.
22293 The @samp{RX200} series does not have a hardware floating-point unit
22294 and so @option{-nofpu} is enabled by default when this type is
22297 @item -mbig-endian-data
22298 @itemx -mlittle-endian-data
22299 @opindex mbig-endian-data
22300 @opindex mlittle-endian-data
22301 Store data (but not code) in the big-endian format. The default is
22302 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
22305 @item -msmall-data-limit=@var{N}
22306 @opindex msmall-data-limit
22307 Specifies the maximum size in bytes of global and static variables
22308 which can be placed into the small data area. Using the small data
22309 area can lead to smaller and faster code, but the size of area is
22310 limited and it is up to the programmer to ensure that the area does
22311 not overflow. Also when the small data area is used one of the RX's
22312 registers (usually @code{r13}) is reserved for use pointing to this
22313 area, so it is no longer available for use by the compiler. This
22314 could result in slower and/or larger code if variables are pushed onto
22315 the stack instead of being held in this register.
22317 Note, common variables (variables that have not been initialized) and
22318 constants are not placed into the small data area as they are assigned
22319 to other sections in the output executable.
22321 The default value is zero, which disables this feature. Note, this
22322 feature is not enabled by default with higher optimization levels
22323 (@option{-O2} etc) because of the potentially detrimental effects of
22324 reserving a register. It is up to the programmer to experiment and
22325 discover whether this feature is of benefit to their program. See the
22326 description of the @option{-mpid} option for a description of how the
22327 actual register to hold the small data area pointer is chosen.
22333 Use the simulator runtime. The default is to use the libgloss
22334 board-specific runtime.
22336 @item -mas100-syntax
22337 @itemx -mno-as100-syntax
22338 @opindex mas100-syntax
22339 @opindex mno-as100-syntax
22340 When generating assembler output use a syntax that is compatible with
22341 Renesas's AS100 assembler. This syntax can also be handled by the GAS
22342 assembler, but it has some restrictions so it is not generated by default.
22344 @item -mmax-constant-size=@var{N}
22345 @opindex mmax-constant-size
22346 Specifies the maximum size, in bytes, of a constant that can be used as
22347 an operand in a RX instruction. Although the RX instruction set does
22348 allow constants of up to 4 bytes in length to be used in instructions,
22349 a longer value equates to a longer instruction. Thus in some
22350 circumstances it can be beneficial to restrict the size of constants
22351 that are used in instructions. Constants that are too big are instead
22352 placed into a constant pool and referenced via register indirection.
22354 The value @var{N} can be between 0 and 4. A value of 0 (the default)
22355 or 4 means that constants of any size are allowed.
22359 Enable linker relaxation. Linker relaxation is a process whereby the
22360 linker attempts to reduce the size of a program by finding shorter
22361 versions of various instructions. Disabled by default.
22363 @item -mint-register=@var{N}
22364 @opindex mint-register
22365 Specify the number of registers to reserve for fast interrupt handler
22366 functions. The value @var{N} can be between 0 and 4. A value of 1
22367 means that register @code{r13} is reserved for the exclusive use
22368 of fast interrupt handlers. A value of 2 reserves @code{r13} and
22369 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
22370 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
22371 A value of 0, the default, does not reserve any registers.
22373 @item -msave-acc-in-interrupts
22374 @opindex msave-acc-in-interrupts
22375 Specifies that interrupt handler functions should preserve the
22376 accumulator register. This is only necessary if normal code might use
22377 the accumulator register, for example because it performs 64-bit
22378 multiplications. The default is to ignore the accumulator as this
22379 makes the interrupt handlers faster.
22385 Enables the generation of position independent data. When enabled any
22386 access to constant data is done via an offset from a base address
22387 held in a register. This allows the location of constant data to be
22388 determined at run time without requiring the executable to be
22389 relocated, which is a benefit to embedded applications with tight
22390 memory constraints. Data that can be modified is not affected by this
22393 Note, using this feature reserves a register, usually @code{r13}, for
22394 the constant data base address. This can result in slower and/or
22395 larger code, especially in complicated functions.
22397 The actual register chosen to hold the constant data base address
22398 depends upon whether the @option{-msmall-data-limit} and/or the
22399 @option{-mint-register} command-line options are enabled. Starting
22400 with register @code{r13} and proceeding downwards, registers are
22401 allocated first to satisfy the requirements of @option{-mint-register},
22402 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
22403 is possible for the small data area register to be @code{r8} if both
22404 @option{-mint-register=4} and @option{-mpid} are specified on the
22407 By default this feature is not enabled. The default can be restored
22408 via the @option{-mno-pid} command-line option.
22410 @item -mno-warn-multiple-fast-interrupts
22411 @itemx -mwarn-multiple-fast-interrupts
22412 @opindex mno-warn-multiple-fast-interrupts
22413 @opindex mwarn-multiple-fast-interrupts
22414 Prevents GCC from issuing a warning message if it finds more than one
22415 fast interrupt handler when it is compiling a file. The default is to
22416 issue a warning for each extra fast interrupt handler found, as the RX
22417 only supports one such interrupt.
22419 @item -mallow-string-insns
22420 @itemx -mno-allow-string-insns
22421 @opindex mallow-string-insns
22422 @opindex mno-allow-string-insns
22423 Enables or disables the use of the string manipulation instructions
22424 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
22425 @code{SWHILE} and also the @code{RMPA} instruction. These
22426 instructions may prefetch data, which is not safe to do if accessing
22427 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
22428 for more information).
22430 The default is to allow these instructions, but it is not possible for
22431 GCC to reliably detect all circumstances where a string instruction
22432 might be used to access an I/O register, so their use cannot be
22433 disabled automatically. Instead it is reliant upon the programmer to
22434 use the @option{-mno-allow-string-insns} option if their program
22435 accesses I/O space.
22437 When the instructions are enabled GCC defines the C preprocessor
22438 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
22439 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
22445 Use only (or not only) @code{JSR} instructions to access functions.
22446 This option can be used when code size exceeds the range of @code{BSR}
22447 instructions. Note that @option{-mno-jsr} does not mean to not use
22448 @code{JSR} but instead means that any type of branch may be used.
22451 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
22452 has special significance to the RX port when used with the
22453 @code{interrupt} function attribute. This attribute indicates a
22454 function intended to process fast interrupts. GCC ensures
22455 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
22456 and/or @code{r13} and only provided that the normal use of the
22457 corresponding registers have been restricted via the
22458 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
22461 @node S/390 and zSeries Options
22462 @subsection S/390 and zSeries Options
22463 @cindex S/390 and zSeries Options
22465 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
22469 @itemx -msoft-float
22470 @opindex mhard-float
22471 @opindex msoft-float
22472 Use (do not use) the hardware floating-point instructions and registers
22473 for floating-point operations. When @option{-msoft-float} is specified,
22474 functions in @file{libgcc.a} are used to perform floating-point
22475 operations. When @option{-mhard-float} is specified, the compiler
22476 generates IEEE floating-point instructions. This is the default.
22479 @itemx -mno-hard-dfp
22481 @opindex mno-hard-dfp
22482 Use (do not use) the hardware decimal-floating-point instructions for
22483 decimal-floating-point operations. When @option{-mno-hard-dfp} is
22484 specified, functions in @file{libgcc.a} are used to perform
22485 decimal-floating-point operations. When @option{-mhard-dfp} is
22486 specified, the compiler generates decimal-floating-point hardware
22487 instructions. This is the default for @option{-march=z9-ec} or higher.
22489 @item -mlong-double-64
22490 @itemx -mlong-double-128
22491 @opindex mlong-double-64
22492 @opindex mlong-double-128
22493 These switches control the size of @code{long double} type. A size
22494 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22495 type. This is the default.
22498 @itemx -mno-backchain
22499 @opindex mbackchain
22500 @opindex mno-backchain
22501 Store (do not store) the address of the caller's frame as backchain pointer
22502 into the callee's stack frame.
22503 A backchain may be needed to allow debugging using tools that do not understand
22504 DWARF call frame information.
22505 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
22506 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
22507 the backchain is placed into the topmost word of the 96/160 byte register
22510 In general, code compiled with @option{-mbackchain} is call-compatible with
22511 code compiled with @option{-mmo-backchain}; however, use of the backchain
22512 for debugging purposes usually requires that the whole binary is built with
22513 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
22514 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22515 to build a linux kernel use @option{-msoft-float}.
22517 The default is to not maintain the backchain.
22519 @item -mpacked-stack
22520 @itemx -mno-packed-stack
22521 @opindex mpacked-stack
22522 @opindex mno-packed-stack
22523 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
22524 specified, the compiler uses the all fields of the 96/160 byte register save
22525 area only for their default purpose; unused fields still take up stack space.
22526 When @option{-mpacked-stack} is specified, register save slots are densely
22527 packed at the top of the register save area; unused space is reused for other
22528 purposes, allowing for more efficient use of the available stack space.
22529 However, when @option{-mbackchain} is also in effect, the topmost word of
22530 the save area is always used to store the backchain, and the return address
22531 register is always saved two words below the backchain.
22533 As long as the stack frame backchain is not used, code generated with
22534 @option{-mpacked-stack} is call-compatible with code generated with
22535 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
22536 S/390 or zSeries generated code that uses the stack frame backchain at run
22537 time, not just for debugging purposes. Such code is not call-compatible
22538 with code compiled with @option{-mpacked-stack}. Also, note that the
22539 combination of @option{-mbackchain},
22540 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22541 to build a linux kernel use @option{-msoft-float}.
22543 The default is to not use the packed stack layout.
22546 @itemx -mno-small-exec
22547 @opindex msmall-exec
22548 @opindex mno-small-exec
22549 Generate (or do not generate) code using the @code{bras} instruction
22550 to do subroutine calls.
22551 This only works reliably if the total executable size does not
22552 exceed 64k. The default is to use the @code{basr} instruction instead,
22553 which does not have this limitation.
22559 When @option{-m31} is specified, generate code compliant to the
22560 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
22561 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
22562 particular to generate 64-bit instructions. For the @samp{s390}
22563 targets, the default is @option{-m31}, while the @samp{s390x}
22564 targets default to @option{-m64}.
22570 When @option{-mzarch} is specified, generate code using the
22571 instructions available on z/Architecture.
22572 When @option{-mesa} is specified, generate code using the
22573 instructions available on ESA/390. Note that @option{-mesa} is
22574 not possible with @option{-m64}.
22575 When generating code compliant to the GNU/Linux for S/390 ABI,
22576 the default is @option{-mesa}. When generating code compliant
22577 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
22583 The @option{-mhtm} option enables a set of builtins making use of
22584 instructions available with the transactional execution facility
22585 introduced with the IBM zEnterprise EC12 machine generation
22586 @ref{S/390 System z Built-in Functions}.
22587 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
22593 When @option{-mvx} is specified, generate code using the instructions
22594 available with the vector extension facility introduced with the IBM
22595 z13 machine generation.
22596 This option changes the ABI for some vector type values with regard to
22597 alignment and calling conventions. In case vector type values are
22598 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
22599 command will be added to mark the resulting binary with the ABI used.
22600 @option{-mvx} is enabled by default when using @option{-march=z13}.
22603 @itemx -mno-zvector
22605 @opindex mno-zvector
22606 The @option{-mzvector} option enables vector language extensions and
22607 builtins using instructions available with the vector extension
22608 facility introduced with the IBM z13 machine generation.
22609 This option adds support for @samp{vector} to be used as a keyword to
22610 define vector type variables and arguments. @samp{vector} is only
22611 available when GNU extensions are enabled. It will not be expanded
22612 when requesting strict standard compliance e.g. with @option{-std=c99}.
22613 In addition to the GCC low-level builtins @option{-mzvector} enables
22614 a set of builtins added for compatibility with AltiVec-style
22615 implementations like Power and Cell. In order to make use of these
22616 builtins the header file @file{vecintrin.h} needs to be included.
22617 @option{-mzvector} is disabled by default.
22623 Generate (or do not generate) code using the @code{mvcle} instruction
22624 to perform block moves. When @option{-mno-mvcle} is specified,
22625 use a @code{mvc} loop instead. This is the default unless optimizing for
22632 Print (or do not print) additional debug information when compiling.
22633 The default is to not print debug information.
22635 @item -march=@var{cpu-type}
22637 Generate code that runs on @var{cpu-type}, which is the name of a
22638 system representing a certain processor type. Possible values for
22639 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
22640 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
22641 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
22644 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
22645 @samp{g6} are deprecated and will be removed with future releases.
22647 Specifying @samp{native} as cpu type can be used to select the best
22648 architecture option for the host processor.
22649 @option{-march=native} has no effect if GCC does not recognize the
22652 @item -mtune=@var{cpu-type}
22654 Tune to @var{cpu-type} everything applicable about the generated code,
22655 except for the ABI and the set of available instructions.
22656 The list of @var{cpu-type} values is the same as for @option{-march}.
22657 The default is the value used for @option{-march}.
22660 @itemx -mno-tpf-trace
22661 @opindex mtpf-trace
22662 @opindex mno-tpf-trace
22663 Generate code that adds (does not add) in TPF OS specific branches to trace
22664 routines in the operating system. This option is off by default, even
22665 when compiling for the TPF OS@.
22668 @itemx -mno-fused-madd
22669 @opindex mfused-madd
22670 @opindex mno-fused-madd
22671 Generate code that uses (does not use) the floating-point multiply and
22672 accumulate instructions. These instructions are generated by default if
22673 hardware floating point is used.
22675 @item -mwarn-framesize=@var{framesize}
22676 @opindex mwarn-framesize
22677 Emit a warning if the current function exceeds the given frame size. Because
22678 this is a compile-time check it doesn't need to be a real problem when the program
22679 runs. It is intended to identify functions that most probably cause
22680 a stack overflow. It is useful to be used in an environment with limited stack
22681 size e.g.@: the linux kernel.
22683 @item -mwarn-dynamicstack
22684 @opindex mwarn-dynamicstack
22685 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
22686 arrays. This is generally a bad idea with a limited stack size.
22688 @item -mstack-guard=@var{stack-guard}
22689 @itemx -mstack-size=@var{stack-size}
22690 @opindex mstack-guard
22691 @opindex mstack-size
22692 If these options are provided the S/390 back end emits additional instructions in
22693 the function prologue that trigger a trap if the stack size is @var{stack-guard}
22694 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
22695 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
22696 the frame size of the compiled function is chosen.
22697 These options are intended to be used to help debugging stack overflow problems.
22698 The additionally emitted code causes only little overhead and hence can also be
22699 used in production-like systems without greater performance degradation. The given
22700 values have to be exact powers of 2 and @var{stack-size} has to be greater than
22701 @var{stack-guard} without exceeding 64k.
22702 In order to be efficient the extra code makes the assumption that the stack starts
22703 at an address aligned to the value given by @var{stack-size}.
22704 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
22706 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
22708 If the hotpatch option is enabled, a ``hot-patching'' function
22709 prologue is generated for all functions in the compilation unit.
22710 The funtion label is prepended with the given number of two-byte
22711 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
22712 the label, 2 * @var{post-halfwords} bytes are appended, using the
22713 largest NOP like instructions the architecture allows (maximum
22716 If both arguments are zero, hotpatching is disabled.
22718 This option can be overridden for individual functions with the
22719 @code{hotpatch} attribute.
22722 @node Score Options
22723 @subsection Score Options
22724 @cindex Score Options
22726 These options are defined for Score implementations:
22731 Compile code for big-endian mode. This is the default.
22735 Compile code for little-endian mode.
22739 Disable generation of @code{bcnz} instructions.
22743 Enable generation of unaligned load and store instructions.
22747 Enable the use of multiply-accumulate instructions. Disabled by default.
22751 Specify the SCORE5 as the target architecture.
22755 Specify the SCORE5U of the target architecture.
22759 Specify the SCORE7 as the target architecture. This is the default.
22763 Specify the SCORE7D as the target architecture.
22767 @subsection SH Options
22769 These @samp{-m} options are defined for the SH implementations:
22774 Generate code for the SH1.
22778 Generate code for the SH2.
22781 Generate code for the SH2e.
22785 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
22786 that the floating-point unit is not used.
22788 @item -m2a-single-only
22789 @opindex m2a-single-only
22790 Generate code for the SH2a-FPU, in such a way that no double-precision
22791 floating-point operations are used.
22794 @opindex m2a-single
22795 Generate code for the SH2a-FPU assuming the floating-point unit is in
22796 single-precision mode by default.
22800 Generate code for the SH2a-FPU assuming the floating-point unit is in
22801 double-precision mode by default.
22805 Generate code for the SH3.
22809 Generate code for the SH3e.
22813 Generate code for the SH4 without a floating-point unit.
22815 @item -m4-single-only
22816 @opindex m4-single-only
22817 Generate code for the SH4 with a floating-point unit that only
22818 supports single-precision arithmetic.
22822 Generate code for the SH4 assuming the floating-point unit is in
22823 single-precision mode by default.
22827 Generate code for the SH4.
22831 Generate code for SH4-100.
22833 @item -m4-100-nofpu
22834 @opindex m4-100-nofpu
22835 Generate code for SH4-100 in such a way that the
22836 floating-point unit is not used.
22838 @item -m4-100-single
22839 @opindex m4-100-single
22840 Generate code for SH4-100 assuming the floating-point unit is in
22841 single-precision mode by default.
22843 @item -m4-100-single-only
22844 @opindex m4-100-single-only
22845 Generate code for SH4-100 in such a way that no double-precision
22846 floating-point operations are used.
22850 Generate code for SH4-200.
22852 @item -m4-200-nofpu
22853 @opindex m4-200-nofpu
22854 Generate code for SH4-200 without in such a way that the
22855 floating-point unit is not used.
22857 @item -m4-200-single
22858 @opindex m4-200-single
22859 Generate code for SH4-200 assuming the floating-point unit is in
22860 single-precision mode by default.
22862 @item -m4-200-single-only
22863 @opindex m4-200-single-only
22864 Generate code for SH4-200 in such a way that no double-precision
22865 floating-point operations are used.
22869 Generate code for SH4-300.
22871 @item -m4-300-nofpu
22872 @opindex m4-300-nofpu
22873 Generate code for SH4-300 without in such a way that the
22874 floating-point unit is not used.
22876 @item -m4-300-single
22877 @opindex m4-300-single
22878 Generate code for SH4-300 in such a way that no double-precision
22879 floating-point operations are used.
22881 @item -m4-300-single-only
22882 @opindex m4-300-single-only
22883 Generate code for SH4-300 in such a way that no double-precision
22884 floating-point operations are used.
22888 Generate code for SH4-340 (no MMU, no FPU).
22892 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
22897 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
22898 floating-point unit is not used.
22900 @item -m4a-single-only
22901 @opindex m4a-single-only
22902 Generate code for the SH4a, in such a way that no double-precision
22903 floating-point operations are used.
22906 @opindex m4a-single
22907 Generate code for the SH4a assuming the floating-point unit is in
22908 single-precision mode by default.
22912 Generate code for the SH4a.
22916 Same as @option{-m4a-nofpu}, except that it implicitly passes
22917 @option{-dsp} to the assembler. GCC doesn't generate any DSP
22918 instructions at the moment.
22922 Compile code for the processor in big-endian mode.
22926 Compile code for the processor in little-endian mode.
22930 Align doubles at 64-bit boundaries. Note that this changes the calling
22931 conventions, and thus some functions from the standard C library do
22932 not work unless you recompile it first with @option{-mdalign}.
22936 Shorten some address references at link time, when possible; uses the
22937 linker option @option{-relax}.
22941 Use 32-bit offsets in @code{switch} tables. The default is to use
22946 Enable the use of bit manipulation instructions on SH2A.
22950 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
22951 alignment constraints.
22955 Comply with the calling conventions defined by Renesas.
22958 @opindex mno-renesas
22959 Comply with the calling conventions defined for GCC before the Renesas
22960 conventions were available. This option is the default for all
22961 targets of the SH toolchain.
22964 @opindex mnomacsave
22965 Mark the @code{MAC} register as call-clobbered, even if
22966 @option{-mrenesas} is given.
22972 Control the IEEE compliance of floating-point comparisons, which affects the
22973 handling of cases where the result of a comparison is unordered. By default
22974 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
22975 enabled @option{-mno-ieee} is implicitly set, which results in faster
22976 floating-point greater-equal and less-equal comparisons. The implicit settings
22977 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
22979 @item -minline-ic_invalidate
22980 @opindex minline-ic_invalidate
22981 Inline code to invalidate instruction cache entries after setting up
22982 nested function trampolines.
22983 This option has no effect if @option{-musermode} is in effect and the selected
22984 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
22986 If the selected code generation option does not allow the use of the @code{icbi}
22987 instruction, and @option{-musermode} is not in effect, the inlined code
22988 manipulates the instruction cache address array directly with an associative
22989 write. This not only requires privileged mode at run time, but it also
22990 fails if the cache line had been mapped via the TLB and has become unmapped.
22994 Dump instruction size and location in the assembly code.
22997 @opindex mpadstruct
22998 This option is deprecated. It pads structures to multiple of 4 bytes,
22999 which is incompatible with the SH ABI@.
23001 @item -matomic-model=@var{model}
23002 @opindex matomic-model=@var{model}
23003 Sets the model of atomic operations and additional parameters as a comma
23004 separated list. For details on the atomic built-in functions see
23005 @ref{__atomic Builtins}. The following models and parameters are supported:
23010 Disable compiler generated atomic sequences and emit library calls for atomic
23011 operations. This is the default if the target is not @code{sh*-*-linux*}.
23014 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
23015 built-in functions. The generated atomic sequences require additional support
23016 from the interrupt/exception handling code of the system and are only suitable
23017 for SH3* and SH4* single-core systems. This option is enabled by default when
23018 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
23019 this option also partially utilizes the hardware atomic instructions
23020 @code{movli.l} and @code{movco.l} to create more efficient code, unless
23021 @samp{strict} is specified.
23024 Generate software atomic sequences that use a variable in the thread control
23025 block. This is a variation of the gUSA sequences which can also be used on
23026 SH1* and SH2* targets. The generated atomic sequences require additional
23027 support from the interrupt/exception handling code of the system and are only
23028 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
23029 parameter has to be specified as well.
23032 Generate software atomic sequences that temporarily disable interrupts by
23033 setting @code{SR.IMASK = 1111}. This model works only when the program runs
23034 in privileged mode and is only suitable for single-core systems. Additional
23035 support from the interrupt/exception handling code of the system is not
23036 required. This model is enabled by default when the target is
23037 @code{sh*-*-linux*} and SH1* or SH2*.
23040 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
23041 instructions only. This is only available on SH4A and is suitable for
23042 multi-core systems. Since the hardware instructions support only 32 bit atomic
23043 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
23044 Code compiled with this option is also compatible with other software
23045 atomic model interrupt/exception handling systems if executed on an SH4A
23046 system. Additional support from the interrupt/exception handling code of the
23047 system is not required for this model.
23050 This parameter specifies the offset in bytes of the variable in the thread
23051 control block structure that should be used by the generated atomic sequences
23052 when the @samp{soft-tcb} model has been selected. For other models this
23053 parameter is ignored. The specified value must be an integer multiple of four
23054 and in the range 0-1020.
23057 This parameter prevents mixed usage of multiple atomic models, even if they
23058 are compatible, and makes the compiler generate atomic sequences of the
23059 specified model only.
23065 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
23066 Notice that depending on the particular hardware and software configuration
23067 this can degrade overall performance due to the operand cache line flushes
23068 that are implied by the @code{tas.b} instruction. On multi-core SH4A
23069 processors the @code{tas.b} instruction must be used with caution since it
23070 can result in data corruption for certain cache configurations.
23073 @opindex mprefergot
23074 When generating position-independent code, emit function calls using
23075 the Global Offset Table instead of the Procedure Linkage Table.
23078 @itemx -mno-usermode
23080 @opindex mno-usermode
23081 Don't allow (allow) the compiler generating privileged mode code. Specifying
23082 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
23083 inlined code would not work in user mode. @option{-musermode} is the default
23084 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
23085 @option{-musermode} has no effect, since there is no user mode.
23087 @item -multcost=@var{number}
23088 @opindex multcost=@var{number}
23089 Set the cost to assume for a multiply insn.
23091 @item -mdiv=@var{strategy}
23092 @opindex mdiv=@var{strategy}
23093 Set the division strategy to be used for integer division operations.
23094 @var{strategy} can be one of:
23099 Calls a library function that uses the single-step division instruction
23100 @code{div1} to perform the operation. Division by zero calculates an
23101 unspecified result and does not trap. This is the default except for SH4,
23102 SH2A and SHcompact.
23105 Calls a library function that performs the operation in double precision
23106 floating point. Division by zero causes a floating-point exception. This is
23107 the default for SHcompact with FPU. Specifying this for targets that do not
23108 have a double precision FPU defaults to @code{call-div1}.
23111 Calls a library function that uses a lookup table for small divisors and
23112 the @code{div1} instruction with case distinction for larger divisors. Division
23113 by zero calculates an unspecified result and does not trap. This is the default
23114 for SH4. Specifying this for targets that do not have dynamic shift
23115 instructions defaults to @code{call-div1}.
23119 When a division strategy has not been specified the default strategy is
23120 selected based on the current target. For SH2A the default strategy is to
23121 use the @code{divs} and @code{divu} instructions instead of library function
23124 @item -maccumulate-outgoing-args
23125 @opindex maccumulate-outgoing-args
23126 Reserve space once for outgoing arguments in the function prologue rather
23127 than around each call. Generally beneficial for performance and size. Also
23128 needed for unwinding to avoid changing the stack frame around conditional code.
23130 @item -mdivsi3_libfunc=@var{name}
23131 @opindex mdivsi3_libfunc=@var{name}
23132 Set the name of the library function used for 32-bit signed division to
23134 This only affects the name used in the @samp{call} division strategies, and
23135 the compiler still expects the same sets of input/output/clobbered registers as
23136 if this option were not present.
23138 @item -mfixed-range=@var{register-range}
23139 @opindex mfixed-range
23140 Generate code treating the given register range as fixed registers.
23141 A fixed register is one that the register allocator can not use. This is
23142 useful when compiling kernel code. A register range is specified as
23143 two registers separated by a dash. Multiple register ranges can be
23144 specified separated by a comma.
23146 @item -mbranch-cost=@var{num}
23147 @opindex mbranch-cost=@var{num}
23148 Assume @var{num} to be the cost for a branch instruction. Higher numbers
23149 make the compiler try to generate more branch-free code if possible.
23150 If not specified the value is selected depending on the processor type that
23151 is being compiled for.
23154 @itemx -mno-zdcbranch
23155 @opindex mzdcbranch
23156 @opindex mno-zdcbranch
23157 Assume (do not assume) that zero displacement conditional branch instructions
23158 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
23159 compiler prefers zero displacement branch code sequences. This is
23160 enabled by default when generating code for SH4 and SH4A. It can be explicitly
23161 disabled by specifying @option{-mno-zdcbranch}.
23163 @item -mcbranch-force-delay-slot
23164 @opindex mcbranch-force-delay-slot
23165 Force the usage of delay slots for conditional branches, which stuffs the delay
23166 slot with a @code{nop} if a suitable instruction cannot be found. By default
23167 this option is disabled. It can be enabled to work around hardware bugs as
23168 found in the original SH7055.
23171 @itemx -mno-fused-madd
23172 @opindex mfused-madd
23173 @opindex mno-fused-madd
23174 Generate code that uses (does not use) the floating-point multiply and
23175 accumulate instructions. These instructions are generated by default
23176 if hardware floating point is used. The machine-dependent
23177 @option{-mfused-madd} option is now mapped to the machine-independent
23178 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
23179 mapped to @option{-ffp-contract=off}.
23185 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
23186 and cosine approximations. The option @option{-mfsca} must be used in
23187 combination with @option{-funsafe-math-optimizations}. It is enabled by default
23188 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
23189 approximations even if @option{-funsafe-math-optimizations} is in effect.
23195 Allow or disallow the compiler to emit the @code{fsrra} instruction for
23196 reciprocal square root approximations. The option @option{-mfsrra} must be used
23197 in combination with @option{-funsafe-math-optimizations} and
23198 @option{-ffinite-math-only}. It is enabled by default when generating code for
23199 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
23200 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
23203 @item -mpretend-cmove
23204 @opindex mpretend-cmove
23205 Prefer zero-displacement conditional branches for conditional move instruction
23206 patterns. This can result in faster code on the SH4 processor.
23210 Generate code using the FDPIC ABI.
23214 @node Solaris 2 Options
23215 @subsection Solaris 2 Options
23216 @cindex Solaris 2 options
23218 These @samp{-m} options are supported on Solaris 2:
23221 @item -mclear-hwcap
23222 @opindex mclear-hwcap
23223 @option{-mclear-hwcap} tells the compiler to remove the hardware
23224 capabilities generated by the Solaris assembler. This is only necessary
23225 when object files use ISA extensions not supported by the current
23226 machine, but check at runtime whether or not to use them.
23228 @item -mimpure-text
23229 @opindex mimpure-text
23230 @option{-mimpure-text}, used in addition to @option{-shared}, tells
23231 the compiler to not pass @option{-z text} to the linker when linking a
23232 shared object. Using this option, you can link position-dependent
23233 code into a shared object.
23235 @option{-mimpure-text} suppresses the ``relocations remain against
23236 allocatable but non-writable sections'' linker error message.
23237 However, the necessary relocations trigger copy-on-write, and the
23238 shared object is not actually shared across processes. Instead of
23239 using @option{-mimpure-text}, you should compile all source code with
23240 @option{-fpic} or @option{-fPIC}.
23244 These switches are supported in addition to the above on Solaris 2:
23249 This is a synonym for @option{-pthread}.
23252 @node SPARC Options
23253 @subsection SPARC Options
23254 @cindex SPARC options
23256 These @samp{-m} options are supported on the SPARC:
23259 @item -mno-app-regs
23261 @opindex mno-app-regs
23263 Specify @option{-mapp-regs} to generate output using the global registers
23264 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
23265 global register 1, each global register 2 through 4 is then treated as an
23266 allocable register that is clobbered by function calls. This is the default.
23268 To be fully SVR4 ABI-compliant at the cost of some performance loss,
23269 specify @option{-mno-app-regs}. You should compile libraries and system
23270 software with this option.
23276 With @option{-mflat}, the compiler does not generate save/restore instructions
23277 and uses a ``flat'' or single register window model. This model is compatible
23278 with the regular register window model. The local registers and the input
23279 registers (0--5) are still treated as ``call-saved'' registers and are
23280 saved on the stack as needed.
23282 With @option{-mno-flat} (the default), the compiler generates save/restore
23283 instructions (except for leaf functions). This is the normal operating mode.
23286 @itemx -mhard-float
23288 @opindex mhard-float
23289 Generate output containing floating-point instructions. This is the
23293 @itemx -msoft-float
23295 @opindex msoft-float
23296 Generate output containing library calls for floating point.
23297 @strong{Warning:} the requisite libraries are not available for all SPARC
23298 targets. Normally the facilities of the machine's usual C compiler are
23299 used, but this cannot be done directly in cross-compilation. You must make
23300 your own arrangements to provide suitable library functions for
23301 cross-compilation. The embedded targets @samp{sparc-*-aout} and
23302 @samp{sparclite-*-*} do provide software floating-point support.
23304 @option{-msoft-float} changes the calling convention in the output file;
23305 therefore, it is only useful if you compile @emph{all} of a program with
23306 this option. In particular, you need to compile @file{libgcc.a}, the
23307 library that comes with GCC, with @option{-msoft-float} in order for
23310 @item -mhard-quad-float
23311 @opindex mhard-quad-float
23312 Generate output containing quad-word (long double) floating-point
23315 @item -msoft-quad-float
23316 @opindex msoft-quad-float
23317 Generate output containing library calls for quad-word (long double)
23318 floating-point instructions. The functions called are those specified
23319 in the SPARC ABI@. This is the default.
23321 As of this writing, there are no SPARC implementations that have hardware
23322 support for the quad-word floating-point instructions. They all invoke
23323 a trap handler for one of these instructions, and then the trap handler
23324 emulates the effect of the instruction. Because of the trap handler overhead,
23325 this is much slower than calling the ABI library routines. Thus the
23326 @option{-msoft-quad-float} option is the default.
23328 @item -mno-unaligned-doubles
23329 @itemx -munaligned-doubles
23330 @opindex mno-unaligned-doubles
23331 @opindex munaligned-doubles
23332 Assume that doubles have 8-byte alignment. This is the default.
23334 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
23335 alignment only if they are contained in another type, or if they have an
23336 absolute address. Otherwise, it assumes they have 4-byte alignment.
23337 Specifying this option avoids some rare compatibility problems with code
23338 generated by other compilers. It is not the default because it results
23339 in a performance loss, especially for floating-point code.
23342 @itemx -mno-user-mode
23343 @opindex muser-mode
23344 @opindex mno-user-mode
23345 Do not generate code that can only run in supervisor mode. This is relevant
23346 only for the @code{casa} instruction emitted for the LEON3 processor. This
23349 @item -mfaster-structs
23350 @itemx -mno-faster-structs
23351 @opindex mfaster-structs
23352 @opindex mno-faster-structs
23353 With @option{-mfaster-structs}, the compiler assumes that structures
23354 should have 8-byte alignment. This enables the use of pairs of
23355 @code{ldd} and @code{std} instructions for copies in structure
23356 assignment, in place of twice as many @code{ld} and @code{st} pairs.
23357 However, the use of this changed alignment directly violates the SPARC
23358 ABI@. Thus, it's intended only for use on targets where the developer
23359 acknowledges that their resulting code is not directly in line with
23360 the rules of the ABI@.
23362 @item -mstd-struct-return
23363 @itemx -mno-std-struct-return
23364 @opindex mstd-struct-return
23365 @opindex mno-std-struct-return
23366 With @option{-mstd-struct-return}, the compiler generates checking code
23367 in functions returning structures or unions to detect size mismatches
23368 between the two sides of function calls, as per the 32-bit ABI@.
23370 The default is @option{-mno-std-struct-return}. This option has no effect
23377 Enable Local Register Allocation. This is the default for SPARC since GCC 7
23378 so @option{-mno-lra} needs to be passed to get old Reload.
23380 @item -mcpu=@var{cpu_type}
23382 Set the instruction set, register set, and instruction scheduling parameters
23383 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23384 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
23385 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
23386 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
23387 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
23388 @samp{niagara3}, @samp{niagara4} and @samp{niagara7}.
23390 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
23391 which selects the best architecture option for the host processor.
23392 @option{-mcpu=native} has no effect if GCC does not recognize
23395 Default instruction scheduling parameters are used for values that select
23396 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
23397 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
23399 Here is a list of each supported architecture and their supported
23407 supersparc, hypersparc, leon, leon3
23410 f930, f934, sparclite86x
23416 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7
23419 By default (unless configured otherwise), GCC generates code for the V7
23420 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
23421 additionally optimizes it for the Cypress CY7C602 chip, as used in the
23422 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
23423 SPARCStation 1, 2, IPX etc.
23425 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
23426 architecture. The only difference from V7 code is that the compiler emits
23427 the integer multiply and integer divide instructions which exist in SPARC-V8
23428 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
23429 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
23432 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
23433 the SPARC architecture. This adds the integer multiply, integer divide step
23434 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
23435 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
23436 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
23437 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
23438 MB86934 chip, which is the more recent SPARClite with FPU@.
23440 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
23441 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
23442 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
23443 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
23444 optimizes it for the TEMIC SPARClet chip.
23446 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
23447 architecture. This adds 64-bit integer and floating-point move instructions,
23448 3 additional floating-point condition code registers and conditional move
23449 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
23450 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
23451 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
23452 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
23453 @option{-mcpu=niagara}, the compiler additionally optimizes it for
23454 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
23455 additionally optimizes it for Sun UltraSPARC T2 chips. With
23456 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
23457 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
23458 additionally optimizes it for Sun UltraSPARC T4 chips. With
23459 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
23460 Oracle SPARC M7 chips.
23462 @item -mtune=@var{cpu_type}
23464 Set the instruction scheduling parameters for machine type
23465 @var{cpu_type}, but do not set the instruction set or register set that the
23466 option @option{-mcpu=@var{cpu_type}} does.
23468 The same values for @option{-mcpu=@var{cpu_type}} can be used for
23469 @option{-mtune=@var{cpu_type}}, but the only useful values are those
23470 that select a particular CPU implementation. Those are
23471 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
23472 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
23473 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
23474 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
23475 @samp{niagara4} and @samp{niagara7}. With native Solaris and
23476 GNU/Linux toolchains, @samp{native} can also be used.
23481 @opindex mno-v8plus
23482 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
23483 difference from the V8 ABI is that the global and out registers are
23484 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
23485 mode for all SPARC-V9 processors.
23491 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
23492 Visual Instruction Set extensions. The default is @option{-mno-vis}.
23498 With @option{-mvis2}, GCC generates code that takes advantage of
23499 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
23500 default is @option{-mvis2} when targeting a cpu that supports such
23501 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
23502 also sets @option{-mvis}.
23508 With @option{-mvis3}, GCC generates code that takes advantage of
23509 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
23510 default is @option{-mvis3} when targeting a cpu that supports such
23511 instructions, such as niagara-3 and later. Setting @option{-mvis3}
23512 also sets @option{-mvis2} and @option{-mvis}.
23518 With @option{-mvis4}, GCC generates code that takes advantage of
23519 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
23520 default is @option{-mvis4} when targeting a cpu that supports such
23521 instructions, such as niagara-7 and later. Setting @option{-mvis4}
23522 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
23527 @opindex mno-cbcond
23528 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
23529 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
23530 when targeting a CPU that supports such instructions, such as Niagara-4 and
23537 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
23538 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
23539 when targeting a CPU that supports such instructions, such as Niagara-3 and
23546 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
23547 Population Count instruction. The default is @option{-mpopc}
23548 when targeting a CPU that supports such an instruction, such as Niagara-2 and
23555 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
23556 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
23557 when targeting a CPU that supports such an instruction, such as Niagara-7 and
23561 @opindex mfix-at697f
23562 Enable the documented workaround for the single erratum of the Atmel AT697F
23563 processor (which corresponds to erratum #13 of the AT697E processor).
23566 @opindex mfix-ut699
23567 Enable the documented workarounds for the floating-point errata and the data
23568 cache nullify errata of the UT699 processor.
23571 These @samp{-m} options are supported in addition to the above
23572 on SPARC-V9 processors in 64-bit environments:
23579 Generate code for a 32-bit or 64-bit environment.
23580 The 32-bit environment sets int, long and pointer to 32 bits.
23581 The 64-bit environment sets int to 32 bits and long and pointer
23584 @item -mcmodel=@var{which}
23586 Set the code model to one of
23590 The Medium/Low code model: 64-bit addresses, programs
23591 must be linked in the low 32 bits of memory. Programs can be statically
23592 or dynamically linked.
23595 The Medium/Middle code model: 64-bit addresses, programs
23596 must be linked in the low 44 bits of memory, the text and data segments must
23597 be less than 2GB in size and the data segment must be located within 2GB of
23601 The Medium/Anywhere code model: 64-bit addresses, programs
23602 may be linked anywhere in memory, the text and data segments must be less
23603 than 2GB in size and the data segment must be located within 2GB of the
23607 The Medium/Anywhere code model for embedded systems:
23608 64-bit addresses, the text and data segments must be less than 2GB in
23609 size, both starting anywhere in memory (determined at link time). The
23610 global register %g4 points to the base of the data segment. Programs
23611 are statically linked and PIC is not supported.
23614 @item -mmemory-model=@var{mem-model}
23615 @opindex mmemory-model
23616 Set the memory model in force on the processor to one of
23620 The default memory model for the processor and operating system.
23623 Relaxed Memory Order
23626 Partial Store Order
23632 Sequential Consistency
23635 These memory models are formally defined in Appendix D of the SPARC-V9
23636 architecture manual, as set in the processor's @code{PSTATE.MM} field.
23639 @itemx -mno-stack-bias
23640 @opindex mstack-bias
23641 @opindex mno-stack-bias
23642 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
23643 frame pointer if present, are offset by @minus{}2047 which must be added back
23644 when making stack frame references. This is the default in 64-bit mode.
23645 Otherwise, assume no such offset is present.
23649 @subsection SPU Options
23650 @cindex SPU options
23652 These @samp{-m} options are supported on the SPU:
23656 @itemx -merror-reloc
23657 @opindex mwarn-reloc
23658 @opindex merror-reloc
23660 The loader for SPU does not handle dynamic relocations. By default, GCC
23661 gives an error when it generates code that requires a dynamic
23662 relocation. @option{-mno-error-reloc} disables the error,
23663 @option{-mwarn-reloc} generates a warning instead.
23666 @itemx -munsafe-dma
23668 @opindex munsafe-dma
23670 Instructions that initiate or test completion of DMA must not be
23671 reordered with respect to loads and stores of the memory that is being
23673 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
23674 memory accesses, but that can lead to inefficient code in places where the
23675 memory is known to not change. Rather than mark the memory as volatile,
23676 you can use @option{-msafe-dma} to tell the compiler to treat
23677 the DMA instructions as potentially affecting all memory.
23679 @item -mbranch-hints
23680 @opindex mbranch-hints
23682 By default, GCC generates a branch hint instruction to avoid
23683 pipeline stalls for always-taken or probably-taken branches. A hint
23684 is not generated closer than 8 instructions away from its branch.
23685 There is little reason to disable them, except for debugging purposes,
23686 or to make an object a little bit smaller.
23690 @opindex msmall-mem
23691 @opindex mlarge-mem
23693 By default, GCC generates code assuming that addresses are never larger
23694 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
23695 a full 32-bit address.
23700 By default, GCC links against startup code that assumes the SPU-style
23701 main function interface (which has an unconventional parameter list).
23702 With @option{-mstdmain}, GCC links your program against startup
23703 code that assumes a C99-style interface to @code{main}, including a
23704 local copy of @code{argv} strings.
23706 @item -mfixed-range=@var{register-range}
23707 @opindex mfixed-range
23708 Generate code treating the given register range as fixed registers.
23709 A fixed register is one that the register allocator cannot use. This is
23710 useful when compiling kernel code. A register range is specified as
23711 two registers separated by a dash. Multiple register ranges can be
23712 specified separated by a comma.
23718 Compile code assuming that pointers to the PPU address space accessed
23719 via the @code{__ea} named address space qualifier are either 32 or 64
23720 bits wide. The default is 32 bits. As this is an ABI-changing option,
23721 all object code in an executable must be compiled with the same setting.
23723 @item -maddress-space-conversion
23724 @itemx -mno-address-space-conversion
23725 @opindex maddress-space-conversion
23726 @opindex mno-address-space-conversion
23727 Allow/disallow treating the @code{__ea} address space as superset
23728 of the generic address space. This enables explicit type casts
23729 between @code{__ea} and generic pointer as well as implicit
23730 conversions of generic pointers to @code{__ea} pointers. The
23731 default is to allow address space pointer conversions.
23733 @item -mcache-size=@var{cache-size}
23734 @opindex mcache-size
23735 This option controls the version of libgcc that the compiler links to an
23736 executable and selects a software-managed cache for accessing variables
23737 in the @code{__ea} address space with a particular cache size. Possible
23738 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
23739 and @samp{128}. The default cache size is 64KB.
23741 @item -matomic-updates
23742 @itemx -mno-atomic-updates
23743 @opindex matomic-updates
23744 @opindex mno-atomic-updates
23745 This option controls the version of libgcc that the compiler links to an
23746 executable and selects whether atomic updates to the software-managed
23747 cache of PPU-side variables are used. If you use atomic updates, changes
23748 to a PPU variable from SPU code using the @code{__ea} named address space
23749 qualifier do not interfere with changes to other PPU variables residing
23750 in the same cache line from PPU code. If you do not use atomic updates,
23751 such interference may occur; however, writing back cache lines is
23752 more efficient. The default behavior is to use atomic updates.
23755 @itemx -mdual-nops=@var{n}
23756 @opindex mdual-nops
23757 By default, GCC inserts NOPs to increase dual issue when it expects
23758 it to increase performance. @var{n} can be a value from 0 to 10. A
23759 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
23760 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
23762 @item -mhint-max-nops=@var{n}
23763 @opindex mhint-max-nops
23764 Maximum number of NOPs to insert for a branch hint. A branch hint must
23765 be at least 8 instructions away from the branch it is affecting. GCC
23766 inserts up to @var{n} NOPs to enforce this, otherwise it does not
23767 generate the branch hint.
23769 @item -mhint-max-distance=@var{n}
23770 @opindex mhint-max-distance
23771 The encoding of the branch hint instruction limits the hint to be within
23772 256 instructions of the branch it is affecting. By default, GCC makes
23773 sure it is within 125.
23776 @opindex msafe-hints
23777 Work around a hardware bug that causes the SPU to stall indefinitely.
23778 By default, GCC inserts the @code{hbrp} instruction to make sure
23779 this stall won't happen.
23783 @node System V Options
23784 @subsection Options for System V
23786 These additional options are available on System V Release 4 for
23787 compatibility with other compilers on those systems:
23792 Create a shared object.
23793 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
23797 Identify the versions of each tool used by the compiler, in a
23798 @code{.ident} assembler directive in the output.
23802 Refrain from adding @code{.ident} directives to the output file (this is
23805 @item -YP,@var{dirs}
23807 Search the directories @var{dirs}, and no others, for libraries
23808 specified with @option{-l}.
23810 @item -Ym,@var{dir}
23812 Look in the directory @var{dir} to find the M4 preprocessor.
23813 The assembler uses this option.
23814 @c This is supposed to go with a -Yd for predefined M4 macro files, but
23815 @c the generic assembler that comes with Solaris takes just -Ym.
23818 @node TILE-Gx Options
23819 @subsection TILE-Gx Options
23820 @cindex TILE-Gx options
23822 These @samp{-m} options are supported on the TILE-Gx:
23825 @item -mcmodel=small
23826 @opindex mcmodel=small
23827 Generate code for the small model. The distance for direct calls is
23828 limited to 500M in either direction. PC-relative addresses are 32
23829 bits. Absolute addresses support the full address range.
23831 @item -mcmodel=large
23832 @opindex mcmodel=large
23833 Generate code for the large model. There is no limitation on call
23834 distance, pc-relative addresses, or absolute addresses.
23836 @item -mcpu=@var{name}
23838 Selects the type of CPU to be targeted. Currently the only supported
23839 type is @samp{tilegx}.
23845 Generate code for a 32-bit or 64-bit environment. The 32-bit
23846 environment sets int, long, and pointer to 32 bits. The 64-bit
23847 environment sets int to 32 bits and long and pointer to 64 bits.
23850 @itemx -mlittle-endian
23851 @opindex mbig-endian
23852 @opindex mlittle-endian
23853 Generate code in big/little endian mode, respectively.
23856 @node TILEPro Options
23857 @subsection TILEPro Options
23858 @cindex TILEPro options
23860 These @samp{-m} options are supported on the TILEPro:
23863 @item -mcpu=@var{name}
23865 Selects the type of CPU to be targeted. Currently the only supported
23866 type is @samp{tilepro}.
23870 Generate code for a 32-bit environment, which sets int, long, and
23871 pointer to 32 bits. This is the only supported behavior so the flag
23872 is essentially ignored.
23876 @subsection V850 Options
23877 @cindex V850 Options
23879 These @samp{-m} options are defined for V850 implementations:
23883 @itemx -mno-long-calls
23884 @opindex mlong-calls
23885 @opindex mno-long-calls
23886 Treat all calls as being far away (near). If calls are assumed to be
23887 far away, the compiler always loads the function's address into a
23888 register, and calls indirect through the pointer.
23894 Do not optimize (do optimize) basic blocks that use the same index
23895 pointer 4 or more times to copy pointer into the @code{ep} register, and
23896 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
23897 option is on by default if you optimize.
23899 @item -mno-prolog-function
23900 @itemx -mprolog-function
23901 @opindex mno-prolog-function
23902 @opindex mprolog-function
23903 Do not use (do use) external functions to save and restore registers
23904 at the prologue and epilogue of a function. The external functions
23905 are slower, but use less code space if more than one function saves
23906 the same number of registers. The @option{-mprolog-function} option
23907 is on by default if you optimize.
23911 Try to make the code as small as possible. At present, this just turns
23912 on the @option{-mep} and @option{-mprolog-function} options.
23914 @item -mtda=@var{n}
23916 Put static or global variables whose size is @var{n} bytes or less into
23917 the tiny data area that register @code{ep} points to. The tiny data
23918 area can hold up to 256 bytes in total (128 bytes for byte references).
23920 @item -msda=@var{n}
23922 Put static or global variables whose size is @var{n} bytes or less into
23923 the small data area that register @code{gp} points to. The small data
23924 area can hold up to 64 kilobytes.
23926 @item -mzda=@var{n}
23928 Put static or global variables whose size is @var{n} bytes or less into
23929 the first 32 kilobytes of memory.
23933 Specify that the target processor is the V850.
23937 Specify that the target processor is the V850E3V5. The preprocessor
23938 constant @code{__v850e3v5__} is defined if this option is used.
23942 Specify that the target processor is the V850E3V5. This is an alias for
23943 the @option{-mv850e3v5} option.
23947 Specify that the target processor is the V850E2V3. The preprocessor
23948 constant @code{__v850e2v3__} is defined if this option is used.
23952 Specify that the target processor is the V850E2. The preprocessor
23953 constant @code{__v850e2__} is defined if this option is used.
23957 Specify that the target processor is the V850E1. The preprocessor
23958 constants @code{__v850e1__} and @code{__v850e__} are defined if
23959 this option is used.
23963 Specify that the target processor is the V850ES. This is an alias for
23964 the @option{-mv850e1} option.
23968 Specify that the target processor is the V850E@. The preprocessor
23969 constant @code{__v850e__} is defined if this option is used.
23971 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
23972 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
23973 are defined then a default target processor is chosen and the
23974 relevant @samp{__v850*__} preprocessor constant is defined.
23976 The preprocessor constants @code{__v850} and @code{__v851__} are always
23977 defined, regardless of which processor variant is the target.
23979 @item -mdisable-callt
23980 @itemx -mno-disable-callt
23981 @opindex mdisable-callt
23982 @opindex mno-disable-callt
23983 This option suppresses generation of the @code{CALLT} instruction for the
23984 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
23987 This option is enabled by default when the RH850 ABI is
23988 in use (see @option{-mrh850-abi}), and disabled by default when the
23989 GCC ABI is in use. If @code{CALLT} instructions are being generated
23990 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
23996 Pass on (or do not pass on) the @option{-mrelax} command-line option
24000 @itemx -mno-long-jumps
24001 @opindex mlong-jumps
24002 @opindex mno-long-jumps
24003 Disable (or re-enable) the generation of PC-relative jump instructions.
24006 @itemx -mhard-float
24007 @opindex msoft-float
24008 @opindex mhard-float
24009 Disable (or re-enable) the generation of hardware floating point
24010 instructions. This option is only significant when the target
24011 architecture is @samp{V850E2V3} or higher. If hardware floating point
24012 instructions are being generated then the C preprocessor symbol
24013 @code{__FPU_OK__} is defined, otherwise the symbol
24014 @code{__NO_FPU__} is defined.
24018 Enables the use of the e3v5 LOOP instruction. The use of this
24019 instruction is not enabled by default when the e3v5 architecture is
24020 selected because its use is still experimental.
24024 @opindex mrh850-abi
24026 Enables support for the RH850 version of the V850 ABI. This is the
24027 default. With this version of the ABI the following rules apply:
24031 Integer sized structures and unions are returned via a memory pointer
24032 rather than a register.
24035 Large structures and unions (more than 8 bytes in size) are passed by
24039 Functions are aligned to 16-bit boundaries.
24042 The @option{-m8byte-align} command-line option is supported.
24045 The @option{-mdisable-callt} command-line option is enabled by
24046 default. The @option{-mno-disable-callt} command-line option is not
24050 When this version of the ABI is enabled the C preprocessor symbol
24051 @code{__V850_RH850_ABI__} is defined.
24055 Enables support for the old GCC version of the V850 ABI. With this
24056 version of the ABI the following rules apply:
24060 Integer sized structures and unions are returned in register @code{r10}.
24063 Large structures and unions (more than 8 bytes in size) are passed by
24067 Functions are aligned to 32-bit boundaries, unless optimizing for
24071 The @option{-m8byte-align} command-line option is not supported.
24074 The @option{-mdisable-callt} command-line option is supported but not
24075 enabled by default.
24078 When this version of the ABI is enabled the C preprocessor symbol
24079 @code{__V850_GCC_ABI__} is defined.
24081 @item -m8byte-align
24082 @itemx -mno-8byte-align
24083 @opindex m8byte-align
24084 @opindex mno-8byte-align
24085 Enables support for @code{double} and @code{long long} types to be
24086 aligned on 8-byte boundaries. The default is to restrict the
24087 alignment of all objects to at most 4-bytes. When
24088 @option{-m8byte-align} is in effect the C preprocessor symbol
24089 @code{__V850_8BYTE_ALIGN__} is defined.
24092 @opindex mbig-switch
24093 Generate code suitable for big switch tables. Use this option only if
24094 the assembler/linker complain about out of range branches within a switch
24099 This option causes r2 and r5 to be used in the code generated by
24100 the compiler. This setting is the default.
24102 @item -mno-app-regs
24103 @opindex mno-app-regs
24104 This option causes r2 and r5 to be treated as fixed registers.
24109 @subsection VAX Options
24110 @cindex VAX options
24112 These @samp{-m} options are defined for the VAX:
24117 Do not output certain jump instructions (@code{aobleq} and so on)
24118 that the Unix assembler for the VAX cannot handle across long
24123 Do output those jump instructions, on the assumption that the
24124 GNU assembler is being used.
24128 Output code for G-format floating-point numbers instead of D-format.
24131 @node Visium Options
24132 @subsection Visium Options
24133 @cindex Visium options
24139 A program which performs file I/O and is destined to run on an MCM target
24140 should be linked with this option. It causes the libraries libc.a and
24141 libdebug.a to be linked. The program should be run on the target under
24142 the control of the GDB remote debugging stub.
24146 A program which performs file I/O and is destined to run on the simulator
24147 should be linked with option. This causes libraries libc.a and libsim.a to
24151 @itemx -mhard-float
24153 @opindex mhard-float
24154 Generate code containing floating-point instructions. This is the
24158 @itemx -msoft-float
24160 @opindex msoft-float
24161 Generate code containing library calls for floating-point.
24163 @option{-msoft-float} changes the calling convention in the output file;
24164 therefore, it is only useful if you compile @emph{all} of a program with
24165 this option. In particular, you need to compile @file{libgcc.a}, the
24166 library that comes with GCC, with @option{-msoft-float} in order for
24169 @item -mcpu=@var{cpu_type}
24171 Set the instruction set, register set, and instruction scheduling parameters
24172 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24173 @samp{mcm}, @samp{gr5} and @samp{gr6}.
24175 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
24177 By default (unless configured otherwise), GCC generates code for the GR5
24178 variant of the Visium architecture.
24180 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
24181 architecture. The only difference from GR5 code is that the compiler will
24182 generate block move instructions.
24184 @item -mtune=@var{cpu_type}
24186 Set the instruction scheduling parameters for machine type @var{cpu_type},
24187 but do not set the instruction set or register set that the option
24188 @option{-mcpu=@var{cpu_type}} would.
24192 Generate code for the supervisor mode, where there are no restrictions on
24193 the access to general registers. This is the default.
24196 @opindex muser-mode
24197 Generate code for the user mode, where the access to some general registers
24198 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
24199 mode; on the GR6, only registers r29 to r31 are affected.
24203 @subsection VMS Options
24205 These @samp{-m} options are defined for the VMS implementations:
24208 @item -mvms-return-codes
24209 @opindex mvms-return-codes
24210 Return VMS condition codes from @code{main}. The default is to return POSIX-style
24211 condition (e.g.@ error) codes.
24213 @item -mdebug-main=@var{prefix}
24214 @opindex mdebug-main=@var{prefix}
24215 Flag the first routine whose name starts with @var{prefix} as the main
24216 routine for the debugger.
24220 Default to 64-bit memory allocation routines.
24222 @item -mpointer-size=@var{size}
24223 @opindex mpointer-size=@var{size}
24224 Set the default size of pointers. Possible options for @var{size} are
24225 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
24226 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
24227 The later option disables @code{pragma pointer_size}.
24230 @node VxWorks Options
24231 @subsection VxWorks Options
24232 @cindex VxWorks Options
24234 The options in this section are defined for all VxWorks targets.
24235 Options specific to the target hardware are listed with the other
24236 options for that target.
24241 GCC can generate code for both VxWorks kernels and real time processes
24242 (RTPs). This option switches from the former to the latter. It also
24243 defines the preprocessor macro @code{__RTP__}.
24246 @opindex non-static
24247 Link an RTP executable against shared libraries rather than static
24248 libraries. The options @option{-static} and @option{-shared} can
24249 also be used for RTPs (@pxref{Link Options}); @option{-static}
24256 These options are passed down to the linker. They are defined for
24257 compatibility with Diab.
24260 @opindex Xbind-lazy
24261 Enable lazy binding of function calls. This option is equivalent to
24262 @option{-Wl,-z,now} and is defined for compatibility with Diab.
24266 Disable lazy binding of function calls. This option is the default and
24267 is defined for compatibility with Diab.
24271 @subsection x86 Options
24272 @cindex x86 Options
24274 These @samp{-m} options are defined for the x86 family of computers.
24278 @item -march=@var{cpu-type}
24280 Generate instructions for the machine type @var{cpu-type}. In contrast to
24281 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
24282 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
24283 to generate code that may not run at all on processors other than the one
24284 indicated. Specifying @option{-march=@var{cpu-type}} implies
24285 @option{-mtune=@var{cpu-type}}.
24287 The choices for @var{cpu-type} are:
24291 This selects the CPU to generate code for at compilation time by determining
24292 the processor type of the compiling machine. Using @option{-march=native}
24293 enables all instruction subsets supported by the local machine (hence
24294 the result might not run on different machines). Using @option{-mtune=native}
24295 produces code optimized for the local machine under the constraints
24296 of the selected instruction set.
24299 Original Intel i386 CPU@.
24302 Intel i486 CPU@. (No scheduling is implemented for this chip.)
24306 Intel Pentium CPU with no MMX support.
24309 Intel Lakemont MCU, based on Intel Pentium CPU.
24312 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
24315 Intel Pentium Pro CPU@.
24318 When used with @option{-march}, the Pentium Pro
24319 instruction set is used, so the code runs on all i686 family chips.
24320 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
24323 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
24328 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
24332 Intel Pentium M; low-power version of Intel Pentium III CPU
24333 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
24337 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
24340 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
24344 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
24345 SSE2 and SSE3 instruction set support.
24348 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
24349 instruction set support.
24352 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24353 SSE4.1, SSE4.2 and POPCNT instruction set support.
24356 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24357 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
24360 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24361 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
24364 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24365 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
24366 instruction set support.
24369 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24370 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24371 BMI, BMI2 and F16C instruction set support.
24374 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24375 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24376 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
24379 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24380 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24381 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
24382 XSAVES instruction set support.
24385 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
24386 instruction set support.
24389 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24390 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
24393 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24394 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24395 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
24396 AVX512CD instruction set support.
24398 @item skylake-avx512
24399 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24400 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24401 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
24402 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
24405 AMD K6 CPU with MMX instruction set support.
24409 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
24412 @itemx athlon-tbird
24413 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
24419 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
24420 instruction set support.
24426 Processors based on the AMD K8 core with x86-64 instruction set support,
24427 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
24428 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
24429 instruction set extensions.)
24432 @itemx opteron-sse3
24433 @itemx athlon64-sse3
24434 Improved versions of AMD K8 cores with SSE3 instruction set support.
24438 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
24439 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
24440 instruction set extensions.)
24443 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
24444 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
24445 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
24447 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24448 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
24449 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
24452 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24453 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
24454 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
24455 64-bit instruction set extensions.
24457 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24458 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
24459 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
24460 SSE4.2, ABM and 64-bit instruction set extensions.
24463 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
24464 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
24465 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
24466 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
24467 instruction set extensions.
24470 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
24471 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
24472 instruction set extensions.)
24475 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
24476 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
24477 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
24480 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
24484 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
24485 instruction set support.
24488 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
24489 (No scheduling is implemented for this chip.)
24492 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
24493 (No scheduling is implemented for this chip.)
24496 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24497 (No scheduling is implemented for this chip.)
24500 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
24501 (No scheduling is implemented for this chip.)
24504 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
24505 (No scheduling is implemented for this chip.)
24508 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24509 (No scheduling is implemented for this chip.)
24512 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
24513 (No scheduling is implemented for this chip.)
24516 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
24517 AVX and AVX2 instruction set support.
24518 (No scheduling is implemented for this chip.)
24521 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24522 instruction set support.
24523 (No scheduling is implemented for this chip.)
24526 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24527 instruction set support.
24528 (No scheduling is implemented for this chip.)
24531 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24532 instruction set support.
24533 (No scheduling is implemented for this chip.)
24536 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24537 instruction set support.
24538 (No scheduling is implemented for this chip.)
24541 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24542 instruction set support.
24543 (No scheduling is implemented for this chip.)
24546 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24547 instruction set support.
24548 (No scheduling is implemented for this chip.)
24551 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
24554 @item -mtune=@var{cpu-type}
24556 Tune to @var{cpu-type} everything applicable about the generated code, except
24557 for the ABI and the set of available instructions.
24558 While picking a specific @var{cpu-type} schedules things appropriately
24559 for that particular chip, the compiler does not generate any code that
24560 cannot run on the default machine type unless you use a
24561 @option{-march=@var{cpu-type}} option.
24562 For example, if GCC is configured for i686-pc-linux-gnu
24563 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
24564 but still runs on i686 machines.
24566 The choices for @var{cpu-type} are the same as for @option{-march}.
24567 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
24571 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
24572 If you know the CPU on which your code will run, then you should use
24573 the corresponding @option{-mtune} or @option{-march} option instead of
24574 @option{-mtune=generic}. But, if you do not know exactly what CPU users
24575 of your application will have, then you should use this option.
24577 As new processors are deployed in the marketplace, the behavior of this
24578 option will change. Therefore, if you upgrade to a newer version of
24579 GCC, code generation controlled by this option will change to reflect
24581 that are most common at the time that version of GCC is released.
24583 There is no @option{-march=generic} option because @option{-march}
24584 indicates the instruction set the compiler can use, and there is no
24585 generic instruction set applicable to all processors. In contrast,
24586 @option{-mtune} indicates the processor (or, in this case, collection of
24587 processors) for which the code is optimized.
24590 Produce code optimized for the most current Intel processors, which are
24591 Haswell and Silvermont for this version of GCC. If you know the CPU
24592 on which your code will run, then you should use the corresponding
24593 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
24594 But, if you want your application performs better on both Haswell and
24595 Silvermont, then you should use this option.
24597 As new Intel processors are deployed in the marketplace, the behavior of
24598 this option will change. Therefore, if you upgrade to a newer version of
24599 GCC, code generation controlled by this option will change to reflect
24600 the most current Intel processors at the time that version of GCC is
24603 There is no @option{-march=intel} option because @option{-march} indicates
24604 the instruction set the compiler can use, and there is no common
24605 instruction set applicable to all processors. In contrast,
24606 @option{-mtune} indicates the processor (or, in this case, collection of
24607 processors) for which the code is optimized.
24610 @item -mcpu=@var{cpu-type}
24612 A deprecated synonym for @option{-mtune}.
24614 @item -mfpmath=@var{unit}
24616 Generate floating-point arithmetic for selected unit @var{unit}. The choices
24617 for @var{unit} are:
24621 Use the standard 387 floating-point coprocessor present on the majority of chips and
24622 emulated otherwise. Code compiled with this option runs almost everywhere.
24623 The temporary results are computed in 80-bit precision instead of the precision
24624 specified by the type, resulting in slightly different results compared to most
24625 of other chips. See @option{-ffloat-store} for more detailed description.
24627 This is the default choice for non-Darwin x86-32 targets.
24630 Use scalar floating-point instructions present in the SSE instruction set.
24631 This instruction set is supported by Pentium III and newer chips,
24632 and in the AMD line
24633 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
24634 instruction set supports only single-precision arithmetic, thus the double and
24635 extended-precision arithmetic are still done using 387. A later version, present
24636 only in Pentium 4 and AMD x86-64 chips, supports double-precision
24639 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
24640 or @option{-msse2} switches to enable SSE extensions and make this option
24641 effective. For the x86-64 compiler, these extensions are enabled by default.
24643 The resulting code should be considerably faster in the majority of cases and avoid
24644 the numerical instability problems of 387 code, but may break some existing
24645 code that expects temporaries to be 80 bits.
24647 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
24648 and the default choice for x86-32 targets with the SSE2 instruction set
24649 when @option{-ffast-math} is enabled.
24654 Attempt to utilize both instruction sets at once. This effectively doubles the
24655 amount of available registers, and on chips with separate execution units for
24656 387 and SSE the execution resources too. Use this option with care, as it is
24657 still experimental, because the GCC register allocator does not model separate
24658 functional units well, resulting in unstable performance.
24661 @item -masm=@var{dialect}
24662 @opindex masm=@var{dialect}
24663 Output assembly instructions using selected @var{dialect}. Also affects
24664 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
24665 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
24666 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
24667 not support @samp{intel}.
24670 @itemx -mno-ieee-fp
24672 @opindex mno-ieee-fp
24673 Control whether or not the compiler uses IEEE floating-point
24674 comparisons. These correctly handle the case where the result of a
24675 comparison is unordered.
24680 @opindex mhard-float
24681 Generate output containing 80387 instructions for floating point.
24686 @opindex msoft-float
24687 Generate output containing library calls for floating point.
24689 @strong{Warning:} the requisite libraries are not part of GCC@.
24690 Normally the facilities of the machine's usual C compiler are used, but
24691 this cannot be done directly in cross-compilation. You must make your
24692 own arrangements to provide suitable library functions for
24695 On machines where a function returns floating-point results in the 80387
24696 register stack, some floating-point opcodes may be emitted even if
24697 @option{-msoft-float} is used.
24699 @item -mno-fp-ret-in-387
24700 @opindex mno-fp-ret-in-387
24701 Do not use the FPU registers for return values of functions.
24703 The usual calling convention has functions return values of types
24704 @code{float} and @code{double} in an FPU register, even if there
24705 is no FPU@. The idea is that the operating system should emulate
24708 The option @option{-mno-fp-ret-in-387} causes such values to be returned
24709 in ordinary CPU registers instead.
24711 @item -mno-fancy-math-387
24712 @opindex mno-fancy-math-387
24713 Some 387 emulators do not support the @code{sin}, @code{cos} and
24714 @code{sqrt} instructions for the 387. Specify this option to avoid
24715 generating those instructions. This option is the default on
24716 OpenBSD and NetBSD@. This option is overridden when @option{-march}
24717 indicates that the target CPU always has an FPU and so the
24718 instruction does not need emulation. These
24719 instructions are not generated unless you also use the
24720 @option{-funsafe-math-optimizations} switch.
24722 @item -malign-double
24723 @itemx -mno-align-double
24724 @opindex malign-double
24725 @opindex mno-align-double
24726 Control whether GCC aligns @code{double}, @code{long double}, and
24727 @code{long long} variables on a two-word boundary or a one-word
24728 boundary. Aligning @code{double} variables on a two-word boundary
24729 produces code that runs somewhat faster on a Pentium at the
24730 expense of more memory.
24732 On x86-64, @option{-malign-double} is enabled by default.
24734 @strong{Warning:} if you use the @option{-malign-double} switch,
24735 structures containing the above types are aligned differently than
24736 the published application binary interface specifications for the x86-32
24737 and are not binary compatible with structures in code compiled
24738 without that switch.
24740 @item -m96bit-long-double
24741 @itemx -m128bit-long-double
24742 @opindex m96bit-long-double
24743 @opindex m128bit-long-double
24744 These switches control the size of @code{long double} type. The x86-32
24745 application binary interface specifies the size to be 96 bits,
24746 so @option{-m96bit-long-double} is the default in 32-bit mode.
24748 Modern architectures (Pentium and newer) prefer @code{long double}
24749 to be aligned to an 8- or 16-byte boundary. In arrays or structures
24750 conforming to the ABI, this is not possible. So specifying
24751 @option{-m128bit-long-double} aligns @code{long double}
24752 to a 16-byte boundary by padding the @code{long double} with an additional
24755 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
24756 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
24758 Notice that neither of these options enable any extra precision over the x87
24759 standard of 80 bits for a @code{long double}.
24761 @strong{Warning:} if you override the default value for your target ABI, this
24762 changes the size of
24763 structures and arrays containing @code{long double} variables,
24764 as well as modifying the function calling convention for functions taking
24765 @code{long double}. Hence they are not binary-compatible
24766 with code compiled without that switch.
24768 @item -mlong-double-64
24769 @itemx -mlong-double-80
24770 @itemx -mlong-double-128
24771 @opindex mlong-double-64
24772 @opindex mlong-double-80
24773 @opindex mlong-double-128
24774 These switches control the size of @code{long double} type. A size
24775 of 64 bits makes the @code{long double} type equivalent to the @code{double}
24776 type. This is the default for 32-bit Bionic C library. A size
24777 of 128 bits makes the @code{long double} type equivalent to the
24778 @code{__float128} type. This is the default for 64-bit Bionic C library.
24780 @strong{Warning:} if you override the default value for your target ABI, this
24781 changes the size of
24782 structures and arrays containing @code{long double} variables,
24783 as well as modifying the function calling convention for functions taking
24784 @code{long double}. Hence they are not binary-compatible
24785 with code compiled without that switch.
24787 @item -malign-data=@var{type}
24788 @opindex malign-data
24789 Control how GCC aligns variables. Supported values for @var{type} are
24790 @samp{compat} uses increased alignment value compatible uses GCC 4.8
24791 and earlier, @samp{abi} uses alignment value as specified by the
24792 psABI, and @samp{cacheline} uses increased alignment value to match
24793 the cache line size. @samp{compat} is the default.
24795 @item -mlarge-data-threshold=@var{threshold}
24796 @opindex mlarge-data-threshold
24797 When @option{-mcmodel=medium} is specified, data objects larger than
24798 @var{threshold} are placed in the large data section. This value must be the
24799 same across all objects linked into the binary, and defaults to 65535.
24803 Use a different function-calling convention, in which functions that
24804 take a fixed number of arguments return with the @code{ret @var{num}}
24805 instruction, which pops their arguments while returning. This saves one
24806 instruction in the caller since there is no need to pop the arguments
24809 You can specify that an individual function is called with this calling
24810 sequence with the function attribute @code{stdcall}. You can also
24811 override the @option{-mrtd} option by using the function attribute
24812 @code{cdecl}. @xref{Function Attributes}.
24814 @strong{Warning:} this calling convention is incompatible with the one
24815 normally used on Unix, so you cannot use it if you need to call
24816 libraries compiled with the Unix compiler.
24818 Also, you must provide function prototypes for all functions that
24819 take variable numbers of arguments (including @code{printf});
24820 otherwise incorrect code is generated for calls to those
24823 In addition, seriously incorrect code results if you call a
24824 function with too many arguments. (Normally, extra arguments are
24825 harmlessly ignored.)
24827 @item -mregparm=@var{num}
24829 Control how many registers are used to pass integer arguments. By
24830 default, no registers are used to pass arguments, and at most 3
24831 registers can be used. You can control this behavior for a specific
24832 function by using the function attribute @code{regparm}.
24833 @xref{Function Attributes}.
24835 @strong{Warning:} if you use this switch, and
24836 @var{num} is nonzero, then you must build all modules with the same
24837 value, including any libraries. This includes the system libraries and
24841 @opindex msseregparm
24842 Use SSE register passing conventions for float and double arguments
24843 and return values. You can control this behavior for a specific
24844 function by using the function attribute @code{sseregparm}.
24845 @xref{Function Attributes}.
24847 @strong{Warning:} if you use this switch then you must build all
24848 modules with the same value, including any libraries. This includes
24849 the system libraries and startup modules.
24851 @item -mvect8-ret-in-mem
24852 @opindex mvect8-ret-in-mem
24853 Return 8-byte vectors in memory instead of MMX registers. This is the
24854 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
24855 Studio compilers until version 12. Later compiler versions (starting
24856 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
24857 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
24858 you need to remain compatible with existing code produced by those
24859 previous compiler versions or older versions of GCC@.
24868 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
24869 is specified, the significands of results of floating-point operations are
24870 rounded to 24 bits (single precision); @option{-mpc64} rounds the
24871 significands of results of floating-point operations to 53 bits (double
24872 precision) and @option{-mpc80} rounds the significands of results of
24873 floating-point operations to 64 bits (extended double precision), which is
24874 the default. When this option is used, floating-point operations in higher
24875 precisions are not available to the programmer without setting the FPU
24876 control word explicitly.
24878 Setting the rounding of floating-point operations to less than the default
24879 80 bits can speed some programs by 2% or more. Note that some mathematical
24880 libraries assume that extended-precision (80-bit) floating-point operations
24881 are enabled by default; routines in such libraries could suffer significant
24882 loss of accuracy, typically through so-called ``catastrophic cancellation'',
24883 when this option is used to set the precision to less than extended precision.
24885 @item -mstackrealign
24886 @opindex mstackrealign
24887 Realign the stack at entry. On the x86, the @option{-mstackrealign}
24888 option generates an alternate prologue and epilogue that realigns the
24889 run-time stack if necessary. This supports mixing legacy codes that keep
24890 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
24891 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
24892 applicable to individual functions.
24894 @item -mpreferred-stack-boundary=@var{num}
24895 @opindex mpreferred-stack-boundary
24896 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
24897 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
24898 the default is 4 (16 bytes or 128 bits).
24900 @strong{Warning:} When generating code for the x86-64 architecture with
24901 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
24902 used to keep the stack boundary aligned to 8 byte boundary. Since
24903 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
24904 intended to be used in controlled environment where stack space is
24905 important limitation. This option leads to wrong code when functions
24906 compiled with 16 byte stack alignment (such as functions from a standard
24907 library) are called with misaligned stack. In this case, SSE
24908 instructions may lead to misaligned memory access traps. In addition,
24909 variable arguments are handled incorrectly for 16 byte aligned
24910 objects (including x87 long double and __int128), leading to wrong
24911 results. You must build all modules with
24912 @option{-mpreferred-stack-boundary=3}, including any libraries. This
24913 includes the system libraries and startup modules.
24915 @item -mincoming-stack-boundary=@var{num}
24916 @opindex mincoming-stack-boundary
24917 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
24918 boundary. If @option{-mincoming-stack-boundary} is not specified,
24919 the one specified by @option{-mpreferred-stack-boundary} is used.
24921 On Pentium and Pentium Pro, @code{double} and @code{long double} values
24922 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
24923 suffer significant run time performance penalties. On Pentium III, the
24924 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
24925 properly if it is not 16-byte aligned.
24927 To ensure proper alignment of this values on the stack, the stack boundary
24928 must be as aligned as that required by any value stored on the stack.
24929 Further, every function must be generated such that it keeps the stack
24930 aligned. Thus calling a function compiled with a higher preferred
24931 stack boundary from a function compiled with a lower preferred stack
24932 boundary most likely misaligns the stack. It is recommended that
24933 libraries that use callbacks always use the default setting.
24935 This extra alignment does consume extra stack space, and generally
24936 increases code size. Code that is sensitive to stack space usage, such
24937 as embedded systems and operating system kernels, may want to reduce the
24938 preferred alignment to @option{-mpreferred-stack-boundary=2}.
24995 @itemx -mavx512ifma
24996 @opindex mavx512ifma
24998 @itemx -mavx512vbmi
24999 @opindex mavx512vbmi
25011 @opindex mclfushopt
25028 @itemx -mprefetchwt1
25029 @opindex mprefetchwt1
25088 These switches enable the use of instructions in the MMX, SSE,
25089 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25090 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
25091 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
25092 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, 3DNow!@: or enhanced 3DNow!@:
25093 extended instruction sets. Each has a corresponding @option{-mno-} option
25094 to disable use of these instructions.
25096 These extensions are also available as built-in functions: see
25097 @ref{x86 Built-in Functions}, for details of the functions enabled and
25098 disabled by these switches.
25100 To generate SSE/SSE2 instructions automatically from floating-point
25101 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
25103 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
25104 generates new AVX instructions or AVX equivalence for all SSEx instructions
25107 These options enable GCC to use these extended instructions in
25108 generated code, even without @option{-mfpmath=sse}. Applications that
25109 perform run-time CPU detection must compile separate files for each
25110 supported architecture, using the appropriate flags. In particular,
25111 the file containing the CPU detection code should be compiled without
25114 @item -mdump-tune-features
25115 @opindex mdump-tune-features
25116 This option instructs GCC to dump the names of the x86 performance
25117 tuning features and default settings. The names can be used in
25118 @option{-mtune-ctrl=@var{feature-list}}.
25120 @item -mtune-ctrl=@var{feature-list}
25121 @opindex mtune-ctrl=@var{feature-list}
25122 This option is used to do fine grain control of x86 code generation features.
25123 @var{feature-list} is a comma separated list of @var{feature} names. See also
25124 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
25125 on if it is not preceded with @samp{^}, otherwise, it is turned off.
25126 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
25127 developers. Using it may lead to code paths not covered by testing and can
25128 potentially result in compiler ICEs or runtime errors.
25131 @opindex mno-default
25132 This option instructs GCC to turn off all tunable features. See also
25133 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
25137 This option instructs GCC to emit a @code{cld} instruction in the prologue
25138 of functions that use string instructions. String instructions depend on
25139 the DF flag to select between autoincrement or autodecrement mode. While the
25140 ABI specifies the DF flag to be cleared on function entry, some operating
25141 systems violate this specification by not clearing the DF flag in their
25142 exception dispatchers. The exception handler can be invoked with the DF flag
25143 set, which leads to wrong direction mode when string instructions are used.
25144 This option can be enabled by default on 32-bit x86 targets by configuring
25145 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
25146 instructions can be suppressed with the @option{-mno-cld} compiler option
25150 @opindex mvzeroupper
25151 This option instructs GCC to emit a @code{vzeroupper} instruction
25152 before a transfer of control flow out of the function to minimize
25153 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
25156 @item -mprefer-avx128
25157 @opindex mprefer-avx128
25158 This option instructs GCC to use 128-bit AVX instructions instead of
25159 256-bit AVX instructions in the auto-vectorizer.
25163 This option enables GCC to generate @code{CMPXCHG16B} instructions.
25164 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
25165 (or oword) data types.
25166 This is useful for high-resolution counters that can be updated
25167 by multiple processors (or cores). This instruction is generated as part of
25168 atomic built-in functions: see @ref{__sync Builtins} or
25169 @ref{__atomic Builtins} for details.
25173 This option enables generation of @code{SAHF} instructions in 64-bit code.
25174 Early Intel Pentium 4 CPUs with Intel 64 support,
25175 prior to the introduction of Pentium 4 G1 step in December 2005,
25176 lacked the @code{LAHF} and @code{SAHF} instructions
25177 which are supported by AMD64.
25178 These are load and store instructions, respectively, for certain status flags.
25179 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
25180 @code{drem}, and @code{remainder} built-in functions;
25181 see @ref{Other Builtins} for details.
25185 This option enables use of the @code{movbe} instruction to implement
25186 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
25190 This option enables built-in functions @code{__builtin_ia32_crc32qi},
25191 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
25192 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
25196 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
25197 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
25198 with an additional Newton-Raphson step
25199 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
25200 (and their vectorized
25201 variants) for single-precision floating-point arguments. These instructions
25202 are generated only when @option{-funsafe-math-optimizations} is enabled
25203 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
25204 Note that while the throughput of the sequence is higher than the throughput
25205 of the non-reciprocal instruction, the precision of the sequence can be
25206 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
25208 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
25209 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
25210 combination), and doesn't need @option{-mrecip}.
25212 Also note that GCC emits the above sequence with additional Newton-Raphson step
25213 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
25214 already with @option{-ffast-math} (or the above option combination), and
25215 doesn't need @option{-mrecip}.
25217 @item -mrecip=@var{opt}
25218 @opindex mrecip=opt
25219 This option controls which reciprocal estimate instructions
25220 may be used. @var{opt} is a comma-separated list of options, which may
25221 be preceded by a @samp{!} to invert the option:
25225 Enable all estimate instructions.
25228 Enable the default instructions, equivalent to @option{-mrecip}.
25231 Disable all estimate instructions, equivalent to @option{-mno-recip}.
25234 Enable the approximation for scalar division.
25237 Enable the approximation for vectorized division.
25240 Enable the approximation for scalar square root.
25243 Enable the approximation for vectorized square root.
25246 So, for example, @option{-mrecip=all,!sqrt} enables
25247 all of the reciprocal approximations, except for square root.
25249 @item -mveclibabi=@var{type}
25250 @opindex mveclibabi
25251 Specifies the ABI type to use for vectorizing intrinsics using an
25252 external library. Supported values for @var{type} are @samp{svml}
25253 for the Intel short
25254 vector math library and @samp{acml} for the AMD math core library.
25255 To use this option, both @option{-ftree-vectorize} and
25256 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
25257 ABI-compatible library must be specified at link time.
25259 GCC currently emits calls to @code{vmldExp2},
25260 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
25261 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
25262 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
25263 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
25264 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
25265 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
25266 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
25267 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
25268 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
25269 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
25270 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
25271 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
25272 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
25273 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
25274 when @option{-mveclibabi=acml} is used.
25276 @item -mabi=@var{name}
25278 Generate code for the specified calling convention. Permissible values
25279 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
25280 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
25281 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
25282 You can control this behavior for specific functions by
25283 using the function attributes @code{ms_abi} and @code{sysv_abi}.
25284 @xref{Function Attributes}.
25286 @item -mtls-dialect=@var{type}
25287 @opindex mtls-dialect
25288 Generate code to access thread-local storage using the @samp{gnu} or
25289 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
25290 @samp{gnu2} is more efficient, but it may add compile- and run-time
25291 requirements that cannot be satisfied on all systems.
25294 @itemx -mno-push-args
25295 @opindex mpush-args
25296 @opindex mno-push-args
25297 Use PUSH operations to store outgoing parameters. This method is shorter
25298 and usually equally fast as method using SUB/MOV operations and is enabled
25299 by default. In some cases disabling it may improve performance because of
25300 improved scheduling and reduced dependencies.
25302 @item -maccumulate-outgoing-args
25303 @opindex maccumulate-outgoing-args
25304 If enabled, the maximum amount of space required for outgoing arguments is
25305 computed in the function prologue. This is faster on most modern CPUs
25306 because of reduced dependencies, improved scheduling and reduced stack usage
25307 when the preferred stack boundary is not equal to 2. The drawback is a notable
25308 increase in code size. This switch implies @option{-mno-push-args}.
25312 Support thread-safe exception handling on MinGW. Programs that rely
25313 on thread-safe exception handling must compile and link all code with the
25314 @option{-mthreads} option. When compiling, @option{-mthreads} defines
25315 @option{-D_MT}; when linking, it links in a special thread helper library
25316 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
25318 @item -mms-bitfields
25319 @itemx -mno-ms-bitfields
25320 @opindex mms-bitfields
25321 @opindex mno-ms-bitfields
25323 Enable/disable bit-field layout compatible with the native Microsoft
25326 If @code{packed} is used on a structure, or if bit-fields are used,
25327 it may be that the Microsoft ABI lays out the structure differently
25328 than the way GCC normally does. Particularly when moving packed
25329 data between functions compiled with GCC and the native Microsoft compiler
25330 (either via function call or as data in a file), it may be necessary to access
25333 This option is enabled by default for Microsoft Windows
25334 targets. This behavior can also be controlled locally by use of variable
25335 or type attributes. For more information, see @ref{x86 Variable Attributes}
25336 and @ref{x86 Type Attributes}.
25338 The Microsoft structure layout algorithm is fairly simple with the exception
25339 of the bit-field packing.
25340 The padding and alignment of members of structures and whether a bit-field
25341 can straddle a storage-unit boundary are determine by these rules:
25344 @item Structure members are stored sequentially in the order in which they are
25345 declared: the first member has the lowest memory address and the last member
25348 @item Every data object has an alignment requirement. The alignment requirement
25349 for all data except structures, unions, and arrays is either the size of the
25350 object or the current packing size (specified with either the
25351 @code{aligned} attribute or the @code{pack} pragma),
25352 whichever is less. For structures, unions, and arrays,
25353 the alignment requirement is the largest alignment requirement of its members.
25354 Every object is allocated an offset so that:
25357 offset % alignment_requirement == 0
25360 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
25361 unit if the integral types are the same size and if the next bit-field fits
25362 into the current allocation unit without crossing the boundary imposed by the
25363 common alignment requirements of the bit-fields.
25366 MSVC interprets zero-length bit-fields in the following ways:
25369 @item If a zero-length bit-field is inserted between two bit-fields that
25370 are normally coalesced, the bit-fields are not coalesced.
25377 unsigned long bf_1 : 12;
25379 unsigned long bf_2 : 12;
25384 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
25385 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
25387 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
25388 alignment of the zero-length bit-field is greater than the member that follows it,
25389 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
25410 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
25411 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
25412 bit-field does not affect the alignment of @code{bar} or, as a result, the size
25415 Taking this into account, it is important to note the following:
25418 @item If a zero-length bit-field follows a normal bit-field, the type of the
25419 zero-length bit-field may affect the alignment of the structure as whole. For
25420 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
25421 normal bit-field, and is of type short.
25423 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
25424 still affect the alignment of the structure:
25435 Here, @code{t4} takes up 4 bytes.
25438 @item Zero-length bit-fields following non-bit-field members are ignored:
25450 Here, @code{t5} takes up 2 bytes.
25454 @item -mno-align-stringops
25455 @opindex mno-align-stringops
25456 Do not align the destination of inlined string operations. This switch reduces
25457 code size and improves performance in case the destination is already aligned,
25458 but GCC doesn't know about it.
25460 @item -minline-all-stringops
25461 @opindex minline-all-stringops
25462 By default GCC inlines string operations only when the destination is
25463 known to be aligned to least a 4-byte boundary.
25464 This enables more inlining and increases code
25465 size, but may improve performance of code that depends on fast
25466 @code{memcpy}, @code{strlen},
25467 and @code{memset} for short lengths.
25469 @item -minline-stringops-dynamically
25470 @opindex minline-stringops-dynamically
25471 For string operations of unknown size, use run-time checks with
25472 inline code for small blocks and a library call for large blocks.
25474 @item -mstringop-strategy=@var{alg}
25475 @opindex mstringop-strategy=@var{alg}
25476 Override the internal decision heuristic for the particular algorithm to use
25477 for inlining string operations. The allowed values for @var{alg} are:
25483 Expand using i386 @code{rep} prefix of the specified size.
25487 @itemx unrolled_loop
25488 Expand into an inline loop.
25491 Always use a library call.
25494 @item -mmemcpy-strategy=@var{strategy}
25495 @opindex mmemcpy-strategy=@var{strategy}
25496 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
25497 should be inlined and what inline algorithm to use when the expected size
25498 of the copy operation is known. @var{strategy}
25499 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
25500 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
25501 the max byte size with which inline algorithm @var{alg} is allowed. For the last
25502 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
25503 in the list must be specified in increasing order. The minimal byte size for
25504 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
25507 @item -mmemset-strategy=@var{strategy}
25508 @opindex mmemset-strategy=@var{strategy}
25509 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
25510 @code{__builtin_memset} expansion.
25512 @item -momit-leaf-frame-pointer
25513 @opindex momit-leaf-frame-pointer
25514 Don't keep the frame pointer in a register for leaf functions. This
25515 avoids the instructions to save, set up, and restore frame pointers and
25516 makes an extra register available in leaf functions. The option
25517 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
25518 which might make debugging harder.
25520 @item -mtls-direct-seg-refs
25521 @itemx -mno-tls-direct-seg-refs
25522 @opindex mtls-direct-seg-refs
25523 Controls whether TLS variables may be accessed with offsets from the
25524 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
25525 or whether the thread base pointer must be added. Whether or not this
25526 is valid depends on the operating system, and whether it maps the
25527 segment to cover the entire TLS area.
25529 For systems that use the GNU C Library, the default is on.
25532 @itemx -mno-sse2avx
25534 Specify that the assembler should encode SSE instructions with VEX
25535 prefix. The option @option{-mavx} turns this on by default.
25540 If profiling is active (@option{-pg}), put the profiling
25541 counter call before the prologue.
25542 Note: On x86 architectures the attribute @code{ms_hook_prologue}
25543 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
25545 @item -mrecord-mcount
25546 @itemx -mno-record-mcount
25547 @opindex mrecord-mcount
25548 If profiling is active (@option{-pg}), generate a __mcount_loc section
25549 that contains pointers to each profiling call. This is useful for
25550 automatically patching and out calls.
25553 @itemx -mno-nop-mcount
25554 @opindex mnop-mcount
25555 If profiling is active (@option{-pg}), generate the calls to
25556 the profiling functions as NOPs. This is useful when they
25557 should be patched in later dynamically. This is likely only
25558 useful together with @option{-mrecord-mcount}.
25560 @item -mskip-rax-setup
25561 @itemx -mno-skip-rax-setup
25562 @opindex mskip-rax-setup
25563 When generating code for the x86-64 architecture with SSE extensions
25564 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
25565 register when there are no variable arguments passed in vector registers.
25567 @strong{Warning:} Since RAX register is used to avoid unnecessarily
25568 saving vector registers on stack when passing variable arguments, the
25569 impacts of this option are callees may waste some stack space,
25570 misbehave or jump to a random location. GCC 4.4 or newer don't have
25571 those issues, regardless the RAX register value.
25574 @itemx -mno-8bit-idiv
25575 @opindex m8bit-idiv
25576 On some processors, like Intel Atom, 8-bit unsigned integer divide is
25577 much faster than 32-bit/64-bit integer divide. This option generates a
25578 run-time check. If both dividend and divisor are within range of 0
25579 to 255, 8-bit unsigned integer divide is used instead of
25580 32-bit/64-bit integer divide.
25582 @item -mavx256-split-unaligned-load
25583 @itemx -mavx256-split-unaligned-store
25584 @opindex mavx256-split-unaligned-load
25585 @opindex mavx256-split-unaligned-store
25586 Split 32-byte AVX unaligned load and store.
25588 @item -mstack-protector-guard=@var{guard}
25589 @opindex mstack-protector-guard=@var{guard}
25590 Generate stack protection code using canary at @var{guard}. Supported
25591 locations are @samp{global} for global canary or @samp{tls} for per-thread
25592 canary in the TLS block (the default). This option has effect only when
25593 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
25595 @item -mmitigate-rop
25596 @opindex mmitigate-rop
25597 Try to avoid generating code sequences that contain unintended return
25598 opcodes, to mitigate against certain forms of attack. At the moment,
25599 this option is limited in what it can do and should not be relied
25600 on to provide serious protection.
25602 @item -mgeneral-regs-only
25603 @opindex mgeneral-regs-only
25604 Generate code that uses only the general-purpose registers. This
25605 prevents the compiler from using floating-point, vector, mask and bound
25610 These @samp{-m} switches are supported in addition to the above
25611 on x86-64 processors in 64-bit environments.
25624 Generate code for a 16-bit, 32-bit or 64-bit environment.
25625 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
25627 generates code that runs on any i386 system.
25629 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
25630 types to 64 bits, and generates code for the x86-64 architecture.
25631 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
25632 and @option{-mdynamic-no-pic} options.
25634 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
25636 generates code for the x86-64 architecture.
25638 The @option{-m16} option is the same as @option{-m32}, except for that
25639 it outputs the @code{.code16gcc} assembly directive at the beginning of
25640 the assembly output so that the binary can run in 16-bit mode.
25642 The @option{-miamcu} option generates code which conforms to Intel MCU
25643 psABI. It requires the @option{-m32} option to be turned on.
25645 @item -mno-red-zone
25646 @opindex mno-red-zone
25647 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
25648 by the x86-64 ABI; it is a 128-byte area beyond the location of the
25649 stack pointer that is not modified by signal or interrupt handlers
25650 and therefore can be used for temporary data without adjusting the stack
25651 pointer. The flag @option{-mno-red-zone} disables this red zone.
25653 @item -mcmodel=small
25654 @opindex mcmodel=small
25655 Generate code for the small code model: the program and its symbols must
25656 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
25657 Programs can be statically or dynamically linked. This is the default
25660 @item -mcmodel=kernel
25661 @opindex mcmodel=kernel
25662 Generate code for the kernel code model. The kernel runs in the
25663 negative 2 GB of the address space.
25664 This model has to be used for Linux kernel code.
25666 @item -mcmodel=medium
25667 @opindex mcmodel=medium
25668 Generate code for the medium model: the program is linked in the lower 2
25669 GB of the address space. Small symbols are also placed there. Symbols
25670 with sizes larger than @option{-mlarge-data-threshold} are put into
25671 large data or BSS sections and can be located above 2GB. Programs can
25672 be statically or dynamically linked.
25674 @item -mcmodel=large
25675 @opindex mcmodel=large
25676 Generate code for the large model. This model makes no assumptions
25677 about addresses and sizes of sections.
25679 @item -maddress-mode=long
25680 @opindex maddress-mode=long
25681 Generate code for long address mode. This is only supported for 64-bit
25682 and x32 environments. It is the default address mode for 64-bit
25685 @item -maddress-mode=short
25686 @opindex maddress-mode=short
25687 Generate code for short address mode. This is only supported for 32-bit
25688 and x32 environments. It is the default address mode for 32-bit and
25692 @node x86 Windows Options
25693 @subsection x86 Windows Options
25694 @cindex x86 Windows Options
25695 @cindex Windows Options for x86
25697 These additional options are available for Microsoft Windows targets:
25703 specifies that a console application is to be generated, by
25704 instructing the linker to set the PE header subsystem type
25705 required for console applications.
25706 This option is available for Cygwin and MinGW targets and is
25707 enabled by default on those targets.
25711 This option is available for Cygwin and MinGW targets. It
25712 specifies that a DLL---a dynamic link library---is to be
25713 generated, enabling the selection of the required runtime
25714 startup object and entry point.
25716 @item -mnop-fun-dllimport
25717 @opindex mnop-fun-dllimport
25718 This option is available for Cygwin and MinGW targets. It
25719 specifies that the @code{dllimport} attribute should be ignored.
25723 This option is available for MinGW targets. It specifies
25724 that MinGW-specific thread support is to be used.
25728 This option is available for MinGW-w64 targets. It causes
25729 the @code{UNICODE} preprocessor macro to be predefined, and
25730 chooses Unicode-capable runtime startup code.
25734 This option is available for Cygwin and MinGW targets. It
25735 specifies that the typical Microsoft Windows predefined macros are to
25736 be set in the pre-processor, but does not influence the choice
25737 of runtime library/startup code.
25741 This option is available for Cygwin and MinGW targets. It
25742 specifies that a GUI application is to be generated by
25743 instructing the linker to set the PE header subsystem type
25746 @item -fno-set-stack-executable
25747 @opindex fno-set-stack-executable
25748 This option is available for MinGW targets. It specifies that
25749 the executable flag for the stack used by nested functions isn't
25750 set. This is necessary for binaries running in kernel mode of
25751 Microsoft Windows, as there the User32 API, which is used to set executable
25752 privileges, isn't available.
25754 @item -fwritable-relocated-rdata
25755 @opindex fno-writable-relocated-rdata
25756 This option is available for MinGW and Cygwin targets. It specifies
25757 that relocated-data in read-only section is put into the @code{.data}
25758 section. This is a necessary for older runtimes not supporting
25759 modification of @code{.rdata} sections for pseudo-relocation.
25761 @item -mpe-aligned-commons
25762 @opindex mpe-aligned-commons
25763 This option is available for Cygwin and MinGW targets. It
25764 specifies that the GNU extension to the PE file format that
25765 permits the correct alignment of COMMON variables should be
25766 used when generating code. It is enabled by default if
25767 GCC detects that the target assembler found during configuration
25768 supports the feature.
25771 See also under @ref{x86 Options} for standard options.
25773 @node Xstormy16 Options
25774 @subsection Xstormy16 Options
25775 @cindex Xstormy16 Options
25777 These options are defined for Xstormy16:
25782 Choose startup files and linker script suitable for the simulator.
25785 @node Xtensa Options
25786 @subsection Xtensa Options
25787 @cindex Xtensa Options
25789 These options are supported for Xtensa targets:
25793 @itemx -mno-const16
25795 @opindex mno-const16
25796 Enable or disable use of @code{CONST16} instructions for loading
25797 constant values. The @code{CONST16} instruction is currently not a
25798 standard option from Tensilica. When enabled, @code{CONST16}
25799 instructions are always used in place of the standard @code{L32R}
25800 instructions. The use of @code{CONST16} is enabled by default only if
25801 the @code{L32R} instruction is not available.
25804 @itemx -mno-fused-madd
25805 @opindex mfused-madd
25806 @opindex mno-fused-madd
25807 Enable or disable use of fused multiply/add and multiply/subtract
25808 instructions in the floating-point option. This has no effect if the
25809 floating-point option is not also enabled. Disabling fused multiply/add
25810 and multiply/subtract instructions forces the compiler to use separate
25811 instructions for the multiply and add/subtract operations. This may be
25812 desirable in some cases where strict IEEE 754-compliant results are
25813 required: the fused multiply add/subtract instructions do not round the
25814 intermediate result, thereby producing results with @emph{more} bits of
25815 precision than specified by the IEEE standard. Disabling fused multiply
25816 add/subtract instructions also ensures that the program output is not
25817 sensitive to the compiler's ability to combine multiply and add/subtract
25820 @item -mserialize-volatile
25821 @itemx -mno-serialize-volatile
25822 @opindex mserialize-volatile
25823 @opindex mno-serialize-volatile
25824 When this option is enabled, GCC inserts @code{MEMW} instructions before
25825 @code{volatile} memory references to guarantee sequential consistency.
25826 The default is @option{-mserialize-volatile}. Use
25827 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
25829 @item -mforce-no-pic
25830 @opindex mforce-no-pic
25831 For targets, like GNU/Linux, where all user-mode Xtensa code must be
25832 position-independent code (PIC), this option disables PIC for compiling
25835 @item -mtext-section-literals
25836 @itemx -mno-text-section-literals
25837 @opindex mtext-section-literals
25838 @opindex mno-text-section-literals
25839 These options control the treatment of literal pools. The default is
25840 @option{-mno-text-section-literals}, which places literals in a separate
25841 section in the output file. This allows the literal pool to be placed
25842 in a data RAM/ROM, and it also allows the linker to combine literal
25843 pools from separate object files to remove redundant literals and
25844 improve code size. With @option{-mtext-section-literals}, the literals
25845 are interspersed in the text section in order to keep them as close as
25846 possible to their references. This may be necessary for large assembly
25847 files. Literals for each function are placed right before that function.
25849 @item -mauto-litpools
25850 @itemx -mno-auto-litpools
25851 @opindex mauto-litpools
25852 @opindex mno-auto-litpools
25853 These options control the treatment of literal pools. The default is
25854 @option{-mno-auto-litpools}, which places literals in a separate
25855 section in the output file unless @option{-mtext-section-literals} is
25856 used. With @option{-mauto-litpools} the literals are interspersed in
25857 the text section by the assembler. Compiler does not produce explicit
25858 @code{.literal} directives and loads literals into registers with
25859 @code{MOVI} instructions instead of @code{L32R} to let the assembler
25860 do relaxation and place literals as necessary. This option allows
25861 assembler to create several literal pools per function and assemble
25862 very big functions, which may not be possible with
25863 @option{-mtext-section-literals}.
25865 @item -mtarget-align
25866 @itemx -mno-target-align
25867 @opindex mtarget-align
25868 @opindex mno-target-align
25869 When this option is enabled, GCC instructs the assembler to
25870 automatically align instructions to reduce branch penalties at the
25871 expense of some code density. The assembler attempts to widen density
25872 instructions to align branch targets and the instructions following call
25873 instructions. If there are not enough preceding safe density
25874 instructions to align a target, no widening is performed. The
25875 default is @option{-mtarget-align}. These options do not affect the
25876 treatment of auto-aligned instructions like @code{LOOP}, which the
25877 assembler always aligns, either by widening density instructions or
25878 by inserting NOP instructions.
25881 @itemx -mno-longcalls
25882 @opindex mlongcalls
25883 @opindex mno-longcalls
25884 When this option is enabled, GCC instructs the assembler to translate
25885 direct calls to indirect calls unless it can determine that the target
25886 of a direct call is in the range allowed by the call instruction. This
25887 translation typically occurs for calls to functions in other source
25888 files. Specifically, the assembler translates a direct @code{CALL}
25889 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
25890 The default is @option{-mno-longcalls}. This option should be used in
25891 programs where the call target can potentially be out of range. This
25892 option is implemented in the assembler, not the compiler, so the
25893 assembly code generated by GCC still shows direct call
25894 instructions---look at the disassembled object code to see the actual
25895 instructions. Note that the assembler uses an indirect call for
25896 every cross-file call, not just those that really are out of range.
25899 @node zSeries Options
25900 @subsection zSeries Options
25901 @cindex zSeries options
25903 These are listed under @xref{S/390 and zSeries Options}.
25909 @section Specifying Subprocesses and the Switches to Pass to Them
25912 @command{gcc} is a driver program. It performs its job by invoking a
25913 sequence of other programs to do the work of compiling, assembling and
25914 linking. GCC interprets its command-line parameters and uses these to
25915 deduce which programs it should invoke, and which command-line options
25916 it ought to place on their command lines. This behavior is controlled
25917 by @dfn{spec strings}. In most cases there is one spec string for each
25918 program that GCC can invoke, but a few programs have multiple spec
25919 strings to control their behavior. The spec strings built into GCC can
25920 be overridden by using the @option{-specs=} command-line switch to specify
25923 @dfn{Spec files} are plain-text files that are used to construct spec
25924 strings. They consist of a sequence of directives separated by blank
25925 lines. The type of directive is determined by the first non-whitespace
25926 character on the line, which can be one of the following:
25929 @item %@var{command}
25930 Issues a @var{command} to the spec file processor. The commands that can
25934 @item %include <@var{file}>
25935 @cindex @code{%include}
25936 Search for @var{file} and insert its text at the current point in the
25939 @item %include_noerr <@var{file}>
25940 @cindex @code{%include_noerr}
25941 Just like @samp{%include}, but do not generate an error message if the include
25942 file cannot be found.
25944 @item %rename @var{old_name} @var{new_name}
25945 @cindex @code{%rename}
25946 Rename the spec string @var{old_name} to @var{new_name}.
25950 @item *[@var{spec_name}]:
25951 This tells the compiler to create, override or delete the named spec
25952 string. All lines after this directive up to the next directive or
25953 blank line are considered to be the text for the spec string. If this
25954 results in an empty string then the spec is deleted. (Or, if the
25955 spec did not exist, then nothing happens.) Otherwise, if the spec
25956 does not currently exist a new spec is created. If the spec does
25957 exist then its contents are overridden by the text of this
25958 directive, unless the first character of that text is the @samp{+}
25959 character, in which case the text is appended to the spec.
25961 @item [@var{suffix}]:
25962 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
25963 and up to the next directive or blank line are considered to make up the
25964 spec string for the indicated suffix. When the compiler encounters an
25965 input file with the named suffix, it processes the spec string in
25966 order to work out how to compile that file. For example:
25970 z-compile -input %i
25973 This says that any input file whose name ends in @samp{.ZZ} should be
25974 passed to the program @samp{z-compile}, which should be invoked with the
25975 command-line switch @option{-input} and with the result of performing the
25976 @samp{%i} substitution. (See below.)
25978 As an alternative to providing a spec string, the text following a
25979 suffix directive can be one of the following:
25982 @item @@@var{language}
25983 This says that the suffix is an alias for a known @var{language}. This is
25984 similar to using the @option{-x} command-line switch to GCC to specify a
25985 language explicitly. For example:
25992 Says that .ZZ files are, in fact, C++ source files.
25995 This causes an error messages saying:
25998 @var{name} compiler not installed on this system.
26002 GCC already has an extensive list of suffixes built into it.
26003 This directive adds an entry to the end of the list of suffixes, but
26004 since the list is searched from the end backwards, it is effectively
26005 possible to override earlier entries using this technique.
26009 GCC has the following spec strings built into it. Spec files can
26010 override these strings or create their own. Note that individual
26011 targets can also add their own spec strings to this list.
26014 asm Options to pass to the assembler
26015 asm_final Options to pass to the assembler post-processor
26016 cpp Options to pass to the C preprocessor
26017 cc1 Options to pass to the C compiler
26018 cc1plus Options to pass to the C++ compiler
26019 endfile Object files to include at the end of the link
26020 link Options to pass to the linker
26021 lib Libraries to include on the command line to the linker
26022 libgcc Decides which GCC support library to pass to the linker
26023 linker Sets the name of the linker
26024 predefines Defines to be passed to the C preprocessor
26025 signed_char Defines to pass to CPP to say whether @code{char} is signed
26027 startfile Object files to include at the start of the link
26030 Here is a small example of a spec file:
26033 %rename lib old_lib
26036 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
26039 This example renames the spec called @samp{lib} to @samp{old_lib} and
26040 then overrides the previous definition of @samp{lib} with a new one.
26041 The new definition adds in some extra command-line options before
26042 including the text of the old definition.
26044 @dfn{Spec strings} are a list of command-line options to be passed to their
26045 corresponding program. In addition, the spec strings can contain
26046 @samp{%}-prefixed sequences to substitute variable text or to
26047 conditionally insert text into the command line. Using these constructs
26048 it is possible to generate quite complex command lines.
26050 Here is a table of all defined @samp{%}-sequences for spec
26051 strings. Note that spaces are not generated automatically around the
26052 results of expanding these sequences. Therefore you can concatenate them
26053 together or combine them with constant text in a single argument.
26057 Substitute one @samp{%} into the program name or argument.
26060 Substitute the name of the input file being processed.
26063 Substitute the basename of the input file being processed.
26064 This is the substring up to (and not including) the last period
26065 and not including the directory.
26068 This is the same as @samp{%b}, but include the file suffix (text after
26072 Marks the argument containing or following the @samp{%d} as a
26073 temporary file name, so that that file is deleted if GCC exits
26074 successfully. Unlike @samp{%g}, this contributes no text to the
26077 @item %g@var{suffix}
26078 Substitute a file name that has suffix @var{suffix} and is chosen
26079 once per compilation, and mark the argument in the same way as
26080 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
26081 name is now chosen in a way that is hard to predict even when previously
26082 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
26083 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
26084 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
26085 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
26086 was simply substituted with a file name chosen once per compilation,
26087 without regard to any appended suffix (which was therefore treated
26088 just like ordinary text), making such attacks more likely to succeed.
26090 @item %u@var{suffix}
26091 Like @samp{%g}, but generates a new temporary file name
26092 each time it appears instead of once per compilation.
26094 @item %U@var{suffix}
26095 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
26096 new one if there is no such last file name. In the absence of any
26097 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
26098 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
26099 involves the generation of two distinct file names, one
26100 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
26101 simply substituted with a file name chosen for the previous @samp{%u},
26102 without regard to any appended suffix.
26104 @item %j@var{suffix}
26105 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
26106 writable, and if @option{-save-temps} is not used;
26107 otherwise, substitute the name
26108 of a temporary file, just like @samp{%u}. This temporary file is not
26109 meant for communication between processes, but rather as a junk
26110 disposal mechanism.
26112 @item %|@var{suffix}
26113 @itemx %m@var{suffix}
26114 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
26115 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
26116 all. These are the two most common ways to instruct a program that it
26117 should read from standard input or write to standard output. If you
26118 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
26119 construct: see for example @file{f/lang-specs.h}.
26121 @item %.@var{SUFFIX}
26122 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
26123 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
26124 terminated by the next space or %.
26127 Marks the argument containing or following the @samp{%w} as the
26128 designated output file of this compilation. This puts the argument
26129 into the sequence of arguments that @samp{%o} substitutes.
26132 Substitutes the names of all the output files, with spaces
26133 automatically placed around them. You should write spaces
26134 around the @samp{%o} as well or the results are undefined.
26135 @samp{%o} is for use in the specs for running the linker.
26136 Input files whose names have no recognized suffix are not compiled
26137 at all, but they are included among the output files, so they are
26141 Substitutes the suffix for object files. Note that this is
26142 handled specially when it immediately follows @samp{%g, %u, or %U},
26143 because of the need for those to form complete file names. The
26144 handling is such that @samp{%O} is treated exactly as if it had already
26145 been substituted, except that @samp{%g, %u, and %U} do not currently
26146 support additional @var{suffix} characters following @samp{%O} as they do
26147 following, for example, @samp{.o}.
26150 Substitutes the standard macro predefinitions for the
26151 current target machine. Use this when running @command{cpp}.
26154 Like @samp{%p}, but puts @samp{__} before and after the name of each
26155 predefined macro, except for macros that start with @samp{__} or with
26156 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
26160 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
26161 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
26162 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
26163 and @option{-imultilib} as necessary.
26166 Current argument is the name of a library or startup file of some sort.
26167 Search for that file in a standard list of directories and substitute
26168 the full name found. The current working directory is included in the
26169 list of directories scanned.
26172 Current argument is the name of a linker script. Search for that file
26173 in the current list of directories to scan for libraries. If the file
26174 is located insert a @option{--script} option into the command line
26175 followed by the full path name found. If the file is not found then
26176 generate an error message. Note: the current working directory is not
26180 Print @var{str} as an error message. @var{str} is terminated by a newline.
26181 Use this when inconsistent options are detected.
26183 @item %(@var{name})
26184 Substitute the contents of spec string @var{name} at this point.
26186 @item %x@{@var{option}@}
26187 Accumulate an option for @samp{%X}.
26190 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
26194 Output the accumulated assembler options specified by @option{-Wa}.
26197 Output the accumulated preprocessor options specified by @option{-Wp}.
26200 Process the @code{asm} spec. This is used to compute the
26201 switches to be passed to the assembler.
26204 Process the @code{asm_final} spec. This is a spec string for
26205 passing switches to an assembler post-processor, if such a program is
26209 Process the @code{link} spec. This is the spec for computing the
26210 command line passed to the linker. Typically it makes use of the
26211 @samp{%L %G %S %D and %E} sequences.
26214 Dump out a @option{-L} option for each directory that GCC believes might
26215 contain startup files. If the target supports multilibs then the
26216 current multilib directory is prepended to each of these paths.
26219 Process the @code{lib} spec. This is a spec string for deciding which
26220 libraries are included on the command line to the linker.
26223 Process the @code{libgcc} spec. This is a spec string for deciding
26224 which GCC support library is included on the command line to the linker.
26227 Process the @code{startfile} spec. This is a spec for deciding which
26228 object files are the first ones passed to the linker. Typically
26229 this might be a file named @file{crt0.o}.
26232 Process the @code{endfile} spec. This is a spec string that specifies
26233 the last object files that are passed to the linker.
26236 Process the @code{cpp} spec. This is used to construct the arguments
26237 to be passed to the C preprocessor.
26240 Process the @code{cc1} spec. This is used to construct the options to be
26241 passed to the actual C compiler (@command{cc1}).
26244 Process the @code{cc1plus} spec. This is used to construct the options to be
26245 passed to the actual C++ compiler (@command{cc1plus}).
26248 Substitute the variable part of a matched option. See below.
26249 Note that each comma in the substituted string is replaced by
26253 Remove all occurrences of @code{-S} from the command line. Note---this
26254 command is position dependent. @samp{%} commands in the spec string
26255 before this one see @code{-S}, @samp{%} commands in the spec string
26256 after this one do not.
26258 @item %:@var{function}(@var{args})
26259 Call the named function @var{function}, passing it @var{args}.
26260 @var{args} is first processed as a nested spec string, then split
26261 into an argument vector in the usual fashion. The function returns
26262 a string which is processed as if it had appeared literally as part
26263 of the current spec.
26265 The following built-in spec functions are provided:
26268 @item @code{getenv}
26269 The @code{getenv} spec function takes two arguments: an environment
26270 variable name and a string. If the environment variable is not
26271 defined, a fatal error is issued. Otherwise, the return value is the
26272 value of the environment variable concatenated with the string. For
26273 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
26276 %:getenv(TOPDIR /include)
26279 expands to @file{/path/to/top/include}.
26281 @item @code{if-exists}
26282 The @code{if-exists} spec function takes one argument, an absolute
26283 pathname to a file. If the file exists, @code{if-exists} returns the
26284 pathname. Here is a small example of its usage:
26288 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
26291 @item @code{if-exists-else}
26292 The @code{if-exists-else} spec function is similar to the @code{if-exists}
26293 spec function, except that it takes two arguments. The first argument is
26294 an absolute pathname to a file. If the file exists, @code{if-exists-else}
26295 returns the pathname. If it does not exist, it returns the second argument.
26296 This way, @code{if-exists-else} can be used to select one file or another,
26297 based on the existence of the first. Here is a small example of its usage:
26301 crt0%O%s %:if-exists(crti%O%s) \
26302 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
26305 @item @code{replace-outfile}
26306 The @code{replace-outfile} spec function takes two arguments. It looks for the
26307 first argument in the outfiles array and replaces it with the second argument. Here
26308 is a small example of its usage:
26311 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
26314 @item @code{remove-outfile}
26315 The @code{remove-outfile} spec function takes one argument. It looks for the
26316 first argument in the outfiles array and removes it. Here is a small example
26320 %:remove-outfile(-lm)
26323 @item @code{pass-through-libs}
26324 The @code{pass-through-libs} spec function takes any number of arguments. It
26325 finds any @option{-l} options and any non-options ending in @file{.a} (which it
26326 assumes are the names of linker input library archive files) and returns a
26327 result containing all the found arguments each prepended by
26328 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
26329 intended to be passed to the LTO linker plugin.
26332 %:pass-through-libs(%G %L %G)
26335 @item @code{print-asm-header}
26336 The @code{print-asm-header} function takes no arguments and simply
26337 prints a banner like:
26343 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
26346 It is used to separate compiler options from assembler options
26347 in the @option{--target-help} output.
26351 Substitutes the @code{-S} switch, if that switch is given to GCC@.
26352 If that switch is not specified, this substitutes nothing. Note that
26353 the leading dash is omitted when specifying this option, and it is
26354 automatically inserted if the substitution is performed. Thus the spec
26355 string @samp{%@{foo@}} matches the command-line option @option{-foo}
26356 and outputs the command-line option @option{-foo}.
26359 Like %@{@code{S}@} but mark last argument supplied within as a file to be
26360 deleted on failure.
26363 Substitutes all the switches specified to GCC whose names start
26364 with @code{-S}, but which also take an argument. This is used for
26365 switches like @option{-o}, @option{-D}, @option{-I}, etc.
26366 GCC considers @option{-o foo} as being
26367 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
26368 text, including the space. Thus two arguments are generated.
26371 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
26372 (the order of @code{S} and @code{T} in the spec is not significant).
26373 There can be any number of ampersand-separated variables; for each the
26374 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
26377 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
26380 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
26383 Substitutes @code{X} if one or more switches whose names start with
26384 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
26385 once, no matter how many such switches appeared. However, if @code{%*}
26386 appears somewhere in @code{X}, then @code{X} is substituted once
26387 for each matching switch, with the @code{%*} replaced by the part of
26388 that switch matching the @code{*}.
26390 If @code{%*} appears as the last part of a spec sequence then a space
26391 is added after the end of the last substitution. If there is more
26392 text in the sequence, however, then a space is not generated. This
26393 allows the @code{%*} substitution to be used as part of a larger
26394 string. For example, a spec string like this:
26397 %@{mcu=*:--script=%*/memory.ld@}
26401 when matching an option like @option{-mcu=newchip} produces:
26404 --script=newchip/memory.ld
26408 Substitutes @code{X}, if processing a file with suffix @code{S}.
26411 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
26414 Substitutes @code{X}, if processing a file for language @code{S}.
26417 Substitutes @code{X}, if not processing a file for language @code{S}.
26420 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
26421 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
26422 @code{*} sequences as well, although they have a stronger binding than
26423 the @samp{|}. If @code{%*} appears in @code{X}, all of the
26424 alternatives must be starred, and only the first matching alternative
26427 For example, a spec string like this:
26430 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
26434 outputs the following command-line options from the following input
26435 command-line options:
26440 -d fred.c -foo -baz -boggle
26441 -d jim.d -bar -baz -boggle
26444 @item %@{S:X; T:Y; :D@}
26446 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
26447 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
26448 be as many clauses as you need. This may be combined with @code{.},
26449 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
26454 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
26455 or similar construct can use a backslash to ignore the special meaning
26456 of the character following it, thus allowing literal matching of a
26457 character that is otherwise specially treated. For example,
26458 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
26459 @option{-std=iso9899:1999} option is given.
26461 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
26462 construct may contain other nested @samp{%} constructs or spaces, or
26463 even newlines. They are processed as usual, as described above.
26464 Trailing white space in @code{X} is ignored. White space may also
26465 appear anywhere on the left side of the colon in these constructs,
26466 except between @code{.} or @code{*} and the corresponding word.
26468 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
26469 handled specifically in these constructs. If another value of
26470 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
26471 @option{-W} switch is found later in the command line, the earlier
26472 switch value is ignored, except with @{@code{S}*@} where @code{S} is
26473 just one letter, which passes all matching options.
26475 The character @samp{|} at the beginning of the predicate text is used to
26476 indicate that a command should be piped to the following command, but
26477 only if @option{-pipe} is specified.
26479 It is built into GCC which switches take arguments and which do not.
26480 (You might think it would be useful to generalize this to allow each
26481 compiler's spec to say which switches take arguments. But this cannot
26482 be done in a consistent fashion. GCC cannot even decide which input
26483 files have been specified without knowing which switches take arguments,
26484 and it must know which input files to compile in order to tell which
26487 GCC also knows implicitly that arguments starting in @option{-l} are to be
26488 treated as compiler output files, and passed to the linker in their
26489 proper position among the other output files.
26491 @node Environment Variables
26492 @section Environment Variables Affecting GCC
26493 @cindex environment variables
26495 @c man begin ENVIRONMENT
26496 This section describes several environment variables that affect how GCC
26497 operates. Some of them work by specifying directories or prefixes to use
26498 when searching for various kinds of files. Some are used to specify other
26499 aspects of the compilation environment.
26501 Note that you can also specify places to search using options such as
26502 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
26503 take precedence over places specified using environment variables, which
26504 in turn take precedence over those specified by the configuration of GCC@.
26505 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
26506 GNU Compiler Collection (GCC) Internals}.
26511 @c @itemx LC_COLLATE
26513 @c @itemx LC_MONETARY
26514 @c @itemx LC_NUMERIC
26519 @c @findex LC_COLLATE
26520 @findex LC_MESSAGES
26521 @c @findex LC_MONETARY
26522 @c @findex LC_NUMERIC
26526 These environment variables control the way that GCC uses
26527 localization information which allows GCC to work with different
26528 national conventions. GCC inspects the locale categories
26529 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
26530 so. These locale categories can be set to any value supported by your
26531 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
26532 Kingdom encoded in UTF-8.
26534 The @env{LC_CTYPE} environment variable specifies character
26535 classification. GCC uses it to determine the character boundaries in
26536 a string; this is needed for some multibyte encodings that contain quote
26537 and escape characters that are otherwise interpreted as a string
26540 The @env{LC_MESSAGES} environment variable specifies the language to
26541 use in diagnostic messages.
26543 If the @env{LC_ALL} environment variable is set, it overrides the value
26544 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
26545 and @env{LC_MESSAGES} default to the value of the @env{LANG}
26546 environment variable. If none of these variables are set, GCC
26547 defaults to traditional C English behavior.
26551 If @env{TMPDIR} is set, it specifies the directory to use for temporary
26552 files. GCC uses temporary files to hold the output of one stage of
26553 compilation which is to be used as input to the next stage: for example,
26554 the output of the preprocessor, which is the input to the compiler
26557 @item GCC_COMPARE_DEBUG
26558 @findex GCC_COMPARE_DEBUG
26559 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
26560 @option{-fcompare-debug} to the compiler driver. See the documentation
26561 of this option for more details.
26563 @item GCC_EXEC_PREFIX
26564 @findex GCC_EXEC_PREFIX
26565 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
26566 names of the subprograms executed by the compiler. No slash is added
26567 when this prefix is combined with the name of a subprogram, but you can
26568 specify a prefix that ends with a slash if you wish.
26570 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
26571 an appropriate prefix to use based on the pathname it is invoked with.
26573 If GCC cannot find the subprogram using the specified prefix, it
26574 tries looking in the usual places for the subprogram.
26576 The default value of @env{GCC_EXEC_PREFIX} is
26577 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
26578 the installed compiler. In many cases @var{prefix} is the value
26579 of @code{prefix} when you ran the @file{configure} script.
26581 Other prefixes specified with @option{-B} take precedence over this prefix.
26583 This prefix is also used for finding files such as @file{crt0.o} that are
26586 In addition, the prefix is used in an unusual way in finding the
26587 directories to search for header files. For each of the standard
26588 directories whose name normally begins with @samp{/usr/local/lib/gcc}
26589 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
26590 replacing that beginning with the specified prefix to produce an
26591 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
26592 @file{foo/bar} just before it searches the standard directory
26593 @file{/usr/local/lib/bar}.
26594 If a standard directory begins with the configured
26595 @var{prefix} then the value of @var{prefix} is replaced by
26596 @env{GCC_EXEC_PREFIX} when looking for header files.
26598 @item COMPILER_PATH
26599 @findex COMPILER_PATH
26600 The value of @env{COMPILER_PATH} is a colon-separated list of
26601 directories, much like @env{PATH}. GCC tries the directories thus
26602 specified when searching for subprograms, if it cannot find the
26603 subprograms using @env{GCC_EXEC_PREFIX}.
26606 @findex LIBRARY_PATH
26607 The value of @env{LIBRARY_PATH} is a colon-separated list of
26608 directories, much like @env{PATH}. When configured as a native compiler,
26609 GCC tries the directories thus specified when searching for special
26610 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
26611 using GCC also uses these directories when searching for ordinary
26612 libraries for the @option{-l} option (but directories specified with
26613 @option{-L} come first).
26617 @cindex locale definition
26618 This variable is used to pass locale information to the compiler. One way in
26619 which this information is used is to determine the character set to be used
26620 when character literals, string literals and comments are parsed in C and C++.
26621 When the compiler is configured to allow multibyte characters,
26622 the following values for @env{LANG} are recognized:
26626 Recognize JIS characters.
26628 Recognize SJIS characters.
26630 Recognize EUCJP characters.
26633 If @env{LANG} is not defined, or if it has some other value, then the
26634 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
26635 recognize and translate multibyte characters.
26639 Some additional environment variables affect the behavior of the
26642 @include cppenv.texi
26646 @node Precompiled Headers
26647 @section Using Precompiled Headers
26648 @cindex precompiled headers
26649 @cindex speed of compilation
26651 Often large projects have many header files that are included in every
26652 source file. The time the compiler takes to process these header files
26653 over and over again can account for nearly all of the time required to
26654 build the project. To make builds faster, GCC allows you to
26655 @dfn{precompile} a header file.
26657 To create a precompiled header file, simply compile it as you would any
26658 other file, if necessary using the @option{-x} option to make the driver
26659 treat it as a C or C++ header file. You may want to use a
26660 tool like @command{make} to keep the precompiled header up-to-date when
26661 the headers it contains change.
26663 A precompiled header file is searched for when @code{#include} is
26664 seen in the compilation. As it searches for the included file
26665 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
26666 compiler looks for a precompiled header in each directory just before it
26667 looks for the include file in that directory. The name searched for is
26668 the name specified in the @code{#include} with @samp{.gch} appended. If
26669 the precompiled header file cannot be used, it is ignored.
26671 For instance, if you have @code{#include "all.h"}, and you have
26672 @file{all.h.gch} in the same directory as @file{all.h}, then the
26673 precompiled header file is used if possible, and the original
26674 header is used otherwise.
26676 Alternatively, you might decide to put the precompiled header file in a
26677 directory and use @option{-I} to ensure that directory is searched
26678 before (or instead of) the directory containing the original header.
26679 Then, if you want to check that the precompiled header file is always
26680 used, you can put a file of the same name as the original header in this
26681 directory containing an @code{#error} command.
26683 This also works with @option{-include}. So yet another way to use
26684 precompiled headers, good for projects not designed with precompiled
26685 header files in mind, is to simply take most of the header files used by
26686 a project, include them from another header file, precompile that header
26687 file, and @option{-include} the precompiled header. If the header files
26688 have guards against multiple inclusion, they are skipped because
26689 they've already been included (in the precompiled header).
26691 If you need to precompile the same header file for different
26692 languages, targets, or compiler options, you can instead make a
26693 @emph{directory} named like @file{all.h.gch}, and put each precompiled
26694 header in the directory, perhaps using @option{-o}. It doesn't matter
26695 what you call the files in the directory; every precompiled header in
26696 the directory is considered. The first precompiled header
26697 encountered in the directory that is valid for this compilation is
26698 used; they're searched in no particular order.
26700 There are many other possibilities, limited only by your imagination,
26701 good sense, and the constraints of your build system.
26703 A precompiled header file can be used only when these conditions apply:
26707 Only one precompiled header can be used in a particular compilation.
26710 A precompiled header cannot be used once the first C token is seen. You
26711 can have preprocessor directives before a precompiled header; you cannot
26712 include a precompiled header from inside another header.
26715 The precompiled header file must be produced for the same language as
26716 the current compilation. You cannot use a C precompiled header for a C++
26720 The precompiled header file must have been produced by the same compiler
26721 binary as the current compilation is using.
26724 Any macros defined before the precompiled header is included must
26725 either be defined in the same way as when the precompiled header was
26726 generated, or must not affect the precompiled header, which usually
26727 means that they don't appear in the precompiled header at all.
26729 The @option{-D} option is one way to define a macro before a
26730 precompiled header is included; using a @code{#define} can also do it.
26731 There are also some options that define macros implicitly, like
26732 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
26735 @item If debugging information is output when using the precompiled
26736 header, using @option{-g} or similar, the same kind of debugging information
26737 must have been output when building the precompiled header. However,
26738 a precompiled header built using @option{-g} can be used in a compilation
26739 when no debugging information is being output.
26741 @item The same @option{-m} options must generally be used when building
26742 and using the precompiled header. @xref{Submodel Options},
26743 for any cases where this rule is relaxed.
26745 @item Each of the following options must be the same when building and using
26746 the precompiled header:
26748 @gccoptlist{-fexceptions}
26751 Some other command-line options starting with @option{-f},
26752 @option{-p}, or @option{-O} must be defined in the same way as when
26753 the precompiled header was generated. At present, it's not clear
26754 which options are safe to change and which are not; the safest choice
26755 is to use exactly the same options when generating and using the
26756 precompiled header. The following are known to be safe:
26758 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
26759 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
26760 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
26765 For all of these except the last, the compiler automatically
26766 ignores the precompiled header if the conditions aren't met. If you
26767 find an option combination that doesn't work and doesn't cause the
26768 precompiled header to be ignored, please consider filing a bug report,
26771 If you do use differing options when generating and using the
26772 precompiled header, the actual behavior is a mixture of the
26773 behavior for the options. For instance, if you use @option{-g} to
26774 generate the precompiled header but not when using it, you may or may
26775 not get debugging information for routines in the precompiled header.