1 @c Copyright (C) 1988-2016 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{} 1988-2016 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 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
182 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
183 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
184 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness}
185 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
186 -fsigned-bitfields -fsigned-char @gol
187 -funsigned-bitfields -funsigned-char @gol
188 -trigraphs -traditional -traditional-cpp}
190 @item C++ Language Options
191 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
192 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
193 -fconstexpr-depth=@var{n} -ffriend-injection @gol
194 -fno-elide-constructors @gol
195 -fno-enforce-eh-specs @gol
196 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
197 -fno-implicit-templates @gol
198 -fno-implicit-inline-templates @gol
199 -fno-implement-inlines -fms-extensions @gol
200 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
201 -fno-optional-diags -fpermissive @gol
202 -fno-pretty-templates @gol
203 -frepo -fno-rtti -fsized-deallocation @gol
204 -ftemplate-backtrace-limit=@var{n} @gol
205 -ftemplate-depth=@var{n} @gol
206 -fno-threadsafe-statics -fuse-cxa-atexit @gol
207 -fno-weak -nostdinc++ @gol
208 -fvisibility-inlines-hidden @gol
209 -fvisibility-ms-compat @gol
210 -fext-numeric-literals @gol
211 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
212 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
213 -Wnamespaces -Wnarrowing @gol
214 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
215 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
216 -Wno-non-template-friend -Wold-style-cast @gol
217 -Woverloaded-virtual -Wno-pmf-conversions @gol
218 -Wsign-promo -Wvirtual-inheritance}
220 @item Objective-C and Objective-C++ Language Options
221 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
222 Objective-C and Objective-C++ Dialects}.
223 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
224 -fgnu-runtime -fnext-runtime @gol
225 -fno-nil-receivers @gol
226 -fobjc-abi-version=@var{n} @gol
227 -fobjc-call-cxx-cdtors @gol
228 -fobjc-direct-dispatch @gol
229 -fobjc-exceptions @gol
232 -fobjc-std=objc1 @gol
233 -fno-local-ivars @gol
234 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
235 -freplace-objc-classes @gol
238 -Wassign-intercept @gol
239 -Wno-protocol -Wselector @gol
240 -Wstrict-selector-match @gol
241 -Wundeclared-selector}
243 @item Diagnostic Message Formatting Options
244 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
245 @gccoptlist{-fmessage-length=@var{n} @gol
246 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
247 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
248 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
250 @item Warning Options
251 @xref{Warning Options,,Options to Request or Suppress Warnings}.
252 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
253 -pedantic-errors @gol
254 -w -Wextra -Wall -Waddress -Waggregate-return @gol
255 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
256 -Wno-attributes -Wbool-compare -Wno-builtin-macro-redefined @gol
257 -Wc90-c99-compat -Wc99-c11-compat @gol
258 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
259 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
260 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdate-time -Wdelete-incomplete @gol
261 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
262 -Wdisabled-optimization @gol
263 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
264 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
265 -Wempty-body -Wenum-compare -Wno-endif-labels @gol
266 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
267 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
268 -Wformat-security -Wformat-signedness -Wformat-y2k -Wframe-address @gol
269 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
270 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
271 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
272 -Winit-self -Winline -Wno-int-conversion @gol
273 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
274 -Winvalid-pch -Wlarger-than=@var{len} @gol
275 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
276 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args @gol
277 -Wmisleading-indentation -Wmissing-braces @gol
278 -Wmissing-field-initializers -Wmissing-include-dirs @gol
279 -Wno-multichar -Wnonnull -Wnonnull-compare @gol
280 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
281 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
282 -Woverride-init-side-effects -Woverlength-strings @gol
283 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
284 -Wparentheses -Wno-pedantic-ms-format @gol
285 -Wplacement-new -Wplacement-new=@var{n} @gol
286 -Wpointer-arith -Wno-pointer-to-int-cast @gol
287 -Wno-pragmas -Wredundant-decls -Wno-return-local-addr @gol
288 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
289 -Wshift-overflow -Wshift-overflow=@var{n} @gol
290 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
291 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
292 -Wno-scalar-storage-order @gol
293 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
294 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
295 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
296 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
297 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
298 -Wmissing-format-attribute -Wsubobject-linkage @gol
299 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
300 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
301 -Wtype-limits -Wundef @gol
302 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
303 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
304 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
305 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
306 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
307 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
308 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
309 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
310 -Wzero-as-null-pointer-constant -Whsa}
312 @item C and Objective-C-only Warning Options
313 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
314 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
315 -Wold-style-declaration -Wold-style-definition @gol
316 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
317 -Wdeclaration-after-statement -Wpointer-sign}
319 @item Debugging Options
320 @xref{Debugging Options,,Options for Debugging Your Program}.
321 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
322 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
323 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
324 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
325 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
326 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
327 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
328 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
329 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
330 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
331 -fvar-tracking -fvar-tracking-assignments}
333 @item Optimization Options
334 @xref{Optimize Options,,Options that Control Optimization}.
335 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
336 -falign-jumps[=@var{n}] @gol
337 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
338 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
339 -fauto-inc-dec -fbranch-probabilities @gol
340 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
341 -fbtr-bb-exclusive -fcaller-saves @gol
342 -fcombine-stack-adjustments -fconserve-stack @gol
343 -fcompare-elim -fcprop-registers -fcrossjumping @gol
344 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
345 -fcx-limited-range @gol
346 -fdata-sections -fdce -fdelayed-branch @gol
347 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
348 -fdevirtualize-at-ltrans -fdse @gol
349 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
350 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
351 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
352 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
353 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
354 -fif-conversion2 -findirect-inlining @gol
355 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
356 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
357 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
358 -fira-algorithm=@var{algorithm} @gol
359 -fira-region=@var{region} -fira-hoist-pressure @gol
360 -fira-loop-pressure -fno-ira-share-save-slots @gol
361 -fno-ira-share-spill-slots @gol
362 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
363 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
364 -fkeep-static-consts -flive-range-shrinkage @gol
365 -floop-block -floop-interchange -floop-strip-mine @gol
366 -floop-unroll-and-jam -floop-nest-optimize @gol
367 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
368 -flto-partition=@var{alg} -fmerge-all-constants @gol
369 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
370 -fmove-loop-invariants -fno-branch-count-reg @gol
371 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
372 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
373 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
374 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
375 -fomit-frame-pointer -foptimize-sibling-calls @gol
376 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
377 -fprefetch-loop-arrays @gol
378 -fprofile-correction @gol
379 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
380 -fprofile-reorder-functions @gol
381 -freciprocal-math -free -frename-registers -freorder-blocks @gol
382 -freorder-blocks-algorithm=@var{algorithm} @gol
383 -freorder-blocks-and-partition -freorder-functions @gol
384 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
385 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
386 -fsched-spec-load -fsched-spec-load-dangerous @gol
387 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
388 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
389 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
390 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
391 -fschedule-fusion @gol
392 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
393 -fselective-scheduling -fselective-scheduling2 @gol
394 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
395 -fsemantic-interposition -fshrink-wrap -fsignaling-nans @gol
396 -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
398 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
399 -fstdarg-opt -fstrict-aliasing @gol
400 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
401 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
402 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
403 -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
404 -ftree-loop-if-convert-stores -ftree-loop-im @gol
405 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
406 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
407 -ftree-loop-vectorize @gol
408 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
409 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
410 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
411 -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
412 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
413 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
414 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
415 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
416 --param @var{name}=@var{value}
417 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
419 @item Program Instrumentation Options
420 @xref{Instrumentation Options,,Program Instrumentation Options}.
421 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
422 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
423 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
424 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
425 -fsanitize-undefined-trap-on-error -fbounds-check @gol
426 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
427 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
428 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
429 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
430 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
431 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
432 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
433 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
434 -fchkp-use-wrappers @gol
435 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
436 -fstack-protector-explicit -fstack-check @gol
437 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
438 -fno-stack-limit -fsplit-stack @gol
439 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
440 -fvtv-counts -fvtv-debug @gol
441 -finstrument-functions @gol
442 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
443 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
445 @item Preprocessor Options
446 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
447 @gccoptlist{-A@var{question}=@var{answer} @gol
448 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
449 -C -dD -dI -dM -dN @gol
450 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
451 -idirafter @var{dir} @gol
452 -include @var{file} -imacros @var{file} @gol
453 -iprefix @var{file} -iwithprefix @var{dir} @gol
454 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
455 -imultilib @var{dir} -isysroot @var{dir} @gol
456 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
457 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
458 -remap -trigraphs -undef -U@var{macro} @gol
459 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
461 @item Assembler Option
462 @xref{Assembler Options,,Passing Options to the Assembler}.
463 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
466 @xref{Link Options,,Options for Linking}.
467 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
468 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
469 -s -static -static-libgcc -static-libstdc++ @gol
470 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
471 -static-libmpx -static-libmpxwrappers @gol
472 -shared -shared-libgcc -symbolic @gol
473 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
474 -u @var{symbol} -z @var{keyword}}
476 @item Directory Options
477 @xref{Directory Options,,Options for Directory Search}.
478 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
479 -iquote@var{dir} -L@var{dir} -no-canonical-prefixes -I- @gol
480 --sysroot=@var{dir} --no-sysroot-suffix}
482 @item Code Generation Options
483 @xref{Code Gen Options,,Options for Code Generation Conventions}.
484 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
485 -ffixed-@var{reg} -fexceptions @gol
486 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
487 -fasynchronous-unwind-tables @gol
489 -finhibit-size-directive -fno-common -fno-ident @gol
490 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
491 -fno-jump-tables @gol
492 -frecord-gcc-switches @gol
493 -freg-struct-return -fshort-enums -fshort-wchar @gol
494 -fverbose-asm -fpack-struct[=@var{n}] @gol
495 -fleading-underscore -ftls-model=@var{model} @gol
496 -fstack-reuse=@var{reuse_level} @gol
498 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
499 -fstrict-volatile-bitfields -fsync-libcalls}
501 @item Developer Options
502 @xref{Developer Options,,GCC Developer Options}.
503 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
504 -fchecking -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
505 -fdisable-ipa-@var{pass_name} @gol
506 -fdisable-rtl-@var{pass_name} @gol
507 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
508 -fdisable-tree-@var{pass_name} @gol
509 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
510 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
511 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
512 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
513 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
515 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
516 -fdump-statistics @gol
518 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
519 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
520 -fdump-tree-cfg -fdump-tree-alias @gol
522 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
523 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
524 -fdump-tree-gimple@r{[}-raw@r{]} @gol
525 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
526 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
527 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
528 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
529 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
530 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
531 -fdump-tree-nrv -fdump-tree-vect @gol
532 -fdump-tree-sink @gol
533 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
534 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
535 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
536 -fdump-tree-vtable-verify @gol
537 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
538 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
539 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
540 -fdump-final-insns=@var{file} @gol
541 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
542 -fenable-@var{kind}-@var{pass} @gol
543 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
544 -fira-verbose=@var{n} @gol
545 -flto-report -flto-report-wpa -fmem-report-wpa @gol
546 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
547 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
548 -fprofile-report @gol
549 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
550 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
551 -fstats -fstack-usage -ftime-report @gol
552 -fvar-tracking-assignments-toggle -gtoggle @gol
553 -print-file-name=@var{library} -print-libgcc-file-name @gol
554 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
555 -print-prog-name=@var{program} -print-search-dirs -Q @gol
556 -print-sysroot -print-sysroot-headers-suffix @gol
557 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
559 @item Machine-Dependent Options
560 @xref{Submodel Options,,Machine-Dependent Options}.
561 @c This list is ordered alphanumerically by subsection name.
562 @c Try and put the significant identifier (CPU or system) first,
563 @c so users have a clue at guessing where the ones they want will be.
565 @emph{AArch64 Options}
566 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
567 -mgeneral-regs-only @gol
568 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
570 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
571 -mtls-dialect=desc -mtls-dialect=traditional @gol
572 -mtls-size=@var{size} @gol
573 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
574 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
575 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
576 -mlow-precision-sqrt -mno-low-precision-sqrt@gol
577 -mlow-precision-div -mno-low-precision-div @gol
578 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
580 @emph{Adapteva Epiphany Options}
581 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
582 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
583 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
584 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
585 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
586 -msplit-vecmove-early -m1reg-@var{reg}}
589 @gccoptlist{-mbarrel-shifter @gol
590 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
591 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
592 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
593 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
594 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
595 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
596 -mlong-calls -mmedium-calls -msdata @gol
597 -mucb-mcount -mvolatile-cache @gol
598 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
599 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
600 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
601 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
602 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
603 -mtune=@var{cpu} -mmultcost=@var{num} @gol
604 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
605 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
608 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
609 -mabi=@var{name} @gol
610 -mapcs-stack-check -mno-apcs-stack-check @gol
611 -mapcs-float -mno-apcs-float @gol
612 -mapcs-reentrant -mno-apcs-reentrant @gol
613 -msched-prolog -mno-sched-prolog @gol
614 -mlittle-endian -mbig-endian @gol
615 -mfloat-abi=@var{name} @gol
616 -mfp16-format=@var{name}
617 -mthumb-interwork -mno-thumb-interwork @gol
618 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
619 -mtune=@var{name} -mprint-tune-info @gol
620 -mstructure-size-boundary=@var{n} @gol
621 -mabort-on-noreturn @gol
622 -mlong-calls -mno-long-calls @gol
623 -msingle-pic-base -mno-single-pic-base @gol
624 -mpic-register=@var{reg} @gol
625 -mnop-fun-dllimport @gol
626 -mpoke-function-name @gol
628 -mtpcs-frame -mtpcs-leaf-frame @gol
629 -mcaller-super-interworking -mcallee-super-interworking @gol
630 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
631 -mword-relocations @gol
632 -mfix-cortex-m3-ldrd @gol
633 -munaligned-access @gol
634 -mneon-for-64bits @gol
635 -mslow-flash-data @gol
636 -masm-syntax-unified @gol
641 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
642 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
643 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
645 @emph{Blackfin Options}
646 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
647 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
648 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
649 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
650 -mno-id-shared-library -mshared-library-id=@var{n} @gol
651 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
652 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
653 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
657 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
658 -msim -msdata=@var{sdata-type}}
661 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
662 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
663 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
664 -mstack-align -mdata-align -mconst-align @gol
665 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
666 -melf -maout -melinux -mlinux -sim -sim2 @gol
667 -mmul-bug-workaround -mno-mul-bug-workaround}
670 @gccoptlist{-mmac @gol
671 -mcr16cplus -mcr16c @gol
672 -msim -mint32 -mbit-ops
673 -mdata-model=@var{model}}
675 @emph{Darwin Options}
676 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
677 -arch_only -bind_at_load -bundle -bundle_loader @gol
678 -client_name -compatibility_version -current_version @gol
680 -dependency-file -dylib_file -dylinker_install_name @gol
681 -dynamic -dynamiclib -exported_symbols_list @gol
682 -filelist -flat_namespace -force_cpusubtype_ALL @gol
683 -force_flat_namespace -headerpad_max_install_names @gol
685 -image_base -init -install_name -keep_private_externs @gol
686 -multi_module -multiply_defined -multiply_defined_unused @gol
687 -noall_load -no_dead_strip_inits_and_terms @gol
688 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
689 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
690 -private_bundle -read_only_relocs -sectalign @gol
691 -sectobjectsymbols -whyload -seg1addr @gol
692 -sectcreate -sectobjectsymbols -sectorder @gol
693 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
694 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
695 -segprot -segs_read_only_addr -segs_read_write_addr @gol
696 -single_module -static -sub_library -sub_umbrella @gol
697 -twolevel_namespace -umbrella -undefined @gol
698 -unexported_symbols_list -weak_reference_mismatches @gol
699 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
700 -mkernel -mone-byte-bool}
702 @emph{DEC Alpha Options}
703 @gccoptlist{-mno-fp-regs -msoft-float @gol
704 -mieee -mieee-with-inexact -mieee-conformant @gol
705 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
706 -mtrap-precision=@var{mode} -mbuild-constants @gol
707 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
708 -mbwx -mmax -mfix -mcix @gol
709 -mfloat-vax -mfloat-ieee @gol
710 -mexplicit-relocs -msmall-data -mlarge-data @gol
711 -msmall-text -mlarge-text @gol
712 -mmemory-latency=@var{time}}
715 @gccoptlist{-msmall-model -mno-lsim}
718 @gccoptlist{-msim -mlra -mnodiv}
721 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
722 -mhard-float -msoft-float @gol
723 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
724 -mdouble -mno-double @gol
725 -mmedia -mno-media -mmuladd -mno-muladd @gol
726 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
727 -mlinked-fp -mlong-calls -malign-labels @gol
728 -mlibrary-pic -macc-4 -macc-8 @gol
729 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
730 -moptimize-membar -mno-optimize-membar @gol
731 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
732 -mvliw-branch -mno-vliw-branch @gol
733 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
734 -mno-nested-cond-exec -mtomcat-stats @gol
738 @emph{GNU/Linux Options}
739 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
740 -tno-android-cc -tno-android-ld}
742 @emph{H8/300 Options}
743 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
746 @gccoptlist{-march=@var{architecture-type} @gol
747 -mdisable-fpregs -mdisable-indexing @gol
748 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
749 -mfixed-range=@var{register-range} @gol
750 -mjump-in-delay -mlinker-opt -mlong-calls @gol
751 -mlong-load-store -mno-disable-fpregs @gol
752 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
753 -mno-jump-in-delay -mno-long-load-store @gol
754 -mno-portable-runtime -mno-soft-float @gol
755 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
756 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
757 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
758 -munix=@var{unix-std} -nolibdld -static -threads}
761 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
762 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
763 -mconstant-gp -mauto-pic -mfused-madd @gol
764 -minline-float-divide-min-latency @gol
765 -minline-float-divide-max-throughput @gol
766 -mno-inline-float-divide @gol
767 -minline-int-divide-min-latency @gol
768 -minline-int-divide-max-throughput @gol
769 -mno-inline-int-divide @gol
770 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
771 -mno-inline-sqrt @gol
772 -mdwarf2-asm -mearly-stop-bits @gol
773 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
774 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
775 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
776 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
777 -msched-spec-ldc -msched-spec-control-ldc @gol
778 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
779 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
780 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
781 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
784 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
785 -msign-extend-enabled -muser-enabled}
787 @emph{M32R/D Options}
788 @gccoptlist{-m32r2 -m32rx -m32r @gol
790 -malign-loops -mno-align-loops @gol
791 -missue-rate=@var{number} @gol
792 -mbranch-cost=@var{number} @gol
793 -mmodel=@var{code-size-model-type} @gol
794 -msdata=@var{sdata-type} @gol
795 -mno-flush-func -mflush-func=@var{name} @gol
796 -mno-flush-trap -mflush-trap=@var{number} @gol
800 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
802 @emph{M680x0 Options}
803 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
804 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
805 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
806 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
807 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
808 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
809 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
810 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
814 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
815 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
816 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
817 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
818 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
821 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
822 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
823 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
824 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
827 @emph{MicroBlaze Options}
828 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
829 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
830 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
831 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
832 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
835 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
836 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
837 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
838 -mips16 -mno-mips16 -mflip-mips16 @gol
839 -minterlink-compressed -mno-interlink-compressed @gol
840 -minterlink-mips16 -mno-interlink-mips16 @gol
841 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
842 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
843 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
844 -mno-float -msingle-float -mdouble-float @gol
845 -modd-spreg -mno-odd-spreg @gol
846 -mabs=@var{mode} -mnan=@var{encoding} @gol
847 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
850 -mvirt -mno-virt @gol
852 -mmicromips -mno-micromips @gol
853 -mfpu=@var{fpu-type} @gol
854 -msmartmips -mno-smartmips @gol
855 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
856 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
857 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
858 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
859 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
860 -membedded-data -mno-embedded-data @gol
861 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
862 -mcode-readable=@var{setting} @gol
863 -msplit-addresses -mno-split-addresses @gol
864 -mexplicit-relocs -mno-explicit-relocs @gol
865 -mcheck-zero-division -mno-check-zero-division @gol
866 -mdivide-traps -mdivide-breaks @gol
867 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
868 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
869 -mfix-24k -mno-fix-24k @gol
870 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
871 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
872 -mfix-vr4120 -mno-fix-vr4120 @gol
873 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
874 -mflush-func=@var{func} -mno-flush-func @gol
875 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
876 -mcompact-branches=@var{policy} @gol
877 -mfp-exceptions -mno-fp-exceptions @gol
878 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
879 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
880 -mframe-header-opt -mno-frame-header-opt}
883 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
884 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
885 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
886 -mno-base-addresses -msingle-exit -mno-single-exit}
888 @emph{MN10300 Options}
889 @gccoptlist{-mmult-bug -mno-mult-bug @gol
890 -mno-am33 -mam33 -mam33-2 -mam34 @gol
891 -mtune=@var{cpu-type} @gol
892 -mreturn-pointer-on-d0 @gol
893 -mno-crt0 -mrelax -mliw -msetlb}
896 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
898 @emph{MSP430 Options}
899 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
901 -mcode-region= -mdata-region= @gol
902 -msilicon-errata= -msilicon-errata-warn= @gol
906 @gccoptlist{-mbig-endian -mlittle-endian @gol
907 -mreduced-regs -mfull-regs @gol
908 -mcmov -mno-cmov @gol
909 -mperf-ext -mno-perf-ext @gol
910 -mv3push -mno-v3push @gol
911 -m16bit -mno-16bit @gol
912 -misr-vector-size=@var{num} @gol
913 -mcache-block-size=@var{num} @gol
914 -march=@var{arch} @gol
915 -mcmodel=@var{code-model} @gol
918 @emph{Nios II Options}
919 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
921 -mno-bypass-cache -mbypass-cache @gol
922 -mno-cache-volatile -mcache-volatile @gol
923 -mno-fast-sw-div -mfast-sw-div @gol
924 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
925 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
926 -mcustom-fpu-cfg=@var{name} @gol
927 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
928 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
930 @emph{Nvidia PTX Options}
931 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
933 @emph{PDP-11 Options}
934 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
935 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
936 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
937 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
938 -mbranch-expensive -mbranch-cheap @gol
939 -munix-asm -mdec-asm}
941 @emph{picoChip Options}
942 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
943 -msymbol-as-address -mno-inefficient-warnings}
945 @emph{PowerPC Options}
946 See RS/6000 and PowerPC Options.
949 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
950 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
951 -m64bit-doubles -m32bit-doubles}
953 @emph{RS/6000 and PowerPC Options}
954 @gccoptlist{-mcpu=@var{cpu-type} @gol
955 -mtune=@var{cpu-type} @gol
956 -mcmodel=@var{code-model} @gol
958 -maltivec -mno-altivec @gol
959 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
960 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
961 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
962 -mfprnd -mno-fprnd @gol
963 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
964 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
965 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
966 -malign-power -malign-natural @gol
967 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
968 -msingle-float -mdouble-float -msimple-fpu @gol
969 -mstring -mno-string -mupdate -mno-update @gol
970 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
971 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
972 -mstrict-align -mno-strict-align -mrelocatable @gol
973 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
974 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
975 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
976 -mprioritize-restricted-insns=@var{priority} @gol
977 -msched-costly-dep=@var{dependence_type} @gol
978 -minsert-sched-nops=@var{scheme} @gol
979 -mcall-sysv -mcall-netbsd @gol
980 -maix-struct-return -msvr4-struct-return @gol
981 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
982 -mblock-move-inline-limit=@var{num} @gol
983 -misel -mno-isel @gol
984 -misel=yes -misel=no @gol
986 -mspe=yes -mspe=no @gol
988 -mgen-cell-microcode -mwarn-cell-microcode @gol
989 -mvrsave -mno-vrsave @gol
990 -mmulhw -mno-mulhw @gol
991 -mdlmzb -mno-dlmzb @gol
992 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
993 -mprototype -mno-prototype @gol
994 -msim -mmvme -mads -myellowknife -memb -msdata @gol
995 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
996 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
997 -mno-recip-precision @gol
998 -mveclibabi=@var{type} -mfriz -mno-friz @gol
999 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1000 -msave-toc-indirect -mno-save-toc-indirect @gol
1001 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1002 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1003 -mquad-memory -mno-quad-memory @gol
1004 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1005 -mcompat-align-parm -mno-compat-align-parm @gol
1006 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1007 -mupper-regs -mno-upper-regs @gol
1008 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1012 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1014 -mbig-endian-data -mlittle-endian-data @gol
1017 -mas100-syntax -mno-as100-syntax@gol
1019 -mmax-constant-size=@gol
1022 -mallow-string-insns -mno-allow-string-insns@gol
1024 -mno-warn-multiple-fast-interrupts@gol
1025 -msave-acc-in-interrupts}
1027 @emph{S/390 and zSeries Options}
1028 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1029 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1030 -mlong-double-64 -mlong-double-128 @gol
1031 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1032 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1033 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1034 -mhtm -mvx -mzvector @gol
1035 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1036 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1037 -mhotpatch=@var{halfwords},@var{halfwords}}
1039 @emph{Score Options}
1040 @gccoptlist{-meb -mel @gol
1044 -mscore5 -mscore5u -mscore7 -mscore7d}
1047 @gccoptlist{-m1 -m2 -m2e @gol
1048 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1050 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1051 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1052 -mb -ml -mdalign -mrelax @gol
1053 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1054 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1055 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1056 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1057 -maccumulate-outgoing-args @gol
1058 -matomic-model=@var{atomic-model} @gol
1059 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1060 -mcbranch-force-delay-slot @gol
1061 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1062 -mpretend-cmove -mtas}
1064 @emph{Solaris 2 Options}
1065 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1068 @emph{SPARC Options}
1069 @gccoptlist{-mcpu=@var{cpu-type} @gol
1070 -mtune=@var{cpu-type} @gol
1071 -mcmodel=@var{code-model} @gol
1072 -mmemory-model=@var{mem-model} @gol
1073 -m32 -m64 -mapp-regs -mno-app-regs @gol
1074 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1075 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1076 -mhard-quad-float -msoft-quad-float @gol
1077 -mstack-bias -mno-stack-bias @gol
1078 -mstd-struct-return -mno-std-struct-return @gol
1079 -munaligned-doubles -mno-unaligned-doubles @gol
1080 -muser-mode -mno-user-mode @gol
1081 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1082 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1083 -mcbcond -mno-cbcond @gol
1084 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1085 -mfix-at697f -mfix-ut699}
1088 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1089 -msafe-dma -munsafe-dma @gol
1091 -msmall-mem -mlarge-mem -mstdmain @gol
1092 -mfixed-range=@var{register-range} @gol
1094 -maddress-space-conversion -mno-address-space-conversion @gol
1095 -mcache-size=@var{cache-size} @gol
1096 -matomic-updates -mno-atomic-updates}
1098 @emph{System V Options}
1099 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1101 @emph{TILE-Gx Options}
1102 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1103 -mcmodel=@var{code-model}}
1105 @emph{TILEPro Options}
1106 @gccoptlist{-mcpu=@var{cpu} -m32}
1109 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1110 -mprolog-function -mno-prolog-function -mspace @gol
1111 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1112 -mapp-regs -mno-app-regs @gol
1113 -mdisable-callt -mno-disable-callt @gol
1114 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1115 -mv850e -mv850 -mv850e3v5 @gol
1126 @gccoptlist{-mg -mgnu -munix}
1128 @emph{Visium Options}
1129 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1130 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1133 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1134 -mpointer-size=@var{size}}
1136 @emph{VxWorks Options}
1137 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1138 -Xbind-lazy -Xbind-now}
1141 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1142 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1143 -mfpmath=@var{unit} @gol
1144 -masm=@var{dialect} -mno-fancy-math-387 @gol
1145 -mno-fp-ret-in-387 -msoft-float @gol
1146 -mno-wide-multiply -mrtd -malign-double @gol
1147 -mpreferred-stack-boundary=@var{num} @gol
1148 -mincoming-stack-boundary=@var{num} @gol
1149 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1150 -mrecip -mrecip=@var{opt} @gol
1151 -mvzeroupper -mprefer-avx128 @gol
1152 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1153 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1154 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1155 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1156 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1157 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1158 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1159 -mpku -mthreads @gol
1160 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1161 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1162 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1163 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1164 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1165 -mregparm=@var{num} -msseregparm @gol
1166 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1167 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1168 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1169 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1170 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1171 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1172 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1173 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1176 @emph{x86 Windows Options}
1177 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1178 -mnop-fun-dllimport -mthread @gol
1179 -municode -mwin32 -mwindows -fno-set-stack-executable}
1181 @emph{Xstormy16 Options}
1184 @emph{Xtensa Options}
1185 @gccoptlist{-mconst16 -mno-const16 @gol
1186 -mfused-madd -mno-fused-madd @gol
1188 -mserialize-volatile -mno-serialize-volatile @gol
1189 -mtext-section-literals -mno-text-section-literals @gol
1190 -mauto-litpools -mno-auto-litpools @gol
1191 -mtarget-align -mno-target-align @gol
1192 -mlongcalls -mno-longcalls}
1194 @emph{zSeries Options}
1195 See S/390 and zSeries Options.
1199 @node Overall Options
1200 @section Options Controlling the Kind of Output
1202 Compilation can involve up to four stages: preprocessing, compilation
1203 proper, assembly and linking, always in that order. GCC is capable of
1204 preprocessing and compiling several files either into several
1205 assembler input files, or into one assembler input file; then each
1206 assembler input file produces an object file, and linking combines all
1207 the object files (those newly compiled, and those specified as input)
1208 into an executable file.
1210 @cindex file name suffix
1211 For any given input file, the file name suffix determines what kind of
1212 compilation is done:
1216 C source code that must be preprocessed.
1219 C source code that should not be preprocessed.
1222 C++ source code that should not be preprocessed.
1225 Objective-C source code. Note that you must link with the @file{libobjc}
1226 library to make an Objective-C program work.
1229 Objective-C source code that should not be preprocessed.
1233 Objective-C++ source code. Note that you must link with the @file{libobjc}
1234 library to make an Objective-C++ program work. Note that @samp{.M} refers
1235 to a literal capital M@.
1237 @item @var{file}.mii
1238 Objective-C++ source code that should not be preprocessed.
1241 C, C++, Objective-C or Objective-C++ header file to be turned into a
1242 precompiled header (default), or C, C++ header file to be turned into an
1243 Ada spec (via the @option{-fdump-ada-spec} switch).
1246 @itemx @var{file}.cp
1247 @itemx @var{file}.cxx
1248 @itemx @var{file}.cpp
1249 @itemx @var{file}.CPP
1250 @itemx @var{file}.c++
1252 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1253 the last two letters must both be literally @samp{x}. Likewise,
1254 @samp{.C} refers to a literal capital C@.
1258 Objective-C++ source code that must be preprocessed.
1260 @item @var{file}.mii
1261 Objective-C++ source code that should not be preprocessed.
1265 @itemx @var{file}.hp
1266 @itemx @var{file}.hxx
1267 @itemx @var{file}.hpp
1268 @itemx @var{file}.HPP
1269 @itemx @var{file}.h++
1270 @itemx @var{file}.tcc
1271 C++ header file to be turned into a precompiled header or Ada spec.
1274 @itemx @var{file}.for
1275 @itemx @var{file}.ftn
1276 Fixed form Fortran source code that should not be preprocessed.
1279 @itemx @var{file}.FOR
1280 @itemx @var{file}.fpp
1281 @itemx @var{file}.FPP
1282 @itemx @var{file}.FTN
1283 Fixed form Fortran source code that must be preprocessed (with the traditional
1286 @item @var{file}.f90
1287 @itemx @var{file}.f95
1288 @itemx @var{file}.f03
1289 @itemx @var{file}.f08
1290 Free form Fortran source code that should not be preprocessed.
1292 @item @var{file}.F90
1293 @itemx @var{file}.F95
1294 @itemx @var{file}.F03
1295 @itemx @var{file}.F08
1296 Free form Fortran source code that must be preprocessed (with the
1297 traditional preprocessor).
1302 @c FIXME: Descriptions of Java file types.
1308 @item @var{file}.ads
1309 Ada source code file that contains a library unit declaration (a
1310 declaration of a package, subprogram, or generic, or a generic
1311 instantiation), or a library unit renaming declaration (a package,
1312 generic, or subprogram renaming declaration). Such files are also
1315 @item @var{file}.adb
1316 Ada source code file containing a library unit body (a subprogram or
1317 package body). Such files are also called @dfn{bodies}.
1319 @c GCC also knows about some suffixes for languages not yet included:
1330 @itemx @var{file}.sx
1331 Assembler code that must be preprocessed.
1334 An object file to be fed straight into linking.
1335 Any file name with no recognized suffix is treated this way.
1339 You can specify the input language explicitly with the @option{-x} option:
1342 @item -x @var{language}
1343 Specify explicitly the @var{language} for the following input files
1344 (rather than letting the compiler choose a default based on the file
1345 name suffix). This option applies to all following input files until
1346 the next @option{-x} option. Possible values for @var{language} are:
1348 c c-header cpp-output
1349 c++ c++-header c++-cpp-output
1350 objective-c objective-c-header objective-c-cpp-output
1351 objective-c++ objective-c++-header objective-c++-cpp-output
1352 assembler assembler-with-cpp
1354 f77 f77-cpp-input f95 f95-cpp-input
1360 Turn off any specification of a language, so that subsequent files are
1361 handled according to their file name suffixes (as they are if @option{-x}
1362 has not been used at all).
1365 If you only want some of the stages of compilation, you can use
1366 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1367 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1368 @command{gcc} is to stop. Note that some combinations (for example,
1369 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1374 Compile or assemble the source files, but do not link. The linking
1375 stage simply is not done. The ultimate output is in the form of an
1376 object file for each source file.
1378 By default, the object file name for a source file is made by replacing
1379 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1381 Unrecognized input files, not requiring compilation or assembly, are
1386 Stop after the stage of compilation proper; do not assemble. The output
1387 is in the form of an assembler code file for each non-assembler input
1390 By default, the assembler file name for a source file is made by
1391 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1393 Input files that don't require compilation are ignored.
1397 Stop after the preprocessing stage; do not run the compiler proper. The
1398 output is in the form of preprocessed source code, which is sent to the
1401 Input files that don't require preprocessing are ignored.
1403 @cindex output file option
1406 Place output in file @var{file}. This applies to whatever
1407 sort of output is being produced, whether it be an executable file,
1408 an object file, an assembler file or preprocessed C code.
1410 If @option{-o} is not specified, the default is to put an executable
1411 file in @file{a.out}, the object file for
1412 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1413 assembler file in @file{@var{source}.s}, a precompiled header file in
1414 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1419 Print (on standard error output) the commands executed to run the stages
1420 of compilation. Also print the version number of the compiler driver
1421 program and of the preprocessor and the compiler proper.
1425 Like @option{-v} except the commands are not executed and arguments
1426 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1427 This is useful for shell scripts to capture the driver-generated command lines.
1431 Print (on the standard output) a description of the command-line options
1432 understood by @command{gcc}. If the @option{-v} option is also specified
1433 then @option{--help} is also passed on to the various processes
1434 invoked by @command{gcc}, so that they can display the command-line options
1435 they accept. If the @option{-Wextra} option has also been specified
1436 (prior to the @option{--help} option), then command-line options that
1437 have no documentation associated with them are also displayed.
1440 @opindex target-help
1441 Print (on the standard output) a description of target-specific command-line
1442 options for each tool. For some targets extra target-specific
1443 information may also be printed.
1445 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1446 Print (on the standard output) a description of the command-line
1447 options understood by the compiler that fit into all specified classes
1448 and qualifiers. These are the supported classes:
1451 @item @samp{optimizers}
1452 Display all of the optimization options supported by the
1455 @item @samp{warnings}
1456 Display all of the options controlling warning messages
1457 produced by the compiler.
1460 Display target-specific options. Unlike the
1461 @option{--target-help} option however, target-specific options of the
1462 linker and assembler are not displayed. This is because those
1463 tools do not currently support the extended @option{--help=} syntax.
1466 Display the values recognized by the @option{--param}
1469 @item @var{language}
1470 Display the options supported for @var{language}, where
1471 @var{language} is the name of one of the languages supported in this
1475 Display the options that are common to all languages.
1478 These are the supported qualifiers:
1481 @item @samp{undocumented}
1482 Display only those options that are undocumented.
1485 Display options taking an argument that appears after an equal
1486 sign in the same continuous piece of text, such as:
1487 @samp{--help=target}.
1489 @item @samp{separate}
1490 Display options taking an argument that appears as a separate word
1491 following the original option, such as: @samp{-o output-file}.
1494 Thus for example to display all the undocumented target-specific
1495 switches supported by the compiler, use:
1498 --help=target,undocumented
1501 The sense of a qualifier can be inverted by prefixing it with the
1502 @samp{^} character, so for example to display all binary warning
1503 options (i.e., ones that are either on or off and that do not take an
1504 argument) that have a description, use:
1507 --help=warnings,^joined,^undocumented
1510 The argument to @option{--help=} should not consist solely of inverted
1513 Combining several classes is possible, although this usually
1514 restricts the output so much that there is nothing to display. One
1515 case where it does work, however, is when one of the classes is
1516 @var{target}. For example, to display all the target-specific
1517 optimization options, use:
1520 --help=target,optimizers
1523 The @option{--help=} option can be repeated on the command line. Each
1524 successive use displays its requested class of options, skipping
1525 those that have already been displayed.
1527 If the @option{-Q} option appears on the command line before the
1528 @option{--help=} option, then the descriptive text displayed by
1529 @option{--help=} is changed. Instead of describing the displayed
1530 options, an indication is given as to whether the option is enabled,
1531 disabled or set to a specific value (assuming that the compiler
1532 knows this at the point where the @option{--help=} option is used).
1534 Here is a truncated example from the ARM port of @command{gcc}:
1537 % gcc -Q -mabi=2 --help=target -c
1538 The following options are target specific:
1540 -mabort-on-noreturn [disabled]
1544 The output is sensitive to the effects of previous command-line
1545 options, so for example it is possible to find out which optimizations
1546 are enabled at @option{-O2} by using:
1549 -Q -O2 --help=optimizers
1552 Alternatively you can discover which binary optimizations are enabled
1553 by @option{-O3} by using:
1556 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1557 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1558 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1563 Display the version number and copyrights of the invoked GCC@.
1565 @item -pass-exit-codes
1566 @opindex pass-exit-codes
1567 Normally the @command{gcc} program exits with the code of 1 if any
1568 phase of the compiler returns a non-success return code. If you specify
1569 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1570 the numerically highest error produced by any phase returning an error
1571 indication. The C, C++, and Fortran front ends return 4 if an internal
1572 compiler error is encountered.
1576 Use pipes rather than temporary files for communication between the
1577 various stages of compilation. This fails to work on some systems where
1578 the assembler is unable to read from a pipe; but the GNU assembler has
1581 @item -specs=@var{file}
1583 Process @var{file} after the compiler reads in the standard @file{specs}
1584 file, in order to override the defaults which the @command{gcc} driver
1585 program uses when determining what switches to pass to @command{cc1},
1586 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1587 @option{-specs=@var{file}} can be specified on the command line, and they
1588 are processed in order, from left to right. @xref{Spec Files}, for
1589 information about the format of the @var{file}.
1593 Invoke all subcommands under a wrapper program. The name of the
1594 wrapper program and its parameters are passed as a comma separated
1598 gcc -c t.c -wrapper gdb,--args
1602 This invokes all subprograms of @command{gcc} under
1603 @samp{gdb --args}, thus the invocation of @command{cc1} is
1604 @samp{gdb --args cc1 @dots{}}.
1606 @item -fplugin=@var{name}.so
1608 Load the plugin code in file @var{name}.so, assumed to be a
1609 shared object to be dlopen'd by the compiler. The base name of
1610 the shared object file is used to identify the plugin for the
1611 purposes of argument parsing (See
1612 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1613 Each plugin should define the callback functions specified in the
1616 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1617 @opindex fplugin-arg
1618 Define an argument called @var{key} with a value of @var{value}
1619 for the plugin called @var{name}.
1621 @item -fdump-ada-spec@r{[}-slim@r{]}
1622 @opindex fdump-ada-spec
1623 For C and C++ source and include files, generate corresponding Ada specs.
1624 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1625 GNAT User's Guide}, which provides detailed documentation on this feature.
1627 @item -fada-spec-parent=@var{unit}
1628 @opindex fada-spec-parent
1629 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1630 Ada specs as child units of parent @var{unit}.
1632 @item -fdump-go-spec=@var{file}
1633 @opindex fdump-go-spec
1634 For input files in any language, generate corresponding Go
1635 declarations in @var{file}. This generates Go @code{const},
1636 @code{type}, @code{var}, and @code{func} declarations which may be a
1637 useful way to start writing a Go interface to code written in some
1640 @include @value{srcdir}/../libiberty/at-file.texi
1644 @section Compiling C++ Programs
1646 @cindex suffixes for C++ source
1647 @cindex C++ source file suffixes
1648 C++ source files conventionally use one of the suffixes @samp{.C},
1649 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1650 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1651 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1652 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1653 files with these names and compiles them as C++ programs even if you
1654 call the compiler the same way as for compiling C programs (usually
1655 with the name @command{gcc}).
1659 However, the use of @command{gcc} does not add the C++ library.
1660 @command{g++} is a program that calls GCC and automatically specifies linking
1661 against the C++ library. It treats @samp{.c},
1662 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1663 files unless @option{-x} is used. This program is also useful when
1664 precompiling a C header file with a @samp{.h} extension for use in C++
1665 compilations. On many systems, @command{g++} is also installed with
1666 the name @command{c++}.
1668 @cindex invoking @command{g++}
1669 When you compile C++ programs, you may specify many of the same
1670 command-line options that you use for compiling programs in any
1671 language; or command-line options meaningful for C and related
1672 languages; or options that are meaningful only for C++ programs.
1673 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1674 explanations of options for languages related to C@.
1675 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1676 explanations of options that are meaningful only for C++ programs.
1678 @node C Dialect Options
1679 @section Options Controlling C Dialect
1680 @cindex dialect options
1681 @cindex language dialect options
1682 @cindex options, dialect
1684 The following options control the dialect of C (or languages derived
1685 from C, such as C++, Objective-C and Objective-C++) that the compiler
1689 @cindex ANSI support
1693 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1694 equivalent to @option{-std=c++98}.
1696 This turns off certain features of GCC that are incompatible with ISO
1697 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1698 such as the @code{asm} and @code{typeof} keywords, and
1699 predefined macros such as @code{unix} and @code{vax} that identify the
1700 type of system you are using. It also enables the undesirable and
1701 rarely used ISO trigraph feature. For the C compiler,
1702 it disables recognition of C++ style @samp{//} comments as well as
1703 the @code{inline} keyword.
1705 The alternate keywords @code{__asm__}, @code{__extension__},
1706 @code{__inline__} and @code{__typeof__} continue to work despite
1707 @option{-ansi}. You would not want to use them in an ISO C program, of
1708 course, but it is useful to put them in header files that might be included
1709 in compilations done with @option{-ansi}. Alternate predefined macros
1710 such as @code{__unix__} and @code{__vax__} are also available, with or
1711 without @option{-ansi}.
1713 The @option{-ansi} option does not cause non-ISO programs to be
1714 rejected gratuitously. For that, @option{-Wpedantic} is required in
1715 addition to @option{-ansi}. @xref{Warning Options}.
1717 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1718 option is used. Some header files may notice this macro and refrain
1719 from declaring certain functions or defining certain macros that the
1720 ISO standard doesn't call for; this is to avoid interfering with any
1721 programs that might use these names for other things.
1723 Functions that are normally built in but do not have semantics
1724 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1725 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1726 built-in functions provided by GCC}, for details of the functions
1731 Determine the language standard. @xref{Standards,,Language Standards
1732 Supported by GCC}, for details of these standard versions. This option
1733 is currently only supported when compiling C or C++.
1735 The compiler can accept several base standards, such as @samp{c90} or
1736 @samp{c++98}, and GNU dialects of those standards, such as
1737 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1738 compiler accepts all programs following that standard plus those
1739 using GNU extensions that do not contradict it. For example,
1740 @option{-std=c90} turns off certain features of GCC that are
1741 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1742 keywords, but not other GNU extensions that do not have a meaning in
1743 ISO C90, such as omitting the middle term of a @code{?:}
1744 expression. On the other hand, when a GNU dialect of a standard is
1745 specified, all features supported by the compiler are enabled, even when
1746 those features change the meaning of the base standard. As a result, some
1747 strict-conforming programs may be rejected. The particular standard
1748 is used by @option{-Wpedantic} to identify which features are GNU
1749 extensions given that version of the standard. For example
1750 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1751 comments, while @option{-std=gnu99 -Wpedantic} does not.
1753 A value for this option must be provided; possible values are
1759 Support all ISO C90 programs (certain GNU extensions that conflict
1760 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1762 @item iso9899:199409
1763 ISO C90 as modified in amendment 1.
1769 ISO C99. This standard is substantially completely supported, modulo
1770 bugs and floating-point issues
1771 (mainly but not entirely relating to optional C99 features from
1772 Annexes F and G). See
1773 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1774 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1779 ISO C11, the 2011 revision of the ISO C standard. This standard is
1780 substantially completely supported, modulo bugs, floating-point issues
1781 (mainly but not entirely relating to optional C11 features from
1782 Annexes F and G) and the optional Annexes K (Bounds-checking
1783 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1787 GNU dialect of ISO C90 (including some C99 features).
1791 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1795 GNU dialect of ISO C11. This is the default for C code.
1796 The name @samp{gnu1x} is deprecated.
1800 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1801 additional defect reports. Same as @option{-ansi} for C++ code.
1805 GNU dialect of @option{-std=c++98}.
1809 The 2011 ISO C++ standard plus amendments.
1810 The name @samp{c++0x} is deprecated.
1814 GNU dialect of @option{-std=c++11}.
1815 The name @samp{gnu++0x} is deprecated.
1819 The 2014 ISO C++ standard plus amendments.
1820 The name @samp{c++1y} is deprecated.
1824 GNU dialect of @option{-std=c++14}.
1825 This is the default for C++ code.
1826 The name @samp{gnu++1y} is deprecated.
1829 The next revision of the ISO C++ standard, tentatively planned for
1830 2017. Support is highly experimental, and will almost certainly
1831 change in incompatible ways in future releases.
1834 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1835 and will almost certainly change in incompatible ways in future
1839 @item -fgnu89-inline
1840 @opindex fgnu89-inline
1841 The option @option{-fgnu89-inline} tells GCC to use the traditional
1842 GNU semantics for @code{inline} functions when in C99 mode.
1843 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1844 Using this option is roughly equivalent to adding the
1845 @code{gnu_inline} function attribute to all inline functions
1846 (@pxref{Function Attributes}).
1848 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1849 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1850 specifies the default behavior).
1851 This option is not supported in @option{-std=c90} or
1852 @option{-std=gnu90} mode.
1854 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1855 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1856 in effect for @code{inline} functions. @xref{Common Predefined
1857 Macros,,,cpp,The C Preprocessor}.
1859 @item -aux-info @var{filename}
1861 Output to the given filename prototyped declarations for all functions
1862 declared and/or defined in a translation unit, including those in header
1863 files. This option is silently ignored in any language other than C@.
1865 Besides declarations, the file indicates, in comments, the origin of
1866 each declaration (source file and line), whether the declaration was
1867 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1868 @samp{O} for old, respectively, in the first character after the line
1869 number and the colon), and whether it came from a declaration or a
1870 definition (@samp{C} or @samp{F}, respectively, in the following
1871 character). In the case of function definitions, a K&R-style list of
1872 arguments followed by their declarations is also provided, inside
1873 comments, after the declaration.
1875 @item -fallow-parameterless-variadic-functions
1876 @opindex fallow-parameterless-variadic-functions
1877 Accept variadic functions without named parameters.
1879 Although it is possible to define such a function, this is not very
1880 useful as it is not possible to read the arguments. This is only
1881 supported for C as this construct is allowed by C++.
1885 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1886 keyword, so that code can use these words as identifiers. You can use
1887 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1888 instead. @option{-ansi} implies @option{-fno-asm}.
1890 In C++, this switch only affects the @code{typeof} keyword, since
1891 @code{asm} and @code{inline} are standard keywords. You may want to
1892 use the @option{-fno-gnu-keywords} flag instead, which has the same
1893 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1894 switch only affects the @code{asm} and @code{typeof} keywords, since
1895 @code{inline} is a standard keyword in ISO C99.
1898 @itemx -fno-builtin-@var{function}
1899 @opindex fno-builtin
1900 @cindex built-in functions
1901 Don't recognize built-in functions that do not begin with
1902 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1903 functions provided by GCC}, for details of the functions affected,
1904 including those which are not built-in functions when @option{-ansi} or
1905 @option{-std} options for strict ISO C conformance are used because they
1906 do not have an ISO standard meaning.
1908 GCC normally generates special code to handle certain built-in functions
1909 more efficiently; for instance, calls to @code{alloca} may become single
1910 instructions which adjust the stack directly, and calls to @code{memcpy}
1911 may become inline copy loops. The resulting code is often both smaller
1912 and faster, but since the function calls no longer appear as such, you
1913 cannot set a breakpoint on those calls, nor can you change the behavior
1914 of the functions by linking with a different library. In addition,
1915 when a function is recognized as a built-in function, GCC may use
1916 information about that function to warn about problems with calls to
1917 that function, or to generate more efficient code, even if the
1918 resulting code still contains calls to that function. For example,
1919 warnings are given with @option{-Wformat} for bad calls to
1920 @code{printf} when @code{printf} is built in and @code{strlen} is
1921 known not to modify global memory.
1923 With the @option{-fno-builtin-@var{function}} option
1924 only the built-in function @var{function} is
1925 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1926 function is named that is not built-in in this version of GCC, this
1927 option is ignored. There is no corresponding
1928 @option{-fbuiltin-@var{function}} option; if you wish to enable
1929 built-in functions selectively when using @option{-fno-builtin} or
1930 @option{-ffreestanding}, you may define macros such as:
1933 #define abs(n) __builtin_abs ((n))
1934 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1939 @cindex hosted environment
1941 Assert that compilation targets a hosted environment. This implies
1942 @option{-fbuiltin}. A hosted environment is one in which the
1943 entire standard library is available, and in which @code{main} has a return
1944 type of @code{int}. Examples are nearly everything except a kernel.
1945 This is equivalent to @option{-fno-freestanding}.
1947 @item -ffreestanding
1948 @opindex ffreestanding
1949 @cindex hosted environment
1951 Assert that compilation targets a freestanding environment. This
1952 implies @option{-fno-builtin}. A freestanding environment
1953 is one in which the standard library may not exist, and program startup may
1954 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1955 This is equivalent to @option{-fno-hosted}.
1957 @xref{Standards,,Language Standards Supported by GCC}, for details of
1958 freestanding and hosted environments.
1962 @cindex OpenACC accelerator programming
1963 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1964 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1965 compiler generates accelerated code according to the OpenACC Application
1966 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1967 implies @option{-pthread}, and thus is only supported on targets that
1968 have support for @option{-pthread}.
1970 @item -fopenacc-dim=@var{geom}
1971 @opindex fopenacc-dim
1972 @cindex OpenACC accelerator programming
1973 Specify default compute dimensions for parallel offload regions that do
1974 not explicitly specify. The @var{geom} value is a triple of
1975 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
1976 can be omitted, to use a target-specific default value.
1980 @cindex OpenMP parallel
1981 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1982 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1983 compiler generates parallel code according to the OpenMP Application
1984 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1985 implies @option{-pthread}, and thus is only supported on targets that
1986 have support for @option{-pthread}. @option{-fopenmp} implies
1987 @option{-fopenmp-simd}.
1990 @opindex fopenmp-simd
1993 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1994 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1999 @cindex Enable Cilk Plus
2000 Enable the usage of Cilk Plus language extension features for C/C++.
2001 When the option @option{-fcilkplus} is specified, enable the usage of
2002 the Cilk Plus Language extension features for C/C++. The present
2003 implementation follows ABI version 1.2. This is an experimental
2004 feature that is only partially complete, and whose interface may
2005 change in future versions of GCC as the official specification
2006 changes. Currently, all features but @code{_Cilk_for} have been
2011 When the option @option{-fgnu-tm} is specified, the compiler
2012 generates code for the Linux variant of Intel's current Transactional
2013 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2014 an experimental feature whose interface may change in future versions
2015 of GCC, as the official specification changes. Please note that not
2016 all architectures are supported for this feature.
2018 For more information on GCC's support for transactional memory,
2019 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2020 Transactional Memory Library}.
2022 Note that the transactional memory feature is not supported with
2023 non-call exceptions (@option{-fnon-call-exceptions}).
2025 @item -fms-extensions
2026 @opindex fms-extensions
2027 Accept some non-standard constructs used in Microsoft header files.
2029 In C++ code, this allows member names in structures to be similar
2030 to previous types declarations.
2039 Some cases of unnamed fields in structures and unions are only
2040 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2041 fields within structs/unions}, for details.
2043 Note that this option is off for all targets but x86
2044 targets using ms-abi.
2046 @item -fplan9-extensions
2047 @opindex fplan9-extensions
2048 Accept some non-standard constructs used in Plan 9 code.
2050 This enables @option{-fms-extensions}, permits passing pointers to
2051 structures with anonymous fields to functions that expect pointers to
2052 elements of the type of the field, and permits referring to anonymous
2053 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2054 struct/union fields within structs/unions}, for details. This is only
2055 supported for C, not C++.
2059 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2060 options for strict ISO C conformance) implies @option{-trigraphs}.
2062 @cindex traditional C language
2063 @cindex C language, traditional
2065 @itemx -traditional-cpp
2066 @opindex traditional-cpp
2067 @opindex traditional
2068 Formerly, these options caused GCC to attempt to emulate a pre-standard
2069 C compiler. They are now only supported with the @option{-E} switch.
2070 The preprocessor continues to support a pre-standard mode. See the GNU
2071 CPP manual for details.
2073 @item -fcond-mismatch
2074 @opindex fcond-mismatch
2075 Allow conditional expressions with mismatched types in the second and
2076 third arguments. The value of such an expression is void. This option
2077 is not supported for C++.
2079 @item -flax-vector-conversions
2080 @opindex flax-vector-conversions
2081 Allow implicit conversions between vectors with differing numbers of
2082 elements and/or incompatible element types. This option should not be
2085 @item -funsigned-char
2086 @opindex funsigned-char
2087 Let the type @code{char} be unsigned, like @code{unsigned char}.
2089 Each kind of machine has a default for what @code{char} should
2090 be. It is either like @code{unsigned char} by default or like
2091 @code{signed char} by default.
2093 Ideally, a portable program should always use @code{signed char} or
2094 @code{unsigned char} when it depends on the signedness of an object.
2095 But many programs have been written to use plain @code{char} and
2096 expect it to be signed, or expect it to be unsigned, depending on the
2097 machines they were written for. This option, and its inverse, let you
2098 make such a program work with the opposite default.
2100 The type @code{char} is always a distinct type from each of
2101 @code{signed char} or @code{unsigned char}, even though its behavior
2102 is always just like one of those two.
2105 @opindex fsigned-char
2106 Let the type @code{char} be signed, like @code{signed char}.
2108 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2109 the negative form of @option{-funsigned-char}. Likewise, the option
2110 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2112 @item -fsigned-bitfields
2113 @itemx -funsigned-bitfields
2114 @itemx -fno-signed-bitfields
2115 @itemx -fno-unsigned-bitfields
2116 @opindex fsigned-bitfields
2117 @opindex funsigned-bitfields
2118 @opindex fno-signed-bitfields
2119 @opindex fno-unsigned-bitfields
2120 These options control whether a bit-field is signed or unsigned, when the
2121 declaration does not use either @code{signed} or @code{unsigned}. By
2122 default, such a bit-field is signed, because this is consistent: the
2123 basic integer types such as @code{int} are signed types.
2125 @item -fsso-struct=@var{endianness}
2126 @opindex fsso-struct
2127 Set the default scalar storage order of structures and unions to the
2128 specified endianness. The accepted values are @samp{big-endian} and
2129 @samp{little-endian}. If the option is not passed, the compiler uses
2130 the native endianness of the target. This option is not supported for C++.
2132 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2133 code that is not binary compatible with code generated without it if the
2134 specified endianness is not the native endianness of the target.
2137 @node C++ Dialect Options
2138 @section Options Controlling C++ Dialect
2140 @cindex compiler options, C++
2141 @cindex C++ options, command-line
2142 @cindex options, C++
2143 This section describes the command-line options that are only meaningful
2144 for C++ programs. You can also use most of the GNU compiler options
2145 regardless of what language your program is in. For example, you
2146 might compile a file @file{firstClass.C} like this:
2149 g++ -g -fstrict-enums -O -c firstClass.C
2153 In this example, only @option{-fstrict-enums} is an option meant
2154 only for C++ programs; you can use the other options with any
2155 language supported by GCC@.
2157 Some options for compiling C programs, such as @option{-std}, are also
2158 relevant for C++ programs.
2159 @xref{C Dialect Options,,Options Controlling C Dialect}.
2161 Here is a list of options that are @emph{only} for compiling C++ programs:
2165 @item -fabi-version=@var{n}
2166 @opindex fabi-version
2167 Use version @var{n} of the C++ ABI@. The default is version 0.
2169 Version 0 refers to the version conforming most closely to
2170 the C++ ABI specification. Therefore, the ABI obtained using version 0
2171 will change in different versions of G++ as ABI bugs are fixed.
2173 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2175 Version 2 is the version of the C++ ABI that first appeared in G++
2176 3.4, and was the default through G++ 4.9.
2178 Version 3 corrects an error in mangling a constant address as a
2181 Version 4, which first appeared in G++ 4.5, implements a standard
2182 mangling for vector types.
2184 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2185 attribute const/volatile on function pointer types, decltype of a
2186 plain decl, and use of a function parameter in the declaration of
2189 Version 6, which first appeared in G++ 4.7, corrects the promotion
2190 behavior of C++11 scoped enums and the mangling of template argument
2191 packs, const/static_cast, prefix ++ and --, and a class scope function
2192 used as a template argument.
2194 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2195 builtin type and corrects the mangling of lambdas in default argument
2198 Version 8, which first appeared in G++ 4.9, corrects the substitution
2199 behavior of function types with function-cv-qualifiers.
2201 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2204 Version 10, which first appeared in G++ 6.1, adds mangling of
2205 attributes that affect type identity, such as ia32 calling convention
2206 attributes (e.g. @samp{stdcall}).
2208 See also @option{-Wabi}.
2210 @item -fabi-compat-version=@var{n}
2211 @opindex fabi-compat-version
2212 On targets that support strong aliases, G++
2213 works around mangling changes by creating an alias with the correct
2214 mangled name when defining a symbol with an incorrect mangled name.
2215 This switch specifies which ABI version to use for the alias.
2217 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2218 compatibility). If another ABI version is explicitly selected, this
2219 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2220 use @option{-fabi-compat-version=2}.
2222 If this option is not provided but @option{-Wabi=@var{n}} is, that
2223 version is used for compatibility aliases. If this option is provided
2224 along with @option{-Wabi} (without the version), the version from this
2225 option is used for the warning.
2227 @item -fno-access-control
2228 @opindex fno-access-control
2229 Turn off all access checking. This switch is mainly useful for working
2230 around bugs in the access control code.
2234 Check that the pointer returned by @code{operator new} is non-null
2235 before attempting to modify the storage allocated. This check is
2236 normally unnecessary because the C++ standard specifies that
2237 @code{operator new} only returns @code{0} if it is declared
2238 @code{throw()}, in which case the compiler always checks the
2239 return value even without this option. In all other cases, when
2240 @code{operator new} has a non-empty exception specification, memory
2241 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2242 @samp{new (nothrow)}.
2246 Enable support for the C++ Extensions for Concepts Technical
2247 Specification, ISO 19217 (2015), which allows code like
2250 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2251 template <Addable T> T add (T a, T b) @{ return a + b; @}
2254 @item -fconstexpr-depth=@var{n}
2255 @opindex fconstexpr-depth
2256 Set the maximum nested evaluation depth for C++11 constexpr functions
2257 to @var{n}. A limit is needed to detect endless recursion during
2258 constant expression evaluation. The minimum specified by the standard
2261 @item -fdeduce-init-list
2262 @opindex fdeduce-init-list
2263 Enable deduction of a template type parameter as
2264 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2267 template <class T> auto forward(T t) -> decltype (realfn (t))
2274 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2278 This deduction was implemented as a possible extension to the
2279 originally proposed semantics for the C++11 standard, but was not part
2280 of the final standard, so it is disabled by default. This option is
2281 deprecated, and may be removed in a future version of G++.
2283 @item -ffriend-injection
2284 @opindex ffriend-injection
2285 Inject friend functions into the enclosing namespace, so that they are
2286 visible outside the scope of the class in which they are declared.
2287 Friend functions were documented to work this way in the old Annotated
2288 C++ Reference Manual.
2289 However, in ISO C++ a friend function that is not declared
2290 in an enclosing scope can only be found using argument dependent
2291 lookup. GCC defaults to the standard behavior.
2293 This option is for compatibility, and may be removed in a future
2296 @item -fno-elide-constructors
2297 @opindex fno-elide-constructors
2298 The C++ standard allows an implementation to omit creating a temporary
2299 that is only used to initialize another object of the same type.
2300 Specifying this option disables that optimization, and forces G++ to
2301 call the copy constructor in all cases.
2303 @item -fno-enforce-eh-specs
2304 @opindex fno-enforce-eh-specs
2305 Don't generate code to check for violation of exception specifications
2306 at run time. This option violates the C++ standard, but may be useful
2307 for reducing code size in production builds, much like defining
2308 @code{NDEBUG}. This does not give user code permission to throw
2309 exceptions in violation of the exception specifications; the compiler
2310 still optimizes based on the specifications, so throwing an
2311 unexpected exception results in undefined behavior at run time.
2313 @item -fextern-tls-init
2314 @itemx -fno-extern-tls-init
2315 @opindex fextern-tls-init
2316 @opindex fno-extern-tls-init
2317 The C++11 and OpenMP standards allow @code{thread_local} and
2318 @code{threadprivate} variables to have dynamic (runtime)
2319 initialization. To support this, any use of such a variable goes
2320 through a wrapper function that performs any necessary initialization.
2321 When the use and definition of the variable are in the same
2322 translation unit, this overhead can be optimized away, but when the
2323 use is in a different translation unit there is significant overhead
2324 even if the variable doesn't actually need dynamic initialization. If
2325 the programmer can be sure that no use of the variable in a
2326 non-defining TU needs to trigger dynamic initialization (either
2327 because the variable is statically initialized, or a use of the
2328 variable in the defining TU will be executed before any uses in
2329 another TU), they can avoid this overhead with the
2330 @option{-fno-extern-tls-init} option.
2332 On targets that support symbol aliases, the default is
2333 @option{-fextern-tls-init}. On targets that do not support symbol
2334 aliases, the default is @option{-fno-extern-tls-init}.
2337 @itemx -fno-for-scope
2339 @opindex fno-for-scope
2340 If @option{-ffor-scope} is specified, the scope of variables declared in
2341 a @i{for-init-statement} is limited to the @code{for} loop itself,
2342 as specified by the C++ standard.
2343 If @option{-fno-for-scope} is specified, the scope of variables declared in
2344 a @i{for-init-statement} extends to the end of the enclosing scope,
2345 as was the case in old versions of G++, and other (traditional)
2346 implementations of C++.
2348 If neither flag is given, the default is to follow the standard,
2349 but to allow and give a warning for old-style code that would
2350 otherwise be invalid, or have different behavior.
2352 @item -fno-gnu-keywords
2353 @opindex fno-gnu-keywords
2354 Do not recognize @code{typeof} as a keyword, so that code can use this
2355 word as an identifier. You can use the keyword @code{__typeof__} instead.
2356 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2357 @option{-std=c++98}, @option{-std=c++11}, etc.
2359 @item -fno-implicit-templates
2360 @opindex fno-implicit-templates
2361 Never emit code for non-inline templates that are instantiated
2362 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2363 @xref{Template Instantiation}, for more information.
2365 @item -fno-implicit-inline-templates
2366 @opindex fno-implicit-inline-templates
2367 Don't emit code for implicit instantiations of inline templates, either.
2368 The default is to handle inlines differently so that compiles with and
2369 without optimization need the same set of explicit instantiations.
2371 @item -fno-implement-inlines
2372 @opindex fno-implement-inlines
2373 To save space, do not emit out-of-line copies of inline functions
2374 controlled by @code{#pragma implementation}. This causes linker
2375 errors if these functions are not inlined everywhere they are called.
2377 @item -fms-extensions
2378 @opindex fms-extensions
2379 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2380 int and getting a pointer to member function via non-standard syntax.
2382 @item -fno-nonansi-builtins
2383 @opindex fno-nonansi-builtins
2384 Disable built-in declarations of functions that are not mandated by
2385 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2386 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2389 @opindex fnothrow-opt
2390 Treat a @code{throw()} exception specification as if it were a
2391 @code{noexcept} specification to reduce or eliminate the text size
2392 overhead relative to a function with no exception specification. If
2393 the function has local variables of types with non-trivial
2394 destructors, the exception specification actually makes the
2395 function smaller because the EH cleanups for those variables can be
2396 optimized away. The semantic effect is that an exception thrown out of
2397 a function with such an exception specification results in a call
2398 to @code{terminate} rather than @code{unexpected}.
2400 @item -fno-operator-names
2401 @opindex fno-operator-names
2402 Do not treat the operator name keywords @code{and}, @code{bitand},
2403 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2404 synonyms as keywords.
2406 @item -fno-optional-diags
2407 @opindex fno-optional-diags
2408 Disable diagnostics that the standard says a compiler does not need to
2409 issue. Currently, the only such diagnostic issued by G++ is the one for
2410 a name having multiple meanings within a class.
2413 @opindex fpermissive
2414 Downgrade some diagnostics about nonconformant code from errors to
2415 warnings. Thus, using @option{-fpermissive} allows some
2416 nonconforming code to compile.
2418 @item -fno-pretty-templates
2419 @opindex fno-pretty-templates
2420 When an error message refers to a specialization of a function
2421 template, the compiler normally prints the signature of the
2422 template followed by the template arguments and any typedefs or
2423 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2424 rather than @code{void f(int)}) so that it's clear which template is
2425 involved. When an error message refers to a specialization of a class
2426 template, the compiler omits any template arguments that match
2427 the default template arguments for that template. If either of these
2428 behaviors make it harder to understand the error message rather than
2429 easier, you can use @option{-fno-pretty-templates} to disable them.
2433 Enable automatic template instantiation at link time. This option also
2434 implies @option{-fno-implicit-templates}. @xref{Template
2435 Instantiation}, for more information.
2439 Disable generation of information about every class with virtual
2440 functions for use by the C++ run-time type identification features
2441 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2442 of the language, you can save some space by using this flag. Note that
2443 exception handling uses the same information, but G++ generates it as
2444 needed. The @code{dynamic_cast} operator can still be used for casts that
2445 do not require run-time type information, i.e.@: casts to @code{void *} or to
2446 unambiguous base classes.
2448 @item -fsized-deallocation
2449 @opindex fsized-deallocation
2450 Enable the built-in global declarations
2452 void operator delete (void *, std::size_t) noexcept;
2453 void operator delete[] (void *, std::size_t) noexcept;
2455 as introduced in C++14. This is useful for user-defined replacement
2456 deallocation functions that, for example, use the size of the object
2457 to make deallocation faster. Enabled by default under
2458 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2459 warns about places that might want to add a definition.
2461 @item -fstrict-enums
2462 @opindex fstrict-enums
2463 Allow the compiler to optimize using the assumption that a value of
2464 enumerated type can only be one of the values of the enumeration (as
2465 defined in the C++ standard; basically, a value that can be
2466 represented in the minimum number of bits needed to represent all the
2467 enumerators). This assumption may not be valid if the program uses a
2468 cast to convert an arbitrary integer value to the enumerated type.
2470 @item -ftemplate-backtrace-limit=@var{n}
2471 @opindex ftemplate-backtrace-limit
2472 Set the maximum number of template instantiation notes for a single
2473 warning or error to @var{n}. The default value is 10.
2475 @item -ftemplate-depth=@var{n}
2476 @opindex ftemplate-depth
2477 Set the maximum instantiation depth for template classes to @var{n}.
2478 A limit on the template instantiation depth is needed to detect
2479 endless recursions during template class instantiation. ANSI/ISO C++
2480 conforming programs must not rely on a maximum depth greater than 17
2481 (changed to 1024 in C++11). The default value is 900, as the compiler
2482 can run out of stack space before hitting 1024 in some situations.
2484 @item -fno-threadsafe-statics
2485 @opindex fno-threadsafe-statics
2486 Do not emit the extra code to use the routines specified in the C++
2487 ABI for thread-safe initialization of local statics. You can use this
2488 option to reduce code size slightly in code that doesn't need to be
2491 @item -fuse-cxa-atexit
2492 @opindex fuse-cxa-atexit
2493 Register destructors for objects with static storage duration with the
2494 @code{__cxa_atexit} function rather than the @code{atexit} function.
2495 This option is required for fully standards-compliant handling of static
2496 destructors, but only works if your C library supports
2497 @code{__cxa_atexit}.
2499 @item -fno-use-cxa-get-exception-ptr
2500 @opindex fno-use-cxa-get-exception-ptr
2501 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2502 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2503 if the runtime routine is not available.
2505 @item -fvisibility-inlines-hidden
2506 @opindex fvisibility-inlines-hidden
2507 This switch declares that the user does not attempt to compare
2508 pointers to inline functions or methods where the addresses of the two functions
2509 are taken in different shared objects.
2511 The effect of this is that GCC may, effectively, mark inline methods with
2512 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2513 appear in the export table of a DSO and do not require a PLT indirection
2514 when used within the DSO@. Enabling this option can have a dramatic effect
2515 on load and link times of a DSO as it massively reduces the size of the
2516 dynamic export table when the library makes heavy use of templates.
2518 The behavior of this switch is not quite the same as marking the
2519 methods as hidden directly, because it does not affect static variables
2520 local to the function or cause the compiler to deduce that
2521 the function is defined in only one shared object.
2523 You may mark a method as having a visibility explicitly to negate the
2524 effect of the switch for that method. For example, if you do want to
2525 compare pointers to a particular inline method, you might mark it as
2526 having default visibility. Marking the enclosing class with explicit
2527 visibility has no effect.
2529 Explicitly instantiated inline methods are unaffected by this option
2530 as their linkage might otherwise cross a shared library boundary.
2531 @xref{Template Instantiation}.
2533 @item -fvisibility-ms-compat
2534 @opindex fvisibility-ms-compat
2535 This flag attempts to use visibility settings to make GCC's C++
2536 linkage model compatible with that of Microsoft Visual Studio.
2538 The flag makes these changes to GCC's linkage model:
2542 It sets the default visibility to @code{hidden}, like
2543 @option{-fvisibility=hidden}.
2546 Types, but not their members, are not hidden by default.
2549 The One Definition Rule is relaxed for types without explicit
2550 visibility specifications that are defined in more than one
2551 shared object: those declarations are permitted if they are
2552 permitted when this option is not used.
2555 In new code it is better to use @option{-fvisibility=hidden} and
2556 export those classes that are intended to be externally visible.
2557 Unfortunately it is possible for code to rely, perhaps accidentally,
2558 on the Visual Studio behavior.
2560 Among the consequences of these changes are that static data members
2561 of the same type with the same name but defined in different shared
2562 objects are different, so changing one does not change the other;
2563 and that pointers to function members defined in different shared
2564 objects may not compare equal. When this flag is given, it is a
2565 violation of the ODR to define types with the same name differently.
2569 Do not use weak symbol support, even if it is provided by the linker.
2570 By default, G++ uses weak symbols if they are available. This
2571 option exists only for testing, and should not be used by end-users;
2572 it results in inferior code and has no benefits. This option may
2573 be removed in a future release of G++.
2577 Do not search for header files in the standard directories specific to
2578 C++, but do still search the other standard directories. (This option
2579 is used when building the C++ library.)
2582 In addition, these optimization, warning, and code generation options
2583 have meanings only for C++ programs:
2586 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2589 Warn when G++ it generates code that is probably not compatible with
2590 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2591 ABI with each major release, normally @option{-Wabi} will warn only if
2592 there is a check added later in a release series for an ABI issue
2593 discovered since the initial release. @option{-Wabi} will warn about
2594 more things if an older ABI version is selected (with
2595 @option{-fabi-version=@var{n}}).
2597 @option{-Wabi} can also be used with an explicit version number to
2598 warn about compatibility with a particular @option{-fabi-version}
2599 level, e.g. @option{-Wabi=2} to warn about changes relative to
2600 @option{-fabi-version=2}.
2602 If an explicit version number is provided and
2603 @option{-fabi-compat-version} is not specified, the version number
2604 from this option is used for compatibility aliases. If no explicit
2605 version number is provided with this option, but
2606 @option{-fabi-compat-version} is specified, that version number is
2607 used for ABI warnings.
2609 Although an effort has been made to warn about
2610 all such cases, there are probably some cases that are not warned about,
2611 even though G++ is generating incompatible code. There may also be
2612 cases where warnings are emitted even though the code that is generated
2615 You should rewrite your code to avoid these warnings if you are
2616 concerned about the fact that code generated by G++ may not be binary
2617 compatible with code generated by other compilers.
2619 Known incompatibilities in @option{-fabi-version=2} (which was the
2620 default from GCC 3.4 to 4.9) include:
2625 A template with a non-type template parameter of reference type was
2626 mangled incorrectly:
2629 template <int &> struct S @{@};
2633 This was fixed in @option{-fabi-version=3}.
2636 SIMD vector types declared using @code{__attribute ((vector_size))} were
2637 mangled in a non-standard way that does not allow for overloading of
2638 functions taking vectors of different sizes.
2640 The mangling was changed in @option{-fabi-version=4}.
2643 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2644 qualifiers, and @code{decltype} of a plain declaration was folded away.
2646 These mangling issues were fixed in @option{-fabi-version=5}.
2649 Scoped enumerators passed as arguments to a variadic function are
2650 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2651 On most targets this does not actually affect the parameter passing
2652 ABI, as there is no way to pass an argument smaller than @code{int}.
2654 Also, the ABI changed the mangling of template argument packs,
2655 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2656 a class scope function used as a template argument.
2658 These issues were corrected in @option{-fabi-version=6}.
2661 Lambdas in default argument scope were mangled incorrectly, and the
2662 ABI changed the mangling of @code{nullptr_t}.
2664 These issues were corrected in @option{-fabi-version=7}.
2667 When mangling a function type with function-cv-qualifiers, the
2668 un-qualified function type was incorrectly treated as a substitution
2671 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2674 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2675 unaligned accesses. Note that this did not affect the ABI of a
2676 function with a @code{nullptr_t} parameter, as parameters have a
2679 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2682 Target-specific attributes that affect the identity of a type, such as
2683 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2684 did not affect the mangled name, leading to name collisions when
2685 function pointers were used as template arguments.
2687 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2691 It also warns about psABI-related changes. The known psABI changes at this
2697 For SysV/x86-64, unions with @code{long double} members are
2698 passed in memory as specified in psABI. For example:
2708 @code{union U} is always passed in memory.
2712 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2715 Warn when a type with an ABI tag is used in a context that does not
2716 have that ABI tag. See @ref{C++ Attributes} for more information
2719 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2720 @opindex Wctor-dtor-privacy
2721 @opindex Wno-ctor-dtor-privacy
2722 Warn when a class seems unusable because all the constructors or
2723 destructors in that class are private, and it has neither friends nor
2724 public static member functions. Also warn if there are no non-private
2725 methods, and there's at least one private member function that isn't
2726 a constructor or destructor.
2728 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2729 @opindex Wdelete-non-virtual-dtor
2730 @opindex Wno-delete-non-virtual-dtor
2731 Warn when @code{delete} is used to destroy an instance of a class that
2732 has virtual functions and non-virtual destructor. It is unsafe to delete
2733 an instance of a derived class through a pointer to a base class if the
2734 base class does not have a virtual destructor. This warning is enabled
2737 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2738 @opindex Wliteral-suffix
2739 @opindex Wno-literal-suffix
2740 Warn when a string or character literal is followed by a ud-suffix which does
2741 not begin with an underscore. As a conforming extension, GCC treats such
2742 suffixes as separate preprocessing tokens in order to maintain backwards
2743 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2747 #define __STDC_FORMAT_MACROS
2748 #include <inttypes.h>
2753 printf("My int64: %" PRId64"\n", i64);
2757 In this case, @code{PRId64} is treated as a separate preprocessing token.
2759 This warning is enabled by default.
2761 @item -Wlto-type-mismatch
2762 @opindex Wlto-type-mismatch
2763 @opindex Wno-lto-type-mismatch
2765 During the link-time optimization warn about type mismatches in
2766 global declarations from different compilation units.
2767 Requires @option{-flto} to be enabled. Enabled by default.
2769 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2771 @opindex Wno-narrowing
2772 With @option{-std=gnu++98} or @option{-std=c++98}, warn when a narrowing
2773 conversion prohibited by C++11 occurs within
2777 int i = @{ 2.2 @}; // error: narrowing from double to int
2780 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2782 When a later standard is in effect, e.g. when using @option{-std=c++11},
2783 narrowing conversions are diagnosed by default, as required by the standard.
2784 A narrowing conversion from a constant produces an error,
2785 and a narrowing conversion from a non-constant produces a warning,
2786 but @option{-Wno-narrowing} suppresses the diagnostic.
2787 Note that this does not affect the meaning of well-formed code;
2788 narrowing conversions are still considered ill-formed in SFINAE contexts.
2790 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2792 @opindex Wno-noexcept
2793 Warn when a noexcept-expression evaluates to false because of a call
2794 to a function that does not have a non-throwing exception
2795 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2796 the compiler to never throw an exception.
2798 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2799 @opindex Wnon-virtual-dtor
2800 @opindex Wno-non-virtual-dtor
2801 Warn when a class has virtual functions and an accessible non-virtual
2802 destructor itself or in an accessible polymorphic base class, in which
2803 case it is possible but unsafe to delete an instance of a derived
2804 class through a pointer to the class itself or base class. This
2805 warning is automatically enabled if @option{-Weffc++} is specified.
2807 @item -Wreorder @r{(C++ and Objective-C++ only)}
2809 @opindex Wno-reorder
2810 @cindex reordering, warning
2811 @cindex warning for reordering of member initializers
2812 Warn when the order of member initializers given in the code does not
2813 match the order in which they must be executed. For instance:
2819 A(): j (0), i (1) @{ @}
2824 The compiler rearranges the member initializers for @code{i}
2825 and @code{j} to match the declaration order of the members, emitting
2826 a warning to that effect. This warning is enabled by @option{-Wall}.
2828 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2829 @opindex fext-numeric-literals
2830 @opindex fno-ext-numeric-literals
2831 Accept imaginary, fixed-point, or machine-defined
2832 literal number suffixes as GNU extensions.
2833 When this option is turned off these suffixes are treated
2834 as C++11 user-defined literal numeric suffixes.
2835 This is on by default for all pre-C++11 dialects and all GNU dialects:
2836 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2837 @option{-std=gnu++14}.
2838 This option is off by default
2839 for ISO C++11 onwards (@option{-std=c++11}, ...).
2842 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2845 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2848 Warn about violations of the following style guidelines from Scott Meyers'
2849 @cite{Effective C++} series of books:
2853 Define a copy constructor and an assignment operator for classes
2854 with dynamically-allocated memory.
2857 Prefer initialization to assignment in constructors.
2860 Have @code{operator=} return a reference to @code{*this}.
2863 Don't try to return a reference when you must return an object.
2866 Distinguish between prefix and postfix forms of increment and
2867 decrement operators.
2870 Never overload @code{&&}, @code{||}, or @code{,}.
2874 This option also enables @option{-Wnon-virtual-dtor}, which is also
2875 one of the effective C++ recommendations. However, the check is
2876 extended to warn about the lack of virtual destructor in accessible
2877 non-polymorphic bases classes too.
2879 When selecting this option, be aware that the standard library
2880 headers do not obey all of these guidelines; use @samp{grep -v}
2881 to filter out those warnings.
2883 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2884 @opindex Wstrict-null-sentinel
2885 @opindex Wno-strict-null-sentinel
2886 Warn about the use of an uncasted @code{NULL} as sentinel. When
2887 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2888 to @code{__null}. Although it is a null pointer constant rather than a
2889 null pointer, it is guaranteed to be of the same size as a pointer.
2890 But this use is not portable across different compilers.
2892 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2893 @opindex Wno-non-template-friend
2894 @opindex Wnon-template-friend
2895 Disable warnings when non-templatized friend functions are declared
2896 within a template. Since the advent of explicit template specification
2897 support in G++, if the name of the friend is an unqualified-id (i.e.,
2898 @samp{friend foo(int)}), the C++ language specification demands that the
2899 friend declare or define an ordinary, nontemplate function. (Section
2900 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2901 could be interpreted as a particular specialization of a templatized
2902 function. Because this non-conforming behavior is no longer the default
2903 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2904 check existing code for potential trouble spots and is on by default.
2905 This new compiler behavior can be turned off with
2906 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2907 but disables the helpful warning.
2909 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2910 @opindex Wold-style-cast
2911 @opindex Wno-old-style-cast
2912 Warn if an old-style (C-style) cast to a non-void type is used within
2913 a C++ program. The new-style casts (@code{dynamic_cast},
2914 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2915 less vulnerable to unintended effects and much easier to search for.
2917 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2918 @opindex Woverloaded-virtual
2919 @opindex Wno-overloaded-virtual
2920 @cindex overloaded virtual function, warning
2921 @cindex warning for overloaded virtual function
2922 Warn when a function declaration hides virtual functions from a
2923 base class. For example, in:
2930 struct B: public A @{
2935 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2946 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2947 @opindex Wno-pmf-conversions
2948 @opindex Wpmf-conversions
2949 Disable the diagnostic for converting a bound pointer to member function
2952 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2953 @opindex Wsign-promo
2954 @opindex Wno-sign-promo
2955 Warn when overload resolution chooses a promotion from unsigned or
2956 enumerated type to a signed type, over a conversion to an unsigned type of
2957 the same size. Previous versions of G++ tried to preserve
2958 unsignedness, but the standard mandates the current behavior.
2960 @item -Wtemplates @r{(C++ and Objective-C++ only)}
2962 Warn when a primary template declaration is encountered. Some coding
2963 rules disallow templates, and this may be used to enforce that rule.
2964 The warning is inactive inside a system header file, such as the STL, so
2965 one can still use the STL. One may also instantiate or specialize
2968 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
2969 @opindex Wmultiple-inheritance
2970 Warn when a class is defined with multiple direct base classes. Some
2971 coding rules disallow multiple inheritance, and this may be used to
2972 enforce that rule. The warning is inactive inside a system header file,
2973 such as the STL, so one can still use the STL. One may also define
2974 classes that indirectly use multiple inheritance.
2976 @item -Wvirtual-inheritance
2977 @opindex Wvirtual-inheritance
2978 Warn when a class is defined with a virtual direct base classe. Some
2979 coding rules disallow multiple inheritance, and this may be used to
2980 enforce that rule. The warning is inactive inside a system header file,
2981 such as the STL, so one can still use the STL. One may also define
2982 classes that indirectly use virtual inheritance.
2985 @opindex Wnamespaces
2986 Warn when a namespace definition is opened. Some coding rules disallow
2987 namespaces, and this may be used to enforce that rule. The warning is
2988 inactive inside a system header file, such as the STL, so one can still
2989 use the STL. One may also use using directives and qualified names.
2991 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
2993 @opindex Wno-terminate
2994 Disable the warning about a throw-expression that will immediately
2995 result in a call to @code{terminate}.
2998 @node Objective-C and Objective-C++ Dialect Options
2999 @section Options Controlling Objective-C and Objective-C++ Dialects
3001 @cindex compiler options, Objective-C and Objective-C++
3002 @cindex Objective-C and Objective-C++ options, command-line
3003 @cindex options, Objective-C and Objective-C++
3004 (NOTE: This manual does not describe the Objective-C and Objective-C++
3005 languages themselves. @xref{Standards,,Language Standards
3006 Supported by GCC}, for references.)
3008 This section describes the command-line options that are only meaningful
3009 for Objective-C and Objective-C++ programs. You can also use most of
3010 the language-independent GNU compiler options.
3011 For example, you might compile a file @file{some_class.m} like this:
3014 gcc -g -fgnu-runtime -O -c some_class.m
3018 In this example, @option{-fgnu-runtime} is an option meant only for
3019 Objective-C and Objective-C++ programs; you can use the other options with
3020 any language supported by GCC@.
3022 Note that since Objective-C is an extension of the C language, Objective-C
3023 compilations may also use options specific to the C front-end (e.g.,
3024 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3025 C++-specific options (e.g., @option{-Wabi}).
3027 Here is a list of options that are @emph{only} for compiling Objective-C
3028 and Objective-C++ programs:
3031 @item -fconstant-string-class=@var{class-name}
3032 @opindex fconstant-string-class
3033 Use @var{class-name} as the name of the class to instantiate for each
3034 literal string specified with the syntax @code{@@"@dots{}"}. The default
3035 class name is @code{NXConstantString} if the GNU runtime is being used, and
3036 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3037 @option{-fconstant-cfstrings} option, if also present, overrides the
3038 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3039 to be laid out as constant CoreFoundation strings.
3042 @opindex fgnu-runtime
3043 Generate object code compatible with the standard GNU Objective-C
3044 runtime. This is the default for most types of systems.
3046 @item -fnext-runtime
3047 @opindex fnext-runtime
3048 Generate output compatible with the NeXT runtime. This is the default
3049 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3050 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3053 @item -fno-nil-receivers
3054 @opindex fno-nil-receivers
3055 Assume that all Objective-C message dispatches (@code{[receiver
3056 message:arg]}) in this translation unit ensure that the receiver is
3057 not @code{nil}. This allows for more efficient entry points in the
3058 runtime to be used. This option is only available in conjunction with
3059 the NeXT runtime and ABI version 0 or 1.
3061 @item -fobjc-abi-version=@var{n}
3062 @opindex fobjc-abi-version
3063 Use version @var{n} of the Objective-C ABI for the selected runtime.
3064 This option is currently supported only for the NeXT runtime. In that
3065 case, Version 0 is the traditional (32-bit) ABI without support for
3066 properties and other Objective-C 2.0 additions. Version 1 is the
3067 traditional (32-bit) ABI with support for properties and other
3068 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3069 nothing is specified, the default is Version 0 on 32-bit target
3070 machines, and Version 2 on 64-bit target machines.
3072 @item -fobjc-call-cxx-cdtors
3073 @opindex fobjc-call-cxx-cdtors
3074 For each Objective-C class, check if any of its instance variables is a
3075 C++ object with a non-trivial default constructor. If so, synthesize a
3076 special @code{- (id) .cxx_construct} instance method which runs
3077 non-trivial default constructors on any such instance variables, in order,
3078 and then return @code{self}. Similarly, check if any instance variable
3079 is a C++ object with a non-trivial destructor, and if so, synthesize a
3080 special @code{- (void) .cxx_destruct} method which runs
3081 all such default destructors, in reverse order.
3083 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3084 methods thusly generated only operate on instance variables
3085 declared in the current Objective-C class, and not those inherited
3086 from superclasses. It is the responsibility of the Objective-C
3087 runtime to invoke all such methods in an object's inheritance
3088 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3089 by the runtime immediately after a new object instance is allocated;
3090 the @code{- (void) .cxx_destruct} methods are invoked immediately
3091 before the runtime deallocates an object instance.
3093 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3094 support for invoking the @code{- (id) .cxx_construct} and
3095 @code{- (void) .cxx_destruct} methods.
3097 @item -fobjc-direct-dispatch
3098 @opindex fobjc-direct-dispatch
3099 Allow fast jumps to the message dispatcher. On Darwin this is
3100 accomplished via the comm page.
3102 @item -fobjc-exceptions
3103 @opindex fobjc-exceptions
3104 Enable syntactic support for structured exception handling in
3105 Objective-C, similar to what is offered by C++ and Java. This option
3106 is required to use the Objective-C keywords @code{@@try},
3107 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3108 @code{@@synchronized}. This option is available with both the GNU
3109 runtime and the NeXT runtime (but not available in conjunction with
3110 the NeXT runtime on Mac OS X 10.2 and earlier).
3114 Enable garbage collection (GC) in Objective-C and Objective-C++
3115 programs. This option is only available with the NeXT runtime; the
3116 GNU runtime has a different garbage collection implementation that
3117 does not require special compiler flags.
3119 @item -fobjc-nilcheck
3120 @opindex fobjc-nilcheck
3121 For the NeXT runtime with version 2 of the ABI, check for a nil
3122 receiver in method invocations before doing the actual method call.
3123 This is the default and can be disabled using
3124 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3125 checked for nil in this way no matter what this flag is set to.
3126 Currently this flag does nothing when the GNU runtime, or an older
3127 version of the NeXT runtime ABI, is used.
3129 @item -fobjc-std=objc1
3131 Conform to the language syntax of Objective-C 1.0, the language
3132 recognized by GCC 4.0. This only affects the Objective-C additions to
3133 the C/C++ language; it does not affect conformance to C/C++ standards,
3134 which is controlled by the separate C/C++ dialect option flags. When
3135 this option is used with the Objective-C or Objective-C++ compiler,
3136 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3137 This is useful if you need to make sure that your Objective-C code can
3138 be compiled with older versions of GCC@.
3140 @item -freplace-objc-classes
3141 @opindex freplace-objc-classes
3142 Emit a special marker instructing @command{ld(1)} not to statically link in
3143 the resulting object file, and allow @command{dyld(1)} to load it in at
3144 run time instead. This is used in conjunction with the Fix-and-Continue
3145 debugging mode, where the object file in question may be recompiled and
3146 dynamically reloaded in the course of program execution, without the need
3147 to restart the program itself. Currently, Fix-and-Continue functionality
3148 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3153 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3154 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3155 compile time) with static class references that get initialized at load time,
3156 which improves run-time performance. Specifying the @option{-fzero-link} flag
3157 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3158 to be retained. This is useful in Zero-Link debugging mode, since it allows
3159 for individual class implementations to be modified during program execution.
3160 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3161 regardless of command-line options.
3163 @item -fno-local-ivars
3164 @opindex fno-local-ivars
3165 @opindex flocal-ivars
3166 By default instance variables in Objective-C can be accessed as if
3167 they were local variables from within the methods of the class they're
3168 declared in. This can lead to shadowing between instance variables
3169 and other variables declared either locally inside a class method or
3170 globally with the same name. Specifying the @option{-fno-local-ivars}
3171 flag disables this behavior thus avoiding variable shadowing issues.
3173 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3174 @opindex fivar-visibility
3175 Set the default instance variable visibility to the specified option
3176 so that instance variables declared outside the scope of any access
3177 modifier directives default to the specified visibility.
3181 Dump interface declarations for all classes seen in the source file to a
3182 file named @file{@var{sourcename}.decl}.
3184 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3185 @opindex Wassign-intercept
3186 @opindex Wno-assign-intercept
3187 Warn whenever an Objective-C assignment is being intercepted by the
3190 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3191 @opindex Wno-protocol
3193 If a class is declared to implement a protocol, a warning is issued for
3194 every method in the protocol that is not implemented by the class. The
3195 default behavior is to issue a warning for every method not explicitly
3196 implemented in the class, even if a method implementation is inherited
3197 from the superclass. If you use the @option{-Wno-protocol} option, then
3198 methods inherited from the superclass are considered to be implemented,
3199 and no warning is issued for them.
3201 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3203 @opindex Wno-selector
3204 Warn if multiple methods of different types for the same selector are
3205 found during compilation. The check is performed on the list of methods
3206 in the final stage of compilation. Additionally, a check is performed
3207 for each selector appearing in a @code{@@selector(@dots{})}
3208 expression, and a corresponding method for that selector has been found
3209 during compilation. Because these checks scan the method table only at
3210 the end of compilation, these warnings are not produced if the final
3211 stage of compilation is not reached, for example because an error is
3212 found during compilation, or because the @option{-fsyntax-only} option is
3215 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3216 @opindex Wstrict-selector-match
3217 @opindex Wno-strict-selector-match
3218 Warn if multiple methods with differing argument and/or return types are
3219 found for a given selector when attempting to send a message using this
3220 selector to a receiver of type @code{id} or @code{Class}. When this flag
3221 is off (which is the default behavior), the compiler omits such warnings
3222 if any differences found are confined to types that share the same size
3225 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3226 @opindex Wundeclared-selector
3227 @opindex Wno-undeclared-selector
3228 Warn if a @code{@@selector(@dots{})} expression referring to an
3229 undeclared selector is found. A selector is considered undeclared if no
3230 method with that name has been declared before the
3231 @code{@@selector(@dots{})} expression, either explicitly in an
3232 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3233 an @code{@@implementation} section. This option always performs its
3234 checks as soon as a @code{@@selector(@dots{})} expression is found,
3235 while @option{-Wselector} only performs its checks in the final stage of
3236 compilation. This also enforces the coding style convention
3237 that methods and selectors must be declared before being used.
3239 @item -print-objc-runtime-info
3240 @opindex print-objc-runtime-info
3241 Generate C header describing the largest structure that is passed by
3246 @node Diagnostic Message Formatting Options
3247 @section Options to Control Diagnostic Messages Formatting
3248 @cindex options to control diagnostics formatting
3249 @cindex diagnostic messages
3250 @cindex message formatting
3252 Traditionally, diagnostic messages have been formatted irrespective of
3253 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3254 options described below
3255 to control the formatting algorithm for diagnostic messages,
3256 e.g.@: how many characters per line, how often source location
3257 information should be reported. Note that some language front ends may not
3258 honor these options.
3261 @item -fmessage-length=@var{n}
3262 @opindex fmessage-length
3263 Try to format error messages so that they fit on lines of about
3264 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3265 done; each error message appears on a single line. This is the
3266 default for all front ends.
3268 @item -fdiagnostics-show-location=once
3269 @opindex fdiagnostics-show-location
3270 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3271 reporter to emit source location information @emph{once}; that is, in
3272 case the message is too long to fit on a single physical line and has to
3273 be wrapped, the source location won't be emitted (as prefix) again,
3274 over and over, in subsequent continuation lines. This is the default
3277 @item -fdiagnostics-show-location=every-line
3278 Only meaningful in line-wrapping mode. Instructs the diagnostic
3279 messages reporter to emit the same source location information (as
3280 prefix) for physical lines that result from the process of breaking
3281 a message which is too long to fit on a single line.
3283 @item -fdiagnostics-color[=@var{WHEN}]
3284 @itemx -fno-diagnostics-color
3285 @opindex fdiagnostics-color
3286 @cindex highlight, color
3287 @vindex GCC_COLORS @r{environment variable}
3288 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3289 or @samp{auto}. The default depends on how the compiler has been configured,
3290 it can be any of the above @var{WHEN} options or also @samp{never}
3291 if @env{GCC_COLORS} environment variable isn't present in the environment,
3292 and @samp{auto} otherwise.
3293 @samp{auto} means to use color only when the standard error is a terminal.
3294 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3295 aliases for @option{-fdiagnostics-color=always} and
3296 @option{-fdiagnostics-color=never}, respectively.
3298 The colors are defined by the environment variable @env{GCC_COLORS}.
3299 Its value is a colon-separated list of capabilities and Select Graphic
3300 Rendition (SGR) substrings. SGR commands are interpreted by the
3301 terminal or terminal emulator. (See the section in the documentation
3302 of your text terminal for permitted values and their meanings as
3303 character attributes.) These substring values are integers in decimal
3304 representation and can be concatenated with semicolons.
3305 Common values to concatenate include
3307 @samp{4} for underline,
3309 @samp{7} for inverse,
3310 @samp{39} for default foreground color,
3311 @samp{30} to @samp{37} for foreground colors,
3312 @samp{90} to @samp{97} for 16-color mode foreground colors,
3313 @samp{38;5;0} to @samp{38;5;255}
3314 for 88-color and 256-color modes foreground colors,
3315 @samp{49} for default background color,
3316 @samp{40} to @samp{47} for background colors,
3317 @samp{100} to @samp{107} for 16-color mode background colors,
3318 and @samp{48;5;0} to @samp{48;5;255}
3319 for 88-color and 256-color modes background colors.
3321 The default @env{GCC_COLORS} is
3323 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3326 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3327 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3328 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3329 string disables colors.
3330 Supported capabilities are as follows.
3334 @vindex error GCC_COLORS @r{capability}
3335 SGR substring for error: markers.
3338 @vindex warning GCC_COLORS @r{capability}
3339 SGR substring for warning: markers.
3342 @vindex note GCC_COLORS @r{capability}
3343 SGR substring for note: markers.
3346 @vindex caret GCC_COLORS @r{capability}
3347 SGR substring for caret line.
3350 @vindex locus GCC_COLORS @r{capability}
3351 SGR substring for location information, @samp{file:line} or
3352 @samp{file:line:column} etc.
3355 @vindex quote GCC_COLORS @r{capability}
3356 SGR substring for information printed within quotes.
3359 @item -fno-diagnostics-show-option
3360 @opindex fno-diagnostics-show-option
3361 @opindex fdiagnostics-show-option
3362 By default, each diagnostic emitted includes text indicating the
3363 command-line option that directly controls the diagnostic (if such an
3364 option is known to the diagnostic machinery). Specifying the
3365 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3367 @item -fno-diagnostics-show-caret
3368 @opindex fno-diagnostics-show-caret
3369 @opindex fdiagnostics-show-caret
3370 By default, each diagnostic emitted includes the original source line
3371 and a caret @samp{^} indicating the column. This option suppresses this
3372 information. The source line is truncated to @var{n} characters, if
3373 the @option{-fmessage-length=n} option is given. When the output is done
3374 to the terminal, the width is limited to the width given by the
3375 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3379 @node Warning Options
3380 @section Options to Request or Suppress Warnings
3381 @cindex options to control warnings
3382 @cindex warning messages
3383 @cindex messages, warning
3384 @cindex suppressing warnings
3386 Warnings are diagnostic messages that report constructions that
3387 are not inherently erroneous but that are risky or suggest there
3388 may have been an error.
3390 The following language-independent options do not enable specific
3391 warnings but control the kinds of diagnostics produced by GCC@.
3394 @cindex syntax checking
3396 @opindex fsyntax-only
3397 Check the code for syntax errors, but don't do anything beyond that.
3399 @item -fmax-errors=@var{n}
3400 @opindex fmax-errors
3401 Limits the maximum number of error messages to @var{n}, at which point
3402 GCC bails out rather than attempting to continue processing the source
3403 code. If @var{n} is 0 (the default), there is no limit on the number
3404 of error messages produced. If @option{-Wfatal-errors} is also
3405 specified, then @option{-Wfatal-errors} takes precedence over this
3410 Inhibit all warning messages.
3415 Make all warnings into errors.
3420 Make the specified warning into an error. The specifier for a warning
3421 is appended; for example @option{-Werror=switch} turns the warnings
3422 controlled by @option{-Wswitch} into errors. This switch takes a
3423 negative form, to be used to negate @option{-Werror} for specific
3424 warnings; for example @option{-Wno-error=switch} makes
3425 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3428 The warning message for each controllable warning includes the
3429 option that controls the warning. That option can then be used with
3430 @option{-Werror=} and @option{-Wno-error=} as described above.
3431 (Printing of the option in the warning message can be disabled using the
3432 @option{-fno-diagnostics-show-option} flag.)
3434 Note that specifying @option{-Werror=}@var{foo} automatically implies
3435 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3438 @item -Wfatal-errors
3439 @opindex Wfatal-errors
3440 @opindex Wno-fatal-errors
3441 This option causes the compiler to abort compilation on the first error
3442 occurred rather than trying to keep going and printing further error
3447 You can request many specific warnings with options beginning with
3448 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3449 implicit declarations. Each of these specific warning options also
3450 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3451 example, @option{-Wno-implicit}. This manual lists only one of the
3452 two forms, whichever is not the default. For further
3453 language-specific options also refer to @ref{C++ Dialect Options} and
3454 @ref{Objective-C and Objective-C++ Dialect Options}.
3456 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3457 options, such as @option{-Wunused}, which may turn on further options,
3458 such as @option{-Wunused-value}. The combined effect of positive and
3459 negative forms is that more specific options have priority over less
3460 specific ones, independently of their position in the command-line. For
3461 options of the same specificity, the last one takes effect. Options
3462 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3463 as if they appeared at the end of the command-line.
3465 When an unrecognized warning option is requested (e.g.,
3466 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3467 that the option is not recognized. However, if the @option{-Wno-} form
3468 is used, the behavior is slightly different: no diagnostic is
3469 produced for @option{-Wno-unknown-warning} unless other diagnostics
3470 are being produced. This allows the use of new @option{-Wno-} options
3471 with old compilers, but if something goes wrong, the compiler
3472 warns that an unrecognized option is present.
3479 Issue all the warnings demanded by strict ISO C and ISO C++;
3480 reject all programs that use forbidden extensions, and some other
3481 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3482 version of the ISO C standard specified by any @option{-std} option used.
3484 Valid ISO C and ISO C++ programs should compile properly with or without
3485 this option (though a rare few require @option{-ansi} or a
3486 @option{-std} option specifying the required version of ISO C)@. However,
3487 without this option, certain GNU extensions and traditional C and C++
3488 features are supported as well. With this option, they are rejected.
3490 @option{-Wpedantic} does not cause warning messages for use of the
3491 alternate keywords whose names begin and end with @samp{__}. Pedantic
3492 warnings are also disabled in the expression that follows
3493 @code{__extension__}. However, only system header files should use
3494 these escape routes; application programs should avoid them.
3495 @xref{Alternate Keywords}.
3497 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3498 C conformance. They soon find that it does not do quite what they want:
3499 it finds some non-ISO practices, but not all---only those for which
3500 ISO C @emph{requires} a diagnostic, and some others for which
3501 diagnostics have been added.
3503 A feature to report any failure to conform to ISO C might be useful in
3504 some instances, but would require considerable additional work and would
3505 be quite different from @option{-Wpedantic}. We don't have plans to
3506 support such a feature in the near future.
3508 Where the standard specified with @option{-std} represents a GNU
3509 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3510 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3511 extended dialect is based. Warnings from @option{-Wpedantic} are given
3512 where they are required by the base standard. (It does not make sense
3513 for such warnings to be given only for features not in the specified GNU
3514 C dialect, since by definition the GNU dialects of C include all
3515 features the compiler supports with the given option, and there would be
3516 nothing to warn about.)
3518 @item -pedantic-errors
3519 @opindex pedantic-errors
3520 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3521 requires a diagnostic, in some cases where there is undefined behavior
3522 at compile-time and in some other cases that do not prevent compilation
3523 of programs that are valid according to the standard. This is not
3524 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3525 by this option and not enabled by the latter and vice versa.
3530 This enables all the warnings about constructions that some users
3531 consider questionable, and that are easy to avoid (or modify to
3532 prevent the warning), even in conjunction with macros. This also
3533 enables some language-specific warnings described in @ref{C++ Dialect
3534 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3536 @option{-Wall} turns on the following warning flags:
3538 @gccoptlist{-Waddress @gol
3539 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3541 -Wc++11-compat -Wc++14-compat@gol
3542 -Wchar-subscripts @gol
3544 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3546 -Wimplicit @r{(C and Objective-C only)} @gol
3547 -Wimplicit-int @r{(C and Objective-C only)} @gol
3548 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3549 -Winit-self @r{(only for C++)} @gol
3550 -Wlogical-not-parentheses
3551 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3552 -Wmaybe-uninitialized @gol
3553 -Wmemset-transposed-args @gol
3554 -Wmisleading-indentation @r{(only for C/C++)} @gol
3555 -Wmissing-braces @r{(only for C/ObjC)} @gol
3556 -Wnarrowing @r{(only for C++)} @gol
3558 -Wnonnull-compare @gol
3564 -Wsequence-point @gol
3565 -Wsign-compare @r{(only in C++)} @gol
3566 -Wsizeof-pointer-memaccess @gol
3567 -Wstrict-aliasing @gol
3568 -Wstrict-overflow=1 @gol
3570 -Wtautological-compare @gol
3572 -Wuninitialized @gol
3573 -Wunknown-pragmas @gol
3574 -Wunused-function @gol
3577 -Wunused-variable @gol
3578 -Wvolatile-register-var @gol
3581 Note that some warning flags are not implied by @option{-Wall}. Some of
3582 them warn about constructions that users generally do not consider
3583 questionable, but which occasionally you might wish to check for;
3584 others warn about constructions that are necessary or hard to avoid in
3585 some cases, and there is no simple way to modify the code to suppress
3586 the warning. Some of them are enabled by @option{-Wextra} but many of
3587 them must be enabled individually.
3593 This enables some extra warning flags that are not enabled by
3594 @option{-Wall}. (This option used to be called @option{-W}. The older
3595 name is still supported, but the newer name is more descriptive.)
3597 @gccoptlist{-Wclobbered @gol
3599 -Wignored-qualifiers @gol
3600 -Wmissing-field-initializers @gol
3601 -Wmissing-parameter-type @r{(C only)} @gol
3602 -Wold-style-declaration @r{(C only)} @gol
3603 -Woverride-init @gol
3604 -Wsign-compare @r{(C only)} @gol
3606 -Wuninitialized @gol
3607 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3608 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3609 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3612 The option @option{-Wextra} also prints warning messages for the
3618 A pointer is compared against integer zero with @code{<}, @code{<=},
3619 @code{>}, or @code{>=}.
3622 (C++ only) An enumerator and a non-enumerator both appear in a
3623 conditional expression.
3626 (C++ only) Ambiguous virtual bases.
3629 (C++ only) Subscripting an array that has been declared @code{register}.
3632 (C++ only) Taking the address of a variable that has been declared
3636 (C++ only) A base class is not initialized in a derived class's copy
3641 @item -Wchar-subscripts
3642 @opindex Wchar-subscripts
3643 @opindex Wno-char-subscripts
3644 Warn if an array subscript has type @code{char}. This is a common cause
3645 of error, as programmers often forget that this type is signed on some
3647 This warning is enabled by @option{-Wall}.
3651 @opindex Wno-comment
3652 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3653 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3654 This warning is enabled by @option{-Wall}.
3656 @item -Wno-coverage-mismatch
3657 @opindex Wno-coverage-mismatch
3658 Warn if feedback profiles do not match when using the
3659 @option{-fprofile-use} option.
3660 If a source file is changed between compiling with @option{-fprofile-gen} and
3661 with @option{-fprofile-use}, the files with the profile feedback can fail
3662 to match the source file and GCC cannot use the profile feedback
3663 information. By default, this warning is enabled and is treated as an
3664 error. @option{-Wno-coverage-mismatch} can be used to disable the
3665 warning or @option{-Wno-error=coverage-mismatch} can be used to
3666 disable the error. Disabling the error for this warning can result in
3667 poorly optimized code and is useful only in the
3668 case of very minor changes such as bug fixes to an existing code-base.
3669 Completely disabling the warning is not recommended.
3672 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3674 Suppress warning messages emitted by @code{#warning} directives.
3676 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3677 @opindex Wdouble-promotion
3678 @opindex Wno-double-promotion
3679 Give a warning when a value of type @code{float} is implicitly
3680 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3681 floating-point unit implement @code{float} in hardware, but emulate
3682 @code{double} in software. On such a machine, doing computations
3683 using @code{double} values is much more expensive because of the
3684 overhead required for software emulation.
3686 It is easy to accidentally do computations with @code{double} because
3687 floating-point literals are implicitly of type @code{double}. For
3691 float area(float radius)
3693 return 3.14159 * radius * radius;
3697 the compiler performs the entire computation with @code{double}
3698 because the floating-point literal is a @code{double}.
3701 @itemx -Wformat=@var{n}
3704 @opindex ffreestanding
3705 @opindex fno-builtin
3707 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3708 the arguments supplied have types appropriate to the format string
3709 specified, and that the conversions specified in the format string make
3710 sense. This includes standard functions, and others specified by format
3711 attributes (@pxref{Function Attributes}), in the @code{printf},
3712 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3713 not in the C standard) families (or other target-specific families).
3714 Which functions are checked without format attributes having been
3715 specified depends on the standard version selected, and such checks of
3716 functions without the attribute specified are disabled by
3717 @option{-ffreestanding} or @option{-fno-builtin}.
3719 The formats are checked against the format features supported by GNU
3720 libc version 2.2. These include all ISO C90 and C99 features, as well
3721 as features from the Single Unix Specification and some BSD and GNU
3722 extensions. Other library implementations may not support all these
3723 features; GCC does not support warning about features that go beyond a
3724 particular library's limitations. However, if @option{-Wpedantic} is used
3725 with @option{-Wformat}, warnings are given about format features not
3726 in the selected standard version (but not for @code{strfmon} formats,
3727 since those are not in any version of the C standard). @xref{C Dialect
3728 Options,,Options Controlling C Dialect}.
3735 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3736 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3737 @option{-Wformat} also checks for null format arguments for several
3738 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3739 aspects of this level of format checking can be disabled by the
3740 options: @option{-Wno-format-contains-nul},
3741 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3742 @option{-Wformat} is enabled by @option{-Wall}.
3744 @item -Wno-format-contains-nul
3745 @opindex Wno-format-contains-nul
3746 @opindex Wformat-contains-nul
3747 If @option{-Wformat} is specified, do not warn about format strings that
3750 @item -Wno-format-extra-args
3751 @opindex Wno-format-extra-args
3752 @opindex Wformat-extra-args
3753 If @option{-Wformat} is specified, do not warn about excess arguments to a
3754 @code{printf} or @code{scanf} format function. The C standard specifies
3755 that such arguments are ignored.
3757 Where the unused arguments lie between used arguments that are
3758 specified with @samp{$} operand number specifications, normally
3759 warnings are still given, since the implementation could not know what
3760 type to pass to @code{va_arg} to skip the unused arguments. However,
3761 in the case of @code{scanf} formats, this option suppresses the
3762 warning if the unused arguments are all pointers, since the Single
3763 Unix Specification says that such unused arguments are allowed.
3765 @item -Wno-format-zero-length
3766 @opindex Wno-format-zero-length
3767 @opindex Wformat-zero-length
3768 If @option{-Wformat} is specified, do not warn about zero-length formats.
3769 The C standard specifies that zero-length formats are allowed.
3774 Enable @option{-Wformat} plus additional format checks. Currently
3775 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3778 @item -Wformat-nonliteral
3779 @opindex Wformat-nonliteral
3780 @opindex Wno-format-nonliteral
3781 If @option{-Wformat} is specified, also warn if the format string is not a
3782 string literal and so cannot be checked, unless the format function
3783 takes its format arguments as a @code{va_list}.
3785 @item -Wformat-security
3786 @opindex Wformat-security
3787 @opindex Wno-format-security
3788 If @option{-Wformat} is specified, also warn about uses of format
3789 functions that represent possible security problems. At present, this
3790 warns about calls to @code{printf} and @code{scanf} functions where the
3791 format string is not a string literal and there are no format arguments,
3792 as in @code{printf (foo);}. This may be a security hole if the format
3793 string came from untrusted input and contains @samp{%n}. (This is
3794 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3795 in future warnings may be added to @option{-Wformat-security} that are not
3796 included in @option{-Wformat-nonliteral}.)
3798 @item -Wformat-signedness
3799 @opindex Wformat-signedness
3800 @opindex Wno-format-signedness
3801 If @option{-Wformat} is specified, also warn if the format string
3802 requires an unsigned argument and the argument is signed and vice versa.
3805 @opindex Wformat-y2k
3806 @opindex Wno-format-y2k
3807 If @option{-Wformat} is specified, also warn about @code{strftime}
3808 formats that may yield only a two-digit year.
3813 @opindex Wno-nonnull
3814 Warn about passing a null pointer for arguments marked as
3815 requiring a non-null value by the @code{nonnull} function attribute.
3817 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3818 can be disabled with the @option{-Wno-nonnull} option.
3820 @item -Wnonnull-compare
3821 @opindex Wnonnull-compare
3822 @opindex Wno-nonnull-compare
3823 Warn when comparing an argument marked with the @code{nonnull}
3824 function attribute against null inside the function.
3826 @option{-Wnonnull-compare} is included in @option{-Wall}. It
3827 can be disabled with the @option{-Wno-nonnull-compare} option.
3829 @item -Wnull-dereference
3830 @opindex Wnull-dereference
3831 @opindex Wno-null-dereference
3832 Warn if the compiler detects paths that trigger erroneous or
3833 undefined behavior due to dereferencing a null pointer. This option
3834 is only active when @option{-fdelete-null-pointer-checks} is active,
3835 which is enabled by optimizations in most targets. The precision of
3836 the warnings depends on the optimization options used.
3838 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3840 @opindex Wno-init-self
3841 Warn about uninitialized variables that are initialized with themselves.
3842 Note this option can only be used with the @option{-Wuninitialized} option.
3844 For example, GCC warns about @code{i} being uninitialized in the
3845 following snippet only when @option{-Winit-self} has been specified:
3856 This warning is enabled by @option{-Wall} in C++.
3858 @item -Wimplicit-int @r{(C and Objective-C only)}
3859 @opindex Wimplicit-int
3860 @opindex Wno-implicit-int
3861 Warn when a declaration does not specify a type.
3862 This warning is enabled by @option{-Wall}.
3864 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3865 @opindex Wimplicit-function-declaration
3866 @opindex Wno-implicit-function-declaration
3867 Give a warning whenever a function is used before being declared. In
3868 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3869 enabled by default and it is made into an error by
3870 @option{-pedantic-errors}. This warning is also enabled by
3873 @item -Wimplicit @r{(C and Objective-C only)}
3875 @opindex Wno-implicit
3876 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3877 This warning is enabled by @option{-Wall}.
3879 @item -Wignored-qualifiers @r{(C and C++ only)}
3880 @opindex Wignored-qualifiers
3881 @opindex Wno-ignored-qualifiers
3882 Warn if the return type of a function has a type qualifier
3883 such as @code{const}. For ISO C such a type qualifier has no effect,
3884 since the value returned by a function is not an lvalue.
3885 For C++, the warning is only emitted for scalar types or @code{void}.
3886 ISO C prohibits qualified @code{void} return types on function
3887 definitions, so such return types always receive a warning
3888 even without this option.
3890 This warning is also enabled by @option{-Wextra}.
3892 @item -Wignored-attributes @r{(C and C++ only)}
3893 @opindex Wignored-attributes
3894 @opindex Wno-ignored-attributes
3895 Warn when an attribute is ignored. This is different from the
3896 @option{-Wattributes} option in that it warns whenever the compiler decides
3897 to drop an attribute, not that the attribute is either unknown, used in a
3898 wrong place, etc. This warning is enabled by default.
3903 Warn if the type of @code{main} is suspicious. @code{main} should be
3904 a function with external linkage, returning int, taking either zero
3905 arguments, two, or three arguments of appropriate types. This warning
3906 is enabled by default in C++ and is enabled by either @option{-Wall}
3907 or @option{-Wpedantic}.
3909 @item -Wmisleading-indentation @r{(C and C++ only)}
3910 @opindex Wmisleading-indentation
3911 @opindex Wno-misleading-indentation
3912 Warn when the indentation of the code does not reflect the block structure.
3913 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3914 @code{for} clauses with a guarded statement that does not use braces,
3915 followed by an unguarded statement with the same indentation.
3917 In the following example, the call to ``bar'' is misleadingly indented as
3918 if it were guarded by the ``if'' conditional.
3921 if (some_condition ())
3923 bar (); /* Gotcha: this is not guarded by the "if". */
3926 In the case of mixed tabs and spaces, the warning uses the
3927 @option{-ftabstop=} option to determine if the statements line up
3930 The warning is not issued for code involving multiline preprocessor logic
3931 such as the following example.
3936 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3942 The warning is not issued after a @code{#line} directive, since this
3943 typically indicates autogenerated code, and no assumptions can be made
3944 about the layout of the file that the directive references.
3946 This warning is enabled by @option{-Wall} in C and C++.
3948 @item -Wmissing-braces
3949 @opindex Wmissing-braces
3950 @opindex Wno-missing-braces
3951 Warn if an aggregate or union initializer is not fully bracketed. In
3952 the following example, the initializer for @code{a} is not fully
3953 bracketed, but that for @code{b} is fully bracketed. This warning is
3954 enabled by @option{-Wall} in C.
3957 int a[2][2] = @{ 0, 1, 2, 3 @};
3958 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3961 This warning is enabled by @option{-Wall}.
3963 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3964 @opindex Wmissing-include-dirs
3965 @opindex Wno-missing-include-dirs
3966 Warn if a user-supplied include directory does not exist.
3969 @opindex Wparentheses
3970 @opindex Wno-parentheses
3971 Warn if parentheses are omitted in certain contexts, such
3972 as when there is an assignment in a context where a truth value
3973 is expected, or when operators are nested whose precedence people
3974 often get confused about.
3976 Also warn if a comparison like @code{x<=y<=z} appears; this is
3977 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3978 interpretation from that of ordinary mathematical notation.
3980 Also warn about constructions where there may be confusion to which
3981 @code{if} statement an @code{else} branch belongs. Here is an example of
3996 In C/C++, every @code{else} branch belongs to the innermost possible
3997 @code{if} statement, which in this example is @code{if (b)}. This is
3998 often not what the programmer expected, as illustrated in the above
3999 example by indentation the programmer chose. When there is the
4000 potential for this confusion, GCC issues a warning when this flag
4001 is specified. To eliminate the warning, add explicit braces around
4002 the innermost @code{if} statement so there is no way the @code{else}
4003 can belong to the enclosing @code{if}. The resulting code
4020 Also warn for dangerous uses of the GNU extension to
4021 @code{?:} with omitted middle operand. When the condition
4022 in the @code{?}: operator is a boolean expression, the omitted value is
4023 always 1. Often programmers expect it to be a value computed
4024 inside the conditional expression instead.
4026 This warning is enabled by @option{-Wall}.
4028 @item -Wsequence-point
4029 @opindex Wsequence-point
4030 @opindex Wno-sequence-point
4031 Warn about code that may have undefined semantics because of violations
4032 of sequence point rules in the C and C++ standards.
4034 The C and C++ standards define the order in which expressions in a C/C++
4035 program are evaluated in terms of @dfn{sequence points}, which represent
4036 a partial ordering between the execution of parts of the program: those
4037 executed before the sequence point, and those executed after it. These
4038 occur after the evaluation of a full expression (one which is not part
4039 of a larger expression), after the evaluation of the first operand of a
4040 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4041 function is called (but after the evaluation of its arguments and the
4042 expression denoting the called function), and in certain other places.
4043 Other than as expressed by the sequence point rules, the order of
4044 evaluation of subexpressions of an expression is not specified. All
4045 these rules describe only a partial order rather than a total order,
4046 since, for example, if two functions are called within one expression
4047 with no sequence point between them, the order in which the functions
4048 are called is not specified. However, the standards committee have
4049 ruled that function calls do not overlap.
4051 It is not specified when between sequence points modifications to the
4052 values of objects take effect. Programs whose behavior depends on this
4053 have undefined behavior; the C and C++ standards specify that ``Between
4054 the previous and next sequence point an object shall have its stored
4055 value modified at most once by the evaluation of an expression.
4056 Furthermore, the prior value shall be read only to determine the value
4057 to be stored.''. If a program breaks these rules, the results on any
4058 particular implementation are entirely unpredictable.
4060 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4061 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4062 diagnosed by this option, and it may give an occasional false positive
4063 result, but in general it has been found fairly effective at detecting
4064 this sort of problem in programs.
4066 The standard is worded confusingly, therefore there is some debate
4067 over the precise meaning of the sequence point rules in subtle cases.
4068 Links to discussions of the problem, including proposed formal
4069 definitions, may be found on the GCC readings page, at
4070 @uref{http://gcc.gnu.org/@/readings.html}.
4072 This warning is enabled by @option{-Wall} for C and C++.
4074 @item -Wno-return-local-addr
4075 @opindex Wno-return-local-addr
4076 @opindex Wreturn-local-addr
4077 Do not warn about returning a pointer (or in C++, a reference) to a
4078 variable that goes out of scope after the function returns.
4081 @opindex Wreturn-type
4082 @opindex Wno-return-type
4083 Warn whenever a function is defined with a return type that defaults
4084 to @code{int}. Also warn about any @code{return} statement with no
4085 return value in a function whose return type is not @code{void}
4086 (falling off the end of the function body is considered returning
4087 without a value), and about a @code{return} statement with an
4088 expression in a function whose return type is @code{void}.
4090 For C++, a function without return type always produces a diagnostic
4091 message, even when @option{-Wno-return-type} is specified. The only
4092 exceptions are @code{main} and functions defined in system headers.
4094 This warning is enabled by @option{-Wall}.
4096 @item -Wshift-count-negative
4097 @opindex Wshift-count-negative
4098 @opindex Wno-shift-count-negative
4099 Warn if shift count is negative. This warning is enabled by default.
4101 @item -Wshift-count-overflow
4102 @opindex Wshift-count-overflow
4103 @opindex Wno-shift-count-overflow
4104 Warn if shift count >= width of type. This warning is enabled by default.
4106 @item -Wshift-negative-value
4107 @opindex Wshift-negative-value
4108 @opindex Wno-shift-negative-value
4109 Warn if left shifting a negative value. This warning is enabled by
4110 @option{-Wextra} in C99 and C++11 modes (and newer).
4112 @item -Wshift-overflow
4113 @itemx -Wshift-overflow=@var{n}
4114 @opindex Wshift-overflow
4115 @opindex Wno-shift-overflow
4116 Warn about left shift overflows. This warning is enabled by
4117 default in C99 and C++11 modes (and newer).
4120 @item -Wshift-overflow=1
4121 This is the warning level of @option{-Wshift-overflow} and is enabled
4122 by default in C99 and C++11 modes (and newer). This warning level does
4123 not warn about left-shifting 1 into the sign bit. (However, in C, such
4124 an overflow is still rejected in contexts where an integer constant expression
4127 @item -Wshift-overflow=2
4128 This warning level also warns about left-shifting 1 into the sign bit,
4129 unless C++14 mode is active.
4135 Warn whenever a @code{switch} statement has an index of enumerated type
4136 and lacks a @code{case} for one or more of the named codes of that
4137 enumeration. (The presence of a @code{default} label prevents this
4138 warning.) @code{case} labels outside the enumeration range also
4139 provoke warnings when this option is used (even if there is a
4140 @code{default} label).
4141 This warning is enabled by @option{-Wall}.
4143 @item -Wswitch-default
4144 @opindex Wswitch-default
4145 @opindex Wno-switch-default
4146 Warn whenever a @code{switch} statement does not have a @code{default}
4150 @opindex Wswitch-enum
4151 @opindex Wno-switch-enum
4152 Warn whenever a @code{switch} statement has an index of enumerated type
4153 and lacks a @code{case} for one or more of the named codes of that
4154 enumeration. @code{case} labels outside the enumeration range also
4155 provoke warnings when this option is used. The only difference
4156 between @option{-Wswitch} and this option is that this option gives a
4157 warning about an omitted enumeration code even if there is a
4158 @code{default} label.
4161 @opindex Wswitch-bool
4162 @opindex Wno-switch-bool
4163 Warn whenever a @code{switch} statement has an index of boolean type
4164 and the case values are outside the range of a boolean type.
4165 It is possible to suppress this warning by casting the controlling
4166 expression to a type other than @code{bool}. For example:
4169 switch ((int) (a == 4))
4175 This warning is enabled by default for C and C++ programs.
4177 @item -Wsync-nand @r{(C and C++ only)}
4179 @opindex Wno-sync-nand
4180 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4181 built-in functions are used. These functions changed semantics in GCC 4.4.
4185 @opindex Wno-trigraphs
4186 Warn if any trigraphs are encountered that might change the meaning of
4187 the program (trigraphs within comments are not warned about).
4188 This warning is enabled by @option{-Wall}.
4190 @item -Wunused-but-set-parameter
4191 @opindex Wunused-but-set-parameter
4192 @opindex Wno-unused-but-set-parameter
4193 Warn whenever a function parameter is assigned to, but otherwise unused
4194 (aside from its declaration).
4196 To suppress this warning use the @code{unused} attribute
4197 (@pxref{Variable Attributes}).
4199 This warning is also enabled by @option{-Wunused} together with
4202 @item -Wunused-but-set-variable
4203 @opindex Wunused-but-set-variable
4204 @opindex Wno-unused-but-set-variable
4205 Warn whenever a local variable is assigned to, but otherwise unused
4206 (aside from its declaration).
4207 This warning is enabled by @option{-Wall}.
4209 To suppress this warning use the @code{unused} attribute
4210 (@pxref{Variable Attributes}).
4212 This warning is also enabled by @option{-Wunused}, which is enabled
4215 @item -Wunused-function
4216 @opindex Wunused-function
4217 @opindex Wno-unused-function
4218 Warn whenever a static function is declared but not defined or a
4219 non-inline static function is unused.
4220 This warning is enabled by @option{-Wall}.
4222 @item -Wunused-label
4223 @opindex Wunused-label
4224 @opindex Wno-unused-label
4225 Warn whenever a label is declared but not used.
4226 This warning is enabled by @option{-Wall}.
4228 To suppress this warning use the @code{unused} attribute
4229 (@pxref{Variable Attributes}).
4231 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4232 @opindex Wunused-local-typedefs
4233 Warn when a typedef locally defined in a function is not used.
4234 This warning is enabled by @option{-Wall}.
4236 @item -Wunused-parameter
4237 @opindex Wunused-parameter
4238 @opindex Wno-unused-parameter
4239 Warn whenever a function parameter is unused aside from its declaration.
4241 To suppress this warning use the @code{unused} attribute
4242 (@pxref{Variable Attributes}).
4244 @item -Wno-unused-result
4245 @opindex Wunused-result
4246 @opindex Wno-unused-result
4247 Do not warn if a caller of a function marked with attribute
4248 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4249 its return value. The default is @option{-Wunused-result}.
4251 @item -Wunused-variable
4252 @opindex Wunused-variable
4253 @opindex Wno-unused-variable
4254 Warn whenever a local or static variable is unused aside from its
4255 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4256 but not for C++. This warning is enabled by @option{-Wall}.
4258 To suppress this warning use the @code{unused} attribute
4259 (@pxref{Variable Attributes}).
4261 @item -Wunused-const-variable
4262 @itemx -Wunused-const-variable=@var{n}
4263 @opindex Wunused-const-variable
4264 @opindex Wno-unused-const-variable
4265 Warn whenever a constant static variable is unused aside from its declaration.
4266 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4267 for C, but not for C++. In C this declares variable storage, but in C++ this
4268 is not an error since const variables take the place of @code{#define}s.
4270 To suppress this warning use the @code{unused} attribute
4271 (@pxref{Variable Attributes}).
4274 @item -Wunused-const-variable=1
4275 This is the warning level that is enabled by @option{-Wunused-variable} for
4276 C. It warns only about unused static const variables defined in the main
4277 compilation unit, but not about static const variables declared in any
4280 @item -Wunused-const-variable=2
4281 This warning level also warns for unused constant static variables in
4282 headers (excluding system headers). This is the warning level of
4283 @option{-Wunused-const-variable} and must be explicitly requested since
4284 in C++ this isn't an error and in C it might be harder to clean up all
4288 @item -Wunused-value
4289 @opindex Wunused-value
4290 @opindex Wno-unused-value
4291 Warn whenever a statement computes a result that is explicitly not
4292 used. To suppress this warning cast the unused expression to
4293 @code{void}. This includes an expression-statement or the left-hand
4294 side of a comma expression that contains no side effects. For example,
4295 an expression such as @code{x[i,j]} causes a warning, while
4296 @code{x[(void)i,j]} does not.
4298 This warning is enabled by @option{-Wall}.
4303 All the above @option{-Wunused} options combined.
4305 In order to get a warning about an unused function parameter, you must
4306 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4307 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4309 @item -Wuninitialized
4310 @opindex Wuninitialized
4311 @opindex Wno-uninitialized
4312 Warn if an automatic variable is used without first being initialized
4313 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4314 warn if a non-static reference or non-static @code{const} member
4315 appears in a class without constructors.
4317 If you want to warn about code that uses the uninitialized value of the
4318 variable in its own initializer, use the @option{-Winit-self} option.
4320 These warnings occur for individual uninitialized or clobbered
4321 elements of structure, union or array variables as well as for
4322 variables that are uninitialized or clobbered as a whole. They do
4323 not occur for variables or elements declared @code{volatile}. Because
4324 these warnings depend on optimization, the exact variables or elements
4325 for which there are warnings depends on the precise optimization
4326 options and version of GCC used.
4328 Note that there may be no warning about a variable that is used only
4329 to compute a value that itself is never used, because such
4330 computations may be deleted by data flow analysis before the warnings
4333 @item -Winvalid-memory-model
4334 @opindex Winvalid-memory-model
4335 @opindex Wno-invalid-memory-model
4336 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4337 and the C11 atomic generic functions with a memory consistency argument
4338 that is either invalid for the operation or outside the range of values
4339 of the @code{memory_order} enumeration. For example, since the
4340 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4341 defined for the relaxed, release, and sequentially consistent memory
4342 orders the following code is diagnosed:
4347 __atomic_store_n (i, 0, memory_order_consume);
4351 @option{-Winvalid-memory-model} is enabled by default.
4353 @item -Wmaybe-uninitialized
4354 @opindex Wmaybe-uninitialized
4355 @opindex Wno-maybe-uninitialized
4356 For an automatic variable, if there exists a path from the function
4357 entry to a use of the variable that is initialized, but there exist
4358 some other paths for which the variable is not initialized, the compiler
4359 emits a warning if it cannot prove the uninitialized paths are not
4360 executed at run time. These warnings are made optional because GCC is
4361 not smart enough to see all the reasons why the code might be correct
4362 in spite of appearing to have an error. Here is one example of how
4383 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4384 always initialized, but GCC doesn't know this. To suppress the
4385 warning, you need to provide a default case with assert(0) or
4388 @cindex @code{longjmp} warnings
4389 This option also warns when a non-volatile automatic variable might be
4390 changed by a call to @code{longjmp}. These warnings as well are possible
4391 only in optimizing compilation.
4393 The compiler sees only the calls to @code{setjmp}. It cannot know
4394 where @code{longjmp} will be called; in fact, a signal handler could
4395 call it at any point in the code. As a result, you may get a warning
4396 even when there is in fact no problem because @code{longjmp} cannot
4397 in fact be called at the place that would cause a problem.
4399 Some spurious warnings can be avoided if you declare all the functions
4400 you use that never return as @code{noreturn}. @xref{Function
4403 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4405 @item -Wunknown-pragmas
4406 @opindex Wunknown-pragmas
4407 @opindex Wno-unknown-pragmas
4408 @cindex warning for unknown pragmas
4409 @cindex unknown pragmas, warning
4410 @cindex pragmas, warning of unknown
4411 Warn when a @code{#pragma} directive is encountered that is not understood by
4412 GCC@. If this command-line option is used, warnings are even issued
4413 for unknown pragmas in system header files. This is not the case if
4414 the warnings are only enabled by the @option{-Wall} command-line option.
4417 @opindex Wno-pragmas
4419 Do not warn about misuses of pragmas, such as incorrect parameters,
4420 invalid syntax, or conflicts between pragmas. See also
4421 @option{-Wunknown-pragmas}.
4423 @item -Wstrict-aliasing
4424 @opindex Wstrict-aliasing
4425 @opindex Wno-strict-aliasing
4426 This option is only active when @option{-fstrict-aliasing} is active.
4427 It warns about code that might break the strict aliasing rules that the
4428 compiler is using for optimization. The warning does not catch all
4429 cases, but does attempt to catch the more common pitfalls. It is
4430 included in @option{-Wall}.
4431 It is equivalent to @option{-Wstrict-aliasing=3}
4433 @item -Wstrict-aliasing=n
4434 @opindex Wstrict-aliasing=n
4435 This option is only active when @option{-fstrict-aliasing} is active.
4436 It warns about code that might break the strict aliasing rules that the
4437 compiler is using for optimization.
4438 Higher levels correspond to higher accuracy (fewer false positives).
4439 Higher levels also correspond to more effort, similar to the way @option{-O}
4441 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4443 Level 1: Most aggressive, quick, least accurate.
4444 Possibly useful when higher levels
4445 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4446 false negatives. However, it has many false positives.
4447 Warns for all pointer conversions between possibly incompatible types,
4448 even if never dereferenced. Runs in the front end only.
4450 Level 2: Aggressive, quick, not too precise.
4451 May still have many false positives (not as many as level 1 though),
4452 and few false negatives (but possibly more than level 1).
4453 Unlike level 1, it only warns when an address is taken. Warns about
4454 incomplete types. Runs in the front end only.
4456 Level 3 (default for @option{-Wstrict-aliasing}):
4457 Should have very few false positives and few false
4458 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4459 Takes care of the common pun+dereference pattern in the front end:
4460 @code{*(int*)&some_float}.
4461 If optimization is enabled, it also runs in the back end, where it deals
4462 with multiple statement cases using flow-sensitive points-to information.
4463 Only warns when the converted pointer is dereferenced.
4464 Does not warn about incomplete types.
4466 @item -Wstrict-overflow
4467 @itemx -Wstrict-overflow=@var{n}
4468 @opindex Wstrict-overflow
4469 @opindex Wno-strict-overflow
4470 This option is only active when @option{-fstrict-overflow} is active.
4471 It warns about cases where the compiler optimizes based on the
4472 assumption that signed overflow does not occur. Note that it does not
4473 warn about all cases where the code might overflow: it only warns
4474 about cases where the compiler implements some optimization. Thus
4475 this warning depends on the optimization level.
4477 An optimization that assumes that signed overflow does not occur is
4478 perfectly safe if the values of the variables involved are such that
4479 overflow never does, in fact, occur. Therefore this warning can
4480 easily give a false positive: a warning about code that is not
4481 actually a problem. To help focus on important issues, several
4482 warning levels are defined. No warnings are issued for the use of
4483 undefined signed overflow when estimating how many iterations a loop
4484 requires, in particular when determining whether a loop will be
4488 @item -Wstrict-overflow=1
4489 Warn about cases that are both questionable and easy to avoid. For
4490 example, with @option{-fstrict-overflow}, the compiler simplifies
4491 @code{x + 1 > x} to @code{1}. This level of
4492 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4493 are not, and must be explicitly requested.
4495 @item -Wstrict-overflow=2
4496 Also warn about other cases where a comparison is simplified to a
4497 constant. For example: @code{abs (x) >= 0}. This can only be
4498 simplified when @option{-fstrict-overflow} is in effect, because
4499 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4500 zero. @option{-Wstrict-overflow} (with no level) is the same as
4501 @option{-Wstrict-overflow=2}.
4503 @item -Wstrict-overflow=3
4504 Also warn about other cases where a comparison is simplified. For
4505 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4507 @item -Wstrict-overflow=4
4508 Also warn about other simplifications not covered by the above cases.
4509 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4511 @item -Wstrict-overflow=5
4512 Also warn about cases where the compiler reduces the magnitude of a
4513 constant involved in a comparison. For example: @code{x + 2 > y} is
4514 simplified to @code{x + 1 >= y}. This is reported only at the
4515 highest warning level because this simplification applies to many
4516 comparisons, so this warning level gives a very large number of
4520 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4521 @opindex Wsuggest-attribute=
4522 @opindex Wno-suggest-attribute=
4523 Warn for cases where adding an attribute may be beneficial. The
4524 attributes currently supported are listed below.
4527 @item -Wsuggest-attribute=pure
4528 @itemx -Wsuggest-attribute=const
4529 @itemx -Wsuggest-attribute=noreturn
4530 @opindex Wsuggest-attribute=pure
4531 @opindex Wno-suggest-attribute=pure
4532 @opindex Wsuggest-attribute=const
4533 @opindex Wno-suggest-attribute=const
4534 @opindex Wsuggest-attribute=noreturn
4535 @opindex Wno-suggest-attribute=noreturn
4537 Warn about functions that might be candidates for attributes
4538 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4539 functions visible in other compilation units or (in the case of @code{pure} and
4540 @code{const}) if it cannot prove that the function returns normally. A function
4541 returns normally if it doesn't contain an infinite loop or return abnormally
4542 by throwing, calling @code{abort} or trapping. This analysis requires option
4543 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4544 higher. Higher optimization levels improve the accuracy of the analysis.
4546 @item -Wsuggest-attribute=format
4547 @itemx -Wmissing-format-attribute
4548 @opindex Wsuggest-attribute=format
4549 @opindex Wmissing-format-attribute
4550 @opindex Wno-suggest-attribute=format
4551 @opindex Wno-missing-format-attribute
4555 Warn about function pointers that might be candidates for @code{format}
4556 attributes. Note these are only possible candidates, not absolute ones.
4557 GCC guesses that function pointers with @code{format} attributes that
4558 are used in assignment, initialization, parameter passing or return
4559 statements should have a corresponding @code{format} attribute in the
4560 resulting type. I.e.@: the left-hand side of the assignment or
4561 initialization, the type of the parameter variable, or the return type
4562 of the containing function respectively should also have a @code{format}
4563 attribute to avoid the warning.
4565 GCC also warns about function definitions that might be
4566 candidates for @code{format} attributes. Again, these are only
4567 possible candidates. GCC guesses that @code{format} attributes
4568 might be appropriate for any function that calls a function like
4569 @code{vprintf} or @code{vscanf}, but this might not always be the
4570 case, and some functions for which @code{format} attributes are
4571 appropriate may not be detected.
4574 @item -Wsuggest-final-types
4575 @opindex Wno-suggest-final-types
4576 @opindex Wsuggest-final-types
4577 Warn about types with virtual methods where code quality would be improved
4578 if the type were declared with the C++11 @code{final} specifier,
4580 declared in an anonymous namespace. This allows GCC to more aggressively
4581 devirtualize the polymorphic calls. This warning is more effective with link
4582 time optimization, where the information about the class hierarchy graph is
4585 @item -Wsuggest-final-methods
4586 @opindex Wno-suggest-final-methods
4587 @opindex Wsuggest-final-methods
4588 Warn about virtual methods where code quality would be improved if the method
4589 were declared with the C++11 @code{final} specifier,
4590 or, if possible, its type were
4591 declared in an anonymous namespace or with the @code{final} specifier.
4593 more effective with link time optimization, where the information about the
4594 class hierarchy graph is more complete. It is recommended to first consider
4595 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4598 @item -Wsuggest-override
4599 Warn about overriding virtual functions that are not marked with the override
4602 @item -Warray-bounds
4603 @itemx -Warray-bounds=@var{n}
4604 @opindex Wno-array-bounds
4605 @opindex Warray-bounds
4606 This option is only active when @option{-ftree-vrp} is active
4607 (default for @option{-O2} and above). It warns about subscripts to arrays
4608 that are always out of bounds. This warning is enabled by @option{-Wall}.
4611 @item -Warray-bounds=1
4612 This is the warning level of @option{-Warray-bounds} and is enabled
4613 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4615 @item -Warray-bounds=2
4616 This warning level also warns about out of bounds access for
4617 arrays at the end of a struct and for arrays accessed through
4618 pointers. This warning level may give a larger number of
4619 false positives and is deactivated by default.
4622 @item -Wbool-compare
4623 @opindex Wno-bool-compare
4624 @opindex Wbool-compare
4625 Warn about boolean expression compared with an integer value different from
4626 @code{true}/@code{false}. For instance, the following comparison is
4631 if ((n > 1) == 2) @{ @dots{} @}
4633 This warning is enabled by @option{-Wall}.
4635 @item -Wduplicated-cond
4636 @opindex Wno-duplicated-cond
4637 @opindex Wduplicated-cond
4638 Warn about duplicated conditions in an if-else-if chain. For instance,
4639 warn for the following code:
4641 if (p->q != NULL) @{ @dots{} @}
4642 else if (p->q != NULL) @{ @dots{} @}
4645 @item -Wframe-address
4646 @opindex Wno-frame-address
4647 @opindex Wframe-address
4648 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
4649 is called with an argument greater than 0. Such calls may return indeterminate
4650 values or crash the program. The warning is included in @option{-Wall}.
4652 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4653 @opindex Wno-discarded-qualifiers
4654 @opindex Wdiscarded-qualifiers
4655 Do not warn if type qualifiers on pointers are being discarded.
4656 Typically, the compiler warns if a @code{const char *} variable is
4657 passed to a function that takes a @code{char *} parameter. This option
4658 can be used to suppress such a warning.
4660 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4661 @opindex Wno-discarded-array-qualifiers
4662 @opindex Wdiscarded-array-qualifiers
4663 Do not warn if type qualifiers on arrays which are pointer targets
4664 are being discarded. Typically, the compiler warns if a
4665 @code{const int (*)[]} variable is passed to a function that
4666 takes a @code{int (*)[]} parameter. This option can be used to
4667 suppress such a warning.
4669 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4670 @opindex Wno-incompatible-pointer-types
4671 @opindex Wincompatible-pointer-types
4672 Do not warn when there is a conversion between pointers that have incompatible
4673 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4674 which warns for pointer argument passing or assignment with different
4677 @item -Wno-int-conversion @r{(C and Objective-C only)}
4678 @opindex Wno-int-conversion
4679 @opindex Wint-conversion
4680 Do not warn about incompatible integer to pointer and pointer to integer
4681 conversions. This warning is about implicit conversions; for explicit
4682 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4683 @option{-Wno-pointer-to-int-cast} may be used.
4685 @item -Wno-div-by-zero
4686 @opindex Wno-div-by-zero
4687 @opindex Wdiv-by-zero
4688 Do not warn about compile-time integer division by zero. Floating-point
4689 division by zero is not warned about, as it can be a legitimate way of
4690 obtaining infinities and NaNs.
4692 @item -Wsystem-headers
4693 @opindex Wsystem-headers
4694 @opindex Wno-system-headers
4695 @cindex warnings from system headers
4696 @cindex system headers, warnings from
4697 Print warning messages for constructs found in system header files.
4698 Warnings from system headers are normally suppressed, on the assumption
4699 that they usually do not indicate real problems and would only make the
4700 compiler output harder to read. Using this command-line option tells
4701 GCC to emit warnings from system headers as if they occurred in user
4702 code. However, note that using @option{-Wall} in conjunction with this
4703 option does @emph{not} warn about unknown pragmas in system
4704 headers---for that, @option{-Wunknown-pragmas} must also be used.
4706 @item -Wtautological-compare
4707 @opindex Wtautological-compare
4708 @opindex Wno-tautological-compare
4709 Warn if a self-comparison always evaluates to true or false. This
4710 warning detects various mistakes such as:
4714 if (i > i) @{ @dots{} @}
4716 This warning is enabled by @option{-Wall}.
4719 @opindex Wtrampolines
4720 @opindex Wno-trampolines
4721 Warn about trampolines generated for pointers to nested functions.
4722 A trampoline is a small piece of data or code that is created at run
4723 time on the stack when the address of a nested function is taken, and is
4724 used to call the nested function indirectly. For some targets, it is
4725 made up of data only and thus requires no special treatment. But, for
4726 most targets, it is made up of code and thus requires the stack to be
4727 made executable in order for the program to work properly.
4730 @opindex Wfloat-equal
4731 @opindex Wno-float-equal
4732 Warn if floating-point values are used in equality comparisons.
4734 The idea behind this is that sometimes it is convenient (for the
4735 programmer) to consider floating-point values as approximations to
4736 infinitely precise real numbers. If you are doing this, then you need
4737 to compute (by analyzing the code, or in some other way) the maximum or
4738 likely maximum error that the computation introduces, and allow for it
4739 when performing comparisons (and when producing output, but that's a
4740 different problem). In particular, instead of testing for equality, you
4741 should check to see whether the two values have ranges that overlap; and
4742 this is done with the relational operators, so equality comparisons are
4745 @item -Wtraditional @r{(C and Objective-C only)}
4746 @opindex Wtraditional
4747 @opindex Wno-traditional
4748 Warn about certain constructs that behave differently in traditional and
4749 ISO C@. Also warn about ISO C constructs that have no traditional C
4750 equivalent, and/or problematic constructs that should be avoided.
4754 Macro parameters that appear within string literals in the macro body.
4755 In traditional C macro replacement takes place within string literals,
4756 but in ISO C it does not.
4759 In traditional C, some preprocessor directives did not exist.
4760 Traditional preprocessors only considered a line to be a directive
4761 if the @samp{#} appeared in column 1 on the line. Therefore
4762 @option{-Wtraditional} warns about directives that traditional C
4763 understands but ignores because the @samp{#} does not appear as the
4764 first character on the line. It also suggests you hide directives like
4765 @code{#pragma} not understood by traditional C by indenting them. Some
4766 traditional implementations do not recognize @code{#elif}, so this option
4767 suggests avoiding it altogether.
4770 A function-like macro that appears without arguments.
4773 The unary plus operator.
4776 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4777 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4778 constants.) Note, these suffixes appear in macros defined in the system
4779 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4780 Use of these macros in user code might normally lead to spurious
4781 warnings, however GCC's integrated preprocessor has enough context to
4782 avoid warning in these cases.
4785 A function declared external in one block and then used after the end of
4789 A @code{switch} statement has an operand of type @code{long}.
4792 A non-@code{static} function declaration follows a @code{static} one.
4793 This construct is not accepted by some traditional C compilers.
4796 The ISO type of an integer constant has a different width or
4797 signedness from its traditional type. This warning is only issued if
4798 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4799 typically represent bit patterns, are not warned about.
4802 Usage of ISO string concatenation is detected.
4805 Initialization of automatic aggregates.
4808 Identifier conflicts with labels. Traditional C lacks a separate
4809 namespace for labels.
4812 Initialization of unions. If the initializer is zero, the warning is
4813 omitted. This is done under the assumption that the zero initializer in
4814 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4815 initializer warnings and relies on default initialization to zero in the
4819 Conversions by prototypes between fixed/floating-point values and vice
4820 versa. The absence of these prototypes when compiling with traditional
4821 C causes serious problems. This is a subset of the possible
4822 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4825 Use of ISO C style function definitions. This warning intentionally is
4826 @emph{not} issued for prototype declarations or variadic functions
4827 because these ISO C features appear in your code when using
4828 libiberty's traditional C compatibility macros, @code{PARAMS} and
4829 @code{VPARAMS}. This warning is also bypassed for nested functions
4830 because that feature is already a GCC extension and thus not relevant to
4831 traditional C compatibility.
4834 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4835 @opindex Wtraditional-conversion
4836 @opindex Wno-traditional-conversion
4837 Warn if a prototype causes a type conversion that is different from what
4838 would happen to the same argument in the absence of a prototype. This
4839 includes conversions of fixed point to floating and vice versa, and
4840 conversions changing the width or signedness of a fixed-point argument
4841 except when the same as the default promotion.
4843 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4844 @opindex Wdeclaration-after-statement
4845 @opindex Wno-declaration-after-statement
4846 Warn when a declaration is found after a statement in a block. This
4847 construct, known from C++, was introduced with ISO C99 and is by default
4848 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4853 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4855 @item -Wno-endif-labels
4856 @opindex Wno-endif-labels
4857 @opindex Wendif-labels
4858 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4863 Warn whenever a local variable or type declaration shadows another
4864 variable, parameter, type, class member (in C++), or instance variable
4865 (in Objective-C) or whenever a built-in function is shadowed. Note
4866 that in C++, the compiler warns if a local variable shadows an
4867 explicit typedef, but not if it shadows a struct/class/enum.
4869 @item -Wno-shadow-ivar @r{(Objective-C only)}
4870 @opindex Wno-shadow-ivar
4871 @opindex Wshadow-ivar
4872 Do not warn whenever a local variable shadows an instance variable in an
4875 @item -Wlarger-than=@var{len}
4876 @opindex Wlarger-than=@var{len}
4877 @opindex Wlarger-than-@var{len}
4878 Warn whenever an object of larger than @var{len} bytes is defined.
4880 @item -Wframe-larger-than=@var{len}
4881 @opindex Wframe-larger-than
4882 Warn if the size of a function frame is larger than @var{len} bytes.
4883 The computation done to determine the stack frame size is approximate
4884 and not conservative.
4885 The actual requirements may be somewhat greater than @var{len}
4886 even if you do not get a warning. In addition, any space allocated
4887 via @code{alloca}, variable-length arrays, or related constructs
4888 is not included by the compiler when determining
4889 whether or not to issue a warning.
4891 @item -Wno-free-nonheap-object
4892 @opindex Wno-free-nonheap-object
4893 @opindex Wfree-nonheap-object
4894 Do not warn when attempting to free an object that was not allocated
4897 @item -Wstack-usage=@var{len}
4898 @opindex Wstack-usage
4899 Warn if the stack usage of a function might be larger than @var{len} bytes.
4900 The computation done to determine the stack usage is conservative.
4901 Any space allocated via @code{alloca}, variable-length arrays, or related
4902 constructs is included by the compiler when determining whether or not to
4905 The message is in keeping with the output of @option{-fstack-usage}.
4909 If the stack usage is fully static but exceeds the specified amount, it's:
4912 warning: stack usage is 1120 bytes
4915 If the stack usage is (partly) dynamic but bounded, it's:
4918 warning: stack usage might be 1648 bytes
4921 If the stack usage is (partly) dynamic and not bounded, it's:
4924 warning: stack usage might be unbounded
4928 @item -Wunsafe-loop-optimizations
4929 @opindex Wunsafe-loop-optimizations
4930 @opindex Wno-unsafe-loop-optimizations
4931 Warn if the loop cannot be optimized because the compiler cannot
4932 assume anything on the bounds of the loop indices. With
4933 @option{-funsafe-loop-optimizations} warn if the compiler makes
4936 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4937 @opindex Wno-pedantic-ms-format
4938 @opindex Wpedantic-ms-format
4939 When used in combination with @option{-Wformat}
4940 and @option{-pedantic} without GNU extensions, this option
4941 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4942 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4943 which depend on the MS runtime.
4945 @item -Wplacement-new
4946 @itemx -Wplacement-new=@var{n}
4947 @opindex Wplacement-new
4948 @opindex Wno-placement-new
4949 Warn about placement new expressions with undefined behavior, such as
4950 constructing an object in a buffer that is smaller than the type of
4951 the object. For example, the placement new expression below is diagnosed
4952 because it attempts to construct an array of 64 integers in a buffer only
4958 This warning is enabled by default.
4961 @item -Wplacement-new=1
4962 This is the default warning level of @option{-Wplacement-new}. At this
4963 level the warning is not issued for some strictly undefined constructs that
4964 GCC allows as extensions for compatibility with legacy code. For example,
4965 the following @code{new} expression is not diagnosed at this level even
4966 though it has undefined behavior according to the C++ standard because
4967 it writes past the end of the one-element array.
4969 struct S @{ int n, a[1]; @};
4970 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
4971 new (s->a)int [32]();
4974 @item -Wplacement-new=2
4975 At this level, in addition to diagnosing all the same constructs as at level
4976 1, a diagnostic is also issued for placement new expressions that construct
4977 an object in the last member of structure whose type is an array of a single
4978 element and whose size is less than the size of the object being constructed.
4979 While the previous example would be diagnosed, the following construct makes
4980 use of the flexible member array extension to avoid the warning at level 2.
4982 struct S @{ int n, a[]; @};
4983 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
4984 new (s->a)int [32]();
4989 @item -Wpointer-arith
4990 @opindex Wpointer-arith
4991 @opindex Wno-pointer-arith
4992 Warn about anything that depends on the ``size of'' a function type or
4993 of @code{void}. GNU C assigns these types a size of 1, for
4994 convenience in calculations with @code{void *} pointers and pointers
4995 to functions. In C++, warn also when an arithmetic operation involves
4996 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4999 @opindex Wtype-limits
5000 @opindex Wno-type-limits
5001 Warn if a comparison is always true or always false due to the limited
5002 range of the data type, but do not warn for constant expressions. For
5003 example, warn if an unsigned variable is compared against zero with
5004 @code{<} or @code{>=}. This warning is also enabled by
5007 @item -Wbad-function-cast @r{(C and Objective-C only)}
5008 @opindex Wbad-function-cast
5009 @opindex Wno-bad-function-cast
5010 Warn when a function call is cast to a non-matching type.
5011 For example, warn if a call to a function returning an integer type
5012 is cast to a pointer type.
5014 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5015 @opindex Wc90-c99-compat
5016 @opindex Wno-c90-c99-compat
5017 Warn about features not present in ISO C90, but present in ISO C99.
5018 For instance, warn about use of variable length arrays, @code{long long}
5019 type, @code{bool} type, compound literals, designated initializers, and so
5020 on. This option is independent of the standards mode. Warnings are disabled
5021 in the expression that follows @code{__extension__}.
5023 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5024 @opindex Wc99-c11-compat
5025 @opindex Wno-c99-c11-compat
5026 Warn about features not present in ISO C99, but present in ISO C11.
5027 For instance, warn about use of anonymous structures and unions,
5028 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5029 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5030 and so on. This option is independent of the standards mode. Warnings are
5031 disabled in the expression that follows @code{__extension__}.
5033 @item -Wc++-compat @r{(C and Objective-C only)}
5034 @opindex Wc++-compat
5035 Warn about ISO C constructs that are outside of the common subset of
5036 ISO C and ISO C++, e.g.@: request for implicit conversion from
5037 @code{void *} to a pointer to non-@code{void} type.
5039 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5040 @opindex Wc++11-compat
5041 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5042 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5043 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5044 enabled by @option{-Wall}.
5046 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5047 @opindex Wc++14-compat
5048 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5049 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5053 @opindex Wno-cast-qual
5054 Warn whenever a pointer is cast so as to remove a type qualifier from
5055 the target type. For example, warn if a @code{const char *} is cast
5056 to an ordinary @code{char *}.
5058 Also warn when making a cast that introduces a type qualifier in an
5059 unsafe way. For example, casting @code{char **} to @code{const char **}
5060 is unsafe, as in this example:
5063 /* p is char ** value. */
5064 const char **q = (const char **) p;
5065 /* Assignment of readonly string to const char * is OK. */
5067 /* Now char** pointer points to read-only memory. */
5072 @opindex Wcast-align
5073 @opindex Wno-cast-align
5074 Warn whenever a pointer is cast such that the required alignment of the
5075 target is increased. For example, warn if a @code{char *} is cast to
5076 an @code{int *} on machines where integers can only be accessed at
5077 two- or four-byte boundaries.
5079 @item -Wwrite-strings
5080 @opindex Wwrite-strings
5081 @opindex Wno-write-strings
5082 When compiling C, give string constants the type @code{const
5083 char[@var{length}]} so that copying the address of one into a
5084 non-@code{const} @code{char *} pointer produces a warning. These
5085 warnings help you find at compile time code that can try to write
5086 into a string constant, but only if you have been very careful about
5087 using @code{const} in declarations and prototypes. Otherwise, it is
5088 just a nuisance. This is why we did not make @option{-Wall} request
5091 When compiling C++, warn about the deprecated conversion from string
5092 literals to @code{char *}. This warning is enabled by default for C++
5097 @opindex Wno-clobbered
5098 Warn for variables that might be changed by @code{longjmp} or
5099 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5101 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5102 @opindex Wconditionally-supported
5103 @opindex Wno-conditionally-supported
5104 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5107 @opindex Wconversion
5108 @opindex Wno-conversion
5109 Warn for implicit conversions that may alter a value. This includes
5110 conversions between real and integer, like @code{abs (x)} when
5111 @code{x} is @code{double}; conversions between signed and unsigned,
5112 like @code{unsigned ui = -1}; and conversions to smaller types, like
5113 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5114 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5115 changed by the conversion like in @code{abs (2.0)}. Warnings about
5116 conversions between signed and unsigned integers can be disabled by
5117 using @option{-Wno-sign-conversion}.
5119 For C++, also warn for confusing overload resolution for user-defined
5120 conversions; and conversions that never use a type conversion
5121 operator: conversions to @code{void}, the same type, a base class or a
5122 reference to them. Warnings about conversions between signed and
5123 unsigned integers are disabled by default in C++ unless
5124 @option{-Wsign-conversion} is explicitly enabled.
5126 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5127 @opindex Wconversion-null
5128 @opindex Wno-conversion-null
5129 Do not warn for conversions between @code{NULL} and non-pointer
5130 types. @option{-Wconversion-null} is enabled by default.
5132 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5133 @opindex Wzero-as-null-pointer-constant
5134 @opindex Wno-zero-as-null-pointer-constant
5135 Warn when a literal @samp{0} is used as null pointer constant. This can
5136 be useful to facilitate the conversion to @code{nullptr} in C++11.
5138 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5139 @opindex Wsubobject-linkage
5140 @opindex Wno-subobject-linkage
5141 Warn if a class type has a base or a field whose type uses the anonymous
5142 namespace or depends on a type with no linkage. If a type A depends on
5143 a type B with no or internal linkage, defining it in multiple
5144 translation units would be an ODR violation because the meaning of B
5145 is different in each translation unit. If A only appears in a single
5146 translation unit, the best way to silence the warning is to give it
5147 internal linkage by putting it in an anonymous namespace as well. The
5148 compiler doesn't give this warning for types defined in the main .C
5149 file, as those are unlikely to have multiple definitions.
5150 @option{-Wsubobject-linkage} is enabled by default.
5154 @opindex Wno-date-time
5155 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5156 are encountered as they might prevent bit-wise-identical reproducible
5159 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5160 @opindex Wdelete-incomplete
5161 @opindex Wno-delete-incomplete
5162 Warn when deleting a pointer to incomplete type, which may cause
5163 undefined behavior at runtime. This warning is enabled by default.
5165 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5166 @opindex Wuseless-cast
5167 @opindex Wno-useless-cast
5168 Warn when an expression is casted to its own type.
5171 @opindex Wempty-body
5172 @opindex Wno-empty-body
5173 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5174 while} statement. This warning is also enabled by @option{-Wextra}.
5176 @item -Wenum-compare
5177 @opindex Wenum-compare
5178 @opindex Wno-enum-compare
5179 Warn about a comparison between values of different enumerated types.
5180 In C++ enumeral mismatches in conditional expressions are also
5181 diagnosed and the warning is enabled by default. In C this warning is
5182 enabled by @option{-Wall}.
5184 @item -Wjump-misses-init @r{(C, Objective-C only)}
5185 @opindex Wjump-misses-init
5186 @opindex Wno-jump-misses-init
5187 Warn if a @code{goto} statement or a @code{switch} statement jumps
5188 forward across the initialization of a variable, or jumps backward to a
5189 label after the variable has been initialized. This only warns about
5190 variables that are initialized when they are declared. This warning is
5191 only supported for C and Objective-C; in C++ this sort of branch is an
5194 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5195 can be disabled with the @option{-Wno-jump-misses-init} option.
5197 @item -Wsign-compare
5198 @opindex Wsign-compare
5199 @opindex Wno-sign-compare
5200 @cindex warning for comparison of signed and unsigned values
5201 @cindex comparison of signed and unsigned values, warning
5202 @cindex signed and unsigned values, comparison warning
5203 Warn when a comparison between signed and unsigned values could produce
5204 an incorrect result when the signed value is converted to unsigned.
5205 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5206 also enabled by @option{-Wextra}.
5208 @item -Wsign-conversion
5209 @opindex Wsign-conversion
5210 @opindex Wno-sign-conversion
5211 Warn for implicit conversions that may change the sign of an integer
5212 value, like assigning a signed integer expression to an unsigned
5213 integer variable. An explicit cast silences the warning. In C, this
5214 option is enabled also by @option{-Wconversion}.
5216 @item -Wfloat-conversion
5217 @opindex Wfloat-conversion
5218 @opindex Wno-float-conversion
5219 Warn for implicit conversions that reduce the precision of a real value.
5220 This includes conversions from real to integer, and from higher precision
5221 real to lower precision real values. This option is also enabled by
5222 @option{-Wconversion}.
5224 @item -Wno-scalar-storage-order
5225 @opindex -Wno-scalar-storage-order
5226 @opindex -Wscalar-storage-order
5227 Do not warn on suspicious constructs involving reverse scalar storage order.
5229 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5230 @opindex Wsized-deallocation
5231 @opindex Wno-sized-deallocation
5232 Warn about a definition of an unsized deallocation function
5234 void operator delete (void *) noexcept;
5235 void operator delete[] (void *) noexcept;
5237 without a definition of the corresponding sized deallocation function
5239 void operator delete (void *, std::size_t) noexcept;
5240 void operator delete[] (void *, std::size_t) noexcept;
5242 or vice versa. Enabled by @option{-Wextra} along with
5243 @option{-fsized-deallocation}.
5245 @item -Wsizeof-pointer-memaccess
5246 @opindex Wsizeof-pointer-memaccess
5247 @opindex Wno-sizeof-pointer-memaccess
5248 Warn for suspicious length parameters to certain string and memory built-in
5249 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5250 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5251 but a pointer, and suggests a possible fix, or about
5252 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5255 @item -Wsizeof-array-argument
5256 @opindex Wsizeof-array-argument
5257 @opindex Wno-sizeof-array-argument
5258 Warn when the @code{sizeof} operator is applied to a parameter that is
5259 declared as an array in a function definition. This warning is enabled by
5260 default for C and C++ programs.
5262 @item -Wmemset-transposed-args
5263 @opindex Wmemset-transposed-args
5264 @opindex Wno-memset-transposed-args
5265 Warn for suspicious calls to the @code{memset} built-in function, if the
5266 second argument is not zero and the third argument is zero. This warns e.g.@
5267 about @code{memset (buf, sizeof buf, 0)} where most probably
5268 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5269 is only emitted if the third argument is literal zero. If it is some
5270 expression that is folded to zero, a cast of zero to some type, etc.,
5271 it is far less likely that the user has mistakenly exchanged the arguments
5272 and no warning is emitted. This warning is enabled by @option{-Wall}.
5276 @opindex Wno-address
5277 Warn about suspicious uses of memory addresses. These include using
5278 the address of a function in a conditional expression, such as
5279 @code{void func(void); if (func)}, and comparisons against the memory
5280 address of a string literal, such as @code{if (x == "abc")}. Such
5281 uses typically indicate a programmer error: the address of a function
5282 always evaluates to true, so their use in a conditional usually
5283 indicate that the programmer forgot the parentheses in a function
5284 call; and comparisons against string literals result in unspecified
5285 behavior and are not portable in C, so they usually indicate that the
5286 programmer intended to use @code{strcmp}. This warning is enabled by
5290 @opindex Wlogical-op
5291 @opindex Wno-logical-op
5292 Warn about suspicious uses of logical operators in expressions.
5293 This includes using logical operators in contexts where a
5294 bit-wise operator is likely to be expected. Also warns when
5295 the operands of a logical operator are the same:
5298 if (a < 0 && a < 0) @{ @dots{} @}
5301 @item -Wlogical-not-parentheses
5302 @opindex Wlogical-not-parentheses
5303 @opindex Wno-logical-not-parentheses
5304 Warn about logical not used on the left hand side operand of a comparison.
5305 This option does not warn if the RHS operand is of a boolean type. Its
5306 purpose is to detect suspicious code like the following:
5310 if (!a > 1) @{ @dots{} @}
5313 It is possible to suppress the warning by wrapping the LHS into
5316 if ((!a) > 1) @{ @dots{} @}
5319 This warning is enabled by @option{-Wall}.
5321 @item -Waggregate-return
5322 @opindex Waggregate-return
5323 @opindex Wno-aggregate-return
5324 Warn if any functions that return structures or unions are defined or
5325 called. (In languages where you can return an array, this also elicits
5328 @item -Wno-aggressive-loop-optimizations
5329 @opindex Wno-aggressive-loop-optimizations
5330 @opindex Waggressive-loop-optimizations
5331 Warn if in a loop with constant number of iterations the compiler detects
5332 undefined behavior in some statement during one or more of the iterations.
5334 @item -Wno-attributes
5335 @opindex Wno-attributes
5336 @opindex Wattributes
5337 Do not warn if an unexpected @code{__attribute__} is used, such as
5338 unrecognized attributes, function attributes applied to variables,
5339 etc. This does not stop errors for incorrect use of supported
5342 @item -Wno-builtin-macro-redefined
5343 @opindex Wno-builtin-macro-redefined
5344 @opindex Wbuiltin-macro-redefined
5345 Do not warn if certain built-in macros are redefined. This suppresses
5346 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5347 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5349 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5350 @opindex Wstrict-prototypes
5351 @opindex Wno-strict-prototypes
5352 Warn if a function is declared or defined without specifying the
5353 argument types. (An old-style function definition is permitted without
5354 a warning if preceded by a declaration that specifies the argument
5357 @item -Wold-style-declaration @r{(C and Objective-C only)}
5358 @opindex Wold-style-declaration
5359 @opindex Wno-old-style-declaration
5360 Warn for obsolescent usages, according to the C Standard, in a
5361 declaration. For example, warn if storage-class specifiers like
5362 @code{static} are not the first things in a declaration. This warning
5363 is also enabled by @option{-Wextra}.
5365 @item -Wold-style-definition @r{(C and Objective-C only)}
5366 @opindex Wold-style-definition
5367 @opindex Wno-old-style-definition
5368 Warn if an old-style function definition is used. A warning is given
5369 even if there is a previous prototype.
5371 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5372 @opindex Wmissing-parameter-type
5373 @opindex Wno-missing-parameter-type
5374 A function parameter is declared without a type specifier in K&R-style
5381 This warning is also enabled by @option{-Wextra}.
5383 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5384 @opindex Wmissing-prototypes
5385 @opindex Wno-missing-prototypes
5386 Warn if a global function is defined without a previous prototype
5387 declaration. This warning is issued even if the definition itself
5388 provides a prototype. Use this option to detect global functions
5389 that do not have a matching prototype declaration in a header file.
5390 This option is not valid for C++ because all function declarations
5391 provide prototypes and a non-matching declaration declares an
5392 overload rather than conflict with an earlier declaration.
5393 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5395 @item -Wmissing-declarations
5396 @opindex Wmissing-declarations
5397 @opindex Wno-missing-declarations
5398 Warn if a global function is defined without a previous declaration.
5399 Do so even if the definition itself provides a prototype.
5400 Use this option to detect global functions that are not declared in
5401 header files. In C, no warnings are issued for functions with previous
5402 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5403 missing prototypes. In C++, no warnings are issued for function templates,
5404 or for inline functions, or for functions in anonymous namespaces.
5406 @item -Wmissing-field-initializers
5407 @opindex Wmissing-field-initializers
5408 @opindex Wno-missing-field-initializers
5412 Warn if a structure's initializer has some fields missing. For
5413 example, the following code causes such a warning, because
5414 @code{x.h} is implicitly zero:
5417 struct s @{ int f, g, h; @};
5418 struct s x = @{ 3, 4 @};
5421 This option does not warn about designated initializers, so the following
5422 modification does not trigger a warning:
5425 struct s @{ int f, g, h; @};
5426 struct s x = @{ .f = 3, .g = 4 @};
5429 In C++ this option does not warn either about the empty @{ @}
5430 initializer, for example:
5433 struct s @{ int f, g, h; @};
5437 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5438 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5440 @item -Wno-multichar
5441 @opindex Wno-multichar
5443 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5444 Usually they indicate a typo in the user's code, as they have
5445 implementation-defined values, and should not be used in portable code.
5447 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5448 @opindex Wnormalized=
5449 @opindex Wnormalized
5450 @opindex Wno-normalized
5453 @cindex character set, input normalization
5454 In ISO C and ISO C++, two identifiers are different if they are
5455 different sequences of characters. However, sometimes when characters
5456 outside the basic ASCII character set are used, you can have two
5457 different character sequences that look the same. To avoid confusion,
5458 the ISO 10646 standard sets out some @dfn{normalization rules} which
5459 when applied ensure that two sequences that look the same are turned into
5460 the same sequence. GCC can warn you if you are using identifiers that
5461 have not been normalized; this option controls that warning.
5463 There are four levels of warning supported by GCC@. The default is
5464 @option{-Wnormalized=nfc}, which warns about any identifier that is
5465 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5466 recommended form for most uses. It is equivalent to
5467 @option{-Wnormalized}.
5469 Unfortunately, there are some characters allowed in identifiers by
5470 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5471 identifiers. That is, there's no way to use these symbols in portable
5472 ISO C or C++ and have all your identifiers in NFC@.
5473 @option{-Wnormalized=id} suppresses the warning for these characters.
5474 It is hoped that future versions of the standards involved will correct
5475 this, which is why this option is not the default.
5477 You can switch the warning off for all characters by writing
5478 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5479 only do this if you are using some other normalization scheme (like
5480 ``D''), because otherwise you can easily create bugs that are
5481 literally impossible to see.
5483 Some characters in ISO 10646 have distinct meanings but look identical
5484 in some fonts or display methodologies, especially once formatting has
5485 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5486 LETTER N'', displays just like a regular @code{n} that has been
5487 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5488 normalization scheme to convert all these into a standard form as
5489 well, and GCC warns if your code is not in NFKC if you use
5490 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5491 about every identifier that contains the letter O because it might be
5492 confused with the digit 0, and so is not the default, but may be
5493 useful as a local coding convention if the programming environment
5494 cannot be fixed to display these characters distinctly.
5496 @item -Wno-deprecated
5497 @opindex Wno-deprecated
5498 @opindex Wdeprecated
5499 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5501 @item -Wno-deprecated-declarations
5502 @opindex Wno-deprecated-declarations
5503 @opindex Wdeprecated-declarations
5504 Do not warn about uses of functions (@pxref{Function Attributes}),
5505 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5506 Attributes}) marked as deprecated by using the @code{deprecated}
5510 @opindex Wno-overflow
5512 Do not warn about compile-time overflow in constant expressions.
5517 Warn about One Definition Rule violations during link-time optimization.
5518 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5521 @opindex Wopenm-simd
5522 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5523 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5524 option can be used to relax the cost model.
5526 @item -Woverride-init @r{(C and Objective-C only)}
5527 @opindex Woverride-init
5528 @opindex Wno-override-init
5532 Warn if an initialized field without side effects is overridden when
5533 using designated initializers (@pxref{Designated Inits, , Designated
5536 This warning is included in @option{-Wextra}. To get other
5537 @option{-Wextra} warnings without this one, use @option{-Wextra
5538 -Wno-override-init}.
5540 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5541 @opindex Woverride-init-side-effects
5542 @opindex Wno-override-init-side-effects
5543 Warn if an initialized field with side effects is overridden when
5544 using designated initializers (@pxref{Designated Inits, , Designated
5545 Initializers}). This warning is enabled by default.
5550 Warn if a structure is given the packed attribute, but the packed
5551 attribute has no effect on the layout or size of the structure.
5552 Such structures may be mis-aligned for little benefit. For
5553 instance, in this code, the variable @code{f.x} in @code{struct bar}
5554 is misaligned even though @code{struct bar} does not itself
5555 have the packed attribute:
5562 @} __attribute__((packed));
5570 @item -Wpacked-bitfield-compat
5571 @opindex Wpacked-bitfield-compat
5572 @opindex Wno-packed-bitfield-compat
5573 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5574 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5575 the change can lead to differences in the structure layout. GCC
5576 informs you when the offset of such a field has changed in GCC 4.4.
5577 For example there is no longer a 4-bit padding between field @code{a}
5578 and @code{b} in this structure:
5585 @} __attribute__ ((packed));
5588 This warning is enabled by default. Use
5589 @option{-Wno-packed-bitfield-compat} to disable this warning.
5594 Warn if padding is included in a structure, either to align an element
5595 of the structure or to align the whole structure. Sometimes when this
5596 happens it is possible to rearrange the fields of the structure to
5597 reduce the padding and so make the structure smaller.
5599 @item -Wredundant-decls
5600 @opindex Wredundant-decls
5601 @opindex Wno-redundant-decls
5602 Warn if anything is declared more than once in the same scope, even in
5603 cases where multiple declaration is valid and changes nothing.
5605 @item -Wnested-externs @r{(C and Objective-C only)}
5606 @opindex Wnested-externs
5607 @opindex Wno-nested-externs
5608 Warn if an @code{extern} declaration is encountered within a function.
5610 @item -Wno-inherited-variadic-ctor
5611 @opindex Winherited-variadic-ctor
5612 @opindex Wno-inherited-variadic-ctor
5613 Suppress warnings about use of C++11 inheriting constructors when the
5614 base class inherited from has a C variadic constructor; the warning is
5615 on by default because the ellipsis is not inherited.
5620 Warn if a function that is declared as inline cannot be inlined.
5621 Even with this option, the compiler does not warn about failures to
5622 inline functions declared in system headers.
5624 The compiler uses a variety of heuristics to determine whether or not
5625 to inline a function. For example, the compiler takes into account
5626 the size of the function being inlined and the amount of inlining
5627 that has already been done in the current function. Therefore,
5628 seemingly insignificant changes in the source program can cause the
5629 warnings produced by @option{-Winline} to appear or disappear.
5631 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5632 @opindex Wno-invalid-offsetof
5633 @opindex Winvalid-offsetof
5634 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5635 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5636 to a non-standard-layout type is undefined. In existing C++ implementations,
5637 however, @code{offsetof} typically gives meaningful results.
5638 This flag is for users who are aware that they are
5639 writing nonportable code and who have deliberately chosen to ignore the
5642 The restrictions on @code{offsetof} may be relaxed in a future version
5643 of the C++ standard.
5645 @item -Wno-int-to-pointer-cast
5646 @opindex Wno-int-to-pointer-cast
5647 @opindex Wint-to-pointer-cast
5648 Suppress warnings from casts to pointer type of an integer of a
5649 different size. In C++, casting to a pointer type of smaller size is
5650 an error. @option{Wint-to-pointer-cast} is enabled by default.
5653 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5654 @opindex Wno-pointer-to-int-cast
5655 @opindex Wpointer-to-int-cast
5656 Suppress warnings from casts from a pointer to an integer type of a
5660 @opindex Winvalid-pch
5661 @opindex Wno-invalid-pch
5662 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5663 the search path but can't be used.
5667 @opindex Wno-long-long
5668 Warn if @code{long long} type is used. This is enabled by either
5669 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5670 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5672 @item -Wvariadic-macros
5673 @opindex Wvariadic-macros
5674 @opindex Wno-variadic-macros
5675 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5676 alternate syntax is used in ISO C99 mode. This is enabled by either
5677 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5678 messages, use @option{-Wno-variadic-macros}.
5682 @opindex Wno-varargs
5683 Warn upon questionable usage of the macros used to handle variable
5684 arguments like @code{va_start}. This is default. To inhibit the
5685 warning messages, use @option{-Wno-varargs}.
5687 @item -Wvector-operation-performance
5688 @opindex Wvector-operation-performance
5689 @opindex Wno-vector-operation-performance
5690 Warn if vector operation is not implemented via SIMD capabilities of the
5691 architecture. Mainly useful for the performance tuning.
5692 Vector operation can be implemented @code{piecewise}, which means that the
5693 scalar operation is performed on every vector element;
5694 @code{in parallel}, which means that the vector operation is implemented
5695 using scalars of wider type, which normally is more performance efficient;
5696 and @code{as a single scalar}, which means that vector fits into a
5699 @item -Wno-virtual-move-assign
5700 @opindex Wvirtual-move-assign
5701 @opindex Wno-virtual-move-assign
5702 Suppress warnings about inheriting from a virtual base with a
5703 non-trivial C++11 move assignment operator. This is dangerous because
5704 if the virtual base is reachable along more than one path, it is
5705 moved multiple times, which can mean both objects end up in the
5706 moved-from state. If the move assignment operator is written to avoid
5707 moving from a moved-from object, this warning can be disabled.
5712 Warn if variable length array is used in the code.
5713 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5714 the variable length array.
5716 @item -Wvolatile-register-var
5717 @opindex Wvolatile-register-var
5718 @opindex Wno-volatile-register-var
5719 Warn if a register variable is declared volatile. The volatile
5720 modifier does not inhibit all optimizations that may eliminate reads
5721 and/or writes to register variables. This warning is enabled by
5724 @item -Wdisabled-optimization
5725 @opindex Wdisabled-optimization
5726 @opindex Wno-disabled-optimization
5727 Warn if a requested optimization pass is disabled. This warning does
5728 not generally indicate that there is anything wrong with your code; it
5729 merely indicates that GCC's optimizers are unable to handle the code
5730 effectively. Often, the problem is that your code is too big or too
5731 complex; GCC refuses to optimize programs when the optimization
5732 itself is likely to take inordinate amounts of time.
5734 @item -Wpointer-sign @r{(C and Objective-C only)}
5735 @opindex Wpointer-sign
5736 @opindex Wno-pointer-sign
5737 Warn for pointer argument passing or assignment with different signedness.
5738 This option is only supported for C and Objective-C@. It is implied by
5739 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5740 @option{-Wno-pointer-sign}.
5742 @item -Wstack-protector
5743 @opindex Wstack-protector
5744 @opindex Wno-stack-protector
5745 This option is only active when @option{-fstack-protector} is active. It
5746 warns about functions that are not protected against stack smashing.
5748 @item -Woverlength-strings
5749 @opindex Woverlength-strings
5750 @opindex Wno-overlength-strings
5751 Warn about string constants that are longer than the ``minimum
5752 maximum'' length specified in the C standard. Modern compilers
5753 generally allow string constants that are much longer than the
5754 standard's minimum limit, but very portable programs should avoid
5755 using longer strings.
5757 The limit applies @emph{after} string constant concatenation, and does
5758 not count the trailing NUL@. In C90, the limit was 509 characters; in
5759 C99, it was raised to 4095. C++98 does not specify a normative
5760 minimum maximum, so we do not diagnose overlength strings in C++@.
5762 This option is implied by @option{-Wpedantic}, and can be disabled with
5763 @option{-Wno-overlength-strings}.
5765 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5766 @opindex Wunsuffixed-float-constants
5768 Issue a warning for any floating constant that does not have
5769 a suffix. When used together with @option{-Wsystem-headers} it
5770 warns about such constants in system header files. This can be useful
5771 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5772 from the decimal floating-point extension to C99.
5774 @item -Wno-designated-init @r{(C and Objective-C only)}
5775 Suppress warnings when a positional initializer is used to initialize
5776 a structure that has been marked with the @code{designated_init}
5780 Issue a warning when HSAIL cannot be emitted for the compiled function or
5785 @node Debugging Options
5786 @section Options for Debugging Your Program
5787 @cindex options, debugging
5788 @cindex debugging information options
5790 To tell GCC to emit extra information for use by a debugger, in almost
5791 all cases you need only to add @option{-g} to your other options.
5793 GCC allows you to use @option{-g} with
5794 @option{-O}. The shortcuts taken by optimized code may occasionally
5795 be surprising: some variables you declared may not exist
5796 at all; flow of control may briefly move where you did not expect it;
5797 some statements may not be executed because they compute constant
5798 results or their values are already at hand; some statements may
5799 execute in different places because they have been moved out of loops.
5800 Nevertheless it is possible to debug optimized output. This makes
5801 it reasonable to use the optimizer for programs that might have bugs.
5803 If you are not using some other optimization option, consider
5804 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
5805 With no @option{-O} option at all, some compiler passes that collect
5806 information useful for debugging do not run at all, so that
5807 @option{-Og} may result in a better debugging experience.
5812 Produce debugging information in the operating system's native format
5813 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
5816 On most systems that use stabs format, @option{-g} enables use of extra
5817 debugging information that only GDB can use; this extra information
5818 makes debugging work better in GDB but probably makes other debuggers
5820 refuse to read the program. If you want to control for certain whether
5821 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5822 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5826 Produce debugging information for use by GDB@. This means to use the
5827 most expressive format available (DWARF, stabs, or the native format
5828 if neither of those are supported), including GDB extensions if at all
5832 @itemx -gdwarf-@var{version}
5834 Produce debugging information in DWARF format (if that is supported).
5835 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5836 for most targets is 4. DWARF Version 5 is only experimental.
5838 Note that with DWARF Version 2, some ports require and always
5839 use some non-conflicting DWARF 3 extensions in the unwind tables.
5841 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5842 for maximum benefit.
5844 GCC no longer supports DWARF Version 1, which is substantially
5845 different than Version 2 and later. For historical reasons, some
5846 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
5847 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
5848 in their names, but apply to all currently-supported versions of DWARF.
5852 Produce debugging information in stabs format (if that is supported),
5853 without GDB extensions. This is the format used by DBX on most BSD
5854 systems. On MIPS, Alpha and System V Release 4 systems this option
5855 produces stabs debugging output that is not understood by DBX or SDB@.
5856 On System V Release 4 systems this option requires the GNU assembler.
5860 Produce debugging information in stabs format (if that is supported),
5861 using GNU extensions understood only by the GNU debugger (GDB)@. The
5862 use of these extensions is likely to make other debuggers crash or
5863 refuse to read the program.
5867 Produce debugging information in COFF format (if that is supported).
5868 This is the format used by SDB on most System V systems prior to
5873 Produce debugging information in XCOFF format (if that is supported).
5874 This is the format used by the DBX debugger on IBM RS/6000 systems.
5878 Produce debugging information in XCOFF format (if that is supported),
5879 using GNU extensions understood only by the GNU debugger (GDB)@. The
5880 use of these extensions is likely to make other debuggers crash or
5881 refuse to read the program, and may cause assemblers other than the GNU
5882 assembler (GAS) to fail with an error.
5886 Produce debugging information in Alpha/VMS debug format (if that is
5887 supported). This is the format used by DEBUG on Alpha/VMS systems.
5890 @itemx -ggdb@var{level}
5891 @itemx -gstabs@var{level}
5892 @itemx -gcoff@var{level}
5893 @itemx -gxcoff@var{level}
5894 @itemx -gvms@var{level}
5895 Request debugging information and also use @var{level} to specify how
5896 much information. The default level is 2.
5898 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5901 Level 1 produces minimal information, enough for making backtraces in
5902 parts of the program that you don't plan to debug. This includes
5903 descriptions of functions and external variables, and line number
5904 tables, but no information about local variables.
5906 Level 3 includes extra information, such as all the macro definitions
5907 present in the program. Some debuggers support macro expansion when
5908 you use @option{-g3}.
5910 @option{-gdwarf} does not accept a concatenated debug level, to avoid
5911 confusion with @option{-gdwarf-@var{level}}.
5912 Instead use an additional @option{-g@var{level}} option to change the
5913 debug level for DWARF.
5915 @item -feliminate-unused-debug-symbols
5916 @opindex feliminate-unused-debug-symbols
5917 Produce debugging information in stabs format (if that is supported),
5918 for only symbols that are actually used.
5920 @item -femit-class-debug-always
5921 @opindex femit-class-debug-always
5922 Instead of emitting debugging information for a C++ class in only one
5923 object file, emit it in all object files using the class. This option
5924 should be used only with debuggers that are unable to handle the way GCC
5925 normally emits debugging information for classes because using this
5926 option increases the size of debugging information by as much as a
5929 @item -fno-merge-debug-strings
5930 @opindex fmerge-debug-strings
5931 @opindex fno-merge-debug-strings
5932 Direct the linker to not merge together strings in the debugging
5933 information that are identical in different object files. Merging is
5934 not supported by all assemblers or linkers. Merging decreases the size
5935 of the debug information in the output file at the cost of increasing
5936 link processing time. Merging is enabled by default.
5938 @item -fdebug-prefix-map=@var{old}=@var{new}
5939 @opindex fdebug-prefix-map
5940 When compiling files in directory @file{@var{old}}, record debugging
5941 information describing them as in @file{@var{new}} instead.
5943 @item -fvar-tracking
5944 @opindex fvar-tracking
5945 Run variable tracking pass. It computes where variables are stored at each
5946 position in code. Better debugging information is then generated
5947 (if the debugging information format supports this information).
5949 It is enabled by default when compiling with optimization (@option{-Os},
5950 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5951 the debug info format supports it.
5953 @item -fvar-tracking-assignments
5954 @opindex fvar-tracking-assignments
5955 @opindex fno-var-tracking-assignments
5956 Annotate assignments to user variables early in the compilation and
5957 attempt to carry the annotations over throughout the compilation all the
5958 way to the end, in an attempt to improve debug information while
5959 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5961 It can be enabled even if var-tracking is disabled, in which case
5962 annotations are created and maintained, but discarded at the end.
5963 By default, this flag is enabled together with @option{-fvar-tracking},
5964 except when selective scheduling is enabled.
5967 @opindex gsplit-dwarf
5968 Separate as much DWARF debugging information as possible into a
5969 separate output file with the extension @file{.dwo}. This option allows
5970 the build system to avoid linking files with debug information. To
5971 be useful, this option requires a debugger capable of reading @file{.dwo}
5976 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
5978 @item -ggnu-pubnames
5979 @opindex ggnu-pubnames
5980 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
5981 suitable for conversion into a GDB@ index. This option is only useful
5982 with a linker that can produce GDB@ index version 7.
5984 @item -fdebug-types-section
5985 @opindex fdebug-types-section
5986 @opindex fno-debug-types-section
5987 When using DWARF Version 4 or higher, type DIEs can be put into
5988 their own @code{.debug_types} section instead of making them part of the
5989 @code{.debug_info} section. It is more efficient to put them in a separate
5990 comdat sections since the linker can then remove duplicates.
5991 But not all DWARF consumers support @code{.debug_types} sections yet
5992 and on some objects @code{.debug_types} produces larger instead of smaller
5993 debugging information.
5995 @item -grecord-gcc-switches
5996 @item -gno-record-gcc-switches
5997 @opindex grecord-gcc-switches
5998 @opindex gno-record-gcc-switches
5999 This switch causes the command-line options used to invoke the
6000 compiler that may affect code generation to be appended to the
6001 DW_AT_producer attribute in DWARF debugging information. The options
6002 are concatenated with spaces separating them from each other and from
6003 the compiler version.
6004 It is enabled by default.
6005 See also @option{-frecord-gcc-switches} for another
6006 way of storing compiler options into the object file.
6008 @item -gstrict-dwarf
6009 @opindex gstrict-dwarf
6010 Disallow using extensions of later DWARF standard version than selected
6011 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6012 DWARF extensions from later standard versions is allowed.
6014 @item -gno-strict-dwarf
6015 @opindex gno-strict-dwarf
6016 Allow using extensions of later DWARF standard version than selected with
6017 @option{-gdwarf-@var{version}}.
6019 @item -gz@r{[}=@var{type}@r{]}
6021 Produce compressed debug sections in DWARF format, if that is supported.
6022 If @var{type} is not given, the default type depends on the capabilities
6023 of the assembler and linker used. @var{type} may be one of
6024 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6025 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6026 compression in traditional GNU format). If the linker doesn't support
6027 writing compressed debug sections, the option is rejected. Otherwise,
6028 if the assembler does not support them, @option{-gz} is silently ignored
6029 when producing object files.
6031 @item -feliminate-dwarf2-dups
6032 @opindex feliminate-dwarf2-dups
6033 Compress DWARF debugging information by eliminating duplicated
6034 information about each symbol. This option only makes sense when
6035 generating DWARF debugging information.
6037 @item -femit-struct-debug-baseonly
6038 @opindex femit-struct-debug-baseonly
6039 Emit debug information for struct-like types
6040 only when the base name of the compilation source file
6041 matches the base name of file in which the struct is defined.
6043 This option substantially reduces the size of debugging information,
6044 but at significant potential loss in type information to the debugger.
6045 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6046 See @option{-femit-struct-debug-detailed} for more detailed control.
6048 This option works only with DWARF debug output.
6050 @item -femit-struct-debug-reduced
6051 @opindex femit-struct-debug-reduced
6052 Emit debug information for struct-like types
6053 only when the base name of the compilation source file
6054 matches the base name of file in which the type is defined,
6055 unless the struct is a template or defined in a system header.
6057 This option significantly reduces the size of debugging information,
6058 with some potential loss in type information to the debugger.
6059 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6060 See @option{-femit-struct-debug-detailed} for more detailed control.
6062 This option works only with DWARF debug output.
6064 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6065 @opindex femit-struct-debug-detailed
6066 Specify the struct-like types
6067 for which the compiler generates debug information.
6068 The intent is to reduce duplicate struct debug information
6069 between different object files within the same program.
6071 This option is a detailed version of
6072 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6073 which serves for most needs.
6075 A specification has the syntax@*
6076 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6078 The optional first word limits the specification to
6079 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6080 A struct type is used directly when it is the type of a variable, member.
6081 Indirect uses arise through pointers to structs.
6082 That is, when use of an incomplete struct is valid, the use is indirect.
6084 @samp{struct one direct; struct two * indirect;}.
6086 The optional second word limits the specification to
6087 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6088 Generic structs are a bit complicated to explain.
6089 For C++, these are non-explicit specializations of template classes,
6090 or non-template classes within the above.
6091 Other programming languages have generics,
6092 but @option{-femit-struct-debug-detailed} does not yet implement them.
6094 The third word specifies the source files for those
6095 structs for which the compiler should emit debug information.
6096 The values @samp{none} and @samp{any} have the normal meaning.
6097 The value @samp{base} means that
6098 the base of name of the file in which the type declaration appears
6099 must match the base of the name of the main compilation file.
6100 In practice, this means that when compiling @file{foo.c}, debug information
6101 is generated for types declared in that file and @file{foo.h},
6102 but not other header files.
6103 The value @samp{sys} means those types satisfying @samp{base}
6104 or declared in system or compiler headers.
6106 You may need to experiment to determine the best settings for your application.
6108 The default is @option{-femit-struct-debug-detailed=all}.
6110 This option works only with DWARF debug output.
6112 @item -fno-dwarf2-cfi-asm
6113 @opindex fdwarf2-cfi-asm
6114 @opindex fno-dwarf2-cfi-asm
6115 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6116 instead of using GAS @code{.cfi_*} directives.
6118 @item -fno-eliminate-unused-debug-types
6119 @opindex feliminate-unused-debug-types
6120 @opindex fno-eliminate-unused-debug-types
6121 Normally, when producing DWARF output, GCC avoids producing debug symbol
6122 output for types that are nowhere used in the source file being compiled.
6123 Sometimes it is useful to have GCC emit debugging
6124 information for all types declared in a compilation
6125 unit, regardless of whether or not they are actually used
6126 in that compilation unit, for example
6127 if, in the debugger, you want to cast a value to a type that is
6128 not actually used in your program (but is declared). More often,
6129 however, this results in a significant amount of wasted space.
6132 @node Optimize Options
6133 @section Options That Control Optimization
6134 @cindex optimize options
6135 @cindex options, optimization
6137 These options control various sorts of optimizations.
6139 Without any optimization option, the compiler's goal is to reduce the
6140 cost of compilation and to make debugging produce the expected
6141 results. Statements are independent: if you stop the program with a
6142 breakpoint between statements, you can then assign a new value to any
6143 variable or change the program counter to any other statement in the
6144 function and get exactly the results you expect from the source
6147 Turning on optimization flags makes the compiler attempt to improve
6148 the performance and/or code size at the expense of compilation time
6149 and possibly the ability to debug the program.
6151 The compiler performs optimization based on the knowledge it has of the
6152 program. Compiling multiple files at once to a single output file mode allows
6153 the compiler to use information gained from all of the files when compiling
6156 Not all optimizations are controlled directly by a flag. Only
6157 optimizations that have a flag are listed in this section.
6159 Most optimizations are only enabled if an @option{-O} level is set on
6160 the command line. Otherwise they are disabled, even if individual
6161 optimization flags are specified.
6163 Depending on the target and how GCC was configured, a slightly different
6164 set of optimizations may be enabled at each @option{-O} level than
6165 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6166 to find out the exact set of optimizations that are enabled at each level.
6167 @xref{Overall Options}, for examples.
6174 Optimize. Optimizing compilation takes somewhat more time, and a lot
6175 more memory for a large function.
6177 With @option{-O}, the compiler tries to reduce code size and execution
6178 time, without performing any optimizations that take a great deal of
6181 @option{-O} turns on the following optimization flags:
6184 -fbranch-count-reg @gol
6185 -fcombine-stack-adjustments @gol
6187 -fcprop-registers @gol
6190 -fdelayed-branch @gol
6192 -fforward-propagate @gol
6193 -fguess-branch-probability @gol
6194 -fif-conversion2 @gol
6195 -fif-conversion @gol
6196 -finline-functions-called-once @gol
6197 -fipa-pure-const @gol
6199 -fipa-reference @gol
6200 -fmerge-constants @gol
6201 -fmove-loop-invariants @gol
6202 -freorder-blocks @gol
6204 -fsplit-wide-types @gol
6210 -ftree-coalesce-vars @gol
6211 -ftree-copy-prop @gol
6213 -ftree-dominator-opts @gol
6215 -ftree-forwprop @gol
6225 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6226 where doing so does not interfere with debugging.
6230 Optimize even more. GCC performs nearly all supported optimizations
6231 that do not involve a space-speed tradeoff.
6232 As compared to @option{-O}, this option increases both compilation time
6233 and the performance of the generated code.
6235 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6236 also turns on the following optimization flags:
6237 @gccoptlist{-fthread-jumps @gol
6238 -falign-functions -falign-jumps @gol
6239 -falign-loops -falign-labels @gol
6242 -fcse-follow-jumps -fcse-skip-blocks @gol
6243 -fdelete-null-pointer-checks @gol
6244 -fdevirtualize -fdevirtualize-speculatively @gol
6245 -fexpensive-optimizations @gol
6246 -fgcse -fgcse-lm @gol
6247 -fhoist-adjacent-loads @gol
6248 -finline-small-functions @gol
6249 -findirect-inlining @gol
6251 -fipa-cp-alignment @gol
6254 -fisolate-erroneous-paths-dereference @gol
6256 -foptimize-sibling-calls @gol
6257 -foptimize-strlen @gol
6258 -fpartial-inlining @gol
6260 -freorder-blocks-algorithm=stc @gol
6261 -freorder-blocks-and-partition -freorder-functions @gol
6262 -frerun-cse-after-loop @gol
6263 -fsched-interblock -fsched-spec @gol
6264 -fschedule-insns -fschedule-insns2 @gol
6265 -fstrict-aliasing -fstrict-overflow @gol
6266 -ftree-builtin-call-dce @gol
6267 -ftree-switch-conversion -ftree-tail-merge @gol
6272 Please note the warning under @option{-fgcse} about
6273 invoking @option{-O2} on programs that use computed gotos.
6277 Optimize yet more. @option{-O3} turns on all optimizations specified
6278 by @option{-O2} and also turns on the @option{-finline-functions},
6279 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6280 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
6281 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
6282 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
6283 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
6287 Reduce compilation time and make debugging produce the expected
6288 results. This is the default.
6292 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6293 do not typically increase code size. It also performs further
6294 optimizations designed to reduce code size.
6296 @option{-Os} disables the following optimization flags:
6297 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6298 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
6299 -freorder-blocks-and-partition -fprefetch-loop-arrays}
6303 Disregard strict standards compliance. @option{-Ofast} enables all
6304 @option{-O3} optimizations. It also enables optimizations that are not
6305 valid for all standard-compliant programs.
6306 It turns on @option{-ffast-math} and the Fortran-specific
6307 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6311 Optimize debugging experience. @option{-Og} enables optimizations
6312 that do not interfere with debugging. It should be the optimization
6313 level of choice for the standard edit-compile-debug cycle, offering
6314 a reasonable level of optimization while maintaining fast compilation
6315 and a good debugging experience.
6318 If you use multiple @option{-O} options, with or without level numbers,
6319 the last such option is the one that is effective.
6321 Options of the form @option{-f@var{flag}} specify machine-independent
6322 flags. Most flags have both positive and negative forms; the negative
6323 form of @option{-ffoo} is @option{-fno-foo}. In the table
6324 below, only one of the forms is listed---the one you typically
6325 use. You can figure out the other form by either removing @samp{no-}
6328 The following options control specific optimizations. They are either
6329 activated by @option{-O} options or are related to ones that are. You
6330 can use the following flags in the rare cases when ``fine-tuning'' of
6331 optimizations to be performed is desired.
6334 @item -fno-defer-pop
6335 @opindex fno-defer-pop
6336 Always pop the arguments to each function call as soon as that function
6337 returns. For machines that must pop arguments after a function call,
6338 the compiler normally lets arguments accumulate on the stack for several
6339 function calls and pops them all at once.
6341 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6343 @item -fforward-propagate
6344 @opindex fforward-propagate
6345 Perform a forward propagation pass on RTL@. The pass tries to combine two
6346 instructions and checks if the result can be simplified. If loop unrolling
6347 is active, two passes are performed and the second is scheduled after
6350 This option is enabled by default at optimization levels @option{-O},
6351 @option{-O2}, @option{-O3}, @option{-Os}.
6353 @item -ffp-contract=@var{style}
6354 @opindex ffp-contract
6355 @option{-ffp-contract=off} disables floating-point expression contraction.
6356 @option{-ffp-contract=fast} enables floating-point expression contraction
6357 such as forming of fused multiply-add operations if the target has
6358 native support for them.
6359 @option{-ffp-contract=on} enables floating-point expression contraction
6360 if allowed by the language standard. This is currently not implemented
6361 and treated equal to @option{-ffp-contract=off}.
6363 The default is @option{-ffp-contract=fast}.
6365 @item -fomit-frame-pointer
6366 @opindex fomit-frame-pointer
6367 Don't keep the frame pointer in a register for functions that
6368 don't need one. This avoids the instructions to save, set up and
6369 restore frame pointers; it also makes an extra register available
6370 in many functions. @strong{It also makes debugging impossible on
6373 On some machines, such as the VAX, this flag has no effect, because
6374 the standard calling sequence automatically handles the frame pointer
6375 and nothing is saved by pretending it doesn't exist. The
6376 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6377 whether a target machine supports this flag. @xref{Registers,,Register
6378 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6380 The default setting (when not optimizing for
6381 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
6382 @option{-fomit-frame-pointer}. You can configure GCC with the
6383 @option{--enable-frame-pointer} configure option to change the default.
6385 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6387 @item -foptimize-sibling-calls
6388 @opindex foptimize-sibling-calls
6389 Optimize sibling and tail recursive calls.
6391 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6393 @item -foptimize-strlen
6394 @opindex foptimize-strlen
6395 Optimize various standard C string functions (e.g. @code{strlen},
6396 @code{strchr} or @code{strcpy}) and
6397 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
6399 Enabled at levels @option{-O2}, @option{-O3}.
6403 Do not expand any functions inline apart from those marked with
6404 the @code{always_inline} attribute. This is the default when not
6407 Single functions can be exempted from inlining by marking them
6408 with the @code{noinline} attribute.
6410 @item -finline-small-functions
6411 @opindex finline-small-functions
6412 Integrate functions into their callers when their body is smaller than expected
6413 function call code (so overall size of program gets smaller). The compiler
6414 heuristically decides which functions are simple enough to be worth integrating
6415 in this way. This inlining applies to all functions, even those not declared
6418 Enabled at level @option{-O2}.
6420 @item -findirect-inlining
6421 @opindex findirect-inlining
6422 Inline also indirect calls that are discovered to be known at compile
6423 time thanks to previous inlining. This option has any effect only
6424 when inlining itself is turned on by the @option{-finline-functions}
6425 or @option{-finline-small-functions} options.
6427 Enabled at level @option{-O2}.
6429 @item -finline-functions
6430 @opindex finline-functions
6431 Consider all functions for inlining, even if they are not declared inline.
6432 The compiler heuristically decides which functions are worth integrating
6435 If all calls to a given function are integrated, and the function is
6436 declared @code{static}, then the function is normally not output as
6437 assembler code in its own right.
6439 Enabled at level @option{-O3}.
6441 @item -finline-functions-called-once
6442 @opindex finline-functions-called-once
6443 Consider all @code{static} functions called once for inlining into their
6444 caller even if they are not marked @code{inline}. If a call to a given
6445 function is integrated, then the function is not output as assembler code
6448 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6450 @item -fearly-inlining
6451 @opindex fearly-inlining
6452 Inline functions marked by @code{always_inline} and functions whose body seems
6453 smaller than the function call overhead early before doing
6454 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6455 makes profiling significantly cheaper and usually inlining faster on programs
6456 having large chains of nested wrapper functions.
6462 Perform interprocedural scalar replacement of aggregates, removal of
6463 unused parameters and replacement of parameters passed by reference
6464 by parameters passed by value.
6466 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6468 @item -finline-limit=@var{n}
6469 @opindex finline-limit
6470 By default, GCC limits the size of functions that can be inlined. This flag
6471 allows coarse control of this limit. @var{n} is the size of functions that
6472 can be inlined in number of pseudo instructions.
6474 Inlining is actually controlled by a number of parameters, which may be
6475 specified individually by using @option{--param @var{name}=@var{value}}.
6476 The @option{-finline-limit=@var{n}} option sets some of these parameters
6480 @item max-inline-insns-single
6481 is set to @var{n}/2.
6482 @item max-inline-insns-auto
6483 is set to @var{n}/2.
6486 See below for a documentation of the individual
6487 parameters controlling inlining and for the defaults of these parameters.
6489 @emph{Note:} there may be no value to @option{-finline-limit} that results
6490 in default behavior.
6492 @emph{Note:} pseudo instruction represents, in this particular context, an
6493 abstract measurement of function's size. In no way does it represent a count
6494 of assembly instructions and as such its exact meaning might change from one
6495 release to an another.
6497 @item -fno-keep-inline-dllexport
6498 @opindex fno-keep-inline-dllexport
6499 This is a more fine-grained version of @option{-fkeep-inline-functions},
6500 which applies only to functions that are declared using the @code{dllexport}
6501 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6504 @item -fkeep-inline-functions
6505 @opindex fkeep-inline-functions
6506 In C, emit @code{static} functions that are declared @code{inline}
6507 into the object file, even if the function has been inlined into all
6508 of its callers. This switch does not affect functions using the
6509 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6510 inline functions into the object file.
6512 @item -fkeep-static-functions
6513 @opindex fkeep-static-functions
6514 Emit @code{static} functions into the object file, even if the function
6517 @item -fkeep-static-consts
6518 @opindex fkeep-static-consts
6519 Emit variables declared @code{static const} when optimization isn't turned
6520 on, even if the variables aren't referenced.
6522 GCC enables this option by default. If you want to force the compiler to
6523 check if a variable is referenced, regardless of whether or not
6524 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6526 @item -fmerge-constants
6527 @opindex fmerge-constants
6528 Attempt to merge identical constants (string constants and floating-point
6529 constants) across compilation units.
6531 This option is the default for optimized compilation if the assembler and
6532 linker support it. Use @option{-fno-merge-constants} to inhibit this
6535 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6537 @item -fmerge-all-constants
6538 @opindex fmerge-all-constants
6539 Attempt to merge identical constants and identical variables.
6541 This option implies @option{-fmerge-constants}. In addition to
6542 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6543 arrays or initialized constant variables with integral or floating-point
6544 types. Languages like C or C++ require each variable, including multiple
6545 instances of the same variable in recursive calls, to have distinct locations,
6546 so using this option results in non-conforming
6549 @item -fmodulo-sched
6550 @opindex fmodulo-sched
6551 Perform swing modulo scheduling immediately before the first scheduling
6552 pass. This pass looks at innermost loops and reorders their
6553 instructions by overlapping different iterations.
6555 @item -fmodulo-sched-allow-regmoves
6556 @opindex fmodulo-sched-allow-regmoves
6557 Perform more aggressive SMS-based modulo scheduling with register moves
6558 allowed. By setting this flag certain anti-dependences edges are
6559 deleted, which triggers the generation of reg-moves based on the
6560 life-range analysis. This option is effective only with
6561 @option{-fmodulo-sched} enabled.
6563 @item -fno-branch-count-reg
6564 @opindex fno-branch-count-reg
6565 Avoid running a pass scanning for opportunities to use ``decrement and
6566 branch'' instructions on a count register instead of generating sequences
6567 of instructions that decrement a register, compare it against zero, and
6568 then branch based upon the result. This option is only meaningful on
6569 architectures that support such instructions, which include x86, PowerPC,
6570 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
6571 doesn't remove the decrement and branch instructions from the generated
6572 instruction stream introduced by other optimization passes.
6574 Enabled by default at @option{-O1} and higher.
6576 The default is @option{-fbranch-count-reg}.
6578 @item -fno-function-cse
6579 @opindex fno-function-cse
6580 Do not put function addresses in registers; make each instruction that
6581 calls a constant function contain the function's address explicitly.
6583 This option results in less efficient code, but some strange hacks
6584 that alter the assembler output may be confused by the optimizations
6585 performed when this option is not used.
6587 The default is @option{-ffunction-cse}
6589 @item -fno-zero-initialized-in-bss
6590 @opindex fno-zero-initialized-in-bss
6591 If the target supports a BSS section, GCC by default puts variables that
6592 are initialized to zero into BSS@. This can save space in the resulting
6595 This option turns off this behavior because some programs explicitly
6596 rely on variables going to the data section---e.g., so that the
6597 resulting executable can find the beginning of that section and/or make
6598 assumptions based on that.
6600 The default is @option{-fzero-initialized-in-bss}.
6602 @item -fthread-jumps
6603 @opindex fthread-jumps
6604 Perform optimizations that check to see if a jump branches to a
6605 location where another comparison subsumed by the first is found. If
6606 so, the first branch is redirected to either the destination of the
6607 second branch or a point immediately following it, depending on whether
6608 the condition is known to be true or false.
6610 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6612 @item -fsplit-wide-types
6613 @opindex fsplit-wide-types
6614 When using a type that occupies multiple registers, such as @code{long
6615 long} on a 32-bit system, split the registers apart and allocate them
6616 independently. This normally generates better code for those types,
6617 but may make debugging more difficult.
6619 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6622 @item -fcse-follow-jumps
6623 @opindex fcse-follow-jumps
6624 In common subexpression elimination (CSE), scan through jump instructions
6625 when the target of the jump is not reached by any other path. For
6626 example, when CSE encounters an @code{if} statement with an
6627 @code{else} clause, CSE follows the jump when the condition
6630 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6632 @item -fcse-skip-blocks
6633 @opindex fcse-skip-blocks
6634 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6635 follow jumps that conditionally skip over blocks. When CSE
6636 encounters a simple @code{if} statement with no else clause,
6637 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6638 body of the @code{if}.
6640 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6642 @item -frerun-cse-after-loop
6643 @opindex frerun-cse-after-loop
6644 Re-run common subexpression elimination after loop optimizations are
6647 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6651 Perform a global common subexpression elimination pass.
6652 This pass also performs global constant and copy propagation.
6654 @emph{Note:} When compiling a program using computed gotos, a GCC
6655 extension, you may get better run-time performance if you disable
6656 the global common subexpression elimination pass by adding
6657 @option{-fno-gcse} to the command line.
6659 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6663 When @option{-fgcse-lm} is enabled, global common subexpression elimination
6664 attempts to move loads that are only killed by stores into themselves. This
6665 allows a loop containing a load/store sequence to be changed to a load outside
6666 the loop, and a copy/store within the loop.
6668 Enabled by default when @option{-fgcse} is enabled.
6672 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6673 global common subexpression elimination. This pass attempts to move
6674 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6675 loops containing a load/store sequence can be changed to a load before
6676 the loop and a store after the loop.
6678 Not enabled at any optimization level.
6682 When @option{-fgcse-las} is enabled, the global common subexpression
6683 elimination pass eliminates redundant loads that come after stores to the
6684 same memory location (both partial and full redundancies).
6686 Not enabled at any optimization level.
6688 @item -fgcse-after-reload
6689 @opindex fgcse-after-reload
6690 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6691 pass is performed after reload. The purpose of this pass is to clean up
6694 @item -faggressive-loop-optimizations
6695 @opindex faggressive-loop-optimizations
6696 This option tells the loop optimizer to use language constraints to
6697 derive bounds for the number of iterations of a loop. This assumes that
6698 loop code does not invoke undefined behavior by for example causing signed
6699 integer overflows or out-of-bound array accesses. The bounds for the
6700 number of iterations of a loop are used to guide loop unrolling and peeling
6701 and loop exit test optimizations.
6702 This option is enabled by default.
6704 @item -funsafe-loop-optimizations
6705 @opindex funsafe-loop-optimizations
6706 This option tells the loop optimizer to assume that loop indices do not
6707 overflow, and that loops with nontrivial exit condition are not
6708 infinite. This enables a wider range of loop optimizations even if
6709 the loop optimizer itself cannot prove that these assumptions are valid.
6710 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
6711 if it finds this kind of loop.
6713 @item -funconstrained-commons
6714 @opindex funconstrained-commons
6715 This option tells the compiler that variables declared in common blocks
6716 (e.g. Fortran) may later be overridden with longer trailing arrays. This
6717 prevents certain optimizations that depend on knowing the array bounds.
6719 @item -fcrossjumping
6720 @opindex fcrossjumping
6721 Perform cross-jumping transformation.
6722 This transformation unifies equivalent code and saves code size. The
6723 resulting code may or may not perform better than without cross-jumping.
6725 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6727 @item -fauto-inc-dec
6728 @opindex fauto-inc-dec
6729 Combine increments or decrements of addresses with memory accesses.
6730 This pass is always skipped on architectures that do not have
6731 instructions to support this. Enabled by default at @option{-O} and
6732 higher on architectures that support this.
6736 Perform dead code elimination (DCE) on RTL@.
6737 Enabled by default at @option{-O} and higher.
6741 Perform dead store elimination (DSE) on RTL@.
6742 Enabled by default at @option{-O} and higher.
6744 @item -fif-conversion
6745 @opindex fif-conversion
6746 Attempt to transform conditional jumps into branch-less equivalents. This
6747 includes use of conditional moves, min, max, set flags and abs instructions, and
6748 some tricks doable by standard arithmetics. The use of conditional execution
6749 on chips where it is available is controlled by @option{-fif-conversion2}.
6751 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6753 @item -fif-conversion2
6754 @opindex fif-conversion2
6755 Use conditional execution (where available) to transform conditional jumps into
6756 branch-less equivalents.
6758 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6760 @item -fdeclone-ctor-dtor
6761 @opindex fdeclone-ctor-dtor
6762 The C++ ABI requires multiple entry points for constructors and
6763 destructors: one for a base subobject, one for a complete object, and
6764 one for a virtual destructor that calls operator delete afterwards.
6765 For a hierarchy with virtual bases, the base and complete variants are
6766 clones, which means two copies of the function. With this option, the
6767 base and complete variants are changed to be thunks that call a common
6770 Enabled by @option{-Os}.
6772 @item -fdelete-null-pointer-checks
6773 @opindex fdelete-null-pointer-checks
6774 Assume that programs cannot safely dereference null pointers, and that
6775 no code or data element resides at address zero.
6776 This option enables simple constant
6777 folding optimizations at all optimization levels. In addition, other
6778 optimization passes in GCC use this flag to control global dataflow
6779 analyses that eliminate useless checks for null pointers; these assume
6780 that a memory access to address zero always results in a trap, so
6781 that if a pointer is checked after it has already been dereferenced,
6784 Note however that in some environments this assumption is not true.
6785 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6786 for programs that depend on that behavior.
6788 This option is enabled by default on most targets. On Nios II ELF, it
6789 defaults to off. On AVR and CR16, this option is completely disabled.
6791 Passes that use the dataflow information
6792 are enabled independently at different optimization levels.
6794 @item -fdevirtualize
6795 @opindex fdevirtualize
6796 Attempt to convert calls to virtual functions to direct calls. This
6797 is done both within a procedure and interprocedurally as part of
6798 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
6799 propagation (@option{-fipa-cp}).
6800 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6802 @item -fdevirtualize-speculatively
6803 @opindex fdevirtualize-speculatively
6804 Attempt to convert calls to virtual functions to speculative direct calls.
6805 Based on the analysis of the type inheritance graph, determine for a given call
6806 the set of likely targets. If the set is small, preferably of size 1, change
6807 the call into a conditional deciding between direct and indirect calls. The
6808 speculative calls enable more optimizations, such as inlining. When they seem
6809 useless after further optimization, they are converted back into original form.
6811 @item -fdevirtualize-at-ltrans
6812 @opindex fdevirtualize-at-ltrans
6813 Stream extra information needed for aggressive devirtualization when running
6814 the link-time optimizer in local transformation mode.
6815 This option enables more devirtualization but
6816 significantly increases the size of streamed data. For this reason it is
6817 disabled by default.
6819 @item -fexpensive-optimizations
6820 @opindex fexpensive-optimizations
6821 Perform a number of minor optimizations that are relatively expensive.
6823 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6827 Attempt to remove redundant extension instructions. This is especially
6828 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
6829 registers after writing to their lower 32-bit half.
6831 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
6832 @option{-O3}, @option{-Os}.
6834 @item -fno-lifetime-dse
6835 @opindex fno-lifetime-dse
6836 In C++ the value of an object is only affected by changes within its
6837 lifetime: when the constructor begins, the object has an indeterminate
6838 value, and any changes during the lifetime of the object are dead when
6839 the object is destroyed. Normally dead store elimination will take
6840 advantage of this; if your code relies on the value of the object
6841 storage persisting beyond the lifetime of the object, you can use this
6842 flag to disable this optimization. To preserve stores before the
6843 constructor starts (e.g. because your operator new clears the object
6844 storage) but still treat the object as dead after the destructor you,
6845 can use @option{-flifetime-dse=1}. The default behavior can be
6846 explicitly selected with @option{-flifetime-dse=2}.
6847 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
6849 @item -flive-range-shrinkage
6850 @opindex flive-range-shrinkage
6851 Attempt to decrease register pressure through register live range
6852 shrinkage. This is helpful for fast processors with small or moderate
6855 @item -fira-algorithm=@var{algorithm}
6856 @opindex fira-algorithm
6857 Use the specified coloring algorithm for the integrated register
6858 allocator. The @var{algorithm} argument can be @samp{priority}, which
6859 specifies Chow's priority coloring, or @samp{CB}, which specifies
6860 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
6861 for all architectures, but for those targets that do support it, it is
6862 the default because it generates better code.
6864 @item -fira-region=@var{region}
6865 @opindex fira-region
6866 Use specified regions for the integrated register allocator. The
6867 @var{region} argument should be one of the following:
6872 Use all loops as register allocation regions.
6873 This can give the best results for machines with a small and/or
6874 irregular register set.
6877 Use all loops except for loops with small register pressure
6878 as the regions. This value usually gives
6879 the best results in most cases and for most architectures,
6880 and is enabled by default when compiling with optimization for speed
6881 (@option{-O}, @option{-O2}, @dots{}).
6884 Use all functions as a single region.
6885 This typically results in the smallest code size, and is enabled by default for
6886 @option{-Os} or @option{-O0}.
6890 @item -fira-hoist-pressure
6891 @opindex fira-hoist-pressure
6892 Use IRA to evaluate register pressure in the code hoisting pass for
6893 decisions to hoist expressions. This option usually results in smaller
6894 code, but it can slow the compiler down.
6896 This option is enabled at level @option{-Os} for all targets.
6898 @item -fira-loop-pressure
6899 @opindex fira-loop-pressure
6900 Use IRA to evaluate register pressure in loops for decisions to move
6901 loop invariants. This option usually results in generation
6902 of faster and smaller code on machines with large register files (>= 32
6903 registers), but it can slow the compiler down.
6905 This option is enabled at level @option{-O3} for some targets.
6907 @item -fno-ira-share-save-slots
6908 @opindex fno-ira-share-save-slots
6909 Disable sharing of stack slots used for saving call-used hard
6910 registers living through a call. Each hard register gets a
6911 separate stack slot, and as a result function stack frames are
6914 @item -fno-ira-share-spill-slots
6915 @opindex fno-ira-share-spill-slots
6916 Disable sharing of stack slots allocated for pseudo-registers. Each
6917 pseudo-register that does not get a hard register gets a separate
6918 stack slot, and as a result function stack frames are larger.
6922 Enable CFG-sensitive rematerialization in LRA. Instead of loading
6923 values of spilled pseudos, LRA tries to rematerialize (recalculate)
6924 values if it is profitable.
6926 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6928 @item -fdelayed-branch
6929 @opindex fdelayed-branch
6930 If supported for the target machine, attempt to reorder instructions
6931 to exploit instruction slots available after delayed branch
6934 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6936 @item -fschedule-insns
6937 @opindex fschedule-insns
6938 If supported for the target machine, attempt to reorder instructions to
6939 eliminate execution stalls due to required data being unavailable. This
6940 helps machines that have slow floating point or memory load instructions
6941 by allowing other instructions to be issued until the result of the load
6942 or floating-point instruction is required.
6944 Enabled at levels @option{-O2}, @option{-O3}.
6946 @item -fschedule-insns2
6947 @opindex fschedule-insns2
6948 Similar to @option{-fschedule-insns}, but requests an additional pass of
6949 instruction scheduling after register allocation has been done. This is
6950 especially useful on machines with a relatively small number of
6951 registers and where memory load instructions take more than one cycle.
6953 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6955 @item -fno-sched-interblock
6956 @opindex fno-sched-interblock
6957 Don't schedule instructions across basic blocks. This is normally
6958 enabled by default when scheduling before register allocation, i.e.@:
6959 with @option{-fschedule-insns} or at @option{-O2} or higher.
6961 @item -fno-sched-spec
6962 @opindex fno-sched-spec
6963 Don't allow speculative motion of non-load instructions. This is normally
6964 enabled by default when scheduling before register allocation, i.e.@:
6965 with @option{-fschedule-insns} or at @option{-O2} or higher.
6967 @item -fsched-pressure
6968 @opindex fsched-pressure
6969 Enable register pressure sensitive insn scheduling before register
6970 allocation. This only makes sense when scheduling before register
6971 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6972 @option{-O2} or higher. Usage of this option can improve the
6973 generated code and decrease its size by preventing register pressure
6974 increase above the number of available hard registers and subsequent
6975 spills in register allocation.
6977 @item -fsched-spec-load
6978 @opindex fsched-spec-load
6979 Allow speculative motion of some load instructions. This only makes
6980 sense when scheduling before register allocation, i.e.@: with
6981 @option{-fschedule-insns} or at @option{-O2} or higher.
6983 @item -fsched-spec-load-dangerous
6984 @opindex fsched-spec-load-dangerous
6985 Allow speculative motion of more load instructions. This only makes
6986 sense when scheduling before register allocation, i.e.@: with
6987 @option{-fschedule-insns} or at @option{-O2} or higher.
6989 @item -fsched-stalled-insns
6990 @itemx -fsched-stalled-insns=@var{n}
6991 @opindex fsched-stalled-insns
6992 Define how many insns (if any) can be moved prematurely from the queue
6993 of stalled insns into the ready list during the second scheduling pass.
6994 @option{-fno-sched-stalled-insns} means that no insns are moved
6995 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6996 on how many queued insns can be moved prematurely.
6997 @option{-fsched-stalled-insns} without a value is equivalent to
6998 @option{-fsched-stalled-insns=1}.
7000 @item -fsched-stalled-insns-dep
7001 @itemx -fsched-stalled-insns-dep=@var{n}
7002 @opindex fsched-stalled-insns-dep
7003 Define how many insn groups (cycles) are examined for a dependency
7004 on a stalled insn that is a candidate for premature removal from the queue
7005 of stalled insns. This has an effect only during the second scheduling pass,
7006 and only if @option{-fsched-stalled-insns} is used.
7007 @option{-fno-sched-stalled-insns-dep} is equivalent to
7008 @option{-fsched-stalled-insns-dep=0}.
7009 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7010 @option{-fsched-stalled-insns-dep=1}.
7012 @item -fsched2-use-superblocks
7013 @opindex fsched2-use-superblocks
7014 When scheduling after register allocation, use superblock scheduling.
7015 This allows motion across basic block boundaries,
7016 resulting in faster schedules. This option is experimental, as not all machine
7017 descriptions used by GCC model the CPU closely enough to avoid unreliable
7018 results from the algorithm.
7020 This only makes sense when scheduling after register allocation, i.e.@: with
7021 @option{-fschedule-insns2} or at @option{-O2} or higher.
7023 @item -fsched-group-heuristic
7024 @opindex fsched-group-heuristic
7025 Enable the group heuristic in the scheduler. This heuristic favors
7026 the instruction that belongs to a schedule group. This is enabled
7027 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7028 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7030 @item -fsched-critical-path-heuristic
7031 @opindex fsched-critical-path-heuristic
7032 Enable the critical-path heuristic in the scheduler. This heuristic favors
7033 instructions on the critical path. This is enabled by default when
7034 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7035 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7037 @item -fsched-spec-insn-heuristic
7038 @opindex fsched-spec-insn-heuristic
7039 Enable the speculative instruction heuristic in the scheduler. This
7040 heuristic favors speculative instructions with greater dependency weakness.
7041 This is enabled by default when scheduling is enabled, i.e.@:
7042 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7043 or at @option{-O2} or higher.
7045 @item -fsched-rank-heuristic
7046 @opindex fsched-rank-heuristic
7047 Enable the rank heuristic in the scheduler. This heuristic favors
7048 the instruction belonging to a basic block with greater size or frequency.
7049 This is enabled by default when scheduling is enabled, i.e.@:
7050 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7051 at @option{-O2} or higher.
7053 @item -fsched-last-insn-heuristic
7054 @opindex fsched-last-insn-heuristic
7055 Enable the last-instruction heuristic in the scheduler. This heuristic
7056 favors the instruction that is less dependent on the last instruction
7057 scheduled. This is enabled by default when scheduling is enabled,
7058 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7059 at @option{-O2} or higher.
7061 @item -fsched-dep-count-heuristic
7062 @opindex fsched-dep-count-heuristic
7063 Enable the dependent-count heuristic in the scheduler. This heuristic
7064 favors the instruction that has more instructions depending on it.
7065 This is enabled by default when scheduling is enabled, i.e.@:
7066 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7067 at @option{-O2} or higher.
7069 @item -freschedule-modulo-scheduled-loops
7070 @opindex freschedule-modulo-scheduled-loops
7071 Modulo scheduling is performed before traditional scheduling. If a loop
7072 is modulo scheduled, later scheduling passes may change its schedule.
7073 Use this option to control that behavior.
7075 @item -fselective-scheduling
7076 @opindex fselective-scheduling
7077 Schedule instructions using selective scheduling algorithm. Selective
7078 scheduling runs instead of the first scheduler pass.
7080 @item -fselective-scheduling2
7081 @opindex fselective-scheduling2
7082 Schedule instructions using selective scheduling algorithm. Selective
7083 scheduling runs instead of the second scheduler pass.
7085 @item -fsel-sched-pipelining
7086 @opindex fsel-sched-pipelining
7087 Enable software pipelining of innermost loops during selective scheduling.
7088 This option has no effect unless one of @option{-fselective-scheduling} or
7089 @option{-fselective-scheduling2} is turned on.
7091 @item -fsel-sched-pipelining-outer-loops
7092 @opindex fsel-sched-pipelining-outer-loops
7093 When pipelining loops during selective scheduling, also pipeline outer loops.
7094 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7096 @item -fsemantic-interposition
7097 @opindex fsemantic-interposition
7098 Some object formats, like ELF, allow interposing of symbols by the
7100 This means that for symbols exported from the DSO, the compiler cannot perform
7101 interprocedural propagation, inlining and other optimizations in anticipation
7102 that the function or variable in question may change. While this feature is
7103 useful, for example, to rewrite memory allocation functions by a debugging
7104 implementation, it is expensive in the terms of code quality.
7105 With @option{-fno-semantic-interposition} the compiler assumes that
7106 if interposition happens for functions the overwriting function will have
7107 precisely the same semantics (and side effects).
7108 Similarly if interposition happens
7109 for variables, the constructor of the variable will be the same. The flag
7110 has no effect for functions explicitly declared inline
7111 (where it is never allowed for interposition to change semantics)
7112 and for symbols explicitly declared weak.
7115 @opindex fshrink-wrap
7116 Emit function prologues only before parts of the function that need it,
7117 rather than at the top of the function. This flag is enabled by default at
7118 @option{-O} and higher.
7120 @item -fcaller-saves
7121 @opindex fcaller-saves
7122 Enable allocation of values to registers that are clobbered by
7123 function calls, by emitting extra instructions to save and restore the
7124 registers around such calls. Such allocation is done only when it
7125 seems to result in better code.
7127 This option is always enabled by default on certain machines, usually
7128 those which have no call-preserved registers to use instead.
7130 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7132 @item -fcombine-stack-adjustments
7133 @opindex fcombine-stack-adjustments
7134 Tracks stack adjustments (pushes and pops) and stack memory references
7135 and then tries to find ways to combine them.
7137 Enabled by default at @option{-O1} and higher.
7141 Use caller save registers for allocation if those registers are not used by
7142 any called function. In that case it is not necessary to save and restore
7143 them around calls. This is only possible if called functions are part of
7144 same compilation unit as current function and they are compiled before it.
7146 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7148 @item -fconserve-stack
7149 @opindex fconserve-stack
7150 Attempt to minimize stack usage. The compiler attempts to use less
7151 stack space, even if that makes the program slower. This option
7152 implies setting the @option{large-stack-frame} parameter to 100
7153 and the @option{large-stack-frame-growth} parameter to 400.
7155 @item -ftree-reassoc
7156 @opindex ftree-reassoc
7157 Perform reassociation on trees. This flag is enabled by default
7158 at @option{-O} and higher.
7162 Perform partial redundancy elimination (PRE) on trees. This flag is
7163 enabled by default at @option{-O2} and @option{-O3}.
7165 @item -ftree-partial-pre
7166 @opindex ftree-partial-pre
7167 Make partial redundancy elimination (PRE) more aggressive. This flag is
7168 enabled by default at @option{-O3}.
7170 @item -ftree-forwprop
7171 @opindex ftree-forwprop
7172 Perform forward propagation on trees. This flag is enabled by default
7173 at @option{-O} and higher.
7177 Perform full redundancy elimination (FRE) on trees. The difference
7178 between FRE and PRE is that FRE only considers expressions
7179 that are computed on all paths leading to the redundant computation.
7180 This analysis is faster than PRE, though it exposes fewer redundancies.
7181 This flag is enabled by default at @option{-O} and higher.
7183 @item -ftree-phiprop
7184 @opindex ftree-phiprop
7185 Perform hoisting of loads from conditional pointers on trees. This
7186 pass is enabled by default at @option{-O} and higher.
7188 @item -fhoist-adjacent-loads
7189 @opindex fhoist-adjacent-loads
7190 Speculatively hoist loads from both branches of an if-then-else if the
7191 loads are from adjacent locations in the same structure and the target
7192 architecture has a conditional move instruction. This flag is enabled
7193 by default at @option{-O2} and higher.
7195 @item -ftree-copy-prop
7196 @opindex ftree-copy-prop
7197 Perform copy propagation on trees. This pass eliminates unnecessary
7198 copy operations. This flag is enabled by default at @option{-O} and
7201 @item -fipa-pure-const
7202 @opindex fipa-pure-const
7203 Discover which functions are pure or constant.
7204 Enabled by default at @option{-O} and higher.
7206 @item -fipa-reference
7207 @opindex fipa-reference
7208 Discover which static variables do not escape the
7210 Enabled by default at @option{-O} and higher.
7214 Perform interprocedural pointer analysis and interprocedural modification
7215 and reference analysis. This option can cause excessive memory and
7216 compile-time usage on large compilation units. It is not enabled by
7217 default at any optimization level.
7220 @opindex fipa-profile
7221 Perform interprocedural profile propagation. The functions called only from
7222 cold functions are marked as cold. Also functions executed once (such as
7223 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7224 functions and loop less parts of functions executed once are then optimized for
7226 Enabled by default at @option{-O} and higher.
7230 Perform interprocedural constant propagation.
7231 This optimization analyzes the program to determine when values passed
7232 to functions are constants and then optimizes accordingly.
7233 This optimization can substantially increase performance
7234 if the application has constants passed to functions.
7235 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7237 @item -fipa-cp-clone
7238 @opindex fipa-cp-clone
7239 Perform function cloning to make interprocedural constant propagation stronger.
7240 When enabled, interprocedural constant propagation performs function cloning
7241 when externally visible function can be called with constant arguments.
7242 Because this optimization can create multiple copies of functions,
7243 it may significantly increase code size
7244 (see @option{--param ipcp-unit-growth=@var{value}}).
7245 This flag is enabled by default at @option{-O3}.
7247 @item -fipa-cp-alignment
7248 @opindex -fipa-cp-alignment
7249 When enabled, this optimization propagates alignment of function
7250 parameters to support better vectorization and string operations.
7252 This flag is enabled by default at @option{-O2} and @option{-Os}. It
7253 requires that @option{-fipa-cp} is enabled.
7257 Perform Identical Code Folding for functions and read-only variables.
7258 The optimization reduces code size and may disturb unwind stacks by replacing
7259 a function by equivalent one with a different name. The optimization works
7260 more effectively with link time optimization enabled.
7262 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
7263 works on different levels and thus the optimizations are not same - there are
7264 equivalences that are found only by GCC and equivalences found only by Gold.
7266 This flag is enabled by default at @option{-O2} and @option{-Os}.
7268 @item -fisolate-erroneous-paths-dereference
7269 @opindex fisolate-erroneous-paths-dereference
7270 Detect paths that trigger erroneous or undefined behavior due to
7271 dereferencing a null pointer. Isolate those paths from the main control
7272 flow and turn the statement with erroneous or undefined behavior into a trap.
7273 This flag is enabled by default at @option{-O2} and higher and depends on
7274 @option{-fdelete-null-pointer-checks} also being enabled.
7276 @item -fisolate-erroneous-paths-attribute
7277 @opindex fisolate-erroneous-paths-attribute
7278 Detect paths that trigger erroneous or undefined behavior due a null value
7279 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
7280 attribute. Isolate those paths from the main control flow and turn the
7281 statement with erroneous or undefined behavior into a trap. This is not
7282 currently enabled, but may be enabled by @option{-O2} in the future.
7286 Perform forward store motion on trees. This flag is
7287 enabled by default at @option{-O} and higher.
7289 @item -ftree-bit-ccp
7290 @opindex ftree-bit-ccp
7291 Perform sparse conditional bit constant propagation on trees and propagate
7292 pointer alignment information.
7293 This pass only operates on local scalar variables and is enabled by default
7294 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7298 Perform sparse conditional constant propagation (CCP) on trees. This
7299 pass only operates on local scalar variables and is enabled by default
7300 at @option{-O} and higher.
7302 @item -fssa-backprop
7303 @opindex fssa-backprop
7304 Propagate information about uses of a value up the definition chain
7305 in order to simplify the definitions. For example, this pass strips
7306 sign operations if the sign of a value never matters. The flag is
7307 enabled by default at @option{-O} and higher.
7310 @opindex fssa-phiopt
7311 Perform pattern matching on SSA PHI nodes to optimize conditional
7312 code. This pass is enabled by default at @option{-O} and higher.
7314 @item -ftree-switch-conversion
7315 @opindex ftree-switch-conversion
7316 Perform conversion of simple initializations in a switch to
7317 initializations from a scalar array. This flag is enabled by default
7318 at @option{-O2} and higher.
7320 @item -ftree-tail-merge
7321 @opindex ftree-tail-merge
7322 Look for identical code sequences. When found, replace one with a jump to the
7323 other. This optimization is known as tail merging or cross jumping. This flag
7324 is enabled by default at @option{-O2} and higher. The compilation time
7326 be limited using @option{max-tail-merge-comparisons} parameter and
7327 @option{max-tail-merge-iterations} parameter.
7331 Perform dead code elimination (DCE) on trees. This flag is enabled by
7332 default at @option{-O} and higher.
7334 @item -ftree-builtin-call-dce
7335 @opindex ftree-builtin-call-dce
7336 Perform conditional dead code elimination (DCE) for calls to built-in functions
7337 that may set @code{errno} but are otherwise side-effect free. This flag is
7338 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7341 @item -ftree-dominator-opts
7342 @opindex ftree-dominator-opts
7343 Perform a variety of simple scalar cleanups (constant/copy
7344 propagation, redundancy elimination, range propagation and expression
7345 simplification) based on a dominator tree traversal. This also
7346 performs jump threading (to reduce jumps to jumps). This flag is
7347 enabled by default at @option{-O} and higher.
7351 Perform dead store elimination (DSE) on trees. A dead store is a store into
7352 a memory location that is later overwritten by another store without
7353 any intervening loads. In this case the earlier store can be deleted. This
7354 flag is enabled by default at @option{-O} and higher.
7358 Perform loop header copying on trees. This is beneficial since it increases
7359 effectiveness of code motion optimizations. It also saves one jump. This flag
7360 is enabled by default at @option{-O} and higher. It is not enabled
7361 for @option{-Os}, since it usually increases code size.
7363 @item -ftree-loop-optimize
7364 @opindex ftree-loop-optimize
7365 Perform loop optimizations on trees. This flag is enabled by default
7366 at @option{-O} and higher.
7368 @item -ftree-loop-linear
7369 @itemx -floop-interchange
7370 @itemx -floop-strip-mine
7372 @itemx -floop-unroll-and-jam
7373 @opindex ftree-loop-linear
7374 @opindex floop-interchange
7375 @opindex floop-strip-mine
7376 @opindex floop-block
7377 @opindex floop-unroll-and-jam
7378 Perform loop nest optimizations. Same as
7379 @option{-floop-nest-optimize}. To use this code transformation, GCC has
7380 to be configured with @option{--with-isl} to enable the Graphite loop
7381 transformation infrastructure.
7383 @item -fgraphite-identity
7384 @opindex fgraphite-identity
7385 Enable the identity transformation for graphite. For every SCoP we generate
7386 the polyhedral representation and transform it back to gimple. Using
7387 @option{-fgraphite-identity} we can check the costs or benefits of the
7388 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7389 are also performed by the code generator isl, like index splitting and
7390 dead code elimination in loops.
7392 @item -floop-nest-optimize
7393 @opindex floop-nest-optimize
7394 Enable the isl based loop nest optimizer. This is a generic loop nest
7395 optimizer based on the Pluto optimization algorithms. It calculates a loop
7396 structure optimized for data-locality and parallelism. This option
7399 @item -floop-parallelize-all
7400 @opindex floop-parallelize-all
7401 Use the Graphite data dependence analysis to identify loops that can
7402 be parallelized. Parallelize all the loops that can be analyzed to
7403 not contain loop carried dependences without checking that it is
7404 profitable to parallelize the loops.
7406 @item -ftree-coalesce-vars
7407 @opindex ftree-coalesce-vars
7408 While transforming the program out of the SSA representation, attempt to
7409 reduce copying by coalescing versions of different user-defined
7410 variables, instead of just compiler temporaries. This may severely
7411 limit the ability to debug an optimized program compiled with
7412 @option{-fno-var-tracking-assignments}. In the negated form, this flag
7413 prevents SSA coalescing of user variables. This option is enabled by
7414 default if optimization is enabled, and it does very little otherwise.
7416 @item -ftree-loop-if-convert
7417 @opindex ftree-loop-if-convert
7418 Attempt to transform conditional jumps in the innermost loops to
7419 branch-less equivalents. The intent is to remove control-flow from
7420 the innermost loops in order to improve the ability of the
7421 vectorization pass to handle these loops. This is enabled by default
7422 if vectorization is enabled.
7424 @item -ftree-loop-if-convert-stores
7425 @opindex ftree-loop-if-convert-stores
7426 Attempt to also if-convert conditional jumps containing memory writes.
7427 This transformation can be unsafe for multi-threaded programs as it
7428 transforms conditional memory writes into unconditional memory writes.
7431 for (i = 0; i < N; i++)
7437 for (i = 0; i < N; i++)
7438 A[i] = cond ? expr : A[i];
7440 potentially producing data races.
7442 @item -ftree-loop-distribution
7443 @opindex ftree-loop-distribution
7444 Perform loop distribution. This flag can improve cache performance on
7445 big loop bodies and allow further loop optimizations, like
7446 parallelization or vectorization, to take place. For example, the loop
7463 @item -ftree-loop-distribute-patterns
7464 @opindex ftree-loop-distribute-patterns
7465 Perform loop distribution of patterns that can be code generated with
7466 calls to a library. This flag is enabled by default at @option{-O3}.
7468 This pass distributes the initialization loops and generates a call to
7469 memset zero. For example, the loop
7485 and the initialization loop is transformed into a call to memset zero.
7487 @item -ftree-loop-im
7488 @opindex ftree-loop-im
7489 Perform loop invariant motion on trees. This pass moves only invariants that
7490 are hard to handle at RTL level (function calls, operations that expand to
7491 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7492 operands of conditions that are invariant out of the loop, so that we can use
7493 just trivial invariantness analysis in loop unswitching. The pass also includes
7496 @item -ftree-loop-ivcanon
7497 @opindex ftree-loop-ivcanon
7498 Create a canonical counter for number of iterations in loops for which
7499 determining number of iterations requires complicated analysis. Later
7500 optimizations then may determine the number easily. Useful especially
7501 in connection with unrolling.
7505 Perform induction variable optimizations (strength reduction, induction
7506 variable merging and induction variable elimination) on trees.
7508 @item -ftree-parallelize-loops=n
7509 @opindex ftree-parallelize-loops
7510 Parallelize loops, i.e., split their iteration space to run in n threads.
7511 This is only possible for loops whose iterations are independent
7512 and can be arbitrarily reordered. The optimization is only
7513 profitable on multiprocessor machines, for loops that are CPU-intensive,
7514 rather than constrained e.g.@: by memory bandwidth. This option
7515 implies @option{-pthread}, and thus is only supported on targets
7516 that have support for @option{-pthread}.
7520 Perform function-local points-to analysis on trees. This flag is
7521 enabled by default at @option{-O} and higher.
7525 Perform scalar replacement of aggregates. This pass replaces structure
7526 references with scalars to prevent committing structures to memory too
7527 early. This flag is enabled by default at @option{-O} and higher.
7531 Perform temporary expression replacement during the SSA->normal phase. Single
7532 use/single def temporaries are replaced at their use location with their
7533 defining expression. This results in non-GIMPLE code, but gives the expanders
7534 much more complex trees to work on resulting in better RTL generation. This is
7535 enabled by default at @option{-O} and higher.
7539 Perform straight-line strength reduction on trees. This recognizes related
7540 expressions involving multiplications and replaces them by less expensive
7541 calculations when possible. This is enabled by default at @option{-O} and
7544 @item -ftree-vectorize
7545 @opindex ftree-vectorize
7546 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
7547 and @option{-ftree-slp-vectorize} if not explicitly specified.
7549 @item -ftree-loop-vectorize
7550 @opindex ftree-loop-vectorize
7551 Perform loop vectorization on trees. This flag is enabled by default at
7552 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7554 @item -ftree-slp-vectorize
7555 @opindex ftree-slp-vectorize
7556 Perform basic block vectorization on trees. This flag is enabled by default at
7557 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7559 @item -fvect-cost-model=@var{model}
7560 @opindex fvect-cost-model
7561 Alter the cost model used for vectorization. The @var{model} argument
7562 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
7563 With the @samp{unlimited} model the vectorized code-path is assumed
7564 to be profitable while with the @samp{dynamic} model a runtime check
7565 guards the vectorized code-path to enable it only for iteration
7566 counts that will likely execute faster than when executing the original
7567 scalar loop. The @samp{cheap} model disables vectorization of
7568 loops where doing so would be cost prohibitive for example due to
7569 required runtime checks for data dependence or alignment but otherwise
7570 is equal to the @samp{dynamic} model.
7571 The default cost model depends on other optimization flags and is
7572 either @samp{dynamic} or @samp{cheap}.
7574 @item -fsimd-cost-model=@var{model}
7575 @opindex fsimd-cost-model
7576 Alter the cost model used for vectorization of loops marked with the OpenMP
7577 or Cilk Plus simd directive. The @var{model} argument should be one of
7578 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
7579 have the same meaning as described in @option{-fvect-cost-model} and by
7580 default a cost model defined with @option{-fvect-cost-model} is used.
7584 Perform Value Range Propagation on trees. This is similar to the
7585 constant propagation pass, but instead of values, ranges of values are
7586 propagated. This allows the optimizers to remove unnecessary range
7587 checks like array bound checks and null pointer checks. This is
7588 enabled by default at @option{-O2} and higher. Null pointer check
7589 elimination is only done if @option{-fdelete-null-pointer-checks} is
7593 @opindex fsplit-paths
7594 Split paths leading to loop backedges. This can improve dead code
7595 elimination and common subexpression elimination. This is enabled by
7596 default at @option{-O2} and above.
7598 @item -fsplit-ivs-in-unroller
7599 @opindex fsplit-ivs-in-unroller
7600 Enables expression of values of induction variables in later iterations
7601 of the unrolled loop using the value in the first iteration. This breaks
7602 long dependency chains, thus improving efficiency of the scheduling passes.
7604 A combination of @option{-fweb} and CSE is often sufficient to obtain the
7605 same effect. However, that is not reliable in cases where the loop body
7606 is more complicated than a single basic block. It also does not work at all
7607 on some architectures due to restrictions in the CSE pass.
7609 This optimization is enabled by default.
7611 @item -fvariable-expansion-in-unroller
7612 @opindex fvariable-expansion-in-unroller
7613 With this option, the compiler creates multiple copies of some
7614 local variables when unrolling a loop, which can result in superior code.
7616 @item -fpartial-inlining
7617 @opindex fpartial-inlining
7618 Inline parts of functions. This option has any effect only
7619 when inlining itself is turned on by the @option{-finline-functions}
7620 or @option{-finline-small-functions} options.
7622 Enabled at level @option{-O2}.
7624 @item -fpredictive-commoning
7625 @opindex fpredictive-commoning
7626 Perform predictive commoning optimization, i.e., reusing computations
7627 (especially memory loads and stores) performed in previous
7628 iterations of loops.
7630 This option is enabled at level @option{-O3}.
7632 @item -fprefetch-loop-arrays
7633 @opindex fprefetch-loop-arrays
7634 If supported by the target machine, generate instructions to prefetch
7635 memory to improve the performance of loops that access large arrays.
7637 This option may generate better or worse code; results are highly
7638 dependent on the structure of loops within the source code.
7640 Disabled at level @option{-Os}.
7643 @itemx -fno-peephole2
7644 @opindex fno-peephole
7645 @opindex fno-peephole2
7646 Disable any machine-specific peephole optimizations. The difference
7647 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7648 are implemented in the compiler; some targets use one, some use the
7649 other, a few use both.
7651 @option{-fpeephole} is enabled by default.
7652 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7654 @item -fno-guess-branch-probability
7655 @opindex fno-guess-branch-probability
7656 Do not guess branch probabilities using heuristics.
7658 GCC uses heuristics to guess branch probabilities if they are
7659 not provided by profiling feedback (@option{-fprofile-arcs}). These
7660 heuristics are based on the control flow graph. If some branch probabilities
7661 are specified by @code{__builtin_expect}, then the heuristics are
7662 used to guess branch probabilities for the rest of the control flow graph,
7663 taking the @code{__builtin_expect} info into account. The interactions
7664 between the heuristics and @code{__builtin_expect} can be complex, and in
7665 some cases, it may be useful to disable the heuristics so that the effects
7666 of @code{__builtin_expect} are easier to understand.
7668 The default is @option{-fguess-branch-probability} at levels
7669 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7671 @item -freorder-blocks
7672 @opindex freorder-blocks
7673 Reorder basic blocks in the compiled function in order to reduce number of
7674 taken branches and improve code locality.
7676 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7678 @item -freorder-blocks-algorithm=@var{algorithm}
7679 @opindex freorder-blocks-algorithm
7680 Use the specified algorithm for basic block reordering. The
7681 @var{algorithm} argument can be @samp{simple}, which does not increase
7682 code size (except sometimes due to secondary effects like alignment),
7683 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
7684 put all often executed code together, minimizing the number of branches
7685 executed by making extra copies of code.
7687 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
7688 @samp{stc} at levels @option{-O2}, @option{-O3}.
7690 @item -freorder-blocks-and-partition
7691 @opindex freorder-blocks-and-partition
7692 In addition to reordering basic blocks in the compiled function, in order
7693 to reduce number of taken branches, partitions hot and cold basic blocks
7694 into separate sections of the assembly and @file{.o} files, to improve
7695 paging and cache locality performance.
7697 This optimization is automatically turned off in the presence of
7698 exception handling, for linkonce sections, for functions with a user-defined
7699 section attribute and on any architecture that does not support named
7702 Enabled for x86 at levels @option{-O2}, @option{-O3}.
7704 @item -freorder-functions
7705 @opindex freorder-functions
7706 Reorder functions in the object file in order to
7707 improve code locality. This is implemented by using special
7708 subsections @code{.text.hot} for most frequently executed functions and
7709 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7710 the linker so object file format must support named sections and linker must
7711 place them in a reasonable way.
7713 Also profile feedback must be available to make this option effective. See
7714 @option{-fprofile-arcs} for details.
7716 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7718 @item -fstrict-aliasing
7719 @opindex fstrict-aliasing
7720 Allow the compiler to assume the strictest aliasing rules applicable to
7721 the language being compiled. For C (and C++), this activates
7722 optimizations based on the type of expressions. In particular, an
7723 object of one type is assumed never to reside at the same address as an
7724 object of a different type, unless the types are almost the same. For
7725 example, an @code{unsigned int} can alias an @code{int}, but not a
7726 @code{void*} or a @code{double}. A character type may alias any other
7729 @anchor{Type-punning}Pay special attention to code like this:
7742 The practice of reading from a different union member than the one most
7743 recently written to (called ``type-punning'') is common. Even with
7744 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7745 is accessed through the union type. So, the code above works as
7746 expected. @xref{Structures unions enumerations and bit-fields
7747 implementation}. However, this code might not:
7758 Similarly, access by taking the address, casting the resulting pointer
7759 and dereferencing the result has undefined behavior, even if the cast
7760 uses a union type, e.g.:
7764 return ((union a_union *) &d)->i;
7768 The @option{-fstrict-aliasing} option is enabled at levels
7769 @option{-O2}, @option{-O3}, @option{-Os}.
7771 @item -fstrict-overflow
7772 @opindex fstrict-overflow
7773 Allow the compiler to assume strict signed overflow rules, depending
7774 on the language being compiled. For C (and C++) this means that
7775 overflow when doing arithmetic with signed numbers is undefined, which
7776 means that the compiler may assume that it does not happen. This
7777 permits various optimizations. For example, the compiler assumes
7778 that an expression like @code{i + 10 > i} is always true for
7779 signed @code{i}. This assumption is only valid if signed overflow is
7780 undefined, as the expression is false if @code{i + 10} overflows when
7781 using twos complement arithmetic. When this option is in effect any
7782 attempt to determine whether an operation on signed numbers
7783 overflows must be written carefully to not actually involve overflow.
7785 This option also allows the compiler to assume strict pointer
7786 semantics: given a pointer to an object, if adding an offset to that
7787 pointer does not produce a pointer to the same object, the addition is
7788 undefined. This permits the compiler to conclude that @code{p + u >
7789 p} is always true for a pointer @code{p} and unsigned integer
7790 @code{u}. This assumption is only valid because pointer wraparound is
7791 undefined, as the expression is false if @code{p + u} overflows using
7792 twos complement arithmetic.
7794 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7795 that integer signed overflow is fully defined: it wraps. When
7796 @option{-fwrapv} is used, there is no difference between
7797 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7798 integers. With @option{-fwrapv} certain types of overflow are
7799 permitted. For example, if the compiler gets an overflow when doing
7800 arithmetic on constants, the overflowed value can still be used with
7801 @option{-fwrapv}, but not otherwise.
7803 The @option{-fstrict-overflow} option is enabled at levels
7804 @option{-O2}, @option{-O3}, @option{-Os}.
7806 @item -falign-functions
7807 @itemx -falign-functions=@var{n}
7808 @opindex falign-functions
7809 Align the start of functions to the next power-of-two greater than
7810 @var{n}, skipping up to @var{n} bytes. For instance,
7811 @option{-falign-functions=32} aligns functions to the next 32-byte
7812 boundary, but @option{-falign-functions=24} aligns to the next
7813 32-byte boundary only if this can be done by skipping 23 bytes or less.
7815 @option{-fno-align-functions} and @option{-falign-functions=1} are
7816 equivalent and mean that functions are not aligned.
7818 Some assemblers only support this flag when @var{n} is a power of two;
7819 in that case, it is rounded up.
7821 If @var{n} is not specified or is zero, use a machine-dependent default.
7823 Enabled at levels @option{-O2}, @option{-O3}.
7825 @item -falign-labels
7826 @itemx -falign-labels=@var{n}
7827 @opindex falign-labels
7828 Align all branch targets to a power-of-two boundary, skipping up to
7829 @var{n} bytes like @option{-falign-functions}. This option can easily
7830 make code slower, because it must insert dummy operations for when the
7831 branch target is reached in the usual flow of the code.
7833 @option{-fno-align-labels} and @option{-falign-labels=1} are
7834 equivalent and mean that labels are not aligned.
7836 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7837 are greater than this value, then their values are used instead.
7839 If @var{n} is not specified or is zero, use a machine-dependent default
7840 which is very likely to be @samp{1}, meaning no alignment.
7842 Enabled at levels @option{-O2}, @option{-O3}.
7845 @itemx -falign-loops=@var{n}
7846 @opindex falign-loops
7847 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7848 like @option{-falign-functions}. If the loops are
7849 executed many times, this makes up for any execution of the dummy
7852 @option{-fno-align-loops} and @option{-falign-loops=1} are
7853 equivalent and mean that loops are not aligned.
7855 If @var{n} is not specified or is zero, use a machine-dependent default.
7857 Enabled at levels @option{-O2}, @option{-O3}.
7860 @itemx -falign-jumps=@var{n}
7861 @opindex falign-jumps
7862 Align branch targets to a power-of-two boundary, for branch targets
7863 where the targets can only be reached by jumping, skipping up to @var{n}
7864 bytes like @option{-falign-functions}. In this case, no dummy operations
7867 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7868 equivalent and mean that loops are not aligned.
7870 If @var{n} is not specified or is zero, use a machine-dependent default.
7872 Enabled at levels @option{-O2}, @option{-O3}.
7874 @item -funit-at-a-time
7875 @opindex funit-at-a-time
7876 This option is left for compatibility reasons. @option{-funit-at-a-time}
7877 has no effect, while @option{-fno-unit-at-a-time} implies
7878 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7882 @item -fno-toplevel-reorder
7883 @opindex fno-toplevel-reorder
7884 Do not reorder top-level functions, variables, and @code{asm}
7885 statements. Output them in the same order that they appear in the
7886 input file. When this option is used, unreferenced static variables
7887 are not removed. This option is intended to support existing code
7888 that relies on a particular ordering. For new code, it is better to
7889 use attributes when possible.
7891 Enabled at level @option{-O0}. When disabled explicitly, it also implies
7892 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
7897 Constructs webs as commonly used for register allocation purposes and assign
7898 each web individual pseudo register. This allows the register allocation pass
7899 to operate on pseudos directly, but also strengthens several other optimization
7900 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7901 however, make debugging impossible, since variables no longer stay in a
7904 Enabled by default with @option{-funroll-loops}.
7906 @item -fwhole-program
7907 @opindex fwhole-program
7908 Assume that the current compilation unit represents the whole program being
7909 compiled. All public functions and variables with the exception of @code{main}
7910 and those merged by attribute @code{externally_visible} become static functions
7911 and in effect are optimized more aggressively by interprocedural optimizers.
7913 This option should not be used in combination with @option{-flto}.
7914 Instead relying on a linker plugin should provide safer and more precise
7917 @item -flto[=@var{n}]
7919 This option runs the standard link-time optimizer. When invoked
7920 with source code, it generates GIMPLE (one of GCC's internal
7921 representations) and writes it to special ELF sections in the object
7922 file. When the object files are linked together, all the function
7923 bodies are read from these ELF sections and instantiated as if they
7924 had been part of the same translation unit.
7926 To use the link-time optimizer, @option{-flto} and optimization
7927 options should be specified at compile time and during the final link.
7928 It is recommended that you compile all the files participating in the
7929 same link with the same options and also specify those options at
7934 gcc -c -O2 -flto foo.c
7935 gcc -c -O2 -flto bar.c
7936 gcc -o myprog -flto -O2 foo.o bar.o
7939 The first two invocations to GCC save a bytecode representation
7940 of GIMPLE into special ELF sections inside @file{foo.o} and
7941 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
7942 @file{foo.o} and @file{bar.o}, merges the two files into a single
7943 internal image, and compiles the result as usual. Since both
7944 @file{foo.o} and @file{bar.o} are merged into a single image, this
7945 causes all the interprocedural analyses and optimizations in GCC to
7946 work across the two files as if they were a single one. This means,
7947 for example, that the inliner is able to inline functions in
7948 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7950 Another (simpler) way to enable link-time optimization is:
7953 gcc -o myprog -flto -O2 foo.c bar.c
7956 The above generates bytecode for @file{foo.c} and @file{bar.c},
7957 merges them together into a single GIMPLE representation and optimizes
7958 them as usual to produce @file{myprog}.
7960 The only important thing to keep in mind is that to enable link-time
7961 optimizations you need to use the GCC driver to perform the link step.
7962 GCC then automatically performs link-time optimization if any of the
7963 objects involved were compiled with the @option{-flto} command-line option.
7965 should specify the optimization options to be used for link-time
7966 optimization though GCC tries to be clever at guessing an
7967 optimization level to use from the options used at compile time
7968 if you fail to specify one at link time. You can always override
7969 the automatic decision to do link-time optimization at link time
7970 by passing @option{-fno-lto} to the link command.
7972 To make whole program optimization effective, it is necessary to make
7973 certain whole program assumptions. The compiler needs to know
7974 what functions and variables can be accessed by libraries and runtime
7975 outside of the link-time optimized unit. When supported by the linker,
7976 the linker plugin (see @option{-fuse-linker-plugin}) passes information
7977 to the compiler about used and externally visible symbols. When
7978 the linker plugin is not available, @option{-fwhole-program} should be
7979 used to allow the compiler to make these assumptions, which leads
7980 to more aggressive optimization decisions.
7982 When @option{-fuse-linker-plugin} is not enabled, when a file is
7983 compiled with @option{-flto}, the generated object file is larger than
7984 a regular object file because it contains GIMPLE bytecodes and the usual
7985 final code (see @option{-ffat-lto-objects}. This means that
7986 object files with LTO information can be linked as normal object
7987 files; if @option{-fno-lto} is passed to the linker, no
7988 interprocedural optimizations are applied. Note that when
7989 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
7990 but you cannot perform a regular, non-LTO link on them.
7992 Additionally, the optimization flags used to compile individual files
7993 are not necessarily related to those used at link time. For instance,
7996 gcc -c -O0 -ffat-lto-objects -flto foo.c
7997 gcc -c -O0 -ffat-lto-objects -flto bar.c
7998 gcc -o myprog -O3 foo.o bar.o
8001 This produces individual object files with unoptimized assembler
8002 code, but the resulting binary @file{myprog} is optimized at
8003 @option{-O3}. If, instead, the final binary is generated with
8004 @option{-fno-lto}, then @file{myprog} is not optimized.
8006 When producing the final binary, GCC only
8007 applies link-time optimizations to those files that contain bytecode.
8008 Therefore, you can mix and match object files and libraries with
8009 GIMPLE bytecodes and final object code. GCC automatically selects
8010 which files to optimize in LTO mode and which files to link without
8013 There are some code generation flags preserved by GCC when
8014 generating bytecodes, as they need to be used during the final link
8015 stage. Generally options specified at link time override those
8016 specified at compile time.
8018 If you do not specify an optimization level option @option{-O} at
8019 link time, then GCC uses the highest optimization level
8020 used when compiling the object files.
8022 Currently, the following options and their settings are taken from
8023 the first object file that explicitly specifies them:
8024 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8025 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8026 and all the @option{-m} target flags.
8028 Certain ABI-changing flags are required to match in all compilation units,
8029 and trying to override this at link time with a conflicting value
8030 is ignored. This includes options such as @option{-freg-struct-return}
8031 and @option{-fpcc-struct-return}.
8033 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8034 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8035 are passed through to the link stage and merged conservatively for
8036 conflicting translation units. Specifically
8037 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8038 precedence; and for example @option{-ffp-contract=off} takes precedence
8039 over @option{-ffp-contract=fast}. You can override them at link time.
8041 If LTO encounters objects with C linkage declared with incompatible
8042 types in separate translation units to be linked together (undefined
8043 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8044 issued. The behavior is still undefined at run time. Similar
8045 diagnostics may be raised for other languages.
8047 Another feature of LTO is that it is possible to apply interprocedural
8048 optimizations on files written in different languages:
8053 gfortran -c -flto baz.f90
8054 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8057 Notice that the final link is done with @command{g++} to get the C++
8058 runtime libraries and @option{-lgfortran} is added to get the Fortran
8059 runtime libraries. In general, when mixing languages in LTO mode, you
8060 should use the same link command options as when mixing languages in a
8061 regular (non-LTO) compilation.
8063 If object files containing GIMPLE bytecode are stored in a library archive, say
8064 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8065 are using a linker with plugin support. To create static libraries suitable
8066 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8067 and @command{ranlib};
8068 to show the symbols of object files with GIMPLE bytecode, use
8069 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8070 and @command{nm} have been compiled with plugin support. At link time, use the the
8071 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8072 the LTO optimization process:
8075 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8078 With the linker plugin enabled, the linker extracts the needed
8079 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8080 to make them part of the aggregated GIMPLE image to be optimized.
8082 If you are not using a linker with plugin support and/or do not
8083 enable the linker plugin, then the objects inside @file{libfoo.a}
8084 are extracted and linked as usual, but they do not participate
8085 in the LTO optimization process. In order to make a static library suitable
8086 for both LTO optimization and usual linkage, compile its object files with
8087 @option{-flto} @option{-ffat-lto-objects}.
8089 Link-time optimizations do not require the presence of the whole program to
8090 operate. If the program does not require any symbols to be exported, it is
8091 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8092 the interprocedural optimizers to use more aggressive assumptions which may
8093 lead to improved optimization opportunities.
8094 Use of @option{-fwhole-program} is not needed when linker plugin is
8095 active (see @option{-fuse-linker-plugin}).
8097 The current implementation of LTO makes no
8098 attempt to generate bytecode that is portable between different
8099 types of hosts. The bytecode files are versioned and there is a
8100 strict version check, so bytecode files generated in one version of
8101 GCC do not work with an older or newer version of GCC.
8103 Link-time optimization does not work well with generation of debugging
8104 information. Combining @option{-flto} with
8105 @option{-g} is currently experimental and expected to produce unexpected
8108 If you specify the optional @var{n}, the optimization and code
8109 generation done at link time is executed in parallel using @var{n}
8110 parallel jobs by utilizing an installed @command{make} program. The
8111 environment variable @env{MAKE} may be used to override the program
8112 used. The default value for @var{n} is 1.
8114 You can also specify @option{-flto=jobserver} to use GNU make's
8115 job server mode to determine the number of parallel jobs. This
8116 is useful when the Makefile calling GCC is already executing in parallel.
8117 You must prepend a @samp{+} to the command recipe in the parent Makefile
8118 for this to work. This option likely only works if @env{MAKE} is
8121 @item -flto-partition=@var{alg}
8122 @opindex flto-partition
8123 Specify the partitioning algorithm used by the link-time optimizer.
8124 The value is either @samp{1to1} to specify a partitioning mirroring
8125 the original source files or @samp{balanced} to specify partitioning
8126 into equally sized chunks (whenever possible) or @samp{max} to create
8127 new partition for every symbol where possible. Specifying @samp{none}
8128 as an algorithm disables partitioning and streaming completely.
8129 The default value is @samp{balanced}. While @samp{1to1} can be used
8130 as an workaround for various code ordering issues, the @samp{max}
8131 partitioning is intended for internal testing only.
8132 The value @samp{one} specifies that exactly one partition should be
8133 used while the value @samp{none} bypasses partitioning and executes
8134 the link-time optimization step directly from the WPA phase.
8136 @item -flto-odr-type-merging
8137 @opindex flto-odr-type-merging
8138 Enable streaming of mangled types names of C++ types and their unification
8139 at link time. This increases size of LTO object files, but enables
8140 diagnostics about One Definition Rule violations.
8142 @item -flto-compression-level=@var{n}
8143 @opindex flto-compression-level
8144 This option specifies the level of compression used for intermediate
8145 language written to LTO object files, and is only meaningful in
8146 conjunction with LTO mode (@option{-flto}). Valid
8147 values are 0 (no compression) to 9 (maximum compression). Values
8148 outside this range are clamped to either 0 or 9. If the option is not
8149 given, a default balanced compression setting is used.
8151 @item -fuse-linker-plugin
8152 @opindex fuse-linker-plugin
8153 Enables the use of a linker plugin during link-time optimization. This
8154 option relies on plugin support in the linker, which is available in gold
8155 or in GNU ld 2.21 or newer.
8157 This option enables the extraction of object files with GIMPLE bytecode out
8158 of library archives. This improves the quality of optimization by exposing
8159 more code to the link-time optimizer. This information specifies what
8160 symbols can be accessed externally (by non-LTO object or during dynamic
8161 linking). Resulting code quality improvements on binaries (and shared
8162 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
8163 See @option{-flto} for a description of the effect of this flag and how to
8166 This option is enabled by default when LTO support in GCC is enabled
8167 and GCC was configured for use with
8168 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8170 @item -ffat-lto-objects
8171 @opindex ffat-lto-objects
8172 Fat LTO objects are object files that contain both the intermediate language
8173 and the object code. This makes them usable for both LTO linking and normal
8174 linking. This option is effective only when compiling with @option{-flto}
8175 and is ignored at link time.
8177 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8178 requires the complete toolchain to be aware of LTO. It requires a linker with
8179 linker plugin support for basic functionality. Additionally,
8180 @command{nm}, @command{ar} and @command{ranlib}
8181 need to support linker plugins to allow a full-featured build environment
8182 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8183 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8184 to these tools. With non fat LTO makefiles need to be modified to use them.
8186 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8189 @item -fcompare-elim
8190 @opindex fcompare-elim
8191 After register allocation and post-register allocation instruction splitting,
8192 identify arithmetic instructions that compute processor flags similar to a
8193 comparison operation based on that arithmetic. If possible, eliminate the
8194 explicit comparison operation.
8196 This pass only applies to certain targets that cannot explicitly represent
8197 the comparison operation before register allocation is complete.
8199 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8201 @item -fcprop-registers
8202 @opindex fcprop-registers
8203 After register allocation and post-register allocation instruction splitting,
8204 perform a copy-propagation pass to try to reduce scheduling dependencies
8205 and occasionally eliminate the copy.
8207 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8209 @item -fprofile-correction
8210 @opindex fprofile-correction
8211 Profiles collected using an instrumented binary for multi-threaded programs may
8212 be inconsistent due to missed counter updates. When this option is specified,
8213 GCC uses heuristics to correct or smooth out such inconsistencies. By
8214 default, GCC emits an error message when an inconsistent profile is detected.
8217 @itemx -fprofile-use=@var{path}
8218 @opindex fprofile-use
8219 Enable profile feedback-directed optimizations,
8220 and the following optimizations
8221 which are generally profitable only with profile feedback available:
8222 @option{-fbranch-probabilities}, @option{-fvpt},
8223 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8224 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
8226 Before you can use this option, you must first generate profiling information.
8227 @xref{Optimize Options}, for information about the @option{-fprofile-generate}
8230 By default, GCC emits an error message if the feedback profiles do not
8231 match the source code. This error can be turned into a warning by using
8232 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8235 If @var{path} is specified, GCC looks at the @var{path} to find
8236 the profile feedback data files. See @option{-fprofile-dir}.
8238 @item -fauto-profile
8239 @itemx -fauto-profile=@var{path}
8240 @opindex fauto-profile
8241 Enable sampling-based feedback-directed optimizations,
8242 and the following optimizations
8243 which are generally profitable only with profile feedback available:
8244 @option{-fbranch-probabilities}, @option{-fvpt},
8245 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8246 @option{-ftree-vectorize},
8247 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
8248 @option{-fpredictive-commoning}, @option{-funswitch-loops},
8249 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
8251 @var{path} is the name of a file containing AutoFDO profile information.
8252 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
8254 Producing an AutoFDO profile data file requires running your program
8255 with the @command{perf} utility on a supported GNU/Linux target system.
8256 For more information, see @uref{https://perf.wiki.kernel.org/}.
8260 perf record -e br_inst_retired:near_taken -b -o perf.data \
8264 Then use the @command{create_gcov} tool to convert the raw profile data
8265 to a format that can be used by GCC.@ You must also supply the
8266 unstripped binary for your program to this tool.
8267 See @uref{https://github.com/google/autofdo}.
8271 create_gcov --binary=your_program.unstripped --profile=perf.data \
8276 The following options control compiler behavior regarding floating-point
8277 arithmetic. These options trade off between speed and
8278 correctness. All must be specifically enabled.
8282 @opindex ffloat-store
8283 Do not store floating-point variables in registers, and inhibit other
8284 options that might change whether a floating-point value is taken from a
8287 @cindex floating-point precision
8288 This option prevents undesirable excess precision on machines such as
8289 the 68000 where the floating registers (of the 68881) keep more
8290 precision than a @code{double} is supposed to have. Similarly for the
8291 x86 architecture. For most programs, the excess precision does only
8292 good, but a few programs rely on the precise definition of IEEE floating
8293 point. Use @option{-ffloat-store} for such programs, after modifying
8294 them to store all pertinent intermediate computations into variables.
8296 @item -fexcess-precision=@var{style}
8297 @opindex fexcess-precision
8298 This option allows further control over excess precision on machines
8299 where floating-point registers have more precision than the IEEE
8300 @code{float} and @code{double} types and the processor does not
8301 support operations rounding to those types. By default,
8302 @option{-fexcess-precision=fast} is in effect; this means that
8303 operations are carried out in the precision of the registers and that
8304 it is unpredictable when rounding to the types specified in the source
8305 code takes place. When compiling C, if
8306 @option{-fexcess-precision=standard} is specified then excess
8307 precision follows the rules specified in ISO C99; in particular,
8308 both casts and assignments cause values to be rounded to their
8309 semantic types (whereas @option{-ffloat-store} only affects
8310 assignments). This option is enabled by default for C if a strict
8311 conformance option such as @option{-std=c99} is used.
8314 @option{-fexcess-precision=standard} is not implemented for languages
8315 other than C, and has no effect if
8316 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8317 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8318 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8319 semantics apply without excess precision, and in the latter, rounding
8324 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8325 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8326 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8328 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8330 This option is not turned on by any @option{-O} option besides
8331 @option{-Ofast} since it can result in incorrect output for programs
8332 that depend on an exact implementation of IEEE or ISO rules/specifications
8333 for math functions. It may, however, yield faster code for programs
8334 that do not require the guarantees of these specifications.
8336 @item -fno-math-errno
8337 @opindex fno-math-errno
8338 Do not set @code{errno} after calling math functions that are executed
8339 with a single instruction, e.g., @code{sqrt}. A program that relies on
8340 IEEE exceptions for math error handling may want to use this flag
8341 for speed while maintaining IEEE arithmetic compatibility.
8343 This option is not turned on by any @option{-O} option since
8344 it can result in incorrect output for programs that depend on
8345 an exact implementation of IEEE or ISO rules/specifications for
8346 math functions. It may, however, yield faster code for programs
8347 that do not require the guarantees of these specifications.
8349 The default is @option{-fmath-errno}.
8351 On Darwin systems, the math library never sets @code{errno}. There is
8352 therefore no reason for the compiler to consider the possibility that
8353 it might, and @option{-fno-math-errno} is the default.
8355 @item -funsafe-math-optimizations
8356 @opindex funsafe-math-optimizations
8358 Allow optimizations for floating-point arithmetic that (a) assume
8359 that arguments and results are valid and (b) may violate IEEE or
8360 ANSI standards. When used at link time, it may include libraries
8361 or startup files that change the default FPU control word or other
8362 similar optimizations.
8364 This option is not turned on by any @option{-O} option since
8365 it can result in incorrect output for programs that depend on
8366 an exact implementation of IEEE or ISO rules/specifications for
8367 math functions. It may, however, yield faster code for programs
8368 that do not require the guarantees of these specifications.
8369 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8370 @option{-fassociative-math} and @option{-freciprocal-math}.
8372 The default is @option{-fno-unsafe-math-optimizations}.
8374 @item -fassociative-math
8375 @opindex fassociative-math
8377 Allow re-association of operands in series of floating-point operations.
8378 This violates the ISO C and C++ language standard by possibly changing
8379 computation result. NOTE: re-ordering may change the sign of zero as
8380 well as ignore NaNs and inhibit or create underflow or overflow (and
8381 thus cannot be used on code that relies on rounding behavior like
8382 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8383 and thus may not be used when ordered comparisons are required.
8384 This option requires that both @option{-fno-signed-zeros} and
8385 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8386 much sense with @option{-frounding-math}. For Fortran the option
8387 is automatically enabled when both @option{-fno-signed-zeros} and
8388 @option{-fno-trapping-math} are in effect.
8390 The default is @option{-fno-associative-math}.
8392 @item -freciprocal-math
8393 @opindex freciprocal-math
8395 Allow the reciprocal of a value to be used instead of dividing by
8396 the value if this enables optimizations. For example @code{x / y}
8397 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
8398 is subject to common subexpression elimination. Note that this loses
8399 precision and increases the number of flops operating on the value.
8401 The default is @option{-fno-reciprocal-math}.
8403 @item -ffinite-math-only
8404 @opindex ffinite-math-only
8405 Allow optimizations for floating-point arithmetic that assume
8406 that arguments and results are not NaNs or +-Infs.
8408 This option is not turned on by any @option{-O} option since
8409 it can result in incorrect output for programs that depend on
8410 an exact implementation of IEEE or ISO rules/specifications for
8411 math functions. It may, however, yield faster code for programs
8412 that do not require the guarantees of these specifications.
8414 The default is @option{-fno-finite-math-only}.
8416 @item -fno-signed-zeros
8417 @opindex fno-signed-zeros
8418 Allow optimizations for floating-point arithmetic that ignore the
8419 signedness of zero. IEEE arithmetic specifies the behavior of
8420 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8421 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8422 This option implies that the sign of a zero result isn't significant.
8424 The default is @option{-fsigned-zeros}.
8426 @item -fno-trapping-math
8427 @opindex fno-trapping-math
8428 Compile code assuming that floating-point operations cannot generate
8429 user-visible traps. These traps include division by zero, overflow,
8430 underflow, inexact result and invalid operation. This option requires
8431 that @option{-fno-signaling-nans} be in effect. Setting this option may
8432 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8434 This option should never be turned on by any @option{-O} option since
8435 it can result in incorrect output for programs that depend on
8436 an exact implementation of IEEE or ISO rules/specifications for
8439 The default is @option{-ftrapping-math}.
8441 @item -frounding-math
8442 @opindex frounding-math
8443 Disable transformations and optimizations that assume default floating-point
8444 rounding behavior. This is round-to-zero for all floating point
8445 to integer conversions, and round-to-nearest for all other arithmetic
8446 truncations. This option should be specified for programs that change
8447 the FP rounding mode dynamically, or that may be executed with a
8448 non-default rounding mode. This option disables constant folding of
8449 floating-point expressions at compile time (which may be affected by
8450 rounding mode) and arithmetic transformations that are unsafe in the
8451 presence of sign-dependent rounding modes.
8453 The default is @option{-fno-rounding-math}.
8455 This option is experimental and does not currently guarantee to
8456 disable all GCC optimizations that are affected by rounding mode.
8457 Future versions of GCC may provide finer control of this setting
8458 using C99's @code{FENV_ACCESS} pragma. This command-line option
8459 will be used to specify the default state for @code{FENV_ACCESS}.
8461 @item -fsignaling-nans
8462 @opindex fsignaling-nans
8463 Compile code assuming that IEEE signaling NaNs may generate user-visible
8464 traps during floating-point operations. Setting this option disables
8465 optimizations that may change the number of exceptions visible with
8466 signaling NaNs. This option implies @option{-ftrapping-math}.
8468 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8471 The default is @option{-fno-signaling-nans}.
8473 This option is experimental and does not currently guarantee to
8474 disable all GCC optimizations that affect signaling NaN behavior.
8476 @item -fsingle-precision-constant
8477 @opindex fsingle-precision-constant
8478 Treat floating-point constants as single precision instead of
8479 implicitly converting them to double-precision constants.
8481 @item -fcx-limited-range
8482 @opindex fcx-limited-range
8483 When enabled, this option states that a range reduction step is not
8484 needed when performing complex division. Also, there is no checking
8485 whether the result of a complex multiplication or division is @code{NaN
8486 + I*NaN}, with an attempt to rescue the situation in that case. The
8487 default is @option{-fno-cx-limited-range}, but is enabled by
8488 @option{-ffast-math}.
8490 This option controls the default setting of the ISO C99
8491 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8494 @item -fcx-fortran-rules
8495 @opindex fcx-fortran-rules
8496 Complex multiplication and division follow Fortran rules. Range
8497 reduction is done as part of complex division, but there is no checking
8498 whether the result of a complex multiplication or division is @code{NaN
8499 + I*NaN}, with an attempt to rescue the situation in that case.
8501 The default is @option{-fno-cx-fortran-rules}.
8505 The following options control optimizations that may improve
8506 performance, but are not enabled by any @option{-O} options. This
8507 section includes experimental options that may produce broken code.
8510 @item -fbranch-probabilities
8511 @opindex fbranch-probabilities
8512 After running a program compiled with @option{-fprofile-arcs}
8513 (@pxref{Instrumentation Options}),
8514 you can compile it a second time using
8515 @option{-fbranch-probabilities}, to improve optimizations based on
8516 the number of times each branch was taken. When a program
8517 compiled with @option{-fprofile-arcs} exits, it saves arc execution
8518 counts to a file called @file{@var{sourcename}.gcda} for each source
8519 file. The information in this data file is very dependent on the
8520 structure of the generated code, so you must use the same source code
8521 and the same optimization options for both compilations.
8523 With @option{-fbranch-probabilities}, GCC puts a
8524 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8525 These can be used to improve optimization. Currently, they are only
8526 used in one place: in @file{reorg.c}, instead of guessing which path a
8527 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8528 exactly determine which path is taken more often.
8530 @item -fprofile-values
8531 @opindex fprofile-values
8532 If combined with @option{-fprofile-arcs}, it adds code so that some
8533 data about values of expressions in the program is gathered.
8535 With @option{-fbranch-probabilities}, it reads back the data gathered
8536 from profiling values of expressions for usage in optimizations.
8538 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8540 @item -fprofile-reorder-functions
8541 @opindex fprofile-reorder-functions
8542 Function reordering based on profile instrumentation collects
8543 first time of execution of a function and orders these functions
8546 Enabled with @option{-fprofile-use}.
8550 If combined with @option{-fprofile-arcs}, this option instructs the compiler
8551 to add code to gather information about values of expressions.
8553 With @option{-fbranch-probabilities}, it reads back the data gathered
8554 and actually performs the optimizations based on them.
8555 Currently the optimizations include specialization of division operations
8556 using the knowledge about the value of the denominator.
8558 @item -frename-registers
8559 @opindex frename-registers
8560 Attempt to avoid false dependencies in scheduled code by making use
8561 of registers left over after register allocation. This optimization
8562 most benefits processors with lots of registers. Depending on the
8563 debug information format adopted by the target, however, it can
8564 make debugging impossible, since variables no longer stay in
8565 a ``home register''.
8567 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8569 @item -fschedule-fusion
8570 @opindex fschedule-fusion
8571 Performs a target dependent pass over the instruction stream to schedule
8572 instructions of same type together because target machine can execute them
8573 more efficiently if they are adjacent to each other in the instruction flow.
8575 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8579 Perform tail duplication to enlarge superblock size. This transformation
8580 simplifies the control flow of the function allowing other optimizations to do
8583 Enabled with @option{-fprofile-use}.
8585 @item -funroll-loops
8586 @opindex funroll-loops
8587 Unroll loops whose number of iterations can be determined at compile time or
8588 upon entry to the loop. @option{-funroll-loops} implies
8589 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8590 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8591 a small constant number of iterations). This option makes code larger, and may
8592 or may not make it run faster.
8594 Enabled with @option{-fprofile-use}.
8596 @item -funroll-all-loops
8597 @opindex funroll-all-loops
8598 Unroll all loops, even if their number of iterations is uncertain when
8599 the loop is entered. This usually makes programs run more slowly.
8600 @option{-funroll-all-loops} implies the same options as
8601 @option{-funroll-loops}.
8604 @opindex fpeel-loops
8605 Peels loops for which there is enough information that they do not
8606 roll much (from profile feedback). It also turns on complete loop peeling
8607 (i.e.@: complete removal of loops with small constant number of iterations).
8609 Enabled with @option{-fprofile-use}.
8611 @item -fmove-loop-invariants
8612 @opindex fmove-loop-invariants
8613 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8614 at level @option{-O1}
8616 @item -funswitch-loops
8617 @opindex funswitch-loops
8618 Move branches with loop invariant conditions out of the loop, with duplicates
8619 of the loop on both branches (modified according to result of the condition).
8621 @item -ffunction-sections
8622 @itemx -fdata-sections
8623 @opindex ffunction-sections
8624 @opindex fdata-sections
8625 Place each function or data item into its own section in the output
8626 file if the target supports arbitrary sections. The name of the
8627 function or the name of the data item determines the section's name
8630 Use these options on systems where the linker can perform optimizations
8631 to improve locality of reference in the instruction space. Most systems
8632 using the ELF object format and SPARC processors running Solaris 2 have
8633 linkers with such optimizations. AIX may have these optimizations in
8636 Only use these options when there are significant benefits from doing
8637 so. When you specify these options, the assembler and linker
8638 create larger object and executable files and are also slower.
8639 You cannot use @command{gprof} on all systems if you
8640 specify this option, and you may have problems with debugging if
8641 you specify both this option and @option{-g}.
8643 @item -fbranch-target-load-optimize
8644 @opindex fbranch-target-load-optimize
8645 Perform branch target register load optimization before prologue / epilogue
8647 The use of target registers can typically be exposed only during reload,
8648 thus hoisting loads out of loops and doing inter-block scheduling needs
8649 a separate optimization pass.
8651 @item -fbranch-target-load-optimize2
8652 @opindex fbranch-target-load-optimize2
8653 Perform branch target register load optimization after prologue / epilogue
8656 @item -fbtr-bb-exclusive
8657 @opindex fbtr-bb-exclusive
8658 When performing branch target register load optimization, don't reuse
8659 branch target registers within any basic block.
8662 @opindex fstdarg-opt
8663 Optimize the prologue of variadic argument functions with respect to usage of
8666 @item -fsection-anchors
8667 @opindex fsection-anchors
8668 Try to reduce the number of symbolic address calculations by using
8669 shared ``anchor'' symbols to address nearby objects. This transformation
8670 can help to reduce the number of GOT entries and GOT accesses on some
8673 For example, the implementation of the following function @code{foo}:
8677 int foo (void) @{ return a + b + c; @}
8681 usually calculates the addresses of all three variables, but if you
8682 compile it with @option{-fsection-anchors}, it accesses the variables
8683 from a common anchor point instead. The effect is similar to the
8684 following pseudocode (which isn't valid C):
8689 register int *xr = &x;
8690 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8694 Not all targets support this option.
8696 @item --param @var{name}=@var{value}
8698 In some places, GCC uses various constants to control the amount of
8699 optimization that is done. For example, GCC does not inline functions
8700 that contain more than a certain number of instructions. You can
8701 control some of these constants on the command line using the
8702 @option{--param} option.
8704 The names of specific parameters, and the meaning of the values, are
8705 tied to the internals of the compiler, and are subject to change
8706 without notice in future releases.
8708 In each case, the @var{value} is an integer. The allowable choices for
8712 @item predictable-branch-outcome
8713 When branch is predicted to be taken with probability lower than this threshold
8714 (in percent), then it is considered well predictable. The default is 10.
8716 @item max-rtl-if-conversion-insns
8717 RTL if-conversion tries to remove conditional branches around a block and
8718 replace them with conditionally executed instructions. This parameter
8719 gives the maximum number of instructions in a block which should be
8720 considered for if-conversion. The default is 10, though the compiler will
8721 also use other heuristics to decide whether if-conversion is likely to be
8724 @item max-crossjump-edges
8725 The maximum number of incoming edges to consider for cross-jumping.
8726 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8727 the number of edges incoming to each block. Increasing values mean
8728 more aggressive optimization, making the compilation time increase with
8729 probably small improvement in executable size.
8731 @item min-crossjump-insns
8732 The minimum number of instructions that must be matched at the end
8733 of two blocks before cross-jumping is performed on them. This
8734 value is ignored in the case where all instructions in the block being
8735 cross-jumped from are matched. The default value is 5.
8737 @item max-grow-copy-bb-insns
8738 The maximum code size expansion factor when copying basic blocks
8739 instead of jumping. The expansion is relative to a jump instruction.
8740 The default value is 8.
8742 @item max-goto-duplication-insns
8743 The maximum number of instructions to duplicate to a block that jumps
8744 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8745 passes, GCC factors computed gotos early in the compilation process,
8746 and unfactors them as late as possible. Only computed jumps at the
8747 end of a basic blocks with no more than max-goto-duplication-insns are
8748 unfactored. The default value is 8.
8750 @item max-delay-slot-insn-search
8751 The maximum number of instructions to consider when looking for an
8752 instruction to fill a delay slot. If more than this arbitrary number of
8753 instructions are searched, the time savings from filling the delay slot
8754 are minimal, so stop searching. Increasing values mean more
8755 aggressive optimization, making the compilation time increase with probably
8756 small improvement in execution time.
8758 @item max-delay-slot-live-search
8759 When trying to fill delay slots, the maximum number of instructions to
8760 consider when searching for a block with valid live register
8761 information. Increasing this arbitrarily chosen value means more
8762 aggressive optimization, increasing the compilation time. This parameter
8763 should be removed when the delay slot code is rewritten to maintain the
8766 @item max-gcse-memory
8767 The approximate maximum amount of memory that can be allocated in
8768 order to perform the global common subexpression elimination
8769 optimization. If more memory than specified is required, the
8770 optimization is not done.
8772 @item max-gcse-insertion-ratio
8773 If the ratio of expression insertions to deletions is larger than this value
8774 for any expression, then RTL PRE inserts or removes the expression and thus
8775 leaves partially redundant computations in the instruction stream. The default value is 20.
8777 @item max-pending-list-length
8778 The maximum number of pending dependencies scheduling allows
8779 before flushing the current state and starting over. Large functions
8780 with few branches or calls can create excessively large lists which
8781 needlessly consume memory and resources.
8783 @item max-modulo-backtrack-attempts
8784 The maximum number of backtrack attempts the scheduler should make
8785 when modulo scheduling a loop. Larger values can exponentially increase
8788 @item max-inline-insns-single
8789 Several parameters control the tree inliner used in GCC@.
8790 This number sets the maximum number of instructions (counted in GCC's
8791 internal representation) in a single function that the tree inliner
8792 considers for inlining. This only affects functions declared
8793 inline and methods implemented in a class declaration (C++).
8794 The default value is 400.
8796 @item max-inline-insns-auto
8797 When you use @option{-finline-functions} (included in @option{-O3}),
8798 a lot of functions that would otherwise not be considered for inlining
8799 by the compiler are investigated. To those functions, a different
8800 (more restrictive) limit compared to functions declared inline can
8802 The default value is 40.
8804 @item inline-min-speedup
8805 When estimated performance improvement of caller + callee runtime exceeds this
8806 threshold (in precent), the function can be inlined regardless the limit on
8807 @option{--param max-inline-insns-single} and @option{--param
8808 max-inline-insns-auto}.
8810 @item large-function-insns
8811 The limit specifying really large functions. For functions larger than this
8812 limit after inlining, inlining is constrained by
8813 @option{--param large-function-growth}. This parameter is useful primarily
8814 to avoid extreme compilation time caused by non-linear algorithms used by the
8816 The default value is 2700.
8818 @item large-function-growth
8819 Specifies maximal growth of large function caused by inlining in percents.
8820 The default value is 100 which limits large function growth to 2.0 times
8823 @item large-unit-insns
8824 The limit specifying large translation unit. Growth caused by inlining of
8825 units larger than this limit is limited by @option{--param inline-unit-growth}.
8826 For small units this might be too tight.
8827 For example, consider a unit consisting of function A
8828 that is inline and B that just calls A three times. If B is small relative to
8829 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8830 large units consisting of small inlineable functions, however, the overall unit
8831 growth limit is needed to avoid exponential explosion of code size. Thus for
8832 smaller units, the size is increased to @option{--param large-unit-insns}
8833 before applying @option{--param inline-unit-growth}. The default is 10000.
8835 @item inline-unit-growth
8836 Specifies maximal overall growth of the compilation unit caused by inlining.
8837 The default value is 20 which limits unit growth to 1.2 times the original
8838 size. Cold functions (either marked cold via an attribute or by profile
8839 feedback) are not accounted into the unit size.
8841 @item ipcp-unit-growth
8842 Specifies maximal overall growth of the compilation unit caused by
8843 interprocedural constant propagation. The default value is 10 which limits
8844 unit growth to 1.1 times the original size.
8846 @item large-stack-frame
8847 The limit specifying large stack frames. While inlining the algorithm is trying
8848 to not grow past this limit too much. The default value is 256 bytes.
8850 @item large-stack-frame-growth
8851 Specifies maximal growth of large stack frames caused by inlining in percents.
8852 The default value is 1000 which limits large stack frame growth to 11 times
8855 @item max-inline-insns-recursive
8856 @itemx max-inline-insns-recursive-auto
8857 Specifies the maximum number of instructions an out-of-line copy of a
8858 self-recursive inline
8859 function can grow into by performing recursive inlining.
8861 @option{--param max-inline-insns-recursive} applies to functions
8863 For functions not declared inline, recursive inlining
8864 happens only when @option{-finline-functions} (included in @option{-O3}) is
8865 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
8866 default value is 450.
8868 @item max-inline-recursive-depth
8869 @itemx max-inline-recursive-depth-auto
8870 Specifies the maximum recursion depth used for recursive inlining.
8872 @option{--param max-inline-recursive-depth} applies to functions
8873 declared inline. For functions not declared inline, recursive inlining
8874 happens only when @option{-finline-functions} (included in @option{-O3}) is
8875 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
8878 @item min-inline-recursive-probability
8879 Recursive inlining is profitable only for function having deep recursion
8880 in average and can hurt for function having little recursion depth by
8881 increasing the prologue size or complexity of function body to other
8884 When profile feedback is available (see @option{-fprofile-generate}) the actual
8885 recursion depth can be guessed from probability that function recurses via a
8886 given call expression. This parameter limits inlining only to call expressions
8887 whose probability exceeds the given threshold (in percents).
8888 The default value is 10.
8890 @item early-inlining-insns
8891 Specify growth that the early inliner can make. In effect it increases
8892 the amount of inlining for code having a large abstraction penalty.
8893 The default value is 14.
8895 @item max-early-inliner-iterations
8896 Limit of iterations of the early inliner. This basically bounds
8897 the number of nested indirect calls the early inliner can resolve.
8898 Deeper chains are still handled by late inlining.
8900 @item comdat-sharing-probability
8901 Probability (in percent) that C++ inline function with comdat visibility
8902 are shared across multiple compilation units. The default value is 20.
8904 @item profile-func-internal-id
8905 A parameter to control whether to use function internal id in profile
8906 database lookup. If the value is 0, the compiler uses an id that
8907 is based on function assembler name and filename, which makes old profile
8908 data more tolerant to source changes such as function reordering etc.
8909 The default value is 0.
8911 @item min-vect-loop-bound
8912 The minimum number of iterations under which loops are not vectorized
8913 when @option{-ftree-vectorize} is used. The number of iterations after
8914 vectorization needs to be greater than the value specified by this option
8915 to allow vectorization. The default value is 0.
8917 @item gcse-cost-distance-ratio
8918 Scaling factor in calculation of maximum distance an expression
8919 can be moved by GCSE optimizations. This is currently supported only in the
8920 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8921 is with simple expressions, i.e., the expressions that have cost
8922 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
8923 hoisting of simple expressions. The default value is 10.
8925 @item gcse-unrestricted-cost
8926 Cost, roughly measured as the cost of a single typical machine
8927 instruction, at which GCSE optimizations do not constrain
8928 the distance an expression can travel. This is currently
8929 supported only in the code hoisting pass. The lesser the cost,
8930 the more aggressive code hoisting is. Specifying 0
8931 allows all expressions to travel unrestricted distances.
8932 The default value is 3.
8934 @item max-hoist-depth
8935 The depth of search in the dominator tree for expressions to hoist.
8936 This is used to avoid quadratic behavior in hoisting algorithm.
8937 The value of 0 does not limit on the search, but may slow down compilation
8938 of huge functions. The default value is 30.
8940 @item max-tail-merge-comparisons
8941 The maximum amount of similar bbs to compare a bb with. This is used to
8942 avoid quadratic behavior in tree tail merging. The default value is 10.
8944 @item max-tail-merge-iterations
8945 The maximum amount of iterations of the pass over the function. This is used to
8946 limit compilation time in tree tail merging. The default value is 2.
8948 @item max-unrolled-insns
8949 The maximum number of instructions that a loop may have to be unrolled.
8950 If a loop is unrolled, this parameter also determines how many times
8951 the loop code is unrolled.
8953 @item max-average-unrolled-insns
8954 The maximum number of instructions biased by probabilities of their execution
8955 that a loop may have to be unrolled. If a loop is unrolled,
8956 this parameter also determines how many times the loop code is unrolled.
8958 @item max-unroll-times
8959 The maximum number of unrollings of a single loop.
8961 @item max-peeled-insns
8962 The maximum number of instructions that a loop may have to be peeled.
8963 If a loop is peeled, this parameter also determines how many times
8964 the loop code is peeled.
8966 @item max-peel-times
8967 The maximum number of peelings of a single loop.
8969 @item max-peel-branches
8970 The maximum number of branches on the hot path through the peeled sequence.
8972 @item max-completely-peeled-insns
8973 The maximum number of insns of a completely peeled loop.
8975 @item max-completely-peel-times
8976 The maximum number of iterations of a loop to be suitable for complete peeling.
8978 @item max-completely-peel-loop-nest-depth
8979 The maximum depth of a loop nest suitable for complete peeling.
8981 @item max-unswitch-insns
8982 The maximum number of insns of an unswitched loop.
8984 @item max-unswitch-level
8985 The maximum number of branches unswitched in a single loop.
8988 The minimum cost of an expensive expression in the loop invariant motion.
8990 @item iv-consider-all-candidates-bound
8991 Bound on number of candidates for induction variables, below which
8992 all candidates are considered for each use in induction variable
8993 optimizations. If there are more candidates than this,
8994 only the most relevant ones are considered to avoid quadratic time complexity.
8996 @item iv-max-considered-uses
8997 The induction variable optimizations give up on loops that contain more
8998 induction variable uses.
9000 @item iv-always-prune-cand-set-bound
9001 If the number of candidates in the set is smaller than this value,
9002 always try to remove unnecessary ivs from the set
9003 when adding a new one.
9005 @item scev-max-expr-size
9006 Bound on size of expressions used in the scalar evolutions analyzer.
9007 Large expressions slow the analyzer.
9009 @item scev-max-expr-complexity
9010 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9011 Complex expressions slow the analyzer.
9013 @item vect-max-version-for-alignment-checks
9014 The maximum number of run-time checks that can be performed when
9015 doing loop versioning for alignment in the vectorizer.
9017 @item vect-max-version-for-alias-checks
9018 The maximum number of run-time checks that can be performed when
9019 doing loop versioning for alias in the vectorizer.
9021 @item vect-max-peeling-for-alignment
9022 The maximum number of loop peels to enhance access alignment
9023 for vectorizer. Value -1 means no limit.
9025 @item max-iterations-to-track
9026 The maximum number of iterations of a loop the brute-force algorithm
9027 for analysis of the number of iterations of the loop tries to evaluate.
9029 @item hot-bb-count-ws-permille
9030 A basic block profile count is considered hot if it contributes to
9031 the given permillage (i.e. 0...1000) of the entire profiled execution.
9033 @item hot-bb-frequency-fraction
9034 Select fraction of the entry block frequency of executions of basic block in
9035 function given basic block needs to have to be considered hot.
9037 @item max-predicted-iterations
9038 The maximum number of loop iterations we predict statically. This is useful
9039 in cases where a function contains a single loop with known bound and
9040 another loop with unknown bound.
9041 The known number of iterations is predicted correctly, while
9042 the unknown number of iterations average to roughly 10. This means that the
9043 loop without bounds appears artificially cold relative to the other one.
9045 @item builtin-expect-probability
9046 Control the probability of the expression having the specified value. This
9047 parameter takes a percentage (i.e. 0 ... 100) as input.
9048 The default probability of 90 is obtained empirically.
9050 @item align-threshold
9052 Select fraction of the maximal frequency of executions of a basic block in
9053 a function to align the basic block.
9055 @item align-loop-iterations
9057 A loop expected to iterate at least the selected number of iterations is
9060 @item tracer-dynamic-coverage
9061 @itemx tracer-dynamic-coverage-feedback
9063 This value is used to limit superblock formation once the given percentage of
9064 executed instructions is covered. This limits unnecessary code size
9067 The @option{tracer-dynamic-coverage-feedback} parameter
9068 is used only when profile
9069 feedback is available. The real profiles (as opposed to statically estimated
9070 ones) are much less balanced allowing the threshold to be larger value.
9072 @item tracer-max-code-growth
9073 Stop tail duplication once code growth has reached given percentage. This is
9074 a rather artificial limit, as most of the duplicates are eliminated later in
9075 cross jumping, so it may be set to much higher values than is the desired code
9078 @item tracer-min-branch-ratio
9080 Stop reverse growth when the reverse probability of best edge is less than this
9081 threshold (in percent).
9083 @item tracer-min-branch-probability
9084 @itemx tracer-min-branch-probability-feedback
9086 Stop forward growth if the best edge has probability lower than this
9089 Similarly to @option{tracer-dynamic-coverage} two parameters are
9090 provided. @option{tracer-min-branch-probability-feedback} is used for
9091 compilation with profile feedback and @option{tracer-min-branch-probability}
9092 compilation without. The value for compilation with profile feedback
9093 needs to be more conservative (higher) in order to make tracer
9096 @item max-cse-path-length
9098 The maximum number of basic blocks on path that CSE considers.
9102 The maximum number of instructions CSE processes before flushing.
9103 The default is 1000.
9105 @item ggc-min-expand
9107 GCC uses a garbage collector to manage its own memory allocation. This
9108 parameter specifies the minimum percentage by which the garbage
9109 collector's heap should be allowed to expand between collections.
9110 Tuning this may improve compilation speed; it has no effect on code
9113 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9114 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9115 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9116 GCC is not able to calculate RAM on a particular platform, the lower
9117 bound of 30% is used. Setting this parameter and
9118 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9119 every opportunity. This is extremely slow, but can be useful for
9122 @item ggc-min-heapsize
9124 Minimum size of the garbage collector's heap before it begins bothering
9125 to collect garbage. The first collection occurs after the heap expands
9126 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9127 tuning this may improve compilation speed, and has no effect on code
9130 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9131 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9132 with a lower bound of 4096 (four megabytes) and an upper bound of
9133 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9134 particular platform, the lower bound is used. Setting this parameter
9135 very large effectively disables garbage collection. Setting this
9136 parameter and @option{ggc-min-expand} to zero causes a full collection
9137 to occur at every opportunity.
9139 @item max-reload-search-insns
9140 The maximum number of instruction reload should look backward for equivalent
9141 register. Increasing values mean more aggressive optimization, making the
9142 compilation time increase with probably slightly better performance.
9143 The default value is 100.
9145 @item max-cselib-memory-locations
9146 The maximum number of memory locations cselib should take into account.
9147 Increasing values mean more aggressive optimization, making the compilation time
9148 increase with probably slightly better performance. The default value is 500.
9150 @item max-sched-ready-insns
9151 The maximum number of instructions ready to be issued the scheduler should
9152 consider at any given time during the first scheduling pass. Increasing
9153 values mean more thorough searches, making the compilation time increase
9154 with probably little benefit. The default value is 100.
9156 @item max-sched-region-blocks
9157 The maximum number of blocks in a region to be considered for
9158 interblock scheduling. The default value is 10.
9160 @item max-pipeline-region-blocks
9161 The maximum number of blocks in a region to be considered for
9162 pipelining in the selective scheduler. The default value is 15.
9164 @item max-sched-region-insns
9165 The maximum number of insns in a region to be considered for
9166 interblock scheduling. The default value is 100.
9168 @item max-pipeline-region-insns
9169 The maximum number of insns in a region to be considered for
9170 pipelining in the selective scheduler. The default value is 200.
9173 The minimum probability (in percents) of reaching a source block
9174 for interblock speculative scheduling. The default value is 40.
9176 @item max-sched-extend-regions-iters
9177 The maximum number of iterations through CFG to extend regions.
9178 A value of 0 (the default) disables region extensions.
9180 @item max-sched-insn-conflict-delay
9181 The maximum conflict delay for an insn to be considered for speculative motion.
9182 The default value is 3.
9184 @item sched-spec-prob-cutoff
9185 The minimal probability of speculation success (in percents), so that
9186 speculative insns are scheduled.
9187 The default value is 40.
9189 @item sched-state-edge-prob-cutoff
9190 The minimum probability an edge must have for the scheduler to save its
9192 The default value is 10.
9194 @item sched-mem-true-dep-cost
9195 Minimal distance (in CPU cycles) between store and load targeting same
9196 memory locations. The default value is 1.
9198 @item selsched-max-lookahead
9199 The maximum size of the lookahead window of selective scheduling. It is a
9200 depth of search for available instructions.
9201 The default value is 50.
9203 @item selsched-max-sched-times
9204 The maximum number of times that an instruction is scheduled during
9205 selective scheduling. This is the limit on the number of iterations
9206 through which the instruction may be pipelined. The default value is 2.
9208 @item selsched-insns-to-rename
9209 The maximum number of best instructions in the ready list that are considered
9210 for renaming in the selective scheduler. The default value is 2.
9213 The minimum value of stage count that swing modulo scheduler
9214 generates. The default value is 2.
9216 @item max-last-value-rtl
9217 The maximum size measured as number of RTLs that can be recorded in an expression
9218 in combiner for a pseudo register as last known value of that register. The default
9221 @item max-combine-insns
9222 The maximum number of instructions the RTL combiner tries to combine.
9223 The default value is 2 at @option{-Og} and 4 otherwise.
9225 @item integer-share-limit
9226 Small integer constants can use a shared data structure, reducing the
9227 compiler's memory usage and increasing its speed. This sets the maximum
9228 value of a shared integer constant. The default value is 256.
9230 @item ssp-buffer-size
9231 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
9232 protection when @option{-fstack-protection} is used.
9234 @item min-size-for-stack-sharing
9235 The minimum size of variables taking part in stack slot sharing when not
9236 optimizing. The default value is 32.
9238 @item max-jump-thread-duplication-stmts
9239 Maximum number of statements allowed in a block that needs to be
9240 duplicated when threading jumps.
9242 @item max-fields-for-field-sensitive
9243 Maximum number of fields in a structure treated in
9244 a field sensitive manner during pointer analysis. The default is zero
9245 for @option{-O0} and @option{-O1},
9246 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
9248 @item prefetch-latency
9249 Estimate on average number of instructions that are executed before
9250 prefetch finishes. The distance prefetched ahead is proportional
9251 to this constant. Increasing this number may also lead to less
9252 streams being prefetched (see @option{simultaneous-prefetches}).
9254 @item simultaneous-prefetches
9255 Maximum number of prefetches that can run at the same time.
9257 @item l1-cache-line-size
9258 The size of cache line in L1 cache, in bytes.
9261 The size of L1 cache, in kilobytes.
9264 The size of L2 cache, in kilobytes.
9266 @item min-insn-to-prefetch-ratio
9267 The minimum ratio between the number of instructions and the
9268 number of prefetches to enable prefetching in a loop.
9270 @item prefetch-min-insn-to-mem-ratio
9271 The minimum ratio between the number of instructions and the
9272 number of memory references to enable prefetching in a loop.
9274 @item use-canonical-types
9275 Whether the compiler should use the ``canonical'' type system. By
9276 default, this should always be 1, which uses a more efficient internal
9277 mechanism for comparing types in C++ and Objective-C++. However, if
9278 bugs in the canonical type system are causing compilation failures,
9279 set this value to 0 to disable canonical types.
9281 @item switch-conversion-max-branch-ratio
9282 Switch initialization conversion refuses to create arrays that are
9283 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9284 branches in the switch.
9286 @item max-partial-antic-length
9287 Maximum length of the partial antic set computed during the tree
9288 partial redundancy elimination optimization (@option{-ftree-pre}) when
9289 optimizing at @option{-O3} and above. For some sorts of source code
9290 the enhanced partial redundancy elimination optimization can run away,
9291 consuming all of the memory available on the host machine. This
9292 parameter sets a limit on the length of the sets that are computed,
9293 which prevents the runaway behavior. Setting a value of 0 for
9294 this parameter allows an unlimited set length.
9296 @item sccvn-max-scc-size
9297 Maximum size of a strongly connected component (SCC) during SCCVN
9298 processing. If this limit is hit, SCCVN processing for the whole
9299 function is not done and optimizations depending on it are
9300 disabled. The default maximum SCC size is 10000.
9302 @item sccvn-max-alias-queries-per-access
9303 Maximum number of alias-oracle queries we perform when looking for
9304 redundancies for loads and stores. If this limit is hit the search
9305 is aborted and the load or store is not considered redundant. The
9306 number of queries is algorithmically limited to the number of
9307 stores on all paths from the load to the function entry.
9308 The default maximum number of queries is 1000.
9310 @item ira-max-loops-num
9311 IRA uses regional register allocation by default. If a function
9312 contains more loops than the number given by this parameter, only at most
9313 the given number of the most frequently-executed loops form regions
9314 for regional register allocation. The default value of the
9317 @item ira-max-conflict-table-size
9318 Although IRA uses a sophisticated algorithm to compress the conflict
9319 table, the table can still require excessive amounts of memory for
9320 huge functions. If the conflict table for a function could be more
9321 than the size in MB given by this parameter, the register allocator
9322 instead uses a faster, simpler, and lower-quality
9323 algorithm that does not require building a pseudo-register conflict table.
9324 The default value of the parameter is 2000.
9326 @item ira-loop-reserved-regs
9327 IRA can be used to evaluate more accurate register pressure in loops
9328 for decisions to move loop invariants (see @option{-O3}). The number
9329 of available registers reserved for some other purposes is given
9330 by this parameter. The default value of the parameter is 2, which is
9331 the minimal number of registers needed by typical instructions.
9332 This value is the best found from numerous experiments.
9334 @item lra-inheritance-ebb-probability-cutoff
9335 LRA tries to reuse values reloaded in registers in subsequent insns.
9336 This optimization is called inheritance. EBB is used as a region to
9337 do this optimization. The parameter defines a minimal fall-through
9338 edge probability in percentage used to add BB to inheritance EBB in
9339 LRA. The default value of the parameter is 40. The value was chosen
9340 from numerous runs of SPEC2000 on x86-64.
9342 @item loop-invariant-max-bbs-in-loop
9343 Loop invariant motion can be very expensive, both in compilation time and
9344 in amount of needed compile-time memory, with very large loops. Loops
9345 with more basic blocks than this parameter won't have loop invariant
9346 motion optimization performed on them. The default value of the
9347 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
9349 @item loop-max-datarefs-for-datadeps
9350 Building data dependencies is expensive for very large loops. This
9351 parameter limits the number of data references in loops that are
9352 considered for data dependence analysis. These large loops are no
9353 handled by the optimizations using loop data dependencies.
9354 The default value is 1000.
9356 @item max-vartrack-size
9357 Sets a maximum number of hash table slots to use during variable
9358 tracking dataflow analysis of any function. If this limit is exceeded
9359 with variable tracking at assignments enabled, analysis for that
9360 function is retried without it, after removing all debug insns from
9361 the function. If the limit is exceeded even without debug insns, var
9362 tracking analysis is completely disabled for the function. Setting
9363 the parameter to zero makes it unlimited.
9365 @item max-vartrack-expr-depth
9366 Sets a maximum number of recursion levels when attempting to map
9367 variable names or debug temporaries to value expressions. This trades
9368 compilation time for more complete debug information. If this is set too
9369 low, value expressions that are available and could be represented in
9370 debug information may end up not being used; setting this higher may
9371 enable the compiler to find more complex debug expressions, but compile
9372 time and memory use may grow. The default is 12.
9374 @item min-nondebug-insn-uid
9375 Use uids starting at this parameter for nondebug insns. The range below
9376 the parameter is reserved exclusively for debug insns created by
9377 @option{-fvar-tracking-assignments}, but debug insns may get
9378 (non-overlapping) uids above it if the reserved range is exhausted.
9380 @item ipa-sra-ptr-growth-factor
9381 IPA-SRA replaces a pointer to an aggregate with one or more new
9382 parameters only when their cumulative size is less or equal to
9383 @option{ipa-sra-ptr-growth-factor} times the size of the original
9386 @item sra-max-scalarization-size-Ospeed
9387 @item sra-max-scalarization-size-Osize
9388 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
9389 replace scalar parts of aggregates with uses of independent scalar
9390 variables. These parameters control the maximum size, in storage units,
9391 of aggregate which is considered for replacement when compiling for
9393 (@option{sra-max-scalarization-size-Ospeed}) or size
9394 (@option{sra-max-scalarization-size-Osize}) respectively.
9396 @item tm-max-aggregate-size
9397 When making copies of thread-local variables in a transaction, this
9398 parameter specifies the size in bytes after which variables are
9399 saved with the logging functions as opposed to save/restore code
9400 sequence pairs. This option only applies when using
9403 @item graphite-max-nb-scop-params
9404 To avoid exponential effects in the Graphite loop transforms, the
9405 number of parameters in a Static Control Part (SCoP) is bounded. The
9406 default value is 10 parameters. A variable whose value is unknown at
9407 compilation time and defined outside a SCoP is a parameter of the SCoP.
9409 @item graphite-max-bbs-per-function
9410 To avoid exponential effects in the detection of SCoPs, the size of
9411 the functions analyzed by Graphite is bounded. The default value is
9414 @item loop-block-tile-size
9415 Loop blocking or strip mining transforms, enabled with
9416 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9417 loop in the loop nest by a given number of iterations. The strip
9418 length can be changed using the @option{loop-block-tile-size}
9419 parameter. The default value is 51 iterations.
9421 @item loop-unroll-jam-size
9422 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
9425 @item loop-unroll-jam-depth
9426 Specify the dimension to be unrolled (counting from the most inner loop)
9427 for the @option{-floop-unroll-and-jam}. The default value is 2.
9429 @item ipa-cp-value-list-size
9430 IPA-CP attempts to track all possible values and types passed to a function's
9431 parameter in order to propagate them and perform devirtualization.
9432 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9433 stores per one formal parameter of a function.
9435 @item ipa-cp-eval-threshold
9436 IPA-CP calculates its own score of cloning profitability heuristics
9437 and performs those cloning opportunities with scores that exceed
9438 @option{ipa-cp-eval-threshold}.
9440 @item ipa-cp-recursion-penalty
9441 Percentage penalty the recursive functions will receive when they
9442 are evaluated for cloning.
9444 @item ipa-cp-single-call-penalty
9445 Percentage penalty functions containg a single call to another
9446 function will receive when they are evaluated for cloning.
9449 @item ipa-max-agg-items
9450 IPA-CP is also capable to propagate a number of scalar values passed
9451 in an aggregate. @option{ipa-max-agg-items} controls the maximum
9452 number of such values per one parameter.
9454 @item ipa-cp-loop-hint-bonus
9455 When IPA-CP determines that a cloning candidate would make the number
9456 of iterations of a loop known, it adds a bonus of
9457 @option{ipa-cp-loop-hint-bonus} to the profitability score of
9460 @item ipa-cp-array-index-hint-bonus
9461 When IPA-CP determines that a cloning candidate would make the index of
9462 an array access known, it adds a bonus of
9463 @option{ipa-cp-array-index-hint-bonus} to the profitability
9464 score of the candidate.
9466 @item ipa-max-aa-steps
9467 During its analysis of function bodies, IPA-CP employs alias analysis
9468 in order to track values pointed to by function parameters. In order
9469 not spend too much time analyzing huge functions, it gives up and
9470 consider all memory clobbered after examining
9471 @option{ipa-max-aa-steps} statements modifying memory.
9473 @item lto-partitions
9474 Specify desired number of partitions produced during WHOPR compilation.
9475 The number of partitions should exceed the number of CPUs used for compilation.
9476 The default value is 32.
9478 @item lto-min-partition
9479 Size of minimal partition for WHOPR (in estimated instructions).
9480 This prevents expenses of splitting very small programs into too many
9483 @item lto-max-partition
9484 Size of max partition for WHOPR (in estimated instructions).
9485 to provide an upper bound for individual size of partition.
9486 Meant to be used only with balanced partitioning.
9488 @item cxx-max-namespaces-for-diagnostic-help
9489 The maximum number of namespaces to consult for suggestions when C++
9490 name lookup fails for an identifier. The default is 1000.
9492 @item sink-frequency-threshold
9493 The maximum relative execution frequency (in percents) of the target block
9494 relative to a statement's original block to allow statement sinking of a
9495 statement. Larger numbers result in more aggressive statement sinking.
9496 The default value is 75. A small positive adjustment is applied for
9497 statements with memory operands as those are even more profitable so sink.
9499 @item max-stores-to-sink
9500 The maximum number of conditional store pairs that can be sunk. Set to 0
9501 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9502 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9504 @item allow-store-data-races
9505 Allow optimizers to introduce new data races on stores.
9506 Set to 1 to allow, otherwise to 0. This option is enabled by default
9507 at optimization level @option{-Ofast}.
9509 @item case-values-threshold
9510 The smallest number of different values for which it is best to use a
9511 jump-table instead of a tree of conditional branches. If the value is
9512 0, use the default for the machine. The default is 0.
9514 @item tree-reassoc-width
9515 Set the maximum number of instructions executed in parallel in
9516 reassociated tree. This parameter overrides target dependent
9517 heuristics used by default if has non zero value.
9519 @item sched-pressure-algorithm
9520 Choose between the two available implementations of
9521 @option{-fsched-pressure}. Algorithm 1 is the original implementation
9522 and is the more likely to prevent instructions from being reordered.
9523 Algorithm 2 was designed to be a compromise between the relatively
9524 conservative approach taken by algorithm 1 and the rather aggressive
9525 approach taken by the default scheduler. It relies more heavily on
9526 having a regular register file and accurate register pressure classes.
9527 See @file{haifa-sched.c} in the GCC sources for more details.
9529 The default choice depends on the target.
9531 @item max-slsr-cand-scan
9532 Set the maximum number of existing candidates that are considered when
9533 seeking a basis for a new straight-line strength reduction candidate.
9536 Enable buffer overflow detection for global objects. This kind
9537 of protection is enabled by default if you are using
9538 @option{-fsanitize=address} option.
9539 To disable global objects protection use @option{--param asan-globals=0}.
9542 Enable buffer overflow detection for stack objects. This kind of
9543 protection is enabled by default when using @option{-fsanitize=address}.
9544 To disable stack protection use @option{--param asan-stack=0} option.
9546 @item asan-instrument-reads
9547 Enable buffer overflow detection for memory reads. This kind of
9548 protection is enabled by default when using @option{-fsanitize=address}.
9549 To disable memory reads protection use
9550 @option{--param asan-instrument-reads=0}.
9552 @item asan-instrument-writes
9553 Enable buffer overflow detection for memory writes. This kind of
9554 protection is enabled by default when using @option{-fsanitize=address}.
9555 To disable memory writes protection use
9556 @option{--param asan-instrument-writes=0} option.
9558 @item asan-memintrin
9559 Enable detection for built-in functions. This kind of protection
9560 is enabled by default when using @option{-fsanitize=address}.
9561 To disable built-in functions protection use
9562 @option{--param asan-memintrin=0}.
9564 @item asan-use-after-return
9565 Enable detection of use-after-return. This kind of protection
9566 is enabled by default when using @option{-fsanitize=address} option.
9567 To disable use-after-return detection use
9568 @option{--param asan-use-after-return=0}.
9570 @item asan-instrumentation-with-call-threshold
9571 If number of memory accesses in function being instrumented
9572 is greater or equal to this number, use callbacks instead of inline checks.
9573 E.g. to disable inline code use
9574 @option{--param asan-instrumentation-with-call-threshold=0}.
9576 @item chkp-max-ctor-size
9577 Static constructors generated by Pointer Bounds Checker may become very
9578 large and significantly increase compile time at optimization level
9579 @option{-O1} and higher. This parameter is a maximum nubmer of statements
9580 in a single generated constructor. Default value is 5000.
9582 @item max-fsm-thread-path-insns
9583 Maximum number of instructions to copy when duplicating blocks on a
9584 finite state automaton jump thread path. The default is 100.
9586 @item max-fsm-thread-length
9587 Maximum number of basic blocks on a finite state automaton jump thread
9588 path. The default is 10.
9590 @item max-fsm-thread-paths
9591 Maximum number of new jump thread paths to create for a finite state
9592 automaton. The default is 50.
9594 @item parloops-chunk-size
9595 Chunk size of omp schedule for loops parallelized by parloops. The default
9598 @item parloops-schedule
9599 Schedule type of omp schedule for loops parallelized by parloops (static,
9600 dynamic, guided, auto, runtime). The default is static.
9602 @item max-ssa-name-query-depth
9603 Maximum depth of recursion when querying properties of SSA names in things
9604 like fold routines. One level of recursion corresponds to following a
9607 @item hsa-gen-debug-stores
9608 Enable emission of special debug stores within HSA kernels which are
9609 then read and reported by libgomp plugin. Generation of these stores
9610 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
9613 @item max-speculative-devirt-maydefs
9614 The maximum number of may-defs we analyze when looking for a must-def
9615 specifying the dynamic type of an object that invokes a virtual call
9616 we may be able to devirtualize speculatively.
9620 @node Instrumentation Options
9621 @section Program Instrumentation Options
9622 @cindex instrumentation options
9623 @cindex program instrumentation options
9624 @cindex run-time error checking options
9625 @cindex profiling options
9626 @cindex options, program instrumentation
9627 @cindex options, run-time error checking
9628 @cindex options, profiling
9630 GCC supports a number of command-line options that control adding
9631 run-time instrumentation to the code it normally generates.
9632 For example, one purpose of instrumentation is collect profiling
9633 statistics for use in finding program hot spots, code coverage
9634 analysis, or profile-guided optimizations.
9635 Another class of program instrumentation is adding run-time checking
9636 to detect programming errors like invalid pointer
9637 dereferences or out-of-bounds array accesses, as well as deliberately
9638 hostile attacks such as stack smashing or C++ vtable hijacking.
9639 There is also a general hook which can be used to implement other
9640 forms of tracing or function-level instrumentation for debug or
9641 program analysis purposes.
9644 @cindex @command{prof}
9647 Generate extra code to write profile information suitable for the
9648 analysis program @command{prof}. You must use this option when compiling
9649 the source files you want data about, and you must also use it when
9652 @cindex @command{gprof}
9655 Generate extra code to write profile information suitable for the
9656 analysis program @command{gprof}. You must use this option when compiling
9657 the source files you want data about, and you must also use it when
9660 @item -fprofile-arcs
9661 @opindex fprofile-arcs
9662 Add code so that program flow @dfn{arcs} are instrumented. During
9663 execution the program records how many times each branch and call is
9664 executed and how many times it is taken or returns. When the compiled
9665 program exits it saves this data to a file called
9666 @file{@var{auxname}.gcda} for each source file. The data may be used for
9667 profile-directed optimizations (@option{-fbranch-probabilities}), or for
9668 test coverage analysis (@option{-ftest-coverage}). Each object file's
9669 @var{auxname} is generated from the name of the output file, if
9670 explicitly specified and it is not the final executable, otherwise it is
9671 the basename of the source file. In both cases any suffix is removed
9672 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
9673 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
9674 @xref{Cross-profiling}.
9676 @cindex @command{gcov}
9680 This option is used to compile and link code instrumented for coverage
9681 analysis. The option is a synonym for @option{-fprofile-arcs}
9682 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
9683 linking). See the documentation for those options for more details.
9688 Compile the source files with @option{-fprofile-arcs} plus optimization
9689 and code generation options. For test coverage analysis, use the
9690 additional @option{-ftest-coverage} option. You do not need to profile
9691 every source file in a program.
9694 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
9695 (the latter implies the former).
9698 Run the program on a representative workload to generate the arc profile
9699 information. This may be repeated any number of times. You can run
9700 concurrent instances of your program, and provided that the file system
9701 supports locking, the data files will be correctly updated. Also
9702 @code{fork} calls are detected and correctly handled (double counting
9706 For profile-directed optimizations, compile the source files again with
9707 the same optimization and code generation options plus
9708 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
9709 Control Optimization}).
9712 For test coverage analysis, use @command{gcov} to produce human readable
9713 information from the @file{.gcno} and @file{.gcda} files. Refer to the
9714 @command{gcov} documentation for further information.
9718 With @option{-fprofile-arcs}, for each function of your program GCC
9719 creates a program flow graph, then finds a spanning tree for the graph.
9720 Only arcs that are not on the spanning tree have to be instrumented: the
9721 compiler adds code to count the number of times that these arcs are
9722 executed. When an arc is the only exit or only entrance to a block, the
9723 instrumentation code can be added to the block; otherwise, a new basic
9724 block must be created to hold the instrumentation code.
9727 @item -ftest-coverage
9728 @opindex ftest-coverage
9729 Produce a notes file that the @command{gcov} code-coverage utility
9730 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
9731 show program coverage. Each source file's note file is called
9732 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
9733 above for a description of @var{auxname} and instructions on how to
9734 generate test coverage data. Coverage data matches the source files
9735 more closely if you do not optimize.
9737 @item -fprofile-dir=@var{path}
9738 @opindex fprofile-dir
9740 Set the directory to search for the profile data files in to @var{path}.
9741 This option affects only the profile data generated by
9742 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9743 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9744 and its related options. Both absolute and relative paths can be used.
9745 By default, GCC uses the current directory as @var{path}, thus the
9746 profile data file appears in the same directory as the object file.
9748 @item -fprofile-generate
9749 @itemx -fprofile-generate=@var{path}
9750 @opindex fprofile-generate
9752 Enable options usually used for instrumenting application to produce
9753 profile useful for later recompilation with profile feedback based
9754 optimization. You must use @option{-fprofile-generate} both when
9755 compiling and when linking your program.
9757 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9759 If @var{path} is specified, GCC looks at the @var{path} to find
9760 the profile feedback data files. See @option{-fprofile-dir}.
9762 To optimize the program based on the collected profile information, use
9763 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
9765 @item -fsanitize=address
9766 @opindex fsanitize=address
9767 Enable AddressSanitizer, a fast memory error detector.
9768 Memory access instructions are instrumented to detect
9769 out-of-bounds and use-after-free bugs.
9770 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
9771 more details. The run-time behavior can be influenced using the
9772 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
9773 the available options are shown at startup of the instrumented program. See
9774 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
9775 for a list of supported options.
9777 @item -fsanitize=kernel-address
9778 @opindex fsanitize=kernel-address
9779 Enable AddressSanitizer for Linux kernel.
9780 See @uref{https://github.com/google/kasan/wiki} for more details.
9782 @item -fsanitize=thread
9783 @opindex fsanitize=thread
9784 Enable ThreadSanitizer, a fast data race detector.
9785 Memory access instructions are instrumented to detect
9786 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
9787 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
9788 environment variable; see
9789 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
9792 @item -fsanitize=leak
9793 @opindex fsanitize=leak
9794 Enable LeakSanitizer, a memory leak detector.
9795 This option only matters for linking of executables and if neither
9796 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
9797 case the executable is linked against a library that overrides @code{malloc}
9798 and other allocator functions. See
9799 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
9800 details. The run-time behavior can be influenced using the
9801 @env{LSAN_OPTIONS} environment variable.
9803 @item -fsanitize=undefined
9804 @opindex fsanitize=undefined
9805 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
9806 Various computations are instrumented to detect undefined behavior
9807 at runtime. Current suboptions are:
9811 @item -fsanitize=shift
9812 @opindex fsanitize=shift
9813 This option enables checking that the result of a shift operation is
9814 not undefined. Note that what exactly is considered undefined differs
9815 slightly between C and C++, as well as between ISO C90 and C99, etc.
9817 @item -fsanitize=integer-divide-by-zero
9818 @opindex fsanitize=integer-divide-by-zero
9819 Detect integer division by zero as well as @code{INT_MIN / -1} division.
9821 @item -fsanitize=unreachable
9822 @opindex fsanitize=unreachable
9823 With this option, the compiler turns the @code{__builtin_unreachable}
9824 call into a diagnostics message call instead. When reaching the
9825 @code{__builtin_unreachable} call, the behavior is undefined.
9827 @item -fsanitize=vla-bound
9828 @opindex fsanitize=vla-bound
9829 This option instructs the compiler to check that the size of a variable
9830 length array is positive.
9832 @item -fsanitize=null
9833 @opindex fsanitize=null
9834 This option enables pointer checking. Particularly, the application
9835 built with this option turned on will issue an error message when it
9836 tries to dereference a NULL pointer, or if a reference (possibly an
9837 rvalue reference) is bound to a NULL pointer, or if a method is invoked
9838 on an object pointed by a NULL pointer.
9840 @item -fsanitize=return
9841 @opindex fsanitize=return
9842 This option enables return statement checking. Programs
9843 built with this option turned on will issue an error message
9844 when the end of a non-void function is reached without actually
9845 returning a value. This option works in C++ only.
9847 @item -fsanitize=signed-integer-overflow
9848 @opindex fsanitize=signed-integer-overflow
9849 This option enables signed integer overflow checking. We check that
9850 the result of @code{+}, @code{*}, and both unary and binary @code{-}
9851 does not overflow in the signed arithmetics. Note, integer promotion
9852 rules must be taken into account. That is, the following is not an
9855 signed char a = SCHAR_MAX;
9859 @item -fsanitize=bounds
9860 @opindex fsanitize=bounds
9861 This option enables instrumentation of array bounds. Various out of bounds
9862 accesses are detected. Flexible array members, flexible array member-like
9863 arrays, and initializers of variables with static storage are not instrumented.
9865 @item -fsanitize=bounds-strict
9866 @opindex fsanitize=bounds-strict
9867 This option enables strict instrumentation of array bounds. Most out of bounds
9868 accesses are detected, including flexible array members and flexible array
9869 member-like arrays. Initializers of variables with static storage are not
9872 @item -fsanitize=alignment
9873 @opindex fsanitize=alignment
9875 This option enables checking of alignment of pointers when they are
9876 dereferenced, or when a reference is bound to insufficiently aligned target,
9877 or when a method or constructor is invoked on insufficiently aligned object.
9879 @item -fsanitize=object-size
9880 @opindex fsanitize=object-size
9881 This option enables instrumentation of memory references using the
9882 @code{__builtin_object_size} function. Various out of bounds pointer
9883 accesses are detected.
9885 @item -fsanitize=float-divide-by-zero
9886 @opindex fsanitize=float-divide-by-zero
9887 Detect floating-point division by zero. Unlike other similar options,
9888 @option{-fsanitize=float-divide-by-zero} is not enabled by
9889 @option{-fsanitize=undefined}, since floating-point division by zero can
9890 be a legitimate way of obtaining infinities and NaNs.
9892 @item -fsanitize=float-cast-overflow
9893 @opindex fsanitize=float-cast-overflow
9894 This option enables floating-point type to integer conversion checking.
9895 We check that the result of the conversion does not overflow.
9896 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
9897 not enabled by @option{-fsanitize=undefined}.
9898 This option does not work well with @code{FE_INVALID} exceptions enabled.
9900 @item -fsanitize=nonnull-attribute
9901 @opindex fsanitize=nonnull-attribute
9903 This option enables instrumentation of calls, checking whether null values
9904 are not passed to arguments marked as requiring a non-null value by the
9905 @code{nonnull} function attribute.
9907 @item -fsanitize=returns-nonnull-attribute
9908 @opindex fsanitize=returns-nonnull-attribute
9910 This option enables instrumentation of return statements in functions
9911 marked with @code{returns_nonnull} function attribute, to detect returning
9912 of null values from such functions.
9914 @item -fsanitize=bool
9915 @opindex fsanitize=bool
9917 This option enables instrumentation of loads from bool. If a value other
9918 than 0/1 is loaded, a run-time error is issued.
9920 @item -fsanitize=enum
9921 @opindex fsanitize=enum
9923 This option enables instrumentation of loads from an enum type. If
9924 a value outside the range of values for the enum type is loaded,
9925 a run-time error is issued.
9927 @item -fsanitize=vptr
9928 @opindex fsanitize=vptr
9930 This option enables instrumentation of C++ member function calls, member
9931 accesses and some conversions between pointers to base and derived classes,
9932 to verify the referenced object has the correct dynamic type.
9936 While @option{-ftrapv} causes traps for signed overflows to be emitted,
9937 @option{-fsanitize=undefined} gives a diagnostic message.
9938 This currently works only for the C family of languages.
9940 @item -fno-sanitize=all
9941 @opindex fno-sanitize=all
9943 This option disables all previously enabled sanitizers.
9944 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
9947 @item -fasan-shadow-offset=@var{number}
9948 @opindex fasan-shadow-offset
9949 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
9950 It is useful for experimenting with different shadow memory layouts in
9951 Kernel AddressSanitizer.
9953 @item -fsanitize-sections=@var{s1},@var{s2},...
9954 @opindex fsanitize-sections
9955 Sanitize global variables in selected user-defined sections. @var{si} may
9958 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
9959 @opindex fsanitize-recover
9960 @opindex fno-sanitize-recover
9961 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
9962 mentioned in comma-separated list of @var{opts}. Enabling this option
9963 for a sanitizer component causes it to attempt to continue
9964 running the program as if no error happened. This means multiple
9965 runtime errors can be reported in a single program run, and the exit
9966 code of the program may indicate success even when errors
9967 have been reported. The @option{-fno-sanitize-recover=} option
9968 can be used to alter
9969 this behavior: only the first detected error is reported
9970 and program then exits with a non-zero exit code.
9972 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
9973 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
9974 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
9975 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
9976 For these sanitizers error recovery is turned on by default, except @option{-fsanitize=address},
9977 for which this feature is experimental.
9978 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
9979 accepted, the former enables recovery for all sanitizers that support it,
9980 the latter disables recovery for all sanitizers that support it.
9982 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
9984 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
9987 Similarly @option{-fno-sanitize-recover} is equivalent to
9989 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
9992 @item -fsanitize-undefined-trap-on-error
9993 @opindex fsanitize-undefined-trap-on-error
9994 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
9995 report undefined behavior using @code{__builtin_trap} rather than
9996 a @code{libubsan} library routine. The advantage of this is that the
9997 @code{libubsan} library is not needed and is not linked in, so this
9998 is usable even in freestanding environments.
10000 @item -fsanitize-coverage=trace-pc
10001 @opindex fsanitize-coverage=trace-pc
10002 Enable coverage-guided fuzzing code instrumentation.
10003 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
10005 @item -fbounds-check
10006 @opindex fbounds-check
10007 For front ends that support it, generate additional code to check that
10008 indices used to access arrays are within the declared range. This is
10009 currently only supported by the Java and Fortran front ends, where
10010 this option defaults to true and false respectively.
10012 @item -fcheck-pointer-bounds
10013 @opindex fcheck-pointer-bounds
10014 @opindex fno-check-pointer-bounds
10015 @cindex Pointer Bounds Checker options
10016 Enable Pointer Bounds Checker instrumentation. Each memory reference
10017 is instrumented with checks of the pointer used for memory access against
10018 bounds associated with that pointer.
10021 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
10022 and @option{-mmpx} are required to enable this feature.
10023 MPX-based instrumentation requires
10024 a runtime library to enable MPX in hardware and handle bounds
10025 violation signals. By default when @option{-fcheck-pointer-bounds}
10026 and @option{-mmpx} options are used to link a program, the GCC driver
10027 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
10028 Bounds checking on calls to dynamic libraries requires a linker
10029 with @option{-z bndplt} support; if GCC was configured with a linker
10030 without support for this option (including the Gold linker and older
10031 versions of ld), a warning is given if you link with @option{-mmpx}
10032 without also specifying @option{-static}, since the overall effectiveness
10033 of the bounds checking protection is reduced.
10034 See also @option{-static-libmpxwrappers}.
10036 MPX-based instrumentation
10037 may be used for debugging and also may be included in production code
10038 to increase program security. Depending on usage, you may
10039 have different requirements for the runtime library. The current version
10040 of the MPX runtime library is more oriented for use as a debugging
10041 tool. MPX runtime library usage implies @option{-lpthread}. See
10042 also @option{-static-libmpx}. The runtime library behavior can be
10043 influenced using various @env{CHKP_RT_*} environment variables. See
10044 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
10047 Generated instrumentation may be controlled by various
10048 @option{-fchkp-*} options and by the @code{bnd_variable_size}
10049 structure field attribute (@pxref{Type Attributes}) and
10050 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
10051 (@pxref{Function Attributes}). GCC also provides a number of built-in
10052 functions for controlling the Pointer Bounds Checker. @xref{Pointer
10053 Bounds Checker builtins}, for more information.
10055 @item -fchkp-check-incomplete-type
10056 @opindex fchkp-check-incomplete-type
10057 @opindex fno-chkp-check-incomplete-type
10058 Generate pointer bounds checks for variables with incomplete type.
10059 Enabled by default.
10061 @item -fchkp-narrow-bounds
10062 @opindex fchkp-narrow-bounds
10063 @opindex fno-chkp-narrow-bounds
10064 Controls bounds used by Pointer Bounds Checker for pointers to object
10065 fields. If narrowing is enabled then field bounds are used. Otherwise
10066 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
10067 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
10069 @item -fchkp-first-field-has-own-bounds
10070 @opindex fchkp-first-field-has-own-bounds
10071 @opindex fno-chkp-first-field-has-own-bounds
10072 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
10073 first field in the structure. By default a pointer to the first field has
10074 the same bounds as a pointer to the whole structure.
10076 @item -fchkp-narrow-to-innermost-array
10077 @opindex fchkp-narrow-to-innermost-array
10078 @opindex fno-chkp-narrow-to-innermost-array
10079 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
10080 case of nested static array access. By default this option is disabled and
10081 bounds of the outermost array are used.
10083 @item -fchkp-optimize
10084 @opindex fchkp-optimize
10085 @opindex fno-chkp-optimize
10086 Enables Pointer Bounds Checker optimizations. Enabled by default at
10087 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
10089 @item -fchkp-use-fast-string-functions
10090 @opindex fchkp-use-fast-string-functions
10091 @opindex fno-chkp-use-fast-string-functions
10092 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
10093 by Pointer Bounds Checker. Disabled by default.
10095 @item -fchkp-use-nochk-string-functions
10096 @opindex fchkp-use-nochk-string-functions
10097 @opindex fno-chkp-use-nochk-string-functions
10098 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
10099 by Pointer Bounds Checker. Disabled by default.
10101 @item -fchkp-use-static-bounds
10102 @opindex fchkp-use-static-bounds
10103 @opindex fno-chkp-use-static-bounds
10104 Allow Pointer Bounds Checker to generate static bounds holding
10105 bounds of static variables. Enabled by default.
10107 @item -fchkp-use-static-const-bounds
10108 @opindex fchkp-use-static-const-bounds
10109 @opindex fno-chkp-use-static-const-bounds
10110 Use statically-initialized bounds for constant bounds instead of
10111 generating them each time they are required. By default enabled when
10112 @option{-fchkp-use-static-bounds} is enabled.
10114 @item -fchkp-treat-zero-dynamic-size-as-infinite
10115 @opindex fchkp-treat-zero-dynamic-size-as-infinite
10116 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
10117 With this option, objects with incomplete type whose
10118 dynamically-obtained size is zero are treated as having infinite size
10119 instead by Pointer Bounds
10120 Checker. This option may be helpful if a program is linked with a library
10121 missing size information for some symbols. Disabled by default.
10123 @item -fchkp-check-read
10124 @opindex fchkp-check-read
10125 @opindex fno-chkp-check-read
10126 Instructs Pointer Bounds Checker to generate checks for all read
10127 accesses to memory. Enabled by default.
10129 @item -fchkp-check-write
10130 @opindex fchkp-check-write
10131 @opindex fno-chkp-check-write
10132 Instructs Pointer Bounds Checker to generate checks for all write
10133 accesses to memory. Enabled by default.
10135 @item -fchkp-store-bounds
10136 @opindex fchkp-store-bounds
10137 @opindex fno-chkp-store-bounds
10138 Instructs Pointer Bounds Checker to generate bounds stores for
10139 pointer writes. Enabled by default.
10141 @item -fchkp-instrument-calls
10142 @opindex fchkp-instrument-calls
10143 @opindex fno-chkp-instrument-calls
10144 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
10145 Enabled by default.
10147 @item -fchkp-instrument-marked-only
10148 @opindex fchkp-instrument-marked-only
10149 @opindex fno-chkp-instrument-marked-only
10150 Instructs Pointer Bounds Checker to instrument only functions
10151 marked with the @code{bnd_instrument} attribute
10152 (@pxref{Function Attributes}). Disabled by default.
10154 @item -fchkp-use-wrappers
10155 @opindex fchkp-use-wrappers
10156 @opindex fno-chkp-use-wrappers
10157 Allows Pointer Bounds Checker to replace calls to built-in functions
10158 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
10159 is used to link a program, the GCC driver automatically links
10160 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
10161 Enabled by default.
10163 @item -fstack-protector
10164 @opindex fstack-protector
10165 Emit extra code to check for buffer overflows, such as stack smashing
10166 attacks. This is done by adding a guard variable to functions with
10167 vulnerable objects. This includes functions that call @code{alloca}, and
10168 functions with buffers larger than 8 bytes. The guards are initialized
10169 when a function is entered and then checked when the function exits.
10170 If a guard check fails, an error message is printed and the program exits.
10172 @item -fstack-protector-all
10173 @opindex fstack-protector-all
10174 Like @option{-fstack-protector} except that all functions are protected.
10176 @item -fstack-protector-strong
10177 @opindex fstack-protector-strong
10178 Like @option{-fstack-protector} but includes additional functions to
10179 be protected --- those that have local array definitions, or have
10180 references to local frame addresses.
10182 @item -fstack-protector-explicit
10183 @opindex fstack-protector-explicit
10184 Like @option{-fstack-protector} but only protects those functions which
10185 have the @code{stack_protect} attribute.
10187 @item -fstack-check
10188 @opindex fstack-check
10189 Generate code to verify that you do not go beyond the boundary of the
10190 stack. You should specify this flag if you are running in an
10191 environment with multiple threads, but you only rarely need to specify it in
10192 a single-threaded environment since stack overflow is automatically
10193 detected on nearly all systems if there is only one stack.
10195 Note that this switch does not actually cause checking to be done; the
10196 operating system or the language runtime must do that. The switch causes
10197 generation of code to ensure that they see the stack being extended.
10199 You can additionally specify a string parameter: @samp{no} means no
10200 checking, @samp{generic} means force the use of old-style checking,
10201 @samp{specific} means use the best checking method and is equivalent
10202 to bare @option{-fstack-check}.
10204 Old-style checking is a generic mechanism that requires no specific
10205 target support in the compiler but comes with the following drawbacks:
10209 Modified allocation strategy for large objects: they are always
10210 allocated dynamically if their size exceeds a fixed threshold.
10213 Fixed limit on the size of the static frame of functions: when it is
10214 topped by a particular function, stack checking is not reliable and
10215 a warning is issued by the compiler.
10218 Inefficiency: because of both the modified allocation strategy and the
10219 generic implementation, code performance is hampered.
10222 Note that old-style stack checking is also the fallback method for
10223 @samp{specific} if no target support has been added in the compiler.
10225 @item -fstack-limit-register=@var{reg}
10226 @itemx -fstack-limit-symbol=@var{sym}
10227 @itemx -fno-stack-limit
10228 @opindex fstack-limit-register
10229 @opindex fstack-limit-symbol
10230 @opindex fno-stack-limit
10231 Generate code to ensure that the stack does not grow beyond a certain value,
10232 either the value of a register or the address of a symbol. If a larger
10233 stack is required, a signal is raised at run time. For most targets,
10234 the signal is raised before the stack overruns the boundary, so
10235 it is possible to catch the signal without taking special precautions.
10237 For instance, if the stack starts at absolute address @samp{0x80000000}
10238 and grows downwards, you can use the flags
10239 @option{-fstack-limit-symbol=__stack_limit} and
10240 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
10241 of 128KB@. Note that this may only work with the GNU linker.
10243 You can locally override stack limit checking by using the
10244 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
10246 @item -fsplit-stack
10247 @opindex fsplit-stack
10248 Generate code to automatically split the stack before it overflows.
10249 The resulting program has a discontiguous stack which can only
10250 overflow if the program is unable to allocate any more memory. This
10251 is most useful when running threaded programs, as it is no longer
10252 necessary to calculate a good stack size to use for each thread. This
10253 is currently only implemented for the x86 targets running
10256 When code compiled with @option{-fsplit-stack} calls code compiled
10257 without @option{-fsplit-stack}, there may not be much stack space
10258 available for the latter code to run. If compiling all code,
10259 including library code, with @option{-fsplit-stack} is not an option,
10260 then the linker can fix up these calls so that the code compiled
10261 without @option{-fsplit-stack} always has a large stack. Support for
10262 this is implemented in the gold linker in GNU binutils release 2.21
10265 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
10266 @opindex fvtable-verify
10267 This option is only available when compiling C++ code.
10268 It turns on (or off, if using @option{-fvtable-verify=none}) the security
10269 feature that verifies at run time, for every virtual call, that
10270 the vtable pointer through which the call is made is valid for the type of
10271 the object, and has not been corrupted or overwritten. If an invalid vtable
10272 pointer is detected at run time, an error is reported and execution of the
10273 program is immediately halted.
10275 This option causes run-time data structures to be built at program startup,
10276 which are used for verifying the vtable pointers.
10277 The options @samp{std} and @samp{preinit}
10278 control the timing of when these data structures are built. In both cases the
10279 data structures are built before execution reaches @code{main}. Using
10280 @option{-fvtable-verify=std} causes the data structures to be built after
10281 shared libraries have been loaded and initialized.
10282 @option{-fvtable-verify=preinit} causes them to be built before shared
10283 libraries have been loaded and initialized.
10285 If this option appears multiple times in the command line with different
10286 values specified, @samp{none} takes highest priority over both @samp{std} and
10287 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
10290 @opindex fvtv-debug
10291 When used in conjunction with @option{-fvtable-verify=std} or
10292 @option{-fvtable-verify=preinit}, causes debug versions of the
10293 runtime functions for the vtable verification feature to be called.
10294 This flag also causes the compiler to log information about which
10295 vtable pointers it finds for each class.
10296 This information is written to a file named @file{vtv_set_ptr_data.log}
10297 in the directory named by the environment variable @env{VTV_LOGS_DIR}
10298 if that is defined or the current working directory otherwise.
10300 Note: This feature @emph{appends} data to the log file. If you want a fresh log
10301 file, be sure to delete any existing one.
10304 @opindex fvtv-counts
10305 This is a debugging flag. When used in conjunction with
10306 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
10307 causes the compiler to keep track of the total number of virtual calls
10308 it encounters and the number of verifications it inserts. It also
10309 counts the number of calls to certain run-time library functions
10310 that it inserts and logs this information for each compilation unit.
10311 The compiler writes this information to a file named
10312 @file{vtv_count_data.log} in the directory named by the environment
10313 variable @env{VTV_LOGS_DIR} if that is defined or the current working
10314 directory otherwise. It also counts the size of the vtable pointer sets
10315 for each class, and writes this information to @file{vtv_class_set_sizes.log}
10316 in the same directory.
10318 Note: This feature @emph{appends} data to the log files. To get fresh log
10319 files, be sure to delete any existing ones.
10321 @item -finstrument-functions
10322 @opindex finstrument-functions
10323 Generate instrumentation calls for entry and exit to functions. Just
10324 after function entry and just before function exit, the following
10325 profiling functions are called with the address of the current
10326 function and its call site. (On some platforms,
10327 @code{__builtin_return_address} does not work beyond the current
10328 function, so the call site information may not be available to the
10329 profiling functions otherwise.)
10332 void __cyg_profile_func_enter (void *this_fn,
10334 void __cyg_profile_func_exit (void *this_fn,
10338 The first argument is the address of the start of the current function,
10339 which may be looked up exactly in the symbol table.
10341 This instrumentation is also done for functions expanded inline in other
10342 functions. The profiling calls indicate where, conceptually, the
10343 inline function is entered and exited. This means that addressable
10344 versions of such functions must be available. If all your uses of a
10345 function are expanded inline, this may mean an additional expansion of
10346 code size. If you use @code{extern inline} in your C code, an
10347 addressable version of such functions must be provided. (This is
10348 normally the case anyway, but if you get lucky and the optimizer always
10349 expands the functions inline, you might have gotten away without
10350 providing static copies.)
10352 A function may be given the attribute @code{no_instrument_function}, in
10353 which case this instrumentation is not done. This can be used, for
10354 example, for the profiling functions listed above, high-priority
10355 interrupt routines, and any functions from which the profiling functions
10356 cannot safely be called (perhaps signal handlers, if the profiling
10357 routines generate output or allocate memory).
10359 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
10360 @opindex finstrument-functions-exclude-file-list
10362 Set the list of functions that are excluded from instrumentation (see
10363 the description of @option{-finstrument-functions}). If the file that
10364 contains a function definition matches with one of @var{file}, then
10365 that function is not instrumented. The match is done on substrings:
10366 if the @var{file} parameter is a substring of the file name, it is
10367 considered to be a match.
10372 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
10376 excludes any inline function defined in files whose pathnames
10377 contain @file{/bits/stl} or @file{include/sys}.
10379 If, for some reason, you want to include letter @samp{,} in one of
10380 @var{sym}, write @samp{\,}. For example,
10381 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
10382 (note the single quote surrounding the option).
10384 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
10385 @opindex finstrument-functions-exclude-function-list
10387 This is similar to @option{-finstrument-functions-exclude-file-list},
10388 but this option sets the list of function names to be excluded from
10389 instrumentation. The function name to be matched is its user-visible
10390 name, such as @code{vector<int> blah(const vector<int> &)}, not the
10391 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
10392 match is done on substrings: if the @var{sym} parameter is a substring
10393 of the function name, it is considered to be a match. For C99 and C++
10394 extended identifiers, the function name must be given in UTF-8, not
10395 using universal character names.
10400 @node Preprocessor Options
10401 @section Options Controlling the Preprocessor
10402 @cindex preprocessor options
10403 @cindex options, preprocessor
10405 These options control the C preprocessor, which is run on each C source
10406 file before actual compilation.
10408 If you use the @option{-E} option, nothing is done except preprocessing.
10409 Some of these options make sense only together with @option{-E} because
10410 they cause the preprocessor output to be unsuitable for actual
10414 @item -Wp,@var{option}
10416 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10417 and pass @var{option} directly through to the preprocessor. If
10418 @var{option} contains commas, it is split into multiple options at the
10419 commas. However, many options are modified, translated or interpreted
10420 by the compiler driver before being passed to the preprocessor, and
10421 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10422 interface is undocumented and subject to change, so whenever possible
10423 you should avoid using @option{-Wp} and let the driver handle the
10426 @item -Xpreprocessor @var{option}
10427 @opindex Xpreprocessor
10428 Pass @var{option} as an option to the preprocessor. You can use this to
10429 supply system-specific preprocessor options that GCC does not
10432 If you want to pass an option that takes an argument, you must use
10433 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10435 @item -no-integrated-cpp
10436 @opindex no-integrated-cpp
10437 Perform preprocessing as a separate pass before compilation.
10438 By default, GCC performs preprocessing as an integrated part of
10439 input tokenization and parsing.
10440 If this option is provided, the appropriate language front end
10441 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10442 and Objective-C, respectively) is instead invoked twice,
10443 once for preprocessing only and once for actual compilation
10444 of the preprocessed input.
10445 This option may be useful in conjunction with the @option{-B} or
10446 @option{-wrapper} options to specify an alternate preprocessor or
10447 perform additional processing of the program source between
10448 normal preprocessing and compilation.
10451 @include cppopts.texi
10453 @node Assembler Options
10454 @section Passing Options to the Assembler
10456 @c prevent bad page break with this line
10457 You can pass options to the assembler.
10460 @item -Wa,@var{option}
10462 Pass @var{option} as an option to the assembler. If @var{option}
10463 contains commas, it is split into multiple options at the commas.
10465 @item -Xassembler @var{option}
10466 @opindex Xassembler
10467 Pass @var{option} as an option to the assembler. You can use this to
10468 supply system-specific assembler options that GCC does not
10471 If you want to pass an option that takes an argument, you must use
10472 @option{-Xassembler} twice, once for the option and once for the argument.
10477 @section Options for Linking
10478 @cindex link options
10479 @cindex options, linking
10481 These options come into play when the compiler links object files into
10482 an executable output file. They are meaningless if the compiler is
10483 not doing a link step.
10487 @item @var{object-file-name}
10488 A file name that does not end in a special recognized suffix is
10489 considered to name an object file or library. (Object files are
10490 distinguished from libraries by the linker according to the file
10491 contents.) If linking is done, these object files are used as input
10500 If any of these options is used, then the linker is not run, and
10501 object file names should not be used as arguments. @xref{Overall
10505 @opindex fuse-ld=bfd
10506 Use the @command{bfd} linker instead of the default linker.
10508 @item -fuse-ld=gold
10509 @opindex fuse-ld=gold
10510 Use the @command{gold} linker instead of the default linker.
10513 @item -l@var{library}
10514 @itemx -l @var{library}
10516 Search the library named @var{library} when linking. (The second
10517 alternative with the library as a separate argument is only for
10518 POSIX compliance and is not recommended.)
10520 It makes a difference where in the command you write this option; the
10521 linker searches and processes libraries and object files in the order they
10522 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10523 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10524 to functions in @samp{z}, those functions may not be loaded.
10526 The linker searches a standard list of directories for the library,
10527 which is actually a file named @file{lib@var{library}.a}. The linker
10528 then uses this file as if it had been specified precisely by name.
10530 The directories searched include several standard system directories
10531 plus any that you specify with @option{-L}.
10533 Normally the files found this way are library files---archive files
10534 whose members are object files. The linker handles an archive file by
10535 scanning through it for members which define symbols that have so far
10536 been referenced but not defined. But if the file that is found is an
10537 ordinary object file, it is linked in the usual fashion. The only
10538 difference between using an @option{-l} option and specifying a file name
10539 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10540 and searches several directories.
10544 You need this special case of the @option{-l} option in order to
10545 link an Objective-C or Objective-C++ program.
10547 @item -nostartfiles
10548 @opindex nostartfiles
10549 Do not use the standard system startup files when linking.
10550 The standard system libraries are used normally, unless @option{-nostdlib}
10551 or @option{-nodefaultlibs} is used.
10553 @item -nodefaultlibs
10554 @opindex nodefaultlibs
10555 Do not use the standard system libraries when linking.
10556 Only the libraries you specify are passed to the linker, and options
10557 specifying linkage of the system libraries, such as @option{-static-libgcc}
10558 or @option{-shared-libgcc}, are ignored.
10559 The standard startup files are used normally, unless @option{-nostartfiles}
10562 The compiler may generate calls to @code{memcmp},
10563 @code{memset}, @code{memcpy} and @code{memmove}.
10564 These entries are usually resolved by entries in
10565 libc. These entry points should be supplied through some other
10566 mechanism when this option is specified.
10570 Do not use the standard system startup files or libraries when linking.
10571 No startup files and only the libraries you specify are passed to
10572 the linker, and options specifying linkage of the system libraries, such as
10573 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
10575 The compiler may generate calls to @code{memcmp}, @code{memset},
10576 @code{memcpy} and @code{memmove}.
10577 These entries are usually resolved by entries in
10578 libc. These entry points should be supplied through some other
10579 mechanism when this option is specified.
10581 @cindex @option{-lgcc}, use with @option{-nostdlib}
10582 @cindex @option{-nostdlib} and unresolved references
10583 @cindex unresolved references and @option{-nostdlib}
10584 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10585 @cindex @option{-nodefaultlibs} and unresolved references
10586 @cindex unresolved references and @option{-nodefaultlibs}
10587 One of the standard libraries bypassed by @option{-nostdlib} and
10588 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10589 which GCC uses to overcome shortcomings of particular machines, or special
10590 needs for some languages.
10591 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10592 Collection (GCC) Internals},
10593 for more discussion of @file{libgcc.a}.)
10594 In most cases, you need @file{libgcc.a} even when you want to avoid
10595 other standard libraries. In other words, when you specify @option{-nostdlib}
10596 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10597 This ensures that you have no unresolved references to internal GCC
10598 library subroutines.
10599 (An example of such an internal subroutine is @code{__main}, used to ensure C++
10600 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10601 GNU Compiler Collection (GCC) Internals}.)
10605 Produce a position independent executable on targets that support it.
10606 For predictable results, you must also specify the same set of options
10607 used for compilation (@option{-fpie}, @option{-fPIE},
10608 or model suboptions) when you specify this linker option.
10612 Don't produce a position independent executable.
10616 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10617 that support it. This instructs the linker to add all symbols, not
10618 only used ones, to the dynamic symbol table. This option is needed
10619 for some uses of @code{dlopen} or to allow obtaining backtraces
10620 from within a program.
10624 Remove all symbol table and relocation information from the executable.
10628 On systems that support dynamic linking, this prevents linking with the shared
10629 libraries. On other systems, this option has no effect.
10633 Produce a shared object which can then be linked with other objects to
10634 form an executable. Not all systems support this option. For predictable
10635 results, you must also specify the same set of options used for compilation
10636 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
10637 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
10638 needs to build supplementary stub code for constructors to work. On
10639 multi-libbed systems, @samp{gcc -shared} must select the correct support
10640 libraries to link against. Failing to supply the correct flags may lead
10641 to subtle defects. Supplying them in cases where they are not necessary
10644 @item -shared-libgcc
10645 @itemx -static-libgcc
10646 @opindex shared-libgcc
10647 @opindex static-libgcc
10648 On systems that provide @file{libgcc} as a shared library, these options
10649 force the use of either the shared or static version, respectively.
10650 If no shared version of @file{libgcc} was built when the compiler was
10651 configured, these options have no effect.
10653 There are several situations in which an application should use the
10654 shared @file{libgcc} instead of the static version. The most common
10655 of these is when the application wishes to throw and catch exceptions
10656 across different shared libraries. In that case, each of the libraries
10657 as well as the application itself should use the shared @file{libgcc}.
10659 Therefore, the G++ and GCJ drivers automatically add
10660 @option{-shared-libgcc} whenever you build a shared library or a main
10661 executable, because C++ and Java programs typically use exceptions, so
10662 this is the right thing to do.
10664 If, instead, you use the GCC driver to create shared libraries, you may
10665 find that they are not always linked with the shared @file{libgcc}.
10666 If GCC finds, at its configuration time, that you have a non-GNU linker
10667 or a GNU linker that does not support option @option{--eh-frame-hdr},
10668 it links the shared version of @file{libgcc} into shared libraries
10669 by default. Otherwise, it takes advantage of the linker and optimizes
10670 away the linking with the shared version of @file{libgcc}, linking with
10671 the static version of libgcc by default. This allows exceptions to
10672 propagate through such shared libraries, without incurring relocation
10673 costs at library load time.
10675 However, if a library or main executable is supposed to throw or catch
10676 exceptions, you must link it using the G++ or GCJ driver, as appropriate
10677 for the languages used in the program, or using the option
10678 @option{-shared-libgcc}, such that it is linked with the shared
10681 @item -static-libasan
10682 @opindex static-libasan
10683 When the @option{-fsanitize=address} option is used to link a program,
10684 the GCC driver automatically links against @option{libasan}. If
10685 @file{libasan} is available as a shared library, and the @option{-static}
10686 option is not used, then this links against the shared version of
10687 @file{libasan}. The @option{-static-libasan} option directs the GCC
10688 driver to link @file{libasan} statically, without necessarily linking
10689 other libraries statically.
10691 @item -static-libtsan
10692 @opindex static-libtsan
10693 When the @option{-fsanitize=thread} option is used to link a program,
10694 the GCC driver automatically links against @option{libtsan}. If
10695 @file{libtsan} is available as a shared library, and the @option{-static}
10696 option is not used, then this links against the shared version of
10697 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
10698 driver to link @file{libtsan} statically, without necessarily linking
10699 other libraries statically.
10701 @item -static-liblsan
10702 @opindex static-liblsan
10703 When the @option{-fsanitize=leak} option is used to link a program,
10704 the GCC driver automatically links against @option{liblsan}. If
10705 @file{liblsan} is available as a shared library, and the @option{-static}
10706 option is not used, then this links against the shared version of
10707 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
10708 driver to link @file{liblsan} statically, without necessarily linking
10709 other libraries statically.
10711 @item -static-libubsan
10712 @opindex static-libubsan
10713 When the @option{-fsanitize=undefined} option is used to link a program,
10714 the GCC driver automatically links against @option{libubsan}. If
10715 @file{libubsan} is available as a shared library, and the @option{-static}
10716 option is not used, then this links against the shared version of
10717 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
10718 driver to link @file{libubsan} statically, without necessarily linking
10719 other libraries statically.
10721 @item -static-libmpx
10722 @opindex static-libmpx
10723 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
10724 used to link a program, the GCC driver automatically links against
10725 @file{libmpx}. If @file{libmpx} is available as a shared library,
10726 and the @option{-static} option is not used, then this links against
10727 the shared version of @file{libmpx}. The @option{-static-libmpx}
10728 option directs the GCC driver to link @file{libmpx} statically,
10729 without necessarily linking other libraries statically.
10731 @item -static-libmpxwrappers
10732 @opindex static-libmpxwrappers
10733 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
10734 to link a program without also using @option{-fno-chkp-use-wrappers}, the
10735 GCC driver automatically links against @file{libmpxwrappers}. If
10736 @file{libmpxwrappers} is available as a shared library, and the
10737 @option{-static} option is not used, then this links against the shared
10738 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
10739 option directs the GCC driver to link @file{libmpxwrappers} statically,
10740 without necessarily linking other libraries statically.
10742 @item -static-libstdc++
10743 @opindex static-libstdc++
10744 When the @command{g++} program is used to link a C++ program, it
10745 normally automatically links against @option{libstdc++}. If
10746 @file{libstdc++} is available as a shared library, and the
10747 @option{-static} option is not used, then this links against the
10748 shared version of @file{libstdc++}. That is normally fine. However, it
10749 is sometimes useful to freeze the version of @file{libstdc++} used by
10750 the program without going all the way to a fully static link. The
10751 @option{-static-libstdc++} option directs the @command{g++} driver to
10752 link @file{libstdc++} statically, without necessarily linking other
10753 libraries statically.
10757 Bind references to global symbols when building a shared object. Warn
10758 about any unresolved references (unless overridden by the link editor
10759 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
10762 @item -T @var{script}
10764 @cindex linker script
10765 Use @var{script} as the linker script. This option is supported by most
10766 systems using the GNU linker. On some targets, such as bare-board
10767 targets without an operating system, the @option{-T} option may be required
10768 when linking to avoid references to undefined symbols.
10770 @item -Xlinker @var{option}
10772 Pass @var{option} as an option to the linker. You can use this to
10773 supply system-specific linker options that GCC does not recognize.
10775 If you want to pass an option that takes a separate argument, you must use
10776 @option{-Xlinker} twice, once for the option and once for the argument.
10777 For example, to pass @option{-assert definitions}, you must write
10778 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
10779 @option{-Xlinker "-assert definitions"}, because this passes the entire
10780 string as a single argument, which is not what the linker expects.
10782 When using the GNU linker, it is usually more convenient to pass
10783 arguments to linker options using the @option{@var{option}=@var{value}}
10784 syntax than as separate arguments. For example, you can specify
10785 @option{-Xlinker -Map=output.map} rather than
10786 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
10787 this syntax for command-line options.
10789 @item -Wl,@var{option}
10791 Pass @var{option} as an option to the linker. If @var{option} contains
10792 commas, it is split into multiple options at the commas. You can use this
10793 syntax to pass an argument to the option.
10794 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
10795 linker. When using the GNU linker, you can also get the same effect with
10796 @option{-Wl,-Map=output.map}.
10798 @item -u @var{symbol}
10800 Pretend the symbol @var{symbol} is undefined, to force linking of
10801 library modules to define it. You can use @option{-u} multiple times with
10802 different symbols to force loading of additional library modules.
10804 @item -z @var{keyword}
10806 @option{-z} is passed directly on to the linker along with the keyword
10807 @var{keyword}. See the section in the documentation of your linker for
10808 permitted values and their meanings.
10811 @node Directory Options
10812 @section Options for Directory Search
10813 @cindex directory options
10814 @cindex options, directory search
10815 @cindex search path
10817 These options specify directories to search for header files, for
10818 libraries and for parts of the compiler:
10823 Add the directory @var{dir} to the head of the list of directories to be
10824 searched for header files. This can be used to override a system header
10825 file, substituting your own version, since these directories are
10826 searched before the system header file directories. However, you should
10827 not use this option to add directories that contain vendor-supplied
10828 system header files (use @option{-isystem} for that). If you use more than
10829 one @option{-I} option, the directories are scanned in left-to-right
10830 order; the standard system directories come after.
10832 If a standard system include directory, or a directory specified with
10833 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
10834 option is ignored. The directory is still searched but as a
10835 system directory at its normal position in the system include chain.
10836 This is to ensure that GCC's procedure to fix buggy system headers and
10837 the ordering for the @code{include_next} directive are not inadvertently changed.
10838 If you really need to change the search order for system directories,
10839 use the @option{-nostdinc} and/or @option{-isystem} options.
10841 @item -iplugindir=@var{dir}
10842 @opindex iplugindir=
10843 Set the directory to search for plugins that are passed
10844 by @option{-fplugin=@var{name}} instead of
10845 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
10846 to be used by the user, but only passed by the driver.
10848 @item -iquote@var{dir}
10850 Add the directory @var{dir} to the head of the list of directories to
10851 be searched for header files only for the case of @code{#include
10852 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
10853 otherwise just like @option{-I}.
10857 Add directory @var{dir} to the list of directories to be searched
10860 @item -B@var{prefix}
10862 This option specifies where to find the executables, libraries,
10863 include files, and data files of the compiler itself.
10865 The compiler driver program runs one or more of the subprograms
10866 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
10867 @var{prefix} as a prefix for each program it tries to run, both with and
10868 without @samp{@var{machine}/@var{version}/} for the corresponding target
10869 machine and compiler version.
10871 For each subprogram to be run, the compiler driver first tries the
10872 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
10873 is not specified, the driver tries two standard prefixes,
10874 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
10875 those results in a file name that is found, the unmodified program
10876 name is searched for using the directories specified in your
10877 @env{PATH} environment variable.
10879 The compiler checks to see if the path provided by @option{-B}
10880 refers to a directory, and if necessary it adds a directory
10881 separator character at the end of the path.
10883 @option{-B} prefixes that effectively specify directory names also apply
10884 to libraries in the linker, because the compiler translates these
10885 options into @option{-L} options for the linker. They also apply to
10886 include files in the preprocessor, because the compiler translates these
10887 options into @option{-isystem} options for the preprocessor. In this case,
10888 the compiler appends @samp{include} to the prefix.
10890 The runtime support file @file{libgcc.a} can also be searched for using
10891 the @option{-B} prefix, if needed. If it is not found there, the two
10892 standard prefixes above are tried, and that is all. The file is left
10893 out of the link if it is not found by those means.
10895 Another way to specify a prefix much like the @option{-B} prefix is to use
10896 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
10899 As a special kludge, if the path provided by @option{-B} is
10900 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
10901 9, then it is replaced by @file{[dir/]include}. This is to help
10902 with boot-strapping the compiler.
10904 @item -no-canonical-prefixes
10905 @opindex no-canonical-prefixes
10906 Do not expand any symbolic links, resolve references to @samp{/../}
10907 or @samp{/./}, or make the path absolute when generating a relative
10910 @item --sysroot=@var{dir}
10912 Use @var{dir} as the logical root directory for headers and libraries.
10913 For example, if the compiler normally searches for headers in
10914 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
10915 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
10917 If you use both this option and the @option{-isysroot} option, then
10918 the @option{--sysroot} option applies to libraries, but the
10919 @option{-isysroot} option applies to header files.
10921 The GNU linker (beginning with version 2.16) has the necessary support
10922 for this option. If your linker does not support this option, the
10923 header file aspect of @option{--sysroot} still works, but the
10924 library aspect does not.
10926 @item --no-sysroot-suffix
10927 @opindex no-sysroot-suffix
10928 For some targets, a suffix is added to the root directory specified
10929 with @option{--sysroot}, depending on the other options used, so that
10930 headers may for example be found in
10931 @file{@var{dir}/@var{suffix}/usr/include} instead of
10932 @file{@var{dir}/usr/include}. This option disables the addition of
10937 This option has been deprecated. Please use @option{-iquote} instead for
10938 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
10940 Any directories you specify with @option{-I} options before the @option{-I-}
10941 option are searched only for the case of @code{#include "@var{file}"};
10942 they are not searched for @code{#include <@var{file}>}.
10944 If additional directories are specified with @option{-I} options after
10945 the @option{-I-} option, these directories are searched for all @code{#include}
10946 directives. (Ordinarily @emph{all} @option{-I} directories are used
10949 In addition, the @option{-I-} option inhibits the use of the current
10950 directory (where the current input file came from) as the first search
10951 directory for @code{#include "@var{file}"}. There is no way to
10952 override this effect of @option{-I-}. With @option{-I.} you can specify
10953 searching the directory that is current when the compiler is
10954 invoked. That is not exactly the same as what the preprocessor does
10955 by default, but it is often satisfactory.
10957 @option{-I-} does not inhibit the use of the standard system directories
10958 for header files. Thus, @option{-I-} and @option{-nostdinc} are
10962 @node Code Gen Options
10963 @section Options for Code Generation Conventions
10964 @cindex code generation conventions
10965 @cindex options, code generation
10966 @cindex run-time options
10968 These machine-independent options control the interface conventions
10969 used in code generation.
10971 Most of them have both positive and negative forms; the negative form
10972 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
10973 one of the forms is listed---the one that is not the default. You
10974 can figure out the other form by either removing @samp{no-} or adding
10978 @item -fstack-reuse=@var{reuse-level}
10979 @opindex fstack_reuse
10980 This option controls stack space reuse for user declared local/auto variables
10981 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
10982 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
10983 local variables and temporaries, @samp{named_vars} enables the reuse only for
10984 user defined local variables with names, and @samp{none} disables stack reuse
10985 completely. The default value is @samp{all}. The option is needed when the
10986 program extends the lifetime of a scoped local variable or a compiler generated
10987 temporary beyond the end point defined by the language. When a lifetime of
10988 a variable ends, and if the variable lives in memory, the optimizing compiler
10989 has the freedom to reuse its stack space with other temporaries or scoped
10990 local variables whose live range does not overlap with it. Legacy code extending
10991 local lifetime is likely to break with the stack reuse optimization.
11010 if (*p == 10) // out of scope use of local1
11021 A(int k) : i(k), j(k) @{ @}
11028 void foo(const A& ar)
11035 foo(A(10)); // temp object's lifetime ends when foo returns
11041 ap->i+= 10; // ap references out of scope temp whose space
11042 // is reused with a. What is the value of ap->i?
11047 The lifetime of a compiler generated temporary is well defined by the C++
11048 standard. When a lifetime of a temporary ends, and if the temporary lives
11049 in memory, the optimizing compiler has the freedom to reuse its stack
11050 space with other temporaries or scoped local variables whose live range
11051 does not overlap with it. However some of the legacy code relies on
11052 the behavior of older compilers in which temporaries' stack space is
11053 not reused, the aggressive stack reuse can lead to runtime errors. This
11054 option is used to control the temporary stack reuse optimization.
11058 This option generates traps for signed overflow on addition, subtraction,
11059 multiplication operations.
11060 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11061 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11062 @option{-fwrapv} being effective. Note that only active options override, so
11063 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11064 results in @option{-ftrapv} being effective.
11068 This option instructs the compiler to assume that signed arithmetic
11069 overflow of addition, subtraction and multiplication wraps around
11070 using twos-complement representation. This flag enables some optimizations
11071 and disables others. This option is enabled by default for the Java
11072 front end, as required by the Java language specification.
11073 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11074 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11075 @option{-fwrapv} being effective. Note that only active options override, so
11076 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11077 results in @option{-ftrapv} being effective.
11080 @opindex fexceptions
11081 Enable exception handling. Generates extra code needed to propagate
11082 exceptions. For some targets, this implies GCC generates frame
11083 unwind information for all functions, which can produce significant data
11084 size overhead, although it does not affect execution. If you do not
11085 specify this option, GCC enables it by default for languages like
11086 C++ that normally require exception handling, and disables it for
11087 languages like C that do not normally require it. However, you may need
11088 to enable this option when compiling C code that needs to interoperate
11089 properly with exception handlers written in C++. You may also wish to
11090 disable this option if you are compiling older C++ programs that don't
11091 use exception handling.
11093 @item -fnon-call-exceptions
11094 @opindex fnon-call-exceptions
11095 Generate code that allows trapping instructions to throw exceptions.
11096 Note that this requires platform-specific runtime support that does
11097 not exist everywhere. Moreover, it only allows @emph{trapping}
11098 instructions to throw exceptions, i.e.@: memory references or floating-point
11099 instructions. It does not allow exceptions to be thrown from
11100 arbitrary signal handlers such as @code{SIGALRM}.
11102 @item -fdelete-dead-exceptions
11103 @opindex fdelete-dead-exceptions
11104 Consider that instructions that may throw exceptions but don't otherwise
11105 contribute to the execution of the program can be optimized away.
11106 This option is enabled by default for the Ada front end, as permitted by
11107 the Ada language specification.
11108 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
11110 @item -funwind-tables
11111 @opindex funwind-tables
11112 Similar to @option{-fexceptions}, except that it just generates any needed
11113 static data, but does not affect the generated code in any other way.
11114 You normally do not need to enable this option; instead, a language processor
11115 that needs this handling enables it on your behalf.
11117 @item -fasynchronous-unwind-tables
11118 @opindex fasynchronous-unwind-tables
11119 Generate unwind table in DWARF format, if supported by target machine. The
11120 table is exact at each instruction boundary, so it can be used for stack
11121 unwinding from asynchronous events (such as debugger or garbage collector).
11123 @item -fno-gnu-unique
11124 @opindex fno-gnu-unique
11125 On systems with recent GNU assembler and C library, the C++ compiler
11126 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
11127 of template static data members and static local variables in inline
11128 functions are unique even in the presence of @code{RTLD_LOCAL}; this
11129 is necessary to avoid problems with a library used by two different
11130 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
11131 therefore disagreeing with the other one about the binding of the
11132 symbol. But this causes @code{dlclose} to be ignored for affected
11133 DSOs; if your program relies on reinitialization of a DSO via
11134 @code{dlclose} and @code{dlopen}, you can use
11135 @option{-fno-gnu-unique}.
11137 @item -fpcc-struct-return
11138 @opindex fpcc-struct-return
11139 Return ``short'' @code{struct} and @code{union} values in memory like
11140 longer ones, rather than in registers. This convention is less
11141 efficient, but it has the advantage of allowing intercallability between
11142 GCC-compiled files and files compiled with other compilers, particularly
11143 the Portable C Compiler (pcc).
11145 The precise convention for returning structures in memory depends
11146 on the target configuration macros.
11148 Short structures and unions are those whose size and alignment match
11149 that of some integer type.
11151 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11152 switch is not binary compatible with code compiled with the
11153 @option{-freg-struct-return} switch.
11154 Use it to conform to a non-default application binary interface.
11156 @item -freg-struct-return
11157 @opindex freg-struct-return
11158 Return @code{struct} and @code{union} values in registers when possible.
11159 This is more efficient for small structures than
11160 @option{-fpcc-struct-return}.
11162 If you specify neither @option{-fpcc-struct-return} nor
11163 @option{-freg-struct-return}, GCC defaults to whichever convention is
11164 standard for the target. If there is no standard convention, GCC
11165 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11166 the principal compiler. In those cases, we can choose the standard, and
11167 we chose the more efficient register return alternative.
11169 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11170 switch is not binary compatible with code compiled with the
11171 @option{-fpcc-struct-return} switch.
11172 Use it to conform to a non-default application binary interface.
11174 @item -fshort-enums
11175 @opindex fshort-enums
11176 Allocate to an @code{enum} type only as many bytes as it needs for the
11177 declared range of possible values. Specifically, the @code{enum} type
11178 is equivalent to the smallest integer type that has enough room.
11180 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11181 code that is not binary compatible with code generated without that switch.
11182 Use it to conform to a non-default application binary interface.
11184 @item -fshort-wchar
11185 @opindex fshort-wchar
11186 Override the underlying type for @code{wchar_t} to be @code{short
11187 unsigned int} instead of the default for the target. This option is
11188 useful for building programs to run under WINE@.
11190 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11191 code that is not binary compatible with code generated without that switch.
11192 Use it to conform to a non-default application binary interface.
11195 @opindex fno-common
11196 In C code, controls the placement of uninitialized global variables.
11197 Unix C compilers have traditionally permitted multiple definitions of
11198 such variables in different compilation units by placing the variables
11200 This is the behavior specified by @option{-fcommon}, and is the default
11201 for GCC on most targets.
11202 On the other hand, this behavior is not required by ISO C, and on some
11203 targets may carry a speed or code size penalty on variable references.
11204 The @option{-fno-common} option specifies that the compiler should place
11205 uninitialized global variables in the data section of the object file,
11206 rather than generating them as common blocks.
11207 This has the effect that if the same variable is declared
11208 (without @code{extern}) in two different compilations,
11209 you get a multiple-definition error when you link them.
11210 In this case, you must compile with @option{-fcommon} instead.
11211 Compiling with @option{-fno-common} is useful on targets for which
11212 it provides better performance, or if you wish to verify that the
11213 program will work on other systems that always treat uninitialized
11214 variable declarations this way.
11218 Ignore the @code{#ident} directive.
11220 @item -finhibit-size-directive
11221 @opindex finhibit-size-directive
11222 Don't output a @code{.size} assembler directive, or anything else that
11223 would cause trouble if the function is split in the middle, and the
11224 two halves are placed at locations far apart in memory. This option is
11225 used when compiling @file{crtstuff.c}; you should not need to use it
11228 @item -fverbose-asm
11229 @opindex fverbose-asm
11230 Put extra commentary information in the generated assembly code to
11231 make it more readable. This option is generally only of use to those
11232 who actually need to read the generated assembly code (perhaps while
11233 debugging the compiler itself).
11235 @option{-fno-verbose-asm}, the default, causes the
11236 extra information to be omitted and is useful when comparing two assembler
11239 @item -frecord-gcc-switches
11240 @opindex frecord-gcc-switches
11241 This switch causes the command line used to invoke the
11242 compiler to be recorded into the object file that is being created.
11243 This switch is only implemented on some targets and the exact format
11244 of the recording is target and binary file format dependent, but it
11245 usually takes the form of a section containing ASCII text. This
11246 switch is related to the @option{-fverbose-asm} switch, but that
11247 switch only records information in the assembler output file as
11248 comments, so it never reaches the object file.
11249 See also @option{-grecord-gcc-switches} for another
11250 way of storing compiler options into the object file.
11254 @cindex global offset table
11256 Generate position-independent code (PIC) suitable for use in a shared
11257 library, if supported for the target machine. Such code accesses all
11258 constant addresses through a global offset table (GOT)@. The dynamic
11259 loader resolves the GOT entries when the program starts (the dynamic
11260 loader is not part of GCC; it is part of the operating system). If
11261 the GOT size for the linked executable exceeds a machine-specific
11262 maximum size, you get an error message from the linker indicating that
11263 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11264 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
11265 on the m68k and RS/6000. The x86 has no such limit.)
11267 Position-independent code requires special support, and therefore works
11268 only on certain machines. For the x86, GCC supports PIC for System V
11269 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11270 position-independent.
11272 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
11277 If supported for the target machine, emit position-independent code,
11278 suitable for dynamic linking and avoiding any limit on the size of the
11279 global offset table. This option makes a difference on AArch64, m68k,
11280 PowerPC and SPARC@.
11282 Position-independent code requires special support, and therefore works
11283 only on certain machines.
11285 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
11292 These options are similar to @option{-fpic} and @option{-fPIC}, but
11293 generated position independent code can be only linked into executables.
11294 Usually these options are used when @option{-pie} GCC option is
11295 used during linking.
11297 @option{-fpie} and @option{-fPIE} both define the macros
11298 @code{__pie__} and @code{__PIE__}. The macros have the value 1
11299 for @option{-fpie} and 2 for @option{-fPIE}.
11303 Do not use the PLT for external function calls in position-independent code.
11304 Instead, load the callee address at call sites from the GOT and branch to it.
11305 This leads to more efficient code by eliminating PLT stubs and exposing
11306 GOT loads to optimizations. On architectures such as 32-bit x86 where
11307 PLT stubs expect the GOT pointer in a specific register, this gives more
11308 register allocation freedom to the compiler.
11309 Lazy binding requires use of the PLT;
11310 with @option{-fno-plt} all external symbols are resolved at load time.
11312 Alternatively, the function attribute @code{noplt} can be used to avoid calls
11313 through the PLT for specific external functions.
11315 In position-dependent code, a few targets also convert calls to
11316 functions that are marked to not use the PLT to use the GOT instead.
11318 @item -fno-jump-tables
11319 @opindex fno-jump-tables
11320 Do not use jump tables for switch statements even where it would be
11321 more efficient than other code generation strategies. This option is
11322 of use in conjunction with @option{-fpic} or @option{-fPIC} for
11323 building code that forms part of a dynamic linker and cannot
11324 reference the address of a jump table. On some targets, jump tables
11325 do not require a GOT and this option is not needed.
11327 @item -ffixed-@var{reg}
11329 Treat the register named @var{reg} as a fixed register; generated code
11330 should never refer to it (except perhaps as a stack pointer, frame
11331 pointer or in some other fixed role).
11333 @var{reg} must be the name of a register. The register names accepted
11334 are machine-specific and are defined in the @code{REGISTER_NAMES}
11335 macro in the machine description macro file.
11337 This flag does not have a negative form, because it specifies a
11340 @item -fcall-used-@var{reg}
11341 @opindex fcall-used
11342 Treat the register named @var{reg} as an allocable register that is
11343 clobbered by function calls. It may be allocated for temporaries or
11344 variables that do not live across a call. Functions compiled this way
11345 do not save and restore the register @var{reg}.
11347 It is an error to use this flag with the frame pointer or stack pointer.
11348 Use of this flag for other registers that have fixed pervasive roles in
11349 the machine's execution model produces disastrous results.
11351 This flag does not have a negative form, because it specifies a
11354 @item -fcall-saved-@var{reg}
11355 @opindex fcall-saved
11356 Treat the register named @var{reg} as an allocable register saved by
11357 functions. It may be allocated even for temporaries or variables that
11358 live across a call. Functions compiled this way save and restore
11359 the register @var{reg} if they use it.
11361 It is an error to use this flag with the frame pointer or stack pointer.
11362 Use of this flag for other registers that have fixed pervasive roles in
11363 the machine's execution model produces disastrous results.
11365 A different sort of disaster results from the use of this flag for
11366 a register in which function values may be returned.
11368 This flag does not have a negative form, because it specifies a
11371 @item -fpack-struct[=@var{n}]
11372 @opindex fpack-struct
11373 Without a value specified, pack all structure members together without
11374 holes. When a value is specified (which must be a small power of two), pack
11375 structure members according to this value, representing the maximum
11376 alignment (that is, objects with default alignment requirements larger than
11377 this are output potentially unaligned at the next fitting location.
11379 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11380 code that is not binary compatible with code generated without that switch.
11381 Additionally, it makes the code suboptimal.
11382 Use it to conform to a non-default application binary interface.
11384 @item -fleading-underscore
11385 @opindex fleading-underscore
11386 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11387 change the way C symbols are represented in the object file. One use
11388 is to help link with legacy assembly code.
11390 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11391 generate code that is not binary compatible with code generated without that
11392 switch. Use it to conform to a non-default application binary interface.
11393 Not all targets provide complete support for this switch.
11395 @item -ftls-model=@var{model}
11396 @opindex ftls-model
11397 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11398 The @var{model} argument should be one of @samp{global-dynamic},
11399 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
11400 Note that the choice is subject to optimization: the compiler may use
11401 a more efficient model for symbols not visible outside of the translation
11402 unit, or if @option{-fpic} is not given on the command line.
11404 The default without @option{-fpic} is @samp{initial-exec}; with
11405 @option{-fpic} the default is @samp{global-dynamic}.
11407 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
11408 @opindex fvisibility
11409 Set the default ELF image symbol visibility to the specified option---all
11410 symbols are marked with this unless overridden within the code.
11411 Using this feature can very substantially improve linking and
11412 load times of shared object libraries, produce more optimized
11413 code, provide near-perfect API export and prevent symbol clashes.
11414 It is @strong{strongly} recommended that you use this in any shared objects
11417 Despite the nomenclature, @samp{default} always means public; i.e.,
11418 available to be linked against from outside the shared object.
11419 @samp{protected} and @samp{internal} are pretty useless in real-world
11420 usage so the only other commonly used option is @samp{hidden}.
11421 The default if @option{-fvisibility} isn't specified is
11422 @samp{default}, i.e., make every symbol public.
11424 A good explanation of the benefits offered by ensuring ELF
11425 symbols have the correct visibility is given by ``How To Write
11426 Shared Libraries'' by Ulrich Drepper (which can be found at
11427 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
11428 solution made possible by this option to marking things hidden when
11429 the default is public is to make the default hidden and mark things
11430 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
11431 and @code{__attribute__ ((visibility("default")))} instead of
11432 @code{__declspec(dllexport)} you get almost identical semantics with
11433 identical syntax. This is a great boon to those working with
11434 cross-platform projects.
11436 For those adding visibility support to existing code, you may find
11437 @code{#pragma GCC visibility} of use. This works by you enclosing
11438 the declarations you wish to set visibility for with (for example)
11439 @code{#pragma GCC visibility push(hidden)} and
11440 @code{#pragma GCC visibility pop}.
11441 Bear in mind that symbol visibility should be viewed @strong{as
11442 part of the API interface contract} and thus all new code should
11443 always specify visibility when it is not the default; i.e., declarations
11444 only for use within the local DSO should @strong{always} be marked explicitly
11445 as hidden as so to avoid PLT indirection overheads---making this
11446 abundantly clear also aids readability and self-documentation of the code.
11447 Note that due to ISO C++ specification requirements, @code{operator new} and
11448 @code{operator delete} must always be of default visibility.
11450 Be aware that headers from outside your project, in particular system
11451 headers and headers from any other library you use, may not be
11452 expecting to be compiled with visibility other than the default. You
11453 may need to explicitly say @code{#pragma GCC visibility push(default)}
11454 before including any such headers.
11456 @code{extern} declarations are not affected by @option{-fvisibility}, so
11457 a lot of code can be recompiled with @option{-fvisibility=hidden} with
11458 no modifications. However, this means that calls to @code{extern}
11459 functions with no explicit visibility use the PLT, so it is more
11460 effective to use @code{__attribute ((visibility))} and/or
11461 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
11462 declarations should be treated as hidden.
11464 Note that @option{-fvisibility} does affect C++ vague linkage
11465 entities. This means that, for instance, an exception class that is
11466 be thrown between DSOs must be explicitly marked with default
11467 visibility so that the @samp{type_info} nodes are unified between
11470 An overview of these techniques, their benefits and how to use them
11471 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
11473 @item -fstrict-volatile-bitfields
11474 @opindex fstrict-volatile-bitfields
11475 This option should be used if accesses to volatile bit-fields (or other
11476 structure fields, although the compiler usually honors those types
11477 anyway) should use a single access of the width of the
11478 field's type, aligned to a natural alignment if possible. For
11479 example, targets with memory-mapped peripheral registers might require
11480 all such accesses to be 16 bits wide; with this flag you can
11481 declare all peripheral bit-fields as @code{unsigned short} (assuming short
11482 is 16 bits on these targets) to force GCC to use 16-bit accesses
11483 instead of, perhaps, a more efficient 32-bit access.
11485 If this option is disabled, the compiler uses the most efficient
11486 instruction. In the previous example, that might be a 32-bit load
11487 instruction, even though that accesses bytes that do not contain
11488 any portion of the bit-field, or memory-mapped registers unrelated to
11489 the one being updated.
11491 In some cases, such as when the @code{packed} attribute is applied to a
11492 structure field, it may not be possible to access the field with a single
11493 read or write that is correctly aligned for the target machine. In this
11494 case GCC falls back to generating multiple accesses rather than code that
11495 will fault or truncate the result at run time.
11497 Note: Due to restrictions of the C/C++11 memory model, write accesses are
11498 not allowed to touch non bit-field members. It is therefore recommended
11499 to define all bits of the field's type as bit-field members.
11501 The default value of this option is determined by the application binary
11502 interface for the target processor.
11504 @item -fsync-libcalls
11505 @opindex fsync-libcalls
11506 This option controls whether any out-of-line instance of the @code{__sync}
11507 family of functions may be used to implement the C++11 @code{__atomic}
11508 family of functions.
11510 The default value of this option is enabled, thus the only useful form
11511 of the option is @option{-fno-sync-libcalls}. This option is used in
11512 the implementation of the @file{libatomic} runtime library.
11516 @node Developer Options
11517 @section GCC Developer Options
11518 @cindex developer options
11519 @cindex debugging GCC
11520 @cindex debug dump options
11521 @cindex dump options
11522 @cindex compilation statistics
11524 This section describes command-line options that are primarily of
11525 interest to GCC developers, including options to support compiler
11526 testing and investigation of compiler bugs and compile-time
11527 performance problems. This includes options that produce debug dumps
11528 at various points in the compilation; that print statistics such as
11529 memory use and execution time; and that print information about GCC's
11530 configuration, such as where it searches for libraries. You should
11531 rarely need to use any of these options for ordinary compilation and
11536 @item -d@var{letters}
11537 @itemx -fdump-rtl-@var{pass}
11538 @itemx -fdump-rtl-@var{pass}=@var{filename}
11540 @opindex fdump-rtl-@var{pass}
11541 Says to make debugging dumps during compilation at times specified by
11542 @var{letters}. This is used for debugging the RTL-based passes of the
11543 compiler. The file names for most of the dumps are made by appending
11544 a pass number and a word to the @var{dumpname}, and the files are
11545 created in the directory of the output file. In case of
11546 @option{=@var{filename}} option, the dump is output on the given file
11547 instead of the pass numbered dump files. Note that the pass number is
11548 assigned as passes are registered into the pass manager. Most passes
11549 are registered in the order that they will execute and for these passes
11550 the number corresponds to the pass execution order. However, passes
11551 registered by plugins, passes specific to compilation targets, or
11552 passes that are otherwise registered after all the other passes are
11553 numbered higher than a pass named "final", even if they are executed
11554 earlier. @var{dumpname} is generated from the name of the output
11555 file if explicitly specified and not an executable, otherwise it is
11556 the basename of the source file. These switches may have different
11557 effects when @option{-E} is used for preprocessing.
11559 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
11560 @option{-d} option @var{letters}. Here are the possible
11561 letters for use in @var{pass} and @var{letters}, and their meanings:
11565 @item -fdump-rtl-alignments
11566 @opindex fdump-rtl-alignments
11567 Dump after branch alignments have been computed.
11569 @item -fdump-rtl-asmcons
11570 @opindex fdump-rtl-asmcons
11571 Dump after fixing rtl statements that have unsatisfied in/out constraints.
11573 @item -fdump-rtl-auto_inc_dec
11574 @opindex fdump-rtl-auto_inc_dec
11575 Dump after auto-inc-dec discovery. This pass is only run on
11576 architectures that have auto inc or auto dec instructions.
11578 @item -fdump-rtl-barriers
11579 @opindex fdump-rtl-barriers
11580 Dump after cleaning up the barrier instructions.
11582 @item -fdump-rtl-bbpart
11583 @opindex fdump-rtl-bbpart
11584 Dump after partitioning hot and cold basic blocks.
11586 @item -fdump-rtl-bbro
11587 @opindex fdump-rtl-bbro
11588 Dump after block reordering.
11590 @item -fdump-rtl-btl1
11591 @itemx -fdump-rtl-btl2
11592 @opindex fdump-rtl-btl2
11593 @opindex fdump-rtl-btl2
11594 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
11595 after the two branch
11596 target load optimization passes.
11598 @item -fdump-rtl-bypass
11599 @opindex fdump-rtl-bypass
11600 Dump after jump bypassing and control flow optimizations.
11602 @item -fdump-rtl-combine
11603 @opindex fdump-rtl-combine
11604 Dump after the RTL instruction combination pass.
11606 @item -fdump-rtl-compgotos
11607 @opindex fdump-rtl-compgotos
11608 Dump after duplicating the computed gotos.
11610 @item -fdump-rtl-ce1
11611 @itemx -fdump-rtl-ce2
11612 @itemx -fdump-rtl-ce3
11613 @opindex fdump-rtl-ce1
11614 @opindex fdump-rtl-ce2
11615 @opindex fdump-rtl-ce3
11616 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
11617 @option{-fdump-rtl-ce3} enable dumping after the three
11618 if conversion passes.
11620 @item -fdump-rtl-cprop_hardreg
11621 @opindex fdump-rtl-cprop_hardreg
11622 Dump after hard register copy propagation.
11624 @item -fdump-rtl-csa
11625 @opindex fdump-rtl-csa
11626 Dump after combining stack adjustments.
11628 @item -fdump-rtl-cse1
11629 @itemx -fdump-rtl-cse2
11630 @opindex fdump-rtl-cse1
11631 @opindex fdump-rtl-cse2
11632 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
11633 the two common subexpression elimination passes.
11635 @item -fdump-rtl-dce
11636 @opindex fdump-rtl-dce
11637 Dump after the standalone dead code elimination passes.
11639 @item -fdump-rtl-dbr
11640 @opindex fdump-rtl-dbr
11641 Dump after delayed branch scheduling.
11643 @item -fdump-rtl-dce1
11644 @itemx -fdump-rtl-dce2
11645 @opindex fdump-rtl-dce1
11646 @opindex fdump-rtl-dce2
11647 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
11648 the two dead store elimination passes.
11650 @item -fdump-rtl-eh
11651 @opindex fdump-rtl-eh
11652 Dump after finalization of EH handling code.
11654 @item -fdump-rtl-eh_ranges
11655 @opindex fdump-rtl-eh_ranges
11656 Dump after conversion of EH handling range regions.
11658 @item -fdump-rtl-expand
11659 @opindex fdump-rtl-expand
11660 Dump after RTL generation.
11662 @item -fdump-rtl-fwprop1
11663 @itemx -fdump-rtl-fwprop2
11664 @opindex fdump-rtl-fwprop1
11665 @opindex fdump-rtl-fwprop2
11666 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
11667 dumping after the two forward propagation passes.
11669 @item -fdump-rtl-gcse1
11670 @itemx -fdump-rtl-gcse2
11671 @opindex fdump-rtl-gcse1
11672 @opindex fdump-rtl-gcse2
11673 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
11674 after global common subexpression elimination.
11676 @item -fdump-rtl-init-regs
11677 @opindex fdump-rtl-init-regs
11678 Dump after the initialization of the registers.
11680 @item -fdump-rtl-initvals
11681 @opindex fdump-rtl-initvals
11682 Dump after the computation of the initial value sets.
11684 @item -fdump-rtl-into_cfglayout
11685 @opindex fdump-rtl-into_cfglayout
11686 Dump after converting to cfglayout mode.
11688 @item -fdump-rtl-ira
11689 @opindex fdump-rtl-ira
11690 Dump after iterated register allocation.
11692 @item -fdump-rtl-jump
11693 @opindex fdump-rtl-jump
11694 Dump after the second jump optimization.
11696 @item -fdump-rtl-loop2
11697 @opindex fdump-rtl-loop2
11698 @option{-fdump-rtl-loop2} enables dumping after the rtl
11699 loop optimization passes.
11701 @item -fdump-rtl-mach
11702 @opindex fdump-rtl-mach
11703 Dump after performing the machine dependent reorganization pass, if that
11706 @item -fdump-rtl-mode_sw
11707 @opindex fdump-rtl-mode_sw
11708 Dump after removing redundant mode switches.
11710 @item -fdump-rtl-rnreg
11711 @opindex fdump-rtl-rnreg
11712 Dump after register renumbering.
11714 @item -fdump-rtl-outof_cfglayout
11715 @opindex fdump-rtl-outof_cfglayout
11716 Dump after converting from cfglayout mode.
11718 @item -fdump-rtl-peephole2
11719 @opindex fdump-rtl-peephole2
11720 Dump after the peephole pass.
11722 @item -fdump-rtl-postreload
11723 @opindex fdump-rtl-postreload
11724 Dump after post-reload optimizations.
11726 @item -fdump-rtl-pro_and_epilogue
11727 @opindex fdump-rtl-pro_and_epilogue
11728 Dump after generating the function prologues and epilogues.
11730 @item -fdump-rtl-sched1
11731 @itemx -fdump-rtl-sched2
11732 @opindex fdump-rtl-sched1
11733 @opindex fdump-rtl-sched2
11734 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
11735 after the basic block scheduling passes.
11737 @item -fdump-rtl-ree
11738 @opindex fdump-rtl-ree
11739 Dump after sign/zero extension elimination.
11741 @item -fdump-rtl-seqabstr
11742 @opindex fdump-rtl-seqabstr
11743 Dump after common sequence discovery.
11745 @item -fdump-rtl-shorten
11746 @opindex fdump-rtl-shorten
11747 Dump after shortening branches.
11749 @item -fdump-rtl-sibling
11750 @opindex fdump-rtl-sibling
11751 Dump after sibling call optimizations.
11753 @item -fdump-rtl-split1
11754 @itemx -fdump-rtl-split2
11755 @itemx -fdump-rtl-split3
11756 @itemx -fdump-rtl-split4
11757 @itemx -fdump-rtl-split5
11758 @opindex fdump-rtl-split1
11759 @opindex fdump-rtl-split2
11760 @opindex fdump-rtl-split3
11761 @opindex fdump-rtl-split4
11762 @opindex fdump-rtl-split5
11763 These options enable dumping after five rounds of
11764 instruction splitting.
11766 @item -fdump-rtl-sms
11767 @opindex fdump-rtl-sms
11768 Dump after modulo scheduling. This pass is only run on some
11771 @item -fdump-rtl-stack
11772 @opindex fdump-rtl-stack
11773 Dump after conversion from GCC's ``flat register file'' registers to the
11774 x87's stack-like registers. This pass is only run on x86 variants.
11776 @item -fdump-rtl-subreg1
11777 @itemx -fdump-rtl-subreg2
11778 @opindex fdump-rtl-subreg1
11779 @opindex fdump-rtl-subreg2
11780 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
11781 the two subreg expansion passes.
11783 @item -fdump-rtl-unshare
11784 @opindex fdump-rtl-unshare
11785 Dump after all rtl has been unshared.
11787 @item -fdump-rtl-vartrack
11788 @opindex fdump-rtl-vartrack
11789 Dump after variable tracking.
11791 @item -fdump-rtl-vregs
11792 @opindex fdump-rtl-vregs
11793 Dump after converting virtual registers to hard registers.
11795 @item -fdump-rtl-web
11796 @opindex fdump-rtl-web
11797 Dump after live range splitting.
11799 @item -fdump-rtl-regclass
11800 @itemx -fdump-rtl-subregs_of_mode_init
11801 @itemx -fdump-rtl-subregs_of_mode_finish
11802 @itemx -fdump-rtl-dfinit
11803 @itemx -fdump-rtl-dfinish
11804 @opindex fdump-rtl-regclass
11805 @opindex fdump-rtl-subregs_of_mode_init
11806 @opindex fdump-rtl-subregs_of_mode_finish
11807 @opindex fdump-rtl-dfinit
11808 @opindex fdump-rtl-dfinish
11809 These dumps are defined but always produce empty files.
11812 @itemx -fdump-rtl-all
11814 @opindex fdump-rtl-all
11815 Produce all the dumps listed above.
11819 Annotate the assembler output with miscellaneous debugging information.
11823 Dump all macro definitions, at the end of preprocessing, in addition to
11828 Produce a core dump whenever an error occurs.
11832 Annotate the assembler output with a comment indicating which
11833 pattern and alternative is used. The length of each instruction is
11838 Dump the RTL in the assembler output as a comment before each instruction.
11839 Also turns on @option{-dp} annotation.
11843 Just generate RTL for a function instead of compiling it. Usually used
11844 with @option{-fdump-rtl-expand}.
11847 @item -fdump-noaddr
11848 @opindex fdump-noaddr
11849 When doing debugging dumps, suppress address output. This makes it more
11850 feasible to use diff on debugging dumps for compiler invocations with
11851 different compiler binaries and/or different
11852 text / bss / data / heap / stack / dso start locations.
11855 @opindex freport-bug
11856 Collect and dump debug information into a temporary file if an
11857 internal compiler error (ICE) occurs.
11859 @item -fdump-unnumbered
11860 @opindex fdump-unnumbered
11861 When doing debugging dumps, suppress instruction numbers and address output.
11862 This makes it more feasible to use diff on debugging dumps for compiler
11863 invocations with different options, in particular with and without
11866 @item -fdump-unnumbered-links
11867 @opindex fdump-unnumbered-links
11868 When doing debugging dumps (see @option{-d} option above), suppress
11869 instruction numbers for the links to the previous and next instructions
11872 @item -fdump-translation-unit @r{(C++ only)}
11873 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
11874 @opindex fdump-translation-unit
11875 Dump a representation of the tree structure for the entire translation
11876 unit to a file. The file name is made by appending @file{.tu} to the
11877 source file name, and the file is created in the same directory as the
11878 output file. If the @samp{-@var{options}} form is used, @var{options}
11879 controls the details of the dump as described for the
11880 @option{-fdump-tree} options.
11882 @item -fdump-class-hierarchy @r{(C++ only)}
11883 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
11884 @opindex fdump-class-hierarchy
11885 Dump a representation of each class's hierarchy and virtual function
11886 table layout to a file. The file name is made by appending
11887 @file{.class} to the source file name, and the file is created in the
11888 same directory as the output file. If the @samp{-@var{options}} form
11889 is used, @var{options} controls the details of the dump as described
11890 for the @option{-fdump-tree} options.
11892 @item -fdump-ipa-@var{switch}
11894 Control the dumping at various stages of inter-procedural analysis
11895 language tree to a file. The file name is generated by appending a
11896 switch specific suffix to the source file name, and the file is created
11897 in the same directory as the output file. The following dumps are
11902 Enables all inter-procedural analysis dumps.
11905 Dumps information about call-graph optimization, unused function removal,
11906 and inlining decisions.
11909 Dump after function inlining.
11913 @item -fdump-passes
11914 @opindex fdump-passes
11915 Dump the list of optimization passes that are turned on and off by
11916 the current command-line options.
11918 @item -fdump-statistics-@var{option}
11919 @opindex fdump-statistics
11920 Enable and control dumping of pass statistics in a separate file. The
11921 file name is generated by appending a suffix ending in
11922 @samp{.statistics} to the source file name, and the file is created in
11923 the same directory as the output file. If the @samp{-@var{option}}
11924 form is used, @samp{-stats} causes counters to be summed over the
11925 whole compilation unit while @samp{-details} dumps every event as
11926 the passes generate them. The default with no option is to sum
11927 counters for each function compiled.
11929 @item -fdump-tree-@var{switch}
11930 @itemx -fdump-tree-@var{switch}-@var{options}
11931 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
11932 @opindex fdump-tree
11933 Control the dumping at various stages of processing the intermediate
11934 language tree to a file. The file name is generated by appending a
11935 switch-specific suffix to the source file name, and the file is
11936 created in the same directory as the output file. In case of
11937 @option{=@var{filename}} option, the dump is output on the given file
11938 instead of the auto named dump files. If the @samp{-@var{options}}
11939 form is used, @var{options} is a list of @samp{-} separated options
11940 which control the details of the dump. Not all options are applicable
11941 to all dumps; those that are not meaningful are ignored. The
11942 following options are available
11946 Print the address of each node. Usually this is not meaningful as it
11947 changes according to the environment and source file. Its primary use
11948 is for tying up a dump file with a debug environment.
11950 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
11951 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
11952 use working backward from mangled names in the assembly file.
11954 When dumping front-end intermediate representations, inhibit dumping
11955 of members of a scope or body of a function merely because that scope
11956 has been reached. Only dump such items when they are directly reachable
11957 by some other path.
11959 When dumping pretty-printed trees, this option inhibits dumping the
11960 bodies of control structures.
11962 When dumping RTL, print the RTL in slim (condensed) form instead of
11963 the default LISP-like representation.
11965 Print a raw representation of the tree. By default, trees are
11966 pretty-printed into a C-like representation.
11968 Enable more detailed dumps (not honored by every dump option). Also
11969 include information from the optimization passes.
11971 Enable dumping various statistics about the pass (not honored by every dump
11974 Enable showing basic block boundaries (disabled in raw dumps).
11976 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
11977 dump a representation of the control flow graph suitable for viewing with
11978 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
11979 the file is pretty-printed as a subgraph, so that GraphViz can render them
11980 all in a single plot.
11982 This option currently only works for RTL dumps, and the RTL is always
11983 dumped in slim form.
11985 Enable showing virtual operands for every statement.
11987 Enable showing line numbers for statements.
11989 Enable showing the unique ID (@code{DECL_UID}) for each variable.
11991 Enable showing the tree dump for each statement.
11993 Enable showing the EH region number holding each statement.
11995 Enable showing scalar evolution analysis details.
11997 Enable showing optimization information (only available in certain
12000 Enable showing missed optimization information (only available in certain
12003 Enable other detailed optimization information (only available in
12005 @item =@var{filename}
12006 Instead of an auto named dump file, output into the given file
12007 name. The file names @file{stdout} and @file{stderr} are treated
12008 specially and are considered already open standard streams. For
12012 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
12013 -fdump-tree-pre=stderr file.c
12016 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
12017 output on to @file{stderr}. If two conflicting dump filenames are
12018 given for the same pass, then the latter option overrides the earlier
12022 @opindex fdump-tree-split-paths
12023 Dump each function after splitting paths to loop backedges. The file
12024 name is made by appending @file{.split-paths} to the source file name.
12027 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
12028 and @option{lineno}.
12031 Turn on all optimization options, i.e., @option{optimized},
12032 @option{missed}, and @option{note}.
12035 The following tree dumps are possible:
12039 @opindex fdump-tree-original
12040 Dump before any tree based optimization, to @file{@var{file}.original}.
12043 @opindex fdump-tree-optimized
12044 Dump after all tree based optimization, to @file{@var{file}.optimized}.
12047 @opindex fdump-tree-gimple
12048 Dump each function before and after the gimplification pass to a file. The
12049 file name is made by appending @file{.gimple} to the source file name.
12052 @opindex fdump-tree-cfg
12053 Dump the control flow graph of each function to a file. The file name is
12054 made by appending @file{.cfg} to the source file name.
12057 @opindex fdump-tree-ch
12058 Dump each function after copying loop headers. The file name is made by
12059 appending @file{.ch} to the source file name.
12062 @opindex fdump-tree-ssa
12063 Dump SSA related information to a file. The file name is made by appending
12064 @file{.ssa} to the source file name.
12067 @opindex fdump-tree-alias
12068 Dump aliasing information for each function. The file name is made by
12069 appending @file{.alias} to the source file name.
12072 @opindex fdump-tree-ccp
12073 Dump each function after CCP@. The file name is made by appending
12074 @file{.ccp} to the source file name.
12077 @opindex fdump-tree-storeccp
12078 Dump each function after STORE-CCP@. The file name is made by appending
12079 @file{.storeccp} to the source file name.
12082 @opindex fdump-tree-pre
12083 Dump trees after partial redundancy elimination. The file name is made
12084 by appending @file{.pre} to the source file name.
12087 @opindex fdump-tree-fre
12088 Dump trees after full redundancy elimination. The file name is made
12089 by appending @file{.fre} to the source file name.
12092 @opindex fdump-tree-copyprop
12093 Dump trees after copy propagation. The file name is made
12094 by appending @file{.copyprop} to the source file name.
12096 @item store_copyprop
12097 @opindex fdump-tree-store_copyprop
12098 Dump trees after store copy-propagation. The file name is made
12099 by appending @file{.store_copyprop} to the source file name.
12102 @opindex fdump-tree-dce
12103 Dump each function after dead code elimination. The file name is made by
12104 appending @file{.dce} to the source file name.
12107 @opindex fdump-tree-sra
12108 Dump each function after performing scalar replacement of aggregates. The
12109 file name is made by appending @file{.sra} to the source file name.
12112 @opindex fdump-tree-sink
12113 Dump each function after performing code sinking. The file name is made
12114 by appending @file{.sink} to the source file name.
12117 @opindex fdump-tree-dom
12118 Dump each function after applying dominator tree optimizations. The file
12119 name is made by appending @file{.dom} to the source file name.
12122 @opindex fdump-tree-dse
12123 Dump each function after applying dead store elimination. The file
12124 name is made by appending @file{.dse} to the source file name.
12127 @opindex fdump-tree-phiopt
12128 Dump each function after optimizing PHI nodes into straightline code. The file
12129 name is made by appending @file{.phiopt} to the source file name.
12132 @opindex fdump-tree-backprop
12133 Dump each function after back-propagating use information up the definition
12134 chain. The file name is made by appending @file{.backprop} to the
12138 @opindex fdump-tree-forwprop
12139 Dump each function after forward propagating single use variables. The file
12140 name is made by appending @file{.forwprop} to the source file name.
12143 @opindex fdump-tree-nrv
12144 Dump each function after applying the named return value optimization on
12145 generic trees. The file name is made by appending @file{.nrv} to the source
12149 @opindex fdump-tree-vect
12150 Dump each function after applying vectorization of loops. The file name is
12151 made by appending @file{.vect} to the source file name.
12154 @opindex fdump-tree-slp
12155 Dump each function after applying vectorization of basic blocks. The file name
12156 is made by appending @file{.slp} to the source file name.
12159 @opindex fdump-tree-vrp
12160 Dump each function after Value Range Propagation (VRP). The file name
12161 is made by appending @file{.vrp} to the source file name.
12164 @opindex fdump-tree-oaccdevlow
12165 Dump each function after applying device-specific OpenACC transformations.
12166 The file name is made by appending @file{.oaccdevlow} to the source file name.
12169 @opindex fdump-tree-all
12170 Enable all the available tree dumps with the flags provided in this option.
12174 @itemx -fopt-info-@var{options}
12175 @itemx -fopt-info-@var{options}=@var{filename}
12177 Controls optimization dumps from various optimization passes. If the
12178 @samp{-@var{options}} form is used, @var{options} is a list of
12179 @samp{-} separated option keywords to select the dump details and
12182 The @var{options} can be divided into two groups: options describing the
12183 verbosity of the dump, and options describing which optimizations
12184 should be included. The options from both the groups can be freely
12185 mixed as they are non-overlapping. However, in case of any conflicts,
12186 the later options override the earlier options on the command
12189 The following options control the dump verbosity:
12193 Print information when an optimization is successfully applied. It is
12194 up to a pass to decide which information is relevant. For example, the
12195 vectorizer passes print the source location of loops which are
12196 successfully vectorized.
12198 Print information about missed optimizations. Individual passes
12199 control which information to include in the output.
12201 Print verbose information about optimizations, such as certain
12202 transformations, more detailed messages about decisions etc.
12204 Print detailed optimization information. This includes
12205 @samp{optimized}, @samp{missed}, and @samp{note}.
12208 One or more of the following option keywords can be used to describe a
12209 group of optimizations:
12213 Enable dumps from all interprocedural optimizations.
12215 Enable dumps from all loop optimizations.
12217 Enable dumps from all inlining optimizations.
12219 Enable dumps from all vectorization optimizations.
12221 Enable dumps from all optimizations. This is a superset of
12222 the optimization groups listed above.
12225 If @var{options} is
12226 omitted, it defaults to @samp{optimized-optall}, which means to dump all
12227 info about successful optimizations from all the passes.
12229 If the @var{filename} is provided, then the dumps from all the
12230 applicable optimizations are concatenated into the @var{filename}.
12231 Otherwise the dump is output onto @file{stderr}. Though multiple
12232 @option{-fopt-info} options are accepted, only one of them can include
12233 a @var{filename}. If other filenames are provided then all but the
12234 first such option are ignored.
12236 Note that the output @var{filename} is overwritten
12237 in case of multiple translation units. If a combined output from
12238 multiple translation units is desired, @file{stderr} should be used
12241 In the following example, the optimization info is output to
12250 gcc -O3 -fopt-info-missed=missed.all
12254 outputs missed optimization report from all the passes into
12255 @file{missed.all}, and this one:
12258 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
12262 prints information about missed optimization opportunities from
12263 vectorization passes on @file{stderr}.
12264 Note that @option{-fopt-info-vec-missed} is equivalent to
12265 @option{-fopt-info-missed-vec}.
12267 As another example,
12269 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
12273 outputs information about missed optimizations as well as
12274 optimized locations from all the inlining passes into
12280 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
12284 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
12285 in conflict since only one output file is allowed. In this case, only
12286 the first option takes effect and the subsequent options are
12287 ignored. Thus only @file{vec.miss} is produced which contains
12288 dumps from the vectorizer about missed opportunities.
12290 @item -fsched-verbose=@var{n}
12291 @opindex fsched-verbose
12292 On targets that use instruction scheduling, this option controls the
12293 amount of debugging output the scheduler prints to the dump files.
12295 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
12296 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
12297 For @var{n} greater than one, it also output basic block probabilities,
12298 detailed ready list information and unit/insn info. For @var{n} greater
12299 than two, it includes RTL at abort point, control-flow and regions info.
12300 And for @var{n} over four, @option{-fsched-verbose} also includes
12305 @item -fenable-@var{kind}-@var{pass}
12306 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
12310 This is a set of options that are used to explicitly disable/enable
12311 optimization passes. These options are intended for use for debugging GCC.
12312 Compiler users should use regular options for enabling/disabling
12317 @item -fdisable-ipa-@var{pass}
12318 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
12319 statically invoked in the compiler multiple times, the pass name should be
12320 appended with a sequential number starting from 1.
12322 @item -fdisable-rtl-@var{pass}
12323 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
12324 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
12325 statically invoked in the compiler multiple times, the pass name should be
12326 appended with a sequential number starting from 1. @var{range-list} is a
12327 comma-separated list of function ranges or assembler names. Each range is a number
12328 pair separated by a colon. The range is inclusive in both ends. If the range
12329 is trivial, the number pair can be simplified as a single number. If the
12330 function's call graph node's @var{uid} falls within one of the specified ranges,
12331 the @var{pass} is disabled for that function. The @var{uid} is shown in the
12332 function header of a dump file, and the pass names can be dumped by using
12333 option @option{-fdump-passes}.
12335 @item -fdisable-tree-@var{pass}
12336 @itemx -fdisable-tree-@var{pass}=@var{range-list}
12337 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
12340 @item -fenable-ipa-@var{pass}
12341 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
12342 statically invoked in the compiler multiple times, the pass name should be
12343 appended with a sequential number starting from 1.
12345 @item -fenable-rtl-@var{pass}
12346 @itemx -fenable-rtl-@var{pass}=@var{range-list}
12347 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
12348 description and examples.
12350 @item -fenable-tree-@var{pass}
12351 @itemx -fenable-tree-@var{pass}=@var{range-list}
12352 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
12353 of option arguments.
12357 Here are some examples showing uses of these options.
12361 # disable ccp1 for all functions
12362 -fdisable-tree-ccp1
12363 # disable complete unroll for function whose cgraph node uid is 1
12364 -fenable-tree-cunroll=1
12365 # disable gcse2 for functions at the following ranges [1,1],
12366 # [300,400], and [400,1000]
12367 # disable gcse2 for functions foo and foo2
12368 -fdisable-rtl-gcse2=foo,foo2
12369 # disable early inlining
12370 -fdisable-tree-einline
12371 # disable ipa inlining
12372 -fdisable-ipa-inline
12373 # enable tree full unroll
12374 -fenable-tree-unroll
12380 @opindex fno-checking
12381 Enable internal consistency checking. The default depends on
12382 the compiler configuration.
12384 @item -frandom-seed=@var{string}
12385 @opindex frandom-seed
12386 This option provides a seed that GCC uses in place of
12387 random numbers in generating certain symbol names
12388 that have to be different in every compiled file. It is also used to
12389 place unique stamps in coverage data files and the object files that
12390 produce them. You can use the @option{-frandom-seed} option to produce
12391 reproducibly identical object files.
12393 The @var{string} can either be a number (decimal, octal or hex) or an
12394 arbitrary string (in which case it's converted to a number by
12397 The @var{string} should be different for every file you compile.
12400 @itemx -save-temps=cwd
12401 @opindex save-temps
12402 Store the usual ``temporary'' intermediate files permanently; place them
12403 in the current directory and name them based on the source file. Thus,
12404 compiling @file{foo.c} with @option{-c -save-temps} produces files
12405 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
12406 preprocessed @file{foo.i} output file even though the compiler now
12407 normally uses an integrated preprocessor.
12409 When used in combination with the @option{-x} command-line option,
12410 @option{-save-temps} is sensible enough to avoid over writing an
12411 input source file with the same extension as an intermediate file.
12412 The corresponding intermediate file may be obtained by renaming the
12413 source file before using @option{-save-temps}.
12415 If you invoke GCC in parallel, compiling several different source
12416 files that share a common base name in different subdirectories or the
12417 same source file compiled for multiple output destinations, it is
12418 likely that the different parallel compilers will interfere with each
12419 other, and overwrite the temporary files. For instance:
12422 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
12423 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
12426 may result in @file{foo.i} and @file{foo.o} being written to
12427 simultaneously by both compilers.
12429 @item -save-temps=obj
12430 @opindex save-temps=obj
12431 Store the usual ``temporary'' intermediate files permanently. If the
12432 @option{-o} option is used, the temporary files are based on the
12433 object file. If the @option{-o} option is not used, the
12434 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
12439 gcc -save-temps=obj -c foo.c
12440 gcc -save-temps=obj -c bar.c -o dir/xbar.o
12441 gcc -save-temps=obj foobar.c -o dir2/yfoobar
12445 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
12446 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
12447 @file{dir2/yfoobar.o}.
12449 @item -time@r{[}=@var{file}@r{]}
12451 Report the CPU time taken by each subprocess in the compilation
12452 sequence. For C source files, this is the compiler proper and assembler
12453 (plus the linker if linking is done).
12455 Without the specification of an output file, the output looks like this:
12462 The first number on each line is the ``user time'', that is time spent
12463 executing the program itself. The second number is ``system time'',
12464 time spent executing operating system routines on behalf of the program.
12465 Both numbers are in seconds.
12467 With the specification of an output file, the output is appended to the
12468 named file, and it looks like this:
12471 0.12 0.01 cc1 @var{options}
12472 0.00 0.01 as @var{options}
12475 The ``user time'' and the ``system time'' are moved before the program
12476 name, and the options passed to the program are displayed, so that one
12477 can later tell what file was being compiled, and with which options.
12479 @item -fdump-final-insns@r{[}=@var{file}@r{]}
12480 @opindex fdump-final-insns
12481 Dump the final internal representation (RTL) to @var{file}. If the
12482 optional argument is omitted (or if @var{file} is @code{.}), the name
12483 of the dump file is determined by appending @code{.gkd} to the
12484 compilation output file name.
12486 @item -fcompare-debug@r{[}=@var{opts}@r{]}
12487 @opindex fcompare-debug
12488 @opindex fno-compare-debug
12489 If no error occurs during compilation, run the compiler a second time,
12490 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
12491 passed to the second compilation. Dump the final internal
12492 representation in both compilations, and print an error if they differ.
12494 If the equal sign is omitted, the default @option{-gtoggle} is used.
12496 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
12497 and nonzero, implicitly enables @option{-fcompare-debug}. If
12498 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
12499 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
12502 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
12503 is equivalent to @option{-fno-compare-debug}, which disables the dumping
12504 of the final representation and the second compilation, preventing even
12505 @env{GCC_COMPARE_DEBUG} from taking effect.
12507 To verify full coverage during @option{-fcompare-debug} testing, set
12508 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
12509 which GCC rejects as an invalid option in any actual compilation
12510 (rather than preprocessing, assembly or linking). To get just a
12511 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
12512 not overridden} will do.
12514 @item -fcompare-debug-second
12515 @opindex fcompare-debug-second
12516 This option is implicitly passed to the compiler for the second
12517 compilation requested by @option{-fcompare-debug}, along with options to
12518 silence warnings, and omitting other options that would cause
12519 side-effect compiler outputs to files or to the standard output. Dump
12520 files and preserved temporary files are renamed so as to contain the
12521 @code{.gk} additional extension during the second compilation, to avoid
12522 overwriting those generated by the first.
12524 When this option is passed to the compiler driver, it causes the
12525 @emph{first} compilation to be skipped, which makes it useful for little
12526 other than debugging the compiler proper.
12530 Turn off generation of debug info, if leaving out this option
12531 generates it, or turn it on at level 2 otherwise. The position of this
12532 argument in the command line does not matter; it takes effect after all
12533 other options are processed, and it does so only once, no matter how
12534 many times it is given. This is mainly intended to be used with
12535 @option{-fcompare-debug}.
12537 @item -fvar-tracking-assignments-toggle
12538 @opindex fvar-tracking-assignments-toggle
12539 @opindex fno-var-tracking-assignments-toggle
12540 Toggle @option{-fvar-tracking-assignments}, in the same way that
12541 @option{-gtoggle} toggles @option{-g}.
12545 Makes the compiler print out each function name as it is compiled, and
12546 print some statistics about each pass when it finishes.
12548 @item -ftime-report
12549 @opindex ftime-report
12550 Makes the compiler print some statistics about the time consumed by each
12551 pass when it finishes.
12553 @item -fira-verbose=@var{n}
12554 @opindex fira-verbose
12555 Control the verbosity of the dump file for the integrated register allocator.
12556 The default value is 5. If the value @var{n} is greater or equal to 10,
12557 the dump output is sent to stderr using the same format as @var{n} minus 10.
12560 @opindex flto-report
12561 Prints a report with internal details on the workings of the link-time
12562 optimizer. The contents of this report vary from version to version.
12563 It is meant to be useful to GCC developers when processing object
12564 files in LTO mode (via @option{-flto}).
12566 Disabled by default.
12568 @item -flto-report-wpa
12569 @opindex flto-report-wpa
12570 Like @option{-flto-report}, but only print for the WPA phase of Link
12574 @opindex fmem-report
12575 Makes the compiler print some statistics about permanent memory
12576 allocation when it finishes.
12578 @item -fmem-report-wpa
12579 @opindex fmem-report-wpa
12580 Makes the compiler print some statistics about permanent memory
12581 allocation for the WPA phase only.
12583 @item -fpre-ipa-mem-report
12584 @opindex fpre-ipa-mem-report
12585 @item -fpost-ipa-mem-report
12586 @opindex fpost-ipa-mem-report
12587 Makes the compiler print some statistics about permanent memory
12588 allocation before or after interprocedural optimization.
12590 @item -fprofile-report
12591 @opindex fprofile-report
12592 Makes the compiler print some statistics about consistency of the
12593 (estimated) profile and effect of individual passes.
12595 @item -fstack-usage
12596 @opindex fstack-usage
12597 Makes the compiler output stack usage information for the program, on a
12598 per-function basis. The filename for the dump is made by appending
12599 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
12600 the output file, if explicitly specified and it is not an executable,
12601 otherwise it is the basename of the source file. An entry is made up
12606 The name of the function.
12610 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
12613 The qualifier @code{static} means that the function manipulates the stack
12614 statically: a fixed number of bytes are allocated for the frame on function
12615 entry and released on function exit; no stack adjustments are otherwise made
12616 in the function. The second field is this fixed number of bytes.
12618 The qualifier @code{dynamic} means that the function manipulates the stack
12619 dynamically: in addition to the static allocation described above, stack
12620 adjustments are made in the body of the function, for example to push/pop
12621 arguments around function calls. If the qualifier @code{bounded} is also
12622 present, the amount of these adjustments is bounded at compile time and
12623 the second field is an upper bound of the total amount of stack used by
12624 the function. If it is not present, the amount of these adjustments is
12625 not bounded at compile time and the second field only represents the
12630 Emit statistics about front-end processing at the end of the compilation.
12631 This option is supported only by the C++ front end, and
12632 the information is generally only useful to the G++ development team.
12634 @item -fdbg-cnt-list
12635 @opindex fdbg-cnt-list
12636 Print the name and the counter upper bound for all debug counters.
12639 @item -fdbg-cnt=@var{counter-value-list}
12641 Set the internal debug counter upper bound. @var{counter-value-list}
12642 is a comma-separated list of @var{name}:@var{value} pairs
12643 which sets the upper bound of each debug counter @var{name} to @var{value}.
12644 All debug counters have the initial upper bound of @code{UINT_MAX};
12645 thus @code{dbg_cnt} returns true always unless the upper bound
12646 is set by this option.
12647 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
12648 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
12650 @item -print-file-name=@var{library}
12651 @opindex print-file-name
12652 Print the full absolute name of the library file @var{library} that
12653 would be used when linking---and don't do anything else. With this
12654 option, GCC does not compile or link anything; it just prints the
12657 @item -print-multi-directory
12658 @opindex print-multi-directory
12659 Print the directory name corresponding to the multilib selected by any
12660 other switches present in the command line. This directory is supposed
12661 to exist in @env{GCC_EXEC_PREFIX}.
12663 @item -print-multi-lib
12664 @opindex print-multi-lib
12665 Print the mapping from multilib directory names to compiler switches
12666 that enable them. The directory name is separated from the switches by
12667 @samp{;}, and each switch starts with an @samp{@@} instead of the
12668 @samp{-}, without spaces between multiple switches. This is supposed to
12669 ease shell processing.
12671 @item -print-multi-os-directory
12672 @opindex print-multi-os-directory
12673 Print the path to OS libraries for the selected
12674 multilib, relative to some @file{lib} subdirectory. If OS libraries are
12675 present in the @file{lib} subdirectory and no multilibs are used, this is
12676 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
12677 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
12678 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
12679 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
12681 @item -print-multiarch
12682 @opindex print-multiarch
12683 Print the path to OS libraries for the selected multiarch,
12684 relative to some @file{lib} subdirectory.
12686 @item -print-prog-name=@var{program}
12687 @opindex print-prog-name
12688 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
12690 @item -print-libgcc-file-name
12691 @opindex print-libgcc-file-name
12692 Same as @option{-print-file-name=libgcc.a}.
12694 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
12695 but you do want to link with @file{libgcc.a}. You can do:
12698 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
12701 @item -print-search-dirs
12702 @opindex print-search-dirs
12703 Print the name of the configured installation directory and a list of
12704 program and library directories @command{gcc} searches---and don't do anything else.
12706 This is useful when @command{gcc} prints the error message
12707 @samp{installation problem, cannot exec cpp0: No such file or directory}.
12708 To resolve this you either need to put @file{cpp0} and the other compiler
12709 components where @command{gcc} expects to find them, or you can set the environment
12710 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
12711 Don't forget the trailing @samp{/}.
12712 @xref{Environment Variables}.
12714 @item -print-sysroot
12715 @opindex print-sysroot
12716 Print the target sysroot directory that is used during
12717 compilation. This is the target sysroot specified either at configure
12718 time or using the @option{--sysroot} option, possibly with an extra
12719 suffix that depends on compilation options. If no target sysroot is
12720 specified, the option prints nothing.
12722 @item -print-sysroot-headers-suffix
12723 @opindex print-sysroot-headers-suffix
12724 Print the suffix added to the target sysroot when searching for
12725 headers, or give an error if the compiler is not configured with such
12726 a suffix---and don't do anything else.
12729 @opindex dumpmachine
12730 Print the compiler's target machine (for example,
12731 @samp{i686-pc-linux-gnu})---and don't do anything else.
12734 @opindex dumpversion
12735 Print the compiler version (for example, @code{3.0})---and don't do
12740 Print the compiler's built-in specs---and don't do anything else. (This
12741 is used when GCC itself is being built.) @xref{Spec Files}.
12744 @node Submodel Options
12745 @section Machine-Dependent Options
12746 @cindex submodel options
12747 @cindex specifying hardware config
12748 @cindex hardware models and configurations, specifying
12749 @cindex target-dependent options
12750 @cindex machine-dependent options
12752 Each target machine supported by GCC can have its own options---for
12753 example, to allow you to compile for a particular processor variant or
12754 ABI, or to control optimizations specific to that machine. By
12755 convention, the names of machine-specific options start with
12758 Some configurations of the compiler also support additional target-specific
12759 options, usually for compatibility with other compilers on the same
12762 @c This list is ordered alphanumerically by subsection name.
12763 @c It should be the same order and spelling as these options are listed
12764 @c in Machine Dependent Options
12767 * AArch64 Options::
12768 * Adapteva Epiphany Options::
12772 * Blackfin Options::
12777 * DEC Alpha Options::
12781 * GNU/Linux Options::
12791 * MicroBlaze Options::
12794 * MN10300 Options::
12798 * Nios II Options::
12799 * Nvidia PTX Options::
12801 * picoChip Options::
12802 * PowerPC Options::
12804 * RS/6000 and PowerPC Options::
12806 * S/390 and zSeries Options::
12809 * Solaris 2 Options::
12812 * System V Options::
12813 * TILE-Gx Options::
12814 * TILEPro Options::
12819 * VxWorks Options::
12821 * x86 Windows Options::
12822 * Xstormy16 Options::
12824 * zSeries Options::
12827 @node AArch64 Options
12828 @subsection AArch64 Options
12829 @cindex AArch64 Options
12831 These options are defined for AArch64 implementations:
12835 @item -mabi=@var{name}
12837 Generate code for the specified data model. Permissible values
12838 are @samp{ilp32} for SysV-like data model where int, long int and pointers
12839 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
12840 but long int and pointers are 64 bits.
12842 The default depends on the specific target configuration. Note that
12843 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12844 entire program with the same ABI, and link with a compatible set of libraries.
12847 @opindex mbig-endian
12848 Generate big-endian code. This is the default when GCC is configured for an
12849 @samp{aarch64_be-*-*} target.
12851 @item -mgeneral-regs-only
12852 @opindex mgeneral-regs-only
12853 Generate code which uses only the general-purpose registers. This will prevent
12854 the compiler from using floating-point and Advanced SIMD registers but will not
12855 impose any restrictions on the assembler.
12857 @item -mlittle-endian
12858 @opindex mlittle-endian
12859 Generate little-endian code. This is the default when GCC is configured for an
12860 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12862 @item -mcmodel=tiny
12863 @opindex mcmodel=tiny
12864 Generate code for the tiny code model. The program and its statically defined
12865 symbols must be within 1MB of each other. Programs can be statically or
12866 dynamically linked.
12868 @item -mcmodel=small
12869 @opindex mcmodel=small
12870 Generate code for the small code model. The program and its statically defined
12871 symbols must be within 4GB of each other. Programs can be statically or
12872 dynamically linked. This is the default code model.
12874 @item -mcmodel=large
12875 @opindex mcmodel=large
12876 Generate code for the large code model. This makes no assumptions about
12877 addresses and sizes of sections. Programs can be statically linked only.
12879 @item -mstrict-align
12880 @opindex mstrict-align
12881 Avoid generating memory accesses that may not be aligned on a natural object
12882 boundary as described in the architecture specification.
12884 @item -momit-leaf-frame-pointer
12885 @itemx -mno-omit-leaf-frame-pointer
12886 @opindex momit-leaf-frame-pointer
12887 @opindex mno-omit-leaf-frame-pointer
12888 Omit or keep the frame pointer in leaf functions. The former behavior is the
12891 @item -mtls-dialect=desc
12892 @opindex mtls-dialect=desc
12893 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12894 of TLS variables. This is the default.
12896 @item -mtls-dialect=traditional
12897 @opindex mtls-dialect=traditional
12898 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12901 @item -mtls-size=@var{size}
12903 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
12904 This option requires binutils 2.26 or newer.
12906 @item -mfix-cortex-a53-835769
12907 @itemx -mno-fix-cortex-a53-835769
12908 @opindex mfix-cortex-a53-835769
12909 @opindex mno-fix-cortex-a53-835769
12910 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12911 This involves inserting a NOP instruction between memory instructions and
12912 64-bit integer multiply-accumulate instructions.
12914 @item -mfix-cortex-a53-843419
12915 @itemx -mno-fix-cortex-a53-843419
12916 @opindex mfix-cortex-a53-843419
12917 @opindex mno-fix-cortex-a53-843419
12918 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
12919 This erratum workaround is made at link time and this will only pass the
12920 corresponding flag to the linker.
12922 @item -mlow-precision-recip-sqrt
12923 @item -mno-low-precision-recip-sqrt
12924 @opindex mlow-precision-recip-sqrt
12925 @opindex mno-low-precision-recip-sqrt
12926 Enable or disable the reciprocal square root approximation.
12927 This option only has an effect if @option{-ffast-math} or
12928 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
12929 precision of reciprocal square root results to about 16 bits for
12930 single precision and to 32 bits for double precision.
12932 @item -mlow-precision-sqrt
12933 @item -mno-low-precision-sqrt
12934 @opindex -mlow-precision-sqrt
12935 @opindex -mno-low-precision-sqrt
12936 Enable or disable the square root approximation.
12937 This option only has an effect if @option{-ffast-math} or
12938 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
12939 precision of square root results to about 16 bits for
12940 single precision and to 32 bits for double precision.
12941 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
12943 @item -mlow-precision-div
12944 @item -mno-low-precision-div
12945 @opindex -mlow-precision-div
12946 @opindex -mno-low-precision-div
12947 Enable or disable the division approximation.
12948 This option only has an effect if @option{-ffast-math} or
12949 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
12950 precision of division results to about 16 bits for
12951 single precision and to 32 bits for double precision.
12953 @item -march=@var{name}
12955 Specify the name of the target architecture and, optionally, one or
12956 more feature modifiers. This option has the form
12957 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
12959 The permissible values for @var{arch} are @samp{armv8-a},
12960 @samp{armv8.1-a}, @samp{armv8.2-a} or @var{native}.
12962 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
12963 support for the ARMv8.2-A architecture extensions.
12965 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
12966 support for the ARMv8.1-A architecture extension. In particular, it
12967 enables the @samp{+crc} and @samp{+lse} features.
12969 The value @samp{native} is available on native AArch64 GNU/Linux and
12970 causes the compiler to pick the architecture of the host system. This
12971 option has no effect if the compiler is unable to recognize the
12972 architecture of the host system,
12974 The permissible values for @var{feature} are listed in the sub-section
12975 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12976 Feature Modifiers}. Where conflicting feature modifiers are
12977 specified, the right-most feature is used.
12979 GCC uses @var{name} to determine what kind of instructions it can emit
12980 when generating assembly code. If @option{-march} is specified
12981 without either of @option{-mtune} or @option{-mcpu} also being
12982 specified, the code is tuned to perform well across a range of target
12983 processors implementing the target architecture.
12985 @item -mtune=@var{name}
12987 Specify the name of the target processor for which GCC should tune the
12988 performance of the code. Permissible values for this option are:
12989 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
12990 @samp{cortex-a72}, @samp{cortex-a73}, @samp{exynos-m1}, @samp{qdf24xx},
12991 @samp{thunderx}, @samp{xgene1}, @samp{vulcan}, @samp{cortex-a57.cortex-a53},
12992 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
12993 @samp{cortex-a73.cortex-a53}, @samp{native}.
12995 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
12996 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}
12997 specify that GCC should tune for a big.LITTLE system.
12999 Additionally on native AArch64 GNU/Linux systems the value
13000 @samp{native} tunes performance to the host system. This option has no effect
13001 if the compiler is unable to recognize the processor of the host system.
13003 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
13004 are specified, the code is tuned to perform well across a range
13005 of target processors.
13007 This option cannot be suffixed by feature modifiers.
13009 @item -mcpu=@var{name}
13011 Specify the name of the target processor, optionally suffixed by one
13012 or more feature modifiers. This option has the form
13013 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
13014 the permissible values for @var{cpu} are the same as those available
13015 for @option{-mtune}. The permissible values for @var{feature} are
13016 documented in the sub-section on
13017 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13018 Feature Modifiers}. Where conflicting feature modifiers are
13019 specified, the right-most feature is used.
13021 GCC uses @var{name} to determine what kind of instructions it can emit when
13022 generating assembly code (as if by @option{-march}) and to determine
13023 the target processor for which to tune for performance (as if
13024 by @option{-mtune}). Where this option is used in conjunction
13025 with @option{-march} or @option{-mtune}, those options take precedence
13026 over the appropriate part of this option.
13028 @item -moverride=@var{string}
13030 Override tuning decisions made by the back-end in response to a
13031 @option{-mtune=} switch. The syntax, semantics, and accepted values
13032 for @var{string} in this option are not guaranteed to be consistent
13035 This option is only intended to be useful when developing GCC.
13037 @item -mpc-relative-literal-loads
13038 @opindex mpc-relative-literal-loads
13039 Enable PC-relative literal loads. With this option literal pools are
13040 accessed using a single instruction and emitted after each function. This
13041 limits the maximum size of functions to 1MB. This is enabled by default for
13042 @option{-mcmodel=tiny}.
13046 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
13047 @anchor{aarch64-feature-modifiers}
13048 @cindex @option{-march} feature modifiers
13049 @cindex @option{-mcpu} feature modifiers
13050 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
13051 the following and their inverses @option{no@var{feature}}:
13055 Enable CRC extension. This is on by default for
13056 @option{-march=armv8.1-a}.
13058 Enable Crypto extension. This also enables Advanced SIMD and floating-point
13061 Enable floating-point instructions. This is on by default for all possible
13062 values for options @option{-march} and @option{-mcpu}.
13064 Enable Advanced SIMD instructions. This also enables floating-point
13065 instructions. This is on by default for all possible values for options
13066 @option{-march} and @option{-mcpu}.
13068 Enable Large System Extension instructions. This is on by default for
13069 @option{-march=armv8.1-a}.
13071 Enable FP16 extension. This also enables floating-point instructions.
13075 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
13076 Conversely, @option{nofp} implies @option{nosimd}, which implies
13079 @node Adapteva Epiphany Options
13080 @subsection Adapteva Epiphany Options
13082 These @samp{-m} options are defined for Adapteva Epiphany:
13085 @item -mhalf-reg-file
13086 @opindex mhalf-reg-file
13087 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
13088 That allows code to run on hardware variants that lack these registers.
13090 @item -mprefer-short-insn-regs
13091 @opindex mprefer-short-insn-regs
13092 Preferentially allocate registers that allow short instruction generation.
13093 This can result in increased instruction count, so this may either reduce or
13094 increase overall code size.
13096 @item -mbranch-cost=@var{num}
13097 @opindex mbranch-cost
13098 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13099 This cost is only a heuristic and is not guaranteed to produce
13100 consistent results across releases.
13104 Enable the generation of conditional moves.
13106 @item -mnops=@var{num}
13108 Emit @var{num} NOPs before every other generated instruction.
13110 @item -mno-soft-cmpsf
13111 @opindex mno-soft-cmpsf
13112 For single-precision floating-point comparisons, emit an @code{fsub} instruction
13113 and test the flags. This is faster than a software comparison, but can
13114 get incorrect results in the presence of NaNs, or when two different small
13115 numbers are compared such that their difference is calculated as zero.
13116 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
13117 software comparisons.
13119 @item -mstack-offset=@var{num}
13120 @opindex mstack-offset
13121 Set the offset between the top of the stack and the stack pointer.
13122 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
13123 can be used by leaf functions without stack allocation.
13124 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
13125 Note also that this option changes the ABI; compiling a program with a
13126 different stack offset than the libraries have been compiled with
13127 generally does not work.
13128 This option can be useful if you want to evaluate if a different stack
13129 offset would give you better code, but to actually use a different stack
13130 offset to build working programs, it is recommended to configure the
13131 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
13133 @item -mno-round-nearest
13134 @opindex mno-round-nearest
13135 Make the scheduler assume that the rounding mode has been set to
13136 truncating. The default is @option{-mround-nearest}.
13139 @opindex mlong-calls
13140 If not otherwise specified by an attribute, assume all calls might be beyond
13141 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
13142 function address into a register before performing a (otherwise direct) call.
13143 This is the default.
13145 @item -mshort-calls
13146 @opindex short-calls
13147 If not otherwise specified by an attribute, assume all direct calls are
13148 in the range of the @code{b} / @code{bl} instructions, so use these instructions
13149 for direct calls. The default is @option{-mlong-calls}.
13153 Assume addresses can be loaded as 16-bit unsigned values. This does not
13154 apply to function addresses for which @option{-mlong-calls} semantics
13157 @item -mfp-mode=@var{mode}
13159 Set the prevailing mode of the floating-point unit.
13160 This determines the floating-point mode that is provided and expected
13161 at function call and return time. Making this mode match the mode you
13162 predominantly need at function start can make your programs smaller and
13163 faster by avoiding unnecessary mode switches.
13165 @var{mode} can be set to one the following values:
13169 Any mode at function entry is valid, and retained or restored when
13170 the function returns, and when it calls other functions.
13171 This mode is useful for compiling libraries or other compilation units
13172 you might want to incorporate into different programs with different
13173 prevailing FPU modes, and the convenience of being able to use a single
13174 object file outweighs the size and speed overhead for any extra
13175 mode switching that might be needed, compared with what would be needed
13176 with a more specific choice of prevailing FPU mode.
13179 This is the mode used for floating-point calculations with
13180 truncating (i.e.@: round towards zero) rounding mode. That includes
13181 conversion from floating point to integer.
13183 @item round-nearest
13184 This is the mode used for floating-point calculations with
13185 round-to-nearest-or-even rounding mode.
13188 This is the mode used to perform integer calculations in the FPU, e.g.@:
13189 integer multiply, or integer multiply-and-accumulate.
13192 The default is @option{-mfp-mode=caller}
13194 @item -mnosplit-lohi
13195 @itemx -mno-postinc
13196 @itemx -mno-postmodify
13197 @opindex mnosplit-lohi
13198 @opindex mno-postinc
13199 @opindex mno-postmodify
13200 Code generation tweaks that disable, respectively, splitting of 32-bit
13201 loads, generation of post-increment addresses, and generation of
13202 post-modify addresses. The defaults are @option{msplit-lohi},
13203 @option{-mpost-inc}, and @option{-mpost-modify}.
13205 @item -mnovect-double
13206 @opindex mno-vect-double
13207 Change the preferred SIMD mode to SImode. The default is
13208 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
13210 @item -max-vect-align=@var{num}
13211 @opindex max-vect-align
13212 The maximum alignment for SIMD vector mode types.
13213 @var{num} may be 4 or 8. The default is 8.
13214 Note that this is an ABI change, even though many library function
13215 interfaces are unaffected if they don't use SIMD vector modes
13216 in places that affect size and/or alignment of relevant types.
13218 @item -msplit-vecmove-early
13219 @opindex msplit-vecmove-early
13220 Split vector moves into single word moves before reload. In theory this
13221 can give better register allocation, but so far the reverse seems to be
13222 generally the case.
13224 @item -m1reg-@var{reg}
13226 Specify a register to hold the constant @minus{}1, which makes loading small negative
13227 constants and certain bitmasks faster.
13228 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
13229 which specify use of that register as a fixed register,
13230 and @samp{none}, which means that no register is used for this
13231 purpose. The default is @option{-m1reg-none}.
13236 @subsection ARC Options
13237 @cindex ARC options
13239 The following options control the architecture variant for which code
13242 @c architecture variants
13245 @item -mbarrel-shifter
13246 @opindex mbarrel-shifter
13247 Generate instructions supported by barrel shifter. This is the default
13248 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
13250 @item -mcpu=@var{cpu}
13252 Set architecture type, register usage, and instruction scheduling
13253 parameters for @var{cpu}. There are also shortcut alias options
13254 available for backward compatibility and convenience. Supported
13255 values for @var{cpu} are
13262 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
13267 Compile for ARC601. Alias: @option{-mARC601}.
13273 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
13274 This is the default when configured with @option{--with-cpu=arc700}@.
13278 Compile for ARC EM.
13282 Compile for ARC HS.
13287 @itemx -mdpfp-compact
13288 @opindex mdpfp-compact
13289 FPX: Generate Double Precision FPX instructions, tuned for the compact
13293 @opindex mdpfp-fast
13294 FPX: Generate Double Precision FPX instructions, tuned for the fast
13297 @item -mno-dpfp-lrsr
13298 @opindex mno-dpfp-lrsr
13299 Disable LR and SR instructions from using FPX extension aux registers.
13303 Generate Extended arithmetic instructions. Currently only
13304 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
13305 supported. This is always enabled for @option{-mcpu=ARC700}.
13309 Do not generate mpy instructions for ARC700.
13313 Generate 32x16 bit multiply and mac instructions.
13317 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
13321 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
13326 @itemx -mspfp-compact
13327 @opindex mspfp-compact
13328 FPX: Generate Single Precision FPX instructions, tuned for the compact
13332 @opindex mspfp-fast
13333 FPX: Generate Single Precision FPX instructions, tuned for the fast
13338 Enable generation of ARC SIMD instructions via target-specific
13339 builtins. Only valid for @option{-mcpu=ARC700}.
13342 @opindex msoft-float
13343 This option ignored; it is provided for compatibility purposes only.
13344 Software floating point code is emitted by default, and this default
13345 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
13346 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
13347 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
13351 Generate swap instructions.
13355 This enables Locked Load/Store Conditional extension to implement
13356 atomic memopry built-in functions. Not available for ARC 6xx or ARC
13361 Enable DIV/REM instructions for ARCv2 cores.
13363 @item -mcode-density
13364 @opindex mcode-density
13365 Enable code density instructions for ARC EM, default on for ARC HS.
13369 Enable double load/store operations for ARC HS cores.
13371 @item -mmpy-option=@var{multo}
13372 @opindex mmpy-option
13373 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
13374 the default value. The recognized values for @var{multo} are:
13378 No multiplier available.
13382 The multiply option is set to w: 16x16 multiplier, fully pipelined.
13383 The following instructions are enabled: MPYW, and MPYUW.
13387 The multiply option is set to wlh1: 32x32 multiplier, fully
13388 pipelined (1 stage). The following instructions are additionally
13389 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
13393 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
13394 (2 stages). The following instructions are additionally enabled: MPY,
13395 MPYU, MPYM, MPYMU, and MPY_S.
13399 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
13400 sequential. The following instructions are additionally enabled: MPY,
13401 MPYU, MPYM, MPYMU, and MPY_S.
13405 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
13406 sequential. The following instructions are additionally enabled: MPY,
13407 MPYU, MPYM, MPYMU, and MPY_S.
13411 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
13412 sequential. The following instructions are additionally enabled: MPY,
13413 MPYU, MPYM, MPYMU, and MPY_S.
13417 This option is only available for ARCv2 cores@.
13419 @item -mfpu=@var{fpu}
13421 Enables specific floating-point hardware extension for ARCv2
13422 core. Supported values for @var{fpu} are:
13428 Enables support for single precision floating point hardware
13433 Enables support for double precision floating point hardware
13434 extensions. The single precision floating point extension is also
13435 enabled. Not available for ARC EM@.
13439 Enables support for double precision floating point hardware
13440 extensions using double precision assist instructions. The single
13441 precision floating point extension is also enabled. This option is
13442 only available for ARC EM@.
13446 Enables support for double precision floating point hardware
13447 extensions using double precision assist instructions, and simple
13448 precision square-root and divide hardware extensions. The single
13449 precision floating point extension is also enabled. This option is
13450 only available for ARC EM@.
13454 Enables support for double precision floating point hardware
13455 extensions using double precision assist instructions, and simple
13456 precision fused multiple and add hardware extension. The single
13457 precision floating point extension is also enabled. This option is
13458 only available for ARC EM@.
13462 Enables support for double precision floating point hardware
13463 extensions using double precision assist instructions, and all simple
13464 precision hardware extensions. The single precision floating point
13465 extension is also enabled. This option is only available for ARC EM@.
13469 Enables support for single precision floating point, and single
13470 precision square-root and divide hardware extensions@.
13474 Enables support for double precision floating point, and double
13475 precision square-root and divide hardware extensions. This option
13476 includes option @samp{fpus_div}. Not available for ARC EM@.
13480 Enables support for single precision floating point, and single
13481 precision fused multiple and add hardware extensions@.
13485 Enables support for double precision floating point, and double
13486 precision fused multiple and add hardware extensions. This option
13487 includes option @samp{fpus_fma}. Not available for ARC EM@.
13491 Enables support for all single precision floating point hardware
13496 Enables support for all single and double precision floating point
13497 hardware extensions. Not available for ARC EM@.
13503 The following options are passed through to the assembler, and also
13504 define preprocessor macro symbols.
13506 @c Flags used by the assembler, but for which we define preprocessor
13507 @c macro symbols as well.
13510 @opindex mdsp-packa
13511 Passed down to the assembler to enable the DSP Pack A extensions.
13512 Also sets the preprocessor symbol @code{__Xdsp_packa}.
13516 Passed down to the assembler to enable the dual viterbi butterfly
13517 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
13519 @c ARC700 4.10 extension instruction
13522 Passed down to the assembler to enable the Locked Load/Store
13523 Conditional extension. Also sets the preprocessor symbol
13528 Passed down to the assembler. Also sets the preprocessor symbol
13529 @code{__Xxmac_d16}.
13533 Passed down to the assembler. Also sets the preprocessor symbol
13536 @c ARC700 4.10 extension instruction
13539 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
13540 extension instruction. Also sets the preprocessor symbol
13543 @c ARC700 4.10 extension instruction
13546 Passed down to the assembler to enable the swap byte ordering
13547 extension instruction. Also sets the preprocessor symbol
13551 @opindex mtelephony
13552 Passed down to the assembler to enable dual and single operand
13553 instructions for telephony. Also sets the preprocessor symbol
13554 @code{__Xtelephony}.
13558 Passed down to the assembler to enable the XY Memory extension. Also
13559 sets the preprocessor symbol @code{__Xxy}.
13563 The following options control how the assembly code is annotated:
13565 @c Assembly annotation options
13569 Annotate assembler instructions with estimated addresses.
13571 @item -mannotate-align
13572 @opindex mannotate-align
13573 Explain what alignment considerations lead to the decision to make an
13574 instruction short or long.
13578 The following options are passed through to the linker:
13580 @c options passed through to the linker
13584 Passed through to the linker, to specify use of the @code{arclinux} emulation.
13585 This option is enabled by default in tool chains built for
13586 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
13587 when profiling is not requested.
13589 @item -marclinux_prof
13590 @opindex marclinux_prof
13591 Passed through to the linker, to specify use of the
13592 @code{arclinux_prof} emulation. This option is enabled by default in
13593 tool chains built for @w{@code{arc-linux-uclibc}} and
13594 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
13598 The following options control the semantics of generated code:
13600 @c semantically relevant code generation options
13603 @opindex mlong-calls
13604 Generate call insns as register indirect calls, thus providing access
13605 to the full 32-bit address range.
13607 @item -mmedium-calls
13608 @opindex mmedium-calls
13609 Don't use less than 25 bit addressing range for calls, which is the
13610 offset available for an unconditional branch-and-link
13611 instruction. Conditional execution of function calls is suppressed, to
13612 allow use of the 25-bit range, rather than the 21-bit range with
13613 conditional branch-and-link. This is the default for tool chains built
13614 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
13618 Do not generate sdata references. This is the default for tool chains
13619 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
13623 @opindex mucb-mcount
13624 Instrument with mcount calls as used in UCB code. I.e. do the
13625 counting in the callee, not the caller. By default ARC instrumentation
13626 counts in the caller.
13628 @item -mvolatile-cache
13629 @opindex mvolatile-cache
13630 Use ordinarily cached memory accesses for volatile references. This is the
13633 @item -mno-volatile-cache
13634 @opindex mno-volatile-cache
13635 Enable cache bypass for volatile references.
13639 The following options fine tune code generation:
13640 @c code generation tuning options
13643 @opindex malign-call
13644 Do alignment optimizations for call instructions.
13646 @item -mauto-modify-reg
13647 @opindex mauto-modify-reg
13648 Enable the use of pre/post modify with register displacement.
13650 @item -mbbit-peephole
13651 @opindex mbbit-peephole
13652 Enable bbit peephole2.
13656 This option disables a target-specific pass in @file{arc_reorg} to
13657 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
13658 generation driven by the combiner pass.
13660 @item -mcase-vector-pcrel
13661 @opindex mcase-vector-pcrel
13662 Use pc-relative switch case tables - this enables case table shortening.
13663 This is the default for @option{-Os}.
13665 @item -mcompact-casesi
13666 @opindex mcompact-casesi
13667 Enable compact casesi pattern.
13668 This is the default for @option{-Os}.
13670 @item -mno-cond-exec
13671 @opindex mno-cond-exec
13672 Disable ARCompact specific pass to generate conditional execution instructions.
13673 Due to delay slot scheduling and interactions between operand numbers,
13674 literal sizes, instruction lengths, and the support for conditional execution,
13675 the target-independent pass to generate conditional execution is often lacking,
13676 so the ARC port has kept a special pass around that tries to find more
13677 conditional execution generating opportunities after register allocation,
13678 branch shortening, and delay slot scheduling have been done. This pass
13679 generally, but not always, improves performance and code size, at the cost of
13680 extra compilation time, which is why there is an option to switch it off.
13681 If you have a problem with call instructions exceeding their allowable
13682 offset range because they are conditionalized, you should consider using
13683 @option{-mmedium-calls} instead.
13685 @item -mearly-cbranchsi
13686 @opindex mearly-cbranchsi
13687 Enable pre-reload use of the cbranchsi pattern.
13689 @item -mexpand-adddi
13690 @opindex mexpand-adddi
13691 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
13692 @code{add.f}, @code{adc} etc.
13694 @item -mindexed-loads
13695 @opindex mindexed-loads
13696 Enable the use of indexed loads. This can be problematic because some
13697 optimizers then assume that indexed stores exist, which is not
13701 Enable Local Register Allocation. This is still experimental for ARC,
13702 so by default the compiler uses standard reload
13703 (i.e. @option{-mno-lra}).
13705 @item -mlra-priority-none
13706 @opindex mlra-priority-none
13707 Don't indicate any priority for target registers.
13709 @item -mlra-priority-compact
13710 @opindex mlra-priority-compact
13711 Indicate target register priority for r0..r3 / r12..r15.
13713 @item -mlra-priority-noncompact
13714 @opindex mlra-priority-noncompact
13715 Reduce target register priority for r0..r3 / r12..r15.
13717 @item -mno-millicode
13718 @opindex mno-millicode
13719 When optimizing for size (using @option{-Os}), prologues and epilogues
13720 that have to save or restore a large number of registers are often
13721 shortened by using call to a special function in libgcc; this is
13722 referred to as a @emph{millicode} call. As these calls can pose
13723 performance issues, and/or cause linking issues when linking in a
13724 nonstandard way, this option is provided to turn off millicode call
13728 @opindex mmixed-code
13729 Tweak register allocation to help 16-bit instruction generation.
13730 This generally has the effect of decreasing the average instruction size
13731 while increasing the instruction count.
13735 Enable 'q' instruction alternatives.
13736 This is the default for @option{-Os}.
13740 Enable Rcq constraint handling - most short code generation depends on this.
13741 This is the default.
13745 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13746 This is the default.
13748 @item -msize-level=@var{level}
13749 @opindex msize-level
13750 Fine-tune size optimization with regards to instruction lengths and alignment.
13751 The recognized values for @var{level} are:
13754 No size optimization. This level is deprecated and treated like @samp{1}.
13757 Short instructions are used opportunistically.
13760 In addition, alignment of loops and of code after barriers are dropped.
13763 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13767 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13768 the behavior when this is not set is equivalent to level @samp{1}.
13770 @item -mtune=@var{cpu}
13772 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13773 by @option{-mcpu=}.
13775 Supported values for @var{cpu} are
13779 Tune for ARC600 cpu.
13782 Tune for ARC601 cpu.
13785 Tune for ARC700 cpu with standard multiplier block.
13788 Tune for ARC700 cpu with XMAC block.
13791 Tune for ARC725D cpu.
13794 Tune for ARC750D cpu.
13798 @item -mmultcost=@var{num}
13800 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13801 normal instruction.
13803 @item -munalign-prob-threshold=@var{probability}
13804 @opindex munalign-prob-threshold
13805 Set probability threshold for unaligning branches.
13806 When tuning for @samp{ARC700} and optimizing for speed, branches without
13807 filled delay slot are preferably emitted unaligned and long, unless
13808 profiling indicates that the probability for the branch to be taken
13809 is below @var{probability}. @xref{Cross-profiling}.
13810 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13814 The following options are maintained for backward compatibility, but
13815 are now deprecated and will be removed in a future release:
13817 @c Deprecated options
13825 @opindex mbig-endian
13828 Compile code for big endian targets. Use of these options is now
13829 deprecated. Users wanting big-endian code, should use the
13830 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13831 building the tool chain, for which big-endian is the default.
13833 @item -mlittle-endian
13834 @opindex mlittle-endian
13837 Compile code for little endian targets. Use of these options is now
13838 deprecated. Users wanting little-endian code should use the
13839 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13840 building the tool chain, for which little-endian is the default.
13842 @item -mbarrel_shifter
13843 @opindex mbarrel_shifter
13844 Replaced by @option{-mbarrel-shifter}.
13846 @item -mdpfp_compact
13847 @opindex mdpfp_compact
13848 Replaced by @option{-mdpfp-compact}.
13851 @opindex mdpfp_fast
13852 Replaced by @option{-mdpfp-fast}.
13855 @opindex mdsp_packa
13856 Replaced by @option{-mdsp-packa}.
13860 Replaced by @option{-mea}.
13864 Replaced by @option{-mmac-24}.
13868 Replaced by @option{-mmac-d16}.
13870 @item -mspfp_compact
13871 @opindex mspfp_compact
13872 Replaced by @option{-mspfp-compact}.
13875 @opindex mspfp_fast
13876 Replaced by @option{-mspfp-fast}.
13878 @item -mtune=@var{cpu}
13880 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13881 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13882 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13884 @item -multcost=@var{num}
13886 Replaced by @option{-mmultcost}.
13891 @subsection ARM Options
13892 @cindex ARM options
13894 These @samp{-m} options are defined for the ARM port:
13897 @item -mabi=@var{name}
13899 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13900 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13903 @opindex mapcs-frame
13904 Generate a stack frame that is compliant with the ARM Procedure Call
13905 Standard for all functions, even if this is not strictly necessary for
13906 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13907 with this option causes the stack frames not to be generated for
13908 leaf functions. The default is @option{-mno-apcs-frame}.
13909 This option is deprecated.
13913 This is a synonym for @option{-mapcs-frame} and is deprecated.
13916 @c not currently implemented
13917 @item -mapcs-stack-check
13918 @opindex mapcs-stack-check
13919 Generate code to check the amount of stack space available upon entry to
13920 every function (that actually uses some stack space). If there is
13921 insufficient space available then either the function
13922 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13923 called, depending upon the amount of stack space required. The runtime
13924 system is required to provide these functions. The default is
13925 @option{-mno-apcs-stack-check}, since this produces smaller code.
13927 @c not currently implemented
13929 @opindex mapcs-float
13930 Pass floating-point arguments using the floating-point registers. This is
13931 one of the variants of the APCS@. This option is recommended if the
13932 target hardware has a floating-point unit or if a lot of floating-point
13933 arithmetic is going to be performed by the code. The default is
13934 @option{-mno-apcs-float}, since the size of integer-only code is
13935 slightly increased if @option{-mapcs-float} is used.
13937 @c not currently implemented
13938 @item -mapcs-reentrant
13939 @opindex mapcs-reentrant
13940 Generate reentrant, position-independent code. The default is
13941 @option{-mno-apcs-reentrant}.
13944 @item -mthumb-interwork
13945 @opindex mthumb-interwork
13946 Generate code that supports calling between the ARM and Thumb
13947 instruction sets. Without this option, on pre-v5 architectures, the
13948 two instruction sets cannot be reliably used inside one program. The
13949 default is @option{-mno-thumb-interwork}, since slightly larger code
13950 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13951 configurations this option is meaningless.
13953 @item -mno-sched-prolog
13954 @opindex mno-sched-prolog
13955 Prevent the reordering of instructions in the function prologue, or the
13956 merging of those instruction with the instructions in the function's
13957 body. This means that all functions start with a recognizable set
13958 of instructions (or in fact one of a choice from a small set of
13959 different function prologues), and this information can be used to
13960 locate the start of functions inside an executable piece of code. The
13961 default is @option{-msched-prolog}.
13963 @item -mfloat-abi=@var{name}
13964 @opindex mfloat-abi
13965 Specifies which floating-point ABI to use. Permissible values
13966 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13968 Specifying @samp{soft} causes GCC to generate output containing
13969 library calls for floating-point operations.
13970 @samp{softfp} allows the generation of code using hardware floating-point
13971 instructions, but still uses the soft-float calling conventions.
13972 @samp{hard} allows generation of floating-point instructions
13973 and uses FPU-specific calling conventions.
13975 The default depends on the specific target configuration. Note that
13976 the hard-float and soft-float ABIs are not link-compatible; you must
13977 compile your entire program with the same ABI, and link with a
13978 compatible set of libraries.
13980 @item -mlittle-endian
13981 @opindex mlittle-endian
13982 Generate code for a processor running in little-endian mode. This is
13983 the default for all standard configurations.
13986 @opindex mbig-endian
13987 Generate code for a processor running in big-endian mode; the default is
13988 to compile code for a little-endian processor.
13990 @item -march=@var{name}
13992 This specifies the name of the target ARM architecture. GCC uses this
13993 name to determine what kind of instructions it can emit when generating
13994 assembly code. This option can be used in conjunction with or instead
13995 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13996 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13997 @samp{armv5}, @samp{armv5e}, @samp{armv5t}, @samp{armv5te},
13998 @samp{armv6}, @samp{armv6-m}, @samp{armv6j}, @samp{armv6k},
13999 @samp{armv6kz}, @samp{armv6s-m},
14000 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk},
14001 @samp{armv7}, @samp{armv7-a}, @samp{armv7-m}, @samp{armv7-r}, @samp{armv7e-m},
14002 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
14003 @samp{armv8.1-a+crc}, @samp{armv8-m.base}, @samp{armv8-m.main},
14004 @samp{armv8-m.main+dsp}, @samp{iwmmxt}, @samp{iwmmxt2}.
14006 Architecture revisions older than @samp{armv4t} are deprecated.
14008 @option{-march=armv6s-m} is the @samp{armv6-m} architecture with support for
14009 the (now mandatory) SVC instruction.
14011 @option{-march=armv6zk} is an alias for @samp{armv6kz}, existing for backwards
14014 @option{-march=armv7ve} is the @samp{armv7-a} architecture with virtualization
14017 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
14018 architecture together with the optional CRC32 extensions.
14020 @option{-march=armv8.1-a} enables compiler support for the ARMv8.1-A
14021 architecture. This also enables the features provided by
14022 @option{-march=armv8-a+crc}.
14024 @option{-march=armv8.2-a} enables compiler support for the ARMv8.2-A
14025 architecture. This also enables the features provided by
14026 @option{-march=armv8.1-a}.
14028 @option{-march=armv8.2-a+fp16} enables compiler support for the
14029 ARMv8.2-A architecture with the optional FP16 instructions extension.
14030 This also enables the features provided by @option{-march=armv8.1-a}
14031 and implies @option{-mfp16-format=ieee}.
14033 @option{-march=native} causes the compiler to auto-detect the architecture
14034 of the build computer. At present, this feature is only supported on
14035 GNU/Linux, and not all architectures are recognized. If the auto-detect
14036 is unsuccessful the option has no effect.
14038 @item -mtune=@var{name}
14040 This option specifies the name of the target ARM processor for
14041 which GCC should tune the performance of the code.
14042 For some ARM implementations better performance can be obtained by using
14044 Permissible names are: @samp{arm2}, @samp{arm250},
14045 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
14046 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
14047 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
14048 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
14050 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
14051 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
14052 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
14053 @samp{strongarm1110},
14054 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
14055 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
14056 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
14057 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
14058 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
14059 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
14060 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
14061 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
14062 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
14063 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
14064 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-r4},
14065 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
14073 @samp{cortex-m0plus},
14074 @samp{cortex-m1.small-multiply},
14075 @samp{cortex-m0.small-multiply},
14076 @samp{cortex-m0plus.small-multiply},
14079 @samp{marvell-pj4},
14080 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
14081 @samp{fa526}, @samp{fa626},
14082 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
14085 Additionally, this option can specify that GCC should tune the performance
14086 of the code for a big.LITTLE system. Permissible names are:
14087 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
14088 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14089 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}.
14091 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
14092 performance for a blend of processors within architecture @var{arch}.
14093 The aim is to generate code that run well on the current most popular
14094 processors, balancing between optimizations that benefit some CPUs in the
14095 range, and avoiding performance pitfalls of other CPUs. The effects of
14096 this option may change in future GCC versions as CPU models come and go.
14098 @option{-mtune=native} causes the compiler to auto-detect the CPU
14099 of the build computer. At present, this feature is only supported on
14100 GNU/Linux, and not all architectures are recognized. If the auto-detect is
14101 unsuccessful the option has no effect.
14103 @item -mcpu=@var{name}
14105 This specifies the name of the target ARM processor. GCC uses this name
14106 to derive the name of the target ARM architecture (as if specified
14107 by @option{-march}) and the ARM processor type for which to tune for
14108 performance (as if specified by @option{-mtune}). Where this option
14109 is used in conjunction with @option{-march} or @option{-mtune},
14110 those options take precedence over the appropriate part of this option.
14112 Permissible names for this option are the same as those for
14115 @option{-mcpu=generic-@var{arch}} is also permissible, and is
14116 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
14117 See @option{-mtune} for more information.
14119 @option{-mcpu=native} causes the compiler to auto-detect the CPU
14120 of the build computer. At present, this feature is only supported on
14121 GNU/Linux, and not all architectures are recognized. If the auto-detect
14122 is unsuccessful the option has no effect.
14124 @item -mfpu=@var{name}
14126 This specifies what floating-point hardware (or hardware emulation) is
14127 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
14128 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
14129 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
14130 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
14131 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
14132 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
14134 If @option{-msoft-float} is specified this specifies the format of
14135 floating-point values.
14137 If the selected floating-point hardware includes the NEON extension
14138 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
14139 operations are not generated by GCC's auto-vectorization pass unless
14140 @option{-funsafe-math-optimizations} is also specified. This is
14141 because NEON hardware does not fully implement the IEEE 754 standard for
14142 floating-point arithmetic (in particular denormal values are treated as
14143 zero), so the use of NEON instructions may lead to a loss of precision.
14145 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}).
14147 @item -mfp16-format=@var{name}
14148 @opindex mfp16-format
14149 Specify the format of the @code{__fp16} half-precision floating-point type.
14150 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
14151 the default is @samp{none}, in which case the @code{__fp16} type is not
14152 defined. @xref{Half-Precision}, for more information.
14154 @item -mstructure-size-boundary=@var{n}
14155 @opindex mstructure-size-boundary
14156 The sizes of all structures and unions are rounded up to a multiple
14157 of the number of bits set by this option. Permissible values are 8, 32
14158 and 64. The default value varies for different toolchains. For the COFF
14159 targeted toolchain the default value is 8. A value of 64 is only allowed
14160 if the underlying ABI supports it.
14162 Specifying a larger number can produce faster, more efficient code, but
14163 can also increase the size of the program. Different values are potentially
14164 incompatible. Code compiled with one value cannot necessarily expect to
14165 work with code or libraries compiled with another value, if they exchange
14166 information using structures or unions.
14168 @item -mabort-on-noreturn
14169 @opindex mabort-on-noreturn
14170 Generate a call to the function @code{abort} at the end of a
14171 @code{noreturn} function. It is executed if the function tries to
14175 @itemx -mno-long-calls
14176 @opindex mlong-calls
14177 @opindex mno-long-calls
14178 Tells the compiler to perform function calls by first loading the
14179 address of the function into a register and then performing a subroutine
14180 call on this register. This switch is needed if the target function
14181 lies outside of the 64-megabyte addressing range of the offset-based
14182 version of subroutine call instruction.
14184 Even if this switch is enabled, not all function calls are turned
14185 into long calls. The heuristic is that static functions, functions
14186 that have the @code{short_call} attribute, functions that are inside
14187 the scope of a @code{#pragma no_long_calls} directive, and functions whose
14188 definitions have already been compiled within the current compilation
14189 unit are not turned into long calls. The exceptions to this rule are
14190 that weak function definitions, functions with the @code{long_call}
14191 attribute or the @code{section} attribute, and functions that are within
14192 the scope of a @code{#pragma long_calls} directive are always
14193 turned into long calls.
14195 This feature is not enabled by default. Specifying
14196 @option{-mno-long-calls} restores the default behavior, as does
14197 placing the function calls within the scope of a @code{#pragma
14198 long_calls_off} directive. Note these switches have no effect on how
14199 the compiler generates code to handle function calls via function
14202 @item -msingle-pic-base
14203 @opindex msingle-pic-base
14204 Treat the register used for PIC addressing as read-only, rather than
14205 loading it in the prologue for each function. The runtime system is
14206 responsible for initializing this register with an appropriate value
14207 before execution begins.
14209 @item -mpic-register=@var{reg}
14210 @opindex mpic-register
14211 Specify the register to be used for PIC addressing.
14212 For standard PIC base case, the default is any suitable register
14213 determined by compiler. For single PIC base case, the default is
14214 @samp{R9} if target is EABI based or stack-checking is enabled,
14215 otherwise the default is @samp{R10}.
14217 @item -mpic-data-is-text-relative
14218 @opindex mpic-data-is-text-relative
14219 Assume that the displacement between the text and data segments is fixed
14220 at static link time. This permits using PC-relative addressing
14221 operations to access data known to be in the data segment. For
14222 non-VxWorks RTP targets, this option is enabled by default. When
14223 disabled on such targets, it will enable @option{-msingle-pic-base} by
14226 @item -mpoke-function-name
14227 @opindex mpoke-function-name
14228 Write the name of each function into the text section, directly
14229 preceding the function prologue. The generated code is similar to this:
14233 .ascii "arm_poke_function_name", 0
14236 .word 0xff000000 + (t1 - t0)
14237 arm_poke_function_name
14239 stmfd sp!, @{fp, ip, lr, pc@}
14243 When performing a stack backtrace, code can inspect the value of
14244 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
14245 location @code{pc - 12} and the top 8 bits are set, then we know that
14246 there is a function name embedded immediately preceding this location
14247 and has length @code{((pc[-3]) & 0xff000000)}.
14254 Select between generating code that executes in ARM and Thumb
14255 states. The default for most configurations is to generate code
14256 that executes in ARM state, but the default can be changed by
14257 configuring GCC with the @option{--with-mode=}@var{state}
14260 You can also override the ARM and Thumb mode for each function
14261 by using the @code{target("thumb")} and @code{target("arm")} function attributes
14262 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
14265 @opindex mtpcs-frame
14266 Generate a stack frame that is compliant with the Thumb Procedure Call
14267 Standard for all non-leaf functions. (A leaf function is one that does
14268 not call any other functions.) The default is @option{-mno-tpcs-frame}.
14270 @item -mtpcs-leaf-frame
14271 @opindex mtpcs-leaf-frame
14272 Generate a stack frame that is compliant with the Thumb Procedure Call
14273 Standard for all leaf functions. (A leaf function is one that does
14274 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
14276 @item -mcallee-super-interworking
14277 @opindex mcallee-super-interworking
14278 Gives all externally visible functions in the file being compiled an ARM
14279 instruction set header which switches to Thumb mode before executing the
14280 rest of the function. This allows these functions to be called from
14281 non-interworking code. This option is not valid in AAPCS configurations
14282 because interworking is enabled by default.
14284 @item -mcaller-super-interworking
14285 @opindex mcaller-super-interworking
14286 Allows calls via function pointers (including virtual functions) to
14287 execute correctly regardless of whether the target code has been
14288 compiled for interworking or not. There is a small overhead in the cost
14289 of executing a function pointer if this option is enabled. This option
14290 is not valid in AAPCS configurations because interworking is enabled
14293 @item -mtp=@var{name}
14295 Specify the access model for the thread local storage pointer. The valid
14296 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
14297 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
14298 (supported in the arm6k architecture), and @samp{auto}, which uses the
14299 best available method for the selected processor. The default setting is
14302 @item -mtls-dialect=@var{dialect}
14303 @opindex mtls-dialect
14304 Specify the dialect to use for accessing thread local storage. Two
14305 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
14306 @samp{gnu} dialect selects the original GNU scheme for supporting
14307 local and global dynamic TLS models. The @samp{gnu2} dialect
14308 selects the GNU descriptor scheme, which provides better performance
14309 for shared libraries. The GNU descriptor scheme is compatible with
14310 the original scheme, but does require new assembler, linker and
14311 library support. Initial and local exec TLS models are unaffected by
14312 this option and always use the original scheme.
14314 @item -mword-relocations
14315 @opindex mword-relocations
14316 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
14317 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
14318 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
14321 @item -mfix-cortex-m3-ldrd
14322 @opindex mfix-cortex-m3-ldrd
14323 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
14324 with overlapping destination and base registers are used. This option avoids
14325 generating these instructions. This option is enabled by default when
14326 @option{-mcpu=cortex-m3} is specified.
14328 @item -munaligned-access
14329 @itemx -mno-unaligned-access
14330 @opindex munaligned-access
14331 @opindex mno-unaligned-access
14332 Enables (or disables) reading and writing of 16- and 32- bit values
14333 from addresses that are not 16- or 32- bit aligned. By default
14334 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
14335 ARMv8-M Baseline architectures, and enabled for all other
14336 architectures. If unaligned access is not enabled then words in packed
14337 data structures are accessed a byte at a time.
14339 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
14340 generated object file to either true or false, depending upon the
14341 setting of this option. If unaligned access is enabled then the
14342 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
14345 @item -mneon-for-64bits
14346 @opindex mneon-for-64bits
14347 Enables using Neon to handle scalar 64-bits operations. This is
14348 disabled by default since the cost of moving data from core registers
14351 @item -mslow-flash-data
14352 @opindex mslow-flash-data
14353 Assume loading data from flash is slower than fetching instruction.
14354 Therefore literal load is minimized for better performance.
14355 This option is only supported when compiling for ARMv7 M-profile and
14358 @item -masm-syntax-unified
14359 @opindex masm-syntax-unified
14360 Assume inline assembler is using unified asm syntax. The default is
14361 currently off which implies divided syntax. This option has no impact
14362 on Thumb2. However, this may change in future releases of GCC.
14363 Divided syntax should be considered deprecated.
14365 @item -mrestrict-it
14366 @opindex mrestrict-it
14367 Restricts generation of IT blocks to conform to the rules of ARMv8.
14368 IT blocks can only contain a single 16-bit instruction from a select
14369 set of instructions. This option is on by default for ARMv8 Thumb mode.
14371 @item -mprint-tune-info
14372 @opindex mprint-tune-info
14373 Print CPU tuning information as comment in assembler file. This is
14374 an option used only for regression testing of the compiler and not
14375 intended for ordinary use in compiling code. This option is disabled
14380 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
14381 Development Tools Engineering Specification", which can be found on
14382 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
14386 @subsection AVR Options
14387 @cindex AVR Options
14389 These options are defined for AVR implementations:
14392 @item -mmcu=@var{mcu}
14394 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
14396 The default for this option is@tie{}@samp{avr2}.
14398 GCC supports the following AVR devices and ISAs:
14400 @include avr-mmcu.texi
14402 @item -maccumulate-args
14403 @opindex maccumulate-args
14404 Accumulate outgoing function arguments and acquire/release the needed
14405 stack space for outgoing function arguments once in function
14406 prologue/epilogue. Without this option, outgoing arguments are pushed
14407 before calling a function and popped afterwards.
14409 Popping the arguments after the function call can be expensive on
14410 AVR so that accumulating the stack space might lead to smaller
14411 executables because arguments need not to be removed from the
14412 stack after such a function call.
14414 This option can lead to reduced code size for functions that perform
14415 several calls to functions that get their arguments on the stack like
14416 calls to printf-like functions.
14418 @item -mbranch-cost=@var{cost}
14419 @opindex mbranch-cost
14420 Set the branch costs for conditional branch instructions to
14421 @var{cost}. Reasonable values for @var{cost} are small, non-negative
14422 integers. The default branch cost is 0.
14424 @item -mcall-prologues
14425 @opindex mcall-prologues
14426 Functions prologues/epilogues are expanded as calls to appropriate
14427 subroutines. Code size is smaller.
14431 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
14432 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
14433 and @code{long long} is 4 bytes. Please note that this option does not
14434 conform to the C standards, but it results in smaller code
14437 @item -mn-flash=@var{num}
14439 Assume that the flash memory has a size of
14440 @var{num} times 64@tie{}KiB.
14442 @item -mno-interrupts
14443 @opindex mno-interrupts
14444 Generated code is not compatible with hardware interrupts.
14445 Code size is smaller.
14449 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
14450 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
14451 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
14452 the assembler's command line and the @option{--relax} option to the
14453 linker's command line.
14455 Jump relaxing is performed by the linker because jump offsets are not
14456 known before code is located. Therefore, the assembler code generated by the
14457 compiler is the same, but the instructions in the executable may
14458 differ from instructions in the assembler code.
14460 Relaxing must be turned on if linker stubs are needed, see the
14461 section on @code{EIND} and linker stubs below.
14465 Assume that the device supports the Read-Modify-Write
14466 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
14470 Treat the stack pointer register as an 8-bit register,
14471 i.e.@: assume the high byte of the stack pointer is zero.
14472 In general, you don't need to set this option by hand.
14474 This option is used internally by the compiler to select and
14475 build multilibs for architectures @code{avr2} and @code{avr25}.
14476 These architectures mix devices with and without @code{SPH}.
14477 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
14478 the compiler driver adds or removes this option from the compiler
14479 proper's command line, because the compiler then knows if the device
14480 or architecture has an 8-bit stack pointer and thus no @code{SPH}
14485 Use address register @code{X} in a way proposed by the hardware. This means
14486 that @code{X} is only used in indirect, post-increment or
14487 pre-decrement addressing.
14489 Without this option, the @code{X} register may be used in the same way
14490 as @code{Y} or @code{Z} which then is emulated by additional
14492 For example, loading a value with @code{X+const} addressing with a
14493 small non-negative @code{const < 64} to a register @var{Rn} is
14497 adiw r26, const ; X += const
14498 ld @var{Rn}, X ; @var{Rn} = *X
14499 sbiw r26, const ; X -= const
14503 @opindex mtiny-stack
14504 Only change the lower 8@tie{}bits of the stack pointer.
14507 @opindex nodevicelib
14508 Don't link against AVR-LibC's device specific library @code{libdev.a}.
14510 @item -Waddr-space-convert
14511 @opindex Waddr-space-convert
14512 Warn about conversions between address spaces in the case where the
14513 resulting address space is not contained in the incoming address space.
14516 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
14517 @cindex @code{EIND}
14518 Pointers in the implementation are 16@tie{}bits wide.
14519 The address of a function or label is represented as word address so
14520 that indirect jumps and calls can target any code address in the
14521 range of 64@tie{}Ki words.
14523 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
14524 bytes of program memory space, there is a special function register called
14525 @code{EIND} that serves as most significant part of the target address
14526 when @code{EICALL} or @code{EIJMP} instructions are used.
14528 Indirect jumps and calls on these devices are handled as follows by
14529 the compiler and are subject to some limitations:
14534 The compiler never sets @code{EIND}.
14537 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
14538 instructions or might read @code{EIND} directly in order to emulate an
14539 indirect call/jump by means of a @code{RET} instruction.
14542 The compiler assumes that @code{EIND} never changes during the startup
14543 code or during the application. In particular, @code{EIND} is not
14544 saved/restored in function or interrupt service routine
14548 For indirect calls to functions and computed goto, the linker
14549 generates @emph{stubs}. Stubs are jump pads sometimes also called
14550 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
14551 The stub contains a direct jump to the desired address.
14554 Linker relaxation must be turned on so that the linker generates
14555 the stubs correctly in all situations. See the compiler option
14556 @option{-mrelax} and the linker option @option{--relax}.
14557 There are corner cases where the linker is supposed to generate stubs
14558 but aborts without relaxation and without a helpful error message.
14561 The default linker script is arranged for code with @code{EIND = 0}.
14562 If code is supposed to work for a setup with @code{EIND != 0}, a custom
14563 linker script has to be used in order to place the sections whose
14564 name start with @code{.trampolines} into the segment where @code{EIND}
14568 The startup code from libgcc never sets @code{EIND}.
14569 Notice that startup code is a blend of code from libgcc and AVR-LibC.
14570 For the impact of AVR-LibC on @code{EIND}, see the
14571 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
14574 It is legitimate for user-specific startup code to set up @code{EIND}
14575 early, for example by means of initialization code located in
14576 section @code{.init3}. Such code runs prior to general startup code
14577 that initializes RAM and calls constructors, but after the bit
14578 of startup code from AVR-LibC that sets @code{EIND} to the segment
14579 where the vector table is located.
14581 #include <avr/io.h>
14584 __attribute__((section(".init3"),naked,used,no_instrument_function))
14585 init3_set_eind (void)
14587 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
14588 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
14593 The @code{__trampolines_start} symbol is defined in the linker script.
14596 Stubs are generated automatically by the linker if
14597 the following two conditions are met:
14600 @item The address of a label is taken by means of the @code{gs} modifier
14601 (short for @emph{generate stubs}) like so:
14603 LDI r24, lo8(gs(@var{func}))
14604 LDI r25, hi8(gs(@var{func}))
14606 @item The final location of that label is in a code segment
14607 @emph{outside} the segment where the stubs are located.
14611 The compiler emits such @code{gs} modifiers for code labels in the
14612 following situations:
14614 @item Taking address of a function or code label.
14615 @item Computed goto.
14616 @item If prologue-save function is used, see @option{-mcall-prologues}
14617 command-line option.
14618 @item Switch/case dispatch tables. If you do not want such dispatch
14619 tables you can specify the @option{-fno-jump-tables} command-line option.
14620 @item C and C++ constructors/destructors called during startup/shutdown.
14621 @item If the tools hit a @code{gs()} modifier explained above.
14625 Jumping to non-symbolic addresses like so is @emph{not} supported:
14630 /* Call function at word address 0x2 */
14631 return ((int(*)(void)) 0x2)();
14635 Instead, a stub has to be set up, i.e.@: the function has to be called
14636 through a symbol (@code{func_4} in the example):
14641 extern int func_4 (void);
14643 /* Call function at byte address 0x4 */
14648 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
14649 Alternatively, @code{func_4} can be defined in the linker script.
14652 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
14653 @cindex @code{RAMPD}
14654 @cindex @code{RAMPX}
14655 @cindex @code{RAMPY}
14656 @cindex @code{RAMPZ}
14657 Some AVR devices support memories larger than the 64@tie{}KiB range
14658 that can be accessed with 16-bit pointers. To access memory locations
14659 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
14660 register is used as high part of the address:
14661 The @code{X}, @code{Y}, @code{Z} address register is concatenated
14662 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
14663 register, respectively, to get a wide address. Similarly,
14664 @code{RAMPD} is used together with direct addressing.
14668 The startup code initializes the @code{RAMP} special function
14669 registers with zero.
14672 If a @ref{AVR Named Address Spaces,named address space} other than
14673 generic or @code{__flash} is used, then @code{RAMPZ} is set
14674 as needed before the operation.
14677 If the device supports RAM larger than 64@tie{}KiB and the compiler
14678 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
14679 is reset to zero after the operation.
14682 If the device comes with a specific @code{RAMP} register, the ISR
14683 prologue/epilogue saves/restores that SFR and initializes it with
14684 zero in case the ISR code might (implicitly) use it.
14687 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
14688 If you use inline assembler to read from locations outside the
14689 16-bit address range and change one of the @code{RAMP} registers,
14690 you must reset it to zero after the access.
14694 @subsubsection AVR Built-in Macros
14696 GCC defines several built-in macros so that the user code can test
14697 for the presence or absence of features. Almost any of the following
14698 built-in macros are deduced from device capabilities and thus
14699 triggered by the @option{-mmcu=} command-line option.
14701 For even more AVR-specific built-in macros see
14702 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
14707 Build-in macro that resolves to a decimal number that identifies the
14708 architecture and depends on the @option{-mmcu=@var{mcu}} option.
14709 Possible values are:
14711 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
14712 @code{4}, @code{5}, @code{51}, @code{6}
14714 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
14715 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
14719 @code{100}, @code{102}, @code{104},
14720 @code{105}, @code{106}, @code{107}
14722 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
14723 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
14724 If @var{mcu} specifies a device, this built-in macro is set
14725 accordingly. For example, with @option{-mmcu=atmega8} the macro is
14726 defined to @code{4}.
14728 @item __AVR_@var{Device}__
14729 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
14730 the device's name. For example, @option{-mmcu=atmega8} defines the
14731 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
14732 @code{__AVR_ATtiny261A__}, etc.
14734 The built-in macros' names follow
14735 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
14736 the device name as from the AVR user manual. The difference between
14737 @var{Device} in the built-in macro and @var{device} in
14738 @option{-mmcu=@var{device}} is that the latter is always lowercase.
14740 If @var{device} is not a device but only a core architecture like
14741 @samp{avr51}, this macro is not defined.
14743 @item __AVR_DEVICE_NAME__
14744 Setting @option{-mmcu=@var{device}} defines this built-in macro to
14745 the device's name. For example, with @option{-mmcu=atmega8} the macro
14746 is defined to @code{atmega8}.
14748 If @var{device} is not a device but only a core architecture like
14749 @samp{avr51}, this macro is not defined.
14751 @item __AVR_XMEGA__
14752 The device / architecture belongs to the XMEGA family of devices.
14754 @item __AVR_HAVE_ELPM__
14755 The device has the @code{ELPM} instruction.
14757 @item __AVR_HAVE_ELPMX__
14758 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14759 R@var{n},Z+} instructions.
14761 @item __AVR_HAVE_MOVW__
14762 The device has the @code{MOVW} instruction to perform 16-bit
14763 register-register moves.
14765 @item __AVR_HAVE_LPMX__
14766 The device has the @code{LPM R@var{n},Z} and
14767 @code{LPM R@var{n},Z+} instructions.
14769 @item __AVR_HAVE_MUL__
14770 The device has a hardware multiplier.
14772 @item __AVR_HAVE_JMP_CALL__
14773 The device has the @code{JMP} and @code{CALL} instructions.
14774 This is the case for devices with at least 16@tie{}KiB of program
14777 @item __AVR_HAVE_EIJMP_EICALL__
14778 @itemx __AVR_3_BYTE_PC__
14779 The device has the @code{EIJMP} and @code{EICALL} instructions.
14780 This is the case for devices with more than 128@tie{}KiB of program memory.
14781 This also means that the program counter
14782 (PC) is 3@tie{}bytes wide.
14784 @item __AVR_2_BYTE_PC__
14785 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14786 with up to 128@tie{}KiB of program memory.
14788 @item __AVR_HAVE_8BIT_SP__
14789 @itemx __AVR_HAVE_16BIT_SP__
14790 The stack pointer (SP) register is treated as 8-bit respectively
14791 16-bit register by the compiler.
14792 The definition of these macros is affected by @option{-mtiny-stack}.
14794 @item __AVR_HAVE_SPH__
14796 The device has the SPH (high part of stack pointer) special function
14797 register or has an 8-bit stack pointer, respectively.
14798 The definition of these macros is affected by @option{-mmcu=} and
14799 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14802 @item __AVR_HAVE_RAMPD__
14803 @itemx __AVR_HAVE_RAMPX__
14804 @itemx __AVR_HAVE_RAMPY__
14805 @itemx __AVR_HAVE_RAMPZ__
14806 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14807 @code{RAMPZ} special function register, respectively.
14809 @item __NO_INTERRUPTS__
14810 This macro reflects the @option{-mno-interrupts} command-line option.
14812 @item __AVR_ERRATA_SKIP__
14813 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14814 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14815 instructions because of a hardware erratum. Skip instructions are
14816 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14817 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14820 @item __AVR_ISA_RMW__
14821 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14823 @item __AVR_SFR_OFFSET__=@var{offset}
14824 Instructions that can address I/O special function registers directly
14825 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14826 address as if addressed by an instruction to access RAM like @code{LD}
14827 or @code{STS}. This offset depends on the device architecture and has
14828 to be subtracted from the RAM address in order to get the
14829 respective I/O@tie{}address.
14831 @item __WITH_AVRLIBC__
14832 The compiler is configured to be used together with AVR-Libc.
14833 See the @option{--with-avrlibc} configure option.
14837 @node Blackfin Options
14838 @subsection Blackfin Options
14839 @cindex Blackfin Options
14842 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14844 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14845 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14846 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14847 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14848 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14849 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14850 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14851 @samp{bf561}, @samp{bf592}.
14853 The optional @var{sirevision} specifies the silicon revision of the target
14854 Blackfin processor. Any workarounds available for the targeted silicon revision
14855 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14856 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14857 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14858 hexadecimal digits representing the major and minor numbers in the silicon
14859 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14860 is not defined. If @var{sirevision} is @samp{any}, the
14861 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14862 If this optional @var{sirevision} is not used, GCC assumes the latest known
14863 silicon revision of the targeted Blackfin processor.
14865 GCC defines a preprocessor macro for the specified @var{cpu}.
14866 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14867 provided by libgloss to be linked in if @option{-msim} is not given.
14869 Without this option, @samp{bf532} is used as the processor by default.
14871 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14872 only the preprocessor macro is defined.
14876 Specifies that the program will be run on the simulator. This causes
14877 the simulator BSP provided by libgloss to be linked in. This option
14878 has effect only for @samp{bfin-elf} toolchain.
14879 Certain other options, such as @option{-mid-shared-library} and
14880 @option{-mfdpic}, imply @option{-msim}.
14882 @item -momit-leaf-frame-pointer
14883 @opindex momit-leaf-frame-pointer
14884 Don't keep the frame pointer in a register for leaf functions. This
14885 avoids the instructions to save, set up and restore frame pointers and
14886 makes an extra register available in leaf functions. The option
14887 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14888 which might make debugging harder.
14890 @item -mspecld-anomaly
14891 @opindex mspecld-anomaly
14892 When enabled, the compiler ensures that the generated code does not
14893 contain speculative loads after jump instructions. If this option is used,
14894 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14896 @item -mno-specld-anomaly
14897 @opindex mno-specld-anomaly
14898 Don't generate extra code to prevent speculative loads from occurring.
14900 @item -mcsync-anomaly
14901 @opindex mcsync-anomaly
14902 When enabled, the compiler ensures that the generated code does not
14903 contain CSYNC or SSYNC instructions too soon after conditional branches.
14904 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14906 @item -mno-csync-anomaly
14907 @opindex mno-csync-anomaly
14908 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14909 occurring too soon after a conditional branch.
14913 When enabled, the compiler is free to take advantage of the knowledge that
14914 the entire program fits into the low 64k of memory.
14917 @opindex mno-low-64k
14918 Assume that the program is arbitrarily large. This is the default.
14920 @item -mstack-check-l1
14921 @opindex mstack-check-l1
14922 Do stack checking using information placed into L1 scratchpad memory by the
14925 @item -mid-shared-library
14926 @opindex mid-shared-library
14927 Generate code that supports shared libraries via the library ID method.
14928 This allows for execute in place and shared libraries in an environment
14929 without virtual memory management. This option implies @option{-fPIC}.
14930 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14932 @item -mno-id-shared-library
14933 @opindex mno-id-shared-library
14934 Generate code that doesn't assume ID-based shared libraries are being used.
14935 This is the default.
14937 @item -mleaf-id-shared-library
14938 @opindex mleaf-id-shared-library
14939 Generate code that supports shared libraries via the library ID method,
14940 but assumes that this library or executable won't link against any other
14941 ID shared libraries. That allows the compiler to use faster code for jumps
14944 @item -mno-leaf-id-shared-library
14945 @opindex mno-leaf-id-shared-library
14946 Do not assume that the code being compiled won't link against any ID shared
14947 libraries. Slower code is generated for jump and call insns.
14949 @item -mshared-library-id=n
14950 @opindex mshared-library-id
14951 Specifies the identification number of the ID-based shared library being
14952 compiled. Specifying a value of 0 generates more compact code; specifying
14953 other values forces the allocation of that number to the current
14954 library but is no more space- or time-efficient than omitting this option.
14958 Generate code that allows the data segment to be located in a different
14959 area of memory from the text segment. This allows for execute in place in
14960 an environment without virtual memory management by eliminating relocations
14961 against the text section.
14963 @item -mno-sep-data
14964 @opindex mno-sep-data
14965 Generate code that assumes that the data segment follows the text segment.
14966 This is the default.
14969 @itemx -mno-long-calls
14970 @opindex mlong-calls
14971 @opindex mno-long-calls
14972 Tells the compiler to perform function calls by first loading the
14973 address of the function into a register and then performing a subroutine
14974 call on this register. This switch is needed if the target function
14975 lies outside of the 24-bit addressing range of the offset-based
14976 version of subroutine call instruction.
14978 This feature is not enabled by default. Specifying
14979 @option{-mno-long-calls} restores the default behavior. Note these
14980 switches have no effect on how the compiler generates code to handle
14981 function calls via function pointers.
14985 Link with the fast floating-point library. This library relaxes some of
14986 the IEEE floating-point standard's rules for checking inputs against
14987 Not-a-Number (NAN), in the interest of performance.
14990 @opindex minline-plt
14991 Enable inlining of PLT entries in function calls to functions that are
14992 not known to bind locally. It has no effect without @option{-mfdpic}.
14995 @opindex mmulticore
14996 Build a standalone application for multicore Blackfin processors.
14997 This option causes proper start files and link scripts supporting
14998 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14999 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
15001 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
15002 selects the one-application-per-core programming model. Without
15003 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
15004 programming model is used. In this model, the main function of Core B
15005 should be named as @code{coreb_main}.
15007 If this option is not used, the single-core application programming
15012 Build a standalone application for Core A of BF561 when using
15013 the one-application-per-core programming model. Proper start files
15014 and link scripts are used to support Core A, and the macro
15015 @code{__BFIN_COREA} is defined.
15016 This option can only be used in conjunction with @option{-mmulticore}.
15020 Build a standalone application for Core B of BF561 when using
15021 the one-application-per-core programming model. Proper start files
15022 and link scripts are used to support Core B, and the macro
15023 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
15024 should be used instead of @code{main}.
15025 This option can only be used in conjunction with @option{-mmulticore}.
15029 Build a standalone application for SDRAM. Proper start files and
15030 link scripts are used to put the application into SDRAM, and the macro
15031 @code{__BFIN_SDRAM} is defined.
15032 The loader should initialize SDRAM before loading the application.
15036 Assume that ICPLBs are enabled at run time. This has an effect on certain
15037 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
15038 are enabled; for standalone applications the default is off.
15042 @subsection C6X Options
15043 @cindex C6X Options
15046 @item -march=@var{name}
15048 This specifies the name of the target architecture. GCC uses this
15049 name to determine what kind of instructions it can emit when generating
15050 assembly code. Permissible names are: @samp{c62x},
15051 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
15054 @opindex mbig-endian
15055 Generate code for a big-endian target.
15057 @item -mlittle-endian
15058 @opindex mlittle-endian
15059 Generate code for a little-endian target. This is the default.
15063 Choose startup files and linker script suitable for the simulator.
15065 @item -msdata=default
15066 @opindex msdata=default
15067 Put small global and static data in the @code{.neardata} section,
15068 which is pointed to by register @code{B14}. Put small uninitialized
15069 global and static data in the @code{.bss} section, which is adjacent
15070 to the @code{.neardata} section. Put small read-only data into the
15071 @code{.rodata} section. The corresponding sections used for large
15072 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
15075 @opindex msdata=all
15076 Put all data, not just small objects, into the sections reserved for
15077 small data, and use addressing relative to the @code{B14} register to
15081 @opindex msdata=none
15082 Make no use of the sections reserved for small data, and use absolute
15083 addresses to access all data. Put all initialized global and static
15084 data in the @code{.fardata} section, and all uninitialized data in the
15085 @code{.far} section. Put all constant data into the @code{.const}
15090 @subsection CRIS Options
15091 @cindex CRIS Options
15093 These options are defined specifically for the CRIS ports.
15096 @item -march=@var{architecture-type}
15097 @itemx -mcpu=@var{architecture-type}
15100 Generate code for the specified architecture. The choices for
15101 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
15102 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
15103 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
15106 @item -mtune=@var{architecture-type}
15108 Tune to @var{architecture-type} everything applicable about the generated
15109 code, except for the ABI and the set of available instructions. The
15110 choices for @var{architecture-type} are the same as for
15111 @option{-march=@var{architecture-type}}.
15113 @item -mmax-stack-frame=@var{n}
15114 @opindex mmax-stack-frame
15115 Warn when the stack frame of a function exceeds @var{n} bytes.
15121 The options @option{-metrax4} and @option{-metrax100} are synonyms for
15122 @option{-march=v3} and @option{-march=v8} respectively.
15124 @item -mmul-bug-workaround
15125 @itemx -mno-mul-bug-workaround
15126 @opindex mmul-bug-workaround
15127 @opindex mno-mul-bug-workaround
15128 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
15129 models where it applies. This option is active by default.
15133 Enable CRIS-specific verbose debug-related information in the assembly
15134 code. This option also has the effect of turning off the @samp{#NO_APP}
15135 formatted-code indicator to the assembler at the beginning of the
15140 Do not use condition-code results from previous instruction; always emit
15141 compare and test instructions before use of condition codes.
15143 @item -mno-side-effects
15144 @opindex mno-side-effects
15145 Do not emit instructions with side effects in addressing modes other than
15148 @item -mstack-align
15149 @itemx -mno-stack-align
15150 @itemx -mdata-align
15151 @itemx -mno-data-align
15152 @itemx -mconst-align
15153 @itemx -mno-const-align
15154 @opindex mstack-align
15155 @opindex mno-stack-align
15156 @opindex mdata-align
15157 @opindex mno-data-align
15158 @opindex mconst-align
15159 @opindex mno-const-align
15160 These options (@samp{no-} options) arrange (eliminate arrangements) for the
15161 stack frame, individual data and constants to be aligned for the maximum
15162 single data access size for the chosen CPU model. The default is to
15163 arrange for 32-bit alignment. ABI details such as structure layout are
15164 not affected by these options.
15172 Similar to the stack- data- and const-align options above, these options
15173 arrange for stack frame, writable data and constants to all be 32-bit,
15174 16-bit or 8-bit aligned. The default is 32-bit alignment.
15176 @item -mno-prologue-epilogue
15177 @itemx -mprologue-epilogue
15178 @opindex mno-prologue-epilogue
15179 @opindex mprologue-epilogue
15180 With @option{-mno-prologue-epilogue}, the normal function prologue and
15181 epilogue which set up the stack frame are omitted and no return
15182 instructions or return sequences are generated in the code. Use this
15183 option only together with visual inspection of the compiled code: no
15184 warnings or errors are generated when call-saved registers must be saved,
15185 or storage for local variables needs to be allocated.
15189 @opindex mno-gotplt
15191 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
15192 instruction sequences that load addresses for functions from the PLT part
15193 of the GOT rather than (traditional on other architectures) calls to the
15194 PLT@. The default is @option{-mgotplt}.
15198 Legacy no-op option only recognized with the cris-axis-elf and
15199 cris-axis-linux-gnu targets.
15203 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
15207 This option, recognized for the cris-axis-elf, arranges
15208 to link with input-output functions from a simulator library. Code,
15209 initialized data and zero-initialized data are allocated consecutively.
15213 Like @option{-sim}, but pass linker options to locate initialized data at
15214 0x40000000 and zero-initialized data at 0x80000000.
15218 @subsection CR16 Options
15219 @cindex CR16 Options
15221 These options are defined specifically for the CR16 ports.
15227 Enable the use of multiply-accumulate instructions. Disabled by default.
15231 @opindex mcr16cplus
15233 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
15238 Links the library libsim.a which is in compatible with simulator. Applicable
15239 to ELF compiler only.
15243 Choose integer type as 32-bit wide.
15247 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
15249 @item -mdata-model=@var{model}
15250 @opindex mdata-model
15251 Choose a data model. The choices for @var{model} are @samp{near},
15252 @samp{far} or @samp{medium}. @samp{medium} is default.
15253 However, @samp{far} is not valid with @option{-mcr16c}, as the
15254 CR16C architecture does not support the far data model.
15257 @node Darwin Options
15258 @subsection Darwin Options
15259 @cindex Darwin options
15261 These options are defined for all architectures running the Darwin operating
15264 FSF GCC on Darwin does not create ``fat'' object files; it creates
15265 an object file for the single architecture that GCC was built to
15266 target. Apple's GCC on Darwin does create ``fat'' files if multiple
15267 @option{-arch} options are used; it does so by running the compiler or
15268 linker multiple times and joining the results together with
15271 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
15272 @samp{i686}) is determined by the flags that specify the ISA
15273 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
15274 @option{-force_cpusubtype_ALL} option can be used to override this.
15276 The Darwin tools vary in their behavior when presented with an ISA
15277 mismatch. The assembler, @file{as}, only permits instructions to
15278 be used that are valid for the subtype of the file it is generating,
15279 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
15280 The linker for shared libraries, @file{/usr/bin/libtool}, fails
15281 and prints an error if asked to create a shared library with a less
15282 restrictive subtype than its input files (for instance, trying to put
15283 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
15284 for executables, @command{ld}, quietly gives the executable the most
15285 restrictive subtype of any of its input files.
15290 Add the framework directory @var{dir} to the head of the list of
15291 directories to be searched for header files. These directories are
15292 interleaved with those specified by @option{-I} options and are
15293 scanned in a left-to-right order.
15295 A framework directory is a directory with frameworks in it. A
15296 framework is a directory with a @file{Headers} and/or
15297 @file{PrivateHeaders} directory contained directly in it that ends
15298 in @file{.framework}. The name of a framework is the name of this
15299 directory excluding the @file{.framework}. Headers associated with
15300 the framework are found in one of those two directories, with
15301 @file{Headers} being searched first. A subframework is a framework
15302 directory that is in a framework's @file{Frameworks} directory.
15303 Includes of subframework headers can only appear in a header of a
15304 framework that contains the subframework, or in a sibling subframework
15305 header. Two subframeworks are siblings if they occur in the same
15306 framework. A subframework should not have the same name as a
15307 framework; a warning is issued if this is violated. Currently a
15308 subframework cannot have subframeworks; in the future, the mechanism
15309 may be extended to support this. The standard frameworks can be found
15310 in @file{/System/Library/Frameworks} and
15311 @file{/Library/Frameworks}. An example include looks like
15312 @code{#include <Framework/header.h>}, where @file{Framework} denotes
15313 the name of the framework and @file{header.h} is found in the
15314 @file{PrivateHeaders} or @file{Headers} directory.
15316 @item -iframework@var{dir}
15317 @opindex iframework
15318 Like @option{-F} except the directory is a treated as a system
15319 directory. The main difference between this @option{-iframework} and
15320 @option{-F} is that with @option{-iframework} the compiler does not
15321 warn about constructs contained within header files found via
15322 @var{dir}. This option is valid only for the C family of languages.
15326 Emit debugging information for symbols that are used. For stabs
15327 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
15328 This is by default ON@.
15332 Emit debugging information for all symbols and types.
15334 @item -mmacosx-version-min=@var{version}
15335 The earliest version of MacOS X that this executable will run on
15336 is @var{version}. Typical values of @var{version} include @code{10.1},
15337 @code{10.2}, and @code{10.3.9}.
15339 If the compiler was built to use the system's headers by default,
15340 then the default for this option is the system version on which the
15341 compiler is running, otherwise the default is to make choices that
15342 are compatible with as many systems and code bases as possible.
15346 Enable kernel development mode. The @option{-mkernel} option sets
15347 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
15348 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
15349 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
15350 applicable. This mode also sets @option{-mno-altivec},
15351 @option{-msoft-float}, @option{-fno-builtin} and
15352 @option{-mlong-branch} for PowerPC targets.
15354 @item -mone-byte-bool
15355 @opindex mone-byte-bool
15356 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
15357 By default @code{sizeof(bool)} is @code{4} when compiling for
15358 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
15359 option has no effect on x86.
15361 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
15362 to generate code that is not binary compatible with code generated
15363 without that switch. Using this switch may require recompiling all
15364 other modules in a program, including system libraries. Use this
15365 switch to conform to a non-default data model.
15367 @item -mfix-and-continue
15368 @itemx -ffix-and-continue
15369 @itemx -findirect-data
15370 @opindex mfix-and-continue
15371 @opindex ffix-and-continue
15372 @opindex findirect-data
15373 Generate code suitable for fast turnaround development, such as to
15374 allow GDB to dynamically load @file{.o} files into already-running
15375 programs. @option{-findirect-data} and @option{-ffix-and-continue}
15376 are provided for backwards compatibility.
15380 Loads all members of static archive libraries.
15381 See man ld(1) for more information.
15383 @item -arch_errors_fatal
15384 @opindex arch_errors_fatal
15385 Cause the errors having to do with files that have the wrong architecture
15388 @item -bind_at_load
15389 @opindex bind_at_load
15390 Causes the output file to be marked such that the dynamic linker will
15391 bind all undefined references when the file is loaded or launched.
15395 Produce a Mach-o bundle format file.
15396 See man ld(1) for more information.
15398 @item -bundle_loader @var{executable}
15399 @opindex bundle_loader
15400 This option specifies the @var{executable} that will load the build
15401 output file being linked. See man ld(1) for more information.
15404 @opindex dynamiclib
15405 When passed this option, GCC produces a dynamic library instead of
15406 an executable when linking, using the Darwin @file{libtool} command.
15408 @item -force_cpusubtype_ALL
15409 @opindex force_cpusubtype_ALL
15410 This causes GCC's output file to have the @samp{ALL} subtype, instead of
15411 one controlled by the @option{-mcpu} or @option{-march} option.
15413 @item -allowable_client @var{client_name}
15414 @itemx -client_name
15415 @itemx -compatibility_version
15416 @itemx -current_version
15418 @itemx -dependency-file
15420 @itemx -dylinker_install_name
15422 @itemx -exported_symbols_list
15425 @itemx -flat_namespace
15426 @itemx -force_flat_namespace
15427 @itemx -headerpad_max_install_names
15430 @itemx -install_name
15431 @itemx -keep_private_externs
15432 @itemx -multi_module
15433 @itemx -multiply_defined
15434 @itemx -multiply_defined_unused
15437 @itemx -no_dead_strip_inits_and_terms
15438 @itemx -nofixprebinding
15439 @itemx -nomultidefs
15441 @itemx -noseglinkedit
15442 @itemx -pagezero_size
15444 @itemx -prebind_all_twolevel_modules
15445 @itemx -private_bundle
15447 @itemx -read_only_relocs
15449 @itemx -sectobjectsymbols
15453 @itemx -sectobjectsymbols
15456 @itemx -segs_read_only_addr
15458 @itemx -segs_read_write_addr
15459 @itemx -seg_addr_table
15460 @itemx -seg_addr_table_filename
15461 @itemx -seglinkedit
15463 @itemx -segs_read_only_addr
15464 @itemx -segs_read_write_addr
15465 @itemx -single_module
15467 @itemx -sub_library
15469 @itemx -sub_umbrella
15470 @itemx -twolevel_namespace
15473 @itemx -unexported_symbols_list
15474 @itemx -weak_reference_mismatches
15475 @itemx -whatsloaded
15476 @opindex allowable_client
15477 @opindex client_name
15478 @opindex compatibility_version
15479 @opindex current_version
15480 @opindex dead_strip
15481 @opindex dependency-file
15482 @opindex dylib_file
15483 @opindex dylinker_install_name
15485 @opindex exported_symbols_list
15487 @opindex flat_namespace
15488 @opindex force_flat_namespace
15489 @opindex headerpad_max_install_names
15490 @opindex image_base
15492 @opindex install_name
15493 @opindex keep_private_externs
15494 @opindex multi_module
15495 @opindex multiply_defined
15496 @opindex multiply_defined_unused
15497 @opindex noall_load
15498 @opindex no_dead_strip_inits_and_terms
15499 @opindex nofixprebinding
15500 @opindex nomultidefs
15502 @opindex noseglinkedit
15503 @opindex pagezero_size
15505 @opindex prebind_all_twolevel_modules
15506 @opindex private_bundle
15507 @opindex read_only_relocs
15509 @opindex sectobjectsymbols
15512 @opindex sectcreate
15513 @opindex sectobjectsymbols
15516 @opindex segs_read_only_addr
15517 @opindex segs_read_write_addr
15518 @opindex seg_addr_table
15519 @opindex seg_addr_table_filename
15520 @opindex seglinkedit
15522 @opindex segs_read_only_addr
15523 @opindex segs_read_write_addr
15524 @opindex single_module
15526 @opindex sub_library
15527 @opindex sub_umbrella
15528 @opindex twolevel_namespace
15531 @opindex unexported_symbols_list
15532 @opindex weak_reference_mismatches
15533 @opindex whatsloaded
15534 These options are passed to the Darwin linker. The Darwin linker man page
15535 describes them in detail.
15538 @node DEC Alpha Options
15539 @subsection DEC Alpha Options
15541 These @samp{-m} options are defined for the DEC Alpha implementations:
15544 @item -mno-soft-float
15545 @itemx -msoft-float
15546 @opindex mno-soft-float
15547 @opindex msoft-float
15548 Use (do not use) the hardware floating-point instructions for
15549 floating-point operations. When @option{-msoft-float} is specified,
15550 functions in @file{libgcc.a} are used to perform floating-point
15551 operations. Unless they are replaced by routines that emulate the
15552 floating-point operations, or compiled in such a way as to call such
15553 emulations routines, these routines issue floating-point
15554 operations. If you are compiling for an Alpha without floating-point
15555 operations, you must ensure that the library is built so as not to call
15558 Note that Alpha implementations without floating-point operations are
15559 required to have floating-point registers.
15562 @itemx -mno-fp-regs
15564 @opindex mno-fp-regs
15565 Generate code that uses (does not use) the floating-point register set.
15566 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
15567 register set is not used, floating-point operands are passed in integer
15568 registers as if they were integers and floating-point results are passed
15569 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
15570 so any function with a floating-point argument or return value called by code
15571 compiled with @option{-mno-fp-regs} must also be compiled with that
15574 A typical use of this option is building a kernel that does not use,
15575 and hence need not save and restore, any floating-point registers.
15579 The Alpha architecture implements floating-point hardware optimized for
15580 maximum performance. It is mostly compliant with the IEEE floating-point
15581 standard. However, for full compliance, software assistance is
15582 required. This option generates code fully IEEE-compliant code
15583 @emph{except} that the @var{inexact-flag} is not maintained (see below).
15584 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
15585 defined during compilation. The resulting code is less efficient but is
15586 able to correctly support denormalized numbers and exceptional IEEE
15587 values such as not-a-number and plus/minus infinity. Other Alpha
15588 compilers call this option @option{-ieee_with_no_inexact}.
15590 @item -mieee-with-inexact
15591 @opindex mieee-with-inexact
15592 This is like @option{-mieee} except the generated code also maintains
15593 the IEEE @var{inexact-flag}. Turning on this option causes the
15594 generated code to implement fully-compliant IEEE math. In addition to
15595 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
15596 macro. On some Alpha implementations the resulting code may execute
15597 significantly slower than the code generated by default. Since there is
15598 very little code that depends on the @var{inexact-flag}, you should
15599 normally not specify this option. Other Alpha compilers call this
15600 option @option{-ieee_with_inexact}.
15602 @item -mfp-trap-mode=@var{trap-mode}
15603 @opindex mfp-trap-mode
15604 This option controls what floating-point related traps are enabled.
15605 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
15606 The trap mode can be set to one of four values:
15610 This is the default (normal) setting. The only traps that are enabled
15611 are the ones that cannot be disabled in software (e.g., division by zero
15615 In addition to the traps enabled by @samp{n}, underflow traps are enabled
15619 Like @samp{u}, but the instructions are marked to be safe for software
15620 completion (see Alpha architecture manual for details).
15623 Like @samp{su}, but inexact traps are enabled as well.
15626 @item -mfp-rounding-mode=@var{rounding-mode}
15627 @opindex mfp-rounding-mode
15628 Selects the IEEE rounding mode. Other Alpha compilers call this option
15629 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
15634 Normal IEEE rounding mode. Floating-point numbers are rounded towards
15635 the nearest machine number or towards the even machine number in case
15639 Round towards minus infinity.
15642 Chopped rounding mode. Floating-point numbers are rounded towards zero.
15645 Dynamic rounding mode. A field in the floating-point control register
15646 (@var{fpcr}, see Alpha architecture reference manual) controls the
15647 rounding mode in effect. The C library initializes this register for
15648 rounding towards plus infinity. Thus, unless your program modifies the
15649 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
15652 @item -mtrap-precision=@var{trap-precision}
15653 @opindex mtrap-precision
15654 In the Alpha architecture, floating-point traps are imprecise. This
15655 means without software assistance it is impossible to recover from a
15656 floating trap and program execution normally needs to be terminated.
15657 GCC can generate code that can assist operating system trap handlers
15658 in determining the exact location that caused a floating-point trap.
15659 Depending on the requirements of an application, different levels of
15660 precisions can be selected:
15664 Program precision. This option is the default and means a trap handler
15665 can only identify which program caused a floating-point exception.
15668 Function precision. The trap handler can determine the function that
15669 caused a floating-point exception.
15672 Instruction precision. The trap handler can determine the exact
15673 instruction that caused a floating-point exception.
15676 Other Alpha compilers provide the equivalent options called
15677 @option{-scope_safe} and @option{-resumption_safe}.
15679 @item -mieee-conformant
15680 @opindex mieee-conformant
15681 This option marks the generated code as IEEE conformant. You must not
15682 use this option unless you also specify @option{-mtrap-precision=i} and either
15683 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
15684 is to emit the line @samp{.eflag 48} in the function prologue of the
15685 generated assembly file.
15687 @item -mbuild-constants
15688 @opindex mbuild-constants
15689 Normally GCC examines a 32- or 64-bit integer constant to
15690 see if it can construct it from smaller constants in two or three
15691 instructions. If it cannot, it outputs the constant as a literal and
15692 generates code to load it from the data segment at run time.
15694 Use this option to require GCC to construct @emph{all} integer constants
15695 using code, even if it takes more instructions (the maximum is six).
15697 You typically use this option to build a shared library dynamic
15698 loader. Itself a shared library, it must relocate itself in memory
15699 before it can find the variables and constants in its own data segment.
15717 Indicate whether GCC should generate code to use the optional BWX,
15718 CIX, FIX and MAX instruction sets. The default is to use the instruction
15719 sets supported by the CPU type specified via @option{-mcpu=} option or that
15720 of the CPU on which GCC was built if none is specified.
15723 @itemx -mfloat-ieee
15724 @opindex mfloat-vax
15725 @opindex mfloat-ieee
15726 Generate code that uses (does not use) VAX F and G floating-point
15727 arithmetic instead of IEEE single and double precision.
15729 @item -mexplicit-relocs
15730 @itemx -mno-explicit-relocs
15731 @opindex mexplicit-relocs
15732 @opindex mno-explicit-relocs
15733 Older Alpha assemblers provided no way to generate symbol relocations
15734 except via assembler macros. Use of these macros does not allow
15735 optimal instruction scheduling. GNU binutils as of version 2.12
15736 supports a new syntax that allows the compiler to explicitly mark
15737 which relocations should apply to which instructions. This option
15738 is mostly useful for debugging, as GCC detects the capabilities of
15739 the assembler when it is built and sets the default accordingly.
15742 @itemx -mlarge-data
15743 @opindex msmall-data
15744 @opindex mlarge-data
15745 When @option{-mexplicit-relocs} is in effect, static data is
15746 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
15747 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
15748 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
15749 16-bit relocations off of the @code{$gp} register. This limits the
15750 size of the small data area to 64KB, but allows the variables to be
15751 directly accessed via a single instruction.
15753 The default is @option{-mlarge-data}. With this option the data area
15754 is limited to just below 2GB@. Programs that require more than 2GB of
15755 data must use @code{malloc} or @code{mmap} to allocate the data in the
15756 heap instead of in the program's data segment.
15758 When generating code for shared libraries, @option{-fpic} implies
15759 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15762 @itemx -mlarge-text
15763 @opindex msmall-text
15764 @opindex mlarge-text
15765 When @option{-msmall-text} is used, the compiler assumes that the
15766 code of the entire program (or shared library) fits in 4MB, and is
15767 thus reachable with a branch instruction. When @option{-msmall-data}
15768 is used, the compiler can assume that all local symbols share the
15769 same @code{$gp} value, and thus reduce the number of instructions
15770 required for a function call from 4 to 1.
15772 The default is @option{-mlarge-text}.
15774 @item -mcpu=@var{cpu_type}
15776 Set the instruction set and instruction scheduling parameters for
15777 machine type @var{cpu_type}. You can specify either the @samp{EV}
15778 style name or the corresponding chip number. GCC supports scheduling
15779 parameters for the EV4, EV5 and EV6 family of processors and
15780 chooses the default values for the instruction set from the processor
15781 you specify. If you do not specify a processor type, GCC defaults
15782 to the processor on which the compiler was built.
15784 Supported values for @var{cpu_type} are
15790 Schedules as an EV4 and has no instruction set extensions.
15794 Schedules as an EV5 and has no instruction set extensions.
15798 Schedules as an EV5 and supports the BWX extension.
15803 Schedules as an EV5 and supports the BWX and MAX extensions.
15807 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15811 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15814 Native toolchains also support the value @samp{native},
15815 which selects the best architecture option for the host processor.
15816 @option{-mcpu=native} has no effect if GCC does not recognize
15819 @item -mtune=@var{cpu_type}
15821 Set only the instruction scheduling parameters for machine type
15822 @var{cpu_type}. The instruction set is not changed.
15824 Native toolchains also support the value @samp{native},
15825 which selects the best architecture option for the host processor.
15826 @option{-mtune=native} has no effect if GCC does not recognize
15829 @item -mmemory-latency=@var{time}
15830 @opindex mmemory-latency
15831 Sets the latency the scheduler should assume for typical memory
15832 references as seen by the application. This number is highly
15833 dependent on the memory access patterns used by the application
15834 and the size of the external cache on the machine.
15836 Valid options for @var{time} are
15840 A decimal number representing clock cycles.
15846 The compiler contains estimates of the number of clock cycles for
15847 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15848 (also called Dcache, Scache, and Bcache), as well as to main memory.
15849 Note that L3 is only valid for EV5.
15855 @subsection FR30 Options
15856 @cindex FR30 Options
15858 These options are defined specifically for the FR30 port.
15862 @item -msmall-model
15863 @opindex msmall-model
15864 Use the small address space model. This can produce smaller code, but
15865 it does assume that all symbolic values and addresses fit into a
15870 Assume that runtime support has been provided and so there is no need
15871 to include the simulator library (@file{libsim.a}) on the linker
15877 @subsection FT32 Options
15878 @cindex FT32 Options
15880 These options are defined specifically for the FT32 port.
15886 Specifies that the program will be run on the simulator. This causes
15887 an alternate runtime startup and library to be linked.
15888 You must not use this option when generating programs that will run on
15889 real hardware; you must provide your own runtime library for whatever
15890 I/O functions are needed.
15894 Enable Local Register Allocation. This is still experimental for FT32,
15895 so by default the compiler uses standard reload.
15899 Do not use div and mod instructions.
15904 @subsection FRV Options
15905 @cindex FRV Options
15911 Only use the first 32 general-purpose registers.
15916 Use all 64 general-purpose registers.
15921 Use only the first 32 floating-point registers.
15926 Use all 64 floating-point registers.
15929 @opindex mhard-float
15931 Use hardware instructions for floating-point operations.
15934 @opindex msoft-float
15936 Use library routines for floating-point operations.
15941 Dynamically allocate condition code registers.
15946 Do not try to dynamically allocate condition code registers, only
15947 use @code{icc0} and @code{fcc0}.
15952 Change ABI to use double word insns.
15957 Do not use double word instructions.
15962 Use floating-point double instructions.
15965 @opindex mno-double
15967 Do not use floating-point double instructions.
15972 Use media instructions.
15977 Do not use media instructions.
15982 Use multiply and add/subtract instructions.
15985 @opindex mno-muladd
15987 Do not use multiply and add/subtract instructions.
15992 Select the FDPIC ABI, which uses function descriptors to represent
15993 pointers to functions. Without any PIC/PIE-related options, it
15994 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15995 assumes GOT entries and small data are within a 12-bit range from the
15996 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15997 are computed with 32 bits.
15998 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16001 @opindex minline-plt
16003 Enable inlining of PLT entries in function calls to functions that are
16004 not known to bind locally. It has no effect without @option{-mfdpic}.
16005 It's enabled by default if optimizing for speed and compiling for
16006 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
16007 optimization option such as @option{-O3} or above is present in the
16013 Assume a large TLS segment when generating thread-local code.
16018 Do not assume a large TLS segment when generating thread-local code.
16023 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
16024 that is known to be in read-only sections. It's enabled by default,
16025 except for @option{-fpic} or @option{-fpie}: even though it may help
16026 make the global offset table smaller, it trades 1 instruction for 4.
16027 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
16028 one of which may be shared by multiple symbols, and it avoids the need
16029 for a GOT entry for the referenced symbol, so it's more likely to be a
16030 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
16032 @item -multilib-library-pic
16033 @opindex multilib-library-pic
16035 Link with the (library, not FD) pic libraries. It's implied by
16036 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
16037 @option{-fpic} without @option{-mfdpic}. You should never have to use
16041 @opindex mlinked-fp
16043 Follow the EABI requirement of always creating a frame pointer whenever
16044 a stack frame is allocated. This option is enabled by default and can
16045 be disabled with @option{-mno-linked-fp}.
16048 @opindex mlong-calls
16050 Use indirect addressing to call functions outside the current
16051 compilation unit. This allows the functions to be placed anywhere
16052 within the 32-bit address space.
16054 @item -malign-labels
16055 @opindex malign-labels
16057 Try to align labels to an 8-byte boundary by inserting NOPs into the
16058 previous packet. This option only has an effect when VLIW packing
16059 is enabled. It doesn't create new packets; it merely adds NOPs to
16062 @item -mlibrary-pic
16063 @opindex mlibrary-pic
16065 Generate position-independent EABI code.
16070 Use only the first four media accumulator registers.
16075 Use all eight media accumulator registers.
16080 Pack VLIW instructions.
16085 Do not pack VLIW instructions.
16088 @opindex mno-eflags
16090 Do not mark ABI switches in e_flags.
16093 @opindex mcond-move
16095 Enable the use of conditional-move instructions (default).
16097 This switch is mainly for debugging the compiler and will likely be removed
16098 in a future version.
16100 @item -mno-cond-move
16101 @opindex mno-cond-move
16103 Disable the use of conditional-move instructions.
16105 This switch is mainly for debugging the compiler and will likely be removed
16106 in a future version.
16111 Enable the use of conditional set instructions (default).
16113 This switch is mainly for debugging the compiler and will likely be removed
16114 in a future version.
16119 Disable the use of conditional set instructions.
16121 This switch is mainly for debugging the compiler and will likely be removed
16122 in a future version.
16125 @opindex mcond-exec
16127 Enable the use of conditional execution (default).
16129 This switch is mainly for debugging the compiler and will likely be removed
16130 in a future version.
16132 @item -mno-cond-exec
16133 @opindex mno-cond-exec
16135 Disable the use of conditional execution.
16137 This switch is mainly for debugging the compiler and will likely be removed
16138 in a future version.
16140 @item -mvliw-branch
16141 @opindex mvliw-branch
16143 Run a pass to pack branches into VLIW instructions (default).
16145 This switch is mainly for debugging the compiler and will likely be removed
16146 in a future version.
16148 @item -mno-vliw-branch
16149 @opindex mno-vliw-branch
16151 Do not run a pass to pack branches into VLIW instructions.
16153 This switch is mainly for debugging the compiler and will likely be removed
16154 in a future version.
16156 @item -mmulti-cond-exec
16157 @opindex mmulti-cond-exec
16159 Enable optimization of @code{&&} and @code{||} in conditional execution
16162 This switch is mainly for debugging the compiler and will likely be removed
16163 in a future version.
16165 @item -mno-multi-cond-exec
16166 @opindex mno-multi-cond-exec
16168 Disable optimization of @code{&&} and @code{||} in conditional execution.
16170 This switch is mainly for debugging the compiler and will likely be removed
16171 in a future version.
16173 @item -mnested-cond-exec
16174 @opindex mnested-cond-exec
16176 Enable nested conditional execution optimizations (default).
16178 This switch is mainly for debugging the compiler and will likely be removed
16179 in a future version.
16181 @item -mno-nested-cond-exec
16182 @opindex mno-nested-cond-exec
16184 Disable nested conditional execution optimizations.
16186 This switch is mainly for debugging the compiler and will likely be removed
16187 in a future version.
16189 @item -moptimize-membar
16190 @opindex moptimize-membar
16192 This switch removes redundant @code{membar} instructions from the
16193 compiler-generated code. It is enabled by default.
16195 @item -mno-optimize-membar
16196 @opindex mno-optimize-membar
16198 This switch disables the automatic removal of redundant @code{membar}
16199 instructions from the generated code.
16201 @item -mtomcat-stats
16202 @opindex mtomcat-stats
16204 Cause gas to print out tomcat statistics.
16206 @item -mcpu=@var{cpu}
16209 Select the processor type for which to generate code. Possible values are
16210 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
16211 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
16215 @node GNU/Linux Options
16216 @subsection GNU/Linux Options
16218 These @samp{-m} options are defined for GNU/Linux targets:
16223 Use the GNU C library. This is the default except
16224 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
16225 @samp{*-*-linux-*android*} targets.
16229 Use uClibc C library. This is the default on
16230 @samp{*-*-linux-*uclibc*} targets.
16234 Use the musl C library. This is the default on
16235 @samp{*-*-linux-*musl*} targets.
16239 Use Bionic C library. This is the default on
16240 @samp{*-*-linux-*android*} targets.
16244 Compile code compatible with Android platform. This is the default on
16245 @samp{*-*-linux-*android*} targets.
16247 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
16248 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
16249 this option makes the GCC driver pass Android-specific options to the linker.
16250 Finally, this option causes the preprocessor macro @code{__ANDROID__}
16253 @item -tno-android-cc
16254 @opindex tno-android-cc
16255 Disable compilation effects of @option{-mandroid}, i.e., do not enable
16256 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
16257 @option{-fno-rtti} by default.
16259 @item -tno-android-ld
16260 @opindex tno-android-ld
16261 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
16262 linking options to the linker.
16266 @node H8/300 Options
16267 @subsection H8/300 Options
16269 These @samp{-m} options are defined for the H8/300 implementations:
16274 Shorten some address references at link time, when possible; uses the
16275 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
16276 ld, Using ld}, for a fuller description.
16280 Generate code for the H8/300H@.
16284 Generate code for the H8S@.
16288 Generate code for the H8S and H8/300H in the normal mode. This switch
16289 must be used either with @option{-mh} or @option{-ms}.
16293 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
16297 Extended registers are stored on stack before execution of function
16298 with monitor attribute. Default option is @option{-mexr}.
16299 This option is valid only for H8S targets.
16303 Extended registers are not stored on stack before execution of function
16304 with monitor attribute. Default option is @option{-mno-exr}.
16305 This option is valid only for H8S targets.
16309 Make @code{int} data 32 bits by default.
16312 @opindex malign-300
16313 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
16314 The default for the H8/300H and H8S is to align longs and floats on
16316 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
16317 This option has no effect on the H8/300.
16321 @subsection HPPA Options
16322 @cindex HPPA Options
16324 These @samp{-m} options are defined for the HPPA family of computers:
16327 @item -march=@var{architecture-type}
16329 Generate code for the specified architecture. The choices for
16330 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
16331 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
16332 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
16333 architecture option for your machine. Code compiled for lower numbered
16334 architectures runs on higher numbered architectures, but not the
16337 @item -mpa-risc-1-0
16338 @itemx -mpa-risc-1-1
16339 @itemx -mpa-risc-2-0
16340 @opindex mpa-risc-1-0
16341 @opindex mpa-risc-1-1
16342 @opindex mpa-risc-2-0
16343 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
16345 @item -mjump-in-delay
16346 @opindex mjump-in-delay
16347 This option is ignored and provided for compatibility purposes only.
16349 @item -mdisable-fpregs
16350 @opindex mdisable-fpregs
16351 Prevent floating-point registers from being used in any manner. This is
16352 necessary for compiling kernels that perform lazy context switching of
16353 floating-point registers. If you use this option and attempt to perform
16354 floating-point operations, the compiler aborts.
16356 @item -mdisable-indexing
16357 @opindex mdisable-indexing
16358 Prevent the compiler from using indexing address modes. This avoids some
16359 rather obscure problems when compiling MIG generated code under MACH@.
16361 @item -mno-space-regs
16362 @opindex mno-space-regs
16363 Generate code that assumes the target has no space registers. This allows
16364 GCC to generate faster indirect calls and use unscaled index address modes.
16366 Such code is suitable for level 0 PA systems and kernels.
16368 @item -mfast-indirect-calls
16369 @opindex mfast-indirect-calls
16370 Generate code that assumes calls never cross space boundaries. This
16371 allows GCC to emit code that performs faster indirect calls.
16373 This option does not work in the presence of shared libraries or nested
16376 @item -mfixed-range=@var{register-range}
16377 @opindex mfixed-range
16378 Generate code treating the given register range as fixed registers.
16379 A fixed register is one that the register allocator cannot use. This is
16380 useful when compiling kernel code. A register range is specified as
16381 two registers separated by a dash. Multiple register ranges can be
16382 specified separated by a comma.
16384 @item -mlong-load-store
16385 @opindex mlong-load-store
16386 Generate 3-instruction load and store sequences as sometimes required by
16387 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
16390 @item -mportable-runtime
16391 @opindex mportable-runtime
16392 Use the portable calling conventions proposed by HP for ELF systems.
16396 Enable the use of assembler directives only GAS understands.
16398 @item -mschedule=@var{cpu-type}
16400 Schedule code according to the constraints for the machine type
16401 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
16402 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
16403 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
16404 proper scheduling option for your machine. The default scheduling is
16408 @opindex mlinker-opt
16409 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
16410 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
16411 linkers in which they give bogus error messages when linking some programs.
16414 @opindex msoft-float
16415 Generate output containing library calls for floating point.
16416 @strong{Warning:} the requisite libraries are not available for all HPPA
16417 targets. Normally the facilities of the machine's usual C compiler are
16418 used, but this cannot be done directly in cross-compilation. You must make
16419 your own arrangements to provide suitable library functions for
16422 @option{-msoft-float} changes the calling convention in the output file;
16423 therefore, it is only useful if you compile @emph{all} of a program with
16424 this option. In particular, you need to compile @file{libgcc.a}, the
16425 library that comes with GCC, with @option{-msoft-float} in order for
16430 Generate the predefine, @code{_SIO}, for server IO@. The default is
16431 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
16432 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
16433 options are available under HP-UX and HI-UX@.
16437 Use options specific to GNU @command{ld}.
16438 This passes @option{-shared} to @command{ld} when
16439 building a shared library. It is the default when GCC is configured,
16440 explicitly or implicitly, with the GNU linker. This option does not
16441 affect which @command{ld} is called; it only changes what parameters
16442 are passed to that @command{ld}.
16443 The @command{ld} that is called is determined by the
16444 @option{--with-ld} configure option, GCC's program search path, and
16445 finally by the user's @env{PATH}. The linker used by GCC can be printed
16446 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
16447 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
16451 Use options specific to HP @command{ld}.
16452 This passes @option{-b} to @command{ld} when building
16453 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
16454 links. It is the default when GCC is configured, explicitly or
16455 implicitly, with the HP linker. This option does not affect
16456 which @command{ld} is called; it only changes what parameters are passed to that
16458 The @command{ld} that is called is determined by the @option{--with-ld}
16459 configure option, GCC's program search path, and finally by the user's
16460 @env{PATH}. The linker used by GCC can be printed using @samp{which
16461 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
16462 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
16465 @opindex mno-long-calls
16466 Generate code that uses long call sequences. This ensures that a call
16467 is always able to reach linker generated stubs. The default is to generate
16468 long calls only when the distance from the call site to the beginning
16469 of the function or translation unit, as the case may be, exceeds a
16470 predefined limit set by the branch type being used. The limits for
16471 normal calls are 7,600,000 and 240,000 bytes, respectively for the
16472 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
16475 Distances are measured from the beginning of functions when using the
16476 @option{-ffunction-sections} option, or when using the @option{-mgas}
16477 and @option{-mno-portable-runtime} options together under HP-UX with
16480 It is normally not desirable to use this option as it degrades
16481 performance. However, it may be useful in large applications,
16482 particularly when partial linking is used to build the application.
16484 The types of long calls used depends on the capabilities of the
16485 assembler and linker, and the type of code being generated. The
16486 impact on systems that support long absolute calls, and long pic
16487 symbol-difference or pc-relative calls should be relatively small.
16488 However, an indirect call is used on 32-bit ELF systems in pic code
16489 and it is quite long.
16491 @item -munix=@var{unix-std}
16493 Generate compiler predefines and select a startfile for the specified
16494 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
16495 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
16496 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
16497 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
16498 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
16501 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
16502 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
16503 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
16504 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
16505 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
16506 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
16508 It is @emph{important} to note that this option changes the interfaces
16509 for various library routines. It also affects the operational behavior
16510 of the C library. Thus, @emph{extreme} care is needed in using this
16513 Library code that is intended to operate with more than one UNIX
16514 standard must test, set and restore the variable @code{__xpg4_extended_mask}
16515 as appropriate. Most GNU software doesn't provide this capability.
16519 Suppress the generation of link options to search libdld.sl when the
16520 @option{-static} option is specified on HP-UX 10 and later.
16524 The HP-UX implementation of setlocale in libc has a dependency on
16525 libdld.sl. There isn't an archive version of libdld.sl. Thus,
16526 when the @option{-static} option is specified, special link options
16527 are needed to resolve this dependency.
16529 On HP-UX 10 and later, the GCC driver adds the necessary options to
16530 link with libdld.sl when the @option{-static} option is specified.
16531 This causes the resulting binary to be dynamic. On the 64-bit port,
16532 the linkers generate dynamic binaries by default in any case. The
16533 @option{-nolibdld} option can be used to prevent the GCC driver from
16534 adding these link options.
16538 Add support for multithreading with the @dfn{dce thread} library
16539 under HP-UX@. This option sets flags for both the preprocessor and
16543 @node IA-64 Options
16544 @subsection IA-64 Options
16545 @cindex IA-64 Options
16547 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16551 @opindex mbig-endian
16552 Generate code for a big-endian target. This is the default for HP-UX@.
16554 @item -mlittle-endian
16555 @opindex mlittle-endian
16556 Generate code for a little-endian target. This is the default for AIX5
16562 @opindex mno-gnu-as
16563 Generate (or don't) code for the GNU assembler. This is the default.
16564 @c Also, this is the default if the configure option @option{--with-gnu-as}
16570 @opindex mno-gnu-ld
16571 Generate (or don't) code for the GNU linker. This is the default.
16572 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16577 Generate code that does not use a global pointer register. The result
16578 is not position independent code, and violates the IA-64 ABI@.
16580 @item -mvolatile-asm-stop
16581 @itemx -mno-volatile-asm-stop
16582 @opindex mvolatile-asm-stop
16583 @opindex mno-volatile-asm-stop
16584 Generate (or don't) a stop bit immediately before and after volatile asm
16587 @item -mregister-names
16588 @itemx -mno-register-names
16589 @opindex mregister-names
16590 @opindex mno-register-names
16591 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16592 the stacked registers. This may make assembler output more readable.
16598 Disable (or enable) optimizations that use the small data section. This may
16599 be useful for working around optimizer bugs.
16601 @item -mconstant-gp
16602 @opindex mconstant-gp
16603 Generate code that uses a single constant global pointer value. This is
16604 useful when compiling kernel code.
16608 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16609 This is useful when compiling firmware code.
16611 @item -minline-float-divide-min-latency
16612 @opindex minline-float-divide-min-latency
16613 Generate code for inline divides of floating-point values
16614 using the minimum latency algorithm.
16616 @item -minline-float-divide-max-throughput
16617 @opindex minline-float-divide-max-throughput
16618 Generate code for inline divides of floating-point values
16619 using the maximum throughput algorithm.
16621 @item -mno-inline-float-divide
16622 @opindex mno-inline-float-divide
16623 Do not generate inline code for divides of floating-point values.
16625 @item -minline-int-divide-min-latency
16626 @opindex minline-int-divide-min-latency
16627 Generate code for inline divides of integer values
16628 using the minimum latency algorithm.
16630 @item -minline-int-divide-max-throughput
16631 @opindex minline-int-divide-max-throughput
16632 Generate code for inline divides of integer values
16633 using the maximum throughput algorithm.
16635 @item -mno-inline-int-divide
16636 @opindex mno-inline-int-divide
16637 Do not generate inline code for divides of integer values.
16639 @item -minline-sqrt-min-latency
16640 @opindex minline-sqrt-min-latency
16641 Generate code for inline square roots
16642 using the minimum latency algorithm.
16644 @item -minline-sqrt-max-throughput
16645 @opindex minline-sqrt-max-throughput
16646 Generate code for inline square roots
16647 using the maximum throughput algorithm.
16649 @item -mno-inline-sqrt
16650 @opindex mno-inline-sqrt
16651 Do not generate inline code for @code{sqrt}.
16654 @itemx -mno-fused-madd
16655 @opindex mfused-madd
16656 @opindex mno-fused-madd
16657 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16658 instructions. The default is to use these instructions.
16660 @item -mno-dwarf2-asm
16661 @itemx -mdwarf2-asm
16662 @opindex mno-dwarf2-asm
16663 @opindex mdwarf2-asm
16664 Don't (or do) generate assembler code for the DWARF line number debugging
16665 info. This may be useful when not using the GNU assembler.
16667 @item -mearly-stop-bits
16668 @itemx -mno-early-stop-bits
16669 @opindex mearly-stop-bits
16670 @opindex mno-early-stop-bits
16671 Allow stop bits to be placed earlier than immediately preceding the
16672 instruction that triggered the stop bit. This can improve instruction
16673 scheduling, but does not always do so.
16675 @item -mfixed-range=@var{register-range}
16676 @opindex mfixed-range
16677 Generate code treating the given register range as fixed registers.
16678 A fixed register is one that the register allocator cannot use. This is
16679 useful when compiling kernel code. A register range is specified as
16680 two registers separated by a dash. Multiple register ranges can be
16681 specified separated by a comma.
16683 @item -mtls-size=@var{tls-size}
16685 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16688 @item -mtune=@var{cpu-type}
16690 Tune the instruction scheduling for a particular CPU, Valid values are
16691 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16692 and @samp{mckinley}.
16698 Generate code for a 32-bit or 64-bit environment.
16699 The 32-bit environment sets int, long and pointer to 32 bits.
16700 The 64-bit environment sets int to 32 bits and long and pointer
16701 to 64 bits. These are HP-UX specific flags.
16703 @item -mno-sched-br-data-spec
16704 @itemx -msched-br-data-spec
16705 @opindex mno-sched-br-data-spec
16706 @opindex msched-br-data-spec
16707 (Dis/En)able data speculative scheduling before reload.
16708 This results in generation of @code{ld.a} instructions and
16709 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16710 The default setting is disabled.
16712 @item -msched-ar-data-spec
16713 @itemx -mno-sched-ar-data-spec
16714 @opindex msched-ar-data-spec
16715 @opindex mno-sched-ar-data-spec
16716 (En/Dis)able data speculative scheduling after reload.
16717 This results in generation of @code{ld.a} instructions and
16718 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16719 The default setting is enabled.
16721 @item -mno-sched-control-spec
16722 @itemx -msched-control-spec
16723 @opindex mno-sched-control-spec
16724 @opindex msched-control-spec
16725 (Dis/En)able control speculative scheduling. This feature is
16726 available only during region scheduling (i.e.@: before reload).
16727 This results in generation of the @code{ld.s} instructions and
16728 the corresponding check instructions @code{chk.s}.
16729 The default setting is disabled.
16731 @item -msched-br-in-data-spec
16732 @itemx -mno-sched-br-in-data-spec
16733 @opindex msched-br-in-data-spec
16734 @opindex mno-sched-br-in-data-spec
16735 (En/Dis)able speculative scheduling of the instructions that
16736 are dependent on the data speculative loads before reload.
16737 This is effective only with @option{-msched-br-data-spec} enabled.
16738 The default setting is enabled.
16740 @item -msched-ar-in-data-spec
16741 @itemx -mno-sched-ar-in-data-spec
16742 @opindex msched-ar-in-data-spec
16743 @opindex mno-sched-ar-in-data-spec
16744 (En/Dis)able speculative scheduling of the instructions that
16745 are dependent on the data speculative loads after reload.
16746 This is effective only with @option{-msched-ar-data-spec} enabled.
16747 The default setting is enabled.
16749 @item -msched-in-control-spec
16750 @itemx -mno-sched-in-control-spec
16751 @opindex msched-in-control-spec
16752 @opindex mno-sched-in-control-spec
16753 (En/Dis)able speculative scheduling of the instructions that
16754 are dependent on the control speculative loads.
16755 This is effective only with @option{-msched-control-spec} enabled.
16756 The default setting is enabled.
16758 @item -mno-sched-prefer-non-data-spec-insns
16759 @itemx -msched-prefer-non-data-spec-insns
16760 @opindex mno-sched-prefer-non-data-spec-insns
16761 @opindex msched-prefer-non-data-spec-insns
16762 If enabled, data-speculative instructions are chosen for schedule
16763 only if there are no other choices at the moment. This makes
16764 the use of the data speculation much more conservative.
16765 The default setting is disabled.
16767 @item -mno-sched-prefer-non-control-spec-insns
16768 @itemx -msched-prefer-non-control-spec-insns
16769 @opindex mno-sched-prefer-non-control-spec-insns
16770 @opindex msched-prefer-non-control-spec-insns
16771 If enabled, control-speculative instructions are chosen for schedule
16772 only if there are no other choices at the moment. This makes
16773 the use of the control speculation much more conservative.
16774 The default setting is disabled.
16776 @item -mno-sched-count-spec-in-critical-path
16777 @itemx -msched-count-spec-in-critical-path
16778 @opindex mno-sched-count-spec-in-critical-path
16779 @opindex msched-count-spec-in-critical-path
16780 If enabled, speculative dependencies are considered during
16781 computation of the instructions priorities. This makes the use of the
16782 speculation a bit more conservative.
16783 The default setting is disabled.
16785 @item -msched-spec-ldc
16786 @opindex msched-spec-ldc
16787 Use a simple data speculation check. This option is on by default.
16789 @item -msched-control-spec-ldc
16790 @opindex msched-spec-ldc
16791 Use a simple check for control speculation. This option is on by default.
16793 @item -msched-stop-bits-after-every-cycle
16794 @opindex msched-stop-bits-after-every-cycle
16795 Place a stop bit after every cycle when scheduling. This option is on
16798 @item -msched-fp-mem-deps-zero-cost
16799 @opindex msched-fp-mem-deps-zero-cost
16800 Assume that floating-point stores and loads are not likely to cause a conflict
16801 when placed into the same instruction group. This option is disabled by
16804 @item -msel-sched-dont-check-control-spec
16805 @opindex msel-sched-dont-check-control-spec
16806 Generate checks for control speculation in selective scheduling.
16807 This flag is disabled by default.
16809 @item -msched-max-memory-insns=@var{max-insns}
16810 @opindex msched-max-memory-insns
16811 Limit on the number of memory insns per instruction group, giving lower
16812 priority to subsequent memory insns attempting to schedule in the same
16813 instruction group. Frequently useful to prevent cache bank conflicts.
16814 The default value is 1.
16816 @item -msched-max-memory-insns-hard-limit
16817 @opindex msched-max-memory-insns-hard-limit
16818 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16819 disallowing more than that number in an instruction group.
16820 Otherwise, the limit is ``soft'', meaning that non-memory operations
16821 are preferred when the limit is reached, but memory operations may still
16827 @subsection LM32 Options
16828 @cindex LM32 options
16830 These @option{-m} options are defined for the LatticeMico32 architecture:
16833 @item -mbarrel-shift-enabled
16834 @opindex mbarrel-shift-enabled
16835 Enable barrel-shift instructions.
16837 @item -mdivide-enabled
16838 @opindex mdivide-enabled
16839 Enable divide and modulus instructions.
16841 @item -mmultiply-enabled
16842 @opindex multiply-enabled
16843 Enable multiply instructions.
16845 @item -msign-extend-enabled
16846 @opindex msign-extend-enabled
16847 Enable sign extend instructions.
16849 @item -muser-enabled
16850 @opindex muser-enabled
16851 Enable user-defined instructions.
16856 @subsection M32C Options
16857 @cindex M32C options
16860 @item -mcpu=@var{name}
16862 Select the CPU for which code is generated. @var{name} may be one of
16863 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16864 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16865 the M32C/80 series.
16869 Specifies that the program will be run on the simulator. This causes
16870 an alternate runtime library to be linked in which supports, for
16871 example, file I/O@. You must not use this option when generating
16872 programs that will run on real hardware; you must provide your own
16873 runtime library for whatever I/O functions are needed.
16875 @item -memregs=@var{number}
16877 Specifies the number of memory-based pseudo-registers GCC uses
16878 during code generation. These pseudo-registers are used like real
16879 registers, so there is a tradeoff between GCC's ability to fit the
16880 code into available registers, and the performance penalty of using
16881 memory instead of registers. Note that all modules in a program must
16882 be compiled with the same value for this option. Because of that, you
16883 must not use this option with GCC's default runtime libraries.
16887 @node M32R/D Options
16888 @subsection M32R/D Options
16889 @cindex M32R/D options
16891 These @option{-m} options are defined for Renesas M32R/D architectures:
16896 Generate code for the M32R/2@.
16900 Generate code for the M32R/X@.
16904 Generate code for the M32R@. This is the default.
16906 @item -mmodel=small
16907 @opindex mmodel=small
16908 Assume all objects live in the lower 16MB of memory (so that their addresses
16909 can be loaded with the @code{ld24} instruction), and assume all subroutines
16910 are reachable with the @code{bl} instruction.
16911 This is the default.
16913 The addressability of a particular object can be set with the
16914 @code{model} attribute.
16916 @item -mmodel=medium
16917 @opindex mmodel=medium
16918 Assume objects may be anywhere in the 32-bit address space (the compiler
16919 generates @code{seth/add3} instructions to load their addresses), and
16920 assume all subroutines are reachable with the @code{bl} instruction.
16922 @item -mmodel=large
16923 @opindex mmodel=large
16924 Assume objects may be anywhere in the 32-bit address space (the compiler
16925 generates @code{seth/add3} instructions to load their addresses), and
16926 assume subroutines may not be reachable with the @code{bl} instruction
16927 (the compiler generates the much slower @code{seth/add3/jl}
16928 instruction sequence).
16931 @opindex msdata=none
16932 Disable use of the small data area. Variables are put into
16933 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16934 @code{section} attribute has been specified).
16935 This is the default.
16937 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16938 Objects may be explicitly put in the small data area with the
16939 @code{section} attribute using one of these sections.
16941 @item -msdata=sdata
16942 @opindex msdata=sdata
16943 Put small global and static data in the small data area, but do not
16944 generate special code to reference them.
16947 @opindex msdata=use
16948 Put small global and static data in the small data area, and generate
16949 special instructions to reference them.
16953 @cindex smaller data references
16954 Put global and static objects less than or equal to @var{num} bytes
16955 into the small data or BSS sections instead of the normal data or BSS
16956 sections. The default value of @var{num} is 8.
16957 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16958 for this option to have any effect.
16960 All modules should be compiled with the same @option{-G @var{num}} value.
16961 Compiling with different values of @var{num} may or may not work; if it
16962 doesn't the linker gives an error message---incorrect code is not
16967 Makes the M32R-specific code in the compiler display some statistics
16968 that might help in debugging programs.
16970 @item -malign-loops
16971 @opindex malign-loops
16972 Align all loops to a 32-byte boundary.
16974 @item -mno-align-loops
16975 @opindex mno-align-loops
16976 Do not enforce a 32-byte alignment for loops. This is the default.
16978 @item -missue-rate=@var{number}
16979 @opindex missue-rate=@var{number}
16980 Issue @var{number} instructions per cycle. @var{number} can only be 1
16983 @item -mbranch-cost=@var{number}
16984 @opindex mbranch-cost=@var{number}
16985 @var{number} can only be 1 or 2. If it is 1 then branches are
16986 preferred over conditional code, if it is 2, then the opposite applies.
16988 @item -mflush-trap=@var{number}
16989 @opindex mflush-trap=@var{number}
16990 Specifies the trap number to use to flush the cache. The default is
16991 12. Valid numbers are between 0 and 15 inclusive.
16993 @item -mno-flush-trap
16994 @opindex mno-flush-trap
16995 Specifies that the cache cannot be flushed by using a trap.
16997 @item -mflush-func=@var{name}
16998 @opindex mflush-func=@var{name}
16999 Specifies the name of the operating system function to call to flush
17000 the cache. The default is @samp{_flush_cache}, but a function call
17001 is only used if a trap is not available.
17003 @item -mno-flush-func
17004 @opindex mno-flush-func
17005 Indicates that there is no OS function for flushing the cache.
17009 @node M680x0 Options
17010 @subsection M680x0 Options
17011 @cindex M680x0 options
17013 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
17014 The default settings depend on which architecture was selected when
17015 the compiler was configured; the defaults for the most common choices
17019 @item -march=@var{arch}
17021 Generate code for a specific M680x0 or ColdFire instruction set
17022 architecture. Permissible values of @var{arch} for M680x0
17023 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
17024 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
17025 architectures are selected according to Freescale's ISA classification
17026 and the permissible values are: @samp{isaa}, @samp{isaaplus},
17027 @samp{isab} and @samp{isac}.
17029 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
17030 code for a ColdFire target. The @var{arch} in this macro is one of the
17031 @option{-march} arguments given above.
17033 When used together, @option{-march} and @option{-mtune} select code
17034 that runs on a family of similar processors but that is optimized
17035 for a particular microarchitecture.
17037 @item -mcpu=@var{cpu}
17039 Generate code for a specific M680x0 or ColdFire processor.
17040 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
17041 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
17042 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
17043 below, which also classifies the CPUs into families:
17045 @multitable @columnfractions 0.20 0.80
17046 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
17047 @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}
17048 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
17049 @item @samp{5206e} @tab @samp{5206e}
17050 @item @samp{5208} @tab @samp{5207} @samp{5208}
17051 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
17052 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
17053 @item @samp{5216} @tab @samp{5214} @samp{5216}
17054 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
17055 @item @samp{5225} @tab @samp{5224} @samp{5225}
17056 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
17057 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
17058 @item @samp{5249} @tab @samp{5249}
17059 @item @samp{5250} @tab @samp{5250}
17060 @item @samp{5271} @tab @samp{5270} @samp{5271}
17061 @item @samp{5272} @tab @samp{5272}
17062 @item @samp{5275} @tab @samp{5274} @samp{5275}
17063 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
17064 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
17065 @item @samp{5307} @tab @samp{5307}
17066 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
17067 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
17068 @item @samp{5407} @tab @samp{5407}
17069 @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}
17072 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
17073 @var{arch} is compatible with @var{cpu}. Other combinations of
17074 @option{-mcpu} and @option{-march} are rejected.
17076 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
17077 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
17078 where the value of @var{family} is given by the table above.
17080 @item -mtune=@var{tune}
17082 Tune the code for a particular microarchitecture within the
17083 constraints set by @option{-march} and @option{-mcpu}.
17084 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
17085 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
17086 and @samp{cpu32}. The ColdFire microarchitectures
17087 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
17089 You can also use @option{-mtune=68020-40} for code that needs
17090 to run relatively well on 68020, 68030 and 68040 targets.
17091 @option{-mtune=68020-60} is similar but includes 68060 targets
17092 as well. These two options select the same tuning decisions as
17093 @option{-m68020-40} and @option{-m68020-60} respectively.
17095 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
17096 when tuning for 680x0 architecture @var{arch}. It also defines
17097 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
17098 option is used. If GCC is tuning for a range of architectures,
17099 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
17100 it defines the macros for every architecture in the range.
17102 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
17103 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
17104 of the arguments given above.
17110 Generate output for a 68000. This is the default
17111 when the compiler is configured for 68000-based systems.
17112 It is equivalent to @option{-march=68000}.
17114 Use this option for microcontrollers with a 68000 or EC000 core,
17115 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
17119 Generate output for a 68010. This is the default
17120 when the compiler is configured for 68010-based systems.
17121 It is equivalent to @option{-march=68010}.
17127 Generate output for a 68020. This is the default
17128 when the compiler is configured for 68020-based systems.
17129 It is equivalent to @option{-march=68020}.
17133 Generate output for a 68030. This is the default when the compiler is
17134 configured for 68030-based systems. It is equivalent to
17135 @option{-march=68030}.
17139 Generate output for a 68040. This is the default when the compiler is
17140 configured for 68040-based systems. It is equivalent to
17141 @option{-march=68040}.
17143 This option inhibits the use of 68881/68882 instructions that have to be
17144 emulated by software on the 68040. Use this option if your 68040 does not
17145 have code to emulate those instructions.
17149 Generate output for a 68060. This is the default when the compiler is
17150 configured for 68060-based systems. It is equivalent to
17151 @option{-march=68060}.
17153 This option inhibits the use of 68020 and 68881/68882 instructions that
17154 have to be emulated by software on the 68060. Use this option if your 68060
17155 does not have code to emulate those instructions.
17159 Generate output for a CPU32. This is the default
17160 when the compiler is configured for CPU32-based systems.
17161 It is equivalent to @option{-march=cpu32}.
17163 Use this option for microcontrollers with a
17164 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
17165 68336, 68340, 68341, 68349 and 68360.
17169 Generate output for a 520X ColdFire CPU@. This is the default
17170 when the compiler is configured for 520X-based systems.
17171 It is equivalent to @option{-mcpu=5206}, and is now deprecated
17172 in favor of that option.
17174 Use this option for microcontroller with a 5200 core, including
17175 the MCF5202, MCF5203, MCF5204 and MCF5206.
17179 Generate output for a 5206e ColdFire CPU@. The option is now
17180 deprecated in favor of the equivalent @option{-mcpu=5206e}.
17184 Generate output for a member of the ColdFire 528X family.
17185 The option is now deprecated in favor of the equivalent
17186 @option{-mcpu=528x}.
17190 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
17191 in favor of the equivalent @option{-mcpu=5307}.
17195 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
17196 in favor of the equivalent @option{-mcpu=5407}.
17200 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
17201 This includes use of hardware floating-point instructions.
17202 The option is equivalent to @option{-mcpu=547x}, and is now
17203 deprecated in favor of that option.
17207 Generate output for a 68040, without using any of the new instructions.
17208 This results in code that can run relatively efficiently on either a
17209 68020/68881 or a 68030 or a 68040. The generated code does use the
17210 68881 instructions that are emulated on the 68040.
17212 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
17216 Generate output for a 68060, without using any of the new instructions.
17217 This results in code that can run relatively efficiently on either a
17218 68020/68881 or a 68030 or a 68040. The generated code does use the
17219 68881 instructions that are emulated on the 68060.
17221 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
17225 @opindex mhard-float
17227 Generate floating-point instructions. This is the default for 68020
17228 and above, and for ColdFire devices that have an FPU@. It defines the
17229 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
17230 on ColdFire targets.
17233 @opindex msoft-float
17234 Do not generate floating-point instructions; use library calls instead.
17235 This is the default for 68000, 68010, and 68832 targets. It is also
17236 the default for ColdFire devices that have no FPU.
17242 Generate (do not generate) ColdFire hardware divide and remainder
17243 instructions. If @option{-march} is used without @option{-mcpu},
17244 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
17245 architectures. Otherwise, the default is taken from the target CPU
17246 (either the default CPU, or the one specified by @option{-mcpu}). For
17247 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
17248 @option{-mcpu=5206e}.
17250 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
17254 Consider type @code{int} to be 16 bits wide, like @code{short int}.
17255 Additionally, parameters passed on the stack are also aligned to a
17256 16-bit boundary even on targets whose API mandates promotion to 32-bit.
17260 Do not consider type @code{int} to be 16 bits wide. This is the default.
17263 @itemx -mno-bitfield
17264 @opindex mnobitfield
17265 @opindex mno-bitfield
17266 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
17267 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
17271 Do use the bit-field instructions. The @option{-m68020} option implies
17272 @option{-mbitfield}. This is the default if you use a configuration
17273 designed for a 68020.
17277 Use a different function-calling convention, in which functions
17278 that take a fixed number of arguments return with the @code{rtd}
17279 instruction, which pops their arguments while returning. This
17280 saves one instruction in the caller since there is no need to pop
17281 the arguments there.
17283 This calling convention is incompatible with the one normally
17284 used on Unix, so you cannot use it if you need to call libraries
17285 compiled with the Unix compiler.
17287 Also, you must provide function prototypes for all functions that
17288 take variable numbers of arguments (including @code{printf});
17289 otherwise incorrect code is generated for calls to those
17292 In addition, seriously incorrect code results if you call a
17293 function with too many arguments. (Normally, extra arguments are
17294 harmlessly ignored.)
17296 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
17297 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
17301 Do not use the calling conventions selected by @option{-mrtd}.
17302 This is the default.
17305 @itemx -mno-align-int
17306 @opindex malign-int
17307 @opindex mno-align-int
17308 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
17309 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
17310 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
17311 Aligning variables on 32-bit boundaries produces code that runs somewhat
17312 faster on processors with 32-bit busses at the expense of more memory.
17314 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
17315 aligns structures containing the above types differently than
17316 most published application binary interface specifications for the m68k.
17320 Use the pc-relative addressing mode of the 68000 directly, instead of
17321 using a global offset table. At present, this option implies @option{-fpic},
17322 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
17323 not presently supported with @option{-mpcrel}, though this could be supported for
17324 68020 and higher processors.
17326 @item -mno-strict-align
17327 @itemx -mstrict-align
17328 @opindex mno-strict-align
17329 @opindex mstrict-align
17330 Do not (do) assume that unaligned memory references are handled by
17334 Generate code that allows the data segment to be located in a different
17335 area of memory from the text segment. This allows for execute-in-place in
17336 an environment without virtual memory management. This option implies
17339 @item -mno-sep-data
17340 Generate code that assumes that the data segment follows the text segment.
17341 This is the default.
17343 @item -mid-shared-library
17344 Generate code that supports shared libraries via the library ID method.
17345 This allows for execute-in-place and shared libraries in an environment
17346 without virtual memory management. This option implies @option{-fPIC}.
17348 @item -mno-id-shared-library
17349 Generate code that doesn't assume ID-based shared libraries are being used.
17350 This is the default.
17352 @item -mshared-library-id=n
17353 Specifies the identification number of the ID-based shared library being
17354 compiled. Specifying a value of 0 generates more compact code; specifying
17355 other values forces the allocation of that number to the current
17356 library, but is no more space- or time-efficient than omitting this option.
17362 When generating position-independent code for ColdFire, generate code
17363 that works if the GOT has more than 8192 entries. This code is
17364 larger and slower than code generated without this option. On M680x0
17365 processors, this option is not needed; @option{-fPIC} suffices.
17367 GCC normally uses a single instruction to load values from the GOT@.
17368 While this is relatively efficient, it only works if the GOT
17369 is smaller than about 64k. Anything larger causes the linker
17370 to report an error such as:
17372 @cindex relocation truncated to fit (ColdFire)
17374 relocation truncated to fit: R_68K_GOT16O foobar
17377 If this happens, you should recompile your code with @option{-mxgot}.
17378 It should then work with very large GOTs. However, code generated with
17379 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
17380 the value of a global symbol.
17382 Note that some linkers, including newer versions of the GNU linker,
17383 can create multiple GOTs and sort GOT entries. If you have such a linker,
17384 you should only need to use @option{-mxgot} when compiling a single
17385 object file that accesses more than 8192 GOT entries. Very few do.
17387 These options have no effect unless GCC is generating
17388 position-independent code.
17392 @node MCore Options
17393 @subsection MCore Options
17394 @cindex MCore options
17396 These are the @samp{-m} options defined for the Motorola M*Core
17402 @itemx -mno-hardlit
17404 @opindex mno-hardlit
17405 Inline constants into the code stream if it can be done in two
17406 instructions or less.
17412 Use the divide instruction. (Enabled by default).
17414 @item -mrelax-immediate
17415 @itemx -mno-relax-immediate
17416 @opindex mrelax-immediate
17417 @opindex mno-relax-immediate
17418 Allow arbitrary-sized immediates in bit operations.
17420 @item -mwide-bitfields
17421 @itemx -mno-wide-bitfields
17422 @opindex mwide-bitfields
17423 @opindex mno-wide-bitfields
17424 Always treat bit-fields as @code{int}-sized.
17426 @item -m4byte-functions
17427 @itemx -mno-4byte-functions
17428 @opindex m4byte-functions
17429 @opindex mno-4byte-functions
17430 Force all functions to be aligned to a 4-byte boundary.
17432 @item -mcallgraph-data
17433 @itemx -mno-callgraph-data
17434 @opindex mcallgraph-data
17435 @opindex mno-callgraph-data
17436 Emit callgraph information.
17439 @itemx -mno-slow-bytes
17440 @opindex mslow-bytes
17441 @opindex mno-slow-bytes
17442 Prefer word access when reading byte quantities.
17444 @item -mlittle-endian
17445 @itemx -mbig-endian
17446 @opindex mlittle-endian
17447 @opindex mbig-endian
17448 Generate code for a little-endian target.
17454 Generate code for the 210 processor.
17458 Assume that runtime support has been provided and so omit the
17459 simulator library (@file{libsim.a)} from the linker command line.
17461 @item -mstack-increment=@var{size}
17462 @opindex mstack-increment
17463 Set the maximum amount for a single stack increment operation. Large
17464 values can increase the speed of programs that contain functions
17465 that need a large amount of stack space, but they can also trigger a
17466 segmentation fault if the stack is extended too much. The default
17472 @subsection MeP Options
17473 @cindex MeP options
17479 Enables the @code{abs} instruction, which is the absolute difference
17480 between two registers.
17484 Enables all the optional instructions---average, multiply, divide, bit
17485 operations, leading zero, absolute difference, min/max, clip, and
17491 Enables the @code{ave} instruction, which computes the average of two
17494 @item -mbased=@var{n}
17496 Variables of size @var{n} bytes or smaller are placed in the
17497 @code{.based} section by default. Based variables use the @code{$tp}
17498 register as a base register, and there is a 128-byte limit to the
17499 @code{.based} section.
17503 Enables the bit operation instructions---bit test (@code{btstm}), set
17504 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17505 test-and-set (@code{tas}).
17507 @item -mc=@var{name}
17509 Selects which section constant data is placed in. @var{name} may
17510 be @samp{tiny}, @samp{near}, or @samp{far}.
17514 Enables the @code{clip} instruction. Note that @option{-mclip} is not
17515 useful unless you also provide @option{-mminmax}.
17517 @item -mconfig=@var{name}
17519 Selects one of the built-in core configurations. Each MeP chip has
17520 one or more modules in it; each module has a core CPU and a variety of
17521 coprocessors, optional instructions, and peripherals. The
17522 @code{MeP-Integrator} tool, not part of GCC, provides these
17523 configurations through this option; using this option is the same as
17524 using all the corresponding command-line options. The default
17525 configuration is @samp{default}.
17529 Enables the coprocessor instructions. By default, this is a 32-bit
17530 coprocessor. Note that the coprocessor is normally enabled via the
17531 @option{-mconfig=} option.
17535 Enables the 32-bit coprocessor's instructions.
17539 Enables the 64-bit coprocessor's instructions.
17543 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17547 Causes constant variables to be placed in the @code{.near} section.
17551 Enables the @code{div} and @code{divu} instructions.
17555 Generate big-endian code.
17559 Generate little-endian code.
17561 @item -mio-volatile
17562 @opindex mio-volatile
17563 Tells the compiler that any variable marked with the @code{io}
17564 attribute is to be considered volatile.
17568 Causes variables to be assigned to the @code{.far} section by default.
17572 Enables the @code{leadz} (leading zero) instruction.
17576 Causes variables to be assigned to the @code{.near} section by default.
17580 Enables the @code{min} and @code{max} instructions.
17584 Enables the multiplication and multiply-accumulate instructions.
17588 Disables all the optional instructions enabled by @option{-mall-opts}.
17592 Enables the @code{repeat} and @code{erepeat} instructions, used for
17593 low-overhead looping.
17597 Causes all variables to default to the @code{.tiny} section. Note
17598 that there is a 65536-byte limit to this section. Accesses to these
17599 variables use the @code{%gp} base register.
17603 Enables the saturation instructions. Note that the compiler does not
17604 currently generate these itself, but this option is included for
17605 compatibility with other tools, like @code{as}.
17609 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17613 Link the simulator run-time libraries.
17617 Link the simulator runtime libraries, excluding built-in support
17618 for reset and exception vectors and tables.
17622 Causes all functions to default to the @code{.far} section. Without
17623 this option, functions default to the @code{.near} section.
17625 @item -mtiny=@var{n}
17627 Variables that are @var{n} bytes or smaller are allocated to the
17628 @code{.tiny} section. These variables use the @code{$gp} base
17629 register. The default for this option is 4, but note that there's a
17630 65536-byte limit to the @code{.tiny} section.
17634 @node MicroBlaze Options
17635 @subsection MicroBlaze Options
17636 @cindex MicroBlaze Options
17641 @opindex msoft-float
17642 Use software emulation for floating point (default).
17645 @opindex mhard-float
17646 Use hardware floating-point instructions.
17650 Do not optimize block moves, use @code{memcpy}.
17652 @item -mno-clearbss
17653 @opindex mno-clearbss
17654 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17656 @item -mcpu=@var{cpu-type}
17658 Use features of, and schedule code for, the given CPU.
17659 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17660 where @var{X} is a major version, @var{YY} is the minor version, and
17661 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17662 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17664 @item -mxl-soft-mul
17665 @opindex mxl-soft-mul
17666 Use software multiply emulation (default).
17668 @item -mxl-soft-div
17669 @opindex mxl-soft-div
17670 Use software emulation for divides (default).
17672 @item -mxl-barrel-shift
17673 @opindex mxl-barrel-shift
17674 Use the hardware barrel shifter.
17676 @item -mxl-pattern-compare
17677 @opindex mxl-pattern-compare
17678 Use pattern compare instructions.
17680 @item -msmall-divides
17681 @opindex msmall-divides
17682 Use table lookup optimization for small signed integer divisions.
17684 @item -mxl-stack-check
17685 @opindex mxl-stack-check
17686 This option is deprecated. Use @option{-fstack-check} instead.
17689 @opindex mxl-gp-opt
17690 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17692 @item -mxl-multiply-high
17693 @opindex mxl-multiply-high
17694 Use multiply high instructions for high part of 32x32 multiply.
17696 @item -mxl-float-convert
17697 @opindex mxl-float-convert
17698 Use hardware floating-point conversion instructions.
17700 @item -mxl-float-sqrt
17701 @opindex mxl-float-sqrt
17702 Use hardware floating-point square root instruction.
17705 @opindex mbig-endian
17706 Generate code for a big-endian target.
17708 @item -mlittle-endian
17709 @opindex mlittle-endian
17710 Generate code for a little-endian target.
17713 @opindex mxl-reorder
17714 Use reorder instructions (swap and byte reversed load/store).
17716 @item -mxl-mode-@var{app-model}
17717 Select application model @var{app-model}. Valid models are
17720 normal executable (default), uses startup code @file{crt0.o}.
17723 for use with Xilinx Microprocessor Debugger (XMD) based
17724 software intrusive debug agent called xmdstub. This uses startup file
17725 @file{crt1.o} and sets the start address of the program to 0x800.
17728 for applications that are loaded using a bootloader.
17729 This model uses startup file @file{crt2.o} which does not contain a processor
17730 reset vector handler. This is suitable for transferring control on a
17731 processor reset to the bootloader rather than the application.
17734 for applications that do not require any of the
17735 MicroBlaze vectors. This option may be useful for applications running
17736 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17739 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17740 @option{-mxl-mode-@var{app-model}}.
17745 @subsection MIPS Options
17746 @cindex MIPS options
17752 Generate big-endian code.
17756 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17759 @item -march=@var{arch}
17761 Generate code that runs on @var{arch}, which can be the name of a
17762 generic MIPS ISA, or the name of a particular processor.
17764 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17765 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17766 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17767 @samp{mips64r5} and @samp{mips64r6}.
17768 The processor names are:
17769 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17770 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17771 @samp{5kc}, @samp{5kf},
17773 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17774 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17775 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17776 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17777 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17780 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17782 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17783 @samp{m5100}, @samp{m5101},
17784 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17787 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17788 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17789 @samp{rm7000}, @samp{rm9000},
17790 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17793 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17794 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17795 @samp{xlr} and @samp{xlp}.
17796 The special value @samp{from-abi} selects the
17797 most compatible architecture for the selected ABI (that is,
17798 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17800 The native Linux/GNU toolchain also supports the value @samp{native},
17801 which selects the best architecture option for the host processor.
17802 @option{-march=native} has no effect if GCC does not recognize
17805 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17806 (for example, @option{-march=r2k}). Prefixes are optional, and
17807 @samp{vr} may be written @samp{r}.
17809 Names of the form @samp{@var{n}f2_1} refer to processors with
17810 FPUs clocked at half the rate of the core, names of the form
17811 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17812 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17813 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17814 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17815 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17816 accepted as synonyms for @samp{@var{n}f1_1}.
17818 GCC defines two macros based on the value of this option. The first
17819 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17820 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17821 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17822 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17823 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17825 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17826 above. In other words, it has the full prefix and does not
17827 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17828 the macro names the resolved architecture (either @code{"mips1"} or
17829 @code{"mips3"}). It names the default architecture when no
17830 @option{-march} option is given.
17832 @item -mtune=@var{arch}
17834 Optimize for @var{arch}. Among other things, this option controls
17835 the way instructions are scheduled, and the perceived cost of arithmetic
17836 operations. The list of @var{arch} values is the same as for
17839 When this option is not used, GCC optimizes for the processor
17840 specified by @option{-march}. By using @option{-march} and
17841 @option{-mtune} together, it is possible to generate code that
17842 runs on a family of processors, but optimize the code for one
17843 particular member of that family.
17845 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17846 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17847 @option{-march} ones described above.
17851 Equivalent to @option{-march=mips1}.
17855 Equivalent to @option{-march=mips2}.
17859 Equivalent to @option{-march=mips3}.
17863 Equivalent to @option{-march=mips4}.
17867 Equivalent to @option{-march=mips32}.
17871 Equivalent to @option{-march=mips32r3}.
17875 Equivalent to @option{-march=mips32r5}.
17879 Equivalent to @option{-march=mips32r6}.
17883 Equivalent to @option{-march=mips64}.
17887 Equivalent to @option{-march=mips64r2}.
17891 Equivalent to @option{-march=mips64r3}.
17895 Equivalent to @option{-march=mips64r5}.
17899 Equivalent to @option{-march=mips64r6}.
17904 @opindex mno-mips16
17905 Generate (do not generate) MIPS16 code. If GCC is targeting a
17906 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17908 MIPS16 code generation can also be controlled on a per-function basis
17909 by means of @code{mips16} and @code{nomips16} attributes.
17910 @xref{Function Attributes}, for more information.
17912 @item -mflip-mips16
17913 @opindex mflip-mips16
17914 Generate MIPS16 code on alternating functions. This option is provided
17915 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17916 not intended for ordinary use in compiling user code.
17918 @item -minterlink-compressed
17919 @item -mno-interlink-compressed
17920 @opindex minterlink-compressed
17921 @opindex mno-interlink-compressed
17922 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17923 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17925 For example, code using the standard ISA encoding cannot jump directly
17926 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17927 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17928 knows that the target of the jump is not compressed.
17930 @item -minterlink-mips16
17931 @itemx -mno-interlink-mips16
17932 @opindex minterlink-mips16
17933 @opindex mno-interlink-mips16
17934 Aliases of @option{-minterlink-compressed} and
17935 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17936 and are retained for backwards compatibility.
17948 Generate code for the given ABI@.
17950 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17951 generates 64-bit code when you select a 64-bit architecture, but you
17952 can use @option{-mgp32} to get 32-bit code instead.
17954 For information about the O64 ABI, see
17955 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17957 GCC supports a variant of the o32 ABI in which floating-point registers
17958 are 64 rather than 32 bits wide. You can select this combination with
17959 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17960 and @code{mfhc1} instructions and is therefore only supported for
17961 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17963 The register assignments for arguments and return values remain the
17964 same, but each scalar value is passed in a single 64-bit register
17965 rather than a pair of 32-bit registers. For example, scalar
17966 floating-point values are returned in @samp{$f0} only, not a
17967 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17968 remains the same in that the even-numbered double-precision registers
17971 Two additional variants of the o32 ABI are supported to enable
17972 a transition from 32-bit to 64-bit registers. These are FPXX
17973 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17974 The FPXX extension mandates that all code must execute correctly
17975 when run using 32-bit or 64-bit registers. The code can be interlinked
17976 with either FP32 or FP64, but not both.
17977 The FP64A extension is similar to the FP64 extension but forbids the
17978 use of odd-numbered single-precision registers. This can be used
17979 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17980 processors and allows both FP32 and FP64A code to interlink and
17981 run in the same process without changing FPU modes.
17984 @itemx -mno-abicalls
17986 @opindex mno-abicalls
17987 Generate (do not generate) code that is suitable for SVR4-style
17988 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17993 Generate (do not generate) code that is fully position-independent,
17994 and that can therefore be linked into shared libraries. This option
17995 only affects @option{-mabicalls}.
17997 All @option{-mabicalls} code has traditionally been position-independent,
17998 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17999 as an extension, the GNU toolchain allows executables to use absolute
18000 accesses for locally-binding symbols. It can also use shorter GP
18001 initialization sequences and generate direct calls to locally-defined
18002 functions. This mode is selected by @option{-mno-shared}.
18004 @option{-mno-shared} depends on binutils 2.16 or higher and generates
18005 objects that can only be linked by the GNU linker. However, the option
18006 does not affect the ABI of the final executable; it only affects the ABI
18007 of relocatable objects. Using @option{-mno-shared} generally makes
18008 executables both smaller and quicker.
18010 @option{-mshared} is the default.
18016 Assume (do not assume) that the static and dynamic linkers
18017 support PLTs and copy relocations. This option only affects
18018 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
18019 has no effect without @option{-msym32}.
18021 You can make @option{-mplt} the default by configuring
18022 GCC with @option{--with-mips-plt}. The default is
18023 @option{-mno-plt} otherwise.
18029 Lift (do not lift) the usual restrictions on the size of the global
18032 GCC normally uses a single instruction to load values from the GOT@.
18033 While this is relatively efficient, it only works if the GOT
18034 is smaller than about 64k. Anything larger causes the linker
18035 to report an error such as:
18037 @cindex relocation truncated to fit (MIPS)
18039 relocation truncated to fit: R_MIPS_GOT16 foobar
18042 If this happens, you should recompile your code with @option{-mxgot}.
18043 This works with very large GOTs, although the code is also
18044 less efficient, since it takes three instructions to fetch the
18045 value of a global symbol.
18047 Note that some linkers can create multiple GOTs. If you have such a
18048 linker, you should only need to use @option{-mxgot} when a single object
18049 file accesses more than 64k's worth of GOT entries. Very few do.
18051 These options have no effect unless GCC is generating position
18056 Assume that general-purpose registers are 32 bits wide.
18060 Assume that general-purpose registers are 64 bits wide.
18064 Assume that floating-point registers are 32 bits wide.
18068 Assume that floating-point registers are 64 bits wide.
18072 Do not assume the width of floating-point registers.
18075 @opindex mhard-float
18076 Use floating-point coprocessor instructions.
18079 @opindex msoft-float
18080 Do not use floating-point coprocessor instructions. Implement
18081 floating-point calculations using library calls instead.
18085 Equivalent to @option{-msoft-float}, but additionally asserts that the
18086 program being compiled does not perform any floating-point operations.
18087 This option is presently supported only by some bare-metal MIPS
18088 configurations, where it may select a special set of libraries
18089 that lack all floating-point support (including, for example, the
18090 floating-point @code{printf} formats).
18091 If code compiled with @option{-mno-float} accidentally contains
18092 floating-point operations, it is likely to suffer a link-time
18093 or run-time failure.
18095 @item -msingle-float
18096 @opindex msingle-float
18097 Assume that the floating-point coprocessor only supports single-precision
18100 @item -mdouble-float
18101 @opindex mdouble-float
18102 Assume that the floating-point coprocessor supports double-precision
18103 operations. This is the default.
18106 @itemx -mno-odd-spreg
18107 @opindex modd-spreg
18108 @opindex mno-odd-spreg
18109 Enable the use of odd-numbered single-precision floating-point registers
18110 for the o32 ABI. This is the default for processors that are known to
18111 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
18115 @itemx -mabs=legacy
18117 @opindex mabs=legacy
18118 These options control the treatment of the special not-a-number (NaN)
18119 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
18120 @code{neg.@i{fmt}} machine instructions.
18122 By default or when @option{-mabs=legacy} is used the legacy
18123 treatment is selected. In this case these instructions are considered
18124 arithmetic and avoided where correct operation is required and the
18125 input operand might be a NaN. A longer sequence of instructions that
18126 manipulate the sign bit of floating-point datum manually is used
18127 instead unless the @option{-ffinite-math-only} option has also been
18130 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
18131 this case these instructions are considered non-arithmetic and therefore
18132 operating correctly in all cases, including in particular where the
18133 input operand is a NaN. These instructions are therefore always used
18134 for the respective operations.
18137 @itemx -mnan=legacy
18139 @opindex mnan=legacy
18140 These options control the encoding of the special not-a-number (NaN)
18141 IEEE 754 floating-point data.
18143 The @option{-mnan=legacy} option selects the legacy encoding. In this
18144 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
18145 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
18146 by the first bit of their trailing significand field being 1.
18148 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
18149 this case qNaNs are denoted by the first bit of their trailing
18150 significand field being 1, whereas sNaNs are denoted by the first bit of
18151 their trailing significand field being 0.
18153 The default is @option{-mnan=legacy} unless GCC has been configured with
18154 @option{--with-nan=2008}.
18160 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
18161 implement atomic memory built-in functions. When neither option is
18162 specified, GCC uses the instructions if the target architecture
18165 @option{-mllsc} is useful if the runtime environment can emulate the
18166 instructions and @option{-mno-llsc} can be useful when compiling for
18167 nonstandard ISAs. You can make either option the default by
18168 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
18169 respectively. @option{--with-llsc} is the default for some
18170 configurations; see the installation documentation for details.
18176 Use (do not use) revision 1 of the MIPS DSP ASE@.
18177 @xref{MIPS DSP Built-in Functions}. This option defines the
18178 preprocessor macro @code{__mips_dsp}. It also defines
18179 @code{__mips_dsp_rev} to 1.
18185 Use (do not use) revision 2 of the MIPS DSP ASE@.
18186 @xref{MIPS DSP Built-in Functions}. This option defines the
18187 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
18188 It also defines @code{__mips_dsp_rev} to 2.
18191 @itemx -mno-smartmips
18192 @opindex msmartmips
18193 @opindex mno-smartmips
18194 Use (do not use) the MIPS SmartMIPS ASE.
18196 @item -mpaired-single
18197 @itemx -mno-paired-single
18198 @opindex mpaired-single
18199 @opindex mno-paired-single
18200 Use (do not use) paired-single floating-point instructions.
18201 @xref{MIPS Paired-Single Support}. This option requires
18202 hardware floating-point support to be enabled.
18208 Use (do not use) MIPS Digital Media Extension instructions.
18209 This option can only be used when generating 64-bit code and requires
18210 hardware floating-point support to be enabled.
18215 @opindex mno-mips3d
18216 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
18217 The option @option{-mips3d} implies @option{-mpaired-single}.
18220 @itemx -mno-micromips
18221 @opindex mmicromips
18222 @opindex mno-mmicromips
18223 Generate (do not generate) microMIPS code.
18225 MicroMIPS code generation can also be controlled on a per-function basis
18226 by means of @code{micromips} and @code{nomicromips} attributes.
18227 @xref{Function Attributes}, for more information.
18233 Use (do not use) MT Multithreading instructions.
18239 Use (do not use) the MIPS MCU ASE instructions.
18245 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
18251 Use (do not use) the MIPS Virtualization Application Specific instructions.
18257 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
18261 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
18262 an explanation of the default and the way that the pointer size is
18267 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
18269 The default size of @code{int}s, @code{long}s and pointers depends on
18270 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
18271 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
18272 32-bit @code{long}s. Pointers are the same size as @code{long}s,
18273 or the same size as integer registers, whichever is smaller.
18279 Assume (do not assume) that all symbols have 32-bit values, regardless
18280 of the selected ABI@. This option is useful in combination with
18281 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
18282 to generate shorter and faster references to symbolic addresses.
18286 Put definitions of externally-visible data in a small data section
18287 if that data is no bigger than @var{num} bytes. GCC can then generate
18288 more efficient accesses to the data; see @option{-mgpopt} for details.
18290 The default @option{-G} option depends on the configuration.
18292 @item -mlocal-sdata
18293 @itemx -mno-local-sdata
18294 @opindex mlocal-sdata
18295 @opindex mno-local-sdata
18296 Extend (do not extend) the @option{-G} behavior to local data too,
18297 such as to static variables in C@. @option{-mlocal-sdata} is the
18298 default for all configurations.
18300 If the linker complains that an application is using too much small data,
18301 you might want to try rebuilding the less performance-critical parts with
18302 @option{-mno-local-sdata}. You might also want to build large
18303 libraries with @option{-mno-local-sdata}, so that the libraries leave
18304 more room for the main program.
18306 @item -mextern-sdata
18307 @itemx -mno-extern-sdata
18308 @opindex mextern-sdata
18309 @opindex mno-extern-sdata
18310 Assume (do not assume) that externally-defined data is in
18311 a small data section if the size of that data is within the @option{-G} limit.
18312 @option{-mextern-sdata} is the default for all configurations.
18314 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
18315 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
18316 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
18317 is placed in a small data section. If @var{Var} is defined by another
18318 module, you must either compile that module with a high-enough
18319 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
18320 definition. If @var{Var} is common, you must link the application
18321 with a high-enough @option{-G} setting.
18323 The easiest way of satisfying these restrictions is to compile
18324 and link every module with the same @option{-G} option. However,
18325 you may wish to build a library that supports several different
18326 small data limits. You can do this by compiling the library with
18327 the highest supported @option{-G} setting and additionally using
18328 @option{-mno-extern-sdata} to stop the library from making assumptions
18329 about externally-defined data.
18335 Use (do not use) GP-relative accesses for symbols that are known to be
18336 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
18337 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
18340 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
18341 might not hold the value of @code{_gp}. For example, if the code is
18342 part of a library that might be used in a boot monitor, programs that
18343 call boot monitor routines pass an unknown value in @code{$gp}.
18344 (In such situations, the boot monitor itself is usually compiled
18345 with @option{-G0}.)
18347 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
18348 @option{-mno-extern-sdata}.
18350 @item -membedded-data
18351 @itemx -mno-embedded-data
18352 @opindex membedded-data
18353 @opindex mno-embedded-data
18354 Allocate variables to the read-only data section first if possible, then
18355 next in the small data section if possible, otherwise in data. This gives
18356 slightly slower code than the default, but reduces the amount of RAM required
18357 when executing, and thus may be preferred for some embedded systems.
18359 @item -muninit-const-in-rodata
18360 @itemx -mno-uninit-const-in-rodata
18361 @opindex muninit-const-in-rodata
18362 @opindex mno-uninit-const-in-rodata
18363 Put uninitialized @code{const} variables in the read-only data section.
18364 This option is only meaningful in conjunction with @option{-membedded-data}.
18366 @item -mcode-readable=@var{setting}
18367 @opindex mcode-readable
18368 Specify whether GCC may generate code that reads from executable sections.
18369 There are three possible settings:
18372 @item -mcode-readable=yes
18373 Instructions may freely access executable sections. This is the
18376 @item -mcode-readable=pcrel
18377 MIPS16 PC-relative load instructions can access executable sections,
18378 but other instructions must not do so. This option is useful on 4KSc
18379 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
18380 It is also useful on processors that can be configured to have a dual
18381 instruction/data SRAM interface and that, like the M4K, automatically
18382 redirect PC-relative loads to the instruction RAM.
18384 @item -mcode-readable=no
18385 Instructions must not access executable sections. This option can be
18386 useful on targets that are configured to have a dual instruction/data
18387 SRAM interface but that (unlike the M4K) do not automatically redirect
18388 PC-relative loads to the instruction RAM.
18391 @item -msplit-addresses
18392 @itemx -mno-split-addresses
18393 @opindex msplit-addresses
18394 @opindex mno-split-addresses
18395 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
18396 relocation operators. This option has been superseded by
18397 @option{-mexplicit-relocs} but is retained for backwards compatibility.
18399 @item -mexplicit-relocs
18400 @itemx -mno-explicit-relocs
18401 @opindex mexplicit-relocs
18402 @opindex mno-explicit-relocs
18403 Use (do not use) assembler relocation operators when dealing with symbolic
18404 addresses. The alternative, selected by @option{-mno-explicit-relocs},
18405 is to use assembler macros instead.
18407 @option{-mexplicit-relocs} is the default if GCC was configured
18408 to use an assembler that supports relocation operators.
18410 @item -mcheck-zero-division
18411 @itemx -mno-check-zero-division
18412 @opindex mcheck-zero-division
18413 @opindex mno-check-zero-division
18414 Trap (do not trap) on integer division by zero.
18416 The default is @option{-mcheck-zero-division}.
18418 @item -mdivide-traps
18419 @itemx -mdivide-breaks
18420 @opindex mdivide-traps
18421 @opindex mdivide-breaks
18422 MIPS systems check for division by zero by generating either a
18423 conditional trap or a break instruction. Using traps results in
18424 smaller code, but is only supported on MIPS II and later. Also, some
18425 versions of the Linux kernel have a bug that prevents trap from
18426 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
18427 allow conditional traps on architectures that support them and
18428 @option{-mdivide-breaks} to force the use of breaks.
18430 The default is usually @option{-mdivide-traps}, but this can be
18431 overridden at configure time using @option{--with-divide=breaks}.
18432 Divide-by-zero checks can be completely disabled using
18433 @option{-mno-check-zero-division}.
18438 @opindex mno-memcpy
18439 Force (do not force) the use of @code{memcpy} for non-trivial block
18440 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
18441 most constant-sized copies.
18444 @itemx -mno-long-calls
18445 @opindex mlong-calls
18446 @opindex mno-long-calls
18447 Disable (do not disable) use of the @code{jal} instruction. Calling
18448 functions using @code{jal} is more efficient but requires the caller
18449 and callee to be in the same 256 megabyte segment.
18451 This option has no effect on abicalls code. The default is
18452 @option{-mno-long-calls}.
18458 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
18459 instructions, as provided by the R4650 ISA@.
18465 Enable (disable) use of the @code{madd} and @code{msub} integer
18466 instructions. The default is @option{-mimadd} on architectures
18467 that support @code{madd} and @code{msub} except for the 74k
18468 architecture where it was found to generate slower code.
18471 @itemx -mno-fused-madd
18472 @opindex mfused-madd
18473 @opindex mno-fused-madd
18474 Enable (disable) use of the floating-point multiply-accumulate
18475 instructions, when they are available. The default is
18476 @option{-mfused-madd}.
18478 On the R8000 CPU when multiply-accumulate instructions are used,
18479 the intermediate product is calculated to infinite precision
18480 and is not subject to the FCSR Flush to Zero bit. This may be
18481 undesirable in some circumstances. On other processors the result
18482 is numerically identical to the equivalent computation using
18483 separate multiply, add, subtract and negate instructions.
18487 Tell the MIPS assembler to not run its preprocessor over user
18488 assembler files (with a @samp{.s} suffix) when assembling them.
18493 @opindex mno-fix-24k
18494 Work around the 24K E48 (lost data on stores during refill) errata.
18495 The workarounds are implemented by the assembler rather than by GCC@.
18498 @itemx -mno-fix-r4000
18499 @opindex mfix-r4000
18500 @opindex mno-fix-r4000
18501 Work around certain R4000 CPU errata:
18504 A double-word or a variable shift may give an incorrect result if executed
18505 immediately after starting an integer division.
18507 A double-word or a variable shift may give an incorrect result if executed
18508 while an integer multiplication is in progress.
18510 An integer division may give an incorrect result if started in a delay slot
18511 of a taken branch or a jump.
18515 @itemx -mno-fix-r4400
18516 @opindex mfix-r4400
18517 @opindex mno-fix-r4400
18518 Work around certain R4400 CPU errata:
18521 A double-word or a variable shift may give an incorrect result if executed
18522 immediately after starting an integer division.
18526 @itemx -mno-fix-r10000
18527 @opindex mfix-r10000
18528 @opindex mno-fix-r10000
18529 Work around certain R10000 errata:
18532 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18533 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18536 This option can only be used if the target architecture supports
18537 branch-likely instructions. @option{-mfix-r10000} is the default when
18538 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18542 @itemx -mno-fix-rm7000
18543 @opindex mfix-rm7000
18544 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18545 workarounds are implemented by the assembler rather than by GCC@.
18548 @itemx -mno-fix-vr4120
18549 @opindex mfix-vr4120
18550 Work around certain VR4120 errata:
18553 @code{dmultu} does not always produce the correct result.
18555 @code{div} and @code{ddiv} do not always produce the correct result if one
18556 of the operands is negative.
18558 The workarounds for the division errata rely on special functions in
18559 @file{libgcc.a}. At present, these functions are only provided by
18560 the @code{mips64vr*-elf} configurations.
18562 Other VR4120 errata require a NOP to be inserted between certain pairs of
18563 instructions. These errata are handled by the assembler, not by GCC itself.
18566 @opindex mfix-vr4130
18567 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18568 workarounds are implemented by the assembler rather than by GCC,
18569 although GCC avoids using @code{mflo} and @code{mfhi} if the
18570 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18571 instructions are available instead.
18574 @itemx -mno-fix-sb1
18576 Work around certain SB-1 CPU core errata.
18577 (This flag currently works around the SB-1 revision 2
18578 ``F1'' and ``F2'' floating-point errata.)
18580 @item -mr10k-cache-barrier=@var{setting}
18581 @opindex mr10k-cache-barrier
18582 Specify whether GCC should insert cache barriers to avoid the
18583 side-effects of speculation on R10K processors.
18585 In common with many processors, the R10K tries to predict the outcome
18586 of a conditional branch and speculatively executes instructions from
18587 the ``taken'' branch. It later aborts these instructions if the
18588 predicted outcome is wrong. However, on the R10K, even aborted
18589 instructions can have side effects.
18591 This problem only affects kernel stores and, depending on the system,
18592 kernel loads. As an example, a speculatively-executed store may load
18593 the target memory into cache and mark the cache line as dirty, even if
18594 the store itself is later aborted. If a DMA operation writes to the
18595 same area of memory before the ``dirty'' line is flushed, the cached
18596 data overwrites the DMA-ed data. See the R10K processor manual
18597 for a full description, including other potential problems.
18599 One workaround is to insert cache barrier instructions before every memory
18600 access that might be speculatively executed and that might have side
18601 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18602 controls GCC's implementation of this workaround. It assumes that
18603 aborted accesses to any byte in the following regions does not have
18608 the memory occupied by the current function's stack frame;
18611 the memory occupied by an incoming stack argument;
18614 the memory occupied by an object with a link-time-constant address.
18617 It is the kernel's responsibility to ensure that speculative
18618 accesses to these regions are indeed safe.
18620 If the input program contains a function declaration such as:
18626 then the implementation of @code{foo} must allow @code{j foo} and
18627 @code{jal foo} to be executed speculatively. GCC honors this
18628 restriction for functions it compiles itself. It expects non-GCC
18629 functions (such as hand-written assembly code) to do the same.
18631 The option has three forms:
18634 @item -mr10k-cache-barrier=load-store
18635 Insert a cache barrier before a load or store that might be
18636 speculatively executed and that might have side effects even
18639 @item -mr10k-cache-barrier=store
18640 Insert a cache barrier before a store that might be speculatively
18641 executed and that might have side effects even if aborted.
18643 @item -mr10k-cache-barrier=none
18644 Disable the insertion of cache barriers. This is the default setting.
18647 @item -mflush-func=@var{func}
18648 @itemx -mno-flush-func
18649 @opindex mflush-func
18650 Specifies the function to call to flush the I and D caches, or to not
18651 call any such function. If called, the function must take the same
18652 arguments as the common @code{_flush_func}, that is, the address of the
18653 memory range for which the cache is being flushed, the size of the
18654 memory range, and the number 3 (to flush both caches). The default
18655 depends on the target GCC was configured for, but commonly is either
18656 @code{_flush_func} or @code{__cpu_flush}.
18658 @item mbranch-cost=@var{num}
18659 @opindex mbranch-cost
18660 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18661 This cost is only a heuristic and is not guaranteed to produce
18662 consistent results across releases. A zero cost redundantly selects
18663 the default, which is based on the @option{-mtune} setting.
18665 @item -mbranch-likely
18666 @itemx -mno-branch-likely
18667 @opindex mbranch-likely
18668 @opindex mno-branch-likely
18669 Enable or disable use of Branch Likely instructions, regardless of the
18670 default for the selected architecture. By default, Branch Likely
18671 instructions may be generated if they are supported by the selected
18672 architecture. An exception is for the MIPS32 and MIPS64 architectures
18673 and processors that implement those architectures; for those, Branch
18674 Likely instructions are not be generated by default because the MIPS32
18675 and MIPS64 architectures specifically deprecate their use.
18677 @item -mcompact-branches=never
18678 @itemx -mcompact-branches=optimal
18679 @itemx -mcompact-branches=always
18680 @opindex mcompact-branches=never
18681 @opindex mcompact-branches=optimal
18682 @opindex mcompact-branches=always
18683 These options control which form of branches will be generated. The
18684 default is @option{-mcompact-branches=optimal}.
18686 The @option{-mcompact-branches=never} option ensures that compact branch
18687 instructions will never be generated.
18689 The @option{-mcompact-branches=always} option ensures that a compact
18690 branch instruction will be generated if available. If a compact branch
18691 instruction is not available, a delay slot form of the branch will be
18694 This option is supported from MIPS Release 6 onwards.
18696 The @option{-mcompact-branches=optimal} option will cause a delay slot
18697 branch to be used if one is available in the current ISA and the delay
18698 slot is successfully filled. If the delay slot is not filled, a compact
18699 branch will be chosen if one is available.
18701 @item -mfp-exceptions
18702 @itemx -mno-fp-exceptions
18703 @opindex mfp-exceptions
18704 Specifies whether FP exceptions are enabled. This affects how
18705 FP instructions are scheduled for some processors.
18706 The default is that FP exceptions are
18709 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18710 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18713 @item -mvr4130-align
18714 @itemx -mno-vr4130-align
18715 @opindex mvr4130-align
18716 The VR4130 pipeline is two-way superscalar, but can only issue two
18717 instructions together if the first one is 8-byte aligned. When this
18718 option is enabled, GCC aligns pairs of instructions that it
18719 thinks should execute in parallel.
18721 This option only has an effect when optimizing for the VR4130.
18722 It normally makes code faster, but at the expense of making it bigger.
18723 It is enabled by default at optimization level @option{-O3}.
18728 Enable (disable) generation of @code{synci} instructions on
18729 architectures that support it. The @code{synci} instructions (if
18730 enabled) are generated when @code{__builtin___clear_cache} is
18733 This option defaults to @option{-mno-synci}, but the default can be
18734 overridden by configuring GCC with @option{--with-synci}.
18736 When compiling code for single processor systems, it is generally safe
18737 to use @code{synci}. However, on many multi-core (SMP) systems, it
18738 does not invalidate the instruction caches on all cores and may lead
18739 to undefined behavior.
18741 @item -mrelax-pic-calls
18742 @itemx -mno-relax-pic-calls
18743 @opindex mrelax-pic-calls
18744 Try to turn PIC calls that are normally dispatched via register
18745 @code{$25} into direct calls. This is only possible if the linker can
18746 resolve the destination at link time and if the destination is within
18747 range for a direct call.
18749 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18750 an assembler and a linker that support the @code{.reloc} assembly
18751 directive and @option{-mexplicit-relocs} is in effect. With
18752 @option{-mno-explicit-relocs}, this optimization can be performed by the
18753 assembler and the linker alone without help from the compiler.
18755 @item -mmcount-ra-address
18756 @itemx -mno-mcount-ra-address
18757 @opindex mmcount-ra-address
18758 @opindex mno-mcount-ra-address
18759 Emit (do not emit) code that allows @code{_mcount} to modify the
18760 calling function's return address. When enabled, this option extends
18761 the usual @code{_mcount} interface with a new @var{ra-address}
18762 parameter, which has type @code{intptr_t *} and is passed in register
18763 @code{$12}. @code{_mcount} can then modify the return address by
18764 doing both of the following:
18767 Returning the new address in register @code{$31}.
18769 Storing the new address in @code{*@var{ra-address}},
18770 if @var{ra-address} is nonnull.
18773 The default is @option{-mno-mcount-ra-address}.
18775 @item -mframe-header-opt
18776 @itemx -mno-frame-header-opt
18777 @opindex mframe-header-opt
18778 Enable (disable) frame header optimization in the o32 ABI. When using the
18779 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
18780 function to write out register arguments. When enabled, this optimization
18781 will suppress the allocation of the frame header if it can be determined that
18784 This optimization is off by default at all optimization levels.
18789 @subsection MMIX Options
18790 @cindex MMIX Options
18792 These options are defined for the MMIX:
18796 @itemx -mno-libfuncs
18798 @opindex mno-libfuncs
18799 Specify that intrinsic library functions are being compiled, passing all
18800 values in registers, no matter the size.
18803 @itemx -mno-epsilon
18805 @opindex mno-epsilon
18806 Generate floating-point comparison instructions that compare with respect
18807 to the @code{rE} epsilon register.
18809 @item -mabi=mmixware
18811 @opindex mabi=mmixware
18813 Generate code that passes function parameters and return values that (in
18814 the called function) are seen as registers @code{$0} and up, as opposed to
18815 the GNU ABI which uses global registers @code{$231} and up.
18817 @item -mzero-extend
18818 @itemx -mno-zero-extend
18819 @opindex mzero-extend
18820 @opindex mno-zero-extend
18821 When reading data from memory in sizes shorter than 64 bits, use (do not
18822 use) zero-extending load instructions by default, rather than
18823 sign-extending ones.
18826 @itemx -mno-knuthdiv
18828 @opindex mno-knuthdiv
18829 Make the result of a division yielding a remainder have the same sign as
18830 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18831 remainder follows the sign of the dividend. Both methods are
18832 arithmetically valid, the latter being almost exclusively used.
18834 @item -mtoplevel-symbols
18835 @itemx -mno-toplevel-symbols
18836 @opindex mtoplevel-symbols
18837 @opindex mno-toplevel-symbols
18838 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18839 code can be used with the @code{PREFIX} assembly directive.
18843 Generate an executable in the ELF format, rather than the default
18844 @samp{mmo} format used by the @command{mmix} simulator.
18846 @item -mbranch-predict
18847 @itemx -mno-branch-predict
18848 @opindex mbranch-predict
18849 @opindex mno-branch-predict
18850 Use (do not use) the probable-branch instructions, when static branch
18851 prediction indicates a probable branch.
18853 @item -mbase-addresses
18854 @itemx -mno-base-addresses
18855 @opindex mbase-addresses
18856 @opindex mno-base-addresses
18857 Generate (do not generate) code that uses @emph{base addresses}. Using a
18858 base address automatically generates a request (handled by the assembler
18859 and the linker) for a constant to be set up in a global register. The
18860 register is used for one or more base address requests within the range 0
18861 to 255 from the value held in the register. The generally leads to short
18862 and fast code, but the number of different data items that can be
18863 addressed is limited. This means that a program that uses lots of static
18864 data may require @option{-mno-base-addresses}.
18866 @item -msingle-exit
18867 @itemx -mno-single-exit
18868 @opindex msingle-exit
18869 @opindex mno-single-exit
18870 Force (do not force) generated code to have a single exit point in each
18874 @node MN10300 Options
18875 @subsection MN10300 Options
18876 @cindex MN10300 options
18878 These @option{-m} options are defined for Matsushita MN10300 architectures:
18883 Generate code to avoid bugs in the multiply instructions for the MN10300
18884 processors. This is the default.
18886 @item -mno-mult-bug
18887 @opindex mno-mult-bug
18888 Do not generate code to avoid bugs in the multiply instructions for the
18889 MN10300 processors.
18893 Generate code using features specific to the AM33 processor.
18897 Do not generate code using features specific to the AM33 processor. This
18902 Generate code using features specific to the AM33/2.0 processor.
18906 Generate code using features specific to the AM34 processor.
18908 @item -mtune=@var{cpu-type}
18910 Use the timing characteristics of the indicated CPU type when
18911 scheduling instructions. This does not change the targeted processor
18912 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18913 @samp{am33-2} or @samp{am34}.
18915 @item -mreturn-pointer-on-d0
18916 @opindex mreturn-pointer-on-d0
18917 When generating a function that returns a pointer, return the pointer
18918 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18919 only in @code{a0}, and attempts to call such functions without a prototype
18920 result in errors. Note that this option is on by default; use
18921 @option{-mno-return-pointer-on-d0} to disable it.
18925 Do not link in the C run-time initialization object file.
18929 Indicate to the linker that it should perform a relaxation optimization pass
18930 to shorten branches, calls and absolute memory addresses. This option only
18931 has an effect when used on the command line for the final link step.
18933 This option makes symbolic debugging impossible.
18937 Allow the compiler to generate @emph{Long Instruction Word}
18938 instructions if the target is the @samp{AM33} or later. This is the
18939 default. This option defines the preprocessor macro @code{__LIW__}.
18943 Do not allow the compiler to generate @emph{Long Instruction Word}
18944 instructions. This option defines the preprocessor macro
18949 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18950 instructions if the target is the @samp{AM33} or later. This is the
18951 default. This option defines the preprocessor macro @code{__SETLB__}.
18955 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18956 instructions. This option defines the preprocessor macro
18957 @code{__NO_SETLB__}.
18961 @node Moxie Options
18962 @subsection Moxie Options
18963 @cindex Moxie Options
18969 Generate big-endian code. This is the default for @samp{moxie-*-*}
18974 Generate little-endian code.
18978 Generate mul.x and umul.x instructions. This is the default for
18979 @samp{moxiebox-*-*} configurations.
18983 Do not link in the C run-time initialization object file.
18987 @node MSP430 Options
18988 @subsection MSP430 Options
18989 @cindex MSP430 Options
18991 These options are defined for the MSP430:
18997 Force assembly output to always use hex constants. Normally such
18998 constants are signed decimals, but this option is available for
18999 testsuite and/or aesthetic purposes.
19003 Select the MCU to target. This is used to create a C preprocessor
19004 symbol based upon the MCU name, converted to upper case and pre- and
19005 post-fixed with @samp{__}. This in turn is used by the
19006 @file{msp430.h} header file to select an MCU-specific supplementary
19009 The option also sets the ISA to use. If the MCU name is one that is
19010 known to only support the 430 ISA then that is selected, otherwise the
19011 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
19012 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
19013 name selects the 430X ISA.
19015 In addition an MCU-specific linker script is added to the linker
19016 command line. The script's name is the name of the MCU with
19017 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
19018 command line defines the C preprocessor symbol @code{__XXX__} and
19019 cause the linker to search for a script called @file{xxx.ld}.
19021 This option is also passed on to the assembler.
19024 @itemx -mno-warn-mcu
19026 @opindex mno-warn-mcu
19027 This option enables or disables warnings about conflicts between the
19028 MCU name specified by the @option{-mmcu} option and the ISA set by the
19029 @option{-mcpu} option and/or the hardware multiply support set by the
19030 @option{-mhwmult} option. It also toggles warnings about unrecognized
19031 MCU names. This option is on by default.
19035 Specifies the ISA to use. Accepted values are @samp{msp430},
19036 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
19037 @option{-mmcu=} option should be used to select the ISA.
19041 Link to the simulator runtime libraries and linker script. Overrides
19042 any scripts that would be selected by the @option{-mmcu=} option.
19046 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
19050 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
19054 This option is passed to the assembler and linker, and allows the
19055 linker to perform certain optimizations that cannot be done until
19060 Describes the type of hardware multiply supported by the target.
19061 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
19062 for the original 16-bit-only multiply supported by early MCUs.
19063 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
19064 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
19065 A value of @samp{auto} can also be given. This tells GCC to deduce
19066 the hardware multiply support based upon the MCU name provided by the
19067 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
19068 the MCU name is not recognized then no hardware multiply support is
19069 assumed. @code{auto} is the default setting.
19071 Hardware multiplies are normally performed by calling a library
19072 routine. This saves space in the generated code. When compiling at
19073 @option{-O3} or higher however the hardware multiplier is invoked
19074 inline. This makes for bigger, but faster code.
19076 The hardware multiply routines disable interrupts whilst running and
19077 restore the previous interrupt state when they finish. This makes
19078 them safe to use inside interrupt handlers as well as in normal code.
19082 Enable the use of a minimum runtime environment - no static
19083 initializers or constructors. This is intended for memory-constrained
19084 devices. The compiler includes special symbols in some objects
19085 that tell the linker and runtime which code fragments are required.
19087 @item -mcode-region=
19088 @itemx -mdata-region=
19089 @opindex mcode-region
19090 @opindex mdata-region
19091 These options tell the compiler where to place functions and data that
19092 do not have one of the @code{lower}, @code{upper}, @code{either} or
19093 @code{section} attributes. Possible values are @code{lower},
19094 @code{upper}, @code{either} or @code{any}. The first three behave
19095 like the corresponding attribute. The fourth possible value -
19096 @code{any} - is the default. It leaves placement entirely up to the
19097 linker script and how it assigns the standard sections
19098 (@code{.text}, @code{.data}, etc) to the memory regions.
19100 @item -msilicon-errata=
19101 @opindex msilicon-errata
19102 This option passes on a request to assembler to enable the fixes for
19103 the named silicon errata.
19105 @item -msilicon-errata-warn=
19106 @opindex msilicon-errata-warn
19107 This option passes on a request to the assembler to enable warning
19108 messages when a silicon errata might need to be applied.
19112 @node NDS32 Options
19113 @subsection NDS32 Options
19114 @cindex NDS32 Options
19116 These options are defined for NDS32 implementations:
19121 @opindex mbig-endian
19122 Generate code in big-endian mode.
19124 @item -mlittle-endian
19125 @opindex mlittle-endian
19126 Generate code in little-endian mode.
19128 @item -mreduced-regs
19129 @opindex mreduced-regs
19130 Use reduced-set registers for register allocation.
19133 @opindex mfull-regs
19134 Use full-set registers for register allocation.
19138 Generate conditional move instructions.
19142 Do not generate conditional move instructions.
19146 Generate performance extension instructions.
19148 @item -mno-perf-ext
19149 @opindex mno-perf-ext
19150 Do not generate performance extension instructions.
19154 Generate v3 push25/pop25 instructions.
19157 @opindex mno-v3push
19158 Do not generate v3 push25/pop25 instructions.
19162 Generate 16-bit instructions.
19165 @opindex mno-16-bit
19166 Do not generate 16-bit instructions.
19168 @item -misr-vector-size=@var{num}
19169 @opindex misr-vector-size
19170 Specify the size of each interrupt vector, which must be 4 or 16.
19172 @item -mcache-block-size=@var{num}
19173 @opindex mcache-block-size
19174 Specify the size of each cache block,
19175 which must be a power of 2 between 4 and 512.
19177 @item -march=@var{arch}
19179 Specify the name of the target architecture.
19181 @item -mcmodel=@var{code-model}
19183 Set the code model to one of
19186 All the data and read-only data segments must be within 512KB addressing space.
19187 The text segment must be within 16MB addressing space.
19188 @item @samp{medium}
19189 The data segment must be within 512KB while the read-only data segment can be
19190 within 4GB addressing space. The text segment should be still within 16MB
19193 All the text and data segments can be within 4GB addressing space.
19197 @opindex mctor-dtor
19198 Enable constructor/destructor feature.
19202 Guide linker to relax instructions.
19206 @node Nios II Options
19207 @subsection Nios II Options
19208 @cindex Nios II options
19209 @cindex Altera Nios II options
19211 These are the options defined for the Altera Nios II processor.
19217 @cindex smaller data references
19218 Put global and static objects less than or equal to @var{num} bytes
19219 into the small data or BSS sections instead of the normal data or BSS
19220 sections. The default value of @var{num} is 8.
19222 @item -mgpopt=@var{option}
19227 Generate (do not generate) GP-relative accesses. The following
19228 @var{option} names are recognized:
19233 Do not generate GP-relative accesses.
19236 Generate GP-relative accesses for small data objects that are not
19237 external, weak, or uninitialized common symbols.
19238 Also use GP-relative addressing for objects that
19239 have been explicitly placed in a small data section via a @code{section}
19243 As for @samp{local}, but also generate GP-relative accesses for
19244 small data objects that are external, weak, or common. If you use this option,
19245 you must ensure that all parts of your program (including libraries) are
19246 compiled with the same @option{-G} setting.
19249 Generate GP-relative accesses for all data objects in the program. If you
19250 use this option, the entire data and BSS segments
19251 of your program must fit in 64K of memory and you must use an appropriate
19252 linker script to allocate them within the addressable range of the
19256 Generate GP-relative addresses for function pointers as well as data
19257 pointers. If you use this option, the entire text, data, and BSS segments
19258 of your program must fit in 64K of memory and you must use an appropriate
19259 linker script to allocate them within the addressable range of the
19264 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
19265 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
19267 The default is @option{-mgpopt} except when @option{-fpic} or
19268 @option{-fPIC} is specified to generate position-independent code.
19269 Note that the Nios II ABI does not permit GP-relative accesses from
19272 You may need to specify @option{-mno-gpopt} explicitly when building
19273 programs that include large amounts of small data, including large
19274 GOT data sections. In this case, the 16-bit offset for GP-relative
19275 addressing may not be large enough to allow access to the entire
19276 small data section.
19282 Generate little-endian (default) or big-endian (experimental) code,
19285 @item -march=@var{arch}
19287 This specifies the name of the target Nios II architecture. GCC uses this
19288 name to determine what kind of instructions it can emit when generating
19289 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
19291 The preprocessor macro @code{__nios2_arch__} is available to programs,
19292 with value 1 or 2, indicating the targeted ISA level.
19294 @item -mbypass-cache
19295 @itemx -mno-bypass-cache
19296 @opindex mno-bypass-cache
19297 @opindex mbypass-cache
19298 Force all load and store instructions to always bypass cache by
19299 using I/O variants of the instructions. The default is not to
19302 @item -mno-cache-volatile
19303 @itemx -mcache-volatile
19304 @opindex mcache-volatile
19305 @opindex mno-cache-volatile
19306 Volatile memory access bypass the cache using the I/O variants of
19307 the load and store instructions. The default is not to bypass the cache.
19309 @item -mno-fast-sw-div
19310 @itemx -mfast-sw-div
19311 @opindex mno-fast-sw-div
19312 @opindex mfast-sw-div
19313 Do not use table-based fast divide for small numbers. The default
19314 is to use the fast divide at @option{-O3} and above.
19318 @itemx -mno-hw-mulx
19322 @opindex mno-hw-mul
19324 @opindex mno-hw-mulx
19326 @opindex mno-hw-div
19328 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
19329 instructions by the compiler. The default is to emit @code{mul}
19330 and not emit @code{div} and @code{mulx}.
19336 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
19337 CDX (code density) instructions. Enabling these instructions also
19338 requires @option{-march=r2}. Since these instructions are optional
19339 extensions to the R2 architecture, the default is not to emit them.
19341 @item -mcustom-@var{insn}=@var{N}
19342 @itemx -mno-custom-@var{insn}
19343 @opindex mcustom-@var{insn}
19344 @opindex mno-custom-@var{insn}
19345 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
19346 custom instruction with encoding @var{N} when generating code that uses
19347 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
19348 instruction 253 for single-precision floating-point add operations instead
19349 of the default behavior of using a library call.
19351 The following values of @var{insn} are supported. Except as otherwise
19352 noted, floating-point operations are expected to be implemented with
19353 normal IEEE 754 semantics and correspond directly to the C operators or the
19354 equivalent GCC built-in functions (@pxref{Other Builtins}).
19356 Single-precision floating point:
19359 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
19360 Binary arithmetic operations.
19366 Unary absolute value.
19368 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
19369 Comparison operations.
19371 @item @samp{fmins}, @samp{fmaxs}
19372 Floating-point minimum and maximum. These instructions are only
19373 generated if @option{-ffinite-math-only} is specified.
19375 @item @samp{fsqrts}
19376 Unary square root operation.
19378 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
19379 Floating-point trigonometric and exponential functions. These instructions
19380 are only generated if @option{-funsafe-math-optimizations} is also specified.
19384 Double-precision floating point:
19387 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
19388 Binary arithmetic operations.
19394 Unary absolute value.
19396 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
19397 Comparison operations.
19399 @item @samp{fmind}, @samp{fmaxd}
19400 Double-precision minimum and maximum. These instructions are only
19401 generated if @option{-ffinite-math-only} is specified.
19403 @item @samp{fsqrtd}
19404 Unary square root operation.
19406 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
19407 Double-precision trigonometric and exponential functions. These instructions
19408 are only generated if @option{-funsafe-math-optimizations} is also specified.
19414 @item @samp{fextsd}
19415 Conversion from single precision to double precision.
19417 @item @samp{ftruncds}
19418 Conversion from double precision to single precision.
19420 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
19421 Conversion from floating point to signed or unsigned integer types, with
19422 truncation towards zero.
19425 Conversion from single-precision floating point to signed integer,
19426 rounding to the nearest integer and ties away from zero.
19427 This corresponds to the @code{__builtin_lroundf} function when
19428 @option{-fno-math-errno} is used.
19430 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
19431 Conversion from signed or unsigned integer types to floating-point types.
19435 In addition, all of the following transfer instructions for internal
19436 registers X and Y must be provided to use any of the double-precision
19437 floating-point instructions. Custom instructions taking two
19438 double-precision source operands expect the first operand in the
19439 64-bit register X. The other operand (or only operand of a unary
19440 operation) is given to the custom arithmetic instruction with the
19441 least significant half in source register @var{src1} and the most
19442 significant half in @var{src2}. A custom instruction that returns a
19443 double-precision result returns the most significant 32 bits in the
19444 destination register and the other half in 32-bit register Y.
19445 GCC automatically generates the necessary code sequences to write
19446 register X and/or read register Y when double-precision floating-point
19447 instructions are used.
19452 Write @var{src1} into the least significant half of X and @var{src2} into
19453 the most significant half of X.
19456 Write @var{src1} into Y.
19458 @item @samp{frdxhi}, @samp{frdxlo}
19459 Read the most or least (respectively) significant half of X and store it in
19463 Read the value of Y and store it into @var{dest}.
19466 Note that you can gain more local control over generation of Nios II custom
19467 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
19468 and @code{target("no-custom-@var{insn}")} function attributes
19469 (@pxref{Function Attributes})
19470 or pragmas (@pxref{Function Specific Option Pragmas}).
19472 @item -mcustom-fpu-cfg=@var{name}
19473 @opindex mcustom-fpu-cfg
19475 This option enables a predefined, named set of custom instruction encodings
19476 (see @option{-mcustom-@var{insn}} above).
19477 Currently, the following sets are defined:
19479 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
19480 @gccoptlist{-mcustom-fmuls=252 @gol
19481 -mcustom-fadds=253 @gol
19482 -mcustom-fsubs=254 @gol
19483 -fsingle-precision-constant}
19485 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
19486 @gccoptlist{-mcustom-fmuls=252 @gol
19487 -mcustom-fadds=253 @gol
19488 -mcustom-fsubs=254 @gol
19489 -mcustom-fdivs=255 @gol
19490 -fsingle-precision-constant}
19492 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19493 @gccoptlist{-mcustom-floatus=243 @gol
19494 -mcustom-fixsi=244 @gol
19495 -mcustom-floatis=245 @gol
19496 -mcustom-fcmpgts=246 @gol
19497 -mcustom-fcmples=249 @gol
19498 -mcustom-fcmpeqs=250 @gol
19499 -mcustom-fcmpnes=251 @gol
19500 -mcustom-fmuls=252 @gol
19501 -mcustom-fadds=253 @gol
19502 -mcustom-fsubs=254 @gol
19503 -mcustom-fdivs=255 @gol
19504 -fsingle-precision-constant}
19506 Custom instruction assignments given by individual
19507 @option{-mcustom-@var{insn}=} options override those given by
19508 @option{-mcustom-fpu-cfg=}, regardless of the
19509 order of the options on the command line.
19511 Note that you can gain more local control over selection of a FPU
19512 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19513 function attribute (@pxref{Function Attributes})
19514 or pragma (@pxref{Function Specific Option Pragmas}).
19518 These additional @samp{-m} options are available for the Altera Nios II
19519 ELF (bare-metal) target:
19525 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19526 startup and termination code, and is typically used in conjunction with
19527 @option{-msys-crt0=} to specify the location of the alternate startup code
19528 provided by the HAL BSP.
19532 Link with a limited version of the C library, @option{-lsmallc}, rather than
19535 @item -msys-crt0=@var{startfile}
19537 @var{startfile} is the file name of the startfile (crt0) to use
19538 when linking. This option is only useful in conjunction with @option{-mhal}.
19540 @item -msys-lib=@var{systemlib}
19542 @var{systemlib} is the library name of the library that provides
19543 low-level system calls required by the C library,
19544 e.g. @code{read} and @code{write}.
19545 This option is typically used to link with a library provided by a HAL BSP.
19549 @node Nvidia PTX Options
19550 @subsection Nvidia PTX Options
19551 @cindex Nvidia PTX options
19552 @cindex nvptx options
19554 These options are defined for Nvidia PTX:
19562 Generate code for 32-bit or 64-bit ABI.
19565 @opindex mmainkernel
19566 Link in code for a __main kernel. This is for stand-alone instead of
19567 offloading execution.
19571 Apply partitioned execution optimizations. This is the default when any
19572 level of optimization is selected.
19576 @node PDP-11 Options
19577 @subsection PDP-11 Options
19578 @cindex PDP-11 Options
19580 These options are defined for the PDP-11:
19585 Use hardware FPP floating point. This is the default. (FIS floating
19586 point on the PDP-11/40 is not supported.)
19589 @opindex msoft-float
19590 Do not use hardware floating point.
19594 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19598 Return floating-point results in memory. This is the default.
19602 Generate code for a PDP-11/40.
19606 Generate code for a PDP-11/45. This is the default.
19610 Generate code for a PDP-11/10.
19612 @item -mbcopy-builtin
19613 @opindex mbcopy-builtin
19614 Use inline @code{movmemhi} patterns for copying memory. This is the
19619 Do not use inline @code{movmemhi} patterns for copying memory.
19625 Use 16-bit @code{int}. This is the default.
19631 Use 32-bit @code{int}.
19634 @itemx -mno-float32
19636 @opindex mno-float32
19637 Use 64-bit @code{float}. This is the default.
19640 @itemx -mno-float64
19642 @opindex mno-float64
19643 Use 32-bit @code{float}.
19647 Use @code{abshi2} pattern. This is the default.
19651 Do not use @code{abshi2} pattern.
19653 @item -mbranch-expensive
19654 @opindex mbranch-expensive
19655 Pretend that branches are expensive. This is for experimenting with
19656 code generation only.
19658 @item -mbranch-cheap
19659 @opindex mbranch-cheap
19660 Do not pretend that branches are expensive. This is the default.
19664 Use Unix assembler syntax. This is the default when configured for
19665 @samp{pdp11-*-bsd}.
19669 Use DEC assembler syntax. This is the default when configured for any
19670 PDP-11 target other than @samp{pdp11-*-bsd}.
19673 @node picoChip Options
19674 @subsection picoChip Options
19675 @cindex picoChip options
19677 These @samp{-m} options are defined for picoChip implementations:
19681 @item -mae=@var{ae_type}
19683 Set the instruction set, register set, and instruction scheduling
19684 parameters for array element type @var{ae_type}. Supported values
19685 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19687 @option{-mae=ANY} selects a completely generic AE type. Code
19688 generated with this option runs on any of the other AE types. The
19689 code is not as efficient as it would be if compiled for a specific
19690 AE type, and some types of operation (e.g., multiplication) do not
19691 work properly on all types of AE.
19693 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19694 for compiled code, and is the default.
19696 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19697 option may suffer from poor performance of byte (char) manipulation,
19698 since the DSP AE does not provide hardware support for byte load/stores.
19700 @item -msymbol-as-address
19701 Enable the compiler to directly use a symbol name as an address in a
19702 load/store instruction, without first loading it into a
19703 register. Typically, the use of this option generates larger
19704 programs, which run faster than when the option isn't used. However, the
19705 results vary from program to program, so it is left as a user option,
19706 rather than being permanently enabled.
19708 @item -mno-inefficient-warnings
19709 Disables warnings about the generation of inefficient code. These
19710 warnings can be generated, for example, when compiling code that
19711 performs byte-level memory operations on the MAC AE type. The MAC AE has
19712 no hardware support for byte-level memory operations, so all byte
19713 load/stores must be synthesized from word load/store operations. This is
19714 inefficient and a warning is generated to indicate
19715 that you should rewrite the code to avoid byte operations, or to target
19716 an AE type that has the necessary hardware support. This option disables
19721 @node PowerPC Options
19722 @subsection PowerPC Options
19723 @cindex PowerPC options
19725 These are listed under @xref{RS/6000 and PowerPC Options}.
19728 @subsection RL78 Options
19729 @cindex RL78 Options
19735 Links in additional target libraries to support operation within a
19744 Specifies the type of hardware multiplication and division support to
19745 be used. The simplest is @code{none}, which uses software for both
19746 multiplication and division. This is the default. The @code{g13}
19747 value is for the hardware multiply/divide peripheral found on the
19748 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
19749 the multiplication and division instructions supported by the RL78/G14
19750 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
19751 the value @code{mg10} is an alias for @code{none}.
19753 In addition a C preprocessor macro is defined, based upon the setting
19754 of this option. Possible values are: @code{__RL78_MUL_NONE__},
19755 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
19762 Specifies the RL78 core to target. The default is the G14 core, also
19763 known as an S3 core or just RL78. The G13 or S2 core does not have
19764 multiply or divide instructions, instead it uses a hardware peripheral
19765 for these operations. The G10 or S1 core does not have register
19766 banks, so it uses a different calling convention.
19768 If this option is set it also selects the type of hardware multiply
19769 support to use, unless this is overridden by an explicit
19770 @option{-mmul=none} option on the command line. Thus specifying
19771 @option{-mcpu=g13} enables the use of the G13 hardware multiply
19772 peripheral and specifying @option{-mcpu=g10} disables the use of
19773 hardware multiplications altogether.
19775 Note, although the RL78/G14 core is the default target, specifying
19776 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
19777 change the behavior of the toolchain since it also enables G14
19778 hardware multiply support. If these options are not specified on the
19779 command line then software multiplication routines will be used even
19780 though the code targets the RL78 core. This is for backwards
19781 compatibility with older toolchains which did not have hardware
19782 multiply and divide support.
19784 In addition a C preprocessor macro is defined, based upon the setting
19785 of this option. Possible values are: @code{__RL78_G10__},
19786 @code{__RL78_G13__} or @code{__RL78_G14__}.
19796 These are aliases for the corresponding @option{-mcpu=} option. They
19797 are provided for backwards compatibility.
19801 Allow the compiler to use all of the available registers. By default
19802 registers @code{r24..r31} are reserved for use in interrupt handlers.
19803 With this option enabled these registers can be used in ordinary
19806 @item -m64bit-doubles
19807 @itemx -m32bit-doubles
19808 @opindex m64bit-doubles
19809 @opindex m32bit-doubles
19810 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19811 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19812 @option{-m32bit-doubles}.
19816 @node RS/6000 and PowerPC Options
19817 @subsection IBM RS/6000 and PowerPC Options
19818 @cindex RS/6000 and PowerPC Options
19819 @cindex IBM RS/6000 and PowerPC Options
19821 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19823 @item -mpowerpc-gpopt
19824 @itemx -mno-powerpc-gpopt
19825 @itemx -mpowerpc-gfxopt
19826 @itemx -mno-powerpc-gfxopt
19829 @itemx -mno-powerpc64
19833 @itemx -mno-popcntb
19835 @itemx -mno-popcntd
19844 @itemx -mno-hard-dfp
19845 @opindex mpowerpc-gpopt
19846 @opindex mno-powerpc-gpopt
19847 @opindex mpowerpc-gfxopt
19848 @opindex mno-powerpc-gfxopt
19849 @opindex mpowerpc64
19850 @opindex mno-powerpc64
19854 @opindex mno-popcntb
19856 @opindex mno-popcntd
19862 @opindex mno-mfpgpr
19864 @opindex mno-hard-dfp
19865 You use these options to specify which instructions are available on the
19866 processor you are using. The default value of these options is
19867 determined when configuring GCC@. Specifying the
19868 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19869 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19870 rather than the options listed above.
19872 Specifying @option{-mpowerpc-gpopt} allows
19873 GCC to use the optional PowerPC architecture instructions in the
19874 General Purpose group, including floating-point square root. Specifying
19875 @option{-mpowerpc-gfxopt} allows GCC to
19876 use the optional PowerPC architecture instructions in the Graphics
19877 group, including floating-point select.
19879 The @option{-mmfcrf} option allows GCC to generate the move from
19880 condition register field instruction implemented on the POWER4
19881 processor and other processors that support the PowerPC V2.01
19883 The @option{-mpopcntb} option allows GCC to generate the popcount and
19884 double-precision FP reciprocal estimate instruction implemented on the
19885 POWER5 processor and other processors that support the PowerPC V2.02
19887 The @option{-mpopcntd} option allows GCC to generate the popcount
19888 instruction implemented on the POWER7 processor and other processors
19889 that support the PowerPC V2.06 architecture.
19890 The @option{-mfprnd} option allows GCC to generate the FP round to
19891 integer instructions implemented on the POWER5+ processor and other
19892 processors that support the PowerPC V2.03 architecture.
19893 The @option{-mcmpb} option allows GCC to generate the compare bytes
19894 instruction implemented on the POWER6 processor and other processors
19895 that support the PowerPC V2.05 architecture.
19896 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19897 general-purpose register instructions implemented on the POWER6X
19898 processor and other processors that support the extended PowerPC V2.05
19900 The @option{-mhard-dfp} option allows GCC to generate the decimal
19901 floating-point instructions implemented on some POWER processors.
19903 The @option{-mpowerpc64} option allows GCC to generate the additional
19904 64-bit instructions that are found in the full PowerPC64 architecture
19905 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19906 @option{-mno-powerpc64}.
19908 @item -mcpu=@var{cpu_type}
19910 Set architecture type, register usage, and
19911 instruction scheduling parameters for machine type @var{cpu_type}.
19912 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19913 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19914 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19915 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19916 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19917 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19918 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19919 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19920 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19921 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
19922 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
19925 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
19926 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
19927 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
19928 architecture machine types, with an appropriate, generic processor
19929 model assumed for scheduling purposes.
19931 The other options specify a specific processor. Code generated under
19932 those options runs best on that processor, and may not run at all on
19935 The @option{-mcpu} options automatically enable or disable the
19938 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19939 -mpopcntb -mpopcntd -mpowerpc64 @gol
19940 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19941 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19942 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
19943 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
19945 The particular options set for any particular CPU varies between
19946 compiler versions, depending on what setting seems to produce optimal
19947 code for that CPU; it doesn't necessarily reflect the actual hardware's
19948 capabilities. If you wish to set an individual option to a particular
19949 value, you may specify it after the @option{-mcpu} option, like
19950 @option{-mcpu=970 -mno-altivec}.
19952 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19953 not enabled or disabled by the @option{-mcpu} option at present because
19954 AIX does not have full support for these options. You may still
19955 enable or disable them individually if you're sure it'll work in your
19958 @item -mtune=@var{cpu_type}
19960 Set the instruction scheduling parameters for machine type
19961 @var{cpu_type}, but do not set the architecture type or register usage,
19962 as @option{-mcpu=@var{cpu_type}} does. The same
19963 values for @var{cpu_type} are used for @option{-mtune} as for
19964 @option{-mcpu}. If both are specified, the code generated uses the
19965 architecture and registers set by @option{-mcpu}, but the
19966 scheduling parameters set by @option{-mtune}.
19968 @item -mcmodel=small
19969 @opindex mcmodel=small
19970 Generate PowerPC64 code for the small model: The TOC is limited to
19973 @item -mcmodel=medium
19974 @opindex mcmodel=medium
19975 Generate PowerPC64 code for the medium model: The TOC and other static
19976 data may be up to a total of 4G in size.
19978 @item -mcmodel=large
19979 @opindex mcmodel=large
19980 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19981 in size. Other data and code is only limited by the 64-bit address
19985 @itemx -mno-altivec
19987 @opindex mno-altivec
19988 Generate code that uses (does not use) AltiVec instructions, and also
19989 enable the use of built-in functions that allow more direct access to
19990 the AltiVec instruction set. You may also need to set
19991 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19994 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19995 @option{-maltivec=be}, the element order for AltiVec intrinsics such
19996 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19997 match array element order corresponding to the endianness of the
19998 target. That is, element zero identifies the leftmost element in a
19999 vector register when targeting a big-endian platform, and identifies
20000 the rightmost element in a vector register when targeting a
20001 little-endian platform.
20004 @opindex maltivec=be
20005 Generate AltiVec instructions using big-endian element order,
20006 regardless of whether the target is big- or little-endian. This is
20007 the default when targeting a big-endian platform.
20009 The element order is used to interpret element numbers in AltiVec
20010 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
20011 @code{vec_insert}. By default, these match array element order
20012 corresponding to the endianness for the target.
20015 @opindex maltivec=le
20016 Generate AltiVec instructions using little-endian element order,
20017 regardless of whether the target is big- or little-endian. This is
20018 the default when targeting a little-endian platform. This option is
20019 currently ignored when targeting a big-endian platform.
20021 The element order is used to interpret element numbers in AltiVec
20022 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
20023 @code{vec_insert}. By default, these match array element order
20024 corresponding to the endianness for the target.
20029 @opindex mno-vrsave
20030 Generate VRSAVE instructions when generating AltiVec code.
20032 @item -mgen-cell-microcode
20033 @opindex mgen-cell-microcode
20034 Generate Cell microcode instructions.
20036 @item -mwarn-cell-microcode
20037 @opindex mwarn-cell-microcode
20038 Warn when a Cell microcode instruction is emitted. An example
20039 of a Cell microcode instruction is a variable shift.
20042 @opindex msecure-plt
20043 Generate code that allows @command{ld} and @command{ld.so}
20044 to build executables and shared
20045 libraries with non-executable @code{.plt} and @code{.got} sections.
20047 32-bit SYSV ABI option.
20051 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
20053 requires @code{.plt} and @code{.got}
20054 sections that are both writable and executable.
20055 This is a PowerPC 32-bit SYSV ABI option.
20061 This switch enables or disables the generation of ISEL instructions.
20063 @item -misel=@var{yes/no}
20064 This switch has been deprecated. Use @option{-misel} and
20065 @option{-mno-isel} instead.
20069 Enable Local Register Allocation. This is still experimental for PowerPC,
20070 so by default the compiler uses standard reload
20071 (i.e. @option{-mno-lra}).
20077 This switch enables or disables the generation of SPE simd
20083 @opindex mno-paired
20084 This switch enables or disables the generation of PAIRED simd
20087 @item -mspe=@var{yes/no}
20088 This option has been deprecated. Use @option{-mspe} and
20089 @option{-mno-spe} instead.
20095 Generate code that uses (does not use) vector/scalar (VSX)
20096 instructions, and also enable the use of built-in functions that allow
20097 more direct access to the VSX instruction set.
20102 @opindex mno-crypto
20103 Enable the use (disable) of the built-in functions that allow direct
20104 access to the cryptographic instructions that were added in version
20105 2.07 of the PowerPC ISA.
20107 @item -mdirect-move
20108 @itemx -mno-direct-move
20109 @opindex mdirect-move
20110 @opindex mno-direct-move
20111 Generate code that uses (does not use) the instructions to move data
20112 between the general purpose registers and the vector/scalar (VSX)
20113 registers that were added in version 2.07 of the PowerPC ISA.
20119 Enable (disable) the use of the built-in functions that allow direct
20120 access to the Hardware Transactional Memory (HTM) instructions that
20121 were added in version 2.07 of the PowerPC ISA.
20123 @item -mpower8-fusion
20124 @itemx -mno-power8-fusion
20125 @opindex mpower8-fusion
20126 @opindex mno-power8-fusion
20127 Generate code that keeps (does not keeps) some integer operations
20128 adjacent so that the instructions can be fused together on power8 and
20131 @item -mpower8-vector
20132 @itemx -mno-power8-vector
20133 @opindex mpower8-vector
20134 @opindex mno-power8-vector
20135 Generate code that uses (does not use) the vector and scalar
20136 instructions that were added in version 2.07 of the PowerPC ISA. Also
20137 enable the use of built-in functions that allow more direct access to
20138 the vector instructions.
20140 @item -mquad-memory
20141 @itemx -mno-quad-memory
20142 @opindex mquad-memory
20143 @opindex mno-quad-memory
20144 Generate code that uses (does not use) the non-atomic quad word memory
20145 instructions. The @option{-mquad-memory} option requires use of
20148 @item -mquad-memory-atomic
20149 @itemx -mno-quad-memory-atomic
20150 @opindex mquad-memory-atomic
20151 @opindex mno-quad-memory-atomic
20152 Generate code that uses (does not use) the atomic quad word memory
20153 instructions. The @option{-mquad-memory-atomic} option requires use of
20156 @item -mupper-regs-df
20157 @itemx -mno-upper-regs-df
20158 @opindex mupper-regs-df
20159 @opindex mno-upper-regs-df
20160 Generate code that uses (does not use) the scalar double precision
20161 instructions that target all 64 registers in the vector/scalar
20162 floating point register set that were added in version 2.06 of the
20163 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
20164 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
20165 @option{-mvsx} options.
20167 @item -mupper-regs-sf
20168 @itemx -mno-upper-regs-sf
20169 @opindex mupper-regs-sf
20170 @opindex mno-upper-regs-sf
20171 Generate code that uses (does not use) the scalar single precision
20172 instructions that target all 64 registers in the vector/scalar
20173 floating point register set that were added in version 2.07 of the
20174 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
20175 use either of the @option{-mcpu=power8}, @option{-mpower8-vector}, or
20176 @option{-mcpu=power9} options.
20179 @itemx -mno-upper-regs
20180 @opindex mupper-regs
20181 @opindex mno-upper-regs
20182 Generate code that uses (does not use) the scalar
20183 instructions that target all 64 registers in the vector/scalar
20184 floating point register set, depending on the model of the machine.
20186 If the @option{-mno-upper-regs} option is used, it turns off both
20187 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
20190 @itemx -mno-float128
20192 @opindex mno-float128
20193 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
20194 and use either software emulation for IEEE 128-bit floating point or
20195 hardware instructions.
20197 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
20198 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
20199 option. The @option{-mfloat128} option only works on PowerPC 64-bit
20202 If you use the ISA 3.0 instruction set (@option{-mcpu=power9}), the
20203 @option{-mfloat128} option will also enable the generation of ISA 3.0
20204 IEEE 128-bit floating point instructions. Otherwise, IEEE 128-bit
20205 floating point will be done with software emulation.
20207 @item -mfloat128-hardware
20208 @itemx -mno-float128-hardware
20209 @opindex mfloat128-hardware
20210 @opindex mno-float128-hardware
20211 Enable/disable using ISA 3.0 hardware instructions to support the
20212 @var{__float128} data type.
20214 If you use @option{-mfloat128-hardware}, it will enable the option
20215 @option{-mfloat128} as well.
20217 If you select ISA 3.0 instructions with @option{-mcpu=power9}, but do
20218 not use either @option{-mfloat128} or @option{-mfloat128-hardware},
20219 the IEEE 128-bit floating point support will not be enabled.
20221 @item -mfloat-gprs=@var{yes/single/double/no}
20222 @itemx -mfloat-gprs
20223 @opindex mfloat-gprs
20224 This switch enables or disables the generation of floating-point
20225 operations on the general-purpose registers for architectures that
20228 The argument @samp{yes} or @samp{single} enables the use of
20229 single-precision floating-point operations.
20231 The argument @samp{double} enables the use of single and
20232 double-precision floating-point operations.
20234 The argument @samp{no} disables floating-point operations on the
20235 general-purpose registers.
20237 This option is currently only available on the MPC854x.
20243 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
20244 targets (including GNU/Linux). The 32-bit environment sets int, long
20245 and pointer to 32 bits and generates code that runs on any PowerPC
20246 variant. The 64-bit environment sets int to 32 bits and long and
20247 pointer to 64 bits, and generates code for PowerPC64, as for
20248 @option{-mpowerpc64}.
20251 @itemx -mno-fp-in-toc
20252 @itemx -mno-sum-in-toc
20253 @itemx -mminimal-toc
20255 @opindex mno-fp-in-toc
20256 @opindex mno-sum-in-toc
20257 @opindex mminimal-toc
20258 Modify generation of the TOC (Table Of Contents), which is created for
20259 every executable file. The @option{-mfull-toc} option is selected by
20260 default. In that case, GCC allocates at least one TOC entry for
20261 each unique non-automatic variable reference in your program. GCC
20262 also places floating-point constants in the TOC@. However, only
20263 16,384 entries are available in the TOC@.
20265 If you receive a linker error message that saying you have overflowed
20266 the available TOC space, you can reduce the amount of TOC space used
20267 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
20268 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
20269 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
20270 generate code to calculate the sum of an address and a constant at
20271 run time instead of putting that sum into the TOC@. You may specify one
20272 or both of these options. Each causes GCC to produce very slightly
20273 slower and larger code at the expense of conserving TOC space.
20275 If you still run out of space in the TOC even when you specify both of
20276 these options, specify @option{-mminimal-toc} instead. This option causes
20277 GCC to make only one TOC entry for every file. When you specify this
20278 option, GCC produces code that is slower and larger but which
20279 uses extremely little TOC space. You may wish to use this option
20280 only on files that contain less frequently-executed code.
20286 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
20287 @code{long} type, and the infrastructure needed to support them.
20288 Specifying @option{-maix64} implies @option{-mpowerpc64},
20289 while @option{-maix32} disables the 64-bit ABI and
20290 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
20293 @itemx -mno-xl-compat
20294 @opindex mxl-compat
20295 @opindex mno-xl-compat
20296 Produce code that conforms more closely to IBM XL compiler semantics
20297 when using AIX-compatible ABI@. Pass floating-point arguments to
20298 prototyped functions beyond the register save area (RSA) on the stack
20299 in addition to argument FPRs. Do not assume that most significant
20300 double in 128-bit long double value is properly rounded when comparing
20301 values and converting to double. Use XL symbol names for long double
20304 The AIX calling convention was extended but not initially documented to
20305 handle an obscure K&R C case of calling a function that takes the
20306 address of its arguments with fewer arguments than declared. IBM XL
20307 compilers access floating-point arguments that do not fit in the
20308 RSA from the stack when a subroutine is compiled without
20309 optimization. Because always storing floating-point arguments on the
20310 stack is inefficient and rarely needed, this option is not enabled by
20311 default and only is necessary when calling subroutines compiled by IBM
20312 XL compilers without optimization.
20316 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
20317 application written to use message passing with special startup code to
20318 enable the application to run. The system must have PE installed in the
20319 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
20320 must be overridden with the @option{-specs=} option to specify the
20321 appropriate directory location. The Parallel Environment does not
20322 support threads, so the @option{-mpe} option and the @option{-pthread}
20323 option are incompatible.
20325 @item -malign-natural
20326 @itemx -malign-power
20327 @opindex malign-natural
20328 @opindex malign-power
20329 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
20330 @option{-malign-natural} overrides the ABI-defined alignment of larger
20331 types, such as floating-point doubles, on their natural size-based boundary.
20332 The option @option{-malign-power} instructs GCC to follow the ABI-specified
20333 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
20335 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
20339 @itemx -mhard-float
20340 @opindex msoft-float
20341 @opindex mhard-float
20342 Generate code that does not use (uses) the floating-point register set.
20343 Software floating-point emulation is provided if you use the
20344 @option{-msoft-float} option, and pass the option to GCC when linking.
20346 @item -msingle-float
20347 @itemx -mdouble-float
20348 @opindex msingle-float
20349 @opindex mdouble-float
20350 Generate code for single- or double-precision floating-point operations.
20351 @option{-mdouble-float} implies @option{-msingle-float}.
20354 @opindex msimple-fpu
20355 Do not generate @code{sqrt} and @code{div} instructions for hardware
20356 floating-point unit.
20358 @item -mfpu=@var{name}
20360 Specify type of floating-point unit. Valid values for @var{name} are
20361 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
20362 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
20363 @samp{sp_full} (equivalent to @option{-msingle-float}),
20364 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
20367 @opindex mxilinx-fpu
20368 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
20371 @itemx -mno-multiple
20373 @opindex mno-multiple
20374 Generate code that uses (does not use) the load multiple word
20375 instructions and the store multiple word instructions. These
20376 instructions are generated by default on POWER systems, and not
20377 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
20378 PowerPC systems, since those instructions do not work when the
20379 processor is in little-endian mode. The exceptions are PPC740 and
20380 PPC750 which permit these instructions in little-endian mode.
20385 @opindex mno-string
20386 Generate code that uses (does not use) the load string instructions
20387 and the store string word instructions to save multiple registers and
20388 do small block moves. These instructions are generated by default on
20389 POWER systems, and not generated on PowerPC systems. Do not use
20390 @option{-mstring} on little-endian PowerPC systems, since those
20391 instructions do not work when the processor is in little-endian mode.
20392 The exceptions are PPC740 and PPC750 which permit these instructions
20393 in little-endian mode.
20398 @opindex mno-update
20399 Generate code that uses (does not use) the load or store instructions
20400 that update the base register to the address of the calculated memory
20401 location. These instructions are generated by default. If you use
20402 @option{-mno-update}, there is a small window between the time that the
20403 stack pointer is updated and the address of the previous frame is
20404 stored, which means code that walks the stack frame across interrupts or
20405 signals may get corrupted data.
20407 @item -mavoid-indexed-addresses
20408 @itemx -mno-avoid-indexed-addresses
20409 @opindex mavoid-indexed-addresses
20410 @opindex mno-avoid-indexed-addresses
20411 Generate code that tries to avoid (not avoid) the use of indexed load
20412 or store instructions. These instructions can incur a performance
20413 penalty on Power6 processors in certain situations, such as when
20414 stepping through large arrays that cross a 16M boundary. This option
20415 is enabled by default when targeting Power6 and disabled otherwise.
20418 @itemx -mno-fused-madd
20419 @opindex mfused-madd
20420 @opindex mno-fused-madd
20421 Generate code that uses (does not use) the floating-point multiply and
20422 accumulate instructions. These instructions are generated by default
20423 if hardware floating point is used. The machine-dependent
20424 @option{-mfused-madd} option is now mapped to the machine-independent
20425 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20426 mapped to @option{-ffp-contract=off}.
20432 Generate code that uses (does not use) the half-word multiply and
20433 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
20434 These instructions are generated by default when targeting those
20441 Generate code that uses (does not use) the string-search @samp{dlmzb}
20442 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
20443 generated by default when targeting those processors.
20445 @item -mno-bit-align
20447 @opindex mno-bit-align
20448 @opindex mbit-align
20449 On System V.4 and embedded PowerPC systems do not (do) force structures
20450 and unions that contain bit-fields to be aligned to the base type of the
20453 For example, by default a structure containing nothing but 8
20454 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
20455 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
20456 the structure is aligned to a 1-byte boundary and is 1 byte in
20459 @item -mno-strict-align
20460 @itemx -mstrict-align
20461 @opindex mno-strict-align
20462 @opindex mstrict-align
20463 On System V.4 and embedded PowerPC systems do not (do) assume that
20464 unaligned memory references are handled by the system.
20466 @item -mrelocatable
20467 @itemx -mno-relocatable
20468 @opindex mrelocatable
20469 @opindex mno-relocatable
20470 Generate code that allows (does not allow) a static executable to be
20471 relocated to a different address at run time. A simple embedded
20472 PowerPC system loader should relocate the entire contents of
20473 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
20474 a table of 32-bit addresses generated by this option. For this to
20475 work, all objects linked together must be compiled with
20476 @option{-mrelocatable} or @option{-mrelocatable-lib}.
20477 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
20479 @item -mrelocatable-lib
20480 @itemx -mno-relocatable-lib
20481 @opindex mrelocatable-lib
20482 @opindex mno-relocatable-lib
20483 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
20484 @code{.fixup} section to allow static executables to be relocated at
20485 run time, but @option{-mrelocatable-lib} does not use the smaller stack
20486 alignment of @option{-mrelocatable}. Objects compiled with
20487 @option{-mrelocatable-lib} may be linked with objects compiled with
20488 any combination of the @option{-mrelocatable} options.
20494 On System V.4 and embedded PowerPC systems do not (do) assume that
20495 register 2 contains a pointer to a global area pointing to the addresses
20496 used in the program.
20499 @itemx -mlittle-endian
20501 @opindex mlittle-endian
20502 On System V.4 and embedded PowerPC systems compile code for the
20503 processor in little-endian mode. The @option{-mlittle-endian} option is
20504 the same as @option{-mlittle}.
20507 @itemx -mbig-endian
20509 @opindex mbig-endian
20510 On System V.4 and embedded PowerPC systems compile code for the
20511 processor in big-endian mode. The @option{-mbig-endian} option is
20512 the same as @option{-mbig}.
20514 @item -mdynamic-no-pic
20515 @opindex mdynamic-no-pic
20516 On Darwin and Mac OS X systems, compile code so that it is not
20517 relocatable, but that its external references are relocatable. The
20518 resulting code is suitable for applications, but not shared
20521 @item -msingle-pic-base
20522 @opindex msingle-pic-base
20523 Treat the register used for PIC addressing as read-only, rather than
20524 loading it in the prologue for each function. The runtime system is
20525 responsible for initializing this register with an appropriate value
20526 before execution begins.
20528 @item -mprioritize-restricted-insns=@var{priority}
20529 @opindex mprioritize-restricted-insns
20530 This option controls the priority that is assigned to
20531 dispatch-slot restricted instructions during the second scheduling
20532 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20533 or @samp{2} to assign no, highest, or second-highest (respectively)
20534 priority to dispatch-slot restricted
20537 @item -msched-costly-dep=@var{dependence_type}
20538 @opindex msched-costly-dep
20539 This option controls which dependences are considered costly
20540 by the target during instruction scheduling. The argument
20541 @var{dependence_type} takes one of the following values:
20545 No dependence is costly.
20548 All dependences are costly.
20550 @item @samp{true_store_to_load}
20551 A true dependence from store to load is costly.
20553 @item @samp{store_to_load}
20554 Any dependence from store to load is costly.
20557 Any dependence for which the latency is greater than or equal to
20558 @var{number} is costly.
20561 @item -minsert-sched-nops=@var{scheme}
20562 @opindex minsert-sched-nops
20563 This option controls which NOP insertion scheme is used during
20564 the second scheduling pass. The argument @var{scheme} takes one of the
20572 Pad with NOPs any dispatch group that has vacant issue slots,
20573 according to the scheduler's grouping.
20575 @item @samp{regroup_exact}
20576 Insert NOPs to force costly dependent insns into
20577 separate groups. Insert exactly as many NOPs as needed to force an insn
20578 to a new group, according to the estimated processor grouping.
20581 Insert NOPs to force costly dependent insns into
20582 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20586 @opindex mcall-sysv
20587 On System V.4 and embedded PowerPC systems compile code using calling
20588 conventions that adhere to the March 1995 draft of the System V
20589 Application Binary Interface, PowerPC processor supplement. This is the
20590 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20592 @item -mcall-sysv-eabi
20594 @opindex mcall-sysv-eabi
20595 @opindex mcall-eabi
20596 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20598 @item -mcall-sysv-noeabi
20599 @opindex mcall-sysv-noeabi
20600 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20602 @item -mcall-aixdesc
20604 On System V.4 and embedded PowerPC systems compile code for the AIX
20608 @opindex mcall-linux
20609 On System V.4 and embedded PowerPC systems compile code for the
20610 Linux-based GNU system.
20612 @item -mcall-freebsd
20613 @opindex mcall-freebsd
20614 On System V.4 and embedded PowerPC systems compile code for the
20615 FreeBSD operating system.
20617 @item -mcall-netbsd
20618 @opindex mcall-netbsd
20619 On System V.4 and embedded PowerPC systems compile code for the
20620 NetBSD operating system.
20622 @item -mcall-openbsd
20623 @opindex mcall-netbsd
20624 On System V.4 and embedded PowerPC systems compile code for the
20625 OpenBSD operating system.
20627 @item -maix-struct-return
20628 @opindex maix-struct-return
20629 Return all structures in memory (as specified by the AIX ABI)@.
20631 @item -msvr4-struct-return
20632 @opindex msvr4-struct-return
20633 Return structures smaller than 8 bytes in registers (as specified by the
20636 @item -mabi=@var{abi-type}
20638 Extend the current ABI with a particular extension, or remove such extension.
20639 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
20640 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
20641 @samp{elfv1}, @samp{elfv2}@.
20645 Extend the current ABI with SPE ABI extensions. This does not change
20646 the default ABI, instead it adds the SPE ABI extensions to the current
20650 @opindex mabi=no-spe
20651 Disable Book-E SPE ABI extensions for the current ABI@.
20653 @item -mabi=ibmlongdouble
20654 @opindex mabi=ibmlongdouble
20655 Change the current ABI to use IBM extended-precision long double.
20656 This is a PowerPC 32-bit SYSV ABI option.
20658 @item -mabi=ieeelongdouble
20659 @opindex mabi=ieeelongdouble
20660 Change the current ABI to use IEEE extended-precision long double.
20661 This is a PowerPC 32-bit Linux ABI option.
20664 @opindex mabi=elfv1
20665 Change the current ABI to use the ELFv1 ABI.
20666 This is the default ABI for big-endian PowerPC 64-bit Linux.
20667 Overriding the default ABI requires special system support and is
20668 likely to fail in spectacular ways.
20671 @opindex mabi=elfv2
20672 Change the current ABI to use the ELFv2 ABI.
20673 This is the default ABI for little-endian PowerPC 64-bit Linux.
20674 Overriding the default ABI requires special system support and is
20675 likely to fail in spectacular ways.
20678 @itemx -mno-prototype
20679 @opindex mprototype
20680 @opindex mno-prototype
20681 On System V.4 and embedded PowerPC systems assume that all calls to
20682 variable argument functions are properly prototyped. Otherwise, the
20683 compiler must insert an instruction before every non-prototyped call to
20684 set or clear bit 6 of the condition code register (@code{CR}) to
20685 indicate whether floating-point values are passed in the floating-point
20686 registers in case the function takes variable arguments. With
20687 @option{-mprototype}, only calls to prototyped variable argument functions
20688 set or clear the bit.
20692 On embedded PowerPC systems, assume that the startup module is called
20693 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20694 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20699 On embedded PowerPC systems, assume that the startup module is called
20700 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20705 On embedded PowerPC systems, assume that the startup module is called
20706 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20709 @item -myellowknife
20710 @opindex myellowknife
20711 On embedded PowerPC systems, assume that the startup module is called
20712 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20717 On System V.4 and embedded PowerPC systems, specify that you are
20718 compiling for a VxWorks system.
20722 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
20723 header to indicate that @samp{eabi} extended relocations are used.
20729 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20730 Embedded Applications Binary Interface (EABI), which is a set of
20731 modifications to the System V.4 specifications. Selecting @option{-meabi}
20732 means that the stack is aligned to an 8-byte boundary, a function
20733 @code{__eabi} is called from @code{main} to set up the EABI
20734 environment, and the @option{-msdata} option can use both @code{r2} and
20735 @code{r13} to point to two separate small data areas. Selecting
20736 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20737 no EABI initialization function is called from @code{main}, and the
20738 @option{-msdata} option only uses @code{r13} to point to a single
20739 small data area. The @option{-meabi} option is on by default if you
20740 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20743 @opindex msdata=eabi
20744 On System V.4 and embedded PowerPC systems, put small initialized
20745 @code{const} global and static data in the @code{.sdata2} section, which
20746 is pointed to by register @code{r2}. Put small initialized
20747 non-@code{const} global and static data in the @code{.sdata} section,
20748 which is pointed to by register @code{r13}. Put small uninitialized
20749 global and static data in the @code{.sbss} section, which is adjacent to
20750 the @code{.sdata} section. The @option{-msdata=eabi} option is
20751 incompatible with the @option{-mrelocatable} option. The
20752 @option{-msdata=eabi} option also sets the @option{-memb} option.
20755 @opindex msdata=sysv
20756 On System V.4 and embedded PowerPC systems, put small global and static
20757 data in the @code{.sdata} section, which is pointed to by register
20758 @code{r13}. Put small uninitialized global and static data in the
20759 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
20760 The @option{-msdata=sysv} option is incompatible with the
20761 @option{-mrelocatable} option.
20763 @item -msdata=default
20765 @opindex msdata=default
20767 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20768 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20769 same as @option{-msdata=sysv}.
20772 @opindex msdata=data
20773 On System V.4 and embedded PowerPC systems, put small global
20774 data in the @code{.sdata} section. Put small uninitialized global
20775 data in the @code{.sbss} section. Do not use register @code{r13}
20776 to address small data however. This is the default behavior unless
20777 other @option{-msdata} options are used.
20781 @opindex msdata=none
20783 On embedded PowerPC systems, put all initialized global and static data
20784 in the @code{.data} section, and all uninitialized data in the
20785 @code{.bss} section.
20787 @item -mblock-move-inline-limit=@var{num}
20788 @opindex mblock-move-inline-limit
20789 Inline all block moves (such as calls to @code{memcpy} or structure
20790 copies) less than or equal to @var{num} bytes. The minimum value for
20791 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20792 targets. The default value is target-specific.
20796 @cindex smaller data references (PowerPC)
20797 @cindex .sdata/.sdata2 references (PowerPC)
20798 On embedded PowerPC systems, put global and static items less than or
20799 equal to @var{num} bytes into the small data or BSS sections instead of
20800 the normal data or BSS section. By default, @var{num} is 8. The
20801 @option{-G @var{num}} switch is also passed to the linker.
20802 All modules should be compiled with the same @option{-G @var{num}} value.
20805 @itemx -mno-regnames
20807 @opindex mno-regnames
20808 On System V.4 and embedded PowerPC systems do (do not) emit register
20809 names in the assembly language output using symbolic forms.
20812 @itemx -mno-longcall
20814 @opindex mno-longcall
20815 By default assume that all calls are far away so that a longer and more
20816 expensive calling sequence is required. This is required for calls
20817 farther than 32 megabytes (33,554,432 bytes) from the current location.
20818 A short call is generated if the compiler knows
20819 the call cannot be that far away. This setting can be overridden by
20820 the @code{shortcall} function attribute, or by @code{#pragma
20823 Some linkers are capable of detecting out-of-range calls and generating
20824 glue code on the fly. On these systems, long calls are unnecessary and
20825 generate slower code. As of this writing, the AIX linker can do this,
20826 as can the GNU linker for PowerPC/64. It is planned to add this feature
20827 to the GNU linker for 32-bit PowerPC systems as well.
20829 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20830 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20831 addresses represent the callee and the branch island. The
20832 Darwin/PPC linker prefers the first address and generates a @code{bl
20833 callee} if the PPC @code{bl} instruction reaches the callee directly;
20834 otherwise, the linker generates @code{bl L42} to call the branch
20835 island. The branch island is appended to the body of the
20836 calling function; it computes the full 32-bit address of the callee
20839 On Mach-O (Darwin) systems, this option directs the compiler emit to
20840 the glue for every direct call, and the Darwin linker decides whether
20841 to use or discard it.
20843 In the future, GCC may ignore all longcall specifications
20844 when the linker is known to generate glue.
20846 @item -mtls-markers
20847 @itemx -mno-tls-markers
20848 @opindex mtls-markers
20849 @opindex mno-tls-markers
20850 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20851 specifying the function argument. The relocation allows the linker to
20852 reliably associate function call with argument setup instructions for
20853 TLS optimization, which in turn allows GCC to better schedule the
20858 Adds support for multithreading with the @dfn{pthreads} library.
20859 This option sets flags for both the preprocessor and linker.
20864 This option enables use of the reciprocal estimate and
20865 reciprocal square root estimate instructions with additional
20866 Newton-Raphson steps to increase precision instead of doing a divide or
20867 square root and divide for floating-point arguments. You should use
20868 the @option{-ffast-math} option when using @option{-mrecip} (or at
20869 least @option{-funsafe-math-optimizations},
20870 @option{-ffinite-math-only}, @option{-freciprocal-math} and
20871 @option{-fno-trapping-math}). Note that while the throughput of the
20872 sequence is generally higher than the throughput of the non-reciprocal
20873 instruction, the precision of the sequence can be decreased by up to 2
20874 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20877 @item -mrecip=@var{opt}
20878 @opindex mrecip=opt
20879 This option controls which reciprocal estimate instructions
20880 may be used. @var{opt} is a comma-separated list of options, which may
20881 be preceded by a @code{!} to invert the option:
20886 Enable all estimate instructions.
20889 Enable the default instructions, equivalent to @option{-mrecip}.
20892 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20895 Enable the reciprocal approximation instructions for both
20896 single and double precision.
20899 Enable the single-precision reciprocal approximation instructions.
20902 Enable the double-precision reciprocal approximation instructions.
20905 Enable the reciprocal square root approximation instructions for both
20906 single and double precision.
20909 Enable the single-precision reciprocal square root approximation instructions.
20912 Enable the double-precision reciprocal square root approximation instructions.
20916 So, for example, @option{-mrecip=all,!rsqrtd} enables
20917 all of the reciprocal estimate instructions, except for the
20918 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20919 which handle the double-precision reciprocal square root calculations.
20921 @item -mrecip-precision
20922 @itemx -mno-recip-precision
20923 @opindex mrecip-precision
20924 Assume (do not assume) that the reciprocal estimate instructions
20925 provide higher-precision estimates than is mandated by the PowerPC
20926 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20927 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20928 The double-precision square root estimate instructions are not generated by
20929 default on low-precision machines, since they do not provide an
20930 estimate that converges after three steps.
20932 @item -mveclibabi=@var{type}
20933 @opindex mveclibabi
20934 Specifies the ABI type to use for vectorizing intrinsics using an
20935 external library. The only type supported at present is @samp{mass},
20936 which specifies to use IBM's Mathematical Acceleration Subsystem
20937 (MASS) libraries for vectorizing intrinsics using external libraries.
20938 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20939 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20940 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20941 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20942 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20943 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20944 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20945 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20946 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20947 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20948 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20949 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20950 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20951 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20952 for power7. Both @option{-ftree-vectorize} and
20953 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20954 libraries must be specified at link time.
20959 Generate (do not generate) the @code{friz} instruction when the
20960 @option{-funsafe-math-optimizations} option is used to optimize
20961 rounding of floating-point values to 64-bit integer and back to floating
20962 point. The @code{friz} instruction does not return the same value if
20963 the floating-point number is too large to fit in an integer.
20965 @item -mpointers-to-nested-functions
20966 @itemx -mno-pointers-to-nested-functions
20967 @opindex mpointers-to-nested-functions
20968 Generate (do not generate) code to load up the static chain register
20969 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20970 systems where a function pointer points to a 3-word descriptor giving
20971 the function address, TOC value to be loaded in register @code{r2}, and
20972 static chain value to be loaded in register @code{r11}. The
20973 @option{-mpointers-to-nested-functions} is on by default. You cannot
20974 call through pointers to nested functions or pointers
20975 to functions compiled in other languages that use the static chain if
20976 you use @option{-mno-pointers-to-nested-functions}.
20978 @item -msave-toc-indirect
20979 @itemx -mno-save-toc-indirect
20980 @opindex msave-toc-indirect
20981 Generate (do not generate) code to save the TOC value in the reserved
20982 stack location in the function prologue if the function calls through
20983 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20984 saved in the prologue, it is saved just before the call through the
20985 pointer. The @option{-mno-save-toc-indirect} option is the default.
20987 @item -mcompat-align-parm
20988 @itemx -mno-compat-align-parm
20989 @opindex mcompat-align-parm
20990 Generate (do not generate) code to pass structure parameters with a
20991 maximum alignment of 64 bits, for compatibility with older versions
20994 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20995 structure parameter on a 128-bit boundary when that structure contained
20996 a member requiring 128-bit alignment. This is corrected in more
20997 recent versions of GCC. This option may be used to generate code
20998 that is compatible with functions compiled with older versions of
21001 The @option{-mno-compat-align-parm} option is the default.
21005 @subsection RX Options
21008 These command-line options are defined for RX targets:
21011 @item -m64bit-doubles
21012 @itemx -m32bit-doubles
21013 @opindex m64bit-doubles
21014 @opindex m32bit-doubles
21015 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21016 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21017 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
21018 works on 32-bit values, which is why the default is
21019 @option{-m32bit-doubles}.
21025 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
21026 floating-point hardware. The default is enabled for the RX600
21027 series and disabled for the RX200 series.
21029 Floating-point instructions are only generated for 32-bit floating-point
21030 values, however, so the FPU hardware is not used for doubles if the
21031 @option{-m64bit-doubles} option is used.
21033 @emph{Note} If the @option{-fpu} option is enabled then
21034 @option{-funsafe-math-optimizations} is also enabled automatically.
21035 This is because the RX FPU instructions are themselves unsafe.
21037 @item -mcpu=@var{name}
21039 Selects the type of RX CPU to be targeted. Currently three types are
21040 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
21041 the specific @samp{RX610} CPU. The default is @samp{RX600}.
21043 The only difference between @samp{RX600} and @samp{RX610} is that the
21044 @samp{RX610} does not support the @code{MVTIPL} instruction.
21046 The @samp{RX200} series does not have a hardware floating-point unit
21047 and so @option{-nofpu} is enabled by default when this type is
21050 @item -mbig-endian-data
21051 @itemx -mlittle-endian-data
21052 @opindex mbig-endian-data
21053 @opindex mlittle-endian-data
21054 Store data (but not code) in the big-endian format. The default is
21055 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
21058 @item -msmall-data-limit=@var{N}
21059 @opindex msmall-data-limit
21060 Specifies the maximum size in bytes of global and static variables
21061 which can be placed into the small data area. Using the small data
21062 area can lead to smaller and faster code, but the size of area is
21063 limited and it is up to the programmer to ensure that the area does
21064 not overflow. Also when the small data area is used one of the RX's
21065 registers (usually @code{r13}) is reserved for use pointing to this
21066 area, so it is no longer available for use by the compiler. This
21067 could result in slower and/or larger code if variables are pushed onto
21068 the stack instead of being held in this register.
21070 Note, common variables (variables that have not been initialized) and
21071 constants are not placed into the small data area as they are assigned
21072 to other sections in the output executable.
21074 The default value is zero, which disables this feature. Note, this
21075 feature is not enabled by default with higher optimization levels
21076 (@option{-O2} etc) because of the potentially detrimental effects of
21077 reserving a register. It is up to the programmer to experiment and
21078 discover whether this feature is of benefit to their program. See the
21079 description of the @option{-mpid} option for a description of how the
21080 actual register to hold the small data area pointer is chosen.
21086 Use the simulator runtime. The default is to use the libgloss
21087 board-specific runtime.
21089 @item -mas100-syntax
21090 @itemx -mno-as100-syntax
21091 @opindex mas100-syntax
21092 @opindex mno-as100-syntax
21093 When generating assembler output use a syntax that is compatible with
21094 Renesas's AS100 assembler. This syntax can also be handled by the GAS
21095 assembler, but it has some restrictions so it is not generated by default.
21097 @item -mmax-constant-size=@var{N}
21098 @opindex mmax-constant-size
21099 Specifies the maximum size, in bytes, of a constant that can be used as
21100 an operand in a RX instruction. Although the RX instruction set does
21101 allow constants of up to 4 bytes in length to be used in instructions,
21102 a longer value equates to a longer instruction. Thus in some
21103 circumstances it can be beneficial to restrict the size of constants
21104 that are used in instructions. Constants that are too big are instead
21105 placed into a constant pool and referenced via register indirection.
21107 The value @var{N} can be between 0 and 4. A value of 0 (the default)
21108 or 4 means that constants of any size are allowed.
21112 Enable linker relaxation. Linker relaxation is a process whereby the
21113 linker attempts to reduce the size of a program by finding shorter
21114 versions of various instructions. Disabled by default.
21116 @item -mint-register=@var{N}
21117 @opindex mint-register
21118 Specify the number of registers to reserve for fast interrupt handler
21119 functions. The value @var{N} can be between 0 and 4. A value of 1
21120 means that register @code{r13} is reserved for the exclusive use
21121 of fast interrupt handlers. A value of 2 reserves @code{r13} and
21122 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
21123 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
21124 A value of 0, the default, does not reserve any registers.
21126 @item -msave-acc-in-interrupts
21127 @opindex msave-acc-in-interrupts
21128 Specifies that interrupt handler functions should preserve the
21129 accumulator register. This is only necessary if normal code might use
21130 the accumulator register, for example because it performs 64-bit
21131 multiplications. The default is to ignore the accumulator as this
21132 makes the interrupt handlers faster.
21138 Enables the generation of position independent data. When enabled any
21139 access to constant data is done via an offset from a base address
21140 held in a register. This allows the location of constant data to be
21141 determined at run time without requiring the executable to be
21142 relocated, which is a benefit to embedded applications with tight
21143 memory constraints. Data that can be modified is not affected by this
21146 Note, using this feature reserves a register, usually @code{r13}, for
21147 the constant data base address. This can result in slower and/or
21148 larger code, especially in complicated functions.
21150 The actual register chosen to hold the constant data base address
21151 depends upon whether the @option{-msmall-data-limit} and/or the
21152 @option{-mint-register} command-line options are enabled. Starting
21153 with register @code{r13} and proceeding downwards, registers are
21154 allocated first to satisfy the requirements of @option{-mint-register},
21155 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
21156 is possible for the small data area register to be @code{r8} if both
21157 @option{-mint-register=4} and @option{-mpid} are specified on the
21160 By default this feature is not enabled. The default can be restored
21161 via the @option{-mno-pid} command-line option.
21163 @item -mno-warn-multiple-fast-interrupts
21164 @itemx -mwarn-multiple-fast-interrupts
21165 @opindex mno-warn-multiple-fast-interrupts
21166 @opindex mwarn-multiple-fast-interrupts
21167 Prevents GCC from issuing a warning message if it finds more than one
21168 fast interrupt handler when it is compiling a file. The default is to
21169 issue a warning for each extra fast interrupt handler found, as the RX
21170 only supports one such interrupt.
21172 @item -mallow-string-insns
21173 @itemx -mno-allow-string-insns
21174 @opindex mallow-string-insns
21175 @opindex mno-allow-string-insns
21176 Enables or disables the use of the string manipulation instructions
21177 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
21178 @code{SWHILE} and also the @code{RMPA} instruction. These
21179 instructions may prefetch data, which is not safe to do if accessing
21180 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
21181 for more information).
21183 The default is to allow these instructions, but it is not possible for
21184 GCC to reliably detect all circumstances where a string instruction
21185 might be used to access an I/O register, so their use cannot be
21186 disabled automatically. Instead it is reliant upon the programmer to
21187 use the @option{-mno-allow-string-insns} option if their program
21188 accesses I/O space.
21190 When the instructions are enabled GCC defines the C preprocessor
21191 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
21192 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
21198 Use only (or not only) @code{JSR} instructions to access functions.
21199 This option can be used when code size exceeds the range of @code{BSR}
21200 instructions. Note that @option{-mno-jsr} does not mean to not use
21201 @code{JSR} but instead means that any type of branch may be used.
21204 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
21205 has special significance to the RX port when used with the
21206 @code{interrupt} function attribute. This attribute indicates a
21207 function intended to process fast interrupts. GCC ensures
21208 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
21209 and/or @code{r13} and only provided that the normal use of the
21210 corresponding registers have been restricted via the
21211 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
21214 @node S/390 and zSeries Options
21215 @subsection S/390 and zSeries Options
21216 @cindex S/390 and zSeries Options
21218 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
21222 @itemx -msoft-float
21223 @opindex mhard-float
21224 @opindex msoft-float
21225 Use (do not use) the hardware floating-point instructions and registers
21226 for floating-point operations. When @option{-msoft-float} is specified,
21227 functions in @file{libgcc.a} are used to perform floating-point
21228 operations. When @option{-mhard-float} is specified, the compiler
21229 generates IEEE floating-point instructions. This is the default.
21232 @itemx -mno-hard-dfp
21234 @opindex mno-hard-dfp
21235 Use (do not use) the hardware decimal-floating-point instructions for
21236 decimal-floating-point operations. When @option{-mno-hard-dfp} is
21237 specified, functions in @file{libgcc.a} are used to perform
21238 decimal-floating-point operations. When @option{-mhard-dfp} is
21239 specified, the compiler generates decimal-floating-point hardware
21240 instructions. This is the default for @option{-march=z9-ec} or higher.
21242 @item -mlong-double-64
21243 @itemx -mlong-double-128
21244 @opindex mlong-double-64
21245 @opindex mlong-double-128
21246 These switches control the size of @code{long double} type. A size
21247 of 64 bits makes the @code{long double} type equivalent to the @code{double}
21248 type. This is the default.
21251 @itemx -mno-backchain
21252 @opindex mbackchain
21253 @opindex mno-backchain
21254 Store (do not store) the address of the caller's frame as backchain pointer
21255 into the callee's stack frame.
21256 A backchain may be needed to allow debugging using tools that do not understand
21257 DWARF call frame information.
21258 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
21259 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
21260 the backchain is placed into the topmost word of the 96/160 byte register
21263 In general, code compiled with @option{-mbackchain} is call-compatible with
21264 code compiled with @option{-mmo-backchain}; however, use of the backchain
21265 for debugging purposes usually requires that the whole binary is built with
21266 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
21267 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21268 to build a linux kernel use @option{-msoft-float}.
21270 The default is to not maintain the backchain.
21272 @item -mpacked-stack
21273 @itemx -mno-packed-stack
21274 @opindex mpacked-stack
21275 @opindex mno-packed-stack
21276 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
21277 specified, the compiler uses the all fields of the 96/160 byte register save
21278 area only for their default purpose; unused fields still take up stack space.
21279 When @option{-mpacked-stack} is specified, register save slots are densely
21280 packed at the top of the register save area; unused space is reused for other
21281 purposes, allowing for more efficient use of the available stack space.
21282 However, when @option{-mbackchain} is also in effect, the topmost word of
21283 the save area is always used to store the backchain, and the return address
21284 register is always saved two words below the backchain.
21286 As long as the stack frame backchain is not used, code generated with
21287 @option{-mpacked-stack} is call-compatible with code generated with
21288 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
21289 S/390 or zSeries generated code that uses the stack frame backchain at run
21290 time, not just for debugging purposes. Such code is not call-compatible
21291 with code compiled with @option{-mpacked-stack}. Also, note that the
21292 combination of @option{-mbackchain},
21293 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21294 to build a linux kernel use @option{-msoft-float}.
21296 The default is to not use the packed stack layout.
21299 @itemx -mno-small-exec
21300 @opindex msmall-exec
21301 @opindex mno-small-exec
21302 Generate (or do not generate) code using the @code{bras} instruction
21303 to do subroutine calls.
21304 This only works reliably if the total executable size does not
21305 exceed 64k. The default is to use the @code{basr} instruction instead,
21306 which does not have this limitation.
21312 When @option{-m31} is specified, generate code compliant to the
21313 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
21314 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
21315 particular to generate 64-bit instructions. For the @samp{s390}
21316 targets, the default is @option{-m31}, while the @samp{s390x}
21317 targets default to @option{-m64}.
21323 When @option{-mzarch} is specified, generate code using the
21324 instructions available on z/Architecture.
21325 When @option{-mesa} is specified, generate code using the
21326 instructions available on ESA/390. Note that @option{-mesa} is
21327 not possible with @option{-m64}.
21328 When generating code compliant to the GNU/Linux for S/390 ABI,
21329 the default is @option{-mesa}. When generating code compliant
21330 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
21336 The @option{-mhtm} option enables a set of builtins making use of
21337 instructions available with the transactional execution facility
21338 introduced with the IBM zEnterprise EC12 machine generation
21339 @ref{S/390 System z Built-in Functions}.
21340 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
21346 When @option{-mvx} is specified, generate code using the instructions
21347 available with the vector extension facility introduced with the IBM
21348 z13 machine generation.
21349 This option changes the ABI for some vector type values with regard to
21350 alignment and calling conventions. In case vector type values are
21351 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
21352 command will be added to mark the resulting binary with the ABI used.
21353 @option{-mvx} is enabled by default when using @option{-march=z13}.
21356 @itemx -mno-zvector
21358 @opindex mno-zvector
21359 The @option{-mzvector} option enables vector language extensions and
21360 builtins using instructions available with the vector extension
21361 facility introduced with the IBM z13 machine generation.
21362 This option adds support for @samp{vector} to be used as a keyword to
21363 define vector type variables and arguments. @samp{vector} is only
21364 available when GNU extensions are enabled. It will not be expanded
21365 when requesting strict standard compliance e.g. with @option{-std=c99}.
21366 In addition to the GCC low-level builtins @option{-mzvector} enables
21367 a set of builtins added for compatibility with AltiVec-style
21368 implementations like Power and Cell. In order to make use of these
21369 builtins the header file @file{vecintrin.h} needs to be included.
21370 @option{-mzvector} is disabled by default.
21376 Generate (or do not generate) code using the @code{mvcle} instruction
21377 to perform block moves. When @option{-mno-mvcle} is specified,
21378 use a @code{mvc} loop instead. This is the default unless optimizing for
21385 Print (or do not print) additional debug information when compiling.
21386 The default is to not print debug information.
21388 @item -march=@var{cpu-type}
21390 Generate code that runs on @var{cpu-type}, which is the name of a
21391 system representing a certain processor type. Possible values for
21392 @var{cpu-type} are @samp{z900}, @samp{z990}, @samp{z9-109},
21393 @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12}, and @samp{z13}.
21394 The default is @option{-march=z900}. @samp{g5} and @samp{g6} are
21395 deprecated and will be removed with future releases.
21397 @item -mtune=@var{cpu-type}
21399 Tune to @var{cpu-type} everything applicable about the generated code,
21400 except for the ABI and the set of available instructions.
21401 The list of @var{cpu-type} values is the same as for @option{-march}.
21402 The default is the value used for @option{-march}.
21405 @itemx -mno-tpf-trace
21406 @opindex mtpf-trace
21407 @opindex mno-tpf-trace
21408 Generate code that adds (does not add) in TPF OS specific branches to trace
21409 routines in the operating system. This option is off by default, even
21410 when compiling for the TPF OS@.
21413 @itemx -mno-fused-madd
21414 @opindex mfused-madd
21415 @opindex mno-fused-madd
21416 Generate code that uses (does not use) the floating-point multiply and
21417 accumulate instructions. These instructions are generated by default if
21418 hardware floating point is used.
21420 @item -mwarn-framesize=@var{framesize}
21421 @opindex mwarn-framesize
21422 Emit a warning if the current function exceeds the given frame size. Because
21423 this is a compile-time check it doesn't need to be a real problem when the program
21424 runs. It is intended to identify functions that most probably cause
21425 a stack overflow. It is useful to be used in an environment with limited stack
21426 size e.g.@: the linux kernel.
21428 @item -mwarn-dynamicstack
21429 @opindex mwarn-dynamicstack
21430 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
21431 arrays. This is generally a bad idea with a limited stack size.
21433 @item -mstack-guard=@var{stack-guard}
21434 @itemx -mstack-size=@var{stack-size}
21435 @opindex mstack-guard
21436 @opindex mstack-size
21437 If these options are provided the S/390 back end emits additional instructions in
21438 the function prologue that trigger a trap if the stack size is @var{stack-guard}
21439 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
21440 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
21441 the frame size of the compiled function is chosen.
21442 These options are intended to be used to help debugging stack overflow problems.
21443 The additionally emitted code causes only little overhead and hence can also be
21444 used in production-like systems without greater performance degradation. The given
21445 values have to be exact powers of 2 and @var{stack-size} has to be greater than
21446 @var{stack-guard} without exceeding 64k.
21447 In order to be efficient the extra code makes the assumption that the stack starts
21448 at an address aligned to the value given by @var{stack-size}.
21449 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
21451 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
21453 If the hotpatch option is enabled, a ``hot-patching'' function
21454 prologue is generated for all functions in the compilation unit.
21455 The funtion label is prepended with the given number of two-byte
21456 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
21457 the label, 2 * @var{post-halfwords} bytes are appended, using the
21458 largest NOP like instructions the architecture allows (maximum
21461 If both arguments are zero, hotpatching is disabled.
21463 This option can be overridden for individual functions with the
21464 @code{hotpatch} attribute.
21467 @node Score Options
21468 @subsection Score Options
21469 @cindex Score Options
21471 These options are defined for Score implementations:
21476 Compile code for big-endian mode. This is the default.
21480 Compile code for little-endian mode.
21484 Disable generation of @code{bcnz} instructions.
21488 Enable generation of unaligned load and store instructions.
21492 Enable the use of multiply-accumulate instructions. Disabled by default.
21496 Specify the SCORE5 as the target architecture.
21500 Specify the SCORE5U of the target architecture.
21504 Specify the SCORE7 as the target architecture. This is the default.
21508 Specify the SCORE7D as the target architecture.
21512 @subsection SH Options
21514 These @samp{-m} options are defined for the SH implementations:
21519 Generate code for the SH1.
21523 Generate code for the SH2.
21526 Generate code for the SH2e.
21530 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
21531 that the floating-point unit is not used.
21533 @item -m2a-single-only
21534 @opindex m2a-single-only
21535 Generate code for the SH2a-FPU, in such a way that no double-precision
21536 floating-point operations are used.
21539 @opindex m2a-single
21540 Generate code for the SH2a-FPU assuming the floating-point unit is in
21541 single-precision mode by default.
21545 Generate code for the SH2a-FPU assuming the floating-point unit is in
21546 double-precision mode by default.
21550 Generate code for the SH3.
21554 Generate code for the SH3e.
21558 Generate code for the SH4 without a floating-point unit.
21560 @item -m4-single-only
21561 @opindex m4-single-only
21562 Generate code for the SH4 with a floating-point unit that only
21563 supports single-precision arithmetic.
21567 Generate code for the SH4 assuming the floating-point unit is in
21568 single-precision mode by default.
21572 Generate code for the SH4.
21576 Generate code for SH4-100.
21578 @item -m4-100-nofpu
21579 @opindex m4-100-nofpu
21580 Generate code for SH4-100 in such a way that the
21581 floating-point unit is not used.
21583 @item -m4-100-single
21584 @opindex m4-100-single
21585 Generate code for SH4-100 assuming the floating-point unit is in
21586 single-precision mode by default.
21588 @item -m4-100-single-only
21589 @opindex m4-100-single-only
21590 Generate code for SH4-100 in such a way that no double-precision
21591 floating-point operations are used.
21595 Generate code for SH4-200.
21597 @item -m4-200-nofpu
21598 @opindex m4-200-nofpu
21599 Generate code for SH4-200 without in such a way that the
21600 floating-point unit is not used.
21602 @item -m4-200-single
21603 @opindex m4-200-single
21604 Generate code for SH4-200 assuming the floating-point unit is in
21605 single-precision mode by default.
21607 @item -m4-200-single-only
21608 @opindex m4-200-single-only
21609 Generate code for SH4-200 in such a way that no double-precision
21610 floating-point operations are used.
21614 Generate code for SH4-300.
21616 @item -m4-300-nofpu
21617 @opindex m4-300-nofpu
21618 Generate code for SH4-300 without in such a way that the
21619 floating-point unit is not used.
21621 @item -m4-300-single
21622 @opindex m4-300-single
21623 Generate code for SH4-300 in such a way that no double-precision
21624 floating-point operations are used.
21626 @item -m4-300-single-only
21627 @opindex m4-300-single-only
21628 Generate code for SH4-300 in such a way that no double-precision
21629 floating-point operations are used.
21633 Generate code for SH4-340 (no MMU, no FPU).
21637 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21642 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21643 floating-point unit is not used.
21645 @item -m4a-single-only
21646 @opindex m4a-single-only
21647 Generate code for the SH4a, in such a way that no double-precision
21648 floating-point operations are used.
21651 @opindex m4a-single
21652 Generate code for the SH4a assuming the floating-point unit is in
21653 single-precision mode by default.
21657 Generate code for the SH4a.
21661 Same as @option{-m4a-nofpu}, except that it implicitly passes
21662 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21663 instructions at the moment.
21667 Compile code for the processor in big-endian mode.
21671 Compile code for the processor in little-endian mode.
21675 Align doubles at 64-bit boundaries. Note that this changes the calling
21676 conventions, and thus some functions from the standard C library do
21677 not work unless you recompile it first with @option{-mdalign}.
21681 Shorten some address references at link time, when possible; uses the
21682 linker option @option{-relax}.
21686 Use 32-bit offsets in @code{switch} tables. The default is to use
21691 Enable the use of bit manipulation instructions on SH2A.
21695 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21696 alignment constraints.
21700 Comply with the calling conventions defined by Renesas.
21703 @opindex mno-renesas
21704 Comply with the calling conventions defined for GCC before the Renesas
21705 conventions were available. This option is the default for all
21706 targets of the SH toolchain.
21709 @opindex mnomacsave
21710 Mark the @code{MAC} register as call-clobbered, even if
21711 @option{-mrenesas} is given.
21717 Control the IEEE compliance of floating-point comparisons, which affects the
21718 handling of cases where the result of a comparison is unordered. By default
21719 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21720 enabled @option{-mno-ieee} is implicitly set, which results in faster
21721 floating-point greater-equal and less-equal comparisons. The implicit settings
21722 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21724 @item -minline-ic_invalidate
21725 @opindex minline-ic_invalidate
21726 Inline code to invalidate instruction cache entries after setting up
21727 nested function trampolines.
21728 This option has no effect if @option{-musermode} is in effect and the selected
21729 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21731 If the selected code generation option does not allow the use of the @code{icbi}
21732 instruction, and @option{-musermode} is not in effect, the inlined code
21733 manipulates the instruction cache address array directly with an associative
21734 write. This not only requires privileged mode at run time, but it also
21735 fails if the cache line had been mapped via the TLB and has become unmapped.
21739 Dump instruction size and location in the assembly code.
21742 @opindex mpadstruct
21743 This option is deprecated. It pads structures to multiple of 4 bytes,
21744 which is incompatible with the SH ABI@.
21746 @item -matomic-model=@var{model}
21747 @opindex matomic-model=@var{model}
21748 Sets the model of atomic operations and additional parameters as a comma
21749 separated list. For details on the atomic built-in functions see
21750 @ref{__atomic Builtins}. The following models and parameters are supported:
21755 Disable compiler generated atomic sequences and emit library calls for atomic
21756 operations. This is the default if the target is not @code{sh*-*-linux*}.
21759 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21760 built-in functions. The generated atomic sequences require additional support
21761 from the interrupt/exception handling code of the system and are only suitable
21762 for SH3* and SH4* single-core systems. This option is enabled by default when
21763 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21764 this option also partially utilizes the hardware atomic instructions
21765 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21766 @samp{strict} is specified.
21769 Generate software atomic sequences that use a variable in the thread control
21770 block. This is a variation of the gUSA sequences which can also be used on
21771 SH1* and SH2* targets. The generated atomic sequences require additional
21772 support from the interrupt/exception handling code of the system and are only
21773 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21774 parameter has to be specified as well.
21777 Generate software atomic sequences that temporarily disable interrupts by
21778 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21779 in privileged mode and is only suitable for single-core systems. Additional
21780 support from the interrupt/exception handling code of the system is not
21781 required. This model is enabled by default when the target is
21782 @code{sh*-*-linux*} and SH1* or SH2*.
21785 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21786 instructions only. This is only available on SH4A and is suitable for
21787 multi-core systems. Since the hardware instructions support only 32 bit atomic
21788 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21789 Code compiled with this option is also compatible with other software
21790 atomic model interrupt/exception handling systems if executed on an SH4A
21791 system. Additional support from the interrupt/exception handling code of the
21792 system is not required for this model.
21795 This parameter specifies the offset in bytes of the variable in the thread
21796 control block structure that should be used by the generated atomic sequences
21797 when the @samp{soft-tcb} model has been selected. For other models this
21798 parameter is ignored. The specified value must be an integer multiple of four
21799 and in the range 0-1020.
21802 This parameter prevents mixed usage of multiple atomic models, even if they
21803 are compatible, and makes the compiler generate atomic sequences of the
21804 specified model only.
21810 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21811 Notice that depending on the particular hardware and software configuration
21812 this can degrade overall performance due to the operand cache line flushes
21813 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21814 processors the @code{tas.b} instruction must be used with caution since it
21815 can result in data corruption for certain cache configurations.
21818 @opindex mprefergot
21819 When generating position-independent code, emit function calls using
21820 the Global Offset Table instead of the Procedure Linkage Table.
21823 @itemx -mno-usermode
21825 @opindex mno-usermode
21826 Don't allow (allow) the compiler generating privileged mode code. Specifying
21827 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21828 inlined code would not work in user mode. @option{-musermode} is the default
21829 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21830 @option{-musermode} has no effect, since there is no user mode.
21832 @item -multcost=@var{number}
21833 @opindex multcost=@var{number}
21834 Set the cost to assume for a multiply insn.
21836 @item -mdiv=@var{strategy}
21837 @opindex mdiv=@var{strategy}
21838 Set the division strategy to be used for integer division operations.
21839 @var{strategy} can be one of:
21844 Calls a library function that uses the single-step division instruction
21845 @code{div1} to perform the operation. Division by zero calculates an
21846 unspecified result and does not trap. This is the default except for SH4,
21847 SH2A and SHcompact.
21850 Calls a library function that performs the operation in double precision
21851 floating point. Division by zero causes a floating-point exception. This is
21852 the default for SHcompact with FPU. Specifying this for targets that do not
21853 have a double precision FPU defaults to @code{call-div1}.
21856 Calls a library function that uses a lookup table for small divisors and
21857 the @code{div1} instruction with case distinction for larger divisors. Division
21858 by zero calculates an unspecified result and does not trap. This is the default
21859 for SH4. Specifying this for targets that do not have dynamic shift
21860 instructions defaults to @code{call-div1}.
21864 When a division strategy has not been specified the default strategy is
21865 selected based on the current target. For SH2A the default strategy is to
21866 use the @code{divs} and @code{divu} instructions instead of library function
21869 @item -maccumulate-outgoing-args
21870 @opindex maccumulate-outgoing-args
21871 Reserve space once for outgoing arguments in the function prologue rather
21872 than around each call. Generally beneficial for performance and size. Also
21873 needed for unwinding to avoid changing the stack frame around conditional code.
21875 @item -mdivsi3_libfunc=@var{name}
21876 @opindex mdivsi3_libfunc=@var{name}
21877 Set the name of the library function used for 32-bit signed division to
21879 This only affects the name used in the @samp{call} division strategies, and
21880 the compiler still expects the same sets of input/output/clobbered registers as
21881 if this option were not present.
21883 @item -mfixed-range=@var{register-range}
21884 @opindex mfixed-range
21885 Generate code treating the given register range as fixed registers.
21886 A fixed register is one that the register allocator can not use. This is
21887 useful when compiling kernel code. A register range is specified as
21888 two registers separated by a dash. Multiple register ranges can be
21889 specified separated by a comma.
21891 @item -mbranch-cost=@var{num}
21892 @opindex mbranch-cost=@var{num}
21893 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21894 make the compiler try to generate more branch-free code if possible.
21895 If not specified the value is selected depending on the processor type that
21896 is being compiled for.
21899 @itemx -mno-zdcbranch
21900 @opindex mzdcbranch
21901 @opindex mno-zdcbranch
21902 Assume (do not assume) that zero displacement conditional branch instructions
21903 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21904 compiler prefers zero displacement branch code sequences. This is
21905 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21906 disabled by specifying @option{-mno-zdcbranch}.
21908 @item -mcbranch-force-delay-slot
21909 @opindex mcbranch-force-delay-slot
21910 Force the usage of delay slots for conditional branches, which stuffs the delay
21911 slot with a @code{nop} if a suitable instruction can't be found. By default
21912 this option is disabled. It can be enabled to work around hardware bugs as
21913 found in the original SH7055.
21916 @itemx -mno-fused-madd
21917 @opindex mfused-madd
21918 @opindex mno-fused-madd
21919 Generate code that uses (does not use) the floating-point multiply and
21920 accumulate instructions. These instructions are generated by default
21921 if hardware floating point is used. The machine-dependent
21922 @option{-mfused-madd} option is now mapped to the machine-independent
21923 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21924 mapped to @option{-ffp-contract=off}.
21930 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21931 and cosine approximations. The option @option{-mfsca} must be used in
21932 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21933 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21934 approximations even if @option{-funsafe-math-optimizations} is in effect.
21940 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21941 reciprocal square root approximations. The option @option{-mfsrra} must be used
21942 in combination with @option{-funsafe-math-optimizations} and
21943 @option{-ffinite-math-only}. It is enabled by default when generating code for
21944 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21945 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21948 @item -mpretend-cmove
21949 @opindex mpretend-cmove
21950 Prefer zero-displacement conditional branches for conditional move instruction
21951 patterns. This can result in faster code on the SH4 processor.
21955 Generate code using the FDPIC ABI.
21959 @node Solaris 2 Options
21960 @subsection Solaris 2 Options
21961 @cindex Solaris 2 options
21963 These @samp{-m} options are supported on Solaris 2:
21966 @item -mclear-hwcap
21967 @opindex mclear-hwcap
21968 @option{-mclear-hwcap} tells the compiler to remove the hardware
21969 capabilities generated by the Solaris assembler. This is only necessary
21970 when object files use ISA extensions not supported by the current
21971 machine, but check at runtime whether or not to use them.
21973 @item -mimpure-text
21974 @opindex mimpure-text
21975 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21976 the compiler to not pass @option{-z text} to the linker when linking a
21977 shared object. Using this option, you can link position-dependent
21978 code into a shared object.
21980 @option{-mimpure-text} suppresses the ``relocations remain against
21981 allocatable but non-writable sections'' linker error message.
21982 However, the necessary relocations trigger copy-on-write, and the
21983 shared object is not actually shared across processes. Instead of
21984 using @option{-mimpure-text}, you should compile all source code with
21985 @option{-fpic} or @option{-fPIC}.
21989 These switches are supported in addition to the above on Solaris 2:
21994 Add support for multithreading using the POSIX threads library. This
21995 option sets flags for both the preprocessor and linker. This option does
21996 not affect the thread safety of object code produced by the compiler or
21997 that of libraries supplied with it.
22001 This is a synonym for @option{-pthreads}.
22004 @node SPARC Options
22005 @subsection SPARC Options
22006 @cindex SPARC options
22008 These @samp{-m} options are supported on the SPARC:
22011 @item -mno-app-regs
22013 @opindex mno-app-regs
22015 Specify @option{-mapp-regs} to generate output using the global registers
22016 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
22017 global register 1, each global register 2 through 4 is then treated as an
22018 allocable register that is clobbered by function calls. This is the default.
22020 To be fully SVR4 ABI-compliant at the cost of some performance loss,
22021 specify @option{-mno-app-regs}. You should compile libraries and system
22022 software with this option.
22028 With @option{-mflat}, the compiler does not generate save/restore instructions
22029 and uses a ``flat'' or single register window model. This model is compatible
22030 with the regular register window model. The local registers and the input
22031 registers (0--5) are still treated as ``call-saved'' registers and are
22032 saved on the stack as needed.
22034 With @option{-mno-flat} (the default), the compiler generates save/restore
22035 instructions (except for leaf functions). This is the normal operating mode.
22038 @itemx -mhard-float
22040 @opindex mhard-float
22041 Generate output containing floating-point instructions. This is the
22045 @itemx -msoft-float
22047 @opindex msoft-float
22048 Generate output containing library calls for floating point.
22049 @strong{Warning:} the requisite libraries are not available for all SPARC
22050 targets. Normally the facilities of the machine's usual C compiler are
22051 used, but this cannot be done directly in cross-compilation. You must make
22052 your own arrangements to provide suitable library functions for
22053 cross-compilation. The embedded targets @samp{sparc-*-aout} and
22054 @samp{sparclite-*-*} do provide software floating-point support.
22056 @option{-msoft-float} changes the calling convention in the output file;
22057 therefore, it is only useful if you compile @emph{all} of a program with
22058 this option. In particular, you need to compile @file{libgcc.a}, the
22059 library that comes with GCC, with @option{-msoft-float} in order for
22062 @item -mhard-quad-float
22063 @opindex mhard-quad-float
22064 Generate output containing quad-word (long double) floating-point
22067 @item -msoft-quad-float
22068 @opindex msoft-quad-float
22069 Generate output containing library calls for quad-word (long double)
22070 floating-point instructions. The functions called are those specified
22071 in the SPARC ABI@. This is the default.
22073 As of this writing, there are no SPARC implementations that have hardware
22074 support for the quad-word floating-point instructions. They all invoke
22075 a trap handler for one of these instructions, and then the trap handler
22076 emulates the effect of the instruction. Because of the trap handler overhead,
22077 this is much slower than calling the ABI library routines. Thus the
22078 @option{-msoft-quad-float} option is the default.
22080 @item -mno-unaligned-doubles
22081 @itemx -munaligned-doubles
22082 @opindex mno-unaligned-doubles
22083 @opindex munaligned-doubles
22084 Assume that doubles have 8-byte alignment. This is the default.
22086 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
22087 alignment only if they are contained in another type, or if they have an
22088 absolute address. Otherwise, it assumes they have 4-byte alignment.
22089 Specifying this option avoids some rare compatibility problems with code
22090 generated by other compilers. It is not the default because it results
22091 in a performance loss, especially for floating-point code.
22094 @itemx -mno-user-mode
22095 @opindex muser-mode
22096 @opindex mno-user-mode
22097 Do not generate code that can only run in supervisor mode. This is relevant
22098 only for the @code{casa} instruction emitted for the LEON3 processor. This
22101 @item -mfaster-structs
22102 @itemx -mno-faster-structs
22103 @opindex mfaster-structs
22104 @opindex mno-faster-structs
22105 With @option{-mfaster-structs}, the compiler assumes that structures
22106 should have 8-byte alignment. This enables the use of pairs of
22107 @code{ldd} and @code{std} instructions for copies in structure
22108 assignment, in place of twice as many @code{ld} and @code{st} pairs.
22109 However, the use of this changed alignment directly violates the SPARC
22110 ABI@. Thus, it's intended only for use on targets where the developer
22111 acknowledges that their resulting code is not directly in line with
22112 the rules of the ABI@.
22114 @item -mstd-struct-return
22115 @itemx -mno-std-struct-return
22116 @opindex mstd-struct-return
22117 @opindex mno-std-struct-return
22118 With @option{-mstd-struct-return}, the compiler generates checking code
22119 in functions returning structures or unions to detect size mismatches
22120 between the two sides of function calls, as per the 32-bit ABI@.
22122 The default is @option{-mno-std-struct-return}. This option has no effect
22125 @item -mcpu=@var{cpu_type}
22127 Set the instruction set, register set, and instruction scheduling parameters
22128 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22129 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
22130 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
22131 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
22132 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
22133 @samp{niagara3}, @samp{niagara4} and @samp{niagara7}.
22135 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
22136 which selects the best architecture option for the host processor.
22137 @option{-mcpu=native} has no effect if GCC does not recognize
22140 Default instruction scheduling parameters are used for values that select
22141 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
22142 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
22144 Here is a list of each supported architecture and their supported
22152 supersparc, hypersparc, leon, leon3
22155 f930, f934, sparclite86x
22161 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7
22164 By default (unless configured otherwise), GCC generates code for the V7
22165 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
22166 additionally optimizes it for the Cypress CY7C602 chip, as used in the
22167 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
22168 SPARCStation 1, 2, IPX etc.
22170 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
22171 architecture. The only difference from V7 code is that the compiler emits
22172 the integer multiply and integer divide instructions which exist in SPARC-V8
22173 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
22174 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
22177 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
22178 the SPARC architecture. This adds the integer multiply, integer divide step
22179 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
22180 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
22181 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
22182 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
22183 MB86934 chip, which is the more recent SPARClite with FPU@.
22185 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
22186 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
22187 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
22188 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
22189 optimizes it for the TEMIC SPARClet chip.
22191 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
22192 architecture. This adds 64-bit integer and floating-point move instructions,
22193 3 additional floating-point condition code registers and conditional move
22194 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
22195 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
22196 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
22197 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
22198 @option{-mcpu=niagara}, the compiler additionally optimizes it for
22199 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
22200 additionally optimizes it for Sun UltraSPARC T2 chips. With
22201 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
22202 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
22203 additionally optimizes it for Sun UltraSPARC T4 chips. With
22204 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
22205 Oracle SPARC M7 chips.
22207 @item -mtune=@var{cpu_type}
22209 Set the instruction scheduling parameters for machine type
22210 @var{cpu_type}, but do not set the instruction set or register set that the
22211 option @option{-mcpu=@var{cpu_type}} does.
22213 The same values for @option{-mcpu=@var{cpu_type}} can be used for
22214 @option{-mtune=@var{cpu_type}}, but the only useful values are those
22215 that select a particular CPU implementation. Those are
22216 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
22217 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
22218 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
22219 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
22220 @samp{niagara4} and @samp{niagara7}. With native Solaris and
22221 GNU/Linux toolchains, @samp{native} can also be used.
22226 @opindex mno-v8plus
22227 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
22228 difference from the V8 ABI is that the global and out registers are
22229 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
22230 mode for all SPARC-V9 processors.
22236 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
22237 Visual Instruction Set extensions. The default is @option{-mno-vis}.
22243 With @option{-mvis2}, GCC generates code that takes advantage of
22244 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
22245 default is @option{-mvis2} when targeting a cpu that supports such
22246 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
22247 also sets @option{-mvis}.
22253 With @option{-mvis3}, GCC generates code that takes advantage of
22254 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
22255 default is @option{-mvis3} when targeting a cpu that supports such
22256 instructions, such as niagara-3 and later. Setting @option{-mvis3}
22257 also sets @option{-mvis2} and @option{-mvis}.
22263 With @option{-mvis4}, GCC generates code that takes advantage of
22264 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
22265 default is @option{-mvis4} when targeting a cpu that supports such
22266 instructions, such as niagara-7 and later. Setting @option{-mvis4}
22267 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
22272 @opindex mno-cbcond
22273 With @option{-mcbcond}, GCC generates code that takes advantage of
22274 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
22275 The default is @option{-mcbcond} when targeting a cpu that supports such
22276 instructions, such as niagara-4 and later.
22282 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
22283 population count instruction. The default is @option{-mpopc}
22284 when targeting a cpu that supports such instructions, such as Niagara-2 and
22291 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
22292 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
22293 when targeting a cpu that supports such instructions, such as Niagara-3 and
22297 @opindex mfix-at697f
22298 Enable the documented workaround for the single erratum of the Atmel AT697F
22299 processor (which corresponds to erratum #13 of the AT697E processor).
22302 @opindex mfix-ut699
22303 Enable the documented workarounds for the floating-point errata and the data
22304 cache nullify errata of the UT699 processor.
22307 These @samp{-m} options are supported in addition to the above
22308 on SPARC-V9 processors in 64-bit environments:
22315 Generate code for a 32-bit or 64-bit environment.
22316 The 32-bit environment sets int, long and pointer to 32 bits.
22317 The 64-bit environment sets int to 32 bits and long and pointer
22320 @item -mcmodel=@var{which}
22322 Set the code model to one of
22326 The Medium/Low code model: 64-bit addresses, programs
22327 must be linked in the low 32 bits of memory. Programs can be statically
22328 or dynamically linked.
22331 The Medium/Middle code model: 64-bit addresses, programs
22332 must be linked in the low 44 bits of memory, the text and data segments must
22333 be less than 2GB in size and the data segment must be located within 2GB of
22337 The Medium/Anywhere code model: 64-bit addresses, programs
22338 may be linked anywhere in memory, the text and data segments must be less
22339 than 2GB in size and the data segment must be located within 2GB of the
22343 The Medium/Anywhere code model for embedded systems:
22344 64-bit addresses, the text and data segments must be less than 2GB in
22345 size, both starting anywhere in memory (determined at link time). The
22346 global register %g4 points to the base of the data segment. Programs
22347 are statically linked and PIC is not supported.
22350 @item -mmemory-model=@var{mem-model}
22351 @opindex mmemory-model
22352 Set the memory model in force on the processor to one of
22356 The default memory model for the processor and operating system.
22359 Relaxed Memory Order
22362 Partial Store Order
22368 Sequential Consistency
22371 These memory models are formally defined in Appendix D of the Sparc V9
22372 architecture manual, as set in the processor's @code{PSTATE.MM} field.
22375 @itemx -mno-stack-bias
22376 @opindex mstack-bias
22377 @opindex mno-stack-bias
22378 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
22379 frame pointer if present, are offset by @minus{}2047 which must be added back
22380 when making stack frame references. This is the default in 64-bit mode.
22381 Otherwise, assume no such offset is present.
22385 @subsection SPU Options
22386 @cindex SPU options
22388 These @samp{-m} options are supported on the SPU:
22392 @itemx -merror-reloc
22393 @opindex mwarn-reloc
22394 @opindex merror-reloc
22396 The loader for SPU does not handle dynamic relocations. By default, GCC
22397 gives an error when it generates code that requires a dynamic
22398 relocation. @option{-mno-error-reloc} disables the error,
22399 @option{-mwarn-reloc} generates a warning instead.
22402 @itemx -munsafe-dma
22404 @opindex munsafe-dma
22406 Instructions that initiate or test completion of DMA must not be
22407 reordered with respect to loads and stores of the memory that is being
22409 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
22410 memory accesses, but that can lead to inefficient code in places where the
22411 memory is known to not change. Rather than mark the memory as volatile,
22412 you can use @option{-msafe-dma} to tell the compiler to treat
22413 the DMA instructions as potentially affecting all memory.
22415 @item -mbranch-hints
22416 @opindex mbranch-hints
22418 By default, GCC generates a branch hint instruction to avoid
22419 pipeline stalls for always-taken or probably-taken branches. A hint
22420 is not generated closer than 8 instructions away from its branch.
22421 There is little reason to disable them, except for debugging purposes,
22422 or to make an object a little bit smaller.
22426 @opindex msmall-mem
22427 @opindex mlarge-mem
22429 By default, GCC generates code assuming that addresses are never larger
22430 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
22431 a full 32-bit address.
22436 By default, GCC links against startup code that assumes the SPU-style
22437 main function interface (which has an unconventional parameter list).
22438 With @option{-mstdmain}, GCC links your program against startup
22439 code that assumes a C99-style interface to @code{main}, including a
22440 local copy of @code{argv} strings.
22442 @item -mfixed-range=@var{register-range}
22443 @opindex mfixed-range
22444 Generate code treating the given register range as fixed registers.
22445 A fixed register is one that the register allocator cannot use. This is
22446 useful when compiling kernel code. A register range is specified as
22447 two registers separated by a dash. Multiple register ranges can be
22448 specified separated by a comma.
22454 Compile code assuming that pointers to the PPU address space accessed
22455 via the @code{__ea} named address space qualifier are either 32 or 64
22456 bits wide. The default is 32 bits. As this is an ABI-changing option,
22457 all object code in an executable must be compiled with the same setting.
22459 @item -maddress-space-conversion
22460 @itemx -mno-address-space-conversion
22461 @opindex maddress-space-conversion
22462 @opindex mno-address-space-conversion
22463 Allow/disallow treating the @code{__ea} address space as superset
22464 of the generic address space. This enables explicit type casts
22465 between @code{__ea} and generic pointer as well as implicit
22466 conversions of generic pointers to @code{__ea} pointers. The
22467 default is to allow address space pointer conversions.
22469 @item -mcache-size=@var{cache-size}
22470 @opindex mcache-size
22471 This option controls the version of libgcc that the compiler links to an
22472 executable and selects a software-managed cache for accessing variables
22473 in the @code{__ea} address space with a particular cache size. Possible
22474 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
22475 and @samp{128}. The default cache size is 64KB.
22477 @item -matomic-updates
22478 @itemx -mno-atomic-updates
22479 @opindex matomic-updates
22480 @opindex mno-atomic-updates
22481 This option controls the version of libgcc that the compiler links to an
22482 executable and selects whether atomic updates to the software-managed
22483 cache of PPU-side variables are used. If you use atomic updates, changes
22484 to a PPU variable from SPU code using the @code{__ea} named address space
22485 qualifier do not interfere with changes to other PPU variables residing
22486 in the same cache line from PPU code. If you do not use atomic updates,
22487 such interference may occur; however, writing back cache lines is
22488 more efficient. The default behavior is to use atomic updates.
22491 @itemx -mdual-nops=@var{n}
22492 @opindex mdual-nops
22493 By default, GCC inserts nops to increase dual issue when it expects
22494 it to increase performance. @var{n} can be a value from 0 to 10. A
22495 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
22496 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
22498 @item -mhint-max-nops=@var{n}
22499 @opindex mhint-max-nops
22500 Maximum number of nops to insert for a branch hint. A branch hint must
22501 be at least 8 instructions away from the branch it is affecting. GCC
22502 inserts up to @var{n} nops to enforce this, otherwise it does not
22503 generate the branch hint.
22505 @item -mhint-max-distance=@var{n}
22506 @opindex mhint-max-distance
22507 The encoding of the branch hint instruction limits the hint to be within
22508 256 instructions of the branch it is affecting. By default, GCC makes
22509 sure it is within 125.
22512 @opindex msafe-hints
22513 Work around a hardware bug that causes the SPU to stall indefinitely.
22514 By default, GCC inserts the @code{hbrp} instruction to make sure
22515 this stall won't happen.
22519 @node System V Options
22520 @subsection Options for System V
22522 These additional options are available on System V Release 4 for
22523 compatibility with other compilers on those systems:
22528 Create a shared object.
22529 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22533 Identify the versions of each tool used by the compiler, in a
22534 @code{.ident} assembler directive in the output.
22538 Refrain from adding @code{.ident} directives to the output file (this is
22541 @item -YP,@var{dirs}
22543 Search the directories @var{dirs}, and no others, for libraries
22544 specified with @option{-l}.
22546 @item -Ym,@var{dir}
22548 Look in the directory @var{dir} to find the M4 preprocessor.
22549 The assembler uses this option.
22550 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22551 @c the generic assembler that comes with Solaris takes just -Ym.
22554 @node TILE-Gx Options
22555 @subsection TILE-Gx Options
22556 @cindex TILE-Gx options
22558 These @samp{-m} options are supported on the TILE-Gx:
22561 @item -mcmodel=small
22562 @opindex mcmodel=small
22563 Generate code for the small model. The distance for direct calls is
22564 limited to 500M in either direction. PC-relative addresses are 32
22565 bits. Absolute addresses support the full address range.
22567 @item -mcmodel=large
22568 @opindex mcmodel=large
22569 Generate code for the large model. There is no limitation on call
22570 distance, pc-relative addresses, or absolute addresses.
22572 @item -mcpu=@var{name}
22574 Selects the type of CPU to be targeted. Currently the only supported
22575 type is @samp{tilegx}.
22581 Generate code for a 32-bit or 64-bit environment. The 32-bit
22582 environment sets int, long, and pointer to 32 bits. The 64-bit
22583 environment sets int to 32 bits and long and pointer to 64 bits.
22586 @itemx -mlittle-endian
22587 @opindex mbig-endian
22588 @opindex mlittle-endian
22589 Generate code in big/little endian mode, respectively.
22592 @node TILEPro Options
22593 @subsection TILEPro Options
22594 @cindex TILEPro options
22596 These @samp{-m} options are supported on the TILEPro:
22599 @item -mcpu=@var{name}
22601 Selects the type of CPU to be targeted. Currently the only supported
22602 type is @samp{tilepro}.
22606 Generate code for a 32-bit environment, which sets int, long, and
22607 pointer to 32 bits. This is the only supported behavior so the flag
22608 is essentially ignored.
22612 @subsection V850 Options
22613 @cindex V850 Options
22615 These @samp{-m} options are defined for V850 implementations:
22619 @itemx -mno-long-calls
22620 @opindex mlong-calls
22621 @opindex mno-long-calls
22622 Treat all calls as being far away (near). If calls are assumed to be
22623 far away, the compiler always loads the function's address into a
22624 register, and calls indirect through the pointer.
22630 Do not optimize (do optimize) basic blocks that use the same index
22631 pointer 4 or more times to copy pointer into the @code{ep} register, and
22632 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22633 option is on by default if you optimize.
22635 @item -mno-prolog-function
22636 @itemx -mprolog-function
22637 @opindex mno-prolog-function
22638 @opindex mprolog-function
22639 Do not use (do use) external functions to save and restore registers
22640 at the prologue and epilogue of a function. The external functions
22641 are slower, but use less code space if more than one function saves
22642 the same number of registers. The @option{-mprolog-function} option
22643 is on by default if you optimize.
22647 Try to make the code as small as possible. At present, this just turns
22648 on the @option{-mep} and @option{-mprolog-function} options.
22650 @item -mtda=@var{n}
22652 Put static or global variables whose size is @var{n} bytes or less into
22653 the tiny data area that register @code{ep} points to. The tiny data
22654 area can hold up to 256 bytes in total (128 bytes for byte references).
22656 @item -msda=@var{n}
22658 Put static or global variables whose size is @var{n} bytes or less into
22659 the small data area that register @code{gp} points to. The small data
22660 area can hold up to 64 kilobytes.
22662 @item -mzda=@var{n}
22664 Put static or global variables whose size is @var{n} bytes or less into
22665 the first 32 kilobytes of memory.
22669 Specify that the target processor is the V850.
22673 Specify that the target processor is the V850E3V5. The preprocessor
22674 constant @code{__v850e3v5__} is defined if this option is used.
22678 Specify that the target processor is the V850E3V5. This is an alias for
22679 the @option{-mv850e3v5} option.
22683 Specify that the target processor is the V850E2V3. The preprocessor
22684 constant @code{__v850e2v3__} is defined if this option is used.
22688 Specify that the target processor is the V850E2. The preprocessor
22689 constant @code{__v850e2__} is defined if this option is used.
22693 Specify that the target processor is the V850E1. The preprocessor
22694 constants @code{__v850e1__} and @code{__v850e__} are defined if
22695 this option is used.
22699 Specify that the target processor is the V850ES. This is an alias for
22700 the @option{-mv850e1} option.
22704 Specify that the target processor is the V850E@. The preprocessor
22705 constant @code{__v850e__} is defined if this option is used.
22707 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22708 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22709 are defined then a default target processor is chosen and the
22710 relevant @samp{__v850*__} preprocessor constant is defined.
22712 The preprocessor constants @code{__v850} and @code{__v851__} are always
22713 defined, regardless of which processor variant is the target.
22715 @item -mdisable-callt
22716 @itemx -mno-disable-callt
22717 @opindex mdisable-callt
22718 @opindex mno-disable-callt
22719 This option suppresses generation of the @code{CALLT} instruction for the
22720 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22723 This option is enabled by default when the RH850 ABI is
22724 in use (see @option{-mrh850-abi}), and disabled by default when the
22725 GCC ABI is in use. If @code{CALLT} instructions are being generated
22726 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
22732 Pass on (or do not pass on) the @option{-mrelax} command-line option
22736 @itemx -mno-long-jumps
22737 @opindex mlong-jumps
22738 @opindex mno-long-jumps
22739 Disable (or re-enable) the generation of PC-relative jump instructions.
22742 @itemx -mhard-float
22743 @opindex msoft-float
22744 @opindex mhard-float
22745 Disable (or re-enable) the generation of hardware floating point
22746 instructions. This option is only significant when the target
22747 architecture is @samp{V850E2V3} or higher. If hardware floating point
22748 instructions are being generated then the C preprocessor symbol
22749 @code{__FPU_OK__} is defined, otherwise the symbol
22750 @code{__NO_FPU__} is defined.
22754 Enables the use of the e3v5 LOOP instruction. The use of this
22755 instruction is not enabled by default when the e3v5 architecture is
22756 selected because its use is still experimental.
22760 @opindex mrh850-abi
22762 Enables support for the RH850 version of the V850 ABI. This is the
22763 default. With this version of the ABI the following rules apply:
22767 Integer sized structures and unions are returned via a memory pointer
22768 rather than a register.
22771 Large structures and unions (more than 8 bytes in size) are passed by
22775 Functions are aligned to 16-bit boundaries.
22778 The @option{-m8byte-align} command-line option is supported.
22781 The @option{-mdisable-callt} command-line option is enabled by
22782 default. The @option{-mno-disable-callt} command-line option is not
22786 When this version of the ABI is enabled the C preprocessor symbol
22787 @code{__V850_RH850_ABI__} is defined.
22791 Enables support for the old GCC version of the V850 ABI. With this
22792 version of the ABI the following rules apply:
22796 Integer sized structures and unions are returned in register @code{r10}.
22799 Large structures and unions (more than 8 bytes in size) are passed by
22803 Functions are aligned to 32-bit boundaries, unless optimizing for
22807 The @option{-m8byte-align} command-line option is not supported.
22810 The @option{-mdisable-callt} command-line option is supported but not
22811 enabled by default.
22814 When this version of the ABI is enabled the C preprocessor symbol
22815 @code{__V850_GCC_ABI__} is defined.
22817 @item -m8byte-align
22818 @itemx -mno-8byte-align
22819 @opindex m8byte-align
22820 @opindex mno-8byte-align
22821 Enables support for @code{double} and @code{long long} types to be
22822 aligned on 8-byte boundaries. The default is to restrict the
22823 alignment of all objects to at most 4-bytes. When
22824 @option{-m8byte-align} is in effect the C preprocessor symbol
22825 @code{__V850_8BYTE_ALIGN__} is defined.
22828 @opindex mbig-switch
22829 Generate code suitable for big switch tables. Use this option only if
22830 the assembler/linker complain about out of range branches within a switch
22835 This option causes r2 and r5 to be used in the code generated by
22836 the compiler. This setting is the default.
22838 @item -mno-app-regs
22839 @opindex mno-app-regs
22840 This option causes r2 and r5 to be treated as fixed registers.
22845 @subsection VAX Options
22846 @cindex VAX options
22848 These @samp{-m} options are defined for the VAX:
22853 Do not output certain jump instructions (@code{aobleq} and so on)
22854 that the Unix assembler for the VAX cannot handle across long
22859 Do output those jump instructions, on the assumption that the
22860 GNU assembler is being used.
22864 Output code for G-format floating-point numbers instead of D-format.
22867 @node Visium Options
22868 @subsection Visium Options
22869 @cindex Visium options
22875 A program which performs file I/O and is destined to run on an MCM target
22876 should be linked with this option. It causes the libraries libc.a and
22877 libdebug.a to be linked. The program should be run on the target under
22878 the control of the GDB remote debugging stub.
22882 A program which performs file I/O and is destined to run on the simulator
22883 should be linked with option. This causes libraries libc.a and libsim.a to
22887 @itemx -mhard-float
22889 @opindex mhard-float
22890 Generate code containing floating-point instructions. This is the
22894 @itemx -msoft-float
22896 @opindex msoft-float
22897 Generate code containing library calls for floating-point.
22899 @option{-msoft-float} changes the calling convention in the output file;
22900 therefore, it is only useful if you compile @emph{all} of a program with
22901 this option. In particular, you need to compile @file{libgcc.a}, the
22902 library that comes with GCC, with @option{-msoft-float} in order for
22905 @item -mcpu=@var{cpu_type}
22907 Set the instruction set, register set, and instruction scheduling parameters
22908 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22909 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22911 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22913 By default (unless configured otherwise), GCC generates code for the GR5
22914 variant of the Visium architecture.
22916 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22917 architecture. The only difference from GR5 code is that the compiler will
22918 generate block move instructions.
22920 @item -mtune=@var{cpu_type}
22922 Set the instruction scheduling parameters for machine type @var{cpu_type},
22923 but do not set the instruction set or register set that the option
22924 @option{-mcpu=@var{cpu_type}} would.
22928 Generate code for the supervisor mode, where there are no restrictions on
22929 the access to general registers. This is the default.
22932 @opindex muser-mode
22933 Generate code for the user mode, where the access to some general registers
22934 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22935 mode; on the GR6, only registers r29 to r31 are affected.
22939 @subsection VMS Options
22941 These @samp{-m} options are defined for the VMS implementations:
22944 @item -mvms-return-codes
22945 @opindex mvms-return-codes
22946 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22947 condition (e.g.@ error) codes.
22949 @item -mdebug-main=@var{prefix}
22950 @opindex mdebug-main=@var{prefix}
22951 Flag the first routine whose name starts with @var{prefix} as the main
22952 routine for the debugger.
22956 Default to 64-bit memory allocation routines.
22958 @item -mpointer-size=@var{size}
22959 @opindex mpointer-size=@var{size}
22960 Set the default size of pointers. Possible options for @var{size} are
22961 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22962 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22963 The later option disables @code{pragma pointer_size}.
22966 @node VxWorks Options
22967 @subsection VxWorks Options
22968 @cindex VxWorks Options
22970 The options in this section are defined for all VxWorks targets.
22971 Options specific to the target hardware are listed with the other
22972 options for that target.
22977 GCC can generate code for both VxWorks kernels and real time processes
22978 (RTPs). This option switches from the former to the latter. It also
22979 defines the preprocessor macro @code{__RTP__}.
22982 @opindex non-static
22983 Link an RTP executable against shared libraries rather than static
22984 libraries. The options @option{-static} and @option{-shared} can
22985 also be used for RTPs (@pxref{Link Options}); @option{-static}
22992 These options are passed down to the linker. They are defined for
22993 compatibility with Diab.
22996 @opindex Xbind-lazy
22997 Enable lazy binding of function calls. This option is equivalent to
22998 @option{-Wl,-z,now} and is defined for compatibility with Diab.
23002 Disable lazy binding of function calls. This option is the default and
23003 is defined for compatibility with Diab.
23007 @subsection x86 Options
23008 @cindex x86 Options
23010 These @samp{-m} options are defined for the x86 family of computers.
23014 @item -march=@var{cpu-type}
23016 Generate instructions for the machine type @var{cpu-type}. In contrast to
23017 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
23018 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
23019 to generate code that may not run at all on processors other than the one
23020 indicated. Specifying @option{-march=@var{cpu-type}} implies
23021 @option{-mtune=@var{cpu-type}}.
23023 The choices for @var{cpu-type} are:
23027 This selects the CPU to generate code for at compilation time by determining
23028 the processor type of the compiling machine. Using @option{-march=native}
23029 enables all instruction subsets supported by the local machine (hence
23030 the result might not run on different machines). Using @option{-mtune=native}
23031 produces code optimized for the local machine under the constraints
23032 of the selected instruction set.
23035 Original Intel i386 CPU@.
23038 Intel i486 CPU@. (No scheduling is implemented for this chip.)
23042 Intel Pentium CPU with no MMX support.
23045 Intel Lakemont MCU, based on Intel Pentium CPU.
23048 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
23051 Intel Pentium Pro CPU@.
23054 When used with @option{-march}, the Pentium Pro
23055 instruction set is used, so the code runs on all i686 family chips.
23056 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
23059 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
23064 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
23068 Intel Pentium M; low-power version of Intel Pentium III CPU
23069 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
23073 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
23076 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
23080 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
23081 SSE2 and SSE3 instruction set support.
23084 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
23085 instruction set support.
23088 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23089 SSE4.1, SSE4.2 and POPCNT instruction set support.
23092 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23093 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
23096 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23097 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
23100 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23101 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
23102 instruction set support.
23105 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23106 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23107 BMI, BMI2 and F16C instruction set support.
23110 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23111 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23112 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
23115 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23116 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23117 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
23118 XSAVES instruction set support.
23121 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
23122 instruction set support.
23125 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23126 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
23129 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
23130 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23131 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
23132 AVX512CD instruction set support.
23134 @item skylake-avx512
23135 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
23136 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23137 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
23138 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
23141 AMD K6 CPU with MMX instruction set support.
23145 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
23148 @itemx athlon-tbird
23149 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
23155 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
23156 instruction set support.
23162 Processors based on the AMD K8 core with x86-64 instruction set support,
23163 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
23164 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
23165 instruction set extensions.)
23168 @itemx opteron-sse3
23169 @itemx athlon64-sse3
23170 Improved versions of AMD K8 cores with SSE3 instruction set support.
23174 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
23175 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
23176 instruction set extensions.)
23179 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
23180 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
23181 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
23183 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23184 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
23185 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
23188 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23189 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
23190 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
23191 64-bit instruction set extensions.
23193 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23194 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
23195 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
23196 SSE4.2, ABM and 64-bit instruction set extensions.
23199 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
23200 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
23201 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
23202 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
23203 instruction set extensions.
23206 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
23207 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
23208 instruction set extensions.)
23211 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
23212 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
23213 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
23216 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
23220 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
23221 instruction set support.
23224 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
23225 implemented for this chip.)
23228 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
23230 implemented for this chip.)
23233 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
23236 @item -mtune=@var{cpu-type}
23238 Tune to @var{cpu-type} everything applicable about the generated code, except
23239 for the ABI and the set of available instructions.
23240 While picking a specific @var{cpu-type} schedules things appropriately
23241 for that particular chip, the compiler does not generate any code that
23242 cannot run on the default machine type unless you use a
23243 @option{-march=@var{cpu-type}} option.
23244 For example, if GCC is configured for i686-pc-linux-gnu
23245 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
23246 but still runs on i686 machines.
23248 The choices for @var{cpu-type} are the same as for @option{-march}.
23249 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
23253 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
23254 If you know the CPU on which your code will run, then you should use
23255 the corresponding @option{-mtune} or @option{-march} option instead of
23256 @option{-mtune=generic}. But, if you do not know exactly what CPU users
23257 of your application will have, then you should use this option.
23259 As new processors are deployed in the marketplace, the behavior of this
23260 option will change. Therefore, if you upgrade to a newer version of
23261 GCC, code generation controlled by this option will change to reflect
23263 that are most common at the time that version of GCC is released.
23265 There is no @option{-march=generic} option because @option{-march}
23266 indicates the instruction set the compiler can use, and there is no
23267 generic instruction set applicable to all processors. In contrast,
23268 @option{-mtune} indicates the processor (or, in this case, collection of
23269 processors) for which the code is optimized.
23272 Produce code optimized for the most current Intel processors, which are
23273 Haswell and Silvermont for this version of GCC. If you know the CPU
23274 on which your code will run, then you should use the corresponding
23275 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
23276 But, if you want your application performs better on both Haswell and
23277 Silvermont, then you should use this option.
23279 As new Intel processors are deployed in the marketplace, the behavior of
23280 this option will change. Therefore, if you upgrade to a newer version of
23281 GCC, code generation controlled by this option will change to reflect
23282 the most current Intel processors at the time that version of GCC is
23285 There is no @option{-march=intel} option because @option{-march} indicates
23286 the instruction set the compiler can use, and there is no common
23287 instruction set applicable to all processors. In contrast,
23288 @option{-mtune} indicates the processor (or, in this case, collection of
23289 processors) for which the code is optimized.
23292 @item -mcpu=@var{cpu-type}
23294 A deprecated synonym for @option{-mtune}.
23296 @item -mfpmath=@var{unit}
23298 Generate floating-point arithmetic for selected unit @var{unit}. The choices
23299 for @var{unit} are:
23303 Use the standard 387 floating-point coprocessor present on the majority of chips and
23304 emulated otherwise. Code compiled with this option runs almost everywhere.
23305 The temporary results are computed in 80-bit precision instead of the precision
23306 specified by the type, resulting in slightly different results compared to most
23307 of other chips. See @option{-ffloat-store} for more detailed description.
23309 This is the default choice for x86-32 targets.
23312 Use scalar floating-point instructions present in the SSE instruction set.
23313 This instruction set is supported by Pentium III and newer chips,
23314 and in the AMD line
23315 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
23316 instruction set supports only single-precision arithmetic, thus the double and
23317 extended-precision arithmetic are still done using 387. A later version, present
23318 only in Pentium 4 and AMD x86-64 chips, supports double-precision
23321 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
23322 or @option{-msse2} switches to enable SSE extensions and make this option
23323 effective. For the x86-64 compiler, these extensions are enabled by default.
23325 The resulting code should be considerably faster in the majority of cases and avoid
23326 the numerical instability problems of 387 code, but may break some existing
23327 code that expects temporaries to be 80 bits.
23329 This is the default choice for the x86-64 compiler.
23334 Attempt to utilize both instruction sets at once. This effectively doubles the
23335 amount of available registers, and on chips with separate execution units for
23336 387 and SSE the execution resources too. Use this option with care, as it is
23337 still experimental, because the GCC register allocator does not model separate
23338 functional units well, resulting in unstable performance.
23341 @item -masm=@var{dialect}
23342 @opindex masm=@var{dialect}
23343 Output assembly instructions using selected @var{dialect}. Also affects
23344 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
23345 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
23346 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
23347 not support @samp{intel}.
23350 @itemx -mno-ieee-fp
23352 @opindex mno-ieee-fp
23353 Control whether or not the compiler uses IEEE floating-point
23354 comparisons. These correctly handle the case where the result of a
23355 comparison is unordered.
23358 @opindex msoft-float
23359 Generate output containing library calls for floating point.
23361 @strong{Warning:} the requisite libraries are not part of GCC@.
23362 Normally the facilities of the machine's usual C compiler are used, but
23363 this can't be done directly in cross-compilation. You must make your
23364 own arrangements to provide suitable library functions for
23367 On machines where a function returns floating-point results in the 80387
23368 register stack, some floating-point opcodes may be emitted even if
23369 @option{-msoft-float} is used.
23371 @item -mno-fp-ret-in-387
23372 @opindex mno-fp-ret-in-387
23373 Do not use the FPU registers for return values of functions.
23375 The usual calling convention has functions return values of types
23376 @code{float} and @code{double} in an FPU register, even if there
23377 is no FPU@. The idea is that the operating system should emulate
23380 The option @option{-mno-fp-ret-in-387} causes such values to be returned
23381 in ordinary CPU registers instead.
23383 @item -mno-fancy-math-387
23384 @opindex mno-fancy-math-387
23385 Some 387 emulators do not support the @code{sin}, @code{cos} and
23386 @code{sqrt} instructions for the 387. Specify this option to avoid
23387 generating those instructions. This option is the default on
23388 OpenBSD and NetBSD@. This option is overridden when @option{-march}
23389 indicates that the target CPU always has an FPU and so the
23390 instruction does not need emulation. These
23391 instructions are not generated unless you also use the
23392 @option{-funsafe-math-optimizations} switch.
23394 @item -malign-double
23395 @itemx -mno-align-double
23396 @opindex malign-double
23397 @opindex mno-align-double
23398 Control whether GCC aligns @code{double}, @code{long double}, and
23399 @code{long long} variables on a two-word boundary or a one-word
23400 boundary. Aligning @code{double} variables on a two-word boundary
23401 produces code that runs somewhat faster on a Pentium at the
23402 expense of more memory.
23404 On x86-64, @option{-malign-double} is enabled by default.
23406 @strong{Warning:} if you use the @option{-malign-double} switch,
23407 structures containing the above types are aligned differently than
23408 the published application binary interface specifications for the x86-32
23409 and are not binary compatible with structures in code compiled
23410 without that switch.
23412 @item -m96bit-long-double
23413 @itemx -m128bit-long-double
23414 @opindex m96bit-long-double
23415 @opindex m128bit-long-double
23416 These switches control the size of @code{long double} type. The x86-32
23417 application binary interface specifies the size to be 96 bits,
23418 so @option{-m96bit-long-double} is the default in 32-bit mode.
23420 Modern architectures (Pentium and newer) prefer @code{long double}
23421 to be aligned to an 8- or 16-byte boundary. In arrays or structures
23422 conforming to the ABI, this is not possible. So specifying
23423 @option{-m128bit-long-double} aligns @code{long double}
23424 to a 16-byte boundary by padding the @code{long double} with an additional
23427 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
23428 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
23430 Notice that neither of these options enable any extra precision over the x87
23431 standard of 80 bits for a @code{long double}.
23433 @strong{Warning:} if you override the default value for your target ABI, this
23434 changes the size of
23435 structures and arrays containing @code{long double} variables,
23436 as well as modifying the function calling convention for functions taking
23437 @code{long double}. Hence they are not binary-compatible
23438 with code compiled without that switch.
23440 @item -mlong-double-64
23441 @itemx -mlong-double-80
23442 @itemx -mlong-double-128
23443 @opindex mlong-double-64
23444 @opindex mlong-double-80
23445 @opindex mlong-double-128
23446 These switches control the size of @code{long double} type. A size
23447 of 64 bits makes the @code{long double} type equivalent to the @code{double}
23448 type. This is the default for 32-bit Bionic C library. A size
23449 of 128 bits makes the @code{long double} type equivalent to the
23450 @code{__float128} type. This is the default for 64-bit Bionic C library.
23452 @strong{Warning:} if you override the default value for your target ABI, this
23453 changes the size of
23454 structures and arrays containing @code{long double} variables,
23455 as well as modifying the function calling convention for functions taking
23456 @code{long double}. Hence they are not binary-compatible
23457 with code compiled without that switch.
23459 @item -malign-data=@var{type}
23460 @opindex malign-data
23461 Control how GCC aligns variables. Supported values for @var{type} are
23462 @samp{compat} uses increased alignment value compatible uses GCC 4.8
23463 and earlier, @samp{abi} uses alignment value as specified by the
23464 psABI, and @samp{cacheline} uses increased alignment value to match
23465 the cache line size. @samp{compat} is the default.
23467 @item -mlarge-data-threshold=@var{threshold}
23468 @opindex mlarge-data-threshold
23469 When @option{-mcmodel=medium} is specified, data objects larger than
23470 @var{threshold} are placed in the large data section. This value must be the
23471 same across all objects linked into the binary, and defaults to 65535.
23475 Use a different function-calling convention, in which functions that
23476 take a fixed number of arguments return with the @code{ret @var{num}}
23477 instruction, which pops their arguments while returning. This saves one
23478 instruction in the caller since there is no need to pop the arguments
23481 You can specify that an individual function is called with this calling
23482 sequence with the function attribute @code{stdcall}. You can also
23483 override the @option{-mrtd} option by using the function attribute
23484 @code{cdecl}. @xref{Function Attributes}.
23486 @strong{Warning:} this calling convention is incompatible with the one
23487 normally used on Unix, so you cannot use it if you need to call
23488 libraries compiled with the Unix compiler.
23490 Also, you must provide function prototypes for all functions that
23491 take variable numbers of arguments (including @code{printf});
23492 otherwise incorrect code is generated for calls to those
23495 In addition, seriously incorrect code results if you call a
23496 function with too many arguments. (Normally, extra arguments are
23497 harmlessly ignored.)
23499 @item -mregparm=@var{num}
23501 Control how many registers are used to pass integer arguments. By
23502 default, no registers are used to pass arguments, and at most 3
23503 registers can be used. You can control this behavior for a specific
23504 function by using the function attribute @code{regparm}.
23505 @xref{Function Attributes}.
23507 @strong{Warning:} if you use this switch, and
23508 @var{num} is nonzero, then you must build all modules with the same
23509 value, including any libraries. This includes the system libraries and
23513 @opindex msseregparm
23514 Use SSE register passing conventions for float and double arguments
23515 and return values. You can control this behavior for a specific
23516 function by using the function attribute @code{sseregparm}.
23517 @xref{Function Attributes}.
23519 @strong{Warning:} if you use this switch then you must build all
23520 modules with the same value, including any libraries. This includes
23521 the system libraries and startup modules.
23523 @item -mvect8-ret-in-mem
23524 @opindex mvect8-ret-in-mem
23525 Return 8-byte vectors in memory instead of MMX registers. This is the
23526 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
23527 Studio compilers until version 12. Later compiler versions (starting
23528 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
23529 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
23530 you need to remain compatible with existing code produced by those
23531 previous compiler versions or older versions of GCC@.
23540 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
23541 is specified, the significands of results of floating-point operations are
23542 rounded to 24 bits (single precision); @option{-mpc64} rounds the
23543 significands of results of floating-point operations to 53 bits (double
23544 precision) and @option{-mpc80} rounds the significands of results of
23545 floating-point operations to 64 bits (extended double precision), which is
23546 the default. When this option is used, floating-point operations in higher
23547 precisions are not available to the programmer without setting the FPU
23548 control word explicitly.
23550 Setting the rounding of floating-point operations to less than the default
23551 80 bits can speed some programs by 2% or more. Note that some mathematical
23552 libraries assume that extended-precision (80-bit) floating-point operations
23553 are enabled by default; routines in such libraries could suffer significant
23554 loss of accuracy, typically through so-called ``catastrophic cancellation'',
23555 when this option is used to set the precision to less than extended precision.
23557 @item -mstackrealign
23558 @opindex mstackrealign
23559 Realign the stack at entry. On the x86, the @option{-mstackrealign}
23560 option generates an alternate prologue and epilogue that realigns the
23561 run-time stack if necessary. This supports mixing legacy codes that keep
23562 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
23563 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
23564 applicable to individual functions.
23566 @item -mpreferred-stack-boundary=@var{num}
23567 @opindex mpreferred-stack-boundary
23568 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23569 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23570 the default is 4 (16 bytes or 128 bits).
23572 @strong{Warning:} When generating code for the x86-64 architecture with
23573 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
23574 used to keep the stack boundary aligned to 8 byte boundary. Since
23575 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
23576 intended to be used in controlled environment where stack space is
23577 important limitation. This option leads to wrong code when functions
23578 compiled with 16 byte stack alignment (such as functions from a standard
23579 library) are called with misaligned stack. In this case, SSE
23580 instructions may lead to misaligned memory access traps. In addition,
23581 variable arguments are handled incorrectly for 16 byte aligned
23582 objects (including x87 long double and __int128), leading to wrong
23583 results. You must build all modules with
23584 @option{-mpreferred-stack-boundary=3}, including any libraries. This
23585 includes the system libraries and startup modules.
23587 @item -mincoming-stack-boundary=@var{num}
23588 @opindex mincoming-stack-boundary
23589 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
23590 boundary. If @option{-mincoming-stack-boundary} is not specified,
23591 the one specified by @option{-mpreferred-stack-boundary} is used.
23593 On Pentium and Pentium Pro, @code{double} and @code{long double} values
23594 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
23595 suffer significant run time performance penalties. On Pentium III, the
23596 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
23597 properly if it is not 16-byte aligned.
23599 To ensure proper alignment of this values on the stack, the stack boundary
23600 must be as aligned as that required by any value stored on the stack.
23601 Further, every function must be generated such that it keeps the stack
23602 aligned. Thus calling a function compiled with a higher preferred
23603 stack boundary from a function compiled with a lower preferred stack
23604 boundary most likely misaligns the stack. It is recommended that
23605 libraries that use callbacks always use the default setting.
23607 This extra alignment does consume extra stack space, and generally
23608 increases code size. Code that is sensitive to stack space usage, such
23609 as embedded systems and operating system kernels, may want to reduce the
23610 preferred alignment to @option{-mpreferred-stack-boundary=2}.
23667 @itemx -mavx512ifma
23668 @opindex mavx512ifma
23670 @itemx -mavx512vbmi
23671 @opindex mavx512vbmi
23683 @opindex mclfushopt
23700 @itemx -mprefetchwt1
23701 @opindex mprefetchwt1
23757 These switches enable the use of instructions in the MMX, SSE,
23758 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
23759 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
23760 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
23761 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
23762 extended instruction sets. Each has a corresponding @option{-mno-} option
23763 to disable use of these instructions.
23765 These extensions are also available as built-in functions: see
23766 @ref{x86 Built-in Functions}, for details of the functions enabled and
23767 disabled by these switches.
23769 To generate SSE/SSE2 instructions automatically from floating-point
23770 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
23772 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
23773 generates new AVX instructions or AVX equivalence for all SSEx instructions
23776 These options enable GCC to use these extended instructions in
23777 generated code, even without @option{-mfpmath=sse}. Applications that
23778 perform run-time CPU detection must compile separate files for each
23779 supported architecture, using the appropriate flags. In particular,
23780 the file containing the CPU detection code should be compiled without
23783 @item -mdump-tune-features
23784 @opindex mdump-tune-features
23785 This option instructs GCC to dump the names of the x86 performance
23786 tuning features and default settings. The names can be used in
23787 @option{-mtune-ctrl=@var{feature-list}}.
23789 @item -mtune-ctrl=@var{feature-list}
23790 @opindex mtune-ctrl=@var{feature-list}
23791 This option is used to do fine grain control of x86 code generation features.
23792 @var{feature-list} is a comma separated list of @var{feature} names. See also
23793 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
23794 on if it is not preceded with @samp{^}, otherwise, it is turned off.
23795 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
23796 developers. Using it may lead to code paths not covered by testing and can
23797 potentially result in compiler ICEs or runtime errors.
23800 @opindex mno-default
23801 This option instructs GCC to turn off all tunable features. See also
23802 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
23806 This option instructs GCC to emit a @code{cld} instruction in the prologue
23807 of functions that use string instructions. String instructions depend on
23808 the DF flag to select between autoincrement or autodecrement mode. While the
23809 ABI specifies the DF flag to be cleared on function entry, some operating
23810 systems violate this specification by not clearing the DF flag in their
23811 exception dispatchers. The exception handler can be invoked with the DF flag
23812 set, which leads to wrong direction mode when string instructions are used.
23813 This option can be enabled by default on 32-bit x86 targets by configuring
23814 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
23815 instructions can be suppressed with the @option{-mno-cld} compiler option
23819 @opindex mvzeroupper
23820 This option instructs GCC to emit a @code{vzeroupper} instruction
23821 before a transfer of control flow out of the function to minimize
23822 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
23825 @item -mprefer-avx128
23826 @opindex mprefer-avx128
23827 This option instructs GCC to use 128-bit AVX instructions instead of
23828 256-bit AVX instructions in the auto-vectorizer.
23832 This option enables GCC to generate @code{CMPXCHG16B} instructions.
23833 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
23834 (or oword) data types.
23835 This is useful for high-resolution counters that can be updated
23836 by multiple processors (or cores). This instruction is generated as part of
23837 atomic built-in functions: see @ref{__sync Builtins} or
23838 @ref{__atomic Builtins} for details.
23842 This option enables generation of @code{SAHF} instructions in 64-bit code.
23843 Early Intel Pentium 4 CPUs with Intel 64 support,
23844 prior to the introduction of Pentium 4 G1 step in December 2005,
23845 lacked the @code{LAHF} and @code{SAHF} instructions
23846 which are supported by AMD64.
23847 These are load and store instructions, respectively, for certain status flags.
23848 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
23849 @code{drem}, and @code{remainder} built-in functions;
23850 see @ref{Other Builtins} for details.
23854 This option enables use of the @code{movbe} instruction to implement
23855 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
23859 This option enables built-in functions @code{__builtin_ia32_crc32qi},
23860 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
23861 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
23865 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
23866 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
23867 with an additional Newton-Raphson step
23868 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
23869 (and their vectorized
23870 variants) for single-precision floating-point arguments. These instructions
23871 are generated only when @option{-funsafe-math-optimizations} is enabled
23872 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
23873 Note that while the throughput of the sequence is higher than the throughput
23874 of the non-reciprocal instruction, the precision of the sequence can be
23875 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
23877 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
23878 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
23879 combination), and doesn't need @option{-mrecip}.
23881 Also note that GCC emits the above sequence with additional Newton-Raphson step
23882 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
23883 already with @option{-ffast-math} (or the above option combination), and
23884 doesn't need @option{-mrecip}.
23886 @item -mrecip=@var{opt}
23887 @opindex mrecip=opt
23888 This option controls which reciprocal estimate instructions
23889 may be used. @var{opt} is a comma-separated list of options, which may
23890 be preceded by a @samp{!} to invert the option:
23894 Enable all estimate instructions.
23897 Enable the default instructions, equivalent to @option{-mrecip}.
23900 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23903 Enable the approximation for scalar division.
23906 Enable the approximation for vectorized division.
23909 Enable the approximation for scalar square root.
23912 Enable the approximation for vectorized square root.
23915 So, for example, @option{-mrecip=all,!sqrt} enables
23916 all of the reciprocal approximations, except for square root.
23918 @item -mveclibabi=@var{type}
23919 @opindex mveclibabi
23920 Specifies the ABI type to use for vectorizing intrinsics using an
23921 external library. Supported values for @var{type} are @samp{svml}
23922 for the Intel short
23923 vector math library and @samp{acml} for the AMD math core library.
23924 To use this option, both @option{-ftree-vectorize} and
23925 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
23926 ABI-compatible library must be specified at link time.
23928 GCC currently emits calls to @code{vmldExp2},
23929 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
23930 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
23931 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
23932 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
23933 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
23934 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
23935 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
23936 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
23937 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
23938 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
23939 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
23940 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
23941 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
23942 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
23943 when @option{-mveclibabi=acml} is used.
23945 @item -mabi=@var{name}
23947 Generate code for the specified calling convention. Permissible values
23948 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
23949 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
23950 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
23951 You can control this behavior for specific functions by
23952 using the function attributes @code{ms_abi} and @code{sysv_abi}.
23953 @xref{Function Attributes}.
23955 @item -mtls-dialect=@var{type}
23956 @opindex mtls-dialect
23957 Generate code to access thread-local storage using the @samp{gnu} or
23958 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
23959 @samp{gnu2} is more efficient, but it may add compile- and run-time
23960 requirements that cannot be satisfied on all systems.
23963 @itemx -mno-push-args
23964 @opindex mpush-args
23965 @opindex mno-push-args
23966 Use PUSH operations to store outgoing parameters. This method is shorter
23967 and usually equally fast as method using SUB/MOV operations and is enabled
23968 by default. In some cases disabling it may improve performance because of
23969 improved scheduling and reduced dependencies.
23971 @item -maccumulate-outgoing-args
23972 @opindex maccumulate-outgoing-args
23973 If enabled, the maximum amount of space required for outgoing arguments is
23974 computed in the function prologue. This is faster on most modern CPUs
23975 because of reduced dependencies, improved scheduling and reduced stack usage
23976 when the preferred stack boundary is not equal to 2. The drawback is a notable
23977 increase in code size. This switch implies @option{-mno-push-args}.
23981 Support thread-safe exception handling on MinGW. Programs that rely
23982 on thread-safe exception handling must compile and link all code with the
23983 @option{-mthreads} option. When compiling, @option{-mthreads} defines
23984 @option{-D_MT}; when linking, it links in a special thread helper library
23985 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
23987 @item -mms-bitfields
23988 @itemx -mno-ms-bitfields
23989 @opindex mms-bitfields
23990 @opindex mno-ms-bitfields
23992 Enable/disable bit-field layout compatible with the native Microsoft
23995 If @code{packed} is used on a structure, or if bit-fields are used,
23996 it may be that the Microsoft ABI lays out the structure differently
23997 than the way GCC normally does. Particularly when moving packed
23998 data between functions compiled with GCC and the native Microsoft compiler
23999 (either via function call or as data in a file), it may be necessary to access
24002 This option is enabled by default for Microsoft Windows
24003 targets. This behavior can also be controlled locally by use of variable
24004 or type attributes. For more information, see @ref{x86 Variable Attributes}
24005 and @ref{x86 Type Attributes}.
24007 The Microsoft structure layout algorithm is fairly simple with the exception
24008 of the bit-field packing.
24009 The padding and alignment of members of structures and whether a bit-field
24010 can straddle a storage-unit boundary are determine by these rules:
24013 @item Structure members are stored sequentially in the order in which they are
24014 declared: the first member has the lowest memory address and the last member
24017 @item Every data object has an alignment requirement. The alignment requirement
24018 for all data except structures, unions, and arrays is either the size of the
24019 object or the current packing size (specified with either the
24020 @code{aligned} attribute or the @code{pack} pragma),
24021 whichever is less. For structures, unions, and arrays,
24022 the alignment requirement is the largest alignment requirement of its members.
24023 Every object is allocated an offset so that:
24026 offset % alignment_requirement == 0
24029 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
24030 unit if the integral types are the same size and if the next bit-field fits
24031 into the current allocation unit without crossing the boundary imposed by the
24032 common alignment requirements of the bit-fields.
24035 MSVC interprets zero-length bit-fields in the following ways:
24038 @item If a zero-length bit-field is inserted between two bit-fields that
24039 are normally coalesced, the bit-fields are not coalesced.
24046 unsigned long bf_1 : 12;
24048 unsigned long bf_2 : 12;
24053 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
24054 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
24056 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
24057 alignment of the zero-length bit-field is greater than the member that follows it,
24058 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
24079 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
24080 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
24081 bit-field does not affect the alignment of @code{bar} or, as a result, the size
24084 Taking this into account, it is important to note the following:
24087 @item If a zero-length bit-field follows a normal bit-field, the type of the
24088 zero-length bit-field may affect the alignment of the structure as whole. For
24089 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
24090 normal bit-field, and is of type short.
24092 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
24093 still affect the alignment of the structure:
24104 Here, @code{t4} takes up 4 bytes.
24107 @item Zero-length bit-fields following non-bit-field members are ignored:
24119 Here, @code{t5} takes up 2 bytes.
24123 @item -mno-align-stringops
24124 @opindex mno-align-stringops
24125 Do not align the destination of inlined string operations. This switch reduces
24126 code size and improves performance in case the destination is already aligned,
24127 but GCC doesn't know about it.
24129 @item -minline-all-stringops
24130 @opindex minline-all-stringops
24131 By default GCC inlines string operations only when the destination is
24132 known to be aligned to least a 4-byte boundary.
24133 This enables more inlining and increases code
24134 size, but may improve performance of code that depends on fast
24135 @code{memcpy}, @code{strlen},
24136 and @code{memset} for short lengths.
24138 @item -minline-stringops-dynamically
24139 @opindex minline-stringops-dynamically
24140 For string operations of unknown size, use run-time checks with
24141 inline code for small blocks and a library call for large blocks.
24143 @item -mstringop-strategy=@var{alg}
24144 @opindex mstringop-strategy=@var{alg}
24145 Override the internal decision heuristic for the particular algorithm to use
24146 for inlining string operations. The allowed values for @var{alg} are:
24152 Expand using i386 @code{rep} prefix of the specified size.
24156 @itemx unrolled_loop
24157 Expand into an inline loop.
24160 Always use a library call.
24163 @item -mmemcpy-strategy=@var{strategy}
24164 @opindex mmemcpy-strategy=@var{strategy}
24165 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
24166 should be inlined and what inline algorithm to use when the expected size
24167 of the copy operation is known. @var{strategy}
24168 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
24169 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
24170 the max byte size with which inline algorithm @var{alg} is allowed. For the last
24171 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
24172 in the list must be specified in increasing order. The minimal byte size for
24173 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
24176 @item -mmemset-strategy=@var{strategy}
24177 @opindex mmemset-strategy=@var{strategy}
24178 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
24179 @code{__builtin_memset} expansion.
24181 @item -momit-leaf-frame-pointer
24182 @opindex momit-leaf-frame-pointer
24183 Don't keep the frame pointer in a register for leaf functions. This
24184 avoids the instructions to save, set up, and restore frame pointers and
24185 makes an extra register available in leaf functions. The option
24186 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
24187 which might make debugging harder.
24189 @item -mtls-direct-seg-refs
24190 @itemx -mno-tls-direct-seg-refs
24191 @opindex mtls-direct-seg-refs
24192 Controls whether TLS variables may be accessed with offsets from the
24193 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
24194 or whether the thread base pointer must be added. Whether or not this
24195 is valid depends on the operating system, and whether it maps the
24196 segment to cover the entire TLS area.
24198 For systems that use the GNU C Library, the default is on.
24201 @itemx -mno-sse2avx
24203 Specify that the assembler should encode SSE instructions with VEX
24204 prefix. The option @option{-mavx} turns this on by default.
24209 If profiling is active (@option{-pg}), put the profiling
24210 counter call before the prologue.
24211 Note: On x86 architectures the attribute @code{ms_hook_prologue}
24212 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
24214 @item -mrecord-mcount
24215 @itemx -mno-record-mcount
24216 @opindex mrecord-mcount
24217 If profiling is active (@option{-pg}), generate a __mcount_loc section
24218 that contains pointers to each profiling call. This is useful for
24219 automatically patching and out calls.
24222 @itemx -mno-nop-mcount
24223 @opindex mnop-mcount
24224 If profiling is active (@option{-pg}), generate the calls to
24225 the profiling functions as nops. This is useful when they
24226 should be patched in later dynamically. This is likely only
24227 useful together with @option{-mrecord-mcount}.
24229 @item -mskip-rax-setup
24230 @itemx -mno-skip-rax-setup
24231 @opindex mskip-rax-setup
24232 When generating code for the x86-64 architecture with SSE extensions
24233 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
24234 register when there are no variable arguments passed in vector registers.
24236 @strong{Warning:} Since RAX register is used to avoid unnecessarily
24237 saving vector registers on stack when passing variable arguments, the
24238 impacts of this option are callees may waste some stack space,
24239 misbehave or jump to a random location. GCC 4.4 or newer don't have
24240 those issues, regardless the RAX register value.
24243 @itemx -mno-8bit-idiv
24244 @opindex m8bit-idiv
24245 On some processors, like Intel Atom, 8-bit unsigned integer divide is
24246 much faster than 32-bit/64-bit integer divide. This option generates a
24247 run-time check. If both dividend and divisor are within range of 0
24248 to 255, 8-bit unsigned integer divide is used instead of
24249 32-bit/64-bit integer divide.
24251 @item -mavx256-split-unaligned-load
24252 @itemx -mavx256-split-unaligned-store
24253 @opindex mavx256-split-unaligned-load
24254 @opindex mavx256-split-unaligned-store
24255 Split 32-byte AVX unaligned load and store.
24257 @item -mstack-protector-guard=@var{guard}
24258 @opindex mstack-protector-guard=@var{guard}
24259 Generate stack protection code using canary at @var{guard}. Supported
24260 locations are @samp{global} for global canary or @samp{tls} for per-thread
24261 canary in the TLS block (the default). This option has effect only when
24262 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
24264 @item -mmitigate-rop
24265 @opindex mmitigate-rop
24266 Try to avoid generating code sequences that contain unintended return
24267 opcodes, to mitigate against certain forms of attack. At the moment,
24268 this option is limited in what it can do and should not be relied
24269 on to provide serious protection.
24273 These @samp{-m} switches are supported in addition to the above
24274 on x86-64 processors in 64-bit environments.
24287 Generate code for a 16-bit, 32-bit or 64-bit environment.
24288 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
24290 generates code that runs on any i386 system.
24292 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
24293 types to 64 bits, and generates code for the x86-64 architecture.
24294 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
24295 and @option{-mdynamic-no-pic} options.
24297 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
24299 generates code for the x86-64 architecture.
24301 The @option{-m16} option is the same as @option{-m32}, except for that
24302 it outputs the @code{.code16gcc} assembly directive at the beginning of
24303 the assembly output so that the binary can run in 16-bit mode.
24305 The @option{-miamcu} option generates code which conforms to Intel MCU
24306 psABI. It requires the @option{-m32} option to be turned on.
24308 @item -mno-red-zone
24309 @opindex mno-red-zone
24310 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
24311 by the x86-64 ABI; it is a 128-byte area beyond the location of the
24312 stack pointer that is not modified by signal or interrupt handlers
24313 and therefore can be used for temporary data without adjusting the stack
24314 pointer. The flag @option{-mno-red-zone} disables this red zone.
24316 @item -mcmodel=small
24317 @opindex mcmodel=small
24318 Generate code for the small code model: the program and its symbols must
24319 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
24320 Programs can be statically or dynamically linked. This is the default
24323 @item -mcmodel=kernel
24324 @opindex mcmodel=kernel
24325 Generate code for the kernel code model. The kernel runs in the
24326 negative 2 GB of the address space.
24327 This model has to be used for Linux kernel code.
24329 @item -mcmodel=medium
24330 @opindex mcmodel=medium
24331 Generate code for the medium model: the program is linked in the lower 2
24332 GB of the address space. Small symbols are also placed there. Symbols
24333 with sizes larger than @option{-mlarge-data-threshold} are put into
24334 large data or BSS sections and can be located above 2GB. Programs can
24335 be statically or dynamically linked.
24337 @item -mcmodel=large
24338 @opindex mcmodel=large
24339 Generate code for the large model. This model makes no assumptions
24340 about addresses and sizes of sections.
24342 @item -maddress-mode=long
24343 @opindex maddress-mode=long
24344 Generate code for long address mode. This is only supported for 64-bit
24345 and x32 environments. It is the default address mode for 64-bit
24348 @item -maddress-mode=short
24349 @opindex maddress-mode=short
24350 Generate code for short address mode. This is only supported for 32-bit
24351 and x32 environments. It is the default address mode for 32-bit and
24355 @node x86 Windows Options
24356 @subsection x86 Windows Options
24357 @cindex x86 Windows Options
24358 @cindex Windows Options for x86
24360 These additional options are available for Microsoft Windows targets:
24366 specifies that a console application is to be generated, by
24367 instructing the linker to set the PE header subsystem type
24368 required for console applications.
24369 This option is available for Cygwin and MinGW targets and is
24370 enabled by default on those targets.
24374 This option is available for Cygwin and MinGW targets. It
24375 specifies that a DLL---a dynamic link library---is to be
24376 generated, enabling the selection of the required runtime
24377 startup object and entry point.
24379 @item -mnop-fun-dllimport
24380 @opindex mnop-fun-dllimport
24381 This option is available for Cygwin and MinGW targets. It
24382 specifies that the @code{dllimport} attribute should be ignored.
24386 This option is available for MinGW targets. It specifies
24387 that MinGW-specific thread support is to be used.
24391 This option is available for MinGW-w64 targets. It causes
24392 the @code{UNICODE} preprocessor macro to be predefined, and
24393 chooses Unicode-capable runtime startup code.
24397 This option is available for Cygwin and MinGW targets. It
24398 specifies that the typical Microsoft Windows predefined macros are to
24399 be set in the pre-processor, but does not influence the choice
24400 of runtime library/startup code.
24404 This option is available for Cygwin and MinGW targets. It
24405 specifies that a GUI application is to be generated by
24406 instructing the linker to set the PE header subsystem type
24409 @item -fno-set-stack-executable
24410 @opindex fno-set-stack-executable
24411 This option is available for MinGW targets. It specifies that
24412 the executable flag for the stack used by nested functions isn't
24413 set. This is necessary for binaries running in kernel mode of
24414 Microsoft Windows, as there the User32 API, which is used to set executable
24415 privileges, isn't available.
24417 @item -fwritable-relocated-rdata
24418 @opindex fno-writable-relocated-rdata
24419 This option is available for MinGW and Cygwin targets. It specifies
24420 that relocated-data in read-only section is put into the @code{.data}
24421 section. This is a necessary for older runtimes not supporting
24422 modification of @code{.rdata} sections for pseudo-relocation.
24424 @item -mpe-aligned-commons
24425 @opindex mpe-aligned-commons
24426 This option is available for Cygwin and MinGW targets. It
24427 specifies that the GNU extension to the PE file format that
24428 permits the correct alignment of COMMON variables should be
24429 used when generating code. It is enabled by default if
24430 GCC detects that the target assembler found during configuration
24431 supports the feature.
24434 See also under @ref{x86 Options} for standard options.
24436 @node Xstormy16 Options
24437 @subsection Xstormy16 Options
24438 @cindex Xstormy16 Options
24440 These options are defined for Xstormy16:
24445 Choose startup files and linker script suitable for the simulator.
24448 @node Xtensa Options
24449 @subsection Xtensa Options
24450 @cindex Xtensa Options
24452 These options are supported for Xtensa targets:
24456 @itemx -mno-const16
24458 @opindex mno-const16
24459 Enable or disable use of @code{CONST16} instructions for loading
24460 constant values. The @code{CONST16} instruction is currently not a
24461 standard option from Tensilica. When enabled, @code{CONST16}
24462 instructions are always used in place of the standard @code{L32R}
24463 instructions. The use of @code{CONST16} is enabled by default only if
24464 the @code{L32R} instruction is not available.
24467 @itemx -mno-fused-madd
24468 @opindex mfused-madd
24469 @opindex mno-fused-madd
24470 Enable or disable use of fused multiply/add and multiply/subtract
24471 instructions in the floating-point option. This has no effect if the
24472 floating-point option is not also enabled. Disabling fused multiply/add
24473 and multiply/subtract instructions forces the compiler to use separate
24474 instructions for the multiply and add/subtract operations. This may be
24475 desirable in some cases where strict IEEE 754-compliant results are
24476 required: the fused multiply add/subtract instructions do not round the
24477 intermediate result, thereby producing results with @emph{more} bits of
24478 precision than specified by the IEEE standard. Disabling fused multiply
24479 add/subtract instructions also ensures that the program output is not
24480 sensitive to the compiler's ability to combine multiply and add/subtract
24483 @item -mserialize-volatile
24484 @itemx -mno-serialize-volatile
24485 @opindex mserialize-volatile
24486 @opindex mno-serialize-volatile
24487 When this option is enabled, GCC inserts @code{MEMW} instructions before
24488 @code{volatile} memory references to guarantee sequential consistency.
24489 The default is @option{-mserialize-volatile}. Use
24490 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
24492 @item -mforce-no-pic
24493 @opindex mforce-no-pic
24494 For targets, like GNU/Linux, where all user-mode Xtensa code must be
24495 position-independent code (PIC), this option disables PIC for compiling
24498 @item -mtext-section-literals
24499 @itemx -mno-text-section-literals
24500 @opindex mtext-section-literals
24501 @opindex mno-text-section-literals
24502 These options control the treatment of literal pools. The default is
24503 @option{-mno-text-section-literals}, which places literals in a separate
24504 section in the output file. This allows the literal pool to be placed
24505 in a data RAM/ROM, and it also allows the linker to combine literal
24506 pools from separate object files to remove redundant literals and
24507 improve code size. With @option{-mtext-section-literals}, the literals
24508 are interspersed in the text section in order to keep them as close as
24509 possible to their references. This may be necessary for large assembly
24510 files. Literals for each function are placed right before that function.
24512 @item -mauto-litpools
24513 @itemx -mno-auto-litpools
24514 @opindex mauto-litpools
24515 @opindex mno-auto-litpools
24516 These options control the treatment of literal pools. The default is
24517 @option{-mno-auto-litpools}, which places literals in a separate
24518 section in the output file unless @option{-mtext-section-literals} is
24519 used. With @option{-mauto-litpools} the literals are interspersed in
24520 the text section by the assembler. Compiler does not produce explicit
24521 @code{.literal} directives and loads literals into registers with
24522 @code{MOVI} instructions instead of @code{L32R} to let the assembler
24523 do relaxation and place literals as necessary. This option allows
24524 assembler to create several literal pools per function and assemble
24525 very big functions, which may not be possible with
24526 @option{-mtext-section-literals}.
24528 @item -mtarget-align
24529 @itemx -mno-target-align
24530 @opindex mtarget-align
24531 @opindex mno-target-align
24532 When this option is enabled, GCC instructs the assembler to
24533 automatically align instructions to reduce branch penalties at the
24534 expense of some code density. The assembler attempts to widen density
24535 instructions to align branch targets and the instructions following call
24536 instructions. If there are not enough preceding safe density
24537 instructions to align a target, no widening is performed. The
24538 default is @option{-mtarget-align}. These options do not affect the
24539 treatment of auto-aligned instructions like @code{LOOP}, which the
24540 assembler always aligns, either by widening density instructions or
24541 by inserting NOP instructions.
24544 @itemx -mno-longcalls
24545 @opindex mlongcalls
24546 @opindex mno-longcalls
24547 When this option is enabled, GCC instructs the assembler to translate
24548 direct calls to indirect calls unless it can determine that the target
24549 of a direct call is in the range allowed by the call instruction. This
24550 translation typically occurs for calls to functions in other source
24551 files. Specifically, the assembler translates a direct @code{CALL}
24552 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
24553 The default is @option{-mno-longcalls}. This option should be used in
24554 programs where the call target can potentially be out of range. This
24555 option is implemented in the assembler, not the compiler, so the
24556 assembly code generated by GCC still shows direct call
24557 instructions---look at the disassembled object code to see the actual
24558 instructions. Note that the assembler uses an indirect call for
24559 every cross-file call, not just those that really are out of range.
24562 @node zSeries Options
24563 @subsection zSeries Options
24564 @cindex zSeries options
24566 These are listed under @xref{S/390 and zSeries Options}.
24572 @section Specifying Subprocesses and the Switches to Pass to Them
24575 @command{gcc} is a driver program. It performs its job by invoking a
24576 sequence of other programs to do the work of compiling, assembling and
24577 linking. GCC interprets its command-line parameters and uses these to
24578 deduce which programs it should invoke, and which command-line options
24579 it ought to place on their command lines. This behavior is controlled
24580 by @dfn{spec strings}. In most cases there is one spec string for each
24581 program that GCC can invoke, but a few programs have multiple spec
24582 strings to control their behavior. The spec strings built into GCC can
24583 be overridden by using the @option{-specs=} command-line switch to specify
24586 @dfn{Spec files} are plain-text files that are used to construct spec
24587 strings. They consist of a sequence of directives separated by blank
24588 lines. The type of directive is determined by the first non-whitespace
24589 character on the line, which can be one of the following:
24592 @item %@var{command}
24593 Issues a @var{command} to the spec file processor. The commands that can
24597 @item %include <@var{file}>
24598 @cindex @code{%include}
24599 Search for @var{file} and insert its text at the current point in the
24602 @item %include_noerr <@var{file}>
24603 @cindex @code{%include_noerr}
24604 Just like @samp{%include}, but do not generate an error message if the include
24605 file cannot be found.
24607 @item %rename @var{old_name} @var{new_name}
24608 @cindex @code{%rename}
24609 Rename the spec string @var{old_name} to @var{new_name}.
24613 @item *[@var{spec_name}]:
24614 This tells the compiler to create, override or delete the named spec
24615 string. All lines after this directive up to the next directive or
24616 blank line are considered to be the text for the spec string. If this
24617 results in an empty string then the spec is deleted. (Or, if the
24618 spec did not exist, then nothing happens.) Otherwise, if the spec
24619 does not currently exist a new spec is created. If the spec does
24620 exist then its contents are overridden by the text of this
24621 directive, unless the first character of that text is the @samp{+}
24622 character, in which case the text is appended to the spec.
24624 @item [@var{suffix}]:
24625 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
24626 and up to the next directive or blank line are considered to make up the
24627 spec string for the indicated suffix. When the compiler encounters an
24628 input file with the named suffix, it processes the spec string in
24629 order to work out how to compile that file. For example:
24633 z-compile -input %i
24636 This says that any input file whose name ends in @samp{.ZZ} should be
24637 passed to the program @samp{z-compile}, which should be invoked with the
24638 command-line switch @option{-input} and with the result of performing the
24639 @samp{%i} substitution. (See below.)
24641 As an alternative to providing a spec string, the text following a
24642 suffix directive can be one of the following:
24645 @item @@@var{language}
24646 This says that the suffix is an alias for a known @var{language}. This is
24647 similar to using the @option{-x} command-line switch to GCC to specify a
24648 language explicitly. For example:
24655 Says that .ZZ files are, in fact, C++ source files.
24658 This causes an error messages saying:
24661 @var{name} compiler not installed on this system.
24665 GCC already has an extensive list of suffixes built into it.
24666 This directive adds an entry to the end of the list of suffixes, but
24667 since the list is searched from the end backwards, it is effectively
24668 possible to override earlier entries using this technique.
24672 GCC has the following spec strings built into it. Spec files can
24673 override these strings or create their own. Note that individual
24674 targets can also add their own spec strings to this list.
24677 asm Options to pass to the assembler
24678 asm_final Options to pass to the assembler post-processor
24679 cpp Options to pass to the C preprocessor
24680 cc1 Options to pass to the C compiler
24681 cc1plus Options to pass to the C++ compiler
24682 endfile Object files to include at the end of the link
24683 link Options to pass to the linker
24684 lib Libraries to include on the command line to the linker
24685 libgcc Decides which GCC support library to pass to the linker
24686 linker Sets the name of the linker
24687 predefines Defines to be passed to the C preprocessor
24688 signed_char Defines to pass to CPP to say whether @code{char} is signed
24690 startfile Object files to include at the start of the link
24693 Here is a small example of a spec file:
24696 %rename lib old_lib
24699 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
24702 This example renames the spec called @samp{lib} to @samp{old_lib} and
24703 then overrides the previous definition of @samp{lib} with a new one.
24704 The new definition adds in some extra command-line options before
24705 including the text of the old definition.
24707 @dfn{Spec strings} are a list of command-line options to be passed to their
24708 corresponding program. In addition, the spec strings can contain
24709 @samp{%}-prefixed sequences to substitute variable text or to
24710 conditionally insert text into the command line. Using these constructs
24711 it is possible to generate quite complex command lines.
24713 Here is a table of all defined @samp{%}-sequences for spec
24714 strings. Note that spaces are not generated automatically around the
24715 results of expanding these sequences. Therefore you can concatenate them
24716 together or combine them with constant text in a single argument.
24720 Substitute one @samp{%} into the program name or argument.
24723 Substitute the name of the input file being processed.
24726 Substitute the basename of the input file being processed.
24727 This is the substring up to (and not including) the last period
24728 and not including the directory.
24731 This is the same as @samp{%b}, but include the file suffix (text after
24735 Marks the argument containing or following the @samp{%d} as a
24736 temporary file name, so that that file is deleted if GCC exits
24737 successfully. Unlike @samp{%g}, this contributes no text to the
24740 @item %g@var{suffix}
24741 Substitute a file name that has suffix @var{suffix} and is chosen
24742 once per compilation, and mark the argument in the same way as
24743 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
24744 name is now chosen in a way that is hard to predict even when previously
24745 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
24746 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
24747 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
24748 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
24749 was simply substituted with a file name chosen once per compilation,
24750 without regard to any appended suffix (which was therefore treated
24751 just like ordinary text), making such attacks more likely to succeed.
24753 @item %u@var{suffix}
24754 Like @samp{%g}, but generates a new temporary file name
24755 each time it appears instead of once per compilation.
24757 @item %U@var{suffix}
24758 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
24759 new one if there is no such last file name. In the absence of any
24760 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
24761 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
24762 involves the generation of two distinct file names, one
24763 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
24764 simply substituted with a file name chosen for the previous @samp{%u},
24765 without regard to any appended suffix.
24767 @item %j@var{suffix}
24768 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
24769 writable, and if @option{-save-temps} is not used;
24770 otherwise, substitute the name
24771 of a temporary file, just like @samp{%u}. This temporary file is not
24772 meant for communication between processes, but rather as a junk
24773 disposal mechanism.
24775 @item %|@var{suffix}
24776 @itemx %m@var{suffix}
24777 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
24778 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
24779 all. These are the two most common ways to instruct a program that it
24780 should read from standard input or write to standard output. If you
24781 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
24782 construct: see for example @file{f/lang-specs.h}.
24784 @item %.@var{SUFFIX}
24785 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
24786 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
24787 terminated by the next space or %.
24790 Marks the argument containing or following the @samp{%w} as the
24791 designated output file of this compilation. This puts the argument
24792 into the sequence of arguments that @samp{%o} substitutes.
24795 Substitutes the names of all the output files, with spaces
24796 automatically placed around them. You should write spaces
24797 around the @samp{%o} as well or the results are undefined.
24798 @samp{%o} is for use in the specs for running the linker.
24799 Input files whose names have no recognized suffix are not compiled
24800 at all, but they are included among the output files, so they are
24804 Substitutes the suffix for object files. Note that this is
24805 handled specially when it immediately follows @samp{%g, %u, or %U},
24806 because of the need for those to form complete file names. The
24807 handling is such that @samp{%O} is treated exactly as if it had already
24808 been substituted, except that @samp{%g, %u, and %U} do not currently
24809 support additional @var{suffix} characters following @samp{%O} as they do
24810 following, for example, @samp{.o}.
24813 Substitutes the standard macro predefinitions for the
24814 current target machine. Use this when running @command{cpp}.
24817 Like @samp{%p}, but puts @samp{__} before and after the name of each
24818 predefined macro, except for macros that start with @samp{__} or with
24819 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
24823 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
24824 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
24825 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
24826 and @option{-imultilib} as necessary.
24829 Current argument is the name of a library or startup file of some sort.
24830 Search for that file in a standard list of directories and substitute
24831 the full name found. The current working directory is included in the
24832 list of directories scanned.
24835 Current argument is the name of a linker script. Search for that file
24836 in the current list of directories to scan for libraries. If the file
24837 is located insert a @option{--script} option into the command line
24838 followed by the full path name found. If the file is not found then
24839 generate an error message. Note: the current working directory is not
24843 Print @var{str} as an error message. @var{str} is terminated by a newline.
24844 Use this when inconsistent options are detected.
24846 @item %(@var{name})
24847 Substitute the contents of spec string @var{name} at this point.
24849 @item %x@{@var{option}@}
24850 Accumulate an option for @samp{%X}.
24853 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
24857 Output the accumulated assembler options specified by @option{-Wa}.
24860 Output the accumulated preprocessor options specified by @option{-Wp}.
24863 Process the @code{asm} spec. This is used to compute the
24864 switches to be passed to the assembler.
24867 Process the @code{asm_final} spec. This is a spec string for
24868 passing switches to an assembler post-processor, if such a program is
24872 Process the @code{link} spec. This is the spec for computing the
24873 command line passed to the linker. Typically it makes use of the
24874 @samp{%L %G %S %D and %E} sequences.
24877 Dump out a @option{-L} option for each directory that GCC believes might
24878 contain startup files. If the target supports multilibs then the
24879 current multilib directory is prepended to each of these paths.
24882 Process the @code{lib} spec. This is a spec string for deciding which
24883 libraries are included on the command line to the linker.
24886 Process the @code{libgcc} spec. This is a spec string for deciding
24887 which GCC support library is included on the command line to the linker.
24890 Process the @code{startfile} spec. This is a spec for deciding which
24891 object files are the first ones passed to the linker. Typically
24892 this might be a file named @file{crt0.o}.
24895 Process the @code{endfile} spec. This is a spec string that specifies
24896 the last object files that are passed to the linker.
24899 Process the @code{cpp} spec. This is used to construct the arguments
24900 to be passed to the C preprocessor.
24903 Process the @code{cc1} spec. This is used to construct the options to be
24904 passed to the actual C compiler (@command{cc1}).
24907 Process the @code{cc1plus} spec. This is used to construct the options to be
24908 passed to the actual C++ compiler (@command{cc1plus}).
24911 Substitute the variable part of a matched option. See below.
24912 Note that each comma in the substituted string is replaced by
24916 Remove all occurrences of @code{-S} from the command line. Note---this
24917 command is position dependent. @samp{%} commands in the spec string
24918 before this one see @code{-S}, @samp{%} commands in the spec string
24919 after this one do not.
24921 @item %:@var{function}(@var{args})
24922 Call the named function @var{function}, passing it @var{args}.
24923 @var{args} is first processed as a nested spec string, then split
24924 into an argument vector in the usual fashion. The function returns
24925 a string which is processed as if it had appeared literally as part
24926 of the current spec.
24928 The following built-in spec functions are provided:
24931 @item @code{getenv}
24932 The @code{getenv} spec function takes two arguments: an environment
24933 variable name and a string. If the environment variable is not
24934 defined, a fatal error is issued. Otherwise, the return value is the
24935 value of the environment variable concatenated with the string. For
24936 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
24939 %:getenv(TOPDIR /include)
24942 expands to @file{/path/to/top/include}.
24944 @item @code{if-exists}
24945 The @code{if-exists} spec function takes one argument, an absolute
24946 pathname to a file. If the file exists, @code{if-exists} returns the
24947 pathname. Here is a small example of its usage:
24951 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
24954 @item @code{if-exists-else}
24955 The @code{if-exists-else} spec function is similar to the @code{if-exists}
24956 spec function, except that it takes two arguments. The first argument is
24957 an absolute pathname to a file. If the file exists, @code{if-exists-else}
24958 returns the pathname. If it does not exist, it returns the second argument.
24959 This way, @code{if-exists-else} can be used to select one file or another,
24960 based on the existence of the first. Here is a small example of its usage:
24964 crt0%O%s %:if-exists(crti%O%s) \
24965 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
24968 @item @code{replace-outfile}
24969 The @code{replace-outfile} spec function takes two arguments. It looks for the
24970 first argument in the outfiles array and replaces it with the second argument. Here
24971 is a small example of its usage:
24974 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
24977 @item @code{remove-outfile}
24978 The @code{remove-outfile} spec function takes one argument. It looks for the
24979 first argument in the outfiles array and removes it. Here is a small example
24983 %:remove-outfile(-lm)
24986 @item @code{pass-through-libs}
24987 The @code{pass-through-libs} spec function takes any number of arguments. It
24988 finds any @option{-l} options and any non-options ending in @file{.a} (which it
24989 assumes are the names of linker input library archive files) and returns a
24990 result containing all the found arguments each prepended by
24991 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
24992 intended to be passed to the LTO linker plugin.
24995 %:pass-through-libs(%G %L %G)
24998 @item @code{print-asm-header}
24999 The @code{print-asm-header} function takes no arguments and simply
25000 prints a banner like:
25006 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
25009 It is used to separate compiler options from assembler options
25010 in the @option{--target-help} output.
25013 @item %@{@code{S}@}
25014 Substitutes the @code{-S} switch, if that switch is given to GCC@.
25015 If that switch is not specified, this substitutes nothing. Note that
25016 the leading dash is omitted when specifying this option, and it is
25017 automatically inserted if the substitution is performed. Thus the spec
25018 string @samp{%@{foo@}} matches the command-line option @option{-foo}
25019 and outputs the command-line option @option{-foo}.
25021 @item %W@{@code{S}@}
25022 Like %@{@code{S}@} but mark last argument supplied within as a file to be
25023 deleted on failure.
25025 @item %@{@code{S}*@}
25026 Substitutes all the switches specified to GCC whose names start
25027 with @code{-S}, but which also take an argument. This is used for
25028 switches like @option{-o}, @option{-D}, @option{-I}, etc.
25029 GCC considers @option{-o foo} as being
25030 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
25031 text, including the space. Thus two arguments are generated.
25033 @item %@{@code{S}*&@code{T}*@}
25034 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
25035 (the order of @code{S} and @code{T} in the spec is not significant).
25036 There can be any number of ampersand-separated variables; for each the
25037 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
25039 @item %@{@code{S}:@code{X}@}
25040 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
25042 @item %@{!@code{S}:@code{X}@}
25043 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
25045 @item %@{@code{S}*:@code{X}@}
25046 Substitutes @code{X} if one or more switches whose names start with
25047 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
25048 once, no matter how many such switches appeared. However, if @code{%*}
25049 appears somewhere in @code{X}, then @code{X} is substituted once
25050 for each matching switch, with the @code{%*} replaced by the part of
25051 that switch matching the @code{*}.
25053 If @code{%*} appears as the last part of a spec sequence then a space
25054 is added after the end of the last substitution. If there is more
25055 text in the sequence, however, then a space is not generated. This
25056 allows the @code{%*} substitution to be used as part of a larger
25057 string. For example, a spec string like this:
25060 %@{mcu=*:--script=%*/memory.ld@}
25064 when matching an option like @option{-mcu=newchip} produces:
25067 --script=newchip/memory.ld
25070 @item %@{.@code{S}:@code{X}@}
25071 Substitutes @code{X}, if processing a file with suffix @code{S}.
25073 @item %@{!.@code{S}:@code{X}@}
25074 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
25076 @item %@{,@code{S}:@code{X}@}
25077 Substitutes @code{X}, if processing a file for language @code{S}.
25079 @item %@{!,@code{S}:@code{X}@}
25080 Substitutes @code{X}, if not processing a file for language @code{S}.
25082 @item %@{@code{S}|@code{P}:@code{X}@}
25083 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
25084 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
25085 @code{*} sequences as well, although they have a stronger binding than
25086 the @samp{|}. If @code{%*} appears in @code{X}, all of the
25087 alternatives must be starred, and only the first matching alternative
25090 For example, a spec string like this:
25093 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
25097 outputs the following command-line options from the following input
25098 command-line options:
25103 -d fred.c -foo -baz -boggle
25104 -d jim.d -bar -baz -boggle
25107 @item %@{S:X; T:Y; :D@}
25109 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
25110 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
25111 be as many clauses as you need. This may be combined with @code{.},
25112 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
25117 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
25118 construct may contain other nested @samp{%} constructs or spaces, or
25119 even newlines. They are processed as usual, as described above.
25120 Trailing white space in @code{X} is ignored. White space may also
25121 appear anywhere on the left side of the colon in these constructs,
25122 except between @code{.} or @code{*} and the corresponding word.
25124 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
25125 handled specifically in these constructs. If another value of
25126 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
25127 @option{-W} switch is found later in the command line, the earlier
25128 switch value is ignored, except with @{@code{S}*@} where @code{S} is
25129 just one letter, which passes all matching options.
25131 The character @samp{|} at the beginning of the predicate text is used to
25132 indicate that a command should be piped to the following command, but
25133 only if @option{-pipe} is specified.
25135 It is built into GCC which switches take arguments and which do not.
25136 (You might think it would be useful to generalize this to allow each
25137 compiler's spec to say which switches take arguments. But this cannot
25138 be done in a consistent fashion. GCC cannot even decide which input
25139 files have been specified without knowing which switches take arguments,
25140 and it must know which input files to compile in order to tell which
25143 GCC also knows implicitly that arguments starting in @option{-l} are to be
25144 treated as compiler output files, and passed to the linker in their
25145 proper position among the other output files.
25147 @node Environment Variables
25148 @section Environment Variables Affecting GCC
25149 @cindex environment variables
25151 @c man begin ENVIRONMENT
25152 This section describes several environment variables that affect how GCC
25153 operates. Some of them work by specifying directories or prefixes to use
25154 when searching for various kinds of files. Some are used to specify other
25155 aspects of the compilation environment.
25157 Note that you can also specify places to search using options such as
25158 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
25159 take precedence over places specified using environment variables, which
25160 in turn take precedence over those specified by the configuration of GCC@.
25161 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
25162 GNU Compiler Collection (GCC) Internals}.
25167 @c @itemx LC_COLLATE
25169 @c @itemx LC_MONETARY
25170 @c @itemx LC_NUMERIC
25175 @c @findex LC_COLLATE
25176 @findex LC_MESSAGES
25177 @c @findex LC_MONETARY
25178 @c @findex LC_NUMERIC
25182 These environment variables control the way that GCC uses
25183 localization information which allows GCC to work with different
25184 national conventions. GCC inspects the locale categories
25185 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
25186 so. These locale categories can be set to any value supported by your
25187 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
25188 Kingdom encoded in UTF-8.
25190 The @env{LC_CTYPE} environment variable specifies character
25191 classification. GCC uses it to determine the character boundaries in
25192 a string; this is needed for some multibyte encodings that contain quote
25193 and escape characters that are otherwise interpreted as a string
25196 The @env{LC_MESSAGES} environment variable specifies the language to
25197 use in diagnostic messages.
25199 If the @env{LC_ALL} environment variable is set, it overrides the value
25200 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
25201 and @env{LC_MESSAGES} default to the value of the @env{LANG}
25202 environment variable. If none of these variables are set, GCC
25203 defaults to traditional C English behavior.
25207 If @env{TMPDIR} is set, it specifies the directory to use for temporary
25208 files. GCC uses temporary files to hold the output of one stage of
25209 compilation which is to be used as input to the next stage: for example,
25210 the output of the preprocessor, which is the input to the compiler
25213 @item GCC_COMPARE_DEBUG
25214 @findex GCC_COMPARE_DEBUG
25215 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
25216 @option{-fcompare-debug} to the compiler driver. See the documentation
25217 of this option for more details.
25219 @item GCC_EXEC_PREFIX
25220 @findex GCC_EXEC_PREFIX
25221 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
25222 names of the subprograms executed by the compiler. No slash is added
25223 when this prefix is combined with the name of a subprogram, but you can
25224 specify a prefix that ends with a slash if you wish.
25226 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
25227 an appropriate prefix to use based on the pathname it is invoked with.
25229 If GCC cannot find the subprogram using the specified prefix, it
25230 tries looking in the usual places for the subprogram.
25232 The default value of @env{GCC_EXEC_PREFIX} is
25233 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
25234 the installed compiler. In many cases @var{prefix} is the value
25235 of @code{prefix} when you ran the @file{configure} script.
25237 Other prefixes specified with @option{-B} take precedence over this prefix.
25239 This prefix is also used for finding files such as @file{crt0.o} that are
25242 In addition, the prefix is used in an unusual way in finding the
25243 directories to search for header files. For each of the standard
25244 directories whose name normally begins with @samp{/usr/local/lib/gcc}
25245 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
25246 replacing that beginning with the specified prefix to produce an
25247 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
25248 @file{foo/bar} just before it searches the standard directory
25249 @file{/usr/local/lib/bar}.
25250 If a standard directory begins with the configured
25251 @var{prefix} then the value of @var{prefix} is replaced by
25252 @env{GCC_EXEC_PREFIX} when looking for header files.
25254 @item COMPILER_PATH
25255 @findex COMPILER_PATH
25256 The value of @env{COMPILER_PATH} is a colon-separated list of
25257 directories, much like @env{PATH}. GCC tries the directories thus
25258 specified when searching for subprograms, if it can't find the
25259 subprograms using @env{GCC_EXEC_PREFIX}.
25262 @findex LIBRARY_PATH
25263 The value of @env{LIBRARY_PATH} is a colon-separated list of
25264 directories, much like @env{PATH}. When configured as a native compiler,
25265 GCC tries the directories thus specified when searching for special
25266 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
25267 using GCC also uses these directories when searching for ordinary
25268 libraries for the @option{-l} option (but directories specified with
25269 @option{-L} come first).
25273 @cindex locale definition
25274 This variable is used to pass locale information to the compiler. One way in
25275 which this information is used is to determine the character set to be used
25276 when character literals, string literals and comments are parsed in C and C++.
25277 When the compiler is configured to allow multibyte characters,
25278 the following values for @env{LANG} are recognized:
25282 Recognize JIS characters.
25284 Recognize SJIS characters.
25286 Recognize EUCJP characters.
25289 If @env{LANG} is not defined, or if it has some other value, then the
25290 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
25291 recognize and translate multibyte characters.
25295 Some additional environment variables affect the behavior of the
25298 @include cppenv.texi
25302 @node Precompiled Headers
25303 @section Using Precompiled Headers
25304 @cindex precompiled headers
25305 @cindex speed of compilation
25307 Often large projects have many header files that are included in every
25308 source file. The time the compiler takes to process these header files
25309 over and over again can account for nearly all of the time required to
25310 build the project. To make builds faster, GCC allows you to
25311 @dfn{precompile} a header file.
25313 To create a precompiled header file, simply compile it as you would any
25314 other file, if necessary using the @option{-x} option to make the driver
25315 treat it as a C or C++ header file. You may want to use a
25316 tool like @command{make} to keep the precompiled header up-to-date when
25317 the headers it contains change.
25319 A precompiled header file is searched for when @code{#include} is
25320 seen in the compilation. As it searches for the included file
25321 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
25322 compiler looks for a precompiled header in each directory just before it
25323 looks for the include file in that directory. The name searched for is
25324 the name specified in the @code{#include} with @samp{.gch} appended. If
25325 the precompiled header file can't be used, it is ignored.
25327 For instance, if you have @code{#include "all.h"}, and you have
25328 @file{all.h.gch} in the same directory as @file{all.h}, then the
25329 precompiled header file is used if possible, and the original
25330 header is used otherwise.
25332 Alternatively, you might decide to put the precompiled header file in a
25333 directory and use @option{-I} to ensure that directory is searched
25334 before (or instead of) the directory containing the original header.
25335 Then, if you want to check that the precompiled header file is always
25336 used, you can put a file of the same name as the original header in this
25337 directory containing an @code{#error} command.
25339 This also works with @option{-include}. So yet another way to use
25340 precompiled headers, good for projects not designed with precompiled
25341 header files in mind, is to simply take most of the header files used by
25342 a project, include them from another header file, precompile that header
25343 file, and @option{-include} the precompiled header. If the header files
25344 have guards against multiple inclusion, they are skipped because
25345 they've already been included (in the precompiled header).
25347 If you need to precompile the same header file for different
25348 languages, targets, or compiler options, you can instead make a
25349 @emph{directory} named like @file{all.h.gch}, and put each precompiled
25350 header in the directory, perhaps using @option{-o}. It doesn't matter
25351 what you call the files in the directory; every precompiled header in
25352 the directory is considered. The first precompiled header
25353 encountered in the directory that is valid for this compilation is
25354 used; they're searched in no particular order.
25356 There are many other possibilities, limited only by your imagination,
25357 good sense, and the constraints of your build system.
25359 A precompiled header file can be used only when these conditions apply:
25363 Only one precompiled header can be used in a particular compilation.
25366 A precompiled header can't be used once the first C token is seen. You
25367 can have preprocessor directives before a precompiled header; you cannot
25368 include a precompiled header from inside another header.
25371 The precompiled header file must be produced for the same language as
25372 the current compilation. You can't use a C precompiled header for a C++
25376 The precompiled header file must have been produced by the same compiler
25377 binary as the current compilation is using.
25380 Any macros defined before the precompiled header is included must
25381 either be defined in the same way as when the precompiled header was
25382 generated, or must not affect the precompiled header, which usually
25383 means that they don't appear in the precompiled header at all.
25385 The @option{-D} option is one way to define a macro before a
25386 precompiled header is included; using a @code{#define} can also do it.
25387 There are also some options that define macros implicitly, like
25388 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
25391 @item If debugging information is output when using the precompiled
25392 header, using @option{-g} or similar, the same kind of debugging information
25393 must have been output when building the precompiled header. However,
25394 a precompiled header built using @option{-g} can be used in a compilation
25395 when no debugging information is being output.
25397 @item The same @option{-m} options must generally be used when building
25398 and using the precompiled header. @xref{Submodel Options},
25399 for any cases where this rule is relaxed.
25401 @item Each of the following options must be the same when building and using
25402 the precompiled header:
25404 @gccoptlist{-fexceptions}
25407 Some other command-line options starting with @option{-f},
25408 @option{-p}, or @option{-O} must be defined in the same way as when
25409 the precompiled header was generated. At present, it's not clear
25410 which options are safe to change and which are not; the safest choice
25411 is to use exactly the same options when generating and using the
25412 precompiled header. The following are known to be safe:
25414 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
25415 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
25416 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
25421 For all of these except the last, the compiler automatically
25422 ignores the precompiled header if the conditions aren't met. If you
25423 find an option combination that doesn't work and doesn't cause the
25424 precompiled header to be ignored, please consider filing a bug report,
25427 If you do use differing options when generating and using the
25428 precompiled header, the actual behavior is a mixture of the
25429 behavior for the options. For instance, if you use @option{-g} to
25430 generate the precompiled header but not when using it, you may or may
25431 not get debugging information for routines in the precompiled header.